lkml.org 
[lkml]   [2005]   [Apr]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: PATCH [PPC64]: dead processes never reaped
From
Date
On Mon, 2005-04-18 at 14:38 -0500, Linas Vepstas wrote:
>
> Hi,
>
> The patch below appears to fix a problem where a number of dead processes
> linger on the system. On a highly loaded system, dozens of processes
> were found stuck in do_exit(), calling thier very last schedule(), and
> then being lost forever.

Ok, we spent some time with Paul decrypting what _switch_to is supposed
to do. Our understanding at this point is that the current code is
correct on both ppc32 and ppc64, that is:

The "prev" passed in is always "current" and we don't see how it can be
anything else. We use a local variable instead of current in the common
code because accessing current can be slow on some architectures. I
don't see any codepath where prev != current before switch_to.

If we didn't do some black magic that I explain below, _switch_to would
switch the entire context, including stack, and thus including the value
of "prev". Which means that we would always come back with prev beeing
current, which is useless for reaping the old task. What we want is that
this "prev" that was passed to _switch_to() is returned so that we can
rip that previous task despite the change of context, that is basically
prev has to be an invariant vs. the change of context in switch_to.

On ppc & ppc64, we implement that by passing that prev (or it's thread
counterpart) to the assembly context switch code in r3. This code will
preserve it and return it as-is (or re-transformed from thread to task).

So your problem must be somewhere else. I've looked at the need_resched
code path and we always reload prev = current from a non-preemptible
region, so it can't be wrong.

This was verified on 2.6.12-rc2, there might be something else wrong in
an older kernel.

Ben.


-
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: 2005-04-20 07:50    [W:0.064 / U:4.604 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site