OpenZFS 7431 - ZFS Channel Programs

Authored by: Chris Williamson <[email protected]>
Reviewed by: Matthew Ahrens <[email protected]>
Reviewed by: George Wilson <[email protected]>
Reviewed by: John Kennedy <[email protected]>
Reviewed by: Dan Kimmel <[email protected]>
Approved by: Garrett D'Amore <[email protected]>
Ported-by: Don Brady <[email protected]>
Ported-by: John Kennedy <[email protected]>

OpenZFS-issue: https://www.illumos.org/issues/7431
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/dfc11533

Porting Notes:
* The CLI long option arguments for '-t' and '-m' don't parse on linux
* Switched from kmem_alloc to vmem_alloc in zcp_lua_alloc
* Lua implementation is built as its own module (zlua.ko)
* Lua headers consumed directly by zfs code moved to 'include/sys/lua/'
* There is no native setjmp/longjump available in stock Linux kernel.
  Brought over implementations from illumos and FreeBSD
* The get_temporary_prop() was adapted due to VFS platform differences
* Use of inline functions in lua parser to reduce stack usage per C call
* Skip some ZFS Test Suite ZCP tests on sparc64 to avoid stack overflow

5 files changed, 197 insertions, 11 deletions

diff --git a/lib/libzfs/libzfs_dataset.c b/lib/libzfs/libzfs_dataset.c
index 1a4bad18c..b9ba719a0 100644
--- a/lib/libzfs/libzfs_dataset.c
+++ b/lib/libzfs/libzfs_dataset.c
@@ -2480,6 +2480,73 @@ zfs_get_clones_nvl(zfs_handle_t *zhp)
 }
 
 /*
+ * Accepts a property and value and checks that the value
+ * matches the one found by the channel program. If they are
+ * not equal, print both of them.
+ */
+static void
+zcp_check(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t intval,
+    const char *strval)
+{
+	if (!zhp->zfs_hdl->libzfs_prop_debug)
+		return;
+	int error;
+	char *poolname = zhp->zpool_hdl->zpool_name;
+	const char *prop_name = zfs_prop_to_name(prop);
+	const char *program =
+	    "args = ...\n"
+	    "ds = args['dataset']\n"
+	    "prop = args['property']\n"
+	    "value, setpoint = zfs.get_prop(ds, prop)\n"
+	    "return {value=value, setpoint=setpoint}\n";
+	nvlist_t *outnvl;
+	nvlist_t *retnvl;
+	nvlist_t *argnvl = fnvlist_alloc();
+
+	fnvlist_add_string(argnvl, "dataset", zhp->zfs_name);
+	fnvlist_add_string(argnvl, "property", zfs_prop_to_name(prop));
+
+	error = lzc_channel_program(poolname, program,
+	    10 * 1000 * 1000, 10 * 1024 * 1024, argnvl, &outnvl);
+
+	if (error == 0) {
+		retnvl = fnvlist_lookup_nvlist(outnvl, "return");
+		if (zfs_prop_get_type(prop) == PROP_TYPE_NUMBER) {
+			int64_t ans;
+			error = nvlist_lookup_int64(retnvl, "value", &ans);
+			if (error != 0) {
+				(void) fprintf(stderr, "%s: zcp check error: "
+				    "%u\n", prop_name, error);
+				return;
+			}
+			if (ans != intval) {
+				(void) fprintf(stderr, "%s: zfs found %llu, "
+				    "but zcp found %llu\n", prop_name,
+				    (u_longlong_t)intval, (u_longlong_t)ans);
+			}
+		} else {
+			char *str_ans;
+			error = nvlist_lookup_string(retnvl, "value", &str_ans);
+			if (error != 0) {
+				(void) fprintf(stderr, "%s: zcp check error: "
+				    "%u\n", prop_name, error);
+				return;
+			}
+			if (strcmp(strval, str_ans) != 0) {
+				(void) fprintf(stderr,
+				    "%s: zfs found '%s', but zcp found '%s'\n",
+				    prop_name, strval, str_ans);
+			}
+		}
+	} else {
+		(void) fprintf(stderr, "%s: zcp check failed, channel program "
+		    "error: %u\n", prop_name, error);
+	}
+	nvlist_free(argnvl);
+	nvlist_free(outnvl);
+}
+
+/*
  * Retrieve a property from the given object.  If 'literal' is specified, then
  * numbers are left as exact values.  Otherwise, numbers are converted to a
  * human-readable form.
@@ -2526,6 +2593,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 				(void) snprintf(propbuf, proplen, "%llu",
 				    (u_longlong_t)val);
 		}
+		zcp_check(zhp, prop, val, NULL);
 		break;
 
 	case ZFS_PROP_MOUNTPOINT:
@@ -2594,7 +2662,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 			/* 'legacy' or 'none' */
 			(void) strlcpy(propbuf, str, proplen);
 		}
-
+		zcp_check(zhp, prop, 0, propbuf);
 		break;
 
 	case ZFS_PROP_ORIGIN:
@@ -2602,6 +2670,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 		if (str == NULL)
 			return (-1);
 		(void) strlcpy(propbuf, str, proplen);
+		zcp_check(zhp, prop, 0, str);
 		break;
 
 	case ZFS_PROP_CLONES:
@@ -2616,7 +2685,6 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 
 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
 			return (-1);
-
 		/*
 		 * If quota or reservation is 0, we translate this into 'none'
 		 * (unless literal is set), and indicate that it's the default
@@ -2635,6 +2703,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 			else
 				zfs_nicebytes(val, propbuf, proplen);
 		}
+		zcp_check(zhp, prop, val, NULL);
 		break;
 
 	case ZFS_PROP_FILESYSTEM_LIMIT:
@@ -2659,6 +2728,8 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 		} else {
 			zfs_nicenum(val, propbuf, proplen);
 		}
+
+		zcp_check(zhp, prop, val, NULL);
 		break;
 
 	case ZFS_PROP_REFRATIO:
@@ -2673,6 +2744,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 			(void) snprintf(propbuf, proplen, "%llu.%02llux",
 			    (u_longlong_t)(val / 100),
 			    (u_longlong_t)(val % 100));
+		zcp_check(zhp, prop, val, NULL);
 		break;
 
 	case ZFS_PROP_TYPE:
@@ -2693,6 +2765,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 			abort();
 		}
 		(void) snprintf(propbuf, proplen, "%s", str);
+		zcp_check(zhp, prop, 0, propbuf);
 		break;
 
 	case ZFS_PROP_MOUNTED:
@@ -2718,6 +2791,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 		 * consumers.
 		 */
 		(void) strlcpy(propbuf, zhp->zfs_name, proplen);
+		zcp_check(zhp, prop, 0, propbuf);
 		break;
 
 	case ZFS_PROP_MLSLABEL:
@@ -2773,6 +2847,7 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
 			return (-1);
 		(void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val);
+		zcp_check(zhp, prop, val, NULL);
 		break;
 
 	case ZFS_PROP_REFERENCED:
@@ -2784,31 +2859,39 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 	case ZFS_PROP_USEDCHILD:
 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
 			return (-1);
-		if (literal)
+		if (literal) {
 			(void) snprintf(propbuf, proplen, "%llu",
 			    (u_longlong_t)val);
-		else
+		} else {
 			zfs_nicebytes(val, propbuf, proplen);
+		}
+		zcp_check(zhp, prop, val, NULL);
 		break;
 
 	default:
 		switch (zfs_prop_get_type(prop)) {
 		case PROP_TYPE_NUMBER:
 			if (get_numeric_property(zhp, prop, src,
-			    &source, &val) != 0)
+			    &source, &val) != 0) {
 				return (-1);
-			if (literal)
+			}
+
+			if (literal) {
 				(void) snprintf(propbuf, proplen, "%llu",
 				    (u_longlong_t)val);
-			else
+			} else {
 				zfs_nicenum(val, propbuf, proplen);
+			}
+			zcp_check(zhp, prop, val, NULL);
 			break;
 
 		case PROP_TYPE_STRING:
 			str = getprop_string(zhp, prop, &source);
 			if (str == NULL)
 				return (-1);
+
 			(void) strlcpy(propbuf, str, proplen);
+			zcp_check(zhp, prop, 0, str);
 			break;
 
 		case PROP_TYPE_INDEX:
@@ -2817,7 +2900,9 @@ zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
 				return (-1);
 			if (zfs_prop_index_to_string(prop, val, &strval) != 0)
 				return (-1);
+
 			(void) strlcpy(propbuf, strval, proplen);
+			zcp_check(zhp, prop, 0, strval);
 			break;
 
 		default:
diff --git a/lib/libzfs/libzfs_util.c b/lib/libzfs/libzfs_util.c
index e13fee16d..ee8d09b11 100644
--- a/lib/libzfs/libzfs_util.c
+++ b/lib/libzfs/libzfs_util.c
@@ -1053,6 +1053,10 @@ libzfs_init(void)
 	libzfs_mnttab_init(hdl);
 	fletcher_4_init();
 
+	if (getenv("ZFS_PROP_DEBUG") != NULL) {
+		hdl->libzfs_prop_debug = B_TRUE;
+	}
+
 	return (hdl);
 }
 
diff --git a/lib/libzfs_core/Makefile.am b/lib/libzfs_core/Makefile.am
index 8abaebe73..421b8b4bf 100644
--- a/lib/libzfs_core/Makefile.am
+++ b/lib/libzfs_core/Makefile.am
@@ -12,7 +12,8 @@ USER_C = \
 nodist_libzfs_core_la_SOURCES = $(USER_C)
 
 libzfs_core_la_LIBADD = \
-	$(top_builddir)/lib/libnvpair/libnvpair.la
+	$(top_builddir)/lib/libnvpair/libnvpair.la \
+	$(top_builddir)/lib/libuutil/libuutil.la
 
 libzfs_core_la_LDFLAGS = -version-info 1:0:0
 
diff --git a/lib/libzfs_core/libzfs_core.c b/lib/libzfs_core/libzfs_core.c
index 675f4a3de..2fa035e67 100644
--- a/lib/libzfs_core/libzfs_core.c
+++ b/lib/libzfs_core/libzfs_core.c
@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  * Copyright (c) 2017 Datto Inc.
  * Copyright 2017 RackTop Systems.
@@ -156,7 +156,15 @@ lzc_ioctl(zfs_ioc_t ioc, const char *name,
 	}
 
 	while (ioctl(g_fd, ioc, &zc) != 0) {
-		if (errno == ENOMEM && resultp != NULL) {
+		/*
+		 * If ioctl exited with ENOMEM, we retry the ioctl after
+		 * increasing the size of the destination nvlist.
+		 *
+		 * Channel programs that exit with ENOMEM probably ran over the
+		 * lua memory sandbox; they should not be retried.
+		 */
+		if (errno == ENOMEM && resultp != NULL &&
+		    ioc != ZFS_IOC_CHANNEL_PROGRAM) {
 			free((void *)(uintptr_t)zc.zc_nvlist_dst);
 			zc.zc_nvlist_dst_size *= 2;
 			zc.zc_nvlist_dst = (uint64_t)(uintptr_t)
@@ -1051,6 +1059,57 @@ lzc_destroy_bookmarks(nvlist_t *bmarks, nvlist_t **errlist)
 }
 
 /*
+ * Executes a channel program.
+ *
+ * If this function returns 0 the channel program was successfully loaded and
+ * ran without failing. Note that individual commands the channel program ran
+ * may have failed and the channel program is responsible for reporting such
+ * errors through outnvl if they are important.
+ *
+ * This method may also return:
+ *
+ * EINVAL   The program contains syntax errors, or an invalid memory or time
+ *          limit was given. No part of the channel program was executed.
+ *          If caused by syntax errors, 'outnvl' contains information about the
+ *          errors.
+ *
+ * ECHRNG   The program was executed, but encountered a runtime error, such as
+ *          calling a function with incorrect arguments, invoking the error()
+ *          function directly, failing an assert() command, etc. Some portion
+ *          of the channel program may have executed and committed changes.
+ *          Information about the failure can be found in 'outnvl'.
+ *
+ * ENOMEM   The program fully executed, but the output buffer was not large
+ *          enough to store the returned value. No output is returned through
+ *          'outnvl'.
+ *
+ * ENOSPC   The program was terminated because it exceeded its memory usage
+ *          limit. Some portion of the channel program may have executed and
+ *          committed changes to disk. No output is returned through 'outnvl'.
+ *
+ * ETIME    The program was terminated because it exceeded its Lua instruction
+ *          limit. Some portion of the channel program may have executed and
+ *          committed changes to disk. No output is returned through 'outnvl'.
+ */
+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 (error);
+}
+
+/*
  * Performs key management functions
  *
  * crypto_cmd should be a value from zfs_ioc_crypto_cmd_t. If the command
@@ -1110,6 +1169,7 @@ lzc_change_key(const char *fsname, uint64_t crypt_cmd, nvlist_t *props,
 	error = lzc_ioctl(ZFS_IOC_CHANGE_KEY, fsname, ioc_args, NULL);
 	nvlist_free(hidden_args);
 	nvlist_free(ioc_args);
+
 	return (error);
 }
 
diff --git a/lib/libzpool/Makefile.am b/lib/libzpool/Makefile.am
index 06219168b..4ea7961f9 100644
--- a/lib/libzpool/Makefile.am
+++ b/lib/libzpool/Makefile.am
@@ -3,6 +3,7 @@ include $(top_srcdir)/config/Rules.am
 VPATH = \
 	$(top_srcdir)/module/zfs \
 	$(top_srcdir)/module/zcommon \
+	$(top_srcdir)/module/lua \
 	$(top_srcdir)/lib/libzpool
 
 # Suppress unused but set variable warnings often due to ASSERTs
@@ -121,6 +122,11 @@ KERNEL_C = \
 	zap.c \
 	zap_leaf.c \
 	zap_micro.c \
+	zcp.c \
+	zcp_get.c \
+	zcp_global.c \
+	zcp_iter.c \
+	zcp_synctask.c \
 	zfeature.c \
 	zfs_byteswap.c \
 	zfs_debug.c \
@@ -139,9 +145,39 @@ KERNEL_C = \
 	zle.c \
 	zrlock.c
 
+LUA_C = \
+	lapi.c \
+	lauxlib.c \
+	lbaselib.c \
+	lbitlib.c \
+	lcode.c \
+	lcompat.c \
+	lcorolib.c \
+	lctype.c \
+	ldebug.c \
+	ldo.c \
+	ldump.c \
+	lfunc.c \
+	lgc.c \
+	llex.c \
+	lmem.c \
+	lobject.c \
+	lopcodes.c \
+	lparser.c \
+	lstate.c \
+	lstring.c \
+	lstrlib.c \
+	ltable.c \
+	ltablib.c \
+	ltm.c \
+	lundump.c \
+	lvm.c \
+	lzio.c
+
 nodist_libzpool_la_SOURCES = \
 	$(USER_C) \
-	$(KERNEL_C)
+	$(KERNEL_C) \
+	$(LUA_C)
 
 libzpool_la_LIBADD = \
 	$(top_builddir)/lib/libavl/libavl.la \