lkml.org 
[lkml]   [2005]   [Dec]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Recursion bug in -rt
Dinakar,
after further testing and investigation I believe you original
assessment was correct.
The problem you are seeing is not a library problem.
The changes to down_futex need to be reverted. There is a new patch at

http://source.mvista.com/~dsingleton/patch-2.6.15-rc5-rt2-rf2.

that reverts the changes to down_futex.

Thanks for testing this.

David




Dinakar Guniguntala wrote:

>Hi David,
>
>I hit this bug with -rt22-rf11
>
>==========================================
>[ BUG: lock recursion deadlock detected! |
>------------------------------------------
>already locked: [f7abbc94] {futex}
>.. held by: testpi-3: 4595 [f7becdd0, 59]
>... acquired at: futex_wait_robust+0x142/0x1f3
>------------------------------
>| showing all locks held by: | (testpi-3/4595 [f7becdd0, 59]):
>------------------------------
>
>#001: [f7abbc94] {futex}
>... acquired at: futex_wait_robust+0x142/0x1f3
>
>-{current task's backtrace}----------------->
> [<c0103e04>] dump_stack+0x1e/0x20 (20)
> [<c0136bc2>] check_deadlock+0x2d7/0x334 (44)
> [<c01379bc>] task_blocks_on_lock+0x2c/0x224 (36)
> [<c03f29c5>] __down_interruptible+0x37c/0x95d (160)
> [<c013aebf>] down_futex+0xa3/0xe7 (40)
> [<c013ebc5>] futex_wait_robust+0x142/0x1f3 (72)
> [<c013f35c>] do_futex+0x9a/0x109 (40)
> [<c013f4dd>] sys_futex+0x112/0x11e (68)
> [<c0102f03>] sysenter_past_esp+0x54/0x75 (-8116)
>------------------------------
>| showing all locks held by: | (testpi-3/4595 [f7becdd0, 59]):
>------------------------------
>
>#001: [f7abbc94] {futex}
>... acquired at: futex_wait_robust+0x142/0x1f3
>
>---------------------------------------------------------------------
>
>futex.c -> futex_wait_robust
>
> if ((curval & FUTEX_PID) == current->pid) {
> ret = -EAGAIN;
> goto out_unlock;
> }
>
>rt.c -> down_futex
>
> if (!owner_task || owner_task == current) {
> up(sem);
> up_read(&current->mm->mmap_sem);
> return -EAGAIN;
> }
>
>I noticed that both the above checks below have been removed in your
>patch. I do understand that the futex_wait_robust path has been
>made similar to the futex_wait path, but I think we are not taking
>PI into consideration. Basically it looks like we still need to check
>if the current task has become owner. or are we missing a lock somewhere ?
>
>I added the down_futex check above and my test has been
>running for hours without the oops. Without this check it
>used to oops within minutes.
>
>Patch that works for me attached below. Thoughts?
>
> -Dinakar
>
>
>
>
>------------------------------------------------------------------------
>
>Index: linux-2.6.14-rt22-rayrt5/kernel/rt.c
>===================================================================
>--- linux-2.6.14-rt22-rayrt5.orig/kernel/rt.c 2005-12-15 02:15:13.000000000 +0530
>+++ linux-2.6.14-rt22-rayrt5/kernel/rt.c 2005-12-15 02:18:29.000000000 +0530
>@@ -3001,7 +3001,7 @@
> * if the owner can't be found return try again.
> */
>
>- if (!owner_task) {
>+ if (!owner_task || owner_task == current) {
> up(sem);
> up_read(&current->mm->mmap_sem);
> return -EAGAIN;
>
>

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-12-15 20:03    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site