[00:02] <calc> i'm trying to go a git bisect for drm-intel-nightly using mainline-build-one using the information from https://wiki.ubuntu.com/Kernel/KernelBisection but I am getting the following error http://pastebin.ubuntu.com/8274644/ any tips? Do I need to setup a chroot for it to work, I didn't see mention of that in the guide.
[00:08] <calc> looking in the script it looks like it hasn't been updated since precise
[00:31] <Dandel> I have an issue with the kernel panicing due to the ath9k driver and was wondering how i would go about accessing the kernel logs stored in the uefi nvram.
[00:37] <mjg59> Dandel: You'll need to enable the UEFI pstore backend - I think it's turned off by default
[00:37] <mjg59> After that, abrt should automatically pick it up
[00:37] <Dandel> who do i enable it?
[00:38] <Dandel> the syslog, kernlog and others are useless since whenever it freezes or reboots there is spam of null character making it unreadable in text editors.
[00:38] <Dandel> and even after checking the backtrace never got saved to disk
[00:39] <Dandel> also, how would i check how much free memory there is on the efi vars? ( there is a chance my system may brick due to samsung issues, it's a samsung laptop with an amd apu )
[00:40] <mjg59> The current kernel code should prevent that
[00:40] <Dandel> i did not.
[00:40] <Dandel> *it did not
[00:40] <mjg59> ?
[00:40] <Dandel> spamtastic :) ( i get hundreds of lines of wifi spam )
[00:41] <mjg59> I mean, the kernel should prevent the bricking issue
[00:41] <mjg59> Boot with pstore_disable=0
[00:42] <Dandel> k... also, the system has been up less than 40 minutes, and there is already 589 instances where the bandwidth changed
[00:42] <Dandel> ath9k driver leading to a kernel panic in mac80211
[00:43] <Dandel> from what i can read on the picture i took, the hang is in rate_control_set_rates
[00:45] <mjg59> Oh, sorry, ignore what I said about abrt. Missed which channel I was in.
[00:45] <mjg59> After the panic and reboot, there should be some entries in /sys/fs/pstore
[00:45] <mjg59> Just rm them after you've copied them somewhere
[00:46] <Dandel> nothing made it there
[00:47] <mjg59> Have you booted with pstore_disable=0 and then had a crash?
[00:47] <Dandel> no, it takes a few days
[00:47] <mjg59> Ok
[00:48] <mjg59> You need pstore_disable=0 set before the crash
[00:48] <Dandel> It takes about 4 days
[00:48] <Dandel> and that depends on how much streaming of video.
[00:54] <Dandel> how do i enable the pstore now so i can clear it out for next boot?
[00:57] <mjg59> Set pstore_disable=0 when you boot
[00:57] <mjg59> Er
[00:57] <mjg59> efivars.pstore_disable=0
[00:57] <mjg59> Wait, no. Sorry.
[00:58] <Dandel> is there a way to make this permanent where updating grub does not wipe out the config?
[00:58] <mjg59> efi_pstore.pstore_disable=0 pstore.backend=efi
[00:58] <mjg59> I'm afraid I don't know how to configure Ubuntu's grub
[00:58] <Dandel> i know i can modify GRUB_CMDLINE_LINUX_DEFAULT in the file /etc/default/grub but package updates of grub usually wipe it out.
[01:00] <Dandel> mjg59, do you know if the samsung laptops that use amd processors are unaffected by the efi nvram bug you find?
[01:02] <mjg59> I have no idea, I'm afraid
[01:02] <mjg59> But the kernel should prevent nvram filling up now
[01:02] <Dandel> how do i check how much free space is in nvram ?
[01:03] <Dandel> also, do you have any idea on how to boot linux using the windows uefi boot loader? ( grub has a nasty habit of destroying hibernation support on windows )
[01:04] <Dandel> also note, that it's uefi specific, so the requirement is that you haft to be setup to use efi/uefi only.
[01:10] <mjg59> You can't
[01:10] <mjg59> It'll only boot stuff that's signed by Microsoft
[01:14] <Dandel> then what about refind?
[01:17] <Dandel> it's an alternative for grub and does not seem to have the issue with breaking windows hibernation.
[04:34] <infinity> Dandel: Package updates of grub don't wipe out /etc/default/grub, that's the whole point of the file.
[10:27] <Dandel> infinity, then do you have any suggestions at silencing the constant spam?
[10:32] <Dandel> i keep getting the same two messages spammed about 457 times an hour.
[10:46] <infinity> Dandel: I have no idea about you wifi woes, I was just pointing out that your assetion that grub upgrade "wipe out" /etc/default/grub wasn't at all true.  It's a conffile whose sole purpose is for the package to leave it alone so you can configure your stuff.
[10:47] <Dandel> yea, but when the grub package updates it wipes that out.
[10:53] <infinity> ...
[10:53] <infinity> No, it really doesn't.
[10:54] <TJ-> Dandel: Only if the package upgrade is configured to replace it with the maintainers version
[10:55] <infinity> TJ-: Even then, not true, it's not a conffile, but a generated file that we then take great pains to not damage on upgrade.
[10:56] <Dandel> infinity, then explain to me why when using grub on efi, it breaks hibernation on windows. ( not a known issue on non-efi systems )
[10:56] <TJ-> infinity: No, it is shipped in grub2-common, as "/usr/share/grub/default/grub"
[10:59] <infinity> TJ-: Right, but it's not magically replaced.
[10:59] <infinity> TJ-: Unlike a dpkg conffile, which you could configure dpkg to always take the maintainer's version.
[11:00] <infinity> Dandel: That's a bit of a non-sequitor.  How does that relate to your claim that we overwrite your config?
[11:01] <Dandel> it's part of the issues i ran into with the system ( not sure where to report some of it )
[11:03] <infinity> The hibernation issue is https://bugs.launchpad.net/ubuntu/+source/grub2/+bug/1229886
[11:03] <infinity> But if you really are losing /etc/default/grub settings, we'd love a bug and information on how to reproduce it.
[11:03] <infinity> Random assertions that we wipe out your config, however, I can't work with.  Cause we really do everything in our power not to.
[11:06] <Dandel> I generally use the stock package configurations :)
[11:07] <Dandel> although, it was a pain to get dual booting with windows working... I'm thinking of doing a bug report requesting a change in how ubuntu is handled on x86-based efi systems. ( boot loaders )
[11:08] <TJ-> infinity: I just managed to provoke the "replace /etc/default/grub with the package maintainers version" on 14.04 - trying to figure out why it did that now
[11:09] <infinity> TJ-: You mean, without it asking?
[11:16] <Dandel> i think my crash on my wifi may be a user space bug 0o'
[11:55] <TJ-> infinity: Yes. It is an edge-case but it can happen (usually because another sys-admin, or a forgetful sys-admin, is unaware/has forgotten about a previous choice they made with ucf) : http://paste.ubuntu.com/8279350/
[12:11] <Dandel> TJ-, In my case it's more of a forgetful sys-admin, where i set things up once and do not expect to haft to prune overrides every time.
[12:14] <TJ-> Dandel: The only way I know of to cause it to happen every time is to change the UCF configuration itself
[12:15] <Dandel> ucf configuration?
[12:16] <Dandel> the only place i know to modify options for grub is /etc/default/grub and on top of that, the grub search scripts ( and the installer ) do not properly detect windows on efi systems.
[12:16] <Dandel> although i do not have the windows partitions mounted at boot, i mount them later on, but that should not cause issues since i do mount the efi partition.
[12:35] <TJ-> Dandel: Discovery of other OSes is done and controlled by  "/etc/grub.d/30_os-prober"
[16:45] <EQ__> hi there
[16:45] <EQ__> anybody free for some questions about kernel debugging?
[16:45] <EQ__> gdb does not show any function names even with the vmlinux loaded
[16:45] <EQ__> thx
[17:29] <TJ-> EQ__: You need the debug build (unstripped) from the ddeb repository, and possibly the kernel source tree if you wish to dig deeper
[17:34] <EQ__> I have both
[17:34] <EQ__> now I recompiled the source with CONFIG_DEBUG_INFO=y
[17:35] <EQ__> vmlinux is present, same applies here, no function names in the gdb
[17:35] <EQ__> I tried with different 3.13.0-x kernels, 3.11.0-x kernel it was the same
[17:35] <EQ__> but it was working with 3.5.0-x but there was no source on the repository
[17:52] <EQ__> I tried to debug from debian, now I copied the VM with the ubuntu
[17:53] <EQ__> I attached the debugger from ubuntu to ubuntu
[17:53] <EQ__> now it is working, but for me it is weird...
[18:12] <TJ-> EQ__: I've tried to reproduce the issue here, but it works fine. What function name(s) were missing?
[18:30] <EQ__> all of them, are you trying to do it from ubuntu?
[18:31] <EQ__> working with this: GNU gdb (Ubuntu 7.7-0ubuntu3.1) 7.7 - This GDB was configured as "i686-linux-gnu".
[18:32] <EQ__> was not working with this: GNU gdb (GDB) 7.4.1-debian - This GDB was configured as "x86_64-linux-gnu"
[18:32] <EQ__> and yeah, I set the architecture to i386 afterwards
[18:32] <EQ__> when I hit ctrl+c there was no function names in the backtrace
[18:34] <TJ-> EQ__: We could theorize that the Debian gdb is configured differently in terms of where it looks for debug symbols; I'm not sure whether Debian have adopted the Ubuntu ddeb conventions fully as yet - I know there was discussion on doing that
[18:35] <EQ__> I see
[18:44] <EQ__> thanks for the info
[18:44] <EQ__> bye now
[21:19] <Dandel> i managed to decrease the log spam of the wireless, and it seems to involve an edge case brought on by mixing new and old equipment.
[21:19] <Dandel> in particular newer equipment capable of running 450mb/s over the 2.4ghz spectrum.
[21:21] <Dandel> The older hardware is the atheros AR9285 based wireless chip capable of up to 150 mb/s
[21:25] <Dandel> the new router uses newer wireless chips, namely the atheros AR9381 and atheros AR9380 ( the later is for 5ghz )