Handle zap_add() failures in mixed case mode

With "casesensitivity=mixed", zap_add() could fail when the number of
files/directories with the same name (varying in case) exceed the
capacity of the leaf node of a Fatzap. This results in a ASSERT()
failure as zfs_link_create() does not expect zap_add() to fail. The fix
is to handle these failures and rollback the transactions.

Reviewed by: Matt Ahrens <[email protected]>
Reviewed-by: Chunwei Chen <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Sanjeev Bagewadi <[email protected]>
Closes #7011 
Closes #7054 

1 files changed, 56 insertions, 18 deletions

diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
index 5e34cb2ff..e21fe1aca 100644
--- a/module/zfs/zfs_vnops.c
+++ b/module/zfs/zfs_vnops.c
@@ -1429,6 +1429,7 @@ top:
 			dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
 			    0, acl_ids.z_aclp->z_acl_bytes);
 		}
+
 		error = dmu_tx_assign(tx,
 		    (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
 		if (error) {
@@ -1446,10 +1447,22 @@ top:
 		}
 		zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
 
+		error = zfs_link_create(dl, zp, tx, ZNEW);
+		if (error != 0) {
+			/*
+			 * Since, we failed to add the directory entry for it,
+			 * delete the newly created dnode.
+			 */
+			zfs_znode_delete(zp, tx);
+			remove_inode_hash(ZTOI(zp));
+			zfs_acl_ids_free(&acl_ids);
+			dmu_tx_commit(tx);
+			goto out;
+		}
+
 		if (fuid_dirtied)
 			zfs_fuid_sync(zfsvfs, tx);
 
-		(void) zfs_link_create(dl, zp, tx, ZNEW);
 		txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
 		if (flag & FIGNORECASE)
 			txtype |= TX_CI;
@@ -2040,13 +2053,18 @@ top:
 	 */
 	zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
 
-	if (fuid_dirtied)
-		zfs_fuid_sync(zfsvfs, tx);
-
 	/*
 	 * Now put new name in parent dir.
 	 */
-	(void) zfs_link_create(dl, zp, tx, ZNEW);
+	error = zfs_link_create(dl, zp, tx, ZNEW);
+	if (error != 0) {
+		zfs_znode_delete(zp, tx);
+		remove_inode_hash(ZTOI(zp));
+		goto out;
+	}
+
+	if (fuid_dirtied)
+		zfs_fuid_sync(zfsvfs, tx);
 
 	*ipp = ZTOI(zp);
 
@@ -2056,6 +2074,7 @@ top:
 	zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp,
 	    acl_ids.z_fuidp, vap);
 
+out:
 	zfs_acl_ids_free(&acl_ids);
 
 	dmu_tx_commit(tx);
@@ -2065,10 +2084,14 @@ top:
 	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
 		zil_commit(zilog, 0);
 
-	zfs_inode_update(dzp);
-	zfs_inode_update(zp);
+	if (error != 0) {
+		iput(ZTOI(zp));
+	} else {
+		zfs_inode_update(dzp);
+		zfs_inode_update(zp);
+	}
 	ZFS_EXIT(zfsvfs);
-	return (0);
+	return (error);
 }
 
 /*
@@ -3686,6 +3709,13 @@ top:
 				VERIFY3U(zfs_link_destroy(tdl, szp, tx,
 				    ZRENAMING, NULL), ==, 0);
 			}
+		} else {
+			/*
+			 * If we had removed the existing target, subsequent
+			 * call to zfs_link_create() to add back the same entry
+			 * but, the new dnode (szp) should not fail.
+			 */
+			ASSERT(tzp == NULL);
 		}
 	}
 
@@ -3856,14 +3886,18 @@ top:
 	/*
 	 * Insert the new object into the directory.
 	 */
-	(void) zfs_link_create(dl, zp, tx, ZNEW);
-
-	if (flags & FIGNORECASE)
-		txtype |= TX_CI;
-	zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
+	error = zfs_link_create(dl, zp, tx, ZNEW);
+	if (error != 0) {
+		zfs_znode_delete(zp, tx);
+		remove_inode_hash(ZTOI(zp));
+	} else {
+		if (flags & FIGNORECASE)
+			txtype |= TX_CI;
+		zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
 
-	zfs_inode_update(dzp);
-	zfs_inode_update(zp);
+		zfs_inode_update(dzp);
+		zfs_inode_update(zp);
+	}
 
 	zfs_acl_ids_free(&acl_ids);
 
@@ -3871,10 +3905,14 @@ top:
 
 	zfs_dirent_unlock(dl);
 
-	*ipp = ZTOI(zp);
+	if (error == 0) {
+		*ipp = ZTOI(zp);
 
-	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
-		zil_commit(zilog, 0);
+		if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
+			zil_commit(zilog, 0);
+	} else {
+		iput(ZTOI(zp));
+	}
 
 	ZFS_EXIT(zfsvfs);
 	return (error);