**** BEGIN LOGGING AT Sat Oct 13 03:00:02 2012 Oct 13 06:42:47 AndrewX192: ive got ovi maps and don't remember how i got them, just that it most certainly was not via pc Oct 13 06:44:41 oh, it was about the map tiles Oct 13 06:44:51 sorry Oct 13 09:13:13 merlin1991, I updated that package fiasco-image-update-ask because busybox tr and sed applet is buggy Oct 13 09:14:17 I really don't get it why vibra should be of class input in 3.5.4 :( Oct 13 09:14:58 Pali: wanna patches? Oct 13 09:15:30 freemangordon, vibra is still in /drivers/input?? Oct 13 09:15:35 I'd like to see Oct 13 09:15:37 yes :D Oct 13 09:16:03 Pali: vibra is in /sys/class/input Oct 13 09:16:13 on stock it is in /sys/class/leds Oct 13 09:16:26 ah, on maemo kernel it is in leds :D Oct 13 09:16:32 and in upstream in input :D Oct 13 09:16:36 nice :D:D Oct 13 09:17:24 somebody should send patch which move vibra to filesystem subtree :-) Oct 13 09:17:32 freemangordon, I have deb package which patch fiasco flashing to ask user if want to change kernel Oct 13 09:17:40 package is now ready for testing Oct 13 09:19:55 see: http://talk.maemo.org/showthread.php?p=1279967#post1279967 Oct 13 09:20:00 Pali: http://merlin1991.at/~freemangordon/kernel3.5/patches.tar.gz Oct 13 09:20:37 Pali: i really cannot imagine what tipe of input is vibra Oct 13 09:20:40 *type Oct 13 09:21:15 freemangordon, in maemo kernel IR TX is in input too :D Oct 13 09:21:38 hehe Oct 13 09:21:48 this should be in leds if you ask me Oct 13 09:22:37 Pali: if you try those patches, have in mind you'll need to rename startup-pin-query, I don;t know why it hangs Oct 13 09:23:03 freemangordon, are you going to create some git repo for it? Oct 13 09:23:13 maybe after all ssi driver is not compatible with maemo, though I think I got it running once Oct 13 09:23:21 Pali: should I Oct 13 09:23:34 I am only playing with it for the sake om the play Oct 13 09:23:38 *of Oct 13 09:23:45 ok Oct 13 09:24:02 even if i have 3.5.4 perfectly running with maemo, what to do with it? Oct 13 09:24:49 though if i have an answer to the above question, maybe I can do some serious work Oct 13 09:25:13 and bme? Oct 13 09:25:21 what about it? Oct 13 09:25:50 did you deleted it? Oct 13 09:26:01 apt-get uninstall :D Oct 13 09:26:38 and what other did you changed/deleted for upstream kernelM Oct 13 09:26:39 ? Oct 13 09:27:11 used your script, and renamed startup-pin-query and rtcomm-phone-ui Oct 13 09:27:17 that is Oct 13 09:27:23 ok Oct 13 09:27:27 aaah Oct 13 09:28:03 and in /etc/event.../scg i commented some framebuffer trickery Oct 13 09:28:09 *sgx Oct 13 09:28:36 maybe you can create wiki, how to do it... Oct 13 09:28:51 but now when booted, mce and 2 hal pugins use 100% of CPU Oct 13 09:28:59 Pali: what for? Oct 13 09:29:28 all changes in maemo system Oct 13 09:29:31 who and why will use that? Oct 13 09:29:46 I can look at it later for better support Oct 13 09:30:41 Pali: ok, but besides POC what is the benefit of that? Oct 13 09:31:08 I mean, are we going to push 3x kenel in CSSU? I doubtt. Oct 13 09:31:25 for upstreaming charger driver I need to test it on upstream kernel Oct 13 09:31:43 so upstream kernel on maemo will be good for testing... Oct 13 09:31:45 aah, i see Oct 13 09:31:56 we can push upstream kernel to extras-devel Oct 13 09:32:11 well, the patches are enough for it, you have maemo booting Oct 13 09:32:26 Pali: come on, it needs tons of work Oct 13 09:32:45 and I am not sure you can compile it with gcc4.2.1 Oct 13 09:32:55 I am using my thumb toolchain Oct 13 09:33:11 I have on ubuntu: gcc version 4.6.3 (Ubuntu/Linaro 4.6.3-1ubuntu5) Oct 13 09:33:19 (arm cross) Oct 13 09:33:37 yes, but pushing it in extras needs autobuilder approval :D Oct 13 09:33:50 yes :-) Oct 13 09:34:11 and iirc even meego kernel refused to build with 4.2.1 Oct 13 09:34:33 2.6.37 or something Oct 13 09:34:36 what about to create package of new toolchain and push it to extras-devel? Oct 13 09:34:46 just imagine how will 3.5 complain Oct 13 09:34:55 under some other name Oct 13 09:35:24 Pali: to compile on the device? Oct 13 09:35:38 no, to compile it under autobuilder Oct 13 09:35:40 or for scratchbox, i don;t get the idea Oct 13 09:35:44 aah Oct 13 09:36:06 all toolchain executables will have different name (version prefix - name of app) Oct 13 09:36:12 well, there is tgz Oct 13 09:36:36 and if you call that executables with correct name, scratchbox will not rewrite it for internal compiler Oct 13 09:37:04 there is only one problem, compiler in extras-devel in arm must be native compiler Oct 13 09:37:16 so compilation will be very slow Oct 13 09:37:26 ... but who cares in autobuilder? :-) Oct 13 09:37:29 aah, i see. Oct 13 09:38:20 but that means libstdc++ and libgcc in extras too Oct 13 09:38:33 and that is CSSU suff ;) Oct 13 09:38:43 yes... Oct 13 09:38:52 what about static linked compiler? Oct 13 09:39:11 you need those on the device Oct 13 09:39:27 kernel does not need libgcc libc or std++ Oct 13 09:40:03 Pali: this is massive amount of work I still wail to see the benefit of Oct 13 09:40:10 *fail Oct 13 09:40:13 ok :-) Oct 13 09:40:47 btw, can you test fiasco-image-update-ask deb? Oct 13 09:40:48 Pali: have in mind ompa PM in upstream is baaad Oct 13 09:41:01 if exists at all Oct 13 09:41:18 Pali: what I need to do to test it? Oct 13 09:41:56 install that package and then start to install/reinstall -flasher packages Oct 13 09:42:28 when you change kernel (e.g stock --> kernel-power or kernel --> uboot) it should ask you if you wan to flash it Oct 13 09:42:50 and if you going to upgrade (kp51 --> kp52) it should not ask Oct 13 09:43:05 it checking version in /dev/mtd and new version Oct 13 09:43:26 hmm Oct 13 09:43:38 it can extract version from zImage or u-boot or zImage attached in u-boot Oct 13 09:43:50 yeah, got it Oct 13 09:44:00 (now it ignore attached zImage in u-boot) Oct 13 09:44:26 i'd rather leave that to someone else :) Oct 13 09:44:49 ok Oct 13 09:45:31 and I have prapared package kernel-bootimg which contains zImage of stock kernel in /boot/ and item file in /etc/bootmenu.d/ Oct 13 09:45:43 which is needed for building u-boot Oct 13 09:46:15 I will push it to extras-devel Oct 13 09:46:31 ok Oct 13 09:47:22 Pali: any development with romaxa's work? Oct 13 09:47:48 what do you mean? Oct 13 09:48:16 hmm, i thought it will tweak browser rotation Oct 13 09:48:23 s/it/he/ Oct 13 09:48:23 freemangordon meant: hmm, i thought he will tweak browser rotation Oct 13 09:48:43 probably misunderstood something Oct 13 09:49:53 Pali: I am getting "Paly away" messages Oct 13 09:49:59 *Pali Oct 13 11:18:04 greeting Oct 13 11:19:47 I just got myself a Samsung Note 2 mobile. I copied all contacts using csv file transferred by bluetooth. Now I want to transfer the messages. Is there a way to do this? Oct 13 11:24:53 you traitor! Oct 13 11:26:31 smhar: hardly Oct 13 11:27:50 unlike contacts, msgs are generally not considered worthy for ex/import it seems Oct 13 11:28:57 you probably could extract all SMS and IM msgs from events_v1.db Oct 13 11:29:17 with a sqlite frontend Oct 13 11:29:20 kerio, I am not getting rid of my N900, I can not do this. it is just that it is showing its age, in terms of hardware, that I can not depend on it fully for work. Oct 13 11:30:39 kerio, I had the option between iphone and galaxy note and I chose the least evil :-) Oct 13 11:31:42 * kerio doubts the n900 isn't dependable for "work" that's not a physics simulator or something equally silly Oct 13 11:34:27 kerio, I did not say it is not dependable for work. I have been using it for work and personal use for ages. It is just that it has several deep scratches that makes reading hard. Also, its readability in sun -as my work requires- is not very good. Oct 13 11:35:22 kerio, I also need a portable dwg file viewer that is not available in N900 Oct 13 11:36:31 i thought there weren't many transreflective screens for mobiles, but ok Oct 13 11:36:42 having to open autocad projects is a valid concern, i s'pose Oct 13 11:39:01 though Oct 13 11:39:11 opening autocad projects on a mobile? Oct 13 11:39:44 I thought a workstation would be more practical but I wonder Oct 13 11:39:56 any use case for that? Oct 13 11:40:03 FlameReaper-nemo, viewing them, not editing them. it is very feasible in a note 2 size mobile Oct 13 11:40:23 ah Oct 13 11:40:50 FlameReaper-nemo, I need to have a copy of the drawing with me and to check dimensions of items while walking in the site Oct 13 11:41:24 that would be reasonable Oct 13 11:56:42 [general notice] Maemo Council nomination period extended until 28. of this month, PLEASE CONSIDER self-nomination, it doesn't need any super-powers Oct 13 12:13:53 why cant I play music when on a call? Oct 13 12:14:13 I thought pulse audio could do that sort of thing? Oct 13 12:14:31 even symbian could! Oct 13 12:16:33 pulseaudio most definetely can Oct 13 12:16:40 and pulseaudio most definetely doesn't want to Oct 13 12:16:53 because it hates you Oct 13 12:17:38 so why cant I play music when on a call? Oct 13 12:21:30 because pulseaudio hates you Oct 13 12:21:38 or maybe because of some weird nokia policy, ask Pali Oct 13 12:22:21 vi__, you need to patch prolog code which disabling it Oct 13 12:23:14 Pali: does it disable every other sound? Oct 13 12:24:23 kerio, I do not know Oct 13 12:24:49 vi__, decompiled prolog source code is here: https://gitorious.org/community-ssu/policy-settings-rx51/trees/master/rx51/policy Oct 13 12:33:44 Pali: there is prolog used somewhere? :P Oct 13 12:34:00 ZogG_laptop, yes Oct 13 12:34:04 lol Oct 13 12:34:08 auto routing, linux cgroups, ... Oct 13 12:34:27 i had it this semester (visual prolog) - it's wierd language i would say Oct 13 12:34:29 alsa/pulseaudio/phone call policy Oct 13 12:36:14 I had prolog last semester too :-) and this is why I was able to decompile it and fix problems Oct 13 12:36:33 like headset button worked only when phone call was active Oct 13 12:36:53 i missed most lessons and have no idea what's going on with it. still need to make project for that course (pipe ania game) Oct 13 12:37:36 prolog is good for logical programming Oct 13 12:38:25 woot Oct 13 12:38:54 cool detail :) Oct 13 12:42:06 Pali: make it so I can play music while on a call and I will buy you an pizza. Oct 13 13:01:56 damn Prolog. Couldn't they use inuktitut instead of Prolog? So at least google-translate would *maybe* convert it to something meaningful for me Oct 13 13:04:16 prolol. Oct 13 13:05:09 trolog Oct 13 13:05:48 oh yeah, thats better Oct 13 13:05:49 trollog Oct 13 13:05:51 i wonder if the n900 has enough resources to emulate the DS properly Oct 13 13:06:23 DS? Oct 13 13:06:27 nintendo ds Oct 13 13:06:42 nfc Oct 13 13:06:49 67.028MHz arm9 and 33.514MHz arm7 Oct 13 13:06:58 and then some kind of graphical chip Oct 13 13:20:17 while CPU emu is rarely ever a problem, emulating GFX accel is next to impossible in realtime Oct 13 13:21:43 depending on architecture/properties a gfx accel performs 10s to 1000s of times better than a similarly 'fast' CPU on same task Oct 13 13:24:37 and the problem with emulating GFX accel is: the app and libs it uses are all optimized to the very special properties of the gfx accel, so the CPU emulating it has to do rather silly things to emulate some function that helps accel to perform better but costs CPU additional cycles to emulate, while you hardly can optimize much for the real video driven by the CPU Oct 13 13:26:02 DocScrutinizer05: this, assuming you want to emulate the GPU exactly, mame-style Oct 13 13:26:23 it's the reason mame sucks balls for 3d games :) Oct 13 13:28:53 huh, there *is* desmume for the n900 Oct 13 13:34:46 haha i get a whopping 2fps Oct 13 13:35:27 Pali: ZogG_laptop: any URL/pointer for me to read a 30min "Prolog for dummies"? Those "sources" Pali liked to, above, are like hieroglyphs to me Oct 13 13:40:06 DocScrutinizer05: don't go there! that's the road to insanity! Oct 13 13:43:47 DocScrutinizer05: look at http://www.csupomona.edu/~jrfisher/www/prolog_tutorial/contents.html or http://kti.ms.mff.cuni.cz/~bartak/prolog/ Oct 13 14:02:19 EEEW I feel nausea after first 10 lines of code on fist page of introduction: http://www.csupomona.edu/~jrfisher/www/prolog_tutorial/1.html. .oO("WTF comes in 'D'???") Oct 13 14:05:59 and the insanity of closing a large structure (like an "IF" statement) with an almost invisible "." already drove me and all my colleagues mad in early 80s of last century, when we used to substitute "." by ". NOTE dot." in COBOL Oct 13 14:08:10 and COBOL been derided as "compilable prosa" even back then Oct 13 14:08:49 so Prolog seems to be "compilable dadaism poems" Oct 13 14:14:52 aaah, I see >> At the same time, early 1970's, Colmerauer and his group at the University of Marseille-Aix developed a specialized theorem prover... << now that contains two key factoids for me: University of Marseille-Aix, and 1970. Oct 13 14:16:15 adding to that the generally somewhat inverse grammatics of French language (and thinking, it seems), makes for a perfect first characterization of Prolog's origin Oct 13 14:18:02 DocScrutinizer05 in prolog char ',' is && (and) and char ';' is || (or) Oct 13 14:18:21 OMFG Oct 13 14:18:46 next ',' has higher priority then ';' Oct 13 14:19:02 OK, you win! I give up Oct 13 14:19:33 and prolog has unification Oct 13 14:19:52 toldya Oct 13 14:20:10 there is no assign operator Oct 13 14:20:19 '=' is used for unification Oct 13 14:22:04 >>Thus you can load more programs into the database but be careful whether the programs do not use the procedures with the same name. Otherwise, because of cumulating of clauses, these procedures could behave incorrectly.<< SUUUURE Oct 13 14:27:17 you know, one of the minor functions of my first really huge project (>1 manyear) been to overcome the problems in COBOL that originate from COPY statements (== include ), insane way of COBOL defining var structures (05 structure-a. 10 member-of-05 PIC 99,999 DISPLAY.) and the fact that COPY statements can't work inside COPY elements (an include file can't include further files) Oct 13 14:31:11 so I worked 1 year to create a data dictionary to get proper system wide type definitions (e.g for timestamp, consisting of 20 timestamp. 25 year PIC 9999. 25 time. 30 hour PIC 99. 30 min PIC 99. 30 sec PIC 99) into COBOL Oct 13 14:32:30 well, I forgot month and day-of-month var definitions in above copy element Oct 13 14:33:17 anyway PROLOG seems worse than COBOL Oct 13 14:39:58 there are worse things.. like erlang, reusing prologs syntax for a nearly general purpose language Oct 13 14:43:32 eeek Oct 13 14:43:51 http://www.csis.ul.ie/cobol/course/Copy.htm >>A COPY statement can occur anywhere a character-string or a separator can occur except that a COPY statement must not occur within another COPY statement<< Oct 13 14:45:16 >>The text produced as a result of the complete processing of a COPY statement must not contain a COPY statement. << Oct 13 14:47:03 Pali: i'm bored Oct 13 14:47:13 give me something to test Oct 13 14:48:15 kerio, http://talk.maemo.org/showthread.php?p=1279967#post1279967 Oct 13 14:48:24 kerio: testing is boring. fix bluetooth driver and send me teh diffs. Oct 13 14:48:39 Skry: no u Oct 13 14:48:45 also, what's broken in the bt driver? :o Oct 13 14:49:13 its old Oct 13 14:49:35 Pali: hm, how would i test that? i've got kp52 already Oct 13 14:50:01 kerio, reinstall kernel and kernel-flasher packages Oct 13 14:50:24 or reinstall other -flasher package Oct 13 14:51:00 hrmpf, i don't have the KP52 debs anymore Oct 13 14:51:22 I mean to reinstall stock kernel packages Oct 13 14:51:34 Pali: oooooh Oct 13 14:51:53 you should be able to reject flashing Oct 13 14:52:10 ...weird, kernel doesn't depend on kernel-modules Oct 13 14:52:22 yes Oct 13 14:52:25 ist ok Oct 13 14:52:36 kernel is only flashable image Oct 13 14:52:44 kernel-flasher depends on modules Oct 13 14:53:07 Pali: kernel-flasher doesn't depend on modules either Oct 13 14:53:23 then it is bug Oct 13 14:53:28 which is fixed in kernel-power :-) Oct 13 14:53:31 Pali: i'm talking about the stock stuff Oct 13 14:53:50 lol Oct 13 14:53:54 you know, the kind of stuff that has botched deps due to the fucking metapackage Oct 13 14:54:11 ~lart MP Oct 13 14:54:12 * infobot pulls out his louisville slugger and uses MP's head to break the homerun record Oct 13 14:54:12 anyway, "the version of the new kernel isn't similar to the old one" Oct 13 14:54:18 and i was able to abort the flashing Oct 13 14:54:46 its ok Oct 13 14:55:06 Pali: if you want more testing, release KP52 :P Oct 13 14:55:26 I'm still waiting for some patches Oct 13 14:55:46 Pali: canceling the flash could be disastrous if you're doing a HAM system update Oct 13 14:56:05 * kerio looks at cssu-thumb Oct 13 14:56:21 kerio, canceling flash is not fail Oct 13 14:56:31 Pali: that's the whole point Oct 13 14:57:04 it is ok, because it allow us to install kernel-power-flasher without need to flash image to nand (where can be uboot) Oct 13 14:57:27 Pali: but k-p-f depends on the k-p package Oct 13 14:57:40 we already talked about it and this was ony one solution Oct 13 14:57:46 kerio, it is problem? Oct 13 14:57:54 well, you'll have the fiasco image in /boot wasting space Oct 13 14:57:58 no Oct 13 14:58:01 it is deleted too Oct 13 14:58:07 oh Oct 13 14:58:11 eeeeeeeh Oct 13 14:58:16 i suppose it's fine, then Oct 13 14:58:25 if you cancel flash or not it is deleted by shell script by nokia Oct 13 14:58:40 I only patched that script to ask user Oct 13 14:58:43 as long as you don't cancel the flashing of kernel-cssu as you're installing cssu-thumb XD Oct 13 14:59:11 Pali: btw, why can't uboot boot from a fiasco image of a kernel? Oct 13 14:59:43 because fiasco is proprietary format and nobody have time to implement it in uboot Oct 13 14:59:55 uboot using standard mkimage format Oct 13 15:00:11 * DocScrutinizer05 feels this tweak to basic operation principles and concept of kernel flashing needs some more public discussion Oct 13 15:00:55 DocScrutinizer05, I think we already had discussion about it Oct 13 15:01:21 and there is task on garage: https://garage.maemo.org/pm/task.php?func=detailtask&project_task_id=1979&group_id=2300&group_project_id=3864 Oct 13 15:01:42 date 2012-03-14 Oct 13 15:01:51 Pali: btw, don't you have an uboot to test on weird hw revs? Oct 13 15:02:18 kerio, yes testing version of uboot is already here in chanlog Oct 13 15:02:54 http://atrey.karlin.mff.cuni.cz/~pali/u-boot/u-boot-2012.10-rc3-1+zImage-2.6.28-omap1.fiasco Oct 13 15:03:04 Pali: wasn't aware of that, and probably missed the public discussion when it happened Oct 13 15:03:07 thanks Oct 13 15:03:56 Pali: ugh, do you have one with kernel-power attached? :3 Oct 13 15:04:11 kerio, no Oct 13 15:04:12 i don't have the stock modules and i can't boot with it anyway Oct 13 15:06:48 Pali: however a user query probably isn't the solution to what been the task description. It more feels like trying to solve the problem that resulted in defining that task Oct 13 15:07:48 DocScrutinizer05, and what is better? do not ask user and ignore flashing? Oct 13 15:08:28 if you updating kp (51 to 52) then it not ask for flashing Oct 13 15:08:43 Pali: the rationale behind "we need a way to tell which kernel is installed, so we don't do nasty" seems to be we need to implement that check on a higher level, more early and for more than mere flashing of kernel. We also might need to check same result for flashing of kernel modules for example Oct 13 15:08:45 but if you have stock and going to install kp then it ask Oct 13 15:09:40 DocScrutinizer05, now it extract version from /dev/mtd and compare it with new version in fiasco zImage Oct 13 15:10:16 it can detect version of linux kernel in zImage and version of u-boot Oct 13 15:10:49 hmm, that's all fine for solving the task, as long as we get a function-call that can get used in arbitrary situations (like module installation) Oct 13 15:11:40 already implementing a query into kernel flasher is beyond task 1980 Oct 13 15:13:47 Pali: I'm just saying you over-committed task 1980, your solution should get split into the mere check for kernel version (goes to 1980) plus the actual using of it which shouldn't go there but is some other/new task already Oct 13 15:16:48 I admit the wording of task-1980 been a bit fuzzy Oct 13 15:17:52 the >>...to prevent flashing over user kernels.<< part is a comment aiui, not part of the task Oct 13 15:20:17 particularly it's not a complete definition of the implied task to avoid messing with any kernel related stuff. This includes more than "to prevent flashing over user kernels.", it also needs consideration in other related actions like updating kernel modules, installing stuff that depends on certain kernel properties that we don't have a PROVIDES for, etc Oct 13 15:21:38 that's what I meant about "we might need public discussion" Oct 13 15:24:07 1980 should get a solution that looks like "added binary and source for `kernel version ' which will return 0 if version is met, return -1 otherwise" Oct 13 15:24:35 s/el ve/elve/ Oct 13 15:24:35 DocScrutinizer05 meant: 1980 should get a solution that looks like "added binary and source for `kernelversion ' which will return 0 if version is met, return -1 otherwise" Oct 13 15:34:28 Pali: of course this command `kernelversion` probably needs a little bit more smart features, like comparing kernel version against wildcards/regex, checking for kernel version >= or < something, etc. I haven't thought thoroughly about proper spec for that command Oct 13 15:36:14 anyway result of task1980 should be re-usable by others, without them needing to `RE` your .deb or sourcecode Oct 13 15:38:52 Pali: maybe http://bazaar.launchpad.net/~pali/+junk/maemo_fiasco-image-update-ask/view/head:/zimage-get-version.c is exactly that, but then it needs a README to explain how to use it Oct 13 15:44:15 yeah, actually after having a glimpse at sourcecode, it actually is exactly what I asked for (it seems), just I wasn't able to see it as the README is missing as well as any short statement in https://garage.maemo.org/pm/task.php?func=detailtask&project_task_id=1979&group_id=2300&group_project_id=3864 about how it works Oct 13 15:45:39 (funny how poject_task_id in URL seems pragma_base0 aka RL-ID-- ) Oct 13 15:46:52 nevermind, it seems it actually _is_ task 1979 Oct 13 15:50:47 Pali: I guess I got completely fooled by the title of your git/bazaar branch "maemo_fiasco-image-update-ask" - I took that as being specifically for fiasco-updates which it clearly isn't (ment to be) Oct 13 16:00:07 pondering the whole problem, I wonder if we ever need anything beyond >> if [ "b8db739ea8f2aec4da18ddd550791590" == $( md5sum /dev/mtd3ro ) ]; then...<< Oct 13 16:03:19 >> if [ "b8db739ea8f2aec4da18ddd550791590" == $( md5sum /dev/mtd3ro | cut -d " " -f 1) ]; then...<< actually Oct 13 16:05:26 DocScrutinizer05: if md5sum -c /tmp/oldkernelmd5 --status ; then Oct 13 16:06:00 neat, busybox md5sum has -c and -s Oct 13 16:06:39 DocScrutinizer05: i'm not sure you can read from mtd3ro like that and get a sensible result, though Oct 13 16:06:45 i'd nanddump Oct 13 16:11:03 huh? sure you can Oct 13 16:11:27 ooh, you mean because of bad blocks Oct 13 16:11:31 nice catch Oct 13 16:11:35 :-) Oct 13 16:12:35 yeah, nanddump (or mtd_debug) should solve that problem Oct 13 16:32:03 https://docs.google.com/document/d/1r04yjaQoL1c_vuwQr-CdbDQ8EjGVGOL_1V6BNVWaR1c/preview interesting symbian post-mortem Oct 13 16:32:39 ffs, my IroN900 just needed a reboot to re-establsih WLAN connectivity. Though it claimed wlan0 being up and having proper IP, not a single ping went out to AP Oct 13 16:33:06 :-/ Oct 13 16:33:11 wait, the one with the huge uptime? D: Oct 13 16:33:36 not even going offline mode (airplane mode) and back helped to re-establish proper connectivity Oct 13 16:33:41 nah Oct 13 16:33:49 stop icd start icd? Oct 13 16:34:10 t900:~# uptime Oct 13 16:34:11 18:46:12 up 196 days, 19:41, load average: 0.06, 0.01, 0.00 Oct 13 16:34:45 ifconfig wlan0 down ; rmmod wl1152 (or whatever) ; modprobe wl1152 Oct 13 16:34:47 :P Oct 13 16:35:03 * ShadowJK forgets name of wireless module Oct 13 16:35:09 syslog claimed exactly that happened several times this afternoon Oct 13 16:35:59 it never gets unloaded on n900 afaik Oct 13 16:36:08 though on maemo4 it seems like it does Oct 13 16:36:45 Oct 13 15:27:51 IroN900 kernel: [115214.154632] wl1251: 151 tx blocks at 0x3b788, 35 rx blocks at 0x3a780 Oct 13 16:36:46 Oct 13 15:27:51 IroN900 kernel: [115214.170135] wl1251: firmware booted (Rev 4.0.4.3.7) Oct 13 16:36:48 Oct 13 15:27:52 IroN900 wlancond[1606]: Scan issued Oct 13 16:40:23 that's not full unload Oct 13 16:41:28 [15763.458465] wl1251: unloaded Oct 13 16:41:41 [15767.411132] phy1: Selected rate control algorithm 'minstrel' Oct 13 16:41:53 [15767.416992] wl1251: loaded Oct 13 16:42:03 [15767.417510] wl1251: initialized Oct 13 16:42:30 sudo rmmod wl12xx Oct 13 16:42:42 sudo modprobe wl12xx Oct 13 16:42:49 sudo wl1251-cal Oct 13 16:42:55 sudo stop -q wlancond Oct 13 16:42:57 (after that it requests firmware and loads firmware, 2 times, then it does as DocScrutinizer05 pasted) Oct 13 16:43:00 sudo start -q wlancond Oct 13 16:43:04 sudo stop -q icd2 Oct 13 16:43:09 sudo start -q icd2 Oct 13 16:43:22 ^^^ this should help Oct 13 16:45:23 yeah ive done that sequence too :) Oct 13 16:46:58 ShadowJK: http://maemo.cloud-7.de/share-service/xxx Oct 13 16:47:06 Pali: ^^^ Oct 13 16:48:32 I can't umount my nfs mount point anymore Oct 13 16:48:40 something went wrong apparently Oct 13 16:48:46 umount /nfs/music Oct 13 16:48:49 invalid argument Oct 13 16:48:54 if anybody is able to make sense of that and spot who's the culprit for this borked WLAN association, I'd love to hear Oct 13 16:49:31 eh I always just figured thw wlan chip is buggy and needs total reinitialization twice a year Oct 13 16:49:44 grammoboy: stale nfs locks are a notorious problem Oct 13 16:50:33 DocScrutinizer05, pff, how to get rid of it now? Oct 13 16:56:42 heh Oct 13 16:56:52 "invalid argument" is weird Oct 13 16:57:04 usually it says something about busy Oct 13 16:57:11 if there's something using it Oct 13 16:57:23 or hangs if the server went away Oct 13 17:01:40 ShadowJK: the whole friggin sequence doesn't make too much sense to me Oct 13 17:03:54 nfs really can't deal with network issues gracefully Oct 13 17:04:18 meh, grammoboy is offline anyway Oct 13 17:04:58 but my IroN900 is online now again, after eboot. while going to airplane-mode didn't help. Now what's this telling me? Oct 13 17:05:24 reboots are magical Oct 13 17:06:16 I'd blame PSM, since it exposed some friggin delays of several seconds even before it went down completely Oct 13 17:06:33 (delays in ssh session) Oct 13 17:06:44 o my this really seems to be serious Oct 13 17:06:55 hey wb Oct 13 17:07:01 no way to umount my nfs mount point Oct 13 17:07:46 grammoboy: corrupted fs is known to report any inode that it doesn't know any better about, as being a "stale NFS" Oct 13 17:07:58 usually a fsck can fix that Oct 13 17:08:14 on the local device? Oct 13 17:08:16 eg n900 Oct 13 17:08:19 yep Oct 13 17:08:45 NFS links are obviously a special type of inode on the rootfs Oct 13 17:09:05 (or wherever the supposed NFS been supposedly mounted to) Oct 13 17:09:40 do you know how to run fsck? Oct 13 17:09:54 I had stale NFS mounts on devices that evidently never been using NFS Oct 13 17:10:29 grammoboy: depending on the partition you need to check, it's sometimes tricky Oct 13 17:10:46 rootfs can't get umounted, and fsck won't fix mounted fs Oct 13 17:10:56 try BM Oct 13 17:11:05 DocScrutinizer05, its mounted on /nfs/music Oct 13 17:11:05 I think it has an option to fsck Oct 13 17:11:39 df -h /nfs/ Oct 13 17:11:48 will tell you which fs this really is Oct 13 17:12:20 anyway Oct 13 17:12:22 ~bm Oct 13 17:12:22 somebody said backupmenu was http://talk.maemo.org/showthread.php?t=63975 Oct 13 17:12:32 o/ Oct 13 17:12:39 * DocScrutinizer05 afk Oct 13 17:12:58 i'll look into bm Oct 13 17:13:10 first it needs some charging Oct 13 17:13:15 thx Oct 13 17:17:30 grammoboy: better ask somebody who already used NFS and thus has some experience. My advice probably isn't very good Oct 13 17:24:14 "stale nfs mount" error message is like a generic "wtf filesystem disappeared" error that can occur on non-nfs too Oct 13 17:34:53 ShadowJK: so how would you tell the fs with that "stale fs mount" to goddamn umont it and reset that inode? Oct 13 17:35:29 there has to be some why I'd think, other than running a full fsck against the "corrupted" fs Oct 13 17:35:40 s/why/way/ Oct 13 17:35:40 DocScrutinizer05 meant: there has to be some way I'd think, other than running a full fsck against the "corrupted" fs Oct 13 17:36:10 odds are fsck wouldn't even fix the incriminated inode Oct 13 17:39:22 hmmm, http://joelinoff.com/blog/?p=356 http://www.cyberciti.biz/tips/nfs-stale-file-handle-error-and-solution.html Oct 13 17:40:57 http://www.motherboardpoint.com/fix-stale-nfs-file-handle-t204765.html etc, google for "fix stale nfs mount" Oct 13 17:41:30 grammoboy: ^^^ Oct 13 17:41:47 all I can do to help you out Oct 13 17:51:31 Pali, hy Oct 13 17:51:44 panzersajt_, hi Oct 13 17:51:55 thanks for the new u-boot but it would like to work on my device Oct 13 17:51:56 :( Oct 13 17:52:06 *wouldn't Oct 13 17:53:00 panzersajt_, and is zImage without uboot working? Oct 13 17:53:44 it shows this: Oct 13 17:53:45 USER Oct 13 17:53:52 than a lots of numbers and letters Oct 13 17:53:55 than Oct 13 17:54:06 reboot in 2 1 .. Oct 13 17:54:10 and reboots Oct 13 17:54:17 with your u-boot Oct 13 17:54:47 and without uboot? Oct 13 17:55:53 I don't know Oct 13 17:55:57 but it reboots also Oct 13 17:58:03 but the problem may lie in my device or ignorance Oct 13 17:59:18 grammoboy: http://sysunconfig.net/unixtips/stale_nfs.txt Oct 13 17:59:33 panzersajt_, first check if your device is bootable without uboot Oct 13 17:59:41 if not reflash it Oct 13 18:05:49 grammoboy: looks extremely helpful: http://www.linuxquestions.org/questions/linux-embedded-and-single-board-computer-78/stale-nfs-handle-on-embedded-system-that-does-not-have-nfs-running-936884/ Oct 13 18:18:55 Pali, yes it was bootable before installing u-boot Oct 13 18:19:01 not a big deal Oct 13 18:19:09 I can live without mameo Oct 13 18:19:43 grammoboy: http://www.linuxquestions.org/questions/linux-hardware-18/fsck-inode-block-errors-stale-nfs-file-handle-advice-please-930886/ Oct 13 18:20:16 panzersajt_: WUT? Oct 13 18:23:04 panzersajt_: I guess you flashed uBoot over a non-standard kernel, so now when trying to boot into maemo, uBoot loads the stock kernel that it comes bundled with, and that stock kernel doesn't find the matching kernel modules in your rootfs Oct 13 18:25:50 panzersajt_: you should flash the kernel you used before uBoot, via flasher-3.5 (only kernel, not a whole fiasco image!), then your system should boot again, given my assumption been correct. Then you could install proper stock kernel and matching kernel modules, try if it still boots as supposed, and then install uBoot Oct 13 18:26:42 Pali: ^^^ correct? Oct 13 18:28:47 yes, boot maemo with some working kernel, then install & flash stock kernel (+install modules & bootimages of other kernels) Oct 13 18:28:55 check if working and then install uboot Oct 13 18:31:07 it's a widely ignored or even unknown fact that uBoot comes with a kernel attached Oct 13 18:33:37 Pali: silly idea: is it possible to flash a new kernel on the uboot payload? Oct 13 18:34:10 kerio, no because onenand uboot code not working on n900 Oct 13 18:34:22 no, i mean Oct 13 18:34:40 flash the kernel attached to uboot from fremantle Oct 13 18:34:50 leaving uboot there Oct 13 18:35:13 kerio, you can create combined uboot image on device and you can flash it Oct 13 18:35:17 its possible Oct 13 18:35:58 Pali: could uBoot preinstall create a bzImage.bin kernel image in /boot, from /dev/mtd3? Then uBoot has option to boot that kernel from ubifs rather than chainloading attached kernel from /dev/mtd3. which in turn would free up *lots* of space for uBoot in /dev/mtd3 and thus would allow for including ubifs support and a lot of other nice stuff into uBoot Oct 13 18:36:35 DocScrutinizer05, first we have no access to mtd, Oct 13 18:36:46 and second ubifs code is too big Oct 13 18:36:55 Pali: what the fuck Oct 13 18:37:01 it can't be bigger than the whole linux kernel Oct 13 18:37:21 Pali: if uBoot has whole /dev/mtd3 for itself, then I doubt UBI code is too big Oct 13 18:37:23 but if we delete attached kernel image from uboot, then there is space for ubifs driver :-) Oct 13 18:37:52 DocScrutinizer05, problem is that freemangordon and I was not able to fix onenand code to work on n900 Oct 13 18:38:06 DocScrutinizer05: to be fair it should be optified Oct 13 18:38:19 LOL Oct 13 18:38:22 because it's not needed during the fremantle boot process, it's only needed *before* :P Oct 13 18:38:23 DocScrutinizer05, but now you can boot from eMMC Oct 13 18:38:35 Pali: the emmc is accessible, right? Oct 13 18:38:36 and this is already used Oct 13 18:38:39 yes Oct 13 18:38:42 kerio: you're actually not all wrong ;-) Oct 13 18:38:45 kerio, bootmenu is loaded from MyDocs Oct 13 18:38:47 install bzImage.bin in /home/boot Oct 13 18:39:06 Pali: bootmenu happens after the kernel is loaded, doesn't it Oct 13 18:39:10 there are scripts which doing it automatically Oct 13 18:39:12 kerio: but I'd not like to see kernel on yet a third partiton Oct 13 18:39:14 kerio, no Oct 13 18:39:18 I mean uboot bootmenu Oct 13 18:39:50 DocScrutinizer05: i mean, we already abused of the filesystem hierarchy, /home/boot is not even close to /home/opt in terms of nastiness Oct 13 18:40:46 there is script u-boot-update bootmenu which read config files /etc/bootmenu.d/, generate needed uImage in MyDocs/bootmenuimg.d and create MyDocs/bootmenu.cmd script which appear in bootmenu Oct 13 18:41:01 i see Oct 13 18:41:14 kernel-power-bootimg and kernel-bootimg calling this script (if is installed) after install/remove Oct 13 18:41:35 so i wouldn't actually need the bootmenu kludge in preinit... Oct 13 18:41:41 would i? Oct 13 18:42:16 yes Oct 13 18:42:16 Pali: I think it's been an extremely poor idea to use this very path " /etc/bootmenu.d/" which is also used by the real bootmenu which is called after kernel got loaded Oct 13 18:42:33 they are completely independent and unrelated systems Oct 13 18:42:38 DocScrutinizer05: the idea is that it's compatible with *that* bootmenu Oct 13 18:42:57 that's nonsense though Oct 13 18:43:01 DocScrutinizer05, both has same syntax and both show only relevant entries Oct 13 18:43:13 also multiboot has same file format Oct 13 18:43:37 and you really do not want to have one config file in three different locations Oct 13 18:44:23 who cares about file format. You're free to use whatever file format you like, but placing those uBoot config files into a dir that's aslo used by multiboot and bootmenu is an idea begging for trouble and headache Oct 13 18:44:40 no, /etc/bootmenu.d/ holds bootmenu entries Oct 13 18:45:17 kerio is right Oct 13 18:45:27 yeah, i am Oct 13 18:45:27 there are menu entires Oct 13 18:45:28 ~kerio Oct 13 18:45:29 you are probably [...] kerio is correct [...] Oct 13 18:45:49 sorry, I don't get the whole idea Oct 13 18:46:05 bootmenu itself is a completely different concept Oct 13 18:46:08 DocScrutinizer05: Pali's uboot is backwards-compatible with bootmenu Oct 13 18:46:45 multiboot exploited bootmenu due to the braindamaged concept of multiboot to reflash kernel after kernel booted, then boot again Oct 13 18:47:03 kerio: that's nonsense Oct 13 18:47:37 how can a bootloader that selects kernels to load, be compatible with a menu that starts after kernel got loaded? Oct 13 18:47:59 DocScrutinizer05: by loading the kernel and then following the menu entries' specifications Oct 13 18:48:10 eh???? Oct 13 18:48:35 those menu entries specify a script to run and a rootfs, or something like that Oct 13 18:48:44 it's stuff that *should've* been done by the bootloader Oct 13 18:48:47 I repeat: bootmenu itself is something completely different Oct 13 18:49:01 then why does it work? Oct 13 18:49:05 multiboot exploited bootmenu to emulate what uBoot does Oct 13 18:49:48 what you just described is what multiboot does, not what bootmenu does Oct 13 18:50:21 the only different thing with uboot is that the choice is done before loading the kernel and not after Oct 13 18:50:25 bootmenu is more like backupmenu Oct 13 18:50:46 and multiboot is exploiting that to reboot after flashing a new kernel Oct 13 18:50:47 again, then why does pali's bootmenu work? Oct 13 18:51:13 again, what a question is that? Oct 13 18:51:32 there's no such thing like pali's bootmenu Oct 13 18:52:08 thzere's pali's uBoot that has a uboot menu generated from files in /etc/bootmenu.d/ Oct 13 18:52:24 which may or may not have been placed there by multiboot Oct 13 18:52:56 backupmenu also places a file there, do you suggest uBoot can start backupmenu=? Oct 13 18:53:06 that's what Pali suggests, yes Oct 13 18:53:13 that's nonsense Oct 13 18:53:32 and to be fair, starting backupmenu is just a matter of init=/usr/share/backupmenu/BackupMenuLauncher.item or something similar Oct 13 18:54:56 Pali: can uboot start backupmenu as it is? Oct 13 18:54:57 I repeat, /etc/bootmenu.d/ is for bootmenu which is a userland thing not related in any way to bootloader concept at large Oct 13 18:55:12 kerio: uBoot starts kernels Oct 13 18:55:26 backupmenu is no kernel Oct 13 18:55:39 no, but it's a particular init script Oct 13 18:55:47 so what? Oct 13 18:56:02 btw that's even incorrect Oct 13 18:56:17 since backupmenu is not exactly an initscript Oct 13 18:56:26 it gets invoked by an initscript Oct 13 18:56:55 via taadaaaa bootmenu Oct 13 18:57:17 which in turn gets invoked by a hook in sbin/preinit iirc Oct 13 18:59:05 Pali: if you wanna keep number of config files low and non-redundant, and for that purpose want to re-use multiboot crap in /etc/bootmenu.d/, you could symlink them to /etc/uBoot/* Oct 13 19:00:50 but I keep my notion that using same dir for bootmenu, multiboot, and uBoot is a design flaw Oct 13 19:01:55 while multiboot is doing that for obvious reasons (it exploits bootmenu to do its nasty kernel reflashing stuff), uBoot shouldn't link to other packages like that. it desrves its own config file directory Oct 13 19:03:30 I'd even think multiboot and uBoot are mutually exclusive and thus conflicting, and so you probably shouldn't even install uBoot on a system where multiboot is installed Oct 13 19:06:08 it seems that the nfs server is mounted several times Oct 13 19:12:47 check if any process has open filehandles to the server and stop all those processes. Then mount -O the nfs volume again over the original mountpoint, *then* unmount it again Oct 13 19:13:18 this _might_ reset the stale mount Oct 13 19:13:32 if it doesn't, you're in for a fsck I guess Oct 13 19:18:40 Hmm Oct 13 19:19:30 I'm getting more interested to put more Java apps in this little box :D Oct 13 19:21:56 FlameReaper: no you're not >:c Oct 13 19:22:29 kerio: I got Jajuk up and running Oct 13 19:22:53 but the 100% CPU usage @ 950MHz isn't something nice to see :c Oct 13 19:23:07 can't get address for /nfs/music Oct 13 19:23:13 while it is refreshing the libraries Oct 13 19:27:54 kerio: it's an analogy to upstart (uBoot, sorry ;-P) replacing sysV-init (multiboot, sorry to sysV-init). If upstart is using init.d/* and the runlevel dirs to create a setup to keep the system config as compatible and working as feasible then that's still fine. But no way upstart should store its own config under /etc/init.d, and actually it doesn't. there's /etc/events.d/* for that Oct 13 19:28:15 it doesn't store its own config in bootmenu.d goddammit Oct 13 19:28:42 it doesn't even look at it, there's a script that does that and builds a part of the uboot config Oct 13 19:31:11 then I don't get it why everybody says uBoot config has to go into /etc/bootmenu.d Oct 13 19:31:43 if it's not meant that way, then why is anybody suggesting it? Oct 13 19:33:14 this obviously is causing that much confusion for users, even you thought bootmenu and uBoot were similar or related Oct 13 19:34:32 probably the main sin is to do all this during preinstall, instead of shipping a tool to do it on user request Oct 13 19:35:11 (or postinstall, dunno) Oct 13 19:35:22 i *know* what a bootloader is and how bootmenu works Oct 13 19:36:36 if user could invoke `uBoot-import-multiboot` which would come with uBoot installation, then things wre much more clear, and btw more convenient to handle Oct 13 19:37:48 that's what happens, kinda Oct 13 19:37:56 obviously not Oct 13 19:38:07 there's a script, it's invoked by the postinstall of the kernel bootimgs but can also be invoked manually Oct 13 19:38:25 WTF is it getting invoked by postinstall then? Oct 13 19:38:40 idk, ask Pali Oct 13 19:39:40 U-Boot can only load uboot script files (hush) and u-boot-update-bootmenu converting bootmenu format to uboot script Oct 13 19:40:03 :nod: Oct 13 19:40:28 uboot scripting (for uboot bootmenu) is not human job Oct 13 19:40:38 duh Oct 13 19:41:07 also uboot can load uImage kernels (zImage with special header) Oct 13 19:41:40 so that script also convert zImage from /boot (or /opt/boot/) to uImage and store it in MyDocs/bootmenu_subdir/ Oct 13 19:41:53 fsck stops after running shortly in BM Oct 13 19:42:16 grammoboy: it doesn't 'stop' - it just takes ages I guess Oct 13 19:42:23 damn didn't know nfs could give me this kind of trouble Oct 13 19:42:30 DocScrutinizer05, it goes back to the menu Oct 13 19:42:33 ooh Oct 13 19:42:38 Pali: how does it boot something like backupmenu? Oct 13 19:42:40 so it's maybe finished Oct 13 19:43:06 kerio, I do not know if uboot can boot backupmenu Oct 13 19:43:17 oic Oct 13 19:43:20 but it needs something like pass init= Oct 13 19:43:39 BM is not meant to play init process Oct 13 19:44:07 BM depends on bootmenu Oct 13 19:44:22 DocScrutinizer05: to be fair it only needs proc, sys and tmp Oct 13 19:44:24 or something like that Oct 13 19:44:34 (that's what preinit does before bootmenu is loaded) Oct 13 19:45:07 maybe Oct 13 19:45:39 doesn't change the difference of what bootmenu does vs what uBoot does Oct 13 19:46:10 uBoot is unrelated to bootmenu, if anything it is related to then that's multiboot Oct 13 19:46:33 since it obviously tries to replace that Oct 13 19:47:57 yet most users now mix up uBoot and bootmenu, they don't even notice there's a difference Oct 13 19:48:55 and that's caused by instructions suggesting to tweak /etc/bootmenu.d/* entries to edit stuff that actually should get edited in genuine uBoot script Oct 13 19:49:51 man this isn't f*cking windows is it Oct 13 19:50:15 I just can't get rid of the freaking umount problem Oct 13 19:50:35 wtf Oct 13 19:50:49 we're risking to run into compatibility issues for the other direction: tweaking files in /etc/bootmenu.d/* for changing uBoot behaviour may (and iirc actually did already) cause problems for bootmenu Oct 13 19:51:55 grammoboy: usually a reflash is fastest and most painless fix Oct 13 19:52:10 give me a break, my o my Oct 13 19:52:27 create a proper backup with BM, reflash rootfs, the restroe the backup Oct 13 19:52:45 ,then* Oct 13 19:52:52 ,then restore* Oct 13 19:53:16 also backup Mydocs? Oct 13 19:53:46 not needed unless your mountpoint been on /home or below Oct 13 19:54:21 though you should have proper backups of your whole system any time Oct 13 19:54:51 gambling with that stuff isn't worth it Oct 13 19:57:43 he said the mountpoint was /nfs/music Oct 13 19:57:45 so... ubifs error? Oct 13 19:57:46 damn you n900, fucking time waster! Oct 13 19:57:53 :) Oct 13 19:57:59 grammoboy: are you sure you can't just... rm the directory? Oct 13 19:58:25 kerio, hm I thought of that, can try it Oct 13 19:58:54 will cause a dnagling lock in a dangling inode Oct 13 19:59:53 I still dunno what's your actual problem. You say you can't umount the nfs, but why do you want to do that, first instance? Oct 13 20:00:34 I can't get to my files Oct 13 20:00:42 cd /nfs/music is empty Oct 13 20:00:51 sure Oct 13 20:00:54 to restore nfs Oct 13 20:01:01 I need to umount first Oct 13 20:01:07 no Oct 13 20:01:08 cause mounting doesn't work Oct 13 20:01:16 now Oct 13 20:01:21 the nfs that is Oct 13 20:01:23 even not when you try mount -O ? Oct 13 20:01:40 0 Oct 13 20:01:42 zero? Oct 13 20:01:53 and close all processes that have a file handle to this? Oct 13 20:02:15 I've rebooted serveral times Oct 13 20:02:23 then what the hell Oct 13 20:02:33 you don't have a mount anymore, unless you have it automounted Oct 13 20:02:38 just rm the directory Oct 13 20:02:46 http://sysunconfig.net/unixtips/stale_nfs.txt Oct 13 20:03:16 DocScrutinizer05: he rebooted! several times! Oct 13 20:03:21 kerio: so what? Oct 13 20:03:23 in particular, he rebooted at least once Oct 13 20:03:41 go look at e) in that list Oct 13 20:03:56 mount -O overwrites the inode holding the supposed stale nfs lock Oct 13 20:04:14 grammoboy: what's your actual problem, right now? Oct 13 20:04:34 kerio, feeling tired Oct 13 20:04:39 kerio, can't mount nfs Oct 13 20:04:52 oh, so you just can't mount it Oct 13 20:05:02 are nfs-common and portmap installed? Oct 13 20:05:14 sure, see -O Overlay mount. Oct 13 20:05:15 did you do /etc/init.d/nfs-common start and /etc/init.d/portmap start? Oct 13 20:05:36 you can't mount on an allegedly already mounted mountpoint Oct 13 20:05:41 DocScrutinizer05: i'm not even sure he *has* anything mounted right now Oct 13 20:05:58 kerio: that's not the point Oct 13 20:06:09 th einode says he has Oct 13 20:06:33 and since the connection to far end is down, it can't umount Oct 13 20:06:52 but aiui you can mount -O Oct 13 20:06:59 grammoboy: can you pastebin the output of "mount"? Oct 13 20:07:02 on same inode, thus overwriting it Oct 13 20:07:11 and can you rmdir /nfs/music ? Oct 13 20:07:40 when he does that, he doesn't have any way to mount or unmount on that inode anymore Oct 13 20:08:05 and what's the problem? he doesn't have a stale mount anymore, he *rebooted* Oct 13 20:08:13 while the inode stays as it is, since it's considered an open handle Oct 13 20:08:41 is that 0 a zero? Oct 13 20:08:53 grammoboy: why would it be a zero Oct 13 20:08:57 it's a -O Oct 13 20:09:02 hmm, yeah, next boot might result in an arbitrary corrupted inode then Oct 13 20:09:27 -O Overlay mount. Oct 13 20:09:34 nfs, the magical FS with the magical inode corruption powers Oct 13 20:09:35 http://sysunconfig.net/unixtips/stale_nfs.txt Oct 13 20:09:42 nfs is flaky, but not that flaky Oct 13 20:10:02 DocScrutinizer05: you keep posting that, and it clearly lists "reboot" as a way to solve the problem Oct 13 20:10:32 no, but fs is as magic to assume everything it doesn't know is a stale NFS if it's an inode Oct 13 20:11:17 kerio: you keep telling stuff like those were all alternatives rather than a sequence of things to execute Oct 13 20:11:31 -o and -t nfs? Oct 13 20:12:09 O invalid option Oct 13 20:12:11 kerio: while I already explained several times that you can get stale NFS mounts or handles on a system that doesn't even have any nfs drivers Oct 13 20:12:31 too bad, our mount is flaked Oct 13 20:13:40 DocScrutinizer05: fwiw -O is "optlist" on debian's mount Oct 13 20:13:59 :shrug: Oct 13 20:14:45 i'm not going to invest even more time into that, this includes me not reading manpage of busybox mount and finding out how grammoboy made nfs work on maemo at all Oct 13 20:14:58 DocScrutinizer05: i literally just did Oct 13 20:15:11 start nfs-common and portmap, mount -t nfs server:export dir Oct 13 20:16:30 [2012-10-13 21:52:27] create a proper backup with BM, reflash rootfs, the restroe the backup Oct 13 20:17:27 or use hexedit on /dev/mtd5 ;-P Oct 13 20:18:13 ohoho, killing the connection if you didn't mount with -o soft is truly an awful idea Oct 13 20:18:43 kerio: either you've read all the other pages I linked for grammoboy, or you're going thru all that stuff again Oct 13 20:19:02 nah, i'm just trying it right now on the n900 Oct 13 20:19:12 stuff=trial&error and finding out about how nfs and fs in general work Oct 13 20:20:51 if ext2/3/4 fs driver runs into any corrupted inode it doesn't know about at all, it claims "stale NFS" Oct 13 20:21:05 and are you sure it's the same thing that happens with ubifs? Oct 13 20:21:19 quite, yes Oct 13 20:21:52 since afaik ubifs also is using inodes Oct 13 20:22:30 plus I've seen this very effect of stale NFS error on FR Oct 13 20:22:42 which used jffs2 iirc Oct 13 20:22:53 so why would it differ for ubi? Oct 13 20:23:24 hah, "server 192.168.42.42 not responding, still trying" Oct 13 20:23:32 good luck with that, you stupid module Oct 13 20:23:41 i wonder why i can't even umount -fl it Oct 13 20:23:59 because it's braindamaged crap ;-) Oct 13 20:25:10 alright, let's reboot Oct 13 20:25:21 the funny stuff is that mounts seem to actually alter inodes, otherwise how would the "stale NFS" error survive reboots? Oct 13 20:25:42 ah balls Oct 13 20:25:55 screen turned off, i can't turn it on Oct 13 20:25:59 it's still on, isn't it Oct 13 20:26:22 this is a textbook example of bad design, imo Oct 13 20:26:57 there's a reason why nobody touches NFS anymore ;-) Oct 13 20:27:19 meh, it's just that "soft" should really be a default option Oct 13 20:27:41 and i was talking about the n900 in general Oct 13 20:27:48 solves some problems, but others perveil Oct 13 20:28:20 i wonder if a FUSE nfs client would be better, in this regard Oct 13 20:28:40 no, afaik the problems are design-immanent Oct 13 20:28:42 mostly because you can apply some -9 to it to make it stop being a bitch Oct 13 20:28:56 you can't exactly kill the kernel Oct 13 20:29:27 fuse also needs a kernel module afaik Oct 13 20:30:19 as an aside, /me ponders about symlinking /etc/mtab to /proc/mounts Oct 13 20:31:20 DocScrutinizer05: idk, i successfully rmdir'd the mounted directory, after a reboot Oct 13 20:31:39 now for something completely different: how's your boot working, and did the stale NFS vanish? Oct 13 20:31:51 me or grammoboy? Oct 13 20:31:56 you Oct 13 20:32:11 boot went fine after powercycling, no mention of nfs in mtab or mounts or `mount` Oct 13 20:32:19 rmdir'd the directory, no error messages Oct 13 20:32:26 pff Oct 13 20:32:28 sure Oct 13 20:32:46 there's not even an error msg when I rm /sbin/init Oct 13 20:32:55 lol Oct 13 20:33:16 means: on unix you can rm whatever you like, it just unlinks the inode Oct 13 20:33:28 so... i should reboot again? Oct 13 20:34:12 you now probably have a corrupted inode anywhere on your storage medium, and it will show up again whenever somebody tries to use it Oct 13 20:34:29 cool story bro Oct 13 20:34:35 or when you run a fsck against the volume Oct 13 20:34:45 it's ubifs, it's indestructible or something Oct 13 20:34:47 there's no fsck.ubifs Oct 13 20:34:53 haha Oct 13 20:35:40 well, what else can the lack of a checking utility mean? Oct 13 20:36:32 anyway, google says nothing about stale NFS inodes corrupting other inodes Oct 13 20:38:56 kerio: http://lists.infradead.org/pipermail/linux-mtd/2010-February/029017.html Oct 13 20:39:28 DocScrutinizer05: i read that already, i lol'd already, then i had a sad for a bit Oct 13 20:40:15 kerio: also, again: what's your take then about error "stale NFS" on a flash storage in a phone that evidently never seen NFS Oct 13 20:40:31 and those errors surviving reboots Oct 13 20:40:31 well, it's clearly unrelated to nfs Oct 13 20:40:59 and i don't care, unless you have a way to explain how that error can jump from inode to inode when you delete the original file handle Oct 13 20:41:01 ooh sure, and the three letters in the error msg are just for fun? Oct 13 20:41:21 no, "stale NFS inode" or whatever is the error for when there's an inode that's corrupted enough Oct 13 20:41:34 kerio: wtf? jump from inode to inode? it will jump out and bite your nuts Oct 13 20:42:00 i was able to enter the directory that held the mount point, i even listed its contents Oct 13 20:42:01 it was empty Oct 13 20:42:05 i rmdir'd it Oct 13 20:42:54 problem solved, in my opinion Oct 13 20:42:55 honestly, this gets boring Oct 13 20:43:10 you're the one claiming "you have a corrupted inode somewhere, now" Oct 13 20:43:21 first instance you never had any problem, unlike grammoboy Oct 13 20:43:48 and second, I don't give a shit Oct 13 20:43:58 that isn't nice :c Oct 13 20:44:04 I toldya what I've seen and what I concluded Oct 13 20:44:39 and I'm not interested in proving I'm right or wrong, I'm basically not interested in NFS at all Oct 13 20:45:36 and third, I said /me afk some hours ago, and i'm still sitting on this friggin IRC terminal, while my stomach hurts and asks for food Oct 13 20:46:02 i wonder if i get the same situation if i mount it with -o soft Oct 13 20:46:11 and 4th: [2012-10-13 21:52:27] create a proper backup with BM, reflash rootfs, the restroe the backup Oct 13 20:47:16 if NFS handle isn't stored in inode, then WTF didn't reboot help for grammoboy? Oct 13 20:47:35 grammoboy: what's your current error? Oct 13 20:47:54 cya peeps Oct 13 20:48:01 take care doc Oct 13 20:48:22 wow, umount -fl just... didn't umount? :s Oct 13 20:49:12 oh, it did, but it didn't update /etc/mtab Oct 13 20:49:14 this is silly Oct 13 20:49:55 * grammoboy making backup Oct 13 20:50:22 grammoboy: first answer me! Oct 13 20:50:26 what's your error message? Oct 13 20:51:06 kerio, input/output error Oct 13 20:51:07 iirc Oct 13 20:51:16 kerio, umount 'invalid argument' Oct 13 20:51:22 less /proc/mounts Oct 13 20:51:22 kerio: sorry, this been actually not nice. I was impolite Oct 13 20:51:32 DocScrutinizer05: don't worry about it <3 Oct 13 20:51:36 kerio, showmount 'can find address' Oct 13 20:51:38 o/ Oct 13 20:51:47 grammoboy: check /proc/mounts, see if it's actually mounted Oct 13 20:51:52 or if it's just /etc/mtab being silly Oct 13 20:52:09 Pali: can we get some modifications in /proc/mounts so it actually fits what /etc/mtab should have and we can symlink it? Oct 13 20:52:10 first I need to finish the backup Oct 13 20:52:45 DocScrutinizer05, grammoboy: not sure if you're interested, but i got "invalid argument" when trying to umount something that was listed in mtab but not /proc/mounts Oct 13 20:52:53 *but not in Oct 13 20:53:06 haha, nice spotting Oct 13 20:53:21 /etc/mtab is fucking stupid Oct 13 20:53:33 sure thing Oct 13 20:53:45 since afaik it's maintained by mount Oct 13 20:54:26 kerio, hm, if that also gives hope for mount Oct 13 20:54:31 it's maintained by *busybox's mount* Oct 13 20:54:48 kerio, if I can only umount Im not really further Oct 13 20:55:06 i still don't know what your original problem was Oct 13 20:56:18 >>The programs mount and umount maintain a list of currently mounted filesystems in the file /etc/mtab. If no arguments are given to mount, this list is printed.<< Oct 13 20:57:02 * grammoboy q and reboot Oct 13 20:57:06 oh nvm, Invalid argument is just for any path that exists but isn't a mountpoint Oct 13 20:57:10 so obviously if something else (kernel) "umounts" a fs, mount will run into trouble with its private notion of what's mounted Oct 13 20:57:46 DocScrutinizer05: that's why you're somewhat "supposed" to have /etc/mtab as a symlink to /proc/mounts, if your kernel is recent enough Oct 13 20:57:59 it's probably perfectly sane and safe to *edit* /etc7mtab Oct 13 20:58:13 2.6.28.10's /proc/mounts doesn't show bindmounts correctly Oct 13 20:58:51 or just delete this crap Oct 13 20:59:09 yeah, i just deleted the nfs entry that wasn't in /proc/mounts Oct 13 20:59:15 so mount will happily mount everything, but not umount anything after that ;-P Oct 13 20:59:17 which is the official mount list Oct 13 20:59:41 DocScrutinizer05: eeh, i don't even know if mount checks the list to *do* things Oct 13 20:59:58 *shrug* Oct 13 21:00:08 can't hurt to just delete it and test Oct 13 21:00:11 DocScrutinizer05: get off irc and eat something Oct 13 21:00:19 AIESIR Oct 13 21:00:23 AYE even Oct 13 21:01:12 * grammoboy n900 booted Oct 13 21:01:17 DocScrutinizer05: GTFO INTERNET & A BEER. Oct 13 21:01:22 * grammoboy copying backup Oct 13 21:01:48 grammoboy: can you pastebin the contents of your /proc/mounts and your /etc/mtab somewhere? Oct 13 21:10:22 http://ubuntuone.com/4lrYXoZrjoXlK9112VYukq Oct 13 21:10:31 kerio, ^ Oct 13 21:11:15 what about /etc/mtab, or the output of "mount"? Oct 13 21:12:03 oh lol, I managed to get TWO identical entries for /dev/mmcblk1p2 in mtab now Oct 13 21:12:37 kerio, no /nfs/music mount points Oct 13 21:12:50 can you rmdir /nfs/music? Oct 13 21:13:36 kerio, yes Oct 13 21:13:44 then you probably don't have any problems Oct 13 21:14:52 kerio: I umounted /media/mmc1p2/, then copied old version of mtab *with* the /media/mmc1p2/ line over the current one Oct 13 21:15:12 and then you mounted it again? Oct 13 21:15:28 IroN900:~# umount /media/mmc1p2/ Oct 13 21:15:29 umount: cannot umount /media/mmc1p2: Invalid argument Oct 13 21:15:32 see, not even mount cares about /etc/mtab Oct 13 21:15:44 DocScrutinizer05: do umount /etc/fstab Oct 13 21:15:53 IroN900:~# mount /dev/mmcblk1p2 /media/mmc1p2 -t ext3 -o rw,noauto,nodev,exec,nosuid,noatime,nodiratime Oct 13 21:15:59 then there were TWO Oct 13 21:16:10 yeah, it makes sense Oct 13 21:16:10 it has problems mounting the nfs Oct 13 21:16:24 /etc/mtab is there for histerical raisins Oct 13 21:16:29 grammoboy: which problems? Oct 13 21:16:46 did you do /etc/init.d/nfs-common start and /etc/init.d/portmap start? Oct 13 21:17:58 kerio, no and now it seems to be solve Oct 13 21:17:59 d Oct 13 21:18:09 yay Oct 13 21:18:12 o my Oct 13 21:18:21 thx Oct 13 21:18:36 isn't it strange that you've to restart those stuff after reboot? Oct 13 21:18:36 now, umount your fs and remount it with -o soft Oct 13 21:18:47 grammoboy: they should be started at boot, but they're not Oct 13 21:18:47 why -o soft? Oct 13 21:19:08 because otherwise if you lose the connection to the server you have to toy with IP addresses or remove the battery Oct 13 21:19:49 -o soft -t nfs? Oct 13 21:20:13 yeah Oct 13 21:23:28 ok Oct 13 21:24:10 also no rc.local file Oct 13 21:24:13 hm Oct 13 21:24:25 * grammoboy prefers real debian :) Oct 13 21:25:44 rc.local? Oct 13 21:25:55 wuzzat? Oct 13 21:27:55 to start up stuff in debian at boot Oct 13 21:28:00 or right after boot Oct 13 21:28:06 or somewhere in between Oct 13 21:28:16 aaah that silly solitaire moster initscript ;-P Oct 13 21:28:22 monster Oct 13 21:29:44 that 'alterna(t)ive' to proper sysV init Oct 13 21:33:16 DocScrutinizer05, kerio thx for your help guys Oct 13 21:33:37 * grammoboy playing radiohead via nfs Oct 13 21:34:46 the homemade applewine was going to be 11% according to maths, before bottling the direct measurement said 14% Oct 13 21:34:57 tastes more like a vodka than a wine so it might be true heh Oct 13 21:35:34 grammoboy: yw Oct 13 21:35:38 kerio: man 8 mount|grep -A4 "When *the *proc" Oct 13 21:39:46 wow, manpages are really interesting sometimes: >>... s. One can also remount a single file (on a single file). It's also possible to use the bind mount to create a mountpoint from a regular directory, for example: mount --bind foo foo << Oct 13 21:48:58 kerio: >> -f, --fake Causes everything to be done except for the actual system call; if it's not obvious, this ``fakes'' mounting the filesystem. This option is useful in conjunction with the -v flag to determine what the mount command is trying to do. It can also be used to add entries for devices that were mounted earlier with the -n option. The -f option checks for existing record in /etc/mtab and fails when the record already exists Oct 13 21:48:59 (with regular non-fake mount, this check is done by kernel). << Oct 13 21:51:08 kerio: there are 2 interesting details in that: 1) "can also be used to add entries" (I assume also to delete entries, for umount), and 2) "... this check is done by kernel" Oct 14 01:56:49 Why is the dsp profile have POWERSAVE_BIAS=0 ? Oct 14 01:57:18 Isn't it better to have POWERSAVE_BIAS=1 ? Oct 14 02:59:44 please explain what's POWERSAVE_BIAS **** ENDING LOGGING AT Sun Oct 14 03:00:02 2012