[PATCH] cfq-iosched: don't allow sync merges across queues
authorJens Axboe <jens.axboe@oracle.com>
Wed, 20 Dec 2006 10:04:12 +0000 (11:04 +0100)
committerJens Axboe <jens.axboe@oracle.com>
Wed, 20 Dec 2006 10:04:12 +0000 (11:04 +0100)
Currently we allow any merge, even if the io originates from different
processes. This can cause really bad starvation and unfairness, if those
ios happen to be synchronous (reads or direct writes).

So add a allow_merge hook to the io scheduler ops, so an io scheduler can
help decide whether a bio/process combination may be merged with an
existing request.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
block/cfq-iosched.c
block/elevator.c
include/linux/elevator.h

index 533a2938ffd6740f6f41f2c5bd52177e8dde8a0e..9fc5eafa6c0e6283398aafc4bc6a5c584438f10b 100644 (file)
@@ -568,6 +568,38 @@ cfq_merged_requests(request_queue_t *q, struct request *rq,
        cfq_remove_request(next);
 }
 
+static int cfq_allow_merge(request_queue_t *q, struct request *rq,
+                          struct bio *bio)
+{
+       struct cfq_data *cfqd = q->elevator->elevator_data;
+       const int rw = bio_data_dir(bio);
+       struct cfq_queue *cfqq;
+       pid_t key;
+
+       /*
+        * If bio is async or a write, always allow merge
+        */
+       if (!bio_sync(bio) || rw == WRITE)
+               return 1;
+
+       /*
+        * bio is sync. if request is not, disallow.
+        */
+       if (!rq_is_sync(rq))
+               return 0;
+
+       /*
+        * Ok, both bio and request are sync. Allow merge if they are
+        * from the same queue.
+        */
+       key = cfq_queue_pid(current, rw, 1);
+       cfqq = cfq_find_cfq_hash(cfqd, key, current->ioprio);
+       if (cfqq != RQ_CFQQ(rq))
+               return 0;
+
+       return 1;
+}
+
 static inline void
 __cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 {
@@ -2125,6 +2157,7 @@ static struct elevator_type iosched_cfq = {
                .elevator_merge_fn =            cfq_merge,
                .elevator_merged_fn =           cfq_merged_request,
                .elevator_merge_req_fn =        cfq_merged_requests,
+               .elevator_allow_merge_fn =      cfq_allow_merge,
                .elevator_dispatch_fn =         cfq_dispatch_requests,
                .elevator_add_req_fn =          cfq_insert_request,
                .elevator_activate_req_fn =     cfq_activate_request,
index c0063f345c5d620320cd82438394eefe71b442dc..62c7a3069d3a20981067e1865c77be04a3990847 100644 (file)
@@ -50,6 +50,21 @@ static const int elv_hash_shift = 6;
 #define rq_hash_key(rq)                ((rq)->sector + (rq)->nr_sectors)
 #define ELV_ON_HASH(rq)                (!hlist_unhashed(&(rq)->hash))
 
+/*
+ * Query io scheduler to see if the current process issuing bio may be
+ * merged with rq.
+ */
+static int elv_iosched_allow_merge(struct request *rq, struct bio *bio)
+{
+       request_queue_t *q = rq->q;
+       elevator_t *e = q->elevator;
+
+       if (e->ops->elevator_allow_merge_fn)
+               return e->ops->elevator_allow_merge_fn(q, rq, bio);
+
+       return 1;
+}
+
 /*
  * can we safely merge with this request?
  */
@@ -65,12 +80,15 @@ inline int elv_rq_merge_ok(struct request *rq, struct bio *bio)
                return 0;
 
        /*
-        * same device and no special stuff set, merge is ok
+        * must be same device and not a special request
         */
-       if (rq->rq_disk == bio->bi_bdev->bd_disk && !rq->special)
-               return 1;
+       if (rq->rq_disk != bio->bi_bdev->bd_disk || !rq->special)
+               return 0;
 
-       return 0;
+       if (!elv_iosched_allow_merge(rq, bio))
+               return 0;
+
+       return 1;
 }
 EXPORT_SYMBOL(elv_rq_merge_ok);
 
index a24931d244046d72e056fab76f670719256de00a..e88fcbc77f8f381d3c1b801955406f260b0cab4f 100644 (file)
@@ -12,6 +12,8 @@ typedef void (elevator_merge_req_fn) (request_queue_t *, struct request *, struc
 
 typedef void (elevator_merged_fn) (request_queue_t *, struct request *, int);
 
+typedef int (elevator_allow_merge_fn) (request_queue_t *, struct request *, struct bio *);
+
 typedef int (elevator_dispatch_fn) (request_queue_t *, int);
 
 typedef void (elevator_add_req_fn) (request_queue_t *, struct request *);
@@ -33,6 +35,7 @@ struct elevator_ops
        elevator_merge_fn *elevator_merge_fn;
        elevator_merged_fn *elevator_merged_fn;
        elevator_merge_req_fn *elevator_merge_req_fn;
+       elevator_allow_merge_fn *elevator_allow_merge_fn;
 
        elevator_dispatch_fn *elevator_dispatch_fn;
        elevator_add_req_fn *elevator_add_req_fn;