On Mon, Sep 19, 2016 at 10:48:49AM -0700, Omar Sandoval wrote:
> On Sun, Sep 18, 2016 at 09:37:14AM +0200, Alexander Gordeev wrote:
> > Currently maximum number of used hardware queues is limited to
> > number of CPUs in the system. However, using 'nr_cpu_ids' as
> > the limit for (de-)allocations of data structures instead of
> > existing data structures' counters (a) worsens readability and
> > (b) leads to unused memory when number of hardware queues is
> > less than number of CPUs.
> >
> > CC: linux-bl...@vger.kernel.org
> > Signed-off-by: Alexander Gordeev <agord...@redhat.com>
> > ---
> > block/blk-mq.c | 8 ++++----
> > 1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/block/blk-mq.c b/block/blk-mq.c
> > index 276ec7b..2c77b68 100644
> > --- a/block/blk-mq.c
> > +++ b/block/blk-mq.c
> > @@ -2054,8 +2054,8 @@ struct request_queue
> > *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
> > if (!q->queue_ctx)
> > goto err_exit;
> >
> > - q->queue_hw_ctx = kzalloc_node(nr_cpu_ids * sizeof(*(q->queue_hw_ctx)),
> > - GFP_KERNEL, set->numa_node);
> > + q->queue_hw_ctx = kzalloc_node(set->nr_hw_queues *
> > + sizeof(*(q->queue_hw_ctx)), GFP_KERNEL, set->numa_node);
> > if (!q->queue_hw_ctx)
> > goto err_percpu;
> >
> > @@ -2319,7 +2319,7 @@ int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
> > if (set->nr_hw_queues > nr_cpu_ids)
> > set->nr_hw_queues = nr_cpu_ids;
> >
> > - set->tags = kzalloc_node(nr_cpu_ids * sizeof(struct blk_mq_tags *),
> > + set->tags = kzalloc_node(set->nr_hw_queues * sizeof(*set->tags),
> > GFP_KERNEL, set->numa_node);
> > if (!set->tags)
> > return -ENOMEM;
> > @@ -2360,7 +2360,7 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
> > {
> > int i;
> >
> > - for (i = 0; i < nr_cpu_ids; i++) {
> > + for (i = 0; i < set->nr_hw_queues; i++) {
> > if (set->tags[i])
> > blk_mq_free_rq_map(set, set->tags[i], i);
> > }
>
> I don't think this is safe since we might increase the number of
> hardware queues (blk_mq_update_nr_hw_queues()).
It is safe, because blk_mq_update_nr_hw_queues() limits number of
hardware queues to nr_cpu_ids. But still using nr_cpu_ids is wrong,
because (a) set->nr_hw_queues could be less than nr_cpu_ids and
(b) it is set->nr_hw_queues counter that tracks size of the array.
> --
> Omar
