2008-12-10 13:15:28

by Jiri Slaby

[permalink] [raw]
Subject: mmotm 2008-12-09-15-24: e1000e mutex contention

Hi,

during bootup, I got ton of these:

WARNING: at drivers/net/e1000e/ich8lan.c:413
e1000_acquire_swflag_ich8lan+0x43/0xf0()
Hardware name: To Be Filled By O.E.M.
e1000e mutex contention. Owned by pid -1
Modules linked in:
Pid: 1, comm: swapper Tainted: G W 2.6.28-rc7-mm1_64 #490
Call Trace:
[<ffffffff8023e25f>] warn_slowpath+0xaf/0x110
[<ffffffff803c1b1d>] ? vt_console_print+0x26d/0x380
[<ffffffff8023e403>] ? __call_console_drivers+0x73/0x90
[<ffffffff80259954>] ? up+0x34/0x50
[<ffffffff8023e98d>] ? release_console_sem+0x1bd/0x210
[<ffffffff803fe333>] e1000_acquire_swflag_ich8lan+0x43/0xf0
[<ffffffff803fe748>] e1000_read_nvm_ich8lan+0x78/0x130
[<ffffffff80401ad1>] e1000e_read_pba_num+0x31/0x80
[<ffffffff80589609>] e1000_probe+0xcb9/0xd00
[<ffffffff8035d46a>] ? kobject_get+0x1a/0x30
[<ffffffff80370092>] local_pci_probe+0x12/0x20
[<ffffffff80370c50>] pci_device_probe+0x80/0xa0
[<ffffffff803f36eb>] ? driver_sysfs_add+0x6b/0x90
[<ffffffff803f3832>] driver_probe_device+0xa2/0x1e0
[<ffffffff803f3a0b>] __driver_attach+0x9b/0xa0
[<ffffffff803f3970>] ? __driver_attach+0x0/0xa0
[<ffffffff803f2ff3>] bus_for_each_dev+0x63/0x90
[<ffffffff80791560>] ? e1000_init_module+0x0/0x70
[<ffffffff803f367c>] driver_attach+0x1c/0x20
[<ffffffff803f28a8>] bus_add_driver+0x1f8/0x270
[<ffffffff80791560>] ? e1000_init_module+0x0/0x70
[<ffffffff803f3bf0>] driver_register+0x70/0x160
[<ffffffff80791560>] ? e1000_init_module+0x0/0x70
[<ffffffff80370f2a>] __pci_register_driver+0x5a/0xa0
[<ffffffff80791560>] ? e1000_init_module+0x0/0x70
[<ffffffff807915af>] e1000_init_module+0x4f/0x70
[<ffffffff80791360>] ? loop_init+0x0/0x1dc
[<ffffffff8020903c>] _stext+0x3c/0x170
[<ffffffff8027f068>] ? register_irq_proc+0xb8/0xe0
[<ffffffff80300000>] ? load_script+0x70/0x250
[<ffffffff807726ab>] kernel_init+0x166/0x1c2
[<ffffffff8020d5ba>] child_rip+0xa/0x20
[<ffffffff80772545>] ? kernel_init+0x0/0x1c2
[<ffffffff8020d5b0>] ? child_rip+0x0/0x20
---[ end trace 9c09681510f4b604 ]---

I've found few reports of this in the past, not sure if this one is related.
The last known working is some mmotm based on -rc5.


2008-12-10 16:49:56

by Jesse Brandeburg

[permalink] [raw]
Subject: RE: mmotm 2008-12-09-15-24: e1000e mutex contention

Jiri Slaby wrote:
> during bootup, I got ton of these:
>
> WARNING: at drivers/net/e1000e/ich8lan.c:413
> e1000_acquire_swflag_ich8lan+0x43/0xf0()
> Hardware name: To Be Filled By O.E.M.
> e1000e mutex contention. Owned by pid -1

I think these are probably due to akpm's patches yesterday. I think
they've already been dropped. The original issue still exists however, as
reported by Arjan (kerneloops) and others.

We have Dave Graham here working on this, he is not done yet but he is
making progress. We hope to have something to post soon.

Jesse


Attachments:
smime.p7s (6.55 kB)

2008-12-10 17:21:37

by Graham, David

[permalink] [raw]
Subject: RE: mmotm 2008-12-09-15-24: e1000e mutex contention

Yes we've got a lot (about 150 on kernelmoops last time I looked), of these reports now, all warnings. They are different than the double-release of a mutex that was found & fixed Yesterday by Andrew Morton, as these are in-e1000e-driver warnings from a mutex_trylock(), and have not, as far as I know, been flagged by LOCKDEP. I am reviewing all the reports to see if the total set of warnings show that driver is at risk for a deadlock. If so, obviously we'll fix it. If not, we might just remove the trylock warning.

Dave

-----Original Message-----
From: Brandeburg, Jesse
Sent: Wednesday, December 10, 2008 8:50 AM
To: Jiri Slaby; Kirsher, Jeffrey T; David Miller
Cc: Allan, Bruce W; Waskiewicz Jr, Peter P; Ronciak, John; [email protected]; Andrew Morton; Linux kernel mailing list; Graham, David
Subject: RE: mmotm 2008-12-09-15-24: e1000e mutex contention

Jiri Slaby wrote:
> during bootup, I got ton of these:
>
> WARNING: at drivers/net/e1000e/ich8lan.c:413
> e1000_acquire_swflag_ich8lan+0x43/0xf0()
> Hardware name: To Be Filled By O.E.M.
> e1000e mutex contention. Owned by pid -1

I think these are probably due to akpm's patches yesterday. I think
they've already been dropped. The original issue still exists however, as
reported by Arjan (kerneloops) and others.

We have Dave Graham here working on this, he is not done yet but he is
making progress. We hope to have something to post soon.

Jesse
????{.n?+???????+%?????ݶ??w??{.n?+????{??G?????{ay?ʇڙ?,j??f???h?????????z_??(?階?ݢj"???m??????G????????????&???~???iO???z??v?^?m???? ????????I?