Whamcloud - gitweb
LU-14090 mgs: no local logs flag
[fs/lustre-release.git] / lustre / utils / mkfs_lustre.c
index eb5a21a..b1f0f58 100644 (file)
  *
  * You should have received a copy of the GNU General Public License
  * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
  *
  * GPL HEADER END
  */
@@ -27,7 +23,7 @@
  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
  *
- * Copyright (c) 2011, Whamcloud, Inc.
+ * Copyright (c) 2011, 2017, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -53,6 +49,7 @@
 #include <fcntl.h>
 #include <stdarg.h>
 #include <mntent.h>
+#include <stdbool.h>
 
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <getopt.h>
 #include <limits.h>
 #include <ctype.h>
+#include <linux/lnet/nidstr.h>
+#include <linux/lnet/lnetctl.h>
+#include <linux/lustre/lustre_user.h>
+#include <linux/lustre/lustre_ver.h>
 
-#ifdef __linux__
-/* libcfs.h is not really needed here, but on SLES10/PPC, fs.h includes idr.h
- * which requires BITS_PER_LONG to be defined */
-#include <libcfs/libcfs.h>
-#ifndef BLKGETSIZE64
-#include <linux/fs.h> /* for BLKGETSIZE64 */
-#endif
-#include <linux/version.h>
-#endif
-#include <lustre_disk.h>
-#include <lustre_param.h>
-#include <lnet/lnetctl.h>
-#include <lustre_ver.h>
 #include "mount_utils.h"
 
-#ifndef PATH_MAX
-#define PATH_MAX 4096
-#endif
-
-#define MAX_LOOP_DEVICES 16
-#define L_BLOCK_SIZE 4096
-#define INDEX_UNASSIGNED 0xFFFF
-#define MO_IS_LOOP     0x01
-#define MO_FORCEFORMAT 0x02
-
-/* used to describe the options to format the lustre disk, not persistent */
-struct mkfs_opts {
-        struct lustre_disk_data mo_ldd; /* to be written in MOUNT_DATA_FILE */
-        char  mo_device[128];           /* disk device name */
-        char  mo_loopdev[128];          /* in case a loop dev is needed */
-        char  mo_mkfsopts[512];         /* options to the backing-store mkfs */
-        __u64 mo_device_sz;             /* in KB */
-        int   mo_stripe_count;
-        int   mo_flags;
-        int   mo_mgs_failnodes;
-};
-
 char *progname;
 int verbose = 1;
-static int print_only = 0;
-static int failover = 0;
-static int upgrade_to_18 = 0;
+int version;
+static int print_only;
 
-void usage(FILE *out)
-{
-        fprintf(out, "%s v"LUSTRE_VERSION_STRING"\n", progname);
-        fprintf(out, "usage: %s <target types> [options] <device>\n", progname);
-        fprintf(out,
-                "\t<device>:block device or file (e.g /dev/sda or /tmp/ost1)\n"
-                "\ttarget types:\n"
-                "\t\t--ost: object storage, mutually exclusive with mdt,mgs\n"
-                "\t\t--mdt: metadata storage, mutually exclusive with ost\n"
-                "\t\t--mgs: configuration management service - one per site\n"
-                "\toptions (in order of popularity):\n"
-                "\t\t--mgsnode=<nid>[,<...>] : NID(s) of a remote mgs node\n"
-                "\t\t\trequired for all targets other than the mgs node\n"
-                "\t\t--fsname=<filesystem_name> : default is 'lustre'\n"
-                "\t\t--failnode=<nid>[,<...>] : NID(s) of a failover partner\n"
-                "\t\t\tcannot be used with --servicenode\n"
-                "\t\t--servicenode=<nid>[,<...>] : NID(s) of all service partners\n"
-                "\t\t\ttreat all nodes as equal service node, cannot be used with --failnode\n"
-                "\t\t--param <key>=<value> : set a permanent parameter\n"
-                "\t\t\te.g. --param sys.timeout=40\n"
-                "\t\t\t     --param lov.stripesize=2M\n"
-                "\t\t--index=#N : target index (i.e. ost index within lov)\n"
-                "\t\t--comment=<user comment>: arbitrary string (%d bytes)\n"
-                "\t\t--mountfsoptions=<opts> : permanent mount options\n"
-                "\t\t--network=<net>[,<...>] : restrict OST/MDT to network(s)\n"
-#ifndef TUNEFS
-                "\t\t--backfstype=<fstype> : backing fs type (ext3, ldiskfs)\n"
-                "\t\t--device-size=#N(KB) : device size for loop devices\n"
-                "\t\t--mkfsoptions=<opts> : format options\n"
-                "\t\t--reformat: overwrite an existing disk\n"
-                "\t\t--stripe-count-hint=#N : for optimizing MDT inode size\n"
-                "\t\t--iam-dir: use IAM directory format, not ext3 compatible\n"
+#ifdef HAVE_LDISKFS_OSD
+#define FSLIST_LDISKFS "ldiskfs"
+#define HAVE_FSLIST
 #else
-                "\t\t--erase-params : erase all old parameter settings\n"
-                "\t\t--nomgs: turn off MGS service on this MDT\n"
-                "\t\t--writeconf: erase all config logs for this fs.\n"
-#endif
-                "\t\t--dryrun: just report what we would do; "
-                "don't write to disk\n"
-                "\t\t--verbose : e.g. show mkfs progress\n"
-                "\t\t--quiet\n",
-                (int)sizeof(((struct lustre_disk_data *)0)->ldd_userdata));
-        return;
-}
-
-#define vprint if (verbose > 0) printf
-#define verrprint if (verbose >= 0) printf
-
-/*================ utility functions =====================*/
-
-char *strscat(char *dst, char *src, int buflen) {
-        dst[buflen - 1] = 0;
-        if (strlen(dst) + strlen(src) >= buflen) {
-                fprintf(stderr, "string buffer overflow (max %d): '%s' + '%s'"
-                        "\n", buflen, dst, src);
-                exit(EOVERFLOW);
-        }
-        return strcat(dst, src);
-
-}
-
-char *strscpy(char *dst, char *src, int buflen) {
-        dst[0] = 0;
-        return strscat(dst, src, buflen);
-}
-
-inline unsigned int
-dev_major (unsigned long long int __dev)
-{
-        return ((__dev >> 8) & 0xfff) | ((unsigned int) (__dev >> 32) & ~0xfff);
-}
-
-inline unsigned int
-dev_minor (unsigned long long int __dev)
-{
-        return (__dev & 0xff) | ((unsigned int) (__dev >> 12) & ~0xff);
-}
-
-int get_os_version()
-{
-        static int version = 0;
-
-        if (!version) {
-                int fd;
-                char release[4] = "";
-
-                fd = open("/proc/sys/kernel/osrelease", O_RDONLY);
-                if (fd < 0) {
-                        fprintf(stderr, "%s: Warning: Can't resolve kernel "
-                                "version, assuming 2.6\n", progname);
-                } else {
-                        if (read(fd, release, 4) < 0) {
-                                fprintf(stderr, "reading from /proc/sys/kernel"
-                                        "/osrelease: %s\n", strerror(errno));
-                                close(fd);
-                                exit(-1);
-                        }
-                        close(fd);
-                }
-                if (strncmp(release, "2.4.", 4) == 0)
-                        version = 24;
-                else
-                        version = 26;
-        }
-        return version;
-}
-
-static int check_mtab_entry(char *spec)
-{
-        FILE *fp;
-        struct mntent *mnt;
-
-        fp = setmntent(MOUNTED, "r");
-        if (fp == NULL)
-                return(0);
-
-        while ((mnt = getmntent(fp)) != NULL) {
-                if (strcmp(mnt->mnt_fsname, spec) == 0) {
-                        endmntent(fp);
-                        fprintf(stderr, "%s: according to %s %s is "
-                                "already mounted on %s\n",
-                                progname, MOUNTED, spec, mnt->mnt_dir);
-                        return(EEXIST);
-                }
-        }
-        endmntent(fp);
-
-        return(0);
-}
-
-/*============ disk dev functions ===================*/
-
-/* Setup a file in the first unused loop_device */
-int loop_setup(struct mkfs_opts *mop)
-{
-        char loop_base[20];
-        char l_device[64];
-        int i, ret = 0;
-
-        /* Figure out the loop device names */
-        if (!access("/dev/loop0", F_OK | R_OK)) {
-                strcpy(loop_base, "/dev/loop\0");
-        } else if (!access("/dev/loop/0", F_OK | R_OK)) {
-                strcpy(loop_base, "/dev/loop/\0");
-        } else {
-                fprintf(stderr, "%s: can't access loop devices\n", progname);
-                return EACCES;
-        }
-
-        /* Find unused loop device */
-        for (i = 0; i < MAX_LOOP_DEVICES; i++) {
-                char cmd[PATH_MAX];
-                int cmdsz = sizeof(cmd);
-
-                sprintf(l_device, "%s%d", loop_base, i);
-                if (access(l_device, F_OK | R_OK))
-                        break;
-                snprintf(cmd, cmdsz, "losetup %s > /dev/null 2>&1", l_device);
-                ret = system(cmd);
-
-                /* losetup gets 1 (ret=256) for non-set-up device */
-                if (ret) {
-                        /* Set up a loopback device to our file */
-                        snprintf(cmd, cmdsz, "losetup %s %s", l_device,
-                                 mop->mo_device);
-                        ret = run_command(cmd, cmdsz);
-                        if (ret == 256)
-                                /* someone else picked up this loop device
-                                 * behind our back */
-                                continue;
-                        if (ret) {
-                                fprintf(stderr, "%s: error %d on losetup: %s\n",
-                                        progname, ret, strerror(ret));
-                                return ret;
-                        }
-                        strscpy(mop->mo_loopdev, l_device,
-                                sizeof(mop->mo_loopdev));
-                        return ret;
-                }
-        }
-
-        fprintf(stderr, "%s: out of loop devices!\n", progname);
-        return EMFILE;
-}
-
-int loop_cleanup(struct mkfs_opts *mop)
-{
-        char cmd[150];
-        int ret = 1;
-        if ((mop->mo_flags & MO_IS_LOOP) && *mop->mo_loopdev) {
-                sprintf(cmd, "losetup -d %s", mop->mo_loopdev);
-                ret = run_command(cmd, sizeof(cmd));
-        }
-        return ret;
-}
-
-/* return canonicalized absolute pathname, even if the target file does not
- * exist, unlike realpath */
-static char *absolute_path(char *devname)
-{
-        char  buf[PATH_MAX + 1];
-        char *path;
-        char *ptr;
-
-        path = malloc(PATH_MAX + 1);
-        if (path == NULL)
-                return NULL;
-
-        if (devname[0] != '/') {
-                if (getcwd(buf, sizeof(buf) - 1) == NULL)
-                        return NULL;
-                strcat(buf, "/");
-                strcat(buf, devname);
-        } else {
-                strcpy(buf, devname);
-        }
-        /* truncate filename before calling realpath */
-        ptr = strrchr(buf, '/');
-        if (ptr == NULL) {
-                free(path);
-                return NULL;
-        }
-        *ptr = '\0';
-        if (path != realpath(buf, path)) {
-                free(path);
-                return NULL;
-        }
-        /* add the filename back */
-        strcat(path, "/");
-        strcat(path, ptr + 1);
-        return path;
-}
-
-/* Determine if a device is a block device (as opposed to a file) */
-int is_block(char* devname)
-{
-        struct stat st;
-        int         ret = 0;
-        char       *devpath;
-
-        devpath = absolute_path(devname);
-        if (devpath == NULL) {
-                fprintf(stderr, "%s: failed to resolve path to %s\n",
-                        progname, devname);
-                return -1;
-        }
-
-        ret = access(devname, F_OK);
-        if (ret != 0) {
-                if (strncmp(devpath, "/dev/", 5) == 0) {
-                    /* nobody sane wants to create a loopback file under
-                     * /dev. Let's just report the device doesn't exist */
-                    fprintf(stderr, "%s: %s apparently does not exist\n",
-                            progname, devpath);
-                    ret = -1;
-                    goto out;
-                }
-                ret = 0;
-                goto out;
-        }
-        ret = stat(devpath, &st);
-        if (ret != 0) {
-                fprintf(stderr, "%s: cannot stat %s\n", progname, devpath);
-                goto out;
-        }
-        ret = S_ISBLK(st.st_mode);
-out:
-        free(devpath);
-        return ret;
-}
-
-__u64 get_device_size(char* device)
-{
-        int ret, fd;
-        __u64 size = 0;
-
-        fd = open(device, O_RDONLY);
-        if (fd < 0) {
-                fprintf(stderr, "%s: cannot open %s: %s\n",
-                        progname, device, strerror(errno));
-                return 0;
-        }
-
-#ifdef BLKGETSIZE64
-        /* size in bytes. bz5831 */
-        ret = ioctl(fd, BLKGETSIZE64, (void*)&size);
+ #define FSLIST_LDISKFS ""
+#endif /* HAVE_LDISKFS_OSD */
+#ifdef HAVE_ZFS_OSD
+ #ifdef HAVE_FSLIST
+   #define FSLIST_ZFS "|zfs"
+ #else
+  #define FSLIST_ZFS "zfs"
+  #define HAVE_FSLIST
+ #endif
 #else
-        {
-                __u32 lsize = 0;
-                /* size in blocks */
-                ret = ioctl(fd, BLKGETSIZE, (void*)&lsize);
-                size = (__u64)lsize * 512;
-        }
-#endif
-        close(fd);
-        if (ret < 0) {
-                fprintf(stderr, "%s: size ioctl failed: %s\n",
-                        progname, strerror(errno));
-                return 0;
-        }
-
-        vprint("device size = "LPU64"MB\n", size >> 20);
-        /* return value in KB */
-        return size >> 10;
-}
+ #define FSLIST_ZFS ""
+#endif /* HAVE_ZFS_OSD */
 
-int loop_format(struct mkfs_opts *mop)
-{
-        int fd;
-
-        if (mop->mo_device_sz == 0) {
-                fatal();
-                fprintf(stderr, "loop device requires a --device-size= "
-                        "param\n");
-                return EINVAL;
-        }
-
-        fd = creat(mop->mo_device, S_IRUSR|S_IWUSR);
-        if (fd < 0) {
-                fatal();
-                fprintf(stderr, "%s: Unable to create backing store: %s\n",
-                        progname, strerror(errno));
-                return errno;
-        }
-
-        if (ftruncate(fd, mop->mo_device_sz * 1024) != 0) {
-                close(fd);
-                fatal();
-                fprintf(stderr, "%s: Unable to truncate backing store: %s\n",
-                        progname, strerror(errno));
-                return errno;
-        }
-
-        close(fd);
-        return 0;
-}
-
-/* Display the need for the latest e2fsprogs to be installed. make_backfs
- * indicates if the caller is make_lustre_backfs() or not. */
-static void disp_old_e2fsprogs_msg(const char *feature, int make_backfs)
-{
-        static int msg_displayed;
-
-        if (msg_displayed) {
-                fprintf(stderr, "WARNING: %s does not support %s "
-                        "feature.\n\n", E2FSPROGS, feature);
-                return;
-        }
-
-        msg_displayed++;
-
-        fprintf(stderr, "WARNING: The %s package currently installed on "
-                "your system does not support \"%s\" feature.\n",
-                E2FSPROGS, feature);
-#if !(HAVE_LDISKFSPROGS)
-        fprintf(stderr, "Please install the latest version of e2fsprogs from\n"
-                "http://downloads.whamcloud.com/public/e2fsprogs/latest/\n"
-                "to enable this feature.\n");
+#ifndef HAVE_FSLIST
+ #error "no backing OSD types (ldiskfs or ZFS) are configured"
 #endif
-        if (make_backfs)
-                fprintf(stderr, "Feature will not be enabled until %s"
-                        "is updated and '%s -O %s %%{device}' "
-                        "is run.\n\n", E2FSPROGS, TUNE2FS, feature);
-}
 
-/* Check whether the file exists in the device */
-static int file_in_dev(char *file_name, char *dev_name)
-{
-        FILE *fp;
-        char debugfs_cmd[256];
-        unsigned int inode_num;
-        int i;
-
-        /* Construct debugfs command line. */
-        snprintf(debugfs_cmd, sizeof(debugfs_cmd),
-                "%s -c -R 'stat %s' '%s' 2>&1 | egrep '(Inode|unsupported)'",
-                DEBUGFS, file_name, dev_name);
-
-        fp = popen(debugfs_cmd, "r");
-        if (!fp) {
-                fprintf(stderr, "%s: %s\n", progname, strerror(errno));
-                return 0;
-        }
-
-        if (fscanf(fp, "Inode: %u", &inode_num) == 1) { /* exist */
-                pclose(fp);
-                return 1;
-        }
-        i = fread(debugfs_cmd, 1, sizeof(debugfs_cmd), fp);
-        if (i) {
-                debugfs_cmd[i] = 0;
-                fprintf(stderr, "%s", debugfs_cmd);
-                if (strstr(debugfs_cmd, "unsupported feature")) {
-                          disp_old_e2fsprogs_msg("an unknown", 0);
-                }
-                pclose(fp);
-                return -1;
-        }
-        pclose(fp);
-        return 0;
-}
-
-/* Check whether the device has already been used with lustre */
-static int is_lustre_target(struct mkfs_opts *mop)
-{
-        int rc;
+#define FSLIST FSLIST_LDISKFS FSLIST_ZFS
 
-        vprint("checking for existing Lustre data: ");
-
-        if ((rc = file_in_dev(MOUNT_DATA_FILE, mop->mo_device))) {
-                vprint("found %s\n",
-                       (rc == 1) ? MOUNT_DATA_FILE : "extents");
-                 /* in the -1 case, 'extents' means this really IS a lustre
-                    target */
-                return rc;
-        }
-
-        if ((rc = file_in_dev(LAST_RCVD, mop->mo_device))) {
-                vprint("found %s\n", LAST_RCVD);
-                return rc;
-        }
-
-        vprint("not found\n");
-        return 0; /* The device is not a lustre target. */
-}
-
-/* Check if a certain feature is supported by e2fsprogs.
- * Firstly we try to use "debugfs supported_features" command to check if
- * the feature is supported. If this fails we try to set this feature with
- * mke2fs to check for its support. */
-static int is_e2fsprogs_feature_supp(const char *feature)
-{
-        static char supp_features[4096] = "";
-        FILE *fp;
-        char cmd[PATH_MAX];
-        char imgname[] = "/tmp/test-img-XXXXXX";
-        int fd = -1;
-        int ret = 1;
-
-        if (supp_features[0] == '\0') {
-                snprintf(cmd, sizeof(cmd), "%s -c -R supported_features 2>&1",
-                         DEBUGFS);
-
-                /* Using popen() instead of run_command() since debugfs does
-                 * not return proper error code if command is not supported */
-                fp = popen(cmd, "r");
-                if (!fp) {
-                        fprintf(stderr, "%s: %s\n", progname, strerror(errno));
-                        return 0;
-                }
-                ret = fread(supp_features, 1, sizeof(supp_features), fp);
-                fclose(fp);
-        }
-        if (ret > 0 && strstr(supp_features,
-                              strncmp(feature, "-O ", 3) ? feature : feature+3))
-                return 0;
-
-        if ((fd = mkstemp(imgname)) < 0)
-                return -1;
-        else
-                close(fd);
-
-        snprintf(cmd, sizeof(cmd), "%s -F %s %s 100 >/dev/null 2>&1",
-                 MKE2FS, feature, imgname);
-        /* run_command() displays the output of mke2fs when it fails for
-         * some feature, so use system() directly */
-        ret = system(cmd);
-        unlink(imgname);
-
-        return ret;
-}
-
-/**
- * append_unique: append @key or @key=@val pair to @buf only if @key does not
- *                exists
- *      @buf: buffer to hold @key or @key=@val
- *      @prefix: prefix string before @key
- *      @key: key string
- *      @val: value string if it's a @key=@val pair
- */
-static void append_unique(char *buf, char *prefix, char *key, char *val,
-                          size_t maxbuflen)
-{
-        char *anchor, *end;
-        int  len;
-
-        if (key == NULL)
-                return;
-
-        anchor = end = strstr(buf, key);
-        /* try to find exact match string in @buf */
-        while (end && *end != '\0' && *end != ',' && *end != ' ' && *end != '=')
-                ++end;
-        len = end - anchor;
-        if (anchor == NULL || strlen(key) != len ||
-            strncmp(anchor, key, len) != 0) {
-                if (prefix != NULL)
-                        strscat(buf, prefix, maxbuflen);
-
-                strscat(buf, key, maxbuflen);
-                if (val != NULL) {
-                        strscat(buf, "=", maxbuflen);
-                        strscat(buf, val, maxbuflen);
-                }
-        }
-}
-
-static void enable_default_ext4_features(struct mkfs_opts *mop, char *anchor,
-                                         size_t maxbuflen, int user_spec)
-{
-        if (IS_OST(&mop->mo_ldd)) {
-                append_unique(anchor, user_spec ? "," : " -O ",
-                              "extents", NULL, sizeof(mop->mo_mkfsopts));
-                append_unique(anchor, ",", "uninit_bg", NULL, maxbuflen);
-       } else if (IS_MDT(&mop->mo_ldd)) {
-               append_unique(anchor, user_spec ? "," : " -O ",
-                             "dirdata", NULL, maxbuflen);
-               append_unique(anchor, ",", "uninit_bg", NULL, maxbuflen);
-               append_unique(anchor, ",", "^extents", NULL, maxbuflen);
-        } else {
-                append_unique(anchor, user_spec ? "," : " -O ",
-                              "uninit_bg", NULL, maxbuflen);
-        }
-
-        /* Multiple mount protection enabled only if failover node specified */
-        if (failover) {
-                if (is_e2fsprogs_feature_supp("-O mmp") == 0)
-                        append_unique(anchor, ",", "mmp", NULL, maxbuflen);
-                else
-                        disp_old_e2fsprogs_msg("mmp", 1);
-        }
-
-        /* Allow more than 65000 subdirectories */
-        if (is_e2fsprogs_feature_supp("-O dir_nlink") == 0)
-                append_unique(anchor, ",", "dir_nlink", NULL, maxbuflen);
-
-        /* The following options are only valid for ext4-based ldiskfs.
-         * If --backfstype=ext3 is specified, do not enable them. */
-        if (mop->mo_ldd.ldd_mount_type == LDD_MT_EXT3)
-                return;
-
-        /* Allow files larger than 2TB.  Also needs LU-16, but not harmful. */
-        if (is_e2fsprogs_feature_supp("-O huge_file") == 0)
-                append_unique(anchor, ",", "huge_file", NULL, maxbuflen);
-
-        /* Enable large block addresses if the LUN is over 2^32 blocks. */
-        if (mop->mo_device_sz / (L_BLOCK_SIZE >> 10) >= 0x100002000ULL &&
-                    is_e2fsprogs_feature_supp("-O 64bit") == 0)
-                append_unique(anchor, ",", "64bit", NULL, maxbuflen);
-
-        /* Cluster inode/block bitmaps and inode table for more efficient IO.
-         * Align the flex groups on a 1MB boundary for better performance. */
-        /* This -O feature needs to go last, since it adds the "-G" option. */
-        if (is_e2fsprogs_feature_supp("-O flex_bg") == 0) {
-                char tmp_buf[64];
-
-                append_unique(anchor, ",", "flex_bg", NULL, maxbuflen);
-
-                if (IS_OST(&mop->mo_ldd)) {
-                        snprintf(tmp_buf, sizeof(tmp_buf), " -G %u",
-                                 (1 << 20) / L_BLOCK_SIZE);
-                        strscat(anchor, tmp_buf, maxbuflen);
-                }
-        }
-        /* Don't add any more "-O" options here, see last comment above */
-}
-
-/**
- * moveopts_to_end: find the option string, move remaining strings to
- *                  where option string starts, and append the option
- *                  string at the end
- *      @start: where the option string starts before the move
- *      RETURN: where the option string starts after the move
- */
-static char *moveopts_to_end(char *start)
-{
-        char save[512];
-        char *end, *idx;
-
-        /* skip whitespace before options */
-        end = start + 2;
-        while (*end == ' ')
-                ++end;
-
-        /* find end of option characters */
-        while (*end != ' ' && *end != '\0')
-                ++end;
-
-        /* save options */
-        strncpy(save, start, end - start);
-        save[end - start] = '\0';
-
-        /* move remaining options up front */
-        if (*end)
-                memmove(start, end, strlen(end));
-        *(start + strlen(end)) = '\0';
-
-        /* append the specified options */
-        if (*(start + strlen(start) - 1) != ' ')
-                strcat(start, " ");
-        idx = start + strlen(start);
-        strcat(start, save);
-
-        return idx;
-}
-
-/* Build fs according to type */
-int make_lustre_backfs(struct mkfs_opts *mop)
+void usage(FILE *out)
 {
-        __u64 device_sz = mop->mo_device_sz, block_count = 0;
-        char mkfs_cmd[PATH_MAX];
-        char buf[64];
-        char *start;
-        char *dev;
-        int ret = 0, ext_opts = 0;
-        size_t maxbuflen;
-
-        if (!(mop->mo_flags & MO_IS_LOOP)) {
-                mop->mo_device_sz = get_device_size(mop->mo_device);
-
-                if (mop->mo_device_sz == 0)
-                        return ENODEV;
-
-                /* Compare to real size */
-                if (device_sz == 0 || device_sz > mop->mo_device_sz)
-                        device_sz = mop->mo_device_sz;
-                else
-                        mop->mo_device_sz = device_sz;
-        }
-
-        if (mop->mo_device_sz != 0) {
-                if (mop->mo_device_sz < 8096){
-                        fprintf(stderr, "%s: size of filesystem must be larger "
-                                "than 8MB, but is set to %lldKB\n",
-                                progname, (long long)mop->mo_device_sz);
-                        return EINVAL;
-                }
-                block_count = mop->mo_device_sz / (L_BLOCK_SIZE >> 10);
-                /* If the LUN size is just over 2^32 blocks, limit the
-                 * filesystem size to 2^32-1 blocks to avoid problems with
-                 * ldiskfs/mkfs not handling this size.  Bug 22906 */
-                if (block_count > 0xffffffffULL && block_count < 0x100002000ULL)
-                        block_count = 0xffffffffULL;
-        }
-
-        if ((mop->mo_ldd.ldd_mount_type == LDD_MT_EXT3) ||
-            (mop->mo_ldd.ldd_mount_type == LDD_MT_LDISKFS) ||
-            (mop->mo_ldd.ldd_mount_type == LDD_MT_LDISKFS2)) {
-                long inode_size = 0;
-
-                /* Journal size in MB */
-                if (strstr(mop->mo_mkfsopts, "-J") == NULL) {
-                        /* Choose our own default journal size */
-                        long journal_sz = 0, max_sz;
-                        if (device_sz > 1024 * 1024) /* 1GB */
-                                journal_sz = (device_sz / 102400) * 4;
-                        /* cap journal size at 1GB */
-                        if (journal_sz > 1024L)
-                                journal_sz = 1024L;
-                        /* man mkfs.ext3 */
-                        max_sz = (102400 * L_BLOCK_SIZE) >> 20; /* 400MB */
-                        if (journal_sz > max_sz)
-                                journal_sz = max_sz;
-                        if (journal_sz) {
-                                sprintf(buf, " -J size=%ld", journal_sz);
-                                strscat(mop->mo_mkfsopts, buf,
-                                        sizeof(mop->mo_mkfsopts));
-                        }
-                }
-
-                /* Inode size (for extended attributes).  The LOV EA size is
-                 * 32 (EA hdr) + 32 (lov_mds_md) + stripes * 24 (lov_ost_data),
-                 * and we want some margin above that for ACLs, other EAs... */
-                if (strstr(mop->mo_mkfsopts, "-I") == NULL) {
-                        if (IS_MDT(&mop->mo_ldd)) {
-                                if (mop->mo_stripe_count > 72)
-                                        inode_size = 512; /* bz 7241 */
-                                        /* see also "-i" below for EA blocks */
-                                else if (mop->mo_stripe_count > 32)
-                                        inode_size = 2048;
-                                else if (mop->mo_stripe_count > 10)
-                                        inode_size = 1024;
-                                else
-                                        inode_size = 512;
-                        } else if (IS_OST(&mop->mo_ldd)) {
-                                /* We store MDS FID and OST objid in EA on OST
-                                 * we need to make inode bigger as well. */
-                                inode_size = 256;
-                        }
-
-                        if (inode_size > 0) {
-                                sprintf(buf, " -I %ld", inode_size);
-                                strscat(mop->mo_mkfsopts, buf,
-                                        sizeof(mop->mo_mkfsopts));
-                        }
-                }
-
-                /* Bytes_per_inode: disk size / num inodes */
-                if (strstr(mop->mo_mkfsopts, "-i") == NULL &&
-                    strstr(mop->mo_mkfsopts, "-N") == NULL) {
-                        long bytes_per_inode = 0;
-
-                        /* Allocate more inodes on MDT devices.  There is
-                         * no data stored on the MDT, and very little extra
-                         * metadata beyond the inode.  It could go down as
-                         * low as 1024 bytes, but this is conservative.
-                         * Account for external EA blocks for wide striping. */
-                        if (IS_MDT(&mop->mo_ldd)) {
-                                bytes_per_inode = inode_size + 1536;
-
-                                if (mop->mo_stripe_count > 72) {
-                                        int extra = mop->mo_stripe_count * 24;
-                                        extra = ((extra - 1) | 4095) + 1;
-                                        bytes_per_inode += extra;
-                                }
-                        }
-
-                        /* Allocate fewer inodes on large OST devices.  Most
-                         * filesystems can be much more aggressive than even
-                         * this, but it is impossible to know in advance. */
-                        if (IS_OST(&mop->mo_ldd)) {
-                                /* OST > 16TB assume average file size 1MB */
-                                if (device_sz > (16ULL << 30))
-                                        bytes_per_inode = 1024 * 1024;
-                                /* OST > 4TB assume average file size 512kB */
-                                else if (device_sz > (4ULL << 30))
-                                        bytes_per_inode = 512 * 1024;
-                                /* OST > 1TB assume average file size 256kB */
-                                else if (device_sz > (1ULL << 30))
-                                        bytes_per_inode = 256 * 1024;
-                                /* OST > 10GB assume average file size 64kB,
-                                 * plus a bit so that inodes will fit into a
-                                 * 256x flex_bg without overflowing */
-                                else if (device_sz > (10ULL << 20))
-                                        bytes_per_inode = 69905;
-                        }
-
-                        if (bytes_per_inode > 0) {
-                                sprintf(buf, " -i %ld", bytes_per_inode);
-                                strscat(mop->mo_mkfsopts, buf,
-                                        sizeof(mop->mo_mkfsopts));
-                        }
-                }
-
-                if (verbose < 2) {
-                        strscat(mop->mo_mkfsopts, " -q",
-                                sizeof(mop->mo_mkfsopts));
-                }
-
-                /* start handle -O mkfs options */
-                if ((start = strstr(mop->mo_mkfsopts, "-O")) != NULL) {
-                        if (strstr(start + 2, "-O") != NULL) {
-                                fprintf(stderr,
-                                        "%s: don't specify multiple -O options\n",
-                                        progname);
-                                return EINVAL;
-                        }
-                        start = moveopts_to_end(start);
-                        maxbuflen = sizeof(mop->mo_mkfsopts) -
-                                    (start - mop->mo_mkfsopts) - strlen(start);
-                        enable_default_ext4_features(mop, start, maxbuflen, 1);
-                } else {
-                        start = mop->mo_mkfsopts + strlen(mop->mo_mkfsopts),
-                        maxbuflen = sizeof(mop->mo_mkfsopts) -
-                                    strlen(mop->mo_mkfsopts);
-                        enable_default_ext4_features(mop, start, maxbuflen, 0);
-                }
-                /* end handle -O mkfs options */
-
-                /* start handle -E mkfs options */
-                if ((start = strstr(mop->mo_mkfsopts, "-E")) != NULL) {
-                        if (strstr(start + 2, "-E") != NULL) {
-                                fprintf(stderr,
-                                        "%s: don't specify multiple -E options\n",
-                                        progname);
-                                return EINVAL;
-                        }
-                        start = moveopts_to_end(start);
-                        maxbuflen = sizeof(mop->mo_mkfsopts) -
-                                    (start - mop->mo_mkfsopts) - strlen(start);
-                        ext_opts = 1;
-                } else {
-                        start = mop->mo_mkfsopts + strlen(mop->mo_mkfsopts);
-                        maxbuflen = sizeof(mop->mo_mkfsopts) -
-                                    strlen(mop->mo_mkfsopts);
-                }
-
-                /* In order to align the filesystem metadata on 1MB boundaries,
-                 * give a resize value that will reserve a power-of-two group
-                 * descriptor blocks, but leave one block for the superblock.
-                 * Only useful for filesystems with < 2^32 blocks due to resize
-                 * limitations. */
-                if (IS_OST(&mop->mo_ldd) && mop->mo_device_sz > 100 * 1024 &&
-                    mop->mo_device_sz * 1024 / L_BLOCK_SIZE <= 0xffffffffULL) {
-                        unsigned group_blocks = L_BLOCK_SIZE * 8;
-                        unsigned desc_per_block = L_BLOCK_SIZE / 32;
-                        unsigned resize_blks;
-
-                        resize_blks = (1ULL<<32) - desc_per_block*group_blocks;
-                        snprintf(buf, sizeof(buf), "%u", resize_blks);
-                        append_unique(start, ext_opts ? "," : " -E ",
-                                      "resize", buf, maxbuflen);
-                        ext_opts = 1;
-                }
-
-                /* Avoid zeroing out the full journal - speeds up mkfs */
-                if (is_e2fsprogs_feature_supp("-E lazy_journal_init") == 0)
-                        append_unique(start, ext_opts ? "," : " -E ",
-                                      "lazy_journal_init", NULL, maxbuflen);
-                /* end handle -E mkfs options */
-
-                /* Allow reformat of full devices (as opposed to
-                   partitions.)  We already checked for mounted dev. */
-                strscat(mop->mo_mkfsopts, " -F", sizeof(mop->mo_mkfsopts));
-
-                snprintf(mkfs_cmd, sizeof(mkfs_cmd),
-                         "%s -j -b %d -L %s ", MKE2FS, L_BLOCK_SIZE,
-                         mop->mo_ldd.ldd_svname);
-        } else {
-                fprintf(stderr,"%s: unsupported fs type: %d (%s)\n",
-                        progname, mop->mo_ldd.ldd_mount_type,
-                        MT_STR(&mop->mo_ldd));
-                return EINVAL;
-        }
-
-        /* For loop device format the dev, not the filename */
-        dev = mop->mo_device;
-        if (mop->mo_flags & MO_IS_LOOP)
-                dev = mop->mo_loopdev;
-
-        vprint("formatting backing filesystem %s on %s\n",
-               MT_STR(&mop->mo_ldd), dev);
-        vprint("\ttarget name  %s\n", mop->mo_ldd.ldd_svname);
-        vprint("\t4k blocks     "LPU64"\n", block_count);
-        vprint("\toptions       %s\n", mop->mo_mkfsopts);
-
-        /* mkfs_cmd's trailing space is important! */
-        strscat(mkfs_cmd, mop->mo_mkfsopts, sizeof(mkfs_cmd));
-        strscat(mkfs_cmd, " ", sizeof(mkfs_cmd));
-        strscat(mkfs_cmd, dev, sizeof(mkfs_cmd));
-        if (block_count != 0) {
-                sprintf(buf, " "LPU64, block_count);
-                strscat(mkfs_cmd, buf, sizeof(mkfs_cmd));
-        }
-
-        vprint("mkfs_cmd = %s\n", mkfs_cmd);
-        ret = run_command(mkfs_cmd, sizeof(mkfs_cmd));
-        if (ret) {
-                fatal();
-                fprintf(stderr, "Unable to build fs %s (%d)\n", dev, ret);
-        }
-        return ret;
+       fprintf(out, "usage: %s <target type> [--backfstype="FSLIST"] "
+               "--fsname=<filesystem name>\n"
+               "\t--index=<target index> [options] <device>\n", progname);
+#ifdef HAVE_ZFS_OSD
+       fprintf(out, "usage: %s <target type> --backfstype=zfs "
+               "--fsname=<filesystem name> [options]\n"
+               "\t<pool name>/<dataset name>\n"
+               "\t[[<vdev type>] <device> [<device> ...] [vdev type>] ...]\n",
+               progname);
+#endif
+       fprintf(out,
+               "\t<device>:block device or file (e.g /dev/sda or /tmp/ost1)\n"
+#ifdef HAVE_ZFS_OSD
+               "\t<pool name>: name of ZFS pool where target is created "
+                       "(e.g. tank)\n"
+               "\t<dataset name>: name of new dataset, must be unique within "
+                       "pool (e.g. ost1)\n"
+               "\t<vdev type>: type of vdev (mirror, raidz, raidz2, spare, "
+                       "cache, log)\n"
+#endif
+               "\n"
+               "\ttarget types:\n"
+               "\t\t--mgs: configuration management service\n"
+               "\t\t--nomgs: turn off MGS service on this MDT\n"
+#ifndef TUNEFS
+               "\t\t--mdt: metadata storage, mutually exclusive with ost\n"
+               "\t\t--ost: object storage, mutually exclusive with mdt, mgs\n"
+#endif
+               "\toptions (in order of popularity):\n"
+               "\t\t--index=#N: numerical target index (0..N)\n"
+               "\t\t\trequired for all targets other than the MGS,\n"
+               "\t\t\ttarget index may either be a decimal number or\n"
+               "\t\t\thexadecimal number starting with '0x'\n"
+               "\t\t--fsname=<8_char_filesystem_name>: fs targets belong to\n"
+               "\t\t\trequired for all targets other than MGS\n"
+               "\t\t--mgsnode=<nid>[,<...>]: NID(s) of remote MGS\n"
+               "\t\t\trequired for all targets other than MGS\n"
+               "\t\t--mountfsoptions=<opts>: permanent Lustre mount options\n"
+               "\t\t--backfs-mount-opts=<opts>: backing fs mount options\n"
+               "\t\t--failnode=<nid>[,<...>]: NID(s) of backup failover node\n"
+               "\t\t\tmutually exclusive with --servicenode\n"
+               "\t\t--servicenode=<nid>[,<...>]: NID(s) of service partners\n"
+               "\t\t\ttreat nodes as equal service node, mutually exclusive "
+                       "with --failnode\n"
+               "\t\t--param <key>=<value>: set a permanent parameter\n"
+               "\t\t\te.g. --param sys.timeout=40\n"
+               "\t\t\t     --param lov.stripesize=2M\n"
+               "\t\t--network=<net>[,<...>]: restrict OST/MDT to network(s)\n"
+#ifndef TUNEFS
+               "\t\t--backfstype=<fstype>: backing fs type (ldiskfs, zfs)\n"
+               "\t\t--device-size=#N(KB): device size for loop devices\n"
+               "\t\t--mkfsoptions=<opts>: format options\n"
+               "\t\t--reformat: overwrite an existing disk\n"
+               "\t\t--replace: replace an old target with the same index\n"
+               "\t\t--stripe-count-hint=#N: for optimizing MDT inode size\n"
+#else
+               "\t\t--erase-param <key>: erase all instances of a parameter\n"
+               "\t\t--erase-params: erase all old parameter settings\n"
+               "\t\t--writeconf: erase all config logs for this fs.\n"
+               "\t\t--nolocallogs: use logs from MGS, not local ones.\n"
+               "\t\t--quota: enable space accounting on old 2.x device.\n"
+               "\t\t--rename: rename the filesystem name\n"
+#endif
+               "\t\t--comment=<user comment>: arbitrary string (%d bytes)\n"
+               "\t\t--dryrun: report what we would do; don't write to disk\n"
+               "\t\t--verbose: e.g. show mkfs progress\n"
+               "\t\t--force-nohostid: Ignore hostid requirement for ZFS "
+                       "import\n"
+               "\t\t-V|--version: output build version of the utility and\n"
+               "\t\t\texit\n"
+               "\t\t--quiet\n",
+               (int)sizeof(((struct lustre_disk_data *)0)->ldd_userdata));
 }
 
 /* ==================== Lustre config functions =============*/
 
-void print_ldd(char *str, struct lustre_disk_data *ldd)
-{
-        printf("\n   %s:\n", str);
-        printf("Target:     %s\n", ldd->ldd_svname);
-        if (ldd->ldd_svindex == INDEX_UNASSIGNED)
-                printf("Index:      unassigned\n");
-        else
-                printf("Index:      %d\n", ldd->ldd_svindex);
-        if (ldd->ldd_uuid[0])
-                printf("UUID:       %s\n", (char *)ldd->ldd_uuid);
-        printf("Lustre FS:  %s\n", ldd->ldd_fsname);
-        printf("Mount type: %s\n", MT_STR(ldd));
-        printf("Flags:      %#x\n", ldd->ldd_flags);
-        printf("              (%s%s%s%s%s%s%s%s%s%s)\n",
-               IS_MDT(ldd) ? "MDT ":"",
-               IS_OST(ldd) ? "OST ":"",
-               IS_MGS(ldd) ? "MGS ":"",
-               ldd->ldd_flags & LDD_F_NEED_INDEX ? "needs_index ":"",
-               ldd->ldd_flags & LDD_F_VIRGIN     ? "first_time ":"",
-               ldd->ldd_flags & LDD_F_UPDATE     ? "update ":"",
-               ldd->ldd_flags & LDD_F_WRITECONF  ? "writeconf ":"",
-               ldd->ldd_flags & LDD_F_IAM_DIR  ? "IAM_dir_format ":"",
-               ldd->ldd_flags & LDD_F_NO_PRIMNODE? "no_primnode ":"",
-               ldd->ldd_flags & LDD_F_UPGRADE14  ? "upgrade1.4 ":"");
-        printf("Persistent mount opts: %s\n", ldd->ldd_mount_opts);
-        printf("Parameters:%s\n", ldd->ldd_params);
-        if (ldd->ldd_userdata[0])
-                printf("Comment: %s\n", ldd->ldd_userdata);
-        printf("\n");
-}
-
-static int touch_file(char *filename)
-{
-        int fd;
-
-        if (filename == NULL) {
-                return 1;
-        }
-
-        fd = open(filename, O_CREAT | O_TRUNC, 0600);
-        if (fd < 0) {
-                return 1;
-        } else {
-                close(fd);
-                return 0;
-        }
-}
-
-/* keep it less than LL_FID_NAMELEN */
-#define DUMMY_FILE_NAME_LEN             25
-#define EXT3_DIRENT_SIZE                DUMMY_FILE_NAME_LEN
-
-/* Need to add these many entries to this directory to make HTREE dir. */
-#define MIN_ENTRIES_REQ_FOR_HTREE       ((L_BLOCK_SIZE / EXT3_DIRENT_SIZE))
-
-static int add_dummy_files(char *dir)
+void print_ldd(char *str, struct mkfs_opts *mop)
 {
-        char fpname[PATH_MAX];
-        int i;
-        int rc;
-
-        for (i = 0; i < MIN_ENTRIES_REQ_FOR_HTREE; i++) {
-                snprintf(fpname, PATH_MAX, "%s/%0*d", dir,
-                         DUMMY_FILE_NAME_LEN, i);
-
-                rc = touch_file(fpname);
-                if (rc && rc != -EEXIST) {
-                        fprintf(stderr,
-                                "%s: Can't create dummy file %s: %s\n",
-                                progname, fpname , strerror(errno));
-                        return rc;
-                }
-        }
-        return 0;
+       struct lustre_disk_data *ldd = &mop->mo_ldd;
+
+       printf("\n   %s:\n", str);
+       printf("Target:     %s\n", ldd->ldd_svname);
+       if (ldd->ldd_svindex == INDEX_UNASSIGNED)
+               printf("Index:      unassigned\n");
+       else
+               printf("Index:      %d\n", ldd->ldd_svindex);
+       if (ldd->ldd_uuid[0])
+               printf("UUID:       %s\n", (char *)ldd->ldd_uuid);
+       printf("Lustre FS:  %s\n", ldd->ldd_fsname);
+       printf("Mount type: %s\n", MT_STR(ldd));
+       printf("Flags:      %#x\n", ldd->ldd_flags);
+       printf("              (%s%s%s%s%s%s%s%s%s)\n",
+              IS_MDT(ldd) ? "MDT " : "",
+              IS_OST(ldd) ? "OST " : "",
+              IS_MGS(ldd) ? "MGS " : "",
+              ldd->ldd_flags & LDD_F_NEED_INDEX ? "needs_index " : "",
+              ldd->ldd_flags & LDD_F_VIRGIN     ? "first_time " : "",
+              ldd->ldd_flags & LDD_F_UPDATE     ? "update " : "",
+              ldd->ldd_flags & LDD_F_WRITECONF  ? "writeconf " : "",
+              ldd->ldd_flags & LDD_F_NO_PRIMNODE ? "no_primnode " : "",
+              ldd->ldd_flags & LDD_F_NO_LOCAL_LOGS ? "nolocallogs " : "");
+       printf("Persistent mount opts: %s\n", ldd->ldd_mount_opts);
+       osd_print_ldd_params(mop);
+       if (ldd->ldd_userdata[0])
+               printf("Comment: %s\n", ldd->ldd_userdata);
+       printf("\n");
 }
 
-static int __l_mkdir(char * filepnm, int mode , struct mkfs_opts *mop)
-{
-        int ret;
-
-        ret = mkdir(filepnm, mode);
-        if (ret && ret != -EEXIST)
-                return ret;
-
-        /* IAM mode supports ext3 directories of HTREE type only. So add dummy
-         * entries to new directory to create htree type of container for
-         * this directory. */
-        if (mop->mo_ldd.ldd_flags & LDD_F_IAM_DIR)
-                return add_dummy_files(filepnm);
-        return 0;
-}
-
-/* Write the server config files */
-int write_local_files(struct mkfs_opts *mop)
+void set_defaults(struct mkfs_opts *mop)
 {
-        char mntpt[] = "/tmp/mntXXXXXX";
-        char filepnm[128];
-        char *dev;
-        FILE *filep;
-        int ret = 0;
-        size_t num;
-
-        /* Mount this device temporarily in order to write these files */
-        if (!mkdtemp(mntpt)) {
-                fprintf(stderr, "%s: Can't create temp mount point %s: %s\n",
-                        progname, mntpt, strerror(errno));
-                return errno;
-        }
-
-        dev = mop->mo_device;
-        if (mop->mo_flags & MO_IS_LOOP)
-                dev = mop->mo_loopdev;
-
-        ret = mount(dev, mntpt, MT_STR(&mop->mo_ldd), 0,
-                    mop->mo_ldd.ldd_mount_opts);
-        if (ret) {
-                fprintf(stderr, "%s: Unable to mount %s: %s\n",
-                        progname, dev, strerror(errno));
-                ret = errno;
-                if (errno == ENODEV) {
-                        fprintf(stderr, "Is the %s module available?\n",
-                                MT_STR(&mop->mo_ldd));
-                }
-                goto out_rmdir;
-        }
-
-        /* Set up initial directories */
-        sprintf(filepnm, "%s/%s", mntpt, MOUNT_CONFIGS_DIR);
-        ret = __l_mkdir(filepnm, 0777, mop);
-        if ((ret != 0) && (errno != EEXIST)) {
-                fprintf(stderr, "%s: Can't make configs dir %s (%s)\n",
-                        progname, filepnm, strerror(errno));
-                goto out_umnt;
-        } else if (errno == EEXIST) {
-                ret = 0;
-        }
-
-        /* Save the persistent mount data into a file. Lustre must pre-read
-           this file to get the real mount options. */
-        vprint("Writing %s\n", MOUNT_DATA_FILE);
-        sprintf(filepnm, "%s/%s", mntpt, MOUNT_DATA_FILE);
-        filep = fopen(filepnm, "w");
-        if (!filep) {
-                fprintf(stderr, "%s: Unable to create %s file: %s\n",
-                        progname, filepnm, strerror(errno));
-                goto out_umnt;
-        }
-        num = fwrite(&mop->mo_ldd, sizeof(mop->mo_ldd), 1, filep);
-        if (num < 1 && ferror(filep)) {
-                fprintf(stderr, "%s: Unable to write to file (%s): %s\n",
-                        progname, filepnm, strerror(errno));
-                goto out_umnt;
-        }
-        fclose(filep);
-        /* COMPAT_146 */
-#ifdef TUNEFS
-        /* Check for upgrade */
-        if ((mop->mo_ldd.ldd_flags & (LDD_F_UPGRADE14 | LDD_F_SV_TYPE_MGS))
-            == (LDD_F_UPGRADE14 | LDD_F_SV_TYPE_MGS)) {
-                char cmd[128];
-                char *term;
-                int cmdsz = sizeof(cmd);
-                vprint("Copying old logs\n");
-
-                /* Copy the old client log to fsname-client */
-                sprintf(filepnm, "%s/%s/%s-client",
-                        mntpt, MOUNT_CONFIGS_DIR, mop->mo_ldd.ldd_fsname);
-                snprintf(cmd, cmdsz, "cp %s/%s/client %s", mntpt, MDT_LOGS_DIR,
-                         filepnm);
-                ret = run_command(cmd, cmdsz);
-                if (ret) {
-                        fprintf(stderr, "%s: Can't copy 1.4 config %s/client "
-                                "(%d)\n", progname, MDT_LOGS_DIR, ret);
-                        fprintf(stderr, "mount -t ldiskfs %s somewhere, "
-                                "find the client log for fs %s and "
-                                "copy it manually into %s/%s-client, "
-                                "then umount.\n",
-                                mop->mo_device,
-                                mop->mo_ldd.ldd_fsname, MOUNT_CONFIGS_DIR,
-                                mop->mo_ldd.ldd_fsname);
-                        goto out_umnt;
-                }
-
-                /* We need to use the old mdt log because otherwise mdt won't
-                   have complete lov if old clients connect before all
-                   servers upgrade. */
-                /* Copy the old mdt log to fsname-MDT0000 (get old
-                   name from mdt_UUID) */
-                ret = 1;
-                strscpy(filepnm, (char *)mop->mo_ldd.ldd_uuid, sizeof(filepnm));
-                term = strstr(filepnm, "_UUID");
-                if (term) {
-                        *term = '\0';
-                        snprintf(cmd, cmdsz, "cp %s/%s/%s %s/%s/%s",
-                                 mntpt, MDT_LOGS_DIR, filepnm,
-                                 mntpt, MOUNT_CONFIGS_DIR,
-                                 mop->mo_ldd.ldd_svname);
-                        ret = run_command(cmd, cmdsz);
-                }
-                if (ret) {
-                        fprintf(stderr, "%s: Can't copy 1.4 config %s/%s "
-                                "(%d)\n", progname, MDT_LOGS_DIR, filepnm, ret);
-                        fprintf(stderr, "mount -t ext3 %s somewhere, "
-                                "find the MDT log for fs %s and "
-                                "copy it manually into %s/%s, "
-                                "then umount.\n",
-                                mop->mo_device,
-                                mop->mo_ldd.ldd_fsname, MOUNT_CONFIGS_DIR,
-                                mop->mo_ldd.ldd_svname);
-                        goto out_umnt;
-                }
-        }
+       mop->mo_ldd.ldd_magic = LDD_MAGIC;
+       mop->mo_ldd.ldd_config_ver = 1;
+       mop->mo_ldd.ldd_flags = LDD_F_NEED_INDEX | LDD_F_UPDATE | LDD_F_VIRGIN;
+#ifdef HAVE_LDISKFS_OSD
+       mop->mo_ldd.ldd_mount_type = LDD_MT_LDISKFS;
+#else
+       mop->mo_ldd.ldd_mount_type = LDD_MT_ZFS;
 #endif
-        /* end COMPAT_146 */
-
-out_umnt:
-        umount(mntpt);
-out_rmdir:
-        rmdir(mntpt);
-        return ret;
-}
-
-int read_local_files(struct mkfs_opts *mop)
-{
-        char tmpdir[] = "/tmp/dirXXXXXX";
-        char cmd[PATH_MAX];
-        char filepnm[128];
-        char *dev;
-        FILE *filep;
-        int ret = 0;
-        int cmdsz = sizeof(cmd);
-
-        /* Make a temporary directory to hold Lustre data files. */
-        if (!mkdtemp(tmpdir)) {
-                fprintf(stderr, "%s: Can't create temporary directory %s: %s\n",
-                        progname, tmpdir, strerror(errno));
-                return errno;
-        }
-
-        dev = mop->mo_device;
-
-        /* TODO: it's worth observing the get_mountdata() function that is
-                 in mount_utils.c for getting the mountdata out of the
-                 filesystem */
-
-        /* Construct debugfs command line. */
-        snprintf(cmd, cmdsz, "%s -c -R 'dump /%s %s/mountdata' '%s'",
-                 DEBUGFS, MOUNT_DATA_FILE, tmpdir, dev);
-
-        ret = run_command(cmd, cmdsz);
-        if (ret)
-                verrprint("%s: Unable to dump %s dir (%d)\n",
-                          progname, MOUNT_CONFIGS_DIR, ret);
-
-        sprintf(filepnm, "%s/mountdata", tmpdir);
-        filep = fopen(filepnm, "r");
-        if (filep) {
-                size_t num_read;
-                vprint("Reading %s\n", MOUNT_DATA_FILE);
-                num_read = fread(&mop->mo_ldd, sizeof(mop->mo_ldd), 1, filep);
-                if (num_read < 1 && ferror(filep)) {
-                        fprintf(stderr, "%s: Unable to read from file %s: %s\n",
-                                progname, filepnm, strerror(errno));
-                        goto out_close;
-                }
-        } else {
-                /* COMPAT_146 */
-                /* Try to read pre-1.6 config from last_rcvd */
-                struct lr_server_data lsd;
-                verrprint("%s: Unable to read %d.%d config %s.\n",
-                          progname, LUSTRE_MAJOR, LUSTRE_MINOR, filepnm);
-
-                verrprint("Trying 1.4 config from last_rcvd\n");
-                sprintf(filepnm, "%s/%s", tmpdir, LAST_RCVD);
-
-                /* Construct debugfs command line. */
-                snprintf(cmd, cmdsz, "%s -c -R 'dump /%s %s' %s",
-                         DEBUGFS, LAST_RCVD, filepnm, dev);
-
-                ret = run_command(cmd, cmdsz);
-                if (ret) {
-                        fprintf(stderr, "%s: Unable to dump %s file (%d)\n",
-                                progname, LAST_RCVD, ret);
-                        goto out_rmdir;
-                }
-
-                filep = fopen(filepnm, "r");
-                if (!filep) {
-                        fprintf(stderr, "%s: Unable to open %s: %s\n",
-                                progname, filepnm, strerror(errno));
-                        ret = errno;
-                        verrprint("Contents of %s:\n", tmpdir);
-                        verbose+=2;
-                        snprintf(cmd, cmdsz, "ls -l %s/", tmpdir);
-                        run_command(cmd, cmdsz);
-                        verrprint("Contents of disk:\n");
-                        snprintf(cmd, cmdsz, "%s -c -R 'ls -l /' %s",
-                                 DEBUGFS, dev);
-                        run_command(cmd, cmdsz);
-
-                        goto out_rmdir;
-                }
-                vprint("Reading %s\n", LAST_RCVD);
-                ret = fread(&lsd, 1, sizeof(lsd), filep);
-                if (ret < sizeof(lsd)) {
-                        fprintf(stderr, "%s: Short read (%d of %d)\n",
-                                progname, ret, (int)sizeof(lsd));
-                        ret = ferror(filep);
-                        if (ret)
-                                goto out_close;
-                }
-                vprint("Feature compat=%x, incompat=%x\n",
-                       lsd.lsd_feature_compat, lsd.lsd_feature_incompat);
-
-                if ((lsd.lsd_feature_compat & OBD_COMPAT_OST) ||
-                    (lsd.lsd_feature_incompat & OBD_INCOMPAT_OST)) {
-                        mop->mo_ldd.ldd_flags = LDD_F_SV_TYPE_OST;
-                        mop->mo_ldd.ldd_svindex = lsd.lsd_ost_index;
-                } else if ((lsd.lsd_feature_compat & OBD_COMPAT_MDT) ||
-                           (lsd.lsd_feature_incompat & OBD_INCOMPAT_MDT)) {
-                        /* We must co-locate so mgs can see old logs.
-                           If user doesn't want this, they can copy the old
-                           logs manually and re-tunefs. */
-                        mop->mo_ldd.ldd_flags =
-                                LDD_F_SV_TYPE_MDT | LDD_F_SV_TYPE_MGS;
-                        mop->mo_ldd.ldd_svindex = lsd.lsd_mdt_index;
-                } else  {
-                        /* If neither is set, we're pre-1.4.6, make a guess. */
-                        /* Construct debugfs command line. */
-                        snprintf(cmd, cmdsz, "%s -c -R 'rdump /%s %s' %s",
-                                 DEBUGFS, MDT_LOGS_DIR, tmpdir, dev);
-                        run_command(cmd, cmdsz);
-
-                        sprintf(filepnm, "%s/%s", tmpdir, MDT_LOGS_DIR);
-                        if (lsd.lsd_ost_index > 0) {
-                                mop->mo_ldd.ldd_flags = LDD_F_SV_TYPE_OST;
-                                mop->mo_ldd.ldd_svindex = lsd.lsd_ost_index;
-                        } else {
-                                /* If there's a LOGS dir, it's an MDT */
-                                if ((ret = access(filepnm, F_OK)) == 0) {
-                                        mop->mo_ldd.ldd_flags =
-                                        LDD_F_SV_TYPE_MDT |
-                                        LDD_F_SV_TYPE_MGS;
-                                        /* Old MDT's are always index 0
-                                           (pre CMD) */
-                                        mop->mo_ldd.ldd_svindex = 0;
-                                } else {
-                                        /* The index may not be correct */
-                                        mop->mo_ldd.ldd_flags =
-                                        LDD_F_SV_TYPE_OST | LDD_F_NEED_INDEX;
-                                        verrprint("OST with unknown index\n");
-                                }
-                        }
-                }
-
-                ret = 0;
-                memcpy(mop->mo_ldd.ldd_uuid, lsd.lsd_uuid,
-                       sizeof(mop->mo_ldd.ldd_uuid));
-                mop->mo_ldd.ldd_flags |= LDD_F_UPGRADE14;
-        }
-        /* end COMPAT_146 */
-out_close:
-        fclose(filep);
-
-out_rmdir:
-        snprintf(cmd, cmdsz, "rm -rf %s", tmpdir);
-        run_command(cmd, cmdsz);
-        if (ret)
-                verrprint("Failed to read old data (%d)\n", ret);
-        return ret;
+       mop->mo_ldd.ldd_svindex = INDEX_UNASSIGNED;
+       mop->mo_mgs_failnodes = 0;
+       mop->mo_stripe_count = 1;
+       mop->mo_pool_vdevs = NULL;
 }
 
-
-void set_defaults(struct mkfs_opts *mop)
+/* Make the mdt/ost server obd name based on the filesystem name */
+static bool server_make_name(__u32 flags, __u16 index, const char *fs,
+                            char *name_buf, size_t name_buf_size)
 {
-        mop->mo_ldd.ldd_magic = LDD_MAGIC;
-        mop->mo_ldd.ldd_config_ver = 1;
-        mop->mo_ldd.ldd_flags = LDD_F_NEED_INDEX | LDD_F_UPDATE | LDD_F_VIRGIN;
-        mop->mo_mgs_failnodes = 0;
-        strcpy(mop->mo_ldd.ldd_fsname, "lustre");
-        mop->mo_ldd.ldd_mount_type = LDD_MT_LDISKFS;
-
-        mop->mo_ldd.ldd_svindex = INDEX_UNASSIGNED;
-        mop->mo_stripe_count = 1;
+       bool invalid_flag = false;
+
+       if (flags & (LDD_F_SV_TYPE_MDT | LDD_F_SV_TYPE_OST)) {
+               if (!(flags & LDD_F_SV_ALL))
+                       snprintf(name_buf, name_buf_size, "%.8s%c%s%04x", fs,
+                               (flags & LDD_F_VIRGIN) ? ':' :
+                               ((flags & LDD_F_WRITECONF) ? '=' :
+                               ((flags & LDD_F_NO_LOCAL_LOGS) ? '+' : '-')),
+                               (flags & LDD_F_SV_TYPE_MDT) ? "MDT" : "OST",
+                               index);
+       } else if (flags & LDD_F_SV_TYPE_MGS) {
+               snprintf(name_buf, name_buf_size, "MGS");
+       } else {
+               fprintf(stderr, "unknown server type %#x\n", flags);
+               invalid_flag = true;
+       }
+       return invalid_flag;
 }
 
 static inline void badopt(const char *opt, char *type)
 {
-        fprintf(stderr, "%s: '--%s' only valid for %s\n",
-                progname, opt, type);
-        usage(stderr);
+       fprintf(stderr, "%s: '--%s' only valid for %s\n",
+               progname, opt, type);
+       usage(stderr);
 }
 
-static int add_param(char *buf, char *key, char *val)
+#ifdef TUNEFS
+/**
+ * Removes all existing instances of the parameter passed in \a param,
+ * which are in the form of "key=<value>", from the buffer at \a buf.
+ *
+ * The parameter can be either in the form of "key" when passed by option
+ * "--erase-param", or in the form of "key=<value>" when passed by option
+ * "--param".
+ *
+ * \param buf    the buffer holding on-disk server parameters.
+ * \param param          the parameter whose instances are to be removed from \a buf.
+ * \param withval true means the parameter is in the form of "key=<value>"
+ *               false means the parameter is in the form of "key"
+ *
+ * \retval 0     success, parameter was erased,
+ * \retval 1     success, parameter was not found, don't need to do erase_ldd,
+ * \retval EINVAL failure, invalid input parameter.
+ */
+static int erase_param(const char *const buf, const char *const param,
+                      bool withval)
 {
-        int end = sizeof(((struct lustre_disk_data *)0)->ldd_params);
-        int start = strlen(buf);
-        int keylen = 0;
-
-        if (key)
-                keylen = strlen(key);
-        if (start + 1 + keylen + strlen(val) >= end) {
-                fprintf(stderr, "%s: params are too long-\n%s %s%s\n",
-                        progname, buf, key ? key : "", val);
-                return 1;
-        }
-
-        sprintf(buf + start, " %s%s", key ? key : "", val);
-        return 0;
+       char    search[PARAM_MAX + 1] = "";
+       char    *buffer = (char *)buf;
+       bool    found = false;
+
+       if (strlen(param) > PARAM_MAX) {
+               fprintf(stderr, "%s: param to erase is too long-\n%s\n",
+                       progname, param);
+               return EINVAL;
+       }
+
+       /* add_param() writes a space as the first character in ldd_params */
+       search[0] = ' ';
+
+       /* "key" or "key=<value>" */
+       if (withval) {
+               char *keyend;
+
+               keyend = strchr(param, '=');
+               if (!keyend)
+                       return EINVAL;
+               strncpy(&search[1], param, keyend - param + 1);
+       } else {
+               snprintf(search + 1, sizeof(search) - 1, "%s=", param);
+       }
+
+       while (1) {
+               char    *space;
+
+               buffer = strstr(buffer, search);
+               if (!buffer)
+                       return found == true ? 0 : 1;
+               found = true;
+               space = strchr(buffer + 1, ' ');
+               if (space) {
+                       memmove(buffer, space, strlen(space) + 1);
+               } else {
+                       *buffer = '\0';
+                       return 0;
+               }
+       }
 }
+#endif
 
 /* from mount_lustre */
 /* Get rid of symbolic hostnames for tcp, since kernel can't do lookups */
 #define MAXNIDSTR 1024
 static char *convert_hostnames(char *s1)
 {
-        char *converted, *s2 = 0, *c, *end, sep;
-        int left = MAXNIDSTR;
-        lnet_nid_t nid;
-
-        converted = malloc(left);
-        if (converted == NULL) {
-                return NULL;
-        }
-
-        end = s1 + strlen(s1);
-        c = converted;
-        while ((left > 0) && (s1 < end)) {
-                s2 = strpbrk(s1, ",:");
-                if (!s2)
-                        s2 = end;
-                sep = *s2;
-                *s2 = '\0';
-                nid = libcfs_str2nid(s1);
-
-                if (nid == LNET_NID_ANY) {
-                        fprintf(stderr, "%s: Can't parse NID '%s'\n",
-                                progname, s1);
-                        free(converted);
-                        return NULL;
-                }
-                if (strncmp(libcfs_nid2str(nid), "127.0.0.1",
-                            strlen("127.0.0.1")) == 0) {
-                        fprintf(stderr, "%s: The NID '%s' resolves to the "
-                                "loopback address '%s'.  Lustre requires a "
-                                "non-loopback address.\n",
-                                progname, s1, libcfs_nid2str(nid));
-                        free(converted);
-                        return NULL;
-                }
-
-                c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep);
-                left = converted + MAXNIDSTR - c;
-                s1 = s2 + 1;
-        }
-        return converted;
+       char *converted, *s2 = 0, *c, *end, sep;
+       int left = MAXNIDSTR;
+       lnet_nid_t nid;
+
+       converted = malloc(left);
+       if (!converted)
+               return NULL;
+
+       end = s1 + strlen(s1);
+       c = converted;
+       while ((left > 0) && (s1 < end)) {
+               s2 = strpbrk(s1, ",:");
+               if (!s2)
+                       s2 = end;
+               sep = *s2;
+               *s2 = '\0';
+               nid = libcfs_str2nid(s1);
+               *s2 = sep;
+
+               if (nid == LNET_NID_ANY) {
+                       fprintf(stderr, "%s: Cannot resolve hostname '%s'.\n",
+                               progname, s1);
+                       free(converted);
+                       return NULL;
+               }
+               if (strncmp(libcfs_nid2str(nid), "127.0.0.1",
+                           strlen("127.0.0.1")) == 0) {
+                       fprintf(stderr,
+                               "%s: The NID '%s' resolves to the loopback address '%s'.  Lustre requires a non-loopback address.\n",
+                               progname, s1, libcfs_nid2str(nid));
+                       free(converted);
+                       return NULL;
+               }
+
+               c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep);
+               left = converted + MAXNIDSTR - c;
+               s1 = s2 + 1;
+       }
+       return converted;
 }
 
 int parse_opts(int argc, char *const argv[], struct mkfs_opts *mop,
-               char **mountopts)
+              char **mountopts, char *old_fsname)
 {
-        static struct option long_opt[] = {
-                {"iam-dir", 0, 0, 'a'},
-                {"backfstype", 1, 0, 'b'},
-                {"stripe-count-hint", 1, 0, 'c'},
-                {"comment", 1, 0, 'u'},
-                {"configdev", 1, 0, 'C'},
-                {"device-size", 1, 0, 'd'},
-                {"dryrun", 0, 0, 'n'},
-                {"erase-params", 0, 0, 'e'},
-                {"failnode", 1, 0, 'f'},
-                {"failover", 1, 0, 'f'},
-                {"mgs", 0, 0, 'G'},
-                {"help", 0, 0, 'h'},
-                {"index", 1, 0, 'i'},
-                {"mkfsoptions", 1, 0, 'k'},
-                {"mgsnode", 1, 0, 'm'},
-                {"mgsnid", 1, 0, 'm'},
-                {"mdt", 0, 0, 'M'},
-                {"fsname",1, 0, 'L'},
-                {"noformat", 0, 0, 'n'},
-                {"nomgs", 0, 0, 'N'},
-                {"mountfsoptions", 1, 0, 'o'},
-                {"ost", 0, 0, 'O'},
-                {"param", 1, 0, 'p'},
-                {"print", 0, 0, 'n'},
-                {"quiet", 0, 0, 'q'},
-                {"reformat", 0, 0, 'r'},
-                {"servicenode", 1, 0, 's'},
-                {"verbose", 0, 0, 'v'},
-                {"writeconf", 0, 0, 'w'},
-                {"upgrade_to_18", 0, 0, 'U'},
-                {"network", 1, 0, 't'},
-                {0, 0, 0, 0}
-        };
-        char *optstring = "b:c:C:d:ef:Ghi:k:L:m:MnNo:Op:Pqrs:t:Uu:vw";
-        int opt;
-        int rc, longidx;
-        int failnode_set = 0, servicenode_set = 0;
-
-        while ((opt = getopt_long(argc, argv, optstring, long_opt, &longidx)) !=
-               EOF) {
-                switch (opt) {
-                case 'a': {
-                        if (IS_MDT(&mop->mo_ldd))
-                                mop->mo_ldd.ldd_flags |= LDD_F_IAM_DIR;
-                        break;
-                }
-                case 'b': {
-                        int i = 0;
-                        while (i < LDD_MT_LAST) {
-                                if (strcmp(optarg, mt_str(i)) == 0) {
-                                        mop->mo_ldd.ldd_mount_type = i;
-                                        break;
-                                }
-                                i++;
-                        }
-                        break;
-                }
-                case 'c':
-                        if (IS_MDT(&mop->mo_ldd)) {
-                                int stripe_count = atol(optarg);
-                                if (stripe_count <= 0) {
-                                        fprintf(stderr, "%s: bad stripe count "
-                                                "%d\n", progname, stripe_count);
-                                        return 1;
-                                }
-                                mop->mo_stripe_count = stripe_count;
-                        } else {
-                                badopt(long_opt[longidx].name, "MDT");
-                                return 1;
-                        }
-                        break;
-                case 'C': /* Configdev */
-                        //FIXME
-                        printf("Configdev not implemented\n");
-                        return 1;
-                case 'd':
-                        mop->mo_device_sz = atol(optarg);
-                        break;
-                case 'e':
-                        mop->mo_ldd.ldd_params[0] = '\0';
-                        /* Must update the mgs logs */
-                        mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
-                        break;
-                case 'f':
-                case 's': {
-                        char *nids;
-
-                        if ((opt == 'f' && servicenode_set)
-                            || (opt == 's' && failnode_set)) {
-                                fprintf(stderr, "%s: %s cannot use with --%s\n",
-                                        progname, long_opt[longidx].name,
-                                        opt == 'f' ? "servicenode" : "failnode");
-                                return 1;
-                        }
-
-                        nids = convert_hostnames(optarg);
-                        if (!nids)
-                                return 1;
-                        rc = add_param(mop->mo_ldd.ldd_params, PARAM_FAILNODE,
-                                       nids);
-                        free(nids);
-                        if (rc)
-                                return rc;
-                        /* Must update the mgs logs */
-                        mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
-                        if (opt == 'f') {
-                                failnode_set = 1;
-                        } else {
-                                mop->mo_ldd.ldd_flags |= LDD_F_NO_PRIMNODE;
-                                servicenode_set = 1;
-                        }
-                        failover = 1;
-                        break;
-                }
-                case 'G':
-                        mop->mo_ldd.ldd_flags |= LDD_F_SV_TYPE_MGS;
-                        break;
-                case 'h':
-                        usage(stdout);
-                        return 1;
-                case 'i':
-                        if (!(mop->mo_ldd.ldd_flags &
-                              (LDD_F_UPGRADE14 | LDD_F_VIRGIN |
-                               LDD_F_WRITECONF))) {
-                                fprintf(stderr, "%s: cannot change the index of"
-                                        " a registered target\n", progname);
-                                return 1;
-                        }
-                        if (IS_MDT(&mop->mo_ldd) || IS_OST(&mop->mo_ldd)) {
-                                mop->mo_ldd.ldd_svindex = atol(optarg);
-                                mop->mo_ldd.ldd_flags &= ~LDD_F_NEED_INDEX;
-                        } else {
-                                badopt(long_opt[longidx].name, "MDT,OST");
-                                return 1;
-                        }
-                        break;
-                case 'k':
-                        strscpy(mop->mo_mkfsopts, optarg,
-                                sizeof(mop->mo_mkfsopts));
-                        break;
-                case 'L': {
-                        char *tmp;
-                        if (!(mop->mo_flags & MO_FORCEFORMAT) &&
-                            (!(mop->mo_ldd.ldd_flags &
-                               (LDD_F_UPGRADE14 | LDD_F_VIRGIN |
-                                LDD_F_WRITECONF)))) {
-                                fprintf(stderr, "%s: cannot change the name of"
-                                        " a registered target\n", progname);
-                                return 1;
-                        }
-                        if ((strlen(optarg) < 1) || (strlen(optarg) > 8)) {
-                                fprintf(stderr, "%s: filesystem name must be "
-                                        "1-8 chars\n", progname);
-                                return 1;
-                        }
-                        if ((tmp = strpbrk(optarg, "/:"))) {
-                                fprintf(stderr, "%s: char '%c' not allowed in "
-                                        "filesystem name\n", progname, *tmp);
-                                return 1;
-                        }
-                        strscpy(mop->mo_ldd.ldd_fsname, optarg,
-                                sizeof(mop->mo_ldd.ldd_fsname));
-                        break;
-                }
-                case 'm': {
-                        char *nids = convert_hostnames(optarg);
-                        if (!nids)
-                                return 1;
-                        rc = add_param(mop->mo_ldd.ldd_params, PARAM_MGSNODE,
-                                       nids);
-                        free(nids);
-                        if (rc)
-                                return rc;
-                        mop->mo_mgs_failnodes++;
-                        break;
-                }
-                case 'M':
-                        mop->mo_ldd.ldd_flags |= LDD_F_SV_TYPE_MDT;
-                        break;
-                case 'n':
-                        print_only++;
-                        break;
-                case 'N':
-                        mop->mo_ldd.ldd_flags &= ~LDD_F_SV_TYPE_MGS;
-                        break;
-                case 'o':
-                        *mountopts = optarg;
-                        break;
-                case 'O':
-                        mop->mo_ldd.ldd_flags |= LDD_F_SV_TYPE_OST;
-                        break;
-                case 'p':
-                        rc = add_param(mop->mo_ldd.ldd_params, NULL, optarg);
-                        if (rc)
-                                return rc;
-                        /* Must update the mgs logs */
-                        mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
-                        break;
-                case 'q':
-                        verbose--;
-                        break;
-                case 'r':
-                        mop->mo_flags |= MO_FORCEFORMAT;
-                        break;
-                case 't':
-                        if (!IS_MDT(&mop->mo_ldd) && !IS_OST(&mop->mo_ldd)) {
-                                badopt(long_opt[longidx].name, "MDT,OST");
-                                return 1;
-                        }
-
-                        if (!optarg)
-                                return 1;
-
-                        rc = add_param(mop->mo_ldd.ldd_params,
-                                       PARAM_NETWORK, optarg);
-                        if (rc != 0)
-                                return rc;
-                        /* Must update the mgs logs */
-                        mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
-                        break;
-                case 'u':
-                        strscpy(mop->mo_ldd.ldd_userdata, optarg,
-                                sizeof(mop->mo_ldd.ldd_userdata));
-                        break;
-                case 'v':
-                        verbose++;
-                        break;
-                case 'w':
-                        mop->mo_ldd.ldd_flags |= LDD_F_WRITECONF;
-                        break;
-                case 'U':
-                        upgrade_to_18 = 1;
-                        break;
-                default:
-                        if (opt != '?') {
-                                fatal();
-                                fprintf(stderr, "Unknown option '%c'\n", opt);
-                        }
-                        return EINVAL;
-                }
-        }//while
-
-        /* Last arg is device */
-        if (optind != argc - 1) {
-                fatal();
-                fprintf(stderr, "Bad argument: %s\n", argv[optind]);
-                return EINVAL;
-        }
-
-        /* single argument: <device> */
-        if (argc == 2)
-                ++print_only;
-
-        return 0;
-}
+       static struct option long_opts[] = {
+       { .val = 'B',   .name =  "backfs-mount-opts",
+                                               .has_arg = required_argument},
+       { .val = 'f',   .name =  "failnode",    .has_arg = required_argument},
+       { .val = 'f',   .name =  "failover",    .has_arg = required_argument},
+       { .val = 'G',   .name =  "mgs",         .has_arg = no_argument},
+       { .val = 'h',   .name =  "help",        .has_arg = no_argument},
+       { .val = 'i',   .name =  "index",       .has_arg = required_argument},
+       { .val = 'L',   .name =  "fsname",      .has_arg = required_argument},
+       { .val = 'm',   .name =  "mgsnode",     .has_arg = required_argument},
+       { .val = 'm',   .name =  "mgsnid",      .has_arg = required_argument},
+       { .val = 'n',   .name =  "dryrun",      .has_arg = no_argument},
+       { .val = 'N',   .name =  "nomgs",       .has_arg = no_argument},
+       { .val = 'o',   .name =  "mountfsoptions",
+                                               .has_arg = required_argument},
+       { .val = 'p',   .name =  "param",       .has_arg = required_argument},
+       { .val = 'q',   .name =  "quiet",       .has_arg = no_argument},
+       { .val = 's',   .name =  "servicenode", .has_arg = required_argument},
+       { .val = 't',   .name =  "network",     .has_arg = required_argument},
+       { .val = 'u',   .name =  "comment",     .has_arg = required_argument},
+       { .val = 'U',   .name =  "force-nohostid",
+                                               .has_arg = no_argument},
+       { .val = 'v',   .name =  "verbose",     .has_arg = no_argument},
+       { .val = 'V',   .name =  "version",     .has_arg = no_argument},
+#ifndef TUNEFS
+       { .val = 'b',   .name =  "backfstype",  .has_arg = required_argument},
+       { .val = 'c',   .name =  "stripe-count-hint",
+                                               .has_arg = required_argument},
+       { .val = 'd',   .name =  "device-size", .has_arg = required_argument},
+       { .val = 'k',   .name =  "mkfsoptions", .has_arg = required_argument},
+       { .val = 'M',   .name =  "mdt",         .has_arg = no_argument},
+       { .val = 'O',   .name =  "ost",         .has_arg = no_argument},
+       { .val = 'r',   .name =  "reformat",    .has_arg = no_argument},
+       { .val = 'R',   .name =  "replace",     .has_arg = no_argument},
+#else
+       { .val = 'E',   .name =  "erase-param", .has_arg = required_argument},
+       { .val = 'e',   .name =  "erase-params",
+                                               .has_arg = no_argument},
+       { .val = 'l',   .name =  "nolocallogs", .has_arg = no_argument},
+       { .val = 'Q',   .name =  "quota",       .has_arg = no_argument},
+       { .val = 'R',   .name =  "rename",      .has_arg = optional_argument},
+       { .val = 'w',   .name =  "writeconf",   .has_arg = no_argument},
+#endif
+       { .name = NULL } };
+       char *short_opts = "B:f:Ghi:L:m:nNo:p:qs:t:u:vV"
+#ifndef TUNEFS
+                         "b:c:d:k:MOrR";
+#else
+                         "E:elQR::w";
+#endif
+       struct lustre_disk_data *ldd = &mop->mo_ldd;
+       char new_fsname[16] = { 0 };
+       int opt;
+       int rc, longidx;
+       int failnode_set = 0, servicenode_set = 0;
+       int replace = 0;
+       bool index_option = false;
 
-/* Search for opt in mntlist, returning true if found.
- */
-static int in_mntlist(char *opt, char *mntlist)
-{
-        char *ml, *mlp, *item, *ctx = NULL;
-
-        if (!(ml = strdup(mntlist))) {
-                fprintf(stderr, "%s: out of memory\n", progname);
-                exit(1);
-        }
-        mlp = ml;
-        while ((item = strtok_r(mlp, ",", &ctx))) {
-                if (!strcmp(opt, item))
-                        break;
-                mlp = NULL;
-        }
-        free(ml);
-        return (item != NULL);
-}
+#ifdef TUNEFS
+       /*
+        * For the right semantics, if '-e'/'--erase-params' is specified,
+        * it must be picked out and all old parameters should be erased
+        * before any other changes are done.
+        */
+       while ((opt = getopt_long(argc, argv, short_opts, long_opts,
+                                 &longidx)) != EOF) {
+               switch (opt) {
+               case 'e':
+                       ldd->ldd_params[0] = '\0';
+                       mop->mo_flags |= MO_ERASE_ALL;
+                       ldd->ldd_flags |= LDD_F_UPDATE;
+                       break;
+               default:
+                       break;
+               }
+               if (mop->mo_flags & MO_ERASE_ALL)
+                       break;
+       }
+       optind = 0;
+#endif
+       while ((opt = getopt_long(argc, argv, short_opts, long_opts,
+                                 &longidx)) != EOF) {
+               switch (opt) {
+               case 'B':
+                       mop->mo_mountopts = optarg;
+                       break;
+               case 'f':
+               case 's': {
+                       char *nids;
+
+                       if ((opt == 'f' && servicenode_set) ||
+                           (opt == 's' && failnode_set)) {
+                               fprintf(stderr, "%s: %s cannot use with --%s\n",
+                                       progname, long_opts[longidx].name,
+                                       opt == 'f' ? "servicenode" :
+                                       "failnode");
+                               return 1;
+                       }
+
+                       nids = convert_hostnames(optarg);
+                       if (!nids)
+                               return 1;
+
+                       rc = append_param(ldd->ldd_params, PARAM_FAILNODE,
+                                         nids, ':');
+                       free(nids);
+                       if (rc != 0)
+                               return rc;
+
+                       /* Must update the mgs logs */
+                       ldd->ldd_flags |= LDD_F_UPDATE;
+                       if (opt == 'f') {
+                               ldd->ldd_flags &= ~LDD_F_NO_PRIMNODE;
+                               failnode_set = 1;
+                       } else {
+                               ldd->ldd_flags |= LDD_F_NO_PRIMNODE;
+                               servicenode_set = 1;
+                       }
+                       mop->mo_flags |= MO_FAILOVER;
+                       break;
+               }
+               case 'G':
+                       ldd->ldd_flags |= LDD_F_SV_TYPE_MGS;
+                       break;
+               case 'h':
+                       usage(stdout);
+                       return 1;
+               case 'i': {
+                       char *endptr = NULL;
+                       int base;
+
+                       index_option = true;
+                       /* LU-2374: check whether it is OST/MDT later */
+                       base = (strlen(optarg) > 1 &&
+                               !strncmp(optarg, "0x", 2)) ? 16 : 10;
+                       /* Allowed input are base 16 and base 10 numbers only */
+                       mop->mo_ldd.ldd_svindex = strtoul(optarg,
+                                                         &endptr, base);
+                       if (*endptr != '\0') {
+                               fprintf(stderr,
+                                       "%s: wrong index %s. Target index must be decimal or hexadecimal.\n",
+                                       progname, optarg);
+                               return 1;
+                       }
+                       if (ldd->ldd_svindex >= INDEX_UNASSIGNED) {
+                               fprintf(stderr,
+                                       "%s: wrong index %u. Target index must be less than %u.\n",
+                                       progname, ldd->ldd_svindex,
+                                       INDEX_UNASSIGNED);
+                               return 1;
+                       }
+
+                       ldd->ldd_flags &= ~LDD_F_NEED_INDEX;
+                       break;
+               }
+               case 'L': {
+                       const char *tmp;
+                       size_t len;
+
+                       len = strlen(optarg);
+                       if (len < 1 || len > LUSTRE_MAXFSNAME) {
+                               fprintf(stderr,
+                                       "%s: filesystem name must be 1-%d chars\n",
+                                       progname, LUSTRE_MAXFSNAME);
+                               return 1;
+                       }
+
+                       for (tmp = optarg; *tmp != '\0'; ++tmp) {
+                               if (isalnum(*tmp) || *tmp == '_' || *tmp == '-')
+                                       continue;
+                               else
+                                       break;
+                       }
+                       if (*tmp != '\0') {
+                               fprintf(stderr,
+                                       "%s: char '%c' not allowed in filesystem name\n",
+                                       progname, *tmp);
+                               return 1;
+                       }
+                       strscpy(new_fsname, optarg, sizeof(new_fsname));
+                       break;
+               }
+               case 'm': {
+                       char *nids = convert_hostnames(optarg);
+
+                       if (!nids)
+                               return 1;
+
+                       rc = append_param(ldd->ldd_params, PARAM_MGSNODE,
+                                         nids, ':');
+                       free(nids);
+                       if (rc != 0)
+                               return rc;
+
+                       mop->mo_mgs_failnodes++;
+                       break;
+               }
+               case 'n':
+                       print_only++;
+                       break;
+               case 'N':
+                       ldd->ldd_flags &= ~LDD_F_SV_TYPE_MGS;
+                       break;
+               case 'o':
+                       *mountopts = optarg;
+                       break;
+               case 'p':
+#ifdef TUNEFS
+                       /*
+                        * Removes all existing instances of the parameter
+                        * before adding new values.
+                        */
+                       rc = erase_param(ldd->ldd_params, optarg, true);
+                       if (rc > 1)
+                               return rc;
+#endif
+                       rc = add_param(ldd->ldd_params, NULL, optarg);
+                       if (rc != 0)
+                               return rc;
+                       /* Must update the mgs logs */
+                       ldd->ldd_flags |= LDD_F_UPDATE;
+                       break;
+               case 'q':
+                       verbose--;
+                       break;
+               case 't':
+                       if (!IS_MDT(ldd) && !IS_OST(ldd)) {
+                               badopt(long_opts[longidx].name, "MDT,OST");
+                               return 1;
+                       }
+
+                       if (!optarg)
+                               return 1;
+
+                       rc = add_param(ldd->ldd_params, PARAM_NETWORK, optarg);
+                       if (rc != 0)
+                               return rc;
+
+                       /* Must update the mgs logs */
+                       ldd->ldd_flags |= LDD_F_UPDATE;
+                       break;
+               case 'u':
+                       strscpy(ldd->ldd_userdata, optarg,
+                               sizeof(ldd->ldd_userdata));
+                       break;
+               case 'U':
+                       mop->mo_flags |= MO_NOHOSTID_CHECK;
+                       break;
+               case 'v':
+                       verbose++;
+                       break;
+               case 'V':
+                       ++version;
+                       fprintf(stdout, "%s %s\n", progname,
+                               LUSTRE_VERSION_STRING);
+                       return 0;
+#ifndef TUNEFS
+               case 'b': {
+                       int i = 0;
+
+                       do {
+                               if (strcmp(optarg, mt_str(i)) == 0) {
+                                       ldd->ldd_mount_type = i;
+                                       break;
+                               }
+                       } while (++i < LDD_MT_LAST);
+
+                       if (i == LDD_MT_LAST) {
+                               fprintf(stderr,
+                                       "%s: invalid backend filesystem type %s\n",
+                                       progname, optarg);
+                               return 1;
+                       }
+                       break;
+               }
+               case 'c':
+                       if (IS_MDT(ldd)) {
+                               int stripe_count = atol(optarg);
+
+                               if (stripe_count <= 0) {
+                                       fprintf(stderr,
+                                               "%s: bad stripe count %s\n",
+                                               progname, optarg);
+                                       return 1;
+                               }
+                               mop->mo_stripe_count = stripe_count;
+                       } else {
+                               badopt(long_opts[longidx].name, "MDT");
+                               return 1;
+                       }
+                       break;
+               case 'd':
+                       mop->mo_device_kb = atol(optarg);
+                       break;
+               case 'k':
+                       strscpy(mop->mo_mkfsopts, optarg,
+                               sizeof(mop->mo_mkfsopts));
+                       break;
+               case 'M':
+                       ldd->ldd_flags |= LDD_F_SV_TYPE_MDT;
+                       break;
+               case 'O':
+                       ldd->ldd_flags |= LDD_F_SV_TYPE_OST;
+                       break;
+               case 'r':
+                       mop->mo_flags |= MO_FORCEFORMAT;
+                       break;
+               case 'R':
+                       replace = 1;
+                       break;
+#else /* TUNEFS */
+               case 'E':
+                       rc = erase_param(ldd->ldd_params, optarg, false);
+                       /*
+                        * (rc == 1) means not found, so don't need to
+                        * call osd_erase_ldd().
+                        */
+                       if (rc > 1)
+                               return rc;
+                       if (!rc) {
+                               rc = osd_erase_ldd(mop, optarg);
+                               if (rc)
+                                       return rc;
+                       }
+                       /* Must update the mgs logs */
+                       ldd->ldd_flags |= LDD_F_UPDATE;
+                       break;
+               case 'e':
+                       /* Already done in the beginning */
+                       break;
+               case 'Q':
+                       mop->mo_flags |= MO_QUOTA;
+                       break;
+               case 'R': {
+                       char *tmp;
+
+                       mop->mo_flags |= MO_RENAME;
+                       if (!optarg) {
+                               if (IS_SEPARATED_MGS(ldd)) {
+                                       fprintf(stderr,
+                                               "%s: must specify the old fsname to be renamed for separated MGS\n",
+                                               progname);
+                                       return 1;
+                               }
+                               break;
+                       }
+
+                       if ((strlen(optarg) < 1) || (strlen(optarg) > 8)) {
+                               fprintf(stderr,
+                                       "%s: filesystem name must be 1-8 chars\n",
+                                       progname);
+                               return 1;
+                       }
+
+                       tmp = strpbrk(optarg, "/:");
+                       if (tmp) {
+                               fprintf(stderr,
+                                       "%s: char '%c' not allowed in filesystem name\n",
+                                       progname, *tmp);
+                               return 1;
+                       }
+
+                       if (IS_SEPARATED_MGS(ldd)) {
+                               strscpy(old_fsname, optarg,
+                                       sizeof(ldd->ldd_fsname));
+                       } else if (strlen(old_fsname) != strlen(optarg) ||
+                                  strcmp(old_fsname, optarg) != 0) {
+                               fprintf(stderr,
+                                       "%s: the given fsname '%s' to be renamed does not exist\n",
+                                       progname, optarg);
+                               return 1;
+                       }
+                       break;
+               }
+               case 'w':
+                       ldd->ldd_flags |= LDD_F_WRITECONF;
+                       break;
+               case 'l':
+                       if (ldd->ldd_flags & (LDD_F_VIRGIN | LDD_F_WRITECONF)) {
+                               fprintf(stderr, "Can not apply nolocallogs to the target that was writeconfed or never been registered\n");
+                               return EINVAL;
+                       }
+                       ldd->ldd_flags |= LDD_F_NO_LOCAL_LOGS;
+                       break;
+#endif /* !TUNEFS */
+               default:
+                       if (opt != '?') {
+                               fatal();
+                               fprintf(stderr, "Unknown option '%c'\n", opt);
+                       }
+                       return EINVAL;
+               }
+       }
+
+       if (strlen(new_fsname) > 0) {
+               if (!(mop->mo_flags & (MO_FORCEFORMAT | MO_RENAME)) &&
+                   (!(ldd->ldd_flags & (LDD_F_VIRGIN | LDD_F_WRITECONF)))) {
+                       fprintf(stderr,
+                               "%s: cannot change the name of a registered target\n",
+                               progname);
+                       return 1;
+               }
+
+               strscpy(ldd->ldd_fsname, new_fsname, sizeof(ldd->ldd_fsname));
+       }
+
+       if (index_option && !(mop->mo_ldd.ldd_flags &
+                             (LDD_F_VIRGIN | LDD_F_WRITECONF))) {
+               fprintf(stderr,
+                       "%s: cannot change the index of a registered target\n",
+                       progname);
+               return 1;
+       }
 
-/* Issue a message on stderr for every item in wanted_mountopts that is not
- * present in mountopts.  The justwarn boolean toggles between error and
- * warning message.  Return an error count.
- */
-static int check_mountfsoptions(char *mountopts, char *wanted_mountopts,
-                                int justwarn)
-{
-        char *ml, *mlp, *item, *ctx = NULL;
-        int errors = 0;
-
-        if (!(ml = strdup(wanted_mountopts))) {
-                fprintf(stderr, "%s: out of memory\n", progname);
-                exit(1);
-        }
-        mlp = ml;
-        while ((item = strtok_r(mlp, ",", &ctx))) {
-                if (!in_mntlist(item, mountopts)) {
-                        fprintf(stderr, "%s: %s mount option `%s' is missing\n",
-                                progname, justwarn ? "Warning: default"
-                                : "Error: mandatory", item);
-                        errors++;
-                }
-                mlp = NULL;
-        }
-        free(ml);
-        return errors;
-}
+#ifdef TUNEFS
+       if (mop->mo_flags & MO_RENAME) {
+               if (new_fsname[0] == '\0') {
+                       fprintf(stderr,
+                               "%s: need to specify new fsname for renaming case\n",
+                               progname);
+                       return 1;
+               }
+
+               if (strcmp(old_fsname, new_fsname) == 0) {
+                       fprintf(stderr,
+                               "%s: cannot rename fsname '%s' to the same name\n",
+                               progname, old_fsname);
+                       return 1;
+               }
+       }
+#endif
 
-/* Trim embedded white space, leading and trailing commas from string s.
- */
-static void trim_mountfsoptions(char *s)
-{
-        char *p;
-
-        for (p = s; *p; ) {
-                if (isspace(*p)) {
-                        memmove(p, p + 1, strlen(p + 1) + 1);
-                        continue;
-                }
-                p++;
-        }
-
-        while (s[0] == ',')
-                memmove(&s[0], &s[1], strlen(&s[1]) + 1);
-
-        p = s + strlen(s) - 1;
-        while (p >= s && *p == ',')
-                *p-- = '\0';
+       /* Need to clear this flag after parsing 'L' and 'i' options. */
+       if (replace)
+               ldd->ldd_flags &= ~LDD_F_VIRGIN;
+
+       if (optind == argc) {
+               /* The user didn't specify device name */
+               fatal();
+               fprintf(stderr,
+                       "Not enough arguments - device name or pool/dataset name not specified.\n");
+               return EINVAL;
+       }
+
+       /*  The device or pool/filesystem name */
+       strscpy(mop->mo_device, argv[optind], sizeof(mop->mo_device));
+
+       /* Followed by optional vdevs */
+       if (optind < argc - 1)
+               mop->mo_pool_vdevs = (char **)&argv[optind + 1];
+
+       return 0;
 }
 
 int main(int argc, char *const argv[])
 {
-        struct mkfs_opts mop;
-        struct lustre_disk_data *ldd;
-        char *mountopts = NULL;
-        char always_mountopts[512] = "";
-        char default_mountopts[512] = "";
-        int ret = 0;
-
-        if ((progname = strrchr(argv[0], '/')) != NULL)
-                progname++;
-        else
-                progname = argv[0];
-
-        if ((argc < 2) || (argv[argc - 1][0] == '-')) {
-                usage(stderr);
-                return(EINVAL);
-        }
-
-        memset(&mop, 0, sizeof(mop));
-        set_defaults(&mop);
-
-        /* device is last arg */
-        strscpy(mop.mo_device, argv[argc - 1], sizeof(mop.mo_device));
-
-        /* Are we using a loop device? */
-        ret = is_block(mop.mo_device);
-        if (ret < 0) {
-                ret = errno;
-                goto out;
-        }
-        if (ret == 0)
-                mop.mo_flags |= MO_IS_LOOP;
+       struct mkfs_opts mop;
+       struct lustre_disk_data *ldd = &mop.mo_ldd;
+       char *mountopts = NULL;
+       char wanted_mountopts[512] = "";
+       char old_fsname[16] = "";
+       unsigned int mount_type;
+       int ret = 0;
+       int ret2 = 0;
+
+       progname = strrchr(argv[0], '/');
+       if (progname)
+               progname++;
+       else
+               progname = argv[0];
+
+       if ((argc < 2) || (argv[argc - 1][0] == '-')) {
+               usage(stderr);
+               return EINVAL;
+       }
+
+       memset(&mop, 0, sizeof(mop));
+       set_defaults(&mop);
+
+       /* device is last arg */
+       strscpy(mop.mo_device, argv[argc - 1], sizeof(mop.mo_device));
+
+       ret = osd_init();
+       if (ret != 0) {
+               fprintf(stderr, "%s: osd_init() failed: %d (%s)\n",
+                       progname, ret, strerror(ret));
+               return ret;
+       }
 
 #ifdef TUNEFS
-        /* For tunefs, we must read in the old values before parsing any
-           new ones. */
-
-        /* Check whether the disk has already been formatted by mkfs.lustre */
-        ret = is_lustre_target(&mop);
-        if (ret == 0) {
-                fatal();
-                fprintf(stderr, "Device %s has not been formatted with "
-                        "mkfs.lustre\n", mop.mo_device);
-                ret = ENODEV;
-                goto out;
-        }
-
-        ret = read_local_files(&mop);
-        if (ret) {
-                fatal();
-                fprintf(stderr, "Failed to read previous Lustre data from %s "
-                        "(%d)\n", mop.mo_device, ret);
-                goto out;
-        }
-        if (strstr(mop.mo_ldd.ldd_params, PARAM_MGSNODE))
-            mop.mo_mgs_failnodes++;
-
-        if (verbose > 0)
-                print_ldd("Read previous values", &(mop.mo_ldd));
+       /*
+        * For tunefs, we must read in the old values before parsing any
+        * new ones.
+        */
+
+       /* Check whether the disk has already been formatted by mkfs.lustre */
+       ret = osd_is_lustre(mop.mo_device, &mount_type);
+       if (ret == 0) {
+               fatal();
+               fprintf(stderr,
+                       "Device %s has not been formatted with mkfs.lustre\n",
+                       mop.mo_device);
+               ret = ENODEV;
+               goto out;
+       }
+       ldd->ldd_mount_type = mount_type;
+
+       ret = osd_read_ldd(mop.mo_device, ldd);
+       if (ret != 0) {
+               fatal();
+               fprintf(stderr,
+                       "Failed to read previous Lustre data from %s (%d)\n",
+                       mop.mo_device, ret);
+               goto out;
+       }
+
+       strscpy(old_fsname, ldd->ldd_fsname, sizeof(ldd->ldd_fsname));
+       ldd->ldd_flags &= ~(LDD_F_WRITECONF | LDD_F_VIRGIN |
+                           LDD_F_NO_LOCAL_LOGS);
+
+       /* svname of the form lustre:OST1234 means never registered */
+       ret = strlen(ldd->ldd_svname);
+       if (ldd->ldd_svname[ret - 8] == ':') {
+               ldd->ldd_svname[ret - 8] = '-';
+               ldd->ldd_flags |= LDD_F_VIRGIN;
+       } else if (ldd->ldd_svname[ret - 8] == '=') {
+               ldd->ldd_svname[ret - 8] = '-';
+               ldd->ldd_flags |= LDD_F_WRITECONF;
+       } else if (ldd->ldd_svname[ret - 8] == '+') {
+               ldd->ldd_svname[ret - 8] = '-';
+               ldd->ldd_flags |= LDD_F_NO_LOCAL_LOGS;
+       }
+
+       if (strstr(ldd->ldd_params, PARAM_MGSNODE))
+               mop.mo_mgs_failnodes++;
+
+       if (verbose > 0)
+               print_ldd("Read previous values", &mop);
+#endif /* TUNEFS */
+
+       ret = parse_opts(argc, argv, &mop, &mountopts, old_fsname);
+       if (ret != 0 || version)
+               goto out;
+
+       if (!IS_MDT(ldd) && !IS_OST(ldd) && !IS_MGS(ldd)) {
+               fatal();
+               fprintf(stderr, "must set target type: MDT,OST,MGS\n");
+               ret = EINVAL;
+               goto out;
+       }
+
+       if (((IS_MDT(ldd) || IS_MGS(ldd))) && IS_OST(ldd)) {
+               fatal();
+               fprintf(stderr, "OST type is exclusive with MDT,MGS\n");
+               ret = EINVAL;
+               goto out;
+       }
+
+       /* Stand alone MGS doesn't need an index */
+       if (!IS_MDT(ldd) && IS_MGS(ldd)) {
+#ifndef TUNEFS
+               /* But if --index was specified flag an error */
+               if (!(ldd->ldd_flags & LDD_F_NEED_INDEX)) {
+                       badopt("index", "MDT,OST");
+                       goto out;
+               }
 #endif
-
-        ret = parse_opts(argc, argv, &mop, &mountopts);
-        if (ret)
-                goto out;
-
-        ldd = &mop.mo_ldd;
-
-        if (!(IS_MDT(ldd) || IS_OST(ldd) || IS_MGS(ldd))) {
-                fatal();
-                fprintf(stderr, "must set target type: MDT,OST,MGS\n");
-                ret = EINVAL;
-                goto out;
-        }
-
-        if (((IS_MDT(ldd) || IS_MGS(ldd))) && IS_OST(ldd)) {
-                fatal();
-                fprintf(stderr, "OST type is exclusive with MDT,MGS\n");
-                ret = EINVAL;
-                goto out;
-        }
-
-        if ((mop.mo_ldd.ldd_flags & (LDD_F_NEED_INDEX | LDD_F_UPGRADE14)) ==
-            (LDD_F_NEED_INDEX | LDD_F_UPGRADE14)) {
-                fatal();
-                fprintf(stderr, "Can't find the target index, "
-                        "specify with --index\n");
-                ret = EINVAL;
-                goto out;
-        }
-#if 0
-        /*
-         * Comment out these 2 checks temporarily, since for multi-MDSes
-         * in single node only 1 mds node could have mgs service
-         */
-        if (IS_MDT(ldd) && !IS_MGS(ldd) && (mop.mo_mgs_failnodes == 0)) {
-                verrprint("No management node specified, adding MGS to this "
-                          "MDT\n");
-                ldd->ldd_flags |= LDD_F_SV_TYPE_MGS;
-        }
-        if (!IS_MGS(ldd) && (mop.mo_mgs_failnodes == 0)) {
-                fatal();
-                if (IS_MDT(ldd))
-                        fprintf(stderr, "Must specify --mgs or --mgsnode=\n");
-                else
-                        fprintf(stderr, "Must specify --mgsnode=\n");
-                ret = EINVAL;
-                goto out;
-        }
+               ldd->ldd_flags &= ~LDD_F_NEED_INDEX;
+       }
+
+       if (ldd->ldd_flags & LDD_F_NEED_INDEX)
+               fprintf(stderr,
+                       "warning: %s: for Lustre 2.4 and later, the target index must be specified with --index\n",
+                       mop.mo_device);
+
+       /* If no index is supplied for MDT by default set index to zero */
+       if (IS_MDT(ldd) && (ldd->ldd_svindex == INDEX_UNASSIGNED)) {
+               ldd->ldd_flags &= ~LDD_F_NEED_INDEX;
+               ldd->ldd_svindex = 0;
+       }
+#ifndef TUNEFS
+       if (!IS_MGS(ldd) && (mop.mo_mgs_failnodes == 0)) {
+#else
+       /*
+        * Don't check --mgs or --mgsnode if print_only is set or
+        * --erase-params is set.
+        */
+       if (!IS_MGS(ldd) && (mop.mo_mgs_failnodes == 0) && !print_only &&
+           !(mop.mo_flags & MO_ERASE_ALL)) {
 #endif
-
-        /* These are the permanent mount options (always included) */
-        switch (ldd->ldd_mount_type) {
-        case LDD_MT_EXT3:
-        case LDD_MT_LDISKFS:
-        case LDD_MT_LDISKFS2:
-                strscat(default_mountopts, ",errors=remount-ro",
-                        sizeof(default_mountopts));
-                if (IS_MDT(ldd) || IS_MGS(ldd))
-                        strscat(always_mountopts, ",user_xattr",
-                                sizeof(always_mountopts));
-                /* NB: Files created while extents are enabled can only be read
-                 * if mounted using the ext4 or ldiskfs filesystem type. */
-                if (IS_OST(ldd) &&
-                    (ldd->ldd_mount_type == LDD_MT_LDISKFS ||
-                     ldd->ldd_mount_type == LDD_MT_LDISKFS2)) {
-                        strscat(default_mountopts, ",extents,mballoc",
-                                sizeof(default_mountopts));
-                }
-                break;
-        default:
-                fatal();
-                fprintf(stderr, "unknown fs type %d '%s'\n",
-                        ldd->ldd_mount_type,
-                        MT_STR(ldd));
-                ret = EINVAL;
-                goto out;
-        }
-
-        if (mountopts) {
-                trim_mountfsoptions(mountopts);
-                (void)check_mountfsoptions(mountopts, default_mountopts, 1);
-                if (check_mountfsoptions(mountopts, always_mountopts, 0)) {
-                        ret = EINVAL;
-                        goto out;
-                }
-                sprintf(ldd->ldd_mount_opts, "%s", mountopts);
-        } else {
+               fatal();
+               if (IS_MDT(ldd))
+                       fprintf(stderr, "Must specify --mgs or --mgsnode\n");
+               else
+                       fprintf(stderr, "Must specify --mgsnode\n");
+               ret = EINVAL;
+               goto out;
+       }
+       if ((IS_MDT(ldd) || IS_OST(ldd)) && ldd->ldd_fsname[0] == '\0') {
+               fatal();
+               fprintf(stderr, "Must specify --fsname for MDT/OST device\n");
+               ret = EINVAL;
+               goto out;
+       }
+
+       /* These are the permanent mount options (always included) */
+       ret = osd_prepare_lustre(&mop,
+                                wanted_mountopts, sizeof(wanted_mountopts));
+       if (ret != 0) {
+               fatal();
+               fprintf(stderr, "unable to prepare backend (%d)\n", ret);
+               goto out;
+       }
+
+       if (mountopts) {
+               trim_mountfsoptions(mountopts);
+               if (check_mountfsoptions(mountopts, wanted_mountopts)) {
+                       ret = EINVAL;
+                       goto out;
+               }
+               snprintf(ldd->ldd_mount_opts, sizeof(ldd->ldd_mount_opts),
+                        "%s", mountopts);
+       } else {
 #ifdef TUNEFS
-                if (ldd->ldd_mount_opts[0] == 0)
-                        /* use the defaults unless old opts exist */
-#endif
-                {
-                        sprintf(ldd->ldd_mount_opts, "%s%s",
-                                always_mountopts, default_mountopts);
-                        trim_mountfsoptions(ldd->ldd_mount_opts);
-                }
-        }
-
-        server_make_name(ldd->ldd_flags, ldd->ldd_svindex,
-                         ldd->ldd_fsname, ldd->ldd_svname);
-
-        if (verbose >= 0)
-                print_ldd("Permanent disk data", ldd);
-
-        if (print_only) {
-                printf("exiting before disk write.\n");
-                goto out;
-        }
-
-        if (check_mtab_entry(mop.mo_device))
-                return(EEXIST);
-
-        /* Create the loopback file */
-        if (mop.mo_flags & MO_IS_LOOP) {
-                ret = access(mop.mo_device, F_OK);
-                if (ret)
-                        ret = errno;
-#ifndef TUNEFS /* mkfs.lustre */
-                /* Reformat the loopback file */
-                if (ret || (mop.mo_flags & MO_FORCEFORMAT)) {
-                        ret = loop_format(&mop);
-                        if (ret)
-                                goto out;
-                }
-#endif
-                if (ret == 0)
-                        ret = loop_setup(&mop);
-                if (ret) {
-                        fatal();
-                        fprintf(stderr, "Loop device setup for %s failed: %s\n",
-                                mop.mo_device, strerror(ret));
-                        goto out;
-                }
-        }
-
-#ifndef TUNEFS /* mkfs.lustre */
-        /* Check whether the disk has already been formatted by mkfs.lustre */
-        if (!(mop.mo_flags & MO_FORCEFORMAT)) {
-                ret = is_lustre_target(&mop);
-                if (ret) {
-                        fatal();
-                        fprintf(stderr, "Device %s was previously formatted "
-                                "for lustre. Use --reformat to reformat it, "
-                                "or tunefs.lustre to modify.\n",
-                                mop.mo_device);
-                        goto out;
-                }
-        }
-
-        /* Format the backing filesystem */
-        ret = make_lustre_backfs(&mop);
-        if (ret != 0) {
-                fatal();
-                fprintf(stderr, "mkfs failed %d\n", ret);
-                goto out;
-        }
+               if (ldd->ldd_mount_opts[0] == 0)
+               /* use the defaults unless old opts exist */
 #endif
+               {
+                       snprintf(ldd->ldd_mount_opts,
+                                sizeof(ldd->ldd_mount_opts),
+                                "%s", wanted_mountopts);
+                       trim_mountfsoptions(ldd->ldd_mount_opts);
+               }
+       }
+
+       ret = osd_fix_mountopts(&mop, ldd->ldd_mount_opts,
+                               sizeof(ldd->ldd_mount_opts));
+       if (ret != 0) {
+               fatal();
+               fprintf(stderr, "unable to fix mountfsoptions (%d)\n", ret);
+               goto out;
+       }
+
+       if (server_make_name(ldd->ldd_flags, ldd->ldd_svindex,
+                            ldd->ldd_fsname, ldd->ldd_svname,
+                            sizeof(ldd->ldd_svname))) {
+               printf("unknown server type %#x\n", ldd->ldd_flags);
+               goto out;
+       }
+
+       if (verbose >= 0)
+               print_ldd("Permanent disk data", &mop);
+
+       if (print_only) {
+               printf("exiting before disk write.\n");
+               goto out;
+       }
+
+       if (check_mtab_entry(mop.mo_device, mop.mo_device, NULL, NULL)) {
+               fprintf(stderr, "%s: is currently mounted, exiting without any change\n",
+                       mop.mo_device);
+               return EEXIST;
+       }
+
+       /* Create the loopback file */
+       if (mop.mo_flags & MO_IS_LOOP) {
+               ret = access(mop.mo_device, F_OK);
+               if (ret != 0)
+                       ret = errno;
 
-        /* Write our config files */
-        ret = write_local_files(&mop);
-        if (ret != 0) {
-                fatal();
-                fprintf(stderr, "failed to write local files\n");
-                goto out;
-        }
+#ifndef TUNEFS
+               /* Reformat the loopback file */
+               if (ret != 0 || (mop.mo_flags & MO_FORCEFORMAT)) {
+                       ret = loop_format(&mop);
+                       if (ret != 0)
+                               goto out;
+               }
+#endif
+               if (ret == 0)
+                       ret = loop_setup(&mop);
+               if (ret != 0) {
+                       fatal();
+                       fprintf(stderr, "Loop device setup for %s failed: %s\n",
+                               mop.mo_device, strerror(ret));
+                       goto out;
+               }
+       }
 
+#ifndef TUNEFS
+       /* Check whether the disk has already been formatted by mkfs.lustre */
+       if (!(mop.mo_flags & MO_FORCEFORMAT)) {
+               ret = osd_is_lustre(mop.mo_device, &mount_type);
+               if (ret != 0) {
+                       fatal();
+                       fprintf(stderr,
+                               "Device %s was previously formatted for lustre. Use --reformat to reformat it, or tunefs.lustre to modify.\n",
+                               mop.mo_device);
+                       goto out;
+               }
+       }
+
+       /* Format the backing filesystem */
+       ret = osd_make_lustre(&mop);
+       if (ret != 0) {
+               fatal();
+               fprintf(stderr, "mkfs failed %d\n", ret);
+               goto out;
+       }
+#else /* TUNEFS */
+       /* update svname with '=' to refresh config */
+       if (ldd->ldd_flags & LDD_F_WRITECONF) {
+               struct mount_opts opts;
+
+               opts.mo_ldd = *ldd;
+               opts.mo_source = mop.mo_device;
+               (void)osd_label_lustre(&opts);
+       }
+
+       /* update svname with '+' to force remote logs */
+       if (ldd->ldd_flags & LDD_F_NO_LOCAL_LOGS) {
+               struct mount_opts opts;
+
+               opts.mo_ldd = *ldd;
+               opts.mo_source = mop.mo_device;
+               (void) osd_label_lustre(&opts);
+       }
+
+       /* Rename filesystem fsname */
+       if (mop.mo_flags & MO_RENAME) {
+               ret = osd_rename_fsname(&mop, old_fsname);
+               if (ret)
+                       goto out;
+       }
+
+       /* Enable quota accounting */
+       if (mop.mo_flags & MO_QUOTA) {
+               ret = osd_enable_quota(&mop);
+               goto out;
+       }
+#endif /* !TUNEFS */
+
+       /* Write our config files */
+       ret = osd_write_ldd(&mop);
+       if (ret != 0) {
+               fatal();
+               fprintf(stderr, "failed to write local files\n");
+               goto out;
+       }
 out:
-        loop_cleanup(&mop);
-
-        /* Fix any crazy return values from system() */
-        if (ret && ((ret & 255) == 0))
-                return (1);
-        if (ret)
-                verrprint("%s: exiting with %d (%s)\n",
-                          progname, ret, strerror(ret));
-        return (ret);
+       osd_fini();
+       ret2 = loop_cleanup(&mop);
+       if (ret == 0)
+               ret = ret2;
+
+       /* Fix any crazy return values from system() */
+       if (ret != 0 && ((ret & 255) == 0))
+               return 1;
+
+       if (ret != 0)
+               verrprint("%s: exiting with %d (%s)\n",
+                         progname, ret, strerror(ret));
+       return ret;
 }