4996.patch2

diff --git a/usr/src/uts/common/fs/zfs/dnode.c b/usr/src/uts/common/fs/zfs/dnode.c
index 7fdb64d..540b23e 100644
--- a/usr/src/uts/common/fs/zfs/dnode.c
+++ b/usr/src/uts/common/fs/zfs/dnode.c
@@ -1552,6 +1552,72 @@ dnode_dirty_l1(dnode_t *dn, uint64_t l1blkid, dmu_tx_t *tx)
 	}
 }
 
+/*
+ * Dirty all the in-core level-1 dbufs in the range specified by start_blkid
+ * and end_blkid.
+ */
+static void
+dnode_dirty_l1range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid,
+    dmu_tx_t *tx)
+{
+       dmu_buf_impl_t db_search;
+       dmu_buf_impl_t *db;
+       avl_index_t where;
+
+       mutex_enter(&dn->dn_dbufs_mtx);
+
+       db_search.db_level = 1;
+       db_search.db_blkid = start_blkid + 1;
+       db_search.db_state = DB_SEARCH;
+       for (;;) {
+
+               db = avl_find(&dn->dn_dbufs, &db_search, &where);
+               if (db == NULL)
+                       db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER);
+
+               if (db == NULL || db->db_level != 1 ||
+                   db->db_blkid >= end_blkid) {
+                       break;
+               }
+
+               /*
+                * Setup the next blkid we want to search for.
+                */
+               db_search.db_blkid = db->db_blkid + 1;
+               ASSERT3U(db->db_blkid, >=, start_blkid);
+
+               /*
+                * If the dbuf transitions to DB_EVICTING while we're trying
+                * to dirty it, then we will be unable to discover it in
+                * the dbuf hash table. This will result in a call to
+                * dbuf_create() which needs to acquire the dn_dbufs_mtx
+                * lock. To avoid a deadlock, we drop the lock before
+                * dirtying the level-1 dbuf.
+                */
+               mutex_exit(&dn->dn_dbufs_mtx);
+               dnode_dirty_l1(dn, db->db_blkid, tx);
+               mutex_enter(&dn->dn_dbufs_mtx);
+       }
+
+#ifdef ZFS_DEBUG
+       /*
+        * Walk all the in-core level-1 dbufs and verify they have been dirtied.
+        */
+       db_search.db_level = 1;
+       db_search.db_blkid = start_blkid + 1;
+       db_search.db_state = DB_SEARCH;
+       db = avl_find(&dn->dn_dbufs, &db_search, &where);
+       if (db == NULL)
+               db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER);
+       for (; db != NULL; db = AVL_NEXT(&dn->dn_dbufs, db)) {
+               if (db->db_level != 1 || db->db_blkid >= end_blkid)
+                       break;
+               ASSERT(db->db_dirtycnt > 0);
+       }
+#endif
+       mutex_exit(&dn->dn_dbufs_mtx);
+}
+
 void
 dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx)
 {
@@ -1704,6 +1770,8 @@ dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx)
 		if (last != first)
 			dnode_dirty_l1(dn, last, tx);
 
+		dnode_dirty_l1range(dn, first, last, tx);
+
 		int shift = dn->dn_datablkshift + dn->dn_indblkshift -
 		    SPA_BLKPTRSHIFT;
 
diff --git a/usr/src/uts/common/fs/zfs/dnode_sync.c b/usr/src/uts/common/fs/zfs/dnode_sync.c
index d4e6502..d2b7b9f 100644
--- a/usr/src/uts/common/fs/zfs/dnode_sync.c
+++ b/usr/src/uts/common/fs/zfs/dnode_sync.c
@@ -237,7 +237,7 @@ free_verify(dmu_buf_impl_t *db, uint64_t start, uint64_t end, dmu_tx_t *tx)
 
 static void
 free_children(dmu_buf_impl_t *db, uint64_t blkid, uint64_t nblks,
-    dmu_tx_t *tx)
+    boolean_t force_free, dmu_tx_t *tx)
 {
 	dnode_t *dn;
 	blkptr_t *bp;
@@ -289,32 +289,21 @@ free_children(dmu_buf_impl_t *db, uint64_t blkid, uint64_t nblks,
 			rw_exit(&dn->dn_struct_rwlock);
 			ASSERT3P(bp, ==, subdb->db_blkptr);
 
-			free_children(subdb, blkid, nblks, tx);
+			free_children(subdb, blkid, nblks, force_free, tx);
 			dbuf_rele(subdb, FTAG);
 		}
 	}
 
-	/* If this whole block is free, free ourself too. */
-	for (i = 0, bp = db->db.db_data; i < 1 << epbs; i++, bp++) {
-		if (!BP_IS_HOLE(bp))
-			break;
-	}
-	if (i == 1 << epbs) {
+	if (force_free) {
+		for (i = 0, bp = db->db.db_data; i < 1 << epbs; i++, bp++)
+			ASSERT(BP_IS_HOLE(bp));
 		/*
-		 * We only found holes. Grab the rwlock to prevent
-		 * anybody from reading the blocks we're about to
-		 * zero out.
+		 * TODO do we need rwlock
 		 */
 		rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
 		bzero(db->db.db_data, db->db.db_size);
 		rw_exit(&dn->dn_struct_rwlock);
 		free_blocks(dn, db->db_blkptr, 1, tx);
-	} else {
-		/*
-		 * Partial block free; must be marked dirty so that it
-		 * will be written out.
-		 */
-		ASSERT(db->db_dirtycnt > 0);
 	}
 
 	DB_DNODE_EXIT(db);
@@ -327,7 +316,7 @@ free_children(dmu_buf_impl_t *db, uint64_t blkid, uint64_t nblks,
  */
 static void
 dnode_sync_free_range_impl(dnode_t *dn, uint64_t blkid, uint64_t nblks,
-    dmu_tx_t *tx)
+    boolean_t force_free, dmu_tx_t *tx)
 {
 	blkptr_t *bp = dn->dn_phys->dn_blkptr;
 	int dnlevel = dn->dn_phys->dn_nlevels;
@@ -367,7 +356,7 @@ dnode_sync_free_range_impl(dnode_t *dn, uint64_t blkid, uint64_t nblks,
 			    TRUE, FALSE, FTAG, &db));
 			rw_exit(&dn->dn_struct_rwlock);
 
-			free_children(db, blkid, nblks, tx);
+			free_children(db, blkid, nblks, force_free, tx);
 			dbuf_rele(db, FTAG);
 		}
 	}
@@ -386,6 +375,7 @@ dnode_sync_free_range_impl(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;
+	boolean_t dsfra_force_free;
 } dnode_sync_free_range_arg_t;
 
 static void
@@ -395,7 +385,8 @@ dnode_sync_free_range(void *arg, uint64_t blkid, uint64_t nblks)
 	dnode_t *dn = dsfra->dsfra_dnode;
 
 	mutex_exit(&dn->dn_mtx);
-	dnode_sync_free_range_impl(dn, blkid, nblks, dsfra->dsfra_tx);
+	dnode_sync_free_range_impl(dn, blkid, nblks, dsfra->dsfra_force_free,
+	    dsfra->dsfra_tx);
 	mutex_enter(&dn->dn_mtx);
 }
 
@@ -672,6 +663,7 @@ dnode_sync(dnode_t *dn, dmu_tx_t *tx)
 		dnode_sync_free_range_arg_t dsfra;
 		dsfra.dsfra_dnode = dn;
 		dsfra.dsfra_tx = tx;
+		dsfra.dsfra_force_free = freeing_dnode;
 		mutex_enter(&dn->dn_mtx);
 		range_tree_vacate(dn->dn_free_ranges[txgoff],
 		    dnode_sync_free_range, &dsfra);
diff --git a/usr/src/uts/common/fs/zfs/zfs_znode.c b/usr/src/uts/common/fs/zfs/zfs_znode.c
index 6304ec8..dd2cc1d 100644
--- a/usr/src/uts/common/fs/zfs/zfs_znode.c
+++ b/usr/src/uts/common/fs/zfs/zfs_znode.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
  */
 
 /* Portions Copyright 2007 Jeremy Teo */
@@ -1590,7 +1590,8 @@ zfs_trunc(znode_t *zp, uint64_t end)
 		return (0);
 	}
 
-	error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end,  -1);
+	error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end,
+	    DMU_OBJECT_END);
 	if (error) {
 		zfs_range_unlock(rl);
 		return (error);