MFV r354377: 10554 Implemented zpool sync command

illumos/illumos-gate@9c2acf00e2
9c2acf00e2

https://www.illumos.org/issues/10554
  During the port of MMP (illumos bug 10499) from ZoL, I found this
  earlier ZoL project is a prerequisite. Here is the original
  description.  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.

Portions contributed by: Jerry Jelinek <jerry.jelinek@joyent.com>
Author: Alek Pinchuk <apinchuk@datto.com>
MFC after:	3 weeks
Relnotes:	possibly

cddl/contrib/opensolaris/cmd/zpool/zpool.8

@@ -193,6 +193,9 @@
 .Ar ...
 .Op Ar interval Op Ar count
 .Nm
+.Cm sync
+.Oo Ar pool Oc Ns ...
+.Nm
 .Cm upgrade
 .Op Fl v
 .Nm
@@ -1915,6 +1918,19 @@ block size or disabled features will not be included.
 .El
 .It Xo
 .Nm
+.Cm sync
+.Oo Ar pool Oc Ns ...
+.Xc
+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,
+.Nm zpool Cm sync
+will sync all pools on the system.
+Otherwise, it will only sync the specified
+.Ar pool .
+.It Xo
+.Nm
 .Cm upgrade
 .Op Fl v
 .Xc

cddl/contrib/opensolaris/cmd/zpool/zpool_main.c

@@ -100,6 +100,8 @@ static int zpool_do_history(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.
@@ -144,6 +146,7 @@ typedef enum {
 	HELP_GET,
 	HELP_SET,
 	HELP_SPLIT,
+	HELP_SYNC,
 	HELP_REGUID,
 	HELP_REOPEN
 } zpool_help_t;
@@ -200,6 +203,7 @@ static zpool_command_t command_table[] = {
 	{ "history",	zpool_do_history,	HELP_HISTORY		},
 	{ "get",	zpool_do_get,		HELP_GET		},
 	{ "set",	zpool_do_set,		HELP_SET		},
+	{ "sync",	zpool_do_sync,		HELP_SYNC		},
 };
 
 #define	NCOMMAND	(sizeof (command_table) / sizeof (command_table[0]))
@@ -285,6 +289,8 @@ 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();
@@ -2648,6 +2654,45 @@ 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 {
 	boolean_t cb_verbose;
 	int cb_namewidth;

cddl/contrib/opensolaris/lib/libzfs/common/libzfs.h

@@ -271,6 +271,8 @@ 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);

cddl/contrib/opensolaris/lib/libzfs/common/libzfs_pool.c

@@ -3515,6 +3515,27 @@ 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.
  */

cddl/contrib/opensolaris/lib/libzfs_core/common/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.
  */
 
 /*
@@ -137,16 +138,17 @@ lzc_ioctl(zfs_ioc_t ioc, const char *name,
 {
 	zfs_cmd_t zc = { 0 };
 	int error = 0;
-	char *packed;
+	char *packed = NULL;
 #ifdef __FreeBSD__
 	nvlist_t *oldsource;
 #endif
-	size_t size;
+	size_t size = 0;
 
 	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));
 
 #ifdef __FreeBSD__
 	if (zfs_ioctl_version == ZFS_IOCVER_UNDEF)
@@ -160,9 +162,11 @@ lzc_ioctl(zfs_ioc_t ioc, const char *name,
 	}
 #endif
 
-	packed = fnvlist_pack(source, &size);
-	zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed;
-	zc.zc_nvlist_src_size = size;
+	if (source != NULL) {
+		packed = fnvlist_pack(source, &size);
+		zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed;
+		zc.zc_nvlist_src_size = size;
+	}
 
 	if (resultp != NULL) {
 		*resultp = NULL;
@@ -462,6 +466,18 @@ 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
@@ -562,11 +578,7 @@ 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));
 }
 
 /*

cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.h

@@ -23,6 +23,7 @@
  * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
  * Copyright (c) 2013 by Martin Matuska <mm@FreeBSD.org>. All rights reserved.
  * Copyright 2017 RackTop Systems.
+ * Copyright (c) 2017 Datto Inc.
  */
 
 #ifndef	_LIBZFS_CORE_H
@@ -91,6 +92,8 @@ boolean_t lzc_exists(const char *);
 int lzc_rollback(const char *, char *, int);
 int lzc_rollback_to(const char *, const char *);
 
+int lzc_sync(const char *, nvlist_t *, nvlist_t **);
+
 int lzc_rename(const char *, const char *);
 int lzc_destroy(const char *);
 

sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c

@@ -5836,6 +5836,7 @@ 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));
 }
 
@@ -6134,6 +6135,44 @@ zfs_ioc_send_space(const char *snapname, nvlist_t *innvl, nvlist_t *outnvl)
 	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];
 
 static void
@@ -6324,6 +6363,10 @@ zfs_ioctl_init(void)
 	    zfs_ioc_pool_initialize, zfs_secpolicy_config, 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,

sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h

@@ -1016,6 +1016,7 @@ typedef enum zfs_ioc {
 	ZFS_IOC_POOL_CHECKPOINT,
 	ZFS_IOC_POOL_DISCARD_CHECKPOINT,
 	ZFS_IOC_POOL_INITIALIZE,
+	ZFS_IOC_POOL_SYNC,
 	ZFS_IOC_LAST
 } zfs_ioc_t;