MFC r328224: MFV r328220: 8677 Open-Context Channel Programs

illumos/illumos-gate@a3b2868063897ff0083dea538f55f9873eec981f

https://www.illumos.org/issues/8677
  We want to be able to run channel programs outside of synching context.
  This would greatly improve performance of channel program that just gather
  information, as we won't have to wait for synching context anymore.

  This feature should introduce the following:
  - A new command line flag in "zfs program" to specify our intention to
  run in open context.
  - A new flag/option within the channel program ioctl which selects the
  context.
  - Appropriate error handling whenever we try a channel program in
  open-context that contains zfs.sync* expressions.
  - Documentation for the new feature in the manual pages.

Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: Chris Williamson <chris.williamson@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Author: Serapheim Dimitropoulos <serapheim@delphix.com>

12 files changed, 232 insertions, 85 deletions

diff --git a/cddl/contrib/opensolaris/cmd/zfs/zfs-program.8 b/cddl/contrib/opensolaris/cmd/zfs/zfs-program.8
index 0ede564..f7587a5 100644
--- a/cddl/contrib/opensolaris/cmd/zfs/zfs-program.8
+++ b/cddl/contrib/opensolaris/cmd/zfs/zfs-program.8
@@ -18,6 +18,7 @@
 .Nd executes ZFS channel programs
 .Sh SYNOPSIS
 .Cm zfs program
+.Op Fl n
 .Op Fl t Ar instruction-limit
 .Op Fl m Ar memory-limit
 .Ar pool
@@ -45,6 +46,14 @@ will be run on
 and any attempts to access or modify other pools will cause an error.
 .Sh OPTIONS
 .Bl -tag -width "-t"
+.It Fl n
+Executes a read-only channel program, which runs faster.
+The program cannot change on-disk state by calling functions from the
+zfs.sync submodule.
+The program can be used to gather information such as properties and
+determining if changes would succeed (zfs.check.*).
+Without this flag, all pending changes must be synced to disk before a
+channel program can complete.
 .It Fl t Ar instruction-limit
 Execution time limit, in number of Lua instructions to execute.
 If a channel program executes more than the specified number of instructions,
diff --git a/cddl/contrib/opensolaris/cmd/zfs/zfs.8 b/cddl/contrib/opensolaris/cmd/zfs/zfs.8
index d52b30f..ff43798 100644
--- a/cddl/contrib/opensolaris/cmd/zfs/zfs.8
+++ b/cddl/contrib/opensolaris/cmd/zfs/zfs.8
@@ -287,6 +287,7 @@
 .Op Ar snapshot Ns | Ns Ar filesystem
 .Nm
 .Cm program
+.Op Fl n
 .Op Fl t Ar timeout
 .Op Fl m Ar memory_limit
 .Ar pool script
@@ -3294,6 +3295,7 @@ Display the path's inode change time as the first column of output.
 .It Xo
 .Nm
 .Cm program
+.Op Fl n
 .Op Fl t Ar timeout
 .Op Fl m Ar memory_limit
 .Ar pool script
@@ -3316,6 +3318,14 @@ For full documentation of the ZFS channel program interface, see the manual
 page for
 .Xr zfs-program 8 .
 .Bl -tag -width indent
+.It Fl n
+Executes a read-only channel program, which runs faster.
+The program cannot change on-disk state by calling functions from
+the zfs.sync submodule.
+The program can be used to gather information such as properties and
+determining if changes would succeed (zfs.check.*).
+Without this flag, all pending changes must be synced to disk before
+a channel program can complete.
 .It Fl t Ar timeout
 Execution time limit, in milliseconds.
 If a channel program executes for longer than the provided timeout, it will
diff --git a/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c b/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c
index 4727b9f..80e5163 100644
--- a/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c
+++ b/cddl/contrib/opensolaris/cmd/zfs/zfs_main.c
@@ -345,7 +345,7 @@ get_usage(zfs_help_t idx)
 	case HELP_BOOKMARK:
 		return (gettext("\tbookmark <snapshot> <bookmark>\n"));
 	case HELP_CHANNEL_PROGRAM:
-		return (gettext("\tprogram [-t <instruction limit>] "
+		return (gettext("\tprogram [-n] [-t <instruction limit>] "
 		    "[-m <memory limit (b)>] <pool> <program file> "
 		    "[lua args...]\n"));
 	}
@@ -7131,11 +7131,12 @@ zfs_do_channel_program(int argc, char **argv)
 	nvlist_t *outnvl;
 	uint64_t instrlimit = ZCP_DEFAULT_INSTRLIMIT;
 	uint64_t memlimit = ZCP_DEFAULT_MEMLIMIT;
+	boolean_t sync_flag = B_TRUE;
 	zpool_handle_t *zhp;
 
 	/* check options */
 	while (-1 !=
-	    (c = getopt(argc, argv, "t:(instr-limit)m:(memory-limit)"))) {
+	    (c = getopt(argc, argv, "nt:(instr-limit)m:(memory-limit)"))) {
 		switch (c) {
 		case 't':
 		case 'm': {
@@ -7173,6 +7174,10 @@ zfs_do_channel_program(int argc, char **argv)
 			}
 			break;
 		}
+		case 'n': {
+			sync_flag = B_FALSE;
+			break;
+		}
 		case '?':
 			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
 			    optopt);
@@ -7244,8 +7249,13 @@ zfs_do_channel_program(int argc, char **argv)
 	nvlist_t *argnvl = fnvlist_alloc();
 	fnvlist_add_string_array(argnvl, ZCP_ARG_CLIARGV, argv + 2, argc - 2);
 
-	ret = lzc_channel_program(poolname, progbuf, instrlimit, memlimit,
-	    argnvl, &outnvl);
+	if (sync_flag) {
+		ret = lzc_channel_program(poolname, progbuf,
+		    instrlimit, memlimit, argnvl, &outnvl);
+	} else {
+		ret = lzc_channel_program_nosync(poolname, progbuf,
+		    instrlimit, memlimit, argnvl, &outnvl);
+	}
 
 	if (ret != 0) {
 		/*
diff --git a/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c b/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c
index b6355b5..d49cb13 100644
--- a/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c
+++ b/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_dataset.c
@@ -2381,7 +2381,7 @@ zcp_check(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t intval,
 	fnvlist_add_string(argnvl, "dataset", zhp->zfs_name);
 	fnvlist_add_string(argnvl, "property", zfs_prop_to_name(prop));
 
-	error = lzc_channel_program(poolname, program,
+	error = lzc_channel_program_nosync(poolname, program,
 	    10 * 1000 * 1000, 10 * 1024 * 1024, argnvl, &outnvl);
 
 	if (error == 0) {
diff --git a/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.c b/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.c
index ec32486..6766083 100644
--- a/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.c
+++ b/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.c
@@ -918,6 +918,25 @@ lzc_destroy_bookmarks(nvlist_t *bmarks, nvlist_t **errlist)
 	return (error);
 }
 
+static int
+lzc_channel_program_impl(const char *pool, const char *program, boolean_t sync,
+    uint64_t instrlimit, uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl)
+{
+	int error;
+	nvlist_t *args;
+
+	args = fnvlist_alloc();
+	fnvlist_add_string(args, ZCP_ARG_PROGRAM, program);
+	fnvlist_add_nvlist(args, ZCP_ARG_ARGLIST, argnvl);
+	fnvlist_add_boolean_value(args, ZCP_ARG_SYNC, sync);
+	fnvlist_add_uint64(args, ZCP_ARG_INSTRLIMIT, instrlimit);
+	fnvlist_add_uint64(args, ZCP_ARG_MEMLIMIT, memlimit);
+	error = lzc_ioctl(ZFS_IOC_CHANNEL_PROGRAM, pool, args, outnvl);
+	fnvlist_free(args);
+
+	return (error);
+}
+
 /*
  * Executes a channel program.
  *
@@ -955,16 +974,26 @@ int
 lzc_channel_program(const char *pool, const char *program, uint64_t instrlimit,
     uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl)
 {
-	int error;
-	nvlist_t *args;
-
-	args = fnvlist_alloc();
-	fnvlist_add_string(args, ZCP_ARG_PROGRAM, program);
-	fnvlist_add_nvlist(args, ZCP_ARG_ARGLIST, argnvl);
-	fnvlist_add_uint64(args, ZCP_ARG_INSTRLIMIT, instrlimit);
-	fnvlist_add_uint64(args, ZCP_ARG_MEMLIMIT, memlimit);
-	error = lzc_ioctl(ZFS_IOC_CHANNEL_PROGRAM, pool, args, outnvl);
-	fnvlist_free(args);
+	return (lzc_channel_program_impl(pool, program, B_TRUE, instrlimit,
+	    memlimit, argnvl, outnvl));
+}
 
-	return (error);
+/*
+ * Executes a read-only channel program.
+ *
+ * A read-only channel program works programmatically the same way as a
+ * normal channel program executed with lzc_channel_program(). The only
+ * difference is it runs exclusively in open-context and therefore can
+ * return faster. The downside to that, is that the program cannot change
+ * on-disk state by calling functions from the zfs.sync submodule.
+ *
+ * The return values of this function (and their meaning) are exactly the
+ * same as the ones described in lzc_channel_program().
+ */
+int
+lzc_channel_program_nosync(const char *pool, const char *program,
+    uint64_t timeout, uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl)
+{
+	return (lzc_channel_program_impl(pool, program, B_FALSE, timeout,
+	    memlimit, argnvl, outnvl));
 }
diff --git a/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.h b/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.h
index 234dda4..c61ad52 100644
--- a/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.h
+++ b/cddl/contrib/opensolaris/lib/libzfs_core/common/libzfs_core.h
@@ -86,8 +86,10 @@ boolean_t lzc_exists(const char *);
 int lzc_rollback(const char *, char *, int);
 int lzc_rollback_to(const char *, const char *);
 
-int lzc_channel_program(const char *, const char *, uint64_t, uint64_t,
-    nvlist_t *, nvlist_t **);
+int lzc_channel_program(const char *, const char *, uint64_t,
+    uint64_t, nvlist_t *, nvlist_t **);
+int lzc_channel_program_nosync(const char *, const char *, uint64_t,
+    uint64_t, nvlist_t *, nvlist_t **);
 
 #ifdef	__cplusplus
 }
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_destroy.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_destroy.c
index d11d8c5..c8cf60d 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_destroy.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_destroy.c
@@ -542,6 +542,7 @@ dsl_destroy_snapshots_nvl(nvlist_t *snaps, boolean_t defer,
 	nvlist_t *result = fnvlist_alloc();
 	int error = zcp_eval(nvpair_name(nvlist_next_nvpair(snaps, NULL)),
 	    program,
+	    B_TRUE,
 	    0,
 	    zfs_lua_max_memlimit,
 	    nvlist_next_nvpair(wrapper, NULL), result);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zcp.h b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zcp.h
index 2e4ad7a..5713a74 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zcp.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zcp.h
@@ -38,20 +38,25 @@ extern uint64_t zfs_lua_max_memlimit;
 
 int zcp_argerror(lua_State *, int, const char *, ...);
 
-int zcp_eval(const char *, const char *, uint64_t, uint64_t, nvpair_t *,
-    nvlist_t *);
+int zcp_eval(const char *, const char *, boolean_t, uint64_t, uint64_t,
+    nvpair_t *, nvlist_t *);
 
 int zcp_load_list_lib(lua_State *);
 
 int zcp_load_synctask_lib(lua_State *, boolean_t);
 
 typedef void (zcp_cleanup_t)(void *);
+typedef struct zcp_cleanup_handler {
+	zcp_cleanup_t *zch_cleanup_func;
+	void *zch_cleanup_arg;
+	list_node_t zch_node;
+} zcp_cleanup_handler_t;
 
 typedef struct zcp_run_info {
 	dsl_pool_t	*zri_pool;
 
 	/*
-	 * An estimate of the total ammount of space consumed by all
+	 * An estimate of the total amount of space consumed by all
 	 * synctasks we have successfully performed so far in this
 	 * channel program. Used to generate ENOSPC errors for syncfuncs.
 	 */
@@ -89,16 +94,21 @@ typedef struct zcp_run_info {
 	boolean_t	zri_timed_out;
 
 	/*
-	 * The currently registered cleanup function, which will be called
-	 * with the stored argument if a fatal error occurs.
+	 * Boolean indicating whether or not we are running in syncing
+	 * context.
 	 */
-	zcp_cleanup_t	*zri_cleanup;
-	void		*zri_cleanup_arg;
+	boolean_t	zri_sync;
+
+	/*
+	 * List of currently registered cleanup handlers, which will be
+	 * triggered in the event of a fatal error.
+	 */
+	list_t		zri_cleanup_handlers;
 } zcp_run_info_t;
 
 zcp_run_info_t *zcp_run_info(lua_State *);
-void zcp_register_cleanup(lua_State *, zcp_cleanup_t, void *);
-void zcp_clear_cleanup(lua_State *);
+zcp_cleanup_handler_t *zcp_register_cleanup(lua_State *, zcp_cleanup_t, void *);
+void zcp_deregister_cleanup(lua_State *, zcp_cleanup_handler_t *);
 void zcp_cleanup(lua_State *);
 
 /*
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp.c
index 56ccbf1..1e76960 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp.c
@@ -137,13 +137,6 @@ typedef struct zcp_eval_arg {
 	uint64_t	ea_instrlimit;
 } zcp_eval_arg_t;
 
-/*ARGSUSED*/
-static int
-zcp_eval_check(void *arg, dmu_tx_t *tx)
-{
-	return (0);
-}
-
 /*
  * The outer-most error callback handler for use with lua_pcall(). On
  * error Lua will call this callback with a single argument that
@@ -187,41 +180,45 @@ zcp_argerror(lua_State *state, int narg, const char *msg, ...)
  *
  * If an error occurs, the cleanup function will be invoked exactly once and
  * then unreigstered.
+ *
+ * Returns the registered cleanup handler so the caller can deregister it
+ * if no error occurs.
  */
-void
+zcp_cleanup_handler_t *
 zcp_register_cleanup(lua_State *state, zcp_cleanup_t cleanfunc, void *cleanarg)
 {
 	zcp_run_info_t *ri = zcp_run_info(state);
-	/*
-	 * A cleanup function should always be explicitly removed before
-	 * installing a new one to avoid accidental clobbering.
-	 */
-	ASSERT3P(ri->zri_cleanup, ==, NULL);
 
-	ri->zri_cleanup = cleanfunc;
-	ri->zri_cleanup_arg = cleanarg;
+	zcp_cleanup_handler_t *zch = kmem_alloc(sizeof (*zch), KM_SLEEP);
+	zch->zch_cleanup_func = cleanfunc;
+	zch->zch_cleanup_arg = cleanarg;
+	list_insert_head(&ri->zri_cleanup_handlers, zch);
+
+	return (zch);
 }
 
 void
-zcp_clear_cleanup(lua_State *state)
+zcp_deregister_cleanup(lua_State *state, zcp_cleanup_handler_t *zch)
 {
 	zcp_run_info_t *ri = zcp_run_info(state);
-
-	ri->zri_cleanup = NULL;
-	ri->zri_cleanup_arg = NULL;
+	list_remove(&ri->zri_cleanup_handlers, zch);
+	kmem_free(zch, sizeof (*zch));
 }
 
 /*
- * If it exists, execute the currently set cleanup function then unregister it.
+ * Execute the currently registered cleanup handlers then free them and
+ * destroy the handler list.
  */
 void
 zcp_cleanup(lua_State *state)
 {
 	zcp_run_info_t *ri = zcp_run_info(state);
 
-	if (ri->zri_cleanup != NULL) {
-		ri->zri_cleanup(ri->zri_cleanup_arg);
-		zcp_clear_cleanup(state);
+	for (zcp_cleanup_handler_t *zch =
+	    list_remove_head(&ri->zri_cleanup_handlers); zch != NULL;
+	    zch = list_remove_head(&ri->zri_cleanup_handlers)) {
+		zch->zch_cleanup_func(zch->zch_cleanup_arg);
+		kmem_free(zch, sizeof (*zch));
 	}
 }
 
@@ -822,19 +819,12 @@ zcp_panic_cb(lua_State *state)
 }
 
 static void
-zcp_eval_sync(void *arg, dmu_tx_t *tx)
+zcp_eval_impl(dmu_tx_t *tx, boolean_t sync, zcp_eval_arg_t *evalargs)
 {
 	int err;
 	zcp_run_info_t ri;
-	zcp_eval_arg_t *evalargs = arg;
 	lua_State *state = evalargs->ea_state;
 
-	/*
-	 * Open context should have setup the stack to contain:
-	 * 1: Error handler callback
-	 * 2: Script to run (converted to a Lua function)
-	 * 3: nvlist input to function (converted to Lua table or nil)
-	 */
 	VERIFY3U(3, ==, lua_gettop(state));
 
 	/*
@@ -847,8 +837,9 @@ zcp_eval_sync(void *arg, dmu_tx_t *tx)
 	ri.zri_cred = evalargs->ea_cred;
 	ri.zri_tx = tx;
 	ri.zri_timed_out = B_FALSE;
-	ri.zri_cleanup = NULL;
-	ri.zri_cleanup_arg = NULL;
+	ri.zri_sync = sync;
+	list_create(&ri.zri_cleanup_handlers, sizeof (zcp_cleanup_handler_t),
+	    offsetof(zcp_cleanup_handler_t, zch_node));
 	ri.zri_curinstrs = 0;
 	ri.zri_maxinstrs = evalargs->ea_instrlimit;
 
@@ -885,10 +876,10 @@ zcp_eval_sync(void *arg, dmu_tx_t *tx)
 
 	/*
 	 * Remove the error handler callback from the stack. At this point,
-	 * if there is a cleanup function registered, then it was registered
-	 * but never run or removed, which should never occur.
+	 * there shouldn't be any cleanup handler registered in the handler
+	 * list (zri_cleanup_handlers), regardless of whether it ran or not.
 	 */
-	ASSERT3P(ri.zri_cleanup, ==, NULL);
+	list_destroy(&ri.zri_cleanup_handlers);
 	lua_remove(state, 1);
 
 	switch (err) {
@@ -970,9 +961,73 @@ zcp_eval_sync(void *arg, dmu_tx_t *tx)
 	}
 }
 
+static void
+zcp_pool_error(zcp_eval_arg_t *evalargs, const char *poolname)
+{
+	evalargs->ea_result = SET_ERROR(ECHRNG);
+	(void) lua_pushfstring(evalargs->ea_state, "Could not open pool: %s",
+	    poolname);
+	zcp_convert_return_values(evalargs->ea_state, evalargs->ea_outnvl,
+	    ZCP_RET_ERROR, evalargs);
+
+}
+
+static void
+zcp_eval_sync(void *arg, dmu_tx_t *tx)
+{
+	zcp_eval_arg_t *evalargs = arg;
+
+	/*
+	 * Open context should have setup the stack to contain:
+	 * 1: Error handler callback
+	 * 2: Script to run (converted to a Lua function)
+	 * 3: nvlist input to function (converted to Lua table or nil)
+	 */
+	VERIFY3U(3, ==, lua_gettop(evalargs->ea_state));
+
+	zcp_eval_impl(tx, B_TRUE, evalargs);
+}
+
+static void
+zcp_eval_open(zcp_eval_arg_t *evalargs, const char *poolname)
+{
+
+	int error;
+	dsl_pool_t *dp;
+	dmu_tx_t *tx;
+
+	/*
+	 * See comment from the same assertion in zcp_eval_sync().
+	 */
+	VERIFY3U(3, ==, lua_gettop(evalargs->ea_state));
+
+	error = dsl_pool_hold(poolname, FTAG, &dp);
+	if (error != 0) {
+		zcp_pool_error(evalargs, poolname);
+		return;
+	}
+
+	/*
+	 * As we are running in open-context, we have no transaction associated
+	 * with the channel program. At the same time, functions from the
+	 * zfs.check submodule need to be associated with a transaction as
+	 * they are basically dry-runs of their counterparts in the zfs.sync
+	 * submodule. These functions should be able to run in open-context.
+	 * Therefore we create a new transaction that we later abort once
+	 * the channel program has been evaluated.
+	 */
+	tx = dmu_tx_create_dd(dp->dp_mos_dir);
+
+	zcp_eval_impl(tx, B_FALSE, evalargs);
+
+	dmu_tx_abort(tx);
+
+	dsl_pool_rele(dp, FTAG);
+}
+
 int
-zcp_eval(const char *poolname, const char *program, uint64_t instrlimit,
-    uint64_t memlimit, nvpair_t *nvarg, nvlist_t *outnvl)
+zcp_eval(const char *poolname, const char *program, boolean_t sync,
+    uint64_t instrlimit, uint64_t memlimit, nvpair_t *nvarg, nvlist_t *outnvl)
 {
 	int err;
 	lua_State *state;
@@ -1083,9 +1138,14 @@ zcp_eval(const char *poolname, const char *program, uint64_t instrlimit,
 	evalargs.ea_outnvl = outnvl;
 	evalargs.ea_result = 0;
 
-	VERIFY0(dsl_sync_task(poolname, zcp_eval_check,
-	    zcp_eval_sync, &evalargs, 0, ZFS_SPACE_CHECK_NONE));
-
+	if (sync) {
+		err = dsl_sync_task(poolname, NULL,
+		    zcp_eval_sync, &evalargs, 0, ZFS_SPACE_CHECK_NONE);
+		if (err != 0)
+			zcp_pool_error(&evalargs, poolname);
+	} else {
+		zcp_eval_open(&evalargs, poolname);
+	}
 	lua_close(state);
 
 	return (evalargs.ea_result);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp_synctask.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp_synctask.c
index f63773b..0812d24 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp_synctask.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zcp_synctask.c
@@ -55,6 +55,10 @@ typedef struct zcp_synctask_info {
  *
  * If 'sync' is false, executes a dry run and returns the error code.
  *
+ * If we are not running in syncing context and we are not doing a dry run
+ * (meaning we are running a zfs.sync function in open-context) then we
+ * return a Lua error.
+ *
  * This function also handles common fatal error cases for channel program
  * library functions. If a fatal error occurs, err_dsname will be the dataset
  * name reported in error messages, if supplied.
@@ -70,6 +74,13 @@ zcp_sync_task(lua_State *state, dsl_checkfunc_t *checkfunc,
 	if (!sync)
 		return (err);
 
+	if (!ri->zri_sync) {
+		return (luaL_error(state, "running functions from the zfs.sync "
+		    "submodule requires passing sync=TRUE to "
+		    "lzc_channel_program() (i.e. do not specify the \"-n\" "
+		    "command line argument)"));
+	}
+
 	if (err == 0) {
 		syncfunc(arg, ri->zri_tx);
 	} else if (err == EIO) {
@@ -234,6 +245,15 @@ zcp_synctask_snapshot(lua_State *state, boolean_t sync, nvlist_t *err_details)
 	zcp_run_info_t *ri = zcp_run_info(state);
 
 	/*
+	 * On old pools, the ZIL must not be active when a snapshot is created,
+	 * but we can't suspend the ZIL because we're already in syncing
+	 * context.
+	 */
+	if (spa_version(ri->zri_pool->dp_spa) < SPA_VERSION_FAST_SNAP) {
+		return (ENOTSUP);
+	}
+
+	/*
 	 * We only allow for a single snapshot rather than a list, so the
 	 * error list output is unnecessary.
 	 */
@@ -243,33 +263,23 @@ zcp_synctask_snapshot(lua_State *state, boolean_t sync, nvlist_t *err_details)
 	ddsa.ddsa_snaps = fnvlist_alloc();
 	fnvlist_add_boolean(ddsa.ddsa_snaps, dsname);
 
-	/*
-	 * On old pools, the ZIL must not be active when a snapshot is created,
-	 * but we can't suspend the ZIL because we're already in syncing
-	 * context.
-	 */
-	if (spa_version(ri->zri_pool->dp_spa) < SPA_VERSION_FAST_SNAP) {
-		return (ENOTSUP);
-	}
+	zcp_cleanup_handler_t *zch = zcp_register_cleanup(state,
+	    (zcp_cleanup_t *)&fnvlist_free, ddsa.ddsa_snaps);
 
 	err = zcp_sync_task(state, dsl_dataset_snapshot_check,
 	    dsl_dataset_snapshot_sync, &ddsa, sync, dsname);
 
+	zcp_deregister_cleanup(state, zch);
 	fnvlist_free(ddsa.ddsa_snaps);
 
 	return (err);
 }
 
-void
-zcp_synctask_wrapper_cleanup(void *arg)
-{
-	fnvlist_free(arg);
-}
-
 static int
 zcp_synctask_wrapper(lua_State *state)
 {
 	int err;
+	zcp_cleanup_handler_t *zch;
 	int num_ret = 1;
 	nvlist_t *err_details = fnvlist_alloc();
 
@@ -277,7 +287,8 @@ zcp_synctask_wrapper(lua_State *state)
 	 * Make sure err_details is properly freed, even if a fatal error is
 	 * thrown during the synctask.
 	 */
-	zcp_register_cleanup(state, &zcp_synctask_wrapper_cleanup, err_details);
+	zch = zcp_register_cleanup(state,
+	    (zcp_cleanup_t *)&fnvlist_free, err_details);
 
 	zcp_synctask_info_t *info = lua_touserdata(state, lua_upvalueindex(1));
 	boolean_t sync = lua_toboolean(state, lua_upvalueindex(2));
@@ -317,7 +328,7 @@ zcp_synctask_wrapper(lua_State *state)
 		num_ret++;
 	}
 
-	zcp_clear_cleanup(state);
+	zcp_deregister_cleanup(state, zch);
 	fnvlist_free(err_details);
 
 	return (num_ret);
diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c
index b38709a..c54cc1c 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_ioctl.c
@@ -3777,11 +3777,15 @@ zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
 {
 	char *program;
 	uint64_t instrlimit, memlimit;
+	boolean_t sync_flag;
 	nvpair_t *nvarg = NULL;
 
 	if (0 != nvlist_lookup_string(innvl, ZCP_ARG_PROGRAM, &program)) {
 		return (EINVAL);
 	}
+	if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
+		sync_flag = B_TRUE;
+	}
 	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
 		instrlimit = ZCP_DEFAULT_INSTRLIMIT;
 	}
@@ -3797,7 +3801,7 @@ zfs_ioc_channel_program(const char *poolname, nvlist_t *innvl,
 	if (memlimit == 0 || memlimit > zfs_lua_max_memlimit)
 		return (EINVAL);
 
-	return (zcp_eval(poolname, program, instrlimit, memlimit,
+	return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
 	    nvarg, outnvl));
 }
 
diff --git a/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h b/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h
index bd5e024..19de812 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h
+++ b/sys/cddl/contrib/opensolaris/uts/common/sys/fs/zfs.h
@@ -984,6 +984,7 @@ typedef enum {
  */
 #define	ZCP_ARG_PROGRAM		"program"
 #define	ZCP_ARG_ARGLIST		"arg"
+#define	ZCP_ARG_SYNC		"sync"
 #define	ZCP_ARG_INSTRLIMIT	"instrlimit"
 #define	ZCP_ARG_MEMLIMIT	"memlimit"