Illumos #4374

4374 dn_free_ranges should use range_tree_t

Reviewed by: George Wilson <[email protected]>
Reviewed by: Max Grossman <[email protected]>
Reviewed by: Christopher Siden <[email protected]
Reviewed by: Garrett D'Amore <[email protected]>
Reviewed by: Dan McDonald <[email protected]>
Approved by: Dan McDonald <[email protected]>

References:
  https://www.illumos.org/issues/4374
  https://github.com/illumos/illumos-gate/commit/bf16b11

Ported by: Tim Chase <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #2531

19 files changed, 128 insertions, 174 deletions

diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c
index b29cddb6e..66b91cd97 100644
--- a/cmd/zdb/zdb.c
+++ b/cmd/zdb/zdb.c
@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <stdio.h>
@@ -2788,7 +2788,8 @@ dump_simulated_ddt(spa_t *spa)
 		dds.dds_ref_psize = zdde->zdde_ref_psize;
 		dds.dds_ref_dsize = zdde->zdde_ref_dsize;
 
-		ddt_stat_add(&ddh_total.ddh_stat[highbit(refcnt) - 1], &dds, 0);
+		ddt_stat_add(&ddh_total.ddh_stat[highbit64(refcnt) - 1],
+		    &dds, 0);
 
 		umem_free(zdde, sizeof (*zdde));
 	}
diff --git a/cmd/zpool/zpool_vdev.c b/cmd/zpool/zpool_vdev.c
index 31ee78c86..686782b07 100644
--- a/cmd/zpool/zpool_vdev.c
+++ b/cmd/zpool/zpool_vdev.c
@@ -740,7 +740,7 @@ make_leaf_vdev(nvlist_t *props, const char *arg, uint64_t is_log)
 		int sector_size;
 
 		if (check_sector_size_database(path, &sector_size) == B_TRUE)
-			ashift = highbit(sector_size) - 1;
+			ashift = highbit64(sector_size) - 1;
 	}
 
 	if (ashift > 0)
diff --git a/include/sys/dnode.h b/include/sys/dnode.h
index 4ca2a8a13..64bf000b6 100644
--- a/include/sys/dnode.h
+++ b/include/sys/dnode.h
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef	_SYS_DNODE_H
@@ -220,7 +220,7 @@ typedef struct dnode {
 	/* protected by dn_mtx: */
 	kmutex_t dn_mtx;
 	list_t dn_dirty_records[TXG_SIZE];
-	avl_tree_t dn_ranges[TXG_SIZE];
+	struct range_tree *dn_free_ranges[TXG_SIZE];
 	uint64_t dn_allocated_txg;
 	uint64_t dn_free_txg;
 	uint64_t dn_assigned_txg;
@@ -302,8 +302,6 @@ void dnode_buf_byteswap(void *buf, size_t size);
 void dnode_verify(dnode_t *dn);
 int dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx);
 void dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx);
-void dnode_clear_range(dnode_t *dn, uint64_t blkid,
-    uint64_t nblks, dmu_tx_t *tx);
 void dnode_diduse_space(dnode_t *dn, int64_t space);
 void dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx);
 void dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx, boolean_t);
diff --git a/include/sys/range_tree.h b/include/sys/range_tree.h
index a41effe4e..9f3ead537 100644
--- a/include/sys/range_tree.h
+++ b/include/sys/range_tree.h
@@ -24,7 +24,7 @@
  */
 
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef _SYS_RANGE_TREE_H
@@ -85,6 +85,7 @@ void range_tree_stat_verify(range_tree_t *rt);
 
 void range_tree_add(void *arg, uint64_t start, uint64_t size);
 void range_tree_remove(void *arg, uint64_t start, uint64_t size);
+void range_tree_clear(range_tree_t *rt, uint64_t start, uint64_t size);
 
 void range_tree_vacate(range_tree_t *rt, range_tree_func_t *func, void *arg);
 void range_tree_walk(range_tree_t *rt, range_tree_func_t *func, void *arg);
diff --git a/include/sys/zfs_context.h b/include/sys/zfs_context.h
index ec2a7dea0..fd3e0b0d4 100644
--- a/include/sys/zfs_context.h
+++ b/include/sys/zfs_context.h
@@ -25,7 +25,7 @@
 /*
  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef _SYS_ZFS_CONTEXT_H
@@ -623,7 +623,7 @@ extern void delay(clock_t ticks);
 
 extern uint64_t physmem;
 
-extern int highbit(ulong_t i);
+extern int highbit64(uint64_t i);
 extern int random_get_bytes(uint8_t *ptr, size_t len);
 extern int random_get_pseudo_bytes(uint8_t *ptr, size_t len);
 
diff --git a/lib/libzpool/kernel.c b/lib/libzpool/kernel.c
index 05bbd06cc..ba232425d 100644
--- a/lib/libzpool/kernel.c
+++ b/lib/libzpool/kernel.c
@@ -1014,17 +1014,15 @@ delay(clock_t ticks)
  * High order bit is 31 (or 63 in _LP64 kernel).
  */
 int
-highbit(ulong_t i)
+highbit64(uint64_t i)
 {
 	register int h = 1;
 
 	if (i == 0)
 		return (0);
-#ifdef _LP64
-	if (i & 0xffffffff00000000ul) {
+	if (i & 0xffffffff00000000ULL) {
 		h += 32; i >>= 32;
 	}
-#endif
 	if (i & 0xffff0000) {
 		h += 16; i >>= 16;
 	}
diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c
index f530e055b..c6e7197b6 100644
--- a/module/zfs/dbuf.c
+++ b/module/zfs/dbuf.c
@@ -21,7 +21,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  */
 
@@ -40,6 +40,7 @@
 #include <sys/dmu_zfetch.h>
 #include <sys/sa.h>
 #include <sys/sa_impl.h>
+#include <sys/range_tree.h>
 
 struct dbuf_hold_impl_data {
 	/* Function arguments */
@@ -1234,7 +1235,10 @@ dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx)
 	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
 	    db->db_blkid != DMU_SPILL_BLKID) {
 		mutex_enter(&dn->dn_mtx);
-		dnode_clear_range(dn, db->db_blkid, 1, tx);
+		if (dn->dn_free_ranges[txgoff] != NULL) {
+			range_tree_clear(dn->dn_free_ranges[txgoff],
+			    db->db_blkid, 1);
+		}
 		mutex_exit(&dn->dn_mtx);
 		db->db_freed_in_flight = FALSE;
 	}
diff --git a/module/zfs/ddt.c b/module/zfs/ddt.c
index 93997588d..d854e5f23 100644
--- a/module/zfs/ddt.c
+++ b/module/zfs/ddt.c
@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -423,7 +423,7 @@ ddt_stat_update(ddt_t *ddt, ddt_entry_t *dde, uint64_t neg)
 
 	ddt_stat_generate(ddt, dde, &dds);
 
-	bucket = highbit(dds.dds_ref_blocks) - 1;
+	bucket = highbit64(dds.dds_ref_blocks) - 1;
 	ASSERT(bucket >= 0);
 
 	ddh = &ddt->ddt_histogram[dde->dde_type][dde->dde_class];
diff --git a/module/zfs/dnode.c b/module/zfs/dnode.c
index 44ec80001..5aee10409 100644
--- a/module/zfs/dnode.c
+++ b/module/zfs/dnode.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -35,8 +35,7 @@
 #include <sys/spa.h>
 #include <sys/zio.h>
 #include <sys/dmu_zfetch.h>
-
-static int free_range_compar(const void *node1, const void *node2);
+#include <sys/range_tree.h>
 
 static kmem_cache_t *dnode_cache;
 /*
@@ -92,9 +91,7 @@ dnode_cons(void *arg, void *unused, int kmflag)
 
 	for (i = 0; i < TXG_SIZE; i++) {
 		list_link_init(&dn->dn_dirty_link[i]);
-		avl_create(&dn->dn_ranges[i], free_range_compar,
-		    sizeof (free_range_t),
-		    offsetof(struct free_range, fr_node));
+		dn->dn_free_ranges[i] = NULL;
 		list_create(&dn->dn_dirty_records[i],
 		    sizeof (dbuf_dirty_record_t),
 		    offsetof(dbuf_dirty_record_t, dr_dirty_node));
@@ -142,7 +139,7 @@ dnode_dest(void *arg, void *unused)
 
 	for (i = 0; i < TXG_SIZE; i++) {
 		ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
-		avl_destroy(&dn->dn_ranges[i]);
+		ASSERT3P(dn->dn_free_ranges[i], ==, NULL);
 		list_destroy(&dn->dn_dirty_records[i]);
 		ASSERT0(dn->dn_next_nblkptr[i]);
 		ASSERT0(dn->dn_next_nlevels[i]);
@@ -313,19 +310,6 @@ dnode_buf_byteswap(void *vbuf, size_t size)
 	}
 }
 
-static int
-free_range_compar(const void *node1, const void *node2)
-{
-	const free_range_t *rp1 = node1;
-	const free_range_t *rp2 = node2;
-
-	if (rp1->fr_blkid < rp2->fr_blkid)
-		return (-1);
-	else if (rp1->fr_blkid > rp2->fr_blkid)
-		return (1);
-	else return (0);
-}
-
 void
 dnode_setbonuslen(dnode_t *dn, int newsize, dmu_tx_t *tx)
 {
@@ -374,7 +358,7 @@ dnode_setdblksz(dnode_t *dn, int size)
 	    1<<(sizeof (dn->dn_phys->dn_datablkszsec) * 8));
 	dn->dn_datablksz = size;
 	dn->dn_datablkszsec = size >> SPA_MINBLOCKSHIFT;
-	dn->dn_datablkshift = ISP2(size) ? highbit(size - 1) : 0;
+	dn->dn_datablkshift = ISP2(size) ? highbit64(size - 1) : 0;
 }
 
 static dnode_t *
@@ -530,7 +514,7 @@ dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
 		ASSERT0(dn->dn_next_blksz[i]);
 		ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
 		ASSERT3P(list_head(&dn->dn_dirty_records[i]), ==, NULL);
-		ASSERT0(avl_numnodes(&dn->dn_ranges[i]));
+		ASSERT3P(dn->dn_free_ranges[i], ==, NULL);
 	}
 
 	dn->dn_type = ot;
@@ -695,7 +679,8 @@ dnode_move_impl(dnode_t *odn, dnode_t *ndn)
 		list_move_tail(&ndn->dn_dirty_records[i],
 		    &odn->dn_dirty_records[i]);
 	}
-	bcopy(&odn->dn_ranges[0], &ndn->dn_ranges[0], sizeof (odn->dn_ranges));
+	bcopy(&odn->dn_free_ranges[0], &ndn->dn_free_ranges[0],
+	    sizeof (odn->dn_free_ranges));
 	ndn->dn_allocated_txg = odn->dn_allocated_txg;
 	ndn->dn_free_txg = odn->dn_free_txg;
 	ndn->dn_assigned_txg = odn->dn_assigned_txg;
@@ -758,8 +743,7 @@ dnode_move_impl(dnode_t *odn, dnode_t *ndn)
 		list_create(&odn->dn_dirty_records[i],
 		    sizeof (dbuf_dirty_record_t),
 		    offsetof(dbuf_dirty_record_t, dr_dirty_node));
-		odn->dn_ranges[i].avl_root = NULL;
-		odn->dn_ranges[i].avl_numnodes = 0;
+		odn->dn_free_ranges[i] = NULL;
 		odn->dn_next_nlevels[i] = 0;
 		odn->dn_next_indblkshift[i] = 0;
 		odn->dn_next_bonustype[i] = 0;
@@ -1463,59 +1447,6 @@ out:
 }
 
 void
-dnode_clear_range(dnode_t *dn, uint64_t blkid, uint64_t nblks, dmu_tx_t *tx)
-{
-	avl_tree_t *tree = &dn->dn_ranges[tx->tx_txg&TXG_MASK];
-	avl_index_t where;
-	free_range_t *rp;
-	free_range_t rp_tofind;
-	uint64_t endblk = blkid + nblks;
-
-	ASSERT(MUTEX_HELD(&dn->dn_mtx));
-	ASSERT(nblks <= UINT64_MAX - blkid); /* no overflow */
-
-	dprintf_dnode(dn, "blkid=%llu nblks=%llu txg=%llu\n",
-	    blkid, nblks, tx->tx_txg);
-	rp_tofind.fr_blkid = blkid;
-	rp = avl_find(tree, &rp_tofind, &where);
-	if (rp == NULL)
-		rp = avl_nearest(tree, where, AVL_BEFORE);
-	if (rp == NULL)
-		rp = avl_nearest(tree, where, AVL_AFTER);
-
-	while (rp && (rp->fr_blkid <= blkid + nblks)) {
-		uint64_t fr_endblk = rp->fr_blkid + rp->fr_nblks;
-		free_range_t *nrp = AVL_NEXT(tree, rp);
-
-		if (blkid <= rp->fr_blkid && endblk >= fr_endblk) {
-			/* clear this entire range */
-			avl_remove(tree, rp);
-			kmem_free(rp, sizeof (free_range_t));
-		} else if (blkid <= rp->fr_blkid &&
-		    endblk > rp->fr_blkid && endblk < fr_endblk) {
-			/* clear the beginning of this range */
-			rp->fr_blkid = endblk;
-			rp->fr_nblks = fr_endblk - endblk;
-		} else if (blkid > rp->fr_blkid && blkid < fr_endblk &&
-		    endblk >= fr_endblk) {
-			/* clear the end of this range */
-			rp->fr_nblks = blkid - rp->fr_blkid;
-		} else if (blkid > rp->fr_blkid && endblk < fr_endblk) {
-			/* clear a chunk out of this range */
-			free_range_t *new_rp =
-			    kmem_alloc(sizeof (free_range_t), KM_PUSHPAGE);
-
-			new_rp->fr_blkid = endblk;
-			new_rp->fr_nblks = fr_endblk - endblk;
-			avl_insert_here(tree, new_rp, rp, AVL_AFTER);
-			rp->fr_nblks = blkid - rp->fr_blkid;
-		}
-		/* there may be no overlap */
-		rp = nrp;
-	}
-}
-
-void
 dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx)
 {
 	dmu_buf_impl_t *db;
@@ -1665,22 +1596,17 @@ done:
 	 * We will finish up this free operation in the syncing phase.
 	 */
 	mutex_enter(&dn->dn_mtx);
-	dnode_clear_range(dn, blkid, nblks, tx);
 	{
-		free_range_t *rp, *found;
-		avl_index_t where;
-		avl_tree_t *tree = &dn->dn_ranges[tx->tx_txg&TXG_MASK];
-
-		/* Add new range to dn_ranges */
-		rp = kmem_alloc(sizeof (free_range_t), KM_PUSHPAGE);
-		rp->fr_blkid = blkid;
-		rp->fr_nblks = nblks;
-		found = avl_find(tree, rp, &where);
-		ASSERT(found == NULL);
-		avl_insert(tree, rp, where);
-		dprintf_dnode(dn, "blkid=%llu nblks=%llu txg=%llu\n",
-		    blkid, nblks, tx->tx_txg);
+	int txgoff = tx->tx_txg & TXG_MASK;
+	if (dn->dn_free_ranges[txgoff] == NULL) {
+		dn->dn_free_ranges[txgoff] =
+		    range_tree_create(NULL, NULL, &dn->dn_mtx);
+	}
+	range_tree_clear(dn->dn_free_ranges[txgoff], blkid, nblks);
+	range_tree_add(dn->dn_free_ranges[txgoff], blkid, nblks);
 	}
+	dprintf_dnode(dn, "blkid=%llu nblks=%llu txg=%llu\n",
+	    blkid, nblks, tx->tx_txg);
 	mutex_exit(&dn->dn_mtx);
 
 	dbuf_free_range(dn, blkid, blkid + nblks - 1, tx);
@@ -1708,7 +1634,6 @@ dnode_spill_freed(dnode_t *dn)
 uint64_t
 dnode_block_freed(dnode_t *dn, uint64_t blkid)
 {
-	free_range_t range_tofind;
 	void *dp = spa_get_dsl(dn->dn_objset->os_spa);
 	int i;
 
@@ -1728,20 +1653,10 @@ dnode_block_freed(dnode_t *dn, uint64_t blkid)
 	if (blkid == DMU_SPILL_BLKID)
 		return (dnode_spill_freed(dn));
 
-	range_tofind.fr_blkid = blkid;
 	mutex_enter(&dn->dn_mtx);
 	for (i = 0; i < TXG_SIZE; i++) {
-		free_range_t *range_found;
-		avl_index_t idx;
-
-		range_found = avl_find(&dn->dn_ranges[i], &range_tofind, &idx);
-		if (range_found) {
-			ASSERT(range_found->fr_nblks > 0);
-			break;
-		}
-		range_found = avl_nearest(&dn->dn_ranges[i], idx, AVL_BEFORE);
-		if (range_found &&
-		    range_found->fr_blkid + range_found->fr_nblks > blkid)
+		if (dn->dn_free_ranges[i] != NULL &&
+		    range_tree_contains(dn->dn_free_ranges[i], blkid, 1))
 			break;
 	}
 	mutex_exit(&dn->dn_mtx);
diff --git a/module/zfs/dnode_sync.c b/module/zfs/dnode_sync.c
index 885dc2dbc..238920060 100644
--- a/module/zfs/dnode_sync.c
+++ b/module/zfs/dnode_sync.c
@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -32,6 +32,7 @@
 #include <sys/dmu_objset.h>
 #include <sys/dsl_dataset.h>
 #include <sys/spa.h>
+#include <sys/range_tree.h>
 #include <sys/zfeature.h>
 
 static void
@@ -322,7 +323,7 @@ free_children(dmu_buf_impl_t *db, uint64_t blkid, uint64_t nblks,
  * and "free" all the blocks contained there.
  */
 static void
-dnode_sync_free_range(dnode_t *dn, uint64_t blkid, uint64_t nblks,
+dnode_sync_free_range_impl(dnode_t *dn, uint64_t blkid, uint64_t nblks,
     dmu_tx_t *tx)
 {
 	blkptr_t *bp = dn->dn_phys->dn_blkptr;
@@ -381,6 +382,22 @@ dnode_sync_free_range(dnode_t *dn, uint64_t blkid, uint64_t nblks,
 	}
 }
 
+typedef struct dnode_sync_free_range_arg {
+	dnode_t *dsfra_dnode;
+	dmu_tx_t *dsfra_tx;
+} dnode_sync_free_range_arg_t;
+
+static void
+dnode_sync_free_range(void *arg, uint64_t blkid, uint64_t nblks)
+{
+	dnode_sync_free_range_arg_t *dsfra = arg;
+	dnode_t *dn = dsfra->dsfra_dnode;
+
+	mutex_exit(&dn->dn_mtx);
+	dnode_sync_free_range_impl(dn, blkid, nblks, dsfra->dsfra_tx);
+	mutex_enter(&dn->dn_mtx);
+}
+
 /*
  * Try to kick all the dnode's dbufs out of the cache...
  */
@@ -542,7 +559,6 @@ dnode_sync_free(dnode_t *dn, dmu_tx_t *tx)
 void
 dnode_sync(dnode_t *dn, dmu_tx_t *tx)
 {
-	free_range_t *rp;
 	dnode_phys_t *dnp = dn->dn_phys;
 	int txgoff = tx->tx_txg & TXG_MASK;
 	list_t *list = &dn->dn_dirty_records[txgoff];
@@ -601,9 +617,9 @@ dnode_sync(dnode_t *dn, dmu_tx_t *tx)
 		    SPA_MINBLOCKSIZE) == 0);
 		ASSERT(BP_IS_HOLE(&dnp->dn_blkptr[0]) ||
 		    dn->dn_maxblkid == 0 || list_head(list) != NULL ||
-		    avl_last(&dn->dn_ranges[txgoff]) ||
 		    dn->dn_next_blksz[txgoff] >> SPA_MINBLOCKSHIFT ==
-		    dnp->dn_datablkszsec);
+		    dnp->dn_datablkszsec ||
+		    range_tree_space(dn->dn_free_ranges[txgoff]) != 0);
 		dnp->dn_datablkszsec =
 		    dn->dn_next_blksz[txgoff] >> SPA_MINBLOCKSHIFT;
 		dn->dn_next_blksz[txgoff] = 0;
@@ -661,13 +677,16 @@ dnode_sync(dnode_t *dn, dmu_tx_t *tx)
 	}
 
 	/* process all the "freed" ranges in the file */
-	while ((rp = avl_last(&dn->dn_ranges[txgoff]))) {
-		dnode_sync_free_range(dn, rp->fr_blkid, rp->fr_nblks, tx);
-		/* grab the mutex so we don't race with dnode_block_freed() */
+	if (dn->dn_free_ranges[txgoff] != NULL) {
+		dnode_sync_free_range_arg_t dsfra;
+		dsfra.dsfra_dnode = dn;
+		dsfra.dsfra_tx = tx;
 		mutex_enter(&dn->dn_mtx);
-		avl_remove(&dn->dn_ranges[txgoff], rp);
+		range_tree_vacate(dn->dn_free_ranges[txgoff],
+		    dnode_sync_free_range, &dsfra);
+		range_tree_destroy(dn->dn_free_ranges[txgoff]);
+		dn->dn_free_ranges[txgoff] = NULL;
 		mutex_exit(&dn->dn_mtx);
-		kmem_free(rp, sizeof (free_range_t));
 	}
 
 	if (freeing_dnode) {
diff --git a/module/zfs/metaslab.c b/module/zfs/metaslab.c
index 6d6ff0914..06d7f7cd2 100644
--- a/module/zfs/metaslab.c
+++ b/module/zfs/metaslab.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  */
 
@@ -703,7 +703,7 @@ metaslab_ff_alloc(metaslab_t *msp, uint64_t size)
 	 * may exist in the same region.
 	 */
 	uint64_t align = size & -size;
-	uint64_t *cursor = &msp->ms_lbas[highbit(align) - 1];
+	uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
 	avl_tree_t *t = &msp->ms_tree->rt_root;
 
 	return (metaslab_block_picker(t, cursor, size, align));
@@ -744,7 +744,7 @@ metaslab_df_alloc(metaslab_t *msp, uint64_t size)
 	 * may exist in the same region.
 	 */
 	uint64_t align = size & -size;
-	uint64_t *cursor = &msp->ms_lbas[highbit(align) - 1];
+	uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
 	range_tree_t *rt = msp->ms_tree;
 	avl_tree_t *t = &rt->rt_root;
 	uint64_t max_size = metaslab_block_maxsize(msp);
@@ -869,7 +869,7 @@ metaslab_ndf_alloc(metaslab_t *msp, uint64_t size)
 	avl_tree_t *t = &msp->ms_tree->rt_root;
 	avl_index_t where;
 	range_seg_t *rs, rsearch;
-	uint64_t hbit = highbit(size);
+	uint64_t hbit = highbit64(size);
 	uint64_t *cursor = &msp->ms_lbas[hbit - 1];
 	uint64_t max_size = metaslab_block_maxsize(msp);
 
@@ -1117,7 +1117,7 @@ metaslab_weight_factor(metaslab_t *msp)
 	if (msp->ms_sm == NULL) {
 		vdev_t *vd = msp->ms_group->mg_vd;
 
-		i = highbit(msp->ms_size) - 1;
+		i = highbit64(msp->ms_size) - 1;
 		sectors = msp->ms_size >> vd->vdev_ashift;
 		return (sectors * i * vd->vdev_ashift);
 	}
diff --git a/module/zfs/range_tree.c b/module/zfs/range_tree.c
index cb4641078..f35d20c08 100644
--- a/module/zfs/range_tree.c
+++ b/module/zfs/range_tree.c
@@ -23,7 +23,7 @@
  * Use is subject to license terms.
  */
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -60,7 +60,7 @@ range_tree_stat_verify(range_tree_t *rt)
 	for (rs = avl_first(&rt->rt_root); rs != NULL;
 	    rs = AVL_NEXT(&rt->rt_root, rs)) {
 		uint64_t size = rs->rs_end - rs->rs_start;
-		int idx	= highbit(size) - 1;
+		int idx	= highbit64(size) - 1;
 
 		hist[idx]++;
 		ASSERT3U(hist[idx], !=, 0);
@@ -79,7 +79,7 @@ static void
 range_tree_stat_incr(range_tree_t *rt, range_seg_t *rs)
 {
 	uint64_t size = rs->rs_end - rs->rs_start;
-	int idx = highbit(size) - 1;
+	int idx = highbit64(size) - 1;
 
 	ASSERT3U(idx, <,
 	    sizeof (rt->rt_histogram) / sizeof (*rt->rt_histogram));
@@ -93,7 +93,7 @@ static void
 range_tree_stat_decr(range_tree_t *rt, range_seg_t *rs)
 {
 	uint64_t size = rs->rs_end - rs->rs_start;
-	int idx = highbit(size) - 1;
+	int idx = highbit64(size) - 1;
 
 	ASSERT3U(idx, <,
 	    sizeof (rt->rt_histogram) / sizeof (*rt->rt_histogram));
@@ -299,10 +299,10 @@ range_tree_remove(void *arg, uint64_t start, uint64_t size)
 }
 
 static range_seg_t *
-range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size,
-    avl_index_t *wherep)
+range_tree_find_impl(range_tree_t *rt, uint64_t start, uint64_t size)
 {
-	range_seg_t rsearch, *rs;
+	avl_index_t where;
+	range_seg_t rsearch;
 	uint64_t end = start + size;
 
 	ASSERT(MUTEX_HELD(rt->rt_lock));
@@ -310,9 +310,14 @@ range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size,
 
 	rsearch.rs_start = start;
 	rsearch.rs_end = end;
-	rs = avl_find(&rt->rt_root, &rsearch, wherep);
+	return (avl_find(&rt->rt_root, &rsearch, &where));
+}
 
-	if (rs != NULL && rs->rs_start <= start && rs->rs_end >= end)
+static range_seg_t *
+range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size)
+{
+	range_seg_t *rs = range_tree_find_impl(rt, start, size);
+	if (rs != NULL && rs->rs_start <= start && rs->rs_end >= start + size)
 		return (rs);
 	return (NULL);
 }
@@ -321,10 +326,9 @@ void
 range_tree_verify(range_tree_t *rt, uint64_t off, uint64_t size)
 {
 	range_seg_t *rs;
-	avl_index_t where;
 
 	mutex_enter(rt->rt_lock);
-	rs = range_tree_find(rt, off, size, &where);
+	rs = range_tree_find(rt, off, size);
 	if (rs != NULL)
 		panic("freeing free block; rs=%p", (void *)rs);
 	mutex_exit(rt->rt_lock);
@@ -333,9 +337,23 @@ range_tree_verify(range_tree_t *rt, uint64_t off, uint64_t size)
 boolean_t
 range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size)
 {
-	avl_index_t where;
+	return (range_tree_find(rt, start, size) != NULL);
+}
 
-	return (range_tree_find(rt, start, size, &where) != NULL);
+/*
+ * Ensure that this range is not in the tree, regardless of whether
+ * it is currently in the tree.
+ */
+void
+range_tree_clear(range_tree_t *rt, uint64_t start, uint64_t size)
+{
+	range_seg_t *rs;
+
+	while ((rs = range_tree_find_impl(rt, start, size)) != NULL) {
+		uint64_t free_start = MAX(rs->rs_start, start);
+		uint64_t free_end = MIN(rs->rs_end, start + size);
+		range_tree_remove(rt, free_start, free_end - free_start);
+	}
 }
 
 void
diff --git a/module/zfs/space_map.c b/module/zfs/space_map.c
index 645a81387..c7a2312b4 100644
--- a/module/zfs/space_map.c
+++ b/module/zfs/space_map.c
@@ -23,7 +23,7 @@
  * Use is subject to license terms.
  */
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -270,7 +270,7 @@ space_map_set_blocksize(space_map_t *sm, uint64_t size, dmu_tx_t *tx)
 		 * adding more blocks. The block size can grow until it
 		 * reaches space_map_max_blksz.
 		 */
-		newsz = ISP2(size) ? size : 1ULL << highbit(size);
+		newsz = ISP2(size) ? size : 1ULL << highbit64(size);
 		if (newsz > space_map_max_blksz)
 			newsz = space_map_max_blksz;
 
diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c
index 69a315317..89df3674e 100644
--- a/module/zfs/vdev.c
+++ b/module/zfs/vdev.c
@@ -22,7 +22,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -1584,7 +1584,7 @@ vdev_metaslab_set_size(vdev_t *vd)
 	/*
 	 * Aim for roughly 200 metaslabs per vdev.
 	 */
-	vd->vdev_ms_shift = highbit(vd->vdev_asize / 200);
+	vd->vdev_ms_shift = highbit64(vd->vdev_asize / 200);
 	vd->vdev_ms_shift = MAX(vd->vdev_ms_shift, SPA_MAXBLOCKSHIFT);
 }
 
diff --git a/module/zfs/vdev_disk.c b/module/zfs/vdev_disk.c
index ee14165f5..9cc6745d1 100644
--- a/module/zfs/vdev_disk.c
+++ b/module/zfs/vdev_disk.c
@@ -23,7 +23,7 @@
  * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
  * Rewritten for Linux by Brian Behlendorf <[email protected]>.
  * LLNL-CODE-403049.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -308,7 +308,7 @@ skip_open:
 	*max_psize = *psize;
 
 	/* Based on the minimum sector size set the block size */
-	*ashift = highbit(MAX(block_size, SPA_MINBLOCKSIZE)) - 1;
+	*ashift = highbit64(MAX(block_size, SPA_MINBLOCKSIZE)) - 1;
 
 	/* Try to set the io scheduler elevator algorithm */
 	(void) vdev_elevator_switch(v, zfs_vdev_scheduler);
diff --git a/module/zfs/zap.c b/module/zfs/zap.c
index cfae26a74..02c92748f 100644
--- a/module/zfs/zap.c
+++ b/module/zfs/zap.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 /*
@@ -84,7 +84,7 @@ fzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags)
 	    &zap->zap_f.zap_phys, zap_evict);
 
 	mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
-	zap->zap_f.zap_block_shift = highbit(zap->zap_dbuf->db_size) - 1;
+	zap->zap_f.zap_block_shift = highbit64(zap->zap_dbuf->db_size) - 1;
 
 	zp = zap->zap_f.zap_phys;
 	/*
@@ -458,7 +458,7 @@ zap_open_leaf(uint64_t blkid, dmu_buf_t *db)
 	rw_init(&l->l_rwlock, NULL, RW_DEFAULT, NULL);
 	rw_enter(&l->l_rwlock, RW_WRITER);
 	l->l_blkid = blkid;
-	l->l_bs = highbit(db->db_size)-1;
+	l->l_bs = highbit64(db->db_size) - 1;
 	l->l_dbuf = db;
 	l->l_phys = NULL;
 
diff --git a/module/zfs/zap_leaf.c b/module/zfs/zap_leaf.c
index 13bc879d7..0e30ef319 100644
--- a/module/zfs/zap_leaf.c
+++ b/module/zfs/zap_leaf.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 /*
@@ -105,16 +105,16 @@ zap_leaf_byteswap(zap_leaf_phys_t *buf, int size)
 {
 	int i;
 	zap_leaf_t l;
-	l.l_bs = highbit(size)-1;
+	l.l_bs = highbit64(size) - 1;
 	l.l_phys = buf;
 
-	buf->l_hdr.lh_block_type = 	BSWAP_64(buf->l_hdr.lh_block_type);
-	buf->l_hdr.lh_prefix = 		BSWAP_64(buf->l_hdr.lh_prefix);
-	buf->l_hdr.lh_magic = 		BSWAP_32(buf->l_hdr.lh_magic);
-	buf->l_hdr.lh_nfree = 		BSWAP_16(buf->l_hdr.lh_nfree);
-	buf->l_hdr.lh_nentries = 	BSWAP_16(buf->l_hdr.lh_nentries);
-	buf->l_hdr.lh_prefix_len = 	BSWAP_16(buf->l_hdr.lh_prefix_len);
-	buf->l_hdr.lh_freelist = 	BSWAP_16(buf->l_hdr.lh_freelist);
+	buf->l_hdr.lh_block_type =	BSWAP_64(buf->l_hdr.lh_block_type);
+	buf->l_hdr.lh_prefix =		BSWAP_64(buf->l_hdr.lh_prefix);
+	buf->l_hdr.lh_magic =		BSWAP_32(buf->l_hdr.lh_magic);
+	buf->l_hdr.lh_nfree =		BSWAP_16(buf->l_hdr.lh_nfree);
+	buf->l_hdr.lh_nentries =	BSWAP_16(buf->l_hdr.lh_nentries);
+	buf->l_hdr.lh_prefix_len =	BSWAP_16(buf->l_hdr.lh_prefix_len);
+	buf->l_hdr.lh_freelist =	BSWAP_16(buf->l_hdr.lh_freelist);
 
 	for (i = 0; i < ZAP_LEAF_HASH_NUMENTRIES(&l); i++)
 		buf->l_hash[i] = BSWAP_16(buf->l_hash[i]);
@@ -157,7 +157,7 @@ zap_leaf_init(zap_leaf_t *l, boolean_t sort)
 {
 	int i;
 
-	l->l_bs = highbit(l->l_dbuf->db_size)-1;
+	l->l_bs = highbit64(l->l_dbuf->db_size) - 1;
 	zap_memset(&l->l_phys->l_hdr, 0, sizeof (struct zap_leaf_header));
 	zap_memset(l->l_phys->l_hash, CHAIN_END, 2*ZAP_LEAF_HASH_NUMENTRIES(l));
 	for (i = 0; i < ZAP_LEAF_NUMCHUNKS(l); i++) {
diff --git a/module/zfs/zap_micro.c b/module/zfs/zap_micro.c
index 73d84b554..68fb74769 100644
--- a/module/zfs/zap_micro.c
+++ b/module/zfs/zap_micro.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zio.h>
@@ -374,7 +374,7 @@ mzap_open(objset_t *os, uint64_t obj, dmu_buf_t *db)
 
 	if (*(uint64_t *)db->db_data != ZBT_MICRO) {
 		mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
-		zap->zap_f.zap_block_shift = highbit(db->db_size) - 1;
+		zap->zap_f.zap_block_shift = highbit64(db->db_size) - 1;
 	} else {
 		zap->zap_ismicro = TRUE;
 	}
diff --git a/module/zfs/zio.c b/module/zfs/zio.c
index f84e2129b..7e440a381 100644
--- a/module/zfs/zio.c
+++ b/module/zfs/zio.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
  */
 
@@ -1367,7 +1367,7 @@ __zio_execute(zio_t *zio)
 		}
 
 		zio->io_stage = stage;
-		rv = zio_pipeline[highbit(stage) - 1](zio);
+		rv = zio_pipeline[highbit64(stage) - 1](zio);
 
 		if (rv == ZIO_PIPELINE_STOP)
 			return;