Illumos #3552, #3564

3552 condensing one space map burns 3 seconds of CPU in spa_sync() thread
3564 spa_sync() spends 5-10% of its time in metaslab_sync() (when not condensing)
Reviewed by: Adam Leventhal <[email protected]>
Reviewed by: Dan Kimmel <[email protected]>
Reviewed by: Matthew Ahrens <[email protected]>
Approved by: Richard Lowe <[email protected]>

References:
  illumos/illumos-gate@16a4a8074274d2d7cc408589cf6359f4a378c861
  https://www.illumos.org/issues/3552
  https://www.illumos.org/issues/3564

Ported-by: Tim Chase <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #1513

1 files changed, 24 insertions, 9 deletions

diff --git a/module/zfs/space_map.c b/module/zfs/space_map.c
index ab7cb7a9a..a031f3a20 100644
--- a/module/zfs/space_map.c
+++ b/module/zfs/space_map.c
@@ -107,6 +107,7 @@ space_map_add(space_map_t *sm, uint64_t start, uint64_t size)
 	int merge_before, merge_after;
 
 	ASSERT(MUTEX_HELD(sm->sm_lock));
+	VERIFY(!sm->sm_condensing);
 	VERIFY(size != 0);
 	VERIFY3U(start, >=, sm->sm_start);
 	VERIFY3U(end, <=, sm->sm_start + sm->sm_size);
@@ -175,6 +176,7 @@ space_map_remove(space_map_t *sm, uint64_t start, uint64_t size)
 	int left_over, right_over;
 
 	ASSERT(MUTEX_HELD(sm->sm_lock));
+	VERIFY(!sm->sm_condensing);
 	VERIFY(size != 0);
 	VERIFY(P2PHASE(start, 1ULL << sm->sm_shift) == 0);
 	VERIFY(P2PHASE(size, 1ULL << sm->sm_shift) == 0);
@@ -244,6 +246,20 @@ space_map_contains(space_map_t *sm, uint64_t start, uint64_t size)
 }
 
 void
+space_map_swap(space_map_t **msrc, space_map_t **mdst)
+{
+	space_map_t *sm;
+
+	ASSERT(MUTEX_HELD((*msrc)->sm_lock));
+	ASSERT0((*mdst)->sm_space);
+	ASSERT0(avl_numnodes(&(*mdst)->sm_root));
+
+	sm = *msrc;
+	*msrc = *mdst;
+	*mdst = sm;
+}
+
+void
 space_map_vacate(space_map_t *sm, space_map_func_t *func, space_map_t *mdest)
 {
 	space_seg_t *ss;
@@ -424,9 +440,9 @@ space_map_sync(space_map_t *sm, uint8_t maptype,
 	space_map_obj_t *smo, objset_t *os, dmu_tx_t *tx)
 {
 	spa_t *spa = dmu_objset_spa(os);
-	void *cookie = NULL;
+	avl_tree_t *t = &sm->sm_root;
 	space_seg_t *ss;
-	uint64_t bufsize, start, size, run_len, delta, sm_space;
+	uint64_t bufsize, start, size, run_len, total, sm_space, nodes;
 	uint64_t *entry, *entry_map, *entry_map_end;
 
 	ASSERT(MUTEX_HELD(sm->sm_lock));
@@ -455,13 +471,14 @@ space_map_sync(space_map_t *sm, uint8_t maptype,
 	    SM_DEBUG_SYNCPASS_ENCODE(spa_sync_pass(spa)) |
 	    SM_DEBUG_TXG_ENCODE(dmu_tx_get_txg(tx));
 
-	delta = 0;
+	total = 0;
+	nodes = avl_numnodes(&sm->sm_root);
 	sm_space = sm->sm_space;
-	while ((ss = avl_destroy_nodes(&sm->sm_root, &cookie)) != NULL) {
+	for (ss = avl_first(t); ss != NULL; ss = AVL_NEXT(t, ss)) {
 		size = ss->ss_end - ss->ss_start;
 		start = (ss->ss_start - sm->sm_start) >> sm->sm_shift;
 
-		delta += size;
+		total += size;
 		size >>= sm->sm_shift;
 
 		while (size) {
@@ -483,7 +500,6 @@ space_map_sync(space_map_t *sm, uint8_t maptype,
 			start += run_len;
 			size -= run_len;
 		}
-		kmem_cache_free(space_seg_cache, ss);
 	}
 
 	if (entry != entry_map) {
@@ -499,12 +515,11 @@ space_map_sync(space_map_t *sm, uint8_t maptype,
 	 * Ensure that the space_map's accounting wasn't changed
 	 * while we were in the middle of writing it out.
 	 */
+	VERIFY3U(nodes, ==, avl_numnodes(&sm->sm_root));
 	VERIFY3U(sm->sm_space, ==, sm_space);
+	VERIFY3U(sm->sm_space, ==, total);
 
 	zio_buf_free(entry_map, bufsize);
-
-	sm->sm_space -= delta;
-	VERIFY0(sm->sm_space);
 }
 
 void