Implemented zpool sync command

This addition will enable us to sync an open TXG to the main pool
on demand. The functionality is similar to 'sync(2)' but 'zpool sync'
will return when data has hit the main storage instead of potentially
just the ZIL as is the case with the 'sync(2)' cmd.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Alek Pinchuk <apinchuk@datto.com>

Author: alek-p

cmd/zpool/zpool_main.c

@@ -27,6 +27,7 @@
  * Copyright (c) 2012 by Cyril Plisko. All rights reserved.
  * Copyright (c) 2013 by Prasad Joshi (sTec). All rights reserved.
  * Copyright 2016 Nexenta Systems, Inc.
+ * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
  * Copyright (c) 2017 Datto Inc.
  */
 
@@ -111,6 +112,8 @@ static int zpool_do_events(int, char **);
 static int zpool_do_get(int, char **);
 static int zpool_do_set(int, char **);
 
+static int zpool_do_sync(int, char **);
+
 /*
  * These libumem hooks provide a reasonable set of defaults for the allocator's
  * debugging facilities.
@@ -156,6 +159,7 @@ typedef enum {
 	HELP_GET,
 	HELP_SET,
 	HELP_SPLIT,
+	HELP_SYNC,
 	HELP_REGUID,
 	HELP_REOPEN
 } zpool_help_t;
@@ -288,6 +292,7 @@ static zpool_command_t command_table[] = {
 	{ NULL },
 	{ "get",	zpool_do_get,		HELP_GET		},
 	{ "set",	zpool_do_set,		HELP_SET		},
+	{ "sync",	zpool_do_sync,		HELP_SYNC		},
 };
 
 #define	NCOMMAND	(ARRAY_SIZE(command_table))
@@ -378,6 +383,8 @@ get_usage(zpool_help_t idx) {
 		    "[<device> ...]\n"));
 	case HELP_REGUID:
 		return (gettext("\treguid <pool>\n"));
+	case HELP_SYNC:
+		return (gettext("\tsync [pool] ...\n"));
 	}
 
 	abort();
@@ -2966,6 +2973,45 @@ zpool_do_import(int argc, char **argv)
 	return (err ? 1 : 0);
 }
 
+/*
+ * zpool sync [-f] [pool] ...
+ *
+ * -f (undocumented) force uberblock (and config including zpool cache file)
+ *    update.
+ *
+ * Sync the specified pool(s).
+ * Without arguments "zpool sync" will sync all pools.
+ * This command initiates TXG sync(s) and will return after the TXG(s) commit.
+ *
+ */
+static int
+zpool_do_sync(int argc, char **argv)
+{
+	int ret;
+	boolean_t force = B_FALSE;
+
+	/* check options */
+	while ((ret  = getopt(argc, argv, "f")) != -1) {
+		switch (ret) {
+		case 'f':
+			force = B_TRUE;
+			break;
+		case '?':
+			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
+			    optopt);
+			usage(B_FALSE);
+		}
+	}
+
+	argc -= optind;
+	argv += optind;
+
+	/* if argc == 0 we will execute zpool_sync_one on all pools */
+	ret = for_each_pool(argc, argv, B_FALSE, NULL, zpool_sync_one, &force);
+
+	return (ret);
+}
+
 typedef struct iostat_cbdata {
 	uint64_t cb_flags;
 	int cb_name_flags;

include/libzfs.h

@@ -295,6 +295,8 @@ extern int zpool_clear(zpool_handle_t *, const char *, nvlist_t *);
 extern int zpool_reguid(zpool_handle_t *);
 extern int zpool_reopen(zpool_handle_t *);
 
+extern int zpool_sync_one(zpool_handle_t *, void *);
+
 extern int zpool_vdev_online(zpool_handle_t *, const char *, int,
     vdev_state_t *);
 extern int zpool_vdev_offline(zpool_handle_t *, const char *, boolean_t);

include/libzfs_core.h

@@ -114,6 +114,8 @@ int lzc_channel_program_nosync(const char *, const char *, uint64_t,
 int lzc_sync(const char *, nvlist_t *, nvlist_t **);
 int lzc_reopen(const char *, boolean_t);
 
+int lzc_sync(const char *, nvlist_t *, nvlist_t **);
+
 #ifdef	__cplusplus
 }
 #endif

include/sys/zfs_ioctl.h

@@ -591,6 +591,8 @@ typedef enum zfs_ioc {
 	ZFS_IOC_POOL_INITIALIZE,
 
 	ZFS_IOC_CHANNEL_PROGRAM,
+	ZFS_IOC_POOL_SYNC,
+
 	/*
 	 * Linux - 3/64 numbers reserved.
 	 */

lib/libzfs/libzfs_pool.c

@@ -3679,6 +3679,27 @@ zpool_reopen(zpool_handle_t *zhp)
 	return (zpool_standard_error(hdl, errno, msg));
 }
 
+/* call into libzfs_core to execute the sync IOCTL per pool */
+int
+zpool_sync_one(zpool_handle_t *zhp, void *data)
+{
+	int ret;
+	libzfs_handle_t *hdl = zpool_get_handle(zhp);
+	const char *pool_name = zpool_get_name(zhp);
+	boolean_t *force = data;
+	nvlist_t *innvl = fnvlist_alloc();
+
+	fnvlist_add_boolean_value(innvl, "force", *force);
+	if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
+		nvlist_free(innvl);
+		return (zpool_standard_error_fmt(hdl, ret,
+		    dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
+	}
+	nvlist_free(innvl);
+
+	return (0);
+}
+
 /*
  * Convert from a devid string to a path.
  */

lib/libzfs_core/libzfs_core.c

@@ -24,6 +24,7 @@
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright 2017 RackTop Systems.
+ * Copyright (c) 2017 Datto Inc.
  */
 
 /*
@@ -135,7 +136,8 @@ lzc_ioctl(zfs_ioc_t ioc, const char *name,
 	ASSERT3S(g_refcount, >, 0);
 	VERIFY3S(g_fd, !=, -1);
 
-	(void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name));
+	if (name != NULL)
+		(void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name));
 
 	if (source != NULL) {
 		packed = fnvlist_pack(source, &size);
@@ -410,6 +412,18 @@ lzc_exists(const char *dataset)
 	return (ioctl(g_fd, ZFS_IOC_OBJSET_STATS, &zc) == 0);
 }
 
+/*
+ * outnvl is unused.
+ * It was added to preserve the function signature in case it is
+ * needed in the future.
+ */
+/*ARGSUSED*/
+int
+lzc_sync(const char *pool_name, nvlist_t *innvl, nvlist_t **outnvl)
+{
+	return (lzc_ioctl(ZFS_IOC_POOL_SYNC, pool_name, innvl, NULL));
+}
+
 /*
  * Create "user holds" on snapshots.  If there is a hold on a snapshot,
  * the snapshot can not be destroyed.  (However, it can be marked for deletion
@@ -510,11 +524,7 @@ lzc_release(nvlist_t *holds, nvlist_t **errlist)
 int
 lzc_get_holds(const char *snapname, nvlist_t **holdsp)
 {
-	int error;
-	nvlist_t *innvl = fnvlist_alloc();
-	error = lzc_ioctl(ZFS_IOC_GET_HOLDS, snapname, innvl, holdsp);
-	fnvlist_free(innvl);
-	return (error);
+	return (lzc_ioctl(ZFS_IOC_GET_HOLDS, snapname, NULL, holdsp));
 }
 
 /*

man/man8/zpool.8

@@ -173,6 +173,9 @@
 .Oo Ar pool Oc Ns ...
 .Op Ar interval Op Ar count
 .Nm
+.Cm sync
+.Oo Ar pool Oc Ns ...
+.Nm
 .Cm upgrade
 .Nm
 .Cm upgrade
@@ -1803,6 +1806,18 @@ Warnings about pools not using the latest on-disk format will not be included.
 .El
 .It Xo
 .Nm
+.Cm sync
+.Oo Ar pool Oc Ns ...
+.Xc
+This command forces all in-core dirty data to be written to the primary pool
+storage and not the ZIL.
+It will also update administrative information including quota reporting.
+Without arguments,
+.Sx zpool sync
+will sync all pools on the system.
+Otherwise, it will sync only the specified pool(s).
+.It Xo
+.Nm
 .Cm upgrade
 .Xc
 Displays pools which do not have all supported features enabled and pools

module/zfs/zfs_ioctl.c

@@ -36,7 +36,8 @@
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright 2016 Toomas Soome <tsoome@me.com>
  * Copyright 2017 RackTop Systems.
- * Copyright (c) 2017 Datto Inc.
+ * Copyright (c) 2016 Actifio, Inc. All rights reserved.
+ * Copyright (c) 2017, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
  */
 
 #define __APPLE_API_PRIVATE
@@ -5315,6 +5316,7 @@ zfs_ioc_hold(const char *pool, nvlist_t *args, nvlist_t *errlist)
 static int
 zfs_ioc_get_holds(const char *snapname, nvlist_t *args, nvlist_t *outnvl)
 {
+	ASSERT3P(args, ==, NULL);
 	return (dsl_dataset_get_holds(snapname, outnvl));
 }
 
@@ -5843,6 +5845,43 @@ zfs_ioc_osx_proxy_dataset(zfs_cmd_t *zc)
 	return (error);
 }
 
+/*
+ * Sync the currently open TXG to disk for the specified pool.
+ * This is somewhat similar to 'zfs_sync()'.
+ * For cases that do not result in error this ioctl will wait for
+ * the currently open TXG to commit before returning back to the caller.
+ *
+ * innvl: {
+ *  "force" -> when true, force uberblock update even if there is no dirty data.
+ *             In addition this will cause the vdev configuration to be written
+ *             out including updating the zpool cache file. (boolean_t)
+ * }
+ *
+ * onvl is unused
+ */
+/* ARGSUSED */
+static int
+zfs_ioc_pool_sync(const char *pool, nvlist_t *innvl, nvlist_t *onvl)
+{
+	int err;
+	boolean_t force;
+	spa_t *spa;
+
+	if ((err = spa_open(pool, &spa, FTAG)) != 0)
+		return (err);
+
+	force = fnvlist_lookup_boolean_value(innvl, "force");
+	if (force) {
+		spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
+		vdev_config_dirty(spa->spa_root_vdev);
+		spa_config_exit(spa, SCL_CONFIG, FTAG);
+	}
+	txg_wait_synced(spa_get_dsl(spa), 0);
+
+	spa_close(spa, FTAG);
+
+	return (err);
+}
 
 static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];
 
@@ -6045,6 +6084,10 @@ zfs_ioctl_init(void)
 	    POOL_NAME, POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_TRUE,
 	    B_TRUE);
 
+	zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
+	    zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
+	    POOL_CHECK_SUSPENDED | POOL_CHECK_READONLY, B_FALSE, B_FALSE);
+
 	/* IOCTLS that use the legacy function signature */
 
 	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,

tests/runfiles/linux.run

@@ -391,6 +391,9 @@ tests = ['zpool_upgrade_001_pos',
     'zpool_upgrade_006_neg', 'zpool_upgrade_008_pos',
     'zpool_upgrade_009_neg']
 
+[tests/functional/cli_root/zpool_sync]
+tests = ['zpool_sync_001_pos', 'zpool_sync_002_neg']
+
 # DISABLED:
 # zfs_share_001_neg - requires additional dependencies
 # zfs_unshare_001_neg - requires additional dependencies

tests/zfs-tests/include/libtest.shlib

@@ -24,6 +24,7 @@
 # Use is subject to license terms.
 # Copyright (c) 2012, 2017 by Delphix. All rights reserved.
 # Copyright 2016 Nexenta Systems, Inc.
+# Copyright (c) 2017 Lawrence Livermore National Security, LLC.
 # Copyright (c) 2017 Datto Inc.
 #
 
@@ -3274,6 +3275,46 @@ function format_zvol # volume_name blockdev mountpoint
 	fi
 }
 
+#
+# Sync data to the pool
+#
+# $1 pool name
+# $2 boolean to force uberblock (and config including zpool cache file) update
+#
+function sync_pool #pool <force>
+{
+	typeset pool=${1:-$TESTPOOL}
+	typeset force=${2:-false}
+
+	if [[ $force == true ]]; then
+		log_must zpool sync -f $pool
+	else
+		log_must zpool sync $pool
+	fi
+
+	return 0
+}
+
+#
+# Sync data to the pool
+#
+# $1 pool name
+# $2 boolean to force uberblock (and config including zpool cache file) update
+#
+function sync_pool #pool <force>
+{
+	typeset pool=${1:-$TESTPOOL}
+	typeset force=${2:-false}
+
+	if [[ $force == true ]]; then
+		log_must zpool sync -f $pool
+	else
+		log_must zpool sync $pool
+	fi
+
+	return 0
+}
+
 function unmount_fs_mountpoint # mountpoint
 {
 	typeset mountpoint=$1