OpenZFS 8677 - Open-Context Channel Programs

Authored by: Serapheim Dimitropoulos <[email protected]>
Reviewed by: Matt Ahrens <[email protected]>
Reviewed by: Chris Williamson <[email protected]>
Reviewed by: Pavel Zakharov <[email protected]>
Approved by: Robert Mustacchi <[email protected]>
Ported-by: Don Brady <[email protected]>

We want to be able to run channel programs outside of synching
context. This would greatly improve performance for channel programs
that just gather information, as they won't have to wait for synching
context anymore.

=== What is implemented?

This feature introduces the following:
- A new command line flag in "zfs program" to specify our intention
  to run in open context. (The -n option)
- 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.

=== How do we handle zfs.sync functions in open context?

When such a function is found by the interpreter and we are running
in open context we abort the script and we spit out a descriptive
runtime error. For example, given the script below ...

arg = ...
fs = arg["argv"][1]
err = zfs.sync.destroy(fs)
msg = "destroying " .. fs .. " err=" .. err
return msg

if we run it in open context, we will get back the following error:

Channel program execution failed:
[string "channel program"]:3: 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)
stack traceback:
            [C]: in function 'destroy'
            [string "channel program"]:3: in main chunk

=== What about testing?

We've introduced new wrappers for all channel program tests that
run each channel program as both (startard & open-context) and
expect the appropriate behavior depending on the program using
the zfs.sync module.

OpenZFS-issue: https://www.illumos.org/issues/8677
OpenZFS-commit: openzfs/openzfs@17a49e15
Closes openzfs#6558

cmd/zfs/zfs_main.c

@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2017 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.
@@ -340,7 +340,7 @@ 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"));
 	case HELP_LOAD_KEY:
@@ -7098,10 +7098,11 @@ 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 ((c = getopt(argc, argv, "t:m:")) != -1) {
+	while ((c = getopt(argc, argv, "nt:m:")) != -1) {
 		switch (c) {
 		case 't':
 		case 'm': {
@@ -7139,6 +7140,10 @@ 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);
@@ -7210,8 +7215,13 @@ 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) {
 		/*

include/libzfs_core.h

@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
  * Copyright (c) 2017 Datto Inc.
  * Copyright 2017 RackTop Systems.
  * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
@@ -101,8 +101,10 @@ 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 **);
 
 int lzc_sync(const char *, nvlist_t *, nvlist_t **);
 int lzc_reopen(const char *, boolean_t);

include/sys/fs/zfs.h

@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2013, 2017 Joyent, Inc. All rights reserved.
  * Copyright (c) 2014 Integros [integros.com]
@@ -1204,6 +1204,7 @@ 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"
 

include/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 *);
 
 /*

lib/libzfs/libzfs_dataset.c

@@ -22,7 +22,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2013, Joyent, Inc. All rights reserved.
- * Copyright (c) 2011, 2016 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
  * Copyright (c) 2012 DEY Storage Systems, Inc.  All rights reserved.
  * Copyright (c) 2012 Pawel Jakub Dawidek <[email protected]>.
  * Copyright (c) 2013 Martin Matuska. All rights reserved.
@@ -2506,7 +2506,7 @@ 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) {

lib/libzfs_core/libzfs_core.c

@@ -1063,6 +1063,25 @@ 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.
  *
@@ -1100,18 +1119,28 @@ 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));
 }
 
 /*

man/man8/zfs-program.8

@@ -17,7 +17,8 @@
 .Nm zfs program
 .Nd executes ZFS channel programs
 .Sh SYNOPSIS
-.Cm zfs program
+.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,

man/man8/zfs.8

@@ -21,7 +21,7 @@
 .\"
 .\" Copyright (c) 2009 Sun Microsystems, Inc. All Rights Reserved.
 .\" Copyright 2011 Joshua M. Clulow <[email protected]>
-.\" Copyright (c) 2011, 2016 by Delphix. All rights reserved.
+.\" Copyright (c) 2011, 2017 by Delphix. All rights reserved.
 .\" Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
 .\" Copyright (c) 2014, Joyent, Inc. All rights reserved.
 .\" Copyright (c) 2014 by Adam Stevko. All rights reserved.
@@ -272,6 +272,7 @@
 .Ar snapshot 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
@@ -4042,6 +4043,7 @@ 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
@@ -4063,6 +4065,14 @@ For full documentation of the ZFS channel program interface, see the manual
 page for
 .Xr zfs-program 8 .
 .Bl -tag -width ""
+.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

module/zfs/dsl_destroy.c

@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  * Copyright (c) 2013 by Joyent, Inc. All rights reserved.
  * Copyright (c) 2016 Actifio, Inc. All rights reserved.
@@ -550,6 +550,7 @@ 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);