Remove fastwrite mutex

The fast write mutex is intended to protect accounting, but it is
redundant because all accounting is performed through atomic operations.
It also serializes all metaslab IO behind a mutex, which introduces a
theoretical scaling regression that the Illumos developers did not like
when we showed this to them. Removing it makes the selection of the
metaslab_group lock free as it is on Illumos. The selection is not quite
the same without the lock because the loop races with IO completions,
but any imbalances caused by this are likely to be corrected by
subsequent metaslab group selections.

Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #3643

diff --git a/include/sys/metaslab_impl.h b/include/sys/metaslab_impl.h
index 88bda071fa739e424aad8518165b3712c686e332..27a53b515fbc48ab5b200e88259df91cb6effe19 100644 (file)
--- a/include/sys/metaslab_impl.h
+++ b/include/sys/metaslab_impl.h
@@ -69,7 +69,6 @@ struct metaslab_class {
        uint64_t                mc_space;       /* total space (alloc + free) */
        uint64_t                mc_dspace;      /* total deflated space */
        uint64_t                mc_histogram[RANGE_TREE_HISTOGRAM_SIZE];
-       kmutex_t                mc_fastwrite_lock;
 };
 
 /*

diff --git a/module/zfs/metaslab.c b/module/zfs/metaslab.c
index 8946f39add3537377723e2d5b0c63c3aa9f367ee..19323cdb8ce5f654038d47cf498d9d9eb4bf0569 100644 (file)
--- a/module/zfs/metaslab.c
+++ b/module/zfs/metaslab.c
@@ -198,7 +198,6 @@ metaslab_class_create(spa_t *spa, metaslab_ops_t *ops)
        mc->mc_spa = spa;
        mc->mc_rotor = NULL;
        mc->mc_ops = ops;
-       mutex_init(&mc->mc_fastwrite_lock, NULL, MUTEX_DEFAULT, NULL);
 
        return (mc);
 }
@@ -212,7 +211,6 @@ metaslab_class_destroy(metaslab_class_t *mc)
        ASSERT(mc->mc_space == 0);
        ASSERT(mc->mc_dspace == 0);
 
-       mutex_destroy(&mc->mc_fastwrite_lock);
        kmem_free(mc, sizeof (metaslab_class_t));
 }
 
@@ -2214,9 +2212,6 @@ metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
        if (psize >= metaslab_gang_bang && (ddi_get_lbolt() & 3) == 0)
                return (SET_ERROR(ENOSPC));
 
-       if (flags & METASLAB_FASTWRITE)
-               mutex_enter(&mc->mc_fastwrite_lock);
-
        /*
         * Start at the rotor and loop through all mgs until we find something.
         * Note that there's no locking on mc_rotor or mc_aliquot because
@@ -2401,7 +2396,6 @@ top:
                        if (flags & METASLAB_FASTWRITE) {
                                atomic_add_64(&vd->vdev_pending_fastwrite,
                                    psize);
-                               mutex_exit(&mc->mc_fastwrite_lock);
                        }
 
                        return (0);
@@ -2425,9 +2419,6 @@ next:
 
        bzero(&dva[d], sizeof (dva_t));
 
-       if (flags & METASLAB_FASTWRITE)
-               mutex_exit(&mc->mc_fastwrite_lock);
-
        return (SET_ERROR(ENOSPC));
 }