RDS: IB: fix the rds_ib_fmr_wq kick call
authorSantosh Shilimkar <santosh.shilimkar@oracle.com>
Mon, 14 Sep 2015 05:34:37 +0000 (22:34 -0700)
committerSantosh Shilimkar <santosh.shilimkar@oracle.com>
Mon, 5 Oct 2015 18:19:01 +0000 (11:19 -0700)
RDS IB mr pool has its own workqueue 'rds_ib_fmr_wq', so we need
to use queue_delayed_work() to kick the work. This was hurting
the performance since pool maintenance was less often triggered
from other path.

Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
net/rds/ib_rdma.c

index 872f523630f29e2ecbfd83b7fd17ea3134346071..b6644fa2ea47214324a3aeb4b6ca8bd9489b9070 100644 (file)
@@ -319,7 +319,7 @@ static struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev)
        int err = 0, iter = 0;
 
        if (atomic_read(&pool->dirty_count) >= pool->max_items / 10)
-               schedule_delayed_work(&pool->flush_worker, 10);
+               queue_delayed_work(rds_ib_fmr_wq, &pool->flush_worker, 10);
 
        while (1) {
                ibmr = rds_ib_reuse_fmr(pool);