3740 Poor ZFS send / receive performance due to snapshot hold / relea…

…se processing

Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Christopher Siden <christopher.siden@delphix.com>

usr/src/cmd/ndmpd/ndmp/ndmpd_chkpnt.c

@@ -1,6 +1,7 @@
 /*
  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
 /*
@@ -203,7 +204,7 @@ snapshot_hold(char *volname, char *snapname, char *jname, boolean_t recursive)
 	}
 
 	p = strchr(snapname, '@') + 1;
-	if (zfs_hold(zhp, p, jname, recursive, B_FALSE, cleanup_fd) != 0) {
+	if (zfs_hold(zhp, p, jname, recursive, cleanup_fd) != 0) {
 		NDMP_LOG(LOG_ERR, "Cannot hold snapshot %s", p);
 		zfs_close(zhp);
 		return (-1);

usr/src/cmd/zfs/zfs_main.c

@@ -25,6 +25,7 @@
  * Copyright (c) 2012 by Delphix. All rights reserved.
  * Copyright 2012 Milan Jurik. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2013 Steven Hartland.  All rights reserved.
  */
 
 #include <assert.h>
@@ -5164,8 +5165,7 @@ zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
 			continue;
 		}
 		if (holding) {
-			if (zfs_hold(zhp, delim+1, tag, recursive,
-			    B_FALSE, -1) != 0)
+			if (zfs_hold(zhp, delim+1, tag, recursive, -1) != 0)
 				++errors;
 		} else {
 			if (zfs_release(zhp, delim+1, tag, recursive) != 0)

usr/src/cmd/zhack/zhack.c

@@ -21,6 +21,7 @@
 
 /*
  * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
 /*
@@ -153,7 +154,7 @@ import_pool(const char *target, boolean_t readonly)
 	g_importargs.poolname = g_pool;
 	pools = zpool_search_import(g_zfs, &g_importargs);
 
-	if (pools == NULL || nvlist_next_nvpair(pools, NULL) == NULL) {
+	if (nvlist_empty(pools)) {
 		if (!g_importargs.can_be_active) {
 			g_importargs.can_be_active = B_TRUE;
 			if (zpool_search_import(g_zfs, &g_importargs) != NULL ||

usr/src/cmd/ztest/ztest.c

@@ -22,6 +22,7 @@
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2012 by Delphix. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
+ * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
 /*
@@ -4711,7 +4712,7 @@ ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id)
 
 	error = user_release_one(fullname, tag);
 	if (error)
-		fatal(0, "user_release_one(%s)", fullname, tag);
+		fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error);
 
 	VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT);
 

usr/src/lib/libzfs/common/libzfs.h

@@ -24,6 +24,7 @@
  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2012 by Delphix. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
 #ifndef	_LIBZFS_H
@@ -594,7 +595,8 @@ extern int zfs_send(zfs_handle_t *, const char *, const char *,
 
 extern int zfs_promote(zfs_handle_t *);
 extern int zfs_hold(zfs_handle_t *, const char *, const char *,
-    boolean_t, boolean_t, int);
+    boolean_t, int);
+extern int zfs_hold_nvl(zfs_handle_t *, int, nvlist_t *);
 extern int zfs_release(zfs_handle_t *, const char *, const char *, boolean_t);
 extern int zfs_get_holds(zfs_handle_t *, nvlist_t **);
 extern uint64_t zvol_volsize_to_reservation(uint64_t, nvlist_t *);

usr/src/lib/libzfs/common/libzfs_dataset.c

@@ -25,6 +25,7 @@
  * Copyright (c) 2012 DEY Storage Systems, Inc.  All rights reserved.
  * Copyright 2012 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2013 Martin Matuska. All rights reserved.
+ * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
 #include <ctype.h>
@@ -3104,18 +3105,14 @@ static int
 zfs_check_snap_cb(zfs_handle_t *zhp, void *arg)
 {
 	struct destroydata *dd = arg;
-	zfs_handle_t *szhp;
 	char name[ZFS_MAXNAMELEN];
 	int rv = 0;
 
 	(void) snprintf(name, sizeof (name),
 	    "%s@%s", zhp->zfs_name, dd->snapname);
 
-	szhp = make_dataset_handle(zhp->zfs_hdl, name);
-	if (szhp) {
+	if (lzc_exists(name))
 		verify(nvlist_add_boolean(dd->nvl, name) == 0);
-		zfs_close(szhp);
-	}
 
 	rv = zfs_iter_filesystems(zhp, zfs_check_snap_cb, dd);
 	zfs_close(zhp);
@@ -3135,7 +3132,7 @@ zfs_destroy_snaps(zfs_handle_t *zhp, char *snapname, boolean_t defer)
 	verify(nvlist_alloc(&dd.nvl, NV_UNIQUE_NAME, 0) == 0);
 	(void) zfs_check_snap_cb(zfs_handle_dup(zhp), &dd);
 
-	if (nvlist_next_nvpair(dd.nvl, NULL) == NULL) {
+	if (nvlist_empty(dd.nvl)) {
 		ret = zfs_standard_error_fmt(zhp->zfs_hdl, ENOENT,
 		    dgettext(TEXT_DOMAIN, "cannot destroy '%s@%s'"),
 		    zhp->zfs_name, snapname);
@@ -3160,7 +3157,7 @@ zfs_destroy_snaps_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, boolean_t defer)
 	if (ret == 0)
 		return (0);
 
-	if (nvlist_next_nvpair(errlist, NULL) == NULL) {
+	if (nvlist_empty(errlist)) {
 		char errbuf[1024];
 		(void) snprintf(errbuf, sizeof (errbuf),
 		    dgettext(TEXT_DOMAIN, "cannot destroy snapshots"));
@@ -4082,18 +4079,14 @@ static int
 zfs_hold_one(zfs_handle_t *zhp, void *arg)
 {
 	struct holdarg *ha = arg;
-	zfs_handle_t *szhp;
 	char name[ZFS_MAXNAMELEN];
 	int rv = 0;
 
 	(void) snprintf(name, sizeof (name),
 	    "%s@%s", zhp->zfs_name, ha->snapname);
 
-	szhp = make_dataset_handle(zhp->zfs_hdl, name);
-	if (szhp) {
+	if (lzc_exists(name))
 		fnvlist_add_string(ha->nvl, name, ha->tag);
-		zfs_close(szhp);
-	}
 
 	if (ha->recursive)
 		rv = zfs_iter_filesystems(zhp, zfs_hold_one, ha);
@@ -4103,41 +4096,55 @@ zfs_hold_one(zfs_handle_t *zhp, void *arg)
 
 int
 zfs_hold(zfs_handle_t *zhp, const char *snapname, const char *tag,
-    boolean_t recursive, boolean_t enoent_ok, int cleanup_fd)
+    boolean_t recursive, int cleanup_fd)
 {
 	int ret;
 	struct holdarg ha;
-	nvlist_t *errors;
-	libzfs_handle_t *hdl = zhp->zfs_hdl;
-	char errbuf[1024];
-	nvpair_t *elem;
 
 	ha.nvl = fnvlist_alloc();
 	ha.snapname = snapname;
 	ha.tag = tag;
 	ha.recursive = recursive;
 	(void) zfs_hold_one(zfs_handle_dup(zhp), &ha);
 
-	if (nvlist_next_nvpair(ha.nvl, NULL) == NULL) {
+	if (nvlist_empty(ha.nvl)) {
+		char errbuf[1024];
+
 		fnvlist_free(ha.nvl);
 		ret = ENOENT;
-		if (!enoent_ok) {
-			(void) snprintf(errbuf, sizeof (errbuf),
-			    dgettext(TEXT_DOMAIN,
-			    "cannot hold snapshot '%s@%s'"),
-			    zhp->zfs_name, snapname);
-			(void) zfs_standard_error(hdl, ret, errbuf);
-		}
+		(void) snprintf(errbuf, sizeof (errbuf),
+		    dgettext(TEXT_DOMAIN,
+		    "cannot hold snapshot '%s@%s'"),
+		    zhp->zfs_name, snapname);
+		(void) zfs_standard_error(zhp->zfs_hdl, ret, errbuf);
 		return (ret);
 	}
 
-	ret = lzc_hold(ha.nvl, cleanup_fd, &errors);
+	ret = zfs_hold_nvl(zhp, cleanup_fd, ha.nvl);
 	fnvlist_free(ha.nvl);
 
-	if (ret == 0)
+	return (ret);
+}
+
+int
+zfs_hold_nvl(zfs_handle_t *zhp, int cleanup_fd, nvlist_t *holds)
+{
+	int ret;
+	nvlist_t *errors;
+	libzfs_handle_t *hdl = zhp->zfs_hdl;
+	char errbuf[1024];
+	nvpair_t *elem;
+
+	errors = NULL;
+	ret = lzc_hold(holds, cleanup_fd, &errors);
+
+	if (ret == 0) {
+		/* There may be errors even in the success case. */
+		fnvlist_free(errors);
 		return (0);
+	}
 
-	if (nvlist_next_nvpair(errors, NULL) == NULL) {
+	if (nvlist_empty(errors)) {
 		/* no hold-specific errors */
 		(void) snprintf(errbuf, sizeof (errbuf),
 		    dgettext(TEXT_DOMAIN, "cannot hold"));
@@ -4177,10 +4184,6 @@ zfs_hold(zfs_handle_t *zhp, const char *snapname, const char *tag,
 		case EEXIST:
 			(void) zfs_error(hdl, EZFS_REFTAG_HOLD, errbuf);
 			break;
-		case ENOENT:
-			if (enoent_ok)
-				return (ENOENT);
-			/* FALLTHROUGH */
 		default:
 			(void) zfs_standard_error(hdl,
 			    fnvpair_value_int32(elem), errbuf);
@@ -4191,30 +4194,21 @@ zfs_hold(zfs_handle_t *zhp, const char *snapname, const char *tag,
 	return (ret);
 }
 
-struct releasearg {
-	nvlist_t *nvl;
-	const char *snapname;
-	const char *tag;
-	boolean_t recursive;
-};
-
 static int
 zfs_release_one(zfs_handle_t *zhp, void *arg)
 {
 	struct holdarg *ha = arg;
-	zfs_handle_t *szhp;
 	char name[ZFS_MAXNAMELEN];
 	int rv = 0;
 
 	(void) snprintf(name, sizeof (name),
 	    "%s@%s", zhp->zfs_name, ha->snapname);
 
-	szhp = make_dataset_handle(zhp->zfs_hdl, name);
-	if (szhp) {
+	if (lzc_exists(name)) {
 		nvlist_t *holds = fnvlist_alloc();
 		fnvlist_add_boolean(holds, ha->tag);
 		fnvlist_add_nvlist(ha->nvl, name, holds);
-		zfs_close(szhp);
+		fnvlist_free(holds);
 	}
 
 	if (ha->recursive)
@@ -4229,7 +4223,7 @@ zfs_release(zfs_handle_t *zhp, const char *snapname, const char *tag,
 {
 	int ret;
 	struct holdarg ha;
-	nvlist_t *errors;
+	nvlist_t *errors = NULL;
 	nvpair_t *elem;
 	libzfs_handle_t *hdl = zhp->zfs_hdl;
 	char errbuf[1024];
@@ -4240,7 +4234,7 @@ zfs_release(zfs_handle_t *zhp, const char *snapname, const char *tag,
 	ha.recursive = recursive;
 	(void) zfs_release_one(zfs_handle_dup(zhp), &ha);
 
-	if (nvlist_next_nvpair(ha.nvl, NULL) == NULL) {
+	if (nvlist_empty(ha.nvl)) {
 		fnvlist_free(ha.nvl);
 		ret = ENOENT;
 		(void) snprintf(errbuf, sizeof (errbuf),
@@ -4254,10 +4248,13 @@ zfs_release(zfs_handle_t *zhp, const char *snapname, const char *tag,
 	ret = lzc_release(ha.nvl, &errors);
 	fnvlist_free(ha.nvl);
 
-	if (ret == 0)
+	if (ret == 0) {
+		/* There may be errors even in the success case. */
+		fnvlist_free(errors);
 		return (0);
+	}
 
-	if (nvlist_next_nvpair(errors, NULL) == NULL) {
+	if (nvlist_empty(errors)) {
 		/* no hold-specific errors */
 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
 		    "cannot release"));