sbitmap: optimize wakeup check

Even if we have no waiters on any of the sbitmap_queue wait states, we
still have to loop every entry to check. We do this for every IO, so
the cost adds up.

Shift a bit of the cost to the slow path, when we actually have waiters.
Wrap prepare_to_wait_exclusive() and finish_wait(), so we can maintain
an internal count of how many are currently active. Then we can simply
check this count in sbq_wake_ptr() and not have to loop if we don't
have any sleepers.

Convert the two users of sbitmap with waiting, blk-mq-tag and iSCSI.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 28 insertions, 0 deletions

diff --git a/lib/sbitmap.c b/lib/sbitmap.c
index f99382e59314..a89fbe7cf6ca 100644
--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -394,6 +394,7 @@ int sbitmap_queue_init_node(struct sbitmap_queue *sbq, unsigned int depth,
         sbq->min_shallow_depth = UINT_MAX;
         sbq->wake_batch = sbq_calc_wake_batch(sbq, depth);
         atomic_set(&sbq->wake_index, 0);
+        atomic_set(&sbq->ws_active, 0);
 
         sbq->ws = kzalloc_node(SBQ_WAIT_QUEUES * sizeof(*sbq->ws), flags, node);
         if (!sbq->ws) {
@@ -509,6 +510,9 @@ static struct sbq_wait_state *sbq_wake_ptr(struct sbitmap_queue *sbq)
 {
         int i, wake_index;
 
+        if (!atomic_read(&sbq->ws_active))
+                return NULL;
+
         wake_index = atomic_read(&sbq->wake_index);
         for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
                 struct sbq_wait_state *ws = &sbq->ws[wake_index];
@@ -634,6 +638,7 @@ void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m)
 
         seq_printf(m, "wake_batch=%u\n", sbq->wake_batch);
         seq_printf(m, "wake_index=%d\n", atomic_read(&sbq->wake_index));
+        seq_printf(m, "ws_active=%d\n", atomic_read(&sbq->ws_active));
 
         seq_puts(m, "ws={\n");
         for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
@@ -649,3 +654,26 @@ void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m)
         seq_printf(m, "min_shallow_depth=%u\n", sbq->min_shallow_depth);
 }
 EXPORT_SYMBOL_GPL(sbitmap_queue_show);
+
+void sbitmap_prepare_to_wait(struct sbitmap_queue *sbq,
+                             struct sbq_wait_state *ws,
+                             struct sbq_wait *sbq_wait, int state)
+{
+        if (!sbq_wait->accounted) {
+                atomic_inc(&sbq->ws_active);
+                sbq_wait->accounted = 1;
+        }
+        prepare_to_wait_exclusive(&ws->wait, &sbq_wait->wait, state);
+}
+EXPORT_SYMBOL_GPL(sbitmap_prepare_to_wait);
+
+void sbitmap_finish_wait(struct sbitmap_queue *sbq, struct sbq_wait_state *ws,
+                         struct sbq_wait *sbq_wait)
+{
+        finish_wait(&ws->wait, &sbq_wait->wait);
+        if (sbq_wait->accounted) {
+                atomic_dec(&sbq->ws_active);
+                sbq_wait->accounted = 0;
+        }
+}
+EXPORT_SYMBOL_GPL(sbitmap_finish_wait);