]> granicus.if.org Git - zfs/blobdiff - lib/libzfs/libzfs_util.c
Race condition between spa async threads and export
[zfs] / lib / libzfs / libzfs_util.c
old mode 100755 (executable)
new mode 100644 (file)
index 2f29027..9dcbb9b
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013, Joyent, Inc. All rights reserved.
- * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2018, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
+ * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
+ * Copyright (c) 2017 Datto Inc.
  */
 
 /*
 #include <stdlib.h>
 #include <strings.h>
 #include <unistd.h>
-#include <ctype.h>
 #include <math.h>
 #include <sys/stat.h>
 #include <sys/mnttab.h>
 #include <sys/mntent.h>
 #include <sys/types.h>
-#include <wait.h>
+#include <sys/wait.h>
 
 #include <libzfs.h>
 #include <libzfs_core.h>
@@ -52,6 +53,8 @@
 #include "zfs_prop.h"
 #include "zfeature_common.h"
 #include <zfs_fletcher.h>
+#include <libzutil.h>
+#include <sys/zfs_sysfs.h>
 
 int
 libzfs_errno(libzfs_handle_t *hdl)
@@ -225,6 +228,9 @@ libzfs_error_description(libzfs_handle_t *hdl)
        case EZFS_NOTSUP:
                return (dgettext(TEXT_DOMAIN, "operation not supported "
                    "on this dataset"));
+       case EZFS_IOC_NOTSUPPORTED:
+               return (dgettext(TEXT_DOMAIN, "operation not supported by "
+                   "zfs kernel module"));
        case EZFS_ACTIVE_SPARE:
                return (dgettext(TEXT_DOMAIN, "pool has active shared spare "
                    "device"));
@@ -245,6 +251,9 @@ libzfs_error_description(libzfs_handle_t *hdl)
        case EZFS_POSTSPLIT_ONLINE:
                return (dgettext(TEXT_DOMAIN, "disk was split from this pool "
                    "into a new one"));
+       case EZFS_SCRUB_PAUSED:
+               return (dgettext(TEXT_DOMAIN, "scrub is paused; "
+                   "use 'zpool scrub' to resume"));
        case EZFS_SCRUBBING:
                return (dgettext(TEXT_DOMAIN, "currently scrubbing; "
                    "use 'zpool scrub -s' to cancel current scrub"));
@@ -256,6 +265,46 @@ libzfs_error_description(libzfs_handle_t *hdl)
                return (dgettext(TEXT_DOMAIN, "invalid diff data"));
        case EZFS_POOLREADONLY:
                return (dgettext(TEXT_DOMAIN, "pool is read-only"));
+       case EZFS_NO_PENDING:
+               return (dgettext(TEXT_DOMAIN, "operation is not "
+                   "in progress"));
+       case EZFS_CHECKPOINT_EXISTS:
+               return (dgettext(TEXT_DOMAIN, "checkpoint exists"));
+       case EZFS_DISCARDING_CHECKPOINT:
+               return (dgettext(TEXT_DOMAIN, "currently discarding "
+                   "checkpoint"));
+       case EZFS_NO_CHECKPOINT:
+               return (dgettext(TEXT_DOMAIN, "checkpoint does not exist"));
+       case EZFS_DEVRM_IN_PROGRESS:
+               return (dgettext(TEXT_DOMAIN, "device removal in progress"));
+       case EZFS_VDEV_TOO_BIG:
+               return (dgettext(TEXT_DOMAIN, "device exceeds supported size"));
+       case EZFS_ACTIVE_POOL:
+               return (dgettext(TEXT_DOMAIN, "pool is imported on a "
+                   "different host"));
+       case EZFS_CRYPTOFAILED:
+               return (dgettext(TEXT_DOMAIN, "encryption failure"));
+       case EZFS_TOOMANY:
+               return (dgettext(TEXT_DOMAIN, "argument list too long"));
+       case EZFS_INITIALIZING:
+               return (dgettext(TEXT_DOMAIN, "currently initializing"));
+       case EZFS_NO_INITIALIZE:
+               return (dgettext(TEXT_DOMAIN, "there is no active "
+                   "initialization"));
+       case EZFS_WRONG_PARENT:
+               return (dgettext(TEXT_DOMAIN, "invalid parent dataset"));
+       case EZFS_TRIMMING:
+               return (dgettext(TEXT_DOMAIN, "currently trimming"));
+       case EZFS_NO_TRIM:
+               return (dgettext(TEXT_DOMAIN, "there is no active trim"));
+       case EZFS_TRIM_NOTSUP:
+               return (dgettext(TEXT_DOMAIN, "trim operations are not "
+                   "supported by this device"));
+       case EZFS_NO_RESILVER_DEFER:
+               return (dgettext(TEXT_DOMAIN, "this action requires the "
+                   "resilver_defer feature"));
+       case EZFS_EXPORT_IN_PROGRESS:
+               return (dgettext(TEXT_DOMAIN, "pool export in progress"));
        case EZFS_UNKNOWN:
                return (dgettext(TEXT_DOMAIN, "unknown error"));
        default:
@@ -418,6 +467,29 @@ zfs_standard_error_fmt(libzfs_handle_t *hdl, int error, const char *fmt, ...)
                    "pool I/O is currently suspended"));
                zfs_verror(hdl, EZFS_POOLUNAVAIL, fmt, ap);
                break;
+       case EREMOTEIO:
+               zfs_verror(hdl, EZFS_ACTIVE_POOL, fmt, ap);
+               break;
+       case ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE:
+       case ZFS_ERR_IOC_CMD_UNAVAIL:
+               zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
+                   "module does not support this operation. A reboot may "
+                   "be required to enable this operation."));
+               zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
+               break;
+       case ZFS_ERR_IOC_ARG_UNAVAIL:
+               zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
+                   "module does not support an option for this operation. "
+                   "A reboot may be required to enable this option."));
+               zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
+               break;
+       case ZFS_ERR_IOC_ARG_REQUIRED:
+       case ZFS_ERR_IOC_ARG_BADTYPE:
+               zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
+               break;
+       case ZFS_ERR_WRONG_PARENT:
+               zfs_verror(hdl, EZFS_WRONG_PARENT, fmt, ap);
+               break;
        default:
                zfs_error_aux(hdl, strerror(error));
                zfs_verror(hdl, EZFS_UNKNOWN, fmt, ap);
@@ -469,6 +541,11 @@ zpool_standard_error_fmt(libzfs_handle_t *hdl, int error, const char *fmt, ...)
                zfs_verror(hdl, EZFS_BUSY, fmt, ap);
                break;
 
+       /* There is no pending operation to cancel */
+       case ENOTACTIVE:
+               zfs_verror(hdl, EZFS_NO_PENDING, fmt, ap);
+               break;
+
        case ENXIO:
                zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
                    "one or more devices is currently unavailable"));
@@ -506,7 +583,43 @@ zpool_standard_error_fmt(libzfs_handle_t *hdl, int error, const char *fmt, ...)
                    "block size out of range or does not match"));
                zfs_verror(hdl, EZFS_BADPROP, fmt, ap);
                break;
-
+       case EREMOTEIO:
+               zfs_verror(hdl, EZFS_ACTIVE_POOL, fmt, ap);
+               break;
+       case ZFS_ERR_CHECKPOINT_EXISTS:
+               zfs_verror(hdl, EZFS_CHECKPOINT_EXISTS, fmt, ap);
+               break;
+       case ZFS_ERR_DISCARDING_CHECKPOINT:
+               zfs_verror(hdl, EZFS_DISCARDING_CHECKPOINT, fmt, ap);
+               break;
+       case ZFS_ERR_NO_CHECKPOINT:
+               zfs_verror(hdl, EZFS_NO_CHECKPOINT, fmt, ap);
+               break;
+       case ZFS_ERR_DEVRM_IN_PROGRESS:
+               zfs_verror(hdl, EZFS_DEVRM_IN_PROGRESS, fmt, ap);
+               break;
+       case ZFS_ERR_VDEV_TOO_BIG:
+               zfs_verror(hdl, EZFS_VDEV_TOO_BIG, fmt, ap);
+               break;
+       case ZFS_ERR_EXPORT_IN_PROGRESS:
+               zfs_verror(hdl, EZFS_EXPORT_IN_PROGRESS, fmt, ap);
+               break;
+       case ZFS_ERR_IOC_CMD_UNAVAIL:
+               zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
+                   "module does not support this operation. A reboot may "
+                   "be required to enable this operation."));
+               zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
+               break;
+       case ZFS_ERR_IOC_ARG_UNAVAIL:
+               zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "the loaded zfs "
+                   "module does not support an option for this operation. "
+                   "A reboot may be required to enable this option."));
+               zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
+               break;
+       case ZFS_ERR_IOC_ARG_REQUIRED:
+       case ZFS_ERR_IOC_ARG_BADTYPE:
+               zfs_verror(hdl, EZFS_IOC_NOTSUPPORTED, fmt, ap);
+               break;
        default:
                zfs_error_aux(hdl, strerror(error));
                zfs_verror(hdl, EZFS_UNKNOWN, fmt, ap);
@@ -593,120 +706,6 @@ zfs_strdup(libzfs_handle_t *hdl, const char *str)
        return (ret);
 }
 
-/*
- * Convert a number to an appropriately human-readable output.
- */
-void
-zfs_nicenum_format(uint64_t num, char *buf, size_t buflen,
-    enum zfs_nicenum_format format)
-{
-       uint64_t n = num;
-       int index = 0;
-       const char *u;
-       const char *units[3][7] = {
-           [ZFS_NICENUM_1024] = {"", "K", "M", "G", "T", "P", "E"},
-           [ZFS_NICENUM_TIME] = {"ns", "us", "ms", "s", "?", "?", "?"}
-       };
-
-       const int units_len[] = {[ZFS_NICENUM_1024] = 6,
-           [ZFS_NICENUM_TIME] = 4};
-
-       const int k_unit[] = {  [ZFS_NICENUM_1024] = 1024,
-           [ZFS_NICENUM_TIME] = 1000};
-
-       double val;
-
-       if (format == ZFS_NICENUM_RAW) {
-               snprintf(buf, buflen, "%llu", (u_longlong_t) num);
-               return;
-       }
-
-
-       while (n >= k_unit[format] && index < units_len[format]) {
-               n /= k_unit[format];
-               index++;
-       }
-
-       u = units[format][index];
-
-       /* Don't print 0ns times */
-       if ((format == ZFS_NICENUM_TIME) && (num == 0)) {
-               (void) snprintf(buf, buflen, "-");
-       } else if ((index == 0) || ((num %
-           (uint64_t) powl(k_unit[format], index)) == 0)) {
-               /*
-                * If this is an even multiple of the base, always display
-                * without any decimal precision.
-                */
-               (void) snprintf(buf, buflen, "%llu%s", (u_longlong_t) n, u);
-
-       } else {
-               /*
-                * We want to choose a precision that reflects the best choice
-                * for fitting in 5 characters.  This can get rather tricky when
-                * we have numbers that are very close to an order of magnitude.
-                * For example, when displaying 10239 (which is really 9.999K),
-                * we want only a single place of precision for 10.0K.  We could
-                * develop some complex heuristics for this, but it's much
-                * easier just to try each combination in turn.
-                */
-               int i;
-               for (i = 2; i >= 0; i--) {
-                       val = (double) num /
-                           (uint64_t) powl(k_unit[format], index);
-
-                       /*
-                        * Don't print floating point values for time.  Note,
-                        * we use floor() instead of round() here, since
-                        * round can result in undesirable results.  For
-                        * example, if "num" is in the range of
-                        * 999500-999999, it will print out "1000us".  This
-                        * doesn't happen if we use floor().
-                        */
-                       if (format == ZFS_NICENUM_TIME) {
-                               if (snprintf(buf, buflen, "%d%s",
-                                   (unsigned int) floor(val), u) <= 5)
-                                       break;
-
-                       } else {
-                               if (snprintf(buf, buflen, "%.*f%s", i,
-                                   val, u) <= 5)
-                                       break;
-                       }
-               }
-       }
-}
-
-/*
- * Convert a number to an appropriately human-readable output.
- */
-void
-zfs_nicenum(uint64_t num, char *buf, size_t buflen)
-{
-       zfs_nicenum_format(num, buf, buflen, ZFS_NICENUM_1024);
-}
-
-/*
- * Convert a time to an appropriately human-readable output.
- * @num:       Time in nanoseconds
- */
-void
-zfs_nicetime(uint64_t num, char *buf, size_t buflen)
-{
-       zfs_nicenum_format(num, buf, buflen, ZFS_NICENUM_TIME);
-}
-
-/*
- * Print out a raw number with correct column spacing
- */
-void
-zfs_niceraw(uint64_t num, char *buf, size_t buflen)
-{
-       zfs_nicenum_format(num, buf, buflen, ZFS_NICENUM_RAW);
-}
-
-
-
 void
 libzfs_print_on_error(libzfs_handle_t *hdl, boolean_t printerr)
 {
@@ -725,43 +724,185 @@ libzfs_module_loaded(const char *module)
        return (access(path, F_OK) == 0);
 }
 
-int
-libzfs_run_process(const char *path, char *argv[], int flags)
+
+/*
+ * Read lines from an open file descriptor and store them in an array of
+ * strings until EOF.  lines[] will be allocated and populated with all the
+ * lines read.  All newlines are replaced with NULL terminators for
+ * convenience.  lines[] must be freed after use with libzfs_free_str_array().
+ *
+ * Returns the number of lines read.
+ */
+static int
+libzfs_read_stdout_from_fd(int fd, char **lines[])
+{
+
+       FILE *fp;
+       int lines_cnt = 0;
+       size_t len = 0;
+       char *line = NULL;
+       char **tmp_lines = NULL, **tmp;
+       char *nl = NULL;
+       int rc;
+
+       fp = fdopen(fd, "r");
+       if (fp == NULL)
+               return (0);
+       while (1) {
+               rc = getline(&line, &len, fp);
+               if (rc == -1)
+                       break;
+
+               tmp = realloc(tmp_lines, sizeof (*tmp_lines) * (lines_cnt + 1));
+               if (tmp == NULL) {
+                       /* Return the lines we were able to process */
+                       break;
+               }
+               tmp_lines = tmp;
+
+               /* Terminate newlines */
+               if ((nl = strchr(line, '\n')) != NULL)
+                       *nl = '\0';
+               tmp_lines[lines_cnt] = line;
+               lines_cnt++;
+               line = NULL;
+       }
+       fclose(fp);
+       *lines = tmp_lines;
+       return (lines_cnt);
+}
+
+static int
+libzfs_run_process_impl(const char *path, char *argv[], char *env[], int flags,
+    char **lines[], int *lines_cnt)
 {
        pid_t pid;
        int error, devnull_fd;
+       int link[2];
+
+       /*
+        * Setup a pipe between our child and parent process if we're
+        * reading stdout.
+        */
+       if ((lines != NULL) && pipe(link) == -1)
+               return (-ESTRPIPE);
 
        pid = vfork();
        if (pid == 0) {
+               /* Child process */
                devnull_fd = open("/dev/null", O_WRONLY);
 
                if (devnull_fd < 0)
                        _exit(-1);
 
-               if (!(flags & STDOUT_VERBOSE))
+               if (!(flags & STDOUT_VERBOSE) && (lines == NULL))
                        (void) dup2(devnull_fd, STDOUT_FILENO);
+               else if (lines != NULL) {
+                       /* Save the output to lines[] */
+                       dup2(link[1], STDOUT_FILENO);
+                       close(link[0]);
+                       close(link[1]);
+               }
 
                if (!(flags & STDERR_VERBOSE))
                        (void) dup2(devnull_fd, STDERR_FILENO);
 
                close(devnull_fd);
 
-               (void) execvp(path, argv);
+               if (flags & NO_DEFAULT_PATH) {
+                       if (env == NULL)
+                               execv(path, argv);
+                       else
+                               execve(path, argv, env);
+               } else {
+                       if (env == NULL)
+                               execvp(path, argv);
+                       else
+                               execvpe(path, argv, env);
+               }
+
                _exit(-1);
        } else if (pid > 0) {
+               /* Parent process */
                int status;
 
                while ((error = waitpid(pid, &status, 0)) == -1 &&
-                       errno == EINTR);
+                   errno == EINTR) { }
                if (error < 0 || !WIFEXITED(status))
                        return (-1);
 
+               if (lines != NULL) {
+                       close(link[1]);
+                       *lines_cnt = libzfs_read_stdout_from_fd(link[0], lines);
+               }
                return (WEXITSTATUS(status));
        }
 
        return (-1);
 }
 
+int
+libzfs_run_process(const char *path, char *argv[], int flags)
+{
+       return (libzfs_run_process_impl(path, argv, NULL, flags, NULL, NULL));
+}
+
+/*
+ * Run a command and store its stdout lines in an array of strings (lines[]).
+ * lines[] is allocated and populated for you, and the number of lines is set in
+ * lines_cnt.  lines[] must be freed after use with libzfs_free_str_array().
+ * All newlines (\n) in lines[] are terminated for convenience.
+ */
+int
+libzfs_run_process_get_stdout(const char *path, char *argv[], char *env[],
+    char **lines[], int *lines_cnt)
+{
+       return (libzfs_run_process_impl(path, argv, env, 0, lines, lines_cnt));
+}
+
+/*
+ * Same as libzfs_run_process_get_stdout(), but run without $PATH set.  This
+ * means that *path needs to be the full path to the executable.
+ */
+int
+libzfs_run_process_get_stdout_nopath(const char *path, char *argv[],
+    char *env[], char **lines[], int *lines_cnt)
+{
+       return (libzfs_run_process_impl(path, argv, env, NO_DEFAULT_PATH,
+           lines, lines_cnt));
+}
+
+/*
+ * Free an array of strings.  Free both the strings contained in the array and
+ * the array itself.
+ */
+void
+libzfs_free_str_array(char **strs, int count)
+{
+       while (--count >= 0)
+               free(strs[count]);
+
+       free(strs);
+}
+
+/*
+ * Returns 1 if environment variable is set to "YES", "yes", "ON", "on", or
+ * a non-zero number.
+ *
+ * Returns 0 otherwise.
+ */
+int
+libzfs_envvar_is_set(char *envvar)
+{
+       char *env = getenv(envvar);
+       if (env && (strtoul(env, NULL, 0) > 0 ||
+           (!strncasecmp(env, "YES", 3) && strnlen(env, 4) == 3) ||
+           (!strncasecmp(env, "ON", 2) && strnlen(env, 3) == 2)))
+               return (1);
+
+       return (0);
+}
+
 /*
  * Verify the required ZFS_DEV device is available and optionally attempt
  * to load the ZFS modules.  Under normal circumstances the modules
@@ -792,13 +933,14 @@ libzfs_load_module(const char *module)
                                load = 0;
                }
 
-               if (load && libzfs_run_process("/sbin/modprobe", argv, 0))
-                       return (ENOEXEC);
-       }
+               if (load) {
+                       if (libzfs_run_process("/sbin/modprobe", argv, 0))
+                               return (ENOEXEC);
+               }
 
-       /* Module loading is synchronous it must be available */
-       if (!libzfs_module_loaded(module))
-               return (ENXIO);
+               if (!libzfs_module_loaded(module))
+                       return (ENXIO);
+       }
 
        /*
         * Device creation by udev is asynchronous and waiting may be
@@ -863,12 +1005,13 @@ libzfs_init(void)
                return (NULL);
        }
 
-       hdl->libzfs_sharetab = fopen("/etc/dfs/sharetab", "r");
+       hdl->libzfs_sharetab = fopen(ZFS_SHARETAB, "r");
 
        if (libzfs_core_init() != 0) {
                (void) close(hdl->libzfs_fd);
                (void) fclose(hdl->libzfs_mnttab);
-               (void) fclose(hdl->libzfs_sharetab);
+               if (hdl->libzfs_sharetab)
+                       (void) fclose(hdl->libzfs_sharetab);
                free(hdl);
                return (NULL);
        }
@@ -879,6 +1022,26 @@ libzfs_init(void)
        libzfs_mnttab_init(hdl);
        fletcher_4_init();
 
+       if (getenv("ZFS_PROP_DEBUG") != NULL) {
+               hdl->libzfs_prop_debug = B_TRUE;
+       }
+
+       /*
+        * For testing, remove some settable properties and features
+        */
+       if (libzfs_envvar_is_set("ZFS_SYSFS_PROP_SUPPORT_TEST")) {
+               zprop_desc_t *proptbl;
+
+               proptbl = zpool_prop_get_table();
+               proptbl[ZPOOL_PROP_COMMENT].pd_zfs_mod_supported = B_FALSE;
+
+               proptbl = zfs_prop_get_table();
+               proptbl[ZFS_PROP_DNODESIZE].pd_zfs_mod_supported = B_FALSE;
+
+               zfeature_info_t *ftbl = spa_feature_table;
+               ftbl[SPA_FEATURE_LARGE_BLOCKS].fi_zfs_mod_supported = B_FALSE;
+       }
+
        return (hdl);
 }
 
@@ -896,7 +1059,6 @@ libzfs_fini(libzfs_handle_t *hdl)
                (void) fclose(hdl->libzfs_sharetab);
        zfs_uninit_libshare(hdl);
        zpool_free_handles(hdl);
-       libzfs_fru_clear(hdl, B_TRUE);
        namespace_clear(hdl);
        libzfs_mnttab_fini(hdl);
        libzfs_core_fini();
@@ -926,7 +1088,7 @@ zfs_get_pool_handle(const zfs_handle_t *zhp)
  * Given a name, determine whether or not it's a valid path
  * (starts with '/' or "./").  If so, walk the mnttab trying
  * to match the device number.  If not, treat the path as an
- * fs/vol/snap name.
+ * fs/vol/snap/bkmark name.
  */
 zfs_handle_t *
 zfs_path_to_zhandle(libzfs_handle_t *hdl, char *path, zfs_type_t argtype)
@@ -970,171 +1132,6 @@ zfs_path_to_zhandle(libzfs_handle_t *hdl, char *path, zfs_type_t argtype)
        return (zfs_open(hdl, entry.mnt_special, ZFS_TYPE_FILESYSTEM));
 }
 
-/*
- * Append partition suffix to an otherwise fully qualified device path.
- * This is used to generate the name the full path as its stored in
- * ZPOOL_CONFIG_PATH for whole disk devices.  On success the new length
- * of 'path' will be returned on error a negative value is returned.
- */
-int
-zfs_append_partition(char *path, size_t max_len)
-{
-       int len = strlen(path);
-
-       if ((strncmp(path, UDISK_ROOT, strlen(UDISK_ROOT)) == 0) ||
-           (strncmp(path, ZVOL_ROOT, strlen(ZVOL_ROOT)) == 0)) {
-               if (len + 6 >= max_len)
-                       return (-1);
-
-               (void) strcat(path, "-part1");
-               len += 6;
-       } else {
-               if (len + 2 >= max_len)
-                       return (-1);
-
-               if (isdigit(path[len-1])) {
-                       (void) strcat(path, "p1");
-                       len += 2;
-               } else {
-                       (void) strcat(path, "1");
-                       len += 1;
-               }
-       }
-
-       return (len);
-}
-
-/*
- * Given a shorthand device name check if a file by that name exists in any
- * of the 'zpool_default_import_path' or ZPOOL_IMPORT_PATH directories.  If
- * one is found, store its fully qualified path in the 'path' buffer passed
- * by the caller and return 0, otherwise return an error.
- */
-int
-zfs_resolve_shortname(const char *name, char *path, size_t len)
-{
-       int i, error = -1;
-       char *dir, *env, *envdup;
-
-       env = getenv("ZPOOL_IMPORT_PATH");
-       errno = ENOENT;
-
-       if (env) {
-               envdup = strdup(env);
-               dir = strtok(envdup, ":");
-               while (dir && error) {
-                       (void) snprintf(path, len, "%s/%s", dir, name);
-                       error = access(path, F_OK);
-                       dir = strtok(NULL, ":");
-               }
-               free(envdup);
-       } else {
-               for (i = 0; i < DEFAULT_IMPORT_PATH_SIZE && error < 0; i++) {
-                       (void) snprintf(path, len, "%s/%s",
-                           zpool_default_import_path[i], name);
-                       error = access(path, F_OK);
-               }
-       }
-
-       return (error ? ENOENT : 0);
-}
-
-/*
- * Given a shorthand device name look for a match against 'cmp_name'.  This
- * is done by checking all prefix expansions using either the default
- * 'zpool_default_import_paths' or the ZPOOL_IMPORT_PATH environment
- * variable.  Proper partition suffixes will be appended if this is a
- * whole disk.  When a match is found 0 is returned otherwise ENOENT.
- */
-static int
-zfs_strcmp_shortname(char *name, char *cmp_name, int wholedisk)
-{
-       int path_len, cmp_len, i = 0, error = ENOENT;
-       char *dir, *env, *envdup = NULL;
-       char path_name[MAXPATHLEN];
-
-       cmp_len = strlen(cmp_name);
-       env = getenv("ZPOOL_IMPORT_PATH");
-
-       if (env) {
-               envdup = strdup(env);
-               dir = strtok(envdup, ":");
-       } else {
-               dir =  zpool_default_import_path[i];
-       }
-
-       while (dir) {
-               /* Trim trailing directory slashes from ZPOOL_IMPORT_PATH */
-               while (dir[strlen(dir)-1] == '/')
-                       dir[strlen(dir)-1] = '\0';
-
-               path_len = snprintf(path_name, MAXPATHLEN, "%s/%s", dir, name);
-               if (wholedisk)
-                       path_len = zfs_append_partition(path_name, MAXPATHLEN);
-
-               if ((path_len == cmp_len) && strcmp(path_name, cmp_name) == 0) {
-                       error = 0;
-                       break;
-               }
-
-               if (env) {
-                       dir = strtok(NULL, ":");
-               } else if (++i < DEFAULT_IMPORT_PATH_SIZE) {
-                       dir = zpool_default_import_path[i];
-               } else {
-                       dir = NULL;
-               }
-       }
-
-       if (env)
-               free(envdup);
-
-       return (error);
-}
-
-/*
- * Given either a shorthand or fully qualified path name look for a match
- * against 'cmp'.  The passed name will be expanded as needed for comparison
- * purposes and redundant slashes stripped to ensure an accurate match.
- */
-int
-zfs_strcmp_pathname(char *name, char *cmp, int wholedisk)
-{
-       int path_len, cmp_len;
-       char path_name[MAXPATHLEN];
-       char cmp_name[MAXPATHLEN];
-       char *dir, *dup;
-
-       /* Strip redundant slashes if one exists due to ZPOOL_IMPORT_PATH */
-       memset(cmp_name, 0, MAXPATHLEN);
-       dup = strdup(cmp);
-       dir = strtok(dup, "/");
-       while (dir) {
-               strlcat(cmp_name, "/", sizeof (cmp_name));
-               strlcat(cmp_name, dir, sizeof (cmp_name));
-               dir = strtok(NULL, "/");
-       }
-       free(dup);
-
-       if (name[0] != '/')
-               return (zfs_strcmp_shortname(name, cmp_name, wholedisk));
-
-       (void) strlcpy(path_name, name, MAXPATHLEN);
-       path_len = strlen(path_name);
-       cmp_len = strlen(cmp_name);
-
-       if (wholedisk) {
-               path_len = zfs_append_partition(path_name, MAXPATHLEN);
-               if (path_len == -1)
-                       return (ENOMEM);
-       }
-
-       if ((path_len != cmp_len) || strcmp(path_name, cmp_name))
-               return (ENOENT);
-
-       return (0);
-}
-
 /*
  * Initialize the zc_nvlist_dst member to prepare for receiving an nvlist from
  * an ioctl().
@@ -1425,6 +1422,10 @@ zprop_print_one_property(const char *name, zprop_get_cbdata_t *cbp,
                        case ZPROP_SRC_RECEIVED:
                                str = "received";
                                break;
+
+                       default:
+                               str = NULL;
+                               assert(!"unhandled zprop_source_t");
                        }
                        break;
 
@@ -1585,6 +1586,8 @@ zprop_parse_value(libzfs_handle_t *hdl, nvpair_t *elem, int prop,
        const char *propname;
        char *value;
        boolean_t isnone = B_FALSE;
+       boolean_t isauto = B_FALSE;
+       int err = 0;
 
        if (type == ZFS_TYPE_POOL) {
                proptype = zpool_prop_get_type(prop);
@@ -1607,7 +1610,12 @@ zprop_parse_value(libzfs_handle_t *hdl, nvpair_t *elem, int prop,
                            "'%s' must be a string"), nvpair_name(elem));
                        goto error;
                }
-               (void) nvpair_value_string(elem, svalp);
+               err = nvpair_value_string(elem, svalp);
+               if (err != 0) {
+                       zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+                           "'%s' is invalid"), nvpair_name(elem));
+                       goto error;
+               }
                if (strlen(*svalp) >= ZFS_MAXPROPLEN) {
                        zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
                            "'%s' is too long"), nvpair_name(elem));
@@ -1620,8 +1628,9 @@ zprop_parse_value(libzfs_handle_t *hdl, nvpair_t *elem, int prop,
                        (void) nvpair_value_string(elem, &value);
                        if (strcmp(value, "none") == 0) {
                                isnone = B_TRUE;
-                       } else if (zfs_nicestrtonum(hdl, value, ivalp)
-                           != 0) {
+                       } else if (strcmp(value, "auto") == 0) {
+                               isauto = B_TRUE;
+                       } else if (zfs_nicestrtonum(hdl, value, ivalp) != 0) {
                                goto error;
                        }
                } else if (datatype == DATA_TYPE_UINT64) {
@@ -1651,6 +1660,31 @@ zprop_parse_value(libzfs_handle_t *hdl, nvpair_t *elem, int prop,
                    prop == ZFS_PROP_SNAPSHOT_LIMIT)) {
                        *ivalp = UINT64_MAX;
                }
+
+               /*
+                * Special handling for setting 'refreservation' to 'auto'.  Use
+                * UINT64_MAX to tell the caller to use zfs_fix_auto_resv().
+                * 'auto' is only allowed on volumes.
+                */
+               if (isauto) {
+                       switch (prop) {
+                       case ZFS_PROP_REFRESERVATION:
+                               if ((type & ZFS_TYPE_VOLUME) == 0) {
+                                       zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+                                           "'%s=auto' only allowed on "
+                                           "volumes"), nvpair_name(elem));
+                                       goto error;
+                               }
+                               *ivalp = UINT64_MAX;
+                               break;
+                       default:
+                               zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+                                   "'auto' is invalid value for '%s'"),
+                                   nvpair_name(elem));
+                               goto error;
+                       }
+               }
+
                break;
 
        case PROP_TYPE_INDEX:
@@ -1916,3 +1950,66 @@ zprop_iter(zprop_func func, void *cb, boolean_t show_all, boolean_t ordered,
 {
        return (zprop_iter_common(func, cb, show_all, ordered, type));
 }
+
+/*
+ * Fill given version buffer with zfs userland version
+ */
+void
+zfs_version_userland(char *version, int len)
+{
+       (void) strlcpy(version, ZFS_META_ALIAS, len);
+}
+
+/*
+ * Fill given version buffer with zfs kernel version read from ZFS_SYSFS_DIR
+ * Returns 0 on success, and -1 on error (with errno set)
+ */
+int
+zfs_version_kernel(char *version, int len)
+{
+       int _errno;
+       int fd;
+       int rlen;
+
+       if ((fd = open(ZFS_SYSFS_DIR "/version", O_RDONLY)) == -1)
+               return (-1);
+
+       if ((rlen = read(fd, version, len)) == -1) {
+               version[0] = '\0';
+               _errno = errno;
+               (void) close(fd);
+               errno = _errno;
+               return (-1);
+       }
+
+       version[rlen-1] = '\0';  /* discard '\n' */
+
+       if (close(fd) == -1)
+               return (-1);
+
+       return (0);
+}
+
+/*
+ * Prints both zfs userland and kernel versions
+ * Returns 0 on success, and -1 on error (with errno set)
+ */
+int
+zfs_version_print(void)
+{
+       char zver_userland[128];
+       char zver_kernel[128];
+
+       if (zfs_version_kernel(zver_kernel, sizeof (zver_kernel)) == -1) {
+               fprintf(stderr, "zfs_version_kernel() failed: %s\n",
+                   strerror(errno));
+               return (-1);
+       }
+
+       zfs_version_userland(zver_userland, sizeof (zver_userland));
+
+       (void) printf("%s\n", zver_userland);
+       (void) printf("zfs-kmod-%s\n", zver_kernel);
+
+       return (0);
+}