lkml.org 
[lkml]   [2020]   [Dec]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [RFC PATCH v2 2/2] blk-mq: Lockout tagset iter when freeing rqs
From
Date
On 12/17/20 3:07 AM, John Garry wrote:
> diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
> index a6df2d5df88a..853ed5b889aa 100644
> --- a/block/blk-mq-tag.c
> +++ b/block/blk-mq-tag.c
> @@ -358,10 +358,19 @@ void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
> {
> int i;
>
> for (i = 0; i < tagset->nr_hw_queues; i++) {
> - if (tagset->tags && tagset->tags[i])
> - __blk_mq_all_tag_iter(tagset->tags[i], fn, priv,
> + if (tagset->tags && tagset->tags[i]) {
> + struct blk_mq_tags *tags = tagset->tags[i];
> +
> + if (!atomic_inc_not_zero(&tags->iter_usage_counter))
> + continue;
> +
> + __blk_mq_all_tag_iter(tags, fn, priv,
> BT_TAG_ITER_STARTED);
> +
> + atomic_dec(&tags->iter_usage_counter);
> + }
> }
> }

Atomic operations are (a) more expensive than rcu_read_lock() /
rcu_read_lock() and (b) do not provide the same guarantees.
rcu_read_lock() has acquire semantics and rcu_read_unlock() has
release semantics. Regular atomic operations do not have these
semantics which means that the CPU is allowed to reorder certain
regular loads and stores against atomic operations. Additionally,
atomic operations are more expensive than the corresponding RCU
primitives. In other words, I would be much happier if this patch
series would use RCU instead of atomics.

Thanks,

Bart.

\
 
 \ /
  Last update: 2020-12-18 02:57    [W:0.235 / U:0.048 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site