Re: [PATCH] mempool_resize and sanity checks

Balbir Singh (balbir_soni@yahoo.com)
Fri, 15 Mar 2002 16:38:34 -0800 (PST)


--0-136897776-1016239114=:84439
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

For some reason, it seems that the diffs did not
make it. Resending them

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

In mempool_resize(), I plan to replace the calls
to spin_unlock_irqrestore() -> pool_free -> spin_lock_
irqsave with the new sequence. The newer code
has fewer calls to spin_lock_irqsave and spin_unlock_
restore(). I think the new code should execute faster.

I have tested this code on a UP, if somebody could
test the patch on an SMP box and let me know if it
works, I would be thankful. I also have a sample test
case that uses mempool_resize().

I have also added the BUG_ON sanity checks in
mempool_create().

New Code

if (new_min_nr < pool->min_nr) {
pool->min_nr = new_min_nr;
/*
* Free possible excess elements.
*/
INIT_LIST_HEAD(&tmp_pool_to_free);
while (pool->curr_nr > pool->min_nr) {
tmp = pool->elements.next;
if (tmp == &pool->elements)
BUG();
list_del(tmp);
list_add(tmp, &tmp_pool_to_free);
pool->curr_nr--;
}
spin_unlock_irqrestore(&pool->lock, flags);

list_for_each(tmp, &tmp_pool_to_free) {
element = tmp;
pool->free(element, pool->pool_data);
}
return;
}

Original Code

if (new_min_nr < pool->min_nr) {
pool->min_nr = new_min_nr;
/*
* Free possible excess elements.
*/
while (pool->curr_nr > pool->min_nr) {
tmp = pool->elements.next;
if (tmp == &pool->elements)
BUG();
list_del(tmp);
element = tmp;
pool->curr_nr--;
spin_unlock_irqrestore(&pool->lock,
flags);

pool->free(element, pool->pool_data);

spin_lock_irqsave(&pool->lock, flags);
}
spin_unlock_irqrestore(&pool->lock, flags);
return;
}

Diffs are attached.

Balbir

__________________________________________________
Do You Yahoo!?
Yahoo! Sports - live college hoops coverage
http://sports.yahoo.com/
--0-136897776-1016239114=:84439
Content-Type: text/plain; name="mempool.c.diff.txt"
Content-Description: mempool.c.diff.txt
Content-Disposition: inline; filename="mempool.c.diff.txt"

--- mempool.c Tue Mar 12 19:21:33 2002
+++ mempool.c.new Tue Mar 5 10:37:19 2002
@@ -34,6 +34,9 @@
mempool_t *pool;
int i;

+ BUG_ON(!alloc_fn);
+ BUG_ON(!free_fn);
+
pool = kmalloc(sizeof(*pool), GFP_KERNEL);
if (!pool)
return NULL;
@@ -97,6 +100,7 @@
void *element;
unsigned long flags;
struct list_head *tmp;
+ struct list_head tmp_pool_to_free;

if (new_min_nr <= 0)
BUG();
@@ -107,20 +111,21 @@
/*
* Free possible excess elements.
*/
+ INIT_LIST_HEAD(&tmp_pool_to_free);
while (pool->curr_nr > pool->min_nr) {
tmp = pool->elements.next;
if (tmp == &pool->elements)
BUG();
list_del(tmp);
- element = tmp;
+ list_add(tmp, &tmp_pool_to_free);
pool->curr_nr--;
- spin_unlock_irqrestore(&pool->lock, flags);
+ }
+ spin_unlock_irqrestore(&pool->lock, flags);

+ list_for_each(tmp, &tmp_pool_to_free) {
+ element = tmp;
pool->free(element, pool->pool_data);
-
- spin_lock_irqsave(&pool->lock, flags);
}
- spin_unlock_irqrestore(&pool->lock, flags);
return;
}
delta = new_min_nr - pool->min_nr;

--0-136897776-1016239114=:84439--
-
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/