Bug 10711

Summary:	BUG: unable to handle kernel paging request - scsi_bus_uevent
Product:	IO/Storage	Reporter:	Rafael J. Wysocki (rjw)
Component:	SCSI	Assignee:	linux-scsi (linux-scsi)
Status:	CLOSED CODE_FIX
Severity:	normal	CC:	sitsofe, zdenek.kabelac
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	2.6.26-rc2	Subsystem:
Regression:	Yes	Bisected commit-id:
Bug Depends on:
Bug Blocks:	10492

Description Rafael J. Wysocki 2008-05-15 15:42:25 UTC

Subject    : BUG: unable to handle kernel paging request - scsi_bus_uevent
Submitter  : "Zdenek Kabelac" <zdenek.kabelac@gmail.com>
Date       : 2008-05-14 11:23
References : http://lkml.org/lkml/2008/5/14/111

This entry is being used for tracking a regression from 2.6.25.  Please don't
close it until the problem is fixed in the mainline.

Comment 1 Zdenek Kabelac 2008-05-20 05:19:57 UTC

Bug happend at least twice on my desktop with 2.6.26-rc3 - thought now I do not have such a backtrace as this deadlock now kills the machine -  I just know that RIP was from scsi_bus_uevent. If it would be worth to add - I could try to add a camera snapshost - but the trace is far from complete (just some bottom lines :( ).

The error happens once per couple reboots - but it is unpredicteble and hard to check. But it is definitely present in -rc3

Comment 2 Sitsofe Wheeler 2008-05-20 14:19:48 UTC

Yes I still see it too despite the patches mentioned in http://groups.google.com/group/fa.linux.kernel/browse_frm/thread/b64dbe94dea882cb/a8f1d83328fcbdc3?#a8f1d83328fcbdc3 . I bisected it down in http://groups.google.com/group/fa.linux.kernel/msg/c36e78f19fc4aaa3 and reverting the commit [dc16f5f2ede8cc2acf8ac22857a7fecf3a4296c2] PNP: make generic pnp_add_dma_resource() semmed to make the problem go away.

Comment 3 Sitsofe Wheeler 2008-05-23 12:45:22 UTC

This seems to have been resolved by James's patch in http://groups.google.com/group/fa.linux.kernel/browse_frm/thread/b64dbe94dea882cb/c60bafa089f89c6b?#c60bafa089f89c6b for me...

Comment 4 Zdenek Kabelac 2008-05-23 14:25:23 UTC

Is this the bugfix - or just the patch hides the real problem ?
I would have assumed that scsi_bus_uevent should not have been called at all for the non-scsi device

Comment 5 Sitsofe Wheeler 2008-05-24 03:26:34 UTC

Many subsystems (like the new PATA/SATA layer and USB block devices) also utilise the SCSI layer.

Comment 6 Sitsofe Wheeler 2008-05-31 12:24:11 UTC

This doesn't happen in the latest -next kernels for me.

Comment 7 Rafael J. Wysocki 2008-05-31 12:28:51 UTC

Does it happen with 2.6.26-rc4-git4?

Comment 8 Zdenek Kabelac 2008-06-02 06:15:33 UTC

I'll had to make more test boots to see if the bug is still present - but I do not have a good testcase where I could easily replicate the bug - it happens randomly during the boot - I think no patch solving this issue was committed to the mainline so far ?

Comment 9 Rafael J. Wysocki 2008-06-16 08:41:03 UTC

On Monday, 16 of June 2008, Zdenek Kabelac wrote:
> Hi
> 
> I've not see this bug over quite a few recent reboots - so I'd would
> have considered this bug also fixed for  me - if the problem reapers
> I'll reopen this bug.
> 
> (Well if there was no commit fixing this bug directly - it's hard to
> be sure the bug is fixed)
> 
> Zdenek

Comment 10 Sitsofe Wheeler 2008-06-16 14:49:31 UTC

There WAS a commit fixing this bug directly - it was carried in the SCSI fixes tree and was commit [1f42ea7bc0ddfadebd9e1c5362b41b53902dbcb1] and started turned up first in the next-20080526 tree and was committed to v2.6.26-rc5 (if gitk is right).