Whamcloud - gitweb
LU-4694 hsm: Cleanup codes about return value
[fs/lustre-release.git] / lustre / utils / lhsmtool_posix.c
index 0709264..2e18351 100644 (file)
  *
  * This particular tool can also import an existing HSM archive.
  */
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+#include <stdio.h>
 #include <stdint.h>
 #include <stdlib.h>
 #include <dirent.h>
@@ -78,6 +83,7 @@ struct options {
        unsigned long long       o_bandwidth;
        size_t                   o_chunk_size;
        enum ct_action           o_action;
+       char                    *o_event_fifo;
        char                    *o_mnt;
        char                    *o_hsm_root;
        char                    *o_src; /* for import, or rebind */
@@ -152,9 +158,10 @@ static void usage(const char *name, int rc)
        "into a Lustre filesystem.\n"
        " Usage:\n"
        "   %s [options] --import <src> <dst> <lustre_mount_point>\n"
-       "      import an archived subtree at\n"
-       "       <src> (relative to hsm_root) into the Lustre filesystem at\n"
-       "       <dst> (absolute)\n"
+       "      import an archived subtree from\n"
+       "       <src> (FID or relative path to hsm_root) into the Lustre\n"
+       "             filesystem at\n"
+       "       <dst> (absolute path)\n"
        "   %s [options] --rebind <old_FID> <new_FID> <lustre_mount_point>\n"
        "      rebind an entry in the HSM to a new FID\n"
        "       <old_FID> old FID the HSM entry is bound to\n"
@@ -164,16 +171,19 @@ static void usage(const char *name, int rc)
        "       each line of <list_file> consists of <old_FID> <new_FID>\n"
        "   %s [options] --max-sequence <fsname>\n"
        "       return the max fid sequence of archived files\n"
-       "   -A, --archive <#>        Archive number (repeatable)\n"
-       "   -p, --hsm-root <path>    Target HSM mount point\n"
-       "   -q, --quiet              Produce less verbose output\n"
-       "   -v, --verbose            Produce more verbose output\n"
-       "   -c, --chunk-size <sz>    I/O size used during data copy\n"
-       "                            (unit can be used, default is MB)\n"
-       "   --abort-on-error         Abort operation on major error\n"
-       "   --dry-run                Don't run, just show what would be done\n"
-       "   --bandwidth <bw>         Limit I/O bandwidth (unit can be used\n,"
-       "                            default is MB)\n",
+       "   --abort-on-error          Abort operation on major error\n"
+       "   -A, --archive <#>         Archive number (repeatable)\n"
+       "   -b, --bandwidth <bw>      Limit I/O bandwidth (unit can be used\n,"
+       "                             default is MB)\n"
+       "   --dry-run                 Don't run, just show what would be done\n"
+       "   -c, --chunk-size <sz>     I/O size used during data copy\n"
+       "                             (unit can be used, default is MB)\n"
+       "   -f, --event-fifo <path>   Write events stream to fifo\n"
+       "   -p, --hsm-root <path>     Target HSM mount point\n"
+       "   -q, --quiet               Produce less verbose output\n"
+       "   -u, --update-interval <s> Interval between progress reports sent\n"
+       "                             to Coordinator\n"
+       "   -v, --verbose             Produce more verbose output\n",
        cmd_name, cmd_name, cmd_name, cmd_name, cmd_name);
 
        exit(rc);
@@ -189,6 +199,8 @@ static int ct_parseopts(int argc, char * const *argv)
                {"chunk-size",     required_argument, NULL,                'c'},
                {"chunk_size",     required_argument, NULL,                'c'},
                {"daemon",         no_argument,       &opt.o_daemonize,     1},
+               {"event-fifo",     required_argument, NULL,                'f'},
+               {"event_fifo",     required_argument, NULL,                'f'},
                {"dry-run",        no_argument,       &opt.o_dry_run,       1},
                {"help",           no_argument,       NULL,                'h'},
                {"hsm-root",       required_argument, NULL,                'p'},
@@ -204,7 +216,8 @@ static int ct_parseopts(int argc, char * const *argv)
                {"no_xattr",       no_argument,       &opt.o_copy_xattrs,   0},
                {"quiet",          no_argument,       NULL,                'q'},
                {"rebind",         no_argument,       NULL,                'r'},
-               {"report",         required_argument, &opt.o_report_int,    0},
+               {"update-interval", required_argument,  NULL,              'u'},
+               {"update_interval", required_argument,  NULL,              'u'},
                {"verbose",        no_argument,       NULL,                'v'},
                {0, 0, 0, 0}
        };
@@ -213,7 +226,7 @@ static int ct_parseopts(int argc, char * const *argv)
        unsigned long long       unit;
 
        optind = 0;
-       while ((c = getopt_long(argc, argv, "A:b:c:hiMp:qrv",
+       while ((c = getopt_long(argc, argv, "A:b:c:f:hiMp:qru:v",
                                long_opts, NULL)) != -1) {
                switch (c) {
                case 'A':
@@ -241,6 +254,9 @@ static int ct_parseopts(int argc, char * const *argv)
                        else
                                opt.o_bandwidth = value;
                        break;
+               case 'f':
+                       opt.o_event_fifo = optarg;
+                       break;
                case 'h':
                        usage(argv[0], 0);
                case 'i':
@@ -258,6 +274,15 @@ static int ct_parseopts(int argc, char * const *argv)
                case 'r':
                        opt.o_action = CA_REBIND;
                        break;
+               case 'u':
+                       opt.o_report_int = atoi(optarg);
+                       if (opt.o_report_int < 0) {
+                               rc = -EINVAL;
+                               CT_ERROR(rc, "bad value for -%c '%s'", c,
+                                        optarg);
+                               return rc;
+                       }
+                       break;
                case 'v':
                        opt.o_verbose++;
                        break;
@@ -341,6 +366,9 @@ static int ct_mkdir_p(const char *path)
        int      rc;
 
        ptr = strdup(path);
+       if (ptr == NULL)
+               return -errno;
+
        saved = ptr;
        while (*ptr == '/')
                ptr++;
@@ -460,8 +488,8 @@ static int ct_restore_stripe(const char *src, const char *dst, int dst_fd,
        rc = fsetxattr(dst_fd, XATTR_LUSTRE_LOV, lovea, lovea_size,
                       XATTR_CREATE);
        if (rc < 0) {
-               CT_ERROR(errno, "cannot set lov EA on '%s'", dst);
                rc = -errno;
+               CT_ERROR(rc, "cannot set lov EA on '%s'", dst);
        }
 
        return rc;
@@ -507,7 +535,7 @@ static int ct_copy_data(struct hsm_copyaction_private *hcp, const char *src,
        __u64                    offset = hai->hai_extent.offset;
        struct stat              src_st;
        struct stat              dst_st;
-       char                    *buf;
+       char                    *buf = NULL;
        __u64                    write_total = 0;
        __u64                    length;
        time_t                   last_print_time = time(NULL);
@@ -515,10 +543,6 @@ static int ct_copy_data(struct hsm_copyaction_private *hcp, const char *src,
 
        CT_TRACE("going to copy data from '%s' to '%s'", src, dst);
 
-       buf = malloc(opt.o_chunk_size);
-       if (buf == NULL)
-               return -ENOMEM;
-
        if (fstat(src_fd, &src_st) < 0) {
                rc = -errno;
                CT_ERROR(rc, "cannot stat '%s'", src);
@@ -545,16 +569,19 @@ static int ct_copy_data(struct hsm_copyaction_private *hcp, const char *src,
 
        rc = lseek(src_fd, hai->hai_extent.offset, SEEK_SET);
        if (rc < 0) {
-               CT_ERROR(errno,
+               rc = -errno;
+               CT_ERROR(rc,
                         "cannot seek for read to "LPU64" (len %jd) in '%s'",
                         hai->hai_extent.offset, (intmax_t)src_st.st_size, src);
-               rc = -errno;
-               goto out;
+               return rc;
        }
 
+       /* Don't read beyond a given extent */
+       length = min(hai->hai_extent.length, src_st.st_size);
+
        he.offset = offset;
        he.length = 0;
-       rc = llapi_hsm_action_progress(hcp, &he, 0);
+       rc = llapi_hsm_action_progress(hcp, &he, length, 0);
        if (rc < 0) {
                /* Action has been canceled or something wrong
                 * is happening. Stop copying data. */
@@ -564,8 +591,12 @@ static int ct_copy_data(struct hsm_copyaction_private *hcp, const char *src,
        }
 
        errno = 0;
-       /* Don't read beyond a given extent */
-       length = min(hai->hai_extent.length, src_st.st_size);
+
+       buf = malloc(opt.o_chunk_size);
+       if (buf == NULL) {
+               rc = -ENOMEM;
+               goto out;
+       }
 
        CT_DEBUG("Going to copy "LPU64" bytes %s -> %s\n", length, src, dst);
 
@@ -604,7 +635,7 @@ static int ct_copy_data(struct hsm_copyaction_private *hcp, const char *src,
                        last_print_time = time(0);
                        CT_TRACE("%%"LPU64" ", 100 * write_total / length);
                        he.length = write_total;
-                       rc = llapi_hsm_action_progress(hcp, &he, 0);
+                       rc = llapi_hsm_action_progress(hcp, &he, length, 0);
                        if (rc < 0) {
                                /* Action has been canceled or something wrong
                                 * is happening. Stop copying data. */
@@ -631,14 +662,14 @@ out:
                rc = ftruncate(dst_fd, src_st.st_size);
                if (rc < 0) {
                        rc = -errno;
-                       CT_ERROR(rc,
-                                "cannot truncate '%s' to size %jd",
+                       CT_ERROR(rc, "cannot truncate '%s' to size %jd",
                                 dst, (intmax_t)src_st.st_size);
                        err_major++;
                }
        }
 
-       free(buf);
+       if (buf != NULL)
+               free(buf);
 
        return rc;
 }
@@ -767,8 +798,9 @@ static int ct_begin(struct hsm_copyaction_private **phcp,
 static int ct_fini(struct hsm_copyaction_private **phcp,
                   const struct hsm_action_item *hai, int hp_flags, int ct_rc)
 {
-       char    lstr[PATH_MAX];
-       int     rc;
+       struct hsm_copyaction_private   *hcp;
+       char                             lstr[PATH_MAX];
+       int                              rc;
 
        CT_TRACE("Action completed, notifying coordinator "
                 "cookie="LPX64", FID="DFID", hp_flags=%d err=%d",
@@ -776,6 +808,17 @@ static int ct_fini(struct hsm_copyaction_private **phcp,
                 hp_flags, -ct_rc);
 
        ct_path_lustre(lstr, sizeof(lstr), opt.o_mnt, &hai->hai_fid);
+
+       if (phcp == NULL || *phcp == NULL) {
+               rc = llapi_hsm_action_begin(&hcp, ctdata, hai, -1, 0, true);
+               if (rc < 0) {
+                       CT_ERROR(rc, "llapi_hsm_action_begin() on '%s' failed",
+                                lstr);
+                       return rc;
+               }
+               phcp = &hcp;
+       }
+
        rc = llapi_hsm_action_end(phcp, &hai->hai_extent, hp_flags, abs(ct_rc));
        if (rc == -ECANCELED)
                CT_ERROR(rc, "completed action on '%s' has been canceled: "
@@ -816,7 +859,7 @@ static int ct_archive(const struct hsm_action_item *hai, const long hal_flags)
        if (hai->hai_extent.length == -1) {
                /* whole file, write it to tmp location and atomically
                 * replace old archived file */
-               strncat(dst, "_tmp", sizeof(dst) - strlen(dst) - 1);
+               strlcat(dst, "_tmp", sizeof(dst));
                /* we cannot rely on the same test because ct_copy_data()
                 * updates hai_extent.length */
                rename_needed = true;
@@ -837,7 +880,7 @@ static int ct_archive(const struct hsm_action_item *hai, const long hal_flags)
 
        src_fd = llapi_hsm_action_get_fd(hcp);
        if (src_fd < 0) {
-               rc = -errno;
+               rc = src_fd;
                CT_ERROR(rc, "cannot open '%s' for read", src);
                goto fini_major;
        }
@@ -1025,8 +1068,7 @@ out:
        if (!(dst_fd < 0))
                close(dst_fd);
 
-       if (hcp != NULL)
-               rc = ct_fini(&hcp, hai, hp_flags, rcf);
+       rc = ct_fini(&hcp, hai, hp_flags, rcf);
 
        return rc;
 }
@@ -1097,6 +1139,11 @@ static int ct_restore(const struct hsm_action_item *hai, const long hal_flags)
        }
 
        dst_fd = llapi_hsm_action_get_fd(hcp);
+       if (dst_fd < 0) {
+               rc = dst_fd;
+               CT_ERROR(rc, "cannot open '%s' for write", dst);
+               goto fini;
+       }
 
        if (set_lovea) {
                /* the layout cannot be allocated through .fid so we have to
@@ -1123,8 +1170,7 @@ static int ct_restore(const struct hsm_action_item *hai, const long hal_flags)
        CT_TRACE("data restore from '%s' to '%s' done", src, dst);
 
 fini:
-       if (hcp != NULL)
-               rc = ct_fini(&hcp, hai, hp_flags, rc);
+       rc = ct_fini(&hcp, hai, hp_flags, rc);
 
        /* object swaping is done by cdt at copy end, so close of volatile file
         * cannot be done before */
@@ -1164,7 +1210,7 @@ static int ct_remove(const struct hsm_action_item *hai, const long hal_flags)
                goto fini;
        }
 
-       strncat(dst, ".lov", sizeof(dst) - strlen(dst) - 1);
+       strlcat(dst, ".lov", sizeof(dst));
        rc = unlink(dst);
        if (rc < 0) {
                rc = -errno;
@@ -1174,23 +1220,7 @@ static int ct_remove(const struct hsm_action_item *hai, const long hal_flags)
        }
 
 fini:
-       if (hcp != NULL)
-               rc = ct_fini(&hcp, hai, 0, rc);
-
-       return rc;
-}
-
-static int ct_report_error(const struct hsm_action_item *hai, int flags,
-                          int errval)
-{
-       struct hsm_copyaction_private   *hcp;
-       int                              rc;
-
-       rc = llapi_hsm_action_begin(&hcp, ctdata, hai, -1, 0, true);
-       if (rc < 0)
-               return rc;
-
-       rc = llapi_hsm_action_end(&hcp, &hai->hai_extent, flags, abs(errval));
+       rc = ct_fini(&hcp, hai, 0, rc);
 
        return rc;
 }
@@ -1243,7 +1273,7 @@ static int ct_process_item(struct hsm_action_item *hai, const long hal_flags)
                CT_ERROR(rc, "unknown action %d, on '%s'", hai->hai_action,
                         opt.o_mnt);
                err_minor++;
-               ct_report_error(hai, 0, rc);
+               ct_fini(NULL, hai, 0, rc);
        }
 
        return 0;
@@ -1359,29 +1389,53 @@ static int ct_import_one(const char *src, const char *dst)
 static char *path_concat(const char *dirname, const char *basename)
 {
        char    *result;
-       int      dirlen = strlen(dirname);
+       int      rc;
 
-       result = malloc(dirlen + strlen(basename) + 2);
-       if (result == NULL)
+       rc = asprintf(&result, "%s/%s", dirname, basename);
+       if (rc < 0)
                return NULL;
 
-       memcpy(result, dirname, dirlen);
-       result[dirlen] = '/';
-       strcpy(result + dirlen + 1, basename);
-
        return result;
 }
 
+static int ct_import_fid(const lustre_fid *import_fid)
+{
+       char    fid_path[PATH_MAX];
+       int     rc;
+
+       ct_path_lustre(fid_path, sizeof(fid_path), opt.o_mnt, import_fid);
+       rc = access(fid_path, F_OK);
+       if (rc == 0 || errno != ENOENT) {
+               rc = (errno == 0) ? -EEXIST : -errno;
+               CT_ERROR(rc, "cannot import '"DFID"'", PFID(import_fid));
+               return rc;
+       }
+
+       ct_path_archive(fid_path, sizeof(fid_path), opt.o_hsm_root,
+                       import_fid);
+
+       CT_TRACE("Resolving "DFID" to %s", PFID(import_fid), fid_path);
+
+       return ct_import_one(fid_path, opt.o_dst);
+}
+
 static int ct_import_recurse(const char *relpath)
 {
        DIR             *dir;
        struct dirent    ent, *cookie = NULL;
        char            *srcpath, *newpath;
+       lustre_fid       import_fid;
        int              rc;
 
        if (relpath == NULL)
                return -EINVAL;
 
+       /* Is relpath a FID? In which case SFID should expand to three
+        * elements. */
+       rc = sscanf(relpath, SFID, RFID(&import_fid));
+       if (rc == 3)
+               return ct_import_fid(&import_fid);
+
        srcpath = path_concat(opt.o_hsm_root, relpath);
        if (srcpath == NULL) {
                err_major++;
@@ -1483,10 +1537,10 @@ static int ct_rebind_one(const lustre_fid *old_fid, const lustre_fid *new_fid)
                        return -errno;
                }
                /* rename lov file */
-               strncat(src, ".lov", sizeof(src) - strlen(src) - 1);
-               strncat(dst, ".lov", sizeof(dst) - strlen(dst) - 1);
+               strlcat(src, ".lov", sizeof(src));
+               strlcat(dst, ".lov", sizeof(dst));
                if (rename(src, dst))
-                       CT_ERROR(errno, "cannot '%s' rename to '%s'", src, dst);
+                       CT_ERROR(errno, "cannot rename '%s' to '%s'", src, dst);
 
        }
        return 0;
@@ -1644,24 +1698,31 @@ out:
 
 static int ct_max_sequence(void)
 {
-       int   rc, i;
-       char  path[PATH_MAX];
-       __u64 seq = 0;
-       __u16 subseq;
+       int     rc, i;
+       char    path[PATH_MAX];
+       __u64   seq = 0;
+       __u16   subseq;
 
-       strncpy(path, opt.o_hsm_root, sizeof(path));
+       strlcpy(path, opt.o_hsm_root, sizeof(path));
        /* FID sequence is stored in top-level directory names:
         * hsm_root/16bits (high weight)/16 bits/16 bits/16 bits (low weight).
         */
        for (i = 0; i < 4; i++) {
+               size_t path_len;
+
                rc = ct_dir_level_max(path, &subseq);
                if (rc != 0)
                        return rc;
                seq |= ((__u64)subseq << ((3 - i) * 16));
-               sprintf(path + strlen(path), "/%04x", subseq);
+               path_len = strlen(path);
+               rc = snprintf(path + path_len, sizeof(path) - path_len,
+                             "/%04x", subseq);
+               if (rc >= (sizeof(path) - path_len))
+                       return -E2BIG;
+               path[sizeof(path) - 1] = '\0';
        }
 
-       printf("max_sequence: %016Lx\n", seq);
+       printf("max_sequence: "LPX64"\n", seq);
 
        return 0;
 }
@@ -1693,6 +1754,15 @@ static int ct_run(void)
 
        setbuf(stdout, NULL);
 
+       if (opt.o_event_fifo != NULL) {
+               rc = llapi_hsm_register_event_fifo(opt.o_event_fifo);
+               if (rc < 0) {
+                       CT_ERROR(rc, "failed to register event fifo");
+                       return rc;
+               }
+               llapi_error_callback_set(llapi_hsm_log_error);
+       }
+
        rc = llapi_hsm_copytool_register(&ctdata, opt.o_mnt, 0,
                                         opt.o_archive_cnt, opt.o_archive_id);
        if (rc < 0) {
@@ -1731,7 +1801,8 @@ static int ct_run(void)
                         hal->hal_fsname, hal->hal_archive_id, hal->hal_count);
 
                if (strcmp(hal->hal_fsname, fs_name) != 0) {
-                       CT_ERROR(EINVAL, "'%s' invalid fs name, expecting: %s",
+                       rc = -EINVAL;
+                       CT_ERROR(rc, "'%s' invalid fs name, expecting: %s",
                                 hal->hal_fsname, fs_name);
                        err_major++;
                        if (opt.o_abort_on_error)
@@ -1759,13 +1830,13 @@ static int ct_run(void)
                        hai = hai_next(hai);
                }
 
-               llapi_hsm_action_list_free(&hal);
-
                if (opt.o_abort_on_error && err_major)
                        break;
        }
 
        llapi_hsm_copytool_unregister(&ctdata);
+       if (opt.o_event_fifo != NULL)
+               llapi_hsm_unregister_event_fifo(opt.o_event_fifo);
 
        return rc;
 }
@@ -1788,7 +1859,7 @@ static int ct_setup(void)
        if (rc < 0) {
                CT_ERROR(rc, "cannot find a Lustre filesystem mounted at '%s'",
                         opt.o_mnt);
-               return -rc;
+               return rc;
        }
 
        return rc;
@@ -1814,14 +1885,16 @@ int main(int argc, char **argv)
 {
        int     rc;
 
-       strncpy(cmd_name, basename(argv[0]), sizeof(cmd_name));
+       strlcpy(cmd_name, basename(argv[0]), sizeof(cmd_name));
        rc = ct_parseopts(argc, argv);
        if (rc < 0) {
                CT_WARN("try '%s --help' for more information", cmd_name);
                return -rc;
        }
 
-       ct_setup();
+       rc = ct_setup();
+       if (rc < 0)
+               goto error_cleanup;
 
        switch (opt.o_action) {
        case CA_IMPORT:
@@ -1843,6 +1916,7 @@ int main(int argc, char **argv)
                         " rc=%d (%s)", err_major, err_minor, rc,
                         strerror(-rc));
 
+error_cleanup:
        ct_cleanup();
 
        return -rc;