**** BEGIN LOGGING AT Thu Jan 26 02:59:57 2012 **** ENDING LOGGING AT Thu Jan 26 03:37:48 2012 **** BEGIN LOGGING AT Thu Jan 26 03:55:12 2012 Jan 26 16:52:28 bluelightning: how would you make builders aware they have to build zaurus-installer (or a bootloader fwiw) with a separate recipe? Jan 26 16:52:42 some oewarn parsing machine files? Jan 26 17:03:39 ant_work: so, I'm a little confused as to why not building it with the image is a good idea Jan 26 17:12:31 first reason is the generation of linux-kexecboot has been easily broken in the past Jan 26 17:12:46 e.g. by changes in oe-core Jan 26 17:12:56 so at least one can build a normal image Jan 26 17:14:18 how do othe rlayers do? do they deploy an u-boot on each bitbake image? Jan 26 17:15:09 where is the documentation suppose to reside in the BSP ? Jan 26 17:15:20 +d Jan 26 17:17:05 and fwiw we can imagine to deploy a package for linux-kexecboot able to self-flash itself (not on collie) Jan 26 17:17:34 and put it in the feeds Jan 26 17:19:19 hmm, ok Jan 26 17:19:28 well bbnote might help Jan 26 17:21:57 hm.. about docs I see Jan 26 17:22:00 This README file contains information on building the meta-crownbay Jan 26 17:22:00 BSP layer Jan 26 17:22:01 ... Jan 26 17:22:17 which layer is that? Jan 26 17:22:24 http://git.yoctoproject.org/cgit/cgit.cgi/meta-intel/tree/meta-crownbay/README Jan 26 17:22:39 im browsing Intel Jan 26 17:22:54 oh right... that sounds reasonable Jan 26 17:23:15 well, one layer for Zaurus, one for ipaq, ... Jan 26 17:23:37 s/layer/subdir/ Jan 26 17:23:39 :p Jan 26 17:23:59 and a /common Jan 26 17:24:30 seems very clean and logic Jan 26 17:25:48 heh.. I'm trying to see how TI did it but they are down :/ Jan 26 17:26:10 well, not TI...angstrom Jan 26 17:29:20 bbl **** ENDING LOGGING AT Thu Jan 26 17:46:57 2012 **** BEGIN LOGGING AT Thu Jan 26 17:48:20 2012 Jan 26 22:25:28 bluelightning: I seemingly can't see how core image-sato adds kernel in boot while core-image-minimal does not Jan 26 22:25:41 with same machine config Jan 26 22:27:04 ant__: maybe some package in core-image-sato pulls in the kernel? Jan 26 22:27:26 ant__: if you enable buildhistory or testlab you can get depgraphs of just what's in the image Jan 26 22:27:37 probable, I was looking to see if core-image-minimal does remove that instead Jan 26 22:30:20 it's put there by pkg_postinst_kernel-base () Jan 26 22:31:53 sure, kernel-base is not installed, that's why Jan 26 22:33:48 what causes that to be installed though? Jan 26 22:39:04 heh, I'm looking in task-base right now Jan 26 22:40:34 is complicatged, it starts from core-image.bbclass Jan 26 22:42:16 o Jan 26 22:42:25 in rootfs I see Jan 26 22:42:27 Installing kernel-3.1.8 (3.1-r0.8) to root... Jan 26 22:42:27 Downloading file:/oe/oe-core/build/tmp-eglibc/deploy/ipk/poodle/kernel-3.1.8_3.1-r0.8_poodle.ipk. Jan 26 22:42:27 Installing kernel-image-3.1.8 (3.1-r0.8) to root... Jan 26 22:42:40 between the modules Jan 26 23:09:37 bluelightning: ah, must be that Jan 26 23:09:40 ./classes/module.bbclass:RDEPENDS += "kernel-image update-modules" Jan 26 23:10:04 the modules are pulling it the kernel Jan 26 23:10:16 hmm ok Jan 26 23:10:18 and the modules are pulled in task-base-extended Jan 26 23:10:41 ah right, task-base-extended... which you don't get with minimal Jan 26 23:10:55 yes Jan 26 23:11:37 ok, what to do? I'd like the kernel in all images Jan 26 23:12:04 and the modules ofc Jan 26 23:12:34 well, core-image-minimal is just a test Jan 26 23:12:47 console-image was more rich Jan 26 23:27:28 btw, last present of Koen: NOTE: multiple providers are available for jpeg (jpeg, libjpeg-turbo) Jan 26 23:28:36 forces to NOTE: consider defining a PREFERRED_PROVIDER entry to match jpeg-native Jan 26 23:39:49 ant__: just set IMAGE_INSTALL_append = " kernel-image" (or kernel, I think that does the same thing) Jan 26 23:40:56 is an IMAGE_ var ok in machine.conf? Jan 26 23:42:28 ant__: no Jan 26 23:42:55 well, just do what I did in h3600.conf if that's what you want... Jan 26 23:43:05 if it's really essential for boot Jan 26 23:43:39 I have to consider the initramfs images case, should not be overpopulated Jan 26 23:43:55 alternatively you could make a new image recipe with the desired items in it Jan 26 23:44:24 ant__: well, task-core-boot shouldn't appear in initramfs images though, so MACHINE_ESSENTIAL_EXTRA_* shouldn't affect it Jan 26 23:57:09 I see Jan 27 00:09:01 wow, I think I've fixed tzdata postinst finally Jan 27 00:09:42 if I don't fall asleep I'll commit it after full rebuild Jan 27 00:27:41 gn **** ENDING LOGGING AT Fri Jan 27 02:59:57 2012 **** BEGIN LOGGING AT Fri Jan 27 02:59:58 2012 Jan 27 09:06:42 hi bluelightning, Jay7 Jan 27 09:07:03 hi ant_work, Jay7 Jan 27 09:07:33 Jay7: I wanted to upload screenshots/fotos but I'd have to change the v.0.6 label ;) What to do? Jan 27 09:07:59 release 0.6 as it is? Jan 27 09:08:26 I have this half-finished patch for the gui+tui Jan 27 09:08:56 morning ppl Jan 27 09:09:33 ant_work: where you want to change v.0.6 label? Jan 27 09:09:46 my picture shows 0.5.9 Jan 27 09:10:11 the web page says 0.6 Jan 27 09:10:43 maybe we change that to 0.5 Jan 27 09:13:50 ah Jan 27 09:13:51 change label on site to 0.5.9 :) Jan 27 09:15:13 ok, actual 0.6 -> 0.5 and upload 0.5.9 for now Jan 27 09:16:01 I'll do a 'family picture' of 3 Z Jan 27 09:17:09 nice :) Jan 27 23:33:31 Jay7: once finished that I'll move to kexecboot again ;) Jan 27 23:33:43 :) Jan 27 23:33:55 * Jay7 is trying to setup ntbackup properly Jan 27 23:34:05 what I miss is, /etc/timezone is not installed by this script if missing Jan 27 23:34:21 http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/sys-libs/timezone-data/timezone-data-2011n.ebuild?diff_format=s&revision=1.2&view=markup Jan 27 23:34:38 only /etc/localtime is recreated Jan 27 23:34:53 I'll let the first part of my previous patch and see Jan 27 23:37:20 Jay7: like this Jan 27 23:37:22 http://paste.debian.net/153839/ Jan 27 23:38:26 I'm a bit out of context, sorry :) Jan 27 23:39:35 I'm here just because meta-oe has old 2011g which alwais fails to fetch on first build :p Jan 28 00:13:00 * Jay7 -> sleep Jan 28 00:25:21 gn **** ENDING LOGGING AT Sat Jan 28 03:00:00 2012 **** BEGIN LOGGING AT Sat Jan 28 03:00:00 2012 Jan 28 21:56:32 Jay7: all3 devices flashed with latest kexecboot Jan 28 21:56:46 I'll fill some partitions and take a picture Jan 28 22:23:51 bah, need a real camera then **** ENDING LOGGING AT Sun Jan 29 02:59:58 2012 **** BEGIN LOGGING AT Sun Jan 29 02:59:58 2012 Jan 29 22:09:50 bluelightning: I'm building opie finally. For the moiment I've got only openobex-1.5 fatch error Jan 29 22:13:37 hi ant__ Jan 29 22:13:47 hmm... that's not right Jan 29 22:13:52 hey Jan 29 22:14:36 btw no distro, preferred-opie-versions-1.2.5.inc Jan 29 22:14:56 right Jan 29 22:15:08 is on krnel.org Jan 29 22:15:08 so openobex is in meta-oe Jan 29 22:15:10 was Jan 29 22:15:12 yes :( Jan 29 22:18:03 that's all, image built Jan 29 22:20:09 seems there is need for a default preferred provider for ntpdate, jpeg, jpeg-native Jan 29 22:22:12 ok, opie-image also lacks kernel & modules Jan 29 22:39:40 I don't know what the correct way of specifying those for distro-less would be Jan 29 22:39:50 (the preferred providers) Jan 29 22:42:42 did you consider setting MACHINE_ESSENTIAL_EXTRA_RDEPENDS ? Jan 29 22:44:38 see, I put it in MACHINE_ESSENTIAL_EXTRA_RRECOMMENDS_append Jan 29 22:44:44 but no result :/ Jan 29 22:45:13 btw, with core-image-sato is even worse Jan 29 22:45:35 for it lacks knowledge about kernel an kernel-image targets Jan 29 22:45:52 I suspect it does not include kernel.bbclass Jan 29 22:46:18 argh Jan 29 22:46:24 s/sato/minimal/ Jan 29 22:47:04 ( I put it in IMAGE_EXTRADEPENDS, my bad) Jan 29 22:47:37 this is build dependency, not runtime Jan 29 22:52:17 look, I'm copying kernel+modules by hand to boot opie :) Jan 29 22:52:39 http://aseigo.blogspot.com/2012/01/reveal.html Jan 29 22:53:24 time for meta-kde :) Jan 29 22:53:38 and for another kexecboot port ;) Jan 29 22:55:21 * Jay7 -> sleep Jan 29 22:55:48 ant__: you need to rebuild task-core-boot if you change it after building that Jan 29 22:56:13 Jay7: I started a kde layer with what we had in OE (i.e. not much) Jan 29 22:56:24 Jay7: I'm not able to photograph Zaurus lcd's with iphone... Jan 29 22:56:40 ant__: np Jan 29 22:57:02 I'll make some shots from qemu once Jan 29 22:57:29 I have to take some serious shots and process with PS Jan 29 22:57:35 it would be great to make Mer buildable on top of OE.. Jan 29 22:57:41 bluelightning: opie booting on poodle Jan 29 22:58:17 ts even calibrate :p Jan 29 22:58:34 big progress :) Jan 29 22:58:35 awesome fast Jan 29 22:58:45 sweet :) Jan 29 22:58:45 32MB ram Jan 29 22:59:35 seems working oob Jan 29 22:59:39 I should be able to run opie on collie then :) Jan 29 23:00:53 it used to work on collie... so it should still :) Jan 29 23:01:06 we might need to update a few things Jan 29 23:01:07 wrong keymap Jan 29 23:01:13 in console Jan 29 23:01:18 hmm :( Jan 29 23:01:40 well.. anyway, good night :) Jan 29 23:01:41 * Jay7 gone Jan 29 23:01:46 night :) Jan 29 23:02:19 bluelightning: at lest /? key works Jan 29 23:02:56 I remember worse situations, w/out slashes Jan 29 23:03:03 heh yeah Jan 29 23:03:12 seems the modifiers are missing Jan 29 23:03:19 no numbers :) Jan 29 23:03:38 but is really fast compared to sato/x11 Jan 29 23:04:19 very responsive, I guess we have really optimized settings Jan 29 23:04:54 well, it was built on the platform designed to run on the hardware :) Jan 29 23:05:41 first opie's with older kernels were slower on c7x0/spitz Jan 29 23:05:46 years ago Jan 29 23:05:59 kernel is really snappy Jan 29 23:06:33 hmm... I know we did have a performance regression somewhere along the line, I forget when Jan 29 23:07:16 hmm Integer 890472 dh FP 1.39 secs Jan 29 23:08:02 now 1.25 secs Jan 29 23:08:31 I remember those tests are a bit liers Jan 29 23:09:21 http://comments.gmane.org/gmane.linux.distributions.angstrom.devel/3193 Jan 29 23:11:56 anyway it's almost unbelievable Jan 29 23:12:35 ah, btw Jan 29 23:13:03 I commented udev's DEVICE_CACHE or what in /etc/default Jan 29 23:15:38 did you need to do that? Jan 29 23:15:59 yes if using meta-oe Jan 29 23:16:48 heh, drawpad is as well responsive Jan 29 23:18:53 now, if poodle only would shut-off Jan 29 23:18:57 :/ Jan 29 23:19:08 we need a way to disable the device cache properly Jan 29 23:19:44 koen behavior is funny, he said 'get rid of it' but then no comments about my patch to make it opt-in Jan 29 23:20:07 just ping, he does... Jan 29 23:20:31 he applied 1/2 and was silent on 2/2 Jan 29 23:20:57 np, I'll bump the thread Jan 29 23:21:27 but again, he's the first saying we don't have any use of that cache Jan 29 23:25:07 btw do hardware buttons work? Jan 29 23:25:10 ok, back to opie, there must have been keymaps in oe-classic Jan 29 23:26:21 mom, I'm flashing it Jan 29 23:26:34 yes Jan 29 23:29:18 i.e Calendar key works Jan 29 23:29:32 Contacts too Jan 29 23:29:50 Menu too Jan 29 23:30:07 Mail no, maybe is not defined Jan 29 23:30:58 Cancel seems doing Suspending... Jan 29 23:31:45 OK opens the 'start' menu Jan 29 23:32:39 central key (between round of cursors) is also not working Jan 29 23:32:53 it works in kexecboot Jan 29 23:33:06 so kernel patch is ok Jan 29 23:35:05 hmm, maybe another keymap thing Jan 29 23:35:16 surely Jan 29 23:35:27 I did find strange stuff with hardware buttons on other devices in more recent kernels Jan 29 23:35:41 our russian friends are the experts (by necessity...) Jan 29 23:35:44 still not sure why Jan 29 23:35:49 hmm yes :) Jan 29 23:38:25 ok, now I got a kernel in core-image-minimal Jan 29 23:39:27 but I lack the symlink provided in kernel-image-postginst hm... Jan 29 23:39:39 ah.. kernel-base postinst Jan 29 23:40:19 I'm using brute force as you said Jan 29 23:40:21 IMAGE_INSTALL_append = " kernel-base kernel-image " Jan 29 23:42:01 if we use IMAGE, then why not Jan 29 23:42:15 EXTRA_IMAGEDEPENDS += "virtual/kernel" Jan 29 23:42:27 once for all? Jan 29 23:49:55 EXTRA_IMAGEDEPENDS shouldn't be necessary... Jan 29 23:51:51 ok, but it does the trick Jan 29 23:52:09 I mean Jan 29 23:52:10 EXTRA_IMAGEDEPENDS += "zaurus-installer" Jan 29 23:52:23 was not nice... Jan 29 23:52:48 while IMAGE_INSTALL_append = " kernel-base kernel-image " does the trick Jan 29 23:53:29 one for the kernel, other for the symlink, .. nowthe modules are still missing Jan 29 23:57:52 I'm wondering if we have to deal with task-base-extended at some point Jan 29 23:58:02 heh Jan 29 23:58:12 that would bring in MACHINE_EXTRA_RDEPENDS Jan 29 23:58:29 including modules not needed for boot Jan 29 23:59:20 I'll try to see how Yocto choices modules basing on BSP and tasks Jan 29 23:59:39 we could maybe converge soon the kernel Jan 29 23:59:57 the problem with task-base-extended is I think it forces a build of a whole bunch of X stuff not needed for opie Jan 30 00:01:38 the point is some devices always need kernel + modules Jan 30 00:01:51 we'd need a knob to easily choice Jan 30 00:02:03 for an easy choice Jan 30 00:02:27 it seems a bit sparse Jan 30 00:07:37 good night Jan 30 00:07:41 night **** ENDING LOGGING AT Mon Jan 30 02:59:59 2012 **** BEGIN LOGGING AT Mon Jan 30 02:59:59 2012 **** ENDING LOGGING AT Tue Jan 31 02:59:58 2012 **** BEGIN LOGGING AT Tue Jan 31 02:59:58 2012 **** BEGIN LOGGING AT Tue Jan 31 17:23:32 2012 **** ENDING LOGGING AT Wed Feb 01 02:59:57 2012 **** BEGIN LOGGING AT Wed Feb 01 02:59:57 2012 Feb 01 16:56:58 another zaurus user in angstrom-distro-users@ Feb 01 16:57:15 complaining on narcissus Feb 01 17:01:00 yeah, I'm guessing angstrom is no longer building for zaurus machines Feb 01 17:01:08 as in, the angstrom autobuilder Feb 01 17:01:26 that process is very opaque unfortunately :/ Feb 01 17:15:41 we should release some images Feb 01 17:31:31 if you release images you need feeds Feb 01 17:31:38 then you might as well start up a proper distro Feb 01 17:31:47 which might be the right thing to do, I don't know **** ENDING LOGGING AT Thu Feb 02 02:59:57 2012 **** BEGIN LOGGING AT Thu Feb 02 02:59:58 2012 **** ENDING LOGGING AT Fri Feb 03 02:59:57 2012 **** BEGIN LOGGING AT Fri Feb 03 02:59:57 2012 Feb 03 23:53:46 ka6sox: don't worry, I'm testing the firewalls **** ENDING LOGGING AT Sat Feb 04 02:59:58 2012 **** BEGIN LOGGING AT Sat Feb 04 02:59:58 2012 Feb 04 20:25:47 hm, is this the wrong channel to ask questions about how to get a crashkernel running? if so, where'd be the best channel to ask otherwise? **** ENDING LOGGING AT Sun Feb 05 02:59:59 2012 **** BEGIN LOGGING AT Sun Feb 05 02:59:59 2012 **** BEGIN LOGGING AT Sun Feb 05 08:38:10 2012 **** ENDING LOGGING AT Mon Feb 06 02:59:58 2012 **** BEGIN LOGGING AT Mon Feb 06 02:59:58 2012 Feb 06 16:45:41 bluelightning: hi, any concrete outcomes out of the meeting? Feb 06 16:46:53 ant_work: it was a pretty informal meeting, a collection of discussions Feb 06 16:47:20 ant_work: the first though was that koen is happy with my proposals on meta-oe splitting as set out in the email Feb 06 16:47:37 I did make some other notes which I need to put in the form of a post to the list Feb 06 16:48:10 great, thx Feb 06 16:49:31 fwiw opie-image is broken atm..dunno why, was ok 1 week ago :/ Feb 06 16:51:00 and we have to find a way to add the modules (all?) to the images Feb 06 16:52:18 finally, thre is some strange rtc bug/mismatch. pls update the kernel to latest 3.1 when you'll have time Feb 06 16:52:41 ant_work: I wonder about "all"... the machines in meta-yocto do that and it means I get things like ocfs2 in sato images which is not really very useful on an embedded device Feb 06 16:53:05 of course you can blame the kernel config but it seems to me we should be selective about which modules we need Feb 06 16:53:07 look at qemu recipes Feb 06 16:53:19 one calls 'kernel-modules' Feb 06 16:53:50 I saw your patches, will apply them tonight Feb 06 16:54:10 opie-image is broken - what happens? Feb 06 16:54:41 server does not start, I get only serial Feb 06 16:54:51 I suspect some udev change :/ Feb 06 16:57:38 *sigh* Feb 06 17:07:53 about modules, I tried to add MACHINE_EXTRA_RRECOMMENDS_append = " kernel-modules" Feb 06 17:08:23 or alike, one that was in core-base Feb 06 17:08:30 still no modules Feb 06 17:08:58 So I need to investigate the qemu tasks Feb 06 17:09:22 (or maybe qemumips.conf is just broken) Feb 06 17:10:17 I think you need task-base-extended for those to be installed Feb 06 17:11:37 there is to say that the modules needed for booting are installed but probably networking isn't so upgrade is painful Feb 06 17:12:04 installed = in kernel Feb 06 17:13:20 finally a word about tzdata: I could not care less now than the old versions have been purged from meta-oe :) Feb 06 17:13:40 *that Feb 06 17:14:27 now the udev batltle... Feb 06 17:14:51 then xserver-nodm-init... Feb 06 17:15:21 after that we should have less issues stacking meta-oe Feb 06 17:16:35 heading home, bbl Feb 06 17:16:36 indeed, thanks for the work you've done so far :) Feb 06 17:16:38 cya Feb 07 00:07:38 bluelightning: seems is not the known udev issue, maybe systemd Feb 07 00:07:46 is breaking opie Feb 07 00:09:55 seems the latest systemd.bbclass Feb 07 00:11:42 rebuilding, will see tomorrow **** ENDING LOGGING AT Tue Feb 07 02:59:58 2012 **** BEGIN LOGGING AT Tue Feb 07 02:59:58 2012 Feb 07 15:49:39 omg what a monster patch in the list Feb 07 15:50:31 let's hope systemd is really separated afterwards Feb 08 00:00:44 if the handoff from one kernel to the next doesn't work correctly (not all hardware is initialized correctly) ... what should I be looking at? Feb 08 00:01:15 later kernel to in clude in the bootmanager? different kernel settings? **** ENDING LOGGING AT Wed Feb 08 02:59:57 2012 **** BEGIN LOGGING AT Wed Feb 08 02:59:57 2012 Feb 08 21:51:19 bluelightning: hi there Feb 08 22:13:16 hi ant____ Feb 08 22:13:40 I'm wondering about meta-oe split Feb 08 22:14:04 I would import in meta-handheld klibc, klibc.bbclass and kexec-tools-klibc as first move Feb 08 22:14:22 then Zaurus would not depend anymore on meta-oe Feb 08 22:14:41 (well, there is still xinput-calibrator to put in oe-core) Feb 08 22:15:15 it is like that: if you use kdrive you have tslib in oe-core Feb 08 22:15:33 if you use xserver you need xinput-calibrator from meta-oe Feb 08 22:15:47 this leads to following RP commit Feb 08 22:16:34 http://patches.openembedded.org/patch/16325/ Feb 08 22:17:35 ant____: the first thing is to make meta-oe reusable Feb 08 22:17:49 ant____: or, more usable in different contexts Feb 08 22:18:10 the plan I've outlined in the email thread is a start for that Feb 08 22:18:11 I tend to forget about it for the moment Feb 08 22:18:21 it's too full of systemd'isms Feb 08 22:19:41 well, that's another problem to address Feb 08 22:19:54 to build core-image-sato with full xserver-xorg we just need xinput-calibrator Feb 08 22:20:05 that's for the shortcoming future Feb 08 22:20:41 or wait for the holy-grail unification of xserver-nodm-init Feb 08 22:21:32 as you see I become a bit more pessimistic ... ;) Feb 08 22:22:20 the reason is simple: people tend to stay away from meta-oe so there are less testers Feb 08 22:22:47 it just gained a bad reputation... Feb 08 22:23:25 well this is it Feb 08 22:23:35 see how there is zero comment from oe-core people to the thread about systemd Feb 08 22:23:40 the X issues are just one aspect, I'm trying to solve the others first Feb 08 22:24:17 that's not entirely true, one of my colleagues Joshua Lock was in that thread I think Feb 08 22:24:26 yes, we still have udev beforehand Feb 08 22:24:50 systemd is on our radar, we just have bigger issues to consider before that Feb 08 22:24:59 btw, about that RP comments, machine shouldn't poke Feb 08 22:25:38 seems we have to remove all providers but virtual/kernel and virtual/xserver Feb 08 22:26:23 if you can make a case that the thing you are setting is machine specific it's ok Feb 08 22:28:09 I'm just wondering about libgl and libx11 Feb 08 22:28:20 iirc RP suggested that settings Feb 08 22:28:27 for Z Feb 08 22:31:34 you can always ask RP for clarification Feb 08 22:36:06 s#@t, opie-image doesn't start ... Feb 08 22:37:51 ant____: has to be something recent; it worked only a week or two ago on h1940 Feb 08 22:37:59 ant____: could be missing devices Feb 08 22:38:00 yes Feb 08 22:38:41 I fear systemd contamination Feb 08 22:39:27 ant____: is the initscript being started? Feb 08 22:40:49 I see udev starting Feb 08 22:41:03 then I get prompt on serial Feb 08 22:41:16 and can't login :/ Feb 08 22:41:35 password is not zapped in opie-image Feb 08 22:42:24 if it was zapped you would not be able to login... so... ? Feb 08 22:42:37 usually is blank Feb 08 22:42:44 yes Feb 08 22:42:48 when using image_features std Feb 08 22:42:55 if debug-tweaks is in IMAGE_FEATURES Feb 08 22:42:58 y Feb 08 22:43:11 not yet Feb 08 22:43:18 not now in opie Feb 08 22:43:41 let me retry a working image and I diff it Feb 08 23:14:45 bluelightning: working image was built early on dec 3nd Feb 08 23:14:54 3rd Feb 08 23:15:03 right, that gives us somewhere to start Feb 08 23:15:18 fwiw is missing utc tmezone, just american Feb 08 23:15:48 strange, /etc/localtime is Universal Feb 08 23:16:04 timezone is ok as well Feb 08 23:16:13 so it'a just opie Feb 08 23:16:49 metadata from dec 2nd night, built on 3rd Feb 08 23:19:08 eh Feb 08 23:19:33 there are udev-180 and various systemd commits Feb 08 23:20:34 the systemd.bbclass one being the worst Feb 08 23:37:44 bluelightning: as far as I see there are minor size diffs (8bytes) in libopietooth.so.1.0.0 and libopieui2.so.1.9.2 Feb 08 23:37:49 maybe random Feb 08 23:38:01 the other binarise in /bin and /Sbin match Feb 08 23:38:31 ant____: maybe try copying the files over, see if that's really significant or not Feb 08 23:41:06 argh Feb 08 23:41:15 in the 'good' opie-image Feb 08 23:41:29 there is /usr/sbin/update-rc.d Feb 08 23:41:43 missing in the latest :/ Feb 08 23:41:57 in fact thgere is no initial 'configuring' ... Feb 08 23:43:07 right, that'll be it then Feb 08 23:45:17 there is more.. Feb 08 23:45:31 in the 'good' image I have Feb 08 23:45:54 this is exactly what buildhistory is good for detecting btw Feb 08 23:45:59 heh Feb 08 23:46:12 Iah, no, it's just size Feb 08 23:46:28 /usr/sbin/chpasswd symlink Feb 08 23:46:36 well, it would have picked up that update-rc.d was missing **** ENDING LOGGING AT Thu Feb 09 00:04:03 2012 **** BEGIN LOGGING AT Thu Feb 09 00:12:17 2012 Feb 09 00:23:03 gn **** ENDING LOGGING AT Thu Feb 09 00:58:30 2012 **** BEGIN LOGGING AT Thu Feb 09 01:10:18 2012 **** ENDING LOGGING AT Thu Feb 09 02:59:57 2012 **** BEGIN LOGGING AT Thu Feb 09 02:59:58 2012 Feb 09 17:23:01 if the handoff from one kernel to the next doesn't work correctly (not all hardware is initialized correctly) ... what should I be looking at? Feb 09 17:23:06 later kernel to include in the bootmanager? different kernel settings? Feb 09 17:25:35 exco: what device you are using? Feb 09 17:25:54 Toshiba Ac100 Feb 09 17:33:11 Jay7, ideas? Feb 09 17:47:40 Tequila: ping Feb 09 17:47:55 Jay7, pong Feb 09 17:47:58 Tequila have AC-100 with working setup Feb 09 17:48:06 hey exco Feb 09 17:48:09 :D Feb 09 17:48:15 Jay7, no he isn't ;-) Feb 09 17:48:19 hey Tequila Feb 09 17:48:59 so I managed to build your branch ... sent you an email why it won't build pulling your rep atm Feb 09 17:49:08 need to track back your request... but sorry I'm fully busy Feb 09 17:49:33 never mind ... just so you know (but you might see it a lot faster) Feb 09 17:49:40 wrong HW initialisation is known problem of ARM/MIPS arch'es Feb 09 17:49:57 that's mostly because kexec is resetting devices Feb 09 17:49:59 yep, wrong shutdown too Feb 09 17:50:20 is there anything I can help with ... or do I need to sit back and wait for kexec to improve? Feb 09 17:50:30 exco, I had a mail from a guy who managed to work-around the problem using a kernel with USB disabled Feb 09 17:50:55 the kernel used for the kexecboot, not the final kernel Feb 09 17:50:56 usb disabled for the ac100 kernel or the bootloader kernel? Feb 09 17:51:00 ah ok Feb 09 17:51:05 bootloader yep Feb 09 17:51:54 that means no usb key can be used with kexecboot Feb 09 17:51:56 we could just enable the minimum stuff needed? and the 2nd kernel would initialize everything as needed? Feb 09 17:52:12 but it should be sufficient to keep functional multiboot Feb 09 17:52:22 try to make noise in some kernel ML Feb 09 17:52:28 yep that's what I guess Feb 09 17:52:37 about USB vs kexec Feb 09 17:53:03 we first need to try more recent kernel of course Feb 09 17:53:10 I can make the noise but I'd rather not since I'm like at the beginning stages of kernel comilation (following howto stuff) Feb 09 17:53:37 ok so I'll try to compile multiboot with a more recent kernel ... that might take some time. Feb 09 17:54:27 there's just too many options in kernel configuration ... and more with each later version Feb 09 17:55:53 first maybe try the current one with usb support disabled Feb 09 17:56:35 yes, that I can handle ... Feb 09 17:57:07 would be nice ... then I can have a working AC100 and move on to working on compiling ics Feb 09 17:57:57 or Ubuntu/2.3 to make use of the video capabilities (as a media player) Feb 09 18:20:56 Tequila, Jay7 I will let you know if I manage to get a step further and then we (you) can make the noise Feb 09 18:20:59 thanks for your help **** ENDING LOGGING AT Fri Feb 10 02:59:57 2012 **** BEGIN LOGGING AT Fri Feb 10 02:59:57 2012 **** BEGIN LOGGING AT Fri Feb 10 07:28:11 2012 **** ENDING LOGGING AT Sat Feb 11 02:59:57 2012 **** BEGIN LOGGING AT Sat Feb 11 02:59:57 2012 Feb 11 22:16:25 Jay7: have you tested kernel on collie recently? **** ENDING LOGGING AT Sun Feb 12 02:59:56 2012 **** BEGIN LOGGING AT Sun Feb 12 02:59:56 2012 Feb 12 12:48:33 ant: no **** ENDING LOGGING AT Mon Feb 13 02:59:58 2012 **** BEGIN LOGGING AT Mon Feb 13 02:59:58 2012 **** ENDING LOGGING AT Tue Feb 14 02:59:56 2012 **** BEGIN LOGGING AT Tue Feb 14 02:59:56 2012 Feb 14 12:04:24 bluelightning: fyi I've fwd'd to you the unanswered mail I sent to DH Feb 14 12:06:49 ant_work: hmm, don't think i received it yet... Feb 14 12:08:47 an answer would have been propedeuthic to the layer split Feb 14 12:10:30 though I can't see reasons against a meta-initramfs layer Feb 14 12:16:57 still, klibc and kexecboot should maybe reside in oe-core Feb 14 13:14:42 ant_work: btw I think he's at ELC so that might be why you haven't had any response Feb 14 13:14:55 hmm, well it was a month ago Feb 14 13:14:59 send a ping Feb 14 13:15:10 or I can follow up with him if you like Feb 14 13:17:21 as you like, ask him if he has particular needs for the layer Feb 14 13:19:00 in one way it is good to have the recipes out of meta-handheld too (some .bbappend needed) Feb 14 13:19:30 I'm specially interested in the modular kernel like linux-yocto Feb 14 23:04:06 emerge -auD system --newuse Feb 14 23:04:33 (wrong window ofc) Feb 14 23:20:56 bluelightning: have you thought about a different hierarchy for meta-handheld ? Feb 14 23:21:08 maybe like meta-intel ? Feb 14 23:21:26 so well' have the right place for the documentation Feb 14 23:22:13 i.e. meta-zaurus, meta-ipaq, ... Feb 14 23:22:42 ant____: I'd like to leave the structure as-is for the moment Feb 14 23:22:57 I don't want to complicate things yet Feb 14 23:23:33 well, the README have yet to be written :) Feb 14 23:26:09 we can always have separate READMEs Feb 14 23:26:17 or have the documentation elsewhere Feb 14 23:27:55 now the questionable thing is the dependency on meta-openembedded Feb 14 23:28:21 that udev .bbappend complicates the things Feb 14 23:30:09 as are doing the recipes in meta-handheld inheriting klibc.bbclass Feb 14 23:32:32 anyway, I'm updating the servers those days so I'll wait for the fix to the systemd breakage Feb 14 23:33:24 before rebuilding Feb 14 23:34:41 (I'm battling with a stupid fw updater for the fujitsu hdd's , 1,5 gb image when uncompressed - Centos ) Feb 14 23:35:08 docs binaries and libs are less than 110 mb Feb 14 23:35:20 some crazy man Feb 14 23:35:29 *not less Feb 14 23:36:29 tonight I got the light after disassembling some lines ..it needs to be connected and download it :) Feb 14 23:37:24 110mb were not enough ;) Feb 15 00:00:55 gn **** ENDING LOGGING AT Wed Feb 15 02:59:57 2012 **** BEGIN LOGGING AT Wed Feb 15 02:59:57 2012 Feb 15 23:37:44 bluelightning: hi Feb 15 23:37:50 wip Feb 15 23:37:52 NOTE: package linux-yocto-tiny-3.0.12+git1+04a52a32cbdf0972033b97b83eaa83eb275dfdc9_1+f389d310965a56091f688b28ea8be6d9cbb7fbbe-r1: task do_fetch: Started Feb 15 23:37:52 WARNING: Failed to fetch URL http://kernel.org/pub/linux/kernel/people/jsipek/guilt/guilt-0.33.tar.gz Feb 15 23:55:10 gn **** ENDING LOGGING AT Thu Feb 16 02:59:57 2012 **** BEGIN LOGGING AT Thu Feb 16 02:59:58 2012 Feb 16 09:44:33 bluelightning: so I've played a bit with linux-yocto and linux-yocto-tiny Feb 16 09:45:26 the first obstacle is to set KMACHINE Feb 16 09:47:16 ant_work: ah ok Feb 16 09:47:33 I used yocto/standard/base Feb 16 09:47:53 I have to admit I've not played around with our kernel stuff much Feb 16 09:48:12 it looks much x86 centric Feb 16 09:48:44 I'm still learning but first impression is I should start with linux-yocto-tiny Feb 16 09:49:15 which declares in SRC_URI the .config blocks Feb 16 09:49:21 ant_work: well, I will say that it's largely been developed by Wind River developers who support a lot more than just x86 Feb 16 09:49:43 oh, I'll check their BSP layer then Feb 16 09:50:02 I got the impression it was Intell stuf Feb 16 09:50:06 -l Feb 16 09:50:10 +f Feb 16 09:50:54 Intel people have worked on it since the Yocto Project started but AIUI it's was something that came from Wind River Feb 16 09:50:59 the point is, the BSP should declare the KERNEL_FEATURES of the machine Feb 16 09:51:19 ain that regard the qemu examples are not immediately clear Feb 16 09:58:35 I think this is the latest doc: http://www.yoctoproject.org/docs/current/kernel-manual/kernel-manual.html Feb 16 10:00:52 yes I think that would be it Feb 16 10:01:45 somehow combined with http://www.yoctoproject.org/docs/current/dev-manual/dev-manual.html#dev-manual-kernel-appendix Feb 16 10:02:08 ok, then I'd say we take the official SRCREV Feb 16 10:04:42 bluelightning: after so many years with vanilla I feel a bit uncomfortably with pre-mangled sources ;) Feb 16 10:05:05 it is an unusual approach, definitely Feb 16 10:07:40 the idea is setting once the basic machine/kernel features for boot from SD/CF/USB and forget about defconfigs Feb 16 10:08:09 and an extended set for the production kernel, with net and so on Feb 16 10:08:39 then linux-kexecboot and linux will need almost zeromaintainance Feb 16 10:08:56 eh, would, maybe... Feb 16 10:10:50 ah, I noticed that the many .cfg pieces always do build in-kernel (foo = y) so the modules load/unload mgmt is not necessary Feb 16 15:32:49 kenws: have you figured out how to abuse of linux-yocto? Feb 16 15:41:11 ant_work: You mean on how to build it using a vexpress defconfig? not really. I think it would work by adding things to the 'meta' branch of the linux-yocto-3.2 tree that has the fragments. But I don't know how to do it in your own layer. However, it's quite easy to provide your own kernel recipe and a defconfig (rather than using fragments). Feb 16 15:42:00 yes, but it is tempting to modularize it Feb 16 15:43:19 think about linux-kexecboot: it needs couple of standard kernel features plus machine specific bits Feb 16 15:43:40 i.e. some devices may boot from SD/CF/USB/NFS thus needing those filesystem Feb 16 15:44:03 another only boots form nand or tftp so it would be awaste to compile all other block devices Feb 16 15:45:08 my impression is that anyway we talk about blocks of options built in kernel, not modules Feb 16 15:45:38 so once we add ata ext2,3, .... we should be served Feb 16 15:46:02 this inflates th ekernel size, though Feb 16 15:47:32 "blocks of options" yeah, the ssc files pull in certain config files. I don't really remember whether there is a way to specify to build that feature as a module or build it into the kernel Feb 16 15:47:44 the whole thing seems to come from windriver folks Feb 16 15:48:39 no, its like 'foo=y' Feb 16 15:48:55 again, I just gave a first glimpse ;) Feb 16 15:48:59 ah, ok. Feb 16 15:49:50 I was playing with tohse http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-kernel/linux/linux-yocto-tiny Feb 16 15:51:24 hm Feb 16 15:51:35 yes, then there are the ones in /meta Feb 16 15:51:39 is this tree being pulled in when building oe-core? Feb 16 15:51:44 i thought it's http://git.yoctoproject.org/cgit/cgit.cgi/linux-yocto-3.2/tree/meta/cfg/kernel-cache/bsp?h=meta Feb 16 15:51:44 y Feb 16 15:52:07 at least for the qemuarm MACHINE Feb 16 15:52:13 I'm hacking around without meta-yocto layer Feb 16 15:52:55 what's your MACHINE? Feb 16 15:53:11 I'm testing with armv47 poodle Feb 16 15:53:17 -7 Feb 16 15:54:20 is "poodle" for real hardware? Feb 16 15:54:35 I have to admit that I've hever heard of it Feb 16 15:54:35 yes, unsupported in oe-core Feb 16 15:54:39 * kenws ducks Feb 16 15:54:41 : ) Feb 16 15:54:44 is one of older Zaurus Feb 16 15:55:20 ah, i reme,ber those but I thought they were armv5 or less Feb 16 15:55:40 o, yes this is even 32MB pxa250 Feb 16 15:55:48 collie is armv4 Feb 16 15:56:06 sry, all those dogs barfing... Feb 16 15:56:49 you see, pxa is very old and rather well supported by vanilla kernel Feb 16 15:57:21 if I start with -yocto is because I'm hoping it will need less mainainance Feb 16 15:57:28 ok. So, how do you change those fragments. My impression was that the meta branch gets pulled in when building the kernel (sort of merged in your kernel tree) Feb 16 15:57:47 yep, that's probably a good idea Feb 16 15:58:11 yes, but the magic happens having proper KMACHINE KTYPE KARCH Feb 16 15:59:14 as I understand it now, the workflow allows building from 'minimal' fragment and extend it Feb 16 15:59:16 OR Feb 16 15:59:25 just using a provided defconfig Feb 16 15:59:44 some sleepless night will tell soon ;) Feb 16 16:01:43 Did you find any documentation on this? I just noticed that the KMACHINE pulls in the fragment for your board (like arm-versatile-926ejs) Feb 16 16:02:39 yes Feb 16 16:02:52 see this log around 11h Feb 16 16:03:04 well, 4 hours ago Feb 16 16:03:53 ah, I see. Feb 16 16:04:51 yesterday night I gave up...I was fetching an unconsistent revision and git was stalled **** ENDING LOGGING AT Thu Feb 16 16:19:53 2012 **** BEGIN LOGGING AT Thu Feb 16 16:20:20 2012 **** ENDING LOGGING AT Thu Feb 16 16:24:35 2012 **** BEGIN LOGGING AT Thu Feb 16 16:26:17 2012 **** ENDING LOGGING AT Thu Feb 16 22:21:34 2012 **** BEGIN LOGGING AT Thu Feb 16 22:22:15 2012 **** ENDING LOGGING AT Fri Feb 17 02:59:57 2012 **** BEGIN LOGGING AT Fri Feb 17 02:59:57 2012 **** ENDING LOGGING AT Sat Feb 18 02:59:58 2012 **** BEGIN LOGGING AT Sat Feb 18 02:59:58 2012 Feb 18 22:59:09 Jay7: ping Feb 18 22:59:15 hi ant__ Feb 18 22:59:41 I seem to have found a too easy way for the tui+gui issue Feb 18 22:59:46 too easy... Feb 18 23:01:31 http://paste.debian.net/156773/ Feb 18 23:02:55 hehe Feb 18 23:03:48 well, I did not add in git the new file ;) Feb 18 23:03:55 anyway inside theme-tui.h all content should be protected with #ifdef USE_TEXTUI Feb 18 23:04:09 so I see no difference Feb 18 23:06:21 it happens to compile ;) Feb 18 23:12:22 is ok even with --enable-fbui=no --enable-textui Feb 18 23:12:48 I'll send you a patch, then decide if adding guards is easier Feb 18 23:14:37 the issue is ./res/theme.h:#define DEFAULT_FONT (&ter_u16n_ascii_font) Feb 18 23:14:37 Binary file ./gui.o matches Feb 18 23:14:37 Binary file ./tui.o matches Feb 18 23:15:45 hmm Feb 18 23:16:01 well, send me your patch Feb 18 23:16:08 I'll check it tomorrow Feb 18 23:16:14 both include theme.h Feb 18 23:17:00 ah, is it decided to go back one hour in Russia? Feb 18 23:17:26 not before new president :) Feb 18 23:17:40 I fear our new president would be our old president... Feb 18 23:17:41 he he Feb 18 23:17:52 hm Feb 18 23:18:55 i.e. Putin again :) Feb 18 23:19:11 too much Putin for one country :) Feb 18 23:19:26 we got 20year Berlusconi Feb 18 23:19:41 too bad Feb 18 23:19:50 too long = too bad Feb 18 23:20:14 yeah... same here.. 12 years as president + 4 years as premier-minister Feb 18 23:20:31 and trying to go for next 5 years Feb 18 23:20:49 or 6.. they are changed it Feb 18 23:24:37 ok, patch sent Feb 18 23:24:45 I see some strangeg tabs in one line Feb 18 23:25:08 got it Feb 18 23:32:48 btw Feb 18 23:32:50 configure: WARNING: unrecognized options: --disable-silent-rules, --with-libtool Feb 18 23:32:51 -sysroot **** ENDING LOGGING AT Sun Feb 19 02:59:58 2012 **** BEGIN LOGGING AT Sun Feb 19 02:59:58 2012 Feb 19 23:15:51 Jay7: I have runtime-tested the 3 linux-kexecboot: Feb 19 23:15:58 tui-only ok Feb 19 23:16:03 gui-only ok Feb 19 23:16:40 gui + tui = heh, tui drawed on the top of gui :) Feb 19 23:18:12 but basically you only see textui, the background is visible only during frame redraw (i.e. pressing left/right cursors) Feb 19 23:19:07 no problem to boot, anyway, it kexecs just fine Feb 19 23:24:14 I've sent you a small video Feb 19 23:25:46 somehow iphone did rotate the video... Feb 19 23:33:36 gn **** ENDING LOGGING AT Mon Feb 20 03:00:00 2012 **** BEGIN LOGGING AT Mon Feb 20 03:00:00 2012 Feb 20 12:31:25 bluelightning: I interrupted my linux-yocto experiments (kernel won't fetch) Feb 20 12:31:50 I see there is a mysterious poky-extras layer mentioned in th ekernel guide I was reading Feb 20 12:32:16 ant_work: hmm... probably best to check with zeddii/dvhart about that Feb 20 12:33:55 sure, as you already know I'm aiming at bare oe-core + BSP so it would be bad being bound to Poky or its extensions Feb 20 12:34:41 btw I see koen is on irc Feb 20 12:35:16 what exactly would you do to put klibc + related recipes together in meta-initramfs? Feb 20 12:35:30 where should the layer reside? in meta-oe? Feb 20 12:36:09 i.e. meta-openembedded/meta-initramfs? Feb 20 12:38:25 seems reasonable to me... Feb 20 12:45:17 koen: mumbling about a new meta-openembedded/meta-initramfs layer to host klibc + class + related static recipes. Thoughts? Feb 20 12:45:18 I'd even add kexecboot and linux-kexecboot there Feb 20 12:45:18 ant_work: sounds like a good idea Feb 20 12:47:11 ant_work: we could even move that to the oe-core repo (not layer) later on Feb 20 12:47:22 koen: ok, great. I'll send a msg to the list with the proposal and some patches. Thx. Feb 20 12:48:01 cool :) Feb 20 12:48:13 bluelightning: now, how to set compatible machine .. ?? Feb 20 12:48:36 why do you need to set it? Feb 20 12:52:19 i.e. linux-kexecboot Feb 20 12:53:43 well, I'll try to group all recipes in /meta-initramfs and make them machine-agnostic Feb 20 12:53:58 then I'll write some .bbappend for meta-handheld Feb 20 12:54:39 one thing I've seen.. some prefs are set in kexecboot.inc. This is not ideal when merging layetrs, isn't? Feb 20 12:54:59 we can't append to th e.inc files iirc Feb 20 13:12:09 ant_work: you can't no, but you can just append all recipes that use the .inc Feb 20 15:54:34 hi ppl Feb 20 15:54:41 ant_work: got your movie :) Feb 20 15:55:01 hi Feb 20 15:58:54 Jay7: gui+tui was only a few kb larger, tui only is much smaller Feb 20 15:59:06 yeah Feb 20 15:59:18 I don't have the numbers here but about 980kb, 1020, 1040 Feb 20 16:00:10 now, I should try to reproduce pfalcon issues Feb 20 16:01:56 it was June 2011 Feb 20 16:03:07 "4. Once I patched it to do fallback described in 2), it started to Feb 20 16:03:07 segfault. Feb 20 16:03:07 5. I traced that to it trying to access icons in params->gui, which is Feb 20 16:03:07 not initialized for text ui." Feb 20 16:03:49 seems fixed with current code, but I always have framebuffer Feb 20 16:05:27 and the TCIFLUSH stuff has been committed Feb 20 16:07:45 so maybe you could add the check for missing framebuffer and fall back to the textui Feb 20 16:07:57 then pfalcon would be happy Feb 20 16:09:06 sure Feb 20 16:09:28 Jay7: ah. maybe with my split I did not include any icons code Feb 20 16:10:14 maybe you could rename in theme-gui.h and theme-textui.h Feb 20 16:10:33 the first and the second part of theme.h Feb 20 16:12:06 I'll check this soon Feb 20 16:12:11 ASAP even :) Feb 20 17:33:58 bbl Feb 20 23:00:59 bluelightning: I've almost done the split Feb 20 23:01:07 now, the issue is the kernel... Feb 20 23:01:10 ant__: the split ? Feb 20 23:01:27 yes, to meta-initramfs Feb 20 23:01:55 our JaMa insists commiting .git kernels which nobody appears to use Feb 20 23:02:09 I'd remove that alltoghether Feb 20 23:02:23 has only been tested on spitz, maybe Feb 20 23:06:06 well, he was keen on having it there only a few months ago Feb 20 23:09:18 now we have 3: 3.1, 3.2, git (pre 3.2) Feb 20 23:09:36 3.2 won't compile, we need new klibc (or patch the existent) Feb 20 23:10:07 is still too new ;) Feb 20 23:17:53 ok, I'll send patch for 3.1 only atm Feb 20 23:18:10 (this is a 'private' meta-handheld thing atm) Feb 20 23:19:30 ok Feb 20 23:39:43 testing now... Feb 20 23:46:01 seems bb can still finish parsing ;) Feb 20 23:47:11 ok, some minor cleanings and I'm done Feb 21 00:05:59 I'm working on this psplash thing btw Feb 21 00:06:24 implementing this is harder than I had imagined... Feb 21 00:08:12 oh, the old suxxer Feb 21 00:18:03 eek Feb 21 00:18:05 ERROR: No recipes available for: Feb 21 00:18:05 /oe/meta-handheld/recipes-core/udev/udev_173.bbappend Feb 21 00:18:14 the last meta-oe hook! Feb 21 00:23:58 ok, Result: 250 <4F05A6650581AD7F> Mail accepted Feb 21 00:27:10 hm.. Feb 21 00:27:55 I'v moved a dir in the wrong layer Feb 21 00:28:28 /linux-kexecboot-3.1 Feb 21 00:28:42 there are defconfigs there Feb 21 00:40:18 bluelightning: well, if I move the a.m. dir from meta-initramfs to meta-handhelds Feb 21 00:40:30 | ERROR: Function failed: Fetcher failure for URL: 'file://defconfig'. Unable to fetch URL from any source. Feb 21 00:41:01 ant__: a.m.? Feb 21 00:41:11 oh right Feb 21 00:41:13 above mentioned Feb 21 00:41:33 maybe FILESPATH_append = ":${FILE_DIRNAME}/linux-${PV}" Feb 21 00:41:55 ah right... you should do that using FILESEXTRAPATHS_prepend := now Feb 21 00:42:10 doh Feb 21 00:42:12 FILESEXTRAPATHS cannot be used, because it puts extra paths before FILESPATH and we need it at the end Feb 21 00:42:12 # only as fallback for patches available only in linux-PV dir and ie defconfigs needs to go from linux-kexecboot-PV Feb 21 00:42:12 FILESPATH_append = ":${FILE_DIRNAME}/linux-${PV}" Feb 21 00:42:36 smthg wrong here Feb 21 00:43:06 it happens to work in meta-handheld Feb 21 00:47:21 bluelightning: ah, wait, is my bad probably Feb 21 00:47:47 in the linux-kexecboot_3.1.bbappend I set ILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:" Feb 21 00:56:54 ok, too late now Feb 21 00:56:56 gn Feb 21 00:58:34 nite **** ENDING LOGGING AT Tue Feb 21 02:59:58 2012 **** BEGIN LOGGING AT Tue Feb 21 02:59:58 2012 Feb 21 13:50:22 bluelightning: how to proceed wrt meta-initramfs maintainance? Koen seems to decline Feb 21 13:50:52 ant_work: I think you may be volunteering ;) Feb 21 13:51:09 ant_work: in which case you would need commit access Feb 21 13:51:24 the alternative is to host it elsewhere Feb 21 13:51:26 yes, and only temporarlyω Feb 21 13:51:51 I mean, if we get more attention this could maybe land in oe-core Feb 21 13:52:23 then I'd say dvhart or you could be maintainers of the layer Feb 21 13:53:49 there are som eother initramfs related pieces in /recipes-core/initrdscripts i.e. initramfs-framework Feb 21 13:54:23 all different but belonging to the same family Feb 21 14:03:57 bluelightning: I don't expect much ativity in that layer, only some version updates and evtl bugfixes Feb 21 14:04:58 the layer itself depends on oe-core, not on meta-oe, so is easy to test/debug Feb 21 14:14:22 hmm, ok... I will ping dvhart about this Feb 21 14:14:27 thanks :) Feb 21 14:25:18 yw Feb 21 14:25:50 I'd say, now I'll fix the patch so that Koen can commit it initially in meta-openembedded Feb 21 14:54:29 bluelightning: do you already have write access on meta-openembedded? Feb 21 14:54:45 ant_work: no, not as far as I am aware Feb 21 14:55:08 maybe Khem has? Feb 21 17:18:35 bluelightning: I've spotted a CMDLINE_DEBUG ?= "${@base_conditional('DISTRO_TYPE', 'release', 'quiet', 'debug',d)}" Feb 21 17:19:19 heh, is just 3 lines above the Feb 21 17:19:20 +KEXECBOOT_LABEL ?= "${@d.getVar('DISTRO', True) or d.getVar('DISTRO_VERSION', True)}-${MACHINE}" Feb 21 17:19:24 ant_work: hmm Feb 21 17:19:29 which we discussed not long ago Feb 21 17:19:40 ant_work: I think we should file a bug to stop using IMAGE_FEATURES in OE-Core for this Feb 21 17:24:06 well, here it happens in our very own kexecboot-cfg_0.1.bb Feb 21 17:25:13 maybe/surely better to fix it before committing to meta-initramfs Feb 21 17:25:21 ok, I'll send a V2 Feb 21 17:25:39 ...in a few hours Feb 21 17:25:40 bbl Feb 21 20:24:27 [Zaurus-devel] [PATCH]Quick&dirty fix that enables offline charging. Feb 21 21:53:39 hi bluelightning Feb 21 21:53:47 hi ant__ Feb 21 21:53:52 I've only a last issue with LOGO_SIZE Feb 21 21:54:03 maybe I take that back to meta-handheld Feb 21 21:54:25 | ERROR: Function failed: Fetcher failure for URL: 'file://./logo_linux_clut224.ppm.bz2'. Unable to fetch URL from any source. Feb 21 22:02:27 in meta-handheld we don't add any reference to LOGO_SIZE except for certain machines for which it should be defined Feb 21 22:02:31 what's going on there? Feb 21 22:02:52 SRC_URI += "file://defconfig \ Feb 21 22:02:52 file://${LOGO_SIZE}/logo_linux_clut224.ppm.bz2 \ Feb 21 22:02:52 " Feb 21 22:03:10 I tried unspecifying LOGO_SIZE Feb 21 22:03:27 then it defaults to . Feb 21 22:04:01 if I add the two files in meta-handheld I think we insulate the problem Feb 21 22:20:18 ok, here we are Feb 21 22:21:24 well, this starts to be annoying... ERROR: No recipes available for: Feb 21 22:21:24 /oe/meta-handheld/recipes-core/udev/udev_173.bbappend Feb 21 22:21:46 rm -rf ./recipes-core Feb 21 22:25:34 gosh Feb 21 22:25:53 if I remove the logos from meta-initramfs those are not found... Feb 21 22:32:22 bluelightning: bingo! Feb 21 22:32:23 FILESEXTRAPATHS_prepend := "${THISDIR}/linux-kexecboot-${PV}:${THISDIR}/files:" Feb 21 22:33:04 it's abit counterintuitive that local ./files is ignored Feb 21 22:33:18 but it's like that :p Feb 21 22:33:20 that shouldn't prevent that... Feb 21 22:34:05 I'll send v3 of the patch 3/3 **** ENDING LOGGING AT Wed Feb 22 02:59:58 2012 **** BEGIN LOGGING AT Wed Feb 22 02:59:59 2012 Feb 22 09:05:33 bluelightning: btw, have opie any instant messenger? Feb 22 09:05:53 Jay7: there's qpe-gaim but it's been unmaintained for a while Feb 22 09:06:18 Jay7: I have an item on my todo list to resurrect it and bring it up to the latest libpurple, but haven't got around to it yet Feb 22 09:06:37 ok, thanks :) Feb 22 09:34:56 bluelightning: Koen did already commit the new layer Feb 22 09:35:27 please commit the remaining 4 patches for meta-handheld http://patches.openembedded.org/project/oe/list/?submitter=193&state=* Feb 22 09:39:46 ah, and then we'll have to fix kexecboot-cfg Feb 22 09:40:15 (I realize now SRC_URI = "file://icon.xpm" is bad, the icon is in meta-handheld) Feb 22 09:41:39 I don't know, maybe we should commit the standard OE icon.xpm and qvga/vga bootlogos in meta-initramfs Feb 22 09:42:22 I'll have to see how you did it for psplash Feb 22 09:42:46 so we have now our own layer? :) Feb 22 09:43:01 yes Feb 22 09:43:11 http://cgit.openembedded.org/meta-openembedded/tree/meta-initramfs Feb 22 09:43:25 (I don't know who has write acess yet ;) Feb 22 09:43:49 I hope bluelightning will self-candidate for that Feb 22 09:49:57 well.. should go now Feb 22 09:53:09 bbl Feb 22 10:01:19 back Feb 22 10:04:23 bluelightning: yes, basically I forgot to add the icon.xpm to meta-initramfs. Icon is already removed from meta-handheld. Feb 22 10:10:25 ant_work: ok cool, will sort out meta-handheld today Feb 22 10:10:41 you can quietly go with the 4 patches, I'll send a new one for meta-initramfs later today Feb 22 10:11:13 there is still that DISTRO_TYPE... Feb 22 10:22:13 ok, I'll default to CMDLINE_DEBUG ?= "quiet" Feb 22 14:00:31 bluelightning: note that having added the .bbappend for kexecboot now meta-handheld depends on meta-initramfs Feb 22 14:01:15 ant_work: right, I will be adding that to the readme Feb 22 14:02:01 it's the same issue wrt udev/meta-oe...there must be a better way to do that! Feb 22 14:03:36 the .bbappend is like a tentacle from an octopus (the layer) :p Feb 22 14:04:43 finally, you'd have to add the new layer to layer-index Feb 22 14:04:52 then I'll let you in peace, promised Feb 22 14:04:56 heh Feb 22 14:04:57 :) Feb 22 14:04:59 no problem :) Feb 22 22:30:54 Jay7: about size Feb 22 22:30:55 β”‚ zImage.bin.both β”‚1008788β”‚ Feb 22 22:30:56 β”‚ zImage.bin.gui β”‚1008392β”‚ Feb 22 22:30:56 β”‚ zImage.bin.tui β”‚ 997156β”‚ Feb 22 22:31:48 and (gui only) zImage-kexecboot-3.2-r2.2-poodle.bin β”‚1023072 Feb 22 22:33:06 hehe.. I see no point to have gui-only then Feb 22 22:33:12 just few bytes Feb 22 22:33:57 sure Feb 22 22:34:20 3.2 is abit larger, though Feb 22 22:36:40 ok, it may be good idea to enable tui always by default Feb 22 22:36:57 (after fixing gui+tui mode) Feb 22 22:37:19 well, time to sleep Feb 22 22:37:23 good night Feb 22 22:37:26 gn Feb 22 23:07:38 bluelightning: we really need to make ipaqs work with oe-core udev rules... Feb 22 23:08:04 ant__: maybe we can get that udev rule accepted upstream Feb 22 23:08:15 this is the only tenctacle of meta-oe :) Feb 22 23:09:00 http://paste.debian.net/157253/ Feb 22 23:09:55 ^ those are of oe-core and meta-initramfs Feb 22 23:11:50 I've tested boot of 3.2.7 on poodle, all ok Feb 22 23:13:56 ah, great :) Feb 22 23:18:34 btw am sorting out meta-handheld now with the meta-initramfs move Feb 22 23:18:46 I want to try to reproduce the parsing issue first though Feb 22 23:19:18 I really don't see that Feb 22 23:20:12 but I removed udev from meta-handheld FWIW Feb 22 23:20:25 this is the only change Feb 22 23:20:42 well, thus I don't depend on meta-oe Feb 22 23:26:12 well, interesting, now I can use console=tty0 Feb 22 23:26:17 ah, got the error :/ Feb 22 23:26:30 somehow Zaurus has tty1 since long timw Feb 22 23:26:36 ah, good Feb 22 23:26:55 if I have meta-handheld (without your recent patches) and not meta-initramfs Feb 22 23:27:03 and with meta-oe Feb 22 23:29:23 must be the latter then Feb 22 23:30:13 actual meta-handheld (2012-02-07) has no knowledge of meta-initramfs Feb 22 23:30:14 maybe, the thing that bothers me most is there's no proper error message so there's no clue as to what has gone wrong :( Feb 22 23:31:14 that's bad Feb 22 23:33:28 btw I've an 'older' bitbake maybe db689a9 Feb 22 23:34:17 I've pulled now Feb 22 23:36:31 anything strange Feb 22 23:37:25 if you really want a weird, verbose, bb error, just place oe-core as last layer Feb 22 23:37:31 :) Feb 22 23:37:52 but is unrelated to the new issue Feb 23 00:06:08 bluelightning: ah, maybe is klibc.bbclass Feb 23 00:06:30 ant__: hmm, will have to check that out Feb 23 00:06:31 has been already removed from meta-oe Feb 23 00:06:34 meta-initramfs: import klibc recipe and bbclass from meta-oe Feb 23 00:06:35 ant__: yes, that is the cause Feb 23 00:06:51 it lives now in meta-initramfs Feb 23 00:06:58 the root cause of the error at least, not how we fail to handle it properly :) Feb 23 00:07:16 right, I'll merge those patches now having found the bitbake issue Feb 23 00:07:55 so just removing the class is enough to reproduce it? Feb 23 00:08:21 or maybe the klibc 'providers' Feb 23 00:08:48 yeah, it's easy to reproduce Feb 23 00:09:04 any ParserError will trigger it Feb 23 00:09:12 nice, we have opened new perspectives :p Feb 23 00:25:17 well, not really... it's a regression introduced at the beginning of this month Feb 23 00:25:24 but still, nice to catch it Feb 23 00:28:49 ok, two last patches as bonus ;) Feb 23 00:32:37 oh, had you read that? http://www.mail-archive.com/angstrom-distro-devel@linuxtogo.org/msg05496.html Feb 23 00:34:23 ant__: he submitted a fix, already merged Feb 23 00:34:25 :) Feb 23 00:34:51 nice, I had an older bitbake checkout and did not suffer Feb 23 00:34:57 luck Feb 23 00:35:40 btw the patch you just sent doesn't explain why the tty1->tty0 change is being made... Feb 23 00:36:16 ah, right. tty0 is the standard Feb 23 00:36:29 somehow Zaurus had custom value Feb 23 00:36:41 I've tested on c7x0 and poodle fwiw Feb 23 00:36:57 spitz just built, testing it as well Feb 23 00:37:33 bluelightning: but you can let those 2 patches wait Feb 23 00:37:47 let JaMA and others comment, pls Feb 23 00:39:06 ok Feb 23 00:39:25 btw the removal patches was just up to v2 right? Feb 23 00:39:28 were Feb 23 00:40:32 no, v3 :/ Feb 23 00:40:50 ah, hm Feb 23 00:40:56 * bluelightning is lost in his emails Feb 23 00:41:00 I archived the superseeded Feb 23 00:41:14 yeah maybe I should just use patchwork eh? :) Feb 23 00:41:26 http://patches.openembedded.org/project/oe/list/?submitter=193 Feb 23 00:41:32 yes, 7 + 2 Feb 23 00:43:46 ant__: so, apart from the 2 most recent I need to apply all currently listed in patchwork, is that correct? Feb 23 00:44:45 yes, please Feb 23 00:48:35 well, not the 2 for meta-initramfs ;) Feb 23 00:52:21 atm I don't have push access for meta-initramfs, so I couldn't really... Feb 23 00:54:50 ok, spitz boots as well Feb 23 00:57:04 heh, I have 3 sato with broken touchscreen calibration. we need xinput-calibrator and pointercal in oe-core Feb 23 00:57:19 3 = c7x0, spitz, poodle Feb 23 00:59:33 ok, done Feb 23 00:59:41 also updated README Feb 23 01:00:06 great, thx Feb 23 01:00:20 wow, is very late now Feb 23 01:00:22 thank you :) Feb 23 01:00:27 yes :( Feb 23 01:00:28 have a good night! Feb 23 01:00:31 and you **** ENDING LOGGING AT Thu Feb 23 02:59:58 2012 **** BEGIN LOGGING AT Thu Feb 23 02:59:58 2012 Feb 23 08:55:59 * Jay7 looked over old bookmarks about zauruses Feb 23 08:56:09 lot of them are pretty dead.. Feb 23 09:02:21 yes :( Feb 23 09:36:29 we would need to flush google and part of the web Feb 23 09:37:23 but at least we could update/remove the pages in linuxtogo Feb 23 09:37:50 btw I plan to add a README.zaurus and put low-level info there Feb 23 09:39:11 bluelightning: I don't know how to respond to JaMa wrt the many kernels and the .inc he added Feb 23 09:51:12 eh, done Feb 23 09:53:00 bluelightning: in which layer-index category would you classify meta-initramfs? Feb 23 09:53:29 base or misc? Feb 23 10:30:12 ant_work: not sure... I think maybe it still qualifies as a software layer Feb 23 10:30:35 ant_work: what is current state of ubifs in klibc? Feb 23 10:30:48 afaik stalled Feb 23 10:31:06 maks said to look at util-linux-ng, though Feb 23 10:38:58 https://gitorious.org/util-linux-ng/util-linux-ng/blobs/master/include/fsprobe.h Feb 23 10:39:57 .. heh.. #include Feb 23 10:42:37 better to look at u-boot imho Feb 23 10:45:50 well, seems all is there: Feb 23 10:45:51 https://gitorious.org/util-linux-ng/util-linux-ng/trees/master/libblkid/src/superblocks Feb 23 10:55:45 Jay7: the definition of superblocks differ between klibc and libblkid Feb 23 10:55:54 I see.. Feb 23 10:55:59 we just need to get the magi I think Feb 23 10:56:06 then, they use uuid's too Feb 23 10:56:31 uuids are good for root=UUID=blablabla Feb 23 10:56:42 ok, but we don't have to mount anything Feb 23 10:56:51 but it's mostly unused on our devices :) Feb 23 12:57:12 bluelightning: [Angstrom-devel] Koen Kooi : layerman: disable opie layer, it breaks qt4 builds Feb 23 12:57:24 Jay7: yes I know, he emailed me Feb 23 12:57:30 Jay7: not sure what's causing it :( Feb 23 12:57:45 Dmitry made a number of changes to meta-opie to avoid interfering with qt4 Feb 23 13:53:23 bluelightning: in case you've missed it http://blogs.coreboot.org/blog/2011/08/24/gsoc-spice-payload-report/ Feb 23 13:54:00 he slaps with "The docs(Getting started wiki page for example) are out dated and you get conducted by the old code base, and trust me, it`s not a good way to get started." Feb 23 13:54:10 and we all agree :/ Feb 23 13:54:20 yes Feb 23 13:54:28 interesting to read though Feb 23 15:10:41 bluelightning: you'll have to help me wrt SIGGEN_EXCLUDERECIPES_ABISAFE ;) Feb 23 15:11:03 once we'll add pointercal Feb 23 15:11:44 add pointercal-xinput to SIGGEN_EXCLUDERECIPES_ABISAFE, even Feb 23 15:12:25 I don't know anything about it yet... Feb 23 15:12:36 heh Feb 23 15:12:52 some new RP hack ;) Feb 24 00:06:21 bluelightning: we are not so far from linux-yocto Feb 24 00:06:49 this would drastically reduce layer size Feb 24 00:07:20 thus maintenance Feb 24 00:10:13 yes I see you have been talking to darren, great stuff :) Feb 24 00:14:21 hm Feb 24 00:14:24 http://pastebin.com/84uXYqza Feb 24 00:14:30 the meta-opie error Feb 24 00:18:58 seems to be Feb 24 00:19:00 PV = "${OPIE_GIT_PV}" Feb 24 00:19:32 ${OPIE_GIT_PV} unexpanded, is now undeclared Feb 24 00:19:50 you did remove a class, isn't? Feb 24 00:21:00 hm.. OPIE_GIT_PV = "${OPIE_VERSION}" Feb 24 00:21:40 seems include/preferred-opie-git-versions.inc is ignored/not included Feb 24 00:28:16 did you include that manually? Feb 24 00:28:20 you need to do so Feb 24 00:29:07 I personally included the 1.25 when building opie Feb 24 00:29:37 ah, hmm Feb 24 00:29:56 perhaps that should set OPIE_GIT_PV to avoid parse errors Feb 24 00:29:58 now is commented and I don't have the layer atm Feb 24 00:30:35 do you suggest 1.25 or .git? Feb 24 00:39:16 -rw-r--r-- 1 andrea users 1570288 Feb 24 01:37 zImage-3.0.12-yocto-standard+ Feb 24 00:41:11 1,5 MB, not bad for starting Feb 24 00:41:46 (this is not linux-kexecboot, I still have to see how to inject the cpio there) Feb 24 00:42:39 gn **** ENDING LOGGING AT Fri Feb 24 02:59:58 2012 **** BEGIN LOGGING AT Fri Feb 24 02:59:58 2012 Feb 24 12:35:38 bluelightning: do you know dvhart timezone? **** ENDING LOGGING AT Sat Feb 25 02:59:59 2012 **** BEGIN LOGGING AT Sat Feb 25 02:59:59 2012 Feb 25 15:37:47 bluelightning: ERROR: Nothing RPROVIDES 'ttf-dejavu-sans' (but /oe/meta-opie/recipes-opie/tasks/task-opie.bb RDEPENDS on or otherwise requires it) Feb 25 16:07:43 ant__: yes, it's in meta-oe Feb 25 16:07:54 meta-opie requires meta-oe, that can't be helped Feb 25 16:09:46 evening Feb 25 16:13:11 hi guys **** ENDING LOGGING AT Sun Feb 26 02:59:59 2012 **** BEGIN LOGGING AT Sun Feb 26 02:59:59 2012 **** ENDING LOGGING AT Mon Feb 27 02:59:58 2012 **** BEGIN LOGGING AT Mon Feb 27 02:59:58 2012 Feb 27 09:51:47 hi bluelightning Feb 27 09:51:55 hi ant_work Feb 27 12:56:49 bluelightning: have you ever gived a try to qt/mame? http://qmc2.arcadehits.net/wordpress/download/ Feb 27 13:07:27 ant_work: I haven't no... arcades were never really my kind of thing Feb 27 13:09:52 neither but I have a customer for this (my young son) Feb 27 13:10:27 ah... the most demanding customers of all :) Feb 27 13:10:41 yes, iphone is not enough... Feb 27 13:10:59 lacks all that nice little keys Feb 27 13:11:12 and th epen! Feb 27 13:11:21 heh Feb 27 13:11:45 I'll install gcc for him, you never know ;) Feb 27 13:12:26 heh :) **** ENDING LOGGING AT Tue Feb 28 02:59:59 2012 **** BEGIN LOGGING AT Tue Feb 28 02:59:59 2012 Feb 28 10:58:41 hello bluelightning Feb 28 10:58:54 hi ant_work Feb 28 10:59:07 good news: core-image-sato boots again (w/out meta-oe) Feb 28 10:59:30 ah good, do you know what changed? Feb 28 11:33:45 no, must have been be some temporary issue fixed afterwards Feb 28 11:34:01 still no touchscreen calibration Feb 28 11:53:02 ah, pls remember to add meta-initramfs in LayerIndex Feb 28 11:53:57 whoops I completely forgot about that Feb 28 11:54:02 will do that now, thanks for the reminder Feb 28 11:56:51 done Feb 28 11:57:15 great, thx Feb 28 11:59:36 I'll have to ask dvhart whether he's updting linux-yocto-tiny to 3.2.7 Feb 28 12:00:26 then we would match with standard kernel and I could seriously analyze the remaining kernel .config diffs Feb 28 14:37:19 wow, klibc_2.0 expected in 2 weeks Feb 28 21:55:33 ant__: hi Feb 28 21:55:44 how 'buildable' is oe today? :) Feb 28 21:56:05 hi Yuri Feb 28 21:56:09 not too bad :) Feb 28 21:56:52 I'll fix klibc QA this evening Feb 28 21:57:06 maybe I'll add the patch to compile with kernel 3.2 Feb 28 21:57:29 but I'll prefer to wait until mid march (announced release date on irc) Feb 28 21:58:23 btw there is some mips work in #kexec Feb 28 22:00:36 hm, good Feb 28 22:01:12 btw, is initramfs layer already pushed? Feb 28 22:01:20 yes Feb 28 22:01:33 I should add it then Feb 28 22:02:20 ant__: what is url for clone initramfs layer? Feb 28 22:02:21 I've built with this Feb 28 22:02:23 http://paste.debian.net/157966/ Feb 28 22:02:56 ah.. is it part of meta-oe? Feb 28 22:03:06 but you have to remove udev from meta-handheld to do this Feb 28 22:03:10 yes Feb 28 22:03:19 well, part of meta-openembedded Feb 28 22:03:36 like meta-oe Feb 28 22:04:46 Jay7: if you add opie you have/had to add in your local.conf Feb 28 22:04:48 require conf/distro/include/preferred-opie-versions-1.2.5.inc Feb 28 22:04:48 #require conf/distro/include/preferred-opie-git-versions.inc Feb 28 22:09:23 seems it easier to use one BBLAYER += ... per layer instead of BBLAYER = "..." with multiple layers Feb 28 22:09:27 easier to manage Feb 28 22:10:14 ah, there is a problem with unquoted line in opie Feb 28 22:10:16 ERROR: ParseError at /oe/meta-opie/recipes-qtopia/scummvm/scummvm_1.1.1.bb:18: unparsed line: 'INSANE_SKIP_${PN} = True' Feb 28 22:10:49 this check is nice but almost all layers need fixes Feb 28 22:11:48 well.. I'll start with some console version :) Feb 28 22:11:57 btw, which image to build for this? Feb 28 22:12:03 Jay7: well, there is a wrapper around bitbake, which swallows all layers at the beginning Feb 28 22:12:08 something with nandlogical :) Feb 28 22:12:12 maybe that could work Feb 28 22:12:48 about images, I've built core-image-sato and opie-image Feb 28 22:13:03 core-image-minimal lacks kernel in /boot & modules Feb 28 22:13:25 wait, kernel is fixed in zaurus.inc, modules not iirc Feb 28 22:13:47 bluelightning is working on that Feb 28 22:13:54 ok, thanks for info Feb 28 22:14:12 one man on hpc.ru erased flash on akita Feb 28 22:14:27 but I see no recovery image for akita.. Feb 28 22:14:39 there are some on poketworkstation.org iirc Feb 28 22:14:44 but for older models Feb 28 22:15:09 I think you can dump with nandlogical Feb 28 22:15:22 http://pocketworkstation.org/files/recover/ Feb 28 22:15:55 we should create recovery images for newer models Feb 28 22:21:02 well.. too late here Feb 28 22:21:08 * Jay7 -> sleep Feb 28 22:21:22 gn **** ENDING LOGGING AT Wed Feb 29 02:59:58 2012 **** BEGIN LOGGING AT Wed Feb 29 02:59:58 2012 Feb 29 08:58:04 alexxy: I was reading in gentoo-dev about the move to /usr Feb 29 08:58:57 what is the problem with initramfs? One could even imagine to embed one in kernel and add a second from bootloader config (with /usr) Feb 29 09:00:50 I can imagine people's fear: like 10yrs ago when to boot scsi/ataraid one had to build an initrd Feb 29 09:31:07 ant_work: there is no problems Feb 29 09:31:14 i have /usr on lvm2 Feb 29 09:31:18 and it works fine Feb 29 09:31:19 =D Feb 29 09:31:28 but new udev moved to /usr Feb 29 09:31:46 so you need initramfs capable to mount /usr before pivot_root Feb 29 09:53:05 this is not trivial, you may want a monolitic kernel or you'll have to play with modules Feb 29 09:54:04 I would do that in C with klibc in an embedded cpio Feb 29 09:55:14 Jay7: a kexecboot fork ;) Feb 29 09:55:55 ant_work: hehe Feb 29 09:56:23 * Jay7 moved some time ago / into /usr and remounted it as / Feb 29 09:56:43 because archlinux have issues with separate /usr and initramfs Feb 29 10:00:45 with gentoo I just left a big / http://pastebin.com/4VAEJysV Feb 29 10:03:32 /dev/mapper/mirror-root 16G 8.3G 7.0G 55% / Feb 29 10:03:34 :) Feb 29 10:06:29 heh, gentoo LAMP server only needs 3.3 GB for root+usr (900M are sources) Feb 29 13:00:36 bluelightning: have you found missing quotes in meta-handheld? Feb 29 13:00:50 I had no errors Feb 29 13:01:07 ant_work: no, didn't see any Feb 29 13:07:55 bluelightning: I'll send you the defconfigs for 3.2 this night. Pls ask dvhart if he plans to commit linux-yocto-tiny_3.2 soon Feb 29 13:15:33 ant_work: he's currently preparing for an internal conference so it might not happen until after next week Feb 29 13:15:47 ah, I see. Feb 29 13:16:25 unfortunately Yocto misses 3.1 kernel, for which we already have ready defconfigs .... Feb 29 13:17:12 so I'll jump on 3.2 wagon. 3.0 seems dead rail Feb 29 13:36:43 oh.wait, what is that? http://git.yoctoproject.org/cgit/cgit.cgi/poky-extras/tree/meta-kernel-dev/recipes-kernel/linux/linux-korg.bb Feb 29 13:37:39 "yoctized" Feb 29 15:30:26 evening Feb 29 15:34:50 hi Feb 29 21:06:54 hi bluelightning Feb 29 21:07:20 would you pls commit the 3.27 -> 3.2.8 patch? Feb 29 21:07:49 I've launched a build right now and will work on defconfigs later Feb 29 21:16:53 hi ant__ Feb 29 21:17:01 sure will look at it now Feb 29 21:17:13 seems I've lured zeddii ;) Feb 29 21:20:11 done Feb 29 21:21:56 cool, good to see some collaboration :) Feb 29 21:22:12 Iheh Feb 29 21:22:42 I have read some threads and afaik this is indeed a bit mysterious Feb 29 21:23:00 dvhart is trying to understand it better as well Feb 29 21:23:18 I happen to have had luck maybe ;) Feb 29 21:24:25 ah, about that tty0 - tty1 nobody remembers why that change Feb 29 21:24:34 I'll test on more Zaurus Feb 29 21:24:42 poodle is ok Mar 01 00:00:53 ok, configs almost done Mar 01 00:00:55 gn **** ENDING LOGGING AT Thu Mar 01 02:59:59 2012 **** BEGIN LOGGING AT Thu Mar 01 02:59:59 2012 Mar 01 22:39:26 bluelightning: hello Mar 01 22:39:36 hi ant__ Mar 01 22:40:09 did we already ask RP to remove zaurusd from oe-core? Mar 01 22:40:18 ant__: I did discuss it with him Mar 01 22:40:32 ok then Mar 01 22:40:41 ant__: he has an idea that it will be turned into something more generic Mar 01 22:41:35 but I think we will still be able to remove it Mar 01 22:41:43 we're planning a round of cleanup in oe-core and meta-oe fairly soon, I think we will remove it as part of that Mar 01 22:44:52 about kernels, we'll move to 3.2 soon. iirc we have the 1940 to check Mar 01 22:46:10 ah yes Mar 01 22:46:23 well, there are lot of patches coming with 3.3 and 3.4 for pxa and sha1100 Mar 01 22:46:42 unfortunately I'm going to a conference next week so I won't be able to do much on this until after that Mar 01 22:47:20 I still wait for the klibc_2 release, now promised mid mai Mar 01 22:47:28 sry, mid march Mar 01 22:47:43 if you want I can ask hpa about it next week in person ;) Mar 01 22:47:52 heh Mar 01 22:47:55 well, at least I assume he is going also Mar 01 22:48:15 he or maks are supposed to start the release testings Mar 01 23:35:32 bluelightning: curiosity about lcd's http://permalink.gmane.org/gmane.comp.handhelds.openembedded.scm/19527 Mar 01 23:35:42 zaurus + ipaqs Mar 01 23:35:57 looks like kdrive stuff Mar 01 23:50:01 hmm, not sure Mar 01 23:51:03 well, we dumped kdrive because it lost calibration, now we have a semi-broken xorg and kdrive got tslib back :/ Mar 01 23:51:42 memory-wise xorg is not much hungry Mar 01 23:51:53 I think we should stay there Mar 02 00:27:21 hmm, I haven't really used either on our devices Mar 02 00:27:29 * bluelightning -> bed Mar 02 00:27:31 night **** ENDING LOGGING AT Fri Mar 02 02:59:58 2012 **** BEGIN LOGGING AT Fri Mar 02 02:59:58 2012 Mar 02 21:21:12 ouch.. collie 3.2.8 http://pastebin.com/h5fMxEZf Mar 02 21:54:07 hm.. Mar 02 21:54:48 there is wip about gpio, merge for next Mar 02 22:21:29 Jay7: "Linux 3.3-rc5 and pxa-gpio possible regression" Mar 02 22:24:31 ah, I have even marked it: "[PATCH] ARM: sa1100: collie: Remove irq_to_gpio from power driver" Mar 02 22:27:00 well, I'll add the patch... Mar 02 22:30:19 nice Mar 02 22:30:35 http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Fstable%2Flinux-stable.git&a=search&h=HEAD&st=commit&s=collie Mar 02 22:30:40 I've sent first results of oe-core PARALLEL_MAKE testing Mar 02 22:54:35 great Mar 02 22:54:50 Jay7: do you have time to flash collie with those 3 patches? Mar 02 22:54:58 just linux-kexecboot I mean Mar 02 22:55:59 ant__: sorry, I'm going to bed now Mar 02 22:56:08 bother me tomorrow Mar 02 22:56:27 nyes, I'll send you that Mar 02 22:56:33 np Mar 02 22:57:27 good night :) Mar 02 22:57:35 gn Mar 02 22:58:11 got your email Mar 02 22:58:43 ant__: do kexecboot there have GUI+TUI patch? Mar 02 22:59:02 no, is just gui Mar 02 22:59:42 ok.. just remember that GUI was never worked on collie after refactoring.. Mar 02 22:59:51 will check it out again Mar 02 23:22:03 nice, core-image-base is a minimal console img (12M tar.gz), not bloated like core-image-minimal **** ENDING LOGGING AT Sat Mar 03 02:59:59 2012 **** BEGIN LOGGING AT Sat Mar 03 02:59:59 2012 Mar 03 22:08:01 Jay7: now I remember why we did want tty1...to avoid the the blinking cursor on tty0 Mar 03 22:08:33 shouldn't happens Mar 03 22:08:44 it does :) Mar 03 22:09:02 on both poodle and c7x0 Mar 03 22:09:03 kexecboot looking at console=ttyvX cmdline tag for tty to disable cursor on Mar 03 22:09:20 s/ttyvX/ttyX/ Mar 03 22:09:43 I remember that with console on tty0 we had some hangs Mar 03 22:09:54 tty0 is not virtual console, maybe because of that Mar 03 22:10:31 problem is that I can't find reliable way to get tty under fb0.. Mar 03 22:10:32 anyway, it does work apart the blinking cursor Mar 03 22:10:50 but we may hardcode it or use configure knob Mar 03 23:15:16 btw in the core-images 'loadkeys /etc/keymap-2.6.map' is not run Mar 03 23:17:31 heh Mar 03 23:17:34 KERNEL_MAJMIN=`uname -r | cut -d '.' -f 1,2` Mar 03 23:17:34 if [ -e /etc/keymap-$KERNEL_MAJMIN.map ]; then Mar 03 23:17:34 loadkeys /etc/keymap-$KERNEL_MAJMIN.map Mar 03 23:17:34 fi Mar 03 23:17:55 kernel 3.2 here :) Mar 03 23:18:07 seems a bug Mar 03 23:18:27 I'll have to inform luelightning Mar 03 23:18:31 +b **** ENDING LOGGING AT Sun Mar 04 02:59:58 2012 **** BEGIN LOGGING AT Sun Mar 04 02:59:59 2012 Mar 04 08:26:59 hm.. Mar 04 08:27:22 core-image-base building systemd.. Mar 04 08:27:33 meta-oe is pure poison.. **** ENDING LOGGING AT Mon Mar 05 02:59:58 2012 **** BEGIN LOGGING AT Mon Mar 05 02:59:58 2012 **** ENDING LOGGING AT Tue Mar 06 02:59:58 2012 **** BEGIN LOGGING AT Tue Mar 06 02:59:58 2012 Mar 06 18:48:35 ant: zImage you have sent to me doesn't boot like before Mar 06 18:48:44 but now with other splash (OE's one) Mar 06 18:49:11 we need another collie owner to test Mar 06 18:49:19 just to be sure that my collie is ok Mar 06 22:45:35 hi Jay7, bluelightning1 Mar 06 23:25:05 hi ant__ Mar 06 23:25:31 how's going at the conference? Mar 06 23:26:12 ant__: not too bad :) feeling a bit unwell though :/ Mar 06 23:26:38 ouch Mar 06 23:27:45 we had some 'resident evil' kind of bacteria at home for one month...kid, wife, me (slightly) Mar 06 23:28:25 now here is warm again and the suxxer get destroyed by temperature :p Mar 06 23:30:26 * ant__ is guessing Paul will go back home with some new tablet from that meeting Mar 06 23:38:33 no, unfortunately no gadgets... there's a conference bag etc. though :) Mar 06 23:41:12 eh, this week any presentation would be uncomfortably shadowed out by the ipad3/hd one Mar 06 23:56:04 btw pls look at the patches for meta-handheld on the list when you'll feel better Mar 06 23:56:42 forget the ones for tty0 atm, I've archived those Mar 06 23:59:59 really? isn't that broken without those? Mar 07 00:00:29 yes I saw the ones you just sent, will try to merge them soon Mar 07 00:29:45 you see, having console on tty0 in linux-kexecboot makes a blinking cursor appear Mar 07 00:30:57 no issues on linux-standard, apart more verbose console Mar 07 00:32:17 btw I'm hacking meta-hh about keymaps but that's not the way Mar 07 00:33:44 ok, too late now, zZZZzz Mar 07 00:33:46 gn **** ENDING LOGGING AT Wed Mar 07 02:59:58 2012 **** BEGIN LOGGING AT Wed Mar 07 02:59:58 2012 Mar 07 10:55:47 ant_work: zImage you have sent to me doesn't boot like before Mar 07 10:56:03 but now with nice OE splash :) Mar 07 10:59:45 heh Mar 07 11:00:32 there are many patches for sha1100/collie in 3.3 and 3.4, lets' hope this is fixed Mar 07 11:00:58 could be some framebuffer issue Mar 07 11:02:17 Jay7: if you apply a simple check in the code (i.e. existence of dev/fb0 or failing ioctl) I'll send you a kernel with gui+tui Mar 07 11:02:42 I'd say just check if fb exists Mar 07 11:02:54 sure Mar 07 11:03:08 we have some holidays again (3 days) Mar 07 11:03:17 this will not solve the 'broken fb driver' situation, though Mar 07 11:03:21 I'll try to implement such check Mar 07 11:03:51 yeah.. Mar 07 11:04:03 fb is initialized but not working Mar 07 11:04:11 heh, seems so Mar 07 11:04:16 or may be working but doing this wrong :) Mar 07 11:04:31 this time it was on tty0 fwiw Mar 07 11:05:37 maybe the defconfig is wrong...but iirc 2.6.38 was booting Mar 07 11:05:44 with fb? Mar 07 11:13:25 it was booting with old kexecboot Mar 07 11:13:32 btw, I should check this again **** ENDING LOGGING AT Thu Mar 08 02:59:58 2012 **** BEGIN LOGGING AT Thu Mar 08 02:59:59 2012 **** ENDING LOGGING AT Fri Mar 09 02:59:58 2012 **** BEGIN LOGGING AT Fri Mar 09 02:59:58 2012 **** ENDING LOGGING AT Sat Mar 10 02:59:58 2012 **** BEGIN LOGGING AT Sat Mar 10 02:59:58 2012 Mar 10 15:00:12 * Jay7 is at home again **** ENDING LOGGING AT Sun Mar 11 02:59:58 2012 **** BEGIN LOGGING AT Sun Mar 11 02:59:58 2012 Mar 11 22:25:34 bluelightning: hi, are you back? Mar 11 22:25:42 hi ant__ Mar 11 22:25:43 yes :) Mar 11 22:26:09 I have a nice log to show you, *unseen* before... Mar 11 22:26:12 wait a min Mar 11 22:27:51 http://paste.debian.net/159374/ Mar 11 22:29:00 Ignoring.pseudo: uh-oh! ? Mar 11 22:29:24 definitely smthg is broken wrt postinst Mar 11 22:34:48 bluelightning: well, so you'll see it in its ascii beauty Mar 11 22:34:53 http://imgbin.org/index.php?page=image&id=7103 Mar 11 22:40:12 well, on the first one I'm not entirely sure what has gone wrong there but it's nothing to do with opie... libnl2 is depended upon by connman Mar 11 22:40:49 not sure about the pseudo one, but it looks nasty Mar 11 22:44:17 heh.. as texture looks sinisterly like spaced invaders http://imgbin.org/index.php?page=image&id=7104 Mar 11 22:44:37 I thought I had been rootkit'ed Mar 11 22:46:41 there are otavio's commits in queue for opkg/rootfs/postinst let see Mar 11 22:49:40 ant__: are permissions blocking user read of /proc/*/exe ? Mar 11 22:50:25 andrea@mizar ~ $ cat /proc/3012/exe Mar 11 22:50:25 cat: /proc/3012/exe: Permission denied Mar 11 22:51:58 root can cat this file, which btw seems to be udev stuff Mar 11 22:57:25 bluelightning: I think some broken postinst is running on do_rootfs, not checking for $D Mar 11 22:58:03 what the point reading /proc? Mar 11 22:58:40 but seems to be pseudo, so... Mar 11 22:58:49 ant__: presumably pseudo needs to... is the process it's complaining about a root process or a user one? Mar 11 22:59:10 well, /proc is empty in image rootfs, kernel is not running... Mar 11 23:00:20 this has to be done as real postinst returning 1, marking as unpacked only, configuration required on target Mar 11 23:03:33 andrea@mizar /oe/oe-core/build/tmp-eglibc/work/poodle-oe-linux-gnueabi/opie-image-1.0-r0/rootfs/etc $ grep -R '/proc/1' . Mar 11 23:03:33 ./init.d/rcS:[ -d "/proc/1" ] || mount /proc Mar 11 23:03:33 ./init.d/bootlogd: if [ -d /proc/1/. ] Mar 11 23:03:33 ./init.d/stop-bootlogd: if [ -d /proc/1/. ] Mar 11 23:03:33 ./rcS.d/S07bootlogd: if [ -d /proc/1/. ] Mar 11 23:03:35 ./rc2.d/S99stop-bootlogd: if [ -d /proc/1/. ] Mar 11 23:03:37 ./rc3.d/S99stop-bootlogd: if [ -d /proc/1/. ] Mar 11 23:03:39 ./rc4.d/S99stop-bootlogd: if [ -d /proc/1/. ] Mar 11 23:03:41 ./rc5.d/S99stop-bootlogd: if [ -d /proc/1/. ] Mar 11 23:05:54 this is the only one checking for proc Mar 11 23:42:45 it's pseudo doing it itself and it's looking at your /proc, not the one in the rootfs Mar 11 23:42:53 ant: ^ Mar 11 23:46:39 I'll rebuild from scratch and see Mar 11 23:50:48 launched, gn **** ENDING LOGGING AT Mon Mar 12 02:59:58 2012 **** BEGIN LOGGING AT Mon Mar 12 02:59:58 2012 Mar 12 22:38:41 hi bluelightning Mar 12 22:38:50 hi ant__ Mar 12 22:40:30 bluelightning: we'd need to fix the keymaps, currently skipped Mar 12 23:25:18 I've revived the trhead Mar 12 23:29:10 gn **** ENDING LOGGING AT Tue Mar 13 02:59:58 2012 **** BEGIN LOGGING AT Tue Mar 13 02:59:58 2012 **** BEGIN LOGGING AT Tue Mar 13 09:48:19 2012 Mar 13 18:42:11 Linux Foundation Collaboration Summit April 3-5 San Francisco Mar 13 18:42:19 invitation-only.. Mar 13 21:25:10 bluelightning: do you have time to pull the floating patches? Mar 13 21:25:38 ant__: I'll try to do that this evening Mar 13 21:25:49 great, thx Mar 13 21:26:02 have you seen the RFC wrt keymaps? Mar 13 21:26:44 unfortunately I'm having difficulties with my email... latest kontact is crashing a lot :( Mar 13 21:27:14 heh, overwhelmed by patch flood ;) Mar 13 21:27:44 * Jay7 switched from kontact long time ago Mar 13 21:27:52 kmail2 is not working at all Mar 13 21:28:08 it was working OK for me in 4.8.0 Mar 13 21:28:11 forget KDE :) Mar 13 21:28:16 4.8.1 has introduced some issues Mar 13 21:28:21 no, I like KDE too much Mar 13 21:28:34 kmail can't deal with my gmail imap account :) Mar 13 21:28:36 recent KDE PIM has been a rough ride though :/ Mar 13 21:28:38 even akonadi, not kmail Mar 13 21:28:54 ok, rfc is http://lists.linuxtogo.org/pipermail/openembedded-core/2012-March/018633.html Mar 13 21:29:25 as hack I'm setting $KERNEL_MAJMIN = "2.6" in our layer... Mar 13 21:30:17 ant__: imho, we may just remove that check Mar 13 21:30:31 Jay7: how was done for ben nanonote? is the patch a full kernel-map? Mar 13 21:30:40 yes iirc Mar 13 21:30:55 ok, that's another way to do that then Mar 13 21:31:11 just non-upstreamed patch which change keymap to working Mar 13 21:31:24 I don't know why this kind of paches can't be upstreamed.. Mar 13 21:32:12 same for matrix-keyboard for corgi...lacking / Mar 13 21:32:37 we can't use loadable keymap with kexecboot Mar 13 21:32:45 well.. we will not :) Mar 13 21:33:01 so general solution is kernel patch Mar 13 21:33:06 surely I will not rely on udev for that... Mar 13 21:55:29 Jay7: this week maks, klibc dev, will discuss phD Mar 13 21:55:43 so next week we should have klibc_2 and sync th efstypes Mar 13 21:55:44 ant__: I hope he will complete :) Mar 13 21:55:53 he said after 2.0 :) Mar 13 21:56:02 ah :) Mar 13 21:56:19 is ubifs scheduled for 2.0? Mar 13 21:56:37 I was talking about ubifs :) Mar 13 21:56:42 2.01 maybe Mar 13 21:56:45 nice **** BEGIN LOGGING AT Tue Mar 13 22:04:38 2012 Mar 13 23:19:51 gn **** BEGIN LOGGING AT Wed Mar 14 00:54:06 2012 **** ENDING LOGGING AT Wed Mar 14 02:59:58 2012 **** BEGIN LOGGING AT Wed Mar 14 02:59:58 2012 **** ENDING LOGGING AT Thu Mar 15 02:59:58 2012 **** BEGIN LOGGING AT Thu Mar 15 02:59:58 2012 Mar 15 19:10:58 * Jay7 got real office job finally.. Mar 15 19:11:24 http://aboutecho.com <- here :) Mar 15 19:23:28 Jay7: congratulations :) Mar 15 19:54:39 bluelightning: thanks :) **** ENDING LOGGING AT Fri Mar 16 02:59:58 2012 **** BEGIN LOGGING AT Fri Mar 16 02:59:58 2012 Mar 16 12:56:04 bluelightning: I've seen meta-smartphone is still using *-2.6 keymaps. I think other layers don't have devices with kb so do not care Mar 16 12:59:05 quickly we can add our keymap.sh script without fixing the real issue in oe-core (maybe is not an issue, I got no feedback :/) Mar 16 13:14:57 I guess like you suggest most people aren't dealing with devices that have special keyboards Mar 16 16:40:27 bluelightning: urgh we have 3.2.11 now Mar 16 16:41:08 youμve played with linux-yocto-tiny recently, isn't? Mar 16 16:42:27 yes, I was just building it though Mar 16 16:42:32 that kernel even boots :) Mar 16 16:42:48 using our defconfigs as base, though Mar 16 16:43:51 I've read your talk with Bruce but did not fully understood it Mar 16 16:44:26 he said that starting with a defconfig it isn't worth to modularize with extra .cfg pieces Mar 16 16:44:35 i I get it correctly Mar 16 16:44:52 if even Mar 16 16:46:36 anyway it looks to me urgent to hook our recipes to linux-yocto(-tiny) and benefit of the upgrades done there Mar 16 16:47:15 I still have to try the linux-yocto-korg variant Mar 16 16:48:18 I agree, the config fragment thing is really useful Mar 16 16:48:33 but yeah, I think it's really either-or Mar 16 16:49:24 in fact there is some mangling done but the defconfig ends up pretty similar to the source Mar 16 16:49:54 but we start with 'heavy' defconfigs, with many fs compiled in kernel Mar 16 17:28:13 bbl Mar 16 23:28:40 bluelightning: would you accept this? Mar 16 23:28:42 http://paste.debian.net/160058/ Mar 16 23:30:20 possibly... it would be better if we could get the needed fix into oe-core, but if that's not going to happen soon... Mar 16 23:30:23 or we could add a sed in some task in the bbappend Mar 16 23:32:12 yes, we can e.g. add a do install_append Mar 16 23:32:48 it would be less invasive I think... Mar 16 23:33:03 let me try Mar 16 23:54:46 * ant__ is digging is sed regexp Mar 17 00:21:16 ok, trying to be elegant... Mar 17 00:53:25 ok, I'm back now Mar 17 00:53:42 bluelightning: this is better http://paste.debian.net/160066/ Mar 17 00:54:11 ant__: looks good Mar 17 00:54:13 :) Mar 17 00:54:39 heh, a beer helped Mar 17 01:07:08 well, maybe it is even better put the sed in do_intall_append Mar 17 01:14:20 ok, do as you think is better Mar 17 01:14:30 have anice week-end Mar 17 01:14:32 gn **** ENDING LOGGING AT Sat Mar 17 02:59:58 2012 **** BEGIN LOGGING AT Sat Mar 17 02:59:59 2012 Mar 18 00:04:58 bluelightning: ping Mar 18 00:32:26 hi ant__ Mar 18 00:33:07 hi, still on screen ? heh Mar 18 00:34:27 yeah Mar 18 00:34:32 I've sent the keymap patch now asking your opinion: add the sed in do_istall_append which has more limited scope? Mar 18 00:34:56 ..but we should sed a file in $D which is not so elegant... Mar 18 00:36:23 * ant__ is reading about "new ipad2" heating issues :/ Mar 18 00:37:10 those multicore chips needs watt, that's the sad rule Mar 18 00:48:12 patch looks ok :) Mar 18 00:48:42 sed probably belongs in do_configure where you have it currently I would think Mar 18 00:51:15 seems logical to me as well Mar 18 00:52:29 and all keymaps in meta-hh are 2.6, not only the ones for the machine listed in do_install_append case Mar 18 00:56:42 yeah Mar 18 00:56:54 gotta get some sleep... will try to apply this tomorrow Mar 18 00:56:57 cya Mar 18 00:57:40 good idea Mar 18 00:57:42 gn **** ENDING LOGGING AT Sun Mar 18 02:59:59 2012 **** BEGIN LOGGING AT Sun Mar 18 02:59:59 2012 Mar 18 23:01:28 hmmm... trying to get my head around the "new" openembedded.... still used to the "classic" setup ... Mar 18 23:26:25 hi there Mar 18 23:29:01 hey Mar 18 23:38:22 any devs around? Mar 18 23:49:36 I saw a post on zaurus-devel that names this channel as the place where general zaurus development is discussed these days Mar 18 23:50:12 i've been working o a new distro for Zaurii machines for a year and a half now Mar 18 23:50:30 Stupid of me to do it by myself, but hey... Mar 18 23:50:37 i've come a long way Mar 18 23:51:09 i mostly focused on performance issues Mar 18 23:51:19 and usability Mar 18 23:53:04 To this end i have made a heavily patched kdrive server based on 1.2.0 release Mar 18 23:53:27 most of the critical code paths have been replaced with arm asm code, including screen rotation Mar 18 23:54:01 most of the asm optimisations where taken from the maemo kdrive server Mar 18 23:54:17 i've also written faster screen rotation code Mar 18 23:54:32 current shadowfb code updates the screen pixel by pixel Mar 18 23:55:19 i've replaced that with iwmmxt asm code that rotates and copies 4x8 blocks of pixels Mar 18 23:55:37 it also has support for usb mice with mousewheel. Mar 18 23:56:06 silkscreen keys are used for mousewheel emulation Mar 18 23:56:59 and other patches, that reduce code size and api optimisations Mar 18 23:57:39 it uses half the memory of the current xserver-xorg-lite and is at least 50% faster, atleast that's what gtkperf says Mar 18 23:57:53 it uses xtscal for calibration Mar 18 23:59:58 i also have patches for the pxafb driver that place the framebuffer in section mapping to reduce/eliminate tlb thrashing Mar 19 00:02:26 i've added support for usb mode switching and x rotation to zaurusd Mar 19 00:03:14 i also have this: http://notaz.gp2x.de/ginge.php Mar 19 00:03:34 ginge basically gives us access to the complete software library of gp2x Mar 19 00:04:26 dromede: which devices have you tested this on? Mar 19 00:05:01 i have an akita Mar 19 00:05:15 * hvontres|home twiddles thumbs as poodle kernel compiles Mar 19 00:05:17 C-1000, that is Mar 19 00:05:46 some of my patches are pxa270 specific, unfortunatley Mar 19 00:06:59 hmm, sounds interesting. Mar 19 00:07:28 i have other things aswell, i just can't remember all of it right now Mar 19 00:07:36 and my oe layer is a mess Mar 19 00:08:06 heh.. I am still getting used to the new OE setup Mar 19 00:08:32 I should figure out how to setup a custom layer for messing with image configs./ Mar 19 00:08:41 i also have a basic port of gmenu2x: http://mtorromeo.github.com/gmenu2x/ Mar 19 00:09:05 * hvontres|home needs a faster build machine..... the old dual Turion is getting long in the tooth Mar 19 00:10:00 gmenu2x could be a great ui alternative for older zaurii devices **** ENDING LOGGING AT Mon Mar 19 02:59:58 2012 **** BEGIN LOGGING AT Mon Mar 19 02:59:58 2012 Mar 19 04:55:40 ~seen ant Mar 19 05:30:19 hvontres|home: hi Mar 19 05:30:35 dromede: hi too :) Mar 19 05:33:56 Jay7: hey Mar 19 05:35:25 BBLAYERS = "/home/oe/openembedded-core/meta" Mar 19 05:35:27 BBLAYERS += " /home/oe/openembedded-core/meta-openembedded/meta-initramfs " Mar 19 05:35:28 BBLAYERS += " /home/oe/openembedded-core/meta-handheld " Mar 19 05:35:30 BBLAYERS += " /home/oe/openembedded-core/meta-openembedded/meta-oe " Mar 19 05:35:47 hvontres|home: I'm using this in bblayers.conf Mar 19 05:37:07 is that to build a minimal image? Mar 19 05:37:20 I am using the stock Angstom config for now Mar 19 05:46:35 ah Mar 19 05:46:43 I'm not using angstrom right now Mar 19 05:47:01 I'm following ant Mar 19 05:47:17 he is concentrating on core minimal image Mar 19 05:47:28 Well, I've been away from this mess for a while, so I figured that would be a good place to start. Mar 19 05:48:32 I've gotten the Angstorom console-image to boot, but I can't figure out how to get the USB networking to start.... It looks like the low level usb-gadget stuff is hosed :( Mar 19 05:48:58 it's very possible Mar 19 05:49:39 well.. I should go to office now Mar 19 05:49:49 ant should appear soon from his office :) Mar 19 05:49:50 And I can't get my Spectrum24 to load the firmware either. I figure getting some sort of local networking up and running will help with user space work... that way we can update pacakges from a local feed insteadk Mar 19 05:50:07 ahh.. just in time for me to go to bed :) Mar 19 05:50:37 timezones are mess :) Mar 19 05:50:37 Jay7: does ant have some sort of webpage we could use as a status board? Mar 19 05:50:58 hvontres|home: no.. seems we should setup some Mar 19 05:51:10 I have VDS under kexecboot.org Mar 19 05:51:48 but I still hope we may host something on LTG Mar 19 05:51:49 speaking of kexecboot, did you know you can boot one kexecboot from another? really helps with testing....:) Mar 19 05:51:59 I'll ask florian as he appears Mar 19 05:52:06 hvontres|home: should be possible Mar 19 05:52:52 I've done it before... That way you can make sure it works before you flash it to the device.... should help prolong the life of the internal flash Mar 19 05:54:06 I was a bit out of OE and kexecboot development some time Mar 19 05:54:22 now I get office job so I hope to back to deal :) Mar 19 05:55:58 yeah... Those office jobs can really get in the way of the fun stuff though.. ): Mar 19 06:02:08 hmmm.... g_ether seems to load, but nothing shows up on the far end under lsusb :( Mar 19 06:07:51 you may raise this in zaurus-devel ML Mar 19 06:17:58 yeah,I want to dig into this a little more first. Mar 19 06:50:34 well, I think the pxa25x_udc driver is somehow not connecting to the outside world, but that is a far as I think i can get for now... (see ML for more details) Mar 19 06:50:38 time for bed Mar 19 08:30:16 morning Mar 19 08:33:38 hi dromede Mar 19 08:36:35 jay7 said earlier that you are working on a core minimal image Mar 19 08:36:50 will it be systemd or sysvinit based? Mar 19 08:36:55 he is testing it :) Mar 19 08:37:10 ant_work: check today's log Mar 19 08:37:19 atm I tend to use oe-core only, thus sysvinit Mar 19 08:37:43 i've managed to boot angstrom's systemd-image Mar 19 08:37:47 on an akita Mar 19 08:38:07 it does work and it does probably boot a bit faster Mar 19 08:38:27 but it seems to me that it uses quite a bit if memory Mar 19 08:38:45 I've tested core-image-minimal, core-image-core, core-image-base, core-image-sato and opie-image Mar 19 08:38:56 boot speed is not important on Z :) Mar 19 08:39:06 most of time it is sleeping Mar 19 08:39:10 i agree Mar 19 08:39:31 my image is also based on core-image-minimal Mar 19 08:39:33 but memory is critical Mar 19 08:40:07 what about touchscreen? Mar 19 08:40:16 that's why i'm not using xserver-xorg-lite Mar 19 08:40:32 it takes about ~20% right for the start Mar 19 08:40:48 so are you back to kdrive? Mar 19 08:40:58 my custom kdrive is about ~885kb big and uses around ~9% of memory Mar 19 08:41:06 yes Mar 19 08:41:42 it's not an ideal solution, but it makes sense if performance and memory usage is critical Mar 19 08:42:12 as is said it'w much faster than xorg-lite, has support for usb mice Mar 19 08:42:25 and touchscreen calibration via xtscal Mar 19 08:42:36 ah, it was broken with 1.7 iirc Mar 19 08:42:52 that's why we moved to xserver Mar 19 08:43:07 i've tested everything from 1.7, all the way back t. 1.1.0 Mar 19 08:43:19 we had a custom 1.3 iirc Mar 19 08:43:19 1.2.0 has proven to be fast and stable Mar 19 08:43:25 I see Mar 19 08:43:33 but Mar 19 08:43:40 the patchset is quite big Mar 19 08:45:35 i have also done a lot of tests to find and optimal gtk version Mar 19 08:46:08 i've found 2.16 to give best performance that would still allow compilation of recent packages Mar 19 08:47:01 newer gtk versions than 2.16 tend to be slower, especially scrolling Mar 19 08:48:32 my current Z desktop is based on pcmanfm-mod and awesomewm Mar 19 08:50:36 atm I think we suffer of a bad integration between oe-core and meta-oe Mar 19 08:51:28 I see issues with udev, xserver-nodm-init and ofc touchscreen Mar 19 08:51:51 using systemd solves some problems, sureω Mar 19 08:52:26 udev 181 is broken, i've tried it yesterday Mar 19 08:52:31 164 works fine Mar 19 08:54:37 I've tried different versions too, images built w/out systemd still sufer the 'udev-cache' plague Mar 19 08:56:01 back to ts, have you done any try about ads7846 kernel driver noise? Mar 19 08:56:09 yes Mar 19 08:56:28 more interpolation? Mar 19 08:56:37 have you read my posts on zaurus-devel on this? Mar 19 08:57:08 back in anuary Mar 19 08:57:14 January Mar 19 08:58:55 I think there is a problem with pxa spi driver Mar 19 08:59:04 screen redraw is slower on corgi Mar 19 09:00:50 my screen often flickers during CF traffic Mar 19 09:01:11 i use CF for root Mar 19 09:01:41 i don't know if it this is related Mar 19 09:02:13 as far as ads7846 noise is concerned, i have come up with a solution that works ok for me Mar 19 09:02:31 i completely disable in-kernel filtering Mar 19 09:02:53 it would be great to create image recipe with all that things :) Mar 19 09:03:08 i reduce ads7846 clock Mar 19 09:03:30 and i have patches for tslib that add low pass filtering an median filtering Mar 19 09:03:43 yes, an image would be nice Mar 19 09:03:52 but a complete overlay is better Mar 19 09:04:06 we have meta-handheld for this Mar 19 09:04:29 also, i don't build everything with oe Mar 19 09:05:04 i build my kernel separately, i have to integrate them with meta-handheld kernel Mar 19 09:05:27 and there are things i'm not allowed to give show Mar 19 09:05:32 like ginge source code Mar 19 09:05:46 i can only provide ginge in binary form Mar 19 09:06:24 the original author wants it to stay closed source Mar 19 09:07:31 np Mar 19 09:08:15 my libsdl and kdrive rely on a kernel patch that places the framebuffer in a section mapping Mar 19 09:08:50 this changes the way clients access the framebuffer Mar 19 09:09:17 i dont use mmap for fb access, i have two ioctls that return start and end adress of the framebuffer Mar 19 09:09:33 hm Mar 19 09:09:44 i mean, mmapping works Mar 19 09:09:47 would it work with accelerated drivers like atiw100 ? Mar 19 09:09:59 but then you dont get the benefit of a section map Mar 19 09:11:02 i don't know Mar 19 09:11:17 it depends on the way w100 uses ram Mar 19 09:12:47 btw, have you got a serial cable? Mar 19 09:12:56 no Mar 19 09:13:05 I have 2 Mar 19 09:13:08 and i can't get one since amazon wont ship to my country Mar 19 09:13:46 do you think I can send it by mail or could it be 'lost' during trasport? Mar 19 09:14:00 where are you from? Mar 19 09:14:04 Italy Mar 19 09:14:10 heh Mar 19 09:14:13 neighbors Mar 19 09:14:18 Croatia here Mar 19 09:16:45 i don't know how reliable our post office is Mar 19 09:17:16 heh Mar 19 09:17:25 i've ordered a few books and some stuff off of ebay and DX, i always got my package Mar 19 09:17:33 sometimes even twice :) Mar 19 09:17:46 I'm 120km form Venice Mar 19 09:17:55 risk seems low... Mar 19 09:18:58 europe is so small :) Mar 19 09:19:06 i'm from Koprivnica, a small city near the Hungarian border Mar 19 09:20:29 I received lot of stuff together with poodle, even a cf-gps Mar 19 09:20:48 tbh never inserted... Mar 19 09:21:08 ant_work: I got one from XorA :) Mar 19 09:21:19 never inserted too :) Mar 19 09:21:29 dromede: do you have the recharger/cabling? Mar 19 09:21:41 i have an original charger Mar 19 09:21:59 and the original micro usb/host usb cables Mar 19 09:22:39 i can't believe pulster still sells serial cables for 50 euros a piece Mar 19 09:23:06 they can be found on amazon for as little as 7 bucks Mar 19 09:23:26 I got one from maslowsky iirc Mar 19 09:23:53 he bought some on ebuy for 5$ per package Mar 19 09:24:21 I'll be away in DE this week, I'll send you one when I'm back Mar 19 09:24:52 great! Mar 19 09:25:24 i could really use one Mar 19 09:25:37 to diagnoes pm issues, for starters Mar 19 09:25:43 and why kexecboot wont boot Mar 19 09:25:58 ? Mar 19 09:26:34 I can boot 3.1 and 3.2 on corgi/poodle/spitz Mar 19 09:26:50 i'm stuck with 2.6.26 kexecboot on an akita Mar 19 09:26:59 anything newer than that won't boot Mar 19 09:27:03 the weird thing is Mar 19 09:27:23 newer versions will boot if i boot them from kexecboot 2.6.26 Mar 19 09:27:38 but they wont but if i flash them with zaurus-updater Mar 19 09:28:08 hm, do you use excessive lzma compression? Mar 19 09:28:19 this was stopping the boot initally Mar 19 09:28:52 tbh we onkly had the issue on 32MB devices like poodle Mar 19 09:29:05 not enough ram to decompress... Mar 19 09:29:20 well, has anyone else tried running newer kexecboot images on an akita? Mar 19 09:29:40 i use meta-oe to build kexecboot-linux Mar 19 09:30:18 the kernel config is exactly the same for akita/spitz Mar 19 09:30:27 i know Mar 19 09:30:45 so if spitz boots there is some problem with kernel, indeed Mar 19 09:30:54 at one point i tried to use gzip compression Mar 19 09:31:03 it didn't matter Mar 19 09:31:55 i also tried flashing with older zaurus updaters Mar 19 09:32:01 also didn't help Mar 19 09:32:12 very strange Mar 19 09:32:47 they boot in qemu aswell Mar 19 09:34:16 i've exported my kernel patchset with git format-patch Mar 19 09:34:35 there's 14 patches, not all of them are mine Mar 19 09:35:14 i don't know where to put them so you guys could take a look Mar 19 09:37:46 I'll try latest linux-kexecboot one time on my akita and tosa Mar 19 09:38:08 I've even charged akita :) Mar 19 09:39:26 i'm surprised that my battery can still hold a good charge Mar 19 09:39:34 it's ~5 years old now Mar 19 09:39:52 and i run my akita @ 524mhz constantly Mar 19 09:40:01 i still get ~4,5 hours Mar 19 09:41:27 about kernel review, best would be to get a review from lumag/stanislav/pawel Mar 19 09:41:41 so I'd say put it in github somewhere Mar 19 09:41:52 and send a msg to zaurus-devel Mar 19 09:42:33 and about the rest of the work (gtk, kdrive) probably this will ve good for ipaqs too Mar 19 09:42:40 bluelightning: ^^ Mar 19 09:42:48 yep Mar 19 09:42:57 hi guys Mar 19 09:43:01 hello Mar 19 09:43:14 reading your discussion, sounds great :) Mar 19 09:43:38 i'll integrate all of my work into my oe-layer and put it on github Mar 19 09:43:47 I'm getting abit crazy about the direction to take... kdrive or xorg Mar 19 09:45:27 yeah.. Mar 19 09:45:47 opie :) Mar 19 09:46:26 ;) Mar 19 09:46:48 i vote gmenu2x for console apps and emulators Mar 19 09:47:07 dromede: nanonote-style ;) Mar 19 09:47:33 something like that, yes Mar 19 09:47:47 I would prefer to have opie + console apps :) Mar 19 09:48:19 what's the status of opie these days? i haven'tried it in years Mar 19 09:48:20 so we need multiple images :) Mar 19 09:48:33 bluelightning says that opie is great ;) Mar 19 09:49:05 dromede: lots of bugs fixed, some features implemented, visually it hasn't changed very much at all though Mar 19 09:50:56 what about apps? Mar 19 09:53:06 no modern browser Mar 19 09:53:20 but anyway we can't have modern browser on Z Mar 19 09:53:24 yes we can Mar 19 09:53:25 no enough RAM Mar 19 09:53:39 opera mini + microemulator + jamvm should work great Mar 19 09:53:47 hm.. Mar 19 09:53:56 i have dillo3 running Mar 19 09:53:58 that's interesting Mar 19 09:54:15 dillo3 starts under 3 seconds :) Mar 19 09:55:01 you know what? i'm gonna tar up my current proto image, and upload it Mar 19 09:55:19 dromede: which version of opera mini you are using? Mar 19 09:55:39 i haven't tried any of that yet, it was just an idea Mar 19 09:55:44 ah Mar 19 09:56:31 http://www.opera.com/developer/tools/mini/ Mar 19 09:56:48 this is a demo of opera mini running through microemulator Mar 19 09:57:03 and i know maemo people have this running on nokia devices Mar 19 09:58:45 bluelightning: haven't you seen any opie-compatible java interfaces? Mar 19 09:59:11 I meah something to allow java programs to draw on top of opie Mar 19 10:05:30 bluelightning: do you foresee kdrive will be abandoned in oe-core any soon? Mar 19 10:09:44 ant_work: it would even better for us imho :) Mar 19 10:10:06 we can then just import it into meta-hh Mar 19 10:14:23 seems we are still at this point: http://lists.linuxtogo.org/pipermail/zaurus-devel/2011-June/000557.html Mar 19 10:14:42 but dromede did some progress :) Mar 19 10:15:12 I was about proposing to add zinput-calibrator to oe-core and have a complete framework Mar 19 10:15:18 *Xinput Mar 19 10:15:49 then we can overlay our kdrive in meta-hh Mar 19 10:16:05 but kdrive needs tslib.. Mar 19 10:17:54 what's wrong with tslib? Mar 19 10:18:11 nothing now, it ha been pulled back Mar 19 10:18:33 problem was "Remove xcalibrate and tslib support" Mar 19 10:18:54 i actually have patches that add evdev touchscreen support to my kdrive Mar 19 10:19:00 but it never worked properly Mar 19 10:19:35 tslib works just fine Mar 19 10:20:02 i have to add support for different resolutions to it Mar 19 10:20:31 I added xinput-calibrator to my core-image-sato and could calibrate the ts of poodle Mar 19 10:20:32 since libsdl console apps mostly use QVGA Mar 19 10:21:04 but no way for corgi/spitz , to much jitter Mar 19 10:21:05 Jay7: there was some java stuff on top of opie a long, long time ago... I'm not sure if it worked very well even back then, though I never actually tried it myself Mar 19 10:22:11 ant_work: I don't know for sure, it's possible it may disappear in the near future due to it being abandoned though Mar 19 10:25:19 i forgot to mention that most usb gadget drivers work Mar 19 10:25:27 i've tried g_ether and g_mass_storage Mar 19 10:25:37 the problem is a misconfigured OP2OCR Mar 19 10:25:46 this was adressed before on zaurus-devel Mar 19 10:26:06 you can use pxaregs for a quick fix Mar 19 10:26:27 pxaregs UP2OCR 0x00020070 Mar 19 10:26:41 this should enable proper usb client mode Mar 19 10:27:49 can a kernel patch solve that? Mar 19 10:28:25 as i said i've added automatic host/client switching to metan's zaurusd Mar 19 10:28:36 UP2OCR is configured for host in the kernel Mar 19 10:28:58 I see Mar 19 10:29:09 i use a kernel patch that exports relevant gpio's through "gpio-keys" Mar 19 10:29:40 oh, have you seen how that has been wrecked in kernel 3.3 ? Mar 19 10:30:45 3.3 kernels refuse to boot on my akita Mar 19 10:30:50 and they will crash qemu Mar 19 10:31:16 http://paste.debian.net/160295/ Mar 19 10:31:44 I have marked >100 mails to read in LAKML referinng Zaurus/pxa/sha1100 Mar 19 10:32:15 wow Mar 19 10:32:20 i.e. Re: [PATCH v3] pxa2xx_spi: fix memory corruption Mar 19 10:33:26 the spi bus master driver is a monster Mar 19 10:33:29 i dare not look at it Mar 19 10:36:34 we better stay with 3.2 for the moment Mar 19 10:37:14 and ideally we should hook our kernel to the linux-yocto Mar 19 10:50:46 what do we gain with linux-yocto? Mar 19 10:51:49 dromede: config fragments is the most useful feature Mar 19 10:53:07 oh Mar 19 10:53:13 yes, read something about that Mar 19 10:53:38 damn, i make a lot of typo's Mar 19 10:53:49 I'm not really used to realtime chat :/ Mar 19 11:06:52 hm Mar 19 11:07:36 kernel built with meta-handheld defconfig wont boot on my akita Mar 19 11:09:53 dromede: I don't know if it's true but iirc the first akita had a different nand Mar 19 11:10:38 how can i check which one i have? Mar 19 11:10:54 http://www.mail-archive.com/angstrom-distro-users@linuxtogo.org/msg03698.html Mar 19 11:12:25 cat proc/mtd Mar 19 11:14:20 mtd0: 006c0000 00800000 "Boot PROM Filesystem" Mar 19 11:14:21 mtd1: 00700000 00020000 "System Area" Mar 19 11:14:21 mtd2: 03a00000 00020000 "Root Filesystem" Mar 19 11:14:21 mtd3: 03f00000 00020000 "Home Filesystem" Mar 19 11:16:35 so i guess mtd0 is different in my case ? Mar 19 11:16:35 erasesize is definitely 0x20000 Mar 19 11:17:08 can't say, iirc I found that pastebin in internet (I have no akita) Mar 19 11:25:33 i compiled 3.2.9 with my config, it works Mar 19 11:47:37 nice, what was the problem? Mar 19 11:50:50 i don't know :) Mar 19 11:51:08 what's the best way to compare configs? Mar 19 11:52:43 diff? Mar 19 11:53:16 yes, but how do i generate a smaller config file? Mar 19 11:53:28 make showconfig iirc Mar 19 11:53:33 ok Mar 19 11:53:42 ant_work may remember better :) Mar 19 11:55:10 it's isnt showconfig Mar 19 11:57:03 savedefconfig Mar 19 11:57:19 yep, got it Mar 19 12:03:21 ah, yes Mar 19 12:03:27 showconfig is from freebsd ports :) Mar 19 12:05:01 dromede: do you think you'll be able to write a kernel patch for reading the mtd size as in kexecboot? Mar 19 12:06:19 could you be more specific? Mar 19 12:07:33 the code in userspace is : http://git.linuxtogo.org/?p=groups/kexecboot/kexecboot.git;a=tree;f=machine Mar 19 12:09:41 dromede: 2.4 has code for reading boot parameters from angel (Zaurus bootloader) Mar 19 12:09:52 it is floating around the internet Mar 19 12:10:05 or that, yes Mar 19 12:10:20 lumag may know more about this Mar 19 12:10:38 main idea - read mtd partitioning from it Mar 19 12:10:47 or from paraminfo area in NAND Mar 19 12:10:57 well Mar 19 12:11:25 i dont have a very good understanding of mtd devices Mar 19 12:11:29 never looked into it Mar 19 12:13:40 this one http://piro.sytes.net/~piro/pukiwiki/pukiwiki.php?zaurus%A1%A1boot%A1%A1loader%A1%A1for%A1%A1C3000 Mar 19 12:13:54 was for 2.6.16 Mar 19 12:19:38 http://paste.debian.net/160307/ Mar 19 12:19:47 this looks relevant to what you want Mar 19 12:21:00 i have an old pdaxrom 2.16.16 kernel, i'll apply the patch and look into it Mar 19 12:35:04 seems like this asm does the trick Mar 19 12:35:12 pls test, thx Mar 19 12:35:33 ok guys, I'm going now. I'll be back on friday Mar 19 12:35:39 see you soon Mar 19 12:35:44 bye **** ENDING LOGGING AT Tue Mar 20 02:59:58 2012 **** BEGIN LOGGING AT Tue Mar 20 02:59:59 2012 **** BEGIN LOGGING AT Tue Mar 20 12:29:04 2012 Mar 20 16:49:02 https://github.com/mkatic/meta-zos Mar 20 16:49:16 This overlay should contain everything i talked about earlier Mar 20 16:53:23 dromede: cool Mar 20 16:53:44 dromede: I guess ideally we would want to break out various pieces and put them somewhere more appropriate Mar 20 16:53:56 yes Mar 20 16:54:43 What's the status of the current zaurusd in meta-handheld? Mar 20 16:54:51 that's the one RP made, right? Mar 20 16:54:54 yeah Mar 20 16:55:08 I'm still a bit puzzled how we ended up with two distinct zaurusds Mar 20 16:55:47 RP would like to see (his) zaurusd turned into something more generic for device configuration Mar 20 16:55:54 well, one seems to be cobbled together with sh scripts Mar 20 16:55:56 so it's not dead... but it hasn't seen much work in a while Mar 20 16:56:06 last fix was from me a few years ago I think Mar 20 16:56:12 the other is written in C Mar 20 16:56:19 in daemon form Mar 20 16:56:40 although,it's probably not well tested Mar 20 16:56:42 i use it daily Mar 20 16:58:43 RP zaurusd has two daemons in it, switchevd and tskeys, as well as the scripts you mention Mar 20 17:00:42 IMHO, tskeys functionality should go into kdrive Mar 20 17:01:01 well, I would still need it for opie Mar 20 17:01:21 right Mar 20 17:01:26 i forgot about that Mar 20 17:02:31 it looks as if metan's zaurusd is only useful if you don't have any other mechanism of controlling the backlight and autosuspend, is that right? Mar 20 17:04:29 currently, yes Mar 20 17:04:45 i think it also has some kind of timekeeping thing, i'm not sure Mar 20 17:09:13 I'm inclined to think we should try to bring the two together although I'm unsure as to how we would do that... Mar 20 17:11:13 I'm looking at RP's zaurusd code now Mar 20 17:11:43 it does seem to be more generic than metan's version Mar 20 17:16:39 zaurusd has some alsa stuff in it as well, and I'm unsure as to how that ties into alsa-state if at all Mar 20 17:16:54 RP zaurusd that is Mar 20 17:31:45 heh Mar 20 17:33:01 now i realise that it would be easier to add usb switching to rp's version Mar 20 17:33:18 i should've looked at it earlier Mar 20 17:35:19 i dont like shell code, but it is much more flexible **** ENDING LOGGING AT Wed Mar 21 02:59:58 2012 **** BEGIN LOGGING AT Wed Mar 21 02:59:58 2012 Mar 21 22:33:41 ho Mar 21 22:34:32 hi ant_mob Mar 21 22:34:40 w much do virtual kb suxx? Mar 21 22:34:49 ho Mar 21 22:34:54 a lot apparently :) Mar 21 22:35:01 e bad Mar 21 22:37:53 bluelightning: I see Koen is starting to become boring Mar 21 22:38:21 protecting meta-ore Mar 21 22:38:57 the latere must be' splitting Mar 21 22:39:08 omg Mar 21 22:39:31 I'm pretty sure he agrees it should be split Mar 21 22:40:01 it's just that moving bits into oe-core right now possibly isn't the way to go Mar 21 22:41:22 then more layers asap Mar 21 22:43:41 I'm thinking about toolchain in meta-oe (4.5) Mar 21 22:43:48 and about X Mar 21 22:44:59 I'm holding / resisting to send a polemic Rfc Mar 21 22:45:58 i.e yes, that layer suxxs as it is now Mar 21 22:47:27 meta-oe shold not be mandatory, in any way Mar 21 22:47:50 well, there will be recipes we need from there... if it was less intrusive that wouldn't be a concern Mar 21 22:48:06 please remember to prune udev from meta-hh btw Mar 21 22:48:23 I'm a bit puzzled though, if we don't have giflib what do we use in oe-core for reading gifs? presumably some other library for that Mar 21 22:48:26 argh Mar 21 22:49:14 meta-graphics? it does exist iirc Mar 21 22:51:01 there is turbojpg iirc too, all that stuff could go in meta-graphics Mar 21 22:51:59 really? never noticed that... Mar 21 22:52:05 (meta-graphics) Mar 21 22:53:57 sry is called meta-multimedia Mar 21 22:54:02 is empty atm Mar 21 22:54:23 yeah I'm not sure what the plans for that are Mar 21 22:55:24 I'd say meta-graphics and meta-music or audio Mar 21 22:55:41 better 2 layers imho Mar 21 22:55:57 well, the first thing is to get the toolchains out of meta-oe Mar 21 22:56:10 agreed Mar 21 22:56:18 that's been definitely agreed but I don't know who is going to do it or when Mar 21 22:56:31 I guess everyone is waiting for khem to do it and he's probably too busy atm Mar 21 22:56:35 don't count on K Mar 21 22:57:11 the other K ;) Mar 21 22:58:03 ah he bumped udev despite the nack Mar 21 22:58:07 heh Mar 21 22:59:06 I just read the ML of today Mar 21 23:16:20 er, udev 17x is still there... Mar 21 23:16:27 in meta-oe Mar 21 23:16:39 181 was replaced with 182, that's all **** ENDING LOGGING AT Thu Mar 22 02:59:58 2012 **** BEGIN LOGGING AT Thu Mar 22 02:59:58 2012 Mar 22 23:35:54 bluelightning: after hard battle I could finally disable the german autocorrection on the iphone ;) Mar 22 23:36:12 ant__: ah, that was the difficulty... Mar 22 23:36:41 yes, then I lost field and got offline oO Mar 22 23:37:03 too near to the Alps Mar 22 23:38:32 I think I did send a msg on the ML which could potentially make Koen only half-bitter Mar 22 23:39:49 probably we'll have to create some new layers and a tool to combine them easily Mar 22 23:40:10 not just 'll' like Angstrom Mar 22 23:40:13 *all Mar 22 23:40:41 maybe extending the dependency concept Mar 22 23:42:31 btw, everybody seems to ignore that atm, after one year, meta-oe will only produce working images with systemd init manager Mar 22 23:42:45 and Angstrom distro Mar 22 23:42:52 heh Mar 22 23:43:16 design flaw? Mar 22 23:44:16 bluelightning: feel free to copy and paste and quote in the next TSC meeting Mar 22 23:47:32 when all the issues will be sorted out we can maybe merge down some Mar 22 23:47:56 layers of common use Mar 22 23:48:37 I don't see other ways to insure isolation Mar 22 23:48:56 and iirc Freescale is doing that (many extra layers) Mar 23 00:03:34 yeah... I think we pretty much agree on what needs to be done, but we haven't actually put stuff into action (although it's good to see the toolchain reorg instigated by koen today) Mar 23 00:04:11 yes, I was too hard judging him Mar 23 00:04:51 probably we can imagine a very fragmented meta-openembedded Mar 23 00:06:30 it remains to see what Yocto decides to do, i.e. moving the X/graphic recipes elsewhere Mar 23 00:07:22 maybe they'll accept to be able to include only some parts of meta-openembedded Mar 23 00:07:49 (like when I include meta-initramfs only) Mar 23 00:14:56 yahwn Mar 23 00:15:01 good night **** ENDING LOGGING AT Fri Mar 23 02:59:58 2012 **** BEGIN LOGGING AT Fri Mar 23 02:59:58 2012 Mar 23 11:10:13 Jay7: did you test new builds of kexecboot on your akita? Mar 23 11:10:34 dromede: long time not doing.. Mar 23 11:13:35 what was the config issue? Mar 23 11:17:02 diff is too big to pinpoint the exact issue Mar 23 11:19:48 my config is based on an older one utx has on his website Mar 23 11:25:00 even with savedefconfig? Mar 23 11:26:21 yes Mar 23 11:27:12 https://github.com/mkatic/meta-zos Mar 23 11:46:33 hm CONFIG_GPIO_PCA953X=y Mar 23 11:46:48 ??? Mar 23 11:46:58 this is necessary for akita Mar 23 11:47:07 akita has a max7310 port expander Mar 23 11:47:23 well not really necessary Mar 23 11:47:29 it will boot if i turn that off Mar 23 11:47:56 ideally you should compare the linux-kexecboot defconfigs Mar 23 11:48:13 here we have all sort of peripheral and network stuff Mar 23 11:48:27 here = production kernel Mar 23 11:48:39 i'm not sure if there's something wrong with the kexecboot config Mar 23 11:48:50 as it will boot with another kexecboot Mar 23 11:48:53 namely 2.6.26 Mar 23 11:50:31 I mean this defconfig http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-kexecboot-3.2/akita/defconfig Mar 23 14:55:17 bluelightning: again about keymaps...are you seriously thinking to use udev or systemd as suggested? Mar 23 14:55:27 :p Mar 23 14:56:01 pls merge the keymaps patch this weekend Mar 23 14:57:03 I honestly wouldn't know how to implement it with udev Mar 23 14:57:15 systemd is out because I don't use it and don't intend to migrate to it in the near future Mar 23 14:57:42 fyi https://wiki.archlinux.org/index.php/Map_scancodes_to_keycodes Mar 23 14:57:51 I'm really failing to merge patches in a timely fashion, sorry about that :/ Mar 23 14:57:53 a new udev rule Mar 23 14:58:03 np Mar 23 14:58:50 ok, interesting... thanks Mar 23 15:00:28 anyway udev should not be assumed as present Mar 23 15:01:20 even, in meta-hh we could probably get rid of it Mar 23 15:01:57 with some machine-specific hardcodings in the bsp Mar 23 15:02:26 but it looks to me more urgent to understand if oe-core will move to systemd Mar 23 15:02:35 I'm wondering about whether udev is the best option for old 2.6.21-hh-bound ipaqs Mar 23 15:03:06 yea Mar 23 15:03:38 if got that kernel building btw Mar 23 15:03:43 wasn't too hard Mar 23 15:03:45 but that kernel lacks devtmpfs so you need the extended table of devices Mar 23 15:03:50 yes Mar 23 15:04:02 well, I need old udev or some replacement Mar 23 15:04:23 iirc 0.92 was the best for old kernels Mar 23 15:04:43 also anything that depends on linux-libc-headers may have to be hacked/downgraded specifically Mar 23 15:05:05 ok, I think that was what was being used in the 2007 angstrom images that worked best on those machines Mar 23 15:05:20 definitely Mar 23 15:09:04 still, opie-image + 3.2 kernel flies on 32MB poodle, much smoother than with 2.4 I'd say Mar 23 15:09:49 we really need more kernel hackers / archeologist ;) Mar 23 15:10:25 cool :) Mar 23 15:10:34 indeed... Mar 24 00:18:46 hi hvontres|home Mar 24 00:20:18 ant__: hey Mar 24 00:20:53 so I finally dusted off the old poodle last weekend...:) Mar 24 00:21:06 heh, I've read Mar 24 00:21:32 we lost 2 years because of that exagerate lzma compression... Mar 24 00:22:12 Any ideas on how to debug the usb gagget driver? Mar 24 00:22:46 hm, iirc dromede posted the fix Mar 24 00:22:55 using pxaregs iirc Mar 24 00:23:06 pls check channel logs Mar 24 00:23:44 Mar 19 10:25:37 the problem is a misconfigured OP2OCR Mar 24 00:23:48 I think that was for pxa27x, but I'll check. I'll need to figure out how to add custom pkgs to the images now. Mar 24 00:24:04 ah, maybe you're right Mar 24 00:24:32 I've been around last two weeks, only checked quickly Mar 24 00:24:41 * hvontres|home loves free wifi on the train :) **** ENDING LOGGING AT Sat Mar 24 02:59:58 2012 **** BEGIN LOGGING AT Sat Mar 24 02:59:59 2012 Mar 24 21:36:48 hi channel Mar 24 21:49:13 hi ant__ Mar 24 21:53:14 what was exactly the issue with udev ts rule? ts not found? Mar 24 21:53:45 (wrt udev in meta-hh) Mar 24 21:57:14 hm.. now I notice psplash must be rotated on poodle... Mar 24 22:02:37 yep, no ts with recent kernels Mar 24 22:03:07 ok, I'll try to do some research about that Mar 24 22:03:19 strange that Zaurus doesnt suffer Mar 24 22:04:27 actually I'd need a bootlog to see how the ts is recognized by the kernel Mar 24 22:44:52 bluelightning: on poodle http://paste.debian.net/160880/ Mar 24 22:47:18 in my case only ATTRS{modalias}=="ads7846" matches, it seems Mar 24 22:48:57 remember ATTRS is for the parent while ATTR is for the input device Mar 24 22:49:38 let see on other models for fun Mar 24 22:51:20 heh, maybe the rule is indeed broken... Mar 24 22:51:45 note the 1 in the string " -e0,1,3, " Mar 24 23:29:18 bluelightning: this is c7x0 http://paste.debian.net/160882/ Mar 24 23:44:16 I used udevadm to be able to come up with that new rule that did match Mar 24 23:49:38 I'd suspect there is a new field Mar 24 23:50:01 or maybe that 1 means parent Mar 24 23:50:07 just guessing Mar 24 23:55:22 hm, maybe is a sorto of shell glob Mar 24 23:57:16 in fact I'm using udev from oe-core atm, so the rule works Mar 24 23:57:22 ;) Mar 25 00:04:14 it's a sort of regex, now I see Mar 25 00:04:41 well, feel free to send a patch for oe-core with the relaxed check Mar 25 00:05:06 I Think this will replace the actual string (with 18) Mar 25 00:05:45 no need for another rule Mar 25 00:08:08 bluelightning: about regex, pls push that keymaps patch ;) Mar 25 00:08:14 cu Mar 25 00:08:16 gn Mar 25 00:08:25 goodnight Mar 25 00:08:29 will do :) **** ENDING LOGGING AT Sun Mar 25 02:59:58 2012 **** BEGIN LOGGING AT Sun Mar 25 02:59:58 2012 **** ENDING LOGGING AT Mon Mar 26 02:59:58 2012 **** BEGIN LOGGING AT Mon Mar 26 02:59:58 2012 Mar 26 21:49:28 hi **** ENDING LOGGING AT Tue Mar 27 02:59:58 2012 **** BEGIN LOGGING AT Tue Mar 27 02:59:58 2012 Mar 27 14:05:22 hi Mar 27 14:05:36 I have netbooted with kexec here: Mar 27 14:05:52 http://www.youtube.com/watch?v=nnuSlQuLeMM Mar 27 14:06:02 I used netbootcd Mar 27 14:06:27 it was for an old laptop that did not had support in the BIOS for the ethernet PCMCIA card Mar 27 14:07:39 I am interested to boot over wifi Mar 27 14:07:51 with either some hardcoded values Mar 27 14:08:00 or with a tool like wicd-curses Mar 27 14:17:11 zoobab: that's interesting **** ENDING LOGGING AT Wed Mar 28 02:59:58 2012 **** BEGIN LOGGING AT Wed Mar 28 02:59:59 2012 **** ENDING LOGGING AT Thu Mar 29 02:59:58 2012 **** BEGIN LOGGING AT Thu Mar 29 02:59:58 2012 Mar 29 14:23:33 dromede_: I'm back in office, I'll send you the cable tomorrow morning Mar 29 14:34:38 bluelightning: I'll try Yocto BSP tools, version 2 soonish Mar 29 22:41:43 bluelightning: are you still on screen ? Mar 29 22:43:02 hi ant__ Mar 29 22:43:04 yes :) Mar 29 22:43:23 heh Mar 29 22:43:52 quickly, I have a doubt wrt linux-yocto and tools Mar 29 22:44:15 it seems we'd need a git checkout for each machine Mar 29 22:45:00 as far as I could customize it, at least, could not find the concept of 'family' like SOC Mar 29 22:45:16 that may be the case yes... probably worth checking with zeddii though Mar 29 22:45:46 clearly that path is not viable for meta-handhelds Mar 29 22:46:13 we can alway abuse of the 'compatibilty' hooks Mar 29 22:46:31 and countinue supplying our defconfigs Mar 29 22:46:45 thus no need for linux.inc Mar 29 22:46:51 in perspective Mar 29 22:47:50 we can still use config fragments even without the repo structure Mar 29 22:47:56 that has value IMHO Mar 29 22:48:00 that sounds good Mar 29 22:48:20 I plan to add the kexec stuff and create linux-.kexecboot of it Mar 29 22:48:39 just four-five CONFIG options Mar 29 22:48:56 (what is in linux-kexecboot.inc) Mar 29 22:49:30 reusing oe-core kernel stuff would be wonderful Mar 29 22:53:06 bluelightning: remember there is that meta-oe .bbappend to remove asap (udev ;) Mar 29 23:04:14 ant__: it needs to go into oe-core before it can be removed... Mar 29 23:04:48 yea, pls send the patch Mar 29 23:05:57 (I can half -promise I'll read the driver and try to underestand the sysfs fingerprint of that ATTRS) Mar 29 23:06:55 maybe something changed in driver or at platform level Mar 29 23:07:58 * ant__ notes 'touchscreen internals' on its to-do Mar 29 23:09:34 *his, heh well, writing in english in third person is not that easy at 1 o'clock Mar 29 23:09:46 see you online tomorrow Mar 29 23:09:48 gn **** ENDING LOGGING AT Fri Mar 30 02:59:58 2012 **** BEGIN LOGGING AT Fri Mar 30 02:59:58 2012 Mar 30 10:02:21 bluelightning: so, I could find some hints about ts not recognized. Mar 30 10:02:38 hi ant_work Mar 30 10:02:42 oh, what did you find? Mar 30 10:02:46 all dates back to udev 124 Mar 30 10:02:52 http://lists.linuxtogo.org/pipermail/openembedded-commits/2009-January/022948.html Mar 30 10:03:48 so I'd look in the specific driver code and see if really modalias is wrong Mar 30 10:06:28 er, but it's not devices that identify as "ads7846" that I'm trying to pick up... Mar 30 10:07:16 sure, but it signals a first mismatch Mar 30 10:07:38 I'm reading input.h now... Mar 30 12:34:52 bluelightning: ah, it seems the same touchscreen of the first GTA01 Mar 30 12:37:40 anyway, after reading http://www.kernel.org/doc/Documentation/sysfs-rules.txt Mar 30 12:38:02 seems that those rules are a rather hacky way to match sysfs entries Mar 30 12:38:35 "Everything else is just a kernel driver-core implementation detail Mar 30 12:38:35 that should not be assumed to be stable across kernel releases." Mar 30 12:39:02 right... ok Mar 30 12:39:41 I suppose the last hurd of multitouch devices changes the matter Mar 30 16:08:58 hm.. Mar 30 16:09:01 7 #define BIT_MASK(nr) (1UL << ((nr) % BITS_PER_LONG)) Mar 30 16:09:32 I'm looking at ads7846.c Mar 30 16:09:48 1292 input_dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_ABS);1293 input_dev->keybit[BIT_WORD(BTN_TOUCH)] = BIT_MASK(BTN_TOUCH); Mar 30 16:10:08 #define EV_KEY 0x01 Mar 30 16:10:18 #define EV_ABS 0x03 Mar 30 16:10:37 and so on Mar 30 16:10:59 I'm trying to understand how that modalias is composed Mar 30 16:12:04 btw #define BTN_TOUCH 0x14a Mar 30 16:12:52 and #define BIT_WORD(nr) ((nr) / BITS_PER_LONG) Mar 30 16:13:15 ok, I'll have to write it down in C, seems easier :) Mar 30 16:13:44 dromede: can you help here? Mar 30 16:14:21 what's the problem) Mar 30 16:14:50 trying to understand the stuff beyond the udev ts rule Mar 30 16:15:45 SUBSYSTEM=="input", KERNEL=="event[0-9]*", ATTRS{modalias}=="input:*-e0*,3,*a0,1,*18,*", SYMLINK+="input/touchscreen0" Mar 30 16:16:12 which devices are affected? Mar 30 16:16:24 this rule does not work for some ipaq Mar 30 16:16:42 and for old GTA01 Mar 30 16:17:17 see the udev files in meta-handheld Mar 30 16:17:33 we're trying to solve the ts probe in oe-core Mar 30 16:18:09 I'm trying to see if e.g. the 18 pattern is regular Mar 30 16:18:49 the (platform)driver gives info in order to populate sysfs, isn't? Mar 30 16:20:10 we end up with 3 strings Mar 30 16:20:44 http://cgit.openembedded.org/meta-handheld/tree/recipes-core/udev/udev/touchscreen.rules Mar 30 16:21:59 i don't even have that ruleset in my rootfs Mar 30 16:22:24 me too, using oe-core standard udev Mar 30 16:22:58 yet, i get proper event* devices and the touchscreen0 symlink Mar 30 16:23:02 those 0,1 hm... #define ABS_X 0x00 Mar 30 16:23:02 #define ABS_Y 0x01 Mar 30 16:23:19 these only define the capabilites of the input device Mar 30 16:23:33 yes, I'm looking at that Mar 30 16:23:41 is it comma separated array? Mar 30 16:24:02 I can't get the whole picture yet (w/out device :) Mar 30 16:25:22 standard oe-core rules: http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-core/udev/udev/local.rules Mar 30 16:26:22 hm.. and that 18 #define ABS_PRESSURE 0x18 Mar 30 16:27:53 well, ok, I'll look at the full string later at home Mar 30 16:28:14 dromede: btw the cable has been sent Mar 30 16:28:22 great, thanks again Mar 30 16:28:52 i can't be of much help with the udev problem Mar 30 16:29:23 it looks to me a very specific regex Mar 30 16:29:32 maybe working only on ads7846 Mar 30 16:29:41 and not with other drivers Mar 30 16:30:00 i suggest using "udevadm monitor" and modprobe, rmmod ads7846 to see what's really happening Mar 30 16:30:23 yea, I already have some verbose logs, thx Mar 30 16:31:35 probably best would be match with the 'real' modalias as it is defined by the driver Mar 30 16:32:01 i.e MODULE_ALIAS("spi:ads7846"); Mar 30 16:32:37 but this baby does much more than ts .... Mar 30 16:33:14 i'm still in favor of the older corgi_ts driver Mar 30 16:33:32 heh Mar 30 16:33:40 it was much smaller and it worked Mar 30 16:33:56 and i doesn't have as much overhead as ads7846 Mar 30 16:34:07 it was on the ssp bus, isn't? Mar 30 16:34:13 i actually tried to port it to current kernel Mar 30 16:34:16 yes Mar 30 16:34:29 all of that was ripped out in favor of the current spi master driver Mar 30 16:34:54 the current solution is more generic Mar 30 16:35:03 but i believe it has more overhead Mar 30 16:35:14 let's hope spi will be better Mar 30 16:35:51 it's ok as it is now Mar 30 16:36:27 pm issues will be my first priority when i get the cable Mar 30 16:37:07 this is the only real problem i have with current kernels Mar 30 16:37:12 btw I think in 3.3 is [PATCH v3] pxa2xx_spi: fix memory corruption Mar 30 16:37:29 i compiled the latest one today Mar 30 16:37:33 haven't tried it yet Mar 30 16:38:39 this weekend I'll work again on defconfigs Mar 30 16:38:51 I'll try to 'yoctize' them Mar 30 16:39:04 fragments and all that? Mar 30 16:39:25 yes, linux-yocto-tiny or maybe linux-yocto-korg Mar 30 16:39:56 i'm working a decent desktop solution for Z Mar 30 16:40:12 great, I'm not yet in userland :) Mar 30 16:40:15 currently, its pcmanfm + awesome Mar 30 16:40:31 awesome is awesome :) Mar 30 16:40:36 very flexible and lightweight Mar 30 16:40:37 ah, awesome was light, I remember Mar 30 16:41:03 look, we are fixing the base images, feel free to put all in a couple of recipes Mar 30 16:41:25 surely interesting for all meta-handheld machines Mar 30 16:41:32 hm Mar 30 16:41:33 'RAM challenged' Mar 30 16:42:02 i have some tslib patches that might be interesting Mar 30 16:42:05 more filters Mar 30 16:42:13 median and lowpass Mar 30 16:42:20 eh, then we have to go back to kdrive :) Mar 30 16:42:25 is not a problem for me Mar 30 16:42:30 oh right, forgot about that Mar 30 16:42:42 as it is is slow on 32MB devices Mar 30 16:43:06 but it is bad configurated, probably Mar 30 16:43:19 I'm not so fond on X Mar 30 16:43:54 i hope to have something presentable in a few weeks Mar 30 16:44:03 if you say that using kdrive 1.2 or 1.3 with tslib we have no issues with kernel 3.x and udev, well, why not? Mar 30 16:44:36 it's a large patchset Mar 30 16:44:41 and mostly specific to pxa27x Mar 30 16:44:51 ok, those pathces could be left out Mar 30 16:45:06 it also requires legacy X headers Mar 30 16:45:13 it is fast, but it's messy Mar 30 16:45:41 I really don't know...since years I hear 'kdrive is dead' Mar 30 16:45:50 but is still in oe-core.... Mar 30 16:45:51 i'll have an image for you guys in a few weeks so you can give it a go Mar 30 16:45:59 true, kdrive is no longer maintained Mar 30 16:46:27 but then again, it's perfect for our primitive devices Mar 30 16:46:59 there's also gmenu2x Mar 30 16:51:09 ok, see you later on Mar 30 16:51:12 bye Mar 30 21:50:57 dromede: I got that modalias string: http://paste.debian.net/161503/ Mar 30 22:23:09 Jay7: are you there? Mar 30 22:38:02 well, it's now clear :)) Mar 30 22:38:18 the letters are the titles in /include/linux/input.h Mar 30 22:38:28 ATTRS{modalias}=="input:b0000v0000p0000e0000-e0,1,3,k14A,ra0,1,18,mlsfw" Mar 30 22:38:40 b0000v0000p0000e0000- board, vendor, product , (etc) Mar 30 22:38:49 e 0,1,3 e = supported events (#define EV_SYN 0x00 #define EV_KEY 0x01 #define EV_ABS 0x03) Mar 30 22:38:50 ant__: hi Mar 30 22:38:57 k 14A k = supported keys/buttons (#define BTN_TOUCH 0x14a) Mar 30 22:39:02 r r = relative axes Mar 30 22:39:07 a 0,1,18 a = absolute axes (#define ABS_X 0x00 #define ABS_Y 0x01 #define ABS_PRESSURE 0x18) Mar 30 22:40:27 m misc events Mar 30 22:40:32 l leds Mar 30 22:40:41 a autorepeat Mar 30 22:40:47 s sounds Mar 30 22:41:47 something like that ;) Mar 30 22:42:10 Jay7: how did we decide it is a touchscreen? Mar 30 22:42:33 ioctlo gives similar info iirc Mar 30 22:42:38 *ioctl Mar 30 22:42:39 ant__: kexecboot just looking at evdev capabilities Mar 30 22:47:39 Jay7: if you look at the logs around 16 h o'clock you'll see Mar 30 22:47:42 input_dev->evbit[0] = BIT_MASK(EV_KEY) | BIT_MASK(EV_ABS);1293 input_dev->keybit[BIT_WORD(BTN_TOUCH)] = BIT_MASK(BTN_TOUCH); Mar 30 22:48:12 now, for arm BITS_PER_LONG is 32 Mar 30 22:48:53 can you verify using this formula ? #define BIT_MASK(nr) (1UL << ((nr) % BITS_PER_LONG)) Mar 30 22:49:32 I'm too tired to understand this all :) Mar 30 22:49:40 heh he Mar 30 22:49:45 let's continue tomorrow :) Mar 30 22:49:50 not now ;) Mar 30 22:50:08 ah, about the Russian clock..how many hours then? Mar 30 22:50:11 I'm out of context now Mar 30 22:50:26 02:50 Mar 30 22:50:30 omg Mar 30 22:50:55 I start feeling sleepy at 00:50 Mar 30 22:51:01 we are not changing to/from daylight saving more btw Mar 30 22:51:20 heh.. I'm feeling sleepy about 3 hours already :) Mar 30 22:51:28 ok, that was the change Mar 30 22:52:30 just need to assemble computer for my former customer Mar 30 22:53:22 because it is not possible while child is not sleeping :) Mar 30 22:53:46 oh, I remember that :/ Mar 30 23:08:50 ok, e0,1,3, k14A, are respectively input_dev->evbit and input_dev->keybit Mar 30 23:10:20 just those BIT_MASK doesn't seeem to corrispond at first sight :/ Mar 30 23:12:30 and also wrong sems #define BIT_WORD(nr) ((nr) / BITS_PER_LONG) Mar 30 23:12:44 bah, maybe tose are internal structures Mar 30 23:12:48 I'll check tomorrow Mar 30 23:16:52 Jay7: dromede: Mar 30 23:17:38 doing poweroff from console (core-image-base) do finally shut off (suspend?) the poodle Mar 30 23:18:03 not from kexecboot, probably need some more powerman in defconfig Mar 30 23:18:42 it does even light the recharge-led ! Mar 30 23:19:48 I had almost lost any hope ;) Mar 30 23:20:24 fwiw it does that with rootfs on nand Mar 30 23:23:04 sadly it did that only once...now is rebooting :/ Mar 30 23:23:34 and recharge led is off Mar 30 23:23:45 strange Mar 30 23:24:21 maybe battery was so dry it could not reboot :) Mar 30 23:26:35 gn **** ENDING LOGGING AT Sat Mar 31 02:59:58 2012 **** BEGIN LOGGING AT Sat Mar 31 02:59:59 2012 Mar 31 10:48:51 poweroff from console works just fine for me Mar 31 15:19:34 here are some gtk2.16.6 vs gtk2.24.8 gtkperf benchmarks Mar 31 15:19:37 http://paste.debian.net/161553/ Mar 31 15:20:06 as i said before 2.16 is faster Mar 31 15:20:12 but i have to say, 2.24 is more stable Mar 31 15:21:10 there are apps that claim that they need =< gtk2.16 but will fail with a segfault in various places Mar 31 15:21:21 this simply doesn't happen with 2.24.8 Mar 31 15:22:34 so i'm still on the fence on which version to use Mar 31 15:23:28 the most notable difference between the two is scrolling, it's much "smoother" on 2.16 Mar 31 15:23:39 especially in gtktreeview widgets Mar 31 23:04:17 well, to end up my dissertation I have calculated evbit[0] = 3 and keybit[10] = 10 Mar 31 23:17:55 bluelightning: I'm pretty sure the 18 in the string is for ts reporting pressure Mar 31 23:18:20 we can quickly add our local_rules in overlay to oe-core's udev Mar 31 23:18:32 but best would be send a patch Mar 31 23:19:31 important is to cut the binding with meta-oe Mar 31 23:42:09 ok, I'll send 1st april patch ;) Mar 31 23:44:03 sent Mar 31 23:44:05 gn **** ENDING LOGGING AT Sun Apr 01 02:59:58 2012 **** BEGIN LOGGING AT Sun Apr 01 02:59:58 2012 **** ENDING LOGGING AT Mon Apr 02 02:59:58 2012 **** BEGIN LOGGING AT Mon Apr 02 02:59:58 2012 Apr 02 08:51:55 dromede: hi, calculating the speed of our postal services and the contrary wind blowing from E/NE you have some little hope to get the cable within this week... Apr 02 08:58:53 ant_work: morning Apr 02 08:59:39 ant_work: since you've explained clearly where the difference came from I wonder if we can just get the udev rule change into OE-Core Apr 02 08:59:42 hello Apr 02 08:59:54 you know what..now I have a doubt... Apr 02 08:59:55 unfortunately now is not a good time :/ Apr 02 09:00:02 heh Apr 02 09:00:08 (i.e. stabilisation period) Apr 02 09:00:16 we can test it in meta-hh for now Apr 02 09:00:47 the point is, I have to be sure that we don't match joysticks or touchpads Apr 02 09:01:26 iircjoystick has a separate driver Apr 02 09:02:15 ant_work: That's ok, i still don't have a usb serial adapter Apr 02 09:02:36 i'm gonna try and make one from an old cellphone cable Apr 02 09:02:40 if i can find it ... Apr 02 09:02:56 so the udev rule is ok now ? Apr 02 09:06:05 I hope so, ideally we would add a couple of rules for each ts_driver, matching the respective modalias Apr 02 09:06:35 which part of the rule was the problem? Apr 02 09:06:52 (for h1940 ts.input->name = "S3C24XX TouchScreen"; ts.input->id.vendor = 0xDEAD; ts.input->id.product = 0xBEEF; ts.input->id.version = 0x0102;) Apr 02 09:07:49 hehe Apr 02 09:07:50 the '18' in the string means pressure and SRC24XX doesn't seem to expose that Apr 02 09:08:48 it's not really used in tslib either Apr 02 09:08:52 pressure i mean Apr 02 09:12:20 I think only kernel 2.4 did make use of ts_pressure, at least one app (Petite Peinture) Apr 02 09:12:35 well, I mean in the 2.4 era Apr 02 09:24:53 bluelightning: yesterday I fell asleep reading the "Angstrom meets Yocto" thread. What's the juice of all that? After stabilization, will poky be split in distro-layer and integration-layer(meta-yocto)? Apr 02 09:25:36 as Khem pointed out, we already have 'defaults' in oe-core which are rather sane Apr 02 09:25:53 yes, which poky already builds on Apr 02 09:25:56 I think ideally meta-yocto would be replaced with meta-poky and separate layers for each of the BSPs in there currently Apr 02 09:41:05 hm, I can't say I immediately understand that Apr 02 09:42:34 well, meta-yocto is termed an "integration layer" but all that's really in it is the distro policy and a BSP for each hardware reference machine Apr 02 09:44:22 ah, ok, I was wondering about the magic glue supposedly provided by meta-yocto Apr 02 09:46:52 so poky/tree/meta-yocto is the 'poky' distro plus support for some machines Apr 02 09:49:38 other than that, cgit.cgi/poky/tree/ seems exactly copy of openembedded-core/tree/ Apr 02 10:08:29 yes, that's because it is :) Apr 02 10:09:09 I think we removed oe-core's local.conf.sample in favour of our own customised version, but that's it Apr 02 10:12:02 well, clearly /meta-yocto/conf/machine belongs to BSP layer(s) Apr 02 10:14:38 and about /poky/tree/meta, it seems to me superfluous Apr 02 10:15:01 why? Apr 02 10:15:02 and the argumentation 'there is no release' is weak Apr 02 10:15:35 the point is you only need to do one git checkout and you have everything you need... Apr 02 10:15:49 again, weak.... sorry Apr 02 10:16:09 we are talking to developer, not granny's Apr 02 10:16:34 one could just refer to a git TAG or SRCREV Apr 02 10:16:42 that's all Apr 02 10:16:50 well, I disagree, I think it saves new users a lot of hassle Apr 02 10:17:00 and there's relatively little disadvantage Apr 02 10:22:59 the advantage is imho minimal (and hides the real layer structure); the disavantage is the confusion it originates (is it a mirror of oe-core, is it 'upstream') Apr 02 10:23:23 what is it finally? is just a checkout of oe-core Apr 02 10:24:24 see, I could even say: then keep only this and shut off openembedded.org/oe-core Apr 02 10:24:36 but two is really confusing Apr 02 10:25:25 same for bitbake Apr 02 10:49:01 please, be serious... Apr 02 10:51:45 a single repository gives us a trivial way to test a specific version of the metadata (OE-Core, BSPs) with a specific version of bitbake Apr 02 10:53:26 we will likely end up doing the tags anyway, so you will have the best of both worlds Apr 02 11:16:10 bluelightning: look, as you know I've zero (commercial) interest being that's just a time-consuming hobby for me. So, I can say that from an 'outsider' POV is not clear what the 2 worlds are Apr 02 11:16:57 I just see two projects, using oe-core + bitbake Apr 02 11:17:23 and, fwiw, I'll insist having sane defaults for building distro-less images Apr 02 11:18:12 I think people have different ideas on how to structure things; ultimately we're still using the same code Apr 02 11:18:18 I'll let RPJ Day eviscerate the issue :) Apr 02 11:18:41 do we not already have sane defaults? Apr 02 11:19:04 I0d say ies, maybe not at 100% but close Apr 02 11:19:35 (because of xorg) Apr 02 11:19:46 right, yes... that one is still not addressed Apr 02 11:20:11 we still have quite a bit of cleanup to do with meta-oe Apr 02 11:20:26 right Apr 02 11:20:59 this one is 'doomed' imho, it could be eventually be 'next' or a dev branch Apr 02 11:21:43 seeing how it has been deployed until now , it could rather be a sort of 'next' Apr 02 11:21:57 or rather a systemd-fork Apr 02 11:22:29 I, for me, don't put much value in it Apr 02 11:22:44 *on it Apr 02 11:23:05 (so does Yocto crowd it seems ;) Apr 02 11:25:01 at the moment we have had very little time to look at it, we have enough to do with maintaining the majority of OE-Core and bitbake... Apr 02 11:25:13 are there any ways/scripts to autogenerate .bb's from source packages? Apr 02 11:25:25 I see value in the recipes within it, whether they are all munged into one layer or not is another question Apr 02 11:25:38 something like autscan in bitbake that spat out initial .bb files Apr 02 11:25:42 *autoscan Apr 02 11:27:01 bluelightning: there are surely some recipes to cherry-pick but the real deal will be the unification of graphic stuff Apr 02 11:27:41 but the again: Koen says we'll bloat oe-core and Yocto people tends to keep all together for stabilization Apr 02 11:28:18 dromede: iirc there was work to produce srpms but I'm not aware of the opposite Apr 02 11:28:24 dromede: there is, yes - there were some patches to add such a script but I don't think they got merged Apr 02 11:29:45 well, it seems to me that alot of the fields in .bb files are already present in configure.ac templates Apr 02 11:30:00 a script like that would be easy to do Apr 02 11:30:28 could you point me to those scripts you mentioned? Apr 02 11:31:21 http://comments.gmane.org/gmane.linux.embedded.poky/7510 Apr 02 11:31:28 thanls Apr 02 11:31:32 *thanks Apr 02 11:31:36 https://bugzilla.yoctoproject.org/show_bug.cgi?id=1656 Apr 02 11:33:12 dromede: well, the easiest way is to take a source-wise 'similar' recipe and modify it Apr 02 11:33:47 i know Apr 02 11:34:23 but it would be cool to have something like "bitbake -c autoscan URL" Apr 02 11:34:31 and you get an initial .bb file Apr 02 11:34:45 like autoscan does for configure scripts Apr 02 11:34:53 that's basically what that script does Apr 02 11:35:50 maybe calling it a script is an understatement Apr 02 11:35:56 since it's ~1800 lines :) Apr 02 11:36:31 well, it all depends on how smart you want it to be... Apr 02 11:39:57 ant_work: the aim is to be able to add meta-oe to your layer config and not have it change anything (or not anything fundamental at least) Apr 02 11:41:31 ok, but why would one need to? Apr 02 11:42:13 same for Distro: why does YP need one if the defaults are sane? Apr 02 11:42:33 that's the underlying question about Poky distro Apr 02 11:42:44 what does it add to the defaults? Apr 02 11:44:55 ok, I already know: versioning, stable releases, tested combinations of bitbake+metadata, .. such things Apr 02 11:45:03 .. Apr 02 11:45:08 I haven't compared them to see exactly but there are minor differences in distro policy Apr 02 11:45:16 stuff we do not necessarily want to force on the world Apr 02 11:45:22 such as OEBasicHash Apr 02 11:46:18 (in the case of OEBasicHash, we wanted to test it ourselves first) Apr 02 11:46:42 although I think Martin and Koen already turned it on for themselves Apr 02 11:47:05 I'm under the impression only me and Khem are testing distroless-oe-core Apr 02 11:52:09 ..and I'm not yet satisfied about its status Apr 02 11:52:20 then I'll move to $distro Apr 02 11:53:18 Another topic now, somehow related Apr 02 11:53:39 linux-yocto-tiny has been 'normalized' and now all resides in a foreign repo Apr 02 11:54:07 don't know anything about that I'm afraid Apr 02 11:54:15 best to talk to darren Apr 02 11:54:18 is a recent commit Apr 02 11:54:33 so I suspect I'll need meta-yocto layer now Apr 02 11:54:42 why? Apr 02 11:55:02 I'll work on the kernel convergence during next holidays Apr 02 11:55:26 because I'd have to add a proper BSP if I want to try a pure yocto kernel Apr 02 11:56:01 ah right Apr 02 11:57:34 hmm, maybe yocto-kernel-tools Apr 02 12:03:02 no, another one for the moment Apr 02 12:03:03 http://git.yoctoproject.org/cgit.cgi//log/?h=tzanussi/yocto-bsp-tools.v2 Apr 02 12:03:17 it's a bit sparse... ;) Apr 02 12:07:07 got it: http://git.yoctoproject.org/cgit.cgi/poky-contrib/log/?h=tzanussi/yocto-bsp-tools.v2 Apr 02 12:09:04 bluelightning: I think this is the up-to-date stuff, which will be demonstrated today in Freemont Apr 02 12:10:26 I'll first try to hack around it and provide the .cfg bits in the recipe Apr 02 12:11:14 but if we ever convert the configs and create proper BSP the work will be done once for all Apr 02 12:11:44 just it will be per-machine Apr 02 12:37:13 bluelightning: oops, I did not rebase the patch, sorry Apr 02 12:37:24 ant_work: no worries, I did it :) Apr 02 12:38:08 btw pls verify if the string *ra0,1,* *a0,1,* would apply to your ts Apr 02 12:38:22 I suspect relative (r) is never set for ts Apr 02 12:38:38 so we would enforce the check vs. joysticks etc. Apr 02 12:40:49 it may take me a while to verify that Apr 02 12:41:16 np, maybe before submitting the patch to oe-core Apr 02 12:42:10 btw, could we have just usend an amend instead of .bbappend for the one-liner? Apr 02 12:42:27 amend? Apr 02 12:42:34 amend.bbclass ? Apr 02 12:42:45 amend.inc is obsolete.. Apr 02 12:42:49 ah, ok Apr 02 12:43:51 anyway much better a .bbappend binding to oe-core than on meta-oe :) Apr 02 12:44:25 put it last on to-do list ;) Apr 02 12:53:45 bluelightning: ehm, in the case you'd add meta-oe then your ts will not be recognized. Better to propose a patch there as well. Apr 02 12:54:00 ant_work: I know Apr 02 12:54:08 ant_work: I don't have time to sort that out atm Apr 02 12:54:19 I relaized it only now... Apr 02 12:54:36 np, I can send a patch for meta-oe anf gather som emore comments Apr 02 13:33:34 bluelightning: heh, I realize now a less invasive patch can be done using sed in e.g. do install_prepend Apr 02 13:34:08 and get rid of the local.rules file Apr 02 15:58:40 wow Apr 02 15:58:52 long conversation here :) Apr 02 16:02:42 heh, we have logging service for free Apr 02 16:03:57 I think all summing up in the last RP reply Apr 02 16:04:44 https://lists.yoctoproject.org/pipermail/yocto/2012-April/007969.html Apr 02 16:05:21 single repo is 'easy'... yet not easier than a tarball :) Apr 02 16:06:34 guys, yocto bsp summit is starting if you're interested Apr 02 16:07:13 thx for reminder Apr 02 16:07:27 #yocto-bsp-summit and dail-in info (including 1-800 which you can use for free via skype) for audio is on the mailing list Apr 02 16:07:36 Iμll have to drive home now... Apr 02 16:07:59 ant_work: ah ok Apr 02 16:08:31 I'll try to keep in touch on iphone ... Apr 02 16:08:38 let see Apr 02 16:11:05 apparently they are recording the sessions as well Apr 02 16:14:16 ok, bbl **** BEGIN LOGGING AT Mon Apr 02 20:24:03 2012 Apr 02 20:40:53 any log of the summit? Apr 02 20:41:42 no log but I can give you a very brief summary so far Apr 02 20:42:09 only if you have time, thx Apr 02 20:42:20 tom gave his presentation, there was a lot of discussion amongst the audience about linux-yocto and related stuff, and the ability to use just parts of it Apr 02 20:42:31 now Denys is talking about meta-ti Apr 02 20:44:03 I see, thx. Tom already answered on the ML Apr 02 21:18:57 bluelightning: should I increase PRINC := "${@int(PRINC) + 1}" to +2 if I touch the recipe? Apr 02 21:20:05 sound strange, we'd have a +2 spread with oe-core recipe Apr 02 21:20:37 ant__: technically I suppose yes Apr 02 21:21:17 ok, the .bbappend will disappear anyway once oe-core will accept the patch Apr 02 21:21:22 if Apr 02 21:21:26 heh Apr 02 21:36:04 ok, while you guys were looking at the summit I sent out the 2 udev patches... Apr 02 21:36:41 third will follow for meta-oe once I get comments **** ENDING LOGGING AT Tue Apr 03 02:59:58 2012 **** BEGIN LOGGING AT Tue Apr 03 02:59:58 2012 Apr 03 09:20:32 bluelightning: good morning Apr 03 09:20:40 hi ant_work Apr 03 09:20:48 I may have discovered a Yocto bug last night... Apr 03 09:22:02 iirc I did already see that: in log_do_patch of linux-yocto-tiny I see "grep failed - out of memory". Can't be more precise, data is @home. Apr 03 09:22:22 ouch Apr 03 09:22:26 it's when it verify the patch list Apr 03 09:22:48 well, I would say please report it to the yocto project bugtracker Apr 03 09:22:53 fwiw that recipe seems not working anymore like did previously Apr 03 09:23:08 i.e. it doesn't respect my defconfig Apr 03 09:23:26 and I get a kernel for versatile... Apr 03 09:23:36 surely is my fault or bad setup Apr 03 09:23:44 but the grep error is worth noting Apr 03 09:24:14 not using your defconfig is probably a bug too Apr 03 09:24:34 it's clear in the logs: noo HW configuration found Apr 03 09:25:58 I thought KCONFIG_MODE = "--allnoconfig" would allow that Apr 03 09:34:22 ok, I'll test do_patch () of kernel-yocto.bbclass again on my system. Apr 03 09:34:54 oom issue seems arising during loop Apr 03 09:36:37 heh, seems having to do with scc, maybe is really my fault Apr 03 10:42:26 btw Denys may also be testing with a distro-less configuration Apr 03 15:32:42 about that, I don't see missing preferred providers building the usual core-images (base,core,sato,minimal) Apr 03 15:32:59 that happens adding meta-oe Apr 03 15:59:16 ah.. reading /kernel/v3.0/ChangeLog-3.2.14 Apr 03 15:59:22 backlight: fix typo in tosa_lcd.c Apr 03 16:00:29 couple of ext4: fix race between ... :/ **** ENDING LOGGING AT Wed Apr 04 02:59:58 2012 **** BEGIN LOGGING AT Wed Apr 04 02:59:59 2012 Apr 04 08:09:26 linux-3.4-rc1 PXA2xx PCMCIA possible regression Apr 04 08:09:38 s/possible/confirmed/ Apr 04 08:10:01 no CF Apr 04 08:10:32 morning Apr 04 08:10:34 doh Apr 04 08:10:35 :/ Apr 04 08:10:42 hello Apr 04 08:11:23 there is already a fix, we suffer from follow-up's after big kernel reworks in 3.3 Apr 04 08:16:54 ok, at least there's a fix Apr 04 08:18:09 Paul discovered it on his hx4700 Apr 04 08:18:59 sadly I did not manage to boot with Paul's last patches... those are quite old by now though Apr 04 08:19:28 :/ Apr 04 08:19:44 hence why I did not merge them into meta-handheld Apr 04 08:20:20 about kernels, I'll try to see what's wrong with linux-yocto(-tiny) those days Apr 04 08:28:19 dvhart has not yet opened a bug about http://paste.debian.net/161947 Apr 04 08:28:33 we had a schort chat yesterday, maybe he found smthg Apr 04 08:54:31 I'll restart from /tools/kgit-meta, after echo "[INFO] validating against known patches (`basename $_meta_series`)" Apr 04 23:02:09 bluelightning: can we fix dependency of meta-handheld? Apr 04 23:02:21 ant__: which dependency? Apr 04 23:02:24 layer: meta-initramfs, meta-oe Apr 04 23:02:29 meta-oe can go Apr 04 23:03:19 ant__: are you just talking about the README? Apr 04 23:03:54 yes, dvhart is following it ;) Apr 04 23:04:07 I told him he won't need meta-oe Apr 04 23:05:23 ok, sure I can fix that Apr 04 23:06:04 btw I had to fork the repo temporarly on github Apr 05 00:45:27 ant__: why was that? **** ENDING LOGGING AT Thu Apr 05 02:59:58 2012 **** BEGIN LOGGING AT Thu Apr 05 02:59:58 2012 Apr 05 08:44:14 'morning bluelightning Apr 05 08:51:22 dvhart asked for a meta-handheld repo to pick up the devel patches wrt linux-yocto-tiny Apr 05 09:21:42 hi ant_work Apr 05 09:22:16 ah right Apr 05 09:28:26 he even read the README and added meta-oe ;) Apr 05 09:32:25 ah I forgot to commit that change Apr 05 09:33:50 done and pushed Apr 05 22:43:12 hey bluelightning Apr 05 22:43:48 I'm falling on kb today...pls consider merging the udev refinement Apr 05 22:44:45 atm bad news wrt linux-yocto-tiny, let Bruce do the magic Apr 05 23:23:18 ant__: ah yes... I'll try to do some stuff over the weekend Apr 05 23:47:32 gn **** ENDING LOGGING AT Fri Apr 06 03:00:00 2012 **** BEGIN LOGGING AT Fri Apr 06 03:00:00 2012 Apr 06 10:05:13 dromede: so, 6-700km in one week are a challenge for our postal services :/ Apr 06 10:05:57 even in EU Apr 06 10:14:09 heh Apr 06 10:14:14 it seems so Apr 06 10:14:26 it's not a problem Apr 06 10:14:36 i still have to wait for the usb serial adapter Apr 06 10:14:56 and we're not in EU yet :) Apr 06 10:15:02 Croatia, that is Apr 06 10:15:16 let's hope the cable did not disappear underway in SL Apr 06 10:16:13 ah, I see, Croatia's membership in the Union will be formalised as of 1 July 2013 Apr 06 10:16:23 yep Apr 06 10:16:26 I thought all was already signed/done Apr 06 10:16:53 did you try booting 3.4 yet? Apr 06 10:17:01 no, not yet Apr 06 12:52:42 heh Apr 06 12:52:48 just got the serial cable Apr 06 17:40:28 bluelightning: gotcha Apr 06 17:40:33 http://paste.debian.net/162345/ Apr 06 17:40:37 'yoctized' Apr 06 17:41:39 I'll send you a first patch reusing full defconfig Apr 06 17:42:04 then, as second step, I'll try to tokenize it and create the machine.cfg Apr 06 17:42:56 this means only .bbappend in layer :) Apr 06 17:43:34 no linux.inc even Apr 06 17:46:42 ant__: cool :) Apr 06 17:46:45 ant__: well done Apr 06 17:47:43 I thought I'd have a go at some Opie stuff again Apr 06 17:48:19 I've fixed one bug and am now reworking the Euroconv app into a full currency converter with rates download from online :) Apr 06 17:48:48 nice Apr 06 17:49:07 I'll have to add the same logic to my biz app (php) Apr 06 17:49:16 now are just tables Apr 06 17:49:35 I stumbled across this: http://openexchangerates.org/ Apr 06 17:49:51 this is what I'll be using for the data Apr 06 17:50:47 I've never touched json Apr 06 17:51:21 I'm having hard time to update my LAMP application to modern php, even Apr 06 17:51:36 I've decided I'l llet it in procedural style Apr 06 17:51:51 (with use of some classes) Apr 06 17:53:13 oh, Jay7 is back ... Apr 06 17:53:20 ant__: hi Apr 06 17:54:54 upgrading php is pain :) Apr 06 17:54:58 hi Jay7 Apr 06 17:55:05 Jay7: will you ever find some time to split gui + tui headers and write the fallback code in case both are configured? Apr 06 17:55:06 * Jay7 migrated one site not too log ago Apr 06 17:55:09 bluelightning: hi Apr 06 17:55:18 ant__: yeah... Apr 06 17:55:23 well... Apr 06 17:55:37 * Jay7 scratching head Apr 06 17:55:52 ant__: I'd never used json either, I just found a nice clean C parser (https://github.com/udp/json-parser) and copied it into my code Apr 06 17:56:12 I suspect there are tons of PHP json parsing examples out there Apr 06 17:57:14 this is what the data in json format looks like: http://openexchangerates.org/latest.json Apr 06 17:57:18 pretty straightforward Apr 06 17:58:15 bbl Apr 06 19:38:37 http://elinux.org/CEWG_Open_Project_Proposal_2012 Apr 06 20:15:52 ant__: ping Apr 06 20:37:51 Proposals for the initial 2012 project list will be accepted until April 23, 2012. Apr 06 20:40:48 * Jay7 -> sleep Apr 06 21:13:15 dromede: pong Apr 06 22:34:48 bluelightning: so should we always use PRINC := "${@int(PRINC) + 1}" in the .bbappend(s) ? Apr 06 22:35:02 ant__: technically I suppose yes Apr 06 22:35:13 ant__: it's a pretty awful construct though :/ Apr 06 22:35:29 ant__: I've been thinking about how to possibly improve that at a bitbake level but haven't come up with anything yet Apr 06 22:35:40 hm, even to linux-yocto* kernels .. ? Apr 06 22:36:25 well it all depends on whether you expect multiple levels of bbappends Apr 06 22:36:53 it seems a little superfluous really Apr 06 22:37:05 for kernel can be confusing **** ENDING LOGGING AT Sat Apr 07 02:59:58 2012 **** BEGIN LOGGING AT Sat Apr 07 02:59:59 2012 **** ENDING LOGGING AT Sun Apr 08 03:00:01 2012 **** BEGIN LOGGING AT Sun Apr 08 03:00:01 2012 Apr 08 21:31:34 hi channel Apr 08 21:33:35 hi dromede_ Apr 08 21:36:37 hi ant Apr 08 21:36:45 i got the cable the other day Apr 08 21:37:01 oh, really surprising :p Apr 08 21:37:18 too fast to be true Apr 08 21:37:24 i assembled and old pc that has serial ports Apr 08 21:37:28 heh Apr 08 21:37:36 turned it into a headless ssh/distcc machine Apr 08 21:37:41 cable works Apr 08 21:37:49 \°/ Apr 08 21:38:07 latest kexecboot decompresses normally Apr 08 21:38:12 I'm refactoring the kernels btw Apr 08 21:38:12 and then i get nothing Apr 08 21:38:29 i also managed to boot 3.4-rc2 Apr 08 21:38:35 I'll push very soon kernel built with linux-yocto fragments Apr 08 21:38:50 nice Apr 08 21:39:15 I still don't see what is wrong with the defconfig of akita Apr 08 21:39:17 but Apr 08 21:39:36 being that yocto kernels can't accept(yet) a mini-defconfig Apr 08 21:39:58 I have re-expanded those Apr 08 21:40:09 so we can rediff with working akita Apr 08 21:40:16 3.4-rc2 craps out at sharpsl_save_param() in arch/arm/common/sharpsl_param.c Apr 08 21:40:28 i have no idea why Apr 08 21:40:50 i commented the function out and then it boots Apr 08 21:41:05 I'll start with normal kernel, I have to update klibc otherwise Apr 08 21:41:17 linux-kexecboot will be a bit delayed Apr 08 21:41:42 (klibc_1.5.25 requires <=3.1) Apr 08 21:42:26 i'll deal with kexecboot on akita and general pm issues on 3.2 Apr 08 21:42:36 I've seen there is already a linux-yocto 3.4 Apr 08 21:43:00 we can let yocto devs do the updating Apr 08 21:43:16 once we have the valid config fragments Apr 08 21:43:43 ever tried running kgdb over serial ? Apr 08 21:44:26 no but I think I've read about that Apr 08 21:44:48 i think i'll need that to debug pm issues Apr 08 21:49:14 yes, ideally wid latest kernels Apr 08 21:49:21 *with Apr 08 22:31:58 dromede_: pls test when you'll have time Apr 08 22:33:01 test what? Apr 08 22:33:14 https://github.com/andrea-adami/meta-handheld Apr 08 22:33:39 atm the defconfig is respected Apr 08 22:33:42 100% Apr 08 22:33:58 but it's easy to use the fragments now Apr 08 22:34:41 you'll have to check in the kernel workdir and look in the /meta dir Apr 08 22:35:38 ok, will do Apr 08 22:39:38 the config warnings/logs are e.g. in Apr 08 22:39:40 andrea@mizar /oe/oe-core/build/tmp-eglibc/work/poodle-oe-linux-gnueabi/linux-yoc Apr 08 22:39:40 to-tiny-3.2.11+git1+59f350ec3794e19fa806c1b73749d851f8ebf364_1+ec236058dc254183dbfb3744bf21f110c37af30b-r0/linux/meta/cfg/standard/tiny $ Apr 08 22:43:05 i.e. akita.cfg should be purged Apr 08 22:43:45 akita.cfg should also have max7310 driver enabled Apr 08 22:44:04 just that? Apr 08 22:44:50 i'm not sure yet Apr 08 22:45:46 np, this is the occasion to check an fix it once for all Apr 08 22:59:57 btw, have you read the wiki of flameman wrt akita? Apr 08 23:01:36 iirc he was the last one doing deep testing Apr 08 23:01:40 http://elinux.org/Flameman/zaurus-akita Apr 08 23:01:58 in fact, Maxim MAX7310 gpio expander seems important... Apr 08 23:03:48 yep Apr 08 23:14:19 what was CONFIG_GPIO_PCA953X=y ? Apr 08 23:16:34 that's the max7310 driver Apr 08 23:16:42 ahh Apr 08 23:16:55 ok, I'll redo the config and push again Apr 08 23:22:36 should I add irq support too? Apr 08 23:22:59 CONFIG_GPIO_PCA953X_IRQ Apr 08 23:23:40 I don't know on akita, is *it* the irq controller? Apr 08 23:23:58 i done have CONFIG_GPIO_PCA953X_IRQ enabled in my defconfig Apr 08 23:24:09 s/done/don't/g Apr 08 23:24:39 maybe i should enable it Apr 08 23:24:41 i'll try Apr 08 23:25:02 i'm running your kernel now Apr 08 23:25:48 zImage is pretty big Apr 08 23:25:55 2.1 mb Apr 08 23:26:11 yes, many filesystem Apr 08 23:26:13 s Apr 08 23:26:42 we really need to rmove the userspace frags which are handled by yocto-tiny Apr 08 23:55:48 ok, I pushed a revised patch Apr 09 00:00:40 gn **** ENDING LOGGING AT Mon Apr 09 02:59:58 2012 **** BEGIN LOGGING AT Mon Apr 09 02:59:58 2012 Apr 09 16:18:30 linux-kexecboot latest is now fixed on akita Apr 09 16:18:57 akita's backlight control is hooked to a max7310 expander Apr 09 16:19:18 and the driver for that expander isn't included in the akita defconfig for linux-kexecboot Apr 09 16:20:12 so adding CONFIG_GPIO_PCA953X=y to akita defconfig will make it work normally Apr 09 16:20:35 i would also suggest that we include printk support by default in all defconfig Apr 09 16:20:57 while this means a larger kernel, it also means easier debugging Apr 09 16:21:50 i also would like to see CONFIG_CC_OPTIMIZE_FOR_SIZE=y included by default in all defconfig's Apr 09 16:21:57 this should reduce the kernel size even more Apr 09 21:27:01 hi ant__ Apr 09 21:27:28 just merging your patches... we need a PR bump on these kernel recipes right? Apr 09 21:29:25 yes, but pls wait some mins Apr 09 21:30:01 I'm adding CC_OPTIMIZE_FOR_SIZE which somehow disappeared... Apr 09 21:30:38 and there is one patch to skip: preferred_provider for kernel Apr 09 21:30:50 let us do some more testing Apr 09 21:44:17 bluelightning: almost done Apr 09 21:46:15 ant__: ok, no problem Apr 09 22:05:51 ok, sent Apr 09 22:07:30 bluelightning: you should have 1udev + 5 kernel patches pending Apr 09 22:08:20 ant__: ok, checking Apr 09 22:08:25 I'll update now the repository on github with the yocto patches. please wait with those Apr 09 22:10:14 argh..the akita one did revert the previuous patch.... Apr 09 22:10:24 ant__: er, your optimise for size patches removed the CONFIG_GPIO_PCA953X=y lines Apr 09 22:10:27 for akita defconfigs Apr 09 22:10:32 righto Apr 09 22:10:33 do you want me to fix that? Apr 09 22:10:39 sure, pls do Apr 09 22:10:42 ok Apr 09 22:13:42 I'm working on too many copies of the repo;) Apr 09 22:31:24 ok, pushed Apr 09 22:32:29 great, thx Apr 09 22:49:19 bluelightning: I've sent the big boy right now Apr 09 22:49:32 still OPT IN Apr 09 22:52:40 so you want me to merge this now? Apr 09 22:53:15 yes, just noticing the PR is maybe wrong.. Apr 09 22:54:04 so should we always use PRINC := "${@int(PRINC) + 1}" in the .bbappend(s) Apr 09 22:54:07 ^^ Apr 09 22:54:29 jeez Apr 09 22:54:32 I'm concerned that since this modifies SRC_URI unconditionally it would break other machines when meta-handheld is in the user's layers Apr 09 22:55:19 maybe itμs better letting it sit some more days in the forked repo.... Apr 09 22:55:40 we are not in hurry, recipe needs much care Apr 09 22:55:54 if you don't mind... Apr 09 22:56:09 this is really great work though Apr 09 22:56:10 yes, pls wait, i'ts better so Apr 09 22:56:57 I get a strange ERROR: Failure expanding variable FILESPATH wrt that patch Apr 09 22:57:14 (did work previously) Apr 09 23:03:32 interesting Apr 09 23:03:35 http://paste.debian.net/162676/ Apr 09 23:05:20 it is that damned PR ;) Apr 09 23:05:24 heh Apr 09 23:06:20 patch v2 follows Apr 09 23:07:47 bluelightning: so, add that PRINC or not? Apr 09 23:08:13 ant__: just leave it out for now I think Apr 09 23:08:33 PR -> oeeror, PRINC, all fine but r1 Apr 09 23:08:47 ok, let it r0 Apr 09 23:09:04 JaMa will add it ;) Apr 09 23:09:43 note this recipe is only triggered if REFERRED_PROVIDER_virtual/kernel = "linux-yocto-tiny" Apr 09 23:10:24 so this should not interfere Apr 09 23:22:39 bluelightning: patch v2 tested. can be merged then Apr 09 23:23:09 yeah I get that... but this recipe is used by other machines Apr 09 23:23:20 which will fail because the logo file etc. does not exist for them Apr 09 23:23:23 well, as long as COMPATIBLE_MACHINE is set for Zaurus only... Apr 09 23:23:37 which again will stop that recipe from working for those machines Apr 09 23:23:46 hm, the parsing? Apr 09 23:23:50 that would be fine if this was a separate recipe, but it's a bbappend... Apr 09 23:25:02 I can try with h1940 Apr 09 23:25:59 think machines outside of meta-handheld Apr 09 23:28:02 twell, I see, people inadvertitely including meta-handheld Apr 09 23:28:12 i.e. Angstrom users Apr 09 23:28:24 in fact, using h1940 is flawless Apr 09 23:28:39 while MACHINE ?= "beagleboard" Apr 09 23:28:44 is a disaster Apr 09 23:30:03 http://paste.debian.net/162678/ Apr 09 23:30:24 Please set a valid MACHINE in your local.conf or environment Apr 09 23:35:04 do you have a layer supporting beagleboard in your bblayers.conf? Apr 09 23:35:11 the output is messy though I agree Apr 09 23:35:22 no, just tried one Apr 09 23:36:22 should I try qemuarm? Apr 09 23:43:42 qemuarm is ok Apr 09 23:43:53 is in oe-core, not in our layer Apr 09 23:44:11 hmm Apr 09 23:44:20 how does it cope with the changes to SRC_URI? Apr 09 23:44:46 http://paste.debian.net/162680/ Apr 09 23:46:13 what about qemux86? Apr 09 23:47:06 since the OE-Core version of the recipe does list that in COMPATIBLE_MACHINE Apr 09 23:47:52 same http://paste.debian.net/162681/ Apr 09 23:48:51 BBLAYERS = " \ Apr 09 23:48:51 /oe/oe-core/meta \ Apr 09 23:48:51 /oe/meta-openembedded/meta-initramfs \ Apr 09 23:48:51 /oe/meta-handheld \ Apr 09 23:48:51 " Apr 09 23:49:10 sry Apr 09 23:49:13 ok, so that's the kind of problem I want to avoid Apr 09 23:49:32 i.e. we mask the recipe Apr 09 23:49:45 now I see Apr 09 23:51:02 why should another machine include our BSP though? Apr 09 23:51:55 well, no hurry Apr 09 23:52:04 I'm falling otk Apr 09 23:52:25 see you, gn **** ENDING LOGGING AT Tue Apr 10 03:00:00 2012 **** BEGIN LOGGING AT Tue Apr 10 03:00:00 2012 Apr 10 08:20:58 hi bluelightning Apr 10 08:21:22 I think [Bug 2250] will be closed Apr 10 08:23:15 ok, I'll have a word with darren later on Apr 10 08:24:13 Now, I'm a bit puzzled by that 'grep memory exhausted' which I could track down to a yocto kernel script Apr 10 08:24:18 kgit-meta Apr 10 08:25:34 http://git.yoctoproject.org/cgit/cgit.cgi/yocto-kernel-tools/tree/tools/kgit-meta Apr 10 08:25:38 # 802 Apr 10 08:26:08 if you add an echo before and after ' wait j' you'll see the grep msg Apr 10 08:26:13 is here Apr 10 08:27:20 I observed the system: it did loop 5 times and I could distincly see the memory amount was increasing in big steps (doubling ?) Apr 10 08:27:24 it reached 3,5 GB Apr 10 08:28:07 (the grep pid) Apr 10 08:29:33 this appears in do.patch.log, just doing "bitbake _do_patch linux-yocto-tiny" Apr 10 08:31:23 for poodle fwiw Apr 10 08:36:06 woah Apr 10 08:36:10 3.5gb... Apr 10 08:36:38 in fact the progress bar dalways stops at 80% Apr 10 08:36:48 bah..today I can't type... Apr 10 08:38:04 probably this happens when *not* using a full BSP with patches in the expected place Apr 10 08:38:25 otherwise Bruce and Darren would have noticed it Apr 10 08:41:38 bluelightning: I suspect grep did exhaust the memory trying to double the 3.5GB ... I have 8GB and almost 1 is used by the system during normal bitbake operations Apr 10 08:42:06 well, I have 8 or 9 GB swap too .... Apr 10 08:47:20 ant_work: maybe you're hitting a grep bug... Apr 10 08:51:55 always possible... Apr 10 08:52:16 but low probability ;) Apr 10 08:54:20 you'd think so, but we managed to hit a tar bug not that long ago that was only fixed fairly recently Apr 10 08:54:29 hence why we had to add tar-replacement-native Apr 10 08:54:40 heh Apr 10 08:55:23 the code is not easy to follow: the processes are in background Apr 10 09:00:46 wrap_meta_series()->do_pending_patches () -> background_op_wait 2 progress_update 500 Apr 10 09:01:35 # wait for the push Apr 10 09:01:36 wait $j Apr 10 09:01:39 here Apr 10 09:08:02 dromede: so akita is fixed now, isn't? Apr 10 09:08:51 (I suspect we kept the wrong defconfig by mistake when we did unify spitz-akita) Apr 10 09:09:38 as for CC_OPTIMIZE_FOR_SIZE iirc it was enabled in the 2.6.3x kernels then did somehow disappear... Apr 10 09:09:49 probably my bad :p Apr 10 10:16:24 bluelightning: I was looking at the meta-intel layer and at their .bbappend for linux-yocto Apr 10 10:16:45 well, they just set i.e. COMPATIBLE_MACHINE_crownbay = "crownbay" Apr 10 10:19:03 to keep the recipe simple I could add the OE logo with a patch for each machine but then it will be harder to override (i.e. by distro) Apr 10 10:23:19 ant_work: yes, linux-kexecboot now works on an akita Apr 10 10:23:46 bluelightning: or easier add the logo with a SRC_URI_append for each machine, expliciting vga/qvga in the path Apr 10 10:23:57 dromede: nice Apr 10 10:25:09 dromede: do you feel browsing kexecboot menu with down/up cursors is slow on akita? Apr 10 10:25:16 extremely slow Apr 10 10:25:22 (it is slower on spitz than on c7x0 and poodle) Apr 10 10:25:41 ok, then we need to adjust the event rate Apr 10 10:25:52 i'll take a look at kexecboot code soonish Apr 10 10:26:05 is a configure option fwiw Apr 10 10:27:05 currently, i'm trying to solve kernel hangs during offline charging Apr 10 10:27:59 charger led and charger related gpio's are handled by the scoop device 1 Apr 10 10:28:18 which is of course suspended when offline charging Apr 10 10:29:02 with older kernels we could use --enable-evdev-rate=1000,250 . check recipes-bsp/kexecboot/kexecboot-handhelds.inc Apr 10 10:29:10 ah, charging ... :/ Apr 10 10:29:27 charging is fine now Apr 10 10:29:55 problems arise if you try to remove or plug in the charger during suspend Apr 10 10:29:58 maybe on spitz/akita...I'm unsure about others Apr 10 10:30:30 or if your battery gets charged during suspend Apr 10 10:31:25 i hope i get this fixed today Apr 10 10:31:37 great Apr 10 10:32:20 or maybe i should say hacked Apr 10 10:32:37 there is no easy fix for the current situation with offline charging Apr 10 11:06:25 hm... Apr 10 11:07:04 how is it done by the firmware? when extracting battery and recabling then led stay lit and recharge goes on Apr 10 11:07:36 it's the kernel which seems unable to initialize/set properly Apr 10 13:26:04 bluelightning: aehm.. I forgot KMACHINE_spitz = "spitz" .... Apr 10 13:26:23 patch v3 for this evening... Apr 10 14:20:31 I'll wrap patches+logol in machine-specific overrides so it should be 'transparent' for other machines Apr 10 14:21:03 still don't understand what is the benefit of declaring COMPATIBLE_MACHINE_crownbay = "crownbay" Apr 10 14:40:35 I'm puzzled by that also... Apr 10 14:41:10 unless crownbay override is also set for a machine other than crownbay Apr 10 14:41:28 there is also crownbay-noemgd, maybe that's what it's intended to exclude? Apr 10 14:43:16 nope, doesn't seem like it... Apr 10 14:46:14 bluelightning: we are gathering supervision of zeddii and dvhart for free ;) Apr 10 14:46:30 #2250 Apr 10 14:52:34 anyway, if I remov logo and only let Apr 10 14:52:37 SRC_URI += "\ Apr 10 14:52:37 file://${KMACHINE}-${LINUX_KERNEL_TYPE}.scc Apr 10 14:52:42 " Apr 10 14:53:01 we can reuse it for linux-yocto.bbappend Apr 10 14:53:29 where LINUXKERNEL_TYPE = "standard" Apr 10 14:54:10 or maybe better expand it under SRC_URI_append_$MACHINE Apr 10 14:54:20 and get rid of all non-obvious vars Apr 10 14:54:43 code will not look better afterwards.... Apr 10 14:55:01 but the interferences will be eliminated Apr 10 14:55:32 (in case some crazy guy would stack this layer and use a machine from another) Apr 10 14:56:55 in fact your point was the unconditional SRC_URI += Apr 10 21:52:40 disabling "--enable-evdev-rate=1000,250" had no effect on the speed of menu navigation in kexecboot Apr 10 21:52:52 still very slow Apr 10 21:56:31 iirc those are the defaults...you'll have to change those values Apr 10 21:58:25 ok, i'll try "10,250" Apr 10 21:59:49 ant__: what kernel are you currently running on your spitz? Apr 10 22:00:10 3.2.11 or 3.2.12, have to check Apr 10 22:00:25 default meta-hh defconfig? Apr 10 22:00:32 yes Apr 10 22:01:12 btw I've refreshed linux-yocto-tiny in the experimental branch on github Apr 10 22:01:24 they are working on 3.4, will be committed soon Apr 10 22:01:47 could you try doing the following on your spitz please: Apr 10 22:01:53 make sure it's online charging Apr 10 22:01:56 then suspend Apr 10 22:02:03 then try removing the ac cable Apr 10 22:02:06 then resume Apr 10 22:02:39 ok, let it charge a bit before, it's dry... Apr 10 22:02:53 ok Apr 10 22:03:22 (I hope I have a bootable distro...) Apr 10 22:03:57 the above scenario fails on my akita, that's what i'm trying to fix Apr 10 22:04:04 i just want to be sure it's not just my device Apr 10 22:05:04 when i say fails i mean it wont resume Apr 10 22:13:26 --enable-evdev-rate=10,250 makes kexecboot unusable Apr 10 22:13:52 it also hangs the complete system Apr 10 22:14:08 ...when entering the menu Apr 10 22:37:59 txtui works fine but there's a lot of flicker Apr 10 22:38:24 I'm back Apr 10 22:38:33 so, booting now on spitz Apr 10 22:39:19 dromede_: so, after booting the kernel the led is off Apr 10 22:39:31 try reconnecting the ac cable Apr 10 22:39:34 I extract and reinsert AC then led is on Apr 10 22:39:43 now suspend and remove the cable Apr 10 22:39:48 kernel is 3.2.8 Apr 10 22:39:58 (this image) Apr 10 22:40:35 it does not resume Apr 10 22:40:36 ok Apr 10 22:40:41 will it resume? Apr 10 22:40:56 pull AC in? Apr 10 22:41:23 well, isn't DC btw ;) Apr 10 22:41:44 AC in or out during suspend, and your device should halt Apr 10 22:42:29 pitz was dead Apr 10 22:42:37 had to remove battery Apr 10 22:42:40 yep Apr 10 22:42:50 so it's not just my device Apr 10 22:43:01 then i know what happens and where Apr 10 22:43:06 i just don't know why Apr 10 22:44:04 during suspend, the kernel hangs in spitz_pm.c, spitz_charge() Apr 10 22:44:08 if you let the power cable inserted it does resume Apr 10 22:44:15 yes Apr 10 22:44:18 i know Apr 10 22:44:26 the led is a signal smthg is wrong :/ Apr 10 22:44:58 during suspend, every time you insert or remove the cable, spitz_charge() gets called t configure the charger circuit Apr 10 22:45:07 btw, there is some output, wait a min Apr 10 22:47:00 I have to redirect stderr, mom Apr 10 22:47:43 in a nutshell, it seems the charger code can't talk to scoop device 1 during suspend for some reason Apr 10 22:47:52 that's what i think is wrong Apr 10 22:48:59 seems voltage readings are wrong... Apr 10 22:49:27 spi -108 error, right? Apr 10 22:51:49 I can't log those... hm... Apr 10 22:54:53 voltage readings are a different problem Apr 10 22:55:26 well, now it did resume w/out AC cable Apr 10 22:55:40 again Apr 10 22:56:39 was the ac led on before you went to suspend? Apr 10 22:57:02 seems not Apr 10 22:57:11 remember, if you boot with the ac cable plugged in, the led will light up and then die Apr 10 22:57:15 I'm rebootiung, try to get those msgs Apr 10 22:57:16 you have to reinsert the cable Apr 10 23:00:12 see.. booted without AC cable it suspends/resume Apr 10 23:00:53 yes, that works, i know Apr 10 23:02:01 it will only fail if you insert the cable while suspended or if you remove the cable during suspend Apr 10 23:02:33 yes, seems confirmed Apr 10 23:03:13 i'll post a lengthy explanation on arm-linux, see if i can get some help Apr 10 23:03:46 btw, --enable-bg-buffer in kexecboot makes a world of difference Apr 10 23:03:50 I'll check on poodle (also pxa) Apr 10 23:04:25 you see, thinbgs have changed very much after 2.6.3 with the new kernel drivers... Apr 10 23:04:35 2.6.3* Apr 10 23:05:31 as far as suspend/resume/charging goes, thing have went downhill since 2.6.26 Apr 10 23:05:33 for kexecboot, those parameters were ok with older 2.6 kernels, even for spitz Apr 10 23:05:45 also this... Apr 10 23:06:11 is probably the spi Apr 10 23:06:33 you mean the pm bug? Apr 10 23:06:47 no, the kb response and kexecboot menu Apr 10 23:07:54 curiosly for c7xo (corgi) and poodle (pxa25x) this did not change so much Apr 10 23:08:27 so we kept the 2.6 kexecboot configuration Apr 10 23:08:35 well, the corgi has the w100 Apr 10 23:08:47 and the poodle is only qvga, right? Apr 10 23:08:54 but poodle is pxafb Apr 10 23:08:59 yes, qvga Apr 10 23:09:18 that's still 4x less pixels to push Apr 10 23:09:30 actually 3x **** ENDING LOGGING AT Wed Apr 11 02:59:58 2012 **** BEGIN LOGGING AT Wed Apr 11 02:59:58 2012 Apr 11 15:31:20 [03:03] btw, --enable-bg-buffer in kexecboot makes a world of difference Apr 11 15:31:28 dromede: hi Apr 11 15:31:58 dromede: is it helps with menu browsing speed? Apr 11 15:33:06 it helps alot, yes Apr 11 15:36:02 hi Jay7, dromede Apr 11 15:36:12 hi ant Apr 11 15:38:24 hm.. Apr 11 15:38:41 then I would say problem is our drawing code Apr 11 15:38:51 as I suspected before Apr 11 15:39:10 we added lot of eye-candy things but price is speed now.. Apr 11 15:39:19 what exactly does enable-bg-buffer do? Apr 11 15:39:29 btw, CELF proposal deadline is 19 Apr iirc Apr 11 15:39:46 dromede: it save pre-drawed background in ram Apr 11 15:40:20 then it need to redraw only menu items Apr 11 15:40:40 what about screen rotation, how is that handled? Apr 11 15:40:54 again, with 2.6.26 kernel is as fast as before Apr 11 15:40:56 by different line drawing functions Apr 11 15:40:58 on spitz Apr 11 15:41:14 imho is a kernel/driver issue, not kexecboot Apr 11 15:41:50 new ts driver, new kb driver, .... Apr 11 15:41:53 there is place for improvements in drawing code too.. Apr 11 15:42:38 ant_work: was there any fb driver changes? Apr 11 15:43:20 iirc, there where Apr 11 15:43:26 maybe, not evident on pxafb of poodle (qvga) Apr 11 15:43:35 changes in how fb memory is allocated Apr 11 15:43:50 hm.. it may affect us too Apr 11 15:44:41 it looks like is pxa270 problem Apr 11 15:44:50 btw, I've changed ram-to-fb copying routine to use 32-bit words (was 16-bit before) Apr 11 15:45:05 ~at same time Apr 11 15:54:29 so the kexecboot menu is drawn to an offscreen buffer and then copied to fb? Apr 11 16:15:10 dromede: I've sent you a movie very explicative Apr 11 16:18:41 bbl Apr 11 16:36:44 dromede: yes Apr 11 16:37:03 + bg-buffer buffering predrawed background Apr 11 16:37:31 hm.. but then problem is RAM/CPU Apr 11 16:37:43 because we are drawing in RAM **** ENDING LOGGING AT Thu Apr 12 02:59:58 2012 **** BEGIN LOGGING AT Thu Apr 12 02:59:58 2012 Apr 12 14:32:08 hello, good afternoon. i'd like to build the latest kexecboot to my c7x0. Apr 12 14:32:28 i've configure the openembedded environment, Apr 12 14:32:44 and it's currently bitbaking virtual/kernel Apr 12 14:33:04 ah, ok, I've read it a bit too late :/ Apr 12 14:33:21 how do i ask it to bitbake kexecboot? Apr 12 14:33:24 you'd use kernel 3.1 for building klibc 1.5.25 Apr 12 14:33:36 just bitbake linux-kexecboot Apr 12 14:34:05 ok.. you've helped me doing it before, but it's been almost an year and i had already forgotten :-) Apr 12 14:34:13 you ar eusing the new oe-corebuildsystem? Apr 12 14:34:54 probably not :-( i went to angrstrom website, and followed the instructions: http://www.angstrom-distribution.org/building-angstrom Apr 12 14:35:14 it seems everything is a bit out of date in angstrom webpage, so i'm probably missing something Apr 12 14:35:20 ok, this will build a working linux-kexecboot Apr 12 14:35:46 with the new layer-based OE you'd have that klibc issue Apr 12 14:36:18 ok.. and the version of kexecboot that i'm about to build, does it have the debug option without requiring the serial cable? Apr 12 14:36:18 (klibc update was promised mid-end march but still...) Apr 12 14:36:36 enough to check the sd/cf, yes Apr 12 14:37:28 i don't know if you remember, but kernels that are built for my c7x0 never boot (it becomes stalled after pressing ok in kexecboot). so i'll have to change the kernel parameters and compile it, as before, i think. Apr 12 14:37:47 and last time i had to find a serial cable, thats why i asked :-) Apr 12 14:38:00 for c7x0 it should just work Apr 12 14:38:21 I left OE-classic one year ago so can't say today Apr 12 14:39:03 i'll try it. maybe with the 3.0 kernel things will be different already. Apr 12 14:39:16 thank you very much! Apr 12 14:39:20 I've booted 3.2.12 Apr 12 14:40:24 bluelightning: maybe we should sum-up the instructions to build for handhelds somewhere Apr 12 14:40:51 ant_work: yes, we definitely should Apr 12 14:41:42 there are benifits in using the newest model? the oe-core? Apr 12 14:42:16 yes, it's the only on ecurrently used to develop for Zaurus Apr 12 14:42:47 the only one actively maintained Apr 12 14:43:27 ok, i've just found instructions on openembedded.org, i'll try to set it up as well Apr 12 14:43:39 bluelightning: you don't have rights to modify http://www.openembedded.org/wiki/Main_Page isn't? Apr 12 14:44:12 ant_work: I think I do now yes Apr 12 14:44:15 well, I do have it seems ... Apr 12 14:44:47 it's 'just' ignoring oe-core one year after Apr 12 14:44:52 I just haven't had the time to do something about the state of the wiki Apr 12 14:45:07 changes need to be approved I think Apr 12 14:45:27 yes... there will be more time for this after 1.2 Apr 12 14:45:34 (for me at least) Apr 12 14:45:44 I'd start grouping the links to layers, oe-core, yocto manual, ... Apr 12 14:46:49 well, the links are mostly in place Apr 12 14:47:42 maybe instead of a note in Getting_started about the new OE I'd link OpenEmbedded-Core page Apr 12 14:48:03 you don't happen to have a binary of kexecboot around do you? :-) seems that the first build will take some time.. Apr 12 14:48:16 hm, maybe in Gmail, let see Apr 12 14:48:49 its for c700. Apr 12 14:51:19 no.. I'l look this evening then Apr 12 14:52:56 ok, thank you very much. Apr 12 15:04:43 one more thing, the appropriate machine selection for oe-core is still C7x0 ? i wish i knew where to lookup these things by my own.. Apr 12 22:53:02 bluelightning: http://paste.debian.net/163070/ Apr 12 22:54:03 seems we can move to linux-yocto-tiny or even to full linux-yocto Apr 12 22:58:26 ant__: cool :) Apr 12 22:58:58 presumably since you're using the defconfig it doesn't make any difference which you use as a base? Apr 12 22:59:32 I'm thinking I could put all in /linux-yocto and reuse just one .scc each machine Apr 12 23:00:12 i.e. LINUX_KERNEL_TYPE ?= "standard" vs. "tiny" Apr 12 23:00:12 Apr 12 23:00:51 or even LINUX_KERNEL_TYPE = "preempt-rt" Apr 12 23:00:51 ;) Apr 12 23:01:17 (prolly won't work offhand) Apr 12 23:20:46 gn **** ENDING LOGGING AT Fri Apr 13 02:59:58 2012 **** BEGIN LOGGING AT Fri Apr 13 02:59:59 2012 Apr 13 10:07:15 bluelightning: ping Apr 13 10:07:23 hi ant_work Apr 13 10:07:27 hello Apr 13 10:08:25 I'm asking your opinion about klibc: the release is delayed and we cannot build the actual 1.5.25 from meta-initramfs Apr 13 10:08:46 now, what if I add a _git version in meta-handheld? Apr 13 10:09:23 I fear this would be masked by the one in meta-initramfs Apr 13 10:09:58 I know I could reorder the layers but... Apr 13 10:10:55 hm...or maybe overpatch the recipe in meta-handheld Apr 13 10:12:54 I would prefer to have only stable releases in meta-initramfs, not git tags Apr 13 10:18:50 ant_work: is 1.5.25 broken in some way? Apr 13 10:20:36 well, after kernel 3.1.x Apr 13 10:20:50 ..and we already moved to 3.2 Apr 13 10:21:29 in fact there is a more deep issue.... Apr 13 10:22:07 klibc depends on virtual/kernel but maybe it could be built with dev headers instead Apr 13 10:22:44 is that an improvement or a bugfix though? Apr 13 10:23:05 different headers Apr 13 10:23:06 ah right would help with avoiding binding to specific kernels Apr 13 10:23:10 yep Apr 13 10:23:18 I'll have to ask khem, last time we talked a kernel ' make headers_install' was necessary to build klibc Apr 13 10:23:30 btw this makes klibc machine-specific :/ Apr 13 10:23:39 why not have a git recipe in meta-initramfs with D_P = -1 and then use PREFERRED_VERSION in meta-handheld? Apr 13 10:24:41 have you seen how split is the recipe ? Apr 13 10:25:04 I haven't no Apr 13 10:25:33 don't do ;) you'll be scared Apr 13 10:26:00 see, I'll try to add a .bbappend and some patches in meta-handheld Apr 13 10:26:11 maybe it's enough temporarly Apr 13 10:27:26 * ant_work wil now start boring people in #klibc Apr 13 10:27:55 heh **** ENDING LOGGING AT Sat Apr 14 02:59:58 2012 **** BEGIN LOGGING AT Sat Apr 14 02:59:58 2012 Apr 14 21:00:25 bluelightning: ping Apr 14 22:14:27 hi ant__ Apr 14 22:15:36 hello Apr 14 22:16:21 I have a strange issue which seems unrelated to the metadata Apr 14 22:17:07 pls. see https://github.com/andrea-adami/meta-handheld Apr 14 22:17:37 what's the issue? Apr 14 22:18:07 now, trying to build linux-yocto-tiny I think bitbake will get the poodle.scc in /linux-yocto-tiny/poodle/ isn't? Apr 14 22:18:23 and the poodle .cfg + patches from Apr 14 22:18:31 /linux-yocto/ dir Apr 14 22:18:54 FILESEXTRAPATHS_prepend := "${THISDIR}/${PN}:${THISDIR}/linux-yocto:${THISDIR}/files:" Apr 14 22:18:56 ^ Apr 14 22:19:23 In fact, it seems to fetch correctly the expected files Apr 14 22:19:55 but... Apr 14 22:20:05 in log of kernel_configcheck I see Apr 14 22:20:07 cat: /oe/oe-core/build/tmp-eglibc/work/poodle-oe-linux-gnueabi/linux-yocto-tiny-3.2.11+git1+135c75bf9615334b5b8bb9108d612fe7dfbdb901_1+ec236058dc254183dbfb3744bf21f110c37af30b-r0/linux/meta/cfg/poodle.cfg: No such file or directory Apr 14 22:20:19 and the config is ignored :/ Apr 14 22:20:29 seems more a bug for Bruce Apr 14 22:21:12 yeah, I guess so Apr 14 22:21:28 what do you think about this layout ? Apr 14 22:21:37 should only impact Zaurus Apr 14 22:22:15 and we get the -tiny almost for free Apr 14 22:23:11 well, no Apr 14 22:23:22 KERNEL_FEATURES = "" Apr 14 22:23:22 KCONFIG_MODE="--alldefconfig" Apr 14 22:23:29 this will changed to Apr 14 22:23:41 KERNEL_FEATURES_akita and so on Apr 14 22:24:18 then inheriting the recipe will be transparent for other machines Apr 14 22:32:31 pushed Apr 14 22:37:10 sounds good Apr 14 22:37:33 unfortunately I'm trying to sort out some problems for 1.2 atm so I can't really look at it too closely, sorry :/ Apr 14 22:37:36 hm, now I still have Apr 14 22:37:37 ERROR: Multiple .bb files are due to be built which each provide virtual/kernel (/oe/oe-core/meta/recipes-kernel/linux/linux-yocto_3.2.bb /oe/oe-core/meta/recipes-kernel/linux/linux-yocto-tiny_3.2.bb). Apr 14 22:37:37 This usually means one provides something the other doesn't and should. Apr 14 22:38:13 why that? I set D_P -1 Apr 14 22:40:24 (this only for testing) Apr 14 22:49:24 well, at the moment we could live with linux-yocto only Apr 14 22:49:49 in fact linux-yocto-tiny would be for linux-kexecboot **** ENDING LOGGING AT Sun Apr 15 02:59:59 2012 **** BEGIN LOGGING AT Sun Apr 15 03:00:00 2012 Apr 15 22:14:04 bluelightning: ping **** ENDING LOGGING AT Mon Apr 16 02:59:58 2012 **** BEGIN LOGGING AT Mon Apr 16 02:59:59 2012 **** ENDING LOGGING AT Tue Apr 17 02:59:59 2012 **** BEGIN LOGGING AT Tue Apr 17 03:00:00 2012 Apr 17 09:17:00 hi bluelightning Apr 17 09:17:09 squashed out the last bugs? Apr 17 09:17:46 s/last/last known/ heh Apr 17 13:05:38 ant_work: still working on a few last-minute ones Apr 17 13:07:41 I have founs a couple of oddities in the linux-yocto recipes but I'll keep those fo 1.3 ;) Apr 17 13:12:51 probably best at this point unless they are showstoppers Apr 17 14:10:41 yes, as always I've hit some horrible corner cases Apr 17 14:11:10 (zeddii words) Apr 17 14:23:07 bluelightning: I have only a last doubt, wrt order of variable evaluation in case of .bbappend Apr 17 14:24:20 ant_work: what specifically? Apr 17 14:24:22 i.e. the linux-yocto recipe sets KERNEL_FEATURES= Apr 17 14:24:29 KERNEL_FEATURES_append=" features/taskstats" Apr 17 14:24:37 and. i.e. KERNEL_FEATURES_append_qemux86=" cfg/sound" Apr 17 14:24:52 in our .bbappend I did Apr 17 14:24:54 KERNEL_FEATURES_akita = "" Apr 17 14:25:14 now, I'm dubious about the _append Apr 17 14:25:45 I'm not sure myself... Apr 17 14:26:02 ideally we should avoid having ambiguities like that Apr 17 14:26:17 in the case of linux-yocto-tiny it's easy: the recipe just sets KERNEL_FEATURES="" Apr 17 14:27:31 no, I'll have to bother Bruce, for those can be added here in the recipe or (probablythe preferred way) in the .scc Apr 17 14:27:37 s/no/now/ Apr 17 14:28:49 i.e. KERNEL_FEATURES_append_qemux86-64=" cfg/sound" Apr 17 14:28:52 can be expressed Apr 17 14:29:04 include cfg/sound.scc Apr 17 14:29:06 afaik Apr 17 22:18:24 bluelightning: still on duty? Apr 17 22:18:42 ant__: yes :( Apr 17 22:18:47 :/ Apr 17 22:19:26 look, I'm doing unhortodoxes things Apr 17 22:19:40 linux-yocto-kexecboot_3.2.bb Apr 17 22:19:40 require recipes-kernel/linux/linux-yocto_3.2.bb Apr 17 22:19:40 INITRAMFS_IMAGE = "initramfs-kexecboot-klibc-image" Apr 17 22:19:50 EOF Apr 17 22:19:53 two lines Apr 17 22:20:12 is it ok to require another .bb ? Apr 17 22:21:08 this happens to work ;) Apr 17 22:21:09 now, I'm trying to include a .bbbappend but this seems impossible Apr 17 22:23:52 I think that's fine Apr 17 22:24:07 but yes, you lose any bbappends Apr 17 22:24:18 I think that's probably not an issue though? Apr 17 22:25:27 I'm trying to avoid to repeat the SRC_URI_zaurus and so on defined in linux-yocto_3.2.bbappend Apr 17 22:25:47 yes, it is not a real / blocker issue, Apr 17 22:26:59 well, I can put the common BSP stuff in an .inc Apr 17 22:34:30 sounds like that might be the way to go Apr 17 23:02:13 well, it seems.. just, no warnings if the 'include' is missing Apr 17 23:02:26 while Apr 17 23:02:27 ERROR: ParseError at /oe/meta-handheld/recipes-kernel/linux/linux-yocto-tiny_3.2.bbappend:5: Could not include required file zaurus-yocto-tiny.inc Apr 17 23:02:34 with require Apr 17 23:02:53 I think a Warning should be spit out Apr 17 23:03:12 not an error, just a warning Apr 17 23:03:27 in case of missing includefile Apr 17 23:03:58 ...so you see it at parsetime and bot after configure ;) Apr 17 23:04:17 and *not* after configure or even build... Apr 17 23:04:24 :p Apr 17 23:07:34 ah, use require not include Apr 17 23:08:26 the thing is, include is already used internally where we do not want an error if the file doesn't exist Apr 17 23:08:46 well, when I say internally, I mean OE's bitbake.conf Apr 17 23:09:59 I see Apr 17 23:10:20 if I use the full path I'm on the safe side, isn't? Apr 17 23:10:28 i.e. require recipes-kernel/linux/zaurus-yocto-standard.inc Apr 17 23:10:46 and not ./zaurus-yocto-standard.inc Apr 17 23:14:10 if the recipe is in the same directory you don't gain anything by doing that **** ENDING LOGGING AT Tue Apr 17 23:46:14 2012 **** BEGIN LOGGING AT Tue Apr 17 23:48:17 2012 **** ENDING LOGGING AT Wed Apr 18 02:59:58 2012 **** BEGIN LOGGING AT Wed Apr 18 02:59:59 2012 Apr 18 07:04:43 morning Apr 18 07:05:06 * Jay7 is trying to collect features to propose for CELF sponsorship Apr 18 07:05:16 s/CELF/CEWG/ Apr 18 08:00:12 hi ant_work Apr 18 08:00:26 hello Yuri Apr 18 08:07:55 Jay7: about features, I don't know if it can count but I've almost finished the linux-kexecboot recipe using yocto kct (kernel configuration tools) Apr 18 08:08:42 and now CELF is Linux Foundation Apr 18 08:08:57 -> Tocto Apr 18 08:09:01 *Yocto ;) Apr 18 08:09:49 how ubifs support is progressing in klibc? Apr 18 08:10:07 we can even (ab)use the linux-yocto-tiny so we meet the 'tiny' project too Apr 18 08:10:24 hm..klibc progress is slow Apr 18 08:19:17 well.. I mean mostly kexecboot features to implement Apr 18 08:19:24 e.g. boot methods Apr 18 08:19:52 further gui optimization and enchancement Apr 18 08:19:57 sure, but you can wrap it in a new shiny 'Yoctoproject' bag Apr 18 08:20:15 may be import some code from qi bootmenu to use efl Apr 18 08:20:30 I don't see that as urgent Apr 18 08:20:39 the ui is nice and light Apr 18 08:21:10 touchscreen support is wanted too :) Apr 18 08:21:23 there are still issues on qvga with lines too long Apr 18 08:21:43 (in debug menu) Apr 18 08:23:35 I think you should apply some minor fixes (headers, falback to textui if both compiled) and release 0.6 Apr 18 08:24:06 then yes, I can try to bother in #klibc Apr 18 08:29:21 Jay7: the code here https://gitorious.org/util-linux-ng/util-linux-ng/blobs/master/libblkid/src/superblocks/ubifs.c Apr 18 08:29:36 "Every UBIFS node starts with this common part. " Apr 18 08:30:01 suggests we can hack and read the superblock like done for jffs2 Apr 18 08:30:36 I'm tempted by libblkid but we only miss a little piece in klibc... Apr 18 08:31:11 blkid must be in klibc now IMHO Apr 18 08:31:25 almost all things are using UUID's now Apr 18 08:34:48 well, I'd try { .magic = "\x31\x18\x10\x06", .len = 4 }, Apr 18 08:41:45 in kernel, ubifs-media.h declares Apr 18 08:41:48 37 #define UBIFS_NODE_MAGIC 0x06101831 Apr 18 08:48:53 ok, I've asked again in #klibc Apr 18 12:55:53 Jay7: probably we should move away from linuxtogo, Apr 18 12:56:05 ah Apr 18 12:56:07 and put src in github Apr 18 12:56:10 may be github Apr 18 12:56:15 yes :) Apr 18 12:56:19 look http://labs.kernelconcepts.de/Projects/LinuxToGo1/ Apr 18 12:56:29 I didn't know we were in a matrix :) Apr 18 12:57:02 well, it's all florian I think Apr 18 13:02:15 Jay7: where should we host th eprecompiled images? How much space is in kexecboot.org? Apr 18 13:02:32 some gigabytes iirc Apr 18 13:02:43 we may host there Apr 18 13:04:20 I'll resent the header patch to the mailing list just to see if it is still working Apr 18 13:04:46 ML's are working Apr 18 13:04:49 ok Apr 18 13:04:59 iirc, angstrom and oe's are hosted by ltg too Apr 18 13:05:26 but we may switch to google groups Apr 18 14:57:26 bbl **** ENDING LOGGING AT Thu Apr 19 02:59:59 2012 **** BEGIN LOGGING AT Thu Apr 19 02:59:59 2012 Apr 19 07:34:04 well, linux-yocto-tiny-kexecboot_3.2 does boot on c7x0 Apr 19 07:34:10 but... Apr 19 07:34:28 stalls on logo Apr 19 07:35:10 so, 2 possible causes: 1) somehow lacks the initramfs, 2) toolchain is broken :/ Apr 19 08:05:18 hi ant_work Apr 19 08:05:29 .. and others :) Apr 19 21:57:06 bluelightning: mystery solved, now linux-yocto-tiny-kexecboot does boot! Apr 19 21:57:18 ant__: what was the issue? Apr 19 21:57:29 [PATCH 1/3] check console device file on fs when booting Apr 19 21:57:50 in the case of initramfs, devtmpfs must be mounted by hand Apr 19 21:58:26 probably this patch pre-dates devtmpfs Apr 19 21:58:55 phew... Apr 19 21:59:06 I can finish the remaining configs now Apr 19 21:59:35 my program is to get rid of most kernels and possibly keep only linux-yocto Apr 19 21:59:39 species Apr 19 22:00:24 cool :) Apr 19 22:00:47 should be doable for ipaqs too Apr 19 22:01:09 I'l lpush soon the new meta-handheld so you'll see Apr 19 22:01:23 I'll be interested to check it out Apr 19 22:01:48 I need some opinions.... Apr 19 22:01:52 :) Apr 19 22:05:05 I'm sorry I haven't had much of a chance to help you out with this Apr 19 22:05:21 been so busy lately :( Apr 19 22:05:42 keep on going, only a few days left... Apr 19 22:06:22 1.2 release is next week, isn't? Apr 19 22:08:41 bluelightning: first task for you would be to clean-out the initramfs images from the new stuff which is bloating Apr 19 22:09:23 we get i.e. /lib/modules /usr/share/run-postinst Apr 19 22:09:51 /etc/rcS.d and many liitle crumbs of cruft Apr 19 22:10:34 (I'm not up-to-date with the newest knobs for populating images) Apr 19 22:41:02 hmm ok, we should definitely fix that Apr 19 23:15:53 ok, I'll have to finish gathering the defconfigs... probably tomorrow I'll refresh the repo Apr 19 23:15:57 see you Apr 19 23:15:59 gn **** ENDING LOGGING AT Fri Apr 20 02:59:58 2012 **** BEGIN LOGGING AT Fri Apr 20 02:59:58 2012 Apr 20 09:51:21 bluelightning: ok, koen committed the meta-initramfs part Apr 20 09:55:13 ant_work: ok, so are your meta-handheld changes ready for merging as well? Apr 20 09:55:50 I think this evening I0ll finish Apr 20 09:56:22 I'll resend the patch to the ML once done Apr 20 09:56:44 or a pull request if you prefer Apr 20 10:07:48 ant_work: if it's just a couple of patches, whichever you prefer; but for more than a few a pull request is easier to deal with Apr 20 10:14:40 logically are 3 patches, linux-yocto, linux-yocto-tiny, linux-yocto-tiny-kexecboot Apr 20 10:14:50 the latter has really a long name..ideas? Apr 20 10:25:17 so are the tiny recipes actually giving us anything right now? Apr 20 10:25:32 since the defconfig is still being used Apr 20 11:09:19 switching to yocto-based kernels gives us 2 immediate benefits tough Apr 20 11:09:54 1) the kernel updates 2) the .cfg is somehow sanitized Apr 20 11:11:13 at the moment I plan to use the same defconfigs (renamed $machine.cfg) for linux-yocto and linux-yocto-tiny and one other for linux-yocto-tiny-kexecboot Apr 20 11:12:20 further improvement will be sculpting the .cfg to the bone Apr 20 11:12:49 adn readd the requested features using the fragment/feaures logic of Yocto Apr 20 11:14:07 actually there are two or three ways to accomplish this, because the kernel tools are just interfaced with bitbake Apr 20 11:14:37 in fact it's a bit counter-intuitive for an oe developer Apr 20 11:16:31 especially wrt fetching things not listed in SRC_URI Apr 20 11:17:12 ant_work: I get all that, I'm asking specifically about the tiny recipes Apr 20 11:17:44 tiny is a better startpoint for linux-kexecboot imho Apr 20 11:17:56 it uses --allnoconfig by default Apr 20 11:18:45 ant_work: but right now if you're supplying a defconfig does that make any difference? Apr 20 11:19:56 if it's the same defconfig no, kernel size is the same Apr 20 11:20:14 ok Apr 20 11:20:22 but the KERNEL_FATURES added by linux-yocto Apr 20 11:21:30 in which case I'd suggest not including the linux-yocto-tiny recipes yet, just linux-yocto... that way people will not build them thinking they will be able to get any benefit Apr 20 11:21:59 once the config is broken up into fragments, that's the right time to add them IMHO Apr 20 11:22:28 I agree the linux-yocto recipes are worth adding right now though Apr 20 11:23:43 ehm.. we have linux-yocto-tiny-kexecboot in meta-initramfs Apr 20 11:26:05 right, but there's no linux-yocto there... Apr 20 11:27:03 I'm suggesting, don't put both recipes in if they produce the same result when users will expect by building the tiny version that they will get a smaller kernel Apr 20 11:28:53 I se, but note in the last version I have, the yocto-tiny is a bit smaller Apr 20 11:29:18 how much is a bit? Apr 20 11:29:49 atm Apr 20 11:29:49 KERNEL_FEATURES="features/netfilter" -> 200kb iirc Apr 20 11:30:12 + KERNEL_FEATURES_append=" features/taskstats" Apr 20 11:30:35 while yocto-tiny has KERNEL_FEATURES = "" and all is defined in the .scc Apr 20 11:31:23 and ehm, I still have to understand how the modules are managed Apr 20 11:31:43 well, OK... it's not ideal, but the plan is to improve these, so let's run with it Apr 20 11:31:43 it looks like the .cfg only add CONFIG_XY=y Apr 20 11:32:56 I'm confident that reorganizing the fragments will be rather quicktask Apr 20 11:33:40 unfortunately the kernel tools 1.2 cannot (yet) swallow the mini-defconfigs we have, the work would be already done Apr 20 11:34:07 what goes wrong there atm? Apr 20 11:35:02 afaik only the kernel Kconfig has knowledge about how to rebuild a full defconfig Apr 20 11:35:37 so it would need a 'make oldconfig' just at the beginning Apr 20 11:36:16 otherwise atm the kernel tools screw and apply a standard arm-versatile config Apr 20 11:36:35 zeddii may add that for 1.3 Apr 20 11:36:57 ...once my work will be done :p Apr 20 11:37:23 as for adding the ipaq's: Apr 20 11:38:00 you'll see I have split (in the latest version) the BSP SRC_URI $ co machine-overrides in two files Apr 20 11:38:21 zaurus-yocto.inc and zaurus-yocto-tiny.inc Apr 20 11:38:53 so it's just mater of cloning those Apr 20 11:39:15 ah, btw Apr 20 11:39:43 could you pls search in the Yocto/Poky ML and grep for 'bootlogo' ? Apr 20 11:40:15 atm I use a do_config_prepend task but zeddiii hinted this was discussed somewhere Apr 20 11:40:20 maybe there is a knob Apr 20 11:40:46 I'm more and more tempted to add the logos as kernel patches Apr 20 11:40:58 oe-qvga-logo and oe-vga-logo Apr 20 11:41:15 ok I'll see if I can find the discussion Apr 20 11:41:25 yes that might be an easier solution Apr 20 11:47:14 most google matches are about psplash... Apr 20 13:38:31 bluelightning: btw, reading the ML, COMPATIBLE_MACHINE = or COMPATIBLE_MACHINE += ? Apr 20 13:39:23 in the .bbappend Apr 20 13:46:23 well ideally you'd want to avoid disturbing the values already there... but I'm not sure how to do that in the regex style we use at the moment Apr 20 14:07:04 heh, we are up to 3.2.15 this week Apr 20 14:07:14 and stable: 3.1.10 (EOL) Apr 20 14:07:26 we'd remove this one soon Apr 20 14:07:52 (and the git recipes too) Apr 20 14:09:12 see, we have 99% probs the kernel is stable enough for our arm devices once tested by Yocto-builder for beagleboard Apr 20 14:09:31 zeddii is already working on 3.4 Apr 20 14:11:28 yeah, it's helpful to be able to take advantage of the work bruce & co do on updating/testing/stabilising the kernel recipes Apr 20 15:13:04 ah, seems we can use COMPATIBLE_MACHINE .= "akita|c7x0|...." Apr 20 15:17:00 .= "| even **** ENDING LOGGING AT Fri Apr 20 18:23:12 2012 **** BEGIN LOGGING AT Fri Apr 20 18:24:16 2012 **** ENDING LOGGING AT Fri Apr 20 18:25:00 2012 **** BEGIN LOGGING AT Fri Apr 20 18:26:19 2012 Apr 20 20:47:07 hi bluelightning Apr 20 21:01:48 I'm preparing the patch for logo, then I refresh again the forked meta-handheld Apr 20 23:40:38 done **** ENDING LOGGING AT Sat Apr 21 02:59:59 2012 **** BEGIN LOGGING AT Sat Apr 21 03:00:00 2012 **** ENDING LOGGING AT Sun Apr 22 03:00:00 2012 **** BEGIN LOGGING AT Sun Apr 22 03:00:01 2012 Apr 22 12:07:23 Jay7: ping Apr 22 12:21:07 Jay7: I've uploaded some images in my profile @ kexecboot.org Apr 22 12:21:38 look and think how to make those kernels available for download Apr 22 12:22:36 bbl Apr 22 19:41:35 ant: I've moved uploaded files to kexecboot folder Apr 22 19:41:56 I think we may just link files by hands Apr 22 19:43:15 i.e. edit Download page and create links Apr 22 21:03:53 ant__: hi Apr 22 21:04:02 hello Apr 22 21:04:10 I'm preparing CEWG project proposal Apr 22 21:04:23 do you know dromede's email? Apr 22 21:04:27 yes Apr 22 21:04:42 I'll add him as possible contractors too Apr 22 21:05:46 is dromede at gmail.com Apr 22 21:06:25 thanks Apr 22 21:07:59 as you've seen I've embedded kexecboot in the linux-yocto-tiny kernel Apr 22 21:08:55 I've tested it on poodle, c7x0 and spitz Apr 22 21:09:03 time to commit it Apr 22 21:09:08 bluelightning: ^^ Apr 22 21:09:23 about files on site - just add links to Download section Apr 22 21:09:36 I've moved files into kexecboot folder Apr 22 21:10:32 well this evening I'll work around klibc Apr 22 21:10:53 I'll try to add the ubifs hackish routines Apr 22 21:11:34 ah, I forgot to change evrate on spitz.... Apr 22 21:11:47 is rather slow but working Apr 22 21:12:45 ant__: ah.. I need dromedes name too :) Apr 22 21:12:50 first/last :) Apr 22 21:13:46 Marko KatiΔ‡ dromede@gmail.com Apr 22 21:14:03 (from zaurus-devel public ML ;) Apr 22 21:14:17 hehe.. didn't thought about it Apr 22 21:18:43 hi guys Apr 22 21:18:59 ant__: so I should merge from your fork? Apr 22 21:19:03 hi Paul Apr 22 21:19:23 I can send the patch , just wanting to ask one thing Apr 22 21:20:01 do you think we can switch virtual/kernel to linux-yocto for Z? Apr 22 21:20:26 for linux-kexecboot yes, linux-yocto-tiny-kexecboot is equivalent Apr 22 21:20:48 for normal kernel I'd say yes as well... Apr 22 21:24:21 bluelightning: one more thing: PRINC ? Apr 22 21:52:08 ant__: I've sent proposal to you Apr 22 21:52:14 check it please Apr 22 21:53:07 sure Apr 22 22:15:09 ant__: I'm going to bed now Apr 22 22:15:20 please send reviewed proposal to me :) Apr 22 22:15:28 tomorrow (23 Apr) is deadline Apr 22 22:20:30 ok, I've sent the patches, will read right now Apr 22 22:20:32 :) Apr 22 22:24:46 Jay7: sounds a bit like the old proposal... Apr 22 22:24:50 :) Apr 22 22:25:20 is NFS boot just a thing of adding kernel modules and cmdline? no userland? Apr 22 22:27:39 bluelightning: ok, pls pull the 4 patches Apr 22 22:28:13 new patchset will follow (tomorrow ?) for klibc + replacing linux-kexecboot Apr 22 22:45:26 ah, PRINC will become actual once we have to update the recipes Apr 22 22:45:42 not now Apr 22 23:03:11 bluelightning: ah, I did not add linux-yocto-tiny finally Apr 22 23:03:28 unnecessary atm **** ENDING LOGGING AT Mon Apr 23 02:59:58 2012 **** BEGIN LOGGING AT Mon Apr 23 02:59:58 2012 Apr 23 08:14:26 morning Apr 23 08:14:44 ant_work: I've added network boot to boot methods item Apr 23 08:14:51 hi Jay7 Apr 23 08:14:57 mail is sent to celinux-dev Apr 23 08:15:05 ok, fine Apr 23 09:37:54 bluelightning: about kernel logo, there is a big problem having it as a patch Apr 23 09:38:25 ant_work: what happens? Apr 23 09:38:33 as far as I could dig, you have to copy the patch in each machine directory and it's big Apr 23 09:38:46 otherwise it is fetched but not found by the kernel tools Apr 23 09:40:25 its' a strange thing: the kernel tree used to build yocto kernels has a /meta dir, inside this one you'll find a symlink pointing to the dir where the recipe was parsed (pwd ?) Apr 23 09:40:48 or better, where the first .scc is Apr 23 09:41:11 hmm, not sure what thats about Apr 23 09:41:21 heh Apr 23 09:42:27 I've committed so that it behaves 'normally' wrt OE standards Apr 23 09:42:37 it would be pretty easy to have a zaurus override Apr 23 09:42:52 just add OVERRIDES += "zaurus" in zaurus.inc Apr 23 09:43:14 then you could use this to enable the logo Apr 23 09:43:14 well, it does a check about file existence Apr 23 09:43:18 or anything else for that matter Apr 23 09:43:26 then SRC_URI is a problem though... Apr 23 09:43:36 if you have staged the file there you need it, isn't? Apr 23 09:43:57 why list the logo in SRC_URI otherwise? Apr 23 09:44:01 "staged" how? Apr 23 09:44:08 fetched Apr 23 09:44:33 under what circumstances though? it has to be conditionally added Apr 23 09:45:35 if [ -e ${WORKDIR}/logo_linux_clut224.ppm ]; then Apr 23 09:46:56 to SRC_URI I meant Apr 23 09:47:27 ah, this is only done with overrides see _akita and so on Apr 23 09:48:02 SRC_URI_append_akita Apr 23 09:48:08 right Apr 23 09:48:32 it's an ugly recipe-style, I know Apr 23 09:48:43 but is for 6 devices Apr 23 09:49:02 you could always do the override as suggested above Apr 23 09:49:06 maybe yes, add zaurus as SOC...there must be some sort of that Apr 23 09:49:13 override Apr 23 09:50:07 anyway now the recipe doesn't use any custom variable like LOGO_SIZE anymore Apr 23 09:50:30 ok, that's good Apr 23 09:50:39 I'll do the merge of your patches this lunchtime Apr 23 09:52:17 one last thing: thiose kernel tools detected a mismatch: we were forcing THUMB for collie Apr 23 09:56:16 "${ARM_INSTRUCTION_SET}" = "thumb" but collie is seen as armv4 Apr 23 09:58:36 hmm, I guess that would have caused some problems Apr 23 09:59:41 I remember something changed with the tune files Apr 23 10:00:07 collie includes tune-strongarm1100.inc Apr 23 10:00:59 which requires arch-armv4.inc Apr 23 10:02:18 where I find AVAILTUNES += "armv4 armv4t" Apr 23 10:02:48 and we do Apr 23 10:02:48 AVAILTUNES += "strongarm" Apr 23 10:05:32 may be INHERIT += "kernel-logo" ? :) Apr 23 10:05:48 it's about logo problem :) Apr 23 10:06:01 or even "bootsplash" Apr 23 10:06:32 hmm Apr 23 10:06:41 ant_work: about collie and thumb Apr 23 10:06:49 is it true for poodle too? Apr 23 10:07:29 this may be reason why I see problems with collie which you have not Apr 23 10:07:46 (with your poodle) Apr 23 10:13:20 no, poodle is armv5te Apr 23 10:13:51 for collie, there is no CONFIG_ARM_THUMB=y Apr 23 10:14:48 in the produced .configs Apr 23 10:18:22 Jay7: btw after login in kexecboot.org Apr 23 10:18:23 Notice: Undefined property: stdClass::$cache in DrupalDatabaseCache->prepareItem() (line 422 of /var/www/dms/site/includes/cache.inc). Apr 23 10:18:35 yes, I know Apr 23 10:18:58 it's known bug in drupal Apr 23 10:19:10 nothing bad really Apr 23 10:19:31 I hope it'll gone with next release Apr 23 10:19:35 hm, how do we refer the kexecboot folder now? Apr 23 10:19:51 use 'insert link' button in UI Apr 23 10:19:58 in editor I mean Apr 23 10:20:06 ah, ok, works with local files Apr 23 10:20:53 write some text, select it, press 'insert link', press button nearly URL field Apr 23 10:21:27 same browser will be opened and you then may choose kexecboot folder from tree at left and files in main view Apr 23 10:23:26 seems it has to be done per-file Apr 23 10:23:36 can we just point to the dir? Apr 23 10:25:57 we need to create a page wit all the links then Apr 23 10:26:14 brb Apr 23 10:59:03 ah Apr 23 10:59:20 well, I'll look for appropriate drupal module Apr 23 12:00:47 Jay7: btw, CONFIG_KLIBC_THUMB is not set so kexecboot should work on collie Apr 23 12:00:55 in any case Apr 23 12:01:15 ant_work: but what about kexecboot itself? Apr 23 12:01:33 ah, the static..hmm Apr 23 12:09:37 iirc, objdump or readelf should show thumb mode on files Apr 23 12:09:47 but not sure Apr 23 12:09:54 or may be even file Apr 23 12:10:27 it looks like in oe-core collie is armv4 so no thumb, neither for kernel Apr 23 21:31:36 bluelightning: klibc is somehow fixed Apr 23 21:32:04 pls loo kat and commit the 3 patches I sent right now Apr 23 21:38:44 ant__: done Apr 23 21:38:46 thanks Apr 23 21:40:13 thank you Apr 23 21:40:28 now formatting the klibc patch Apr 23 21:40:43 for meta-initramfs **** ENDING LOGGING AT Tue Apr 24 02:59:58 2012 **** BEGIN LOGGING AT Tue Apr 24 02:59:58 2012 Apr 24 09:16:43 Jay7: it seems that a download page (10-20 lines in php) is not so easy with drupal Apr 24 09:16:55 thsi seems good: http://drupal.org/node/458744#comment-1573834 Apr 24 09:26:29 http://drupal.org/project/download_file Apr 24 09:26:35 may be this module will help us Apr 24 09:27:33 heh.. just for fetching an array :) Apr 24 09:27:50 (one day I'l have a look at drupal) Apr 24 09:28:37 Jay7: we can maybe remoce the reference to Sources Apr 24 09:28:45 ach.. today I can't type Apr 24 09:29:48 do we have place for the git tree on kexecboot.oerg? Apr 24 09:30:18 argh again...this kb is too small for my fingers Apr 24 09:35:48 well, let it be on github...maybe we should create a new Username like 'kexecboot' or 'kexecboot-devel' Apr 24 09:36:05 Jay7: can you do that? Apr 24 09:36:31 ant_work: you may do it faster :) Apr 24 09:36:43 which username do you prefer? Apr 24 09:36:44 I don't have even login on github Apr 24 09:36:56 ah, it's about creating a new one Apr 24 09:37:14 is it possible to have project with multiple admins? Apr 24 09:37:24 to have e.g. $ git clone git@github.com:username/kexecboot.git Apr 24 09:37:32 hm.. don't know Apr 24 09:37:38 then we may just use kexecboot as project name and every devel may have own name Apr 24 09:38:13 seems you need Organizations to collaborate, let me read Apr 24 09:38:36 check how OE mirror on github is organized Apr 24 09:38:43 (if any) Apr 24 09:42:41 I meant git clone git@github.com:kexecboot/kexecboot.git or alike Apr 24 09:43:36 I think you should keep exclusive commit control over there, like now Apr 24 09:43:51 i.e. we send patches to the ML and you pull Apr 24 09:44:00 seems easy enough Apr 24 09:44:26 (and you could maybe reuse your password, one less :) Apr 24 09:45:53 btw, Go to admin page ( https://github.com/user/repo/admin ) and in the Collaborators tab you can add as many as you want ( since the free ones are public repos) Apr 24 09:46:03 The added user should now be able to push to your repository on GitHub. Apr 24 09:47:16 but again, I think it's ok like it is now Apr 24 09:47:30 maybe once you gain a big development crew ;) Apr 24 09:49:59 Jay7: heh.. I can't create a second free (Username kexecboot) with my email: https://github.com/signup/free Apr 24 12:46:50 Jay7: may I use jay4mail as email?Another one? Apr 24 12:47:00 ant_work: use it Apr 24 12:47:16 it's kind of public email for opensource projects now :) Apr 24 13:00:23 ok, Download page is updated, now we need that DL/Links Apr 24 13:00:28 for Zaurus Apr 24 13:01:26 btw, *you* have to push in that repo with your keys Apr 24 13:01:41 oh.. Apr 24 13:01:47 as first at least Apr 24 13:01:47 well, I'll do from home Apr 24 13:02:48 I think you can add more keys but let's start simply :) Apr 24 22:04:13 ant__: hi Apr 24 22:05:44 how to push existing git repo to github? Apr 24 22:05:51 to save history, etc Apr 24 22:41:21 ah, hm.. I did just with forked repos..let see the docs Apr 24 22:47:27 Jay7: as I did for meta-handheld, you clone it locally then clone Apr 24 22:47:37 http://help.github.com/remotes/ clone Apr 24 22:48:55 then you push it with your ssh key Apr 24 22:50:17 so, you clone the original repo, clone the empty kexecboot then merge them and push Apr 24 22:52:16 I was trying to build klibc with linux-libc-headers.... now I stop with that and do a quick test of ubifs fsscan Apr 24 23:29:31 sh@#t Apr 24 23:29:55 ubi_io_write: error -5 while writing 256 bytes to PEB ... Apr 24 23:30:22 ... have CONFIG_MTD_NAND_VERIFY_WRITE enabled in your kernel, you will need to turn it off. Apr 24 23:30:35 The code does not currently (as of 2.6.26) perform verification of sub-page writes correctly. As UBI is one of the few users of sub-page writes, not much else seems to be affected by this bug. Apr 24 23:30:36 :p Apr 24 23:31:02 can't belive the bug survived from 2.6.26 Apr 24 23:34:21 good news is we have all tools in oe-core-base image Apr 24 23:34:37 ubi tools I mean **** ENDING LOGGING AT Wed Apr 25 02:59:58 2012 **** BEGIN LOGGING AT Wed Apr 25 02:59:58 2012 Apr 25 20:59:44 Jay7: hi Apr 25 20:59:53 hi Apr 25 21:00:00 FS type 'ubifs' detected Apr 25 21:00:15 ..but I can't create volumes on Z :/ Apr 25 21:00:19 smthg wrong Apr 25 21:00:26 1) good! Apr 25 21:00:29 2) bad! :) Apr 25 21:00:30 I'm flashing another image to test Apr 25 21:00:41 someone is using ubi on Z.. Apr 25 21:00:48 now, I've added ubi too, has a different magic Apr 25 21:00:48 may be JaMa Apr 25 21:00:58 I'm confused..is it managed by kernel? Apr 25 21:01:07 the volume I mean Apr 25 21:01:23 I don't know Apr 25 21:01:36 better to find one with Z and ubifs Apr 25 21:02:47 ubi: Apr 25 21:02:49 00000000 55 42 49 23 Apr 25 21:03:04 UBI# Apr 25 21:03:21 ubifs: Apr 25 21:03:23 00000000 31 18 10 06 Apr 25 21:03:40 1...! Apr 25 21:03:45 ^ seems to be volume 1 Apr 25 21:04:07 no idea here, I'll read more docs Apr 25 21:27:20 well, I've downloaded some images from the web ans the strings do match Apr 25 21:28:05 * Jay7 -> sleep Apr 25 21:28:14 gn Apr 25 21:28:16 ant__: good luck in your investigations :) Apr 25 21:28:35 ;) Apr 25 23:34:34 heh Apr 25 23:34:36 We have investigated the UBIFS issue, and found that it’s related to some kernel configuration options. Apr 25 23:34:36 In order to be able to work with UBIFS, you need to changes the following: Apr 25 23:34:36 - Enable “NFTL support” Apr 25 23:34:36 - Enable “write support for NFTL” Apr 25 23:34:36 - Disable “Verify NAND page writes”. Apr 25 23:34:50 I'll retry tomorrow Apr 25 23:34:52 gn **** ENDING LOGGING AT Thu Apr 26 03:00:00 2012 **** BEGIN LOGGING AT Thu Apr 26 03:00:00 2012 Apr 26 07:55:51 hi, am i right in thinking that some of the zaurus kernel hackers (zaurus-devel-lists posters) hang out here? Apr 26 08:04:31 hi , not really, they are in linux arm kernel ML usually Apr 26 08:05:03 sometime one appears here, though :) Apr 26 08:05:26 what's the problem with kernel? Apr 26 08:06:21 ant_work: no problem Apr 26 08:06:29 i have 2 spitz machines Apr 26 08:06:43 they run debian Apr 26 08:06:57 but they have a very old kernel Apr 26 08:07:07 2.6.24 yonggun or something. Apr 26 08:07:22 i spent a long time searching last night, and i found the zaurus devel list Apr 26 08:07:38 which was full of interesting posts about people working on 3.x kernels for zaurus! Apr 26 08:07:50 so i did some more searching and found this place. Apr 26 08:08:25 so far i havent been able to find any kernels to download though. Apr 26 08:09:05 ah, I have a build of last week, for all Z Apr 26 08:09:31 ant_work: whereabouts do you host this stuff? Apr 26 08:09:46 one of my two machines has (i think) kexecboot installed Apr 26 08:10:09 its the funky thing that android had right? where it would find kernels and let you select which one to boot? Apr 26 08:10:15 the kexecboot kernels will be on kexecboot.org site Apr 26 08:11:14 ant_work: im basically trying to get back up to speed with my z's. they have been collecting dust and acting as paperweights for too long. Apr 26 08:11:33 Jay7: there is a size limit for the upload, I could no upload the modules Apr 26 08:11:34 i need to do some research and then im going to try and get them installed with squeeze and a nice new kernel! Apr 26 08:12:22 if you have a decent build machine you could build kernel and images in a couple of hours Apr 26 08:12:45 ant_work: there shouldn't be any Apr 26 08:12:57 what error you receive? Apr 26 08:13:07 ok, I'll retry to upload kernels+modules this evening Apr 26 08:13:12 iirc was size too big Apr 26 08:13:31 hm.. I'll check from home today Apr 26 08:13:36 bother me about that :) Apr 26 08:14:15 HRH_H_Crab: I can send it to you by mail if you prefer (linux-yocto_3.2) Apr 26 08:14:38 Yocto is the new hype :) Apr 26 08:15:37 Jay7: remember to push the github repo Apr 26 08:16:00 yeah.. too much new reading for one evening :) Apr 26 08:16:06 ant_work: dont worry about it too much right now. Apr 26 08:16:10 im at work atm. Apr 26 08:16:43 im going to spend some time doing a bit more research at the weekend and plan what to do next. Apr 26 08:16:59 im just really happy that it looks like there might be a bit more life left in those two spitz! Apr 26 12:15:00 Jay7: http://linux-mtd.infradead.org/doc/ubidesign/ubidesign.pdf Apr 26 12:15:09 pag 8 Apr 26 12:15:24 & 9 Apr 26 12:16:26 so it seems the volume identifier is at bit 256 and should be UBI! Apr 26 12:16:43 but the erasecounter header is at 0 and is UBI# Apr 26 12:17:09 (at bit 256 on Zaurus) Apr 26 12:17:24 I'll check more in depth... Apr 26 16:05:02 ant_work: still about? Apr 26 16:05:24 yes, heading home soon Apr 26 16:05:50 if you can, please mail that kernel to horseshoecrab@gmail.com Apr 26 16:06:06 is this oabi or eabi? Apr 26 16:06:22 im trying to figure out what i need to do with debootstrap. Apr 26 16:06:41 hopefully i can get something working off a sd card using the z that already has kexecboot installed. Apr 26 16:07:04 this is no fun... all the docs end in dead links! :P Apr 26 16:07:16 we have eabi Apr 26 16:07:27 iirc oabi compatibility is also set, musct check Apr 26 16:07:31 o.k. cool. Apr 26 16:07:46 thanks for your help btw. Apr 26 16:08:13 the defconfig is at http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto/spitz/spitz.cfg Apr 26 16:08:19 yw ;) Apr 26 16:08:56 this is then minimally mangled by kernel configuration tools Apr 26 16:09:43 I still have to shrink it to the bone, the other fragments are added separately Apr 26 16:09:47 hmm. im not sure what i'll do with config files just yet. my main aim is just to try and figure out how to get a reasonable debian install working with a prebuilt kernel. but if i ever get that far, i might start playing with compilers Apr 26 16:10:08 CONFIG_AEABI=y Apr 26 16:10:09 CONFIG_OABI_COMPAT=y Apr 26 16:10:24 http://www.penguin.cz/~utx/zaurus/feed/images/spitz/ Apr 26 16:10:34 that stuff looks interesting and potentially useful. Apr 26 16:11:06 yes, this is the 'old way' providing full defconfigs Apr 26 16:12:14 I tested that kernel and booted a minimal shell only image on spitz as well Apr 26 16:12:34 maybe you'll miss some kernel module Apr 26 16:12:44 probably not Apr 26 16:14:08 ugh Apr 26 16:14:17 the horrible part at the moment is the bloody debootstrap. Apr 26 16:16:42 atm I would just unpack a debian imag on CF or SD and boot it with kexecboot Apr 26 16:24:25 well, the only place that seems to have a decent image is yongguns old site and i was hoping for something newer. i think debootstrap is working now though so i'll see how i go. Apr 26 16:24:38 i assume that there is no problem with older versions of kexecboot booting newer kernels? Apr 26 16:31:09 kexecboot after 2.6.32 should behave ok Apr 26 16:31:53 there was an hd* -> sd* change Apr 26 16:32:07 for CF cards and internal hdd Apr 26 16:32:31 I'll send you the latest 3.2 I've tested on my spitz Apr 26 16:35:01 bbl Apr 26 19:45:48 I've pushed kexecboot to github repo Apr 26 19:47:28 now every commit from me shown as kexecboot's one :) Apr 26 20:53:11 Jay7: http://paste.debian.net/165575/ Apr 26 20:53:26 so the first bytes are "UBI#" Apr 26 20:53:45 kind of 'magic' :) Apr 26 20:53:51 but then it depends on nand size: on zaurus corgi the vid is at 256b Apr 26 20:53:58 UBI! Apr 26 20:54:01 ant__: I've increased max upload file size to 8Mb Apr 26 20:54:05 is it enough? Apr 26 20:54:14 on bigger nand can be first bytes of second page Apr 26 20:54:20 2048 + 2048 Apr 26 20:54:52 I still have no images, just ubi container atm Apr 26 20:55:33 if you see in fstype, for lvm2 it reads block 0 and block 1 Apr 26 20:56:38 well, maybe it is enough to detect ubi container, the kernel will ubiattach if proper cmdliner is given Apr 26 20:57:00 (I hope) Apr 26 20:57:47 Jay7: pls make a second directory for normal Zaurus kernels Apr 26 20:57:55 ok Apr 26 20:58:07 I'm playing with downloadfile module Apr 26 20:58:11 I'll upload in my profile now Apr 26 20:59:10 ant__: done (kernels directory) Apr 26 20:59:39 it is possible to create special page type for which you may attach list of files Apr 26 21:00:10 but I can't enable description field.. Apr 26 21:00:15 just list of files now Apr 26 21:00:58 yes, Apr 26 21:01:04 I can upload now Apr 26 21:08:16 done Apr 26 21:12:00 ant__: thanks for the kernel! Apr 26 21:12:48 np, we are finishing the download page ... Apr 26 21:12:53 heheh Apr 26 21:13:03 you worked hard for all these years, Apr 26 21:13:08 and finally you got a user? ;) Apr 26 21:13:27 it's like the ship in the bottle Apr 26 21:13:40 night hobby Apr 26 21:13:50 i think its a cool hobby Apr 26 21:13:57 I'm too old for youpr0n Apr 26 21:14:07 ;) Apr 26 21:14:13 i think the zaurus is a nice piece of electronics. Apr 26 21:14:35 you see, I'm sourrounded by Apple devices at home..wife, kid... Apr 26 21:14:41 i don't see a lot different in keeping a c3200 running nicely in 2012 as being too different from a classic car. Apr 26 21:14:58 not good grammar Apr 26 21:15:02 but you know what i mean Apr 26 21:15:09 yes Apr 26 21:15:26 I have an old vespa Apr 26 21:15:28 :) Apr 26 21:15:55 and in the years I gathered vintage electronic instruments Apr 26 21:16:14 ..and if only I had a bit more money I'd buy an Hammond B3 Apr 26 21:16:22 that's all :) Apr 26 21:17:19 i have a few of those Apr 26 21:17:33 (vintage electronic instruments - not hammonds though) Apr 26 21:18:52 anyway, back to my aston martin, i mean uh, spitz... Apr 26 21:19:39 I'm testing ubi on Zaurus, knowing is a waste of time...but I don't have any other device to test Apr 26 21:19:55 it's nonsense with small nand Apr 26 21:20:17 still mount time is a fractiuon Apr 26 21:20:25 -u Apr 26 21:22:56 ant__: check http://kexecboot.org/node/24 Apr 26 21:23:05 press edit and look inside Apr 26 21:23:57 only ugly thing is one folder per such page Apr 26 21:24:32 well, the zaurus kernels could be hyperlinked when we mention Zaurus. HEre are a bit OT Apr 26 21:24:56 or in the documentation as well Apr 26 21:25:02 it's not a problem to create another page for kernels e.g. Apr 26 21:25:11 or use same folder.. Apr 26 21:25:21 anyway, is it ok for you? Apr 26 21:25:24 yes Apr 26 21:25:42 well, then I'll replace current download page with this Apr 26 21:26:14 all seems fine with git Apr 26 21:26:36 what we will do with LTG repo? Apr 26 21:27:15 hm Apr 26 21:27:49 we'd only need the mailing list there Apr 26 21:29:27 hm.. and seems it not possible to attach already uploaded file to page Apr 26 21:33:04 ant__: I've switched pages Apr 26 21:33:48 seems I should remove all files in kexecboot and kernels folders and allow you to reupload 'em :) Apr 26 21:34:24 or write special module.. Apr 26 21:34:28 ^<^ Apr 26 21:36:55 ant__: I've cleared kexecboot folder :\ Apr 26 21:37:03 please attach files to download page Apr 26 21:37:11 you may provide description there too Apr 26 21:37:24 description will be shown instead of filename Apr 26 21:41:36 can be changed afterwards? Apr 26 21:42:01 yes, imho Apr 26 21:42:12 For security reasons, your upload has been renamed to initramfs-kexecboot-klibc-image-akita-20120421232906.rootfs.cpio_.gz. Apr 26 21:42:16 but not sure about time :) Apr 26 21:42:29 o_O Apr 26 21:43:55 seems I should write simple module which will add block with table of files from folder Apr 26 21:53:21 * Jay7 -> sleep Apr 26 21:54:10 gn Apr 26 21:55:06 ant__: seems drupal will rename all files Apr 26 21:55:20 some crazy pseudo-security Apr 26 21:55:36 well, I'll investigate tomorrow may be Apr 26 21:57:56 it's not too bad, though, if we can rename it later Apr 26 21:58:48 bluelightning: tomorrow is the D Day ? Apr 26 22:21:51 ant__: I think so... I'm on holiday atm though Apr 26 22:22:53 enjoy ! **** ENDING LOGGING AT Fri Apr 27 03:00:00 2012 **** BEGIN LOGGING AT Fri Apr 27 03:00:00 2012 Apr 27 07:13:07 Jay7: ok, I'v come to the point fs is detected but 'no support' in kernel Apr 27 07:13:25 first reson is I detected 'ubi' but you have to mount 'ubifs' Apr 27 07:14:06 hm.. Apr 27 07:14:20 fstype should return name suitable for mount use Apr 27 07:14:23 second reason is that you have to mount e.g. ubi0_0 Apr 27 07:14:29 and not the mtd Apr 27 07:14:34 well.. that's worse Apr 27 07:14:37 id a bit like losetup.... Apr 27 07:14:57 I'm looking how to erxtract volume id Apr 27 07:15:04 can you look into /proc/partitions? Apr 27 07:15:14 are there any ubi* volumes? Apr 27 07:15:16 I was doin reading the sector Apr 27 07:15:58 iirc no partitions but it was late Apr 27 07:16:23 np Apr 27 07:16:55 and, I *hope* ubiattach is done by the kernel or we need the binary... Apr 27 07:17:24 Jay7: look the steps here: http://free-electrons.com/blog/creating-flashing-ubi-ubifs-images/ Apr 27 07:17:43 I could finally create the volume and fill it Apr 27 07:18:02 it looks like lvm usage Apr 27 07:18:20 yes, 2 steps Apr 27 07:18:52 we can override root device by config file anyway Apr 27 07:19:27 ah.. but we should detect fs and mount it before.. Apr 27 07:19:59 afaik the volume id is at 256 or at 512 or at2049 typically Apr 27 07:20:26 2048 even Apr 27 07:20:55 if I give you that you can write the mount string Apr 27 07:21:04 dev/ubiX_Y Apr 27 07:21:23 hm.. may be good idea Apr 27 07:21:24 X is almost always 0 but Y depends on how many volumes (I think) Apr 27 07:21:44 I'l see what is in /proc and /sys Apr 27 07:22:57 Jay7: btw the error I got is in the FAQ.... http://www.linux-mtd.infradead.org/faq/ubi.html#L_subpage_verify_fail Apr 27 07:23:15 I was sure I was booting the latest yocto kernel but I did not :/ Apr 27 07:23:35 CONFIG_MTD_NAND_VERIFY_WRITE must be disabled if we want to use sub-pages Apr 27 07:23:57 (256bytes instead of a full 512) Apr 27 07:24:16 I don't know how risky it can be.... Apr 27 07:24:35 we need to ask someone :) Apr 27 07:24:43 heh Apr 27 07:25:16 or we can use full pages of 512 and re-enable verify (seems also possible) Apr 27 07:43:15 ant_work: well things seem to work nicely. Apr 27 07:43:36 kexecboot looks to be working well and your new kernel boots and works very well. Apr 27 07:43:40 seems a bit faster too. Apr 27 07:43:47 nice Apr 27 07:43:49 hostap working. Apr 27 07:44:13 the only problem is with my debootstrap nonsense - so far i can only boot the existing root file systems. Apr 27 07:44:34 is there some reason why your kernel wouldnt want to boot from sd? Apr 27 07:46:20 maybe wrong cmdline in /boot.cfg ? Apr 27 07:47:46 (/boot of the sdX partition) Apr 27 07:48:07 you need a /boot/boot.cfg Apr 27 07:49:03 ah, and the kernel is meant to be launched by kexecboot (machine id detection is not configured) Apr 27 07:49:20 [o.k.D Apr 27 07:49:22 cool. Apr 27 07:49:26 i need to research this. Apr 27 07:49:27 thanks. Apr 27 07:49:39 look in kexecboot.org Apr 27 07:49:48 yeah will do. Apr 27 07:50:10 unfortunately the networking on my z has decided to crap out. Apr 27 07:50:15 but that is standard behaviour Apr 27 07:50:26 i strongly doubt its related to the new kernel Apr 27 07:50:32 but i can look into that when i get home. Apr 27 07:51:51 can't exclude, I didn't test runtime basically Apr 27 07:52:35 i dont think the buffalo card and my airport extreme are very friendly. Apr 27 07:52:59 at least the extreme is more friendly than my express... Apr 27 07:53:35 I still have older router, 8011 b/g Apr 27 07:53:44 I'll test soon Apr 27 07:58:51 Jay7: there is a link to a git tag on linuxtogo in 'how_to_write_config'..I'll fix that Apr 27 08:00:56 ok Apr 27 09:12:50 Jay7: seems it's all in the cmdline...rootfstype=ubifs ubi.mtd=3 root=ubi0:$VOL_LABEL (many are using 'rootfs' here) Apr 27 09:13:07 no need to ubiattach or mount Apr 27 09:13:13 (for booting) Apr 27 09:13:40 ant_work: but for reading boot.cfg.. Apr 27 09:25:38 I hope to find a /dev/ubi0 created by devtmpfs at boot Apr 27 09:26:02 in kexecboot, after mounting devtmpfs Apr 27 09:30:51 hm.. .g. $ modprobe ubi mtd=3 loads the UBI kernel module and attaches mtd3 Apr 27 09:31:13 If UBI is compiled into the kernel, the mtd device to attach may be specified in the kernel boot parameters, e.g., ubi.mtd=3 command makes UBI attach mtd3 when the kernel is booting, Apr 27 09:31:45 And finally, MTD devices may be attached or detached at any time with the ubiattach and ubidetach utilities Apr 27 09:32:00 ubiattach-klibc necessary ? Apr 27 09:32:06 :/ Apr 27 09:32:55 ..from busybox... Apr 27 09:34:21 hehh Apr 27 09:34:22 https://github.com/radekp/gta04-init/commit/5f6ee2878b8be6b55d740c7976e4361089142fd5 Apr 27 09:34:28 Klibc compiled ubiattach and init Apr 27 09:35:06 static, perfect Apr 27 09:35:32 you'll need to special-case the detection of UBI Apr 27 09:35:44 fs name -> ubifs Apr 27 09:36:02 (would need a proper detection but atm is the only one fs for ubi) Apr 27 09:36:52 ubiattach -p /dev/mtdX Apr 27 09:37:35 then you get the /dev/ubi_ctrl and /dev/ubi0 Apr 27 09:38:36 here we can mount it mount -t ubifs ubi0:volume_name /mount/point Apr 27 09:38:53 in case of two mtd parts Apr 27 09:39:01 you'll get /dev/ubi1 Apr 27 09:40:00 iirc in oe-core volume_name is "$MACHINE_rootfs" Apr 27 09:40:12 when you build the images Apr 27 09:44:21 now that I have a valid volume I'll dump more bytes and see Apr 27 10:06:40 Jay7: thi sstarnge header has all info Apr 27 10:06:46 https://casper.berkeley.edu/svn/trunk/roach/sw/linux/include/mtd/ubi-header.h Apr 27 10:31:33 ant_work: nice Apr 27 10:32:01 but how to get the volume name from any volume? Apr 27 10:43:41 Jay7: volume number 65280 contains info (ubi -layout) Apr 27 10:44:57 http://linux-mtd.infradead.org/doc/ubidesign/ubidesign.pdf page 11 Apr 27 10:45:40 oh Apr 27 10:45:42 ok Apr 27 10:55:30 hm.. The layout volume is an "internal" UBI volume, and the users do not see it and cannot access it. Apr 27 10:56:34 still, Volume table is an on-flash data structure which contain ... Apr 27 10:56:42 if it is on-flash... **** ENDING LOGGING AT Sat Apr 28 03:00:00 2012 **** BEGIN LOGGING AT Sat Apr 28 03:00:01 2012 **** ENDING LOGGING AT Sun Apr 29 02:59:58 2012 **** BEGIN LOGGING AT Sun Apr 29 02:59:58 2012 Apr 29 11:49:44 hi channel Apr 29 15:56:00 -rwxr-xr-x 2 andrea users 34360 Apr 29 17:54 ubiattach Apr 29 15:56:00 -rwxr-xr-x 2 andrea users 34360 Apr 29 17:54 ubidetach Apr 29 15:56:14 file ubiattach Apr 29 15:56:14 ubiattach: ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked, stripped Apr 29 15:56:37 Jay7: bigger than expected Apr 29 15:57:06 well, will compress 50% Apr 29 16:03:40 Jay7: ubiattach-klibc works runtime ;) Apr 29 16:03:56 I'll add ubidetach, seems easy Apr 29 16:31:20 ok, we have ubi-utils-klibc Apr 29 16:31:37 for now I've only built those 2 Apr 29 16:31:56 we don't need more, maybe only ubiattach even Apr 29 16:39:17 ok, compressed size is 16-17 kb Apr 29 16:39:53 bbl Apr 29 16:58:25 jay7:ping Apr 29 17:09:48 /win 21 Apr 29 17:09:50 ops Apr 29 17:09:54 *oops Apr 29 17:29:44 dromede: pong Apr 29 17:30:23 i've read the backlog and saw this celf thing Apr 29 17:30:34 ah, hehe Apr 29 17:31:01 so ok, i'll help if i can Apr 29 17:31:17 i've been away recently Apr 29 17:31:30 nice Apr 29 17:31:35 i have to put all of my hobbies on hold for a while Apr 29 17:31:57 i'm still in college and i have neglected my duties for far too long Apr 29 17:32:13 now it's all piled up, i have to deal with that for now Apr 29 17:32:27 i'll have more time after may Apr 29 17:32:40 * Jay7 is now suspended all hobbies too but because new full time job.. need time to understand how things are working Apr 29 17:33:51 if i undestood correctly, kexecboot was backed by celf last year? Apr 29 17:40:11 we got sponsorship in 2010 Apr 29 17:40:52 anyways, i've looked at the kexecboot source Apr 29 17:41:06 there's room for improvement on the gui side of things Apr 29 17:41:10 i can work on that Apr 29 17:41:21 http://elinux.org/CELF_Project_Proposal/Improve_kexecboot Apr 29 17:41:34 this work was sponsored Apr 29 17:41:55 http://elinux.org/Improve_kexecboot Apr 29 17:42:00 this year proposal Apr 29 17:42:33 did it really take only 65 hours for the first one ? :) Apr 29 17:44:10 iirc. something about Apr 29 17:48:37 ah, 81 hours total Apr 29 17:50:38 close enough Apr 29 18:38:02 Jay7: I'm testing this last headers spli btw http://paste.debian.net/166575/ Apr 29 18:38:11 hi dromede Apr 29 18:43:39 hi ant__ Apr 29 20:24:13 for now Apr 29 20:24:14 PACKAGES =+ "ubiattach-klibc ubidetach-klibc mtdinfo-klibc ubimkvol-klibc \ Apr 29 20:24:15 ubirename-klibc ubirmvol-klibc ubirsvol-klibc" **** ENDING LOGGING AT Mon Apr 30 02:59:58 2012 **** BEGIN LOGGING AT Mon Apr 30 02:59:59 2012 Apr 30 07:58:53 wow Omegamoon ... Apr 30 07:59:17 :) Apr 30 07:59:22 yeah, I know Apr 30 07:59:33 has been a while Apr 30 08:00:28 but how are things here ant_work Apr 30 08:00:48 lo Omegamoon Apr 30 08:00:56 morning ant_work Apr 30 08:01:10 hi there HRH_H_Crab Apr 30 08:01:31 Omegamoon: you may remember me from the comments section of your blog... Apr 30 08:01:40 i think i made one or two posts there... Apr 30 08:02:58 yeah I remember :) Apr 30 08:03:18 I should do something about that blog... Apr 30 08:03:23 * ant_work is filtering a massive spam attack from .br servers Apr 30 08:04:03 I mean... it's so... inactive ;-) Apr 30 08:04:07 it is Apr 30 08:04:11 however, Apr 30 08:04:25 ive been playing with my zaurus' a bit over the last week or so Apr 30 08:04:43 and every time i went to google i ended up at your blog eventually! ;) Apr 30 08:04:46 ant_work: that wasn't me spamming you;-) Apr 30 08:06:10 HRH_H_Crab: haha, well, maybe I have to do something about that then, making sure that the FIRST post in Google leads you to my blog :P Apr 30 08:06:44 i think a lot of it had to do with searching for: Apr 30 08:06:49 "debootstrap zaurus" Apr 30 08:06:51 ;P Apr 30 08:07:09 there are only so many appropriate pages that google is going to suggest... Apr 30 08:07:53 oh yeah, that's a difficult one Apr 30 08:07:58 hello there Apr 30 08:08:14 how are you doing? still in NL? Apr 30 08:08:32 there are some nice bootstrap scripts in the Fedora repositories Apr 30 08:09:10 ant_work: oh yeah, I wouldn't leave NL Apr 30 08:09:18 except for holidays, that is Apr 30 08:09:27 heh Apr 30 08:10:40 have you seen how poky/oe did merge ? Apr 30 08:10:45 queensday today, so that means a day off Apr 30 08:11:48 I saw yocto Apr 30 08:11:58 seems nicely done Apr 30 08:12:35 although I missed reference to the zaurus packages Apr 30 08:12:48 we have a dedicated BSP Apr 30 08:13:16 http://www.openembedded.org/wiki/LayerIndex Apr 30 08:13:38 meta-handheld is the BSP Apr 30 08:13:46 meta-initramfs is for the kexecboot stuff Apr 30 08:13:47 I'll check that out for sure Apr 30 08:13:55 didn't know that Apr 30 08:13:58 we are now using Yocto kernels on Z Apr 30 08:14:12 (validated by the kernel config tools) Apr 30 08:14:30 what version are you using Apr 30 08:14:35 ? Apr 30 08:14:37 there is hope that ipaq will follow Apr 30 08:14:48 yocto-mainline is 3.2 Apr 30 08:15:00 3.4 is also in testing Apr 30 08:15:06 oh great Apr 30 08:15:33 and how is that performing? Apr 30 08:15:43 I'll have to find the time and reduce the zaurus. cfg to the minimal fragments Apr 30 08:16:02 ant_work: i am possibly missing something obvious, Apr 30 08:16:08 but i cant seem to get sound working Apr 30 08:16:15 Omegamoon: infamous power/recharge issues Apr 30 08:16:15 all the modules seem to be present and correct. Apr 30 08:16:34 Omegamoon: good news is are being worked on Apr 30 08:16:37 ouch... Apr 30 08:17:02 there are already first patches Apr 30 08:17:25 HRH_H_Crab: check your alsa config Apr 30 08:17:55 aah Apr 30 08:18:00 iirc there is smthg wrong with alsamixer in the oe/yocto images too Apr 30 08:18:06 HRH_H_Crab: it has been a while, but check the channels and make sure they're not on 'mute' Apr 30 08:18:21 Omegamoon: i think its more fundamental than that. Apr 30 08:19:28 hmm, where are the times when things were simple ;-) Apr 30 08:20:51 Omegamoon: ah, and btw I have tested ubi/ubifs on Zaurus Apr 30 08:21:15 ..and with big big luck right those days a kernel patch appeared :) Apr 30 08:22:08 so we can keep CONFIG_MTD_NAND_VERIFY_WRITE=y Apr 30 08:22:58 (it was not possible for ubi to write subpages of 256b, ecc verify was/is broken for subpages) Apr 30 08:23:40 for kexecboot it means that scanning time is <1/10 for partitions on nand Apr 30 08:24:37 argh, Walmart.com.br strikes again, from another mx :/ Apr 30 09:16:41 Omegamoon: with a quad core and a decent hdd (no raid) you can (cross)build oe-core-sato in two hours and an half Apr 30 09:17:11 build is better parallelized now Apr 30 09:25:57 morning Apr 30 09:27:39 ant_work: quad core... I'm always saving out on cores, so the best I have is dual core currently Apr 30 09:28:01 Jay7: hi there Apr 30 09:28:12 Omegamoon: wow! you are here :) Apr 30 09:28:27 hehe, yeah, it's for real Apr 30 09:28:48 had to happen some day ;-) Apr 30 09:30:43 Jay7: so with ubiattach we are ready to test ubi volumes. I'm thinking that lvm2 is similar, just needs libdevmapper Apr 30 09:30:45 cd Apr 30 09:30:51 oops Apr 30 09:31:30 ant_work: I'm going to update my kexecboot devel VM :) Apr 30 09:32:01 well, VM support will come later :) Apr 30 11:57:43 strange Apr 30 11:57:51 lsmod shows loads of sound modules loaded. Apr 30 11:58:12 snd_soc_core 61150 3 snd_soc_spitz,snd_soc_pxa2xx_i2s,snd_soc_wm8750 Apr 30 11:58:16 looks healthy Apr 30 11:58:17 but: Apr 30 11:58:32 crab@mongkok:/etc/modprobe.d$ aplay -l Apr 30 11:58:32 aplay: device_list:252: no soundcards found... Apr 30 12:01:13 HRH_H_Crab: check /proc/asound/devices Apr 30 12:01:21 and cards Apr 30 12:01:43 and other files there :) Apr 30 12:03:42 yeah Apr 30 12:03:45 theres not much. Apr 30 12:03:54 a timer and a sequencer Apr 30 12:04:23 nothing in cards or devices Apr 30 12:05:04 ls Apr 30 12:12:40 iirc there is smthg in kernel bootlog, check dmesg Apr 30 12:12:55 (this for all zaurus since long time) Apr 30 12:13:12 ...hmmm..or maybe is zaurusd Apr 30 12:13:34 sorry but I stop testing on boot since (too) long Apr 30 12:14:47 Jay7: http://www.kernel.org/doc/Documentation/ABI/stable/sysfs-class-ubi Apr 30 12:15:54 we get easily from here the ubi volume(s) Apr 30 12:18:45 yes, nice link Apr 30 12:19:08 usually we expect ubi0_0 but we could have ubi0_1 or even ubi1_X in case of two ubi mtdparts Apr 30 12:19:35 I think initially we can skip the ubi volume name(s) Apr 30 12:19:43 and just look atthe devices Apr 30 12:20:29 so, if you detect 'UBI#' you'll have to exec ubiattach on that mtd then process the devices Apr 30 12:21:15 seems we don't need ubifs detection in case of sysfs usage Apr 30 12:21:35 atm no, and btw is the *only* supported ubi fs atm :) Apr 30 12:21:52 just look into /sys/class/ubi, attach, mount and check for boot.cfg Apr 30 12:22:05 yes, seems so easy.... Apr 30 12:23:18 now, in userspace the devices are created by udev I think. I suppose devtmpfs will repopulate after ubiattach Apr 30 12:23:40 I see /dev/ubi_ctrl and /dev/ubi0 Apr 30 12:23:57 I'll check better when volumes are mounted Apr 30 12:27:02 may be we should go this way Apr 30 12:27:15 i.e. use /sys as only information source Apr 30 12:27:57 but I'm not sure 2.6.20 kernel have sysfs Apr 30 12:56:45 KernelVersion: 2.6.22 Apr 30 12:57:00 this Artem doc Apr 30 12:57:19 July 2006 Apr 30 12:57:29 well.... Apr 30 16:18:51 bbl **** ENDING LOGGING AT Tue May 01 03:00:01 2012 **** BEGIN LOGGING AT Tue May 01 03:00:03 2012 May 01 08:47:08 well, the good news is that i have finally got a nice fresh debian rootfs. May 01 08:47:31 the bad news is that ive lost ssh connection to the z! ;P **** ENDING LOGGING AT Wed May 02 02:59:58 2012 **** BEGIN LOGGING AT Wed May 02 02:59:58 2012 May 02 07:35:41 jeez...I've spent 2 hours trying to convert scanf() to a getchar() loop ... last I played with that was 20 years ago :) May 02 07:36:53 the idea was the patch should change only a few lines...I'll give up and rewrite the routines :/ May 02 07:37:59 it is for ubiformat-klibc , which is not so important: we only need ubiattach May 02 07:38:03 why you are doing this? :) May 02 07:38:13 I've finished th eubi-utils, all 12 May 02 07:38:14 ah, to use klibc May 02 07:38:38 don't we have some *scanf in klibc? May 02 07:38:57 no, neuther scanf nor fscanf May 02 07:39:08 there is getchar and getc, though May 02 07:40:04 Jay7: btw, it looks like the argv[] is not respected :/ May 02 07:40:17 i.e. ubiformat -O /dev/mtd3 id ok May 02 07:40:33 but ubiformat /dev/mtd3 -> too many mtd defined May 02 07:40:40 but we have read() btw :) May 02 07:40:42 sorry May 02 07:40:47 but ubiformat /dev/mtd3 -O 512-> too many mtd defined May 02 07:41:01 seems device should be last May 02 07:41:15 thi s works with eglibc May 02 07:41:30 with 'normal' ubiformat May 02 07:41:52 ah, then this is klibc's getopt implementation may be May 02 07:42:23 anyway, we may do it with right order :) May 02 07:43:32 see, I only patch scanf() May 02 07:43:33 http://git.infradead.org/mtd-utils.git/blob/HEAD:/ubi-utils/ubiformat.c May 02 07:43:53 I'll read a single char, 'n' or 'y' May 02 07:44:07 to keep the code small May 02 07:45:25 ah May 02 07:45:26 232 else if (optind != argc - 1) May 02 07:45:26 233 return errmsg("more then one MTD device specified (use -h for help)"); May 02 07:46:08 ^^ this looks like the problem May 02 07:47:22 in kexecboot we don't use optind, isn't? May 02 07:47:41 maybe the klibc implementation is buggy May 02 07:51:24 "By default, getopt() permutes the contents of argv as it scans, so that eventually all the nonoptions are at the end." May 02 07:51:28 "The variable optind is the index of the next element to be processed in argv. The system initializes this value to 1" May 02 07:53:20 that is the reason May 02 07:54:23 no option => stop processing => use all parameters left as non-options May 02 07:54:45 heh POSIXLY_CORRECT May 02 07:54:46 If this is set, then option processing stops as soon as a nonoption argument is encountered. May 02 07:55:14 POSIX.2 and POSIX.1-2001, provided the environment variable POSIXLY_CORRECT is set. Otherwise, the elements of argv aren't really const, because we permute them. We pretend they're const in the prototype to be compatible with other systems. May 02 07:55:21 now that is fun May 02 07:55:36 just don't care much about this May 02 07:56:20 I've noticed it with ubiformat, not for ubiattach May 02 08:18:59 Jay7: this is the cleaner code I can find http://opensource.apple.com/source/cvs/cvs-30/cvs/lib/yesno.c May 02 12:22:12 Jay7: finally the patch seems one-liner ;) May 02 12:22:44 - if (scanf("%3s", buf) == EOF) { May 02 12:22:59 + if (fgets(buf,4,stdin) == NULL) May 02 12:23:23 hehe May 02 12:24:12 buf[4] is enough: I was thinking fgets gets the '\n' and adds a '\0' May 02 12:31:12 bluelightning: I can't remember the policy now about empty packages: better to filter them out with custom PACKAGES or let them live ? May 02 12:31:59 ant_work: I'm not sure either to be honest May 02 12:32:09 perhaps we don't have a clearly defined policy May 02 12:32:25 i.e. for uclibc-utils-klibc I left only the $PN-dbg cause a) I removed the docs, b) there is no -dev nor -staticdev c) the utils are packaged one by one May 02 12:33:12 with PACKAGES =+ "xyz-klibc" & FILES_xyz-klibc = "bb May 02 12:33:19 if empty though the package won't be produced though, unless ALLOW_EMPTY_ is set May 02 12:33:40 in fact I see them in package-split in Workdir May 02 12:34:11 ALLOW_EMPTY is per recipe? May 02 12:34:22 I didn't set it May 02 12:34:23 it should be set per-package May 02 12:34:34 you don't need to unless you want the package to be produced even if it's empty May 02 12:34:37 ah, yes, per -package May 02 12:35:07 the fact is $PN is usually packaged if not excluded May 02 12:37:42 similar applies to kexecboot packaging fwiw ;) May 02 12:43:41 bbl May 02 23:30:36 Jay7: root@poodle:~# cat /sys/class/ubi/ubi0_0/name May 02 23:30:37 poodle-rootfs May 02 23:30:37 root@poodle:~# **** ENDING LOGGING AT Thu May 03 03:00:01 2012 **** BEGIN LOGGING AT Thu May 03 03:00:02 2012 May 03 08:32:22 Jay7: hi May 03 08:32:30 hi May 03 08:32:43 so, I've finished to waste time with ubi & klibc :) May 03 08:33:13 now, I'll package the 33kb of ubiattach in the cpio May 03 08:34:19 how would you special-case the detection of mtd's containing ubi ? May 03 08:35:34 if no FS was found by fstype and device is mtd then look into /sys may be May 03 08:35:35 I'm thinking that one day we'll readd as special-case the losetup stuff, this will need different processing though May 03 08:38:36 i.e. DVT_MTD, sizeof("/dev/mtdblock")-1, "/dev/mtdblock" }, May 03 08:38:47 we do not use mtd as block May 03 08:38:56 for UBI May 03 08:40:22 may be even easied May 03 08:40:27 *easier May 03 08:40:53 just look into /sys after processing all devices May 03 08:44:41 well, we obtain UBI# from first erase counter scanning the mtd May 03 08:45:07 even before ubiattach May 03 08:45:39 I've looked yesterday at sys/classes/ubi *after* attaching and mounting a volume May 03 08:46:37 I'm looping chicken-egg here :) May 03 08:52:27 I can add an if in const char *detect_fstype(char *device, struct charlist *fl) May 03 08:52:43 around "/* Check that FS is known */" May 03 08:53:00 and ubiattach here May 03 08:58:21 better to check /sys before attaching :) May 03 08:58:54 I'm already here with the patches May 03 08:59:04 just need to decide if it is called ubi or ubifs May 03 08:59:36 well, it must be as it is in /proc/filesystems May 03 09:00:00 I am at log_msg(lg, "+ FS %s is not supported by kernel", fstype); May 03 09:06:54 cat /proc/filesystems |grep ubi May 03 09:06:55 nodev ubifs May 03 09:10:17 definitely the kernel calls it ^^ May 03 09:10:40 but then we need to check in sysfs to see if there are volumes May 03 09:10:54 because the mount string is different May 03 09:12:04 so yes, we need to gather the volumes *and* their names May 03 09:12:07 "mounts "rootfs" volume of UBI device 0 to /mnt/ubifs ("rootfs" is volume name). This method of specifying UBI volume is more preferable because it does not depend on volume number." May 03 09:12:30 well, we *do parse* the volumes so for us ID could also be an option May 03 09:13:36 Note, if X is not specified, UBIFS assumes 0, i.e., "ubi0:rootfs" and "ubi:rootfs" are equivalent. May 03 09:14:14 so mount -t ubifs ubi0:0 /mnt/XXX should be ok for the first May 03 09:25:30 ah, even simpler May 03 09:25:32 http://www.mjmwired.net/kernel/Documentation/filesystems/ubifs.txt May 03 09:26:03 103 Mount volume 0 on UBI device 0 to /mnt/ubifs: May 03 09:26:03 104 $ mount -t ubifs ubi0_0 /mnt/ubifs May 03 09:28:56 The UBI volume to mount is specified using "ubiX_Y" or "ubiX:NAME" syntax May 03 09:29:30 where "X" is UBI device number, "Y" is UBI volume number, and "NAME" is May 03 09:29:35 UBI volume name. May 03 10:51:50 i.e. just get ubiX_Y from sysfs and mount it May 03 10:52:13 but we need to mount sysfs then.. now we are using only procfs May 03 16:42:45 bluelightning: did I tell you psplash is rotated 90° on qvga (poodle) ? May 03 16:43:05 do we miss some BSP customization? May 03 16:46:08 ant_work: I'm not sure... you'd have to look at psplash's code May 03 16:46:16 it's not particularly complicated May 03 16:46:21 is it ok on ipaqs? May 03 16:46:32 we have reused it in kexecboot at first time :) May 03 16:47:10 ant_work: I don't think my opie recipes include psplash, so I've not really tested it May 03 16:47:11 fb.c was psplash-based before I've broken^Wrewritten it :) May 03 16:47:37 s/recipes/images/ May 03 16:47:44 fwiw it was broken when iy was Yocto instead of OE as well May 03 16:47:53 so no bugs added :) May 03 16:54:41 ah, I see, there is a single psplash-poky-img.h May 03 16:54:50 #define POKY_IMG_WIDTH (300) May 03 16:54:50 #define POKY_IMG_HEIGHT (241) May 03 16:55:14 241? May 03 16:55:15 so it would be ok but needs to be rotated if fb is rotated May 03 16:55:18 odd May 03 16:55:22 ^_^ May 03 16:55:31 http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-core/psplash/files/psplash-poky-img.h May 03 16:55:58 is centered on vga May 03 16:56:05 just fine on corgi/spitz May 03 16:56:17 on qvga the bar surposes th elogo... May 03 16:56:28 is even too big May 03 16:56:38 (the logo) May 03 17:00:02 I have to take pictures of kexecboot for the web site, I'll record a movie of psplash for you ;) May 03 17:00:40 Jay7: btw, no way to take screenshots on qemu on qvga? May 03 17:00:56 lumag was working on collie but I don't know... May 03 17:00:57 no qvga emulation afaik May 03 21:14:24 well, kexecboot hackers, you guys rock. May 03 21:14:29 just thought id mention that. May 03 21:15:35 thanks :) May 03 21:17:27 thanks for a very neat bootloader. May 03 21:17:54 its thanks to you that we can have cool stuff (zaurii with non-prehistoric kernels and up to date os') May 03 21:19:19 indeed, awesome work guys May 03 21:19:29 :) May 03 21:20:32 ant__ is our 'engine' :) May 03 21:20:51 motivator :) May 03 21:30:08 hi there ;) May 03 21:30:44 Jay7: so at boot there is only sys/class/ubi/version May 03 21:31:04 and /dev/ubi_ctrl May 03 21:31:29 so we are obliged to scan the mtd and then ubiattach May 03 22:43:08 ok, now ubifs is recognized: here we get May 03 22:43:10 log_msg(lg, "+ can't mount device: %s", ERRMSG); May 03 22:43:34 invalid... May 03 22:43:37 ofc May 03 22:47:31 gn **** ENDING LOGGING AT Fri May 04 02:59:59 2012 **** BEGIN LOGGING AT Fri May 04 02:59:59 2012 May 04 07:58:39 hi bluelightning May 04 07:58:47 hi ant_work May 04 07:58:53 Jay7: I did boot ubifs from kexecboot :) May 04 07:59:05 ant_work: wow! May 04 07:59:27 all hardcoded :/ May 04 07:59:32 but it bots May 04 07:59:36 +o May 04 07:59:39 that's better than nothing :) May 04 08:00:01 now, up to you to decide where to do the ubiattach May 04 08:00:19 I was doing it in devicescan, after the mount error May 04 08:01:12 if we can know that there is ubifs, we should add explicit checking for 'ubifs' FS before mounting May 04 08:01:24 then do ubiattach and change device name May 04 08:01:42 yes, I hardcoded ubi0_0 May 04 08:01:47 and seems we shouldn't do detaching May 04 08:02:04 the device with ubi headers is mtd3 May 04 08:02:26 in fact we scan the mtdblock3 but it's the same May 04 08:02:32 btw, do you detecting ubifs explicitly or ubi container? May 04 08:02:52 UBI# is the 'eraseblock marker' May 04 08:03:11 for UBI May 04 08:03:31 but I forced it to be ubifs_image May 04 08:03:42 if one ubi container can contain only one ubifs volume then it's enough to detect container May 04 08:03:45 (atm there is only one filesystem for UBI) May 04 08:04:14 or we may detect container then check /sys May 04 08:04:21 yes May 04 08:04:47 and even, I don't know what happens if first block is bad (at the beginning) May 04 08:05:01 i.e. detect container on /dev/mtdX, do ubiattach and check sysfs for actual volumes May 04 08:05:03 maybe we should read two to be sure (see lvml2) May 04 08:05:20 Jay7: yes May 04 08:05:34 same for lvm May 04 08:05:47 but do vgchange -y instead :)( May 04 08:05:58 seems so May 04 08:06:14 sounds like plan May 04 08:06:18 but we should read and find a list of volumes May 04 08:07:02 so maybe we need to add ubi and ubifs to fstype.c May 04 08:07:15 list of volumes is in /sys May 04 08:09:23 practically we could blindly try to ubiattach i.e. ubi0_0 ubi0_1 ..ubin_n May 04 08:09:49 ubiattach should immediatelyfail May 04 08:09:56 yes, for every /dev/mtd* May 04 08:09:57 it takes half a second to attach May 04 08:10:22 only for mtd with fstpe ubifs (UBI#) May 04 08:10:40 why for every ? May 04 08:11:01 well, even better - not for every :) May 04 08:11:55 I meant do ubiattach for every /dev/mtd* and then look into /sys for actual volumes :) May 04 08:12:13 but your idea is better May 04 08:12:19 do you plan to add a function call for ubi/lvml2 ? May 04 08:13:21 I'd prefer to wait for CEWG results May 04 08:13:30 but some background work may be done already May 04 08:15:19 I mean, should we try to do it at the beginning of detect_fstype() so it will not fail or ubiattach before it tries to mount? May 04 08:15:51 log_msg(lg, "+ FS type '%s' detected", fstype); -> ubifs May 04 08:16:26 if I let ubi then it is not supported in kernel May 04 08:17:13 we should check ubifs presence in kernel before May 04 08:17:22 then do ubiattach if it is ubi container May 04 08:19:06 may be at end of devicescan_next() May 04 08:19:18 after detect_fstype() May 04 08:20:20 or in scan_devices() before mount May 04 08:20:32 second is better imho May 04 08:22:16 hm.. but if there may exists multiple volumes in same container (on same mtd) then we need specail processing May 04 08:22:20 another loop around May 04 08:55:18 I think in theory we should run detect_fstype one second time on the volumes May 04 08:55:43 and check for ubifs magic, which is not UBI# or UBI! May 04 08:56:58 this would be the real "ubifs", ubifs_image May 04 09:03:57 so in fact I should add two entries to static struct imagetype images[] = { May 04 09:04:58 one is for the eraseblock marker UBI# -> how do we call it? ubicontainer_image ? which triggers the ubiattach May 04 09:07:12 'ubi' and 'ubifs'? May 04 09:07:19 heh, one would be for the ubi_volumes (UBI! header) May 04 09:07:30 then the real ubifs ubifs_image May 04 09:07:55 #define UBIFS_SUPER_MAGIC 0x24051905 May 04 09:08:18 I think we can process the volumes with sys for simplicity May 04 09:09:15 I don't know.. do you prefer to have the right name or *identify from the first detection? May 04 09:13:54 #define UBIFS_NODE_MAGIC 0x06101831 May 04 09:14:01 argh ... 4 now :/ May 04 09:14:15 better kiss May 04 09:14:30 then if you get the contract you'll redo it properly :) May 04 09:14:45 good idea :) May 04 09:15:22 ok, I'll se nd you a first patch for that May 04 09:15:49 after {0, "jffs2", jffs2_image}, I'll just add {0, "ubifs", ubifs_image}, May 04 09:16:27 then we have to intercept it as discussed before May 04 09:23:07 seems it is possible to have multiple volumes on top of same mtd partition.. May 04 09:23:32 so we can't easily detect all volumes in general case May 04 09:23:43 only by parsing ubi headers May 04 09:23:56 or just ubiattach and look into /sys May 04 09:24:06 the volume appears in /sys/class/ubi only after ubiattach May 04 09:25:18 http://git.denx.de/?p=u-boot.git;a=blob;f=drivers/mtd/ubi/scan.c;h=d5c1d27dcbe83220f2bf60d3d498d527d3e6acd8;hb=HEAD May 04 09:25:27 906 struct ubi_scan_info *ubi_scan(struct ubi_device *ubi) May 04 09:25:34 ^^ this is an alternative.... May 04 09:26:55 it's kernel function May 04 09:27:03 not sure there is way to run it from userland May 04 09:30:16 The modern way of mounting UBIFS is mounting UBI volume character device nodes, e.g.: May 04 09:30:18 $ mount -t ubifs /dev/ubi0_0 /mnt/ubifs May 04 09:30:19 will mount UBIFS to UBI volume 0 on UBI device 0. This is the easiest way to mount UBIFS, but it is supported only in kernels starting from version 2.6.32. May 04 09:30:48 hm.. so on modern kernels after ubiattach /dev/ubiX_Y should appear May 04 09:30:55 it does May 04 09:31:25 no need to mount sysfs :) May 04 09:31:42 if devtmpfs is catching this May 04 09:31:59 hm.. at boot there is only /dev/ubi_ctrl May 04 09:32:24 interesting to look what is there after ubiattach May 04 09:32:35 I should try w/out udev May 04 09:32:59 or in the initramfs, better May 04 09:33:17 add busybox shell there :) May 04 09:33:29 after ubiatatch you get /dev/ubi0 May 04 09:33:37 we may check /bin/sh presence btw May 04 09:33:44 and if there is a volume, ubi0_0 May 04 09:33:48 and show "Run Shell" item in menu May 04 11:09:30 Jay7: http://code.google.com/p/wl-700ge/source/detail?spec=svn188&r=188 May 04 11:09:58 o_O May 04 11:10:09 why not /proc ? May 04 11:10:20 look at the tty, maybe is worth May 04 11:10:45 still I don't see the point... May 04 11:13:04 btw you can add kexecboot can boot Android. Maybe someone doesn't know it May 04 11:16:02 seems they check for ttySX May 04 11:17:58 ant_work: I miss the point too May 04 11:18:14 ttyS check is commenter May 04 11:18:16 *d May 04 11:18:28 ah May 04 11:18:36 ofs++ May 04 11:18:39 hm.. May 04 11:18:57 no, still can't catch :) May 04 11:19:19 malloc(sizeof(str_dev)+strlen(value)) May 04 11:20:53 and even ,we have tested console=tty1 console=ttyS0,115200n8 May 04 11:22:24 ah May 04 11:22:27 value[3] May 04 11:22:42 is S and not a number May 04 11:24:07 i.e. return 0 in line 22 May 04 11:25:02 only thing is strncat vs strcat May 04 11:25:11 I mean (value[3] > '0') returns true for pure case... May 04 11:25:22 may be his gcc complains about strcat usage :) May 04 11:25:42 no, it is partially right (maybe) May 04 11:25:55 ant_work: but value[3]<'9' will break this :) May 04 11:26:55 anyway in that case the check is incorrect May 04 11:26:58 but.. May 04 11:27:19 yes, because we need only ttyX case May 04 11:27:23 you can have ttyA0 ... May 04 11:27:26 for cursor switching May 04 11:27:42 that code is wrong anyway May 04 11:27:44 yes, vt console is tty, sure May 04 11:27:55 but I don't know how to get tty under fb May 04 11:28:35 only partially useful patch is about procfs May 04 11:28:53 but this should be wrapped by #ifdef and configure option May 04 11:29:20 hm.. I know what is wrong with our check May 04 11:29:28 tty0123abc May 04 11:30:06 even just tty0blablabla May 04 11:30:35 with strncat it will copy only tty0 May 04 11:30:55 but tty11 will be broken too May 04 11:31:05 heh May 04 11:32:01 main question is "why this man doesn't mailed patches to us?" May 04 11:32:56 anyway the akefile installs empty /proc :) May 04 11:33:01 *Makefile May 04 11:34:32 seems he just have already mounted procfs May 04 11:34:42 kexecboot is called from script May 04 11:35:22 http://code.google.com/p/wl-700ge/source/browse/trunk/package/kexecboot/files/kexecboot.init May 04 11:36:32 ah, ok, check for presence before May 04 11:38:03 Jay7: and here talks about a bug http://dev.openaos.org/wiki/Gen8Linux2.6.37 May 04 11:38:27 on openAOS Note2: kexecboot has a bug, if right after boot you press 'Power' twice, it hangs. May 04 11:38:44 what is Gen8? May 04 11:39:06 ah.. seems tablet May 04 11:39:44 hm.. powered by kexecboot :)) May 04 11:40:27 http://www.openaos.org/wp-upload/gen8/2011-08-31/dev/ May 04 11:40:36 volume up/down patch May 04 11:41:48 http://dev.openaos.org/wiki/SettingUpMultiRootGen8 May 04 11:42:05 I think we already have the volume patch May 04 11:42:44 btw: ttyACM0: USB ACM device May 04 11:43:24 http://dev.openaos.org/wiki/FBMenu May 04 11:46:09 another one... May 04 11:47:11 just "dialog" replacement May 04 11:47:38 look at note about OpenEmbedded at top :) May 04 11:47:43 about glibc :) May 04 11:48:31 I think that wrt size we are almost at the state-of-the art May 04 11:48:49 it is true, though, the new gui θ icons are heavier May 04 11:49:02 but just some Kb May 04 11:49:20 they have very good compression ratio :) May 04 11:50:51 textui only was 16-17 kb less May 04 11:51:29 btw, remember to split the headers! May 04 11:51:31 textui should be improved so difference will be little lower :) May 04 11:51:43 yes, I remember May 04 12:56:57 Jay7: about ubiattach, I did a fork + execve. Better ideas? May 04 12:57:27 use same way as first kexec is called May 04 13:03:10 ah, yes, fexecw() * (fork, execve and wait) May 04 15:21:17 bluelightning: you'll have harder times wrt the tsc with the new member ;) May 04 15:22:04 ant_work: I hope not... May 04 15:22:27 is just a feeling atm May 04 15:28:18 now, about us, we didn't do Easter Cleanings in meta-hh May 04 15:28:40 what about ipaqs? May 04 15:28:59 can we move the 1940 and 4700 to linux-yocto_3.2 ? May 04 15:29:13 and remove linux 3.1 and 3.2 ? May 04 15:29:34 and .git as well (as for me) May 04 15:30:33 (and kexecboot versions ofc) May 04 15:40:21 heh May 04 15:40:33 collie is on the safe side... May 04 15:40:40 Russel King: May 04 15:40:44 "What I'm basically pointing out is that the majority of my infrastructure May 04 15:40:44 here runs on ARMv4 StrongARM based platforms, and I don't see this May 04 15:40:44 changing any time soon. As long as that is the case, I have a vested May 04 15:40:44 and selfish interest in keeping the kernel buildable for ARMv4. May 04 15:40:44 " May 04 15:40:53 37 mins ago :) May 04 16:08:41 awesome :) May 04 16:09:12 as for the ipaqs, I just need to make some time to do testing May 04 16:09:30 I've lost touch with hx4700 status May 04 16:13:11 spitz tester ready for duty here! May 04 16:18:41 as for the users, we really need a sort of feed, or we just use Angstrom feeds. Honestbly I last built A. one year ago... May 04 16:19:26 yes May 04 16:19:27 afaik there are still a couple of rough edges for Z May 04 16:19:41 udev and xserver for sure May 04 16:20:01 never tried systemd but reading ML is horror-story May 04 16:20:24 unless it's fixed now May 04 16:21:06 as for Narcissus, the scripts basically need just the EXTRA's for jffs2 and ubi May 04 16:21:39 the latter is a bit problematic May 04 16:21:55 I have a pending patch for this but cannot test collie and tosa May 04 16:22:32 the number of nodes differs btw models May 04 16:22:56 the subpages on poodle/corgi are of 256b May 04 16:23:04 still to check on spitz May 04 16:23:45 finally, we should split the spitz family for the SL-C3000 has less nand May 04 16:26:28 just 5120k May 04 16:26:38 for mtd 2 May 04 16:27:28 and borzoi and terrier also differs :/ May 04 16:27:57 same for 750 / and 760 May 04 16:31:18 dunno if it is worth, though. Maybe a well written ZAURUS.UBI is enough ;) May 04 16:31:21 ant_work: i havent bothered to play with X but im using stock debian udev now. May 04 16:31:32 seems to work nicely on my spitz May 04 16:32:28 i made a totally fresh rootfs using debootstrap with no funky stuff. May 04 16:32:37 the images built with OE are not yet sopolished May 04 16:32:45 maybe opie-image May 04 16:33:14 but the X images are still needing ts calibration May 04 16:33:40 bluelightning: maybe it is enough to provide some pointercal values in the BSP May 04 16:34:28 but we need the xinput-calibrator in oe-core May 04 16:34:53 k. was against, he said oe-core should be thinner and not fatter May 04 16:35:32 fun.. we already have xinput in oe-core May 04 16:36:10 I'm sure he'll nak if I propose the patch May 04 16:37:52 ant_work: propose it anyway, then if you feel like it is not being dealt with appropriately it can be brought up at the TSC level May 04 16:38:11 it's a small think, I hope this will not happen May 04 16:39:38 (anyway on Z the calibration is bad but that's another problem) **** ENDING LOGGING AT Sat May 05 03:00:01 2012 **** BEGIN LOGGING AT Sat May 05 03:00:02 2012 May 05 05:40:12 hello May 05 17:53:59 hi May 05 23:46:25 Jay7: we might need to pass a config parameter for ubiattach: "-O" May 05 23:47:22 for some models it sould be -O 512 (poodle, c7x0) and for spitz /akita -O 2048 May 05 23:47:50 so to be safe pass it as config option May 05 23:47:55 gn **** ENDING LOGGING AT Sun May 06 03:00:01 2012 **** BEGIN LOGGING AT Sun May 06 03:00:02 2012 **** ENDING LOGGING AT Mon May 07 02:59:59 2012 **** BEGIN LOGGING AT Mon May 07 02:59:59 2012 May 07 15:31:46 hey May 07 15:52:58 hm.. we have some new people here May 07 15:55:42 found your project via google, looks perfect for what i need to do May 07 15:56:05 sounds good :) May 07 15:56:23 just trying to package together a working initramfs containing it now :P May 07 15:58:20 this shouldn't be hard task :) May 07 15:58:41 it's proving a bit of a pain for reasons i haven't figured out yet May 07 15:58:53 just some dirs, kexec and kexecboot binaries May 07 15:58:59 well, yeah May 07 15:59:08 but the only way to boot the system is via a kernel.img file May 07 15:59:16 so no way to provide a separate initrd May 07 15:59:20 so i've cooked it into the kernel May 07 15:59:24 but i can't even get a busybox shell going May 07 15:59:28 same way we doing for zauruses May 07 16:00:00 this is for a raspberry pi May 07 16:00:07 ah, cool May 07 16:00:22 I want to buy one later :) May 07 16:00:50 yeah, the boot setup by default is really limited May 07 16:00:58 so i figure kexecboot is a cheap way to boot from USB etc May 07 16:01:11 dwc_otg.lpm_enable=0 console=ttyAMA0,115200 kgdboc=ttyAMA0,115200 console=tty1 root=/dev/ram0 May 07 16:01:14 oops, wrong window, sorry May 07 16:01:20 sure.. that if for what it was designed :) May 07 16:01:30 s/if/it/ May 07 16:01:47 sh*t.. s/if/is/ :) May 07 16:04:13 gotta hope kexec works at all i guess May 07 16:07:11 better to check from 'normal' userland May 07 16:07:40 i.e. boot any supplied/known to work environment and try to play with kexec from there May 07 16:08:01 well there's a suppled emergency kernel May 07 16:08:13 which i've kludged the kexec/kexecboot binaries onto May 07 16:08:22 but kexecboot doesn't start, saying it can't read /dev/tty1 May 07 16:08:43 we are going to support other methods (switch_root/losetup+switch_root) but now only kexec May 07 16:08:54 is /dev/tty1 there? May 07 16:09:08 no, which is why that error appears, lemme try mknod May 07 16:09:22 you may change it to use tty0 May 07 16:09:34 iirc, it is read from kernel cmdline May 07 16:09:44 tty0 isn't there either May 07 16:10:04 i did try kexecboot from a known working environment, where i saw the menu May 07 16:10:42 nice.. at least framebuffer is working then :) May 07 16:11:25 anyway, it should be possible to run w/o ttyX May 07 16:12:07 it's kind of ugly code only for switching off cursor on framebuffer console May 07 16:12:10 if i try to use kexec directly, the last thing i see is "Bye!" May 07 16:12:18 and then nothing, which makes me think kexec isn't working May 07 16:12:29 very possible :( May 07 16:12:47 try with uncompressed destination kernel May 07 16:13:05 seems that the kexec I have doesn't support that May 07 16:13:10 says zImage in the supported image types bit May 07 16:13:17 i just grabbed this kexec binary from the debian image May 07 16:13:27 Andrea may know more but he is outta here May 07 16:14:28 what kernel image type RPi is using? May 07 16:14:31 uImage? May 07 16:14:34 Image May 07 16:14:57 well.. you may try to compile own zImage May 07 16:15:24 but you need config and environment May 07 16:15:35 yeah, i have those May 07 16:15:41 i've built a zImage, that's what I tried to boot May 07 16:15:42 (and patches may be) May 07 16:16:10 which compression you are using? May 07 16:16:22 better to use gzip for testing time at least May 07 16:16:24 not lzma May 07 16:16:37 for both - kernel and initramfs May 07 16:16:50 using gzip yep May 07 16:19:10 may be good idea to wait for Andrea to appear here May 07 16:19:16 his nich is ant_* May 07 16:19:23 *nick May 07 16:19:47 he have more experience with kexec debugging on different systems :) May 07 16:19:57 at least in theory :) May 07 16:19:58 alright cool, i'll idle around **** ENDING LOGGING AT Tue May 08 03:00:01 2012 **** BEGIN LOGGING AT Tue May 08 03:00:02 2012 May 08 17:29:01 https://github.com/djwillis/meta-raspberrypi May 08 17:29:21 layer for raspberry pi May 08 17:29:25 indeed May 08 17:29:51 it hasn't been announced because some of the firmware licensing is still being properly documented **** ENDING LOGGING AT Wed May 09 02:59:58 2012 **** BEGIN LOGGING AT Wed May 09 02:59:59 2012 **** ENDING LOGGING AT Thu May 10 03:00:03 2012 **** BEGIN LOGGING AT Thu May 10 03:00:04 2012 May 10 09:18:22 seems Andrea is busy these days.. May 10 11:59:27 ant_work: hi May 10 11:59:51 ant_work: Hexxeh have Raspberry Pi and trying to run kexec[boot] on it May 10 12:00:15 hi there May 10 12:00:17 seems his kexec version is non-working May 10 12:00:28 can you help him with debugging? :) May 10 12:01:32 Jay7: somebody produced a log they got via the serial console May 10 12:01:36 lemme see if i can find it May 10 12:01:52 Hexxeh: log of kexec'ed kernel? May 10 12:01:58 yeah May 10 12:02:31 gah. paste expired. May 10 12:02:41 i'll ask him for the log again once he logs on later today. May 10 12:02:41 it happens May 10 12:12:54 so it's arm11/armv6 May 10 12:13:28 long ago I've seena patch onkexec m/l for arm v6/7 so probably today is fixed May 10 12:14:52 as for us, we are using kexec-tools-klibc_2.0.2.bb May 10 12:15:04 maybe the fix is in 2.0.3 May 10 12:15:32 try with a normal glibc/compiled kexec-tools 2.0.3 May 10 12:16:04 ...but maybe it was a kernel patch for kexec..can't remember ffhand May 10 12:22:35 * ant_work notes kexec-klibc needs updates May 10 12:24:32 ant_work: hi May 10 12:24:42 hi Paul May 10 16:26:24 Jay7: any news from CELF? May 10 16:26:49 ant_work: nothing May 10 16:28:58 Jay7: btw adding ubiattach needs a bit more work than initially thought: we need to add config options for 1) patch 2) args (-O) May 10 16:29:27 s/patch/path/ May 10 16:29:47 much like for kexec May 10 16:41:46 bluelightning: This patch adds support for the HP iPAQ hx4700 to the existing May 10 16:41:46 > pxa2xx-pcmcia driver. May 10 16:43:08 ant_work: which patch? May 10 16:44:07 [PATCH RESEND 1/3 v7] ARM: pxa: hx4700: Add PCMCIA/CF platform device May 10 16:44:13 applied yesterday May 10 16:44:28 probably for 3.4 May 10 16:46:31 yes, I read today in the comments of [PATCH] ARM: mach-pxa: Use REGULATOR_SUPPLY macro May 10 16:46:40 The hx4700 patch works for me on linux-3.4.0-rc6. May 10 16:46:45 Tested-by: Paul Parsons May 10 16:47:11 we have to ask zeddii about 3.4. iirc there is already a recipe somewhere May 10 16:57:03 bluelightning: when doing the refresh, pls try first adding the new machines to the linux-yocto_ kernel May 10 16:57:20 ant_work: ok May 10 16:58:59 once done for some devices, then we'll have to image a sort of parser to translate the machine_features into .cfg fragments May 10 16:59:42 and finally clean out the machine.cfg configs May 10 17:01:26 gotta go now, bbl **** ENDING LOGGING AT Fri May 11 02:59:58 2012 **** BEGIN LOGGING AT Fri May 11 02:59:59 2012 May 11 07:52:06 ok, klibc builds now with linux-libc-headers. All but kexec-tools-klibc is building: I'll try to upgrade and fix it this evening May 11 15:36:34 bluelightning: how good/bad is qt 4.8 today ? May 11 15:37:07 I know folks basing a project on Yocto 1.2 but needing latest QT May 11 15:37:53 QTE even May 11 15:41:31 ant_work: it should be OK, I've not heard too many complaints and I do know people are using it **** ENDING LOGGING AT Sat May 12 02:59:58 2012 **** BEGIN LOGGING AT Sat May 12 02:59:59 2012 May 12 23:03:19 hi ant__ May 12 23:03:43 have been testing spitz today May 12 23:04:09 hello May 12 23:04:36 bunch of things broken unfortunately :/ May 12 23:04:55 X images? May 12 23:05:01 no, I'm testing with Opie May 12 23:05:13 mostly minor stuff though, like certain things not being installed May 12 23:05:24 I have some kernel interface updates to do in Opie as well it seems May 12 23:06:12 iirc the modules were missing May 12 23:06:41 I loaded the latest kexecboot as well May 12 23:06:54 I'm now mostly booting core-image-base for testing... May 12 23:07:15 works OK but the menu responds to key presses quite slowly, any ideas about that? May 12 23:07:19 menu is slow, isn't? May 12 23:07:37 seems a spi bus issue on pxa270 May 12 23:07:38 ah, you've already found that then May 12 23:07:51 is ok on c7x0 and poodle May 12 23:08:04 maybe we have to change th eevent rate May 12 23:08:10 in kexecboot.cfg May 12 23:08:22 in kexecboot-cfg.bb May 12 23:09:37 same params were ok with 2.6x kernels btw. it's indeed the new matrix kb driver on the spi bus May 12 23:09:47 :/ May 12 23:10:20 argh May 12 23:10:27 new kernels always bring new problems May 12 23:11:08 good news are I've seen some related fix months ago, maybe are already in 3.3 May 12 23:11:45 which opie version is to test btw? May 12 23:11:53 latest from git May 12 23:11:57 ok May 12 23:12:16 I have some meta-opie updates locally though May 12 23:12:27 (mainly to make it work with latest opie) May 12 23:13:14 I'm now trying to solve an 'easy' issue: I'm pulling now i686 headers adapting that debian klibc patch :/ May 12 23:14:11 -> Error: bad instruction `bswapl r3'. May 12 23:41:34 \°/ May 12 23:41:37 yea! May 12 23:45:54 ant__: fixed? May 12 23:46:42 yes, now looking for the best variable to use May 13 00:12:20 much better now May 13 00:12:28 will send tomorrow May 13 00:12:30 gn **** ENDING LOGGING AT Sun May 13 02:59:59 2012 **** BEGIN LOGGING AT Sun May 13 02:59:59 2012 May 13 16:44:00 So kexecboot is just an init stepping-stone that allows the user to select the kernel at load-time (not unlike an OS loader)? May 13 16:53:15 yes May 13 16:53:30 'linux as a bootloader' **** ENDING LOGGING AT Mon May 14 02:59:59 2012 **** BEGIN LOGGING AT Mon May 14 02:59:59 2012 May 14 11:11:18 Jay7: any luck with kexec for th eraspberry-pi folks? May 14 11:11:43 ask Hexxeh :) May 14 11:11:56 afaik it is stalled a bit May 14 11:12:47 I'll work on latest kexec-tools later today. I think I'll commit a 2.0.3 and maybe a _git version if necessary May 14 11:13:08 in klibc flavor ofc May 14 11:13:13 he is using some debian-based kexec tool May 14 11:13:27 but I think the problems are libc-independent :/ May 14 11:13:36 may be good idea to find someone who is bulding OE for RP May 14 11:13:50 or build static kexec version May 14 11:13:53 newest May 14 11:14:21 our kexec-tools-klibc are probably overpatched but still do work on armv4/armv5te May 14 11:22:24 ah May 14 11:22:26 http://lists.infradead.org/pipermail/kexec/2011-October/005627.html May 14 11:25:10 seems still open issue, I don't see patches committed May 14 11:28:13 Hexxeh: ^^ maybe worth a try May 14 11:33:28 have kexec some debugging routines inside? May 14 11:34:13 anyway, he said that last line on console was 'Bye' May 14 11:34:22 so kexec tried to start kernel May 14 11:34:41 yes May 14 11:35:16 sounds like the ben-nanonote issue, machine not properly re-initialized May 14 11:36:03 I think the problem is on the kernel side May 14 12:47:11 Jay7: I've seen CELF has a page with links to beagleboard and raspberry-pi May 14 12:47:29 it would'nt be bad if kexecboot was mentioned there ;) May 14 12:50:21 Jay7: http://us.generation-nt.com/answer/patch-fix-kexec-reboot-arm-help-205749271.html May 14 12:50:39 yeah May 14 12:50:44 ^^ Peter Chubb fixed it on beagleboard (after that presentation where it was broken :p) May 14 12:50:57 nice :) May 14 12:51:30 or even complicated like this one: http://forum.xda-developers.com/showthread.php?t=1205288&page=17 May 14 12:53:16 Jay7: well, better to keep it only for armv7 :) May 14 12:53:17 http://www.gossamer-threads.com/lists/linux/kernel/1467521?do=post_view_threaded#1467521 May 14 12:53:31 breaks pre-ARMv6 May 14 13:05:41 we can't keep it for armv7 :) May 14 13:05:46 only May 14 13:05:55 you know why :) May 14 13:18:44 I'm thinking the kernels can easily be patched May 14 14:42:29 Hexxeh: any (good) news wrt kexec ? May 14 14:42:51 nothing :/ May 14 14:43:00 somebody on the channel has an interesting theory as to how to get it working May 14 14:43:05 RaYmAn May 14 14:43:10 pls try th epatch I posted above May 14 14:43:25 http://us.generation-nt.com/answer/patch-fix-kexec-reboot-arm-help-205749271.html May 14 15:06:58 Hexxeh: note the patch applies up to linux 3.2 May 14 15:07:06 3.3 has different source May 14 15:07:07 http://lxr.free-electrons.com/source/arch/arm/kernel/machine_kexec.c?a=arm May 14 15:08:30 somebody posted a log they got via the serial port May 14 15:08:34 it does load the new kernel May 14 15:08:43 but that gets stuck reinitialisating the framebuffer May 14 15:15:20 ah, maybe wrong cmdline May 14 16:24:16 I'm working on armv7 right now... May 14 16:24:21 Kernel 2.6.32.9 May 14 16:26:13 can you kexec another kernel on your machine? May 14 16:27:02 ideally we should test kernel 3.3 after big refactoring of kexec stff May 14 16:27:29 I'm waiting for 3.4... May 14 16:29:11 Jay7: I'm thinking we should first check the fstype for lvm2 and ubi then start the loop of devicescan May 14 16:29:30 ant_work: I wish I could even get the last bit of messages from kexec-tools May 14 16:29:54 afaik, it reboots and jumps to the new kernel May 14 16:30:06 then it is fine :) May 14 16:30:28 ant_work: do you mean do special check before using fstype? May 14 16:30:54 yes, to attach the volumes May 14 16:31:15 first run you only check for those containers May 14 16:31:24 second run for fstype May 14 16:32:15 but really we should not scan /mtdblock but /ubi0 May 14 16:32:33 once we know it's ubi May 14 16:33:36 still thinking about the best approach... May 14 16:33:41 well, that's right idea May 14 16:33:49 in my tests all was hardcoded :) May 14 16:33:57 yes, I can't suggest best approach right now too May 14 16:43:30 Jay7: I'll try to write all in an external function call May 14 16:43:38 then you'll integrate it better :) May 14 16:44:45 this evening I hope to quickly update kexec-tools, then I should have time for kexec May 14 16:45:28 (I could gain the attention of GNUtoo about ubifs kernel patches so I'm free in that regard :) May 14 16:47:34 ok, reminder sent :p May 14 16:47:36 bbl **** ENDING LOGGING AT Tue May 15 02:59:59 2012 **** BEGIN LOGGING AT Tue May 15 02:59:59 2012 May 15 09:42:34 ant_work: btw I notice the hinge stuff is not working anymore in Opie... is that used at all with X-based images on spitz? May 15 09:43:48 hm... wasn't a zaurusd thing? May 15 09:49:05 could be May 15 10:00:26 iirc kernel creates the 3 interfaces (ts, kb, gpio-keys) May 15 10:01:01 EV_SW May 15 10:01:20 to detect lid state **** ENDING LOGGING AT Wed May 16 03:00:01 2012 **** BEGIN LOGGING AT Wed May 16 03:00:01 2012 May 16 16:01:42 Tequila: I've seen you had issues with cmdline length May 16 16:01:50 Jay7: https://gitorious.org/~tequila/ac100/tequilas-ac100-multiboot/blobs/master/patches/kexecboot-122708bc.patch May 16 16:02:15 iirc we had some long discussion about the max size May 16 16:05:08 Tequila: and about "we want to be able to specify May 16 16:05:08 a different root then the one where the boot.cfg file is found." afaik it is possible just editing boot.cfg and adding a new root= there May 16 16:09:06 i.e. you can just keep the kernel on the SD/CF partition and have the rootfs elsewhere **** ENDING LOGGING AT Thu May 17 02:59:59 2012 **** BEGIN LOGGING AT Thu May 17 03:00:00 2012 **** ENDING LOGGING AT Fri May 18 02:59:58 2012 **** BEGIN LOGGING AT Fri May 18 02:59:58 2012 May 18 22:12:59 * ant__ is re-emerging pulseadio with equalizer on its xeon quad just to see..they say it sucks 3-5% cpu on i5 2.3Ghz so maybe it is a bit too heavy for older handheld ;) May 18 22:15:34 ant__: hi May 18 22:15:42 hey May 18 22:15:44 ant__: have you tested sound recently on spitz? May 18 22:15:51 not really May 18 22:16:03 I'm planning to dive is unserland :) May 18 22:16:06 I'm not sure we're building the right modules, at least they don't get installed by default even if we do May 18 22:16:28 iirc I've ehm removed the modules-autoload stuff May 18 22:16:36 from zaurus.inc May 18 22:17:04 have yet to understand how kmod copes with that May 18 22:17:05 oh, I don't remember about that... why was that done? May 18 22:17:22 iirc they were beautifulli ignored May 18 22:17:31 -i +y May 18 22:17:50 lemme see May 18 22:17:53 hmm ok May 18 22:18:01 well somehow the drivers have to get loaded May 18 22:18:50 zaurus.inc: remove module_autoload remnants, now built in kernel May 18 22:20:00 scary FIXME isn't ? May 18 22:20:10 http://cgit.openembedded.org/meta-handheld/commit/conf/machine/include/zaurus.inc?id=08855d4a059a1fa19f77a83978ea8cfb0067069d May 18 22:23:03 bluelightning: I see May 18 22:23:05 http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto/spitz/spitz.cfg May 18 22:23:42 her we should probably write some kind of template for the config fragments May 18 22:23:49 let's call it a policy May 18 22:24:29 I have yet to clean out the .configs.. I'll bother zeddii about the script he was using May 18 22:24:51 it doesn't look too clear, though May 18 22:25:11 there are machine-specific options which ought imho be compiled in kernel May 18 22:25:28 the rest of generic options can be deployed as module May 18 22:26:11 (just talking about sound) May 18 22:27:15 i.e. CONFIG_SND_PXA2XX_SOC_SPITZ=m May 18 22:34:14 right, sounds good May 18 22:36:21 well, this one is hw-specific so the .config is wrong :( May 18 22:39:14 I hope that 'XOR' script will help us to distill a pure configuration file for the devices we have in meta-hh **** ENDING LOGGING AT Sat May 19 02:59:58 2012 **** BEGIN LOGGING AT Sat May 19 02:59:59 2012 **** ENDING LOGGING AT Sun May 20 02:59:58 2012 **** BEGIN LOGGING AT Sun May 20 02:59:59 2012 May 20 20:36:13 hi Jay7 May 20 20:36:31 I've uploaded some new (ugly) screenshots May 20 20:37:56 note how selecting the last item goes out-of-border bottom May 20 20:38:08 just cosmetic **** ENDING LOGGING AT Mon May 21 02:59:59 2012 **** BEGIN LOGGING AT Mon May 21 02:59:59 2012 **** ENDING LOGGING AT Tue May 22 02:59:58 2012 **** BEGIN LOGGING AT Tue May 22 02:59:59 2012 May 22 22:41:39 hi bluelightning May 22 22:41:45 hi ant__ May 22 22:42:09 ant__: some movement on hx4700 btw, I'm about to test a new image with a 3.4 kernel May 22 22:42:20 nice May 22 22:43:05 I'm testing the last things with the klibc guys atm. hpa is back ! May 22 22:43:42 cool :) May 22 22:44:13 about 3.4 (and defconfigs) we have to ping Bruce, also wrt his script May 22 22:44:41 atm I've just updated the old 3.2 linux.inc recipe we had May 22 22:45:07 I'm not sure when linux-yocto 3.4 is arriving but I know bruce is working hard on it May 22 22:45:16 ideally we'd love to remove 3.1 and git May 22 22:45:28 (and the very old stuff ofc) May 22 22:51:11 martin doesn't want git anymore? May 22 22:52:03 I don't know, just don't see any activity May 22 22:53:05 we are at the point the devices do boot, I'd like to polish the remaining bugs instead of chasing the new ones ;) May 22 22:54:30 we have to stop at some point (3.4 ?) and try to fix power management, ts jitter and other outstanding kernel issues May 22 22:59:30 ant__: yes pm is slightly broken **** ENDING LOGGING AT Wed May 23 02:59:59 2012 **** BEGIN LOGGING AT Wed May 23 02:59:59 2012 **** ENDING LOGGING AT Wed May 23 04:35:22 2012 **** BEGIN LOGGING AT Wed May 23 04:36:15 2012 **** ENDING LOGGING AT Thu May 24 02:59:58 2012 **** BEGIN LOGGING AT Thu May 24 02:59:58 2012 **** ENDING LOGGING AT Fri May 25 02:59:59 2012 **** BEGIN LOGGING AT Fri May 25 02:59:59 2012 May 25 04:44:47 bidding for proposals are started May 25 04:44:55 CEWG I mean May 25 04:45:08 s/are/is/ May 25 10:06:43 ant_work: hi May 25 10:07:17 ant_work: any good news about Z userland (x11/opie)? :) May 25 10:07:48 * Jay7 going to lvee.org conference with talk about kexecboot May 25 13:21:21 Jay7: hello May 25 13:22:05 not much time for testing lately...now is better May 25 13:22:28 it seems there are issues with sound / modules not loaded May 25 13:22:46 then the old calibration problem May 25 13:23:00 finally power management May 25 13:23:11 it opie affected by ts calibration problem too? May 25 13:23:44 hm.. I was testing on poodle which has not the issue... May 25 13:23:57 different TS chip iirc May 25 13:24:02 probably not, I think still uses tslib May 25 13:24:06 btw, tosa may be not affected too May 25 13:24:30 I'll check this weekend maybe May 25 13:25:06 iirc bluelightning has some more opie-updates to commit May 25 13:25:37 btw, is meta-oe still required to build some core x11 image? May 25 13:32:29 no, stay away May 25 13:32:43 maybe opie needs it May 25 13:33:14 as it is, udev and xserver of oe-core are starting ok May 25 13:33:34 months ago, adding meta-oe broke horribly May 25 13:33:49 maybe now is fixed, I doubt... **** ENDING LOGGING AT Sat May 26 02:59:57 2012 **** BEGIN LOGGING AT Sat May 26 02:59:58 2012 **** ENDING LOGGING AT Sun May 27 02:59:59 2012 **** BEGIN LOGGING AT Sun May 27 02:59:59 2012 May 27 22:18:10 jeez...old poodle won't find my Sandisk Ultra 4GC CF :/ May 27 22:24:29 wow qt4e-demo-image booting on poodle! May 27 22:24:50 even tslib calibration did work May 27 22:25:12 it's really too small... May 27 22:25:30 I can't read the menu's on the left (it is not rotated btw) May 27 22:48:12 I meant this screen http://doc-snapshot.qt-project.org/4.8/qtdemo.html May 27 22:50:07 ant__: hi May 27 22:50:16 ant__: it's not really designed for low-res screens I think May 27 22:50:37 ant__: in fact I believe it's not well suited for resizing of any kind May 27 22:50:58 well, now I've seen where th elauch link appears :) May 27 22:51:12 *launch I suppose, can't really read May 27 22:51:27 still ,it says the examples have not been built .. ? May 27 22:51:57 236MB no examples? May 27 22:52:33 bluelightning: it's all proportioned (small...) May 27 22:52:52 all fits in the screen (somehow) May 27 22:54:16 "Could not launch the example. Ensure that it has been built." May 27 22:56:01 ... it's all in /usr/bin/qtopia... May 27 22:56:11 launcher seems broken :/ May 27 22:56:19 some path? May 27 23:00:38 hmm, it worked when I last tested it, but that was some time ago May 27 23:01:58 root@poodle:~# /usr/bin/qtopia/demos/chip May 27 23:01:58 -sh: /usr/bin/qtopia/demos/chip: Permission denied May 27 23:02:08 ? May 27 23:02:55 analyzing on host: May 27 23:02:57 file chip May 27 23:02:57 chip: ELF 32-bit LSB executable, ARM, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.16, stripped May 27 23:05:57 ehm... May 27 23:06:10 /chip/chip May 27 23:08:34 oh.. could not start qt assistant May 27 23:11:26 well, I have only 32MB RAM... May 27 23:13:46 Out of memory: Kill process 450 (qtdemoE) score 355 or sacrifice child May 27 23:13:46 Killed process 450 (qtdemoE) total-vm:67000kB, anon-rss:10748kB, file-rss:132kB May 27 23:13:57 that's the end of the experiment :) May 27 23:14:21 funny it could not launch from the menu May 27 23:14:36 I'll retest on 64M Zaurus May 27 23:14:45 vga :) May 27 23:39:08 gn **** ENDING LOGGING AT Mon May 28 02:59:58 2012 **** BEGIN LOGGING AT Mon May 28 02:59:58 2012 May 28 11:00:43 Hexxeh: so I'm reading that finally kexec works May 28 11:07:07 Hexxeh: we'd like to host a kexecboot-kernel for raspberry-pi on kexecboot.org May 28 11:08:12 can you provide one or should I build using that patches? (http://www.bootc.net/archives/2012/05/22/another-new-rpi-kernel/) May 28 21:25:37 ant__: yes pm is slightly broken May 28 21:25:47 slightly too much! May 28 21:26:06 power chord attached on boot = panic :) May 28 21:26:42 attach it afterwards (calibration of qte image) and battery dies...maybe was already to the limit.. May 28 21:26:51 bad luck :/ May 28 21:34:13 oh, and what is that now? May 28 21:34:15 WARNING: Unable to get checksum for linux SRC_URI entry defconfig: file could not be found May 28 21:35:26 ah.. my typo lead to the a.m. warning May 28 21:35:29 my bad May 28 21:35:35 ERROR: Nothing PROVIDES 'qt4-demo-image' May 28 21:42:41 bluelightning: qt4e-demo running smoothly on spitz May 28 21:43:49 ant__: cool :) May 28 21:44:21 maybe some swap could further improve May 28 21:44:34 I'm not sure I had a panic from pm issues but I did experience occasional random freezing, maybe that was just a symptom of a panic May 28 21:44:37 I've tried 4000 chips and gradient demo May 28 21:45:09 abou pm, to have a steady yellow led I have tu pull in, ou, in again May 28 21:45:44 sorry, typing with one hand... May 28 21:46:23 btw qte is in portrait, not rotated May 28 21:49:04 yes I noticed that LED issue as well, I was not sure if the LED not being on indicated it was also not charging; I assumed that was the case May 28 21:49:33 yeah, I'm not surprised that qte 4 knows nothing about the specific device rotation May 28 21:49:48 it took some hacking to get that right in Opie with qte2 May 28 21:50:17 look this: phonon mediaplayer cannot find libgstreamer-gst-plugins-base installed May 28 21:50:22 this was on the ML... May 28 21:50:37 about custom images May 28 21:53:15 ouch..no sound May 28 21:54:02 ah, calibration is also ok on spitz :) May 28 21:57:52 yes I was pleased to see that May 28 22:02:47 I'd like to test sound but it looks like the modules are not loaded May 28 22:02:50 snd-pcm-oss May 28 22:03:07 yeah, I noticed that too May 28 22:03:14 should load the rest May 28 22:03:34 what about the specific device driver(s) for the sound chip? May 28 22:03:46 I think those still have to be loaded manually May 28 22:03:47 those are in kernel, isn't? May 28 22:03:59 are you sure they aren't modules? May 28 22:04:06 if not they probably should be May 28 22:08:17 hm..CONFIG_SND_PXA2XX_SOC_SPITZ=m May 28 22:08:56 I've asked Bruce to send me the script for cleaning the configs May 28 22:09:09 this will be the right occasion to fix that May 28 22:09:53 the userspace part could be done by smthg like #include cfg/sound.scc May 28 22:22:44 would that actually ensure the driver gets loaded at runtime? May 28 22:23:41 to me, it looks like the modules are not of use, almost all options are =y May 28 22:23:58 (in the Yocto kernel practice) May 28 22:27:08 ant__: probably just a configuration choice; if your hardware is more or less fixed then modules aren't really necessary May 28 22:27:49 ant__: but for a zaurus I think we want to stick with modules for most drivers May 28 23:06:10 in order to save memory yes, but then we need a robust autoload May 28 23:07:28 naturally... we used to have that though, I can't see why we would not be able to have it now May 28 23:08:02 we have to see how it's made today wit kmod May 28 23:08:22 all old way are deprecated May 28 23:08:50 (prolly there is a modprobe symlink so anything changes...) May 28 23:09:14 remember that udev loves to load modules too May 28 23:09:40 then no idea about systemd May 28 23:09:58 so I really sit and see where we are going :) May 28 23:11:01 hmm May 28 23:11:50 again, let's start with hw-specific modules compiled in kernel May 28 23:13:01 the rest is realm of metadata configuration May 28 23:13:16 I don't like it :/ May 28 23:13:41 it's just a band-aid on top of not fixing the underlying problem May 28 23:14:53 again, I'm pretty sure recent udev takes care of the modules loading May 28 23:15:11 ah ha May 28 23:15:12 Note: Since udev>=171, the OSS emulation modules (snd_seq_oss, snd_pcm_oss, snd_mixer_oss) are not automatically loaded by default. May 28 23:15:28 that's just oss though May 28 23:15:32 this is archlinux but worth checking... May 28 23:15:37 we're not even getting the alsa modules loaded May 28 23:16:22 we should start and see where the initscript do 'autoload' May 28 23:17:58 I remember having noticed issues long long ago, like having the two mechanisms of modprobe.d and modules-load.d May 28 23:18:39 can't remember offhand which one is used now May 28 23:24:16 anyway the list for zaurus was May 28 23:24:18 http://cgit.openembedded.org/meta-handheld/commit/conf/machine/include/zaurus.inc?id=08855d4a059a1fa19f77a83978ea8cfb0067069d May 28 23:30:38 bitbake -g -u depexp qt4e-demo-image May 28 23:30:42 nice May 28 23:31:35 finalyy the rev-dep like Gentoo's portage :) May 28 23:32:56 I prefer the depgraphs produced by buildhistory (or testlab in the old days) - they're much less complicated May 28 23:33:13 at least for package deps anyway May 28 23:34:16 argh... building opie-image with meta-oe enabled ends up building vala ???!!! May 28 23:34:24 talk about dependency nightmares... May 28 23:34:31 meta-oe ? May 28 23:34:43 yes, meta-opie needs it May 28 23:34:57 I mean some evil commit there? May 28 23:35:10 well, some dependency creep at the very least May 28 23:43:25 gn **** ENDING LOGGING AT Tue May 29 02:59:58 2012 **** BEGIN LOGGING AT Tue May 29 02:59:58 2012 May 29 12:00:01 Jay7: so SanDisk 4GB is not recognized on poodle. fine on spitz.. May 29 12:00:26 maybe increase the delay... May 29 12:00:42 what about rescan? May 29 12:00:56 no luck, reboot and rescan several times May 29 12:01:13 one partition is over 2GB May 29 12:01:31 but kernel is the same... May 29 12:03:09 maybe the combination of old hw + kernel driver is slower on poodle. May 29 12:03:48 strange because I used that on c7x0 wich is pxa25x like poodle. (well, 250 vs 255 iirc) May 29 12:04:20 I remember we tested the delays and discovered spitz was the slowest ;) May 29 12:04:56 btw this happens whith 2 jffs2 images, so long scan time... May 29 12:05:30 strange May 29 12:59:56 bluelightning: have you seen the hx4700 thread on angstrom-users? May 29 13:02:44 ant_work: I've just read it... I don't use the SDG bootloader myself though May 29 13:02:51 ant_work: I use HaRET still May 29 13:03:05 ah, there are two bootloaders even... May 29 13:03:09 although Paul seems to think it has issues May 29 13:03:30 afaik after 2.6.2smthg an atag patch was committed May 29 13:03:47 well, HaRET is a tool you can use to do hardware probing which also supports chainloading linux from within Windows Mobile/PocketPC May 29 13:04:21 bluelightning: afaik th e'old' way of passing bootparams is now deprecated May 29 13:04:34 kexecboot should work, though May 29 13:33:48 alexxy may suggest something on hx4700 May 29 14:14:40 in the worst case they can mangle bootloader's mtdparts :) May 29 14:15:05 in kexecboot ofc May 29 14:17:58 ah, btw, May 29 14:18:00 http://cgit.openembedded.org/openembedded/tree/recipes/initrdscripts/files/01-bootldr-buster.sh?h=org.openembedded.dev May 29 14:34:42 ant_work: interesting... I was not aware of that little gem May 29 14:34:59 we probably need to support bootldr on older ipaqs too... May 29 14:37:24 once you boot with kexecboot the cmdline is external, in /boot so you can fix that. The mtdparts? are passed as they are. May 29 14:38:03 we can define a valid cmdline in kexecboot.cfg May 29 14:38:13 we can define a valid cmdline in kexecboot-cfg.bb May 29 14:38:48 right, if kexecboot is involved its easier **** ENDING LOGGING AT Wed May 30 02:59:58 2012 **** BEGIN LOGGING AT Wed May 30 02:59:58 2012 May 30 08:34:04 Jay7: our friends in #raspberrypi seem oriented to adopt kexecboot May 30 08:34:12 ..and to 'hack' it May 30 08:34:35 so pls split the gui and textui headers before people fork... May 30 08:34:53 (I expect they'll edit gui mostly) May 30 10:30:01 ant_work: hi May 30 10:30:22 I'm writing bid for CEWG :) May 30 10:34:16 :) May 30 10:37:02 oh.. 400 users on #raspberrypi :) May 30 10:37:29 should I make it 401? May 30 10:37:33 :) May 30 10:37:54 bluelightning: now you are 400th :) May 30 10:38:21 I've rounded a bit :) May 30 10:38:50 I see they are using voice.... May 30 10:41:32 ant_work: can you say when there was talk about kexecboot? May 30 10:41:35 I'll read logs May 30 10:42:54 last today :) May 30 10:43:26 btw, why fork? May 30 10:44:03 not enough 'open to hackery' ?!? May 30 10:44:09 I'd prefer to have all work to be mainlined May 30 10:44:13 (they where used to edit a shell script :) May 30 10:44:22 *were May 30 10:44:40 Jay7: btw it seems Tim likes the t-pi very much May 30 10:44:51 r-pi even May 30 10:44:55 Tim Bird? May 30 10:44:57 yes May 30 10:45:15 sh@t! I've send bid proposal already :) May 30 10:45:18 it would be nice to have a downloadable image soon in our site May 30 10:45:32 sometimes I think people not used to the free software world think it's the best idea to just do their own thing... I've noticed that a lot with android kernel stuff May 30 10:45:57 NIH-syndrome :) May 30 10:46:59 well, kind of... sometimes it's about control, other times it's just because the person doesn't really understand that things can work better if you at least try to work upstream May 30 10:49:29 let's try to break it then :) May 30 10:50:09 http://lists.celinuxforum.org/pipermail/celinux-dev/2012-May/000604.html May 30 10:50:10 ^^ May 30 10:50:36 [0:10] Nothing at all May 30 10:50:37 [0:10] It's just not as open to hackery afaik May 30 10:50:40 hehe May 30 10:50:46 ant_work: now I understand :) May 30 10:50:59 heh, shell vs. C May 30 10:56:45 [0:07] ShiftPlusOne: the menu should show all options by default and abort timeout when arrow keys are pressed, too, just sayin' :) May 30 10:57:11 item 5 in CEWG proposal :) May 30 10:59:31 Jay7: I'll be back for the week-end. Remember me to send you the kexecboot patches/hacks for ubifs and klibc. May 30 10:59:40 ant_work: sure May 30 13:04:38 cu, bye **** ENDING LOGGING AT Thu May 31 02:59:58 2012 **** BEGIN LOGGING AT Thu May 31 02:59:58 2012 **** ENDING LOGGING AT Fri Jun 01 02:59:58 2012 **** BEGIN LOGGING AT Fri Jun 01 02:59:58 2012 Jun 01 11:26:44 bluelightning: hi Jun 01 11:27:00 I've noticed strange warnings during parse Jun 01 11:27:25 seems metadata is lacking SRC_URI chksum for 'defconfig' Jun 01 11:28:38 maybe I've b0rked smthg locally... Jun 01 12:28:27 ant_work: it's not SRC_URI checksum it's looking for Jun 01 12:28:51 ant_work: bitbake now will try to get an internal checksum for every local file referenced by SRC_URI Jun 01 12:28:59 so it can determine if it has changed Jun 01 12:29:18 if it can't find it it will show up a warning during parse Jun 01 12:45:45 ah, ok, the issue is I'm not using any 'defconfig' for kernel, maybe busybox uses it Jun 01 12:47:30 and afair the Warnings is not spit out on every bitbake call Jun 01 12:47:59 i.e. I've seen that during -c cleansstate but not rebuilding klibc Jun 01 12:48:28 ..but it was late and I was very tired...I'll have to recheck Jun 01 12:51:47 ant_work: it'll only be printed once unless the recipe changes - it doesn't stop the parsing of the recipe Jun 01 12:52:09 afterwards the recipe data is just read from the cache unless it has changed Jun 01 12:52:35 of course you can just touch conf/local.conf to force a full re-parse Jun 01 12:53:24 ahm, now I get it... Jun 01 12:53:30 it's klibc's defconfig :) Jun 01 12:54:15 it can't be other for I did not pull Jun 01 12:57:06 it does report the recipe (well, the PN)... Jun 01 12:59:53 hm.. Jun 01 13:00:03 maybe I should update bitbake then Jun 01 13:00:16 pretty sure recipe was not named Jun 01 13:03:14 well, thinking again, it can't be klibc's defconfig, it is living in workdir... Jun 01 13:40:01 I wanted to make it report the exact recipe file but unfortunately I don't have access to the data object at the time the warning is generated Jun 01 13:40:09 all I have is PN, so that's what it should be printing Jun 01 13:43:37 the whole idea is nice, I'll keep some logs for you and then will pull Jun 01 20:41:38 bluelightning: so the warnings were Jun 01 20:41:45 WARNING: Unable to get checksum for linux-kexecboot SRC_URI entry defconfig: file could not be found Jun 01 20:42:01 and twice Jun 01 20:42:03 WARNING: Unable to get checksum for linux SRC_URI entry defconfig: file could not be found Jun 01 20:42:03 WARNING: Unable to get checksum for linux SRC_URI entry defconfig: file could not be found Jun 01 20:42:20 in different cooker's logs Jun 01 21:54:53 ant__: ok, I'll have to check into it when I get back **** ENDING LOGGING AT Sat Jun 02 02:59:58 2012 **** BEGIN LOGGING AT Sat Jun 02 02:59:58 2012 **** ENDING LOGGING AT Sun Jun 03 02:59:59 2012 **** BEGIN LOGGING AT Sun Jun 03 02:59:59 2012 Jun 03 14:01:47 jay7:ping Jun 03 17:53:40 dromede__: pong Jun 03 17:53:59 Whats the status on the kexecboot cewg bidding? Jun 03 17:54:10 I've sent my bid Jun 03 17:55:31 waiting for decision **** ENDING LOGGING AT Mon Jun 04 02:59:58 2012 **** BEGIN LOGGING AT Mon Jun 04 02:59:59 2012 Jun 04 09:06:35 not a lot of projects for CEWG 2012 Jun 04 09:07:08 ant_work: yeah Jun 04 09:07:21 today is decision day afaik Jun 04 09:07:26 * Jay7 gone Jun 04 09:07:32 how many did they select last time? 10? Jun 04 09:07:42 can't remember Jun 04 09:08:25 "... at a meeting in Japan on June 5." so late today/tomorrow Jun 04 09:08:39 well, fingers crossed ;) Jun 04 10:29:50 ant_work: I'll be at LVEE.org this time Jun 04 10:29:58 hope to announce sponsorship there :) Jun 04 10:31:51 ant_work: btw, I've tried to run build today and got error in docbook Jun 04 10:31:58 have you seet that? Jun 04 11:05:32 no, I've built with fresh pull of saturday afternoon, though Jun 04 11:05:53 ..but I don't see new commits Jun 04 11:07:48 ah.. may be because I have meta-oe in bblayers Jun 04 11:09:05 heh Jun 04 11:11:15 it i sed that after one year there are still issues adding meta-oe layer Jun 04 11:11:38 the whole systemd change should have been kept in a layer Jun 04 11:11:42 imho Jun 04 11:11:54 ^^ it is sad **** BEGIN LOGGING AT Mon Jun 04 15:56:03 2012 **** ENDING LOGGING AT Tue Jun 05 02:59:57 2012 **** BEGIN LOGGING AT Tue Jun 05 02:59:58 2012 Jun 05 06:27:52 hm... Jun 05 06:28:06 I've installed kexecboot from our site to my akita Jun 05 06:28:14 seems we have problems with CF-cards Jun 05 06:28:37 kexecboot doesn't detect card Jun 05 08:20:48 Jay7: I've a semi-working devicescan.c for you Jun 05 08:21:11 but ideally you'l need a separate function for the ubiattach stuff Jun 05 08:21:49 ant_work: I got problem with CF-card on my akita with kernel from kexecboot.org Jun 05 08:22:29 is it a new CF? Jun 05 08:22:31 seems we don't see CF-card.. Jun 05 08:22:45 no, old, but I've 'flashed' from this card Jun 05 08:22:49 so it is working ok Jun 05 08:22:53 I have issues with poodle and SanDisk 4GB (2-3 years ago) Jun 05 08:23:01 but not on spitz and corgi Jun 05 08:23:06 it's about 256Mb iirc Jun 05 08:23:21 ah, then should just work (if it's not b0rked) Jun 05 08:24:10 my 512MbCF is dying Jun 05 08:24:34 mounts only 1/5 times on PC Jun 05 08:25:07 same for an old 256SD ... years and years old :) Jun 05 08:25:39 imagine we all had concerns with the nand flash...which is just perfect after n^n flashes ... Jun 05 08:26:43 hehe :) Jun 05 08:29:57 back to devicescan, I've injected a if (!strncmp(fstype,"ubifs",5)) at the end of detect_fstype(), just before returning Jun 05 08:32:12 and ubiattach is launched here (paths and argv still hardcoded) Jun 05 08:32:48 the mount was failing, though, because I only create an empty volume Jun 05 08:33:07 I'll retry later with a complete rootfs Jun 05 08:34:23 btw, is perror logged or only displayed? Jun 05 08:36:32 ant_work: perror is only displayed Jun 05 08:36:43 iirc Jun 05 08:36:48 Then I'll have to use log_msg() Jun 05 08:37:22 is to check the fexecw was ok Jun 05 08:37:40 I took the same code as in kexec.c Jun 05 08:37:45 *kexecboot.c Jun 05 08:40:48 Jay7: I'm tempted to try to properly add the configure option for ubiattach path but this will need to be redesigned as well for losetup (like kexec we don't know if there is /usr) Jun 05 08:41:16 same for ubiattach args Jun 05 08:41:46 sure Jun 05 08:42:38 ...so I'll hardcode all and let you sort out the mess :) Jun 05 08:43:29 ok :) Jun 05 08:43:30 btw I've forked the repo on github so you can easily see the progresses (or regressions :D) Jun 05 08:44:11 I'll have just to retest this last patch Jun 05 08:54:49 Jay7: I have yet to see how should we properly transform /dev/mtd* -> /dev/ubi0_0. Maybe check for the presence of the node after ubiattach? Jun 05 08:55:13 yes Jun 05 08:55:44 if mtd* contain ubi then ubiattach it and check for /dev/ubi* Jun 05 08:56:12 always in devscan or later, when mounting? Jun 05 08:56:23 the check I mean Jun 05 08:57:11 imho better is in devscan Jun 05 08:57:31 but I'm out of context right now Jun 05 08:57:50 preparing for trip to LVEE :) Jun 05 08:58:12 ah, is this week-end? Jun 05 08:59:21 yes, I'm starting this evening Jun 05 08:59:48 will be in Moscow tomorrow then going to Belarus Jun 05 09:03:21 ...shame if kexecboot won't work during the demonstration... Jun 05 09:04:17 can you quickly compile a new one with more delay? I could do that in a few hours Jun 05 09:16:04 I can't compile anything because of that docbook error Jun 05 09:27:56 I ahve not yet verifiedthat one, pls remove meta-oe, is only necessary for opie atm Jun 05 09:47:28 I've removed already Jun 05 09:48:54 may be I should install docbook-utils on host Jun 05 09:51:53 hm.. I have it installed Jun 05 09:52:02 strange.. why oe-core trying to build it then/ Jun 05 09:52:39 bitbake -g -u depexp blah.bb Jun 05 09:53:23 you'll see deps and rev-deps Jun 05 09:54:29 ant_work: do you have any usable image to extract on SD? Jun 05 09:55:55 maybe for spitz I have smthg on the CF/SD card Jun 05 10:00:50 started from scratch only for linux-kexecboot Jun 05 10:00:59 much better Jun 05 10:00:59 wait Jun 05 10:01:10 zaurus-installer Jun 05 10:01:24 (is linux-yocto-tiny-kexecboot nowaday) Jun 05 10:01:51 well, it's the same thing after all, minimal diffs Jun 05 10:02:14 so should I build zaurus-installer instead linux-kexecboot now? Jun 05 10:02:30 yes, it's easier Jun 05 10:02:36 hm.. ok Jun 05 10:02:43 l-y-t-k is a bit too long name :) Jun 05 10:03:18 again, the defconfig is the same atm so no big deal Jun 05 10:03:47 my dream is to clean out all the old kernels asap ... Jun 05 10:06:53 what is name for console image in oe-core? Jun 05 10:07:43 core-image-minimal ? **** ENDING LOGGING AT Wed Jun 06 02:59:58 2012 **** BEGIN LOGGING AT Wed Jun 06 02:59:58 2012 **** ENDING LOGGING AT Thu Jun 07 02:59:58 2012 **** BEGIN LOGGING AT Thu Jun 07 02:59:59 2012 Jun 07 21:26:32 great, we have 3.4 Jun 07 21:26:50 in oe-core **** ENDING LOGGING AT Fri Jun 08 02:59:59 2012 **** BEGIN LOGGING AT Fri Jun 08 02:59:59 2012 **** ENDING LOGGING AT Sat Jun 09 02:59:59 2012 **** BEGIN LOGGING AT Sat Jun 09 02:59:59 2012 **** ENDING LOGGING AT Sun Jun 10 03:00:02 2012 **** BEGIN LOGGING AT Sun Jun 10 03:00:03 2012 Jun 10 15:11:05 Jay7: ubifs populates the menu but only launching kexecboot as init Jun 10 15:11:29 I'm trying to add /sys and see why the ubi devices seem missing Jun 10 15:13:51 er.., when launchin *not* as init... Jun 10 21:00:12 Jay7: things are going better now with sysfs Jun 10 21:36:25 hi ant__, all Jun 10 21:37:19 hi bluelightning, wb Jun 10 21:37:28 ant__: thanks :) Jun 10 21:37:43 was on holiday in spain/germany Jun 10 21:38:03 rather long journey... Jun 10 21:39:13 yeah, left previous saturday and got back this evening Jun 10 21:39:40 back and lots of work to do... Jun 10 21:40:04 heh, take it easy Jun 10 23:03:36 bluelightning: think about 3.4 and removing older kernels Jun 10 23:04:18 ant__: if I can get 3.4 working with hx4700 that will be something I can consider... Jun 10 23:04:35 at the moment it seems to be failing right at the point when the kernel starts init Jun 10 23:04:38 and btw linux-kexecboot can be removed from both layers Jun 10 23:04:55 we just keep the yocto- Jun 10 23:04:59 ant__: can you send a patch for that? Jun 10 23:05:15 sure Jun 10 23:06:46 I prefer that because then it's someone who is directly involved in kexecboot development making the change :) Jun 10 23:07:14 np, has to be removed in meta-initramfs before, them it's automatical Jun 10 23:07:21 *then Jun 10 23:19:12 Jay7: in menu it appears /dev/mtdblock3 ubifs Jun 10 23:19:25 which is a blasphemy (char device is used) Jun 10 23:19:43 but I think we can mangle the output or you'll have better ideas... Jun 10 23:42:11 same for cmdline, needing extra tags Jun 11 00:04:41 gn **** ENDING LOGGING AT Mon Jun 11 02:59:58 2012 **** BEGIN LOGGING AT Mon Jun 11 02:59:59 2012 Jun 11 10:57:32 bluelightning, I have a sort of how-to about embedding initramfs in kernel. Now, I can't create pages on oe wiki apparently Jun 11 10:57:56 maybe is worth to put it in Yocto documentation? Jun 11 11:04:04 ant_work: is your user "Ant" ? Jun 11 11:04:23 ant_work: it looks like if you were logged in as that user you should be able to create pages Jun 11 11:06:52 hm.. I can edit, yes Jun 11 11:07:11 imagine I want to add one page in category "User" Jun 11 11:08:06 I never use mediawiki, though I'd expect a 'new' label Jun 11 11:08:45 pls check your e-mail, this doc is better together with kernel, a special case of Jun 11 11:09:19 I'm a bit afraid the wiki will further bitrot Jun 11 11:12:05 btw you can remove that export IMAGE_BASENAME = Jun 11 11:12:16 seems we have it for no reasons in many recipes Jun 11 11:14:25 ant_work: it wouldn't be under User:, that's for user pages Jun 11 11:14:50 ant_work: you should just be able to type a URL to the non-existent page and then select to create it from there Jun 11 11:15:23 ant_work: alternatively you can start with a link to the non-existent page from another page, that way at least it starts linked from somewhere appropriate Jun 11 16:19:15 Jay7: http://imagebin.org/215957 Jun 11 16:29:16 http://kexecboot.org/screenshots Jun 11 16:35:44 bbl **** ENDING LOGGING AT Tue Jun 12 02:59:58 2012 **** BEGIN LOGGING AT Tue Jun 12 02:59:58 2012 Jun 12 06:45:41 morning Jun 12 06:45:46 * Jay7 is back from LVEE Jun 12 07:48:02 Jay7: wb Jun 12 08:48:50 I've updated the screenshot, now with 2 ubi Jun 12 08:50:00 ..picture is too big... Jun 12 09:58:11 well, another user writing to me complaining about Narcissus images Jun 12 09:58:24 bluelightning: will Hob replace Narcissus? Jun 12 09:58:49 how do we give images to the users ? Jun 12 09:58:51 ant_work: that's probably up to koen... he has expressed interest though Jun 12 09:59:17 frankly I still think there's room for pre-built images; I think narcissus has been a step back Jun 12 10:00:01 unfortunately N. images have been regularly broken by udev updates Jun 12 10:42:44 again another thread in A.-users with the same subject Jun 12 10:43:01 yes, we should provide som eimages... Jun 12 10:43:31 at least one core-image-base (instead of A. console-image) Jun 12 10:44:29 we could hostthat in kexecboot site, to give some real images to kexec Jun 12 10:48:04 seems reasonable Jun 12 10:54:48 then we need feeds... Jun 12 14:06:12 Jay7: I have an idea about mangling cmdline without hardcoding Jun 12 14:07:06 to inject ubi.mtd=X root=ubi9_0 (rootfstype=ubifs is already added) Jun 12 14:07:22 but your help is welcome :) Jun 12 14:12:54 bbl Jun 12 20:50:34 Jay7: any news about CEWG? Jun 12 20:51:18 ant__: hi Jun 12 20:51:28 no, still have no news Jun 12 20:52:09 Isee, it took some day last time as well...fingers crossed Jun 12 20:52:23 yeah Jun 12 20:52:40 I've seen new screenshots Jun 12 20:52:55 waiting for patch :) Jun 12 20:53:07 ah, it's on github fork Jun 12 20:53:25 I'm trying now to add the ubi tokens to cmdline Jun 12 20:54:07 you go to github -> kexecboot/kexecboot -> Jun 12 20:54:12 Network Jun 12 20:54:20 ok, I'll look when have time Jun 12 20:54:24 ...atm there is only one fork ;) Jun 12 20:54:50 https://github.com/andrea-adami/kexecboot/commits/master Jun 12 20:56:11 Jay7: note that probably you'll have to verify the allocation I did for the char's Jun 12 20:56:14 :p Jun 12 20:56:30 very probably :D Jun 12 20:57:16 np :) Jun 12 20:57:56 well.. I'm going to bed Jun 12 20:58:07 see you around, gn **** ENDING LOGGING AT Wed Jun 13 02:59:58 2012 **** BEGIN LOGGING AT Wed Jun 13 02:59:58 2012 Jun 13 13:14:35 Jay7: I've done some refinement at home, like having a single 'mount' for both cases Jun 13 13:14:47 same for cmdline processing Jun 13 13:15:10 quick solution is to add the new tags (kernel takes the last root= ) Jun 13 13:15:39 don't forget we have limit on kernel cmdline lenght :) Jun 13 13:15:40 but I'm working at a better solution Jun 13 13:15:47 hm, yes Jun 13 13:16:09 I've added two tags to malloc, I'll have to show you Jun 13 13:16:19 (ubirootfs and ubimtd) Jun 13 13:16:46 root=ubi9_0 mtd.id=X Jun 13 13:17:25 but yes, it's implicit it just works for mtd [0-9] Jun 13 13:17:49 you'll have to extract the numbers from the dev.device Jun 13 13:17:56 for a better solution Jun 13 13:22:19 well.. I got the point Jun 13 13:22:24 will look asap Jun 13 13:23:30 remember there is a design flaw: some options could be passed as constant in configure.ac Jun 13 13:23:54 instead of hardcoding Jun 13 13:24:07 what options you mean? Jun 13 13:24:14 like i.e. the choice to mount ubi9 Jun 13 13:25:21 I've let the code like it is so you'll better see which devices are touched Jun 13 13:26:53 ..and I can't yet imagine the loop for multiple UBI volumes Jun 13 13:27:00 later Jun 13 13:27:06 :) Jun 13 13:28:20 we need another general solution :) Jun 13 13:28:32 like multiple kernels per partition before Jun 13 13:28:57 yes, one boot partition is enough. Who nneds more ;)? Jun 13 13:29:22 per volume I mean Jun 13 13:29:56 anyway, I'm going now. I'll commit some minor changes to the patch and the new patch for cmdline Jun 13 13:30:01 in 4-5 hours Jun 13 13:30:06 see you later **** ENDING LOGGING AT Thu Jun 14 02:59:58 2012 **** BEGIN LOGGING AT Thu Jun 14 02:59:58 2012 Jun 14 09:04:23 Jay7: ok, last night I've done a first refactoring. Now I'll have to do a second pass of minor fixes (double log of ubiattach, some log output is trimmed because of my qvga so I added some \n) Jun 14 09:05:44 Jay7: the boot from ubifs starts but kernel can't find rootfs (0,255)... I suppose I have to add ubi_detach, otherwise /dev/ubi9 is taken by the first ubiattach... Jun 14 09:06:04 Jay7: I hope it's only this... Jun 14 09:07:26 ant_work: nice work :) Jun 14 09:07:41 * Jay7 went to office Jun 14 11:14:33 Jay7: I was thinking that ubiattach takes the first free ubi (usually ubi0 or 1 if taken) so maybe we could return the corresponding mtd to mount Jun 14 11:14:41 and avoid ubidetach... Jun 14 11:16:04 http://www.mjmwired.net/kernel/Documentation/filesystems/ubifs.txt Jun 14 11:17:38 btw seems u-boot CEWG project has similar scope: http://elinux.org/Boot_U-Boot_from_UBI_volume Jun 14 11:18:33 well, even u-boot in BUI Jun 14 11:18:37 *UBI Jun 14 11:20:05 ah, this was the link: http://www.kernel.org/doc/Documentation/ABI/stable/sysfs-class-ubi Jun 14 11:21:35 we could get back: Jun 14 11:21:35 What: /sys/class/ubi/ubiX/ubiX_Y/dev Jun 14 11:21:41 Major and minor numbers of the character device corresponding Jun 14 11:21:42 to this UBI volume (in : format). Jun 14 11:21:51 or mount by name Jun 14 11:21:53 What: /sys/class/ubi/ubiX/ubiX_Y/name Jun 14 11:21:59 Volume name. Jun 14 11:22:31 wait... Jun 14 11:22:36 it's easier than thius Jun 14 11:22:51 we just want to know which /dev/ubiX has been created Jun 14 11:23:09 we should take the higher Jun 14 11:23:14 (the last created) Jun 14 11:23:20 What: /sys/class/ubiX/ Jun 14 11:23:56 with a loop from 0 to /sys/class/ubi/ubiX/max_vol_count Jun 14 11:24:20 *the highest nr. Jun 14 11:24:40 or just volume count -1 (starts from 0) Jun 14 11:24:48 err. +1 Jun 14 11:25:04 yes, this seems the simplest way Jun 14 11:28:15 whithout hardcoding /dev/ubi9, ubiattach will report the last volume id Jun 14 13:10:56 ...and we should showr *this* device in menu list, not mtdblock... Jun 14 13:29:48 Jay7: well, I'm a bit confused... item = params->bootcfg->list[choice] Jun 14 13:29:57 void start_kernel(struct params_t *params, int choice) Jun 14 13:29:57 { Jun 14 13:30:41 what is confusing you? Jun 14 13:30:46 so params is a pointer and it should be possible to change it back (/dev/mtdblockX -> /dev/ubiY) Jun 14 13:31:14 hm.. should we pass /dev/ubiX or /dev/mtdblockX? Jun 14 13:31:19 but for a struct ? Jun 14 13:31:26 I mean *show* Jun 14 13:31:49 ubi lives on char devices Jun 14 13:32:04 in UI we should show ubiX Jun 14 13:32:08 so /dev/mtdX or the attached /dev/ubiY Jun 14 13:32:16 both char Jun 14 13:32:40 no mtdX if ubiX found Jun 14 13:32:55 there is no relation Jun 14 13:33:05 I understand Jun 14 13:33:11 ubi0 is the first free Jun 14 13:33:42 once mounted you can read sysfs and check the pairing mtd<->ubi Jun 14 13:33:52 I mean if we found ubi container on mtd device we should start some special routine to scan all ubi containers on this mtd and show 'em Jun 14 13:33:54 *once attached Jun 14 13:34:07 ideally yes Jun 14 13:34:28 so we will found no FS on mtd finally Jun 14 13:34:56 we scan the mtdblock Jun 14 13:34:59 not the mtd Jun 14 13:35:14 mtd is chardevice Jun 14 13:35:57 well, let's it show mtdblock for now, this is purely cosmethic tthing Jun 14 13:37:27 anyway, I'm pretty sure we can avoid to hardcode to ubi9 Jun 14 13:37:45 ubi_attach needs just more logic Jun 14 13:38:49 but must return /dev/ubiX_Y...welll ubiX_0 because we only take the first atm Jun 14 14:49:23 btw Jun 14 14:49:24 http://www.mjmwired.net/kernel/Documentation/sysfs-rules.txt Jun 14 14:49:40 ...ever depend on the class-specific links back to the /sys/class Jun 14 14:49:40 144 directory. ... Jun 14 14:49:45 never Jun 14 14:56:32 ah ha Jun 14 14:56:34 The enumeration of available volume ids is a simple. 'ls /sys/devices/ubi/0/volumes' . Jun 14 14:57:27 or I see /sys/devices/virtual/ubi/ubi0/ubi0_0 Jun 14 14:57:44 I'll have to check with 3.2 and 3.4 kernels Jun 14 14:58:09 it seems we can trust sys/devices *only* Jun 14 15:12:03 ho boys, Jun 14 15:12:27 Documentation/ABI/stable/sysfs-class-ubi is still in 3.4 Jun 14 15:18:29 ok, we need to loop until we get a value of /sys/class/ubi/ubiX/mtd_num, knowing mtd_num Jun 14 15:18:44 this would be the right ubiX Jun 14 15:20:16 this is somewhere where the kernel could be more helpful :/ Jun 14 15:20:41 code to handle the quirks of finding volumes must be in a bunch of different projects Jun 14 15:23:03 well, I've found more.. there are max 32 ubi devices Jun 14 15:23:04 http://lxr.free-electrons.com/source/drivers/mtd/ubi/ubi.h#L49 Jun 14 15:23:18 so we can loop from 0 to 31 Jun 14 15:23:39 the number of volumes per device is another thing, bvariable Jun 14 15:33:50 well, we can steal some code from ubinfo...this outputs the number of volumes :) Jun 14 15:36:47 so, firts ubi_attach (/dev/mtd) 3, resulting the ubi device id, then ubi_findvolumes on that ubi Jun 14 15:50:37 eh.. and we are handling the simple case of ubifs on ubi volumes Jun 14 15:51:17 about gluebi: "Original idea was to put JFFS2 on top of UBI volumes. Gluebi is doing Jun 14 15:51:17 just that - turns UBI volume into an MTD device, then you have mount it Jun 14 15:51:17 with JFFS2." Jun 14 15:51:19 imagine Jun 14 15:54:16 argh... I see the bug Jun 14 15:54:18 Volume ID 1 Jun 14 15:54:30 not 0 :) Jun 14 15:55:16 it is nice to see the output on desktop... Jun 14 15:55:19 http://docs.blackfin.uclinux.org/doku.php?id=linux-kernel:ubifs Jun 14 15:55:55 now we have all the interesting links logged here...sorry for the noise Jun 14 15:56:06 ;) Jun 14 16:48:56 bbl Jun 14 18:59:04 hello, anyone around who can tell me some details ? i wonder why the Documentation in kexecboot states the binary should be placed in /usr/bin but isnt it meant to replace /sbin/init ? Jun 14 19:00:44 lautriv: you may use kexecboot as init replacement or as just standalone binary Jun 14 19:01:25 btw, we have no requirement about placing kexecboot itself to /usr/bin iirc Jun 14 19:01:35 * Jay7 -> reboot Jun 14 19:04:09 * lautriv -> waiting for Jay7 's reboot Jun 14 19:06:05 here Jun 14 19:07:53 Jay7, ok,2 more questions before i brick a android-phone ;) Jun 14 19:08:05 oh Jun 14 19:08:27 what phone are you using? :) Jun 14 19:08:59 it's a Samsung Galaxy S+ Jun 14 19:09:19 i found the kexecboot is dynamically linked and there is no option for a static build ? Jun 14 19:10:08 hm.. Jun 14 19:10:24 iirc, there was --static or something like Jun 14 19:10:46 does not really matter but the more files i have in initram, the more failures i can get. Jun 14 19:11:34 we are compiling it with klibc to use in initramfs as init Jun 14 19:12:06 maybe it's in ./configure but not on the homepage, will investigate a bit. Jun 14 19:12:10 BUT Jun 14 19:12:14 root@myon500:/usr/src/kexecboot# ldd /usr/local/bin/kexecboot Jun 14 19:12:15 libc.so.6 => /lib/arm-linux-gnueabihf/libc.so.6 (0x40005000) Jun 14 19:12:15 /lib/ld-linux-armhf.so.3 (0x2a000000) Jun 14 19:12:36 let all to be "default" Jun 14 19:14:12 AC_ARG_ENABLE([static-linking],[AS_HELP_STRING([--enable-static-linking],[compiles kexecboot as a static executable @<:@default=no@:>@])],[],[enable_static_linking=no]) Jun 14 19:14:23 check ./configure --help output Jun 14 19:14:45 Jay7, was on my list (where i said i'll investigate a bit) Jun 14 19:14:57 ok, lat one for now........ Jun 14 19:15:08 but you are right, we should add it to Documentation page Jun 14 19:15:26 btw, I'm not sure that option is working :) Jun 14 19:15:35 check result with ldd Jun 14 19:15:57 i found zaurus.h and zaurus.c and assume that is only for "special needs" because kexecboot relies on FS and devices like the kernel already found/mounted/uses ? Jun 14 19:18:49 /usr/local/bin/kexecboot: ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked, for GNU/Linux 2.6.26, BuildID[sha1]=0x6eaee80042629330da3692b2a7d075d5b546d001, not stripped Jun 14 19:18:58 :) Jun 14 19:20:05 ok, on the way to build a boot-image for said phone, would you suggest to rename kexecboot to init ( replace it ?) Jun 14 19:20:38 zaurus.* is for special processing on zaurus machines Jun 14 19:20:58 assumed that. Jun 14 19:21:41 you have 2 options Jun 14 19:22:10 1) add init=/usr/bin/kexecboot (or /sbin/kexecboot or something like) Jun 14 19:22:18 I mean add this to kernel cmdline Jun 14 19:22:45 2) rename init to init.orig and place kexecboot as /sbin/init (or make symlink to it) Jun 14 19:23:04 but be sure to place kexecboot on rootfs, not on separate fs Jun 14 19:26:12 on rootfs ? i assumed i place it in the initramfs which is pulled right after the kernel before pivot_root ing anywhere. Jun 14 19:26:33 it's ok Jun 14 19:26:43 I mean not on separate /usr e.g. Jun 14 19:28:13 does it act after the switch to real_root and i have it to place on the regular / while the other kernel is kexec'd afterwards ? Jun 14 19:29:05 no, it is started from initramfs if you place it there Jun 14 19:29:19 ok Jun 14 19:29:22 Jay7, ok, as far as i got you are somewhat involved in the project's site. because i found one strange thing : i would take some insight to prebuild images which contain updater.sh and a zImage, nothing wrong with that but for some reason it downloads to /root/Downloads instead /home/$logged_in_user/Downloads and updater.sh appears to contain binary data........maybe the site was hacked to spread some evil stuff ? Jun 14 19:29:25 btw, don't forget to place kexec binary there too Jun 14 19:30:05 that updater.sh is specially encoded for zauruses Jun 14 19:30:16 download images are for zauruses now Jun 14 19:30:25 urgh. Jun 14 19:30:40 can't say about download path Jun 14 19:30:59 the ath is strange, must be somehow forced. Jun 14 19:31:17 my FF is asking me where to download file Jun 14 19:31:17 *path Jun 14 19:31:48 may be your browser/download manager was started by root? Jun 14 19:32:04 mine is predefined but usually $stupid_regular_user is not able to write to /root/ Jun 14 19:36:39 Jay7, i forgot a small thingie ( having regular bootloaders on all other equipment let's kexec apear a bit strange ) has the kernel-option for kexec to be set on the kernel executing kexecboot/kexec or on the invoked (new) one ? Jun 14 19:37:36 you should enable kexec options in first (kexecboot) kernel config Jun 14 19:37:48 chained kernel may have no such options iirc Jun 14 19:37:59 Andrea (ant) may know better Jun 14 19:38:30 my only usage so far was on an ipaq running angstrom ;) Jun 14 19:39:13 unfortunately it died yesterday on the way to get a kernel-3.5 tweak :( Jun 14 19:51:29 is andrea from greece or a girl ? Jun 14 21:05:55 hi ant__ , had a small talk with Jay7 and he meant you may be well informed. i'm on the way to use kexecboot on a Samsung phone, any special suggestions ? Jun 14 21:06:40 hi, no idea about kexec on that platform Jun 14 21:07:16 have you tried to kexec a kernel from shell? Jun 14 21:09:02 ant__, right now, i'm rebuilding the original kernel to enable kexec, will see soon. had it on an ipaq which worked well and this phone is also arm but no idea about details so far. Jun 14 21:09:46 I see Jun 14 21:10:12 for kexecboot, we used this options for handhelds : Jun 14 21:10:14 http://cgit.openembedded.org/meta-handheld/tree/recipes-bsp/kexecboot/kexecboot-handhelds.inc Jun 14 21:10:42 you should probably check the event rate Jun 14 21:11:28 i used OE in the old days but oe-core fails everywhere :( Jun 14 21:11:54 not if you don't add meta-oe ... Jun 14 21:12:24 (I did not retest but I'm pretty sure it's still broken setup) Jun 14 21:12:46 kexecboot does not need meta-oe Jun 14 21:12:58 neither linux-yocto-tiny-kexecboot Jun 14 21:13:17 actually i run a debian in chroot on the phone and build the kernel with a native gcc Jun 14 21:13:54 ok, then look at the config options Jun 14 21:15:00 i had a read on the kexecboot's page and the defaults looks reasonable (maybe i'll change resolution afterwards) my only change was to build it static Jun 14 21:17:07 probably i will run in a trap of unaccessible selections because the lack of proper buttons . can kexecboot react on touchscreens by default ? Jun 14 21:23:51 ant__: my last builds before trip was successfull Jun 14 21:24:04 that's about oe-core buildability Jun 14 21:24:17 ok Jun 14 21:24:38 lautriv: iirc on OM GT Jun 14 21:24:45 ts was working Jun 14 21:25:05 Jay7: I'm doing 2nd refactoring... Jun 14 21:25:12 OM GT ? Jun 14 21:25:18 lautriv: no, we doesn't support TS now.. only few models reporting keypress on TS touch Jun 14 21:25:25 openmoko gta ? Jun 14 21:27:11 i planned to have a timeout to a known_working kernel ( fallback to the unmodified original ) maybe it's possible to trick it to the volume/wakeup buttons. Jun 14 21:27:39 what on earth ............... i have the phone on a terminal via ssh and see the build-process, they build a ton of tuners and infrared-keycodes but lack on hardware for it o.O Jun 14 21:31:32 Jay7: ok, we need to loop until we get a value of /sys/class/ubi/ubiX/mtd_num, knowing mtd_num Jun 14 21:31:42 ^ so, we need a sort of DLookup Jun 14 21:32:57 and btw, how do we get Jun 14 21:32:59 #define UBI_MAX_DEVICES 32 Jun 14 21:33:16 I'd just hardcode it Jun 14 21:33:31 like in kernel, w/out extra headers Jun 14 21:33:41 directory scan should be easy Jun 14 21:33:51 but klibc may lack some useful functions Jun 14 21:34:15 now it's better than 6months ago ;) Jun 14 21:34:26 more stream functions Jun 14 21:34:32 io buffered Jun 14 21:34:33 ... Jun 14 21:35:10 I was hoping we could use our parser code Jun 14 21:35:17 feeding it properly Jun 14 21:35:38 can you imagine that? Jun 14 21:42:26 not sure Jun 14 21:42:46 np, I'm writing smthg Jun 14 21:43:05 our parser can scan string/file of key=value pairs Jun 14 21:43:37 exactly, ubiX, mtdY Jun 14 21:48:57 * Jay7 -> sleep Jun 14 22:02:50 hmmm ....... if i take the sources from samsung and config from /proc/config.gz this thingie should just build, right ? **** ENDING LOGGING AT Fri Jun 15 02:59:59 2012 **** BEGIN LOGGING AT Fri Jun 15 02:59:59 2012 Jun 15 06:23:14 Jay7, are you around ? Jun 15 06:31:59 lautriv: yes Jun 15 06:33:51 i found something strange and wonder how it can be......./proc/config.gz contains "init=/bin/init" which is reasonable but the (working) initramfs contains only /init and no symlink, how could that be ? Jun 15 06:35:03 *facepalm* forget it.......it's the part after pivot_root Jun 15 06:36:32 yes, initrafms uses /init iirc Jun 15 06:37:04 well.. then you should place kexecboot as /init or make symlink Jun 15 06:47:44 did that already, was just confused about the init= but this refers to the part after initramfs..........just had one coffee this morning, not enought for my old gear ;) Jun 15 07:53:41 Jay7: so now it boots, but only from ubi0...and I know why. The code in scan_ devices does attach mtd2->ubi0 and mtd3->ubi1. When it tries to attach one third time in start_kernel it should get ubi3 form ubi_ctrl as first free device but the code in ubi_attach returns 0 (ubi0) because it was already attached on mtd0 so it's the first match. Jun 15 07:53:59 Jay7: solution: do not ubi_attach in start_kernel Jun 15 07:54:21 Jay7: (or add ubidetach, which will add some size) Jun 15 07:54:46 ok Jun 15 07:54:52 info accepted :) Jun 15 07:54:55 ^^ it should get ubi2 not 3 btw ;) Jun 15 07:55:41 right way is to not attach it again Jun 15 07:55:45 ok Jun 15 07:55:49 just don't detach it from scan Jun 15 07:55:52 Jun 15 07:56:01 then just reuse Jun 15 07:56:13 yes, it should not be a problem if kernel thinks it is mounting ubi3 Jun 15 07:56:16 may be this will require adding some field to our struct's Jun 15 07:56:17 again..ubi2 Jun 15 07:57:10 ^^ correction.. code in ubi_attach returns 0 (ubi0) because it was already attached on mtd2 so it's Jun 15 07:57:30 anyway it boots :) Jun 15 07:58:11 there is some hardcoding, I'll remove that and add config options for ubi customization Jun 15 07:58:33 ('-O 512', and ',512') Jun 15 08:00:00 Well, I have to split the ubiattach part and the lookup. This latter lookup will be used in start_kernel to get the ubiX back. Jun 15 08:00:16 I'l ldo that this night...no time during day.. yahwnn Jun 15 08:08:18 thank you for this work :) Jun 15 08:08:42 yw, you'll have to rewrite it mostly :p Jun 15 08:09:52 it should be easier than writing from scratch :) Jun 15 08:10:09 well.. I should go to office Jun 15 20:37:02 i wonder if it is possible to test a kexecboot-kernel in quemu ? **** ENDING LOGGING AT Sat Jun 16 02:59:58 2012 **** BEGIN LOGGING AT Sat Jun 16 02:59:59 2012 Jun 16 11:03:39 lautriv: it's not possible to run kexec in qemu when you are starting kernel by qemu Jun 16 12:01:57 Jay7: hi, even after this refactoring it fails to boot from ubi1 Jun 16 12:04:39 cmdline seems ok Jun 16 12:04:41 http://paste.debian.net/174800/ Jun 16 12:27:52 ant__: hm.. Jun 16 12:28:37 ant__: btw, is your github repo up-to-date? Jun 16 12:44:31 yes, is the last status Jun 16 12:45:29 btw I had again the infamous segfault when pressing twice on fisrt item (sys menu, back) Jun 16 12:45:54 I think it has to do with the item 'exit' present only when standalone Jun 16 12:47:16 ah, see, I note smthg strange in the serial log * and* in normal debug-screen Jun 16 12:47:29 map /dev/ubi1 on /dev/mtd3 Jun 16 12:47:34 is logged very late... Jun 16 12:47:46 it wasn't so before, maybe I added some bug Jun 16 12:58:23 so did you have any success on this? Jun 16 12:58:35 have you booted kernel successfully from ubi? Jun 16 12:59:03 yes but only from ubi0_0 Jun 16 12:59:07 on mtd2 Jun 16 12:59:57 I'm starting to think the kernel at boot doesn't have ubi1, maybe it's always ubi0.. Jun 16 13:08:11 maybe Jun 16 13:08:24 docs onlytalks about volumes. not devices Jun 16 13:08:38 " And finally, you should specify which UBI volume has to be mounted on boot using the root= argument. The volume is specified the same way as described above (ubiX_Y or ubiX:NAME)." Jun 16 13:14:29 Jay7: this is boot from ubi0_0 Jun 16 13:14:31 http://paste.debian.net/174808/ Jun 16 13:14:32 bbl Jun 16 13:50:35 I've found that bug seems Jun 16 13:51:03 when you hit twice enter on system menu Jun 16 13:51:06 or space Jun 16 13:51:21 menu->current->current points to 0x0 Jun 16 13:51:45 so you can't get menu->current->current->id Jun 16 14:37:41 well, found and fixed :) Jun 16 14:50:46 heh.. seems fixed only in textui Jun 16 15:36:45 done finally Jun 16 17:45:58 well.. I have batch of improvements planned for this night Jun 16 17:46:07 * Jay7 is on hackathon ;) Jun 16 17:46:35 so.. seems ant will re-merge his fork back with some conflicts :) Jun 16 17:59:03 Jay7, are you around ? Jun 16 17:59:12 lautriv: yes Jun 16 18:01:30 Jay7, ran in a strange problem and like to have a second opinion.............prepared kernel failed to boot, i took /proc/config.gz from the running machine and sources which were meant for this thingie. but i got 4 section mismatches and assume it might be the compiler, what do you think ? Jun 16 18:02:13 I'm not very experienced for such kind of problems :) Jun 16 18:02:30 may be bluelightning is Jun 16 18:03:16 did you tried to boot kernel compiled with config from /proc/config.gz and sources from vendor? Jun 16 18:03:59 Jay7, yes, that is why i wonder about it.......my pnly manual change is to enable kexec. Jun 16 18:04:14 try to compile w/o changes and check Jun 16 18:17:40 omg.. every check for ext4 is traversing /proc/filesystems and modules.dep Jun 16 18:18:08 happily only for ext4 now in kexecboot's fstype() Jun 16 20:42:41 well.. UI fallback is almost done Jun 16 21:51:07 ant__: hi Jun 16 21:51:27 I've fixed that famous menu bug and made UI fallback Jun 16 21:51:38 preparing to push Jun 16 21:57:59 great! Jun 16 21:58:30 I'll try now to boot always root=ubi0_0 and chabge only ubi.mtd= Jun 16 21:59:39 btw we should fix the code and consider last two digits e.g. /dev/mtd17 Jun 16 21:59:56 (max) Jun 16 22:00:46 pushed Jun 16 22:01:54 what does the function get_nn do? Jun 16 22:02:30 ah, I see Jun 16 22:03:04 get some number Jun 16 22:03:09 iirc Jun 16 22:03:19 I need last two numbers of a string Jun 16 22:03:39 better to adjust actual code, maybe Jun 16 22:04:00 is number of numbers fixed and known? Jun 16 22:04:58 no, see mtd0 - mtd17 Jun 16 22:05:09 well, mtdblock0 - 17 Jun 16 22:05:40 last one is always numeric Jun 16 22:05:53 try to use just get_nni("mtdblockXX", NULL) Jun 16 22:05:56 is the one before to check Jun 16 22:06:17 iirc, strtoul* should just skip non-numbers Jun 16 22:06:32 I was hoping that... Jun 16 22:06:39 I'l ltry Jun 16 22:07:02 well Jun 16 22:07:06 I have questions Jun 16 22:07:17 tell me Jun 16 22:07:27 1) about fstype - shouldn't we rename ubifs to ubi like lvm2? Jun 16 22:07:52 as I understand we are looking for ubi container, not ubifs really Jun 16 22:07:52 he, I'm not sure, probably Jun 16 22:08:11 see http://www.google.it/url?sa=t&rct=j&q=ubidesign.pdf&source=web&cd=1&ved=0CFMQFjAA&url=http%3A%2F%2Flinux-mtd.infradead.org%2Fdoc%2Fubidesign%2Fubidesign.pdf&ei=wwPdT8DeNLOO4gTejamyCg&usg=AFQjCNHfedL0uV5y3U6sNo_QXodz52zyQA Jun 16 22:08:14 sry Jun 16 22:08:35 long link.. Jun 16 22:09:15 see page 13-14 Jun 16 22:09:36 we talk about UBI Jun 16 22:10:09 afaik you can have several fs on ubi volumes, not just ubifs Jun 16 22:15:16 yes Jun 16 22:15:23 that's why I asked this Jun 16 22:16:31 what we find with devicescan on first block is the erase counter Jun 16 22:16:57 UBI# is erase counter, yes Jun 16 22:17:31 UBI! is ubi volume (like LV in LVM) Jun 16 22:17:41 yes Jun 16 22:18:00 iirc in the channel log we've already mentioned the 3 magic Jun 16 22:18:19 so imho it's better to rename 'ubifs' there to 'ubi' Jun 16 22:19:08 initially I did that but then you have to mangle the device.fstype Jun 16 22:19:24 so I let it ubifs for simplicity Jun 16 22:19:27 hm.. Jun 16 22:19:29 well, ok Jun 16 22:19:34 mount -t ubifs Jun 16 22:19:45 I'll wait for your experiments results :) Jun 16 22:20:01 then I'll look again and decide Jun 16 22:20:09 I think the kernel can only attach ubi0 on the passed mtd id Jun 16 22:20:12 at boot Jun 16 22:20:40 see Jun 16 22:20:42 Apr 26 12:16:26 so it seems the volume identifier is at bit 256 and should be UBI! Jun 16 22:20:42 Apr 26 12:16:43 but the erasecounter header is at 0 and is UBI# Jun 16 22:20:42 Apr 26 12:17:09 (at bit 256 on Zaurus) Jun 16 22:21:02 it's ok for us Jun 16 22:21:14 anyway we will call ubiattach later Jun 16 22:21:26 May 04 08:55:18 I think in theory we should run detect_fstype one second time on the volumes Jun 16 22:21:26 May 04 08:55:43 and check for ubifs magic, which is not UBI# or UBI! Jun 16 22:21:38 well, later :) Jun 16 22:22:07 * Jay7 is looking on todo list Jun 16 22:22:10 maybe we should ubiattach all the mtd and then run devicescan Jun 16 22:22:32 btw, why not Jun 16 22:22:47 is it possible to attach multiple mtd's by one call? Jun 16 22:22:50 at least , try to, it fails nicely Jun 16 22:22:59 hm, no Jun 16 22:23:05 np Jun 16 22:23:13 you can put it in a for cycle Jun 16 22:23:18 sure Jun 16 22:23:20 0 -17 Jun 16 22:29:48 ok, if it boots with root=ubi0_0 from both mtd all the code in start_kernel is unnecessary... Jun 16 22:29:58 let see Jun 16 22:29:59 nice Jun 16 22:30:26 I think now is good time to make LABEL default delimiter, not KERNEL Jun 16 22:30:33 at other side.. Jun 16 22:30:37 ah, maybe yes Jun 16 22:30:46 like other bootloaders Jun 16 22:31:00 well.. will do now then Jun 16 22:31:15 and next is GUI speedup Jun 16 22:31:29 GUI on akita is terrible slow Jun 16 22:31:59 here on poodle is slow when run from shell, otherwise in kernel is nice Jun 16 22:32:11 (with debug enabled) Jun 16 22:32:27 lot of events output Jun 16 22:38:02 Jay7: it booted from mtd2 ! Jun 16 22:38:08 cool! Jun 16 22:38:09 and mtd3 ! Jun 16 22:38:19 really cool :) Jun 16 22:38:24 it seems it's always ubi0_ on cold boot Jun 16 22:38:48 let's assume this as true :) Jun 16 22:39:07 see Jun 16 22:40:07 http://paste.debian.net/174900/ Jun 16 22:41:11 with dmesg and cmdline: Jun 16 22:41:12 http://paste.debian.net/174901/ Jun 16 22:41:35 it is also logic: you give it the mtd Jun 16 22:41:48 and the volume, if you want Jun 16 22:41:54 but *not* the device Jun 16 22:43:39 ok, if you want I can remove the call to find_attached_ubi_device in start_kernel Jun 16 22:44:01 but it's easy enough for you, so you can fic char and cons char Jun 16 22:44:08 *fix ;) Jun 16 22:46:41 Jay7: you'll have soon to report the UBI: number of good PEBs: 2240 Jun 16 22:46:52 for partitions on collie and tosa Jun 16 22:47:15 si I can extend the conf in meta-handheld Jun 16 22:48:23 remove that call Jun 16 22:48:36 seems it's ok to just use ubi0_0 always Jun 16 22:48:46 yes Jun 16 22:49:20 well, about ubi0_ Jun 16 22:49:55 you could have the other forms, ubi:boot Jun 16 22:50:06 I'd ignore them atm Jun 16 22:50:22 btw Jun 16 22:50:48 I don't see that it's necessary to add ubi volume to kernel cmdline Jun 16 22:51:14 yes, as discussed above Jun 16 22:51:15 ah Jun 16 22:51:23 root Jun 16 22:51:25 start_kernel Jun 16 22:51:37 ok then Jun 16 22:51:54 seems kernel always add this as first ubi volume Jun 16 22:53:15 ah, about const char str_ubimtd_off[] = ",512"; Jun 16 22:53:28 ^^ this is an hack for zaurus Jun 16 22:53:45 this should be really a config option Jun 16 22:57:52 sure, I'll add necessary configure options Jun 16 23:24:32 I was thinking about passing mtd id as integer Jun 16 23:24:48 but finally we do string match later so I don't know... Jun 16 23:29:25 easier is to strip '/dev/mtd' from mtd device name Jun 16 23:29:31 and pass rest of string Jun 16 23:30:07 mtdblock Jun 16 23:30:14 ah Jun 16 23:30:24 well.. numbers are same Jun 16 23:30:33 happily Jun 16 23:33:22 well, making LABEL delimiter Jun 16 23:35:05 do you prefer a new patch for kexecboot.c or can I recommit it? Jun 16 23:35:29 do you asking about your ubi work? Jun 16 23:35:34 yes Jun 16 23:35:45 try to cleanup things in your tree Jun 16 23:35:53 I'll import it later Jun 16 23:36:01 ok, I think this is the last patch :) Jun 16 23:36:02 may be this morning even Jun 16 23:36:13 i.e. don't care about other changes Jun 16 23:36:20 ok Jun 16 23:36:29 just remove unneeded things and add needed :) Jun 16 23:42:19 so.. if there is no config file, kernel file name will be used as label Jun 16 23:42:37 if there is config file then there should be LABEL Jun 16 23:46:21 and LABEL should be first line of section Jun 16 23:52:32 ok, I've pushed the last version Jun 16 23:55:13 hm.. seems you are rebased :) Jun 17 00:04:11 hm.. I did something wrong... no boot now.. Jun 17 00:04:26 strange Jun 17 00:05:41 sh@it..I've added get_nni without testing it... Jun 17 00:06:06 pushed LABEL change Jun 17 00:06:15 ok, I'll fix that now, rebasing again Jun 17 00:19:04 yes, is returning 0 Jun 17 00:19:19 always? Jun 17 00:19:49 I see ubi.mtd=0,512 Jun 17 00:19:53 i.e. it doesn't skip letters before? Jun 17 00:20:39 seems the code is returning 0, maybe is the sprintf.. Jun 17 00:20:53 no, it's more like get_nn* Jun 17 00:21:29 well, then you may just skip lenght of "/dev/mtdblock" chars Jun 17 00:22:25 i.e. something like p = device + sizeof("/dev/mtdblock"); and use p as string Jun 17 00:22:46 yes, if we found ubifs we were scanning mtdblockX Jun 17 00:23:03 e.g. sprintf(buffer, "ubi.mtd=%s,512", p); Jun 17 00:23:36 you may correct it by +/-1 char because of possible \0 Jun 17 00:23:42 easier to check Jun 17 00:27:14 btw, now we can enable textui by default too Jun 17 00:27:25 need to check size difference Jun 17 00:27:53 anyway, we can disable it in our configs Jun 17 00:49:06 ok, another refresh, back to one digit for mtd Jun 17 00:49:11 pls fix that :) Jun 17 00:49:15 ok Jun 17 00:49:21 now it boots Jun 17 00:49:34 I'm doing another test from fresh pull Jun 17 00:49:36 fine Jun 17 01:00:19 well, now the last change (LABEL) seemss have killed all boot Jun 17 01:00:35 from any device Jun 17 01:01:36 heh.. I must update all rootfs's Jun 17 01:01:56 Can't parse keyword 'KERNEL' Jun 17 01:03:51 ok, tomorrow Jun 17 01:03:53 gn Jun 17 01:10:47 ant__: nite Jun 17 01:10:58 hm Jun 17 01:11:30 ant__: check you have LABEL's in your boot.cfg Jun 17 01:11:44 I've tested only with LABEL's Jun 17 01:11:54 should check w/o Jun 17 01:12:32 in OE we have still KERNEL before LABEL Jun 17 01:12:59 I've temporarly reverted the patch and it's ok Jun 17 01:13:04 is just that :) Jun 17 01:13:09 ah Jun 17 01:13:18 I have to catch up ! Jun 17 01:13:18 yes, LABEL should go first in section Jun 17 01:13:29 check it please Jun 17 01:13:49 is kexecboot_cfg recipe Jun 17 01:13:56 I'm now adding long awaited configure options for memory transfers Jun 17 01:14:34 i.e. you can choose now between 32/16/8 bits RAM-to-FB transfers Jun 17 01:14:48 I hope this may help me on collie Jun 17 01:16:03 ok, try to sleep now, don't do as I do ;) Jun 17 01:16:18 see you tomorrow **** ENDING LOGGING AT Sun Jun 17 02:59:59 2012 **** BEGIN LOGGING AT Sun Jun 17 02:59:59 2012 Jun 17 05:26:44 I've found why GUI is so slow now Jun 17 05:27:16 GUI redrawing 6 times after every keypress Jun 17 05:27:26 something going wrong :) Jun 17 06:49:11 well.. going home Jun 17 11:42:08 Jay7: I think it is better to be precise with ubi/ubifs definition Jun 17 11:42:17 I'll patch this Jun 17 21:10:18 ok, first attemt to kexecboot an android-phone failed due to ugly handling of bootup............did anyone ever try to kexec from chroot ( given that root does it ) ? Jun 17 21:29:48 lautriv: what's exactly the issue ? Jun 17 23:05:52 Jay7: I've added one more commit, pls consider it and maybe ask Tim about naming ;) Jun 17 23:34:10 cu, gn **** ENDING LOGGING AT Mon Jun 18 02:59:58 2012 **** BEGIN LOGGING AT Mon Jun 18 02:59:58 2012 Jun 18 10:21:41 bluelightning: any news about ipaqs? I'm preparing the patchset for linux-kexecboot and linux 3.1 and 3.2 removal from meta-handheld (those are Zaurus specific afaik). Same for meta-initramfs, for Koen. Jun 18 10:22:28 I'l add 3.4 Jun 18 10:22:53 later, once I'll clear the config fragments Jun 18 10:44:38 I added 3.4 for hx4700 locally but I still haven't diagnosed the problem I had Jun 18 10:44:42 ant_work: ^ Jun 18 10:45:32 (although I did it using a vanilla 3.4 rather than linux-yocto, since linux-yocto 3.4 wasn't finished when I did it) Jun 18 10:45:52 I see, I'd like to remove 2.6.29 and 3.01 as well... Jun 18 10:46:30 well, _git should also go imho Jun 18 10:46:39 2.6.29 is what we had for h1940 ? Jun 18 10:46:57 target is removing the whole recipes around linux.inc Jun 18 10:47:12 and keeping only -yocto Jun 18 10:47:24 if so then you could ping anarsoul to find out the current state of the kernel for that machine Jun 18 10:48:12 I think it is wiser to concentrate on 3.4 even for older models which may be surprisingly boot Jun 18 10:48:39 -be Jun 18 10:50:26 well, I would say the opposite, if the new kernel doesn't work and the old one does I'm not going to be keen to force the new one... Jun 18 10:50:49 I'm not even sure you build 2.6.29 nowaday Jun 18 10:52:27 I built 2.6.21-hh just a month or two ago so it's possible... Jun 18 10:52:45 nice. klibc would fail though :/ Jun 18 10:53:03 wait, now building against linux-libc-headers so .. just try Jun 18 10:53:25 udev/userspace will be an issue Jun 18 11:03:54 yes... I did have to provide an older version of udev and I still had issues, but it did boot Jun 18 11:18:53 trust me, I tried to keep 2.6.26 and 2.6.3x together but there are simply too many changes to take care of... (hd*->sd*, devtmpfs, cgroups, ...) Jun 18 12:27:02 the alternative is not support a whole bunch of machines because there is nobody working on them with recent kernels Jun 18 12:27:04 e.g. h2200 Jun 18 12:47:02 I don't know, I feel guilty having them sitting there alone... Jun 18 13:55:54 bbl Jun 18 16:20:42 man, what a freakshow ........ got it managed to make the samsung boot a kexec-kernel but lost WiFi ( probably erternal propietary stuff but should be present in ROM ) however i need a SomeTerminal.apk and the new google store won't let me download ( only sending to device) which is not possible via Wlan :( ........ can anyone tell me a d/l location for a terminal beside the non-working version from Jack Palevich ? Jun 18 17:28:50 lautriv: hi Jun 18 17:28:58 lautriv: how is your samsung? Jun 18 20:08:19 Jay7, still around ? Jun 18 20:33:01 hi ant__ Jun 18 20:33:17 hello Jun 18 20:34:10 ant__ , sorry for not answering the last question......my ISP does a cut each 24h, that happens around 7:30 AM but i forget it sometimes. Jun 18 20:34:19 err PM Jun 18 20:34:30 np, I was reading logs Jun 18 20:35:34 are you kexec'ing an official kernel? Jun 18 20:35:55 i.e. does wifi work when booted normally? Jun 18 20:36:16 i would use gigabread+ but can't get the patches so i used the stock-rom from/for samsung. Jun 18 20:36:34 there have been some USB issues after kexec in the past Jun 18 20:36:58 but here wifi is integrated I think Jun 18 20:37:12 no, no wifi but that may have 2 problems a.) another comiler leads to incompatible modules b.) obv. those are not within kernel's sources. Jun 18 20:38:14 yes, old story with proprietart stuff Jun 18 20:38:26 right now, i have a booting device with kexec enabled and searching a download-location for a terminal-emulator. Jun 18 20:39:48 Jay7: I'm not so convinced about my last ubifs->ubi renaming. That would be the patch. I've asked klibc guys about the matter fwiw Jun 18 20:40:10 lautriv: yes Jun 18 20:40:13 ant__: hi Jun 18 20:40:27 ant__: i.e. it should be named ubi? Jun 18 20:40:34 it looks like *practically* there is only ubifs on ubi volumes. the rest are experiments or old stuff (logfs cramfs) Jun 18 20:41:12 you see, in /proc/filesystem there is ubifs..that seems the way to keep Jun 18 20:47:42 will saw your nerves tomorrow......bedtime for an old man ;) Jun 18 20:48:24 lautriv: sorry I can't help you, not yed played with Gingerbread & co. Jun 18 20:48:38 still on the linux-side Jun 18 20:50:23 btw, yes Jun 18 20:50:31 bedtime for old man :) Jun 18 20:50:36 * Jay7 -> sleep Jun 18 20:50:37 he he Jun 18 20:51:00 Jay7: hold on with the last patch. Maybe we can avoid that Jun 18 20:51:21 ant__: I'll wait some days :) Jun 18 20:51:42 I'm going to erlang conference this friday in Moscow Jun 18 20:52:05 oh, Moskow again... Jun 18 20:52:09 yeah Jun 18 21:52:37 Jay7: I notice now that I disabled device recreation during all my tests Jun 18 21:52:48 I'll test the other way **** ENDING LOGGING AT Tue Jun 19 02:59:58 2012 **** BEGIN LOGGING AT Tue Jun 19 02:59:59 2012 Jun 19 08:52:31 Jay7: I don't remember why exactly I disabled device recreation, maybe I was testing ubiattach.. Jun 19 08:52:49 np Jun 19 08:55:16 this shouldn't change, ubi devices are created afterwards Jun 19 08:55:43 so I'm confident it will work in bothcases Jun 19 08:56:08 now, to release ubifs images for OE, I need some info Jun 19 08:57:07 which can only (apparently) appear running mtdinfo -a or ubinfo -a after having ubi-formatted Zaurus mtd Jun 19 08:57:24 I have C860, C3200 and C5600 Jun 19 08:57:32 I need all the remaining... Jun 19 08:58:42 (it is the -c parameter (eraseblock count) Jun 19 08:59:51 ..and I fear thi sis not constant i there are bad eraseblock on device... Jun 19 08:59:56 *if Jun 19 09:09:21 Jay7: btw... http://lists.infradead.org/pipermail/linux-mtd/2012-June/042313.html Jun 19 09:09:30 3.5-rc2 can wait :) Jun 19 09:33:31 Jay7: JaMa thinks it's better to keep the renaming as done with last patch Jun 19 09:37:19 ant_work, not worth to fiddle on 3.5-rc2 it's messed up......just wait a bit longer ;) Jun 19 09:46:42 ant_work: I need working image for my Z's then Jun 19 09:49:18 Jay7: apart ubi, once koen and bluelightning will merge the patches I've sent to oe-dev all will build fine Jun 19 09:49:28 with kexecboot from master Jun 19 09:49:40 heh.. tosa users in angstrom-devel Jun 19 09:49:45 :) Jun 19 09:50:03 I need someone toubiformat tosa and give me the LEB count Jun 19 09:50:07 :) Jun 19 09:50:24 can anyone tell me where kernel's modules are stored in android ? "find" will put the device off . ( probably some /proc on read ) Jun 19 09:51:04 aren't they in /lib/modules?? Jun 19 09:51:32 Jay7, they haven't even FHS Jun 19 09:51:54 oh Jun 19 09:52:10 * Jay7 want back to maemo times Jun 19 09:54:14 ant_work: I'll send you output from my devices Jun 19 09:54:18 Jay7, forget that part, just ugly to browse ( a rooted smartphone still haven't proper pwd ) Jun 19 09:54:24 not sure about collie though Jun 19 09:54:39 yea Jun 19 09:54:48 akita an tosa should be ok Jun 19 10:02:26 meh having a terminal on 5" touchscreen let me wish to smash this crap on a wall :( Jun 19 10:02:51 heh.. try 4x6 chars on qvga 3.3'' Jun 19 10:03:31 I've preferred to have truncated output...and 8x16 fonts Jun 19 10:03:41 Ja7: btw, ^^ Jun 19 10:05:13 ant_work: np Jun 19 10:05:34 was that n the idea of truncating output? Jun 19 10:05:57 yes Jun 19 10:06:00 :) Jun 19 10:06:21 but we may try to search good console font Jun 19 10:06:29 e.g. some terminus Jun 19 10:06:50 to have 40 chars I need 6x fonts Jun 19 10:07:17 for qvga ofc Jun 19 10:08:21 30 chars is ok too Jun 19 10:09:00 i.e. for 8x fonts Jun 19 10:10:04 strange...... i had just 2 modules (bthid and some sec*) right below /lib/modules, trying to modprobe bthid failed because it was not below /lib/modules/$(uname -r) so i did ln -s ./ 2.6.35.7 and now i have 5 other modules but not longer bthid o.O Jun 19 13:28:38 bluelightning: I've seen one strange warning after the layer refactoring. Sorry, can't be more precise now but bitbake iscomplaining... Jun 19 18:11:11 greetings Jun 19 18:25:09 i'm planning to share some larger images with the public but would like to avoid the possible massive amount of traffic on my own server, any recommendations to a site which is good to link and can hold a few files around 50~600M ? Jun 19 18:58:59 https://developer.tizen.org/resources/development-unit-program Jun 19 21:22:07 hi ant__ Jun 19 21:22:49 hi there Jun 19 21:25:02 hi bluelightning Jun 19 21:25:12 hi lautriv Jun 19 21:29:25 * lautriv is just waving around to get not bored about a 5GB git clone process ;D Jun 19 21:58:01 meh, why must git be such a endless crap ? Jun 19 23:28:30 bluelightning: how is it possible ubi-utils-klibc got a more recent mtd/ubi-user.h header than in sysroot? Fatally the same version (with renamed IOCTL) is on my Gentoo box :/ Jun 19 23:30:51 so this morning, after buildfrom-scratch of core-image-base, when I bitbake'd linux-yocto-tiny-klibc ubi-utils-klibc did fail for the renamed ioctl ( renamed in 3.0) Jun 19 23:31:16 now I check in sysroot and see the old version... Jun 19 23:34:01 strange...we have 3.4 headers Jun 19 23:34:33 http://lxr.free-electrons.com/source/include/mtd/ubi-user.h?a=arm Jun 19 23:34:41 189 #define UBI_IOCSETVOLPROP _IOW(UBI_VOL_IOC_MAGIC, 6, \ Jun 19 23:34:41 190 struct ubi_set_vol_prop_req) Jun 19 23:35:08 and in sysroot: Jun 19 23:35:09 #define UBI_IOCSETPROP _IOW(UBI_VOL_IOC_MAGIC, 6, struct ubi_set_prop_req) Jun 19 23:36:13 maybe they get sanitized Jun 19 23:38:19 http://comments.gmane.org/gmane.linux.drivers.mtd/34122 **** ENDING LOGGING AT Wed Jun 20 02:59:59 2012 **** BEGIN LOGGING AT Wed Jun 20 02:59:59 2012 Jun 20 08:45:35 Jay7: I've forgot to add the --enable-fb-transfer-width to the metadata and to the docs... Jun 20 08:46:09 Will do with next git tag. 0.6 ? Jun 20 08:46:27 I'd prefer to have ubifs in 0.6 Jun 20 08:46:44 but we may postpone this to 0.7 Jun 20 08:54:40 as you like. Do you have other pending commits for 0.6? Jun 20 08:54:52 ant_work: may be good idea to rename --enable-fb-transfer-width to just --enable-fb-transfer Jun 20 08:55:02 at other side this may confuse by name Jun 20 08:55:21 ant_work: I should check todo on my laptop Jun 20 08:55:29 but laptop is at home :) Jun 20 08:55:35 ah Jun 20 08:55:37 yes Jun 20 08:55:45 fix refreshes Jun 20 08:55:58 we doing 6 redraws per keypress Jun 20 08:56:05 crazy bug :) Jun 20 08:56:11 he Jun 20 08:56:37 you'll have to add one or two config options for UBI as well Jun 20 08:56:56 but I'm unsure when I'll have time to do this again.. may be next hackaton (about next month) Jun 20 08:57:20 yes, ubiattach-path and something for block size (as I understand) Jun 20 08:57:50 things going to make menuconfig :) Jun 20 09:00:24 let's hope CEWG is interested Jun 20 09:01:31 btw, I think I'll remove the fs-detection hacks from klibc recipe: we will patch our fstype.c in kexecboot instead Jun 20 09:02:49 I've updated the site about config options...dunno you but I always end up editing HTML..drupal is not my best friend... Jun 20 09:07:36 Jay7: one more thing: when built with --enable-debug kexecboot did not kexec any kernel when launched from shell. It did echo instead, like with --enable-debug Jun 20 09:08:05 ^^ like --enable-host-debug Jun 20 09:08:11 ant_work: hm.. Jun 20 09:08:32 I did see the ubi cmdline... Jun 20 09:08:33 it should kexec with --enable-debug Jun 20 09:08:44 but it shouldn't with --enable-host-debug Jun 20 09:08:58 I don't think I've enabled both, pretty sure. I'll retry Jun 20 09:09:13 just when NOT init Jun 20 09:09:35 if it is - it is bug :) Jun 20 09:09:54 btw, I wish disable event logging Jun 20 09:10:04 from evdev's Jun 20 09:10:11 ah, yes, is very noisy and useful only fo special debug Jun 20 09:10:27 we may just disable it or make another option Jun 20 09:10:36 better an option... Jun 20 09:10:57 --enable-debug=2 :) Jun 20 09:11:23 imho better than having special option Jun 20 09:11:39 yes, like log_level Jun 20 09:29:46 ant_work: hello! what kernel can be used in zaurus spitz? Jun 20 09:32:53 ant_work: are we using vanilla kernels? Jun 20 09:33:39 spitz, yes, vanilla Jun 20 09:34:13 we use yocto-kernels nowadays, though, with specific patchset Jun 20 09:36:56 ant_work: I try vanilla kernel from git - versions after 3.2-rc7 just freezing after unpacking. Version 3.2-rc7 booted, but without serial console... Booting via uboot from pdaXrom Jun 20 09:37:04 (sorry for my english) Jun 20 09:37:57 ant_work: can you show my config? Jun 20 09:39:01 ant_work: sorry. May I see you config? :) Jun 20 09:41:15 ant_work: and what kind of patches you using? Jun 20 09:44:28 http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto/spitz/spitz.cfg Jun 20 09:45:15 no patches for spitz, just bootlogo Jun 20 09:46:16 btw this should be the same, shrunk by make_savedefconfig Jun 20 09:46:21 http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-3.2/spitz/defconfig Jun 20 09:46:30 ^^ this is the 'old' recipe for kernel Jun 20 10:09:02 ant_work: thank! I will try to build 3.2.12 with you config Jun 20 10:43:42 ant_work: many thanks! 3.2.12 just booted! Jun 20 10:51:37 cool :) Jun 20 20:50:15 [00:48] oe-core mtd-utils is at ca39eb1d98e736109c64ff9c1aa2a6ecca222d8f Jun 20 20:50:23 [00:48] and the one from meta-initramfs is 995cfe51b0a3cf32f381c140bf72b21bf91cef1b Jun 20 20:50:33 for ant :) Jun 20 20:50:45 look at #yocto logs Jun 20 21:19:57 ant__: check #yocto logs if any Jun 20 21:20:07 about mtd-utils problem Jun 20 21:20:13 linux-headers, etc Jun 20 21:21:00 I've found it Jun 20 21:21:12 it id mtd-utils providing older headers Jun 20 21:21:24 ah Jun 20 21:21:28 and ubi-utils-klibc is a rip-off mtd-utils recipe Jun 20 21:21:32 so... Jun 20 21:21:39 well, ok Jun 20 21:21:39 total chaos Jun 20 21:22:09 bizarre at least..same ppeople doing kernel + userspace :) Jun 20 21:22:42 (Artem Bityutskiy) Jun 20 21:23:37 bluelightning: it seems evident mtd-utils should not rewrite linux-libc-headers in sysroot Jun 20 21:23:48 nice you found this Jun 20 21:23:55 * Jay7 going to bed Jun 20 21:24:05 but should provide its own copy somewhere Jun 20 21:24:12 gn Jay7 Jun 20 21:29:26 ouch.. install -d ${D}${includedir}/mtd/ Jun 20 21:29:47 and all headers Jun 20 21:30:38 our ubi-utils-klibc clone recipe has own do_install() Jun 20 21:31:25 so is not guilty :) Jun 20 21:41:32 ant__: ah, good find... yes that is evil :( Jun 20 21:42:15 I'm looking now if there are new versions Jun 20 21:44:27 seems git.infradead is dead :p Jun 20 22:15:04 bluelightning: so about BBPATH, .= or += ? Seems the same in this case Jun 20 22:15:26 ant__: yeah it doesn't really matter Jun 20 22:15:45 I'd like to bump JaMa's patch for Koen has forgotten Jun 20 22:23:58 well, I can't get rid of those warnings... Jun 20 22:24:40 someone else was complaining they were getting warnings too Jun 20 22:24:48 I'll have to dig into it... Jun 20 22:24:54 ah, the other issue, yes Jun 20 22:25:49 now I'm talking about the WARNING: BBPATH references the current directory Jun 20 22:26:23 jeez, meta-handheld is patched, meta-initramfs done by me...why then Jun 20 23:19:12 gn **** ENDING LOGGING AT Thu Jun 21 02:59:58 2012 **** BEGIN LOGGING AT Thu Jun 21 02:59:58 2012 Jun 21 09:44:23 hi ant_work Jun 21 09:44:50 ant_work: just to clarify, what is MACHINE set to in your case when you see those "file could not be found" warnings? Jun 21 10:15:00 I'm using poodle Jun 21 10:15:17 ah, he's gone Jun 21 10:58:11 Jay7: have you built some images finally? Jun 21 11:20:17 bluelightning: wb Jun 21 11:20:25 I'm building for poodle atm Jun 21 11:20:48 stock local.conf, just added rm_work Jun 21 11:26:11 ant_work: ok, thanks Jun 21 11:27:18 yw Jun 21 11:28:03 btw it doesn't seem an intra-layer issue Jun 21 11:33:31 bluelightning: and about ubi-utils staging its (now outdated) headers, it may well be that in the 2.6.2x era kernel was still missing ubi. Moreover, iirc some jffs2 headers not present in sysroot are added Jun 21 11:33:51 er.. mtd-utils, the original recipe Jun 21 11:33:55 right Jun 21 11:34:11 my suggestion on the latter issue would be to file a bug in the yocto bugzilla Jun 21 11:34:22 in gitweb you see that now and then they've synched with kernel's headers in the past Jun 21 11:34:28 for the former I'm just doing some pre-investigation and will file a bug after that Jun 21 11:34:53 ok, thx Jun 21 11:35:24 evidently mtd-utils is the only user of this headers Jun 21 11:37:13 imagine that I added those mtd headers to the ones installed by klibc when building against libc-headers Jun 21 11:37:39 so ubi-utils-klibc finds the copy of the linux-libc-headers in sysroot Jun 21 11:37:59 though, looking in /lib/klibc/include Jun 21 11:38:35 now, maybe it is better to extend the klibc recipe to a full mtd-utils-klibc Jun 21 11:38:54 it's just atm we only need ubiattach-klibc... Jun 21 11:39:32 so it seems a bit a waste of time....I'll see if it's easy I'll do it though ;) Jun 21 12:25:35 i wonder a bit about the behaviour of that smartphone i try to kexecboot. using the same sources, same compiler same .config and only adding kexec and panic=30 let's it freeze on early boot ? Jun 21 12:41:42 we have seen that on other devices ( mips) Jun 21 12:42:15 it sorted out board conf did not initialize properly the memory Jun 21 12:42:36 so when the bootloader did, all ok, after kexec...ko Jun 21 12:47:32 ant_work, from what i got so far, it looks like there are certain mechanisms to prevent the device to boot anything else android, like vendor-specific partitions with propietary stuff and so on. but there are just 2 modules in the original ROM one is bthid and another some sec*.ko probably to encrypt system-partition. however those are not even touched by me and recompiling with the same .config produces a bunch of more modules, most of t Jun 21 12:47:33 hem are for stress-testing. however i created a signed image which is applied with sucess and still no luck :( Jun 21 12:55:12 hm.. Jun 21 12:55:21 which kind of bootloader? Jun 21 13:00:07 it's a special case, phone has 28 (!) partitions where the first 0-7 are vital to the thingie (containing bootloader IMEI CSC and such) part 08 is an image with zImage/initrd and header with offset of 4k, i do only changes on this part (extract/rebuild) to avoid bricking the device. the early partitions contain also a low-level routine to flash via USB by powering-up with a special key-combo. this is my backdoor for failures, so i have to Jun 21 13:00:07 reflash to factory-defaults after no success on boot. Jun 21 13:01:56 hmm 28 partition.. eek.. I've read mtd can only be 0-17 ... Jun 21 13:03:32 even stranger, they use ext4 on flash o.O Jun 21 13:07:06 ... right now i found something. having a little helper which can recover/update/clean some stuff by starting with another key-combo, i saw my kernel can't mount /system while the factory's ROM can. now even this partition IS encrypded so i assume there is a small part wihin the kernel which is NOT in .config (key-hint) Jun 21 13:07:53 it's is sinisterly similar to Zaurus :/ Jun 21 13:07:55 http://www.h5.dion.ne.jp/~rimemoon/zaurus/memo_006.htm Jun 21 13:08:04 and the funny key-combo's Jun 21 13:08:52 on Zaurus the rootfs was signed but only for the stock provided kernel... Jun 21 13:09:05 the bootloader happily allow to boot unsigned kernels Jun 21 13:09:33 which then does not care about signatures/headers Jun 21 13:09:50 finally, it depends on the bootloader I fear Jun 21 13:10:29 I would not be surprised if you could upload your unsigned kernel with one of the emergency routines Jun 21 13:13:05 there is a part in your link which is very close to my info, mentioned partition8 can handle addidtional specs/userland etc. but they say it's not used right now. but i can't share the opinion that bootloader may be the problem because the /system will be mounted from within kernel/initramfs (later).........i will see if i can loopmount the image. Jun 21 13:16:04 ok, not an issue, mount -t ext4 -o loop was successful on the host :( Jun 21 13:21:37 ant_work, is it possible for the kernel to lie about /proc/config.gz ? i assume if it's enabled there is just a copy of the compressed .config which may be injected. Jun 21 13:23:26 I think it's just a copy, don't know where it is kept, though Jun 21 13:24:21 i will give it a rest and work a bit in the garden.......sometimes helpful ;) Jun 21 13:39:37 Jay7: he he Jun 21 13:39:39 is jffs2 using CONFIG_MTD_NAND_VERIFY_WRITE? can I simply unset it? Jun 21 13:39:39 ant_work, switch it off - it is a debugging feature Jun 21 13:39:46 ^^ Artemy Jun 21 13:46:36 hm, nice Jun 21 13:46:43 well Jun 21 13:46:59 * Jay7 is gone to Moscow Jun 21 14:15:37 we just detect a container Jun 21 14:15:37 yep. to identify ubifs you have to find LEB0 in the UBI volume Jun 21 14:52:13 ant_work: I think I may have figured out why you're getting this warning Jun 21 14:53:07 the SRC_URI chcksums you mean? Jun 21 14:56:34 yes Jun 21 14:56:52 ant_work: I suspect that the error is triggered for e.g. kernel recipes that are not providing a defconfig for the machine you have selected Jun 21 14:57:22 ant_work: e.g. 2.6.29 only supports h3600 Jun 21 14:57:35 ant_work: it's complicated by the fact that it can't currently show you the recipe it's talking about :/ Jun 21 14:58:05 the fact that I have two lines signales it is processed twice I think Jun 21 15:00:49 and after removing linux-kexecboot the third warning disappeared Jun 21 15:00:52 ( WARNING: Unable to get checksum for linux-kexecboot SRC_URI entry defconfig: file could not be found) Jun 21 15:01:40 ant_work: if it were linux 3.0.1 and 2.6.29 that would provide two right? Jun 21 15:01:52 neither of those define COMPATIBLE_MACHINE Jun 21 15:02:11 perhaps they should... Jun 21 15:14:08 ok, I'll try later, thx Jun 21 15:14:10 bbl Jun 21 21:35:03 bluelightning: indeed, compatible machine does the trick Jun 21 21:35:18 right Jun 21 21:35:30 so I'm not sure if I do need to do anything Jun 21 21:35:45 since it's kind of complaining about something legitimate at least in this case Jun 21 21:35:56 I'm still not sure what's happening in Gary's situation though Jun 21 21:35:59 note that this won't work COMPATIBLE_MACHINE .= "|h3600|" Jun 21 21:36:17 we need to restrict in this case COMPATIBLE_MACHINE = "h3600" Jun 21 21:36:40 same for hx4700 Jun 21 21:36:44 I think that's fair enough, these recipes are self-contained within our layer Jun 21 21:42:30 I have the two patches ready if you mind Jun 21 21:46:15 sent Jun 21 21:47:43 as we already noticed, default preference -1 is not enough to mask the recipe Jun 21 21:53:04 bluelightning: about WARNING: BBPATH references the current directory, either through an empty entry, or a '.'. Jun 21 21:53:28 there's some discussion about that on the mailing list already, I guess you noticed Jun 21 21:53:28 andrea@mizar /oe/oe-core/build $ bitbake -e core-image-base | grep -e "^BBPATH=" Jun 21 21:53:28 BBPATH=":/oe/oe-core/meta:/oe/meta-openembedded/meta-initramfs:/oe/meta-handheld" Jun 21 21:54:23 is the first colon ok? Jun 21 21:55:14 it looks it was ok when we did prepend Jun 21 22:09:59 yes, as I thought, is in oe-core Jun 21 22:12:18 it's the very old problem of concatenating empty strings and getting wrong separators Jun 21 22:12:21 :p **** ENDING LOGGING AT Fri Jun 22 02:59:59 2012 **** BEGIN LOGGING AT Fri Jun 22 02:59:59 2012 **** ENDING LOGGING AT Sat Jun 23 02:59:58 2012 **** BEGIN LOGGING AT Sat Jun 23 02:59:58 2012 Jun 23 18:54:23 bluelightning, are you around ? Jun 23 18:54:33 lautriv: yep Jun 23 18:55:47 bluelightning, maybe you got that i attempt to bring linux on an android-phone, and i found some strange details where i need your opinion ........ Jun 23 18:56:09 sure, I can try but no promises... Jun 23 18:56:41 as far as i can see, this thingie does never invoke pivot_root and mounts certain parts like /system within initramfs Jun 23 18:57:20 now, the commandline (kernel built in) says init=/sbin/init which is usual but there is none ....... Jun 23 18:59:06 the given init is much larger (144k instead 26 k) and lives on / within initrd, has also a link from /sbin/ueventd but with this commandline it should fail ? Jun 23 19:00:25 maybe it does fallback to /init because the given is not found ? Jun 23 20:15:18 sorry, not sure I can help with that... Jun 23 21:37:22 bluelightning, looks like /init is a fallback in that case, at least i have 2 reasons to assume that a.) the given situation works b.) other people told me so ...... but then, either people from google made a mistake or try to spread confusion. **** ENDING LOGGING AT Sun Jun 24 02:59:58 2012 **** BEGIN LOGGING AT Sun Jun 24 02:59:58 2012 Jun 24 23:09:24 lautriv_tv: note that some bootoaders permit to have multiple initramfs so one init file could be overwritten Jun 24 23:10:00 i.e. http://en.gentoo-wiki.com/wiki/Initramfs#Multiple_initramfs **** ENDING LOGGING AT Mon Jun 25 02:59:58 2012 **** BEGIN LOGGING AT Mon Jun 25 02:59:59 2012 Jun 25 06:05:58 ant__ ,maybe you see this if you come back : that is indeed the case, i found already the built-in initramfs had nothing else an empty /root and a /dev/console before the additional in the boot.img. unfortunately the bootloader can't be checked/tuned without the danger of a brick. google messed that code a lot ( found obsolete parts which are still present ) and obviously they fear the change to linux . Jun 25 08:23:34 hi ant_work Jun 25 08:25:14 'morning Jun 25 08:25:52 ant_work, i wrote some lines before i got you were out ;) lemme repeat that ...... Jun 25 08:25:55 ant__ ,maybe you see this if you come back : that is indeed the case, i found already the built-in initramfs had nothing else an empty /root and a /dev/console before the additional in the boot.img. unfortunately the bootloader can't be checked/tuned without the danger of a brick. google messed that code a lot ( found obsolete parts which are still present ) and obviously they fear the change to linux . Jun 25 08:28:28 have you tried specifying the initramfs from kexec? Jun 25 08:30:15 to do so, i need first a kernel with CONFIG_KEXEC=y which fails somewhere in the customisation. still on the road to tweak something togeter which is accepted by the propietary bootloader. Jun 25 08:39:23 does th ebootloader allow tftp or some sort of kernel upload from usb/serial/net? Jun 25 08:39:36 maybe as emergency or for upgrading Jun 25 08:40:07 we discovered that on Zaurus when they once got first 'rom' upgrade Jun 25 08:41:40 i have only one choice, a so called download-mode which is feed by a host-app. Jun 25 08:41:46 right now, i'm compiling on the target (chrooted) an independent kernel which will get it's built-in initramfs from debian , if i have luck there is at least a way for a console ( external stuff will fail because module-format of another compiler gcc4.7 > old prebuilt toolchain from android ) Jun 25 08:46:11 there is enought space to optimize, i found the first partition on flash contains 109M of zeroes, also init using yaffs which was unfortunately replaced by ext4 and a ton of ascii within the partitions with lot's of path's from the developer's home-dir structure. looks like the whole thing was punched together in a hurry. Jun 25 08:50:19 heh Jun 25 08:51:19 yes, actually it looks like a headless data-block containing sources but i doubt they recompile on the fly. Jun 25 08:51:35 I doubt Jun 25 08:51:53 sound more like leftover crap but you never know... Jun 25 08:51:54 wanna see a pic ? Jun 25 08:53:16 sure, but unless you can 'talk ' with the bootloader things will be hard Jun 25 08:53:40 maybe there are specific rouitines for flashing in that partitions, as it is on Zaurus Jun 25 08:54:55 well, on Zaurus there is one 'Service Menu' image in PROM, the copy of which resides in nand and can be upgraded Jun 25 08:55:20 ...but I suppose Samsung will use only Jtag for its big batches... Jun 25 08:56:15 have you tried booting while pressing some buttons? Maybe the unit goes in a 'DFU' state or similar Jun 25 08:56:25 the partition containing boot.img ( which is 4k offset + zImage + initramfs + unused second-stage ) can be modified because upgrading and rooting works. the only thing which won't work is changing kernel's .config (add kexec) Jun 25 08:58:22 what is the exact model? Jun 25 08:58:52 Samsung Galaxy S plus / GT-I9001 Jun 25 08:59:36 look there, i see no reason to have this in flash -> http://imgur.com/NEybR Jun 25 09:00:10 (and it's much larger than the screenshot) Jun 25 09:04:23 i have to go, will be back approx 18:00 ;) Jun 25 09:06:18 hm. it looks like there was a 'test mode' Jun 25 09:06:22 http://forum.xda-developers.com/wiki/Samsung_Galaxy_S_Series Jun 25 09:32:31 hi bluelightning Jun 25 09:32:37 hi ant_work Jun 25 09:32:40 heh, busy with R-Pi ;) Jun 25 09:32:53 yes Jun 25 09:33:16 I just wish I could get the damn thing to boot with a kernel I've managed to build :/ Jun 25 09:34:55 afaik the kernel is still unstable and video does not survive to kexec Jun 25 09:35:20 typically those are board file flaws Jun 25 09:36:54 assuming the bootloader/other binary will set up the registers (we've seen that on Nanonote) Jun 25 11:22:03 Jay7: back home now? Jun 25 11:45:53 about ubifs :/ Jun 25 11:45:56 hm...so, the proper detection on ubifs implies full scan of the mtd partition ? Jun 25 11:45:56 yep Jun 25 11:47:36 ant_work: hi, yes, at office now :) Jun 25 11:48:05 Jay7: fstype code reads just 1024 bytes afaik :/ Jun 25 11:48:26 look, we just need to know if this mtd have ubi containers Jun 25 11:48:44 ubiattach does rest of work Jun 25 11:48:56 we should find LEB0 to check for ubifs Jun 25 11:49:19 butthe position of it depends on mtd hardware and on VID offset Jun 25 11:49:37 I can't see how fstype.c could do that Jun 25 11:50:09 how to find position of LEB0 then? Jun 25 11:50:25 the eraseblock contains its offset Jun 25 11:50:31 iirc, we should find some 'superblock', read offset from there and jump Jun 25 11:50:39 well, the offset of first VolumeID Jun 25 11:50:59 i.e. we need to jump through mtd Jun 25 11:51:18 see ubi_ec_hdr Jun 25 11:51:19 http://lxr.free-electrons.com/source/drivers/mtd/ubi/ubi-media.h?a=arm Jun 25 11:51:30 I can't do right now Jun 25 11:51:48 anyway, needs some code from userland, cannot be done in kernel Jun 25 11:53:24 btw, why not to check for UBI eraseblock and let ubiattach do the rest? Jun 25 11:53:41 then just check for volumes in sysfs Jun 25 11:53:43 sure Jun 25 11:53:59 i.e. not to look for LEB0 even Jun 25 11:54:01 but there is no info about filesystem on volume Jun 25 11:54:43 LEB0 contains ubifs magic supposedly Jun 25 11:54:47 atm there is no other FS's on top of ubi volumes Jun 25 11:55:07 so we can try just to mount ubifs from Jun 25 11:55:39 yes, we can do that. it's not correct, though Jun 25 11:55:40 the qustion is: can we assume we have ubifs on ubi volumes? Are there really other fs used in the wild? Jun 25 11:55:41 ant_work: ubifs not the only user of ubi Jun 25 11:55:46 well.. other way is to write special code to 'jumping' through mtd and looking for ubifs Jun 25 11:56:07 yes and populate menu with many fs on the same mtd Jun 25 11:56:16 it's a big rewrite! Jun 25 11:56:32 yes, it's rewrite.. Jun 25 11:56:56 I *suppose* it's sane to assume the first volume is boot-volume Jun 25 11:57:23 so we could ignore following volumes Jun 25 11:58:56 maybe we could even inspect this infamous LEB0 and look for the magic Jun 25 11:59:03 but just for the first Jun 25 11:59:11 and for name Jun 25 11:59:17 to show in UI Jun 25 11:59:19 I'l ldo some test on device before Jun 25 11:59:37 don't we need some wear-leveling on mtd when checking for ubifs? Jun 25 11:59:47 like we reading from zaurus NAND Jun 25 12:02:18 ubi does wear-lev. Jun 25 12:03:47 but we are reading from mtd Jun 25 12:04:09 i.e. eraseblocks have fixed positions always? Jun 25 12:10:44 on nand and nor, yes, on block 0 Jun 25 12:11:04 dunno if it is marked bad... Jun 25 12:11:06 :p Jun 25 12:12:08 the erase counter headers are on the beginning, the volume ID headers have no fix position Jun 25 12:12:31 i.e. can be several Kb beyond Jun 25 12:12:58 there is the outdated ubidesign.pdf illustrating that Jun 25 12:13:35 add to this one could add a custom offset (liek the -O params we used) Jun 25 12:14:50 ok Jun 25 12:15:05 so we shouldn't care about wear-leveling then Jun 25 12:15:07 it's good Jun 25 12:15:14 yep, ubi does for us Jun 25 16:11:58 Jay7: ubifs is for july i fear, as debian initramfs-tools needs very much my Jun 25 16:47:58 Jay7, hey i'm not that involved in your project but more in hardware ....... you may be interrested that new FLASH devices are upcoming which do the entire wear-level ( and additional ECC/CRC/remap badblock) within the chip, that means newer designs have only to deal with desired endpoint-addresses ;) Jun 25 21:15:15 Jay7: I'm rebasing and cleaning a bit the code **** ENDING LOGGING AT Tue Jun 26 02:59:58 2012 **** BEGIN LOGGING AT Tue Jun 26 02:59:58 2012 Jun 26 06:15:06 Jay7, around ? Jun 26 06:16:06 lautriv: busy, a bit later Jun 26 06:16:27 ask your question, I'll answer later Jun 26 06:17:41 just a simple thing, different embedded thingies have different serial consoles like ttyO or ttyDCC ....... can i assume the usb-gadget (g_serial) is always ttyGS0 ? Jun 26 07:48:48 lautriv: http://www.mjmwired.net/kernel/Documentation/usb/gadget_serial.txt Jun 26 07:52:53 Jay7: I've realized the separator change will be problematic. The generic error "cannot parse token 'KERNEL'" can be confusing. Maybe adding "missing LABEL! must be before KERNEL!' Jun 26 07:53:23 at least for some time, until people get used Jun 26 08:47:31 ant_work: yes, good idea Jun 26 08:57:59 no problems with images built with actual OE of course Jun 26 09:36:50 bluelightning: he he, RPJ is getting wild :D Jun 26 09:37:28 heh Jun 26 09:37:33 he's often right though Jun 26 09:37:53 yea, that's an uncomfortable role the one he plays ;) Jun 26 09:38:36 * ant_work is sometimes exactly like him Jun 26 09:39:00 i.e. bad Cassandra Jun 26 12:20:52 well, on the subject, the inclusion of the good bits of meta-oe in oe-core seems slow...we'll ask CEWG sponsorship for 2013 ;) Jun 26 13:57:52 Jay7: you should update git->github in ohloh Jun 26 13:58:10 seems not tracking anymore Jun 26 13:58:29 ant_work: what would we do with ltg repo? Jun 26 13:58:40 drop? Jun 26 13:59:22 well, maybe..I can't see any trace og th eproject there... Jun 26 14:05:08 yes, 404 http://projects.linuxtogo.org/projects/kexecboot/ **** ENDING LOGGING AT Wed Jun 27 02:59:58 2012 **** BEGIN LOGGING AT Wed Jun 27 02:59:58 2012 Jun 27 08:57:27 Jay7: FYI [PATCH] ARM: zImage: Allow DTB command line concatenation with ATAG_CMDLINE Jun 27 20:15:54 some recommendations for an arm disassembler ( preferably with tumb etc ) on linux ? **** ENDING LOGGING AT Thu Jun 28 02:59:58 2012 **** BEGIN LOGGING AT Thu Jun 28 02:59:59 2012 Jun 28 08:46:10 Jay7: I've added --with-ubiattach-binary (path) and --enable-ubi-vid-hdr-offset (bytes) Jun 28 08:46:41 now fixing the code to get the last 2 digit of mtdXY Jun 28 08:46:47 ant_work: cool Jun 28 08:48:23 split in 6 patches will make it easier for you to rewrite/import Jun 28 08:48:54 waiting for next hackaton :) Jun 28 08:50:00 yes, I still have to implement ubifs for Z in OE Jun 28 08:50:21 I'll do the changes with new 3.4 kernel Jun 28 08:51:47 so we'll get rid of that ubi vid headers offset Jun 28 23:28:22 Jay7: latest ubi patchset pushed in the fork Jun 28 23:28:46 there are still rough edges ;) Jun 28 23:29:01 kexecboot Jun 28 23:29:12 still boots :) Jun 28 23:29:27 gn **** ENDING LOGGING AT Fri Jun 29 02:59:58 2012 **** BEGIN LOGGING AT Fri Jun 29 02:59:59 2012 Jun 29 09:28:03 Jay7: i've used isdigit and atoi to get the second to last digit Jun 29 09:28:14 now I'd have to test with mtd10, 11 ... Jun 29 09:28:57 to do that I'd have to hack the mtdparts of my poor Zaurus Jun 29 09:29:29 or maybe lautriv can do that...Galaxy has plenty of mtd's :) Jun 29 09:35:28 ant_work: nice Jun 29 09:35:40 I'm going to another kind of conference :) Jun 29 09:36:10 will back at end of weekend Jun 29 09:36:44 hope is good for biz Jun 29 09:39:35 mostly self-promouting :) Jun 29 12:41:16 new EFI bootloader http://freedesktop.org/wiki/Software/gummiboot Jun 29 17:30:07 Jay7, need some code-testing ? Jun 29 20:49:15 lautriv: we should test boot from ubifs on 2 digits mtd (i.e. /dev/mtd10) Jun 29 20:54:09 I'm planning to rewrite the partition formatter utility anyway **** ENDING LOGGING AT Sat Jun 30 02:59:58 2012 **** BEGIN LOGGING AT Sat Jun 30 02:59:59 2012 Jun 30 22:02:10 bluelightning: one of the last changes to linux-yocto-tiny or the kernel-tools has stopped the compatibility of linux-yocto-tiny-kexecboot Jun 30 22:02:32 ant__: have you talked to bruce about that? Jun 30 22:02:33 i.e. all builds and runs but the config is not checked Jun 30 22:02:41 I'll write to him Jun 30 22:02:51 seems is just -tiny Jun 30 22:03:58 ..seems a new bug... Jun 30 22:03:59 WARNING: Can't find any BSP hardware or required configuration fragments. Jun 30 22:04:00 WARNING: Looked at //cfg///hdw_frags.txt and //cfg///required_frags.txt in directory: //cfg// Jun 30 22:04:00 DEBUG: Python function do_kernel_configcheck finished Jun 30 22:05:02 while it should read Jun 30 22:05:05 This BSP sets 884 kernel options that are possibly non-hardware related. Jun 30 22:05:05 The full list can be found in your kernel src dir at: Jun 30 22:05:05 meta/cfg/standard/poodle/specified_non_hdw.cfg Jun 30 22:05:05 DEBUG: Python function do_kernel_configcheck finished Jun 30 22:05:12 like linux-yocto Jun 30 22:05:39 well, I choosed -tiny because of that heap of options ;) Jun 30 23:18:19 bluelightning: ouch, now in the cpio there is new cruft Jun 30 23:18:40 ant__: what's there? Jun 30 23:19:23 /usr/share/run-postinst/run-postinst.awk Jun 30 23:20:00 ..and /lib/modules .... Jun 30 23:20:11 happily no modules, just aliases Jun 30 23:20:28 dep and symbols Jun 30 23:21:11 well, I think it is time to add some rm_cruft task to the initramfs recipe :) Jun 30 23:21:31 I know it is not the best way Jun 30 23:22:10 argh, /etc/init.d/run-postinst Jun 30 23:22:35 and the scripts in rcS.d Jun 30 23:23:04 just a few kb fwiw but still cruft Jun 30 23:23:23 yes those things really ought to not be there for an initramfs **** ENDING LOGGING AT Sun Jul 01 02:59:58 2012 **** BEGIN LOGGING AT Sun Jul 01 02:59:59 2012 Jul 01 17:50:14 Jay7, hi ...... ever tried to compile android ? Jul 01 17:50:30 lautriv: hi Jul 01 17:50:31 no Jul 01 17:51:28 i wonder how they got any market-share , it's a crapload of sources but nothing glues really together o.O Jul 01 20:35:44 Jay7: I have a pending patch which would simplify the fexecw of ubiattach...unfortunately this breaks :/ Jul 01 20:37:06 smthg like that http://paste.debian.net/177294/ Jul 01 20:38:25 hm.. probably is n = fexecw(UBIATTACH_PATH **** ENDING LOGGING AT Mon Jul 02 02:59:58 2012 **** BEGIN LOGGING AT Mon Jul 02 02:59:58 2012 Jul 02 11:36:02 hello guys Jul 02 11:54:20 hi pwgen Jul 02 12:03:43 yuppieee in short time kexecboot will support asus transformer ....!!! Jul 02 12:06:57 tegra, yes Jul 02 12:08:46 tegra 3 quad core nice when there is no android runnig ... Jul 02 12:16:44 I've seen the video. Why is it so slow to boot? Jul 02 12:17:31 seems around 6-7 seconds Jul 02 12:18:05 seems really too much w/out jffs2 scan ;) Jul 02 12:18:41 (I understand the 'bios' is slow and takes time, talking about kexecboot kernel now) Jul 02 12:20:08 I can't believe Zaurus has better/faster hardware... Jul 02 12:23:13 thats not from me , but this gave me an expression hot to do it. Jul 02 12:23:42 there is no jffs2 there, bit there are 11 partitions on the android device Jul 02 12:24:01 and some partitions are not mounting in a correct manner Jul 02 12:24:07 tell me more.. is there any UBI ? Jul 02 12:24:26 this brings me to a simple quiestion, can i blacklist devices/partitons on cmdline ? Jul 02 12:24:32 no ubi Jul 02 12:24:38 a very strange thing. Jul 02 12:24:40 yes, you can redefine mtdparts= Jul 02 12:25:07 some crypted secure boot system, that is rootable bu an application from asus Jul 02 12:25:26 ouch Jul 02 12:25:44 which bootloader? Jul 02 12:25:46 then you have 8 MB to get your kernel + initrd Jul 02 12:26:23 and you have and additional 8 MB space for a backup system ( holding down the volume - button while switching on ) Jul 02 12:26:52 i dont know somthinge like nvidia boot aor arm secure boot, Jul 02 12:28:02 there is a strange tool named blobtools to create a blob you can flash over the serial wire Jul 02 12:28:26 I've read about EFI secure boot couple of days ago... Jul 02 12:28:33 Win8 etc. Jul 02 12:28:42 soundsd really bad... Jul 02 12:29:09 it's the new trend, like Apple, Samsung and others are doing Jul 02 12:29:21 M$oft ofc ;) Jul 02 12:29:22 these bootmechanism seems to need a signatured image unless its unrooted, and these crypt mechanism is deep covered inside the arm cpu Jul 02 12:30:38 iirc you can expect the possibility to unlock it on X86 but not on arm Jul 02 12:30:45 (Microsoft patent iirc) Jul 02 12:31:09 likewise asus had decided to allow people writing to these bootarea, so i can now make an kexecboot initramfs, put it in the bootloader space, and then i can start playing around and do some oe images, btw the e17 image is allready working Jul 02 12:31:21 Ubuntu and Fedora are reacting in different ways Jul 02 12:43:56 pwgen: hi Jul 02 12:44:25 pwgen: ubiformat all partitions then ;) Jul 02 12:44:36 blacklisting some devices may be good idea, but we are reading configs from devices anyway :) Jul 02 12:45:04 so detect FS - mount - read config - blacklist this device Jul 02 12:45:13 looks a bit insane :) Jul 02 12:46:31 Jay7: can be --with-hacked-partitions .. Jul 02 12:46:54 or device-specific Jul 02 12:47:08 or boot.cfg from initramfs Jul 02 12:47:36 yes, first parsed (if present) Jul 02 12:47:37 or kernel cmdline Jul 02 12:48:05 but not every device have access to change kernel cmdline Jul 02 12:48:54 sure, there are patches around to merge atags + devicetree Jul 02 12:48:55 [PATCH] ARM: zImage: Allow DTB command line concatenation with ATAG_CMDLINE Jul 02 12:49:51 some arch have flexible cmdline: part is in firmware/bootloader and the rest in kernel/devicetree Jul 02 12:49:59 or appended Jul 02 12:50:13 iirc there are now 3 Kconfig options Jul 02 12:52:11 is there a trick to make initramfs-kexec-image ? Jul 02 12:54:35 bitbake initramfs-kexecboot-klibc-image Jul 02 12:58:48 or you mean by hand? Jul 02 12:59:27 ideally for th eintegration with OpenEmbedded we just need a BSP layer hosting th ekernel recipe and a few other things Jul 02 13:10:53 http://eric-weiss.de/oe/oe-core/meta-erics/ there is the machine.conf for the tf201 .. Jul 02 13:11:18 and kernel recipes ... and a little more . Jul 02 13:15:39 ok, it looks like you just need the meta-initramfs layer Jul 02 13:15:50 kexecboot is there Jul 02 13:16:25 ok .. will try it later this night Jul 02 13:16:39 hm, see you have it there, flattened Jul 02 13:18:05 i have to patch the kexecboot to make the kexecall with the --hardboot option Jul 02 13:19:56 ah, I've seen the patches for armv7 Jul 02 13:21:09 its doinig shutting down additional hardware inside the linux kernel to allow a sane reboot Jul 02 13:23:35 brb later- traveling home .. Jul 02 22:16:35 pwgen: once you get a kexecboot kernel, pls let it be downloaded on kexecboot site/precompiled images Jul 02 22:21:51 bluelightning: about psplash Jul 02 22:21:58 seems we need 2 fixes: Jul 02 22:22:17 1) qvga logo (rotated, 240x 320y) Jul 02 22:22:38 2) /etc/rotation Jul 02 22:22:41 http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-core/psplash/files/psplash-init Jul 02 22:22:42 vs. Jul 02 22:22:50 http://cgit.openembedded.org/openembedded/tree/recipes/psplash/files/psplash-init?h=org.openembedded.dev Jul 02 22:23:18 maybe a single patch like in oe-classic ? Jul 02 22:23:44 well whatever the solution if we can avoid any ugly hacks that would be best Jul 02 22:23:55 unfortunately I don't have time to look into it atm :( Jul 02 22:24:30 I'll do some experiments Jul 02 22:24:44 later, now busy with kernel frags Jul 02 22:24:51 (moving target :/) Jul 02 22:25:11 btw we'd need 3.4, which has rewritten support for kernel audit **** ENDING LOGGING AT Tue Jul 03 02:59:57 2012 **** BEGIN LOGGING AT Tue Jul 03 02:59:58 2012 Jul 03 07:47:58 damn, looks like there are nowhere proper sources for android-kernels whichever source i get it fails on bad code :( Jul 03 07:49:24 wicht target machine ? Jul 03 07:49:33 wich target machine ? Jul 03 07:50:53 pwgen SGS+, using the msm-7x30 as target Jul 03 07:51:44 these android thingies are often depending on the android sdk ( gcc 4.3 i think ) and they often do bad stuff in kconfig Jul 03 07:51:58 letting out dependencies. Jul 03 07:53:24 pwgen , i am using the prebuild toolchain from the SDK (actually arm-eabi- 4.4 ) with a config from /proc/config.gz from the target itself. Jul 03 07:53:26 and android is NOT linux in common sense, where a fbconsole or at least a sereial console is expected ... Jul 03 07:53:57 LOL . yea ... i "like" android Jul 03 07:54:57 pwgen , the plan is to get rid of it but i need to build a kernel which is matching the given env AND has CONFIG_KEXEC set. Jul 03 07:55:31 i found out that CONFIG_KEXEC mage ughe problems Jul 03 07:56:39 pwgen , yes, whenever i have something built with kexec, it fails to boot. tried already plain samsung, cranium and cyanogen. non is able to build. Jul 03 07:56:48 on my tf201, there is a HARDBOOT patch required and for some other devices, that is doing additional shutdown on hardware Jul 03 07:58:18 lol. the official kernel from asus failed, but i had luck and some guys where sucessfully porting to 3.1. Jul 03 08:00:38 grr. i fear i will be faced with the same problem .. i got a galaxy s3 ... and i fear the day i want to make my own kernel for this ...(:-(( Jul 03 08:06:06 pwgen , maybe they did something against kexec to force people to stay on android ;) it's all about money and market-share. actually i did grep -ri kexec ./ >/tmpX and will see if there is some evil "#ifdef CONFIG_KEXEC" inside. Jul 03 08:08:18 I was looking for that --hardboot Jul 03 08:08:34 seems old stuff Jul 03 08:08:39 https://github.com/Rodderik/android_kernel_epic4g_gb_official/commits/master Jul 03 08:08:53 here some more kexec works for Android Jul 03 08:08:55 https://github.com/CyanogenMod/android_kernel_samsung_victory/commits/gingerbread Jul 03 08:09:12 it looks to me, all those fixes should be done in kernel Jul 03 08:09:31 I mean, it's a specific issue for some arm soc Jul 03 08:10:08 last I've read, kexec was fixed on armv6/7 Jul 03 08:10:16 in kernel 3.x Jul 03 08:10:27 ..maybe I'm wrong... Jul 03 08:10:53 probably but i guess a vanilla won't work and it'S hard to get details (even a datasheet) Jul 03 08:11:35 let's see what the latest 3.5-rc5 gives on my .config ....... Jul 03 08:13:59 btw, don't trust the actual kexec-klibc-static we have in meta-initramfs. It's an older overpatched version, I'm pretty sure there is some smp hack so won't work on smp .. :/ Jul 03 08:14:15 I started with the latest now Jul 03 08:14:47 heh, kexec-tools are *the beast* when compiling klibc-static :) Jul 03 08:16:18 ah, there might be a problem ....... in arch/arm/kernel/setup.c is some routine attemting to get the total amount of MEM with "total = max_low_pfn - min_low_pfn;" but the galaxy has some restricted area for MEM outside the OS ( kernel is set to mem=88M while i assume it has 128M ) Jul 03 08:27:35 ok now kexecboot and kexec native compiled on tf201 ... Jul 03 08:27:54 with sato-sdk .. (:-)) Jul 03 08:28:46 wth is sato ? this android stuff drives me crazy. already pulled 14G of sources and still not complete for ONE bloody target. Jul 03 08:29:18 openemdedded target Jul 03 08:30:37 i was on OE before the new oe-core. not longer useful for me now :( Jul 03 08:30:47 core-image-sato-sdk.bb builds the gcc4.7 and other stuff for an native running angstrom Jul 03 08:30:53 why ? Jul 03 08:32:53 i have an older PDA, which was the preferred target in past (hx4700) but the change to oe-core went to be even worse and anything related to git looks like my hdd explode but never has a proper directory-structure. Jul 03 08:42:20 : i have a build system on my root server that uses "only" 100 G and the meta stuff is the smallest. most space is used for building and sources Jul 03 08:43:25 and the layer stuff removed a lot of unmaintained packages .. ( ok i am missing a lot, but with small work i mostly could add them ) Jul 03 08:45:26 ant_work: any suggestions where i can add "--hardboot --mem-min=0x9800000 mem-max=0xc0000000 " as option to the kexec cmd ? Jul 03 08:45:28 lautriv: hx4700 will be the next puppy in meta-handheld ;) We have an almost-done 3.2 kernel Jul 03 08:45:34 pwgen, i don't care about the space, have plenty of it but the structure is ugly. why do i have some /oe/needless/subdir/with/some/target/specific/setting/referring/to/another crap/of/link/which/is/supposed/to/be/found/in/a/build/var/somewhere/else/fuck.config ? Jul 03 08:46:46 pwgen: is it for kexec -e call? Jul 03 08:46:52 or for kexec -l? Jul 03 08:47:11 i was searching for packages with a find ../oe -name "*findme*.bb" now its find ../meta* -name "*findme*.bb" no big difference .. Jul 03 08:47:38 https://github.com/kexecboot/kexecboot/blob/master/kexecboot.c#L199 Jul 03 08:47:43 ./kexec --load-hardboot --mem-min=0x98000000 --mem-max=0xC0000000 --command-line="tegraid=30.1.3.0.0 mem=1024M@2048M vmalloc=128M androidboot.serialno=015d0787943c0004 video=tegrafb no_console_suspend=1 console=tty debug_uartport=hsport usbcore.old_scheme_first=1 lp0_vec=8192@0xbcbed000 tegra_fbmem=4098560@0xbdc15000 core_edp_mv=1300 audio_codec=wm8903 tegraboot=sdmmc gpt gpt_sector=62078975 root=/dev/sda1 rootdelay=8 halloEric" -l /boot/zImage-3.1 Jul 03 08:47:43 .10.ew --initrd=/boot/initrd.img Jul 03 08:47:52 and the the kexec -e Jul 03 08:47:57 ant_work, had mine already on 3.5-rc2 (handcrafted kernel patches) but i did a shutdown and it would never again wake up. this is no brick but rather a hardware-issue ( looks like the power-control broke) . Jul 03 08:47:57 https://github.com/kexecboot/kexecboot/blob/master/kexecboot.c#L199 Jul 03 08:48:31 *G* THX ... Jul 03 08:48:32 pwgen: quick fix is add next after "-l" Jul 03 08:48:39 and correct idx Jul 03 08:48:56 now it is 2 (next NULL after "-l") Jul 03 08:49:06 set it to index after your options Jul 03 08:49:30 ja ja ... schon klar ... yes yes seems clear ... Jul 03 08:49:53 pwgen: gut :) Jul 03 08:50:33 pwgen: don't forget to report success here ;) Jul 03 08:50:46 Jay7: btw, I have big issues when changing argv for ubi_attach :/ Jul 03 08:51:01 ant_work: what exactly? Jul 03 08:51:36 i have also write a description how the android kids can use kexeboot .. otherwise kexecboot.org will get a lots of what is a patition questions Jul 03 08:52:51 first, the getopt issue: all the option must be before the 'device' i.e. last arg must be "/dev/ubi_ctrl" Jul 03 08:53:06 (klibc only, ofc) Jul 03 08:54:08 maybe its better , in future , to add a new option to boot.cfg for setting the --mem-min --mem-max values .. Jul 03 08:54:16 Jay7: https://github.com/andrea-adami/kexecboot/commit/0cdea21e08e87cbf926e64afb1c36a982f1d5e38 Jul 03 08:54:35 if I remove ubiattach_argv[3] = NULL; Jul 03 08:54:56 and ubiattach_argv[4] = "/dev/ubi_ctrl"; (put in the declaration) Jul 03 08:55:09 pwgen: easy way is to specify another set of configure knobs Jul 03 08:55:26 Jay7: and remove one arg ( i.e. just 3) Jul 03 08:55:35 ant_work: well, it's mostly tech problem Jul 03 08:55:53 does fexecw expect some extra NULL? Jul 03 08:56:00 just no easy way to play with cmd line then :) Jul 03 08:56:04 doesn't look like... Jul 03 08:56:45 it's easy to find in man execve (fexecw is using it) Jul 03 08:57:14 Jay7: the biggest issue is I do test too late at night ;) Jul 03 08:57:36 well.. use some hardcoded string then Jul 03 08:57:39 I'll fix it later Jul 03 08:57:51 I see no big problem here Jul 03 08:57:54 just need some time :) Jul 03 08:58:18 Jay7: ok, as it is it works. I tried to clean that ubi_attach ugly code :/ Jul 03 08:58:41 ..at least putting the declarations on the beginning :p Jul 03 08:58:55 thank you for doing this anyway :) Jul 03 08:59:25 btw, we still have no our logo and splash Jul 03 08:59:38 our own (kexecboot's) Jul 03 09:00:23 true Jul 03 09:05:21 just as a sidenote, are the write-cycles on recent flash (sdcards) somewhat improoved these days ? Jul 03 09:12:11 am i visible ? Jul 03 09:13:25 disconnected :( Jul 03 09:14:34 so i got a brand-new ΅-SD with 64G (sandisk extreme) and fdisk tells me "you have to set cylinders" ? Jul 03 09:15:20 o_O Jul 03 09:16:59 ah, that is SDXC not SDHC Jul 03 09:28:10 hmm, fdisk can't handle this thingie ....... what alternatives may i try to be compatible to some android-phone ? Jul 03 09:45:53 nvm, card-reader was unable to handle it, another one works ;) Jul 03 09:46:42 grmpf .. "Can't parse keyword 'KERNEL'" error Jul 03 09:56:23 pwgen, where ? Jul 03 09:57:43 i am testing the modified kexecboot and when i start it , its complaining with "Can't parse keyword 'KERNEL'" on my boot.cfg Jul 03 09:58:05 now i hardcoded it to /boot/zImage ... Jul 03 09:58:51 iirr that is default if no option found Jul 03 10:11:24 pwgen, you tf201 runs actually android ? Jul 03 10:11:40 no efl dom image and android Jul 03 10:12:40 i wonder if there is no alternate bootloader like u-boot or grub for arm which can replace this propietary crap . Jul 03 10:12:46 a huge collection of merge oe-core images from external sd , and android , and hopefully in future kexecboot loader Jul 03 10:13:08 tegra secure boot ... a very ugly thingie Jul 03 10:16:58 pwgen: pls edit boot.cfg and add LABEL before KERNEL Jul 03 10:17:08 is a very recent change Jul 03 10:17:12 klingeling ... Jul 03 10:17:16 syslinux - style Jul 03 10:17:42 http://kexecboot.org/documentation/how_to_write_config Jul 03 10:17:54 (sorry, we knew it would break :) Jul 03 10:19:58 so we do this before release ;) Jul 03 10:20:03 Jay7: second victim ^^ Jul 03 10:20:10 (first was me :p) Jul 03 10:20:34 I hope this is last victim :) Jul 03 10:21:12 btw I'll add INITRD to the example so we cover all Jul 03 10:25:20 just as a sidenote, are the write-cycles on recent flash (sdcards) somewhat improoved these days ? Jul 03 10:27:55 lautriv: nobody knows :) Jul 03 10:28:29 some SD cards are absolutely dire in terms of reliability Jul 03 10:28:57 see the work done by Arnd Bergmann (incl presentation at ELCE last year) Jul 03 10:30:39 Jay7, bluelightning that is why i wonder about a.) how far i should use logging on such media b.) why on earth does android use ext4 there ? Jul 03 10:31:38 Arnd has some recommendations, probably worth checking his stuff out Jul 03 10:33:41 http://lwn.net/Articles/428584/ Jul 03 11:01:35 uhh thats ugly the load-hardboot kexec need " " around its command-line Jul 03 11:13:42 is there a limitation for command line lenght ? Jul 03 11:16:06 #ifndef COMMAND_LINE_SIZE Jul 03 11:16:07 #define COMMAND_LINE_SIZE 255 Jul 03 11:16:17 somebody already complained Jul 03 11:16:33 ahh char line[COMMAND_LINE_SIZE]; Jul 03 11:16:34 util.h Jul 03 11:21:41 bluelightning: I though twe were talking about raw flash device. About reliability look at this thread (courage !) Jul 03 11:21:44 http://lists.infradead.org/pipermail/linux-mtd/2011-February/034089.html Jul 03 11:22:05 I've researched 'cause I'd like to disable CONFIG_MTD_NAND_VERIFY_WRITE Jul 03 11:22:18 which can make UBI a pain Jul 03 11:22:50 there are NASA researches..bad results! Jul 03 11:24:05 FTL devices are cheaper so I think results are even worse Jul 03 11:31:32 ahhhh cin cfgparser.c char line[256] ... >[:-<] Jul 03 11:32:09 this breaks cfglines longer then 255 characters my APPEND is longer .... Jul 03 11:35:29 pwgen: Jul 03 11:35:30 May 16 16:01:42 Tequila: I've seen you had issues with cmdline length Jul 03 11:35:30 May 16 16:01:50 Jay7: https://gitorious.org/~tequila/ac100/tequilas-ac100-multiboot/blobs/master/patches/kexecboot-122708bc.patch Jul 03 11:35:41 (from our precious logs) Jul 03 11:35:56 hi ant_work Jul 03 11:36:02 * Tequila is around Jul 03 11:36:18 hi there Jul 03 11:36:32 Jay7: seems time to fix that Jul 03 11:36:37 jippie i kexecbooted to my first image .... to to do serois work no Jul 03 11:36:40 brb . Jul 03 11:36:48 \o/ Jul 03 11:36:54 just need to choose right number Jul 03 11:37:01 42 Jul 03 11:37:03 :) Jul 03 11:37:43 don't remember exactly where but I still saw something really larger somewhere :P Jul 03 11:39:39 we should rethink a bit the whole thing: cmdline now can be composed together with parts from fw/bootloader and other tags provided by kernel/devicetree Jul 03 11:40:17 I can't imagine passing 9 mtd one by one in cmdline... Jul 03 11:41:58 probably all hardcoded and CONFIG_CMDLINE_MTDPARTS/TAGS stuff is disabled. No idea about devicetrees Jul 03 12:35:16 is it thinkable that google injects improper information to /proc/config.gz ? Jul 03 12:35:44 no Jul 03 12:37:02 pwgen, but then, why fails a build from zcat /proc/config.gz >.config while a make *defconfig works ( at least it compiles much longer ) Jul 03 12:38:19 wrong compiler, additional compiler settings, .... and many more Jul 03 12:38:32 sorry must leave .. Jul 03 12:38:52 excluded all possibilites already, talking about the same compiler/sources in both cases. Jul 03 14:15:43 Jay7: FWIW Jul 03 14:15:46 http://cgit.openembedded.org/openembedded-core/tree/meta/classes/image_types.bbclass#n182 Jul 03 14:16:01 echo vol_id=0 >> ubinize.cfg Jul 03 14:16:22 it's hardcoded here as well :) Jul 03 23:53:22 bitbake -c cleansstate psplash Jul 03 23:53:35 err Jul 03 23:54:22 bluelightning: the size of psplash logo in oe-core differs with the one in oe-classic Jul 03 23:54:55 ant__: probably because I re-created it from scratch Jul 03 23:54:58 the latter is 220w x 177h Jul 03 23:55:14 I'm testing rotating it **** ENDING LOGGING AT Wed Jul 04 02:59:58 2012 **** BEGIN LOGGING AT Wed Jul 04 02:59:58 2012 Jul 04 08:39:40 hi bluelightning Jul 04 08:39:48 hi ant_work Jul 04 08:40:12 psplash is fixed :) Jul 04 08:40:51 I'd send you a rotation-patch covering the zaurus+ipaqs models as in oe-classics Jul 04 08:41:55 ideally we should create an /etc/rotation for each device but I think it's easier with a single,common patch Jul 04 08:42:40 about the size of the logo, unfortunately is slightly too wide for qvga Jul 04 08:42:56 I'll restore the old one from oe-classic to be used in meta-handheld Jul 04 08:48:00 if it works better let's use it in OE-core... Jul 04 08:50:21 yes, is maybe small on toda's tablets but ok for qvga/vga Jul 04 08:51:23 really I did not pay too much attention to the size, I just took the highest res logo I could find and then converted it Jul 04 08:52:27 np, that was already better than the old yocto logo Jul 04 08:53:23 hey, don't knock our logo ;) Jul 04 08:53:51 meaning the yocto one was horribly cropped on qvga ;) Jul 04 08:54:57 Jay7: I'm giving up with ubi code refactoring...too warm here at night Jul 04 08:56:38 maybe the beers, but I could not find any reason wy ubiattach works with 5 args (one null) and not with 4 :/ Jul 04 08:56:47 const char *ubiattach_argv[] = { NULL, NULL, NULL, NULL, "/dev/ubi_ctrl" }; Jul 04 08:56:53 + ubiattach_argv[3] = NULL; Jul 04 08:57:12 ^^^ If I remove this line ubiattach fails... Jul 04 08:57:19 strange..is defined NULL Jul 04 08:59:24 so for the sake of conformity I have also left ubiattach_argv[1] = NULL in case UBI_VID_HDR_OFFSET is undef Jul 04 08:59:27 bah Jul 04 09:00:00 as it is, the thing boots...I'll not break it :) Jul 04 13:00:18 bluelightning: have you clear how the kernel modules have to be loaded nowadays ? udev ? Jul 04 13:00:31 I'd like to fix sound Jul 04 13:01:28 hm, the last player who joined is kmod Jul 04 13:01:52 no sorry, still not clear on that myself Jul 04 13:02:03 :) Jul 04 15:41:50 hi ant_work / bluelightning Jul 04 15:46:45 hey lautriv Jul 04 15:53:17 ant_work, do you know a disassembler for arm which doesn't require ELF format ? Jul 04 15:54:11 hm.. so no readelf and no objdump... Jul 04 15:55:05 ah, sorry Jul 04 15:55:54 objdump I'd say Jul 04 15:57:32 never used so far, does it handle non-native code ? e.g. if i objdump -d -m arm on a x86_64. Jul 04 16:01:50 well, binutils must be native :) Jul 04 16:02:08 you'd need an arm-..-objdump Jul 04 16:02:54 you'd need to crosscompile binutils for target arm Jul 04 16:03:14 i have it native on the phone ;) Jul 04 16:03:22 use that one Jul 04 16:07:38 :( FIleformat not recognized. Jul 04 16:08:58 try file xy and see what it says Jul 04 16:09:49 did so already, says just "data" Jul 04 16:17:41 do you know what exactly you are disassembling? Jul 04 16:23:53 ant_work, it's about that said phone, looks like they don't like to have the kernel changed. actually i'm trying to get some init-routines from the internal flash. Jul 04 16:25:27 hm, maybe the code is in a PROM or in the GPU and you only have the env-data on that Jul 04 16:26:49 that is what i have to clear, i wonder how they would boot from a blockdrive without some routines before. however, people say the bootloader is there. Jul 04 16:28:59 at least, the first part contains only a bunch of partition-info and device-name plus 109M of zeroes o.O Jul 04 18:55:33 hi Jay7 , do you have some nerves left to talk about some kernel-issues ? Jul 04 18:56:02 hi, lautriv Jul 04 18:56:14 I'm not very experienced in kernel issues debugging :) Jul 04 18:56:36 it's not debugging but some general behaviour i wonder about. Jul 04 19:13:03 HOLY SHIT...........I GOT IT BAKED :) Jul 04 19:13:09 root@node500:/# zcat /proc/config.gz | grep -i kexec Jul 04 19:13:09 CONFIG_KEXEC=y Jul 04 19:13:50 just needed to collect 15G of crappy sources and spend a half brain to drugs ............ Jul 04 19:16:25 Jay7, ok, looks like this thingie is ready to go further ........ you and ant was talking about a test from mtd with 2 digits, tell me what you need to have prepared and i can test it. Jul 04 19:18:27 do you have more that 10 mtd partitions on device? Jul 04 19:19:43 Jay7, the internal has 28 which i must hold temporarly ( avoid brick) but i can attach a second one with 64G and as much parts as you like. Jul 04 19:20:25 omg :) Jul 04 19:20:31 no, 28 is enough Jul 04 19:21:46 nah, this 28 must stay ( android ) but since the kernel is running with kexec, i can e.g. make 12 on the other and boot from #12 . Jul 04 19:24:39 samsung is wasting a lot of space, the first one contains a handful of bytes and 109M of zeroes ..... Jul 04 19:36:50 Jay7, some url to pull sources from ? Jul 04 19:37:34 https://github.com/kexecboot/kexecboot Jul 04 19:37:42 w/o ubifs support here Jul 04 19:37:51 andrea's fork have ubifs Jul 04 19:38:54 Jay7, some special wishes like ./configure --debug-all ? Jul 04 19:39:10 --enable-debug would be useful :) Jul 04 19:40:05 Jay7, ok, resulting binary will/should replace /sbin/init from within initramfs ....... something else ? Jul 04 19:40:37 kexec should reside in /usr/bin/kexec iirc Jul 04 19:40:43 check ./configure --help Jul 04 19:41:57 i assume i can copy that part from kexec-tools from the given install. Jul 04 19:45:13 one thing ....... the given repo is already the one which like to have LABEL before KERNEL ? Jul 04 20:00:07 Jay7, still around ? Jul 04 20:00:38 yes Jul 04 20:00:52 yes, LABEL is delimiter Jul 04 20:00:58 check supplied boot.cfg sample Jul 04 20:01:19 compiled that thingie and got also zaurus.o but default is to NOT bult anything for zaurus....is that ok ? Jul 04 20:12:34 yes, it's just file w/o any code :) Jul 04 20:12:38 in your case Jul 04 20:12:45 all code is #ifdef'ed Jul 04 20:15:24 we need some autotools guru to cleanup our messy configure.ac and Makefile.am :) Jul 04 20:17:03 so i have hopefully the very last question before i can flash and fire-up ........... my debian holds kexec in /sbin while you suggested /usr/bin which one is proper ? Jul 04 20:20:47 any Jul 04 20:21:05 just point kexecboot via ./configre option where it is Jul 04 20:21:12 or make symlink Jul 04 20:21:44 i will check the defaults and place it there ... Jul 04 20:22:33 ok, final build, preparation and flash will take approx 15 minutes, will tell you soon Jul 04 20:28:25 I'll read your answer tomorrow :) Jul 04 20:28:28 * Jay7 -> sleep Jul 04 20:41:00 meh, i forgot something :( Jul 04 21:11:13 Jay7: lautriv: it's not clear to me the mtd partitions limit Jul 04 21:11:22 http://www.kernel.org/doc/Documentation/devices.txt Jul 04 21:11:42 ^^ here seems 16 Jul 04 21:11:53 googling I've read 18 Jul 04 21:12:06 now you have 28 Jul 04 21:12:08 :p Jul 04 21:13:06 right, also if you have a look at the FS-type that is NOT the expected value ( e.g. 0x83 where i know it's ext4 and so on ) Jul 04 21:13:52 I've asked in #mtd Jul 04 21:15:20 i'll try a GPT and produce 12 parts on the external ΅SD (known as /dev/block/mmcblk1) and assume the kernel will handle it the regular way ;) Jul 04 21:17:09 I'm just a bit dubious about the code..should work but you never know Jul 04 21:18:42 all that android-stuff is rather crappy, got tons of non-working stuff and even the final build did spit a lot of re-definitions but somehow works. Jul 04 21:18:55 * ant__ lost hours on constant char pointers to array of char pointers... Jul 04 21:20:22 i guess it will be even harder to get all that fancy stuff up on a regular linux, framebuffers/wifi and friends will hurt me a lot :( Jul 04 21:21:07 well, there are lot of patches for Samsung soc ready to be merged Jul 04 21:22:13 i had a read qualcomm offers some kernel-stuff but still no userland in sight. also right now i haven't a light idea if that FSO stack is useable (phone) Jul 04 21:25:30 oh, wasn't that tool for GPT-partitions on commandline gparted ? did an apt-get and it will install a lot of GUI stuff Jul 04 21:26:17 obv. just parted ;) Jul 04 21:28:49 I have read about an Odin tool , Samsung BML partitions... rings a bell? Jul 04 21:29:09 to use the MTD partition layout instead of Samsung's BML partition layout, you must use ODIN or Heimdall with a full firmware including a PIT file and repartition. Jul 04 21:30:18 ah, your device is different Jul 04 21:30:26 i use odin ( windows in VM ) to flash new roms but they are predefined using some layer-table called ARiesVE.ops this might be the right part. Jul 04 21:33:01 however, i was just curious about that flash, linux will be on a removeable card. but for further investigations it might be interresting because they waste a lot of the internal flash. Jul 04 21:33:09 This device is essentially not even a Galaxy S phone. It uses a Qualcomm Scorpion processor and Adreno 205 GPU. It has far more in common with the HTC Desire HD than the Galaxy S line. Jul 04 21:33:29 ouch Jul 04 21:33:46 yes, one reason why most sources failed to build. more CPU-Powyb Jul 04 21:33:54 er and less GPU poer Jul 04 21:34:20 keyboard seems to die, reasonable after 1b Jul 04 21:34:23 pm5 years Jul 04 22:28:27 gn **** ENDING LOGGING AT Thu Jul 05 02:59:58 2012 **** BEGIN LOGGING AT Thu Jul 05 02:59:58 2012 Jul 05 07:42:01 morning ant_work , sorry for not answering last night, had to replace my keyboard. Jul 05 07:45:39 'morning Jul 05 07:56:23 seems GPT is a bit weird on a ΅SD card. i was able to create partitions ( 13 parts for that test ), could mkfs and copy like usual but if i unplug/replug my host tells me that is no valid block-device (dev/sdc12 for /boot) Jul 05 08:03:28 hm, I'm using an hybrid GPT+MBR on one of my disks Jul 05 08:03:38 (OSX + Linux + Win) Jul 05 08:03:59 ..but is x86 and have only 4 partitions on there Jul 05 08:04:27 iirc I did even test GPT on Zaurus Jul 05 08:05:45 my host (where i created that partitions ) is x86_64, i could understand that behaviour if i had no kernel-support for GPT but then i could not use it in first attempt.......will change to mdsos with a bunch of logicals. Jul 05 08:08:24 lautriv: you did enable CONFIG_EFI_PARTITION isn't? Jul 05 08:08:55 host is not EFI, will check anyway ... Jul 05 08:09:47 #CONFIG_EFI is not set Jul 05 08:13:21 you need that :) Jul 05 08:13:27 see this excellent tutorial Jul 05 08:13:29 http://www.ibm.com/developerworks/linux/library/l-gpt/index.html Jul 05 08:14:53 will have a read but i wonder why it works on regular hdd. Jul 05 08:16:28 ah, news.. Jul 05 08:16:42 The Linux kernel's EFI stub loader—Since version 3.3.0, the Linux kernel has included included direct EFI boot support, Jul 05 08:16:53 http://www.rodsbooks.com/gdisk/booting.html Jul 05 08:17:59 i had also a system-upgrade these days, since then my block-devices are marked sticky which was not the case before. maybe i'm just too actual ;) Jul 05 08:20:02 I've seen smthgf similar whit Ubuntu's gparted iirc Jul 05 08:20:26 only root can access SD Jul 05 08:21:22 user root and group disk, no ? Jul 05 08:24:25 iirc is a checkbox in gparted 'take ownership' can't remember Jul 05 08:24:41 I spend most of my time on Gentoo...on cmdline... Jul 05 08:25:05 never happens doing that by hand :) Jul 05 08:25:41 gparted did f@ck my mbr several times :p Jul 05 08:31:07 reinstalled parted yesterday just for this purpose, still fine with fdisk :) Jul 05 08:35:53 then again, in your interest, move to oe-core asap Jul 05 08:37:47 i left oe a few weeks ago, getting tons of weird subdirs with partial configs spread over the whole which ends in a no-go. Jul 05 08:37:51 sorry, wrong window ;) Jul 05 08:38:59 I looped for a while in the gpt tools inferno..gptsyn, gpt (the big one), gparted... all partially incompatible at the time Jul 05 08:39:24 with the busted hybrid mbr required by Snow Leo Jul 05 08:39:54 that was couple of years ago...hopefully today things have improved :) Jul 05 08:41:58 some things become better, some even worse e.g. GUI's take a lot of bloat just because lazy next-gen pseudo-coders. Jul 05 08:44:15 ant_work, now another small thingie ....... i made an initramfs where /sbin/init was replaced by kexecboot. how does it load firmware ? ( kernel points to userspace-firmware-loader ) Jul 05 08:45:02 wait, you mean wifi firmware? Jul 05 08:45:55 that part is interresting later. i triedd to start and it seems to work but no console where i assume fbdev has also some firmware. Jul 05 08:46:56 no serial console? Jul 05 08:47:57 note that the kexecboot kernel has been designed w/out modules, for simplicity Jul 05 08:48:11 we expect all compiled, ready to boot from block devices Jul 05 08:48:25 framebuffer is ofc compiled in kernel :) Jul 05 08:48:50 not that easy on android, regular kernel says console=ttyDCC0 but before the kernel, there is this bootloader which also acts as battery-load-control and controls keys for debug-mode where the mode switches behaviour of USB. Jul 05 08:49:53 ok, the bootloader will probably pass part of the commandline to the kernel via atags Jul 05 08:50:19 you can change the console= of th ekexec-booted kernel, though, overriding cmdline Jul 05 08:50:28 in boot.cfg Jul 05 08:51:04 framebuffer is compiled into kernel, not sure about firmware. android (as far as i could see mounts /system within initramfs where firmware resides partially in /etc/firmware Jul 05 08:53:03 ok, you can add that second initramfs on the top of the in-kernel-image , passing it as INITRD in boot.cfg Jul 05 08:53:08 that boot.cfg is meant for the kernel's to be kexecbooted but i can't see anything on my replaced boot.img (partition containing kexec-kernel and initramfs tuned to select something ) Jul 05 08:55:38 you mean kexecboot doesn't find anything to boot? Jul 05 08:57:20 no. i boot a kernel with kexec=y and initramfs containing kexecboot in replacement of the stock-kernel. but that boot.cfg is meant to be on the alternate /boot where i switch to, or not ? Jul 05 08:57:55 yes, one boot.cfg each bootable partition Jul 05 08:58:30 have that, and may be recognized but first kerrnel has no screen. Jul 05 08:58:49 ok, and did you redefine its console= ? Jul 05 08:59:13 what is it APPENDing? Jul 05 08:59:28 not the kexec-kernel, just the linux one (external /boot) Jul 05 09:00:12 original --> CONFIG_CMDLINE="init=/sbin/init root=/dev/ram rw initrd=0x11000000,16M console=ttyDCC0 mem=88M ip=dhcp" Jul 05 09:01:00 boot.cfg -> APPEND="init=/sbin/init root=/dev/block/mmcblk1p13 rw initrd=0x11000000,16M console=fbdev mem=88M ip=dhcp" Jul 05 09:02:56 where the original works with the stock initramfs but not with the kexecboot one. Jul 05 09:03:07 seems bizarre to me.. moment Jul 05 09:06:38 APPEND="rw initrd=0x11000000,16M console=tty0 console=ttyDCC0,115200,n8 mem=88M ip=dhcp" Jul 05 09:06:45 maybe smthg lik ethat Jul 05 09:07:32 root= is passed by kexecboot: you can ofc re-direct boot Jul 05 09:08:55 ok, will give it a shot even it's not clear to me where ttyDCC ends, will see what lsusb tells me then.... Jul 05 09:09:23 Google..always helps :9 Jul 05 09:09:38 PDF] Jul 05 09:09:38 A Complete Linux Debugging Solution: Kernel, Driver, and - ARIUM Jul 05 09:09:46 ports, expecting devices defined starting with ttyDCC0 through ttyDCC7. ... first two ports, /dev/ttyDCC0 and /dev/ttyDCC1, can be used as a Console ports. Jul 05 09:10:00 sounds similar enough Jul 05 09:11:38 note that for linux, the master console is the last declared Jul 05 09:11:48 on the master you have usually stderr Jul 05 09:12:08 for kexecbot we have silenced the console... Jul 05 09:12:35 for debugging you need a verbose kernel instead... Jul 05 09:18:32 ok, device appears for a second in the bootloader-stage and in rescue-mode but not while booting :( Jul 05 09:19:49 i'm really sure they do some ugly things with their homebrew init. Jul 05 09:36:53 s/really/pretty ;) Jul 05 09:46:15 doing something in my garden to get rid of that 1000 fans sawing my nerves ..... Jul 05 10:39:58 ok, no garden -->storm :( but i found android-gadget which is obv. part of the problem . Jul 05 10:50:51 lautriv: obouosly those serial-over-usb are headhaches... Jul 05 10:54:28 ant_work, know that from the past, worked fine on 2.6.21 && hx4700 but not since 3.x i found also their android-gadget is very buggy and contains other parts in the .c they have declarations in there and routines for touchscreen and BT o.O fortunately Samsung did another one which compiled ;) Jul 05 10:57:24 the whole project is crappy, lazy coders trying to trick around and the best is my given /proc/config.gz won't compile et all regardless of stock/cranium/cyanogen and nobody knows what and how was patched. Jul 05 11:00:42 lautriv: what HW you are experimenting with? :) Jul 05 11:00:49 JFMI :) Jul 05 11:01:18 Jay7, Samsung Galaxy S plus, using a qualcomm MSM8255T Jul 05 11:17:26 ok, definetively that freakin' bootloader i said panic=300 and get a boot-loop with ~ 5 seconds :( Jul 05 11:20:17 * lautriv --> eat and calm down Jul 05 14:22:17 well... Jul 05 14:22:47 what do you think would be th eargument count of: Jul 05 14:23:07 ubiattach -O 512 -m 3 /dev/ubi_ctrl Jul 05 14:24:41 6 isn't? Jul 05 14:25:56 yep, 6 Jul 05 14:27:39 but the last is argv[5] Jul 05 14:28:04 I'm wondering now the code works after all.. Jul 05 14:28:23 https://github.com/andrea-adami/kexecboot/blob/master/util.c#L397 Jul 05 14:31:13 probably is skipping the last arg, which is in fact the default Jul 05 14:31:33 now I think I know where th eproblem lies... Jul 05 14:42:16 well, it seems the last char* in the array must be NULL, like for main Jul 05 14:43:17 "For example, a C "string" is just a pointer to the first character; the string is terminated by an ASCII NUL ('\0') character. This is also commonly done when you have an array of pointers; the last is the null pointer." Jul 05 14:43:32 ah ha.. Jul 05 14:45:37 (this is the second method, the first method is to pass a count along with the array. This is what memcpy() does) Jul 05 15:13:59 well... char *argv[4] = { "foo", "bar", NULL }; Jul 05 20:14:23 hhhhhhhhhhhhhhhelp Jul 05 20:15:05 tried to manually kexec and got "Cannot open /proc/iomem: No such file or directory" isn't that standard ? Jul 05 20:16:47 nvm, that ugly chroot-script does not always mount /proc ;) Jul 05 20:22:23 Jay7, around ? Jul 05 20:43:37 lautriv, in OE we do pupulate the initramfs with empty /dev /proc /sys Jul 05 20:44:51 devtmpfs and kernel do fill that Jul 05 20:45:54 and /sys is really only needed when using UBI Jul 05 20:46:24 ant__, sure, i have those. only thing i add is /dev/console. actually i try to manually kexec from a chrooted linux but my command is somehow taking forever. Jul 05 20:46:44 i did this --> root@node500:/# kexec -l /boot/zImage --append=init=/sbin/init root=/dev/mmcblk1p1 rw initrd=0x11000000,16M console=fbdev mem=88M ip=dhcp --initrd=/boot/lin.img Jul 05 20:46:45 console should be added by devtmpfs Jul 05 20:46:53 we start with empty /dev Jul 05 20:47:28 before devtmpfs we were using a minimum devices file Jul 05 20:48:39 ant__, because that android does some strange things, i was not sure if they follow the proper way. ( regarding to that serial-over-usb-console i found a android-gadget which contains headers and several parts of other vital code NOT neccessary for the console ) Jul 05 20:51:14 wait, what is rootfstype? Jul 05 20:51:35 ext2 Jul 05 20:51:51 you'd add it to the above string Jul 05 20:52:17 since when ? Jul 05 20:52:31 ehm..we do that since forever... Jul 05 20:52:59 kexecboot is a wrapper around kexec .. this is one of the reasons Jul 05 20:53:12 we detect fstype before Jul 05 20:54:54 meh, disconnected.........literally rootfs=ext2 ? Jul 05 20:55:00 yes Jul 05 20:55:06 wait Jul 05 20:55:12 rootfstype Jul 05 20:55:13 root= and rootfstype= Jul 05 20:55:52 kexecboot Jul 05 20:56:07 do add that automagically Jul 05 20:56:14 *does Jul 05 20:56:35 strange, i installed and booted linux on all possible platforms and medias except S390 since 2 decades and never needed rootfstype. Jul 05 20:58:55 still loading forever ....... Jul 05 21:00:50 is zImage supported ootb with kexec ? Jul 05 21:01:02 yes, zImage and uImage Jul 05 21:01:29 I've tested both on armv5te Jul 05 21:01:48 uImage was tested on beagleboard Jul 05 21:01:52 because an uncompressed vmlinux like above has 60M on a machine with 88M it might be hard ;) Jul 05 21:02:10 ah, we did test also vmlinux for ben-nanonote (mips) Jul 05 21:02:49 i took the vmlinux because the man of kexec doesn't mention any other. Jul 05 21:03:16 check the commit log, zImage since many years, uImage at least 3 years Jul 05 21:03:48 there are still rough edges, mostly in the kernel part Jul 05 21:03:55 i remember the old SGI would also prefer vmlinux, was something with the arch because they fake DMA by software. Jul 05 21:03:58 for new SOC and for mips Jul 05 21:05:42 we had issues when the kernel lzma compression was too hard and failed to decompress on 32MB devices Jul 05 21:06:30 if you use .gz this is not an issue Jul 05 21:07:28 if this one will work sometime, i have it a bit easier on the next one, a Galaxy note GT-N7000. haz 1G RAM ;) Jul 05 21:08:17 * ant__ back fixing the damned pointers to array of pointers and the missing final null pointer Jul 05 21:09:13 * ant__ drinking Weissbier before... Jul 05 21:09:17 * lautriv hands ant__ a big \0 Jul 05 21:09:44 no, a pointer to it :) Jul 05 21:09:59 urgh, Weissbier you know we brew also proper beer ? Jul 05 21:10:14 <-- german Jul 05 21:10:18 I was Muenchner for 2 years Jul 05 21:10:47 nice location, probably the best in Germany but not that beer. Jul 05 21:11:13 well, Helles is not bes altogether... Jul 05 21:11:26 there are many good beers in DE :) Jul 05 21:11:59 what I'm liking less and less are the bitter Pilsener Jul 05 21:12:17 Koelsh ist wie Trinkwasser... Jul 05 21:12:35 there are just too many Beers, like Wines in IT Jul 05 21:13:12 if you get any chance, try a "Pils" from Fόrstenberg ( swabian one, close to the border of swiss) Jul 05 21:13:34 I'll do Jul 05 21:14:03 I'm most traveling in Mchn or Koeln, sometimes in Suedwest Jul 05 21:14:45 from italy, it's not that far from the way to koelln Jul 05 21:14:54 not from Verona, no Jul 05 21:15:27 today is cheaper by plane anyway Jul 05 21:16:08 i found no verbose option to kexec, it won't come back from preload but ctrl-C works immediately o.O Jul 05 21:16:38 there is another way to use kexec, together with kdump Jul 05 21:16:45 the way itself is the half hollyday, much nice locations. Jul 05 21:16:50 I'm unsure arm will work, though Jul 05 21:17:27 I'm exactly South-Gardasee Jul 05 21:18:26 much germans go there Jul 05 21:19:56 yes, man looked a bit intel-specific but i assumed it won't really matter because it's generated from regular C code and generic handling of memory is similar. Jul 05 21:21:04 on the other hand, there is no dependency of CONFIG_KEXEC depends on x86 || x86_64 Jul 05 21:24:15 the issue (I've read in the ML) was the address relocation, where to store the second kernel Jul 05 21:24:55 this was an issue on latest arm soc needing some memory "holes" for the DSP, ... Jul 05 21:26:28 i see, ugly enought to have mem=88M in commandline while the next poser of 2 would be 120M, so something unallocated used for other things anyway. Jul 05 21:26:39 s/poser/power Jul 05 21:26:59 fu... kbd, replaced and even worse. Jul 05 21:27:12 s/120/128 Jul 05 21:27:13 yes, wireless kb do suXX Jul 05 21:28:25 usually not that much but a lot of disturbing parts around ( testing BT/WiFi of 4 different Smartphones and PDA's ) Jul 05 21:30:06 ok, now i have probably to cancel the arm-kexec part and try to dig on the propietary bootloader where one failure leads to a brick :( Jul 05 21:31:15 and there is obv. no disassembler for non-ELF formats around. Jul 05 21:33:00 tomorrow then, with more elan ;) n8 Jul 05 21:33:18 gn Jul 05 21:47:02 lautriv: Jul 05 21:47:04 hmm objdump can dump some formats Jul 05 21:47:04 but not all Jul 05 21:47:04 e.g. it can dump ELF COFF a.out and b.out Jul 05 23:50:03 bluelightning: I think the best patch for psplash is to sed the rotation=0 value out conditionally per-machine like done for udev ts rule Jul 05 23:50:53 ant__: sounds reasonable Jul 05 23:52:06 looking at psplash logo, I can't notice how the kernel bootlogo is much better, antialiased... a tad smaller, though Jul 05 23:52:40 yes, the best bootlogo is the kernel one for qvga/vga Jul 06 00:03:02 time for sleep, goodnight **** ENDING LOGGING AT Fri Jul 06 02:59:58 2012 **** BEGIN LOGGING AT Fri Jul 06 02:59:58 2012 Jul 06 07:27:39 Jay7: I've further simplified that damned argv[] Jul 06 08:48:39 ok, th emax num of mtd char devices is 15 | Jul 06 08:48:55 I post here th elinks for future memory Jul 06 08:49:59 http://www.kernel.org/doc/Documentation/devices.txt Jul 06 08:50:14 http://etutorials.org/Linux+systems/embedded+linux+systems/Chapter+7.+Storage+Device+Manipulation/7.1+MTD-Supported+Devices/ Jul 06 08:50:38 this for Linux at least... Jul 06 11:50:18 ant_work, around ? Jul 06 11:56:48 yes Jul 06 12:09:54 ant_work, so i'm not really done with kexec on arm, got some infos a.) works on omap b.) some patches were around to fit ARM where i have to check if they are applied on debian. c.) no idea why my commandline contains mem=88M because i have in fact 368M ( the whole software looks like made for an elder thingie ) Jul 06 12:10:37 a) already bekannt Jul 06 12:11:56 b) I've seen some around Jul 06 12:12:14 c) eek Jul 06 12:14:14 there is one thing courious : people talk about atags ( i have CONFIG_ATAGS_PROC=y ) where some offset is to be set in R2 and i have no idea which address or even how to check if that is accessible/suppressed by android. Jul 06 12:15:03 ah, ok, I know what 's about Jul 06 12:15:26 atag is the arm way to pass the bootparameters to the kernel Jul 06 12:15:38 the old method has been deprecated Jul 06 12:15:51 it has been introduced with 2.6.26 iirc Jul 06 12:16:11 the problem is, there are older bootloader which still ignore that Jul 06 12:16:19 example: Zaurus :) Jul 06 12:16:46 now, I imagine your 201x bootloader is atag compliant Jul 06 12:16:51 ..but... Jul 06 12:17:04 in the meanwhile arm has got devicetree support Jul 06 12:17:24 this is an alternative method for passing params (from FW) to the kernel Jul 06 12:17:42 i found a kernel-option to enable the old way, however doing kexec doesn't care about the bootloader but do i have to send something via atags instead appending to kexec -l ? Jul 06 12:17:43 so you should get more info about your bootloader Jul 06 12:18:42 the point is, kexecboot kernel should receive/read the params Jul 06 12:20:16 more info about bootloader --> need disassembler which can work on plain images without known headers (propietary part somewhere in the first 8 mtd, no documentation ) Jul 06 12:20:19 now, I can suspect you could miss some param in /proc/cmdline if booting an unofficial kernel Jul 06 12:22:15 you have still to append your cmdline to kexec Jul 06 12:22:59 my cmdline is something complete different to the default kernel-built-in, obv. using some other propietary part --> console=null androidboot.hardware=qcom androidboot.emmc=true hw=6 Jul 06 12:25:58 I have no idea about that proprietary options/modules Jul 06 12:26:29 if you see those in /proc/cmdline they have been a) hardcoded in kernel b) passed by atags Jul 06 12:28:01 c) Jul 06 12:28:01 Remove label Jul 06 12:28:01 Jul 06 12:28:01 Jul 06 12:28:01 Jul 06 12:28:02 Jul 06 12:28:04 Jul 06 12:28:06 Jul 06 12:28:08 More Jul 06 12:28:10 Jul 06 12:28:12 292 of 479 Jul 06 12:28:14 Jul 06 12:28:16 [PATCH] ARM: zImage: Allow DTB command line concatenation with ATAG_CMDLINE Jul 06 12:28:18 sry :/ Jul 06 12:28:30 * ant_work hates Gmail interface now... Jul 06 12:31:43 lol Jul 06 12:34:46 must be passed then, initial and original kernel defconfig contains "init=/sbin/init root=/dev/ram rw initrd=0x11000000,16M console=ttyDCC0 mem=88M ip=dhcp" but even more curious building the sources delivered from samsung with content from /proc/config.gz of the running device fails because of crappy code while a make ariesve_rev00_defconfig works. this thingie is patched as hell and nobody can tell me how and where. Jul 06 12:39:07 ok, i did a cat /proc/atags which said me (beside some ugly characters) the same like /proc/cmdline ....... i assume that atags were inherited from initramfs/custom init. Jul 06 12:41:36 * lautriv feels running in a circle Jul 06 12:44:36 it was the same with Zaurus..custom 2.4 kernel could read params from the old bootloader. Vanilla kernel could not Jul 06 12:45:01 but here there is an initramfs in the middle...and here dirty thing do happen Jul 06 12:46:03 even worse, samsung has some special partitions for that (hence 0..28 ) they used to set flags there e.g. for reboot recovery and hardware-setups :( Jul 06 12:50:19 still a small light on the end of the tunnel : qualcomm announced to deliver some kernel-stuff :) Jul 06 12:53:53 I was reading this http://elinux.org/Android_Booting Jul 06 12:55:04 I'm wondering now about the ramdisk necessity... Jul 06 12:55:10 initrd=0x11000000,16M Jul 06 13:02:17 ok, had a read on your link, basically what i found manually.........that 16M initrd is because they never leave it and mount /system (and others) inside so no pivot_root. Jul 06 13:06:41 see that ugly startup and vomit --> http://pastebin.com/gRTZ3gfe however if i will boot linux i must not care about that stuff, just getting hardware up. Jul 06 13:08:16 omg..thi si an init? Jul 06 13:08:41 yes and all the node/dir creation happens on each boot. Jul 06 13:08:43 more than 1000 lines of init.. Jul 06 13:11:04 HA i took the defconfig from the original 2.6.35.7 and put it into 3.5-rc5, it produced a .config without complaining about anything Jul 06 13:12:11 hm...strange...there have been changes...I do often make oldconfig to check Jul 06 13:12:45 have you maybe done 'make savedefconfig' ? Jul 06 13:13:52 this will keep only the minimum defconfig to rebuild a kernel, the options are re-expanded runtime Jul 06 13:14:03 so you'll have less diff Jul 06 13:20:04 this is on the target, just d/l'ed vanilla and copied defconfig. but i assume when i make something_defconfig and it appears some CONFIG_UNKNOWN_THINGIE the make would complain about ? Jul 06 13:23:10 <-- some steps with the dogs, then work. back @ approx 21h Jul 06 13:23:18 well, if you rename defconfig -> .config, copy it in SRC and do 'make oldconfig' it will show you the issues Jul 06 13:23:42 (usually new modules added, lcd, usb,...) Jul 06 13:49:31 Jay7: he he.. I've got it Jul 06 13:49:34 { "ubiattach", "-m", "3", "-O", "512", NULL } Jul 06 19:14:59 boah, what porks. android formatted my SD-card where the linux chroot was on with NO interaction by me o.O Jul 06 22:03:32 Jay7: UBI pain officially ended Jul 06 22:03:51 now I'll commit the kernel changes to avoid offsets **** ENDING LOGGING AT Sat Jul 07 02:59:58 2012 **** BEGIN LOGGING AT Sat Jul 07 02:59:58 2012 Jul 07 22:44:44 bluelightning: hm.. if we sed/patch psplash rotation at build time the package becomes machine-specific Jul 07 22:45:07 maybe that's why check was added at runtime...as ugly as it is **** ENDING LOGGING AT Sun Jul 08 02:59:58 2012 **** BEGIN LOGGING AT Sun Jul 08 02:59:59 2012 Jul 08 09:37:18 Jay7, do you have an idea why a kexec-l would loop to load a partial kernel again and again ? Jul 08 09:38:27 i did a strace and got this (repeating) -> http://pastebin.com/j4nnxvmz Jul 08 14:20:53 hi bluelightning ;) could you imagine what may lead kexec to run in a loop repeating the same part loading again and again ? Jul 08 14:22:44 hi lautriv Jul 08 14:22:48 no, no idea sorry Jul 08 22:13:22 lautriv: no idea about qcrypto Jul 08 22:13:30 google revealed 2 promising links Jul 08 22:13:37 http://source.firehak.com/samsung-replenish-kernel/src/67573f491ff6/SPH-M580_Kernel/Documentation/crypto/msm/qcrypto.txt Jul 08 22:14:19 xda-dev links about qcrypto.ko **** ENDING LOGGING AT Mon Jul 09 02:59:58 2012 **** BEGIN LOGGING AT Mon Jul 09 02:59:58 2012 Jul 09 08:20:29 good morning bluelightning Jul 09 08:20:37 morning ant_work Jul 09 08:36:54 bluelightning: so, even though overheated, I'm thinking bout this damned psplash Jul 09 08:37:26 are you ok having specific qvga/vga logo as we have for kernel in meta-hh ? Jul 09 08:38:30 is it acceptable to make PACKAGE_ARCH= $MACHINE_ARCH and patch the recipe at build time? Jul 09 08:43:45 back to the logo, in oe-classic we have the best-looking polished kernel-bootlogos for qvga and vga. I'm pretty sure psplash would "crop" the vga logo and not center it on qvga devices. I'll test that and in this case we could just replace the logo in oe-core with the kernel vga OE-logo.ppm we have in meta-hh. Jul 09 08:47:29 ant_work: I'm not sure... it seems like we prefer not to make things machine-specific if we can avoid it Jul 09 08:48:35 then we should 1) patch for runtime 2) use a single logo, in oe-core Jul 09 08:49:17 sounds reasonable Jul 09 08:49:33 I did ppm > png > .h and images looks nice Jul 09 08:50:12 the vga logo has the same size of the qvga, just bigger white border Jul 09 08:50:46 lets' hope it looks ok on 240x320 Jul 09 08:51:22 fwiw Angstrom did use a single logo for vga/qvga in oe-classic root/recipes/psplash/files/angstrom/psplash-poky-img.h Jul 09 09:24:02 hi ant_work ;) about your last hint ( qcrypto.ko ) i assumed that is just the first part of the required load where some mem-allocation problem jumps in,do you assume the qcrypto itself points to a problem ? Jul 09 09:28:20 i wonder why that strace points to a8400000-a840ffff where my whole vmallocinfo is in the range of 0xbf000000-0xecb43000 Jul 09 15:00:33 Jay7: http://www.ohloh.net/p/kexecboot ping Jul 09 20:27:18 :) Jul 09 21:32:49 hi lautriv Jul 09 21:32:59 welcome daal Jul 09 21:43:21 bluelightning: excellent news! psplash-logo vga can be rotated for qvga :) Jul 09 21:43:33 ant__: cool :) Jul 09 22:11:01 rebased and sent Jul 09 22:20:57 bluelightning: now this ugly runtime patch...or maybe we deploy that /etc/rotation elsewhere? Jul 09 22:21:40 i.e. whit zaurusd on Z :) Jul 09 22:21:51 ant__: hmm... not sure Jul 09 22:22:10 or as post-install hack Jul 09 22:24:42 I suppose that if we patch the recipe and add a .bbappend in meta-hh we must then set a PR_INC x.1 Jul 09 22:24:59 that for the joy of packaging feeds Jul 09 22:25:37 yes we probably should Jul 09 22:25:43 this version would the be preferred over the x Jul 09 22:25:49 from oe-core ... Jul 09 22:26:16 for anyone appending the layer Jul 09 22:26:42 if instead we do it machine-specific we have a single recipe and no hassle for others Jul 09 22:27:26 dunno by upgrading... Jul 09 22:28:28 jeez...all this brainstorming for psplash :P Jul 09 22:28:57 * ant__ heading to the beer-fridge Jul 09 23:35:10 bluelightning: patchset for meta-hh sent, pls give a check and commit if sane enough Jul 09 23:35:22 ant__: just saw, will do tomorrow Jul 09 23:35:27 * bluelightning needs sleep Jul 09 23:35:30 cya later Jul 09 23:35:46 sure Jul 09 23:35:48 gn **** ENDING LOGGING AT Tue Jul 10 02:59:58 2012 **** BEGIN LOGGING AT Tue Jul 10 02:59:58 2012 Jul 10 13:59:03 is kexec able to proper reload also initrd assumed it's never pivot-rooting rootfs ? **** ENDING LOGGING AT Wed Jul 11 02:59:58 2012 **** BEGIN LOGGING AT Wed Jul 11 02:59:58 2012 Jul 11 07:58:10 ping ant_work Jul 11 08:00:23 morning Jul 11 08:02:54 bluelightning: smthg like that http://pastebin.com/VH4099K2 Jul 11 08:03:04 my tf201 is booting now with kexecboot !!! yuppie !!! Jul 11 08:03:26 \°/ Jul 11 08:03:41 but i can not see ext4dev partitions ? kexeboot is ignoring them any ideas ? Jul 11 08:04:31 I somehow remember at first we had this 'wanna be ext4' code in klibc fstype.c Jul 11 08:04:39 nowadays should be full ext4 Jul 11 08:05:25 pwgen, what is/was on that ext4 parts ? Jul 11 08:05:56 http://git.kernel.org/?p=libs%2Fklibc%2Fklibc.git&a=search&h=HEAD&st=commit&s=ext4 Jul 11 08:06:27 there are two ext4dev partitions on the internal mmc drives , they are acessible/used from the android installation Jul 11 08:07:13 pwgen, android is used to encrypt the partitions. Jul 11 08:07:15 pwgen: we use that code from klibc (+ patches for vfat, jfs2 and finally ubi) Jul 11 08:07:43 honestly I have only tested 'prope'r ext4 Jul 11 08:08:32 pwgen: check kexecboot's kernel config for ext4dev support Jul 11 08:08:52 hi Jay7 :) Jul 11 08:08:56 ant_work: hi Jul 11 08:11:22 lautriv: btw I found that http://www.epiccm.org/2012/07/kexec-and-proof-of-concept-source-built.html (S III though). Maybe the bootloader is similer? Jul 11 08:13:37 i can read the /dev/mmcblk0p8 and /dev/mmcblk0p2 partition when i boot linux with the 3.1.10 kernel , and i think kexecboot is denying it because /proc/filesystems only contains etx4 and not ext4dev Jul 11 08:18:48 pwgen: it's very possible.. Jul 11 08:19:11 you may do quick hack Jul 11 08:19:34 return ext4 from fstype detection routine even for ext4dev Jul 11 08:19:52 like that https://github.com/andrea-adami/kexecboot/commit/65eb3fbc89d0d67be22c7dac2685fbcd77df8102 Jul 11 08:20:17 it's another quick hack :) Jul 11 08:20:46 iirc ext4dev has been obsoleted... Jul 11 08:21:01 THX . i like quick help .. Jul 11 08:21:22 ... yes i know , but tell this these android kids . Jul 11 08:21:48 .. android is NOT linux , in their point of view .. Jul 11 08:24:40 seems code from 2007 :) Jul 11 08:24:52 "will probably once get renamed to ext4, but for now module is ext4dev" Jul 11 09:34:57 :( Jul 11 10:29:18 bluelightning, hi. i'm stuck on kexec, maybe you know some channel where someone may have a deeper insight ? Jul 11 10:46:03 lautriv: only place is here really Jul 11 10:46:25 kexec-tools ML Jul 11 10:46:33 ant_work know the place :) Jul 11 10:48:14 * lautriv dislikes mailing-lists. Jul 11 10:53:57 can't really get away from them in open source Jul 11 11:29:19 kexec@lists.infradead.org Jul 11 13:14:15 bluelightning: ah, now I se what you were cooking Jul 11 13:14:59 ant_work: yes... kind of a tricky one Jul 11 21:39:34 bluelightning: now, after the cleaning, linux-yocto*_3.4 (with cfg frags) very soon Jul 11 21:40:00 ant____: awesome! :) Jul 11 21:42:51 btw, mind to ack the OE-logo patch in oe-core? Strangely is still sitting there. Jul 11 21:43:37 ant____: I think RP is taking the cautious approach with patches atm since we've recently had some pretty heavy breakage Jul 11 21:43:53 ah, this one is really harmless :) Jul 11 21:44:59 ok, now the eye-candy is done. We need to readd some sound :/ Jul 11 21:48:14 ant____: on that note (no pun intended), are you OK with enabling OSS emulation in our kernel configs? Jul 11 21:48:27 I haven't checked recently but I think we currently have it disabled Jul 11 21:48:54 the idea is to use the yocto frags for multimedia Jul 11 21:49:06 where possible Jul 11 21:49:40 and finally it seems Bruce has committed all updates he had pending Jul 11 21:49:59 but all is targeted for 3.4 iirc Jul 11 21:50:30 so I'm tempted to commit 3.4 and to update to frags afterwards... Jul 11 21:50:38 what do you think ? Jul 11 21:53:05 seems reasonable... is 3.4 OK on zaurus devices? Jul 11 21:53:40 I did boot early 3.4 on poodle + spitz. Ban't say more Jul 11 21:53:50 *Can't Jul 11 21:59:10 bluelightning: check http://git.yoctoproject.org/cgit/cgit.cgi/yocto-kernel-cache/tree/cfg Jul 11 21:59:25 sound.cfg and sound.scc do enable OSS Jul 11 21:59:29 :) Jul 11 22:00:17 ant____: great, problem solved :) Jul 11 22:00:39 ant____: I ask because Opie's sound support relies upon it and I'm too lazy to rewrite it :) Jul 11 22:01:50 hm.. kconf non-hardware sound.cfg but then in the .cfg there is a couple of hw modules? Jul 11 22:01:54 CONFIG_SND_INTEL8X0=m Jul 11 22:02:14 CONFIG_SND_ENS1370=m Jul 11 22:02:59 well, I'll run Bruce's script to split out the actual configs Jul 11 22:03:21 then we'll ask him :) Jul 11 22:03:59 he's very happy to help..I feel we are a kind of guinea pigs :p Jul 11 22:05:42 yeah he's a great guy Jul 11 22:07:50 hm... interesting.. now that /etc/rotation is not created :/ Jul 11 22:08:11 I have to admit I did not test it, I assumed you had... Jul 11 22:08:23 jeex .. is empty .. why now? Typo ? Jul 11 22:08:52 I did test it even on my shell on mailserver :) Jul 11 22:09:19 but in fact I did an edit by hand... :/ Jul 11 22:12:42 ant____: er, won't the file always be nonzero in size? Jul 11 22:12:44 yes.. typo...forgot a * Jul 11 22:13:00 sed -n 's/^[ \t]*DISPLAY_ORIENTATION[ \t]*//p' ${S}/machconfig | tr -dc '[0-9]' > ${S}/rotation Jul 11 22:13:08 before DISPLAY Jul 11 22:13:16 ant____: since you append a \n unconditionally Jul 11 22:13:47 it is 1 byte in fact Jul 11 22:14:00 right, one linefeed no doubt Jul 11 22:14:23 I added thi sparanoid check for leading white spaces..and failed copying it at home :p Jul 11 22:14:33 now we catch Jul 11 22:14:45 DISPLAY_ORIENTATION = Jul 11 22:14:46 DISPLAY_ORIENTATION= Jul 11 22:14:56 DISPLAY_ORIENTATION = Jul 11 22:15:01 DISPLAY_ORIENTATION= Jul 11 22:15:33 so people are free to abuse of the indents :) Jul 11 22:15:58 robustness is good Jul 11 22:17:34 pr bump or not? Jul 11 22:18:19 yeah we probably should technically Jul 11 22:18:38 I hope RP does get ridof those very soon Jul 11 22:19:11 don't worry, he wants to more than anyone I think :) Jul 11 22:23:26 he he Jul 11 22:26:03 hm. .what? -c cleanall formfactor and rebuild core-image-base... formfactor is not rebuilt? Jul 11 22:27:06 doesn't seem right Jul 11 22:27:25 neither Jul 11 22:34:00 ok, patch sent, at least formfactor creates the file now Jul 11 22:40:23 ant____: so again, what's the point of checking if [ -s ... if the file is always going to be > 0 in size? Jul 11 22:41:21 I just took the code from parent recipe Jul 11 22:41:53 my bad idea was probably to terminbate the line Jul 11 22:42:10 in fact, 'read' can swallow both Jul 11 22:43:44 hmm, well I'll take the patch as is especially given the current state is broken, but still... Jul 11 22:44:01 no, fel free to comment the printf, please Jul 11 22:44:28 now we have a much bigger problem with do_rootfs it seems :/ Jul 11 22:55:06 bluelightning: what does it mean Jul 11 22:55:07 Found /oe/oe-core/build/tmp-eglibc/deploy/ipk/poodle/formfactor_0.0-r23_poodle.ipk in Packages, but mtime differs - re-reading Jul 11 22:55:39 ant____: I don't know, is that something opkg is reporting? Jul 11 22:56:11 is log do_rootfs of core-image-base Jul 11 22:56:58 hmm, probably is... Jul 11 22:57:34 are you using OEBasic or OEBasicHash? Jul 11 22:57:48 stock, defaults Jul 11 22:58:02 ah, in OE-core that's OEBasic still I think Jul 11 22:59:03 is there a stamp existing to tell bitbake that formfactor is already built? Jul 11 22:59:54 well, yes byt stamps are fresh Jul 11 23:00:26 hm.. wait, there are the stamps of older version Jul 11 23:00:44 ah, just cleansstate :) Jul 11 23:00:56 I thought clean would have gotten rid of those Jul 11 23:01:25 something funny going on there... can you reproduce the situation? Jul 11 23:04:02 aehmmm Jul 11 23:04:27 core-image-base doesn't need formfactor apparently :/ just -sato Jul 11 23:05:02 and there is indeed a file with lenght 1 byte, for I added the newline Jul 11 23:05:16 now I'm rebuilding -sato Jul 11 23:06:14 hmm Jul 11 23:06:19 that might explain it then Jul 11 23:06:36 I really need to tidy up those images :/ Jul 11 23:06:54 even just sorting out the naming would be helpful Jul 11 23:09:40 at least we have to add a dependency on formfactor in psplash Jul 11 23:10:10 but I think it belongs to the 'base files', BSP Jul 11 23:10:44 hmm, I don't know if that's really necessary Jul 11 23:11:23 ok, now I have /etc/rotation, 4 bytes Jul 11 23:12:11 bluelightning: it di happen now..we have psplash but not formfactor Jul 11 23:12:40 is psplash machine-specific? Jul 11 23:12:47 no Jul 11 23:13:39 that would make it difficult to add a dependency on formfactor which is then... Jul 11 23:13:46 you might be able to get away with RRECOMMENDS Jul 11 23:13:59 which might be more appropriate anyway given it might not be available for every machine Jul 11 23:14:18 ah, right RRECOMMENDS_$PN Jul 11 23:19:10 ok, patch v2 Jul 11 23:25:29 one last thing, if recommended, will the needed file be in the image when psplash is built? Jul 11 23:26:49 it'll be in the image if it exists when psplash is installed in the image Jul 11 23:27:03 too late... Jul 11 23:27:18 s/it exists/formfactor package exists/ Jul 11 23:31:57 well, is really late for me.. I hardly follow. Finally I *think* it is ok to recommend formfactor in the psplash recipe Jul 11 23:32:15 but this is for oe-core then Jul 11 23:33:06 if you mind committing it, I'll send you a v2 patch without '\n' Jul 11 23:36:26 ant____: sure, or it can wait until tommorrow, up to you Jul 11 23:41:02 sent Jul 11 23:42:18 ant____: ok, and it can definitely read the file without \n ? Jul 11 23:42:49 I tested 'read' on my shell, on my Gentoo server Jul 11 23:43:08 it did ignore newlines in my test .sh script Jul 11 23:47:39 omg.. it has been nemed meta-initramfs :/ Jul 11 23:47:46 too late indeed :) Jul 11 23:50:59 good nite! **** ENDING LOGGING AT Thu Jul 12 02:59:58 2012 **** BEGIN LOGGING AT Thu Jul 12 02:59:58 2012 Jul 12 08:45:11 bluelightning: g'morning Jul 12 08:45:14 yawn Jul 12 08:45:19 morning ant_work Jul 12 08:45:48 sory for the mess... :) Jul 12 08:48:16 I've thought again about formfactor and I think we should add e.g. in zaurus.inc the rrecommend Jul 12 08:50:11 again, it's going to be hard to do that... if psplash were already machine specific it would be guaranteed to work Jul 12 08:50:45 imho those bsp packages are MACHINE_ESSENTIAL* Jul 12 08:51:31 MACHINE_ESSENTIAL_* stuff is technically for things required to boot the machine, which this isn't... Jul 12 08:52:43 how do you provide the dpi and other params then? Jul 12 08:55:28 ah, btw formfactor 'config' assumes /usr/sbin/fbset is installed Jul 12 08:55:38 it's rather a mess Jul 12 08:56:44 ok, I'll start removing the duplicates declaration we inherited from oe-classic, now in $machine.conf Jul 12 08:56:56 e.g. MACHINE_DISPLAY_ORIENTATION MACHINE_GUI_CLASS Jul 12 08:57:38 my point is, if formfactor is needed only for X images we have a problem ! Jul 12 09:06:27 yeah the duplication is not ideal Jul 12 09:06:33 probably should be eliminated Jul 12 09:41:48 bluelightning: about modules, a link: http://elinux.org/images/8/89/Managing_Kernel_Modules_With_kmod.pdf Jul 12 11:35:35 boah, what porks. actually it happend 2 times when i try to kexec linux android formats my SD-Card where the linux-chroot lives :( Jul 12 12:31:46 that demonstrates the superiority of Android, it's AI: survival instinct Jul 12 12:31:59 ^_^ Jul 12 12:32:37 google/samsung know what to fear ;) Jul 12 13:31:18 ant_work: nice presentation... I didn't see a lot about autoloading but I guess that's distro-specific Jul 12 16:51:45 afaik tehre is the option to deploy the m.i.t. compatibility-files **** ENDING LOGGING AT Fri Jul 13 02:59:58 2012 **** BEGIN LOGGING AT Fri Jul 13 02:59:59 2012 Jul 13 12:08:23 ant_work, are you around ? Jul 13 12:22:01 ok, i'll bother all a bit ;=) i found my self-compiled kernel (android sources) does not copy the most modules in /lib/modules/$(uname-r)/* obv. because they are not part of the mainstream. what is the proper way to get them bundled ( some list within the regular Makefiles/config might be completet ) ? Jul 13 12:50:47 modules_install is a make target Jul 13 12:52:25 i.e. on Gentoo I run make && make modules_install Jul 13 12:52:38 in OE it is wrapped Jul 13 12:57:30 ant_work, i'm aware of that, using a native (on target) compiler how can i tell the make_modules target to include the already built modules to the depmod and /lib/modules path ? Jul 13 13:18:30 maybe i need to rephrase this....... some modules enter /lib/modules/$(uname-r)/kernel/drivers/somewhere/ like expected. mainly some flash-testing modules. and there are others like the earlier mentioned qcomm.co and friends which were built fine but are not included in modules_install. i know the path but should tell the dependencies and proper install-path to the system. Jul 13 13:18:45 s/co/ko **** ENDING LOGGING AT Sat Jul 14 02:59:58 2012 **** BEGIN LOGGING AT Sat Jul 14 02:59:59 2012 **** ENDING LOGGING AT Sun Jul 15 02:59:59 2012 **** BEGIN LOGGING AT Sun Jul 15 02:59:59 2012 **** ENDING LOGGING AT Mon Jul 16 02:59:57 2012 **** BEGIN LOGGING AT Mon Jul 16 02:59:58 2012 **** ENDING LOGGING AT Tue Jul 17 02:59:58 2012 **** BEGIN LOGGING AT Tue Jul 17 02:59:58 2012 Jul 17 08:08:45 Jay7: http://www.linuxfoundation.org/collaborate/workgroups/ceag/2012-contract-work Jul 17 08:09:11 ant_work: wow! Jul 17 08:09:26 congrats :) Jul 17 08:09:45 July 12 Jul 17 08:09:57 why they still not contacted winners then? :) Jul 17 08:10:06 anyway, that's good Jul 17 08:10:07 thanks Jul 17 08:10:35 I came on that page googling around... Jul 17 08:10:48 I found that for lautriv Jul 17 08:10:49 http://www.droid-life.com/2012/07/16/cm9-inches-closer-to-verizons-locked-galaxy-s3-kexec-boot-sequence-created/ Jul 17 08:11:06 "Using kexec, which has become synonymous with signed bootloader devices, ..." Jul 17 08:11:09 :) Jul 17 08:11:42 ant_work: cool :) Jul 17 08:12:33 I think they still use some script but kexecboot could be "adapted" for Android Jul 17 08:14:17 well.. I should plan my time to include that work then :) Jul 17 08:14:29 next hackathon I wish integrate your ubifs work Jul 17 08:23:18 try to find some time for tosa...I can send you the kernels. I don't trust qemu so much... Jul 17 10:22:16 ant_work, thanks for digging that out. unfortunately the S+ (mine) is a complete different thing. however in this article you can see what a mess android is and how much efford they spend to not become kexec'd ;) Jul 17 10:23:35 heh, ugly encrypted blobs **** ENDING LOGGING AT Wed Jul 18 02:59:58 2012 **** BEGIN LOGGING AT Wed Jul 18 02:59:59 2012 Jul 18 08:05:58 hi bluelightning ;) Jul 18 08:06:16 hi lautriv Jul 18 08:07:24 bluelightning, if i git clone oe-core and need other layers like oe-meta and something from the layer-index, do i clone them beside or below oe-core ? Jul 18 08:08:41 lautriv: I would say beside, but it doesn't really matter since you put the full path in bblayers.conf Jul 18 08:09:38 ok, will see.... Jul 18 08:12:33 another thingie....... debian has some parts for phones, if i select meta-angstrom (lighter and i was used to it on my dead hx4700) are there some apps for the phone-part ? Jul 18 08:22:22 what "parts" does debian have? Jul 18 08:22:41 we have ofono and FSO Jul 18 08:24:06 i meant FSO which i found in meta-smartphone. however i have no bblayers.conf after pulling oe-core,meta-oe,meta-angstrom and meta-smartphone o.O Jul 18 08:32:32 lautriv: it gets created when you run the environment setup script... Jul 18 08:32:50 lautriv: if you're using angstrom I suggest you follow their setup instructions, that does the cloning for you Jul 18 08:34:09 following the wiki, i have to . ./oe-init-build-env and since oe-core, angstrom does a sh** on real targets except beaglebone. Jul 18 08:50:09 can't comment on that really Jul 18 08:50:14 yes on the former Jul 18 08:50:42 if you use that script then bblayers.conf can be found in conf/ under the newly created build directory (which it cd's into) Jul 18 09:00:52 i wonder why i waste always that much time for incomplete thingies, running in the same trap again and again :( Jul 18 09:02:07 I think maybe your expectations are too high... Jul 18 09:03:49 lautriv: what exactly is happening now? Jul 18 09:07:42 bluelightning, just forget it, sorry for wasting your time too. Jul 18 09:08:05 lautriv: you're not wasting my time if you don't give up... Jul 18 09:08:15 happy to help if I can Jul 18 09:13:29 bluelightning, something like oe should have a common ./conf.d where each layer has it's defined place relative to /oe and should deliver a skeleton. anything else is a mess. Jul 18 09:14:23 lautriv: what do you mean by skeleton? Jul 18 09:16:20 basic conig-file which may be tuned on demand. e.g. ./conf.d/meta-angstrom.conf and a simple include ./config.d/* Jul 18 09:22:39 lautriv: the thing is layers do not really supply a ton of options to be configured like other such systems that do use that structure (e.g. apache modules) Jul 18 09:22:46 lautriv: so I'm unsure what would be gained by doing it like that Jul 18 09:23:28 lautriv: you usually need just 3 or 4 layers Jul 18 09:23:42 imho the A. scripts are messy Jul 18 09:24:03 you end up with ton of unverified stuff Jul 18 09:24:32 best is to withe bblayers.conf by hand, imho Jul 18 09:25:30 s/white/write the/ Jul 18 09:25:59 bluelightning, the main problem is i miss a structure. one needs a half day to figure out what and where is to be configured. and at the end, you get something undesired because something else wasn't set. Jul 18 09:26:31 lautriv: angstrom provides such an automatic structure... Jul 18 09:27:53 angstrom provides nothing since they switched to oe-core, only qemu-arm and beagleboard are supported and narcissus broke a year ago. Jul 18 09:33:24 doing it by hand is not turn-key I agree, but it's not exactly a huge pain either Jul 18 09:41:39 let me give you a simple example : before my hx4700 died unexpected, i used oe-classic and had a look on a ton of configs. because this target could only build for 2.6.21, i set that kernel and got 2.6.38. after koen stated he runs 3.0 on this target, i tried that switch and got nothing. you may remember we talked about the neccesary patches which i adoped to 3.5 and was able to build a kernel from scratch but never with oe. the "new" way Jul 18 09:41:39 of oe is even worse and i bet it will take just a few month for the developers to see another no-go to maintain it. Jul 18 09:46:53 it takes work to maintain support for those old machines and frankly nobody is doing that work on a regular basis Jul 18 09:49:28 i don't longer care about the hx4700 because after a shutdown it won't ever come back ( looks like power-distribution broke ) that was an example from the past and even recent machines are the same mess. Jul 18 09:54:29 lautriv: recent machines, for example...? Jul 18 09:57:27 lautriv: fundamentally the kernel recipes have not changed migrating to oe-core. Those using linux.inc I mean. Jul 18 10:00:13 bluelightning: almost done with yocto-kernels and fragments. Found some issues with Bruce's scripts , though. He'll help me to debug later today. Jul 18 10:01:39 ant_work, my statement is rather why must there be some /other/subdir/somewhere/in/my/fs/with/another/partial/config/or/receipe/shit.conf Jul 18 10:03:24 ant_work: cool Jul 18 10:04:00 lautriv: there's one set of configs for basic builds and that's in build/conf/ Jul 18 10:13:29 bbl **** ENDING LOGGING AT Thu Jul 19 02:59:58 2012 **** BEGIN LOGGING AT Thu Jul 19 02:59:59 2012 Jul 19 21:09:42 Jay7: any official mail? Jul 19 21:10:01 ant__: still no **** ENDING LOGGING AT Fri Jul 20 02:59:59 2012 **** BEGIN LOGGING AT Fri Jul 20 02:59:59 2012 Jul 20 21:41:32 hi everyone Jul 20 21:42:54 I am playing around with kexecboot with ubi support on borzoi at the moment. Jul 20 21:45:07 does kexecboot support ubi on spitz/borzoi ? Jul 20 21:53:52 yes but you need patches Jul 20 21:54:27 pls check the news on kexecboot site, there is a link Jul 20 21:55:58 hi, I am using the sources from the fork Jul 20 21:56:06 then you probably want a kernel w/out CONFIG_MTD_NAND_VERIFY_WRITE Jul 20 21:56:23 I've tested both ways fwiw Jul 20 21:56:30 I also have that option disabled Jul 20 21:57:15 ok, then you can boot ubi images Jul 20 21:57:29 how do you build the .ubifs ? Jul 20 21:57:41 I am not a that stage yet Jul 20 21:57:59 I can execute the forked kexecboot binary on the zaurus Jul 20 21:58:01 ok, np, just follow http://free-electrons.com/blog/creating-flashing-ubi-ubifs-images/ Jul 20 21:58:11 ubiformat and so on Jul 20 21:58:40 but it does not on qemu Jul 20 21:59:05 or when I put it in the cpio Jul 20 21:59:10 no idea, I test on real hw Jul 20 22:00:07 if I put the "normal" kexecboot binary in the cpio it does boot up, not with the forked one Jul 20 22:00:24 I will try without qemu Jul 20 22:00:51 very strange, though the code is primitive Jul 20 22:01:09 yeah, I am not sure what is the reason Jul 20 22:01:13 I've tested ubi on spitz,corgi,poodle Jul 20 22:01:21 what kernel ? Jul 20 22:01:29 3.2 linux-yocto Jul 20 22:01:40 working now on 3.4 serie Jul 20 22:01:41 hm, I am on 3.5.0-rc7 vanilla Jul 20 22:01:48 ah, go back Jul 20 22:02:31 hm Jul 20 22:02:40 zaurus is well supported upstream so I've decided to be lazy and follow linux-yocto Jul 20 22:02:56 but I can boot "normal" kexecboot on 3.5.0-rc7 Jul 20 22:03:29 there have been bugs and fixes after 3.3. I did not follow all Jul 20 22:04:32 I will try now on real hw and it might be a qemu issue Jul 20 22:05:29 let me know, if you discover the issue, thx Jul 20 22:06:22 sure, no worries Jul 20 22:06:56 btw I made a patch that gives back life to the hdd led Jul 20 22:07:35 based on http://dev.gentoo.org/~josejx/ata.patch Jul 20 22:07:49 * Jay7 need to merge ant__'s changes back to upstream :) Jul 20 22:08:13 I don't remember who/when but it was decided to bind led to nand Jul 20 22:08:29 access Jul 20 22:08:36 hi Jay7 Jul 20 22:08:56 note there might be a new bug in my obfuscated code :) Jul 20 22:09:12 or qemu, or kernel... Jul 20 22:10:21 * ant__ opened ssh to zeddi for fix/polishing of kernel fragments cleaning-code Jul 20 22:10:34 still not here... Jul 20 22:11:35 I hit some strange issues with akita,spitz,tosa,collie...mmc and mtd are purged :/ Jul 20 22:14:12 are any zaurus specific patches in the yocto sources ? Jul 20 22:14:22 no Jul 20 22:14:30 we used vanilla until 3.2 Jul 20 22:14:36 ok Jul 20 22:16:51 there was/is a problem with sound on zaurus. Is there a fix yet ? Jul 20 22:17:27 iirc we don't autoload some module atm in OE Jul 20 22:17:43 I'm still testing pre Jul 20 22:17:55 /boot environment :/ Jul 20 22:18:16 not much tome for holidays in userland :) Jul 20 22:18:21 *time Jul 20 22:18:30 ah, I tried loading modules manually but it did not fix it, however if sound modules are compiled in the kernel not as module sound works Jul 20 22:18:42 well I got some free time on my hand atm Jul 20 22:19:08 look, the idea is to modularize the kernel config to have it more manageable Jul 20 22:19:22 this is what the kernel-tools are for Jul 20 22:19:45 now, I'm testing some borderline-case with Zaurus... Jul 20 22:20:07 in fact we abuse of the Yocto framework somehow :) Jul 20 22:20:35 we provide atm a full defconfig (I'm testing th epatches) Jul 20 22:21:31 for 3.2 we are using http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto/spitz/spitz.cfg Jul 20 22:22:36 note CONFIG_SND_PXA2XX_SOC_SPITZ=m Jul 20 22:23:14 well, as a module it will not work atm, udev might make it wrong somewhere Jul 20 22:23:34 we ahave recently moved to kmod Jul 20 22:23:49 I have yet to check the interaction with udev Jul 20 22:23:49 ah, sorry, I am not using OE Jul 20 22:24:25 fwiw snd as module is a legacy choice Jul 20 22:24:58 I mean, it was like that 10yrs ago and wasn't changed Jul 20 22:26:22 I would like to have it as module, too. Just saying it works if in kernel.. Jul 20 22:27:18 interesting, good to know Jul 20 22:30:40 something I noticed, sometimes (randomly?) kexecboot does not like my sd card inserted. (I am booting of CF) Jul 20 22:32:07 spitz needs more delay than other zaurus for detection Jul 20 22:32:34 I think I set 3 for delay, is that enough ? Jul 20 22:33:14 It could be some issue with the card, too Jul 20 22:33:36 it was with the CF/SD I have testewd :) Jul 20 22:33:56 iI have some very old CF, slow Jul 20 22:34:14 but modern card are much bigger thus may need more sleep Jul 20 22:34:39 I am using the original microdrive CF still Jul 20 22:35:02 ah, I did only format it once Jul 20 22:35:34 I'm mostly developing on other Zaurus models Jul 20 22:36:04 btw : booting on real hw still stalled the forked kexecboot. I will look into this. might be trivial config thing somewhere Jul 20 22:36:35 see, I'll build on eimage and see Jul 20 22:43:48 Is it possible to overclock the zaurus using pxa2xx_maxfreq=624000 ? Jul 20 22:44:43 sorry for all these questions, but you guys might tried before Jul 20 22:47:11 iirc to overclock they used pxareg or smthg Jul 20 22:47:52 they = pdaxrom / cacko Jul 20 22:49:11 yeah, back then it worked, but that was without cpufreq support I think Jul 20 22:52:58 ups Jul 20 23:02:12 I can not recommend freenode javascript webchat :) **** ENDING LOGGING AT Sat Jul 21 02:59:58 2012 **** BEGIN LOGGING AT Sat Jul 21 02:59:58 2012 Jul 21 22:21:09 Jay7: http://paste.debian.net/180151/ Jul 21 23:42:44 bluelightning: hi Jul 21 23:42:54 hi ant__ Jul 21 23:43:16 I've discovered a new bug in meta-hh for collie :/ Jul 21 23:43:32 | sumtool: option '--pad' doesn't allow an argument Jul 21 23:43:32 ah, what's up there? Jul 21 23:43:39 hello Jul 21 23:43:41 EXTRA_IMAGECMD_jffs2 = "--pad=14680064 --little-endian --eraseblock=0x20000 -n" Jul 21 23:43:43 welcome :) Jul 21 23:43:46 thank you Jul 21 23:43:49 hi BrandonB Jul 21 23:43:53 hello Jul 21 23:43:59 ant__: BrandonB is attempting to install on spitz Jul 21 23:44:08 bluelightning: is in collie.conf since eons Jul 21 23:44:08 did not work out as i expected Jul 21 23:44:35 any suggestions? Jul 21 23:44:38 ant__: hmm, how come it wasn't fixed ages ago? Jul 21 23:44:47 or do i have a bricked zaurus Jul 21 23:44:50 BrandonB: at what point did you get the file/directory not found errors? Jul 21 23:44:57 there has been some recent change to jffs2 iirc Jul 21 23:44:59 upon install Jul 21 23:45:05 no, I don't think you can brick the zaurus really Jul 21 23:45:18 you mean during the flashing process? Jul 21 23:45:20 the FS was fine, i was upgrading from the original 1.01 Jul 21 23:45:23 yes Jul 21 23:45:41 only pwgen could get one with bad eraseblock 0 :) Jul 21 23:45:43 it was like it was trying to find files that were not there, all four files in the howto are on the CF card Jul 21 23:45:55 BrandonB: can you take a picture of the screen? Jul 21 23:46:03 sure i can Jul 21 23:46:07 let me reflash Jul 21 23:46:30 these little things are so neat Jul 21 23:46:40 i have wanted one for years and finally found one for a decent price Jul 21 23:46:53 yeah I really like the spitz, it's a nice device Jul 21 23:47:03 i paid 60 shipped and its near mint condition Jul 21 23:47:07 I got mine from Japan, was a bit pricey :/ Jul 21 23:47:16 cool :) Jul 21 23:48:15 it is still a nice dev testbed for a new arm-linux developer Jul 21 23:49:43 i am a sucker for clamshell devices that can be useful in some way Jul 21 23:49:53 it's a shame the clie ux50 was not taken advantage of any more than it was Jul 21 23:50:50 bluelightning: it is that sumtool does not accept same arg as mkfs.jffs2 Jul 21 23:51:11 ant__: ah, I see, that makes sense Jul 21 23:51:11 we could actually get rid of summary now that we have ubi Jul 21 23:51:28 I thinbk I've been the only user :) Jul 21 23:52:21 now that I look at collie.conf I see the old MACHINE_GUI vars btw Jul 21 23:53:03 fun is I've found that diffeent dpi have been calculated for the same devices :] Jul 21 23:53:38 maybe X did prefer rounded values at the time Jul 21 23:54:34 BrandonB: I'm still thinking Zaurus c-860 is bettar than spitz, though whit slightly slower cpu Jul 21 23:54:35 i have captured the problem Jul 21 23:54:49 you open it to reflash in just 2 secsonds... Jul 21 23:54:53 :) Jul 21 23:55:13 spitz is rather fragile Jul 21 23:55:58 I always fear the plastic will break under my fingers Jul 21 23:56:07 when extracting battery Jul 21 23:57:51 lol Jul 21 23:57:54 I've not managed to break mine yet, but I know what you mean Jul 21 23:58:03 seems like a pretty tough little device to me- it's unscathed Jul 21 23:58:06 I dropped mine recently, still works though Jul 21 23:59:51 here we go Jul 21 23:59:59 http://img.photobucket.com/albums/v160/stampede_dude/IMG_0305.jpg Jul 22 00:00:05 http://img.photobucket.com/albums/v160/stampede_dude/IMG_0304.jpg Jul 22 00:00:12 http://img.photobucket.com/albums/v160/stampede_dude/IMG_0298.jpg Jul 22 00:00:18 http://img.photobucket.com/albums/v160/stampede_dude/IMG_0296.jpg Jul 22 00:00:24 http://img.photobucket.com/albums/v160/stampede_dude/IMG_0291.jpg Jul 22 00:02:35 bluelightning: in fact we have a bug in image_types.bbclass being we reuse EXTRA_IMAGE_CMD whith sumtool. Collie is the only case I know about using --pad=foo Jul 22 00:03:16 no issues for other Z Jul 22 00:03:25 ant__: is that option still needed with sumtool? Jul 22 00:03:31 any idea what the problem is for my machine? Jul 22 00:03:44 ant__: have you seen ^ before? Jul 22 00:04:23 BrandonB: seems a mismatch with regard to your hdd partitioning Jul 22 00:04:27 oh, "no space left on device... " Jul 22 00:04:41 interesting Jul 22 00:04:50 the previous owner said something about a "partial conversion" Jul 22 00:05:03 do you think they might have resized the partitions and i need to restore them? Jul 22 00:05:17 it looks like it yes Jul 22 00:05:26 so what do i need to do from here? Jul 22 00:05:46 look, you can better do the install steps by hand Jul 22 00:06:09 this gnu-tar is running in the installer shell Jul 22 00:06:09 i am a very amateur linux user, i apologize for my lack of knowledge Jul 22 00:06:32 you can boot from SD/CF/NAND and partition/untar as you like Jul 22 00:06:38 the hdd Jul 22 00:06:58 or, you repartition as 'requested' Jul 22 00:07:02 i need to remove the hard drive Jul 22 00:07:04 ? Jul 22 00:07:05 lemme see Jul 22 00:07:45 maybe outdated Jul 22 00:07:46 http://www.linuxtogo.org/~leggewie/index.php?title=Spitz/Install&redirect=no Jul 22 00:08:59 i will try that image Jul 22 00:09:23 the issue is the partitions you have may be too small for the images you're using Jul 22 00:09:38 how can i resize them without a root prompt Jul 22 00:09:45 what does 'df -h' say? Jul 22 00:09:54 i do not have access the the root prompt Jul 22 00:10:02 that kernel panic does not let me proceed Jul 22 00:10:34 easy solution is booting from SD/CF Jul 22 00:10:43 how do i do that Jul 22 00:11:20 you need an image to boot, though Jul 22 00:11:37 i am extremely lost Jul 22 00:11:48 i am sorry to be so troublesome, i have never worked with this device Jul 22 00:12:09 unfortunately atm the autobuilder for users is 'down' Jul 22 00:12:25 so you'd have to build your image yoursewlf Jul 22 00:12:25 if i am not mistaken i have already downloaded an image from it somewhere Jul 22 00:12:47 you are lucky I'm building a roundup of images, spitz included Jul 22 00:12:49 :) Jul 22 00:13:01 wait, let see if it's already built Jul 22 00:13:17 that would be greatly appreciated Jul 22 00:13:57 yes, about 13MB, I'll upload it Jul 22 00:14:06 i would greatly appreciate it Jul 22 00:14:28 so best way to go about it would be to boot from CF, then use root prompt to reformat partitions to suit the latest image that will not install at the present moment? Jul 22 00:14:41 yes, exactly Jul 22 00:14:55 i wonder why the partitions are screwed up on a c3000 Jul 22 00:14:57 at least to verify partitioning Jul 22 00:16:10 i do not want to take this unit apart, that is for sure Jul 22 00:25:26 ok, you first need to flash the new linux-kexecboot here Jul 22 00:25:28 http://www.megafileupload.com/en/file/361597/installkit-spitz-tar-gz.html Jul 22 00:25:42 then you can untar the image Jul 22 00:25:52 http://www.megafileupload.com/en/file/361596/core-image-base-spitz-20120721120815-rootfs-tar-gz.html Jul 22 00:26:46 microsoft SE does not like that site for some reason Jul 22 00:27:04 ah, prolly 'cause pr0n Jul 22 00:27:21 you wouldn't be willig to uploading them to mediafire would you? Jul 22 00:27:22 do you have any preferred upload site? Jul 22 00:27:27 ok Jul 22 00:27:55 hm. .sign up Jul 22 00:28:17 wait, I have a public ip atm Jul 22 00:28:34 looks like i have successfully flashed the latest japanese image Jul 22 00:28:42 can i do the partition check from therre? Jul 22 00:28:57 i know i had a console open from the old image Jul 22 00:30:14 i am at bash prompt Jul 22 00:30:52 what can i do to check partition Jul 22 00:33:19 fdisk Jul 22 00:33:25 then p to print partition table Jul 22 00:33:35 i did df -h Jul 22 00:33:42 i have many partitions Jul 22 00:34:17 or so it looks like Jul 22 00:34:48 p is not a valid switch Jul 22 00:35:14 er, sorry Jul 22 00:35:24 fdisk /dev/hda I think Jul 22 00:35:41 then you should be able to use p Jul 22 00:48:38 moment of truth now... Jul 22 00:49:02 jeez, battling with those toys routers at home... Jul 22 00:49:18 routers? what kind? Jul 22 00:49:29 do you see http at 79.7.223.176 Jul 22 00:49:33 ? Jul 22 00:49:40 yes Jul 22 00:49:45 "It Works!" Jul 22 00:49:57 now to telnet... Jul 22 00:50:02 just kidding Jul 22 00:50:57 try to dowload the two files above (e.g. installkit-spitz.tar.gz ) Jul 22 00:51:07 from that IP add? Jul 22 00:51:22 yes, if it's open.. Jul 22 00:51:37 installkit is workin' Jul 22 00:51:48 what else Jul 22 00:52:01 core-image-base-spitz-20120721120815.rootfs.tar.gz Jul 22 00:52:02 the other has - instead of . so i am not sure which is which Jul 22 00:52:16 also working, great Jul 22 00:52:23 nice Jul 22 00:52:42 what did you do, start a tiny http server and host them on your own machine Jul 22 00:52:43 ? Jul 22 00:52:55 I have apache installed Jul 22 00:53:14 but is for private network only Jul 22 00:53:20 (packages feeds) Jul 22 00:53:29 oh okay Jul 22 00:53:40 so what do i do with those two files? how do i boot from CF card directly? Jul 22 00:53:47 soon we'll have the autobuilder back Jul 22 00:54:02 you have to install the kernel in the installer Jul 22 00:54:12 how do i go about that? Jul 22 00:54:52 do i put the contents of both in the root of my CF? Jul 22 00:55:04 first the installer.tar.gz Jul 22 00:55:31 it has just updater.sh and a zImage (and gnu-tar for spitz) Jul 22 00:55:42 only those 2 files pls Jul 22 00:55:46 ok Jul 22 00:55:58 do the option 4 then CF option then Y to install, then what? Jul 22 00:56:04 then you erase them and untar on SD/CF the 13MB image Jul 22 00:56:32 to the root of the CF? then what? Jul 22 00:56:39 http://kexecboot.org/screenshots Jul 22 00:56:53 finally the image will be found Jul 22 00:57:23 sda and sdb are CF and micro-hdd Jul 22 00:57:28 so it will boot from the CF card? Jul 22 00:57:34 SD card is mmcblk Jul 22 00:57:48 yes Jul 22 00:57:53 is there not a way to boot from microdrive? Jul 22 00:57:55 j/w Jul 22 00:57:58 sure Jul 22 00:58:15 this is not as easily done as i first expected Jul 22 00:58:22 first you need to install an image in that partition Jul 22 00:58:34 this can be done by updater.sh Jul 22 00:58:42 and gnu-tar Jul 22 00:58:53 ..but the partitions must have space :) Jul 22 00:59:14 brb Jul 22 01:00:00 autobuilder just worked for me, i downloaded an image Jul 22 01:02:41 so i just extract the image to CF card and attempt a boot with it inserted? Jul 22 01:06:05 trying your images now ant. i appreciate your help. Jul 22 01:07:41 :O Jul 22 01:07:45 openembedded now! Jul 22 01:07:50 something is working, thank you so much!! Jul 22 01:08:43 "No boor devices found" Jul 22 01:11:17 is kexecboot the equivalent of GRUB for zaurus? Jul 22 01:12:36 alike Jul 22 01:13:02 are you the creator? Jul 22 01:13:04 it expects a /boot/boot.cfg file, as provided in the tar.gz image Jul 22 01:13:16 okay Jul 22 01:18:23 i have the files copying on a single core pentium m so it may be a while Jul 22 01:19:59 I have to go, it's too late here (UTC+2) Jul 22 01:20:27 alright Jul 22 01:20:30 let us know, I'll read backlogs Jul 22 01:20:33 thank you for your help, it is much appreciated Jul 22 01:20:41 yw, see you Jul 22 01:20:45 i think i can go from here if your method can boot from CF card Jul 22 01:20:50 might just work with that until tomorrow Jul 22 01:20:51 yea Jul 22 01:27:40 did not work...no boot found Jul 22 01:27:46 :( Jul 22 01:42:31 night all... sorry I wasn't much help, I was working hard on rpm rootfs issues (mostly solved, yay) Jul 22 01:42:52 no problem, thank you for the help you did profve Jul 22 01:42:59 *provide Jul 22 02:20:09 anyone here? **** ENDING LOGGING AT Sun Jul 22 02:59:59 2012 **** BEGIN LOGGING AT Sun Jul 22 02:59:59 2012 **** ENDING LOGGING AT Mon Jul 23 02:59:58 2012 **** BEGIN LOGGING AT Mon Jul 23 02:59:58 2012 Jul 23 15:18:59 hi pravarth Jul 23 15:19:16 hi Jul 23 15:43:20 bluelightning: any time for tests on collie? Jul 23 15:43:34 ant_work: not right now unfortunately no :( Jul 23 15:43:48 possibly tomorrow night though Jul 23 15:44:03 I have to commit the mkfs.ubifs, I'm unsure about collie nand Jul 23 15:44:17 I've read it's 14,125 MiB Jul 23 15:44:26 (the rw part we can use) Jul 23 15:45:23 thereis the jffs2 to fix too Jul 23 15:45:43 I'll send you the two patches later Jul 23 15:53:50 core-image-base should barely fit... Jul 23 15:56:36 mabe not, .tar.gz it's 13MiB iirc and the partition uses zlib so won't do miracles Jul 23 22:22:53 bluelightning: http://paste.debian.net/180373/ Jul 23 22:36:42 then we need an EXTRA_IMAGECMD_sum.jffs2 specific for collie **** ENDING LOGGING AT Tue Jul 24 02:59:58 2012 **** BEGIN LOGGING AT Tue Jul 24 02:59:58 2012 **** ENDING LOGGING AT Wed Jul 25 02:59:58 2012 **** BEGIN LOGGING AT Wed Jul 25 02:59:58 2012 **** ENDING LOGGING AT Thu Jul 26 02:59:58 2012 **** BEGIN LOGGING AT Thu Jul 26 02:59:58 2012 **** ENDING LOGGING AT Fri Jul 27 02:59:58 2012 **** BEGIN LOGGING AT Fri Jul 27 02:59:59 2012 **** ENDING LOGGING AT Sat Jul 28 02:59:59 2012 **** BEGIN LOGGING AT Sat Jul 28 02:59:59 2012 **** ENDING LOGGING AT Sun Jul 29 02:59:58 2012 **** BEGIN LOGGING AT Sun Jul 29 02:59:59 2012 Jul 29 21:30:53 hm.. now linux-kexecboot on spitz doesn't detect ubifs... Jul 29 21:32:17 ah ha, is ok from shell Jul 29 21:32:47 then is indeed the l-k kernel Jul 29 21:33:00 why then... Jul 29 21:51:51 Jay7: seems sysfs is populated with delay :/ Jul 29 21:51:56 on spitz... Jul 29 21:52:06 ant__: add delay there to check Jul 29 21:53:39 I've uncommented Jul 29 21:53:40 https://github.com/andrea-adami/kexecboot/blob/master/util.c#L442 Jul 29 21:53:53 missing ubi0 in sysfs: Jul 29 21:53:57 missing ubi1 in sysfs: Jul 29 21:54:01 missing ubi... in sysfs: Jul 29 21:55:52 in fact, we return -1 and a check here is not bad :) Jul 29 21:57:01 so sleep after fexcecw ? Jul 29 21:57:14 strange ... sounds bad ... Jul 29 21:57:41 maybe it is just this 3.2 kernel... Jul 29 22:13:45 sleep(3) did not help :/ Jul 29 23:23:49 strange, all ok from shell uing ubiattach-klibc binary Jul 29 23:23:52 http://paste.debian.net/181125/ Jul 29 23:24:02 will see tomorrow... Jul 29 23:24:05 gn **** ENDING LOGGING AT Mon Jul 30 02:59:58 2012 **** BEGIN LOGGING AT Mon Jul 30 02:59:59 2012 Jul 30 09:11:09 Jay7: ubiattach returns a strange code :/ Jul 30 09:11:20 n = fexecw(ubiattach_ precisely Jul 30 09:11:38 ant_work: which one? Jul 30 09:11:42 but not -1 Jul 30 09:11:49 hm.. I have th elogs on #mtd Jul 30 09:11:55 seems like 640080 Jul 30 09:12:08 should be corrisponding to 255 Jul 30 09:12:09 hm.. Jul 30 09:12:34 Artemy says sysfs is out of question Jul 30 09:13:13 it is something 'missing' in the initramfs, launched from shell of full 3.2 kernel is ok Jul 30 09:13:26 (same ubiattach-klibc) Jul 30 09:13:58 i.e. when running from other image with same kexecboot and ubiattach binaries all is ok? Jul 30 09:14:13 yes, as standalone bin is ok Jul 30 09:14:22 same binaries... Jul 30 09:14:36 may be good idea then compile shell into our image Jul 30 09:14:44 I suspect the .config of spitz/akita might be missing something Jul 30 09:14:50 and run it as init instead of kexecboot Jul 30 09:15:12 then run kexecboot or ubiattach by hands and look Jul 30 09:15:31 ah, but you need mount then as well.. Jul 30 09:16:06 even better - insert fexecw call into kexecboot Jul 30 09:16:17 which will run shell after mounting things Jul 30 09:17:37 I see Jul 30 09:18:53 I just can't understand how linux-yocto_3.2 and linux-yocto-tiny_kexecboot can differ at this point... Jul 30 09:22:47 is it working with other kernel versions? Jul 30 09:23:18 no idea, we have one single -kexecboot kernel, 3.2 Jul 30 09:24:31 the point is the same ubiattach binary is used by other pxa Zaurus Jul 30 09:25:06 maybe gcc/optimize for size is breaking pxa27x kernels Jul 30 09:34:16 Jay7: I think we should get the exit status of wait() isn't? Jul 30 09:34:49 can't remember right now Jul 30 09:35:07 but iirc, wait should return status of called proc Jul 30 09:35:24 then we can expect 16bit values Jul 30 09:36:04 8 for the child and 8 for the parent Jul 30 09:36:55 may be wrong endianess.. Jul 30 09:37:21 hm, I have no #mtd logs.. grr... Jul 30 09:38:05 there should be macro WEXITSTATUS to get exit code only Jul 30 09:38:10 http://linux.die.net/man/2/wait Jul 30 09:38:11 i.e. low-order 8 bits Jul 30 09:38:50 I've not implemented that Jul 30 09:39:01 just returning value as is from fexecw Jul 30 09:39:20 ok, it is not -1 so I'd say 'success' Jul 30 09:39:25 ? Jul 30 09:44:18 if fexecw returned 640080 (dec) then low byte is 0x50 (hex) or 80 (dec) Jul 30 09:44:39 not sure about endianess Jul 30 09:45:10 I'm sorry..it could be haven been 680084 ... I remember a pattern of 008 Jul 30 09:45:26 better to change return status to return WEXITSTATUS(status) Jul 30 09:45:46 even w/o checking for WIFEXITED() Jul 30 09:46:10 0x94 Jul 30 09:46:18 148 dec Jul 30 09:48:31 OMG Jul 30 09:48:42 I think I know why... Jul 30 09:48:51 http://paste.debian.net/181125/ Jul 30 09:49:00 mtdparts=sharpsl-nand:7168k(smf),44032k(root),-(home) Jul 30 09:49:16 http://lxr.free-electrons.com/source/arch/arm/mach-pxa/spitz.c#L796 Jul 30 09:49:37 sl-c3200 is 'terrier' and has 44032k ! Jul 30 09:49:51 s@it! Jul 30 09:50:28 64M of NAND? Jul 30 09:50:45 actually 128-7 Jul 30 09:51:05 ah, yes Jul 30 09:51:39 can't see what is wrong Jul 30 09:51:41 I don't remember havint flashed jffs2 on spitz, will test this also Jul 30 09:51:56 kexecboot gets the wrong mtdparts Jul 30 09:52:03 or am I wrong? Jul 30 09:52:38 + [zaurus root] size of /dev/mtdblock2 will be changed from 44032 to 44032 Jul 30 09:52:55 ubi is 44384256 bytes Jul 30 09:53:18 btw Jul 30 09:53:32 size is changed after ubifs detection Jul 30 09:54:06 ah.. just visible info change Jul 30 09:54:06 I'm unsure th efirst linux-kexecboot kernel has the right mtdparts Jul 30 09:54:22 it is like if I had resized, imagine Jul 30 09:54:37 the bootloader has right data but the first kernel no Jul 30 09:54:46 thats known problem Jul 30 09:55:43 I hav to take logs of first linux-kexecboot run, adding printk and so on Jul 30 09:57:09 I will test it easier redefining spitz_nand_partitions[1].size Jul 30 09:57:51 it's just unfortunate we still have a blid kernel, incapable of reading mtd :/ Jul 30 09:57:55 *blind Jul 30 09:58:27 but even if I apply, it would take years before I'll write a kernel driver :p Jul 30 10:01:15 yes, it could be just that Jul 30 10:02:46 will test jffs2, should also fail to boot Jul 30 10:03:02 s/to boot/to be detected/ Jul 30 10:11:17 boot kernel which will run code to read right mtdparts from NAND then kexec same kernel again to run kexecboot :) Jul 30 10:11:44 with right mtdparts :) Jul 30 10:12:22 what do we need is only ugly hack to read mtdparts right from kernel.. Jul 30 10:12:37 we may deploy this as our own local patch Jul 30 11:01:18 fwiw Jul 30 11:01:29 http://lists.infradead.org/pipermail/linux-mtd/2010-July/031200.html Utility for Runtime Partioning Support Jul 30 11:21:38 ant_work: was it upstreamed? Jul 30 11:24:33 checking Jul 30 11:24:36 patch here Jul 30 11:24:37 http://permalink.gmane.org/gmane.linux.drivers.mtd/30951 Jul 30 11:27:32 btw Jul 30 11:27:33 This is a runtime utility. Currently the redefined partition table will exist in current power cycle. Jul 30 11:27:55 I'l ask in #mtd later Jul 30 11:28:08 for us would be enough Jul 30 11:29:27 there are 2 patches around... Jul 30 11:30:06 np Jul 30 11:30:21 we need to change mtd partitioning info in runtime only for us Jul 30 11:30:34 next kernel will see correct mtdparts Jul 30 11:30:39 http://www.digipedia.pl/usenet/thread/18514/19409/ Jul 30 11:30:54 seems rejected : Jul 30 11:42:40 ah, no Jul 30 11:42:44 one is committed Jul 30 11:42:45 http://lists.infradead.org/pipermail/linux-mtd-cvs/2010-October/007058.html Jul 30 11:43:16 and the 007057 Jul 30 11:43:37 so yes, kernel part is done Jul 30 11:45:42 well Jul 30 11:46:06 then we can call some ioctl's and repartition our mtd according to info from bootloader Jul 30 11:46:25 s/bootloader/paraminfo/ Jul 30 11:47:11 ant_work: is artem or roman on irc? Jul 30 11:47:21 artem = dedekind Jul 30 11:47:27 on #mtd Jul 30 11:47:33 ok, that's good :) Jul 30 11:48:08 I'll add this to my todo list :) Jul 30 11:48:10 he was there list night (UCT+2) at 1° clock Jul 30 11:48:24 so I think he's in Russia Jul 30 11:48:34 TZ Jul 30 11:48:42 may be finland Jul 30 11:48:49 no, he's Intel now Jul 30 11:49:10 at least I've seen @intel Jul 30 11:49:16 nokia have hard times.. Jul 30 11:49:23 :/ Jul 30 11:49:30 E51 was a great device Jul 30 11:50:03 I have E50 still :) Jul 30 11:50:16 all batteries I had died :/ Jul 30 11:50:39 I have replaced my this winter Jul 30 11:50:46 Vodafone insists giving me an iphone each year... Jul 30 11:50:54 now I have 3 :/ Jul 30 11:51:21 3gs (great for kids) 4 (wife) 4s (me) Jul 30 11:51:52 I can't choose new phone Jul 30 11:51:58 all I see are bad Jul 30 11:52:12 ergonomically, yes Jul 30 11:52:23 I'll wait for some phones with meego/maemo :) Jul 30 11:53:02 today situation is, you can easily crash your car trying to pair the bluetooth Jul 30 11:53:28 Apple put BT 3 levels down in menu... Jul 30 11:53:38 I have to hack it btw... Jul 30 11:55:01 Jay7: btw Artem did puke seeying if ( -1 == n ) things :D Jul 30 11:55:40 hehe.. any suggestions? :) Jul 30 11:55:52 if (n < 0)? Jul 30 11:55:57 I told him code is still at debug stage Jul 30 11:56:37 no I mean he just expected (n == -1) Jul 30 11:57:08 ah, wait... Jul 30 11:57:12 you can't mistype -1 = n, but you can mistype n = -1 :) Jul 30 11:57:12 is it a pointer? Jul 30 11:57:26 maybe I misread his accent.. Jul 30 11:57:42 no, n is an int Jul 30 11:58:18 you can left/right assign, no changes Jul 30 11:58:51 n = fexecw(ubiattach_path, (char *const *)ubiattach_argv, envp); Jul 30 11:58:51 if (-1 == n) { Jul 30 12:02:11 ok, back to repartition, there is no runtime util in the adopted solution, just an extended use of Jul 30 12:02:21 BLKPG ioctrl Jul 30 12:03:04 so it is a bit of code to add to zaurus* files Jul 30 12:09:49 shouldn't be hard to implement, imho Jul 30 12:11:10 but now we are calling code at menu creation stage Jul 30 12:11:28 with repartitioning we should call it before device scanning Jul 30 12:11:42 then devices will have right sizes automatically Jul 30 12:12:42 right Jul 30 12:13:03 immediately after having mounted th devices Jul 30 12:17:34 best would be a kernel fix, though... Jul 30 12:26:21 brb Jul 30 22:07:52 Jay7: I think I've seen the syntax Jul 30 22:07:58 https://bugs.launchpad.net/ubuntu/+source/gnome-disk-utility/+bug/666038 Jul 30 22:18:01 fyi fexecw(ubiattach_ was returning 65280 Jul 30 22:58:49 Jay7: heh, what's scary is, ubifs is detected even while standard l-k-kernel reports 32 + 89 (correct are 43 + 78) Jul 30 22:59:31 it just finds UBI# in each eraseblock.... Jul 30 23:03:47 ok, patched kernel does the right thing... Jul 30 23:04:18 I'll upload this **** ENDING LOGGING AT Tue Jul 31 02:59:58 2012 **** BEGIN LOGGING AT Tue Jul 31 02:59:59 2012 Jul 31 08:23:27 bluelightning: hello Jul 31 08:23:47 hi ant_work Jul 31 08:24:28 I have an overpatched zaurus.inc to commit. I'll split in 3 patches Jul 31 08:24:58 one is about pulling formfactor in core-image-base (x-less) Jul 31 08:25:07 I ended up doing Jul 31 08:25:10 MACHINE_ESSENTIAL_EXTRA_RRECOMMENDS_append = " kexecboot-cfg " Jul 31 08:25:16 MACHINE_ESSENTIAL_EXTRA_RRECOMMENDS_append = " kexecboot-cfg formfactor " Jul 31 08:25:54 now, I remember we did agree it's not the righ tplace... Jul 31 08:28:02 in alternative I could expand MACHINE_ESSENTIAL_EXTRA_RDEPENDS_append = " kernel-base kernel-image " Jul 31 08:28:11 or add a new variable... Jul 31 08:28:58 iirc there isn't a skeleton for that Jul 31 08:29:12 some standardization would be welcomed Jul 31 08:30:17 RDEPENDS is OK assuming the packages always exist for every machine inheriting zaurus.inc Jul 31 08:30:50 I wonder about formfactor though... could that not be a dependency of whatever actually needs it? Jul 31 08:32:07 our usecase is psplash but generally only X starting scripts do rely on it Jul 31 08:32:35 right, if psplash is reading the file then psplash needs the RDEPENDS it seems to me... Jul 31 08:33:00 imho formfactor is essential Jul 31 08:33:11 in that itdefines things like HAVE_KEYBOARD=1 Jul 31 08:34:17 otherwise we go back to the now unused variables in machine.conf (MACHINE_*) Jul 31 08:38:47 ah, fwiw the Yocto BSP Guide states "here are machine-specific settings used during the build that are defined by the machconfig files" Jul 31 08:38:59 so at build time Jul 31 08:39:58 but..we did not want to patch psplash at build time, this would render it machine-specific Jul 31 21:32:17 Jay7: http://git.infradead.org/linux-mtd.git/commit/d27769ec3df1a8de9ca450d2dcd72d1ab259ba32 Jul 31 22:02:23 ant__: hi Jul 31 22:02:49 btw, it might be possible to merge your ubifs work this weekend Jul 31 22:03:00 we have next hackaton planned Jul 31 22:03:36 but there is another task I may join Jul 31 22:04:58 anyway, will you be free from 4 to 5 Aug? :) Jul 31 22:06:50 on 4 aug I'll have to take family to airport then I'm free Jul 31 22:07:25 wait, on 5th :) Jul 31 22:08:04 ok, will catch you here :) Jul 31 22:08:18 I'm now playing with that ioctl Jul 31 22:08:28 any success? Jul 31 22:08:41 or new info may be Jul 31 22:08:52 I'm still looking for comments, the use seems very low Jul 31 22:09:22 from python, in a module of DeviceKit Jul 31 22:11:13 * Jay7 -> sleep Jul 31 22:11:17 good night Jul 31 22:11:19 see you Jul 31 22:11:30 'nite Jul 31 22:24:55 he he...I think I've found it Jul 31 22:24:57 http://lists.alioth.debian.org/pipermail/parted-devel/2010-March/003543.html Jul 31 22:25:56 nice ideas about max nr of partitions btw Jul 31 22:35:58 oh, there could be even a RESIZE (seems not accepted upstream) Jul 31 22:36:18 http://lists.alioth.debian.org/pipermail/parted-devel/2010-March/003543.html Jul 31 22:42:04 last version http://www.spinics.net/lists/dm-devel/msg17545.html Jul 31 22:43:22 and userspace util :) Jul 31 22:43:24 http://www.spinics.net/lists/dm-devel/msg17546.html Jul 31 22:44:00 we can just resize, it need a kernel patch, though Jul 31 22:44:19 add/del are already mainstream Jul 31 22:44:45 well, tomorrow Jul 31 22:44:46 gn **** ENDING LOGGING AT Wed Aug 01 02:59:59 2012 **** BEGIN LOGGING AT Wed Aug 01 02:59:59 2012 Aug 01 08:48:37 Jay7: an easy enough example Aug 01 08:48:39 http://lists.gnu.org/archive/html/bug-parted/2001-09/msg00086.html Aug 01 08:49:16 ant_work: hi Aug 01 08:49:22 resize is just del + add :) Aug 01 08:49:57 we can just drop all partitions on start and create again with right sizes Aug 01 08:49:58 we could avoid parsing alltogether and just add Aug 01 08:50:07 right Aug 01 08:50:21 well, not all, but rootfs and home Aug 01 08:50:54 right, mtd0 and mtd1 should remain untouched possibly Aug 01 08:51:02 sure Aug 01 08:51:12 * Jay7 gone to office Aug 01 11:52:28 compragarantida.net<---><------>554 Spam not tolerated here, take a hike ! Aug 01 11:52:35 oops sorry.. Aug 01 11:53:33 * ant_work subjected to Brasilian spam flood Aug 01 13:23:12 Jay7: heh Aug 01 13:23:25 on #mtd Aug 01 13:23:26 ant_work> seems there isn't any userspace tool, I'll have to write it Aug 01 13:23:36 yeah, there is no a standard tool, unfortunately Aug 01 13:23:36 but if you add an 'mtdpart' or something like this to mtd-utils, no one will object Aug 01 13:23:51 hehe Aug 01 13:24:03 just take old one and rewrite :) Aug 01 13:24:25 he said the ioctls for block dev are different, forget about resize Aug 01 13:25:18 del+add is enough for me Aug 01 13:25:18 there is the code I posted above (for block I guess) or this Aug 01 13:25:21 http://lists.infradead.org/pipermail/linux-mtd/2010-July/031199.html Aug 01 13:25:44 from the rejected patchset, we could maybe adapt it Aug 01 13:25:59 i.e. should we operate char devices? (/dev/mtd*)? Aug 01 13:26:06 yes Aug 01 13:27:13 btw I found another funny thing in mtd-abi.h Aug 01 13:27:38 * Get basic MTD characteristics info (better to use sysfs) */ Aug 01 13:27:44 #define MEMGETINFO _IOR('M', 1, struct mtd_info_user) Aug 01 13:27:52 ^^ we are using MEMGETINFO Aug 01 13:28:04 better to use sysfs Aug 01 13:29:34 still, I see mtd_info_t meminfo Aug 01 13:29:52 i spopulated by this ioctl Aug 01 13:31:02 what do we getting from there? Aug 01 13:31:24 i.e. meminfo.oobsize == 16 && meminfo.writesize Aug 01 13:32:09 ah Aug 01 13:32:12 I think it's easier to hack directly zaurus.c Aug 01 13:32:15 sysfs is harder to work with.. Aug 01 13:32:27 yes, next CEWG sposorship Aug 01 13:32:29 i.e. you need to open file, read it, parse.. Aug 01 13:32:29 :) Aug 01 13:32:41 may be sysctl interface will help more Aug 01 13:33:02 anyway let's leave it as is up to deprecation :) Aug 01 13:33:25 yes, it is just for Zaurus Aug 01 13:33:45 don't touch, "it just works" Aug 01 13:36:28 anyway we have calculated the size and the offset of each mtd and we have start addr Aug 01 13:36:46 now it is just to understand how the structure is expected Aug 01 13:37:43 http://lxr.free-electrons.com/source/block/ioctl.c Aug 01 13:58:27 Hexxeh: btw, I've seen that the hack to boot Android involves fake mtd partitioning... Aug 01 14:09:35 bbl Aug 01 21:46:14 Jay7: ping Aug 01 21:51:09 ant__: hi Aug 01 21:54:50 Iμ, trying to write some C code but I already have doubts about structures Aug 01 21:55:46 look at this: seems generic code for block devices http://lists.gnu.org/archive/html/bug-parted/2001-09/msg00086.html Aug 01 21:56:37 here the struct seems different http://lists.infradead.org/pipermail/linux-mtd-cvs/2010-October/007057.html Aug 01 21:59:16 but finally we should call the ioctl as it needs: http://lxr.free-electrons.com/source/block/ioctl.c Aug 01 22:45:51 Jay7: seems really easier to hack and adapt the other deprecated runtime util which was needing extra ioctls Aug 01 22:47:01 if you need this code before 4-5 august, well, ask lumag :p Aug 01 22:47:25 just point it to this logs Aug 01 22:48:08 thius channel is becoming a collection of links like a wiki Aug 01 22:48:14 (my bad) Aug 01 22:54:50 it's better than just loose that links :) Aug 01 22:55:13 I need some instructions how to create ubifs on akita or tosa Aug 01 22:55:37 hope to merge your work and do repartitioning code this weekend Aug 01 22:57:16 but right now I need to sleep :) Aug 01 22:57:22 good night Aug 01 23:04:22 well, I'm doing smthg Aug 01 23:04:32 now just need to add the header for blk Aug 01 23:04:35 error: 'BLKPG_ADD_PARTITION' undeclared Aug 01 23:05:02 seems promising way, I've removed the sanity check/permission and merge options Aug 01 23:05:16 so only 2 ioctls are needed Aug 01 23:05:28 they are surely called in the wrong way... Aug 01 23:28:58 it compiles! Aug 01 23:34:34 *mtdpart β”‚ 8876 Aug 01 23:36:35 well, trying to compile klibc I have the usual Aug 01 23:36:41 | mtdpart.c:(.text.startup+0xec): undefined reference to `fscanf' Aug 01 23:36:51 I think I'll remove any user input :)) Aug 01 23:38:52 http://paste.debian.net/181581/ Aug 01 23:39:23 ^^^ the ioctl call is horribly wrong, I'm sure Aug 01 23:39:32 gn **** ENDING LOGGING AT Thu Aug 02 02:59:59 2012 **** BEGIN LOGGING AT Thu Aug 02 02:59:59 2012 Aug 02 08:06:33 Jay7: good morning Aug 02 08:06:59 ant_work: hi Aug 02 08:07:00 I've sent the necessary patch to build ubifs images to the ML Aug 02 08:07:08 cool Aug 02 08:07:17 I have an akt akernel ready fwiw Aug 02 08:07:20 *akita Aug 02 08:07:25 and a tosa one too Aug 02 08:07:43 with latest kexecboot+ubi Aug 02 08:08:25 do you have image to boot to create ubifs on NAND? Aug 02 08:08:39 some console image with mtd-utils Aug 02 08:08:52 yes, any image can do it. I have core-image-base for all models Aug 02 08:09:13 nice Aug 02 08:09:20 just follow the steps here: Aug 02 08:09:23 http://free-electrons.com/blog/creating-flashing-ubi-ubifs-images/ Aug 02 08:09:39 (Solution A – On a booted Linux system) Aug 02 08:10:11 fwiw I do ubimkvol /dev/ubi0 -N rootfs -m Aug 02 08:10:22 -m means full size Aug 02 08:12:43 i.e. w/o home partition/ Aug 02 08:13:04 ah, no, I repeat the steps for mtd2 and mtd3 Aug 02 08:13:30 I've calibrated the images to be max 121Mb Aug 02 08:14:42 but yes, once we can repartition on the fly we can try with e.g. 120 root and 1 home Aug 02 08:15:22 hm.. iirc there was another hack to have a single partition Aug 02 08:18:28 we may add code to deal with 0-size home Aug 02 08:19:01 i.e. just don't add it Aug 02 08:20:34 oh, yes, the issue was only with original bootloader and 2.4 kernels Aug 02 08:20:46 once booted we don't care... Aug 02 08:21:32 "The pdaXii13 installer also allows you to create a single partition " Aug 02 08:21:37 ^^ will look at this Aug 02 08:21:51 and at the repartition tool Aug 02 08:21:54 fs_resize iirc :) Aug 02 08:22:00 right Aug 02 08:22:54 I'm now trying to gather attention/suggestions/ready code in #mtd :p Aug 02 08:23:06 about the on-the-fly resizing Aug 02 11:11:55 ok, first objections .. __u32 num_partitions; Aug 02 11:12:24 "You need to use u64 to avoid problems in Aug 02 11:12:24 64-bit kernel + 32-bit user-space environments." Aug 02 11:12:38 ok, ok, fwiw :) Aug 02 11:22:44 good catch Aug 02 11:55:37 I'm not convinced that code can be paired with the new ioctls... Aug 02 11:56:12 I still have to understand what is the 'master' and why that runtime talks about 'removing parent' Aug 02 11:57:26 textual: "Ex: If partitions are to be created on mtd4,then mtd4 itself is opened and Aug 02 11:57:26 Partitions are created and then mtd4 is deleted." Aug 02 11:57:54 doesn't make sense at first reading. Aug 02 12:13:19 +1 Aug 02 12:13:35 is it possible to have nested partitions? Aug 02 12:13:41 afaik, no Aug 02 12:14:05 but may be you need to have at least one partition to open and do ioctls on it? Aug 02 12:15:56 as far as I see, he opens mtd0 Aug 02 12:16:54 see last link Aug 02 12:16:56 http://lists.infradead.org/pipermail/linux-mtd/2011-June/036430.html Aug 02 12:17:39 I hope it's just a language misunderstanding :p Aug 02 12:24:18 /dev/mtd0 is used as entry point, hence this node should be untouched Aug 02 12:24:31 we could change that to mtd1 Aug 02 12:25:22 but I feel we are really drifting from the code in kernel. iirc a 'master' devices is added as *last* Aug 02 12:25:31 so, no way it could be mtd0... Aug 02 12:27:37 well, the prototypes are Aug 02 12:27:38 +int mtd_is_master(struct mtd_info *mtd); Aug 02 12:27:38 +int mtd_add_partition(struct mtd_info *master, char *name, Aug 02 12:27:38 + long long offset, long long length); Aug 02 12:27:38 +int mtd_del_partition(struct mtd_info *master, int partno); Aug 02 12:29:58 sory for the pastebin but it's easier Aug 02 12:30:00 +int mtd_is_master(struct mtd_info *mtd) Aug 02 12:30:00 +{ Aug 02 12:30:00 + struct mtd_part *part; Aug 02 12:30:00 + int nopart = 0; Aug 02 12:30:00 + Aug 02 12:30:01 + mutex_lock(&mtd_partitions_mutex); Aug 02 12:30:03 + list_for_each_entry(part, &mtd_partitions, list) Aug 02 12:30:05 + if (&part->mtd == mtd) { Aug 02 12:30:07 + nopart = 1; Aug 02 12:30:09 + break; Aug 02 12:30:11 + } Aug 02 12:30:13 + mutex_unlock(&mtd_partitions_mutex); Aug 02 12:30:17 + Aug 02 12:30:19 + return nopart; Aug 02 12:30:21 +} Aug 02 12:31:22 ah, printk(KERN_NOTICE "Creating %d MTD partitions on \"%s\":\n", nbparts, master->name); Aug 02 12:31:22 Aug 02 12:32:01 looks like kernel-code Aug 02 12:33:04 I don't understand what does if (&part->mtd == mtd Aug 02 12:33:25 part.mtd == mtd Aug 02 12:33:29 kind of Aug 02 12:35:08 list_for_each_entry() seems macro which is calling following code for every item in mtd_partitions Aug 02 12:35:24 following code is if () {} Aug 02 12:35:44 but I'm unsure Aug 02 12:35:58 ah, no Aug 02 12:36:08 it just iterate over mtd_partitions Aug 02 12:36:17 it seem sit just checks parts have same id... Aug 02 12:36:17 looking for last item Aug 02 12:36:37 then checking that mtd is the same as last item Aug 02 12:36:41 ah, incrementing you mean, last item then Aug 02 12:36:44 you were right Aug 02 12:36:58 master is last item Aug 02 12:37:01 :/ Aug 02 12:37:16 but we have mtd0/1/2/3 from start Aug 02 12:37:34 how to delete mtd2 and mtd3 if mtd3 is master? Aug 02 12:40:20 I suppose master is mtd4 in that case Aug 02 12:40:38 i.e. mtd_is_master must return 0 Aug 02 12:49:24 anyway, for our purpose we'd just need few lines of code without user input Aug 02 12:49:48 I'll try first to make the runtime util working Aug 02 12:50:11 it's just 8k shared but really we don't need that Aug 02 12:51:29 Jay7: "Roman Tereshonkov wrote: Aug 02 12:51:29 > > > Adding mtd master device after all partitions should provide Aug 02 12:51:29 > > > minimum effect on the created partition enumeration." Aug 02 12:51:58 this was written in the previous form of his patch Aug 02 12:52:52 hm.. Aug 02 12:52:56 i.e. mtd4? Aug 02 12:54:36 see text here http://lists.infradead.org/pipermail/linux-mtd/2010-July/031016.html Aug 02 12:55:21 I think we are missing some private correspondance. Talking with Roman would be great Aug 02 12:58:03 thi sis the final version of the patch Aug 02 12:58:05 http://patchwork.ozlabs.org/patch/65054/ Aug 02 12:58:14 + /* Only master mtd device must be used to control partitions */ Aug 02 12:58:14 + if (!mtd_is_master(mtd)) Aug 02 12:58:14 + return -EACCES; Aug 02 12:59:09 I'm more confused than before Aug 02 13:06:38 Jay7: he he Aug 02 13:06:40 http://lists.infradead.org/pipermail/linux-mtd/2010-November/033175.html Aug 02 13:06:46 [PATCH] mtd: mtdchar/mtdpart: fix master device identification for mtd repartition Aug 02 21:41:41 Jay7: I'm uploading new files on k. server. Just akita atm? Aug 02 21:42:00 +tosa if you have Aug 02 21:42:22 ok. about that did it miss touchscrees? Aug 02 21:42:30 and may you publish some image to boot from sd and make ubifs Aug 02 21:43:02 s/$/?/ Aug 02 21:43:06 about boot, there is fastmap coming. huge improvement for attach time :) Aug 02 21:43:22 (for ubi) Aug 02 21:44:22 wrt OE, there is just one patch pending plus th ekexecboot update ofc Aug 02 21:44:42 bluelightning will merge it soon Aug 02 21:44:48 nice Aug 02 21:44:57 for kexecboot, I wait that you update master Aug 02 21:45:34 http://patches.openembedded.org/patch/33629/ Aug 02 21:47:24 about tosa, I have a note... Aug 02 21:47:26 # CONFIG_TOUCHSCREEN_WM97XX is not set Aug 02 21:48:00 was old lumag patch Aug 02 21:48:34 I'm pretty sure you had issues with tosa, can't remember which ones :) Aug 02 22:07:46 ant__: merged Aug 02 22:23:07 bluelightning: hi, thx Aug 02 22:23:18 Jay7: slowly uploading in /kernels Aug 02 22:23:54 Jay7: maybe we can write a shell script for flashing images runtime Aug 02 22:24:08 both jffs2 and ubi Aug 02 22:25:29 i.e. rewrite updater.sh around nandlogical nandwrite and ubiformat Aug 02 22:25:50 but at runtime :) Aug 02 22:26:41 well, the ubi utils needed are 4 in fact Aug 02 22:48:28 ant__: I'll look tomorrow Aug 02 22:48:32 now sleep Aug 02 22:49:01 I don't see the uploaded files :/ Aug 02 22:49:27 where are the images? Aug 02 22:49:54 can't remember Aug 02 22:50:08 iirc, in kexecboot subfolder Aug 02 22:50:11 ..don't tell me it did silently abort because of the size... Aug 02 22:50:34 I was in /kernels, where I uploaded those Aug 02 22:50:45 kernels are there Aug 02 22:53:42 Files must be less than 8 MB. Aug 02 22:53:56 from web UI Aug 02 22:55:30 can't change it right now Aug 02 22:55:45 upload files to some file sharing service Aug 02 22:55:50 I'll fetch tomorrow Aug 02 22:56:00 * Jay7 gone Aug 02 22:56:31 'nite Aug 03 00:17:36 http://www.crocko.com/A084E6C2563E4A8B9CC86EA10F2FF162/akita.tar.gz Aug 03 00:24:06 http://www.crocko.com/1FD973BF4056441EBB2C1CF8797615C7/c7x0.tar.gz **** ENDING LOGGING AT Fri Aug 03 02:59:58 2012 **** BEGIN LOGGING AT Fri Aug 03 02:59:59 2012 Aug 03 06:14:56 http://www.crocko.com/1FD973BF4056441EBB2C1CF8797615C7/c7x0.tar.gz Aug 03 06:15:27 http://www.crocko.com/4B3AA3EA82214494876C5338CE3E4AAB/poodle.tar.gz Aug 03 06:15:45 http://www.crocko.com/0629985413154A24BE81E220E37CDD2A/spitz.tar.gz Aug 03 06:15:59 http://www.crocko.com/C446430B0E1D4E1CB73358F9BA762B05/tosa.tar.gz Aug 03 06:39:47 ant__: hi Aug 03 06:40:07 'morning Aug 03 06:40:08 It's very hard to download things from crocko.com Aug 03 06:40:13 :/ Aug 03 06:41:08 I've tried for akita and first time get 'no available slots', next time it asked me to wait 30min before Aug 03 06:41:24 heh.. 17 min now Aug 03 06:41:30 let me see, I should have apache installed here Aug 03 06:45:27 I'll fix collie later Aug 03 06:45:29 now I got Aug 03 06:45:31 | Error: max_leb_cnt too low (1061 needed) Aug 03 06:51:40 now Aug 03 06:51:41 | Error: max_leb_cnt too low (1062 needed) Aug 03 06:51:45 ? Aug 03 06:51:52 ok, I'll create one Aug 03 06:54:25 started tosa download Aug 03 06:56:01 ah, seems collie has eraseblock 0x20000 .... Aug 03 06:56:18 ok Aug 03 06:57:46 Iμll have to check that Aug 03 06:59:42 http://linux.derkeiler.com/Mailing-Lists/Kernel/2005-11/4771.html Aug 03 07:01:13 but I see pad=14680064 --little-endian --eraseblock=0x20000 Aug 03 07:16:06 bbl Aug 03 08:11:50 :) Aug 03 08:11:59 collie bootlog you pasted is still there Aug 03 08:12:01 http://pastebin.com/se3BMrmU Aug 03 08:18:39 about collie, the partition seems 0x000e20000 with erasesize 0x00020000 -> 14811136 / erasesize 128KiB Aug 03 08:19:25 then collie has 113 PEB's Aug 03 08:20:44 (I don't know why I found different values, like --pad=14680064 Aug 03 10:16:45 bluelightning: you'll have to pull collie from the drawer Aug 03 10:20:36 ant_work: ok, I have some time this weekend I think Aug 03 10:21:49 great, I've uploaded the kernels Aug 03 10:22:03 now Jay7 should suggest how to make them available Aug 03 10:22:43 if I only coul dreach the raw html editor... Aug 03 10:26:24 I'm really n00b wrt Drupal Aug 03 10:55:00 ant_work: i.e. you see uploaded files now? Aug 03 11:11:17 I have uploaded new installkits right now Aug 03 11:14:57 hm.. why it fails yesterday then? Aug 03 11:15:09 php limit 8M Aug 03 11:15:25 how you worked this around then? :) Aug 03 11:15:36 I'm uploading just kernels :) Aug 03 11:15:40 ah Aug 03 11:15:43 well, ok Aug 03 11:15:54 remind me this night to fix limit plz Aug 03 11:17:09 btw, I'd need your assistance to update my build env to current state Aug 03 11:17:19 seems this night too Aug 03 11:18:51 or I need a way to test ubifs on x86 Aug 03 11:38:48 the only change I have is I'm pointing to my kexecboot git tree Aug 03 11:40:37 i.e. this version can do with ubifs? Aug 03 11:41:00 the files in kexecboot.org, yes Aug 03 11:41:32 wait, you mean kexecboot version? Aug 03 11:42:07 I'm pointing to the last commit in the forked tree Aug 03 11:42:34 ant_work: I mean your version with ubifs patches Aug 03 11:42:44 yes, no changes, seems rather stable Aug 03 11:43:03 nice, I hope we will merge this weekend Aug 03 11:52:32 maybe you could print ubiattach exitstatus if it fails..see last troubles :) Aug 03 11:53:24 atm we have mount failing trying to attach ubi-1 Aug 03 20:27:54 ant____: hi Aug 03 20:33:26 I've updated by OE env and started test build of minimal image Aug 03 20:36:41 great Aug 03 20:37:14 just ask me if you enounter issues Aug 03 20:37:32 I'm fixing collie rootfs's now Aug 03 20:42:57 remind me what to bitbake to get kexecboot kernel? Aug 03 20:45:01 zaurus-installer is the shortest Aug 03 20:45:17 or linux-yocto-tiny-kexecboot Aug 03 20:45:32 ---- initramfs-kexecboot-klibc-image Aug 03 20:45:40 --------kexecboot-klibc Aug 03 20:45:51 --------ubiattach-klibc Aug 03 20:46:03 --------kexec-klibc Aug 03 20:46:39 zaurus installer depends on zaurus-updater and linux-yocto-tiny-kexecboot Aug 03 20:46:44 ..like a Matrioska Aug 03 20:49:13 I have almost finished a doc for kexecboot site wrt building with OE Aug 03 20:51:09 nice Aug 03 20:51:47 ant____: what upload size is suitable for you? Aug 03 20:51:59 we have 5Gb free Aug 03 20:52:16 the images can vary... Aug 03 20:52:17 maximum upload file size I mean Aug 03 20:52:37 hm.. btw, I think it may be good idea to upload via sftp Aug 03 20:52:44 not via http Aug 03 20:53:02 or that Aug 03 20:53:13 give me your ssh public key Aug 03 20:53:20 I'll add your account on that server Aug 03 20:53:37 ok Aug 03 20:53:50 ouch..collie...| Error: max_leb_cnt too low (130 needed) Aug 03 20:54:38 I'm really confused..what was output of mtd / partitions ? Aug 03 20:57:26 http://paste.debian.net/181980/ Aug 03 20:57:34 I'm trying ^^ Aug 03 21:00:17 why not set it to 130 as it want? Aug 03 21:04:54 well, maybe, this is the theorical max size Aug 03 21:05:38 but collie has exactly 14.125 MiB to use (exactly less, the last 128KiB are Angel stuff so padding is exactly at 14MiB) Aug 03 21:06:02 heh.. I've used exactly 3 times to say I'm in doubt :) Aug 03 21:06:19 :) Aug 03 21:06:31 try to create it by hands Aug 03 21:06:34 collie was like that Aug 03 21:07:18 http://www.linuxtogo.org/gowiki/ZaurusCollie Aug 03 21:10:10 http://linux.derkeiler.com/Mailing-Lists/Kernel/2005-11/3642.html Aug 03 21:11:06 +http://linux.derkeiler.com/Mailing-Lists/Kernel/2005-11/4927.html Aug 03 21:11:53 mtd->erasesize = 0x10000 * width / 2; Aug 03 21:11:57 width is 4 Aug 03 21:12:09 so erasesaze remains 0x20000 Aug 03 21:12:48 as confirmed here http://linux.derkeiler.com/Mailing-Lists/Kernel/2005-11/3649.html Aug 03 21:13:10 now..what about subpages? Aug 03 21:17:01 Jay7: guess what? Aug 03 21:17:14 ? Aug 03 21:17:22 image-size exceeds 14.125MiB :p Aug 03 21:17:32 heh Aug 03 21:17:41 what about compression? Aug 03 21:17:52 now is zlib Aug 03 21:18:12 i.e. it exceeds limit even with compression/ Aug 03 21:18:39 core-image-base-collie-20120803204447.rootfs.ubifs β”‚ 16380K Aug 03 21:18:53 seems need some tweaks.. Aug 03 21:19:22 I try now core-image-minimal but this one was 40MiB Aug 03 21:19:47 I think I've read is fixed by now Aug 03 21:21:56 β”‚ core-image-minimal-collie-20120803212055.rootfs.ubifs β”‚6967296β”‚ Aug 03 21:22:07 yes, is fixed Aug 03 21:22:25 ok, I'll commit like it is Aug 03 21:22:35 bluelightning: ping Aug 03 21:23:19 bluelightning: shouldn't ROOT_FLASH_SIZE = "14" in collie.conf avoid the creation of oversized images? Aug 03 21:23:52 like for kernel's do_sizecheck Aug 03 21:28:57 * Jay7 -> sleep Aug 03 21:29:58 hackaton started about 6 UTC tomorrow Aug 03 21:30:25 about 2-3 hrs for others to choose project Aug 03 21:30:37 so, I'll start about 8-9 UTC :) Aug 03 21:30:49 I'll be around Aug 03 21:30:55 nice Aug 03 21:34:14 ant____: that sounds reasonable Aug 03 21:34:34 bluelightning: collie fix sent, pls commit and test if possible Aug 03 21:34:48 I have the prebuilt image if you need Aug 03 21:36:47 bluelightning: moreover, we hint ROOT_FLASH_SIZE can be adjusted in local.conf for e.g. corgi and spitz Aug 03 21:36:59 now, we need a weak assignment in machine.conf then Aug 03 21:51:08 http://www.crocko.com/C1A6D112C9B5495DA447E4194A8BA971/core-image-minimal-collie.ubifs Aug 03 21:52:02 http://www.crocko.com/FD347B1E492342E49B2BDB04B5B35525/core-image-minimal-collie.sum.jffs2 Aug 03 21:52:48 http://www.crocko.com/4F4CFF62BF7D4E9FA9E1D007D4850850/core-image-minimal-collie.tar.gz Aug 03 21:53:12 reboot the router to change ip if the site asks you to wait too long :p **** ENDING LOGGING AT Sat Aug 04 02:59:58 2012 **** BEGIN LOGGING AT Sat Aug 04 02:59:58 2012 Aug 04 17:39:16 ant____: ping Aug 04 17:39:29 * Jay7 gave up on jabber-bot and switched to kexecboot Aug 04 17:39:46 hi Jay7 Aug 04 17:39:58 I've remembered I've found bug with redraw last time :) Aug 04 17:40:13 yes, mostlly impacting spitz here Aug 04 17:40:19 anyway, need to merge your code at first Aug 04 17:41:13 have you flashed akita? Aug 04 17:43:38 no Aug 04 17:43:41 will do soon Aug 04 17:45:39 you said redraw is done 6-7 times iirc Aug 04 17:55:35 btw, if (p[0] == 0x55 && p[1] == 0x42 && p[2] == 0x49 && p[3] == 0x23) { Aug 04 17:55:49 may be rewritten as if (p[0] == 'U' .... Aug 04 18:04:07 * bluelightning is doing a collie build finally Aug 04 18:04:26 unfortunately my home build machine died this week, fortunately I have another machine to use... Aug 04 18:04:36 ouch Aug 04 18:05:06 it did give me the impetus to upgrade however... I've ordered new i7 + bits :D Aug 04 18:07:11 well, merged first 5 patches as is :) Aug 04 18:07:24 last patch need some review before Aug 04 18:09:46 ah, about that, I've seen the ubi rootfs produced by oe-core have as well _0 hardcoded Aug 04 18:10:00 maybe put a config knob... Aug 04 18:12:13 http://cgit.openembedded.org/openembedded-core/tree/meta/classes/image_types.bbclass#n197 Aug 04 18:12:25 echo vol_id=0 >> ubinize.cfg Aug 04 18:20:54 for the rest, pls check twice sizeof and memory allocation Aug 04 18:22:45 trying to review Aug 04 18:23:10 * ant____ eating, brb Aug 04 18:26:32 btw. iirc, we decided to remove fbcon parameter passing to next kernel Aug 04 18:51:19 well, we add it to cmdline in kexecboot.cfg Aug 04 18:51:31 don't know where it is best Aug 04 18:51:39 place for it Aug 04 18:53:04 ant____: about ubifs Aug 04 18:53:26 can you remind me how should look cmdline with ubifs params? Aug 04 18:54:27 Kernel command line: root=ubi0_0 ubi.mtd=3,512 rootfstype=ubifs rootwait mtdparts=sharpsl-nand:7168k(smf),22528k(root),-(home) fbcon=rotate:1 console=ttyS0,115200n8 console=tty1 fbcon=rotate:1 quiet Aug 04 18:54:40 well, ok Aug 04 19:58:02 ant____: I'm getting an error at the end of do_rootfs: Aug 04 19:58:03 Error: max_leb_cnt too low (208 needed) Aug 04 19:58:54 this means the image size is too big Aug 04 19:59:13 collie has just 14 MiB available Aug 04 19:59:43 i.e. 113 eraseblocks x 128 Aug 04 19:59:58 ok... hmm Aug 04 20:00:11 even opie-image has obviously got too big :/ Aug 04 20:00:31 now, I'm wondering about compressing it more Aug 04 20:01:34 fwiw I had one advice but I think it was a joke Aug 04 20:01:36 run it through gzip, blob it on there anyway Aug 04 20:01:40 from #mtd Aug 04 20:02:41 unfortunately this error blocks buildhistory image data collection :/ Aug 04 20:02:51 guess I need to hack it temporarily Aug 04 20:04:09 for kernels, decompressed by bootloader, we could just gzip it but I can't see how to mount a compressed rootfs w/ouit loopback Aug 04 20:05:25 i.e. uImage.gz can be booted but afaik no way for rootfs.ubifs.gz Aug 04 20:06:48 isn't ubifs compressed anyway? Aug 04 20:07:12 sure, like jffs2 there are knobs to favor lzo or zlib Aug 04 20:07:39 right, so gzipping probably isn't going to help much, so the guy was probably joking as you say Aug 04 20:09:33 hmm Aug 04 20:09:34 http://git.uclibc.org/buildroot/commit/?id=b71d24d032e258e9343108e77fde06c6f5187113 Aug 04 20:12:33 ant____: http://pastebin.com/wpNTqia7 Aug 04 20:12:48 some unfortunately large items in there but I don't see anything obvious to cut out... Aug 04 20:14:57 heh... pciutils-ids/usbutils-ids... those could probably be removed Aug 04 20:19:19 ant____: are current kernels on download page latest with your patches? Aug 04 20:19:34 yes Aug 04 20:20:00 ok Aug 04 20:20:08 going to flash :) Aug 04 20:21:37 bluelightning: maybe compresssing the kernel with xz/lzma Aug 04 20:21:51 could give 3-400 KiB back Aug 04 20:21:51 surely we already do that as well? Aug 04 20:22:01 iirc only kexecboot Aug 04 20:22:14 hmm... yes, that might help Aug 04 20:22:42 CONFIG_KERNEL_GZIP=y Aug 04 20:22:42 # CONFIG_KERNEL_LZMA is not set Aug 04 20:23:01 depends on what the downside is... slower boot? more RAM usage? Aug 04 20:23:15 should just be slower to decompress once Aug 04 20:23:43 then is kept in ram Aug 04 20:24:22 right, and the kernel image is in RAM normally so that's no different I guess Aug 04 20:24:42 exactly, it's not XIP Aug 04 20:25:08 well, it can't hurt to try it... Aug 04 20:25:30 hmm, why is opie-sysinfo so big? device icons perhaps? Aug 04 20:28:16 yep, that's the cause :/ Aug 04 20:28:50 and btw 1288 KiB ttf-dejavu-sans Aug 04 20:28:59 liberation? other free? Aug 04 20:29:09 yeah that is rather large :/ Aug 04 20:29:20 +600 KiB ttf-dejavu-sans-mono Aug 04 20:35:34 hmm, liberation is still large but maybe smaller than those two combined Aug 04 20:35:49 although the liberation-fonts recipe in OE-Core wants to install fontconfig-utils :/ Aug 04 20:36:42 I'm rebuilding collie lzma kernel Aug 04 20:40:48 ant____: kernel splash on your core-image looks like qvga scaled to vga Aug 04 20:40:58 but it switched to psplash quickly Aug 04 20:41:44 ok, so akita still boots Aug 04 20:42:28 cool, I have working console on akita Aug 04 20:42:40 can't remember when it was working ok last time :) Aug 04 20:42:51 now even keymap is ok :) Aug 04 20:42:58 bluelightning: zImage-3.2.18+git1+486f7aec824b4127e91ef53228823e996b36~e8469eed93c23e9675a8eeb5-r4.0-collie-20120804203425.binβ”‚1543224β”‚Aug 4 22:42 Aug 04 20:43:11 almost 500KiB Aug 04 20:43:27 ok, that's a pretty decent reduction Aug 04 20:44:09 Jay7: those core-image-base are pretty rich of features Aug 04 20:45:27 hmm Aug 04 20:46:03 wow Aug 04 20:46:16 it even detects my non-standart NAND partitioning :) Aug 04 20:46:38 how that? Aug 04 20:46:40 right mtdparts passed to kexec'ed kernel Aug 04 20:46:49 ah, booting from SD/CF Aug 04 20:46:59 100Mb root + 20Mb home Aug 04 20:47:22 now try to flash jffs2 or ubifs :/ Aug 04 20:47:53 I'm sure it fails because of non-std partitioning Aug 04 20:48:00 I need revert it back Aug 04 20:48:09 I'll work on that today and tomorrow.. Aug 04 20:48:22 we have no fs_resize in this image :( Aug 04 20:48:23 I've written to stanislav and lumag Aug 04 20:48:30 cool Aug 04 20:49:10 about fsresize, you could just copy or extend nandlogical.bb Aug 04 20:50:04 I'm at work Aug 04 20:50:09 have no build env here Aug 04 20:50:17 but I can build at home and fetch.. Aug 04 20:50:30 or write bytes directly Aug 04 20:51:20 or change mtdparts from boot.cfg Aug 04 20:53:30 it's easier I send you a kernel to download Aug 04 20:53:48 what one? w/o zaurus support? Aug 04 20:53:53 are you sure 100 exactly? Aug 04 20:54:16 ah, with same partitioning/ Aug 04 20:54:19 moment Aug 04 20:54:44 102400k(root) Aug 04 20:54:50 -(home) Aug 04 20:55:32 104857600 bytes Aug 04 20:55:50 22020096 bytes home Aug 04 20:56:37 http://lxr.free-electrons.com/source/arch/arm/mach-pxa/spitz.c#L800 Aug 04 20:57:14 set 100 there Aug 04 20:57:48 if this is rootfs size Aug 04 21:06:38 pls try this http://www.crocko.com/E8B3ED34F2414FE2BAF84C8BC81A5E5E/zImage.gz Aug 04 21:06:41 brb Aug 04 21:06:51 hm... resume seems broken Aug 04 21:06:54 or suspend Aug 04 21:07:16 ah, you'r doing too deep tests ;) Aug 04 21:08:20 it can't suspend seems Aug 04 21:08:35 returning to same state but keyboard is unusable Aug 04 21:08:53 may be serial is working Aug 04 21:09:35 bluelightning: http://www.crocko.com/EBF287BBA70D4025B5F18726F0B145CD/zImage-3.2.18+git1+486f7aec824b4127e91ef53228823e996b3696f0_1+7cc31a952f78b8f8e8469eed93c23e9675a8eeb5-r4.0-collie-20120804203425.bin.gz Aug 04 21:10:59 you can add both to kexecboot menu and compare boot time Aug 04 21:11:53 ant____: is it kexecboot kernel? Aug 04 21:12:02 that you send me Aug 04 21:16:01 ant____: your kernel found cacko in my nand! Aug 04 21:16:30 ah.. seems it was just boot.cfg :) Aug 04 21:16:34 it hangs Aug 04 21:19:24 ok, now I can try to flash ubifs Aug 04 21:26:05 ok, I'm back Aug 04 21:27:29 ant____: have you ubiformat cmd around? Aug 04 21:27:37 with all right options Aug 04 21:27:43 yes, link on desktop :) Aug 04 21:27:53 http://free-electrons.com/blog/creating-flashing-ubi-ubifs-images/ Aug 04 21:27:54 good idea :) Aug 04 21:28:09 I just change Aug 04 21:28:14 -ubimkvol /dev/ubi0 -N volume_name -s 64MiB Aug 04 21:28:25 +ubimkvol /dev/ubi0 -N volume_name -m Aug 04 21:28:44 I wish to write ubifs image Aug 04 21:28:50 w/o creating volume by hands Aug 04 21:29:02 in theory, ubiformat -f Aug 04 21:29:34 is .ubifs file suitable for writing by ubiformat? Aug 04 21:29:49 i.e. is it ubinized file? Aug 04 21:31:01 " - a tool to format MTD devices and flash UBI images"; Aug 04 21:31:49 iirc I had issues, then I did the 3 staps by hand, from shell Aug 04 21:32:10 ah Aug 04 21:32:13 I copied .ubifs on the SD card from where I booted Aug 04 21:32:25 file size is not multiple of eraseblock size Aug 04 21:32:37 eraseblock size is 128k Aug 04 21:33:31 MKUBIFS_ARGS_akita = "-m 2048 -e 129024 -c 968 -x zlib" Aug 04 21:33:31 UBINIZE_ARGS_akita = "-m 2048 -p 131072 -s 512" Aug 04 21:34:11 # 121 x 1024 : 128 = 968 PEBs Aug 04 21:39:03 Jay7: don't say 'wow, it is fast'...fastmap seems reducing over 10x Aug 04 21:39:14 :) Aug 04 21:39:19 the ubiattach Aug 04 21:39:36 is now in discussion upstream Aug 04 21:41:40 kexecboot is pretty fast with zeroed mtd2/3 :) Aug 04 21:46:13 ubiattach seems hang Aug 04 21:46:29 ubiattach /dev/ubi_ctrl -m 2 Aug 04 21:46:30 is the green led blinking on mtd use on akita? Aug 04 21:46:34 no Aug 04 21:47:02 no reaction on power even Aug 04 21:47:57 any issue on ubiformat? Aug 04 21:48:26 in desesperate case, flash_eraseall before ubiformat can help Aug 04 21:48:37 ubiformat fails with error about sizes mismatch Aug 04 21:48:56 I've done just flash_erase /dev/mtd[23] 0 0 Aug 04 21:49:42 try ubiformat -O 2048 (strange, though) Aug 04 21:52:15 it should say that kernel has different idea about io size an subpages and ask you to continue Aug 04 21:53:03 mtdinfo -a Aug 04 21:53:08 ubinfo -a Aug 04 21:53:10 can help Aug 04 21:58:18 but really, with CONFIG_MTD_NAND_VERIFY_WRITE disabled you don't need ofsets (-O) Aug 04 21:59:03 I know well ubiformat since I cbeated it to compile with klibc :) Aug 04 21:59:25 no luck Aug 04 21:59:59 what does mtdinfo -a say? Aug 04 22:00:18 file ... (size 17676288 bytes) is not multiple of eraseblock size (131072 bytes) Aug 04 22:00:45 ah, ubiformat ... -f Aug 04 22:02:24 # ubiformat /dev/mtd2 Aug 04 22:02:37 ubiattach -m 2 Aug 04 22:03:00 ubimkvol /dev/ubi0 -N rootfs -m Aug 04 22:03:19 ubiupdatevol /dev/ubi0_0 ./my.ubifs Aug 04 22:03:51 please try that Aug 04 22:04:21 yes, ubiformat /dev/mtd2 + ubiattach -m 2 works! Aug 04 22:04:40 I did not investigate further ;) Aug 04 22:06:14 btw, green led is not blinking when accessing nand Aug 04 22:06:28 anyway, ubiupdatevol is done Aug 04 22:06:32 same on spitz, hdd was routed instead Aug 04 22:06:49 but c7x0 and poodle do have it blinking on nand access Aug 04 22:07:10 seem akita needs kernel patch Aug 04 22:10:39 ok, flashed Aug 04 22:10:58 let's reboot and try to boot from ubifs Aug 04 22:11:53 he, it's booted Aug 04 22:11:58 :) Aug 04 22:12:11 ant____: you are good hacker ;) Aug 04 22:13:22 well.. I'll mainline last patch Aug 04 22:13:41 now, you have 100MiB and could even create 2 volumes Aug 04 22:13:56 I've created ubi1:home already :) Aug 04 22:14:01 ah Aug 04 22:14:06 you mean 2 vol on same mtd Aug 04 22:14:17 then add the missing code to boot from _1 :) Aug 04 22:14:23 exactly Aug 04 22:14:29 yes, good idea Aug 04 22:14:41 ubimkvol /dev/ubi0 -N volume_name -s 64MiB Aug 04 22:15:03 for the first Aug 04 22:15:10 i.e. Aug 04 22:16:14 ubi0_0 and ubi0_1 Aug 04 22:35:59 Jay7: check out _device_get_partiion_range Aug 04 22:36:02 http://lists.alioth.debian.org/pipermail/parted-devel/2010-March/003543.html Aug 04 22:36:48 dunno id mtd appears in /sys/block/DEV/range Aug 04 22:36:56 s/if/id/ Aug 04 22:37:18 er... you understand Aug 04 22:39:23 iirc mtd char and ntdblock deviced are mapped with the same id Aug 04 22:39:36 mtd2 -> mtdblock2 Aug 04 22:44:51 no, doesn't work Aug 04 22:45:11 I have /sys/block/sda/range -> 16 Aug 04 22:45:52 I have /sys/block/mtdblock[0,1,2,3]/range -> 1 Aug 04 22:46:43 forget, I have checked in the kernel docs for the range Aug 04 23:24:41 https://github.com/andrea-adami/kexecboot/commit/6ed370cedbf89122c020530f12504415c864b200#L0R432 Aug 04 23:24:50 if(isdigit(atoi(dev.device+strlen(dev.device)-2))) { Aug 04 23:25:04 isdigit() is for char Aug 04 23:25:27 but here integer is passed from atoi() Aug 05 00:16:19 well.. I've made some fixes in 5th patch Aug 05 00:16:36 but latest (6th) is still need more love Aug 05 00:16:46 I'll push first 5 Aug 05 00:17:53 pushed Aug 05 00:21:15 I'm going to have some sleep Aug 05 00:21:34 hope will resume tomorrow **** ENDING LOGGING AT Sun Aug 05 02:59:59 2012 **** BEGIN LOGGING AT Sun Aug 05 02:59:59 2012 Aug 05 04:23:24 morning Aug 05 07:15:42 morning Jay7 Aug 05 07:16:21 pls rewrite that function to get mtd id with 2 digits. It's used twice Aug 05 07:37:38 s/that function/that as a function/ Aug 05 07:44:27 I'll Aug 05 07:48:42 now I have to cut the grass (and feed mosquitos), then I'll attack that ioctl Aug 05 07:48:57 I'll have some sleep again :) Aug 05 07:49:30 yea, see you later around Aug 05 10:22:07 hi ant____ Aug 05 10:22:31 hello Aug 05 10:23:03 I had a go building opie-image-16mb, fixed one minor issue; however it's still too large Aug 05 10:23:11 grr Aug 05 10:23:16 max_leb_cnt too low (128 needed) Aug 05 10:23:46 I'm not familiar with these numbers; how much more do we have to cut out before it will fit? Aug 05 10:24:36 the available size is 14.125 MiB but we don't use the last 128KiB so you have exactly 14MiB Aug 05 10:24:57 that's compressed though right? Aug 05 10:25:09 that's the size of the image to flash Aug 05 10:25:41 the image itself is compressed, yes Aug 05 10:26:07 you don't 'untar' it Aug 05 10:26:12 no, of course Aug 05 10:26:24 it's all in ram, on the fly Aug 05 10:26:27 so, the jffs2 image is 15MB, does that mean we're kind of close? Aug 05 10:26:33 1MiB Aug 05 10:26:46 right, so it could be possible, that's good :) Aug 05 10:26:48 have you used the lzma-kernel? Aug 05 10:26:59 no, not yet Aug 05 10:27:10 that's half of that Aug 05 10:27:15 actually what I really ought to do is use core-image-minimal, just verify the thing boots first Aug 05 10:27:31 sure Aug 05 10:27:58 but I don't see how using lzma instead of gzip would break the kernel Aug 05 10:28:13 (now that compressors are stable :) Aug 05 10:28:44 it shouldn't, no Aug 05 10:28:53 we need to take timings, I feel on small kernels the difference is negligible Aug 05 10:32:11 bbl Aug 05 11:28:54 ant____: I downloaded the prebuilt kexecboot image from the website, unfortunately it has the same glitchy display as it did previously :( Aug 05 12:00:27 lzma may break boot if compression is > 7e Aug 05 12:00:34 bluelightning: what exactly? Aug 05 13:29:42 irc we used textui on collie Aug 05 13:43:45 it starts with psplash Aug 05 13:43:59 I can take a picture if you like but I'm not sure it'll be much help... Aug 05 13:45:03 the kernel logo is the same as psplash. Do you see both correctly on qvga? Aug 05 13:45:14 (I do on poodle) Aug 05 13:53:58 ok, I could be confusing the two then Aug 05 13:54:15 all I see is corrupted version of OE splash screen Aug 05 13:54:37 the colours are OK, it looks like maybe it has the resolution incorrect Aug 05 14:19:52 I've just filmed boot on poodle with *phone Aug 05 14:20:57 video is rotated but you see what to expect... Aug 05 14:21:30 http://www.crocko.com/F459CE10C9F64F8BA037F9FBDBA2ABF0/IMG_0202.MOV Aug 05 14:34:16 I can't download that file :/ Aug 05 14:34:24 claims there's already a download from my IP Aug 05 14:38:42 mom Aug 05 14:40:50 try that Aug 05 14:44:06 ant____: ok got it... looks very smooth Aug 05 14:44:14 unfortunately doesn't work that way on collie here :( Aug 05 14:44:23 :/ Aug 05 14:45:03 I see the garbled splash screen, then after a short time it is replaced by an almost completely grey screen with a dark blue band towards the top Aug 05 14:45:19 it does not change in response to keypresses Aug 05 14:45:41 so I guess it is booting, it's just a display issue (+ probably input, but we can deal with that later) Aug 05 14:46:19 the issue is, we fall back to text-ui if fb is absent but not if it's broken/misconfigured Aug 05 14:46:40 right Aug 05 14:46:59 could you make a version with forced text mode? Aug 05 14:47:03 sure Aug 05 14:47:24 ok, maybe worth trying Aug 05 14:51:17 bluelightning: I've same trouble on collie Aug 05 14:51:25 strange looking OE splash Aug 05 14:51:34 I'll do now Aug 05 14:51:35 EXTRA_OECONF_collie = "--enable-textui=yes --enable-fbui=no --enable-delay=2 --enable-evdev-rate=1000,250" Aug 05 14:52:08 and yes, later it fade to light gray Aug 05 14:52:48 ant____: I suggest to make text ui default for collie at least for now Aug 05 14:53:06 I'll replace it Aug 05 14:53:24 and announce this on site please Aug 05 14:53:56 I'm pretty sure lumag has something ready for collie Aug 05 14:54:01 had... Aug 05 14:54:37 I really don't know the machine much, it's different Aug 05 14:54:48 maybe is just th ekernel config Aug 05 14:55:04 I'll spend some time on it.. during next hackaton ;) Aug 05 14:56:21 well.. I should merge last ubifs patch.. Aug 05 15:01:03 bluelightning: http://kexecboot.org/sites/kexecboot.org/files/kexecboot/installkit-collie.tar.gz Aug 05 15:02:02 bluelightning: you have a C3200 isn't? Aug 05 15:02:13 only JAma has C3000 iirc. Aug 05 15:03:10 ah, btw, I need to check tosa Aug 05 15:04:43 I have a c3000 and c3100 (latter borrowed from work) Aug 05 15:06:05 great, pls test both kernels on 3000 then :) Aug 05 15:06:44 http://lxr.free-electrons.com/source/arch/arm/mach-pxa/spitz.c#L800 Aug 05 15:07:05 both should create a mtd of 5MiB Aug 05 15:07:08 mtd2 Aug 05 15:08:37 Jay7: about the repartitioning... now it is too warm...I'll take a bath in the see for the moment Aug 05 15:08:58 here is too warm too Aug 05 15:09:02 see you later on, let me know about collie Aug 05 15:09:02 +30 Aug 05 15:09:13 34 (optimistic) Aug 05 15:09:14 I'll check it later today Aug 05 15:10:03 hang on, do I have to repartition in order to use this on spitz? Aug 05 15:10:21 no, not if it is standard Aug 05 15:10:23 afaik, no Aug 05 15:10:55 the issue is kexecboot-kernel just takes the defaults from spitz.c Aug 05 15:11:30 we are trying to make it correctly-redetect repartitioned mtd Aug 05 15:12:42 we are able to read the mtd values from userland but the kernel just ignores that and nobody has written a driver :) Aug 05 15:13:32 so we'll have kexecboot raw-read th emtdparts and then provide a runtime-map for the kernel with the behalf of BLKPG ioctrl Aug 05 15:13:50 (and much work, beers and a dose of luck) Aug 05 15:13:57 heh ok Aug 05 15:14:07 well, it all seemed to work previously at least Aug 05 15:15:21 no, kexecboot from mtd is only possible on C3000 and C3100, you need a modified kernel for 3200 Aug 05 15:15:32 same for C860 & co. Aug 05 15:16:17 in practice mtd was ignored by 'spitz' owners Aug 05 15:19:57 ah right, ok Aug 05 15:20:56 btw C869 and C760 have both 53MiB mtd2, I'll fix the comments in kexecboot Aug 05 15:21:10 what a mess, Sharp! Aug 05 15:30:28 ok, going now, pls test that text-only kernel for collie Aug 05 15:30:29 bye Aug 05 15:45:00 ant____: will do, cya Aug 05 15:53:34 nope, same boot logo and then the screen just fades to white very slowly Aug 05 16:21:41 hm.. that's bad Aug 05 16:22:01 seems time to find serial cable :) Aug 05 16:30:06 hmm, I need to try to get hold of one of those Aug 05 16:32:55 * Jay7 have cable Aug 05 16:33:55 just should remember where it is :) Aug 05 19:01:00 ah, bad news then Aug 05 19:01:19 * ant____ is doublechecking the upload Aug 05 19:08:18 /* #undef USE_FBMENU */ Aug 05 19:08:30 #define USE_TEXTUI 1 Aug 05 19:08:35 seems correct Aug 05 19:14:02 Jay7: the screenshot shows 'ubifs' detected but now it reads 'ubi' Aug 05 19:14:41 it subject to change Aug 05 19:15:13 even, it's improper to show /dev/mtdblock Aug 05 19:15:18 sure Aug 05 19:15:24 should be /dev/ubi Aug 05 19:15:49 ant____: btw, about collie Aug 05 19:16:07 compile please image with text ui and debug Aug 05 19:16:19 ok Aug 05 19:16:22 I'll check it with serial cable Aug 05 19:16:38 you need printk I guess... Aug 05 19:16:45 and a verbose kernel Aug 05 19:16:46 yes Aug 05 19:16:52 ok Aug 05 19:16:53 and --enable-debug in kexecboot Aug 05 19:17:03 can't remember about master console Aug 05 19:17:16 should it be changed to ttyS? Aug 05 19:32:48 we have MDLINE_collie = "console=ttySA0,115200n8 console=tty1 mem=64M fbcon=rotate:1" Aug 05 19:36:07 Jay7: if I enable BUG I have to disable some filesystem... Aug 05 19:36:44 do you mean in kernel? Aug 05 19:36:59 yes, I fear we don't have the extra 100Kb Aug 05 19:37:21 disable anything except jffs and ext3 Aug 05 19:37:24 and fat Aug 05 19:37:54 if we have others :) Aug 05 19:38:45 ext4 and unionfs Aug 05 19:39:03 ubifs ? Aug 05 19:39:27 let disable it atm Aug 05 19:39:54 it's ok to disable it now for collie Aug 05 19:43:29 compiling Aug 05 19:45:26 hm.. Aug 05 19:45:31 tosa doesn't starting up Aug 05 19:45:36 D+M menu is ok Aug 05 19:45:45 but no reaction on power/power+ok Aug 05 19:46:43 wait, battery is exhausted Aug 05 19:47:30 hm.. it is on external power Aug 05 19:47:38 but seems you are right Aug 05 19:47:48 aha, got it Aug 05 19:48:49 that sharp logo is simple and pretty Aug 05 19:49:05 every time I see it I like it :) Aug 05 19:49:17 is copyrighted :/ Aug 05 19:49:34 I know.. Aug 05 19:49:40 even, I think we consumers have the right to attach it Aug 05 19:49:43 we need something alike for kexecboot kernels :) Aug 05 19:50:01 I have the apple sticker on my hackintosh, was in Snow Leo Aug 05 19:50:09 I bought it... Aug 05 19:50:23 tosa have wrong keymap still Aug 05 19:50:32 :/ Aug 05 19:51:32 it's very cool we are operable even with 2 working buttons :) Aug 05 19:51:45 how is fb? Aug 05 19:51:45 left is down and ok is enter :) Aug 05 19:51:52 it's ok but dark Aug 05 19:52:04 maybe backlight is disabled? Aug 05 19:52:12 or at very low level Aug 05 19:53:45 hm.. Aug 05 19:53:59 looking in log on tosa Aug 05 19:54:10 + FS type 'vfat' detected Aug 05 19:54:26 + can't mount /dev/sda1: Invalid argument Aug 05 19:54:54 ah, I've never seriously tried boot from vfat :) Aug 05 19:54:55 it's CF with installkit Aug 05 19:55:09 ah, may be we have no vfat support there Aug 05 19:55:17 irc I had just kernel on fat Aug 05 19:55:22 no.. then it will say about this Aug 05 19:55:41 because seems vfat is in /proc/filesystems Aug 05 19:55:53 yes, fat/vfat are compiled Aug 05 20:01:15 oops..sorry guys, the collie debug kernel is silent...loglevel=3 Aug 05 20:01:21 I'll redo Aug 05 20:01:23 :) Aug 05 20:03:59 hm.. tosa can't boot with power inserted Aug 05 20:04:05 w/o power is ok Aug 05 20:04:39 iirc spitz had similar issues Aug 05 20:05:07 note that kexecboot kernel has very limited power management, no voltage regulators, freq, ... Aug 05 20:05:28 keymap is completely unusable Aug 05 20:06:03 remember tosa is vanilla Aug 05 20:08:34 Jay7: bluelightning pls redownload Aug 05 20:09:00 some i2c errors after resume Aug 05 20:09:31 timeout waiting for bus free Aug 05 20:09:46 well.. seems time to write something to lumag Aug 05 20:18:05 we could move to 3.4 right now, I don't think the situation could be worse ;) Aug 05 20:18:24 we'd get support for >3.4, if any Aug 05 20:18:46 [GIT PULL][for 3.7 v3] pull request from dt_mmp branch in pxa git tree Aug 05 20:18:51 ^^ seen today... Aug 05 20:19:22 ant____: what speed is set on serial? Aug 05 20:19:30 ah, 115200, I see Aug 05 20:20:23 Kernel panic - not syncing: Attempted to kill init! Aug 05 20:21:04 http://pastebin.com/dAY21NUh Aug 05 20:21:50 Warning: unable to open an initial console. Aug 05 20:21:52 is it ok/ Aug 05 20:22:05 I don't see the mtd Aug 05 20:22:31 console=tty1 console=ttySA0,115200n8 fbcon=rotate:1 mem=64M debug Aug 05 20:22:40 no mtdparts in default cmdline Aug 05 20:23:23 compare with this Aug 05 20:23:24 http://paste.debian.net/182259/ Aug 05 20:25:29 something wrong in kernel config? Aug 05 20:26:55 seems the ATA stack or some driver Aug 05 20:26:57 http://pastebin.com/se3BMrmU Aug 05 20:27:23 2.6.38 ^ collie Aug 05 20:29:56 but there is no nand info as well Aug 05 20:31:06 iirc collie uses now CFI driver Aug 05 20:31:15 maybe it's not verbose Aug 05 20:35:37 ok, beer and rechecking of collie.c Aug 05 21:26:22 * Jay7 sleep Aug 05 21:33:37 remember to finish ubifs boot! Aug 05 21:35:12 maybe add the 'ubi_boot_volume_id' default=0 Aug 05 21:35:45 just one for the moment Aug 05 21:36:40 we could get full fsdetection in klibc any soon :) **** ENDING LOGGING AT Mon Aug 06 02:59:58 2012 **** BEGIN LOGGING AT Mon Aug 06 02:59:58 2012 Aug 06 08:36:07 ant_work: morning Aug 06 08:37:00 ant_work: FYI, with the debug image, the text is a bit hard to read because it's spread over different areas and obscured; but I'm certain it's reporting "not syncing; attempted to kill init" Aug 06 08:37:28 interestingly this comes up almost immediately after the logo appears, I would have expected a delay Aug 06 08:38:07 in the 2.6.38 logs at this point pcmcia /ATA/SCSI were loaded Aug 06 08:38:55 maybe during the transition to 3.2 new configure options have been added Aug 06 08:39:24 for sha1100 Aug 06 08:39:46 I'm pretty sure I did enable all locomo stuff I could find Aug 06 08:41:38 bluelightning: maybe you missed it http://pastebin.com/dAY21NUh Aug 06 08:42:15 ant_work: where was that from? Aug 06 08:42:20 from Jay7 Aug 06 08:42:24 ah, ok Aug 06 08:42:28 so it's confirmed at least Aug 06 08:42:32 this was 2.6.38 Aug 06 08:42:33 http://pastebin.com/se3BMrmU Aug 06 08:42:42 bluelightning: yes Aug 06 08:42:46 hi ppl Aug 06 08:43:28 seems kexecboot crashed very early Aug 06 08:43:37 no output Aug 06 08:43:56 may be something wrong with compiler options? Aug 06 08:44:33 possibly related? http://e2e.ti.com/support/embedded/linux/f/354/t/185635.aspx Aug 06 08:44:50 the user does not say whether the option on or off caused the issue... Aug 06 08:45:16 oh wait, there's a backtrace there, ignore... Aug 06 08:46:38 "Warning: unable to open an initial console." doesn't sound good... Aug 06 08:49:31 I'd try straight 3.4 Aug 06 08:49:43 unfortunately -tiny is at 3.2 Aug 06 09:17:16 well, maybe Aug 06 09:17:24 there is a lumag's patch Aug 06 09:18:13 we need some stuff of pxa2xx on collie Aug 06 09:18:24 http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto-tiny-kexecboot/collie/0001-pcmcia-pxa2xx_sharpsl-retain-ops-structure-on-collie.patch Aug 06 09:18:32 ^^ maybe I've disabled it for collie Aug 06 09:21:26 oh, and MFD Aug 06 09:21:28 include Aug 06 10:42:27 Jay7: are you sure "ubiattach_argv[4] = UBI_VID_HDR_OFFSET;" is not empty? Aug 06 10:43:34 isn't AC_DEFINE_UNQUOTED just autoconf macro? Aug 06 10:44:27 it is in #ifdef Aug 06 10:44:50 so UBI_VID_HDR_OFFSET is defined Aug 06 10:45:16 + const char ubiattach_path[] = UBIATTACH_PATH; Aug 06 10:45:35 ah, here too Aug 06 10:45:45 UBIATTACH_PATH should be defined by default Aug 06 10:46:03 anyway I'll wrap all this code by USE_UBIFS Aug 06 10:46:16 ant_work: any ideas about collie? Aug 06 10:46:57 I've quicly checked the collie-kexecboot.cfg and compared with the old collie.defconfig in kernel tree Aug 06 10:47:07 no evident errors... Aug 06 10:48:02 well.. do you know way how to say to gcc to gerenate very generic code Aug 06 10:48:04 just a doubt about lumag's pcmcia patch Aug 06 10:48:05 ? Aug 06 10:48:22 we can start disabling optimize for size maybe Aug 06 10:48:23 i.e. to disable all extensions/optimizations/etc Aug 06 10:48:45 just to exclude gcc from possible sources Aug 06 10:48:51 I would not touch gcc, just kernel Aug 06 10:49:01 but yes, it's a possibility Aug 06 10:49:21 I fear it may do something wrong for strongarm Aug 06 10:49:39 the kernel would not boot I suppose Aug 06 10:49:51 at all Aug 06 10:50:01 hm.. may be, if it doesn't override gcc flags Aug 06 10:50:37 but easiest source of problem is some bug in kexecboot code :) Aug 06 10:50:51 somewhere near start Aug 06 10:51:07 you had CF inserted? Aug 06 10:51:21 when you logged? Aug 06 10:51:25 I've tried w/o any cards and power Aug 06 10:51:48 may be only with SD Aug 06 10:51:54 we have few patches, one is keymap and the other is for pcmcia Aug 06 10:51:56 can't remember exactly Aug 06 10:52:26 well.. seems I should do some builds with full debug Aug 06 10:52:29 maybe the infrastructure change and the patch breaks things, calling th epxa2xx driver Aug 06 10:52:56 we can remove this patch temporarily Aug 06 10:53:42 it sems like you had no block devices Aug 06 11:01:19 here some patches for 2.6.31 http://www.linuxtogo.org/gowiki/CollieEnhancements?action=AttachFile&do=get&target=kernelreplacement-2.6.31.12-r10-rep3.tgz Aug 06 11:03:18 iirc after talking with lumag we only keptt one patch Aug 06 11:04:41 definitely need to check if patched pxa2xx_sharpsl.c is built Aug 06 11:05:13 was this thread Aug 06 11:05:14 http://lists.linuxtogo.org/pipermail/angstrom-distro-users/2011-June/003682.html Aug 06 12:01:50 it shouldn't fail silently if we have no block devices imho Aug 06 12:06:34 the guy there says w/out patch you get kernel panic Aug 06 12:07:03 CompactFlash cards usable without kernel panics Aug 06 12:22:20 patch has not nereb committed Aug 06 12:22:30 http://lists.infradead.org/pipermail/linux-pcmcia/2011-August/007847.html Aug 06 12:22:36 ? Aug 06 12:30:26 bluelightning: taken apart collie quirks, I'd say next urgent thing o fix are the modules, e.g. alsa Aug 06 12:30:48 iirc rtc is loaded by udev but maybe I'm wrong Aug 06 12:31:04 assuming we can get a successful boot with input working, sure... Aug 06 12:31:21 pls try on one of your spitz Aug 06 12:31:25 3200 is ok Aug 06 12:32:35 tosa seems in better shape than collie :) Aug 06 12:32:54 that's indeed strange! Aug 06 12:38:40 Jay7: so it needs just keymap Aug 06 12:42:18 ant_work: it's less broken I'd say :) Aug 06 12:42:28 yes, keymap is must have Aug 06 12:43:05 what happens once you load it in shell? Aug 06 12:43:15 I don't know :) Aug 06 12:43:25 need to go through serial console Aug 06 12:43:50 do we have old'n'working keymaps for tosa? Aug 06 12:44:07 well.. known to work with older kernels in 2.6? Aug 06 13:12:47 yes, we have in keymaps package Aug 06 13:13:00 but we still lack the initial for initramfs Aug 06 13:48:43 image you have sent to me have wrong keymap too Aug 06 13:48:58 akita's one was ok Aug 06 13:59:26 akita,spitz,corgi are fixed in kernel Aug 06 13:59:36 colie + poodle have a patch Aug 06 13:59:48 tosa *had* probably 2.6 patch Aug 06 14:00:10 just need to find the latest version Aug 06 14:00:45 all should have proper keymaps in shell, though Aug 06 19:53:17 Jay7: I can't see obvious missing features in colle .config of linux-y-t-kexecboot Aug 06 19:53:43 urgh... Aug 06 19:53:47 Value requested for CONFIG_ARM_THUMB not in final ".config" Aug 06 19:53:53 what's that? Aug 06 19:56:18 ah, I see, yocto fragments Aug 06 19:56:22 # Begin: meta/cfg/kernel-cache/arch/arm/arm.cfg Aug 06 19:56:22 CONFIG_ARM_THUMB=y Aug 06 20:04:38 anyway, I see pxa2xx_sharpsl.o Aug 06 20:06:02 yes, seems all the modules I see in collie.c have been built Aug 06 20:23:23 ant____: well.. no easy fix then.. Aug 06 20:23:37 some deeper inverstigations needed Aug 06 20:32:03 'm fubious about the other patches for 2.6.31 above Aug 06 20:32:11 *dubious **** ENDING LOGGING AT Tue Aug 07 02:59:58 2012 **** BEGIN LOGGING AT Tue Aug 07 02:59:59 2012 Aug 07 08:18:14 bluelightning: I did a bit of research and in fact in the past we had Aug 07 08:18:16 "[yocto] Kernel Panics on armv4t with gcc.4.5.1" Aug 07 08:19:32 in the meanwhile, pls remember to merge the jffs2 + ubifs images fix for collie Aug 07 08:19:56 but yes, worth asking if someone is building armv4t Aug 07 08:26:34 I guess I had hoped to be able to confirm everything was OK before doing that, but since it seems it was broken before... Aug 07 08:28:20 ..now I found gcc-4.7.0-arm-unbreak-eabi-armv4t.patch Aug 07 08:28:27 ?!? Aug 07 08:29:10 We need to ask Khem and JaMa (om-gta02 is also armv4t) Aug 07 08:29:34 gta-01 fwiw Aug 07 08:32:08 ok, I don't see that patch in oe-core at first glimpse Aug 07 08:32:14 http://repository.timesys.com/buildsources/g/gcc/gcc-4.7.0/ Aug 07 08:32:52 we compile userspace EABI Aug 07 08:36:55 btw it seems linaro 12.05 release does "Includes soft float support for ARMv4T and later" Aug 07 08:37:34 Linaro GCC 4.7 2012.05 is the second release in the 4.7 series. Based off the latest GCC 4.7.0+svn187448 release Aug 07 08:37:34 Read more: http://www.cnx-software.com/2012/06/01/linaro-12-05-release-with-kernel-3-4-and-android-4-0-4/#ixzz22qcmgtjZ Aug 07 08:37:55 somebody has tested it :) Aug 07 08:39:35 don't we have older gcc versions in oe-core? Aug 07 08:57:16 I found that http://repository.timesys.com/buildsources/g/gcc/gcc-4.7.0/gcc-4.7.0-arm-unbreak-eabi-armv4t.patch Aug 07 08:58:29 (wrong windows :) Aug 07 09:08:08 bluelightning: worth to check whether gold is default linker Aug 07 09:08:57 * ant_work hopes we have not opened another can of worms... Aug 07 09:13:19 looking at gcc master Aug 07 09:13:25 http://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/config/arm/linux-eabi.h;h=46d3bc675fb4d00a9ce8875a13fd700da8edb874;hb=HEAD Aug 07 09:13:36 Default to armv5t so that thumb shared libraries work. Aug 07 09:13:38 ... Aug 07 09:14:06 well, the patch does still apply Aug 07 09:14:30 ant_work: AIUI gold is only enabled if ld-is-gold is in DISTRO_FEATURES Aug 07 09:14:36 which I don't think it is by default Aug 07 11:54:00 ok, let's log this second patch Aug 07 11:54:03 http://git.freesmartphone.org/?p=linux-2.6.git;a=commit;h=0333777365bef05b336d4f4a06bb3b37f4ada838 Aug 07 12:02:23 bbl Aug 07 12:58:22 [16:49] ant_work: yes, correct Aug 07 12:58:32 for logging purposes :) Aug 07 14:21:07 ok, I'l lclean and rebuild the toolchain for collie with that couple of extra patches Aug 07 14:33:51 core-image-minimal for collie started Aug 07 14:33:53 bbl Aug 07 19:05:32 Jay7: I'll upload collie kernel in a few mins Aug 07 19:38:41 kernel + image built from scratch applying http://repository.timesys.com/buildsources/g/gcc/gcc-4.7.0/gcc-4.7.0-arm-unbreak-eabi-armv4t.patch Aug 07 19:39:27 Kernel panic - not syncing: Attempted to kill init! Aug 07 19:40:05 http://paste.debian.net/182613/ Aug 07 19:40:40 this console issue is strange Aug 07 19:46:42 hm.. the armv4t binary should run on armv5te, isn't? Aug 07 19:46:53 le try it on other zaurus Aug 07 19:48:01 you may try it on poodle e.g. Aug 07 19:48:11 or c860 Aug 07 19:55:40 :) Aug 07 19:55:47 ok on poodle, text-ui only Aug 07 19:55:54 it's the kernel Aug 07 19:57:15 in fact, klibc does not enable thumb atm Aug 07 19:57:28 so kexecboot-klibc is not affected in any way Aug 07 19:59:16 try yourself on akita/spitz Aug 07 20:03:52 now I'll do a last try integrating JaMa's kernel patch Aug 07 20:10:26 (-mno-unaligned-access) Aug 07 20:18:05 you may try with -O0 -g Aug 07 20:18:18 to disable optimisation and enable debugging info Aug 07 20:20:12 interesting, issues on do_patch now :/ Aug 07 20:21:00 http://paste.debian.net/182624/ Aug 07 20:23:55 ok, fingers crossed Aug 07 20:29:26 no luck Aug 07 20:29:50 ok, needs really a full-debug kernel Aug 07 20:29:51 well, let's suspect problem in kexecboot or kernel then Aug 07 20:30:06 I'd say kexecboot is not guilty :) Aug 07 20:30:29 try yourself the binary on akita: it will show text only Aug 07 20:30:38 or on tosa fwiw Aug 07 20:33:04 I extracted the binary from the cpio, it's exactly the same as in collie kernel Aug 07 20:33:58 it's pretty useful to have image in akita's nand :) Aug 07 20:34:45 sound is broken and probably more but is a good emergency image Aug 07 20:35:16 got the kernel we'll polish it in few weeks Aug 07 20:35:30 it = the image, userland Aug 07 20:36:00 kexecboot.collie is working ok on akita Aug 07 20:43:49 ant__: is it hard to build kernel + static shell in initramfs instead of kexecboot? Aug 07 20:44:07 on collie, yes, only 1MiB Aug 07 20:47:30 we have to see with klibc utils and shell, though seems very limited Aug 07 20:47:49 I can try to add dash Aug 07 20:47:58 try please Aug 07 20:48:18 iirc, we had klibc shell some time ago Aug 07 20:48:23 around kexecboot-0.3 Aug 07 20:48:54 I can't remember shell Aug 07 20:49:11 only before rewriting in C Aug 07 20:49:17 yes, I mean that times Aug 07 20:49:51 I don't know what it was - dash/ash/busybox sh Aug 07 20:51:20 I can try some klibc utils Aug 07 20:51:36 sh is 85KiB Aug 07 20:51:56 do you need kinit? Aug 07 20:55:12 http://paste.debian.net/182629/ Aug 07 20:55:20 we have 150 KiB Aug 07 20:55:27 no, just start /bin/sh as init Aug 07 20:55:45 ok, I add it to cmdline Aug 07 20:56:22 -rw-r--r-- 1 andrea users 45K Aug 7 22:50 klibc-static-utils-sh_2.0-r0.0_armv4.ipk Aug 07 20:56:26 is it shell? Aug 07 20:57:41 yes Aug 07 20:57:56 it's dash iirc, I have to upgrade klibc btw Aug 07 20:58:20 [klibc] dash: sync with latest git Aug 07 20:58:25 y Aug 07 20:58:29 ok Aug 07 20:58:41 we don't have latest, yet Aug 07 20:58:45 np Aug 07 20:59:52 may be good idea to add ls and mount commands too Aug 07 21:01:10 +++ b/meta-initramfs/recipes-bsp/images/initramfs-kexecboot-klibc-image.bb Aug 07 21:01:10 @@ -1,5 +1,5 @@ Aug 07 21:01:10 require initramfs-kexecboot-image.bb Aug 07 21:01:10 Aug 07 21:01:10 -IMAGE_INSTALL = "kexec-klibc kexecboot-klibc ubiattach-klibc" Aug 07 21:01:11 +IMAGE_INSTALL = "kexec-klibc kexecboot-klibc ubiattach-klibc klibc-static-utils-sh klibc-static-utils-ls klibc-static-utils-mount" Aug 07 21:01:23 nice Aug 07 21:02:28 Jay7: btw, I pasted and copied this: Aug 07 21:03:12 http://paste.debian.net/182631/ Aug 07 21:03:39 you already had that insane idea... Aug 07 21:04:00 ah, yes, crunchgen Aug 07 21:04:26 I've used it when doing FreeBSD-based routers "firmware" :) Aug 07 21:04:55 ok, just add it to your to-do Aug 07 21:05:07 but I didn't know it is available on linux Aug 07 21:05:25 we were on #klibc Aug 07 21:06:01 well, ok Aug 07 21:06:07 any ETA on image with shell? Aug 07 21:06:40 mom Summary: There was 1 ERROR message shown, returning a non-zero exit code. Aug 07 21:06:55 * opkg_install_cmd: Cannot install package klibc-static-utils-sh. Aug 07 21:06:59 ? Aug 07 21:07:07 wtf? Aug 07 21:07:41 typo Aug 07 21:07:43 klibc-static-utils-sh_2.0- Aug 07 21:12:42 no, it's smthg different Aug 07 21:12:44 | * satisfy_dependencies_for: Cannot satisfy the following dependencies for klibc-static-utils-sh: Aug 07 21:12:44 | * klibc-static-utils * Aug 07 21:12:54 this seems a new bug ;) Aug 07 21:14:11 not long ago I did add losetup-static for a while then removed Aug 07 21:14:37 seems a packaging issue Aug 07 21:20:02 ah, ok, I let generate the empty klibc-utils-static package Aug 07 21:25:01 still battling Aug 07 21:25:04 now * opkg_install_cmd: Cannot install package klibc-static-utils-mount. Aug 07 21:25:38 well, I'll go to bed now then Aug 07 21:25:46 better ;) Aug 07 21:25:53 leave me link Aug 07 21:25:59 I'll test at morning Aug 07 21:26:12 ok, thx Aug 07 21:26:13 or send by email Aug 07 21:26:14 gn Aug 07 23:23:02 well, opkg is addingg Depends: klibc-static-utils on each static ipk... Aug 07 23:23:10 too bad **** ENDING LOGGING AT Wed Aug 08 02:59:58 2012 **** BEGIN LOGGING AT Wed Aug 08 02:59:58 2012 Aug 08 09:21:24 bluelightning: hi there. I may have discovered a latent bug in klibc recipes or maybe in opkg Aug 08 09:21:40 ant_work: hi Aug 08 09:21:47 ant_work: ah, ok, what have you found? Aug 08 09:22:34 inspecting the ipk's of klibc-static-utils-foo, the control file always has a line "Depends: klibc-static-utils" Aug 08 09:23:22 this is runtime depends Aug 08 09:23:36 but is a bogus for static utils Aug 08 09:24:03 so if I try to add one util to th eimage I get Aug 08 09:24:35 | * satisfy_dependencies_for: Cannot satisfy the following dependencies for klibc-static-utils-sh: | * klibc-static-utils * Aug 08 09:26:00 now, if you look at the recipes, I've added FILES_${PN} = "" to get rid of the empty klibc-static-utils.bb Aug 08 09:26:33 it seems something wrong with PACKAGES_DYNAMIC Aug 08 09:27:07 in that it adds the runtime dependency Aug 08 09:29:35 http://lists.linuxtogo.org/pipermail/openembedded-devel/2011-March/030726.html Aug 08 09:43:42 ant_work: PACKAGES_DYNAMIC won't be what's doing that... it just satisfies existing dependencies of other recipes, it doesn't create new ones Aug 08 09:53:14 looking at commit histtory Aug 08 09:53:16 http://cgit.openembedded.org/openembedded/commit/recipes/klibc?h=org.openembedded.dev&id=e16a751548cb72020bdfc33f8c2b03496dbd85e9 Aug 08 09:53:25 RDEPENDS cleanings for the (shared) utils Aug 08 09:59:52 I'm pretty sure I had a discussion before removing the explicit RDEPENDS Aug 08 10:01:20 with 1.5.26 Aug 08 10:02:11 ant_work: unrelated but I think THIS_KLIBC ought to be using EXTENDPKGV Aug 08 10:02:37 as the default RDEPENDS_${PN}-dev does Aug 08 10:03:23 the point is, the shared lib code should add rdepends automatically for klibc-shared-utils Aug 08 10:03:30 and it does Aug 08 10:03:51 but not for the 'static' flavour Aug 08 10:04:43 ant_work: you're sure it's the shared library dependencies code that's adding that dependency? Aug 08 10:04:58 so I have two options: build an empty bogus package and let rdepend on it OR fix the rdepends for each static util Aug 08 10:05:38 there is any RDEPENDS in the recipes Aug 08 10:05:59 in the latest incarnation Aug 08 10:08:09 I see in some cases there is the need to empty PACKAGES_DYNAMIC Aug 08 10:08:25 http://patches.openembedded.org/patch/1551/ Aug 08 10:08:52 ant_work: I suspect the culprit is do_split_packages Aug 08 10:09:12 ys, I've seen it accepts EXTRA dependencies but cannot empty that Aug 08 10:10:44 ant_work: looking at the code, it does an "if extra_depends == None:" check rather than just "if extra_depends:" so you should be able to set it to '' Aug 08 10:10:55 and that should clear it Aug 08 10:11:02 hm, wait, let me understand Aug 08 10:12:20 we really need to document that function :/ Aug 08 10:13:37 if extra_depends == None: Aug 08 10:13:38 extra_depends = d.getVar("PN", True Aug 08 10:13:51 it's what is happening, it adds PN Aug 08 10:14:09 as default Aug 08 10:14:27 right, but '' != None Aug 08 10:14:44 so if you specify extra_depends='' in the call to do_split_packages, it will skip that Aug 08 10:16:08 this was rather poorly implemented... the default should be '${PN}' and we should just expand the value of it... Aug 08 10:16:33 probably quite old code though Aug 08 10:17:12 ok, I'll try later modifying do_split_packages calls Aug 08 10:18:05 in fact, a number of recipes use extra_depends='' (e.g. connman, ncurses, ...) Aug 08 10:18:05 in the hope this will not break shared utils :) Aug 08 10:18:22 it will not affect shlibdeps Aug 08 10:18:52 but if there is no actual shared library dependency and you still want the runtime dependency anyway, that will be removed... Aug 08 10:19:00 not sure if that's going to be an issue here though Aug 08 10:19:51 I'll try building the initramfs images: klibc-utils-ls needs libklibc while klibc-static-utils-ls doesn't Aug 08 10:22:28 let's add a shell for the poor collie :) Aug 08 10:22:37 ant_work: if you're interested in monitoring the change, buildhistory can be useful for that Aug 08 10:23:09 I've asked for help several times: this recipe is a packaging nightmare :) Aug 08 10:24:57 (and it's not yet 100% being we build it as -native in fact...) Aug 08 10:25:17 no way to have an -sdk atm **** ENDING LOGGING AT Wed Aug 08 20:13:49 2012 **** BEGIN LOGGING AT Wed Aug 08 20:14:16 2012 Aug 08 20:19:53 bluelightning_: extra_depends='' did the job Aug 08 20:20:09 I'll send a patch for meta-initramfs Aug 08 20:20:18 Jay7: kernel coming Aug 08 20:26:10 static bins are: sh, mount, ls, kexec, ubiattach, kexecboot, init=/bin/sh Aug 08 20:29:14 same host like yesterday, file is called zImage **** ENDING LOGGING AT Wed Aug 08 20:41:31 2012 **** BEGIN LOGGING AT Wed Aug 08 20:42:22 2012 Aug 08 20:57:16 ant__: hi Aug 08 20:57:56 downloading zImage Aug 08 21:19:00 seems kexecboot site is out of reach :/ Aug 08 21:30:00 Jay7: btw, [15:14:08] NOTE: klibc uses 8bit transfers that breaks image on Aug 08 21:30:12 hpa> I think it is still true on non-x86 platforms... I don't have optimized memcpy for a lot of platforms Aug 08 21:30:31 it's this code https://github.com/kexecboot/kexecboot/blob/master/fb.c#L490 Aug 08 21:30:32 I don't understand why it would break anything though Aug 08 21:32:51 bluelightning_: unwanted side effect is, no depedencies in control file of shared utils :/ Aug 08 21:35:23 those do depend on libklibc Aug 08 21:38:15 ant__: so, the shared library dependency is missing? Aug 08 21:40:04 yes, not in the control file Aug 08 21:40:23 I can try to add a dummy var, '' in case of -static Aug 08 21:41:16 I'm still building the test image, just checking the ipkg's Aug 08 21:41:39 btw I've spotted a new bug in initramfs-kexecboot-image (not klibc) Aug 08 21:42:12 nothing provides 'ubiattach' for we only have the per-package-split in meta-initramfs Aug 08 21:42:33 we ought to add the full mtd.utils Aug 08 21:43:03 iirc there was a patch proposing to split things a bit, at least jffs2 and ubi Aug 08 21:43:39 I think I'll just replace it with ubiattach-klibc for the moment... Aug 08 21:49:20 ant__: I'm not sure I understand hpa's phrase Aug 08 21:49:55 I was asking if it's still actual Aug 08 21:50:00 ah Aug 08 21:50:09 i.e. we are doing this right? Aug 08 21:50:35 iirc thosa had yellow lines with slower transfers Aug 08 21:52:36 bluelightning_: no, the shared lib is not added to the image Aug 08 21:52:56 (as expected seeing the ipk's) Aug 08 21:53:31 oki, I'll circumvent the issue tomorrow Aug 08 21:53:47 Jay7: got the image? booting? Aug 08 21:55:14 ant__: got image, not tried yet Aug 08 21:55:51 bluelightning_: would you d/l it as well? Aug 08 22:06:05 gn, see you tomorrow Aug 08 22:08:49 nite Aug 08 23:11:51 ant__: which? **** ENDING LOGGING AT Thu Aug 09 02:59:59 2012 **** BEGIN LOGGING AT Thu Aug 09 02:59:59 2012 Aug 09 06:43:31 ant: http://paste.debian.net/182874/ Aug 09 06:43:47 seems /bin/sh is killed too Aug 09 06:46:58 but I'd prefer to do another test - with only /bin/sh placed as /sbin/init in initramfs Aug 09 06:49:06 anyway, I fear we have broken compiler Aug 09 06:49:11 or linker **** ENDING LOGGING AT Fri Aug 10 02:59:58 2012 **** BEGIN LOGGING AT Fri Aug 10 02:59:59 2012 Aug 10 10:11:38 bluelightning: about collie, khem sent me a patch but this makes do_compile fail :/ I'll talk with him late today Aug 10 10:12:34 bluelightning: in the meanwhile pls fix zaurus.inc wrt collie jffs2/ubifs. The erasesize for ubi is wrong right now... Aug 10 10:12:50 ant_work: will do Aug 10 10:13:00 ant_work: sorry have been quite busy lately Aug 10 10:13:08 np, same here :) **** ENDING LOGGING AT Sat Aug 11 02:59:58 2012 **** BEGIN LOGGING AT Sat Aug 11 02:59:58 2012 **** ENDING LOGGING AT Sun Aug 12 02:59:59 2012 **** BEGIN LOGGING AT Sun Aug 12 02:59:59 2012 Aug 12 11:42:50 bluelightning: khem was not around :/ collie-gcc still stuck Aug 12 11:43:09 it seems code for armv5t is generated Aug 12 11:43:45 that would explain kernel panic and the fact binary runs on armv5 Aug 12 11:44:04 I let that to khem... Aug 12 11:44:54 for you, pls remember http://patches.openembedded.org/patch/33765/ Aug 12 11:45:17 (sorry to bother again, now fixing meta-initramfs :) Aug 12 12:17:15 ant__: merged Aug 12 12:17:32 I just sent out a meta-handheld patch to avoid the procps dependency as well Aug 12 12:17:47 didn't have a chance to test it yet due to the collie issues Aug 12 14:23:12 bluelightning: thx Aug 12 14:23:27 now, about conditional expr with python.. Aug 12 14:23:56 sure, what's up? Aug 12 14:24:07 anything better than base_conditional for extending the splt_packages? Aug 12 14:24:10 i.e. Aug 12 14:24:12 extra_depends='' Aug 12 14:24:41 when ! KLIBC_UTILS_VARIANT == "shared" Aug 12 14:25:14 well this is a python function, you can put whatever code you need in it... Aug 12 14:26:05 I'm still not understanding why if it's built in a shared configuration, the system can't detect that the libraries are needed Aug 12 14:26:48 hm, yes Aug 12 14:29:43 see do_package Aug 12 14:30:27 http://paste.debian.net/183234/ Aug 12 14:32:25 DEBUG: LIBNAMES: pkg klibc-utils libs 0 bins 0 sonames [] Aug 12 14:32:32 is that correct? Aug 12 14:35:59 it's strange klibc was no built... Aug 12 14:36:12 I'm building explicitely Aug 12 14:50:05 the lib is in libklibc Aug 12 14:50:54 still I see Depends: klibc-utils Aug 12 14:51:02 whiuch is obviously wrong Aug 12 15:01:07 yes, therein lies the problem... Aug 12 15:08:44 now, some older version of the recipe did declare the RDEPENDS for each util Aug 12 15:09:11 I'm pretty sure that moving to packages dynamic was working, at the beginning... Aug 12 15:10:15 but I could not get easily rid of this dummy klibc-utils package Aug 12 15:10:55 PACKAGES_${PN} = "${PN}" Aug 12 15:10:55 FILES_${PN} = "" Aug 12 15:11:09 now, I'll retry with PACKAGES_${PN} = "" Aug 12 15:14:56 same, Depends: klibc-utils Aug 12 15:15:22 with the variant that now install will fail (broken dep) Aug 12 15:15:50 what about adding the rdep to klibc to the klibc-util package then? Aug 12 15:15:55 none of that influences shlibdeps Aug 12 15:20:14 hm..see this Aug 12 15:20:16 http://cgit.openembedded.org/openembedded/commit/recipes/klibc?h=org.openembedded.dev&id=e16a751548cb72020bdfc33f8c2b03496dbd85e9 Aug 12 15:27:14 trying now adding RDEPENDS_${PN} = "libklibc" in klibc-utils_2.0.bb Aug 12 15:27:32 first I want the shlibs back, then I'l lclean the -static Aug 12 15:36:56 bluelightning: note klibc outputs like that Aug 12 15:37:00 ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked, stripped Aug 12 15:37:18 ^static Aug 12 15:37:22 ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked (uses shared libs), stripped Aug 12 15:37:26 ^shared Aug 12 15:37:55 maybe statically confuses the shlibs code? Aug 12 15:38:04 not sure Aug 12 15:38:26 I'm a bit puzzled why you changed the packaging for debian naming, that's supposed to be handled internally Aug 12 15:38:48 yea, I'm pretty sure I talked with khem and pb about that :/ Aug 12 15:39:29 I think you'll find FILES_ will not do anything at all Aug 12 15:43:20 β”‚ libklibc-dev_2.0-r0.0_armv5te.ipk β”‚ 165356β”‚Aug 12 17:30β”‚β”‚ klibc.inc β”‚ 1574β”‚Jun 20 00:55β”‚ Aug 12 15:43:20 β”‚ libklibc-staticdev_2.0-r0.0_armv5te.ipk β”‚ 75386β”‚Aug 12 17:30β”‚β”‚ klibc_2.0.bb β”‚ 783β”‚Jun 20 00:55β”‚ Aug 12 15:43:20 β”‚ libklibc_2.0-r0.0_armv5te.ipk β”‚ 40462β”‚Aug 12 17:3 Aug 12 15:44:10 ^ Aug 12 15:44:12 *klibc--v28AKWgOZ-aPpboyW9RZ7V6kvQ.so β”‚ 76880 Aug 12 15:44:41 the file is there and is not empty Aug 12 15:45:46 file ./klibc--v28AKWgOZ-aPpboyW9RZ7V6kvQ.so Aug 12 15:45:46 ./klibc--v28AKWgOZ-aPpboyW9RZ7V6kvQ.so: ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked, stripped Aug 12 15:45:55 is ok Aug 12 15:47:23 see, I'll readd one-by-one the rdeps Aug 12 15:47:28 then that's likely to be a result of the default value of FILES Aug 12 16:07:31 heh, better now.. Aug 12 16:07:32 Depends: ${THIS_LIBKLIBC} Aug 12 16:07:39 almost ;) Aug 12 16:10:36 Depends: libklibc (= 2.0-r0.0) Aug 12 16:10:55 now the INC of this recipe :/ Aug 12 16:24:40 seems done, now trying to remove empty packages Aug 12 16:28:20 bluelightning: http://paste.debian.net/183243/ Aug 12 16:30:02 no way to get rid of -dev and -dbg :/ Aug 12 16:31:27 hm, no, still smthg wrong Aug 12 16:31:49 some recipes do correctly (r)depend on libklibc Aug 12 16:31:56 some on klibc-utils :/ Aug 12 16:32:08 some on anything (static?) Aug 12 16:32:34 I'm talking about *shared* klibc-utils Aug 12 16:33:30 the 'static' are fixed Aug 12 16:49:54 ok, I'll send the patch Aug 12 16:50:07 thing are much better Aug 12 18:46:18 morning Aug 12 18:50:03 Jay7: ping Aug 12 18:50:09 alexxy: hi Aug 12 18:50:19 does kexecboot website down? Aug 12 18:50:43 btw hx4700 now fully supported by 3.5 mainline kernel Aug 12 19:07:37 alexxy: ah, kexecboot.org ns is down Aug 12 19:07:49 will fix in next hours Aug 12 19:08:14 alexxy: cool news about hx4700 Aug 12 19:08:24 did you tried latest kexecboot on it? :) Aug 12 19:08:53 no Aug 12 19:09:00 i just wanna build it Aug 12 19:09:05 but site is down Aug 12 19:09:28 can you give me some instructions how to build it manualy Aug 12 19:16:40 kexecboot itself is easy to build Aug 12 19:16:46 check ./configure --help :) Aug 12 19:17:03 but you need kexec binary in initramfs Aug 12 19:17:19 ubifs support coming soon Aug 12 19:17:28 latest patch is still under review Aug 12 19:17:44 I'll bring site to usable state in following hours Aug 12 19:18:11 Jay7: what do i need in initramfs? Aug 12 19:18:20 mdev/udev/? Aug 12 19:18:32 minimal init system? Aug 12 19:19:17 Jay7: btw if you want i may add kexecboot to gentoo tree Aug 12 19:19:34 alexxy: it would be great Aug 12 19:19:46 alexxy: kexecboot may act as init itself Aug 12 19:19:52 aha Aug 12 19:19:58 what else i need? Aug 12 19:20:02 devtmpfs? Aug 12 19:20:04 you need just some minimal set of devices in /dev iirc Aug 12 19:20:12 or yes, devtmpfs Aug 12 19:20:19 ant__ may say more about this Aug 12 19:20:22 aha then i need only kernel kexecboot and devtmpfs Aug 12 19:20:31 iirc, we are using just devtmpfs now Aug 12 19:21:40 ant__: seems we need to write some docs about initramfs building Aug 12 19:21:53 i.e. initramfs tree and kernel options Aug 12 19:22:55 hello Aug 12 19:23:25 in OE we have IMAGE_INSTALL = "kexec-klibc kexecboot-klibc ubiattach-klibc" Aug 12 19:23:40 this and devtmpfs + suitable kernel Aug 12 19:24:14 you can do it also with shared utils if you don't have size limits Aug 12 19:24:57 I've switched dns back now Aug 12 19:25:23 site will be accessible after dns some timeout (about 3-24 hours) :) Aug 12 19:25:27 Jay7: try to push that last commit for ubifs..I have to commit somem oe patches Aug 12 19:26:12 ant__: sure, just need to review this finally Aug 12 19:26:38 I've sent email to qi-hardware ML about kexecboot on Ben NN again Aug 12 19:27:21 one user need to boot from ext3/ext4 Aug 12 19:27:36 seems BNN's uboot can't do this Aug 12 19:30:40 probably the kernel is ok now Aug 12 19:32:02 someone need to test kexec on NN again :) Aug 12 19:39:11 ant__: any new ideas about collie? Aug 12 19:40:15 hm..RCPT address has non-existant domain Aug 12 19:44:45 RCPT address has non-existant domain Aug 12 19:44:52 can't send patches :/ Aug 12 19:46:13 ant__: where is this smtp-server? Aug 12 19:46:28 lists.openembedded.org has address 140.211.169.152 Aug 12 19:46:35 lists.linuxtogo.org is an alias for www.linuxtogo.org. Aug 12 19:48:21 --smtp-server=out.alice.it Aug 12 19:58:01 ok, sent using gmail ssl Aug 12 20:04:09 Jay7: about collie khem sent me a patch but this breaks build :/ Aug 12 20:05:15 http://paste.debian.net/183270/ Aug 12 20:05:46 (it seems code for armv5t is generated) Aug 12 20:11:42 Jay7: where can i get kexecboot sources? Aug 12 20:11:51 * alexxy building kernel for hx4700 Aug 12 20:11:57 https://github.com/kexecboot/kexecboot Aug 12 20:12:11 cool Aug 12 20:18:21 bah 554 554 5.7.1 : Relay access denied (state 13). Aug 12 21:30:24 alexxy: we are still at 3.4 with linux-yocto Aug 12 21:30:54 and only 3.2 for linux-yocto-tiny, which is the base for our kexecboot kernel Aug 12 21:31:11 so, if you get 3.5 working we can pre-commit it Aug 12 21:31:12 ant__: its not a problem Aug 12 21:31:27 i can build all i need manualy Aug 12 21:31:35 in lxc arm container Aug 12 21:31:46 sure but it would be great to have it in meta-handheld Aug 12 21:42:20 Jay7: so, for collie, waiting for khem, full debug + dash linked as init ? Aug 12 21:42:39 ant__: something like Aug 12 21:43:13 I hope khem will fix armv4 compilation Aug 12 21:43:28 yes... Aug 12 21:43:39 I can't imagine how long it is broken Aug 12 21:43:53 at least for collie Aug 12 21:44:18 anyway it's strange, I still think we have a kernel issue Aug 12 21:44:42 no console? Aug 12 21:45:59 btw I've played a bit with the runtime mtdpart utility Aug 12 21:46:29 can't remove partitions..still listed in /proc/mtd Aug 12 21:46:50 I have to study that ioctl but is hard stuff, undocumented :/ Aug 12 21:47:38 good news are the utils compiles and runs gracefully :) Aug 12 21:48:04 at least the framework is done Aug 12 21:57:47 can you create new partition? Aug 12 21:57:57 no Aug 12 21:58:12 I have to remove it before Aug 12 21:58:41 well.. seems you need to ask on #mtd about this Aug 12 21:59:05 it seems the way to call it is Aug 12 21:59:07 http://paste.debian.net/183280/ Aug 12 21:59:55 but if you look, the old code was doing other things Aug 12 21:59:57 http://paste.debian.net/183282/ Aug 12 22:00:12 I just replaced the ioctl, not fixed the call ... Aug 12 22:01:46 ask on #mtd :) Aug 12 22:01:52 it's easier than guessing Aug 12 22:02:07 oh, too late here Aug 12 22:02:11 * Jay7 -> sleep Aug 12 22:02:15 gn **** ENDING LOGGING AT Mon Aug 13 02:59:58 2012 **** BEGIN LOGGING AT Mon Aug 13 02:59:59 2012 Aug 13 08:13:47 Jay7: found another confirmation Aug 13 08:13:49 http://tinyurl.com/cnnuxxh Aug 13 11:00:59 Jay7: today I'll have to hack the mtdparts at any cost...I have only 2 remaining hale-free days Aug 13 11:01:08 *half Aug 13 11:01:16 * ant_work heading home Aug 13 11:01:20 bbl Aug 13 12:57:41 Jay7: can you share initramfs image built for armv5te Aug 13 13:19:53 alexxy: not now Aug 13 13:20:12 it may be faster to request one from ant Aug 13 15:25:20 alexxy: http://www.crocko.com/CBC6131C7DB94D2EAD9FEE83C65FC6E5/initramfs-kexecboot-klibc-image-poodle-20120813152308.rootfs.cpio.gz Aug 13 15:25:58 but kexecboot has extra zaurus patches...should work Aug 13 15:26:27 bluelightning: note the cruft in that image: we just installed kexecboot,kexec,ubiattach Aug 13 15:30:16 ant__: without me downloading it what else do you get installed? Aug 13 15:32:36 andrea@mizar /oe/tt/etc $ ls -al Aug 13 15:32:36 total 24 Aug 13 15:32:36 drwxr-xr-x 4 andrea users 4096 Aug 13 17:24 . Aug 13 15:32:36 drwxr-xr-x 10 andrea users 4096 Aug 13 17:32 .. Aug 13 15:32:36 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 init.d Aug 13 15:32:37 -rw-r--r-- 1 andrea users 48 Aug 13 17:24 ld.so.cache Aug 13 15:32:39 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 rcS.d Aug 13 15:32:41 -rw-r--r-- 1 andrea users 13 Aug 13 17:24 version Aug 13 15:33:36 andrea@mizar /oe/tt/lib $ ls -aRl Aug 13 15:33:36 .: Aug 13 15:33:36 total 12 Aug 13 15:33:36 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 . Aug 13 15:33:36 drwxr-xr-x 10 andrea users 4096 Aug 13 17:32 .. Aug 13 15:33:37 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 modules Aug 13 15:33:39 ./modules: Aug 13 15:33:41 total 12 Aug 13 15:33:43 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 . Aug 13 15:33:45 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 .. Aug 13 15:33:47 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 3.2.18-yocto-standard Aug 13 15:33:49 ./modules/3.2.18-yocto-standard: Aug 13 15:33:51 total 36 Aug 13 15:33:53 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 . Aug 13 15:33:55 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 .. Aug 13 15:33:57 -rw-r--r-- 1 andrea users 45 Aug 13 17:24 modules.alias Aug 13 15:33:59 -rw-r--r-- 1 andrea users 12 Aug 13 17:24 modules.alias.bin Aug 13 15:34:01 -rw-r--r-- 1 andrea users 0 Aug 13 17:24 modules.builtin.bin Aug 13 15:34:05 -rw-r--r-- 1 andrea users 0 Aug 13 17:24 modules.dep Aug 13 15:34:07 -rw-r--r-- 1 andrea users 12 Aug 13 17:24 modules.dep.bin Aug 13 15:34:09 -rw-r--r-- 1 andrea users 52 Aug 13 17:24 modules.devname Aug 13 15:34:11 -rw-r--r-- 1 andrea users 131 Aug 13 17:24 modules.softdep Aug 13 15:34:13 -rw-r--r-- 1 andrea users 49 Aug 13 17:24 modules.symbols Aug 13 15:34:15 -rw-r--r-- 1 andrea users 12 Aug 13 17:24 modules.symbols.bin Aug 13 15:34:17 andrea@mizar /oe/tt/lib $ Aug 13 15:34:19 andrea@mizar /oe/tt/lib $ Aug 13 15:34:21 andrea@mizar /oe/tt/usr/share $ ls -aRl Aug 13 15:34:23 .: Aug 13 15:34:25 total 12 Aug 13 15:34:27 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 . Aug 13 15:34:29 drwxr-xr-x 5 andrea users 4096 Aug 13 17:24 .. Aug 13 15:34:31 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 run-postinsts Aug 13 15:34:35 ./run-postinsts: Aug 13 15:34:37 total 12 Aug 13 15:34:39 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 . Aug 13 15:34:41 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 .. Aug 13 15:34:43 -rw-r--r-- 1 andrea users 582 Aug 13 01:38 run-postinsts.awk Aug 13 15:34:45 andrea@mizar /oe/tt/usr/share $ Aug 13 15:34:47 andrea@mizar /oe/tt/var $ ls -aRl Aug 13 15:34:49 .: Aug 13 15:34:51 total 12 Aug 13 15:34:53 drwxr-xr-x 3 andrea users 4096 Aug 13 17:24 . Aug 13 15:34:55 drwxr-xr-x 10 andrea users 4096 Aug 13 17:32 .. Aug 13 15:34:57 drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 lib Aug 13 15:35:15 bluelightning: that's a load of cruft! Aug 13 15:35:27 hmm Aug 13 15:36:46 do those come from the kernel package or some other package pulled in by it? Aug 13 15:37:15 MACHINE_ESSENTIAL_EXTRA_RDEPENDS_append = " kernel-base kernel-image" you mean? Aug 13 15:37:39 well, I'm interested specifically where those came from... Aug 13 15:38:19 most instructive would be the buildhistory output... Aug 13 15:40:04 I see it's attached to INSTALL_PACKAGES_NORMAL_IPK=' Aug 13 15:40:47 so it adds postins in local 'package_to_install= kexec-klibc run-postinsts ubiattach-klibc kexecboot-klibc' Aug 13 15:42:10 then it goes with + INSTALLED_PKGS='kexec-klibc Aug 13 15:42:10 kexecboot-klibc Aug 13 15:42:11 run-postinsts Aug 13 15:42:11 ubiattach-klibc Aug 13 15:42:11 update-rc.d' Aug 13 15:45:01 ant__: I think you just may need to set IMAGE_FEATURES = "" Aug 13 15:45:18 oh wait, you already do that Aug 13 15:45:54 maybe the license task Aug 13 15:46:14 has to do with rc-update Aug 13 15:46:24 ups update-rc.d Aug 13 15:47:08 what happens if you do: bitbake -e | grep run-postinsts Aug 13 15:48:12 I'll try now, I was looking at both package.manifest and license.manifest Aug 13 15:48:29 of initramfs-kexecboot-klibc-image Aug 13 15:49:33 bluelightning: empty result Aug 13 15:50:12 oh, erm... what about: bitbake -e initramfs-kexecboot-klibc-image | grep run-postinsts Aug 13 15:50:31 :) yes Aug 13 15:52:30 http://paste.debian.net/183389/ Aug 13 15:53:23 what's the sdk stuff? Aug 13 15:54:15 note I have EXTRA_IMAGE_FEATURES = "debug-tweaks" Aug 13 15:56:38 the sdk stuff is just for being able to produce an SDK from the image; nothing to be concerned about Aug 13 15:56:53 what about bitbake -e initramfs-kexecboot-klibc-image | grep PACKAGE_INSTALL Aug 13 16:51:34 ant__: do i need some extra kernel parameters? Aug 13 16:55:33 ant__: also can you ompload it Aug 13 16:55:43 and dont use shareware filehosting Aug 13 17:02:10 Jay7: ^^^ Aug 13 17:40:59 hehe basic initramfs seems work Aug 13 17:41:13 now i'll try to boot something with kexecboot Aug 13 18:36:40 alexxy: booted? Aug 13 18:37:33 note you need /boot/boot.cfg where you can set your cmdline and other bootloader stuff Aug 13 18:46:02 bluelightning: Aug 13 18:46:04 andrea@mizar /oe/oe-core/build $ bitbake -e initramfs-kexecboot-klibc-image | grep PACKAGE_INSTALL Aug 13 18:46:04 # TOOLCHAIN_TARGET_TASK=task-core-standalone-sdk-target task-core-standalone-sdk-target-dbg ${PACKAGE_INSTALL} Aug 13 18:46:04 # TOOLCHAIN_TARGET_TASK_ATTEMPTONLY= ${PACKAGE_INSTALL_ATTEMPTONLY} Aug 13 18:46:04 # PACKAGE_INSTALL=${IMAGE_INSTALL} ${ROOTFS_BOOTSTRAP_INSTALL} ${FEATURE_INSTALL} Aug 13 18:46:05 export PACKAGE_INSTALL="kexec-klibc kexecboot-klibc ubiattach-klibc run-postinsts" Aug 13 18:46:07 # PACKAGE_INSTALL_ATTEMPTONLY=${FEATURE_INSTALL_OPTIONAL} Aug 13 18:46:09 andrea@mizar /oe/oe-core/build $ Aug 13 18:46:35 ant__: what about FEATURE_INSTALL ? Aug 13 18:47:01 andrea@mizar /oe/oe-core/build $ bitbake -e initramfs-kexecboot-klibc-image | grep FEATURE_INSTALL Aug 13 18:47:01 # RDEPENDS= ${IMAGE_INSTALL} ${LINGUAS_INSTALL} ${NORMAL_FEATURE_INSTALL} ${ROOTFS_BOOTSTRAP_INSTALL} Aug 13 18:47:01 # NORMAL_FEATURE_INSTALL_OPTIONAL=${@' '.join(oe.packagegroup.optional_packages(normal_groups(d), d))} Aug 13 18:47:01 # FEATURE_INSTALL_OPTIONAL=${@' '.join(oe.packagegroup.optional_packages(oe.data.typed_value('IMAGE_FEATURES', d), d))} Aug 13 18:47:01 # NORMAL_FEATURE_INSTALL=${@' '.join(oe.packagegroup.required_packages(normal_groups(d), d))} Aug 13 18:47:02 # RRECOMMENDS= ${NORMAL_FEATURE_INSTALL_OPTIONAL} Aug 13 18:47:04 # FEATURE_INSTALL=${@' '.join(oe.packagegroup.required_packages(oe.data.typed_value('IMAGE_FEATURES', d), d))} Aug 13 18:47:07 # PACKAGE_INSTALL=${IMAGE_INSTALL} ${ROOTFS_BOOTSTRAP_INSTALL} ${FEATURE_INSTALL} Aug 13 18:47:09 # PACKAGE_INSTALL_ATTEMPTONLY=${FEATURE_INSTALL_OPTIONAL} Aug 13 18:47:11 andrea@mizar /oe/oe-core/build $ Aug 13 19:02:13 hmm, so that's not set Aug 13 19:03:01 hang on it has to be coming from ROOTFS_BOOTSTRAP_INSTALL, but that implies IMAGE_FEATURES has to have package-management in it... Aug 13 19:03:21 what do you get if you grep for IMAGE_FEATURES ? Aug 13 19:03:28 ant__: ^ Aug 13 19:07:58 mom Aug 13 19:08:07 andrea@mizar /oe/oe-core/build $ bitbake -e initramfs-kexecboot-klibc-image | grep IMAGE_FEATURES Aug 13 19:08:07 # IMAGE_FEATURES= Aug 13 19:08:07 # FEATURE_INSTALL_OPTIONAL=${@' '.join(oe.packagegroup.optional_packages(oe.data.typed_value('IMAGE_FEATURES', d), d))} Aug 13 19:08:07 # ROOTFS_BOOTSTRAP_INSTALL=${@base_contains("IMAGE_FEATURES", "package-management", "", "${ROOTFS_PKGMANAGE_BOOTSTRAP}",d)} Aug 13 19:08:08 # FEATURE_INSTALL=${@' '.join(oe.packagegroup.required_packages(oe.data.typed_value('IMAGE_FEATURES', d), d))} Aug 13 19:08:11 # SDKIMAGE_FEATURES=dev-pkgs dbg-pkgs Aug 13 19:08:13 SDKIMAGE_FEATURES="dev-pkgs dbg-pkgs" Aug 13 19:08:15 # IMAGE_INSTALL_COMPLEMENTARY=${@complementary_globs("IMAGE_FEATURES", d)} Aug 13 19:08:17 # EXTRA_IMAGE_FEATURES=debug-tweaks Aug 13 19:08:19 EXTRA_IMAGE_FEATURES="debug-tweaks" Aug 13 19:08:23 # SDKIMAGE_INSTALL_COMPLEMENTARY=${@complementary_globs("SDKIMAGE_FEATURES", d)} Aug 13 19:08:25 """Return all the IMAGE_FEATURES, with the exception of our special package groups""" Aug 13 19:08:27 features = set(oe.data.typed_value('IMAGE_FEATURES', d)) Aug 13 19:08:29 andrea@mizar /oe/oe-core/build $ Aug 13 19:08:43 hm.. package mgmt... Aug 13 19:09:15 we set ONLINE_PACKAGE_MANAGEMENT = "none" Aug 13 19:09:16 ant__: yep Aug 13 19:09:27 with vanilla 3.5 on hx4700 Aug 13 19:09:57 very nice! Aug 13 19:10:11 bluelightning: ^^ Aug 13 19:10:37 ONLINE_PACKAGE_MANAGEMENT isn't used anymore AFAIK Aug 13 19:10:56 this is puzzling Aug 13 19:11:09 the real question is how is that value getting into PACKAGE_INSTALL Aug 13 19:11:10 grepping... Aug 13 19:12:04 no matches Aug 13 19:13:15 maybe ./classes/image.bbclass:export PACKAGE_INSTALL ?= "${IMAGE_INSTALL} ${ROOTFS_BOOTSTRAP_INSTALL} ${FEATURE_INSTALL}" Aug 13 19:13:23 see Aug 13 19:13:25 andrea@mizar /oe/oe-core/meta $ grep -R PACKAGE_INSTALL . Aug 13 19:13:25 ./classes/image.bbclass:TOOLCHAIN_TARGET_TASK += "${PACKAGE_INSTALL}" Aug 13 19:13:25 ./classes/image.bbclass:TOOLCHAIN_TARGET_TASK_ATTEMPTONLY += "${PACKAGE_INSTALL_ATTEMPTONLY}" Aug 13 19:13:25 ./classes/image.bbclass:export PACKAGE_INSTALL ?= "${IMAGE_INSTALL} ${ROOTFS_BOOTSTRAP_INSTALL} ${FEATURE_INSTALL}" Aug 13 19:13:25 ./classes/image.bbclass:PACKAGE_INSTALL_ATTEMPTONLY ?= "${FEATURE_INSTALL_OPTIONAL}" Aug 13 19:13:27 ./classes/image.bbclass: runtime_mapping_rename("PACKAGE_INSTALL", d) Aug 13 19:13:29 ./classes/image.bbclass: runtime_mapping_rename("PACKAGE_INSTALL_ATTEMPTONLY", d) Aug 13 19:13:31 ./classes/multilib.bbclass: clsextend.map_depends_variable("PACKAGE_INSTALL") Aug 13 19:13:33 ./classes/multilib.bbclass: pinstall = d.getVar("LINGUAS_INSTALL", True) + " " + d.getVar("PACKAGE_INSTALL", True) Aug 13 19:13:36 ./classes/multilib.bbclass: d.setVar("PACKAGE_INSTALL", pinstall) Aug 13 19:13:38 ./classes/multilib.bbclass: d.setVar("PACKAGE_INSTALL_ATTEMPTONLY", "") Aug 13 19:13:40 ./classes/multilib.bbclass: clsextend.map_variable("PACKAGE_INSTALL") Aug 13 19:13:42 ./classes/rootfs_deb.bbclass: export INSTALL_PACKAGES_NORMAL_DEB="${PACKAGE_INSTALL}" Aug 13 19:13:44 ./classes/rootfs_deb.bbclass: export INSTALL_PACKAGES_ATTEMPTONLY_DEB="${PACKAGE_INSTALL_ATTEMPTONLY}" Aug 13 19:13:46 ./classes/rootfs_ipk.bbclass: export INSTALL_PACKAGES_ATTEMPTONLY_IPK="${PACKAGE_INSTALL_ATTEMPTONLY}" Aug 13 19:13:48 ./classes/rootfs_ipk.bbclass: export INSTALL_PACKAGES_IPK="${PACKAGE_INSTALL}" Aug 13 19:13:50 ./classes/rootfs_rpm.bbclass:IMAGE_ROOTFS_EXTRA_SPACE_append = "${@base_contains("PACKAGE_INSTALL", "zypper", " + 51200", "" ,d)}" Aug 13 19:13:55 ./classes/rootfs_rpm.bbclass: export INSTALL_PACKAGES_RPM="${PACKAGE_INSTALL}" Aug 13 19:13:57 ./classes/rootfs_rpm.bbclass: export Aug 13 19:13:59 INSTALL_PACKAGES_ATTEMPTONLY_RPM="${PACKAGE_INSTALL_ATTEMPTONLY}" Aug 13 19:14:11 ant__: how should i specify path? Aug 13 19:14:48 in each bootable partition, put /boot/boot.cfg and /boot/zImage Aug 13 19:15:14 then edit it http://kexecboot.org/documentation/how_to_write_config Aug 13 19:18:15 mine is Aug 13 19:18:17 # First kernel stanza. Aug 13 19:18:17 # Show this label in kexecboot menu. Aug 13 19:18:17 LABEL=oe-core.0-poodle Aug 13 19:18:17 # Specify full kernel path on target. Aug 13 19:18:17 KERNEL=/boot/zImage Aug 13 19:18:18 # Aug 13 19:18:20 # Append this tags to the kernel cmdline. Aug 13 19:18:24 APPEND=console=ttyS0,115200n8 console=tty1 fbcon=rotate:1 quiet Aug 13 19:19:13 alexxy: we just need the kernel .config to integrate the device in meta-handheld Aug 13 19:19:26 alexxy: all the rest comes as bonus :) Aug 13 19:21:12 Jay7: once we clean the cpio from the cruft I'll upload some armv4, armv5 and x86 initramfs Aug 13 19:22:38 now that bluelightning is hiding I'll hack badly the images :) Aug 13 19:38:28 ant__: how long does it take to boot kernel? Aug 13 19:38:38 or seems it cannot boot Aug 13 19:43:46 or seems it cannot kexec anything Aug 13 19:45:58 alexxy: you may check kexec with any other kernel and rootfs known to work Aug 13 19:46:32 Jay7: ok Aug 13 19:56:50 it takes 1-2 seconds, no more Aug 13 19:57:23 depends on rootfstype and on kernel compression Aug 13 19:58:13 alexxy: do cursor keys work? Aug 13 19:58:25 can you use fb menu? Aug 13 19:58:28 yep Aug 13 19:58:34 ok Aug 13 19:58:35 it works Aug 13 19:58:43 but kexec seems not Aug 13 19:58:56 Jay7: on poodle the #[1-9] boot doesn't work Aug 13 19:59:19 alexxy: try on another rootfs with a tested kexec Aug 13 19:59:28 ant__: may be it needs some keymap hack too Aug 13 19:59:28 I have to upgrade kexec-klibc Aug 13 19:59:32 i will try Aug 13 19:59:40 Jay7: I have keymap patch :] Aug 13 19:59:53 cool Aug 13 20:00:01 Jay7: I'll need to test on shell Aug 13 20:00:19 because in oe images we load another map Aug 13 20:00:34 that's why kernel patch is needed Aug 13 20:00:42 like on ben n-n Aug 13 20:01:01 initramfs is perfect to test Aug 13 20:01:18 so it's good idea to have special initramfs with shell :) Aug 13 20:01:36 well, it should not be necessary Aug 13 20:01:44 if kexecboot works... Aug 13 20:03:47 (no space on collie btw) Aug 13 20:09:08 alexxy: ipaq is pxa270 isn't? We tested kexec binary on it Aug 13 20:10:01 yep Aug 13 20:10:05 tested on pxa250, pxa255, pxa270 32/64 MiB RAM Aug 13 20:10:22 but there was at least 4 versions of kexec for ipaq pxa Aug 13 20:10:39 i dont know what is mainline version Aug 13 20:11:02 also mine has 128M ram Aug 13 20:11:25 this is kexec-tools_2.0.2 Aug 13 20:11:37 compiled with some hacks klibc-static Aug 13 20:11:53 seems working beautifully on pxa Aug 13 20:12:19 well, i'll try with glibc version Aug 13 20:12:29 do you have it? Aug 13 20:26:45 Jay7: btw Aug 13 20:26:53 [OE-core] [PATCH 1/1] runqemu: support booting ramfs-based images Aug 13 20:27:00 Added support for booting ramfs-based images (e.g, poky-tiny) which Aug 13 20:27:00 do not have block device support. Aug 13 20:29:10 ant__: where kexecboot expect boot.cfg? Aug 13 20:29:14 in / Aug 13 20:29:21 or in /boot/boot.cfg? Aug 13 20:29:38 seems like it see it only in / Aug 13 20:30:14 in /boot Aug 13 20:30:52 check in the debug mnu Aug 13 20:31:01 you'll see if it is found Aug 13 20:33:17 it see it only in . Aug 13 20:33:42 at compile time there is Aug 13 20:33:45 --enable-zimage enable looking for zImage kernel when no boot.cfg is found [default=yes] Aug 13 20:34:06 so it finds the kernel Aug 13 20:34:33 I suppose the menu lists 'zImage' Aug 13 20:34:45 no Aug 13 20:34:51 i wrote boot.cfg Aug 13 20:35:01 and it shows its label Aug 13 20:35:44 but it hangs after i try to boot anything Aug 13 20:36:35 do you have serial cable? Aug 13 20:37:10 try to add 'debug' to the APPEND line in boot.cfg Aug 13 20:37:42 is kexec enabled in the kexecboot-kernel you're running? Aug 13 20:38:43 anyway Aug 13 20:38:44 https://github.com/kexecboot/kexecboot/blob/master/cfgparser.h#L26 Aug 13 20:39:08 is hardcoded :) Aug 13 20:56:05 kexec enabled Aug 13 20:56:32 i domy have serial for this device Aug 13 20:56:48 *dont Aug 13 21:11:21 pls check in 'Show debug info' in System Menu Aug 13 21:18:55 you should see Aug 13 21:19:01 + config file found Aug 13 21:21:29 alexxy: ouch, I realize now I've sent you a qvga 240x320 version :) Aug 13 21:21:42 wait a couple of mins Aug 13 21:23:14 I'll give you fullscreen 480x640 Aug 13 21:24:58 alexxy: does the kernel you'r e kexec-booting support boot from that device/fs? Aug 13 21:26:06 a.k.a. no-modules Aug 13 21:33:25 alexxy: see PM for the link **** ENDING LOGGING AT Tue Aug 14 02:59:58 2012 **** BEGIN LOGGING AT Tue Aug 14 02:59:58 2012 Aug 14 09:21:10 bluelightning: hello Aug 14 09:21:37 about image pollution, I have a candidate USER_CLASSES ?= "buildstats image-mklibs image-prelink" Aug 14 09:21:47 hi ant_work Aug 14 09:22:14 how does that cause the issue? Aug 14 09:22:32 btw I was under the impression buildhistory is enabled by default, isn't Aug 14 09:22:42 no, it isn't... Aug 14 09:23:18 it increases build time so it's best that it only be enabled for those that need it Aug 14 09:23:46 I've tried zapping all the variables showed in my greps, no luck, maybe I should have rebuilt the tasks Aug 14 09:24:13 now, looking for buildhisory, I see ths USER_CLASSES Aug 14 09:35:12 bluelightning: then in image.bbclass, #166 fakeroot do_rootfs Aug 14 09:37:28 this for /var/lib Aug 14 09:41:01 USER_CLASSES just tells the system to inherit from the specified bbclasses Aug 14 09:43:14 well, I'm checking the classes now and see strange 0 bytes files Aug 14 09:43:32 (chasing for the modules) Aug 14 09:43:51 kernel.bbclass -> inherit linux-kernel-base module_strip Aug 14 09:43:57 module_strip.bbclass is empty Aug 14 09:44:24 yes I'm not entirely sure what that's about Aug 14 09:44:26 (and what is image-empty.bbclass, also 0b ?) Aug 14 09:44:32 heh Aug 14 09:44:47 I'll do some grep once at home :) Aug 14 09:45:42 that's so that we can inherit inherit image-${IMAGE_TYPE_live} and set IMAGE_TYPE_live to "empty" if we want to disable that Aug 14 09:46:15 O_o Aug 14 09:47:36 the code for handling live and vmdk images is somewhat of a mess Aug 14 09:50:20 anyway, it looks we trigger http://cgit.openembedded.org/openembedded-core/tree/meta/classes/image.bbclass#n205 Aug 14 09:50:48 I see mkdir -p ${IMAGE_ROOTFS}/lib/modules/$KERNEL_VERSION Aug 14 09:51:26 i.e. I get drwxr-xr-x 2 andrea users 4096 Aug 13 17:24 3.2.18-yocto-standard Aug 14 09:52:48 so yes, modules pollution happens in fakeroot_do_rootfs Aug 14 09:54:51 and being we have ld.so.cache I think ldconfig is also run Aug 14 09:55:05 see image.bbclass#198 Aug 14 09:56:07 basically we should amend fakeroot do_rootfs () { Aug 14 09:56:24 :/ Aug 14 09:57:13 (note that with a postinst script we could sinply remove unnecessary stuff Aug 14 09:57:16 ) Aug 14 09:57:48 but yes, it's just a band-aid Aug 14 10:07:54 I'd rather we fix whatever needs fixing in image.bbclass... Aug 14 10:08:19 lots of people want to be able to build clean initramfs images Aug 14 10:11:04 I think it's anything new, I did complain lot ago Aug 14 10:11:24 probably that code is there since the origins of oe-core Aug 14 10:11:49 afaik we in oe-classic we had lean images lately Aug 14 10:26:22 we're open to fixing things if you can be specific about what needs fixing... Aug 14 10:32:26 sure, I am narrowing down the investigation fields Aug 14 10:34:30 for what I see, there are rather poor knobs to tweak Aug 14 10:34:38 if [ "x${LDCONFIGDEPEND}" != "x" ]; then Aug 14 10:34:45 for the ldconfig Aug 14 10:34:52 and Aug 14 10:34:53 if [ -e ${STAGING_KERNEL_DIR}/kernel-abiversion ]; then Aug 14 10:35:03 for the modules stuff Aug 14 10:35:13 but I need to test-build Aug 14 10:36:44 the first can be settled setting this new var LDCONFIGDEPEND Aug 14 10:37:12 the second case is more problematic, it just checks if you have built a kernel as far as I see Aug 14 10:37:54 again, unsure about "kernel-abiversion" file, maybe it's not created by kernel recipe Aug 14 10:47:04 bluelightning: I see now why the code was added Aug 14 10:47:26 ..if image do not contain modules at all...unable to open modules.dep Aug 14 10:48:02 my Gentoo servers have no modules, I get this 'unable..' msg and I beautifully ignore it :) Aug 14 10:48:49 Aug 14 09:46:11 [modprobe] FATAL: Could not load /lib/modules/3.3.8-gentoo/modules.dep: No such file or directory_ Aug 14 10:48:59 :) Aug 14 10:50:59 now, in the case of *kexecboot-image we don't have modules nor initscripts complaining Aug 14 10:51:12 but other general-purpose initramfs may need that... **** ENDING LOGGING AT Wed Aug 15 02:59:59 2012 **** BEGIN LOGGING AT Wed Aug 15 02:59:59 2012 Aug 15 20:08:48 Jay7: http://paste.debian.net/183724/ Aug 15 20:09:01 still failing with mtd device... Aug 15 20:14:01 this is what the kernel thinks, until remount Aug 15 20:21:25 the only diff I can see is Aug 15 20:21:43 block/ioctl.c static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user *arg) Aug 15 20:22:21 mtd/mtdchar.c ret = mtdchar_blkpg_ioctl(mtd, (struct blkpg_ioctl_arg __user *)arg); Aug 15 20:23:33 well, the proto is static int mtdchar_blkpg_ioctl(struct mtd_info *mtd, struct blkpg_ioctl_arg __user *arg) Aug 15 20:42:42 I'm so near, that struct mtd_info is filled by another ioctl Aug 15 20:43:02 mtd_info_t mtd_info; Aug 15 20:45:27 my doubt is, i fill the struct with i.e. fd = open("/dev/mtd3", O_RDONLY); ioctl(fd, MEMGETINFO, &mtd_info); **** ENDING LOGGING AT Thu Aug 16 02:59:58 2012 **** BEGIN LOGGING AT Thu Aug 16 02:59:58 2012 **** BEGIN LOGGING AT Thu Aug 16 03:18:28 2012 Aug 16 08:02:09 bluelightning: great job exploding tasks! Aug 16 08:02:33 ant_work: thanks, I hope that we have some more discussion on it Aug 16 08:19:04 * ant_work is hardly following the kernel->userspace interface of mtd char :/ Aug 16 14:50:07 bluelightning: I see kernel.bbclass: remove default module_autoload and module_conf values Aug 16 14:50:39 ant_work: yes, I guess some of them ought to come back in our layer if that mechanism still works Aug 16 14:50:49 now it's the time to reintroduce those. The settings were in zaurus.conf but maybe we could create a generic hh file Aug 16 14:52:14 those + the other I killed with my patch :) Aug 16 14:52:34 (sound and rtc iirc) Aug 16 20:52:04 Jay7: it seems the irc log does not rotate Aug 16 20:52:29 yeah Aug 16 20:52:41 ask ka6sox about this Aug 16 20:53:19 * Jay7 -> sleep Aug 16 20:55:10 'nite Aug 16 22:44:40 bluelightning: time for testing a collie kernel ? Aug 16 22:48:56 I'll pm to you both the d/l link Aug 16 22:51:58 gn **** ENDING LOGGING AT Fri Aug 17 02:59:58 2012 **** BEGIN LOGGING AT Fri Aug 17 02:59:58 2012 Aug 17 07:52:19 bluelightning: hello Aug 17 07:52:27 hi ant_work Aug 17 07:53:24 bluelightning: you know what, I casually discovered that if the batter-cover switch is open the poodle does poweroff. It was restarting instead.. Aug 17 07:53:46 There is some wrong gpio :/ Aug 17 07:53:50 hmm, some devices do pay attention to it Aug 17 07:54:02 my collie doesn't, but my spitz definitely does Aug 17 07:54:13 however the spitz one might actually be wired into the supply line, not sure Aug 17 07:54:41 I've quickly tracked the boards code from 2.6 to 3.2 and there have been heavy changes Aug 17 07:55:04 most is shared as 'pxa_xx.gpio' Aug 17 07:55:24 I've sent an email to eric miao about that Aug 17 07:56:49 this happens in shell and from within kexecboot. Seems that gpio could be sending a WAKE signal Aug 17 07:57:39 btw have you had the etime to try that collie kernel? If it boots I've compiled a rootfs as well overnight Aug 17 07:57:54 I added 3 linaro patches Aug 17 07:58:08 for gcc Aug 17 08:03:24 I downloaded it late last night but did not have time to try flashing it Aug 17 08:04:30 no worries Aug 17 08:04:59 it's encouraging there is a similar issue for armv4 and ppc Aug 17 08:05:29 "Unable to get serial console login prompt" Aug 17 08:05:46 khem patch should fix both **** ENDING LOGGING AT Sat Aug 18 03:00:00 2012 **** BEGIN LOGGING AT Sat Aug 18 03:00:00 2012 **** ENDING LOGGING AT Sun Aug 19 03:00:01 2012 **** BEGIN LOGGING AT Sun Aug 19 03:00:01 2012 **** ENDING LOGGING AT Mon Aug 20 02:59:59 2012 **** BEGIN LOGGING AT Mon Aug 20 02:59:59 2012 Aug 20 08:04:31 hello bluelightning Aug 20 08:04:42 Ive tested your patch for zaurusd Aug 20 08:05:13 pidof symlinks to killall5 Aug 20 08:05:41 no new breakages, usual complaints about alsa mixer Aug 20 08:05:54 so it's ok then? Aug 20 08:06:19 well, yes, still no clean log runtime Aug 20 08:06:35 I'll try readding the snd modules Aug 20 08:07:10 hmm ok... if you don't have any luck then let me know, I'll have a look into it Aug 20 08:07:23 btw I was trying to identify the process with behalf of 'ps' Aug 20 08:07:41 ..bu the binary seems not accepting aux options Aug 20 08:09:15 bluelightning: new issues I've seen are faling postinst scripts Aug 20 08:09:25 not many iirc, 2-3 Aug 20 08:09:42 no, busybox ps will not accept any options, that's why I had to fix up the scripts to handle that Aug 20 08:09:43 one was eglibc locale en_GB :) Aug 20 08:09:50 argh Aug 20 08:10:18 btw I tried that new kernel image you sent, same issue (mangled logo and then very slow fade to white) Aug 20 08:10:45 ouch Aug 20 08:10:55 still killing init? Aug 20 08:11:24 I assume so, you can see the message in this version Aug 20 08:12:10 I'l inform khem :/ Aug 20 08:13:55 I added linaro 830, 840 Aug 20 08:14:13 seems there are missing fixes for the linker Aug 20 08:20:56 ant_work: we should probably disable the kernel logo for collie, it kind of just gets in the way Aug 20 08:21:12 at least temporarily while we're trying to sort out these problems Aug 20 08:21:12 right Aug 20 08:21:43 I'll compile one with more debug but without jffs2/ubi to stay in the size Aug 20 08:21:55 let say just ext2/ext3 Aug 20 08:22:23 we have to see what is killing the serial console Aug 20 08:32:23 hmm, still no serial cables on ebay... Aug 20 23:34:32 bluelightning: this cleans the cpio of the /lib and /etc/ld.so.cache Aug 20 23:34:40 http://paste.debian.net/184804/ Aug 20 23:35:36 now I have to remove the postinst stuff Aug 20 23:36:07 like /etc/init.d/run.postinsts Aug 20 23:37:10 and /usr/share/run-postinsts/run-postinsts.awk Aug 20 23:37:46 then we get an empty /var/lib but is really the last one to catch :) **** ENDING LOGGING AT Tue Aug 21 02:59:58 2012 **** BEGIN LOGGING AT Tue Aug 21 02:59:58 2012 Aug 21 07:47:41 bluelightning: hello, pls check channel logs Aug 21 07:47:55 ant_work: hi, ok, will do Aug 21 07:49:44 ant_work: you mean your info about how to remove the unnecessary files? Aug 21 07:49:55 yes, I could not find a good way to sanitize image.bbclass but you get the idea Aug 21 07:51:40 hmm, I'll see if I can figure out a proper conditional there Aug 21 07:54:16 I was thinking to check for INITRAMFS_IMAGE but in theory one could need that stuff in the cpio **** ENDING LOGGING AT Wed Aug 22 03:00:00 2012 **** BEGIN LOGGING AT Wed Aug 22 03:00:00 2012 **** BEGIN LOGGING AT Wed Aug 22 17:34:23 2012 Aug 22 21:19:00 ant__: where to get new kernel? Aug 22 21:19:30 the one khem posted is not the kexecboot Aug 22 21:19:44 he did some kernel.bbclass chenges Aug 22 21:23:31 Jay7: if you hang around I'll rebuild one kexecboot-kernel Aug 22 21:23:44 let's do this tomorrow Aug 22 21:23:49 I'm busy today **** ENDING LOGGING AT Thu Aug 23 02:59:59 2012 **** BEGIN LOGGING AT Thu Aug 23 03:00:00 2012 Aug 23 07:24:30 Jay7: new bug? http://lxtek.de/pandora/kexecboot/ Aug 23 07:47:02 Jay7: no, it's just broken kexec on pandora :) Aug 23 07:47:17 forget about it Aug 23 08:33:24 hehe Aug 23 08:46:46 we should write that in the documentation: pls check kernel + kexec from shell Aug 23 08:46:49 ;) Aug 23 08:47:12 ..before coming to the conclusion kexecboot is buggy :) Aug 23 08:51:17 yes.. some big disclaimer at top of every page :) Aug 23 13:24:38 Jay7, bluelightning: I'll upload the new zImage for collie in a couple of hours Aug 23 13:28:33 Jay7: btw, about rootfstype= http://git.denx.de/?p=u-boot.git;a=commit;h=10f3bdd36e328b0bf4f8065e2d650dd8e9fa7976 Aug 23 13:29:06 ant_work: khem sent me a kernel/rootfs, haven't tried them yet Aug 23 13:29:16 ant_work: hm.. interesting Aug 23 13:29:22 we need to check how real is it Aug 23 13:29:51 and from which kernel version this is true Aug 23 13:29:59 * Jay7 gone Aug 23 14:52:46 bluelightning: I've only built the kexecboot-kernel Aug 23 14:53:01 that's sufficient to test the current issues Aug 23 14:53:33 in fact the kexecboot kernel is more useful than a bare one because it gives you some functionality without having to also flash a rootfs Aug 23 14:54:17 and is immune of network/usb issues (not compiled :) Aug 23 14:54:40 hmm.. immune for? Aug 23 14:55:28 bummer...from / against Aug 23 14:56:28 I have to vary my english-diet, atm just tech stuff :p Aug 23 14:57:01 immune from Aug 23 14:57:04 yes Aug 23 14:57:06 :) Aug 23 20:42:14 jay7: hi Aug 23 20:42:33 you have a collie, am i right? Aug 23 20:47:09 dromede: yes Aug 23 20:48:27 i'm working on the latest kernel, 3.6-rc3 Aug 23 20:48:32 vanilla kernel wont boot Aug 23 20:48:54 i have patch ready, but i need to test it on a collie first before commiting it Aug 23 20:51:09 here's the patch: Aug 23 20:51:10 http://paste.debian.net/185318 Aug 23 20:51:21 it would be great if you could test it when you have time Aug 23 21:16:18 dromede: if you build one for me I can test it Aug 23 21:17:10 ok, will do Aug 23 21:23:38 hello Aug 23 21:24:45 hi, ant__ Aug 23 21:25:29 I come later at home 'cause server showed nasty ipmi msgs Aug 23 21:30:18 dromede: any idea about blkpg ioctl? Aug 23 21:30:42 sorry, no Aug 23 21:32:03 about the patch for collie, we don't know yet if we are hitting toolchain bug or kernel bug :/ Aug 23 21:32:10 3.2 doesn't find console Aug 23 21:32:19 kill init, panic Aug 23 21:33:10 Jay7 is about testing an experimental kernel, with patched oe-toolchain Aug 23 21:33:34 if that boots on collie, I'd move all to 3.6 Aug 23 21:34:48 dromede: btw, how does the kernel get the params from nand? Aug 23 21:35:22 iirc kernel was not handling fw 'params' Aug 23 21:35:42 you mean mtd params? Aug 23 21:35:45 yep Aug 23 21:35:59 I mean, brightness, lcd phase,... Aug 23 21:36:04 oh Aug 23 21:36:18 http://lxr.free-electrons.com/source/arch/arm/common/sharpsl_param.c Aug 23 21:36:34 just a memcpy from a fixed address into a struct Aug 23 21:36:49 well, then we could read the mtdparts in the same way... Aug 23 21:37:00 fixed address Aug 23 21:37:23 mtdparts are defined statically in machine init code Aug 23 21:37:41 I know :/ lazy devs! Aug 23 21:38:16 the bootloader still loads the params in ram Aug 23 21:38:30 but modern kernels do ignore that pre-atag blob Aug 23 21:38:44 params and mtdparts Aug 23 21:43:33 dromede: ah, now I see. kernel reads bootloader params from ram addresses Aug 23 21:43:36 I meant http://www.h5.dion.ne.jp/~rimemoon/zaurus/memo_006.htm Aug 23 21:44:15 * ant__ is wondering wether mtdparts is copied as well in ram Aug 23 21:44:22 ..and where Aug 23 21:45:20 RIchard Purdie might know Aug 23 21:45:32 he he Aug 23 21:45:33 he's the guy who did the 2.6 port Aug 23 21:45:43 he's too busy for Zaurus nowadays :) Aug 23 21:46:14 we have marex and lumag Aug 23 21:46:24 ..and you Aug 23 21:46:51 right now, i'm trying to get all the bugs out of 3.6 kernel series Aug 23 21:47:05 sounds good Aug 23 21:47:32 3.6 has some good surprises for our machines Aug 23 21:47:39 one of them is cpufreq Aug 23 21:47:54 what about offline charging? :) Aug 23 21:47:56 i did some tests and it seems that it's finally stable Aug 23 21:48:01 as for me, I can't tolerate anymore the wrong mtdparts for the poor linux-kexecboot Aug 23 21:48:06 that's war ! Aug 23 21:48:11 suspend/resume, usb net Aug 23 21:48:35 i talked about suspend/resume before Aug 23 21:48:42 and the bugs currently associated with it Aug 23 21:48:48 well, there's one actually Aug 23 21:49:19 if you suspend the machine with the ac plugged in Aug 23 21:49:32 and if you then remove the ac plug while suspended Aug 23 21:49:35 the machine will hang Aug 23 21:49:49 same goes for plugging in the ac plug while suspended Aug 23 21:50:11 it will also hang if the offline charger finishes charging while suspended Aug 23 21:50:25 this one is not easy to track down Aug 23 21:50:34 the weird thing is Aug 23 21:50:47 even if i completely disable the AC IN interrupt Aug 23 21:50:49 fwiw there are nasty ubifs bugs with 3.6-rc3 Aug 23 21:50:55 the kernel would still hang Aug 23 21:50:58 lot of patches around Aug 23 21:51:48 usb gadgets work with metan's zaurusd and my patches Aug 23 21:52:40 dromede: imagine that I casually discovewred that zaurus poodle actually does 'shut off' if I open the battery-cover switch. Otherwise it restarts :/ Aug 23 21:52:56 some wrong gpio? Aug 23 21:53:19 this sounds familiar, like i read a post somewhere about it Aug 23 21:53:55 I've sent the prom content to Dmitry, who will disassemble the bootloader to find the 'magic' Aug 23 21:54:12 arm does never 'poweroff' Aug 23 21:54:27 it's always a suspend state and a combination of registers Aug 23 21:54:36 i think so, yes Aug 23 21:54:38 I did not know it's so complicated... Aug 23 21:55:20 I'll fwd you the e-mail, maybe you get some new idea about power man Aug 23 21:55:45 i have NO idea what's going on, that's problem Aug 23 21:56:42 ok, Dmitry held a lesson :) Aug 23 21:59:04 dromede: try to communicate with stanislav brabec and marex Aug 23 21:59:23 i'll ask for help on the arm linux list Aug 23 21:59:32 I remember some patches for battery in zaurus-devel made by someone Aug 23 21:59:34 Eric is the pxa maintainer Aug 23 21:59:43 i just need to write a detailed description of the problem Aug 23 22:00:35 yes, there's a battery monitoring driver written by utx Aug 23 22:00:40 it's not in mainline Aug 23 22:09:59 dromede: maybe I was looking for that ? Aug 23 22:10:00 http://piro.sytes.net/~piro/zaurus/bootloader/patch-2.6.16.16-z.diff Aug 23 22:12:44 the patch looks like something similar to kexec() Aug 23 22:12:54 yes, is a bootloader Aug 23 22:13:00 i guess you're interested in cmdline parameters Aug 23 22:13:08 ^_^ Aug 23 22:13:25 just mtdparts Aug 23 22:14:09 this would save the hassle of fake repartitioning with BLKPG Aug 23 22:15:04 actually Aug 23 22:15:35 now i remembered that old pdaxrom installers had an option to change mtd partition sizes Aug 23 22:15:45 yes Aug 23 22:15:51 nand-repartitionig Aug 23 22:16:12 we have the sources Aug 23 22:16:32 fsro_resize? Aug 23 22:16:37 is that it? Aug 23 22:16:47 yes Aug 23 22:17:00 and some other dev stuff Aug 23 22:17:47 actually we just lack the ability to read Aug 23 22:19:39 (from kernel) **** ENDING LOGGING AT Fri Aug 24 03:00:04 2012 **** BEGIN LOGGING AT Fri Aug 24 03:00:04 2012 Aug 24 06:24:11 q Aug 24 12:02:24 bluelightning: Iwrt image-pollution 've seen new contermeasures, like # remove not needed ipkg informations Aug 24 12:02:24 ROOTFS_POSTPROCESS_COMMAND += "remove_packaging_data_files ; " Aug 24 12:04:10 so, yes, there is an issue with the basic images Aug 24 12:06:51 and this is in the classes, one step before tasks and images Aug 24 13:21:49 jay7: The kernels are ready Aug 24 13:21:53 get them here: Aug 24 13:22:00 https://rapidshare.com/files/3175804256/3.6-rc3.tar.bz2 Aug 24 13:22:06 one is vanilla 3.6-rc3 Aug 24 13:22:24 the other is patched Aug 24 13:22:40 hopefully the patched one should but and vanilla shouldn't Aug 24 13:23:06 both were built with old 3.2 defconfigs taken from meta-handheld Aug 24 13:23:13 i tried running the kernel with qemu Aug 24 13:23:18 dromede: should I need to flash that kernel into NAND or boot via kexecboot? Aug 24 13:23:25 kexecboot is fine Aug 24 13:23:37 both failed to boot under qemu Aug 24 13:23:49 any rootfs included? Aug 24 13:23:49 but then again, the kexecboot kernel for collie also failed to boot Aug 24 13:23:53 nope Aug 24 13:24:21 not really needed Aug 24 13:24:29 ok, I'll check ant's kexecboot at first Aug 24 13:24:43 as i said both failed under qemu 0.15.1 Aug 24 13:24:58 but then again so did the kexecboot kernel from kexecboot.org Aug 24 13:25:09 probably means that collie is broken in qemu 0.15.1 Aug 24 14:01:36 dromede: khem tested qemu and said the same Aug 24 14:02:09 we have a supposedly 'working' kernel provided by one user, though Aug 24 14:04:29 maybe it works in qemu_git Aug 24 14:04:36 i only have 0.15.1 running Aug 24 14:04:38 link is somehwhere in the logs Aug 24 14:04:50 hm, no, khem said he used latest Aug 24 14:04:56 ah, ok Aug 24 14:05:04 it seems Dmitry did not finish the port Aug 24 14:05:37 it would still be interesting to test a known working kernel Aug 24 14:06:00 strongarm_ppc_write: Bad offset 0x00000030 Aug 24 14:06:00 pflash_write: Unimplemented flash cmd sequence (offset 00000000, wcycle 0x0 cmd 0x0 value 0xf0f0f0f0) Aug 24 14:06:17 that's what happens Aug 24 14:06:30 so yeah, it looks like an unfinished port Aug 24 14:09:02 ok, trythis Aug 24 14:09:04 Aug 06 11:01:19 here some patches for 2.6.31 http://www.linuxtogo.org/gowiki/CollieEnhancements?action=AttachFile&do=get&target=kernelreplacement-2.6.31.12-r10-rep3.tgz Aug 24 14:09:28 thred was http://lists.linuxtogo.org/pipermail/angstrom-distro-users/2011-June/003682.html Aug 24 14:12:32 nope, this too wont boot Aug 24 14:12:46 here a log of vanilla 2.6.38-c3 booting: http://pastebin.com/se3BMrmU Aug 24 14:13:11 Jay7: was the last one successfully booting :p Aug 24 14:13:34 wait a minute Aug 24 14:13:45 2.6.38 is the last kernel that worked on collie? Aug 24 14:13:58 no, we can boot 3.2 up to freeing init Aug 24 14:14:03 * Jay7 can't remember which one was Aug 24 14:14:04 mom Aug 24 14:14:17 ah, ok Aug 24 14:14:32 3.6-rc3 should fail much earlier Aug 24 14:15:02 Jay7: http://paste.debian.net/183724/ Aug 24 14:15:12 ouch Aug 24 14:15:28 i can't open the link Aug 24 14:17:53 all links are gone Aug 24 14:18:04 Jay7 should have the logs Aug 24 14:18:14 anyway, it was panicing after Aug 24 14:18:15 Warning: unable to open an initial console. Aug 24 14:18:48 as I said, we do have toolchain issues for armv4 Aug 24 14:19:14 but it could also be a new kernel bug Aug 24 14:21:54 dromede: we are adding debug and a shell for kexecboot kernel. We'll know soon Aug 24 14:22:30 you could also use kgdb Aug 24 14:22:54 in the meanwhile I hope one of the last kernels (from Khem itself and from me with less patched toolchain) does boot Aug 24 14:23:16 hopefully Jay7 and bluelightning will have time to test in the weekend Aug 24 14:23:23 ant_work: should do Aug 24 14:24:42 dromede: we don't have 3.6 yet in OE, we can test 3.4 anyway Aug 24 14:25:21 I hope/suppose this would also boot until switching to userspace Aug 24 14:25:38 like 3.2 does Aug 24 14:25:39 it probably won't Aug 24 14:25:48 :/ Aug 24 14:25:55 the sharpsl_param bug was introduced in 3.3-rc1 Aug 24 14:26:32 it affects all zaurus pxa2x machines Aug 24 14:26:37 and probably collie Aug 24 14:26:48 that's why i want jay7 to test my patch Aug 24 14:27:46 btw the same kexecboot binary failing on collie in the initramfs does run ok on armv5 zaurus Aug 24 14:28:19 so we reallysuspect miscompiled code, though klibc config file does not enable thumb Aug 24 14:28:54 as far as I understand, the toolchain is defaulting to armv5 (the linker) Aug 24 14:29:18 which gcc are you using? Aug 24 14:34:02 i use several Aug 24 14:34:14 mostly the one that comes with yocto Aug 24 14:34:27 an old codesourcery gcc4.1.1 based Aug 24 14:35:07 and open2x toolchain for gp2x for some emulators Aug 24 14:35:39 actually the one from codesourcery is only used for compiling mplayer Aug 24 14:36:23 gcc 4.7.2 cant compile iwmmxt code in mplayer Aug 24 14:45:31 we are exactloy using 4.7.2 Aug 24 14:46:30 (4.7.1 + patches) Aug 24 14:49:31 dromede: about params, is this code still in 3.6? Aug 24 14:49:32 http://lists.infradead.org/pipermail/linux-arm-kernel/2010-July/019691.html Aug 24 14:49:47 I think it has been removed in 3. iirc Aug 24 14:51:06 I've found that redboot keeps partitions info on flash Aug 24 14:51:16 maybe this could be adapted Aug 24 14:51:18 http://lxr.free-electrons.com/source/drivers/mtd/redboot.c Aug 24 14:51:42 i think it still in mainline Aug 24 14:52:07 i see it in menuconfig under "kernel features" Aug 24 14:53:08 this links to the piro's bootloader code I showed yesterday Aug 24 14:53:24 expecting params at 0xa00000 smthg Aug 24 14:53:41 on the other side, the redboot approach is much better Aug 24 14:54:10 circumventing the bootloader Aug 24 14:54:37 we could even read the lcd params you're reading copied in ram Aug 24 14:54:54 and machine id ofc Aug 24 16:05:57 well, yes, there are already mtd partition parser to start from Aug 24 16:06:25 i.e. http://lxr.free-electrons.com/source/drivers/mtd/bcm63xxpart.c#L217 Aug 24 16:06:41 or http://lxr.free-electrons.com/source/drivers/mtd/ar7part.c#L137 Aug 24 16:07:11 redboot is more complicated Aug 24 16:07:50 problem is, I have never written a kernel module .... Aug 24 16:10:44 * ant_work comes again to /include/linux/mtd/partitions.h Aug 24 16:11:14 noticing the two mtd_add_partition and mtd_del_partition hooks for the BLKPG ioctl... Aug 24 16:11:56 I'd like to know in advance which option is less time-expensive to implement :) Aug 24 16:13:11 1) mtd_part parser 2) BLKPG_(ADD!DEL)_PARTITION for fake runtime repartitioning **** ENDING LOGGING AT Sat Aug 25 03:00:04 2012 **** BEGIN LOGGING AT Sat Aug 25 03:00:05 2012 Aug 25 21:44:25 so, any good news about collie? **** ENDING LOGGING AT Sun Aug 26 03:00:00 2012 **** BEGIN LOGGING AT Sun Aug 26 03:00:00 2012 **** ENDING LOGGING AT Mon Aug 27 02:59:57 2012 **** BEGIN LOGGING AT Mon Aug 27 02:59:58 2012 Aug 27 14:28:41 jay7: had any time to test the kernels i sent you? Aug 27 16:11:58 dromede: sorry, no Aug 27 16:12:01 still busy **** ENDING LOGGING AT Tue Aug 28 03:00:03 2012