**** BEGIN LOGGING AT Wed Nov 07 03:00:02 2012 Nov 07 12:56:12 GNUtoo: how to make a git local commit/format-patch in shr-chroot for linux-openmoko ? Nov 07 12:56:35 ie I sent raw patches as I failed to do the former Nov 07 13:06:51 GNUtoo: nvm I found my previous local clone of linux-openmoko in shr-chroot ... should do fine . Got checks to do to diff it vs the workdir version from bitbake to check I did not miss anything Nov 07 13:33:35 prahal2, hi Nov 07 13:34:11 prahal2, try to clone linux kernel in shr git Nov 07 13:34:17 hi. Nov 07 13:34:39 http://git.shr-project.org/git/?p=linux.git;a=summary Nov 07 13:35:24 om-gta02/2.6.39/master is the branch to use Nov 07 13:35:29 sorry for the confusion . I just found the clone of the kernel I had forgotten about .) Nov 07 13:35:41 ok Nov 07 13:37:06 hum mine is upstream github Nov 07 13:37:13 prahal2, I think it works too Nov 07 13:37:16 I believe JaMa told me so back then Nov 07 13:37:25 note that all patches in git format applied fine here Nov 07 13:37:34 your diff applied fine too Nov 07 13:37:43 so I just need signed-off-by Nov 07 13:37:57 and a patch in git format for the diff Nov 07 13:38:01 with signed-off too Nov 07 13:39:41 agreed but it is easer for me to have all patches in an up to date single clone (so if I forget something it is a matter of "git log" to get a picture) Nov 07 13:40:02 ok Nov 07 13:40:39 so I will do any diffing with the bitbake devshell version ... will take a little more time but in the long run it will save time Nov 07 13:41:18 ok Nov 07 13:41:54 prahal2, note that I tested your patches, they compile and I didn't spot issues yet Nov 07 13:42:16 (GSM still works, suspend appear to work etc...) Nov 07 13:44:52 indeed the "avoid spinlock in interrupt context" patch is merely working around the samsung gpio api , ie I cut n paste the lowlevel non spinlocked calls . So not real new code nor features . Only reordering of code Nov 07 13:46:43 ok Nov 07 15:08:46 JaMa, hi Nov 07 15:11:52 hi Nov 07 15:12:08 signed-off-by added to http://wiki.shr-project.org/trac/ticket/2065 Nov 07 15:21:54 JaMa, how long is boot under GTA02 with 3.2? Nov 07 15:22:29 and under what filesystem also(ubifs, jffs2 or ext3/4) Nov 07 15:22:52 GNUtoo: http://build.shr-project.org/tests/jama/bootchart/systemd-20120525/om-gta02-3.2.18.a.svg Nov 07 15:22:59 ext3 Nov 07 15:25:38 I suspect ubifs threads for my issue Nov 07 15:26:00 now with linux-next it boots in about 3min30 from the point where the kernel loads init Nov 07 15:26:09 init beeing systemd Nov 07 15:26:46 here you have 25s Nov 07 15:26:49 generating ssh keys or already generated? Nov 07 15:26:50 there is a huge gap Nov 07 15:26:55 already generated Nov 07 15:27:05 http://build.shr-project.org/tests/jama/bootchart/systemd-20120525/om-gta02-3.2.18.a.systemd-analyze-blame.txt Nov 07 15:27:23 ok I'll try to login Nov 07 15:27:33 I ran htop when logged Nov 07 15:27:40 and I saw no ecessive consumation Nov 07 15:27:45 maybe hdd access is slow Nov 07 15:27:48 s/hdd/flash/ Nov 07 15:27:49 GNUtoo meant: maybe flash access is slow Nov 07 15:27:55 btw do you have image from latest shr branches? Nov 07 15:28:01 no Nov 07 15:28:11 here it has issues with systemd-udev on gta04 Nov 07 15:28:16 here too Nov 07 15:28:26 I guess it's fixed in jansa branch Nov 07 15:28:34 there was a commit about that Nov 07 15:28:42 like udev-extraconf Nov 07 15:28:49 what are your issues? Nov 07 15:28:59 because there is 3 or 4 failling stuff Nov 07 15:29:05 one is as usual libsamsung-ipc Nov 07 15:29:11 I must fix fsogsmd Nov 07 15:29:53 GNUtoo: no that's not for that Nov 07 15:30:13 ok Nov 07 15:30:17 GNUtoo: it's because of https://bugzilla.yoctoproject.org/show_bug.cgi?id=3390 Nov 07 15:30:21 but is not enough Nov 07 15:30:41 ah ok Nov 07 15:37:25 [ 93.320000] UBIFS: deferred recovery completed is suspicious Nov 07 15:37:35 after that it get faster Nov 07 15:37:38 but not fast enough Nov 07 15:37:56 and most of the times shell don't appear Nov 07 15:38:03 shr 20120224 om-gta02 ttySAC2 Nov 07 15:38:05 and nothing Nov 07 15:39:41 cannot you test on uSD to eliminate ubifs changes? Nov 07 16:21:09 JaMa, systemd-analyze gives that with 2.6.39: Nov 07 16:21:12 dbus.exceptions.DBusException: org.freedesktop.DBus.Error.ServiceUnknown: The name org.freedesktop.systemd1 was not provided by any .service files Nov 07 16:21:39 weird, I haven't seen this Nov 07 16:21:43 do you have timeout beforehand ? Nov 07 16:21:55 but I was also using older systemd when doing those tests Nov 07 16:22:06 it's with shr-2012.01 Nov 07 16:22:31 2012.01 was without systemd iirc Nov 07 16:22:52 ah indeed Nov 07 16:22:55 and it's slow too Nov 07 16:23:04 with linux-next? Nov 07 16:23:09 yes Nov 07 16:23:17 on uSD? Nov 07 16:23:31 ubifs Nov 07 16:23:38 I'll try jffs2 Nov 07 16:55:35 * JaMa just got SIGILL with latest image on om-gta02 too Nov 07 16:55:42 but not from systemd/systemd-udev Nov 07 16:55:56 ok Nov 07 16:57:35 and also wizard crash on gta04, but not from shr-wizard afaik, and on 2nd try it works Nov 07 16:59:49 JaMa, do you have an initramfs image ready somewhere? Nov 07 17:00:05 initramfs image? Nov 07 17:00:11 yes Nov 07 17:00:46 for instance the fso image in cpio.gz format Nov 07 17:02:50 then no.. I have only smaller initramfs image with kexecboot Nov 07 18:41:41 JaMa, for instance bash is broken Nov 07 18:41:48 with the newest kernels Nov 07 18:41:55 you do init=/bin/bash Nov 07 18:41:57 then ls Nov 07 18:42:05 it lists flies Nov 07 18:42:08 and blocks at the end Nov 07 18:42:26 after group_exit(0) of ls Nov 07 18:46:57 I tried that in chroot and it works Nov 07 18:47:08 I've found issue in mount.sh from udev-extraconf Nov 07 18:47:47 broken only in images without coreutils Nov 07 19:03:58 heyho Nov 07 19:26:04 Oh my, jffs2 doesn't allow "cp -p" and hence debootstrapping fails. Nov 07 19:27:29 (trying to debootstrap debian unstable armhf to gta04's nand) Nov 07 19:28:26 Wrapper script, heh: if [ $1 = "-p" ]; then shift; fi; cp.bin "$@" Nov 07 19:34:39 morphis, hi Nov 07 19:34:57 radekp: btw, opendns.org hijacks nxdomain, which google's 8.8.8.8 free public anycast (different servers in different world locations, the nearest one is choosen by BGP) doesn't. Nov 07 19:36:26 PaulFertser: hi, so you think opendns is better choice? Nov 07 19:37:39 PaulFertser: btw as for the jffs2 and debian - jffs2 does not support mmap which dpkg is using, so jffs2 is not very good choice for debian Nov 07 19:39:07 radekp: no, i think opendns is worse than 8.8.8.8 Nov 07 19:39:18 radekp: but it seems to be running on jffs2 even without mmap Nov 07 19:39:56 PaulFertser: yup it runs, but dpkg is not working - but it can be workarounded with tmpfs mount in /var/cache/apt Nov 07 19:40:40 PaulFertser: i was using it in QtMoko before i switched to ubifs, no idea how hw validation distro handles this... Nov 07 19:41:53 radekp: for me it seems to be working without any tricks Nov 07 19:42:39 hmm strange - maybe they fixed it in debian - IIRC last debian i was using on jffs2 was lenny Nov 07 19:43:02 I want to try to have compatibility with the hw validation kernel Nov 07 19:43:28 PaulFertser: why not just let dhcp configure dns? Nov 07 19:45:37 lindi-: i guess that config we're talking about is meant to be used only for statically configured usb networking so not to force the end-user to run dnsmasq on that interface. Nov 07 19:45:43 Or dhcpd Nov 07 19:46:17 PaulFertser: hmm, I just find it bad that google knows all hosts I communicate with Nov 07 19:46:44 lindi-: do you run your own bind validating dnssec along the way? Nov 07 19:46:56 I do find it bad too, of course. Nov 07 19:47:02 PaulFertser: at home I use my ISP's DNS server Nov 07 19:47:25 PaulFertser: at work we run our own bind and I think it has dnssec enabled nowadays Nov 07 19:47:58 lindi-: btw, have you heard about DANE already? Nov 07 19:48:09 PaulFertser: nope Nov 07 19:49:06 fwiw: I trust google a bit more then my ISP (which can see dns trafic too if they really want), at least google has much more data and I'll be statistically insignificant Nov 07 19:49:35 JaMa: well since I am a customer of my ISP and they are in the same country I have much more leverage against them Nov 07 19:50:01 I don't think I can complain about some gratis service offered by google outside my country Nov 07 19:50:58 ok fair point, my contract with ISP doesn't say anyting about DNS logging etc.. Nov 07 19:51:13 lindi-: that stuff looks like a nice way to kick out those dirty CAs like verisign. Nov 07 19:52:25 PaulFertser: radekp: afaik jffs2 got some limited mmap support Nov 07 19:52:44 read-only Nov 07 19:54:01 but it's true that dpkg didn't work on jffs2 before (I was also using tmpfs or uSD for /var/cache/apt) Nov 07 20:05:14 Hm, gta04's u-boot fails to expose ACM console over usb :( Nov 07 20:08:38 PaulFertser, you mean that it supports ACM but has a bug? Nov 07 20:08:48 because I don't think it has support for ACM Nov 07 20:09:07 GNUtoo: what does it support at all then? Nov 07 20:09:33 serial Nov 07 20:09:40 and boot.scr Nov 07 20:10:05 meh :/ Nov 07 20:27:37 Hm, apparently my nand is kinda bad, fw_printenv/fw_setenv fails sometimes Nov 07 20:35:29 :q Nov 07 20:35:30 oops Nov 07 20:36:49 [ 298.027496] uncorrectable error : uncorrectable error : Nov 07 20:37:32 ouch Nov 07 20:37:38 what kernel? Nov 07 20:37:42 compiled it yourself? Nov 07 20:37:50 Nope, that's radekp's Nov 07 20:37:53 ok Nov 07 20:38:06 what happened with SHR and replicant kernel is that NAND was bad once Nov 07 20:38:12 and good the other boot Nov 07 20:38:16 so it's not reliable Nov 07 20:38:28 one time you boot and it's reliable and you get no errors Nov 07 20:38:35 and the other time it doesn't even boot Nov 07 20:39:11 PaulFertser: well i have given up on NAND on GTA04 - it always showed errors like this Nov 07 20:39:33 PaulFertser: with ubifs i could boot like 1/3 attempts Nov 07 20:39:51 PaulFertser: fortunately SD card works perfect, so i am using just SD card Nov 07 20:40:52 radekp: eh, but eh, that shouldn't be that way you know... Nov 07 20:41:07 What does Neil say about it? Nov 07 20:41:22 PaulFertser: yup, i think he uses SD card too... Nov 07 20:41:25 I complained about NAND, nobody cared Nov 07 20:42:03 :((( Nov 07 20:42:04 PaulFertser: i dont have any sane reason to use NAND anyways - so it's very low priority... Nov 07 20:42:40 PaulFertser: at least for me - i have 32GB class 10SD card, which beats NAND capacity and speed Nov 07 20:43:48 PaulFertser: i wonder if this is problem of 3.x kernels - because HW validation works good for me, at least i havent hit any problem yet with it Nov 07 20:43:57 radekp: i thought of nand as an easy backup boot solution but probably you're right. I guess they should somehow warn folks nand is not working. Nov 07 20:44:46 radekp: do you mean it's Neil who's doing kernel work now but somebody else who maintains hw validation kernel? Nov 07 20:44:55 PaulFertser: with HW validation it's quite good backup boot solution, you can repair/format SD card from there Nov 07 20:45:27 PaulFertser: Neil Brown does 3.x kernel, 2.6.32 (hw validation) is done by Nikolaus Nov 07 20:46:10 i'd much more appreciate to get power management fixed so that i dont end up every second day with discharged phone :( Nov 07 20:46:56 but anything kernel related (except make menuconfig) is beyond my skills... Nov 07 20:49:47 Damn, now i need to prepare the uSD according to that twisted special OMAP requirements to boot from it to fix my u-boot, lol Nov 07 20:51:15 radekp: i made my hw validation unbootable while upgrading Debian to the latest and then i tried to fix it by mounting from qtmoko and it faced nand errors so i deboostrapped new Debian the way i liked and then i tried to boot it but the kernel output was directed to serial so i started modifying u-boot env and then i was screwed Nov 07 20:52:09 PaulFertser: hmm the special uSD should not be a big problem Nov 07 20:52:52 PaulFertser, do you have the serial cable? Nov 07 20:52:59 i am keeping the default distro in NAND, but booted it only a few times Nov 07 20:53:35 Well, it's not a big problem but another hassle, i somehow happen to not have spare uSDs laying around so i need to save my working qtmoko install somewhere and then i might happen to have no space on laptop's harddrive etc etc. Yes, i'm doing it wrong :/ Nov 07 20:54:06 GNUtoo-x60: the debug board? No but i have quite some equipment which accepts 3.3V UART just fine. Nov 07 20:54:20 GNUtoo-x60: btw are you using _OPROFILE=2 now? Nov 07 20:54:37 no Nov 07 20:54:42 _OPROFILE=3 Nov 07 20:54:51 PaulFertser, no, the gta04 serial cable Nov 07 20:54:59 PaulFertser: i have 4 working rootfs on my uSD card now :) Nov 07 20:55:04 the connector is different (better?) from the gta02 Nov 07 20:55:11 GNUtoo-x60: hm, ok Nov 07 20:55:31 GNUtoo-x60: does it work good? ie no echo and loud sound? Nov 07 20:55:34 radekp: is there qtmoko release without gta04-init now? Nov 07 20:55:48 radekp: am i right in assuming that if i didn't know about that special OMAP requirements when i was creating the card i'd need to recreate all the partitions anyway? Nov 07 20:55:56 radekp, look at my commit in cornucopia Nov 07 20:56:00 it explains everything Nov 07 20:56:07 and no it doesn't removes all echos Nov 07 20:56:13 only one kind of echo Nov 07 20:56:26 JaMa: not yet, i didnt have time to look at it yet, but i still find the gta04-init much superior to the new uboot Nov 07 20:56:51 ok Nov 07 20:56:53 GNUtoo-x60: oki, thanks Nov 07 20:57:32 http://git.freesmartphone.org/?p=cornucopia.git;a=commit;h=098d6811b99f4628414ffb4cf6091e6b2ac00ab1 Nov 07 20:58:10 PaulFertser: hmm if your first partition is small FAT it could work, but i am not sure if it will work Nov 07 20:58:25 radekp: gta04-init was removed from SHR kernel by slyon so I was testing new uboot and you're right gta04-init was superior, but uboot works for me too Nov 07 20:58:38 radekp: do you happen to know where the requirements come from? http://omappedia.org/wiki/SD_Configuration sorta sucks in explaining _what_ is needed and why. Nov 07 20:59:43 GNUtoo-x60: (eliminating local echo) cool stuff! Nov 07 21:00:55 now you don't ear the echo anymore but....the remote party still ears an echo Nov 07 21:03:01 GNUtoo-x60: i'd very much expect it to have the same cause as on gta02: acoustic coupling due to the front panel. Nov 07 21:03:26 ok Nov 07 21:04:15 PaulFertser: i think i have followed this http://elinux.org/BeagleBoardBeginners#Formatting_the_SD_card_via_fdisk_.22Expert_mode.22 when fomating the card Nov 07 21:05:43 Yes, same, no reference at all. Nov 07 21:05:45 JaMa: i think the biggest problem with uboot is that you cant easily reboot to the partition you are on Nov 07 21:06:49 yes and chrooting to any /distro/foo was easier to try some new distro then extra partition Nov 07 21:07:43 but with new uboot by default on new gta04 it's weird to have both Nov 07 21:09:11 JaMa: sure, i will try to finish release now and then i hope i will have some time to try it.. Nov 07 21:10:53 https://wiki.linaro.org/WorkingGroups/Kernel/Projects/FlashCardSurvey?action=show&redirect=WorkingGroups%2FKernelConsolidation%2FProjects%2FFlashCardSurvey suggests the card should better not be repartitioned at all... Nov 07 21:11:05 One of the reasons i would prefer internal NAND Nov 07 21:11:51 Night, folks, and thanks! Nov 07 21:12:01 * radekp goes to bed too, cu **** ENDING LOGGING AT Thu Nov 08 03:00:01 2012