2005-11-12 02:20:55

by Bob Picco

Subject: [PATCH] cpuset - fix return without releasing semaphore

It seems wrong to acquire the semaphore and then return from
cpuset_zone_allowed without releasing it. This was only compile tested.

bob

Signed-off-by: Bob Picco <[email protected]>

kernel/cpuset.c | 5 +++--

1 files changed, 3 insertions(+), 2 deletions(-)

Index: linux-2.6.14-mm1/kernel/cpuset.c
===================================================================
--- linux-2.6.14-mm1.orig/kernel/cpuset.c 2005-11-07 07:00:13.000000000 -0500
+++ linux-2.6.14-mm1/kernel/cpuset.c 2005-11-09 17:56:28.000000000 -0500
@@ -2036,11 +2036,12 @@ int cpuset_zone_allowed(struct zone *z,
if (gfp_mask & __GFP_HARDWALL) /* If hardwall request, stop here */
return 0;

+ if (current->flags & PF_EXITING) /* Let dying task have memory */
+ return 1;
+
/* Not hardwall and node outside mems_allowed: scan up cpusets */
down(&callback_sem);

- if (current->flags & PF_EXITING) /* Let dying task have memory */
- return 1;
task_lock(current);
cs = nearest_exclusive_ancestor(current->cpuset);
task_unlock(current);

2005-11-12 18:17:24

by Paul Jackson

Subject: Re: [PATCH] cpuset - fix return without releasing semaphore

Bob wrote:
> It seems wrong to acquire the semaphore and then return from
> cpuset_zone_allowed without releasing it. This was only compile tested.

Acked-by: Paul Jackson <[email protected]>

Nice catch, Bob. Thanks.

--
I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <[email protected]> 1.925.600.0401

2005-11-12 19:26:53

by Paul Jackson

Subject: Re: [PATCH] cpuset - fix return without releasing semaphore

> This was only compile tested.

I boot, function and stress tested it on an ia64 SN2 system.
As expected, works fine, just as it did before.

--
I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <[email protected]> 1.925.600.0401