-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
* GPL HEADER START
*
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* 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
*/
/*
- * Copyright 2008 Sun Microsystems, Inc. All rights reserved
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
*
* lustre/utils/mkfs_lustre.c
*
#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>
-#include <linux/fs.h> /* for BLKGETSIZE64 */
-#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_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;
-};
-
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--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"
- "\t\t--iam-dir: make use of IAM directory format on backfs, incompatible with ext3.\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;
-}
-
-/* Determine if a device is a block device (as opposed to a file) */
-int is_block(char* devname)
-{
- struct stat st;
- int ret = 0;
-
- 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);
-}
-
-__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;
-}
-
-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);
- if (ret < 0) {
- ret = errno;
- fprintf(stderr, "%s: Unable to create backing store: %d\n",
- progname, ret);
- } else {
- close(ret);
- }
-
- ret = truncate(mop->mo_device, mop->mo_device_sz * 1024);
- if (ret != 0) {
- ret = errno;
- fprintf(stderr, "%s: Unable to truncate backing store: %d\n",
- progname, ret);
- }
-
- return ret;
-}
+ #define FSLIST_ZFS ""
+#endif /* HAVE_ZFS_OSD */
-/* 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.lustre.org/public/tools/e2fsprogs/\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;
-
- 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;
- }
+#define FSLIST FSLIST_LDISKFS FSLIST_ZFS
- 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)
-{
- FILE *fp;
- char cmd[PATH_MAX];
- char imgname[] = "/tmp/test-img-XXXXXX";
- int fd = -1;
- int ret = 0;
-
- snprintf(cmd, sizeof(cmd), "%s -c -R \"supported_features %s\" 2>&1",
- DEBUGFS, feature);
-
- /* 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(cmd, 1, sizeof(cmd), fp);
- if (ret > 0) {
- if (strstr(cmd, feature) && !(strstr(cmd, "Unknown")))
- return 0;
- }
-
- if ((fd = mkstemp(imgname)) < 0)
- return -1;
-
- snprintf(cmd, sizeof(cmd), "%s -F -O %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);
- if (fd >= 0)
- remove(imgname);
-
- return ret;
-}
-
-static void disp_old_kernel_msg(char *feature)
-{
- fprintf(stderr, "WARNING: ldiskfs filesystem does not support \"%s\" "
- "feature.\n\n", feature);
-}
-
-static void enable_default_backfs_features(struct mkfs_opts *mop)
-{
- struct utsname uts;
- int maj_high, maj_low, min;
- int ret;
-
- strscat(mop->mo_mkfsopts, " -O dir_index,extents", sizeof(mop->mo_mkfsopts));
-
- /* Upstream e2fsprogs called our uninit_groups feature uninit_bg,
- * check for both of them when testing e2fsprogs features. */
- if (is_e2fsprogs_feature_supp("uninit_groups") == 0)
- strscat(mop->mo_mkfsopts, ",uninit_groups",
- sizeof(mop->mo_mkfsopts));
- else if (is_e2fsprogs_feature_supp("uninit_bg") == 0)
- strscat(mop->mo_mkfsopts, ",uninit_bg",
- sizeof(mop->mo_mkfsopts));
- else
- disp_old_e2fsprogs_msg("uninit_bg", 1);
-
- ret = uname(&uts);
- if (ret)
- return;
-
- sscanf(uts.release, "%d.%d.%d", &maj_high, &maj_low, &min);
- printf("%d %d %d\n", maj_high, maj_low, min);
-
- /* Multiple mount protection is enabled only if failover node is
- * specified and if kernel version is higher than 2.6.9 */
- if (failover) {
- if (KERNEL_VERSION(maj_high, maj_low, min) >=
- KERNEL_VERSION(2,6,9)) {
- if (is_e2fsprogs_feature_supp("mmp") == 0)
- strscat(mop->mo_mkfsopts, ",mmp",
- sizeof(mop->mo_mkfsopts));
- else
- disp_old_e2fsprogs_msg("mmp", 1);
- } else {
- disp_old_kernel_msg("mmp");
- }
- }
-}
-/* Build fs according to type */
-int make_lustre_backfs(struct mkfs_opts *mop)
+void usage(FILE *out)
{
- char mkfs_cmd[PATH_MAX];
- char buf[64];
- char *dev;
- int ret = 0;
- int block_count = 0;
-
- 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 ((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)) {
- __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;
- /* 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));
- }
- }
-
- /* 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 > 100000000)))
- bytes_per_inode = 16384; /* > 100 Gb device */
-
-
- if (bytes_per_inode > 0) {
- sprintf(buf, " -i %ld", bytes_per_inode);
- 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) {
- 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);
- strscat(mop->mo_mkfsopts, buf,
- sizeof(mop->mo_mkfsopts));
- }
- }
-
- if (verbose < 2) {
- strscat(mop->mo_mkfsopts, " -q",
- sizeof(mop->mo_mkfsopts));
- }
-
- if (strstr(mop->mo_mkfsopts, "-O") == NULL)
- enable_default_backfs_features(mop);
-
- /* 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 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);
- strscat(mop->mo_mkfsopts, buf,
- sizeof(mop->mo_mkfsopts));
- }
- snprintf(mkfs_cmd, sizeof(mkfs_cmd), "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! */
- 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, " %d", 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)\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_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)
-{
- 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;
-}
-
-static int __l_mkdir(char * filepnm, int mode , struct mkfs_opts *mop)
+void print_ldd(char *str, 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;
+ 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");
}
-/* 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, 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 = __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;
+ mop->mo_ldd.ldd_svindex = INDEX_UNASSIGNED;
+ mop->mo_mgs_failnodes = 0;
+ mop->mo_stripe_count = 1;
+ mop->mo_pool_vdevs = NULL;
}
-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;
-}
-
-
-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");
- 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;
+ 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 + 8] = "";
+ 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'},
- {"verbose", 0, 0, 'v'},
- {"writeconf", 0, 0, 'w'},
- {"upgrade_to_18", 0, 0, 'U'},
- {0, 0, 0, 0}
- };
- char *optstring = "b:c:C:d:ef:Ghi:k:L:m:MnNo:Op:Pqru:vw";
- int opt;
- int rc, longidx;
-
- 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': {
- char *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;
- 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 '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)
- 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, ",iopen_nopriv,user_xattr",
- sizeof(always_mountopts));
- if ((get_os_version() == 24) && IS_OST(ldd))
- strscat(always_mountopts, ",asyncdel",
- sizeof(always_mountopts));
- /* NB: Files created while extents are enabled cannot be read
- if mounted with a kernel that doesn't include the Lustre ldiskfs
- patches! */
- 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;
- }
- 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;
- }
- }
-
- 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);
-#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;
}