2008/10/20/#ubuntu-kernel.txt

[06:07] <NCommander> amitk, ping
[07:30] <NCommander> How does updateconfigs work? (i.e., I don't get how the common config is generated ...)
[08:37] <amitk> NCommander: if you add a new Kconfig option, running updateconfigs will prompt you for that config option for all flavours
[08:38] <NCommander> amitk, I got that far :-)
[08:38] <NCommander> amitk, its more figuring out what the configuration options should be on the new ones :-)
[08:39]  * NCommander is already mostly done, just finishing sparc config
[08:39] <amitk> you just have to know that. updateconfigs can't read developers minds yet :)
[08:39] <NCommander> amitk, Its tedious work
[08:39]  * amitk nods
[08:40] <NCommander> Once I finish 2.6.27, I'll start doing test builds
[08:40]  * NCommander has access to all the port architectures :-)!
[08:40] <amitk> sweet
[08:41] <NCommander> amitk, find out about the kernel.u.c. account?
[08:41]  * NCommander would like a place to hang his tree
[08:45] <amitk> NCommander: will shoot off email today
[08:45] <NCommander> \o\ /o/
[08:46] <NCommander> amitk, generally, if something can be built as a module, it gets built as that. That's the rule of thumb I followed for kernel configuration updating (Linus added at least 30-40 new config options 25 -> 27 ;.;)
[08:52] <amitk> NCommander: probably fine for now, for Jaunty we might turn that rule over its head to improve on boot times.
[08:52] <NCommander> amitk, I already put a note that we need to sanity check the config files
[08:52] <NCommander> (make sure no insane options for an architecture is there)
[08:54] <NCommander> WOOOOOO
[08:54] <NCommander> finally
[08:54] <NCommander> rebased
[08:57] <NCommander> amitk, stupid question, how do I generate a new abi folder? (its pretty clear that 2.6.25 -> 2.6.27 will break an ABI)
[09:01] <abogani> Nafallo: Build process generate it automatically
[09:01] <abogani> Nafallo: Sorry
[09:01] <abogani> I meant NCommander :-)
[09:01] <NCommander> ah
[09:02] <NCommander> so now I get to see how miserable the ABI broke in two releases
[09:02] <NCommander> woow
[09:02] <abogani> ;)
[09:02] <NCommander> so whats sauce?
[09:21] <NCommander> amitk, ^
[09:27] <amitk> NCommander: any patch that isn't upstream yet (those that _should_ go upstream + those that will never go upstream)
[09:28] <NCommander> amitk, oh, that makes sense, anyway, 2.6.27 is building via debuild, so I haven't completely destoried the world yet ;-)
=== mdz_ is now known as mdz
=== _ruben_ is now known as _ruben
[10:39] <NCommander> amitk, hola
[10:55] <mdz> it looks like 263059 may have been caused by ftrace
=== asac_ is now known as asac
[11:17] <NCommander> amitk, I just discovered linux-ports-meta doesn't depend on linux-firmware. Care to help fix that?
=== D2_WON is now known as giovannigiusti
[13:39] <NCommander> hey BenC 
[13:53]  * NCommander now has a building linux-port 2.6.27 kernel :-)!
[13:56] <BenC> NCommander: hey
[13:56] <NCommander> Morning BenC :-)
[13:56] <NCommander> I need some help resolving one last issue with linux-ports-meta before intrepid goes. I discovered this morning that linux-ports-meta doesn't depend on linux-firmware
[13:58] <NCommander> BenC, care to help?
[14:17] <NCommander> BenC, poke?
[14:21] <Keybuk> rtg: which ftrace options did you change?
[14:23] <rtg> Keybuk: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-intrepid.git;a=commit;h=67d9b90a1c844bf1c6daaffd2c60561fc8c445f7
[14:24] <BenC> NCommander: Ok
[14:24]  * NCommander has a git patch, but is somewhat confused by the current linux-ports-meta package
[14:25] <Keybuk> the explanation doesn't really make sense though
[14:25] <Keybuk> it blames a bug on module _unload_
[14:25] <Keybuk> when we're not unloading modules
[14:25] <Keybuk> (both for e1000 and iwl3945)
[14:25] <mjg59> Keybuk: also discarding of init sections
[14:25] <Keybuk> mjg59: -v ?
[14:25] <mjg59> Keybuk: The ftrace issue also arises when the __init sections of modules are discarded after module load
[14:27] <Keybuk> I guess that certainly fits the pattern of where the iwl3945 hang was occurring
[14:29] <rtg> Keybuk: can you reproduce it anymore?
[14:29] <Keybuk> nope
[14:30] <rtg> Keybuk: well, I guess that's proof positive.
[14:30] <Keybuk> something about the way iwl3945 works was causing corruption of ioremapped memory if another module's __init was discarded while it was initialising ??
[14:32] <rtg> Keybuk: got me. I didn't dig into it nearly that deep.
[14:36] <Keybuk> me neither, I don't really understand this level of stuff ;)
[14:36] <Keybuk> I could go and learn, but I'm more inclined to nod sweetly, back away, and stay in my comfort zone :p
[14:37] <rtg> Keybuk: that's pretty much my attitude. I'm buried with other issues.
[14:37] <rtg> there seems to be plenty of confirmation that this fixes it.
[14:37]  * NCommander needs someone who has no fear to help fix the -ports-meta package
[14:38] <Keybuk> obviously someone should post to lkml that disabling ftrace has fixed another issue with another Intel driver ;)
[14:39] <rtg> Keybuk: i bugged the wireless folks last week about it, so I'll at least let them know what the solution was.
[14:43] <BenC> NCommander: A Git patch?
[14:48] <NCommander> BenC, linux-ports-meta is maintained in git
[14:48] <NCommander> (aka git format-patchs)
=== lamont` is now known as lamont
[14:50] <BenC> NCommander: hmm...not sure why, but ok
[14:51] <Keybuk> rtg: is there a bug about the iwl3945 kill switch state not changing in userspace?
[14:51] <rtg> Keybuk: yep. hang on...
[14:52] <rtg> Keybuk: probably this one: https://bugs.edge.launchpad.net/ubuntu/intrepid/+source/linux/+bug/193970
[14:52] <NCommander> BenC, I just made a realization; the ports kernel still has the firmwares compiled in as blobs so that change is unnecessary :-)
[14:52] <rtg> Keybuk: I think slangasek uploaded something in pm-utils to accommodate a sysfs change.
[14:53] <BenC> NCommander: that doesn't pull in the linux-firmware stuff
[14:53] <NCommander> BenC, I don't think its needed unless I'm mistaken
[14:54] <BenC> NCommander: probably not...what about qla or aic scsi firmware?
[14:54] <NCommander> BenC, TBH, I dunno. My powerpc doesn't have any devices that requires firmware
[14:55] <NCommander> BenC, I have the patch: http://paste.ubuntu.com/60124/
[15:02] <BenC> NCommander: You don't want to add it to linux
[15:02] <BenC> NCommander: It needs to be added to linux-image-* if anything
[15:02] <NCommander> BenC, that's how it was changed in linux-lpia ...
[15:03] <BenC> NCommander: then linux-lpia did it wrong :) Check linux-meta
[15:03] <NCommander> Ah
[15:03] <NCommander> I see
[15:03] <NCommander> BenC, in that case, care to sponsor two commits, and two sponsors ;-)?
[15:06] <NCommander> BenC, http://paste.ubuntu.com/60126/
[15:22] <NCommander> Woo, linux-ports 2.6.27 on PowerPC builds!
[15:44] <NCommander> BenC, is my patch ok?
[15:49] <BenC> NCommander: best to ask someone who is actually handling that tree...I don't know enough about it
[15:51]  * NCommander looks up Stefan Bader's IRC name
[15:52] <rtg> NCommander: smb, but he doesn't appear to be around this morning.
[15:53] <NCommander> argh
[15:53]  * BenC wonders where he is
[15:53]  * NCommander turns on the smb light
[15:54]  * NCommander hopes he'll see the Critical bug on linux-ports-meta
[16:04] <NCommander> BenC, out of curosity, what are the requirements of getting a kernel.u.c account
[16:34] <BenC> NCommander: Not much...I'll get your account processed today or tomorrow
[16:34] <BenC> NCommander: inbox is flooded from weekend
[17:05] <G__81> i am interested in being part of kernel team. How can i help 
[17:05] <G__81> i have some linux internal knowledge on how networking subsystem works to an extent. I have one system and have installed Ibex 8.10 in Ibex 
[17:13] <rtg> G__81: start here https://wiki.ubuntu.com/KernelTeam and here https://wiki.ubuntu.com/KernelMaintenance. Also subscribe to the kernel team mailing list.
[17:14] <rtg> here are the list of bugs that need working on: http://people.ubuntu.com/~ogasawara/intrepid-buglist.html
[17:18] <G__81> Yeah thanks i would look into it 
=== thegodfather is now known as fabbione
[21:01] <mnemo> if you cherry pick this 2.6.28 patch for intrepid you will enable users with intel G45 motherboards to boot the ibex --> https://bugs.launchpad.net/ubuntu/+source/linux-meta/+bug/285572
[21:02] <mnemo> i've tested this patch applied to intrepid's .27 kernel and it works perfectly (instead of freeze on xorg login I get full compiz with no xorg.conf hacks)
[21:02] <mnemo> I tested it on a "Gigabyte GA-EG45M-DS2" board but I'm fairly sure it applies to all the G45 desktop motherboards
[21:03] <mnemo> the patch itself is very non-intrusive and doesn't modify any code on non-G45 code paths
[21:38] <Ng> rtg: I think my boot crash is sorted now, just did 31 reboots with 2.6.27-7.12 and it didn't happen
[21:40] <Ng> did the weird 3965 crash disappear as well?
[23:52] <doctormo> Do you guys know about the regression in 2.6.24-21-generic? involves snd_hda_intel and snd_hda_codec missing symbols.
[23:54] <doctormo> lamont: because it's a bitch to have to recompile the kernel to fix the problem.