2005-09-27 22:20:20

by Nate Diller

Subject: [PATCH 1/3] block cleanups: Fix iosched module refcount leak

If the requested I/O scheduler is already in place, elevator_switch simply leaves the queue
alone, and returns. However, it forgets to call elevator_put, so

'echo [current_sched] > /sys/block/[dev]/queue/scheduler'

will leak a reference, causing the current_sched module to be permanently pinned in memory.

This patchset is against 2.6.14-rc2-mm1, but should apply to anything recent.

NATE

Signed-off-by: Nate Diller <[email protected]>

--- a/drivers/block/elevator.c 2005-09-26 19:17:59.000000000 -0700
+++ b/drivers/block/elevator.c 2005-09-27 11:13:15.000000000 -0700
@@ -692,8 +692,10 @@ ssize_t elv_iosched_store(request_queue_
return -EINVAL;
}

- if (!strcmp(elevator_name, q->elevator->elevator_type->elevator_name))
+ if (!strcmp(elevator_name, q->elevator->elevator_type->elevator_name)) {
+ elevator_put(e);
return count;
+ }

elevator_switch(q, e);
return count;

2005-09-27 22:40:13

by Jens Axboe

Subject: Re: [PATCH 1/3] block cleanups: Fix iosched module refcount leak

On Tue, Sep 27 2005, Nate Diller wrote:
> If the requested I/O scheduler is already in place, elevator_switch simply
> leaves the queue alone, and returns. However, it forgets to call
> elevator_put, so
>
> 'echo [current_sched] > /sys/block/[dev]/queue/scheduler'
>
> will leak a reference, causing the current_sched module to be permanently
> pinned in memory.
>
> This patchset is against 2.6.14-rc2-mm1, but should apply to anything
> recent.

Thanks, looks good.

Signed-off-by: Jens Axboe <[email protected]>

--
Jens Axboe