* Deadlocks with 2.6.1 + -mm sysfs patches
@ 2004-01-10 20:03 Frédéric L. W. Meunier
2004-01-10 22:55 ` Mathieu Segaud
2004-01-12 2:32 ` Frédéric L. W. Meunier
0 siblings, 2 replies; 3+ messages in thread
From: Frédéric L. W. Meunier @ 2004-01-10 20:03 UTC (permalink / raw)
To: linux-hotplug
Well, maybe one of those caused it. It doesn't happen with
plain 2.6.1-rc2. Patches taken from
http://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.1/2.6.1-mm2/broken-out/
. I didn't apply sysfs-add-video-class.patch because I use
matroxfb-2.6.0.gz.
I also didn't apply CONFIG_SYSFS-boot-from-disk-fix.patch and
sysfs-pin-kobject.patch. Am I missing other patches ? Using
latest hotplug and udev.
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-simple-class-device-support.patch
patching file drivers/base/class_simple.c
patching file drivers/base/Makefile
patching file include/linux/device.h
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-remove-tty-class-device-logic.patch
patching file drivers/char/tty_io.c
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-misc-class.patch
patching file drivers/char/misc.c
patching file include/linux/miscdevice.h
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-mem-device-support.patch
patching file drivers/char/mem.c
Hunk #2 succeeded at 677 (offset 1 line).
Hunk #3 succeeded at 688 (offset 1 line).
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-input-class-support.patch
patching file drivers/input/evdev.c
patching file drivers/input/joydev.c
patching file drivers/input/mousedev.c
Hunk #1 succeeded at 203 (offset -22 lines).
Hunk #2 succeeded at 465 (offset -22 lines).
Hunk #3 succeeded at 546 (offset -22 lines).
Hunk #4 succeeded at 566 (offset -22 lines).
patching file drivers/input/tsdev.c
Hunk #1 succeeded at 119 (offset -10 lines).
Hunk #2 succeeded at 334 (offset -10 lines).
patching file drivers/usb/input/hid-input.c
patching file include/linux/input.h
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/devfs-d_revalidate-oops-fix.patch
patching file fs/devfs/base.c
Hunk #1 succeeded at 2185 (offset 14 lines).
Hunk #2 succeeded at 2204 (offset 14 lines).
Hunk #3 succeeded at 2222 (offset 14 lines).
Hunk #4 succeeded at 2235 (offset 14 lines).
Hunk #5 succeeded at 2267 (offset 14 lines).
Hunk #6 succeeded at 2321 (offset 14 lines).
Hunk #7 succeeded at 2357 (offset 14 lines).
Hunk #8 succeeded at 2601 (offset 14 lines).
patching file fs/namei.c
patching file include/linux/fs.h
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/devfs-joystick-fix.patch
patching file drivers/input/joydev.c
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-oss-class.patch
patching file sound/oss/soundcard.c
patching file sound/sound_core.c
% patch -p1 < /home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-alsa-class.patch
patching file include/sound/core.h
patching file sound/core/sound.c
patching file sound/pci/intel8x0.c
Hunk #1 succeeded at 2524 (offset 2 lines).
First, I didn't notice anything from
/home/ftp/pub/src/Linux/kernel/patches/2.6/applied/mm/sysfs-add-mem-device-support.patch,
like /dev/mem.
I guess it was because the deadlocks happened ? ps aux reported
2 /etc/hotplug/input.agent input or something (I'm sure about
the input part). /dev/event1 and /dev/mouse0 were created, but
the mouse didn't work.
I also noticed
Jan 10 16:36:24 pervalidus kernel: usb 2-2: control timeout on ep0in
Jan 10 16:36:39 pervalidus last message repeated 7 times
which don't show up on plain 2.6.1-rc2.
--
http://www.pervalidus.net/contact.html
-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Deadlocks with 2.6.1 + -mm sysfs patches
2004-01-10 20:03 Deadlocks with 2.6.1 + -mm sysfs patches Frédéric L. W. Meunier
@ 2004-01-10 22:55 ` Mathieu Segaud
2004-01-12 2:32 ` Frédéric L. W. Meunier
1 sibling, 0 replies; 3+ messages in thread
From: Mathieu Segaud @ 2004-01-10 22:55 UTC (permalink / raw)
To: linux-hotplug
Frédéric L. W. Meunier <1@pervalidus.net> writes:
> Well, maybe one of those caused it. It doesn't happen with
> plain 2.6.1-rc2. Patches taken from
> http://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.1/2.6.1-mm2/broken-out/
> . I didn't apply sysfs-add-video-class.patch because I use
> matroxfb-2.6.0.gz.
>
> I also didn't apply CONFIG_SYSFS-boot-from-disk-fix.patch and
> sysfs-pin-kobject.patch. Am I missing other patches ? Using
> latest hotplug and udev.
I think you should apply sysfs-pin-kobject.patch
It fixes some use-after-free races in sysfs code (however, still got
one of those ealier today when unloading alsa snd_powermac driver on my iBook)
[snip]
In fact, without this patch, code hits a BUG() in dget() and with it,
it oopses a little later in sysfs_remove_dir(), just before dput()-ing
the dentry (it oopses when putting d pointer in r3 register just
before calling dput(), so seems d is still freed to early)
I am not sure about my interpretation; here are the oopses (with and
without this patch)
1) without sysfs-pin-kobjetc.patch
Jan 8 23:48:30 rivendell kernel: kernel BUG in dget at include/linux/dcache.h:2
74!
Jan 8 23:48:30 rivendell kernel: Oops: Exception in kernel mode, sig: 5 [#1]
Jan 8 23:48:30 rivendell kernel: NIP: C008E318 LR: C00C14B0 SP: C76A1D00 REGS:
c76a1c50 TRAP: 0700 Not tainted
Jan 8 23:48:30 rivendell kernel: MSR: 00029032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 1
1
Jan 8 23:48:30 rivendell kernel: TASK = d72adca0[926] 'modprobe' Last syscall:
129
Jan 8 23:48:30 rivendell kernel: GPR00: 00000000 C76A1D00 D72ADCA0 D5F9EF74 D7F
FC400 00C4899F 00000000 00000000
Jan 8 23:48:30 rivendell kernel: GPR08: 00000015 00000014 D7FFC410 00000050 60B
A5D56 1001D4C8 00000000 10016078
Jan 8 23:48:30 rivendell kernel: GPR16: 00000000 00000000 00000000 10000000 000
00001 10010000 00000000 00000000
Jan 8 23:48:30 rivendell kernel: GPR24: 00000002 10001118 00000000 DB0B7FCC DB0
B8018 DB0B802C D6AEA080 D5F9EF74
Jan 8 23:48:30 rivendell kernel: Call trace:
Jan 8 23:48:30 rivendell kernel: [c00c14b0] kobject_del+0x70/0x9c
Jan 8 23:48:30 rivendell kernel: [c00e7558] class_device_del+0xe0/0x140
Jan 8 23:48:30 rivendell kernel: [c00e75d0] class_device_unregister+0x18/0x34
Jan 8 23:48:30 rivendell kernel: [c00e7ad0] simple_remove_class_device+0x8c/0x9c
Jan 8 23:48:30 rivendell kernel: [db0b6260] sound_remove_unit+0x2c/0x48 [soundcore]
Jan 8 23:48:30 rivendell kernel: [db0d1098] snd_unregister_oss_device+0x60/0x154 [snd]
Jan 8 23:48:30 rivendell kernel: [db3baa30] snd_mixer_oss_notify_handler+0x88/0x244 [snd_mixer_oss]
Jan 8 23:48:30 rivendell kernel: [db3bac4c] alsa_mixer_oss_exit+0x60/0xa84 [snd_mixer_oss]
Jan 8 23:48:30 rivendell kernel: [c002f958] sys_delete_module+0x1a0/0x234
Jan 8 23:48:30 rivendell kernel: [c0005e7c] ret_from_syscall+0x0/0x44
2)with sysfs-pin-kobject.patch
Jan 10 15:23:05 rivendell kernel: Oops: kernel access of bad area, sig: 11 [#1]
Jan 10 15:23:05 rivendell kernel: NIP: C0072034 LR: C008CF90 SP: D7201DB0 REGS:
d7201d00 TRAP: 0301 Not tainted
Jan 10 15:23:05 rivendell kernel: MSR: 00009032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 1
1
Jan 10 15:23:05 rivendell kernel: DAR: 0000002C, DSISR: 40000000
Jan 10 15:23:05 rivendell kernel: TASK = d5614e40[1183] 'modprobe' Last syscall:
129
Jan 10 15:23:05 rivendell kernel: GPR00: C008CF90 D7201DB0 D5614E40 D6212BC0 D62
12BC0 00000000 C02B0000 D7FF8400
Jan 10 15:23:05 rivendell kernel: GPR08: 0000002F 0000002E D7FFDC70 00000000 695
687AE
Jan 10 15:23:05 rivendell kernel: Call trace:
Jan 10 15:23:05 rivendell kernel: [c008cf90] sysfs_remove_dir+0xf0/0x104
Jan 10 15:23:05 rivendell kernel: [c00bf554] kobject_del+0x70/0x9c
Jan 10 15:23:05 rivendell kernel: [c00e58a8] class_device_del+0xe0/0x140
Jan 10 15:23:05 rivendell kernel: [c00e5920] class_device_unregister+0x18/0x34
Jan 10 15:23:05 rivendell kernel: [c00e5e20] simple_remove_class_device+0x8c/0x9c
Jan 10 15:23:05 rivendell kernel: [db0b6260] sound_remove_unit+0x2c/0x48 [soundcore]
Jan 10 15:23:05 rivendell kernel: [db0d2be8] snd_unregister_oss_device+0x60/0x154 [snd]
Jan 10 15:23:05 rivendell kernel: [db58a244] unregister_device+0x68/0x104 [snd_seq_oss]
Jan 10 15:23:05 rivendell kernel: [c002f840] sys_delete_module+0x1a0/0x234
Jan 10 15:23:05 rivendell kernel: [c0005e7c] ret_from_syscall+0x0/0x44
Was not able to track where or when it is freed (this crash does not
occur each time I unload snd-powermac, it seems quite random)
--
Mathieu Segaud
HP LaserJetIII wrote:
> How to turn off faucet?
>
Now that's a good one! Somebody's mucking with my print-server.
Sorry. I'm gonna get my gun....
- Richard Johnson on linux-kernel
-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Deadlocks with 2.6.1 + -mm sysfs patches
2004-01-10 20:03 Deadlocks with 2.6.1 + -mm sysfs patches Frédéric L. W. Meunier
2004-01-10 22:55 ` Mathieu Segaud
@ 2004-01-12 2:32 ` Frédéric L. W. Meunier
1 sibling, 0 replies; 3+ messages in thread
From: Frédéric L. W. Meunier @ 2004-01-12 2:32 UTC (permalink / raw)
To: linux-hotplug
On Sat, 10 Jan 2004, Mathieu Segaud wrote:
> Frédéric L. W. Meunier writes:
>
> > Well, maybe one of those caused it. It doesn't happen with
> > plain 2.6.1-rc2. Patches taken from
> > http://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.1/2.6.1-mm2/broken-out/
> > . I didn't apply sysfs-add-video-class.patch because I use
> > matroxfb-2.6.0.gz.
> >
> > I also didn't apply CONFIG_SYSFS-boot-from-disk-fix.patch and
> > sysfs-pin-kobject.patch. Am I missing other patches ? Using
> > latest hotplug and udev.
>
> I think you should apply sysfs-pin-kobject.patch
> It fixes some use-after-free races in sysfs code (however, still got
> one of those ealier today when unloading alsa snd_powermac driver on my iBook)
OK, I applied the patch. I'm not sure if it changed anything.
I may have rebooted too fast last time.
I noticed my load avarage was increasing up to 1,50 or
something. ps reported:
root 534 0.0 0.2 2396 1204 ? S< 23:59 0:00 /bin/sh /sbin/hotplug input
root 545 0.0 0.2 2396 1204 ? S< 23:59 0:00 /bin/sh /sbin/hotplug input
root 578 0.0 0.1 1552 584 ? D< 23:59 0:00 /etc/hotplug.d/default/udev.hotplug input
root 579 0.0 0.1 1552 584 ? D< 23:59 0:00 /etc/hotplug.d/default/udev.hotplug input
After a few minutes I noticed it dropped and the processes
weren't here anymore.
And again there were:
Jan 11 23:59:16 pervalidus kernel: usb 2-2: control timeout on ep0in
Jan 11 23:59:46 pervalidus last message repeated 11 times
--
http://www.pervalidus.net/contact.html
-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2004-01-12 2:32 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-01-10 20:03 Deadlocks with 2.6.1 + -mm sysfs patches Frédéric L. W. Meunier
2004-01-10 22:55 ` Mathieu Segaud
2004-01-12 2:32 ` Frédéric L. W. Meunier
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).