Whamcloud - gitweb
LU-5817 clio: Do not allow group locks with gid 0
[fs/lustre-release.git] / lustre / utils / lfs.c
index 1c0b2fe..47860f7 100644 (file)
@@ -54,6 +54,7 @@
 #include <err.h>
 #include <pwd.h>
 #include <grp.h>
+#include <sys/quota.h>
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <fcntl.h>
 # include <sys/quota.h>
 #endif
 
-/* For dirname() */
-#include <libgen.h>
-
-#include <lnet/lnetctl.h>
-
 #include <libcfs/libcfs.h>
-#include <lustre/lustre_idl.h>
-#include <lustre/lustreapi.h>
-
 #include <libcfs/libcfsutil.h>
-#include "obdctl.h"
+#include <lustre/lustreapi.h>
+#include <lustre_ver.h>
 
 /* all functions */
 static int lfs_setstripe(int argc, char **argv);
@@ -127,13 +121,21 @@ static int lfs_mv(int argc, char **argv);
        "                 [--stripe-size|-S <stripe_size>]\n"\
        "                 [--pool|-p <pool_name>]\n"\
        "                 [--block|-b] "_tgt"\n"\
+       "                 [--ost-list|-o <ost_indices>]\n"\
        "\tstripe_size:  Number of bytes on each OST (0 filesystem default)\n"\
        "\t              Can be specified with k, m or g (in KB, MB and GB\n"\
        "\t              respectively)\n"\
        "\tstart_ost_idx: OST index of first stripe (-1 default)\n"\
        "\tstripe_count: Number of OSTs to stripe over (0 default, -1 all)\n"\
        "\tpool_name:    Name of OST pool to use (default none)\n"\
-       "\tblock:        Block file access during data migration"
+       "\tblock:        Block file access during data migration\n"\
+       "\tost_indices:  List of OST indices, can be repeated multiple times\n"\
+       "\t              Indices be specified in a format of:\n"\
+       "\t                -o <ost_1>,<ost_i>-<ost_j>,<ost_n>\n"\
+       "\t              Or:\n"\
+       "\t                -o <ost_1> -o <ost_i>-<ost_j> -o <ost_n>\n"\
+       "\t              If --pool is set with --ost-list, then the OSTs\n"\
+       "\t              must be the members of the pool."
 
 /* all avaialable commands */
 command_t cmdlist[] = {
@@ -154,14 +156,18 @@ command_t cmdlist[] = {
         "                 [--layout|-L]\n"
         "                 <directory|filename> ..."},
        {"setdirstripe", lfs_setdirstripe, 0,
-        "To create a remote directory on a specified MDT.\n"
+        "To create a striped directory on a specified MDT. This can only\n"
+        "be done on MDT0 with the right of administrator.\n"
         "usage: setdirstripe <--count|-c stripe_count>\n"
-        "[--index|-i mdt_index] [--hash-type|-t hash_type]\n"
-        "[--default_stripe|-D ] <dir>\n"
+        "              [--index|-i mdt_index] [--hash-type|-t hash_type]\n"
+        "              [--default_stripe|-D ] [--mode|-m mode] <dir>\n"
         "\tstripe_count: stripe count of the striped directory\n"
         "\tmdt_index:  MDT index of first stripe\n"
-        "\thash_type:  hash type of the striped directory\n"
-        "\tdefault_stripe: set default dirstripe of the directory\n"},
+        "\thash_type:  hash type of the striped directory. Hash types:\n"
+        "      fnv_1a_64 FNV-1a hash algorithm (default)\n"
+        "      all_char  sum of characters % MDT_COUNT (not recommended)\n"
+        "\tdefault_stripe: set default dirstripe of the directory\n"
+        "\tmode: the mode of the directory\n"},
        {"getdirstripe", lfs_getdirstripe, 0,
         "To list the striping info for a given directory\n"
         "or recursively for all directories in a directory tree.\n"
@@ -169,10 +175,18 @@ command_t cmdlist[] = {
         "               [--count|-c ] [--index|-i ] [--raw|-R]\n"
         "               [--recursive | -r] [ --default_stripe | -D ] <dir> "},
        {"mkdir", lfs_setdirstripe, 0,
-        "To create a remote directory on a specified MDT. And this can only\n"
-        "be done on MDT0 by administrator.\n"
-        "usage: mkdir <--index|-i mdt_index> <dir>\n"
-        "\tmdt_index:    MDT index of the remote directory.\n"},
+        "To create a striped directory on a specified MDT. This can only\n"
+        "be done on MDT0 with the right of administrator.\n"
+        "usage: mkdir <--count|-c stripe_count>\n"
+        "              [--index|-i mdt_index] [--hash-type|-t hash_type]\n"
+        "              [--default_stripe|-D ] [--mode|-m mode] <dir>\n"
+        "\tstripe_count: stripe count of the striped directory\n"
+        "\tmdt_index:  MDT index of first stripe\n"
+        "\thash_type:  hash type of the striped directory. Hash types:\n"
+        "      fnv_1a_64 FNV-1a hash algorithm (default)\n"
+        "      all_char  sum of characters % MDT_COUNT (not recommended)\n"
+        "\tdefault_stripe: set default dirstripe of the directory\n"
+        "\tmode: the mode of the directory\n"},
        {"rm_entry", lfs_rmentry, 0,
         "To remove the name entry of the remote directory. Note: This\n"
         "command will only delete the name entry, i.e. the remote directory\n"
@@ -292,7 +306,7 @@ command_t cmdlist[] = {
         "usage: fid2path [--link <linkno>] <fsname|rootpath> <fid> ..."
                /* [ --rec <recno> ] */ },
        {"path2fid", lfs_path2fid, 0, "Display the fid(s) for a given path(s).\n"
-        "usage: path2fid <path> ..."},
+        "usage: path2fid [--parents] <path> ..."},
        {"data_version", lfs_data_version, 0, "Display file data version for "
         "a given path.\n" "usage: data_version -[n|r|w] <path>"},
        {"hsm_state", lfs_hsm_state, 0, "Display the HSM information (states, "
@@ -334,18 +348,54 @@ command_t cmdlist[] = {
        {"help", Parser_help, 0, "help"},
        {"exit", Parser_quit, 0, "quit"},
        {"quit", Parser_quit, 0, "quit"},
+       {"--version", Parser_version, 0,
+        "output build version of the utility and exit"},
        { 0, 0, 0, NULL }
 };
 
+/* Generate a random id for the grouplock */
+static int random_group_id(int *gid)
+{
+       int     fd;
+       int     rc;
+       size_t  sz = sizeof(*gid);
+
+       fd = open("/dev/urandom", O_RDONLY);
+       if (fd < 0) {
+               rc = -errno;
+               fprintf(stderr, "cannot open /dev/urandom: %s\n",
+                       strerror(-rc));
+               goto out;
+       }
+
+retry:
+       rc = read(fd, gid, sz);
+       if (rc < sz) {
+               rc = -errno;
+               fprintf(stderr, "cannot read %zu bytes from /dev/urandom: %s\n",
+                       sz, strerror(-rc));
+               goto out;
+       }
+
+       /* gids must be non-zero */
+       if (*gid == 0)
+               goto retry;
+
+out:
+       if (fd >= 0)
+               close(fd);
+
+       return rc;
+}
+
 #define MIGRATION_BLOCKS 1
 
-static int lfs_migrate(char *name, unsigned long long stripe_size,
-                      int stripe_offset, int stripe_count,
-                      int stripe_pattern, char *pool_name,
-                      __u64 migration_flags)
+static int lfs_migrate(char *name, __u64 migration_flags,
+                      struct llapi_stripe_param *param)
 {
        int                      fd, fdv;
-       char                     volatile_file[PATH_MAX];
+       char                     volatile_file[PATH_MAX +
+                                               LUSTRE_VOLATILE_HDR_LEN + 4];
        char                     parent[PATH_MAX];
        char                    *ptr;
        int                      rc;
@@ -356,7 +406,7 @@ static int lfs_migrate(char *name, unsigned long long stripe_size,
        void                    *buf = NULL;
        int                      rsize, wsize;
        __u64                    rpos, wpos, bufoff;
-       int                      gid = 0, sz;
+       int                      gid;
        int                      have_gl = 0;
        struct stat              st, stv;
 
@@ -384,21 +434,10 @@ static int lfs_migrate(char *name, unsigned long long stripe_size,
        }
 
        if (migration_flags & MIGRATION_BLOCKS) {
-               /* generate a random id for the grouplock */
-               fd = open("/dev/urandom", O_RDONLY);
-               if (fd == -1) {
-                       rc = -errno;
-                       fprintf(stderr, "cannot open /dev/urandom (%s)\n",
-                               strerror(-rc));
-                       goto free;
-               }
-               sz = sizeof(gid);
-               rc = read(fd, &gid, sz);
-               close(fd);
-               if (rc < sz) {
-                       rc = -errno;
-                       fprintf(stderr, "cannot read %d bytes from"
-                               " /dev/urandom (%s)\n", sz, strerror(-rc));
+               rc = random_group_id(&gid);
+               if (rc < 0) {
+                       fprintf(stderr, "%s: cannot get random group ID: %s\n",
+                               name, strerror(-rc));
                        goto free;
                }
        }
@@ -421,14 +460,18 @@ static int lfs_migrate(char *name, unsigned long long stripe_size,
                else
                        *ptr = '\0';
        }
-       sprintf(volatile_file, "%s/%s::", parent, LUSTRE_VOLATILE_HDR);
+       rc = snprintf(volatile_file, sizeof(volatile_file), "%s/%s::", parent,
+                     LUSTRE_VOLATILE_HDR);
+       if (rc >= sizeof(volatile_file)) {
+               rc = -E2BIG;
+               goto free;
+       }
 
        /* create, open a volatile file, use caching (ie no directio) */
        /* exclusive create is not needed because volatile files cannot
         * conflict on name by construction */
-       fdv = llapi_file_open_pool(volatile_file, O_CREAT | O_WRONLY,
-                                  0644, stripe_size, stripe_offset,
-                                  stripe_count, stripe_pattern, pool_name);
+       fdv = llapi_file_open_param(volatile_file, O_CREAT | O_WRONLY, 0644,
+                                   param);
        if (fdv < 0) {
                rc = fdv;
                fprintf(stderr, "cannot create volatile file in %s (%s)\n",
@@ -489,8 +532,8 @@ static int lfs_migrate(char *name, unsigned long long stripe_size,
                 * be implemented (see LU-2919) */
                /* group lock is taken after data version read because it
                 * blocks data version call */
-               if (ioctl(fd, LL_IOC_GROUP_LOCK, gid) == -1) {
-                       rc = -errno;
+               rc = llapi_group_lock(fd, gid);
+               if (rc < 0) {
                        fprintf(stderr, "cannot get group lock on %s (%s)\n",
                                name, strerror(-rc));
                        goto error;
@@ -537,11 +580,10 @@ static int lfs_migrate(char *name, unsigned long long stripe_size,
 
        if (migration_flags & MIGRATION_BLOCKS) {
                /* give back group lock */
-               if (ioctl(fd, LL_IOC_GROUP_UNLOCK, gid) == -1) {
-                       rc = -errno;
+               rc = llapi_group_unlock(fd, gid);
+               if (rc < 0)
                        fprintf(stderr, "cannot put group lock on %s (%s)\n",
                                name, strerror(-rc));
-               }
                have_gl = 0;
        }
 
@@ -566,11 +608,14 @@ static int lfs_migrate(char *name, unsigned long long stripe_size,
 
 error:
        /* give back group lock */
-       if ((migration_flags & MIGRATION_BLOCKS) && have_gl &&
-           (ioctl(fd, LL_IOC_GROUP_UNLOCK, gid) == -1)) {
-               /* we keep in rc the original error */
-               fprintf(stderr, "cannot put group lock on %s (%s)\n",
-                       name, strerror(-errno));
+       if ((migration_flags & MIGRATION_BLOCKS) && have_gl) {
+               int rc2;
+
+               /* we keep the original error in rc */
+               rc2 = llapi_group_unlock(fd, gid);
+               if (rc2 < 0)
+                       fprintf(stderr, "cannot put group lock on %s (%s)\n",
+                               name, strerror(-rc2));
        }
 
        close(fdv);
@@ -583,23 +628,110 @@ free:
        return rc;
 }
 
+/**
+ * Parse a string containing an OST index list into an array of integers.
+ *
+ * The input string contains a comma delimited list of individual
+ * indices and ranges, for example "1,2-4,7". Add the indices into the
+ * \a osts array and remove duplicates.
+ *
+ * \param[out] osts    array to store indices in
+ * \param[in] size     size of \a osts array
+ * \param[in] offset   starting index in \a osts
+ * \param[in] arg      string containing OST index list
+ *
+ * \retval positive    number of indices in \a osts
+ * \retval -EINVAL     unable to parse \a arg
+ */
+static int parse_targets(__u32 *osts, int size, int offset, char *arg)
+{
+       int rc;
+       int nr = offset;
+       int slots = size - offset;
+       char *ptr = NULL;
+       bool end_of_loop;
+
+       if (arg == NULL)
+               return -EINVAL;
+
+       end_of_loop = false;
+       while (!end_of_loop) {
+               int start_index;
+               int end_index;
+               int i;
+               char *endptr = NULL;
+
+               rc = -EINVAL;
+
+               ptr = strchrnul(arg, ',');
+
+               end_of_loop = *ptr == '\0';
+               *ptr = '\0';
+
+               start_index = strtol(arg, &endptr, 0);
+               if (endptr == arg) /* no data at all */
+                       break;
+               if (*endptr != '-' && *endptr != '\0') /* has invalid data */
+                       break;
+               if (start_index < 0)
+                       break;
+
+               end_index = start_index;
+               if (*endptr == '-') {
+                       end_index = strtol(endptr + 1, &endptr, 0);
+                       if (*endptr != '\0')
+                               break;
+                       if (end_index < start_index)
+                               break;
+               }
+
+               for (i = start_index; i <= end_index && slots > 0; i++) {
+                       int j;
+
+                       /* remove duplicate */
+                       for (j = 0; j < offset; j++) {
+                               if (osts[j] == i)
+                                       break;
+                       }
+                       if (j == offset) { /* no duplicate */
+                               osts[nr++] = i;
+                               --slots;
+                       }
+               }
+               if (slots == 0 && i < end_index)
+                       break;
+
+               *ptr = ',';
+               arg = ++ptr;
+               offset = nr;
+               rc = 0;
+       }
+       if (!end_of_loop && ptr != NULL)
+               *ptr = ',';
+
+       return rc < 0 ? rc : nr;
+}
+
 /* functions */
 static int lfs_setstripe(int argc, char **argv)
 {
-       char                    *fname;
-       int                      result;
-       unsigned long long       st_size;
-       int                      st_offset, st_count;
-       char                    *end;
-       int                      c;
-       int                      delete = 0;
-       char                    *stripe_size_arg = NULL;
-       char                    *stripe_off_arg = NULL;
-       char                    *stripe_count_arg = NULL;
-       char                    *pool_name_arg = NULL;
-       unsigned long long       size_units = 1;
-       int                      migrate_mode = 0;
-       __u64                    migration_flags = 0;
+       struct llapi_stripe_param       *param;
+       char                            *fname;
+       int                              result;
+       unsigned long long               st_size;
+       int                              st_offset, st_count;
+       char                            *end;
+       int                              c;
+       int                              delete = 0;
+       char                            *stripe_size_arg = NULL;
+       char                            *stripe_off_arg = NULL;
+       char                            *stripe_count_arg = NULL;
+       char                            *pool_name_arg = NULL;
+       unsigned long long               size_units = 1;
+       bool                             migrate_mode = false;
+       __u64                            migration_flags = 0;
+       __u32                            osts[LOV_MAX_STRIPE_COUNT] = { 0 };
+       int                              nr_osts = 0;
 
        struct option            long_opts[] = {
                /* valid only in migrate mode */
@@ -621,12 +753,8 @@ static int lfs_setstripe(int argc, char **argv)
 #endif
                {"stripe-index", required_argument, 0, 'i'},
                {"stripe_index", required_argument, 0, 'i'},
-#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0)
-               /* This formerly implied "stripe-index", but was confusing
-                * with "file offset" (which will eventually be needed for
-                * with different layouts by offset), so deprecate it. */
-               {"offset",       required_argument, 0, 'o'},
-#endif
+               {"ost-list",     required_argument, 0, 'o'},
+               {"ost_list",     required_argument, 0, 'o'},
                {"pool",         required_argument, 0, 'p'},
 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0)
                /* This formerly implied "--stripe-size", but was confusing
@@ -639,12 +767,12 @@ static int lfs_setstripe(int argc, char **argv)
                {0, 0, 0, 0}
        };
 
-        st_size = 0;
-        st_offset = -1;
-        st_count = 0;
+       st_size = 0;
+       st_offset = -1;
+       st_count = 0;
 
        if (strcmp(argv[0], "migrate") == 0)
-               migrate_mode = 1;
+               migrate_mode = true;
 
        optind = 0;
        while ((c = getopt_long(argc, argv, "c:di:o:p:s:S:",
@@ -654,7 +782,7 @@ static int lfs_setstripe(int argc, char **argv)
                        /* Long options. */
                        break;
                case 'b':
-                       if (migrate_mode == 0) {
+                       if (!migrate_mode) {
                                fprintf(stderr, "--block is valid only for"
                                                " migrate mode");
                                return CMD_HELP;
@@ -673,11 +801,19 @@ static int lfs_setstripe(int argc, char **argv)
                        /* delete the default striping pattern */
                        delete = 1;
                        break;
-#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0)
                case 'o':
-                       fprintf(stderr, "warning: '--offset|-o' deprecated, "
-                               "use '--stripe-index|-i' instead\n");
-#endif
+                       nr_osts = parse_targets(osts, ARRAY_SIZE(osts), nr_osts,
+                                               optarg);
+                       if (nr_osts < 0) {
+                               fprintf(stderr,
+                                       "error: %s: bad OST indices '%s'\n",
+                                       argv[0], optarg);
+                               return CMD_HELP;
+                       }
+
+                       if (st_offset == -1) /* first in the command line */
+                               st_offset = osts[0];
+                       break;
                case 'i':
 #if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0)
                        if (strcmp(argv[optind - 1], "--index") == 0)
@@ -750,15 +886,44 @@ static int lfs_setstripe(int argc, char **argv)
                 }
         }
 
+       /* initialize stripe parameters */
+       param = calloc(1, offsetof(typeof(*param), lsp_osts[nr_osts]));
+       if (param == NULL) {
+               fprintf(stderr, "error: %s: run out of memory\n", argv[0]);
+               return CMD_HELP;
+       }
+
+       param->lsp_stripe_size = st_size;
+       param->lsp_stripe_offset = st_offset;
+       param->lsp_stripe_count = st_count;
+       param->lsp_stripe_pattern = 0;
+       param->lsp_pool = pool_name_arg;
+       param->lsp_is_specific = false;
+       if (nr_osts > 0) {
+               if (st_count > 0 && nr_osts != st_count) {
+                       fprintf(stderr, "error: %s: stripe count '%d' doesn't "
+                               "match the number of OSTs: %d\n",
+                               argv[0], st_count, nr_osts);
+                       return CMD_HELP;
+               }
+
+               param->lsp_is_specific = true;
+               param->lsp_stripe_count = nr_osts;
+               memcpy(param->lsp_osts, osts, sizeof(*osts) * nr_osts);
+       }
+
        do {
-               if (migrate_mode)
-                       result = lfs_migrate(fname, st_size, st_offset,
-                                            st_count, 0, pool_name_arg,
-                                            migration_flags);
-               else
-                       result = llapi_file_create_pool(fname, st_size,
-                                                       st_offset, st_count,
-                                                       0, pool_name_arg);
+               if (!migrate_mode) {
+                       result = llapi_file_open_param(fname,
+                                                      O_CREAT | O_WRONLY,
+                                                      0644, param);
+                       if (result >= 0) {
+                               close(result);
+                               result = 0;
+                       }
+               } else {
+                       result = lfs_migrate(fname, migration_flags, param);
+               }
                if (result) {
                        fprintf(stderr,
                                "error: %s: %s stripe file '%s' failed\n",
@@ -769,6 +934,7 @@ static int lfs_setstripe(int argc, char **argv)
                fname = argv[++optind];
        } while (fname != NULL);
 
+       free(param);
        return result;
 }
 
@@ -884,7 +1050,8 @@ static int name2layout(__u32 *layout, char *name)
 #define FIND_POOL_OPT 3
 static int lfs_find(int argc, char **argv)
 {
-        int c, ret;
+       int c, rc;
+       int ret = 0;
         time_t t;
        struct find_param param = {
                .fp_max_depth = -1,
@@ -986,14 +1153,14 @@ static int lfs_find(int argc, char **argv)
                                xsign = &param.fp_msign;
                                param.fp_exclude_mtime = !!neg_opt;
                        }
-                        ret = set_time(&t, xtime, optarg);
-                        if (ret == INT_MAX) {
-                                ret = -1;
-                                goto err;
-                        }
-                        if (ret)
-                                *xsign = ret;
-                        break;
+                       rc = set_time(&t, xtime, optarg);
+                       if (rc == INT_MAX) {
+                               ret = -1;
+                               goto err;
+                       }
+                       if (rc)
+                               *xsign = rc;
+                       break;
                 case 'c':
                         if (optarg[0] == '+') {
                                 param.stripecount_sign = -1;
@@ -1018,8 +1185,8 @@ static int lfs_find(int argc, char **argv)
                        break;
                case 'g':
                case 'G':
-                       ret = name2id(&param.fp_gid, optarg, GROUP);
-                       if (ret) {
+                       rc = name2id(&param.fp_gid, optarg, GROUP);
+                       if (rc) {
                                param.fp_gid = strtoul(optarg, &endptr, 10);
                                 if (*endptr != '\0') {
                                         fprintf(stderr, "Group/GID: %s cannot "
@@ -1040,8 +1207,8 @@ static int lfs_find(int argc, char **argv)
                        break;
                 case 'u':
                 case 'U':
-                       ret = name2id(&param.fp_uid, optarg, USER);
-                       if (ret) {
+                       rc = name2id(&param.fp_uid, optarg, USER);
+                       if (rc) {
                                param.fp_uid = strtoul(optarg, &endptr, 10);
                                 if (*endptr != '\0') {
                                         fprintf(stderr, "User/UID: %s cannot "
@@ -1102,8 +1269,11 @@ static int lfs_find(int argc, char **argv)
                                 tmp = realloc(param.mdtuuid,
                                               param.num_alloc_mdts *
                                               sizeof(*param.mdtuuid));
-                                if (tmp == NULL)
-                                        GOTO(err_free, ret = -ENOMEM);
+                               if (tmp == NULL) {
+                                       ret = -ENOMEM;
+                                       goto err_free;
+                               }
+
                                 param.mdtuuid = tmp;
                         } else {
                                 param.exclude_obd = !!neg_opt;
@@ -1111,8 +1281,11 @@ static int lfs_find(int argc, char **argv)
                                 tmp = realloc(param.obduuid,
                                               param.num_alloc_obds *
                                               sizeof(*param.obduuid));
-                                if (tmp == NULL)
-                                        GOTO(err_free, ret = -ENOMEM);
+                               if (tmp == NULL) {
+                                       ret = -ENOMEM;
+                                       goto err_free;
+                               }
+
                                 param.obduuid = tmp;
                         }
                         for (token = buf; token && *token; token = next) {
@@ -1130,8 +1303,12 @@ static int lfs_find(int argc, char **argv)
                                         *p = 0;
                                         next = p+1;
                                 }
-                               if (strlen(token) > sizeof(puuid->uuid)-1)
-                                       GOTO(err_free, ret = -E2BIG);
+
+                               if (strlen(token) > sizeof(puuid->uuid) - 1) {
+                                       ret = -E2BIG;
+                                       goto err_free;
+                               }
+
                                strncpy(puuid->uuid, token,
                                        sizeof(puuid->uuid));
                         }
@@ -1230,9 +1407,11 @@ err_free:
                 pathend = argc;
         }
 
-        do {
-                ret = llapi_find(argv[pathstart], &param);
-        } while (++pathstart < pathend && !ret);
+       do {
+               rc = llapi_find(argv[pathstart], &param);
+               if (rc != 0 && ret == 0)
+                       ret = rc;
+       } while (++pathstart < pathend);
 
         if (ret)
                 fprintf(stderr, "error: %s failed for %s.\n",
@@ -1485,19 +1664,24 @@ static int lfs_setdirstripe(int argc, char **argv)
        char                    *stripe_offset_opt = NULL;
        char                    *stripe_count_opt = NULL;
        char                    *stripe_hash_opt = NULL;
+       char                    *mode_opt = NULL;
        int                     default_stripe = 0;
+       mode_t                  mode = S_IRWXU | S_IRWXG | S_IRWXO;
+       mode_t                  previous_mode = 0;
 
        struct option long_opts[] = {
                {"count",       required_argument, 0, 'c'},
                {"index",       required_argument, 0, 'i'},
+               {"mode",        required_argument, 0, 'm'},
                {"hash-type",   required_argument, 0, 't'},
-               {"default_stripe", required_argument, 0, 'D'},
+               {"default_stripe", no_argument, 0, 'D'},
                {0, 0, 0, 0}
        };
 
        optind = 0;
 
-       while ((c = getopt_long(argc, argv, "c:Di:t:", long_opts, NULL)) >= 0) {
+       while ((c = getopt_long(argc, argv, "c:Di:m:t:", long_opts,
+                               NULL)) >= 0) {
                switch (c) {
                case 0:
                        /* Long options. */
@@ -1511,6 +1695,9 @@ static int lfs_setdirstripe(int argc, char **argv)
                case 'i':
                        stripe_offset_opt = optarg;
                        break;
+               case 'm':
+                       mode_opt = optarg;
+                       break;
                case 't':
                        stripe_hash_opt = optarg;
                        break;
@@ -1544,6 +1731,16 @@ static int lfs_setdirstripe(int argc, char **argv)
                }
        }
 
+       if (mode_opt != NULL) {
+               mode = strtoul(mode_opt, &end, 8);
+               if (*end != '\0') {
+                       fprintf(stderr, "error: %s: bad mode '%s'\n",
+                               argv[0], mode_opt);
+                       return CMD_HELP;
+               }
+               previous_mode = umask(0);
+       }
+
        if (stripe_hash_opt == NULL ||
            strcmp(stripe_hash_opt, LMV_HASH_NAME_FNV_1A_64) == 0) {
                hash_type = LMV_HASH_TYPE_FNV_1A_64;
@@ -1572,7 +1769,8 @@ static int lfs_setdirstripe(int argc, char **argv)
                                                    stripe_offset, stripe_count,
                                                    hash_type, NULL);
                } else {
-                       result = llapi_dir_create_pool(dname, 0, stripe_offset,
+                       result = llapi_dir_create_pool(dname, mode,
+                                                      stripe_offset,
                                                       stripe_count, hash_type,
                                                       NULL);
                }
@@ -1585,6 +1783,9 @@ static int lfs_setdirstripe(int argc, char **argv)
                dname = argv[++optind];
        } while (dname != NULL);
 
+       if (mode_opt != NULL)
+               umask(previous_mode);
+
        return result;
 }
 
@@ -1704,7 +1905,10 @@ static int showdf(char *mntdir, struct obd_statfs *stat,
         double ratio = 0;
         char *suffix = "KMGTPEZY";
         /* Note if we have >2^64 bytes/fs these buffers will need to be grown */
-        char tbuf[20], ubuf[20], abuf[20], rbuf[20];
+       char tbuf[3 * sizeof(__u64)];
+       char ubuf[3 * sizeof(__u64)];
+       char abuf[3 * sizeof(__u64)];
+       char rbuf[3 * sizeof(__u64)];
 
         if (!uuid || !stat)
                 return -EINVAL;
@@ -1828,6 +2032,9 @@ static int mntdf(char *mntdir, char *fsname, char *pool, int ishow,
                         if (rc == -ENODEV)
                                 break;
 
+                       if (rc == -EAGAIN)
+                               continue;
+
                         if (poolname && tp->st_op == LL_STATFS_LOV &&
                             llapi_search_ost(fsname, poolname,
                                              obd_uuid2str(&uuid_buf)) != 1)
@@ -2006,9 +2213,7 @@ static int lfs_check(int argc, char **argv)
                 return rc;
         }
 
-        rc = llapi_target_iterate(num_types, obd_types,
-                                  mntdir, llapi_ping_target);
-
+       rc = llapi_target_check(num_types, obd_types, mntdir);
         if (rc)
                 fprintf(stderr, "error: %s: %s status failed\n",
                                 argv[0],argv[1]);
@@ -2676,7 +2881,7 @@ static void print_quota(char *mnt, struct if_quotactl *qctl, int type,
                char strbuf[32];
 
                 if (dqb->dqb_bhardlimit &&
-                    toqb(dqb->dqb_curspace) >= dqb->dqb_bhardlimit) {
+                   lustre_stoqb(dqb->dqb_curspace) >= dqb->dqb_bhardlimit) {
                         bover = 1;
                 } else if (dqb->dqb_bsoftlimit && dqb->dqb_btime) {
                         if (dqb->dqb_btime > now) {
@@ -2706,7 +2911,7 @@ static void print_quota(char *mnt, struct if_quotactl *qctl, int type,
                if (bover)
                        diff2str(dqb->dqb_btime, timebuf, now);
 
-               kbytes2str(toqb(dqb->dqb_curspace), strbuf, h);
+               kbytes2str(lustre_stoqb(dqb->dqb_curspace), strbuf, h);
                if (rc == -EREMOTEIO)
                        sprintf(numbuf[0], "%s*", strbuf);
                else
@@ -2840,7 +3045,7 @@ static int lfs_quota(int argc, char **argv)
                         break;
                 case 'o':
                         valid = qctl.qc_valid = QC_UUID;
-                        strncpy(obd_uuid, optarg, sizeof(qctl.obd_uuid));
+                       strlcpy(obd_uuid, optarg, sizeof(qctl.obd_uuid));
                         break;
                 case 'i':
                         valid = qctl.qc_valid = QC_MDTIDX;
@@ -2992,10 +3197,10 @@ static int flushctx_ioctl(char *mp)
 
 static int lfs_flushctx(int argc, char **argv)
 {
-        int     kdestroy = 0, c;
-       FILE   *proc = NULL;
-        char    procline[PATH_MAX], *line;
-        int     rc = 0;
+       int     kdestroy = 0, c;
+       char    mntdir[PATH_MAX] = {'\0'};
+       int     index = 0;
+       int     rc = 0;
 
         optind = 0;
         while ((c = getopt(argc, argv, "k")) != -1) {
@@ -3011,46 +3216,24 @@ static int lfs_flushctx(int argc, char **argv)
         }
 
         if (kdestroy) {
-            int rc;
             if ((rc = system("kdestroy > /dev/null")) != 0) {
                 rc = WEXITSTATUS(rc);
                 fprintf(stderr, "error destroying tickets: %d, continuing\n", rc);
             }
         }
 
-        if (optind >= argc) {
-                /* flush for all mounted lustre fs. */
-                proc = fopen("/proc/mounts", "r");
-                if (!proc) {
-                        fprintf(stderr, "error: %s: can't open /proc/mounts\n",
-                                argv[0]);
-                        return -1;
-                }
-
-                while ((line = fgets(procline, PATH_MAX, proc)) != NULL) {
-                        char dev[PATH_MAX];
-                        char mp[PATH_MAX];
-                        char fs[PATH_MAX];
+       if (optind >= argc) {
+               /* flush for all mounted lustre fs. */
+               while (!llapi_search_mounts(NULL, index++, mntdir, NULL)) {
+                       /* Check if we have a mount point */
+                       if (mntdir[0] == '\0')
+                               continue;
 
-                        if (sscanf(line, "%s %s %s", dev, mp, fs) != 3) {
-                                fprintf(stderr, "%s: unexpected format in "
-                                                "/proc/mounts\n",
-                                        argv[0]);
+                       if (flushctx_ioctl(mntdir))
                                rc = -1;
-                               goto out;
-                        }
 
-                        if (strcmp(fs, "lustre") != 0)
-                                continue;
-                        /* we use '@' to determine it's a client. are there
-                         * any other better way?
-                         */
-                        if (strchr(dev, '@') == NULL)
-                                continue;
-
-                        if (flushctx_ioctl(mp))
-                                rc = -1;
-                }
+                       mntdir[0] = '\0'; /* avoid matching in next loop */
+               }
         } else {
                 /* flush fs as specified */
                 while (optind < argc) {
@@ -3058,10 +3241,6 @@ static int lfs_flushctx(int argc, char **argv)
                                 rc = -1;
                 }
         }
-
-out:
-       if (proc != NULL)
-               fclose(proc);
         return rc;
 }
 
@@ -3102,7 +3281,7 @@ static int lfs_ls(int argc, char **argv)
 static int lfs_changelog(int argc, char **argv)
 {
         void *changelog_priv;
-       struct changelog_ext_rec *rec;
+       struct changelog_rec *rec;
         long long startrec = 0, endrec = 0;
         char *mdd;
         struct option long_opts[] = {
@@ -3136,15 +3315,16 @@ static int lfs_changelog(int argc, char **argv)
         if (argc > optind)
                 endrec = strtoll(argv[optind++], NULL, 10);
 
-        rc = llapi_changelog_start(&changelog_priv,
-                                   CHANGELOG_FLAG_BLOCK |
-                                   (follow ? CHANGELOG_FLAG_FOLLOW : 0),
-                                   mdd, startrec);
-        if (rc < 0) {
-                fprintf(stderr, "Can't start changelog: %s\n",
-                        strerror(errno = -rc));
-                return rc;
-        }
+       rc = llapi_changelog_start(&changelog_priv,
+                                  CHANGELOG_FLAG_BLOCK |
+                                  CHANGELOG_FLAG_JOBID |
+                                  (follow ? CHANGELOG_FLAG_FOLLOW : 0),
+                                  mdd, startrec);
+       if (rc < 0) {
+               fprintf(stderr, "Can't start changelog: %s\n",
+                       strerror(errno = -rc));
+               return rc;
+       }
 
         while ((rc = llapi_changelog_recv(changelog_priv, &rec)) == 0) {
                 time_t secs;
@@ -3159,27 +3339,42 @@ static int lfs_changelog(int argc, char **argv)
                         continue;
                 }
 
-                secs = rec->cr_time >> 30;
-                gmtime_r(&secs, &ts);
-                printf(LPU64" %02d%-5s %02d:%02d:%02d.%06d %04d.%02d.%02d "
-                       "0x%x t="DFID, rec->cr_index, rec->cr_type,
-                       changelog_type2str(rec->cr_type),
-                       ts.tm_hour, ts.tm_min, ts.tm_sec,
-                       (int)(rec->cr_time & ((1<<30) - 1)),
-                       ts.tm_year + 1900, ts.tm_mon + 1, ts.tm_mday,
-                       rec->cr_flags & CLF_FLAGMASK, PFID(&rec->cr_tfid));
-                if (rec->cr_namelen)
-                        /* namespace rec includes parent and filename */
+               secs = rec->cr_time >> 30;
+               gmtime_r(&secs, &ts);
+               printf(LPU64" %02d%-5s %02d:%02d:%02d.%06d %04d.%02d.%02d "
+                      "0x%x t="DFID, rec->cr_index, rec->cr_type,
+                      changelog_type2str(rec->cr_type),
+                      ts.tm_hour, ts.tm_min, ts.tm_sec,
+                      (int)(rec->cr_time & ((1<<30) - 1)),
+                      ts.tm_year + 1900, ts.tm_mon + 1, ts.tm_mday,
+                      rec->cr_flags & CLF_FLAGMASK, PFID(&rec->cr_tfid));
+
+               if (rec->cr_flags & CLF_JOBID) {
+                       struct changelog_ext_jobid *jid =
+                               changelog_rec_jobid(rec);
+
+                       if (jid->cr_jobid[0] != '\0')
+                               printf(" j=%s", jid->cr_jobid);
+               }
+
+               if (rec->cr_namelen)
                        printf(" p="DFID" %.*s", PFID(&rec->cr_pfid),
-                               rec->cr_namelen, rec->cr_name);
-               if (fid_is_sane(&rec->cr_sfid))
-                       printf(" s="DFID" sp="DFID" %.*s",
-                               PFID(&rec->cr_sfid), PFID(&rec->cr_spfid),
-                               changelog_rec_snamelen(rec),
-                               changelog_rec_sname(rec));
+                              rec->cr_namelen, changelog_rec_name(rec));
+
+               if (rec->cr_flags & CLF_RENAME) {
+                       struct changelog_ext_rename *rnm =
+                               changelog_rec_rename(rec);
+
+                       if (!fid_is_zero(&rnm->cr_sfid))
+                               printf(" s="DFID" sp="DFID" %.*s",
+                                      PFID(&rnm->cr_sfid),
+                                      PFID(&rnm->cr_spfid),
+                                      (int)changelog_rec_snamelen(rec),
+                                      changelog_rec_sname(rec));
+               }
                printf("\n");
 
-                llapi_changelog_free(&rec);
+               llapi_changelog_free(&rec);
         }
 
         llapi_changelog_fini(&changelog_priv);
@@ -3251,6 +3446,10 @@ static int lfs_fid2path(int argc, char **argv)
 
        device = argv[optind++];
        path = calloc(1, PATH_MAX);
+       if (path == NULL) {
+               fprintf(stderr, "error: Not enough memory\n");
+               return -errno;
+       }
 
        rc = 0;
        while (optind < argc) {
@@ -3297,33 +3496,75 @@ static int lfs_fid2path(int argc, char **argv)
 
 static int lfs_path2fid(int argc, char **argv)
 {
-       char **path;
-       const char *sep = "";
-       lustre_fid fid;
-       int rc = 0;
+       struct option     long_opts[] = {
+               {"parents", no_argument, 0, 'p'},
+               {0, 0, 0, 0}
+       };
+       char            **path;
+       const char        short_opts[] = "p";
+       const char       *sep = "";
+       lustre_fid        fid;
+       int               rc = 0;
+       bool              show_parents = false;
 
-       if (argc < 2)
+       optind = 0;
+       while ((rc = getopt_long(argc, argv, short_opts,
+                                long_opts, NULL)) != -1) {
+               switch (rc) {
+               case 'p':
+                       show_parents = true;
+                       break;
+               default:
+                       fprintf(stderr, "error: %s: option '%s' unrecognized\n",
+                               argv[0], argv[optind - 1]);
+                       return CMD_HELP;
+               }
+       }
+
+       if (optind > argc - 1)
                return CMD_HELP;
-       else if (argc > 2)
+       else if (optind < argc - 1)
                sep = ": ";
 
-       path = argv + 1;
-       while (*path != NULL) {
-               int err = llapi_path2fid(*path, &fid);
+       rc = 0;
+       for (path = argv + optind; *path != NULL; path++) {
+               int err = 0;
+               if (!show_parents) {
+                       err = llapi_path2fid(*path, &fid);
+                       if (!err)
+                               printf("%s%s"DFID"\n",
+                                      *sep != '\0' ? *path : "", sep,
+                                      PFID(&fid));
+               } else {
+                       char            name[NAME_MAX + 1];
+                       unsigned int    linkno = 0;
+
+                       while ((err = llapi_path2parent(*path, linkno, &fid,
+                                               name, sizeof(name))) == 0) {
+                               if (*sep != '\0' && linkno == 0)
+                                       printf("%s%s", *path, sep);
+
+                               printf("%s"DFID"/%s", linkno != 0 ? "\t" : "",
+                                      PFID(&fid), name);
+                               linkno++;
+                       }
+
+                       /* err == -ENODATA is end-of-loop */
+                       if (linkno > 0 && err == -ENODATA) {
+                               printf("\n");
+                               err = 0;
+                       }
+               }
 
                if (err) {
-                       fprintf(stderr, "%s: can't get fid for %s: %s\n",
-                               argv[0], *path, strerror(-err));
+                       fprintf(stderr, "%s: can't get %sfid for %s: %s\n",
+                               argv[0], show_parents ? "parent " : "", *path,
+                               strerror(-err));
                        if (rc == 0) {
                                rc = err;
                                errno = -err;
                        }
-                       goto out;
                }
-               printf("%s%s"DFID"\n", *sep != '\0' ? *path : "", sep,
-                      PFID(&fid));
-out:
-               path++;
        }
 
        return rc;
@@ -3546,7 +3787,7 @@ static int lfs_hsm_action(int argc, char **argv)
                if ((hps == HPS_RUNNING) &&
                    (hua == HUA_ARCHIVE || hua == HUA_RESTORE))
                        printf("("LPX64 " bytes moved)\n", he.length);
-               else if ((he.offset + he.length) == OBD_OBJECT_EOF)
+               else if ((he.offset + he.length) == LUSTRE_EOF)
                        printf("(from "LPX64 " to EOF)\n", he.offset);
                else
                        printf("(from "LPX64 " to "LPX64")\n",
@@ -3587,6 +3828,13 @@ static int lfs_hsm_prepare_file(char *file, struct lu_fid *fid,
                fprintf(stderr, "Cannot stat %s: %s\n", file, strerror(errno));
                return -errno;
        }
+       /* Checking for regular file as archiving as posix copytool
+        * rejects archiving files other than regular files
+        */
+       if (!S_ISREG(st.st_mode)) {
+               fprintf(stderr, "error: \"%s\" is not a regular file\n", file);
+               return CMD_HELP;
+       }
        /* A request should be ... */
        if (*last_dev != st.st_dev && *last_dev != 0) {
                fprintf(stderr, "All files should be "
@@ -3720,12 +3968,13 @@ static int lfs_hsm_request(int argc, char **argv, int action)
                        /* If allocated buffer was too small, gets something
                         * bigger */
                        if (nbfile_alloc <= hur->hur_request.hr_itemcount) {
+                               ssize_t size;
                                nbfile_alloc = nbfile_alloc * 2 + 1;
                                oldhur = hur;
                                hur = llapi_hsm_user_request_alloc(nbfile_alloc,
                                                                   opaque_len);
                                if (hur == NULL) {
-                                       fprintf(stderr, "Cannot allocate "
+                                       fprintf(stderr, "hsm: cannot allocate "
                                                "the request: %s\n",
                                                strerror(errno));
                                        hur = oldhur;
@@ -3733,7 +3982,19 @@ static int lfs_hsm_request(int argc, char **argv, int action)
                                        fclose(fp);
                                        goto out_free;
                                }
-                               memcpy(hur, oldhur, hur_len(oldhur));
+                               size = hur_len(oldhur);
+                               if (size < 0) {
+                                       fprintf(stderr, "hsm: cannot allocate "
+                                               "%u files + %u bytes data\n",
+                                           oldhur->hur_request.hr_itemcount,
+                                           oldhur->hur_request.hr_data_len);
+                                       free(hur);
+                                       hur = oldhur;
+                                       rc = -E2BIG;
+                                       fclose(fp);
+                                       goto out_free;
+                               }
+                               memcpy(hur, oldhur, size);
                                free(oldhur);
                        }
 
@@ -3825,11 +4086,7 @@ int main(int argc, char **argv)
 
         setlinebuf(stdout);
 
-        ptl_initialize(argc, argv);
-        if (obd_initialize(argc, argv) < 0)
-                exit(2);
-
-        Parser_init("lfs > ", cmdlist);
+       Parser_init("lfs > ", cmdlist);
 
         if (argc > 1) {
                 rc = Parser_execarg(argc - 1, argv + 1, cmdlist);
@@ -3837,7 +4094,10 @@ int main(int argc, char **argv)
                 rc = Parser_commands();
         }
 
-        obd_finalize(argc, argv);
         return rc < 0 ? -rc : rc;
 }
 
+#ifdef _LUSTRE_IDL_H_
+/* Everything we need here should be included by lustreapi.h. */
+# error "lfs should not depend on lustre_idl.h"
+#endif /* _LUSTRE_IDL_H_ */