X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Futils%2Fmount_lustre.c;h=7ed149c8c1b8f76a6eaa32db60c1445a9ab0a60d;hp=fe1920ff03fe0784182b7192239b5c4d87303efe;hb=5315db3f1066619d6effe4f778d2df3ad1ba738f;hpb=54fe9796ec837698a27420c8a92d9493c733b6a9 diff --git a/lustre/utils/mount_lustre.c b/lustre/utils/mount_lustre.c index fe1920f..7ed149c 100644 --- a/lustre/utils/mount_lustre.c +++ b/lustre/utils/mount_lustre.c @@ -15,11 +15,7 @@ * * You should have received a copy of the GNU General Public License * version 2 along with this program; If not, see - * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ @@ -27,7 +23,7 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, 2012, Whamcloud, Inc. + * Copyright (c) 2011, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -42,91 +38,124 @@ #ifndef _GNU_SOURCE #define _GNU_SOURCE #endif -#include "mount_utils.h" +#include +#include +#include #include #include #include #include #include #include -#include "obdctl.h" -#include +#include +#include +#include #include #include +#include +#include + +#include "obdctl.h" +#include "mount_utils.h" + +#ifdef HAVE_LIBMOUNT +# define WITH_LIBMOUNT "(libmount)" +#else +# define WITH_LIBMOUNT "" +#endif -#define MAXOPT 4096 #define MAX_RETRIES 99 -int verbose = 0; -char *progname = NULL; +int verbose; +int version; +char *progname; void usage(FILE *out) { - fprintf(out, "%s v"LUSTRE_VERSION_STRING"\n", progname); - fprintf(out, "\nThis mount helper should only be invoked via the " - "mount (8) command,\ne.g. mount -t lustre dev dir\n\n"); - fprintf(out, "usage: %s [-fhnv] [-o ] \n", - progname); - fprintf(out, - "\t: the disk device, or for a client:\n" - "\t\t[:...]:/-client\n" - "\t: name of the Lustre filesystem (e.g. lustre1)\n" - "\t: filesystem mountpoint (e.g. /mnt/lustre)\n" - "\t-f|--fake: fake mount (updates /etc/mtab)\n" - "\t-o force|--force: force mount even if already in /etc/mtab\n" - "\t-h|--help: print this usage message\n" - "\t-n|--nomtab: do not update /etc/mtab after mount\n" - "\t-v|--verbose: print verbose config settings\n" - "\t: one or more comma separated of:\n" - "\t\t(no)flock,(no)user_xattr,(no)acl\n" - "\t\tabort_recov: abort server recovery handling\n" - "\t\tnosvc: only start MGC/MGS obds\n" - "\t\tnomgs: only start target obds, using existing MGS\n" - "\t\texclude=[:] : colon-separated list of " - "inactive OSTs (e.g. lustre-OST0001)\n" - "\t\tretry=: number of times mount is retried by client\n" - "\t\tmd_stripe_cache_size=: set the raid stripe cache " - "size for the underlying raid if present\n" - ); - exit((out != stdout) ? EINVAL : 0); + fprintf(out, "\nThis mount helper should only be invoked via the " + "mount (8) command,\ne.g. mount -t lustre dev dir\n\n"); + fprintf(out, "usage: %s [-fhnvV] [-o ] \n", + progname); + fprintf(out, "usage: %s [-fhnvV] [-o ] \n", + progname); + fprintf(out, + "\t: the local disk device when mounting a server\n" + "\t: the server MGS and filesystem for a client:\n" + "\t\t[:...]:/[/]\n" + "\t\t\t: MGS LNet Node Identifier (e.g. mgs01@o2ib)\n" + "\t\t\t: Lustre filesystem name (e.g. lustre1)\n" + "\t\t\t: subdirectory of the filesystem to mount\n" + "\t: filesystem mountpoint (e.g. /mnt/lustre)\n" + "\t-f|--fake: fake mount (only update /etc/mtab)\n" + "\t-o force|--force: force mount even if already in /etc/mtab\n" + "\t-h|--help: print this usage message\n" + "\t-n|--nomtab: do not update /etc/mtab after mount\n" + "\t-v|--verbose: print verbose config settings\n" + "\t-V|--version: output build version of the utility and exit\n" + "\tdefault options are marked below with '*'\n" + "\t\t(no)flock: disable* or enable POSIX flock support\n" + "\t\t(no)user_xattr: disable or enable* user xattr namespace\n" + "\t: one or more comma separated server options:\n" + "\t\t(no)acl: disable or enable* POSIX ACL support completely\n" + "\t\tabort_recov: abort server recovery handling\n" + "\t\tnosvc: only start MGC/MGS without starting MDS/OSS\n" + "\t\tnomgs: only start target MDS/OSS, using existing MGS\n" + "\t\tnoscrub: do NOT auto start OI scrub unless requested\n" + "\t\tskip_lfsck: do NOT auto resume paused/crashed LFSCK\n" + "\t\tmd_stripe_cache_size=: set MD RAID device stripe cache size\n" + "\t: one or more comma separated client options:\n" + "\t\texclude=[:]: list of inactive OSTs (e.g. lustre-OST0001)\n" + "\t\tlocalflock: enable POSIX flock only on local client\n" + "\t\tretry=: number of times mount is retried by client\n" +#ifdef HAVE_GSS + "\t\tskpath=: path of keys to load into kernel keyring\n" +#endif + "\t\t(no)user_fid2path: disable* or enable user $MOUNT/.lustre/fid access\n" + "\t\t(no)checksum: disable or enable* data checksums\n" + "\t\t(no)lruresize: disable or enable* LDLM dynamic LRU size\n" + "\t\t(no)lazystatfs: disable or enable* statfs to work if OST is unavailable\n" + "\t\t32bitapi: return only 32-bit inode numbers to userspace\n" + "\t\t(no)verbose: disable or enable* messages at filesystem (un,re)mount\n" + ); + exit((out != stdout) ? EINVAL : 0); } /* Get rid of symbolic hostnames for tcp, since kernel can't do lookups */ #define MAXNIDSTR 1024 static char *convert_hostnames(char *s1) { - char *converted, *s2 = 0, *c; - char sep; - int left = MAXNIDSTR; - lnet_nid_t nid; - - converted = malloc(left); - if (converted == NULL) { - fprintf(stderr, "out of memory: needed %d bytes\n", - MAXNIDSTR); - return NULL; - } - c = converted; - while ((left > 0) && (*s1 != '/')) { - s2 = strpbrk(s1, ",:"); - if (!s2) - goto out_free; - sep = *s2; - *s2 = '\0'; - nid = libcfs_str2nid(s1); - *s2 = sep; /* back to original string */ - if (nid == LNET_NID_ANY) - goto out_free; - c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep); - left = converted + MAXNIDSTR - c; - s1 = s2 + 1; - } - snprintf(c, left, "%s", s1); - return converted; + char *converted, *s2 = 0, *c; + char sep; + int left = MAXNIDSTR; + lnet_nid_t nid; + + converted = malloc(left); + if (converted == NULL) { + fprintf(stderr, "out of memory: needed %d bytes\n", + MAXNIDSTR); + return NULL; + } + c = converted; + while ((left > 0) && (*s1 != '/')) { + s2 = strpbrk(s1, ",:"); + if (!s2) + goto out_free; + sep = *s2; + *s2 = '\0'; + nid = libcfs_str2nid(s1); + *s2 = sep; /* back to original string */ + if (nid == LNET_NID_ANY) + goto out_free; + c += scnprintf(c, left, "%s%c", libcfs_nid2str(nid), sep); + left = converted + MAXNIDSTR - c; + s1 = s2 + 1; + } + snprintf(c, left, "%s", s1); + return converted; out_free: - fprintf(stderr, "%s: Can't parse NID '%s'\n", progname, s1); - free(converted); - return NULL; + fprintf(stderr, "%s: Can't parse NID '%s'\n", progname, s1); + free(converted); + return NULL; } /***************************************************************************** @@ -143,42 +172,45 @@ struct opt_map { }; static const struct opt_map opt_map[] = { - /*"optname", inv,ms_mask */ - /* These flags are parsed by mount, not lustre */ - { "defaults", 0, 0 }, /* default options */ - { "remount", 0, MS_REMOUNT}, /* remount with different options */ - { "rw", 1, MS_RDONLY }, /* read-write */ - { "ro", 0, MS_RDONLY }, /* read-only */ - { "exec", 1, MS_NOEXEC }, /* permit execution of binaries */ - { "noexec", 0, MS_NOEXEC }, /* don't execute binaries */ - { "suid", 1, MS_NOSUID }, /* honor suid executables */ - { "nosuid", 0, MS_NOSUID }, /* don't honor suid executables */ - { "dev", 1, MS_NODEV }, /* interpret device files */ - { "nodev", 0, MS_NODEV }, /* don't interpret devices */ - { "sync", 0, MS_SYNCHRONOUS}, /* synchronous I/O */ - { "async", 1, MS_SYNCHRONOUS}, /* asynchronous I/O */ - { "atime", 1, MS_NOATIME }, /* set file access time on read */ - { "noatime", 0, MS_NOATIME }, /* do not set file access time on read */ +/* these flags are parsed by mount, not Lustre */ +{ .opt = "async", .mask = MS_SYNCHRONOUS, .inv = 1 }, /* asynchronous I/O */ +{ .opt = "atime", .mask = MS_NOATIME, .inv = 1 }, /* set access time */ +{ .opt = "auto" }, /* allow auto mount */ +{ .opt = "defaults" }, /* default options */ +{ .opt = "dev", .mask = MS_NODEV, .inv = 1 }, /* interpret devs */ +{ .opt = "exec", .mask = MS_NOEXEC, .inv = 1 }, /* allow execution */ +{ .opt = "loop" }, +{ .opt = "noatime", .mask = MS_NOATIME }, /* do not set atime */ +{ .opt = "noauto" }, /* mount explicitly */ +{ .opt = "nodev", .mask = MS_NODEV }, /* no interpret devs */ +{ .opt = "noowner", .inv = 1 }, /* no special privs */ +{ .opt = "nosuid", .mask = MS_NOSUID }, /* do not honor suid */ +{ .opt = "nouser", .inv = 1 }, /* users cannot mount */ +{ .opt = "nousers", .inv = 1 }, /* users cannot mount */ +{ .opt = "_netdev" }, /* network only */ +{ .opt = "noexec", .mask = MS_NOEXEC }, /* no execute */ +{ .opt = "remount", .mask = MS_REMOUNT }, /* remount */ +{ .opt = "ro", .mask = MS_RDONLY }, /* read-only */ +{ .opt = "rw", .mask = MS_RDONLY, .inv = 1 }, /* read-write */ +{ .opt = "suid", .mask = MS_NOSUID, .inv = 1 }, /* honor suid */ +{ .opt = "sync", .mask = MS_SYNCHRONOUS }, /* synchronous I/O */ #ifdef MS_NODIRATIME - { "diratime", 1, MS_NODIRATIME }, /* set file access time on read */ - { "nodiratime",0,MS_NODIRATIME }, /* do not set file access time on read */ +{ .opt = "diratime", /* set access time */ + .mask = MS_NODIRATIME, .inv = 1 }, /* on read */ +{ .opt = "nodiratime", /* do not set access */ + .mask = MS_NODIRATIME }, /* time on read */ #endif #ifdef MS_RELATIME - { "relatime", 0, MS_RELATIME }, /* set file access time on read */ - { "norelatime",1,MS_RELATIME }, /* do not set file access time on read */ +{ .opt = "norelatime", /* do not set rel */ + .mask = MS_RELATIME, .inv = 1 }, /* access time */ +{ .opt = "relatime", /* set relative */ + .mask = MS_RELATIME }, /* access time */ #endif #ifdef MS_STRICTATIME - { "strictatime",0,MS_STRICTATIME }, /* update access time strictly */ +{ .opt = "strictatime", + .mask = MS_STRICTATIME }, /* strict access time */ #endif - { "auto", 0, 0 }, /* Can be mounted using -a */ - { "noauto", 0, 0 }, /* Can only be mounted explicitly */ - { "nousers", 1, 0 }, /* Forbid ordinary user to mount */ - { "nouser", 1, 0 }, /* Forbid ordinary user to mount */ - { "noowner", 1, 0 }, /* Device owner has no special privs */ - { "_netdev", 0, 0 }, /* Device accessible only via network */ - { "loop", 0, 0 }, - { NULL, 0, 0 } -}; +{ .opt = NULL } }; /****************************************************************************/ /* 1 = don't pass on to lustre @@ -203,73 +235,147 @@ static int parse_one_option(const char *check, int *flagp) return 0; } -static void append_option(char *options, const char *one) +static size_t merge_strings(char *dst, const char *src, size_t size) +{ + size_t dsize = strlen(dst); + size_t len = strlen(src); + size_t ret = dsize + len; + + dst += dsize; + size -= dsize; + if (len >= size) + len = size - 1; + memcpy(dst, src, len); + dst[len] = '\0'; + return ret; +} + +static int append_option(char *options, size_t options_len, + const char *param, const char *value) { - if (*options) - strcat(options, ","); - strcat(options, one); + int rc; + + if (options[0] != '\0') { + rc = merge_strings(options, ",", options_len); + if (rc >= options_len) + goto out_err; + } + + rc = merge_strings(options, param, options_len); + if (rc >= options_len) + goto out_err; + + if (value != NULL) { + rc = merge_strings(options, value, options_len); + if (rc >= options_len) + goto out_err; + } + return 0; +out_err: + fprintf(stderr, "error: mount options %s%s too long\n", param, value); + return E2BIG; } /* Replace options with subset of Lustre-specific options, and fill in mount flags */ -int parse_options(struct mount_opts *mop, char *orig_options, int *flagp) +int parse_options(struct mount_opts *mop, char *orig_options, + int *flagp, size_t options_len) { - char *options, *opt, *nextopt, *arg, *val; - - options = calloc(strlen(orig_options) + 1, 1); - *flagp = 0; - nextopt = orig_options; - while ((opt = strsep(&nextopt, ","))) { - if (!*opt) - /* empty option */ - continue; - - /* Handle retries in a slightly different - * manner */ - arg = opt; - val = strchr(opt, '='); - /* please note that some ldiskfs mount options are also in the form - * of param=value. We should pay attention not to remove those - * mount options, see bug 22097. */ - if (val && strncmp(arg, "md_stripe_cache_size", 20) == 0) { + char *options, *opt, *nextopt, *arg, *val; + int rc = 0; + + options = calloc(strlen(orig_options) + 1, 1); + if (options == NULL) + return ENOMEM; + + *flagp = 0; + nextopt = orig_options; + while ((opt = strsep(&nextopt, ","))) { + if (!*opt) + /* empty option */ + continue; + + /* Handle retries in a slightly different manner */ + arg = opt; + val = strchr(opt, '='); + /* please note that some ldiskfs mount options are also in + * the form of param=value. We should pay attention not to + * remove those mount options, see bug 22097. */ + if (val && strncmp(arg, "max_sectors_kb", 14) == 0) { + mop->mo_max_sectors_kb = atoi(val + 1); + } else if (val && + strncmp(arg, "md_stripe_cache_size", 20) == 0) { mop->mo_md_stripe_cache_size = atoi(val + 1); - } else if (val && strncmp(arg, "retry", 5) == 0) { + } else if (val && strncmp(arg, "retry", 5) == 0) { mop->mo_retry = atoi(val + 1); if (mop->mo_retry > MAX_RETRIES) mop->mo_retry = MAX_RETRIES; else if (mop->mo_retry < 0) mop->mo_retry = 0; - } else if (val && strncmp(arg, "mgssec", 6) == 0) { - append_option(options, opt); + } else if (val && strncmp(arg, "mgssec", 6) == 0) { + rc = append_option(options, options_len, opt, NULL); + if (rc != 0) + goto out_options; } else if (strncmp(arg, "nosvc", 5) == 0) { mop->mo_nosvc = 1; - append_option(options, opt); - } else if (strcmp(opt, "force") == 0) { - //XXX special check for 'force' option + rc = append_option(options, options_len, opt, NULL); + if (rc != 0) + goto out_options; + } else if (strcmp(opt, "force") == 0) { + /* XXX special check for 'force' option */ ++mop->mo_force; printf("force: %d\n", mop->mo_force); - } else if (parse_one_option(opt, flagp) == 0) { - /* pass this on as an option */ - append_option(options, opt); - } - } +#ifdef HAVE_GSS + } else if (val && strncmp(opt, "skpath=", 7) == 0) { + if (strlen(val) + 1 >= sizeof(mop->mo_skpath)) { + fprintf(stderr, + "%s: shared key path too long\n", + progname); + free(options); + return EINVAL; + } + strncpy(mop->mo_skpath, val + 1, + sizeof(mop->mo_skpath) - 1); +#endif + } else if (parse_one_option(opt, flagp) == 0) { + /* pass this on as an option */ + rc = append_option(options, options_len, opt, NULL); + if (rc != 0) + goto out_options; + } + } #ifdef MS_STRICTATIME - /* set strictatime to default if NOATIME or RELATIME - not given explicit */ - if (!(*flagp & (MS_NOATIME | MS_RELATIME))) - *flagp |= MS_STRICTATIME; +#if LUSTRE_VERSION_CODE > OBD_OCD_VERSION(3, 2, 53, 0) + /* + * LU-1783 + * In the future when upstream fixes land in all supported kernels + * we should stop forcing MS_STRICTATIME in lustre mounts. + * We override the kernel level default of MS_RELATIME for now + * due to a kernel vfs level bug in atime updates that fails + * to reset timestamps from the future. + */ +#warn "remove MS_STRICTATIME override if kernel updates atime from the future" #endif - strcpy(orig_options, options); - free(options); - return 0; + /* set strictatime to default if NOATIME or RELATIME + not given explicit */ + if (!(*flagp & (MS_NOATIME | MS_RELATIME))) + *flagp |= MS_STRICTATIME; +#endif + strcpy(orig_options, options); + +out_options: + free(options); + return rc; } +#ifdef HAVE_SERVER_SUPPORT /* Add mgsnids from ldd params */ static int add_mgsnids(struct mount_opts *mop, char *options, - const char *params) + const char *params, size_t options_len) { char *ptr = (char *)params; char tmp, *sep; + int rc = 0; while ((ptr = strstr(ptr, PARAM_MGSNODE)) != NULL) { sep = strchr(ptr, ' '); @@ -277,7 +383,9 @@ static int add_mgsnids(struct mount_opts *mop, char *options, tmp = *sep; *sep = '\0'; } - append_option(options, ptr); + rc = append_option(options, options_len, ptr, NULL); + if (rc != 0) + goto out; mop->mo_have_mgsnid++; if (sep) { *sep = tmp; @@ -287,14 +395,77 @@ static int add_mgsnids(struct mount_opts *mop, char *options, } } - return 0; +out: + return rc; } -static int parse_ldd(char *source, struct mount_opts *mop, char *options) +static int clear_update_ondisk(char *source, struct lustre_disk_data *ldd) +{ + char wanted_mountopts[512] = ""; + struct mkfs_opts mkop; + int ret; + int ret2; + + memset(&mkop, 0, sizeof(mkop)); + mkop.mo_ldd = *ldd; + mkop.mo_ldd.ldd_flags &= ~LDD_F_UPDATE; + mkop.mo_flags = MO_NOHOSTID_CHECK; /* Ignore missing hostid */ + if (strlen(source) > sizeof(mkop.mo_device)-1) { + fatal(); + fprintf(stderr, "Device name too long: %s\n", source); + return -E2BIG; + } + strncpy(mkop.mo_device, source, sizeof(mkop.mo_device)); + + ret = osd_prepare_lustre(&mkop, + wanted_mountopts, sizeof(wanted_mountopts)); + if (ret) { + fatal(); + fprintf(stderr, "Can't prepare device %s: %s\n", + source, strerror(ret)); + return ret; + } + + /* Create the loopback file */ + if (mkop.mo_flags & MO_IS_LOOP) { + ret = access(mkop.mo_device, F_OK); + if (ret) { + ret = errno; + fatal(); + fprintf(stderr, "Can't access device %s: %s\n", + source, strerror(ret)); + return ret; + } + + ret = loop_setup(&mkop); + if (ret) { + fatal(); + fprintf(stderr, "Loop device setup for %s failed: %s\n", + mkop.mo_device, strerror(ret)); + return ret; + } + } + ret = osd_write_ldd(&mkop); + if (ret != 0) { + fatal(); + fprintf(stderr, "failed to write local files: %s\n", + strerror(ret)); + } + + ret2 = loop_cleanup(&mkop); + if (ret == 0) + ret = ret2; + + return ret; +} + +static int parse_ldd(char *source, struct mount_opts *mop, + char *options, size_t options_len) { struct lustre_disk_data *ldd = &mop->mo_ldd; char *cur, *start; - int rc; + char *temp_options; + int rc = 0; rc = osd_is_lustre(source, &ldd->ldd_mount_type); if (rc == 0) { @@ -307,50 +478,71 @@ static int parse_ldd(char *source, struct mount_opts *mop, char *options) rc = osd_read_ldd(source, ldd); if (rc) { fprintf(stderr, "%s: %s failed to read permanent mount" - " data: %s\n", progname, source, strerror(rc)); + " data: %s\n", progname, source, + rc >= 0 ? strerror(rc) : ""); return rc; } - if (ldd->ldd_flags & LDD_F_NEED_INDEX) { + if ((IS_MDT(ldd) || IS_OST(ldd)) && + (ldd->ldd_flags & LDD_F_NEED_INDEX)) { fprintf(stderr, "%s: %s has no index assigned " "(probably formatted with old mkfs)\n", progname, source); return EINVAL; } - if (ldd->ldd_flags & LDD_F_UPGRADE14) { - fprintf(stderr, "%s: we cannot upgrade %s from this (very old) " - "Lustre version\n", progname, source); - return EINVAL; - } + if (ldd->ldd_flags & LDD_F_UPDATE) + clear_update_ondisk(source, ldd); /* Since we never rewrite ldd, ignore temp flags */ - ldd->ldd_flags &= ~(LDD_F_VIRGIN | LDD_F_UPDATE | LDD_F_WRITECONF); + ldd->ldd_flags &= ~(LDD_F_VIRGIN | LDD_F_WRITECONF); + + /* This is to make sure default options go first */ + temp_options = strdup(options); + if (temp_options == NULL) { + fprintf(stderr, "%s: can't allocate memory for temp_options\n", + progname); + return ENOMEM; + } + strncpy(options, ldd->ldd_mount_opts, options_len); + rc = append_option(options, options_len, temp_options, NULL); + free(temp_options); + if (rc != 0) + return rc; /* svname of the form lustre:OST1234 means never registered */ rc = strlen(ldd->ldd_svname); - if (ldd->ldd_svname[rc - 8] == ':') { - ldd->ldd_svname[rc - 8] = '-'; - ldd->ldd_flags |= LDD_F_VIRGIN; - } else if (ldd->ldd_svname[rc - 8] == '=') { - ldd->ldd_svname[rc - 8] = '-'; - ldd->ldd_flags |= LDD_F_WRITECONF; + if (strcmp(ldd->ldd_svname, "MGS") != 0) { + if (rc < 8) { + fprintf(stderr, "%s: invalid name '%s'\n", + progname, ldd->ldd_svname); + return EINVAL; + } else if (ldd->ldd_svname[rc - 8] == ':') { + ldd->ldd_svname[rc - 8] = '-'; + ldd->ldd_flags |= LDD_F_VIRGIN; + } else if (ldd->ldd_svname[rc - 8] == '=') { + ldd->ldd_svname[rc - 8] = '-'; + ldd->ldd_flags |= LDD_F_WRITECONF; + } } - /* backend osd type */ - append_option(options, "osd="); - strcat(options, mt_type(ldd->ldd_mount_type)); - - append_option(options, ldd->ldd_mount_opts); + rc = append_option(options, options_len, "osd=", + mt_type(ldd->ldd_mount_type)); + if (rc != 0) + return rc; if (!mop->mo_have_mgsnid) { /* Only use disk data if mount -o mgsnode=nid wasn't * specified */ if (ldd->ldd_flags & LDD_F_SV_TYPE_MGS) { - append_option(options, "mgs"); + rc = append_option(options, options_len, "mgs", NULL); + if (rc != 0) + return rc; mop->mo_have_mgsnid++; } else { - add_mgsnids(mop, options, ldd->ldd_params); + if (add_mgsnids(mop, options, ldd->ldd_params, + options_len)) + return E2BIG; } } /* Better have an mgsnid by now */ @@ -360,12 +552,26 @@ static int parse_ldd(char *source, struct mount_opts *mop, char *options) return EINVAL; } - if (ldd->ldd_flags & LDD_F_VIRGIN) - append_option(options, "virgin"); - if (ldd->ldd_flags & LDD_F_WRITECONF) - append_option(options, "writeconf"); - if (ldd->ldd_flags & LDD_F_NO_PRIMNODE) - append_option(options, "noprimnode"); + if (ldd->ldd_flags & LDD_F_VIRGIN) { + rc = append_option(options, options_len, "virgin", NULL); + if (rc != 0) + return rc; + } + if (ldd->ldd_flags & LDD_F_UPDATE) { + rc = append_option(options, options_len, "update", NULL); + if (rc != 0) + return rc; + } + if (ldd->ldd_flags & LDD_F_WRITECONF) { + rc = append_option(options, options_len, "writeconf", NULL); + if (rc != 0) + return rc; + } + if (ldd->ldd_flags & LDD_F_NO_PRIMNODE) { + rc = append_option(options, options_len, "noprimnode", NULL); + if (rc != 0) + return rc; + } /* prefix every lustre parameter with param= so that in-kernel * mount can recognize them properly and send to MGS at registration */ @@ -380,16 +586,17 @@ static int parse_ldd(char *source, struct mount_opts *mop, char *options) *start = '\0'; start++; } - append_option(options, "param="); - strcat(options, cur); + rc = append_option(options, options_len, "param=", cur); + if (rc != 0) + return rc; } /* svname must be last option */ - append_option(options, "svname="); - strcat(options, ldd->ldd_svname); + rc = append_option(options, options_len, "svname=", ldd->ldd_svname); - return 0; + return rc; } +#endif /* HAVE_SERVER_SUPPORT */ static void set_defaults(struct mount_opts *mop) { @@ -404,19 +611,20 @@ static void set_defaults(struct mount_opts *mop) mop->mo_md_stripe_cache_size = 16384; mop->mo_orig_options = ""; mop->mo_nosvc = 0; + mop->mo_max_sectors_kb = -1; } static int parse_opts(int argc, char *const argv[], struct mount_opts *mop) { - static struct option long_opt[] = { - {"fake", 0, 0, 'f'}, - {"force", 0, 0, 1}, - {"help", 0, 0, 'h'}, - {"nomtab", 0, 0, 'n'}, - {"options", 1, 0, 'o'}, - {"verbose", 0, 0, 'v'}, - {0, 0, 0, 0} - }; + static struct option long_opts[] = { + { .val = 1, .name = "force", .has_arg = no_argument }, + { .val = 'f', .name = "fake", .has_arg = no_argument }, + { .val = 'h', .name = "help", .has_arg = no_argument }, + { .val = 'n', .name = "nomtab", .has_arg = no_argument }, + { .val = 'o', .name = "options", .has_arg = required_argument }, + { .val = 'v', .name = "verbose", .has_arg = no_argument }, + { .val = 'V', .name = "version", .has_arg = no_argument }, + { .name = NULL } }; char real_path[PATH_MAX] = {'\0'}; FILE *f; char path[256], name[256]; @@ -424,8 +632,8 @@ static int parse_opts(int argc, char *const argv[], struct mount_opts *mop) char *ptr; int opt, rc; - while ((opt = getopt_long(argc, argv, "fhno:v", - long_opt, NULL)) != EOF){ + while ((opt = getopt_long(argc, argv, "fhno:vV", + long_opts, NULL)) != EOF){ switch (opt) { case 1: ++mop->mo_force; @@ -448,6 +656,11 @@ static int parse_opts(int argc, char *const argv[], struct mount_opts *mop) case 'v': ++verbose; break; + case 'V': + ++version; + fprintf(stdout, "%s %s %s\n", progname, + LUSTRE_VERSION_STRING, WITH_LIBMOUNT); + return 0; default: fprintf(stderr, "%s: unknown option '%c'\n", progname, opt); @@ -505,44 +718,91 @@ static int parse_opts(int argc, char *const argv[], struct mount_opts *mop) return 0; } +#ifdef HAVE_SERVER_SUPPORT +/* change label from : to + * - to indicate the device has + * been registered. only if the label is + * supposed to be changed and target service + * is supposed to start */ +static void label_lustre(struct mount_opts *mop) +{ + if (mop->mo_nosvc) + return; + + if (mop->mo_ldd.ldd_flags & (LDD_F_VIRGIN | LDD_F_WRITECONF)) { + (void)osd_label_lustre(mop); + } else { + struct lustre_disk_data ldd; + int rc; + + /* device label could be changed after journal recovery, + * it should also be relabeled for mount has succeeded. */ + memset(&ldd, 0, sizeof(ldd)); + ldd.ldd_mount_type = mop->mo_ldd.ldd_mount_type; + rc = osd_read_ldd(mop->mo_source, &ldd); + if (rc == 0) { + rc = strlen(ldd.ldd_svname); + if (rc >= 8 && ldd.ldd_svname[rc - 8] != '-') + (void)osd_label_lustre(mop); + } + } +} +#endif /* HAVE_SERVER_SUPPORT */ + int main(int argc, char *const argv[]) { struct mount_opts mop; char *options; - int i, rc, flags; + int i, flags; + int rc; + bool client; + size_t maxopt_len; + size_t g_pagesize; progname = strrchr(argv[0], '/'); progname = progname ? progname + 1 : argv[0]; set_defaults(&mop); - rc = osd_init(); - if (rc) + g_pagesize = sysconf(_SC_PAGESIZE); + if (g_pagesize == -1) { + rc = errno; + printf("error: %d failed to get page size.\n", rc); return rc; + } + maxopt_len = MIN(g_pagesize, 64 * 1024); rc = parse_opts(argc, argv, &mop); - if (rc) + if (rc || version) return rc; - if (verbose) { - for (i = 0; i < argc; i++) - printf("arg[%d] = %s\n", i, argv[i]); + if (verbose) { + for (i = 0; i < argc; i++) + printf("arg[%d] = %s\n", i, argv[i]); printf("source = %s (%s), target = %s\n", mop.mo_usource, mop.mo_source, mop.mo_target); printf("options = %s\n", mop.mo_orig_options); - } + } + + options = malloc(maxopt_len); + if (options == NULL) { + fprintf(stderr, "can't allocate memory for options\n"); + rc = ENOMEM; + goto out_mo_source; + } + + if (strlen(mop.mo_orig_options) >= maxopt_len) { + fprintf(stderr, "error: mount options too long\n"); + rc = E2BIG; + goto out_options; + } - options = malloc(MAXOPT); - if (options == NULL) { - fprintf(stderr, "can't allocate memory for options\n"); - return -1; - } strcpy(options, mop.mo_orig_options); - rc = parse_options(&mop, options, &flags); + rc = parse_options(&mop, options, &flags, maxopt_len); if (rc) { fprintf(stderr, "%s: can't parse options: %s\n", progname, options); - return(EINVAL); + goto out_options; } if (!mop.mo_force) { @@ -552,57 +812,83 @@ int main(int argc, char *const argv[]) fprintf(stderr, "%s: according to %s %s is " "already mounted on %s\n", progname, MOUNTED, mop.mo_usource, mop.mo_target); - return(EEXIST); + rc = EEXIST; + goto out_options; } if (!rc && (flags & MS_REMOUNT)) { fprintf(stderr, "%s: according to %s %s is " "not already mounted on %s\n", progname, MOUNTED, mop.mo_usource, mop.mo_target); - return(ENOENT); + rc = ENOENT; + goto out_options; } } if (flags & MS_REMOUNT) mop.mo_nomtab++; rc = access(mop.mo_target, F_OK); - if (rc) { - rc = errno; + if (rc) { + rc = errno; fprintf(stderr, "%s: %s inaccessible: %s\n", progname, mop.mo_target, strerror(errno)); - return rc; - } + goto out_options; + } - if (!strstr(mop.mo_usource, ":/")) { - rc = parse_ldd(mop.mo_source, &mop, options); + client = (strstr(mop.mo_usource, ":/") != NULL); + if (!client) { +#ifdef HAVE_SERVER_SUPPORT + rc = osd_init(); if (rc) - return rc; + goto out_options; + + rc = parse_ldd(mop.mo_source, &mop, options, maxopt_len); + if (rc) + goto out_osd; +#else + rc = -EINVAL; + goto out_options; +#endif } - /* In Linux 2.4, the target device doesn't get passed to any of our - functions. So we'll stick it on the end of the options. */ - append_option(options, "device="); - strcat(options, mop.mo_source); + /* In Linux 2.4, the target device doesn't get passed to any of our + functions. So we'll stick it on the end of the options. */ + rc = append_option(options, maxopt_len, "device=", mop.mo_source); + if (rc != 0) + goto out_osd; - if (verbose) - printf("mounting device %s at %s, flags=%#x options=%s\n", + if (verbose) + printf("mounting device %s at %s, flags=%#x options=%s\n", mop.mo_source, mop.mo_target, flags, options); - if (!strstr(mop.mo_usource, ":/") && - osd_tune_lustre(mop.mo_source, &mop)) { +#ifdef HAVE_SERVER_SUPPORT + if (!client && osd_tune_lustre(mop.mo_source, &mop)) { if (verbose) fprintf(stderr, "%s: unable to set tunables for %s" " (may cause reduced IO performance)\n", argv[0], mop.mo_source); } +#endif +#ifdef HAVE_GSS + if (mop.mo_skpath[0] != '\0') { + /* Treat shared key failures as fatal */ + rc = load_shared_keys(&mop); + if (rc) { + fprintf(stderr, + "%s: Error loading shared keys: %s\n", + progname, strerror(rc)); + goto out_osd; + } + } +#endif /* HAVE_GSS */ if (!mop.mo_fake) { - /* flags and target get to lustre_get_sb, but not - lustre_fill_super. Lustre ignores the flags, but mount - does not. */ - for (i = 0, rc = -EAGAIN; i <= mop.mo_retry && rc != 0; i++) { + /* flags and target get to lustre_get_sb(), but not + * lustre_fill_super(). Lustre ignores the flags, but mount + * does not. */ + for (i = 0, rc = -EAGAIN; i <= mop.mo_retry && rc != 0; i++) { rc = mount(mop.mo_source, mop.mo_target, "lustre", flags, (void *)options); - if (rc) { + if (rc != 0) { if (verbose) { fprintf(stderr, "%s: mount %s at %s " "failed: %s retries left: " @@ -612,15 +898,21 @@ int main(int argc, char *const argv[]) mop.mo_retry - i); } - if (mop.mo_retry) { - sleep(1 << max((i/2), 5)); - } - else { - rc = errno; - } - } - } - } + if (mop.mo_retry) { + int limit = i/2 > 5 ? i/2 : 5; + + sleep(1 << limit); + } else { + rc = errno; + } +#ifdef HAVE_SERVER_SUPPORT + } else { + if (!client) + label_lustre(&mop); +#endif + } + } + } if (rc) { char *cli; @@ -685,25 +977,29 @@ int main(int argc, char *const argv[]) rc = WEXITSTATUS(ret); } - } else if (!mop.mo_nomtab) { - rc = update_mtab_entry(mop.mo_usource, mop.mo_target, "lustre", - mop.mo_orig_options, 0,0,0); - - /* change label from : to - - * to indicate the device has been registered. - * only if the label is supposed to be changed and - * target service is supposed to start */ - if (mop.mo_ldd.ldd_flags & (LDD_F_VIRGIN | LDD_F_WRITECONF)) { - if (mop.mo_nosvc == 0 ) - (void) osd_label_lustre(&mop); + } else { + /* Deal with utab just for client. Note that we ignore + * the return value here since it is not worth to fail + * mount by prevent some rare cases */ + if (strstr(mop.mo_usource, ":/") != NULL) + update_utab_entry(&mop); + if (!mop.mo_nomtab) { + rc = update_mtab_entry(mop.mo_usource, mop.mo_target, + "lustre", mop.mo_orig_options, + 0, 0, 0); } - } + } +out_osd: +#ifdef HAVE_SERVER_SUPPORT + if (!client) + osd_fini(); +#endif +out_options: free(options); + +out_mo_source: /* mo_usource should be freed, but we can rely on the kernel */ free(mop.mo_source); - - osd_fini(); - - return rc; + return rc; }