* in the LICENSE file that accompanied this code).
*
* You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see [sun.com URL with a
- * copy of GPLv2].
+ * 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
* 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.
*/
/*
/* 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 <string.h>
#include <getopt.h>
#include <limits.h>
+#include <ctype.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 */
+/* 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>
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 */
+ char mo_mkfsopts[512]; /* options to the backing-store mkfs */
__u64 mo_device_sz; /* in KB */
int mo_stripe_count;
int mo_flags;
int verbose = 1;
static int print_only = 0;
static int failover = 0;
+static int upgrade_to_18 = 0;
void usage(FILE *out)
{
"\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--index=#N : target index (i.e. ost index within lov)\n"
+ "\t\t--comment=<user comment>: arbitrary string (%d bytes)\n"
"\t\t--mountfsoptions=<opts> : permanent mount options\n"
+ "\t\t--network=<net>[,<...>] : restrict OST/MDT to network(s)\n"
#ifndef TUNEFS
"\t\t--backfstype=<fstype> : backing fs type (ext3, ldiskfs)\n"
"\t\t--device-size=#N(KB) : device size for loop devices\n"
"\t\t--mkfsoptions=<opts> : format options\n"
"\t\t--reformat: overwrite an existing disk\n"
- "\t\t--stripe-count-hint=#N : used for optimizing MDT inode size\n"
+ "\t\t--stripe-count-hint=#N : for optimizing MDT inode size\n"
+ "\t\t--iam-dir: use IAM directory format, not ext3 compatible\n"
#else
"\t\t--erase-params : erase all old parameter settings\n"
"\t\t--nomgs: turn off MGS service on this MDT\n"
char release[4] = "";
fd = open("/proc/sys/kernel/osrelease", O_RDONLY);
- if (fd < 0)
+ if (fd < 0) {
fprintf(stderr, "%s: Warning: Can't resolve kernel "
"version, assuming 2.6\n", progname);
- else {
- read(fd, release, 4);
+ } 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)
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));
static int msg_displayed;
if (msg_displayed) {
- fprintf(stderr, "WARNING: e2fsprogs does not support %s "
- "feature.\n\n", feature);
+ fprintf(stderr, "WARNING: %s does not support %s "
+ "feature.\n\n", E2FSPROGS, feature);
return;
}
msg_displayed++;
- fprintf(stderr, "WARNING: The e2fsprogs package currently installed on "
- "your system does not support \"%s\" feature.\nPlease install "
- "the latest version of e2fsprogs from http://downloads.lustre.org"
- "/public/tools/e2fsprogs/\nto enable this feature.\n", feature);
-
+ 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");
+#endif
if (make_backfs)
- fprintf(stderr, "Feature will not be enabled until e2fsprogs "
- "is updated and 'tune2fs -O %s %%{device}' "
- "is run.\n\n", feature);
+ 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 */
/* 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);
+ "%s -c -R 'stat %s' '%s' 2>&1 | egrep '(Inode|unsupported)'",
+ DEBUGFS, file_name, dev_name);
fp = popen(debugfs_cmd, "r");
if (!fp) {
* mke2fs to check for its support. */
static int is_e2fsprogs_feature_supp(const char *feature)
{
+ static char supp_features[4096] = "";
FILE *fp;
char cmd[PATH_MAX];
char imgname[] = "/tmp/test-img-XXXXXX";
int fd = -1;
- int ret = 0;
+ int ret = 1;
- snprintf(cmd, sizeof(cmd),
- "debugfs -c -R \"supported_features %s\" 2>&1", feature);
+ if (supp_features[0] == '\0') {
+ snprintf(cmd, sizeof(cmd), "%s -c -R supported_features 2>&1",
+ DEBUGFS);
- /* Using popen() instead of run_command() since debugfs does not return
- * proper error code if command is not supported */
- fp = popen(cmd, "r");
- if (!fp) {
- fprintf(stderr, "%s: %s\n", progname, strerror(errno));
- return 0;
- }
- ret = fread(cmd, 1, sizeof(cmd), fp);
- if (ret > 0) {
- if (strstr(cmd, feature) && !(strstr(cmd, "Unknown")))
+ /* Using popen() instead of run_command() since debugfs does
+ * not return proper error code if command is not supported */
+ fp = popen(cmd, "r");
+ if (!fp) {
+ fprintf(stderr, "%s: %s\n", progname, strerror(errno));
return 0;
+ }
+ ret = fread(supp_features, 1, sizeof(supp_features), fp);
+ fclose(fp);
}
+ if (ret > 0 && strstr(supp_features,
+ strncmp(feature, "-O ", 3) ? feature : feature+3))
+ return 0;
if ((fd = mkstemp(imgname)) < 0)
return -1;
+ else
+ close(fd);
- snprintf(cmd, sizeof(cmd), "mke2fs -F -O %s %s 100 >/dev/null 2>&1",
- feature, imgname);
+ snprintf(cmd, sizeof(cmd), "%s -F %s %s 100 >/dev/null 2>&1",
+ MKE2FS, feature, imgname);
/* run_command() displays the output of mke2fs when it fails for
* some feature, so use system() directly */
ret = system(cmd);
- if (fd >= 0)
- remove(imgname);
+ unlink(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)
+static void enable_default_ext4_features(struct mkfs_opts *mop)
{
- struct utsname uts;
- int maj_high, maj_low, min;
- int ret;
-
- strscat(mop->mo_mkfsopts, " -O dir_index", 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",
+ if (IS_OST(&mop->mo_ldd))
+ strscat(mop->mo_mkfsopts, " -O extents,uninit_bg",
sizeof(mop->mo_mkfsopts));
- else if (is_e2fsprogs_feature_supp("uninit_bg") == 0)
- strscat(mop->mo_mkfsopts, ",uninit_bg",
+ else if (IS_MDT(&mop->mo_ldd))
+ strscat(mop->mo_mkfsopts, " -O dirdata,uninit_bg",
sizeof(mop->mo_mkfsopts));
else
- disp_old_e2fsprogs_msg("uninit_bg", 1);
+ strscat(mop->mo_mkfsopts, " -O uninit_bg",
+ sizeof(mop->mo_mkfsopts));
- ret = uname(&uts);
- if (ret)
+ /* Multiple mount protection enabled only if failover node specified */
+ if (failover) {
+ if (is_e2fsprogs_feature_supp("-O mmp") == 0)
+ strscat(mop->mo_mkfsopts, ",mmp",
+ sizeof(mop->mo_mkfsopts));
+ else
+ disp_old_e2fsprogs_msg("mmp", 1);
+ }
+
+ /* Allow more than 65000 subdirectories */
+ if (is_e2fsprogs_feature_supp("-O dir_nlink") == 0)
+ strscat(mop->mo_mkfsopts,",dir_nlink",sizeof(mop->mo_mkfsopts));
+
+#ifdef HAVE_EXT4_LDISKFS
+ /* The following options are only valid for ext4-based ldiskfs.
+ * If --backfstype=ext3 is specified, do not enable them. */
+ if (mop->mo_ldd.ldd_mount_type == LDD_MT_EXT3)
return;
- sscanf(uts.release, "%d.%d.%d", &maj_high, &maj_low, &min);
- printf("%d %d %d\n", maj_high, maj_low, min);
+ /* Allow files larger than 2TB. Also needs LU-16, but not harmful. */
+ if (is_e2fsprogs_feature_supp("-O huge_file") == 0)
+ strscat(mop->mo_mkfsopts,",huge_file",sizeof(mop->mo_mkfsopts));
- /* 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");
+ /* Cluster inode/block bitmaps and inode table for more efficient IO.
+ * Align the flex groups on a 1MB boundary for better performance.
+ * This -O feature needs to go last, since it adds an extra option. */
+ if (is_e2fsprogs_feature_supp("-O flex_bg") == 0) {
+ char tmp_buf[64];
+
+ strscat(mop->mo_mkfsopts, ",flex_bg", sizeof(mop->mo_mkfsopts));
+
+ if (IS_OST(&mop->mo_ldd)) {
+ snprintf(tmp_buf, sizeof(tmp_buf), " -G %u",
+ (1 << 20) / L_BLOCK_SIZE);
+ strscat(mop->mo_mkfsopts, tmp_buf,
+ sizeof(mop->mo_mkfsopts));
}
}
+#endif
}
+
/* Build fs according to type */
int make_lustre_backfs(struct mkfs_opts *mop)
{
+ __u64 device_sz = mop->mo_device_sz, block_count = 0;
char mkfs_cmd[PATH_MAX];
char buf[64];
char *dev;
int ret = 0;
- int block_count = 0;
+
+ if (!(mop->mo_flags & MO_IS_LOOP)) {
+ mop->mo_device_sz = get_device_size(mop->mo_device);
+
+ if (mop->mo_device_sz == 0)
+ return ENODEV;
+
+ /* Compare to real size */
+ if (device_sz == 0 || device_sz > mop->mo_device_sz)
+ device_sz = mop->mo_device_sz;
+ else
+ mop->mo_device_sz = device_sz;
+ }
if (mop->mo_device_sz != 0) {
if (mop->mo_device_sz < 8096){
return EINVAL;
}
block_count = mop->mo_device_sz / (L_BLOCK_SIZE >> 10);
+ /* If the LUN size is just over 2^32 blocks, limit the
+ * filesystem size to 2^32-1 blocks to avoid problems with
+ * ldiskfs/mkfs not handling this size. Bug 22906 */
+ if (block_count > 0xffffffffULL && block_count < 0x100002000ULL)
+ block_count = 0xffffffffULL;
}
if ((mop->mo_ldd.ldd_mount_type == LDD_MT_EXT3) ||
(mop->mo_ldd.ldd_mount_type == LDD_MT_LDISKFS) ||
(mop->mo_ldd.ldd_mount_type == LDD_MT_LDISKFS2)) {
- __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;
- }
+ long inode_size = 0;
/* Journal size in MB */
if (strstr(mop->mo_mkfsopts, "-J") == NULL) {
if (journal_sz > 1024L)
journal_sz = 1024L;
/* man mkfs.ext3 */
- max_sz = (256000 * L_BLOCK_SIZE) >> 20; /* 1GB */
+ max_sz = (102400 * L_BLOCK_SIZE) >> 20; /* 400MB */
if (journal_sz > max_sz)
journal_sz = max_sz;
if (journal_sz) {
}
}
- /* 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) */
+ /* see also "-i" below for EA blocks */
else if (mop->mo_stripe_count > 32)
inode_size = 2048;
else if (mop->mo_stripe_count > 10)
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 */
+ /* We store MDS FID and OST objid in EA on OST
+ * we need to make inode bigger as well. */
inode_size = 256;
}
}
}
+ /* Bytes_per_inode: disk size / num inodes */
+ if (strstr(mop->mo_mkfsopts, "-i") == NULL &&
+ strstr(mop->mo_mkfsopts, "-N") == NULL) {
+ long bytes_per_inode = 0;
+
+ /* Allocate more inodes on MDT devices. There is
+ * no data stored on the MDT, and very little extra
+ * metadata beyond the inode. It could go down as
+ * low as 1024 bytes, but this is conservative.
+ * Account for external EA blocks for wide striping. */
+ if (IS_MDT(&mop->mo_ldd)) {
+ bytes_per_inode = inode_size + 1536;
+
+ if (mop->mo_stripe_count > 72) {
+ int extra = mop->mo_stripe_count * 24;
+ extra = ((extra - 1) | 4095) + 1;
+ bytes_per_inode += extra;
+ }
+ }
+
+ /* Allocate fewer inodes on large OST devices. Most
+ * filesystems can be much more aggressive than even
+ * this, but it is impossible to know in advance. */
+ if (IS_OST(&mop->mo_ldd)) {
+ /* OST > 8TB assume average file size 1MB */
+ if (device_sz >= (8ULL << 30))
+ bytes_per_inode = 1024 * 1024;
+ /* OST > 1TB assume average file size 256kB */
+ else if (device_sz >= (1ULL << 30))
+ bytes_per_inode = 256 * 1024;
+ /* OST > 100GB assume average file size 64kB,
+ * plus a bit so that inodes will fit into a
+ * 256x flex_bg without overflowing */
+ else if (device_sz >= (10ULL << 20))
+ bytes_per_inode = 69905;
+ }
+
+
+ if (bytes_per_inode > 0) {
+ sprintf(buf, " -i %ld", bytes_per_inode);
+ strscat(mop->mo_mkfsopts, buf,
+ sizeof(mop->mo_mkfsopts));
+ }
+ }
+
if (verbose < 2) {
strscat(mop->mo_mkfsopts, " -q",
sizeof(mop->mo_mkfsopts));
}
if (strstr(mop->mo_mkfsopts, "-O") == NULL)
- enable_default_backfs_features(mop);
+ enable_default_ext4_features(mop);
+
+ /* In order to align the filesystem metadata on 1MB boundaries,
+ * give a resize value that will reserve a power-of-two group
+ * descriptor blocks, but leave one block for the superblock.
+ * Only useful for filesystems with < 2^32 blocks due to resize
+ * limitations. */
+ if (IS_OST(&mop->mo_ldd) && mop->mo_device_sz > 100 * 1024 &&
+ mop->mo_device_sz / L_BLOCK_SIZE <= 0xffffffff) {
+ unsigned group_blocks = L_BLOCK_SIZE * 8;
+ unsigned desc_per_block = L_BLOCK_SIZE / 32;
+ unsigned resize_blks;
+
+ resize_blks = (1ULL<<32) - desc_per_block*group_blocks;
+ snprintf(buf, sizeof(buf)," -E resize=%u,",resize_blks);
+ } else {
+ strncpy(buf, " -E ", sizeof(buf));
+ }
+
+ /* Avoid zeroing out the full journal - speeds up mkfs */
+ if (is_e2fsprogs_feature_supp("-E lazy_journal_init") == 0)
+ strscat(buf, "lazy_journal_init,", sizeof(buf));
+
+ if (strlen(buf) > strlen(" -E "))
+ strscat(mop->mo_mkfsopts, buf,sizeof(mop->mo_mkfsopts));
/* 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),
- "mkfs.ext2 -j -b %d -L %s ", L_BLOCK_SIZE,
+ "%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,
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("\t4k blocks "LPU64"\n", block_count);
vprint("\toptions %s\n", mop->mo_mkfsopts);
/* mkfs_cmd's trailing space is important! */
strscat(mkfs_cmd, " ", sizeof(mkfs_cmd));
strscat(mkfs_cmd, dev, sizeof(mkfs_cmd));
if (block_count != 0) {
- sprintf(buf, " %d", block_count);
+ sprintf(buf, " "LPU64, block_count);
strscat(mkfs_cmd, buf, sizeof(mkfs_cmd));
}
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",
+ 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_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);
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)
+{
+ int ret;
+
+ ret = mkdir(filepnm, mode);
+ if (ret && ret != -EEXIST)
+ return ret;
+
+ /* IAM mode supports ext3 directories of HTREE type only. So add dummy
+ * entries to new directory to create htree type of container for
+ * this directory. */
+ if (mop->mo_ldd.ldd_flags & LDD_F_IAM_DIR)
+ return add_dummy_files(filepnm);
+ return 0;
+}
+
/* Write the server config files */
int write_local_files(struct mkfs_opts *mop)
{
char *dev;
FILE *filep;
int ret = 0;
+ size_t num;
/* Mount this device temporarily in order to write these files */
if (!mkdtemp(mntpt)) {
if (mop->mo_flags & MO_IS_LOOP)
dev = mop->mo_loopdev;
- ret = mount(dev, mntpt, MT_STR(&mop->mo_ldd), 0, NULL);
+ ret = mount(dev, mntpt, MT_STR(&mop->mo_ldd), 0,
+ mop->mo_ldd.ldd_mount_opts);
if (ret) {
fprintf(stderr, "%s: Unable to mount %s: %s\n",
progname, dev, strerror(errno));
/* Set up initial directories */
sprintf(filepnm, "%s/%s", mntpt, MOUNT_CONFIGS_DIR);
- ret = mkdir(filepnm, 0777);
+ 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));
ret = 0;
}
- sprintf(filepnm, "%s/%s", mntpt, "ROOT");
- ret = mkdir(filepnm, 0777);
- if ((ret != 0) && (errno != EEXIST)) {
- fprintf(stderr, "%s: Can't make ROOT 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);
progname, filepnm, strerror(errno));
goto out_umnt;
}
- fwrite(&mop->mo_ldd, sizeof(mop->mo_ldd), 1, filep);
+ 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 */
#endif
/* end COMPAT_146 */
-
out_umnt:
umount(mntpt);
out_rmdir:
filesystem */
/* Construct debugfs command line. */
- snprintf(cmd, cmdsz, "debugfs -c -R 'dump /%s %s/mountdata' %s",
- MOUNT_DATA_FILE, tmpdir, dev);
+ snprintf(cmd, cmdsz, "%s -c -R 'dump /%s %s/mountdata' '%s'",
+ DEBUGFS, MOUNT_DATA_FILE, tmpdir, dev);
ret = run_command(cmd, cmdsz);
if (ret)
sprintf(filepnm, "%s/mountdata", tmpdir);
filep = fopen(filepnm, "r");
if (filep) {
+ size_t num_read;
vprint("Reading %s\n", MOUNT_DATA_FILE);
- fread(&mop->mo_ldd, sizeof(mop->mo_ldd), 1, filep);
+ 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 */
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);
+ snprintf(cmd, cmdsz, "%s -c -R 'dump /%s %s' %s",
+ DEBUGFS, LAST_RCVD, filepnm, dev);
ret = run_command(cmd, cmdsz);
if (ret) {
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);
+ snprintf(cmd, cmdsz, "%s -c -R 'ls -l /' %s",
+ DEBUGFS, dev);
run_command(cmd, cmdsz);
goto out_rmdir;
} 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);
+ 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);
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_mount_type = LDD_MT_LDISKFS;
mop->mo_ldd.ldd_svindex = INDEX_UNASSIGNED;
mop->mo_stripe_count = 1;
sep = *s2;
*s2 = '\0';
nid = libcfs_str2nid(s1);
-
+
if (nid == LNET_NID_ANY) {
- fprintf(stderr, "%s: Can't parse NID '%s'\n", progname, s1);
+ fprintf(stderr, "%s: Can't parse NID '%s'\n",
+ progname, s1);
free(converted);
return NULL;
}
free(converted);
return NULL;
}
-
+
c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep);
left = converted + MAXNIDSTR - c;
s1 = s2 + 1;
char **mountopts)
{
static struct option long_opt[] = {
+ {"iam-dir", 0, 0, 'a'},
{"backfstype", 1, 0, 'b'},
{"stripe-count-hint", 1, 0, 'c'},
{"comment", 1, 0, 'u'},
{"reformat", 0, 0, 'r'},
{"verbose", 0, 0, 'v'},
{"writeconf", 0, 0, 'w'},
+ {"upgrade_to_18", 0, 0, 'U'},
+ {"network", 1, 0, 't'},
{0, 0, 0, 0}
};
char *optstring = "b:c:C:d:ef:Ghi:k:L:m:MnNo:Op:Pqru:vw";
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) {
case 'r':
mop->mo_flags |= MO_FORCEFORMAT;
break;
+ case 't':
+ if (!IS_MDT(&mop->mo_ldd) && !IS_OST(&mop->mo_ldd)) {
+ badopt(long_opt[longidx].name, "MDT,OST");
+ return 1;
+ }
+
+ if (!optarg)
+ return 1;
+
+ rc = add_param(mop->mo_ldd.ldd_params,
+ PARAM_NETWORK, optarg);
+ if (rc != 0)
+ return rc;
+ /* Must update the mgs logs */
+ mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
+ break;
case 'u':
strscpy(mop->mo_ldd.ldd_userdata, optarg,
sizeof(mop->mo_ldd.ldd_userdata));
case 'w':
mop->mo_ldd.ldd_flags |= LDD_F_WRITECONF;
break;
+ case 'U':
+ upgrade_to_18 = 1;
+ break;
default:
if (opt != '?') {
fatal();
return EINVAL;
}
- return 0;
-}
-
-#include <lustre/libiam.h>
+ /* single argument: <device> */
+ if (argc == 2)
+ ++print_only;
-#define LDISKFS_IOC_GETVERSION _IOR('f', 3, long)
-
-#ifndef TUNEFS /* mkfs.lustre */
-static int mkfs_iam_insert(int key_need_convert, char *keybuf,
- int rec_need_convert, char *recbuf, char *filename)
-{
- int fd;
- int ret;
- struct iam_uapi_info ua;
-
- fd = iam_open(filename, &ua);
- if (fd < 0) {
- fprintf(stderr, "failed to iam_open %s\n", filename);
- return 1;
- }
-
- ret = iam_insert(fd, &ua,
- key_need_convert, keybuf,
- rec_need_convert, recbuf);
- iam_close(fd);
- if (ret) {
- fprintf(stderr, "failed to iam_insert %s\n", filename);
- return 1;
- } else {
- return 0;
- }
+ return 0;
}
-static int touch_file(char *filename)
+/* Search for opt in mntlist, returning true if found.
+ */
+static int in_mntlist(char *opt, char *mntlist)
{
- int fd;
+ char *ml, *mlp, *item, *ctx = NULL;
- if (filename == NULL) {
- return 1;
+ if (!(ml = strdup(mntlist))) {
+ fprintf(stderr, "%s: out of memory\n", progname);
+ exit(1);
}
-
- fd = open(filename, O_CREAT | O_TRUNC, 0600);
- if (fd < 0) {
- return 1;
- } else {
- close(fd);
- return 0;
+ mlp = ml;
+ while ((item = strtok_r(mlp, ",", &ctx))) {
+ if (!strcmp(opt, item))
+ break;
+ mlp = NULL;
}
+ free(ml);
+ return (item != NULL);
}
-static int get_generation(char *filename, unsigned long *result)
+/* 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)
{
- int fd;
- int ret;
-
- if (filename == NULL) {
- return 1;
- }
-
- fd = open(filename, O_RDONLY);
- if (fd < 0) {
- fprintf(stderr, "%s: failed to open %s\n",
- __FUNCTION__, filename);
- return 1;
+ 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;
}
-
- ret = ioctl(fd, LDISKFS_IOC_GETVERSION, result);
- close(fd);
-
- return ((ret < 0) ? ret : 0);
+ free(ml);
+ return errors;
}
-static int mkfs_mdt(struct mkfs_opts *mop)
+/* Trim embedded white space, leading and trailing commas from string s.
+ */
+static void trim_mountfsoptions(char *s)
{
- char mntpt[] = "/tmp/mntXXXXXX";
- char fstype[] = "ldiskfs";
- char filepnm[128];
- char recbuf[64];
- char *source;
- int ret;
- unsigned long generation;
- struct stat st;
+ char *p;
- source = mop->mo_device;
- if (mop->mo_flags & MO_IS_LOOP) {
- source = mop->mo_loopdev;
- }
-
- if ((source == NULL) || (*source == 0)) {
- return 1;
- }
-
- if (!mkdtemp(mntpt)) {
- fprintf(stderr, "%s: failed to mkdtemp %s\n",
- __FUNCTION__, mntpt);
- return errno;
- }
-
- ret = mount(source, mntpt, fstype, 0, NULL);
- if (ret) {
- goto out_rmdir;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "seq_ctl");
- ret = touch_file(filepnm);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "seq_srv");
- ret = touch_file(filepnm);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "last_received");
- ret = touch_file(filepnm);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "lov_objid");
- ret = touch_file(filepnm);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "root");
- ret = iam_creat(filepnm, FMT_LVAR, L_BLOCK_SIZE, 4, 17, 4);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "fld");
- ret = iam_creat(filepnm, FMT_LFIX, L_BLOCK_SIZE, 8, 8, 4);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "orphans");
- ret = iam_creat(filepnm, FMT_LFIX, L_BLOCK_SIZE, 20, 8, 4);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "oi.16");
- ret = iam_creat(filepnm, FMT_LFIX, L_BLOCK_SIZE, 16, 8, 4);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "oi.5");
- ret = iam_creat(filepnm, FMT_LFIX, L_BLOCK_SIZE, 5, 8, 4);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, CAPA_KEYS);
- ret = touch_file(filepnm);
- if (ret) {
- goto out_umount;
- }
-
- umount(mntpt);
- ret = mount(source, mntpt, fstype, 0, NULL);
- if (ret) {
- goto out_rmdir;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "root");
- ret = iam_polymorph(filepnm, 040755);
- if (ret) {
- perror("IAM_IOC_POLYMORPH");
- goto out_umount;
- }
-
- umount(mntpt);
- ret = mount(source, mntpt, fstype, 0, NULL);
- if (ret) {
- goto out_rmdir;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "fld");
- ret = mkfs_iam_insert(1, "0000000000000002", 1, "0000000000000000", filepnm);
- if (ret) {
- goto out_umount;
- }
-
- ret = mkfs_iam_insert(1, "0000000000000001", 1, "0000000000000000", filepnm);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(filepnm, sizeof(filepnm) - 1, "%s/%s", mntpt, "root");
- ret = stat(filepnm, &st);
- if (ret) {
- goto out_umount;
- }
-
- ret = get_generation(filepnm, &generation);
- if (ret) {
- goto out_umount;
- }
-
- snprintf(recbuf, sizeof(recbuf) - 1, "110000000000000001%8.8x%8.8x",
- (unsigned int)st.st_ino, (unsigned int)generation);
- ret = mkfs_iam_insert(0, ".", 1, recbuf, filepnm);
- if (ret) {
- goto out_umount;
+ for (p = s; *p; ) {
+ if (isspace(*p)) {
+ memmove(p, p + 1, strlen(p + 1) + 1);
+ continue;
+ }
+ p++;
}
- ret = mkfs_iam_insert(0, "..", 1, recbuf, filepnm);
- if (ret) {
- goto out_umount;
- }
+ while (s[0] == ',')
+ memmove(&s[0], &s[1], strlen(&s[1]) + 1);
-out_umount:
- umount(mntpt);
-out_rmdir:
- rmdir(mntpt);
- return ret;
+ p = s + strlen(s) - 1;
+ while (p >= s && *p == ',')
+ *p-- = '\0';
}
-#endif
int main(int argc, char *const argv[])
{
switch (ldd->ldd_mount_type) {
case LDD_MT_EXT3:
case LDD_MT_LDISKFS:
- case LDD_MT_LDISKFS2: {
- sprintf(always_mountopts, "errors=remount-ro");
+ 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",
+ strscat(always_mountopts, ",user_xattr",
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! */
+ /* NB: Files created while extents are enabled can only be read
+ * if mounted using the ext4 or ldiskfs filesystem type. */
if (IS_OST(ldd) &&
(ldd->ldd_mount_type == LDD_MT_LDISKFS ||
ldd->ldd_mount_type == LDD_MT_LDISKFS2)) {
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: {
+ default:
fatal();
fprintf(stderr, "unknown fs type %d '%s'\n",
ldd->ldd_mount_type,
ret = EINVAL;
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)
{
sprintf(ldd->ldd_mount_opts, "%s%s",
always_mountopts, default_mountopts);
+ trim_mountfsoptions(ldd->ldd_mount_opts);
}
}
goto out;
}
-#ifndef TUNEFS /* mkfs.lustre */
- if (IS_MDT(ldd)) {
- ret = mkfs_mdt(&mop);
- if (ret != 0) {
- fprintf(stderr, "failed to mkfs_mdt\n");
- goto out;
- }
- }
-#endif
-
out:
loop_cleanup(&mop);
git://git.whamcloud.com / fs / lustre-release.git / blobdiff