[12:36] <zul> BenC: http://ozlabs.org/~rusty/paravirt/
[12:38] <zul> ill have a look but it looks like it touches alot of stock files as well
[02:23] <crimsun> missing 'h' in chuck's patch (last hunk) for ATI RS690 HDMI
[10:46] <radone> greetings. I am trying to compile project and I got:  /lib/modules/2.6.15-27-386/build: No such file or directory
[10:46] <radone> linux sources are already installed, but how can I create build directory?
[10:47] <crimsun> you need linux-headers-$(uname -r), not linux-source-2.6.15
[10:54] <radone> crimsun: installed, but without any effect - /lib/modules/2.6.15-27-386/ exists, but sub-directory build/ not :-(
[10:55] <crimsun> well, what's uname -r ?
[10:56] <radone> uname -r: 2.6.15-27-386
[10:56] <crimsun> and what's the dir listing of /usr/src/ ?
[10:56] <crimsun> we should migrate to #ubuntu, btw
[10:58] <radone> linux-headers-2.6.15-27 is included in /usr/src/
[10:58] <crimsun> bingo, you're missing linux-headers-2.6.15-27-386
[10:59] <radone> crimsun: thanks for help. I have overlooked this :)
[11:13] <user__> where is the diff.gz of 2.6.19-5.7? there is no 2.6.19-5 on packages.ubuntu
[11:14] <crimsun> that's because you misread.
[11:14] <crimsun> there is no diff.gz, because it's native.
[11:14] <crimsun> 929c6c7855932231f20b8148235b1d17 59289186 devel optional linux-source-2.6.19_2.6.19-5.7.tar.gz
[11:14] <crimsun> https://lists.ubuntu.com/archives/feisty-changes/2006-November/000206.html
[11:14] <user__> why is there no 2.6.18?
[11:15] <crimsun> because .19 is the target
[12:18] <user__> crimsun, is 2.6.17 using readahead patches?
[12:19] <user__> and where can one obtain the brokenout? of the patchset?
[12:48] <user__> oot@vulaogdx:/usr/src/linux# fakeroot make-kpkg --initrd
[12:48] <user__> exec debian/rules  DEBIAN_REVISION=2.6.18-allesineenpc-1-10.00.Custom  INITRD=YES 
[12:48] <user__> is not working
[12:48] <user__> where did it go wrong?
[12:49] <user__> running edgy
[01:39] <makx> user__ try to specifiy a target also you don't need to be root to build your kernel
[01:40] <makx> user__ also use your home to build something aka ~/src
[01:40] <user__> can I do a request? could ubuntu distribute ck kerenels? 
[01:40] <makx> the ck patchset is useless
[01:40] <makx> mainline is much better
[01:40] <user__> howcome?
[01:40] <makx> if you want hard rt you need to use -rt
[01:40] <user__> what makes it useless?
[01:41] <makx> the O1 scheduler
[01:41] <makx> on 2.4 the -ck patchset was very interesting
[01:41] <user__> is there a -rt(realtime??) patchset?
[01:42] <makx> also requests on irc are lost in the noise so not very effictive :-P
[01:42] <makx> yes ingo takes care of the -rt effort
[01:42] <makx> anyway need to work now
[01:42] <user__> ok last question,what patches also need to be in for good performance?
[01:42] <user__> latency ..
[01:44] <makx> depends on your hardware
[01:46] <user__> can I combine rt with hz patch or other patches from ck? like the hz default patch or the swap prefetch swappiness?
[01:46] <user__> would it benefit ?
[01:46] <makx> rt has dynticks much cooler
[01:48] <user__> so none of the patches in ck would benefit if  one would ues -rt
[06:25] <kylem> Keybuk, ping. what kernel version were you seeing the waitid problems?
[06:27] <Keybuk> kylem: 2.6.17
[06:27] <kylem> k
[06:28] <Keybuk> the code looks identical in 2.6.19
[06:28] <Keybuk>         set_fs(KERNEL_DS);
[06:28] <Keybuk>         ret = sys_waitid(which, pid, (siginfo_t __user *)&info, options,
[06:28] <Keybuk>                          uru ? (struct rusage __user *)&ru : NULL);
[06:28] <Keybuk>         set_fs(old_fs);
[06:28] <Keybuk>         if ((ret < 0) || (info.si_signo == 0))
[06:28] <Keybuk>                 return ret;
[06:28] <Keybuk> note that if ret < 0 or no signal received, it just returns
[06:28] <Keybuk> without ever calling copy_siginto_to_user32
[06:28] <Keybuk> oh, hmm, there's a memset above now
[06:29] <Keybuk> ah
[06:29] <Keybuk> no
[06:29] <Keybuk> that memset is for the 64-bit sigingo
[06:29] <kylem> right.
[06:29] <Keybuk> uinfo is never cleared, and can be returned as passed
[06:29] <kylem> er, shouldn't there have been a copy_siginfo_from_user?
[06:30] <Keybuk> no, the struct is not used
[06:30] <Keybuk> ie. it's only to return data
[06:30] <Keybuk> not to pass in
[06:30] <kylem> ahh. righto.
[06:33] <Keybuk> (note that do_wait deliberately clears the struct when returning certain conditions
[06:33] <Keybuk> so that sys_waitid gets a cleared struct
[06:34] <Keybuk> but because compat_sys_waitid exits early, it never copies the cleared struct into the wider one)