Re: [uml-devel] Re: Diagnosed and repeatable kernel mode panic in schedule() for 2.6!

[Jeff Dike <jdike@addtoit.com>]

blaisorblade_spam@yahoo.it said:
> By the way, would you check if this is needed in the 2.4 kernel, too?
> I've  checked the 2.4.24-1 uml patch and it seems to miss that line
> (and your  reason seem to apply to 2.4 too). 

Yup, I think so too.

> What is more strange is that the panic  happens deterministically, and
> in the same situation (same root fs and same  kernel) I get it at the
> same addresses! Since what happens seem to change  with each single
> kernel, but is easily repeatable, I would say that it is not  a race
> condition, but something with a crazy pointer outside of the core
> code.

I'm seeing two things, which don't happen deterministically :
	an exited process scheduling to itself, causing the exit.c BUG - I
don't see what prevents a TASK_ZOMBIE process from being schedulable offhand.
It doesn't seem to be dequeued from any runqueues as a result of calling exit().

	a segfault in schedule() caused by there being not bits set in the
queue bitmap (except for bit 140 which I guess is there to prevent ffs from
running into nowhereland).  And bit 140 doesn't refer to a valid queue, so
the task pulled of it is bogus.

> So it seems that both active.queue[140] and  active.queue.nr_active
> are corrupted.

Repeating the above, it looks like #140 isn't supposed to be valid.

> However, even in expired.queue[125] there is a problem, since in the
> list  there is one list_head (the one at 0xa08e40a0) whose next and
> prev pointers  point to itself (i.e. that is an empty list); however
> this "empty list" can  be reached by expired.queue[125]. 

Yup, interesting.  What I would do is stick some code in that looks for
this condition (and if you can look for it being created, that's even better).

I was looking for something that seemed like an invariant being violated,
and came up empty.  This looks like a good one, though.  If we can figure
out what's causing it, that gets us one step closer to the bug.

				Jeff



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel