[00:42] rsalveti: ping, hit a kernel crash on boot, have you seen it before? bug 820129 [00:42] Launchpad bug 820129 in linux-linaro "boot crash - kernel BUG at /build/buildd/linux-ti-omap4-2.6.38/kernel/workqueue.c:1029!" [Undecided,New] https://launchpad.net/bugs/820129 [00:57] mahmoh: That could be due to the memory hole on the panda. I notice that you don't have the mem= lines on your kernel cmdline. === robbiew-afk is now known as robbiew [01:06] Also, why are you filing this against linux-linaro? The source is linux-ti-omap4. [01:43] GrueMaster: it wouldn't let me choose that "package" for some reason, could you switch it for me pls? I thought the kernel knew about that hole already and the map wasn't necessary any longer [01:53] after much pain and aggravation it now worked === Jack87 is now known as Jack87|Away [10:52] persia, let me know when you are around [10:52] lilstevie, I'm around [10:53] ok cool [10:53] lets try and get this thing together :) [10:54] Right. [10:54] * persia reviews the instructions [10:56] ok, well I have the tarball decompressed [10:57] but I am thinking maybe I should just clone it on to this [10:57] btw, I am on the tf this time [10:57] trying to avoid the cross_compile issue sneaking in [10:59] Looks like most of the cross compilation stuff comes from step 11.1 [10:59] heh ok [11:00] If you've merged the Ubuntu sauce (which you'll need), then you ought be able to skip step 8.1 [11:00] I haven't merged the ubuntu stuff [11:00] And you certainly don't want to do step 9.1 [11:00] Aha! You'll want to do that. There's a sauce branch in the Ubuntu master kernel repo at kernel.ubuntu.com [11:01] It has stuff like aufs, etc., which we expect to have in any Ubuntu kernel. [11:01] ok [11:01] *don't* run step 11.2 to build the kernel. [11:01] and will it apply cleanly over any kernel [11:01] and will it apply cleanly over any kernel;0 [11:01] Instead, run `debuild -S -us -uc`. Save the artifacts this produces as precious. [11:01] ok [11:02] -us -uc instead of -i -I [11:02] ? [11:03] Once you have saved the artifacts, you can run a build with `debuild -b` which will try to build binaries *without* polluting the source. You can recover pristine source by removing the working directory, and running `dpkg-source -x foo.dsc`. Be aware that this will wipe your git directories: you can also recover the pristine state with git. [11:03] eh I am not fussed about killing git out [11:05] `debuild -S -us -uc -i -I` works too. In detail, '-S' says "make only a source package: don't compile", '-uc' means "don't sign the changes file", '-us' means "don't sign the source package", '-i' means "apply the default filter to remove some files from the source package", and '-I' means "exclude the default set of files from the source package". [11:06] ok cool [11:08] If you succeed at building the source, let me have the precious artifacts, and I'll submit it to my build harness (I'm using sbuild, but I'm unsure if you want to set that up), and verify it works as expected, then upload, if it has all the right binary packages. [11:09] ok, well looking at the ubuntu prime git [11:26] persia, just trying to find the spot that has 2.6.36 [11:48] persia, I can't find 2.6.36 [11:49] Grumble. [11:49] Theoretically, when moving from 2.6.35 to 2.6.38, one ought import the VCS history, including tags. [11:49] And that would have been from the maverick->natty switch. [11:50] hm [11:50] Maybe someone in #ubuntu-kernel would know? If not, try importing the sauce from 2.6.38: much of it is out-of-tree modules, and may well build anyway. [11:53] what is it that is wanted anyway?' [11:53] aufs [11:58] Hello, i have problems with my cross compiled kernel for my beagleboard xm (rev.A). i have downloaded the 2.6.39 kernel on my desktop pc, make the uImage and the modules, copy the 2.6.39 Image folder back to my beagle and make install_modules, after that sudo update-initramfs -k 2.6.39 -c , but when i want to boot the beagle with my new kernel, i get errors modprobe: FATAL: Could not load /lib/modules/2.6.39/modules.dep: No such fil [12:07] lilstevie, I don't know the complete set of what is needed: whatever makes the enforcer not complain. At least aufs. === chuck_ is now known as zul [12:07] aufs isn't on the enforcer list [12:09] the only 2 things that fail are security_smack and security_yama [12:10] Hrm. *some* sort of union file system ought be in the enforce list. Otherwise it's hard to make live images, etc. [12:11] We also use union filesystems for some developer operations (like speeding test-building, etc.) [12:29] hm [13:21] persia, are you free? [13:29] persia: well I have in the workplace development in a few hours so I will have to look at this later, but as a minimum I will patch in aufs and yama and smack [13:29] ugh, why ?!? [13:30] aufs is a super intrusive bit, is there anything you need it for ? [13:31] ogra_: just lilstevie, I don't know the complete set of what is needed: whatever makes the enforcer not complain. At least aufs. [13:33] i never used aufs anywhere in my kernels [13:33] ogra_: the only things enforcer complain about is security_yama and security_smack [13:33] and never had an "enforcer" (whatever that is) complain [13:33] never heard of that [13:34] whats that "engforcer" thing ? [13:34] ogra_: it is part of the mechanism that parses the config [13:34] then its wrong [13:34] aufs is non standard crap you shouldnt be enforced to have to add to your kernel to have a package [13:34] it takes debian.linaro/config/enforce and makes sure they are enabled in kernel config [13:34] linaro ? [13:35] well I don't have any aufs stuff [13:35] right [13:35] fix the file then [13:35] and neither does the enforce [13:35] its nonsense to enforce it [13:35] heh [13:36] what about security_yama and security_smack [13:37] no idea, your kernel doesnt have it ? [13:37] not at all [13:38] well, for that ask in #ubuntu-kernel or the security team, i'm not sure if oneiric has userspace using it [13:38] I have selinux and apparmor enabled in kernel config though [13:38] if not, i would ignore it too [13:38] right, thats a typical ubuntu default [13:39] and apparmor is set default [13:41] right [13:41] with selinux being off [13:42] yep [13:43] is there anything else that I may have missed that is handy to havee? [13:43] have* [13:45] well, what i usually do is to have two terminals open with the omap4 config and the new config i want to write and compare them [13:48] heh === nslu2-log_ is now known as nslu2-log === prpplague is now known as prpplague^2 [15:08] why is my board rebooting?! http://pastebin.ubuntu.com/657985/ [15:08] ignore the first seven lines === bambee_ is now known as bambee [15:12] zul: what was the kernel we ran on the tegras? [15:12] zul: you can run 'lxc-checkconfig' to get the full list [15:13] i dont think its tegra [15:13] (ignore 'namespace cgroup' not being there, we don't want it) [15:13] hallyn: when i tried to mount cgroups invalid fs [15:13] well whatever, but it ws arm and lxc was working :) [15:13] so who's in charge of omap4 kernel? [15:13] (!! config) [15:14] hallyn: oh wait i can mount it nm... [15:15] zul: so does lxc-checkconfig show any errors? [15:15] havent gotten that far but this is with libvirt [15:17] although this line in the /etc/fstab throws errors: http://pastebin.ubuntu.com/657992/ [15:18] zul: is this oneiric? [15:18] i'd just install cgroup-bin at this point... [15:18] hallyn: yep [15:21] zul: oh, well what does cat /proc/cgroups show? [15:21] http://pastebin.ubuntu.com/657993/ [15:24] zul: no freezer or blkio [15:24] yeah just figured that out [15:24] blkio is not bad, but freezer really should be enabled [17:07] GrueMaster: ping [17:07] prpplague: pong [17:08] GrueMaster: getting closer, looks like it may be related to some frequency scaling [17:08] interesting. [17:10] GrueMaster: on a L27 .35 kernel with the governor set to performance, the problem goes away [17:11] Very interesting. [17:11] I'll see if I can check it out on our latest images with the 3.0 kernel. [17:17] GrueMaster: do you have full governor support in your 3.0 ubuntu kernel? [17:18] GrueMaster, our governor should default to preformance on boot, there is the /etc/init.d/ondemand script that sets it after booting, just comment the command in there [17:18] that way it should just stay on performance [17:24] Hrm. No scaling_governor on 3.0 kernel. [17:33] I have no way of knowing what the current setting is, or how to change it. There is no /sys/device/cpu/cpu*/cpufreq/* [17:34] And I am not seeing anything in the kernel config to indicate either way. [17:36] Hrm. cat /sys/devices/system/cpu/cpuidle/current_governor_ro [17:36] menu [17:48] <[7]> GrueMaster: it's "menu" on my 2.6.38-1208 kernel as well [17:49] <[7]> whatever that means... [17:51] menu ?? [17:51] weird [17:52] Yea for outdated kernel docs. [17:52] <[7]> theseven@panda:~$ cat /sys/devices/system/cpu/cpuidle/current_governor_ro [17:52] <[7]> menu [17:52] <[7]> theseven@panda:~$ uname -a [17:52] <[7]> Linux panda 2.6.38-1208-omap4 #11-Ubuntu SMP PREEMPT Fri Apr 15 16:34:35 UTC 2011 armv7l armv7l armv7l GNU/Linux [17:53] http://forum.soft32.com/linux/RFC-PATCH-add-menu-cpuidle-governor-ftopict341954.html [17:53] From 2007. [17:54] And nothing in linux-docs. [17:54] fun [17:58] I'm placing bets that this will end up being the cause of our grief. [18:34] I don't even know where to look for the different driver states. Currently, /sys/devices/system/cpu/cpuidle/current_driver is reporting none. [18:52] GrueMaster: /sys/devices/system/cpu/cpu0/cpufreq # cat scaling_governor [18:52] i bet amitk could tell you, but i also guess its beer o clock in cambridge now [18:52] prpplague, cpuidle ... :) [18:53] seems the menu governor belongs to it [18:53] yea [18:53] so the only built that doesn't seem affected is the latest 2.6.35 for L27 [18:54] prpplague: That path doesn't exist in the 3.0 kernel. [18:55] GrueMaster: right, i was testing a 2.6.39 kernel [18:55] GrueMaster: ti kernel [19:15] hi all [19:16] what means if boot there is no terminal output after the message now booting kernel? [19:17] I have a beagleboard xm rev C [19:17] Uncropressing Linux... done, booting the kernel. [19:17] then notthing [19:22] <[7]> FxIII: which bootargs? [19:22] [7]: how can I check? [19:22] <[7]> shoul be printed somewhere above in the uboot output [19:22] any luck prpplague? [19:22] <[7]> or see the boot.scr file [19:22] MrCurious: appears to be related to the governors [19:23] did i miss hear that one could set to max perf and make it go away as a short cut? [19:23] MrCurious: on some versions of the kernel [19:23] bootargs [19:23] i think this bugs hours are numbered... :) [19:24] * [7] hopes that as well :) [19:24] <[7]> it's kind of a shame that nobody caught that earlier, would have improved the board's speed a lot [19:25] ro elevator=noop vram=12M omapfb.mode=dvi:1280x720MR-19@60 mpurate=500 root=UUID=2202437....3453 fixrtc quiet splash [19:25] <[7]> remove "quiet splash", add "console=/dev/ttyO2,11500n8" [19:25] <[7]> assuming it's the 3rd UART on the XM as well [19:25] this is a binary file... [19:26] <[7]> you'd need to edit it in /boot/boot.script and re-run flash-kernel [19:26] [7]: i did run a precompiled image... [19:28] my I use my serial connection to do boot? [19:28] <[7]> yes, you can edit the bootargs in uboot [19:28] mmm how? [19:28] <[7]> just cancel automatic boot and copy and paste the non-binary contents of the fixed boot.scr there [19:29] i have my omap3 beagleboard.rog # prompt [19:29] ok ok [19:30] FxIII, which iamge did you use ? [19:30] the netbook image *needs* a display, there is no way around it [19:30] if you want to use a serial enabled image use the headless image [19:31] ogra_: i need to have the video output [19:31] (and no, you cant just change boot.scr on a netbook image it would leave you with a completely misconfigured system) [19:31] i followed this https://wiki.ubuntu.com/ARM/OmapNetbook [19:31] including the last part [19:31] <[7]> you can enable video on the headless image as well [19:32] sure [19:32] you just cant use a netbook one without video [19:34] ok where I can find it? [19:34] hedaless ? [19:35] https://wiki.ubuntu.com/ARM/OMAP [19:35] FxIII: Reading the backscroll, you are looking for some kernel output on the netbook image beyond booting the kernel? [19:35] hmm, that wikipage got uglified [19:35] since when doe we have that weird header ? [19:36] ogra_: Since last week wneh we all revised the wiki. DO NOT CHANGE. [19:36] it looks really ugly, i wonder if it could be re-arranged to better fit the theme [19:37] It is the same as the desktop (actually copied from there). [19:37] really ? [19:37] i did not understant one thing, why i cant boot this image? using a monitor i mean [19:37] * ogra_ cant imagine the design team doesnt scream and shout :) [19:38] It's uuuuuugly [19:38] *bleaugh* [19:38] ogra_: https://wiki.ubuntu.com/DesktopTeam [19:38] Oh well :) [19:39] FxIII: Ok, back to stuff that matters... [19:39] GrueMaster, yeah, they only have three items that roughly have the same width [19:39] You have the netbook image for natty. Did you follow the update for the beagleXM Rev B & C? [19:39] it looks less weird (still hurts my eye though) [19:40] <[7]> FxIII: so what's your root problem? [19:40] re [19:40] <[7]> just a blank screen? [19:40] ogra_: you want pain and suffering, go look at my workitem list, pick a few items, and start researching how to implement them. [19:40] black screen on video and no activity on the serial console [19:41] GrueMaster, after i have sorted mine i will do that [19:41] You won't have anything on the serial console past booting the kernel. [19:41] But you need to update the bootloader & u-boot to have video. [19:42] GrueMaster: what update means? [19:42] Follow the instructions at https://wiki.ubuntu.com/ARM/OmapNetbook#Update_for_BeagleXM_Rev_B_.26_Rev_C [19:42] dont that image use some video output on dvi? [19:43] You have to patch the SD image prior to booting it. Otherwise you get nothing. [19:43] i did that [19:43] no results [19:43] no video output for long time [19:43] over ten minutes [19:44] When you copied the new uImage to the boot partition, did you overwrite the existing one? [19:44] Same with the vmlinuz? [19:45] yep [19:46] my sd was a fat [19:46] after dd i got two partitions [19:46] Yes, that is correct. [19:46] i mount the first one and copied uimage [19:46] the same for the second [19:46] on the boot directory [19:46] the umount both [19:47] After DD, you should have untared the updated kernel and copied uImage to partition 1 and vmlinuz to partition 2. [19:47] yes. [19:47] (bear with me, irc delays). [19:47] no problem [19:47] Make sure you run "sync" after copying and before unmounting. [19:50] wtf [19:50] i dont remember... [19:51] ok i can try to redo... [19:51] but umount does not write the pending write? [19:51] anyway i will do it [19:52] It does, but on some systems/SD readers it can be slow. I have to make sure the led stops blinking before removing my card. [19:59] sorry i fall [20:10] GrueMaster: i does not changes [20:10] no output on hdmi no activity on serial [20:11] Try the headless image. It at least spits out more on the serial console. [20:11] I wonder if you have a board with further undocumented changes that aren't in Natty. [20:18] wow, our preinstalled images grew quite a lot. I just noticed it [20:19] janimo: which ones? Desktop is about the same. Server on the otherhand now has the preinstalled pool. [20:21] GrueMaster, desktop. natty used to be smaller IIRC [20:24] libreoffice is the main difference. [20:24] But the images are still smaller than x86. [20:25] launchpad is being painfully slow. [21:49] jani [21:49] hmm bad screen [21:50] * janimo is confused by monitor being wired to panda but keyboard to workstation [21:50] * [7] knows that feeling rather well :) [21:51] <[7]> two keyboards on one desk are really awful [22:06] I use a usb switch for the keyboard and an HDMI switch for the monitor. [22:17] * [7] does something similar, but his monitor just has two inputs === Jack87|Away is now known as Jack87