diff mbox

[v2,11/13] dm-mpath: Micro-optimize the hot path

Message ID 20170427171126.26814-12-bart.vanassche@sandisk.com (mailing list archive)
State Accepted, archived
Delegated to: Mike Snitzer
Headers show

Commit Message

Bart Van Assche April 27, 2017, 5:11 p.m. UTC
Instead of checking MPATHF_QUEUE_IF_NO_PATH,
MPATHF_SAVED_QUEUE_IF_NO_PATH and the no_flush flag to decide whether
or not to push back a request if there are no paths available, only
clear MPATHF_QUEUE_IF_NO_PATH in queue_if_no_path() if no_flush has
not been set. The result is that only a single bit has to be tested
in the hot path to decide whether or not a request must be pushed
back and also that m->lock does not have to be taken in the hot path.

Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
---
 drivers/md/dm-mpath.c | 70 ++++++++-------------------------------------------
 1 file changed, 11 insertions(+), 59 deletions(-)

Comments

Mike Snitzer April 27, 2017, 8:39 p.m. UTC | #1
On Thu, Apr 27 2017 at  1:11P -0400,
Bart Van Assche <bart.vanassche@sandisk.com> wrote:

> Instead of checking MPATHF_QUEUE_IF_NO_PATH,
> MPATHF_SAVED_QUEUE_IF_NO_PATH and the no_flush flag to decide whether
> or not to push back a request if there are no paths available, only
> clear MPATHF_QUEUE_IF_NO_PATH in queue_if_no_path() if no_flush has
> not been set. The result is that only a single bit has to be tested
> in the hot path to decide whether or not a request must be pushed
> back and also that m->lock does not have to be taken in the hot path.
> 
> Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
> Reviewed-by: Hannes Reinecke <hare@suse.com>

This patch was very demanding to review.. all that old must_push_back()
code was very tedious.  Happy to see it go.  BUT I'm left nervous that
something might regress.  Not because of something overlooked (though
that could happen).. just that its a particularly delicate portion of
the mpath target's suspend/resume handling.

Kudos to you for tackling this.  Staged in dm-4.12 for 4.12 inclusion
(will push it out once I go back over what I've staged)

--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
diff mbox

Patch

diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 286d74593b2c..5a3200332eb1 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -442,47 +442,6 @@  static struct pgpath *choose_pgpath(struct multipath *m, size_t nr_bytes)
 }
 
 /*
- * Check whether bios must be queued in the device-mapper core rather
- * than here in the target.
- *
- * If m->queue_if_no_path and m->saved_queue_if_no_path hold the
- * same value then we are not between multipath_presuspend()
- * and multipath_resume() calls and we have no need to check
- * for the DMF_NOFLUSH_SUSPENDING flag.
- */
-static bool __must_push_back(struct multipath *m)
-{
-	return ((test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) !=
-		 test_bit(MPATHF_SAVED_QUEUE_IF_NO_PATH, &m->flags)) &&
-		dm_noflush_suspending(m->ti));
-}
-
-static bool must_push_back_rq(struct multipath *m)
-{
-	bool r;
-	unsigned long flags;
-
-	spin_lock_irqsave(&m->lock, flags);
-	r = (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) ||
-	     __must_push_back(m));
-	spin_unlock_irqrestore(&m->lock, flags);
-
-	return r;
-}
-
-static bool must_push_back_bio(struct multipath *m)
-{
-	bool r;
-	unsigned long flags;
-
-	spin_lock_irqsave(&m->lock, flags);
-	r = __must_push_back(m);
-	spin_unlock_irqrestore(&m->lock, flags);
-
-	return r;
-}
-
-/*
  * Map cloned requests (request-based multipath)
  */
 static int multipath_clone_and_map(struct dm_target *ti, struct request *rq,
@@ -503,7 +462,7 @@  static int multipath_clone_and_map(struct dm_target *ti, struct request *rq,
 		pgpath = choose_pgpath(m, nr_bytes);
 
 	if (!pgpath) {
-		if (must_push_back_rq(m)) {
+		if (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) {
 			pr_debug("no path - requeueing\n");
 			return DM_MAPIO_DELAY_REQUEUE;
 		}
@@ -581,9 +540,9 @@  static int __multipath_map_bio(struct multipath *m, struct bio *bio, struct dm_m
 	}
 
 	if (!pgpath) {
-		if (!must_push_back_bio(m))
-			return -EIO;
-		return DM_MAPIO_REQUEUE;
+		if (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags))
+			return DM_MAPIO_REQUEUE;
+		return -EIO;
 	}
 
 	mpio->pgpath = pgpath;
@@ -675,7 +634,7 @@  static int queue_if_no_path(struct multipath *m, bool queue_if_no_path,
 		else
 			clear_bit(MPATHF_SAVED_QUEUE_IF_NO_PATH, &m->flags);
 	}
-	if (queue_if_no_path)
+	if (queue_if_no_path || dm_noflush_suspending(m->ti))
 		set_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags);
 	else
 		clear_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags);
@@ -1526,12 +1485,9 @@  static int do_end_io(struct multipath *m, struct request *clone,
 	if (mpio->pgpath)
 		fail_path(mpio->pgpath);
 
-	if (!atomic_read(&m->nr_valid_paths)) {
-		if (!test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) {
-			if (!must_push_back_rq(m))
-				r = -EIO;
-		}
-	}
+	if (atomic_read(&m->nr_valid_paths) == 0 &&
+	    !test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags))
+		r = -EIO;
 
 	return r;
 }
@@ -1572,13 +1528,9 @@  static int do_end_io_bio(struct multipath *m, struct bio *clone,
 	if (mpio->pgpath)
 		fail_path(mpio->pgpath);
 
-	if (!atomic_read(&m->nr_valid_paths)) {
-		if (!test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) {
-			if (!must_push_back_bio(m))
-				return -EIO;
-			return DM_ENDIO_REQUEUE;
-		}
-	}
+	if (atomic_read(&m->nr_valid_paths) == 0 &&
+	    !test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags))
+		return -EIO;
 
 	/* Queue for the daemon to resubmit */
 	dm_bio_restore(get_bio_details_from_bio(clone), clone);