bug in 2.6.26-rc9: sleeping function called from invalid context at kernel/rwsem:21

bug in 2.6.26-rc9: sleeping function called from invalid context at kernel/rwsem:21

[Giacomo A. Catenazzi]

Hello,

yesterday at shutdown I had a crash. Unfortunately it seems
that there was one (or more) oops missing. The bug appeared
at unmounting partition at shutdown, so there are no logs
of the previous messages.

A photo of the screen is in:
http://cateee.net/files/oops/20080709010.jpg

The oopes was in the git tree of two days ago.  I've not
seen this crash on last kernel, but OTOH I don't know if
the bug was reproducible.

ciao
	cate


PS: unfortunately I had not time for further tests and
better report, but I hope the trace is enough for such
bugs.

Re: bug in 2.6.26-rc9: sleeping function called from invalid context at kernel/rwsem:21

[Vegard Nossum]

On Thu, Jul 10, 2008 at 10:19 AM, Giacomo A. Catenazzi <cate@cateee.net> wrote:
> Hello,
>
> yesterday at shutdown I had a crash. Unfortunately it seems
> that there was one (or more) oops missing. The bug appeared
> at unmounting partition at shutdown, so there are no logs
> of the previous messages.

Thanks for the report.

It would be really useful to know which filesystems you are/were
using, including virtual filesystems like sysfs. Maybe you are using
ecryptfs or FUSE?

(It would also be interesting to see if you could run
/etc/rc0.d/S40umountfs while the machine is up, to see if you can get
the segfault but keep the machine usable enough to get the whole log.)


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036

Re: bug in 2.6.26-rc9: sleeping function called from invalid context at kernel/rwsem:21

[Giacomo A. Catenazzi]

Vegard Nossum wrote:
> On Thu, Jul 10, 2008 at 10:19 AM, Giacomo A. Catenazzi <cate@cateee.net> wrote:
>> Hello,
>>
>> yesterday at shutdown I had a crash. Unfortunately it seems
>> that there was one (or more) oops missing. The bug appeared
>> at unmounting partition at shutdown, so there are no logs
>> of the previous messages.
> 
> Thanks for the report.
> 
> It would be really useful to know which filesystems you are/were
> using, including virtual filesystems like sysfs. Maybe you are using
> ecryptfs or FUSE?
> 
> (It would also be interesting to see if you could run
> /etc/rc0.d/S40umountfs while the machine is up, to see if you can get
> the segfault but keep the machine usable enough to get the whole log.)

At home, my disks are only ext2.
sysfs, procfs and other system fs and /dev/* mount points are not
umounted in /etc/rc0.d/S40umountfs.

By a quick look (I'm not at home), it seems that on my home computer
only ext2 fs are umounted by /etc/rc0.d/S40umountfs.

As I said, the bug doesn't seems reproducible,
but I'll try harder later.

At subsequent boot, the unclean fs were: / /var and /usr,
Checking was done without error messages, so I think no busy
or opened files at time of crash.

Note: at the boot, one filesystem was checked (after 30+x),
I think it was /home. But I really think it is unrelated
(but it means that some of the other fs have a boot count > 30)

ciao
	cate