OpenZFS 9738 - Fix third block copy allocations, broken at 9112.

Use METASLAB_WEIGHT_CLAIM weight to allocate tertiary blocks.
Previous use of METASLAB_WEIGHT_SECONDARY for that caused errors
later on metaslab_activate_allocator() call, leading to massive
load of unneeded metaslabs and write freezes.

Authored by: mav <mav@FreeBSD.org>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
OpenZFS-issue: https://illumos.org/issues/9738
FreeBSD-commit: https://github.com/freebsd/freebsd/commit/63e7138
Closes #7858

diff --git a/module/zfs/metaslab.c b/module/zfs/metaslab.c
index c1e32884f53a7d586a7b480bf99849c396bb1c44..12fe5890a0837362776857103b0c0c11afaafad8 100644 (file)
--- a/module/zfs/metaslab.c
+++ b/module/zfs/metaslab.c
@@ -3035,7 +3035,6 @@ metaslab_group_alloc_normal(metaslab_group_t *mg, zio_alloc_list_t *zal,
        metaslab_t *msp = NULL;
        uint64_t offset = -1ULL;
        uint64_t activation_weight;
-       boolean_t tertiary = B_FALSE;
 
        activation_weight = METASLAB_WEIGHT_PRIMARY;
        for (int i = 0; i < d; i++) {
@@ -3044,7 +3043,7 @@ metaslab_group_alloc_normal(metaslab_group_t *mg, zio_alloc_list_t *zal,
                        activation_weight = METASLAB_WEIGHT_SECONDARY;
                } else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
                    DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
-                       tertiary = B_TRUE;
+                       activation_weight = METASLAB_WEIGHT_CLAIM;
                        break;
                }
        }
@@ -3053,10 +3052,8 @@ metaslab_group_alloc_normal(metaslab_group_t *mg, zio_alloc_list_t *zal,
         * If we don't have enough metaslabs active to fill the entire array, we
         * just use the 0th slot.
         */
-       if (mg->mg_ms_ready < mg->mg_allocators * 2) {
-               tertiary = B_FALSE;
+       if (mg->mg_ms_ready < mg->mg_allocators * 3)
                allocator = 0;
-       }
 
        ASSERT3U(mg->mg_vd->vdev_ms_count, >=, 2);
 
@@ -3082,7 +3079,7 @@ metaslab_group_alloc_normal(metaslab_group_t *mg, zio_alloc_list_t *zal,
                        msp = mg->mg_primaries[allocator];
                        was_active = B_TRUE;
                } else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
-                   mg->mg_secondaries[allocator] != NULL && !tertiary) {
+                   mg->mg_secondaries[allocator] != NULL) {
                        msp = mg->mg_secondaries[allocator];
                        was_active = B_TRUE;
                } else {
@@ -3125,7 +3122,8 @@ metaslab_group_alloc_normal(metaslab_group_t *mg, zio_alloc_list_t *zal,
                        continue;
                }
 
-               if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
+               if (msp->ms_weight & METASLAB_WEIGHT_CLAIM &&
+                   activation_weight != METASLAB_WEIGHT_CLAIM) {
                        metaslab_passivate(msp, msp->ms_weight &
                            ~METASLAB_WEIGHT_CLAIM);
                        mutex_exit(&msp->ms_lock);