lkml.org 
[lkml]   [2013]   [Apr]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH -next] ipc: make refcounter atomic (was Re: linux-next: Tree for Apr 23 [ Call-Traces: lib/debugobjects.c | kernel/rcupdate.c | kernel/rcutree.c ])
From
On Thu, Apr 25, 2013 at 1:38 AM, Davidlohr Bueso <davidlohr.bueso@hp.com> wrote:
> On Thu, 2013-04-25 at 01:05 +0200, Sedat Dilek wrote:
>> On Thu, Apr 25, 2013 at 12:16 AM, Davidlohr Bueso
>> <davidlohr.bueso@hp.com> wrote:
>> > From: Davidlohr Bueso <davidlohr.bueso@hp.com>
>> >
>> > Sedat reported an issue leading to a NULL dereference in update_queue():
>> >
>> > [ 178.490583] BUG: spinlock bad magic on CPU#1, sh/8066
>> > [ 178.490595] lock: 0xffff88008b53ea18, .magic: 6b6b6b6b, .owner: make/8068, .owner_cpu: 3
>> > [ 178.490599] BUG: unable to handle kernel NULL pointer dereference at (null)
>> > [ 178.490608] IP: [<ffffffff812bacd0>] update_queue+0x70/0x210
>> > [ 178.490610] PGD 0
>> > [ 178.490612] Oops: 0000 [#1] SMP
>> > ...
>> > [ 178.490704] Call Trace:
>> > [ 178.490710] [<ffffffff812baf51>] do_smart_update+0xe1/0x140
>> > [ 178.490713] [<ffffffff812bd6e1>] exit_sem+0x2b1/0x350
>> > [ 178.490718] [<ffffffff8105de80>] do_exit+0x290/0xa70
>> > [ 178.490721] [<ffffffff8105e6f4>] do_group_exit+0x44/0xa0
>> > [ 178.490724] [<ffffffff8105e767>] SyS_exit_group+0x17/0x20
>> > [ 178.490728] [<ffffffff816ce15d>] system_call_fastpath+0x1a/0x1f
>> >
>> > Linus pin-pointed the problem to a race in the reference counter. To quote:
>> >
>> > "That dmesg spew very much implies that the same RCU head got added twice to the RCU
>> > freeing list, and the only way that happens is if the refcount goes to
>> > zero twice. Which implies that either we increment a zero, or we lack
>> > locking and the coherency of the non-atomic access goes away."
>> >
>> > This patch converts the IPC RCU header's reference counter to atomic_t. The return of
>> > ipc_rcu_getref() is modified to inform the callers if it actually succeeded.
>> >
>> > Now all callers return -EIDRM upon failure and abort the current operation. Two exceptions are
>> > in semaphore code where sem_getref_and_unlock() and sem_getref() trigger a warning but proceed
>> > to freeing up any held locks.
>> >
>> > Signed-off-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
>> > Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
>> > CC: Rik van Riel <riel@surriel.com>
>> > CC: Paul McKenney <paul.mckenney@linaro.org>
>> > CC: Sedat Dilek <sedat.dilek@gmail.com>
>> > CC: Emmanuel Benisty <benisty.e@gmail.com>
>> > CC: Andrew Morton <akpm@linux-foundation.org>
>>
>> Missing my Reported-by ...!
>
> Not trying to take away credit or efforts from you, just wanted you to
> reconfirm that *this* actual patch fixes things for you :)
>

No, I am not of those "bad guys" in OSS.
I would not have invested so much time in helping to get this fixed.

Did my usual test-case:
A kernel-rebuild within same kernel-environment with no breakage or
abnormalities in the logs.

Hope I could help.

- Sedat -

> Thanks,
> Davidlohr
>


\
 
 \ /
  Last update: 2013-04-25 02:01    [W:0.050 / U:0.100 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site