Re: 2.6.9-rc2-mm4 ps hang ?

[Andrew Morton]

Badari Pulavarty <pbadari@xxxxxxxxxx> wrote:
>
> 
>  On Fri, 2004-10-01 at 16:49, Andrew Morton wrote:
>  > Badari Pulavarty <pbadari@xxxxxxxxxx> wrote:
>  > >
>  > > Here is the full sysrq-t output.
>  > 
>  > What's this guy up to?
>  ...
>  > 
>  > Something is seriously screwed up if it's stuck in try_to_wake_up().  Tried
>  > generating a few extra traces?
>  > 
>  > Then again, maybe we're missing an up_read() somewhere.  hrm, I'll check.
>  > 
> 
>  It reproduced again. I think this is the one causing all the troubles..
> 
>  db2fmcd       D ffffffff80132e2a     0 10854   7636                    
>  (NOTLB)
>  00000101bae85ef8 0000000000000002 0000020800000018 00000101d9ddd550
>         0000000100000084 000001016d490e20 000001016d491158
>  00000101d9ddd550
>         0000000000000206 ffffffff801353cb
>  Call Trace:<ffffffff801353cb>{try_to_wake_up+971}
>  <ffffffff804455f0>{__down_write+128}
>         <ffffffff80125e6f>{sys32_mmap+143}
>  <ffffffff80124af1>{ia32_sysret+0}
> 
>  when I tried looking at /proc/10854 - it hung.

OK, so maybe that CPU is spinning in try_to_wake_up().  Can you tell if one
CPU is busy when this happens?

Or you could try Peter's suggestion:

--- 25/kernel/sched.c~a	2004-10-01 17:43:32.500700488 -0700
+++ 25-akpm/kernel/sched.c	2004-10-01 17:43:34.754357880 -0700
@@ -1606,7 +1606,7 @@ out_set_cpu:
 		task_rq_unlock(rql, &flags);
 		/* might preempt at this point */
 		rql = task_rq_lock(p, &flags);
-		adjust_sched_timestamp(p, old_rq);
+//		adjust_sched_timestamp(p, old_rq);
 		old_state = p->state;
 		if (!(old_state & state))
 			goto out;
_

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