View previous topic :: View next topic |
Author |
Message |
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sat Feb 26, 2011 9:08 pm Post subject: |
|
|
Nope, just the 2.6.37.2 tag from git with that on top. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Sat Feb 26, 2011 9:12 pm Post subject: |
|
|
Ant_P wrote: | Nope, just the 2.6.37.2 tag from git with that on top. |
yes, it's some extra work
but could you try whether you get the same error with the tarball from kernel.org ?
linux-2.6.37.2.tar.bz2
perhaps patch or anything gets confused by git ?
md5sum linux-2.6.37.2.tar.bz2 wrote: |
89f681bc7c917a84aa7470da6eed5101 linux-2.6.37.2.tar.bz2 |
I just went through some heavy patch marathon and it seems patch is not so clever after all,
it often leaves stuff out (or in) during applies & reverts
edit:
how badly do you want THP (Transparent Huge Pages), btw ?
edit2:
I didn't even know that the transparent huge pages are already included in 2.6.38 - thought they were planned for 2.6.39
so that's the reason 2.6.38 got that boost: THP & Nick's VFS optimizations
edit3:
weird, I just compiled (2.6.37.2_plus_v1_BFS_no-compaction_zcache.patch) and it worked fine - no error message
like already written above: please try to apply it to the plain tarball - whether that makes a difference
could you check the md5sum whether it's correct ?
Quote: | efb62a16edae2b678d54597864c93503 ../2.6.37.2_plus_v1_BFS_no-compaction_zcache.patch |
_________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sun Feb 27, 2011 1:54 pm Post subject: |
|
|
OK, I gave it a try on top of a plain tarball. Same error.
EDIT: it works with BFS disabled... but if I do that I might as well use vanilla 2.6.38.
EDIT2: It worked once I enabled the "hot-plug CPUs" option. Dunno why it breaks without that... |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Sun Feb 27, 2011 2:27 pm Post subject: |
|
|
Ant_P wrote: | OK, I gave it a try on top of a plain tarball. Same error.
EDIT: it works with BFS disabled... but if I do that I might as well use vanilla 2.6.38.
EDIT2: It worked once I enabled the "hot-plug CPUs" option. Dunno why it breaks without that... |
I don't know what you are talking about
take a closer look at the first page - especially at the top
or above your post (the 2nd)
there you clearly see that THP (transparent hugepages) patchsets are available
TRY THEM
edit:
ok, I'm compile-checking those builds again - perhaps I missed something _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sun Feb 27, 2011 3:29 pm Post subject: |
|
|
Yes, I also got the error with the -thp patch. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sun Feb 27, 2011 10:20 pm Post subject: |
|
|
Works. BFS+THP. fglrx is broken with THP but I don't care about that. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Sun Feb 27, 2011 10:27 pm Post subject: |
|
|
Ant_P wrote: | Works. BFS+THP. fglrx is broken with THP but I don't care about that. |
awesome!
I'm also not using fglrx - so I can't say anything about that
xf86-video-ati + Evergreen (5850) works great with it
albright wrote: | maybe this is getting off topic ...
but I spoke too soon - it's very strange but every other
boot, the kerneloftruth patched 2.6.37.2 fails; it's hard
to see any of the boot messages but the first error
appears to be something like:
kernel bug at vmalloc.c
It goes downhill from there
If I use sysrq to reboot, the next time it works ... and so on |
what exact kernel patch are you using ? _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
albright Advocate
Joined: 16 Nov 2003 Posts: 2588 Location: Near Toronto
|
Posted: Fri Mar 04, 2011 1:19 pm Post subject: |
|
|
with the 2.6.37.2_plus_v1_BFS_compaction_TOI.patch I am
getting a weird problem:
every other boot (more or less - it is kind of random) I get
a "kernel bug at vmalloc.c" message and the boot process
is totally corrupted (modules won't load, udev fails etc)
The error occurs - it seems to me - exactly when the kernel
should switch on KMS.
Normally, at that point the console resolution changes and boot
proceeds.
And like I say, about half the time this works, the other half it
doesn't.
Anyone have any idea what could cause this?
My hardware is thinkpad x300,
Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 0c) _________________ .... there is nothing - absolutely nothing - half so much worth
doing as simply messing about with Linux ...
(apologies to Kenneth Graeme) |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Fri Mar 04, 2011 3:52 pm Post subject: |
|
|
albright wrote: | with the 2.6.37.2_plus_v1_BFS_compaction_TOI.patch I am
getting a weird problem:
every other boot (more or less - it is kind of random) I get
a "kernel bug at vmalloc.c" message and the boot process
is totally corrupted (modules won't load, udev fails etc)
The error occurs - it seems to me - exactly when the kernel
should switch on KMS.
Normally, at that point the console resolution changes and boot
proceeds.
And like I say, about half the time this works, the other half it
doesn't.
Anyone have any idea what could cause this?
My hardware is thinkpad x300,
Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller (rev 0c) |
well, you could try to do it like me:
not compiling in the KMS mode to be default and then manually loading (or let it load) the intel kernel module (dunno what its name is)
that way I don't get the error you mentioned (using radeon)
afaik I also got it when KMS was enabled by default and the kernel would even hardlock at boot - never would make it into init _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Fri Mar 04, 2011 6:40 pm Post subject: |
|
|
This THP kernel's been working well, everything runs fine apart from one thing... Firefox. And it does this:
Code: | [ 3652.396391] ------------[ cut here ]------------
[ 3652.396482] kernel BUG at mm/memory.c:1380!
[ 3652.396555] invalid opcode: 0000 [#1] PREEMPT SMP
[ 3652.396647] last sysfs file: /sys/devices/system/cpu/cpu3/cache/index2/shared_cpu_map
[ 3652.396774] CPU 1
[ 3652.396809] Modules linked in: cpufreq_ondemand snd_aloop fbcon font bitblit softcursor radeon ttm drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea cfbimgblt cfbfillrect snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm snd_timer snd soundcore snd_page_alloc
[ 3652.397309]
[ 3652.397340] Pid: 2253, comm: firefox Not tainted 2.6.37.2 #3 Gigabyte Technology Co., Ltd. GA-MA770-UD3/GA-MA770-UD3
[ 3652.397529] RIP: 0010:[<ffffffff810caa4c>] [<ffffffff810caa4c>] follow_page+0x3dc/0x3e0
[ 3652.397673] RSP: 0018:ffff88017cded9c8 EFLAGS: 00010202
[ 3652.397760] RAX: 80000001710002e7 RBX: 00007fb8b3fff000 RCX: 0000000000000000
[ 3652.397877] RDX: ffff88017a906cf8 RSI: ffff88017688d998 RDI: ffff8801aca92380
[ 3652.397993] RBP: ffff88017cdeda28 R08: 0000000000000007 R09: ffff88017cdedb90
[ 3652.398108] R10: ffff88017cdedb90 R11: 0000000000000007 R12: ffff8801aca92300
[ 3652.398223] R13: 0000000000000007 R14: ffff880176811630 R15: 0000000000000001
[ 3652.398340] FS: 00007fb8d724d720(0000) GS:ffff8800cfa80000(0000) knlGS:0000000000000000
[ 3652.398470] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[ 3652.398564] CR2: 00007fb8c9e709c8 CR3: 000000017cc41000 CR4: 00000000000006e0
[ 3652.398680] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 3652.398795] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 3652.398912] Process firefox (pid: 2253, threadinfo ffff88017cdec000, task ffff88017abdbde0)
[ 3652.399045] Stack:
[ 3652.399080] ffff88017cded9e8 ffff880000000000 00003ffffffff000 ffff88017a906cf8
[ 3652.399214] 0000000000000000 ffff8801aca92380 ffff88017cdeda28 ffff880176811630
[ 3652.399347] ffff88017abdbde0 ffff8801aca92300 0000000000000007 0000000000000000
[ 3652.399481] Call Trace:
[ 3652.399529] [<ffffffff810cc9fe>] __get_user_pages+0x11e/0x450
[ 3652.399629] [<ffffffff810ccdc4>] get_user_pages+0x44/0x50
[ 3652.399643] [<ffffffff81058177>] get_user_pages_fast+0x147/0x190
[ 3652.399643] [<ffffffff8108e3fe>] get_futex_key+0x9e/0x1f0
[ 3652.399643] [<ffffffff811034f0>] ? __pollwait+0x0/0xf0
[ 3652.399643] [<ffffffff8108e882>] futex_wait_setup+0x52/0x100
[ 3652.400027] [<ffffffff81336b91>] ? skb_dequeue+0x61/0x80
[ 3652.400027] [<ffffffff8108eafd>] futex_wait+0x1cd/0x360
[ 3652.400027] [<ffffffff8132e7c8>] ? sock_aio_read+0x108/0x110
[ 3652.400027] [<ffffffff8109091d>] do_futex+0x11d/0xb20
[ 3652.400027] [<ffffffff810f2ab2>] ? do_sync_read+0xd2/0x110
[ 3652.400027] [<ffffffff81091396>] sys_futex+0x76/0x170
[ 3652.400027] [<ffffffff81104a77>] ? sys_poll+0x77/0x100
[ 3652.400027] [<ffffffff8102dd7b>] system_call_fastpath+0x16/0x1b
[ 3652.400027] Code: ef 48 8b 7d c8 48 89 55 b8 e8 31 bf 31 00 48 8b 55 b8 48 8b 02 e9 24 fe ff ff 0f ae f0 48 8b 02 a9 80 02 00 00 0f 84 13 fe ff ff <0f> 0b 66 90 55 b8 da ff ff ff 48 89 e5 48 83 ec 40 48 89 5d d8
[ 3652.400027] RIP [<ffffffff810caa4c>] follow_page+0x3dc/0x3e0
[ 3652.400027] RSP <ffff88017cded9c8>
[ 3652.422227] ---[ end trace 4f512a163ea6819c ]---
|
|
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
Posted: Fri Mar 04, 2011 7:42 pm Post subject: |
|
|
Ant_P wrote: | This THP kernel's been working well, everything runs fine apart from one thing... Firefox. And it does this:
Code: | [ 3652.396391] ------------[ cut here ]------------
[ 3652.396482] kernel BUG at mm/memory.c:1380!
[ 3652.396555] invalid opcode: 0000 [#1] PREEMPT SMP
[ 3652.396647] last sysfs file: /sys/devices/system/cpu/cpu3/cache/index2/shared_cpu_map
[ 3652.396774] CPU 1
[ 3652.396809] Modules linked in: cpufreq_ondemand snd_aloop fbcon font bitblit softcursor radeon ttm drm_kms_helper drm fb fbdev i2c_algo_bit cfbcopyarea cfbimgblt cfbfillrect snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm snd_timer snd soundcore snd_page_alloc
[ 3652.397309]
[ 3652.397340] Pid: 2253, comm: firefox Not tainted 2.6.37.2 #3 Gigabyte Technology Co., Ltd. GA-MA770-UD3/GA-MA770-UD3
[ 3652.397529] RIP: 0010:[<ffffffff810caa4c>] [<ffffffff810caa4c>] follow_page+0x3dc/0x3e0
[ 3652.397673] RSP: 0018:ffff88017cded9c8 EFLAGS: 00010202
[ 3652.397760] RAX: 80000001710002e7 RBX: 00007fb8b3fff000 RCX: 0000000000000000
[ 3652.397877] RDX: ffff88017a906cf8 RSI: ffff88017688d998 RDI: ffff8801aca92380
[ 3652.397993] RBP: ffff88017cdeda28 R08: 0000000000000007 R09: ffff88017cdedb90
[ 3652.398108] R10: ffff88017cdedb90 R11: 0000000000000007 R12: ffff8801aca92300
[ 3652.398223] R13: 0000000000000007 R14: ffff880176811630 R15: 0000000000000001
[ 3652.398340] FS: 00007fb8d724d720(0000) GS:ffff8800cfa80000(0000) knlGS:0000000000000000
[ 3652.398470] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[ 3652.398564] CR2: 00007fb8c9e709c8 CR3: 000000017cc41000 CR4: 00000000000006e0
[ 3652.398680] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 3652.398795] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 3652.398912] Process firefox (pid: 2253, threadinfo ffff88017cdec000, task ffff88017abdbde0)
[ 3652.399045] Stack:
[ 3652.399080] ffff88017cded9e8 ffff880000000000 00003ffffffff000 ffff88017a906cf8
[ 3652.399214] 0000000000000000 ffff8801aca92380 ffff88017cdeda28 ffff880176811630
[ 3652.399347] ffff88017abdbde0 ffff8801aca92300 0000000000000007 0000000000000000
[ 3652.399481] Call Trace:
[ 3652.399529] [<ffffffff810cc9fe>] __get_user_pages+0x11e/0x450
[ 3652.399629] [<ffffffff810ccdc4>] get_user_pages+0x44/0x50
[ 3652.399643] [<ffffffff81058177>] get_user_pages_fast+0x147/0x190
[ 3652.399643] [<ffffffff8108e3fe>] get_futex_key+0x9e/0x1f0
[ 3652.399643] [<ffffffff811034f0>] ? __pollwait+0x0/0xf0
[ 3652.399643] [<ffffffff8108e882>] futex_wait_setup+0x52/0x100
[ 3652.400027] [<ffffffff81336b91>] ? skb_dequeue+0x61/0x80
[ 3652.400027] [<ffffffff8108eafd>] futex_wait+0x1cd/0x360
[ 3652.400027] [<ffffffff8132e7c8>] ? sock_aio_read+0x108/0x110
[ 3652.400027] [<ffffffff8109091d>] do_futex+0x11d/0xb20
[ 3652.400027] [<ffffffff810f2ab2>] ? do_sync_read+0xd2/0x110
[ 3652.400027] [<ffffffff81091396>] sys_futex+0x76/0x170
[ 3652.400027] [<ffffffff81104a77>] ? sys_poll+0x77/0x100
[ 3652.400027] [<ffffffff8102dd7b>] system_call_fastpath+0x16/0x1b
[ 3652.400027] Code: ef 48 8b 7d c8 48 89 55 b8 e8 31 bf 31 00 48 8b 55 b8 48 8b 02 e9 24 fe ff ff 0f ae f0 48 8b 02 a9 80 02 00 00 0f 84 13 fe ff ff <0f> 0b 66 90 55 b8 da ff ff ff 48 89 e5 48 83 ec 40 48 89 5d d8
[ 3652.400027] RIP [<ffffffff810caa4c>] follow_page+0x3dc/0x3e0
[ 3652.400027] RSP <ffff88017cded9c8>
[ 3652.422227] ---[ end trace 4f512a163ea6819c ]---
|
|
sorry about that
I get the same with seamonkey
seems like that revision of transparent hugepages isn't so stable (probably the reason about the issues with 2.6.38* which surely got fixed meanwhile)
I've for now reverted back to the version with compaction since there also appearing some problem (similar) when compiling android & enabling overcommit from time to time
there's also the possibility to use THP on top of vanilla 2.6.37.2 and combining other parts - if you'd like to:
1_2.6.37.2_thp33.patch (258.84 KB)
that's the version I'm using & ported from 2.6.37-rc5 _________________ https://github.com/kernelOfTruth/ZFS-for-SystemRescueCD/tree/ZFS-for-SysRescCD-4.9.0
https://github.com/kernelOfTruth/pulseaudio-equalizer-ladspa
Hardcore Gentoo Linux user since 2004 |
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Fri Mar 04, 2011 8:30 pm Post subject: |
|
|
No problem, Chromium's my main browser anyway. It's worth the risk for a few % speedup as I've been doing a lot of stuff with large (600Mpx+) images... |
|
Back to top |
|
|
raj7095 n00b
Joined: 08 Feb 2011 Posts: 59
|
Posted: Tue Mar 08, 2011 1:30 am Post subject: |
|
|
kerneloftruth, are you still going to continue your patchset into version 2.6.38 once it comes out? |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
bingyuan n00b
Joined: 20 Feb 2011 Posts: 13
|
Posted: Tue Mar 08, 2011 3:43 pm Post subject: |
|
|
kernelOfTruth wrote: | probably not - or I'm only going to push out 1 release
currently I need to focus more on studying
thanks |
What a pity... |
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Sat Mar 12, 2011 4:49 pm Post subject: |
|
|
OK, actual THP numbers on Folding@Home 6.34:
BFS no THP: 7m35s per 1%
BFS with THP: 7m19s per 1%
3% or so faster with no manual intervention... and that's on top of the 25% speedup BFS had over CFS already. Not bad. |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
bingyuan n00b
Joined: 20 Feb 2011 Posts: 13
|
Posted: Fri Apr 15, 2011 2:14 pm Post subject: |
|
|
@kernelOfTruth, Welcome back. So you have time now? When for 2.6.38? |
|
Back to top |
|
|
kernelOfTruth Watchman
Joined: 20 Dec 2005 Posts: 6111 Location: Vienna, Austria; Germany; hello world :)
|
|
Back to top |
|
|
bingyuan n00b
Joined: 20 Feb 2011 Posts: 13
|
Posted: Sat Apr 16, 2011 3:19 pm Post subject: |
|
|
@kernelOfTruth,
that's OK. I have patched Linux-2.6.38 with BFQ-v2, reiser4-2.6.38 and BFS-0.400 myself, now works fine, . |
|
Back to top |
|
|
mbar Veteran
Joined: 19 Jan 2005 Posts: 1990 Location: Poland
|
Posted: Thu Apr 21, 2011 8:36 am Post subject: |
|
|
Seems that I have some trouble compiling on .38:
Code: | CC kernel/sched.o
In file included from kernel/sched.c:2:0:
kernel/sched_bfs.c:3074:5: error: conflicting types for ‘mutex_spin_on_owner’
include/linux/sched.h:368:12: note: previous declaration of ‘mutex_spin_on_owner’ was here
kernel/sched_bfs.c: In function ‘mutex_spin_on_owner’:
kernel/sched_bfs.c:3111:19: warning: comparison of distinct pointer types lacks a cast
make[1]: *** [kernel/sched.o] Błąd 1
make: *** [kernel] Błąd 2 |
But the BFS 400 patch applied cleanly. |
|
Back to top |
|
|
|