[PATCH v4 1/2] block: skip elevator data initialization for flushrequests

[Mike Snitzer]

Set REQ_SORTED, in the @rw_flags passed to the request allocator, for
all requests that may be put on IO scheduler.  REQ_SORTED is not set for
flush requests because they are never put on the IO scheduler.

Skip elevator data initialization during request allocation if
REQ_SORTED is not set in @rw_flags passed to get_request().

Signed-off-by: Mike Snitzer <snitzer@xxxxxxxxxx>
Acked-by: Vivek Goyal <vgoyal@xxxxxxxxxx>
---
 block/blk-core.c |   33 ++++++++++++++++++++++++++++-----
 1 file changed, 28 insertions(+), 5 deletions(-)

v4: fixed bug where REQ_SORTED wasn't cleared on entry to get_request
    - and Jens, yes I agree this is still a hack

Index: linux-2.6/block/blk-core.c
===================================================================
--- linux-2.6.orig/block/blk-core.c
+++ linux-2.6/block/blk-core.c
@@ -754,6 +754,9 @@ static void freed_request(struct request
 
 /*
  * Get a free request, queue_lock must be held.
+ * @rw_flags: may be overloaded to convey additional request features;
+ * any overloaded feature flags must be cleared immediately.
+ *
  * Returns NULL on failure, with queue_lock held.
  * Returns !NULL on success, with queue_lock *not held*.
  */
@@ -764,7 +767,11 @@ static struct request *get_request(struc
 	struct request_list *rl = &q->rq;
 	struct io_context *ioc = NULL;
 	const bool is_sync = rw_is_sync(rw_flags) != 0;
-	int may_queue, priv;
+	const bool init_elv_data = !!(rw_flags & REQ_SORTED);
+	int may_queue, priv = 0;
+
+	if (init_elv_data)
+		rw_flags &= ~REQ_SORTED;
 
 	may_queue = elv_may_queue(q, rw_flags);
 	if (may_queue == ELV_MQUEUE_NO)
@@ -808,9 +815,14 @@ static struct request *get_request(struc
 	rl->count[is_sync]++;
 	rl->starved[is_sync] = 0;
 
-	priv = !test_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
-	if (priv)
-		rl->elvpriv++;
+	/*
+	 * Only initialize elevator data if IO scheduler may be used.
+	 */
+	if (init_elv_data) {
+		priv = !test_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
+		if (priv)
+			rl->elvpriv++;
+	}
 
 	if (blk_queue_io_stat(q))
 		rw_flags |= REQ_IO_STAT;
@@ -1197,6 +1209,7 @@ static int __make_request(struct request
 	const unsigned short prio = bio_prio(bio);
 	const bool sync = !!(bio->bi_rw & REQ_SYNC);
 	const bool unplug = !!(bio->bi_rw & REQ_UNPLUG);
+	const bool flush = !!(bio->bi_rw & (REQ_FLUSH | REQ_FUA));
 	const unsigned long ff = bio->bi_rw & REQ_FAILFAST_MASK;
 	int where = ELEVATOR_INSERT_SORT;
 	int rw_flags;
@@ -1210,7 +1223,7 @@ static int __make_request(struct request
 
 	spin_lock_irq(q->queue_lock);
 
-	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) {
+	if (flush) {
 		where = ELEVATOR_INSERT_FLUSH;
 		goto get_rq;
 	}
@@ -1293,6 +1306,16 @@ get_rq:
 		rw_flags |= REQ_SYNC;
 
 	/*
+	 * Set REQ_SORTED for all requests that may be put on IO scheduler.
+	 * The request allocator's IO scheduler initialization will be skipped
+	 * if REQ_SORTED is not set -- elv_set_request() is avoided so that
+	 * that the allocated request's elevator_private pointers are not
+	 * initialized and that space can be used by flush request data.
+	 */
+	if (!flush)
+		rw_flags |= REQ_SORTED;
+
+	/*
 	 * Grab a free request. This is might sleep but can not fail.
 	 * Returns with the queue unlocked.
 	 */
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/