-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
+/*
+ * GPL HEADER START
*
- * Copyright (C) 2006 Cluster File Systems, Inc.
- * Author: Nathan Rutman <nathan@clusterfs.com>
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * This file is part of Lustre, http://www.lustre.org.
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
*
- * Lustre is free software; you can redistribute it and/or
- * modify it under the terms of version 2 of the GNU General Public
- * License as published by the Free Software Foundation.
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
*
- * Lustre is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
+ * 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
*
- * You should have received a copy of the GNU General Public License
- * along with Lustre; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ * 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.
*
+ * GPL HEADER END
*/
- /* This source file is compiled into both mkfs.lustre and tunefs.lustre */
+/*
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, Whamcloud, Inc.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
+ *
+ * lustre/utils/mkfs_lustre.c
+ *
+ * Author: Nathan Rutman <nathan@clusterfs.com>
+*/
+/* This source file is compiled into both mkfs.lustre and tunefs.lustre */
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif /* HAVE_CONFIG_H */
+
+#ifndef _GNU_SOURCE
#define _GNU_SOURCE
+#endif
+#include "mount_utils.h"
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mount.h>
+#include <sys/utsname.h>
#include <string.h>
#include <getopt.h>
#include <limits.h>
+#include <ctype.h>
#ifdef __linux__
-# include <linux/fs.h> /* for BLKGETSIZE64 */
+/* 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>
#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_mount_type_string[20]; /* "ext3", "ldiskfs", ... */
- char mo_device[128]; /* disk device name */
- char mo_mkfsopts[128]; /* options to the backing-store mkfs */
- char mo_loopdev[128]; /* in case a loop dev is needed */
- __u64 mo_device_sz; /* in KB */
- int mo_stripe_count;
- int mo_flags;
- int mo_mgs_failnodes;
-};
-
-static char *progname;
-static int verbose = 1;
+char *progname;
+int verbose = 1;
static int print_only = 0;
+static int upgrade_to_18 = 0;
-
-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--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 the lov)\n"
- /* FIXME implement 1.6.x
- "\t\t--configdev=<altdevice|file>: store configuration info\n"
- "\t\t\tfor this device on an alternate device\n"
- */
- "\t\t--comment=<user comment>: arbitrary user string (%d bytes)\n"
- "\t\t--mountfsoptions=<opts> : permanent mount options\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 : used for optimizing MDT inode size\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--noformat: 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
-
-static void fatal(void)
-{
- verbose = 0;
- fprintf(stderr, "\n%s FATAL: ", progname);
-}
-
-/*================ utility functions =====================*/
-
-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 {
- read(fd, release, 4);
- close(fd);
- }
- if (strncmp(release, "2.4.", 4) == 0)
- version = 24;
- else
- version = 26;
- }
- return version;
-}
-
-int run_command(char *cmd, int cmdsz)
-{
- char log[] = "/tmp/mkfs_logXXXXXX";
- int fd = -1, rc;
-
- if ((cmdsz - strlen(cmd)) < 6) {
- fatal();
- fprintf(stderr, "Command buffer overflow: %.*s...\n",
- cmdsz, cmd);
- return ENOMEM;
- }
-
- if (verbose > 1) {
- printf("cmd: %s\n", cmd);
- } else {
- if ((fd = mkstemp(log)) >= 0) {
- close(fd);
- strcat(cmd, " >");
- strcat(cmd, log);
- }
- }
- strcat(cmd, " 2>&1");
-
- /* Can't use popen because we need the rv of the command */
- rc = system(cmd);
- if (rc && (fd >= 0)) {
- char buf[128];
- FILE *fp;
- fp = fopen(log, "r");
- if (fp) {
- while (fgets(buf, sizeof(buf), fp) != NULL) {
- printf(" %s", buf);
- }
- fclose(fp);
- }
- }
- if (fd >= 0)
- remove(log);
- return rc;
-}
-
-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) {
- fprintf(stderr, "%s: error %d on losetup: %s\n",
- progname, ret, strerror(ret));
- return ret;
- }
- strcpy(mop->mo_loopdev, l_device);
- return ret;
- }
- }
-
- fprintf(stderr, "%s: out of loop devices!\n", progname);
- return EMFILE;
-}
-
-int loop_cleanup(struct mkfs_opts *mop)
-{
- char cmd[128];
- 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;
-}
+ #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
+ #define FSLIST_ZFS ""
+#endif /* HAVE_ZFS_OSD */
-/* Determine if a device is a block device (as opposed to a file) */
-int is_block(char* devname)
-{
- struct stat st;
- int ret = 0;
+#ifndef HAVE_FSLIST
+ #error "no backing OSD types (ldiskfs or ZFS) are configured"
+#endif
- ret = access(devname, F_OK);
- if (ret != 0)
- return 0;
- ret = stat(devname, &st);
- if (ret != 0) {
- fprintf(stderr, "%s: cannot stat %s\n", progname, devname);
- return -1;
- }
- return S_ISBLK(st.st_mode);
-}
+#define FSLIST FSLIST_LDISKFS FSLIST_ZFS
-__u64 get_device_size(char* device)
+void usage(FILE *out)
{
- 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);
+ fprintf(out, "%s v"LUSTRE_VERSION_STRING"\n", progname);
+ 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--mdt: metadata storage, mutually exclusive with ost\n"
+ "\t\t--ost: object storage, mutually exclusive with mdt, mgs\n"
+ "\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--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 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 (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"
#else
- {
- __u32 lsize = 0;
- /* size in blocks */
- ret = ioctl(fd, BLKGETSIZE, (void*)&lsize);
- size = (__u64)lsize * 512;
- }
+ "\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"
+ "\t\t--quota: enable space accounting on old 2.x device.\n"
#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;
-}
-
-int loop_format(struct mkfs_opts *mop)
-{
- int ret = 0;
-
- if (mop->mo_device_sz == 0) {
- fatal();
- fprintf(stderr, "loop device requires a --device-size= "
- "param\n");
- return EINVAL;
- }
-
- ret = creat(mop->mo_device, S_IRUSR|S_IWUSR);
- ret = truncate(mop->mo_device, mop->mo_device_sz * 1024);
- if (ret != 0) {
- ret = errno;
- fprintf(stderr, "%s: Unable to create backing store: %d\n",
- progname, ret);
- }
-
- return ret;
-}
-
-/* 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),
- "debugfs -c -R 'stat %s' %s 2>&1 | egrep '(Inode|unsupported)'",
- 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")) {
- fprintf(stderr, "In all likelihood, the "
- "'unsupported feature' is 'extents', which "
- "older debugfs does not understand.\n"
- "Use e2fsprogs-1.38-cfs1 or later, available "
- "from ftp://ftp.lustre.org/pub/lustre/other/"
- "e2fsprogs/\n");
- }
- 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;
- 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. */
-}
-
-/* Build fs according to type */
-int make_lustre_backfs(struct mkfs_opts *mop)
-{
- char mkfs_cmd[PATH_MAX];
- char buf[40];
- char *dev;
- int ret = 0;
- int block_count = 0;
- int left = sizeof(mkfs_cmd);
-
- 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 %lluKB\n",
- progname, (long long)mop->mo_device_sz);
- return EINVAL;
- }
- block_count = mop->mo_device_sz / (L_BLOCK_SIZE >> 10);
- }
-
- if ((mop->mo_ldd.ldd_mount_type == LDD_MT_EXT3) ||
- (mop->mo_ldd.ldd_mount_type == LDD_MT_LDISKFS)) {
- __u64 device_sz = mop->mo_device_sz;
-
- /* we really need the size */
- if (device_sz == 0) {
- device_sz = get_device_size(mop->mo_device);
- if (device_sz == 0)
- return ENODEV;
- }
-
- /* 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;
- /* 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);
- strcat(mop->mo_mkfsopts, buf);
- }
- }
-
- /* bytes_per_inode: disk size / num inodes */
- if (strstr(mop->mo_mkfsopts, "-i") == NULL) {
- long bytes_per_inode = 0;
-
- if (IS_MDT(&mop->mo_ldd))
- bytes_per_inode = 4096;
-
- /* Allocate fewer inodes on large OST devices. Most
- filesystems can be much more aggressive than even
- this. */
- if ((IS_OST(&mop->mo_ldd) && (device_sz > 1000000)))
- bytes_per_inode = 16384;
-
- if (bytes_per_inode > 0) {
- sprintf(buf, " -i %ld", bytes_per_inode);
- strcat(mop->mo_mkfsopts, buf);
- }
- }
-
- /* Inode size (for extended attributes) */
- if (strstr(mop->mo_mkfsopts, "-I") == NULL) {
- long inode_size = 0;
- if (IS_MDT(&mop->mo_ldd)) {
- if (mop->mo_stripe_count > 72)
- inode_size = 512; /* bz 7241 */
- /* cray stripes across all osts (>60) */
- 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)) {
- /* now as we store fids in EA on OST we need
- to make inode bigger */
- inode_size = 256;
- }
-
- if (inode_size > 0) {
- sprintf(buf, " -I %ld", inode_size);
- strcat(mop->mo_mkfsopts, buf);
- }
-
- }
-
- if (verbose < 2) {
- strcat(mop->mo_mkfsopts, " -q");
- }
-
- /* Enable hashed b-tree directory lookup in large dirs bz6224 */
- if (strstr(mop->mo_mkfsopts, "-O") == NULL) {
- strcat(mop->mo_mkfsopts, " -O dir_index");
- }
-
- /* Allow reformat of full devices (as opposed to
- partitions.) We already checked for mounted dev. */
- strcat(mop->mo_mkfsopts, " -F");
-
- left -= snprintf(mkfs_cmd, left,
- "mkfs.ext2 -j -b %d -L %s ", L_BLOCK_SIZE,
- mop->mo_ldd.ldd_svname);
-
- } else if (mop->mo_ldd.ldd_mount_type == LDD_MT_REISERFS) {
- long journal_sz = 0; /* FIXME default journal size */
- if (journal_sz > 0) {
- sprintf(buf, " --journal_size %ld", journal_sz);
- strcat(mop->mo_mkfsopts, buf);
- }
- left -= snprintf(mkfs_cmd, left, "mkreiserfs -ff ");
-
- } 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 %d\n", block_count);
- vprint("\toptions %s\n", mop->mo_mkfsopts);
-
- /* mkfs_cmd's trailing space is important! */
- strncat(mkfs_cmd, mop->mo_mkfsopts, left);
- left = sizeof(mkfs_cmd) - strlen(mkfs_cmd) - 1;
- strncat(mkfs_cmd, " ", left);
- left = sizeof(mkfs_cmd) - strlen(mkfs_cmd) - 1;
- strncat(mkfs_cmd, dev, left);
- left = sizeof(mkfs_cmd) - strlen(mkfs_cmd) - 1;
- if (block_count != 0) {
- sprintf(buf, " %d", block_count);
- strncat(mkfs_cmd, buf, left);
- left = sizeof(mkfs_cmd) - strlen(mkfs_cmd) - 1;
- }
-
- 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;
+ "\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--quiet\n",
+ (int)sizeof(((struct lustre_disk_data *)0)->ldd_userdata));
+ return;
}
/* ==================== Lustre config functions =============*/
{
printf("\n %s:\n", str);
printf("Target: %s\n", ldd->ldd_svname);
- if (ldd->ldd_svindex == INDEX_UNASSIGNED)
+ if (ldd->ldd_svindex == INDEX_UNASSIGNED)
printf("Index: unassigned\n");
else
printf("Index: %d\n", ldd->ldd_svindex);
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)\n",
- IS_MDT(ldd) ? "MDT ":"",
+ 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_UPGRADE14 ? "upgrade1.4 ":"");
printf("Persistent mount opts: %s\n", ldd->ldd_mount_opts);
printf("Parameters:%s\n", ldd->ldd_params);
printf("\n");
}
-/* Write the server config files */
-int write_local_files(struct mkfs_opts *mop)
-{
- char mntpt[] = "/tmp/mntXXXXXX";
- char filepnm[128];
- char *dev;
- FILE *filep;
- int ret = 0;
-
- /* 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, NULL);
- 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 = mkdir(filepnm, 0777);
- 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;
- }
- fwrite(&mop->mo_ldd, sizeof(mop->mo_ldd), 1, filep);
- 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;
- strcpy(filepnm, mop->mo_ldd.ldd_uuid);
- 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;
- }
- }
-#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;
-
- /* Construct debugfs command line. */
- snprintf(cmd, cmdsz, "debugfs -c -R 'dump /%s %s/mountdata' %s",
- 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) {
- vprint("Reading %s\n", MOUNT_DATA_FILE);
- fread(&mop->mo_ldd, sizeof(mop->mo_ldd), 1, filep);
- } else {
- /* COMPAT_146 */
- /* Try to read pre-1.6 config from last_rcvd */
- struct lr_server_data lsd;
- verrprint("%s: Unable to read %s (%s).\n",
- progname, filepnm, strerror(errno));
-
- verrprint("Trying last_rcvd\n");
- sprintf(filepnm, "%s/%s", tmpdir, LAST_RCVD);
-
- /* Construct debugfs command line. */
- snprintf(cmd, cmdsz, "debugfs -c -R 'dump /%s %s' %s",
- 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, "debugfs -c -R 'ls -l /' %s", 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, "debugfs -c -R 'rdump /%s %s' %s",
- 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;
-}
-
-
void set_defaults(struct mkfs_opts *mop)
{
- 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");
- if (get_os_version() == 24)
- mop->mo_ldd.ldd_mount_type = LDD_MT_EXT3;
- else
- mop->mo_ldd.ldd_mount_type = LDD_MT_LDISKFS;
-
- mop->mo_ldd.ldd_svindex = INDEX_UNASSIGNED;
- mop->mo_stripe_count = 1;
+ 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
+ mop->mo_ldd.ldd_svindex = INDEX_UNASSIGNED;
+ mop->mo_mgs_failnodes = 0;
+ mop->mo_stripe_count = 1;
+ mop->mo_pool_vdevs = NULL;
}
static inline void badopt(const char *opt, char *type)
usage(stderr);
}
-static int add_param(char *buf, char *key, char *val)
-{
- 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;
-}
-
/* 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;
+ 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) && ((s2 = strsep(&s1, ",: \0")))) {
- nid = libcfs_str2nid(s2);
+ 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) {
- if (*s2 == '/')
- /* end of nids */
- break;
- fprintf(stderr, "%s: Can't parse NID '%s'\n",
- progname, s2);
+ fprintf(stderr, "%s: Can't parse NID '%s'\n",
+ progname, s1);
free(converted);
return NULL;
}
-
- if (strncmp(libcfs_nid2str(nid), "127.0.0.1",
+ 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, s2, libcfs_nid2str(nid));
+ "non-loopback address.\n",
+ progname, s1, libcfs_nid2str(nid));
free(converted);
return NULL;
}
- c += snprintf(c, left, "%s,", libcfs_nid2str(nid));
+ c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep);
left = converted + MAXNIDSTR - c;
+ s1 = s2 + 1;
}
- *(c - 1) = '\0';
return converted;
}
{"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'},
{"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'},
+ {"quota", 0, 0, 'Q'},
{0, 0, 0, 0}
};
- char *optstring = "b:c:C:d:ef:Ghi:k:L:m:MnNo:Op:Pqru:vw";
+ 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)) !=
+ while ((opt = getopt_long(argc, argv, optstring, long_opt, &longidx)) !=
EOF) {
switch (opt) {
case 'b': {
}
i++;
}
+ if (i == LDD_MT_LAST) {
+ fprintf(stderr, "%s: invalid backend filesystem"
+ " type %s\n", progname, optarg);
+ return 1;
+ }
break;
}
case 'c':
printf("Configdev not implemented\n");
return 1;
case 'd':
- mop->mo_device_sz = atol(optarg);
+ 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': {
- char *nids = convert_hostnames(optarg);
- if (!nids)
+ 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;
- rc = add_param(mop->mo_ldd.ldd_params, PARAM_FAILNODE,
- nids);
- /* Combo needs to add MDT failnodes as MGS failnodes
- as well */
- if (!rc && IS_MGS(&mop->mo_ldd)) {
- rc = add_param(mop->mo_ldd.ldd_params,
- PARAM_MGSNODE, nids);
}
+
+ nids = convert_hostnames(optarg);
+ if (!nids)
+ return 1;
+ rc = add_param(mop->mo_ldd.ldd_params, PARAM_FAILNODE,
+ nids);
free(nids);
- if (rc)
+ 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;
+ }
+ mop->mo_flags |= MO_FAILOVER;
break;
}
case 'G':
usage(stdout);
return 1;
case 'i':
- if (!(mop->mo_ldd.ldd_flags &
+ if (!(mop->mo_ldd.ldd_flags &
(LDD_F_UPGRADE14 | LDD_F_VIRGIN |
LDD_F_WRITECONF))) {
fprintf(stderr, "%s: cannot change the index of"
}
break;
case 'k':
- strncpy(mop->mo_mkfsopts, optarg,
- sizeof(mop->mo_mkfsopts) - 1);
+ 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 &
+ (!(mop->mo_ldd.ldd_flags &
(LDD_F_UPGRADE14 | LDD_F_VIRGIN |
LDD_F_WRITECONF)))) {
fprintf(stderr, "%s: cannot change the name of"
"filesystem name\n", progname, *tmp);
return 1;
}
- strncpy(mop->mo_ldd.ldd_fsname, optarg,
- sizeof(mop->mo_ldd.ldd_fsname) - 1);
+ strscpy(mop->mo_ldd.ldd_fsname, optarg,
+ sizeof(mop->mo_ldd.ldd_fsname));
break;
}
case 'm': {
char *nids = convert_hostnames(optarg);
- if (!nids)
+ if (!nids)
return 1;
- rc = add_param(mop->mo_ldd.ldd_params, PARAM_MGSNODE,
- nids);
+ rc = add_param(mop->mo_ldd.ldd_params, PARAM_MGSNODE,
+ nids);
free(nids);
- if (rc)
+ if (rc)
return rc;
mop->mo_mgs_failnodes++;
break;
break;
case 'p':
rc = add_param(mop->mo_ldd.ldd_params, NULL, optarg);
- if (rc)
+ if (rc)
return rc;
/* Must update the mgs logs */
mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
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':
- strncpy(mop->mo_ldd.ldd_userdata, optarg,
+ strscpy(mop->mo_ldd.ldd_userdata, optarg,
sizeof(mop->mo_ldd.ldd_userdata));
- mop->mo_ldd.ldd_userdata[
- sizeof(mop->mo_ldd.ldd_userdata) - 1] = 0;
break;
case 'v':
verbose++;
case 'w':
mop->mo_ldd.ldd_flags |= LDD_F_WRITECONF;
break;
+ case 'U':
+ upgrade_to_18 = 1;
+ break;
+ case 'Q':
+ mop->mo_flags |= MO_QUOTA;
+ break;
default:
if (opt != '?') {
fatal();
fprintf(stderr, "Unknown option '%c'\n", opt);
}
- usage(stderr);
- return 1;
+ return EINVAL;
}
}//while
- if (optind >= argc) {
- fatal();
- fprintf(stderr, "Bad arguments\n");
- usage(stderr);
- return 1;
- }
+
+ 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;
+ } else {
+ /* 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;
}
char *mountopts = NULL;
char always_mountopts[512] = "";
char default_mountopts[512] = "";
- int ret = 0;
+ unsigned mount_type;
+ int ret = 0;
if ((progname = strrchr(argv[0], '/')) != NULL)
progname++;
set_defaults(&mop);
/* device is last arg */
- strcpy(mop.mo_device, argv[argc - 1]);
+ 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)
- goto out;
- if (ret == 0)
- mop.mo_flags |= MO_IS_LOOP;
+ ret = osd_init();
+ if (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);
+ ret = osd_is_lustre(mop.mo_device, &mount_type);
if (ret == 0) {
fatal();
fprintf(stderr, "Device %s has not been formatted with "
ret = ENODEV;
goto out;
}
+ mop.mo_ldd.ldd_mount_type = mount_type;
- ret = read_local_files(&mop);
+ ret = osd_read_ldd(mop.mo_device, &mop.mo_ldd);
if (ret) {
fatal();
fprintf(stderr, "Failed to read previous Lustre data from %s "
"(%d)\n", mop.mo_device, ret);
goto out;
}
+ mop.mo_ldd.ldd_flags &= ~(LDD_F_WRITECONF | LDD_F_VIRGIN);
+
if (strstr(mop.mo_ldd.ldd_params, PARAM_MGSNODE))
mop.mo_mgs_failnodes++;
- if (verbose > 0)
+ if (verbose > 0)
print_ldd("Read previous values", &(mop.mo_ldd));
#endif
ret = parse_opts(argc, argv, &mop, &mountopts);
- if (ret)
+ 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");
goto out;
}
+ /* Stand alone MGS doesn't need a index */
+ if (!IS_MDT(ldd) && IS_MGS(ldd))
+ mop.mo_ldd.ldd_flags &= ~LDD_F_NEED_INDEX;
+
if ((mop.mo_ldd.ldd_flags & (LDD_F_NEED_INDEX | LDD_F_UPGRADE14)) ==
(LDD_F_NEED_INDEX | LDD_F_UPGRADE14)) {
fatal();
goto out;
}
+ if (mop.mo_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)) {
+ mop.mo_ldd.ldd_flags &= ~LDD_F_NEED_INDEX;
+ mop.mo_ldd.ldd_svindex = 0;
+ }
+
+#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))
+ if (IS_MDT(ldd))
fprintf(stderr, "Must specify --mgs or --mgsnode=\n");
- else
+ else
fprintf(stderr, "Must specify --mgsnode=\n");
ret = EINVAL;
goto out;
}
-
- /* These are the permanent mount options (always included) */
- switch (ldd->ldd_mount_type) {
- case LDD_MT_EXT3:
- case LDD_MT_LDISKFS: {
- sprintf(always_mountopts, "errors=remount-ro");
- if (IS_MDT(ldd) || IS_MGS(ldd))
- strcat(always_mountopts,
- ",iopen_nopriv,user_xattr");
- if ((get_os_version() == 24) && IS_OST(ldd))
- strcat(always_mountopts, ",asyncdel");
- /* NB: Files created while extents are enabled cannot be read
- if mounted with a kernel that doesn't include the CFS
- patches! */
- if (IS_OST(ldd) &&
- ldd->ldd_mount_type == LDD_MT_LDISKFS) {
- strcat(default_mountopts, ",extents,mballoc");
- }
- break;
- }
- case LDD_MT_SMFS: {
- mop.mo_flags |= MO_IS_LOOP;
- sprintf(always_mountopts, "type=ext3,dev=%s",
- mop.mo_device);
- break;
- }
- default: {
- fatal();
- fprintf(stderr, "unknown fs type %d '%s'\n",
- ldd->ldd_mount_type,
- MT_STR(ldd));
- ret = EINVAL;
- goto out;
- }
- }
+#endif
+ if ((IS_MDT(ldd) || IS_OST(ldd)) && mop.mo_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,
+ default_mountopts, sizeof(default_mountopts),
+ always_mountopts, sizeof(always_mountopts));
+ if (ret) {
+ fatal();
+ fprintf(stderr, "unable to prepare backend (%d)\n", ret);
+ goto out;
+ }
if (mountopts) {
- /* If user specifies mount opts, don't use defaults,
- but always use always_mountopts */
- sprintf(ldd->ldd_mount_opts, "%s,%s",
- always_mountopts, 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 {
#ifdef TUNEFS
- if (ldd->ldd_mount_opts[0] == 0)
+ if (ldd->ldd_mount_opts[0] == 0)
/* use the defaults unless old opts exist */
#endif
{
- sprintf(ldd->ldd_mount_opts, "%s%s",
+ sprintf(ldd->ldd_mount_opts, "%s%s",
always_mountopts, default_mountopts);
+ trim_mountfsoptions(ldd->ldd_mount_opts);
}
}
goto out;
}
- if (check_mtab_entry(mop.mo_device))
- return(EEXIST);
+ if (check_mtab_entry(mop.mo_device, mop.mo_device, NULL, NULL))
+ return(EEXIST);
/* Create the loopback file */
if (mop.mo_flags & MO_IS_LOOP) {
ret = access(mop.mo_device, F_OK);
- if (ret)
+ if (ret)
ret = errno;
#ifndef TUNEFS /* mkfs.lustre */
/* Reformat the loopback file */
- if (ret || (mop.mo_flags & MO_FORCEFORMAT))
+ if (ret || (mop.mo_flags & MO_FORCEFORMAT)) {
ret = loop_format(&mop);
+ if (ret)
+ goto out;
+ }
#endif
- if (ret == 0)
+ if (ret == 0)
ret = loop_setup(&mop);
if (ret) {
fatal();
#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);
+ ret = osd_is_lustre(mop.mo_device, &mount_type);
if (ret) {
fatal();
- fprintf(stderr, "Device %s was previously formatted "
+ fprintf(stderr, "Device %s was previously formatted "
"for lustre. Use --reformat to reformat it, "
"or tunefs.lustre to modify.\n",
mop.mo_device);
}
/* Format the backing filesystem */
- ret = make_lustre_backfs(&mop);
+ ret = osd_make_lustre(&mop);
if (ret != 0) {
fatal();
fprintf(stderr, "mkfs failed %d\n", ret);
goto out;
}
+#else
+ /* update svname with '=' to refresh config */
+ if (mop.mo_ldd.ldd_flags & LDD_F_WRITECONF) {
+ struct mount_opts opts;
+ opts.mo_ldd = mop.mo_ldd;
+ opts.mo_source = mop.mo_device;
+ (void) osd_label_lustre(&opts);
+ }
+
+ /* Enable quota accounting */
+ if (mop.mo_flags & MO_QUOTA) {
+ ret = osd_enable_quota(&mop);
+ goto out;
+ }
+
#endif
/* Write our config files */
- ret = write_local_files(&mop);
+ ret = osd_write_ldd(&mop);
if (ret != 0) {
fatal();
fprintf(stderr, "failed to write local files\n");
}
out:
- loop_cleanup(&mop);
+ loop_cleanup(&mop);
+ osd_fini();
- /* Fix any crazy return values from system() */
+ /* Fix any crazy return values from system() */
if (ret && ((ret & 255) == 0))
return (1);
if (ret)