X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Futils%2Flfs.c;h=47860f7ac9a1ebfe87bff72e2f80108bd2376afa;hp=07e3ed77f4f154da80fcb48f41da2b11d96473c3;hb=a1307d41a3a6682f60c91df4b581f21fb2980265;hpb=65701b4a30efdb695776bcf690a2b3cabc928da1 diff --git a/lustre/utils/lfs.c b/lustre/utils/lfs.c index 07e3ed7..47860f7 100644 --- a/lustre/utils/lfs.c +++ b/lustre/utils/lfs.c @@ -1,6 +1,4 @@ -/* -*- 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. @@ -29,7 +27,7 @@ * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, 2012, Whamcloud, Inc. + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -53,8 +51,10 @@ #include #include #include +#include #include #include +#include #include #include #include @@ -65,41 +65,29 @@ # include #endif -/* For dirname() */ -#include - -#include -#include - -#include -#include -#include -#include -#include - +#include #include -#include "obdctl.h" - -unsigned int libcfs_subsystem_debug = 0; +#include +#include /* all functions */ static int lfs_setstripe(int argc, char **argv); static int lfs_find(int argc, char **argv); static int lfs_getstripe(int argc, char **argv); +static int lfs_getdirstripe(int argc, char **argv); +static int lfs_setdirstripe(int argc, char **argv); +static int lfs_rmentry(int argc, char **argv); static int lfs_osts(int argc, char **argv); static int lfs_mdts(int argc, char **argv); static int lfs_df(int argc, char **argv); static int lfs_getname(int argc, char **argv); static int lfs_check(int argc, char **argv); -static int lfs_catinfo(int argc, char **argv); #ifdef HAVE_SYS_QUOTA_H -static int lfs_quotachown(int argc, char **argv); static int lfs_quotacheck(int argc, char **argv); static int lfs_quotaon(int argc, char **argv); static int lfs_quotaoff(int argc, char **argv); static int lfs_setquota(int argc, char **argv); static int lfs_quota(int argc, char **argv); -static int lfs_quotainv(int argc, char **argv); #endif static int lfs_flushctx(int argc, char **argv); static int lfs_join(int argc, char **argv); @@ -114,57 +102,121 @@ static int lfs_changelog(int argc, char **argv); static int lfs_changelog_clear(int argc, char **argv); static int lfs_fid2path(int argc, char **argv); static int lfs_path2fid(int argc, char **argv); +static int lfs_data_version(int argc, char **argv); +static int lfs_hsm_state(int argc, char **argv); +static int lfs_hsm_set(int argc, char **argv); +static int lfs_hsm_clear(int argc, char **argv); +static int lfs_hsm_action(int argc, char **argv); +static int lfs_hsm_archive(int argc, char **argv); +static int lfs_hsm_restore(int argc, char **argv); +static int lfs_hsm_release(int argc, char **argv); +static int lfs_hsm_remove(int argc, char **argv); +static int lfs_hsm_cancel(int argc, char **argv); +static int lfs_swap_layouts(int argc, char **argv); +static int lfs_mv(int argc, char **argv); + +#define SETSTRIPE_USAGE(_cmd, _tgt) \ + "usage: "_cmd" [--stripe-count|-c ]\n"\ + " [--stripe-index|-i ]\n"\ + " [--stripe-size|-S ]\n"\ + " [--pool|-p ]\n"\ + " [--block|-b] "_tgt"\n"\ + " [--ost-list|-o ]\n"\ + "\tstripe_size: Number of bytes on each OST (0 filesystem default)\n"\ + "\t Can be specified with k, m or g (in KB, MB and GB\n"\ + "\t respectively)\n"\ + "\tstart_ost_idx: OST index of first stripe (-1 default)\n"\ + "\tstripe_count: Number of OSTs to stripe over (0 default, -1 all)\n"\ + "\tpool_name: Name of OST pool to use (default none)\n"\ + "\tblock: Block file access during data migration\n"\ + "\tost_indices: List of OST indices, can be repeated multiple times\n"\ + "\t Indices be specified in a format of:\n"\ + "\t -o ,-,\n"\ + "\t Or:\n"\ + "\t -o -o - -o \n"\ + "\t If --pool is set with --ost-list, then the OSTs\n"\ + "\t must be the members of the pool." /* all avaialable commands */ command_t cmdlist[] = { - {"setstripe", lfs_setstripe, 0, - "Create a new file with a specific striping pattern or\n" - "set the default striping pattern on an existing directory or\n" - "delete the default striping pattern from an existing directory\n" - "usage: setstripe [--size|-s stripe_size] [--count|-c stripe_count]\n" - " [--index|-i|--offset|-o start_ost_index]\n" - " [--pool|-p ] \n" - " or \n" - " setstripe -d (to delete default striping)\n" - "\tstripe_size: Number of bytes on each OST (0 filesystem default)\n" - "\t Can be specified with k, m or g (in KB, MB and GB\n" - "\t respectively)\n" - "\tstart_ost_index: OST index of first stripe (-1 default)\n" - "\tstripe_count: Number of OSTs to stripe over (0 default, -1 all)\n" - "\tpool: Name of OST pool to use (default none)"}, - {"getstripe", lfs_getstripe, 0, - "To list the striping info for a given file or files in a\n" - "directory or recursively for all files in a directory tree.\n" - "usage: getstripe [--obd|-O ] [--quiet | -q] [--verbose | -v]\n" - " [--count | -c ] [--index | -i | --offset | -o]\n" - " [--size | -s ] [--pool | -p ] [--directory | -d]\n" - " [--mdt | -M] [--recursive | -r] [--raw | -R]\n" - " ..."}, + {"setstripe", lfs_setstripe, 0, + "Create a new file with a specific striping pattern or\n" + "set the default striping pattern on an existing directory or\n" + "delete the default striping pattern from an existing directory\n" + "usage: setstripe -d (to delete default striping)\n"\ + " or\n" + SETSTRIPE_USAGE("setstripe", "")}, + {"getstripe", lfs_getstripe, 0, + "To list the striping info for a given file or files in a\n" + "directory or recursively for all files in a directory tree.\n" + "usage: getstripe [--ost|-O ] [--quiet | -q] [--verbose | -v]\n" + " [--stripe-count|-c] [--stripe-index|-i]\n" + " [--pool|-p] [--stripe-size|-S] [--directory|-d]\n" + " [--mdt-index|-M] [--recursive|-r] [--raw|-R]\n" + " [--layout|-L]\n" + " ..."}, + {"setdirstripe", lfs_setdirstripe, 0, + "To create a striped directory on a specified MDT. This can only\n" + "be done on MDT0 with the right of administrator.\n" + "usage: setdirstripe <--count|-c stripe_count>\n" + " [--index|-i mdt_index] [--hash-type|-t hash_type]\n" + " [--default_stripe|-D ] [--mode|-m mode] \n" + "\tstripe_count: stripe count of the striped directory\n" + "\tmdt_index: MDT index of first stripe\n" + "\thash_type: hash type of the striped directory. Hash types:\n" + " fnv_1a_64 FNV-1a hash algorithm (default)\n" + " all_char sum of characters % MDT_COUNT (not recommended)\n" + "\tdefault_stripe: set default dirstripe of the directory\n" + "\tmode: the mode of the directory\n"}, + {"getdirstripe", lfs_getdirstripe, 0, + "To list the striping info for a given directory\n" + "or recursively for all directories in a directory tree.\n" + "usage: getdirstripe [--obd|-O ] [--quiet|-q] [--verbose|-v]\n" + " [--count|-c ] [--index|-i ] [--raw|-R]\n" + " [--recursive | -r] [ --default_stripe | -D ] "}, + {"mkdir", lfs_setdirstripe, 0, + "To create a striped directory on a specified MDT. This can only\n" + "be done on MDT0 with the right of administrator.\n" + "usage: mkdir <--count|-c stripe_count>\n" + " [--index|-i mdt_index] [--hash-type|-t hash_type]\n" + " [--default_stripe|-D ] [--mode|-m mode] \n" + "\tstripe_count: stripe count of the striped directory\n" + "\tmdt_index: MDT index of first stripe\n" + "\thash_type: hash type of the striped directory. Hash types:\n" + " fnv_1a_64 FNV-1a hash algorithm (default)\n" + " all_char sum of characters % MDT_COUNT (not recommended)\n" + "\tdefault_stripe: set default dirstripe of the directory\n" + "\tmode: the mode of the directory\n"}, + {"rm_entry", lfs_rmentry, 0, + "To remove the name entry of the remote directory. Note: This\n" + "command will only delete the name entry, i.e. the remote directory\n" + "will become inaccessable after this command. This can only be done\n" + "by the administrator\n" + "usage: rm_entry \n"}, {"pool_list", lfs_poollist, 0, "List pools or pool OSTs\n" "usage: pool_list [.] | \n"}, {"find", lfs_find, 0, - "To find files that match given parameters recursively in a directory tree.\n" + "find files matching given attributes recursively in directory tree.\n" "usage: find ...\n" - " [[!] --atime|-A [+-]N] [[!] --mtime|-M [+-]N] [[!] --ctime|-C [+-]N]\n" - " [--maxdepth|-D N] [[!] --name|-n ] [--print0|-P]\n" - " [--print|-p] [[!] --obd|-O ] [[!] --mdt|-m ]\n" + " [[!] --atime|-A [+-]N] [[!] --ctime|-C [+-]N]\n" + " [[!] --mtime|-M [+-]N] [[!] --mdt|-m ]\n" + " [--maxdepth|-D N] [[!] --name|-n ]\n" + " [[!] --ost|-O ] [--print|-p] [--print0|-P]\n" + " [[!] --size|-s [+-]N[bkMGTPE]]\n" + " [[!] --stripe-count|-c [+-]]\n" + " [[!] --stripe-index|-i ]\n" + " [[!] --stripe-size|-S [+-]N[kMGT]] [[!] --type|-t ]\n" " [[!] --gid|-g|--group|-G |]\n" - " [[!] --uid|-u|--user|-U |]\n" - " [[!] --pool ]\n" - "\t !: used before an option indicates 'NOT' the requested attribute\n" - "\t -: used before an value indicates 'AT MOST' the requested value\n" - "\t +: used before an option indicates 'AT LEAST' the requested value\n"}, + " [[!] --uid|-u|--user|-U |] [[!] --pool ]\n" + " [[!] --layout|-L released,raid0]\n" + "\t !: used before an option indicates 'NOT' requested attribute\n" + "\t -: used before a value indicates 'AT MOST' requested value\n" + "\t +: used before a value indicates 'AT LEAST' requested value\n"}, {"check", lfs_check, 0, "Display the status of MDS or OSTs (as specified in the command)\n" "or all the servers (MDS and OSTs).\n" "usage: check "}, - {"catinfo", lfs_catinfo, 0, - "Show information of specified type logs.\n" - "usage: catinfo {keyword} [node name]\n" - "\tkeywords are one of followings: config, deletions.\n" - "\tnode name must be provided when use keyword config."}, {"join", lfs_join, 0, "join two lustre files into one.\n" "obsolete, HEAD does not support it anymore.\n"}, @@ -175,28 +227,25 @@ command_t cmdlist[] = { {"df", lfs_df, 0, "report filesystem disk space usage or inodes usage" "of each MDS and all OSDs or a batch belonging to a specific pool .\n" - "Usage: df [-i] [-h] [--pool|-p [.] [path]"}, + "Usage: df [-i] [-h] [--lazy|-l] [--pool|-p [.] [path]"}, {"getname", lfs_getname, 0, "list instances and specified mount points " "[for specified path only]\n" "Usage: getname [-h]|[path ...] "}, #ifdef HAVE_SYS_QUOTA_H - {"quotachown",lfs_quotachown, 0, - "Change files' owner or group on the specified filesystem.\n" - "usage: quotachown [-i] \n" - "\t-i: ignore error if file is not exist\n"}, {"quotacheck", lfs_quotacheck, 0, "Scan the specified filesystem for disk usage, and create,\n" - "or update quota files.\n" + "or update quota files. Deprecated as of 2.4.0.\n" "usage: quotacheck [ -ug ] "}, - {"quotaon", lfs_quotaon, 0, "Turn filesystem quotas on.\n" + {"quotaon", lfs_quotaon, 0, "Turn filesystem" + " quotas on. Deprecated as of 2.4.0.\n" "usage: quotaon [ -ugf ] "}, - {"quotaoff", lfs_quotaoff, 0, "Turn filesystem quotas off.\n" + {"quotaoff", lfs_quotaoff, 0, "Turn filesystem" + " quotas off. Deprecated as of 2.4.0.\n" "usage: quotaoff [ -ug ] "}, {"setquota", lfs_setquota, 0, "Set filesystem quotas.\n" "usage: setquota <-u|-g> |||\n" " -b -B \n" " -i -I \n" - " setquota -t <-u|-g> \n" " setquota <-u|--user|-g|--group> |||\n" " [--block-softlimit ]\n" " [--block-hardlimit ]\n" @@ -208,13 +257,19 @@ command_t cmdlist[] = { " -b can be used instead of --block-softlimit/--block-grace\n" " -B can be used instead of --block-hardlimit\n" " -i can be used instead of --inode-softlimit/--inode-grace\n" - " -I can be used instead of --inode-hardlimit"}, + " -I can be used instead of --inode-hardlimit\n\n" + "Note: The total quota space will be split into many qunits and\n" + " balanced over all server targets, the minimal qunit size is\n" + " 1M bytes for block space and 1K inodes for inode space.\n\n" + " Quota space rebalancing process will stop when this mininum\n" + " value is reached. As a result, quota exceeded can be returned\n" + " while many targets still have 1MB or 1K inodes of spare\n" + " quota space."}, {"quota", lfs_quota, 0, "Display disk usage and limits.\n" - "usage: quota [-q] [-v] [-o |-i |-I ]\n" + "usage: quota [-q] [-v] [-h] [-o |-i |-I " + "]\n" " [<-u|-g> |||] \n" " quota [-o |-i |-I ] -t <-u|-g> "}, - {"quotainv", lfs_quotainv, 0, "Invalidate quota data.\n" - "usage: quotainv [-u|-g] "}, #endif {"flushctx", lfs_flushctx, 0, "Flush security context for current user.\n" "usage: flushctx [-k] [mountpoint...]"}, @@ -244,137 +299,574 @@ command_t cmdlist[] = { "interest to consumer , allowing the system to free up space.\n" "An of 0 means all records.\n" "usage: changelog_clear "}, - {"fid2path", lfs_fid2path, 0, - "Resolve the full path to a given FID. For a specific hardlink " - "specify link number .\n" - /* "For a historical name, specify changelog record .\n" */ - "usage: fid2path [--link ]" - /*[--rec ]*/}, - {"path2fid", lfs_path2fid, 0, "Display the fid for a given path.\n" - "usage: path2fid "}, - {"help", Parser_help, 0, "help"}, - {"exit", Parser_quit, 0, "quit"}, - {"quit", Parser_quit, 0, "quit"}, - { 0, 0, 0, NULL } + {"fid2path", lfs_fid2path, 0, + "Resolve the full path(s) for given FID(s). For a specific hardlink " + "specify link number .\n" + /* "For a historical link name, specify changelog record .\n" */ + "usage: fid2path [--link ] ..." + /* [ --rec ] */ }, + {"path2fid", lfs_path2fid, 0, "Display the fid(s) for a given path(s).\n" + "usage: path2fid [--parents] ..."}, + {"data_version", lfs_data_version, 0, "Display file data version for " + "a given path.\n" "usage: data_version -[n|r|w] "}, + {"hsm_state", lfs_hsm_state, 0, "Display the HSM information (states, " + "undergoing actions) for given files.\n usage: hsm_state ..."}, + {"hsm_set", lfs_hsm_set, 0, "Set HSM user flag on specified files.\n" + "usage: hsm_set [--norelease] [--noarchive] [--dirty] [--exists] " + "[--archived] [--lost] ..."}, + {"hsm_clear", lfs_hsm_clear, 0, "Clear HSM user flag on specified " + "files.\n" + "usage: hsm_clear [--norelease] [--noarchive] [--dirty] [--exists] " + "[--archived] [--lost] ..."}, + {"hsm_action", lfs_hsm_action, 0, "Display current HSM request for " + "given files.\n" "usage: hsm_action ..."}, + {"hsm_archive", lfs_hsm_archive, 0, + "Archive file to external storage.\n" + "usage: hsm_archive [--filelist FILELIST] [--data DATA] [--archive NUM] " + " ..."}, + {"hsm_restore", lfs_hsm_restore, 0, + "Restore file from external storage.\n" + "usage: hsm_restore [--filelist FILELIST] [--data DATA] ..."}, + {"hsm_release", lfs_hsm_release, 0, + "Release files from Lustre.\n" + "usage: hsm_release [--filelist FILELIST] [--data DATA] ..."}, + {"hsm_remove", lfs_hsm_remove, 0, + "Remove file copy from external storage.\n" + "usage: hsm_remove [--filelist FILELIST] [--data DATA] ..."}, + {"hsm_cancel", lfs_hsm_cancel, 0, + "Cancel requests related to specified files.\n" + "usage: hsm_cancel [--filelist FILELIST] [--data DATA] ..."}, + {"swap_layouts", lfs_swap_layouts, 0, "Swap layouts between 2 files.\n" + "usage: swap_layouts "}, + {"migrate", lfs_setstripe, 0, "migrate file from one OST layout to " + "another (may be not safe with concurent writes).\n" + SETSTRIPE_USAGE("migrate ", "")}, + {"mv", lfs_mv, 0, + "To move directories between MDTs.\n" + "usage: mv [--mdt-index|-M] " + "[--verbose|-v]\n"}, + {"help", Parser_help, 0, "help"}, + {"exit", Parser_quit, 0, "quit"}, + {"quit", Parser_quit, 0, "quit"}, + {"--version", Parser_version, 0, + "output build version of the utility and exit"}, + { 0, 0, 0, NULL } }; -static int isnumber(const char *str) +/* Generate a random id for the grouplock */ +static int random_group_id(int *gid) { - const char *ptr; - - if (str[0] != '-' && !isdigit(str[0])) - return 0; + int fd; + int rc; + size_t sz = sizeof(*gid); + + fd = open("/dev/urandom", O_RDONLY); + if (fd < 0) { + rc = -errno; + fprintf(stderr, "cannot open /dev/urandom: %s\n", + strerror(-rc)); + goto out; + } + +retry: + rc = read(fd, gid, sz); + if (rc < sz) { + rc = -errno; + fprintf(stderr, "cannot read %zu bytes from /dev/urandom: %s\n", + sz, strerror(-rc)); + goto out; + } + + /* gids must be non-zero */ + if (*gid == 0) + goto retry; - for (ptr = str + 1; *ptr != '\0'; ptr++) { - if (!isdigit(*ptr)) - return 0; - } +out: + if (fd >= 0) + close(fd); - return 1; + return rc; } -/* functions */ -static int lfs_setstripe(int argc, char **argv) +#define MIGRATION_BLOCKS 1 + +static int lfs_migrate(char *name, __u64 migration_flags, + struct llapi_stripe_param *param) { - char *fname; - int result; - unsigned long long st_size; - int st_offset, st_count; - char *end; - int c; - int delete = 0; - char *stripe_size_arg = NULL; - char *stripe_off_arg = NULL; - char *stripe_count_arg = NULL; - char *pool_name_arg = NULL; - unsigned long long size_units = 1; + int fd, fdv; + char volatile_file[PATH_MAX + + LUSTRE_VOLATILE_HDR_LEN + 4]; + char parent[PATH_MAX]; + char *ptr; + int rc; + __u64 dv1; + struct lov_user_md *lum = NULL; + int lumsz; + int bufsz; + void *buf = NULL; + int rsize, wsize; + __u64 rpos, wpos, bufoff; + int gid; + int have_gl = 0; + struct stat st, stv; + + /* find the right size for the IO and allocate the buffer */ + lumsz = lov_user_md_size(LOV_MAX_STRIPE_COUNT, LOV_USER_MAGIC_V3); + lum = malloc(lumsz); + if (lum == NULL) { + rc = -ENOMEM; + goto free; + } + + rc = llapi_file_get_stripe(name, lum); + /* failure can come from may case and some may be not real error + * (eg: no stripe) + * in case of a real error, a later call will failed with a better + * error management */ + if (rc < 0) + bufsz = 1024*1024; + else + bufsz = lum->lmm_stripe_size; + rc = posix_memalign(&buf, getpagesize(), bufsz); + if (rc != 0) { + rc = -rc; + goto free; + } + + if (migration_flags & MIGRATION_BLOCKS) { + rc = random_group_id(&gid); + if (rc < 0) { + fprintf(stderr, "%s: cannot get random group ID: %s\n", + name, strerror(-rc)); + goto free; + } + } + + /* search for file directory pathname */ + if (strlen(name) > sizeof(parent)-1) { + rc = -E2BIG; + goto free; + } + strncpy(parent, name, sizeof(parent)); + ptr = strrchr(parent, '/'); + if (ptr == NULL) { + if (getcwd(parent, sizeof(parent)) == NULL) { + rc = -errno; + goto free; + } + } else { + if (ptr == parent) + strcpy(parent, "/"); + else + *ptr = '\0'; + } + rc = snprintf(volatile_file, sizeof(volatile_file), "%s/%s::", parent, + LUSTRE_VOLATILE_HDR); + if (rc >= sizeof(volatile_file)) { + rc = -E2BIG; + goto free; + } + + /* create, open a volatile file, use caching (ie no directio) */ + /* exclusive create is not needed because volatile files cannot + * conflict on name by construction */ + fdv = llapi_file_open_param(volatile_file, O_CREAT | O_WRONLY, 0644, + param); + if (fdv < 0) { + rc = fdv; + fprintf(stderr, "cannot create volatile file in %s (%s)\n", + parent, strerror(-rc)); + goto free; + } + + /* open file, direct io */ + /* even if the file is only read, WR mode is nedeed to allow + * layout swap on fd */ + fd = open(name, O_RDWR | O_DIRECT); + if (fd == -1) { + rc = -errno; + fprintf(stderr, "cannot open %s (%s)\n", name, strerror(-rc)); + close(fdv); + goto free; + } + + /* Not-owner (root?) special case. + * Need to set owner/group of volatile file like original. + * This will allow to pass related check during layout_swap. + */ + rc = fstat(fd, &st); + if (rc != 0) { + rc = -errno; + fprintf(stderr, "cannot stat %s (%s)\n", name, + strerror(errno)); + goto error; + } + rc = fstat(fdv, &stv); + if (rc != 0) { + rc = -errno; + fprintf(stderr, "cannot stat %s (%s)\n", volatile_file, + strerror(errno)); + goto error; + } + if (st.st_uid != stv.st_uid || st.st_gid != stv.st_gid) { + rc = fchown(fdv, st.st_uid, st.st_gid); + if (rc != 0) { + rc = -errno; + fprintf(stderr, "cannot chown %s (%s)\n", name, + strerror(errno)); + goto error; + } + } + + /* get file data version */ + rc = llapi_get_data_version(fd, &dv1, LL_DV_RD_FLUSH); + if (rc != 0) { + fprintf(stderr, "cannot get dataversion on %s (%s)\n", + name, strerror(-rc)); + goto error; + } + + if (migration_flags & MIGRATION_BLOCKS) { + /* take group lock to limit concurent access + * this will be no more needed when exclusive access will + * be implemented (see LU-2919) */ + /* group lock is taken after data version read because it + * blocks data version call */ + rc = llapi_group_lock(fd, gid); + if (rc < 0) { + fprintf(stderr, "cannot get group lock on %s (%s)\n", + name, strerror(-rc)); + goto error; + } + have_gl = 1; + } + + /* copy data */ + rpos = 0; + wpos = 0; + bufoff = 0; + rsize = -1; + do { + /* read new data only if we have written all + * previously read data */ + if (wpos == rpos) { + rsize = read(fd, buf, bufsz); + if (rsize < 0) { + rc = -errno; + fprintf(stderr, "read failed on %s" + " (%s)\n", name, + strerror(-rc)); + goto error; + } + rpos += rsize; + bufoff = 0; + } + /* eof ? */ + if (rsize == 0) + break; + wsize = write(fdv, buf + bufoff, rpos - wpos); + if (wsize < 0) { + rc = -errno; + fprintf(stderr, "write failed on volatile" + " for %s (%s)\n", name, strerror(-rc)); + goto error; + } + wpos += wsize; + bufoff += wsize; + } while (1); + + /* flush data */ + fsync(fdv); + + if (migration_flags & MIGRATION_BLOCKS) { + /* give back group lock */ + rc = llapi_group_unlock(fd, gid); + if (rc < 0) + fprintf(stderr, "cannot put group lock on %s (%s)\n", + name, strerror(-rc)); + have_gl = 0; + } + + /* swap layouts + * for a migration we need to: + * - check data version on file did not change + * - keep file mtime + * - keep file atime + */ + rc = llapi_fswap_layouts(fd, fdv, dv1, 0, + SWAP_LAYOUTS_CHECK_DV1 | + SWAP_LAYOUTS_KEEP_MTIME | + SWAP_LAYOUTS_KEEP_ATIME); + if (rc == -EAGAIN) { + fprintf(stderr, "%s: dataversion changed during copy, " + "migration aborted\n", name); + goto error; + } + if (rc != 0) + fprintf(stderr, "%s: swap layout to new file failed: %s\n", + name, strerror(-rc)); - struct option long_opts[] = { - {"count", required_argument, 0, 'c'}, - {"delete", no_argument, 0, 'd'}, - {"index", required_argument, 0, 'i'}, - {"offset", required_argument, 0, 'o'}, - {"pool", required_argument, 0, 'p'}, - {"size", required_argument, 0, 's'}, - {0, 0, 0, 0} - }; +error: + /* give back group lock */ + if ((migration_flags & MIGRATION_BLOCKS) && have_gl) { + int rc2; + + /* we keep the original error in rc */ + rc2 = llapi_group_unlock(fd, gid); + if (rc2 < 0) + fprintf(stderr, "cannot put group lock on %s (%s)\n", + name, strerror(-rc2)); + } + + close(fdv); + close(fd); +free: + if (lum) + free(lum); + if (buf) + free(buf); + return rc; +} - st_size = 0; - st_offset = -1; - st_count = 0; +/** + * Parse a string containing an OST index list into an array of integers. + * + * The input string contains a comma delimited list of individual + * indices and ranges, for example "1,2-4,7". Add the indices into the + * \a osts array and remove duplicates. + * + * \param[out] osts array to store indices in + * \param[in] size size of \a osts array + * \param[in] offset starting index in \a osts + * \param[in] arg string containing OST index list + * + * \retval positive number of indices in \a osts + * \retval -EINVAL unable to parse \a arg + */ +static int parse_targets(__u32 *osts, int size, int offset, char *arg) +{ + int rc; + int nr = offset; + int slots = size - offset; + char *ptr = NULL; + bool end_of_loop; + + if (arg == NULL) + return -EINVAL; + + end_of_loop = false; + while (!end_of_loop) { + int start_index; + int end_index; + int i; + char *endptr = NULL; + + rc = -EINVAL; + + ptr = strchrnul(arg, ','); + + end_of_loop = *ptr == '\0'; + *ptr = '\0'; + + start_index = strtol(arg, &endptr, 0); + if (endptr == arg) /* no data at all */ + break; + if (*endptr != '-' && *endptr != '\0') /* has invalid data */ + break; + if (start_index < 0) + break; + + end_index = start_index; + if (*endptr == '-') { + end_index = strtol(endptr + 1, &endptr, 0); + if (*endptr != '\0') + break; + if (end_index < start_index) + break; + } + + for (i = start_index; i <= end_index && slots > 0; i++) { + int j; + + /* remove duplicate */ + for (j = 0; j < offset; j++) { + if (osts[j] == i) + break; + } + if (j == offset) { /* no duplicate */ + osts[nr++] = i; + --slots; + } + } + if (slots == 0 && i < end_index) + break; + + *ptr = ','; + arg = ++ptr; + offset = nr; + rc = 0; + } + if (!end_of_loop && ptr != NULL) + *ptr = ','; + + return rc < 0 ? rc : nr; +} -#if LUSTRE_VERSION < OBD_OCD_VERSION(2,4,50,0) - if (argc == 5 && argv[1][0] != '-' && - isnumber(argv[2]) && isnumber(argv[3]) && isnumber(argv[4])) { - fprintf(stderr, "error: obsolete usage of setstripe " - "positional parameters. Use -c, -i, -s instead.\n"); - return CMD_HELP; - } else -#else -#warning "remove obsolete positional parameter code" +/* functions */ +static int lfs_setstripe(int argc, char **argv) +{ + struct llapi_stripe_param *param; + char *fname; + int result; + unsigned long long st_size; + int st_offset, st_count; + char *end; + int c; + int delete = 0; + char *stripe_size_arg = NULL; + char *stripe_off_arg = NULL; + char *stripe_count_arg = NULL; + char *pool_name_arg = NULL; + unsigned long long size_units = 1; + bool migrate_mode = false; + __u64 migration_flags = 0; + __u32 osts[LOV_MAX_STRIPE_COUNT] = { 0 }; + int nr_osts = 0; + + struct option long_opts[] = { + /* valid only in migrate mode */ + {"block", no_argument, 0, 'b'}, +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "stripe-count", but was explicitly + * made "stripe-count" for consistency with other options, + * and to separate it from "mdt-count" when DNE arrives. */ + {"count", required_argument, 0, 'c'}, #endif - { - optind = 0; - while ((c = getopt_long(argc, argv, "c:di:o:p:s:", - long_opts, NULL)) >= 0) { - switch (c) { - case 0: - /* Long options. */ - break; - case 'c': - stripe_count_arg = optarg; - break; - case 'd': - /* delete the default striping pattern */ - delete = 1; - break; - case 'i': - case 'o': - stripe_off_arg = optarg; - break; - case 's': - stripe_size_arg = optarg; - break; - case 'p': - pool_name_arg = optarg; - break; - case '?': - return CMD_HELP; - default: - fprintf(stderr, "error: %s: option '%s' " - "unrecognized\n", - argv[0], argv[optind - 1]); - return CMD_HELP; - } - } - - fname = argv[optind]; - - if (delete && - (stripe_size_arg != NULL || stripe_off_arg != NULL || - stripe_count_arg != NULL || pool_name_arg != NULL)) { - fprintf(stderr, "error: %s: cannot specify -d with " - "-s, -c -o or -p options\n", - argv[0]); - return CMD_HELP; - } - } - - if (optind == argc) { - fprintf(stderr, "error: %s: missing filename|dirname\n", - argv[0]); - return CMD_HELP; - } - - /* get the stripe size */ - if (stripe_size_arg != NULL) { - result = parse_size(stripe_size_arg, &st_size, &size_units, 0); - if (result) { - fprintf(stderr, "error: %s: bad size '%s'\n", - argv[0], stripe_size_arg); - return result; - } - } + {"stripe-count", required_argument, 0, 'c'}, + {"stripe_count", required_argument, 0, 'c'}, + {"delete", no_argument, 0, 'd'}, +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "stripe-index", but was explicitly + * made "stripe-index" for consistency with other options, + * and to separate it from "mdt-index" when DNE arrives. */ + {"index", required_argument, 0, 'i'}, +#endif + {"stripe-index", required_argument, 0, 'i'}, + {"stripe_index", required_argument, 0, 'i'}, + {"ost-list", required_argument, 0, 'o'}, + {"ost_list", required_argument, 0, 'o'}, + {"pool", required_argument, 0, 'p'}, +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "--stripe-size", but was confusing + * with "lfs find --size|-s", which means "file size", so use + * the consistent "--stripe-size|-S" for all commands. */ + {"size", required_argument, 0, 's'}, +#endif + {"stripe-size", required_argument, 0, 'S'}, + {"stripe_size", required_argument, 0, 'S'}, + {0, 0, 0, 0} + }; + + st_size = 0; + st_offset = -1; + st_count = 0; + + if (strcmp(argv[0], "migrate") == 0) + migrate_mode = true; + + optind = 0; + while ((c = getopt_long(argc, argv, "c:di:o:p:s:S:", + long_opts, NULL)) >= 0) { + switch (c) { + case 0: + /* Long options. */ + break; + case 'b': + if (!migrate_mode) { + fprintf(stderr, "--block is valid only for" + " migrate mode"); + return CMD_HELP; + } + migration_flags |= MIGRATION_BLOCKS; + break; + case 'c': +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0) + if (strcmp(argv[optind - 1], "--count") == 0) + fprintf(stderr, "warning: '--count' deprecated" + ", use '--stripe-count' instead\n"); +#endif + stripe_count_arg = optarg; + break; + case 'd': + /* delete the default striping pattern */ + delete = 1; + break; + case 'o': + nr_osts = parse_targets(osts, ARRAY_SIZE(osts), nr_osts, + optarg); + if (nr_osts < 0) { + fprintf(stderr, + "error: %s: bad OST indices '%s'\n", + argv[0], optarg); + return CMD_HELP; + } + + if (st_offset == -1) /* first in the command line */ + st_offset = osts[0]; + break; + case 'i': +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0) + if (strcmp(argv[optind - 1], "--index") == 0) + fprintf(stderr, "warning: '--index' deprecated" + ", use '--stripe-index' instead\n"); +#endif + stripe_off_arg = optarg; + break; +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + case 's': +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0) + fprintf(stderr, "warning: '--size|-s' deprecated, " + "use '--stripe-size|-S' instead\n"); +#endif +#endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) */ + case 'S': + stripe_size_arg = optarg; + break; + case 'p': + pool_name_arg = optarg; + break; + default: + return CMD_HELP; + } + } + + fname = argv[optind]; + + if (delete && + (stripe_size_arg != NULL || stripe_off_arg != NULL || + stripe_count_arg != NULL || pool_name_arg != NULL)) { + fprintf(stderr, "error: %s: cannot specify -d with " + "-s, -c, -o, or -p options\n", + argv[0]); + return CMD_HELP; + } + + if (optind == argc) { + fprintf(stderr, "error: %s: missing filename|dirname\n", + argv[0]); + return CMD_HELP; + } + + /* get the stripe size */ + if (stripe_size_arg != NULL) { + result = llapi_parse_size(stripe_size_arg, &st_size, + &size_units, 0); + if (result) { + fprintf(stderr, "error: %s: bad stripe size '%s'\n", + argv[0], stripe_size_arg); + return result; + } + } /* get the stripe offset */ if (stripe_off_arg != NULL) { st_offset = strtol(stripe_off_arg, &end, 0); @@ -394,18 +886,56 @@ static int lfs_setstripe(int argc, char **argv) } } - do { - result = llapi_file_create_pool(fname, st_size, st_offset, - st_count, 0, pool_name_arg); - if (result) { - fprintf(stderr,"error: %s: create stripe file '%s' " - "failed\n", argv[0], fname); - break; - } - fname = argv[++optind]; - } while (fname != NULL); - - return result; + /* initialize stripe parameters */ + param = calloc(1, offsetof(typeof(*param), lsp_osts[nr_osts])); + if (param == NULL) { + fprintf(stderr, "error: %s: run out of memory\n", argv[0]); + return CMD_HELP; + } + + param->lsp_stripe_size = st_size; + param->lsp_stripe_offset = st_offset; + param->lsp_stripe_count = st_count; + param->lsp_stripe_pattern = 0; + param->lsp_pool = pool_name_arg; + param->lsp_is_specific = false; + if (nr_osts > 0) { + if (st_count > 0 && nr_osts != st_count) { + fprintf(stderr, "error: %s: stripe count '%d' doesn't " + "match the number of OSTs: %d\n", + argv[0], st_count, nr_osts); + return CMD_HELP; + } + + param->lsp_is_specific = true; + param->lsp_stripe_count = nr_osts; + memcpy(param->lsp_osts, osts, sizeof(*osts) * nr_osts); + } + + do { + if (!migrate_mode) { + result = llapi_file_open_param(fname, + O_CREAT | O_WRONLY, + 0644, param); + if (result >= 0) { + close(result); + result = 0; + } + } else { + result = lfs_migrate(fname, migration_flags, param); + } + if (result) { + fprintf(stderr, + "error: %s: %s stripe file '%s' failed\n", + argv[0], migrate_mode ? "migrate" : "create", + fname); + break; + } + fname = argv[++optind]; + } while (fname != NULL); + + free(param); + return result; } static int lfs_poollist(int argc, char **argv) @@ -498,36 +1028,62 @@ static int id2name(char **name, unsigned int id, int type) return 0; } +static int name2layout(__u32 *layout, char *name) +{ + char *ptr, *lyt; + + *layout = 0; + for (ptr = name; ; ptr = NULL) { + lyt = strtok(ptr, ","); + if (lyt == NULL) + break; + if (strcmp(lyt, "released") == 0) + *layout |= LOV_PATTERN_F_RELEASED; + else if (strcmp(lyt, "raid0") == 0) + *layout |= LOV_PATTERN_RAID0; + else + return -1; + } + return 0; +} + #define FIND_POOL_OPT 3 static int lfs_find(int argc, char **argv) { - int new_fashion = 1; - int c, ret; + int c, rc; + int ret = 0; time_t t; - struct find_param param = { .maxdepth = -1, .size_units = 0 }; + struct find_param param = { + .fp_max_depth = -1, + .quiet = 1, + }; struct option long_opts[] = { - {"atime", required_argument, 0, 'A'}, - {"ctime", required_argument, 0, 'C'}, - {"maxdepth", required_argument, 0, 'D'}, - {"gid", required_argument, 0, 'g'}, - {"group", required_argument, 0, 'G'}, - {"mtime", required_argument, 0, 'M'}, - {"mdt", required_argument, 0, 'm'}, - {"name", required_argument, 0, 'n'}, - /* --obd is considered as a new option. */ - {"obd", required_argument, 0, 'O'}, - {"ost", required_argument, 0, 'O'}, + {"atime", required_argument, 0, 'A'}, + {"stripe-count", required_argument, 0, 'c'}, + {"stripe_count", required_argument, 0, 'c'}, + {"ctime", required_argument, 0, 'C'}, + {"maxdepth", required_argument, 0, 'D'}, + {"gid", required_argument, 0, 'g'}, + {"group", required_argument, 0, 'G'}, + {"stripe-index", required_argument, 0, 'i'}, + {"stripe_index", required_argument, 0, 'i'}, + {"layout", required_argument, 0, 'L'}, + {"mdt", required_argument, 0, 'm'}, + {"mtime", required_argument, 0, 'M'}, + {"name", required_argument, 0, 'n'}, + /* reserve {"or", no_argument, , 0, 'o'}, to match find(1) */ + {"obd", required_argument, 0, 'O'}, + {"ost", required_argument, 0, 'O'}, /* no short option for pool, p/P already used */ - {"pool", required_argument, 0, FIND_POOL_OPT}, - {"print0", no_argument, 0, 'p'}, - {"print", no_argument, 0, 'P'}, - {"quiet", no_argument, 0, 'q'}, - {"recursive", no_argument, 0, 'r'}, - {"size", required_argument, 0, 's'}, - {"type", required_argument, 0, 't'}, - {"uid", required_argument, 0, 'u'}, - {"user", required_argument, 0, 'U'}, - {"verbose", no_argument, 0, 'v'}, + {"pool", required_argument, 0, FIND_POOL_OPT}, + {"print0", no_argument, 0, 'p'}, + {"print", no_argument, 0, 'P'}, + {"size", required_argument, 0, 's'}, + {"stripe-size", required_argument, 0, 'S'}, + {"stripe_size", required_argument, 0, 'S'}, + {"type", required_argument, 0, 't'}, + {"uid", required_argument, 0, 'u'}, + {"user", required_argument, 0, 'U'}, {0, 0, 0, 0} }; int pathstart = -1; @@ -540,10 +1096,11 @@ static int lfs_find(int argc, char **argv) time(&t); - optind = 0; - /* when getopt_long_only() hits '!' it returns 1, puts "!" in optarg */ - while ((c = getopt_long_only(argc, argv, "-A:C:D:g:G:M:m:n:O:" - "Ppqrs:t:u:U:v", long_opts, NULL)) >= 0) { + optind = 0; + /* when getopt_long_only() hits '!' it returns 1, puts "!" in optarg */ + while ((c = getopt_long_only(argc, argv, + "-A:c:C:D:g:G:i:L:m:M:n:O:Ppqrs:S:t:u:U:v", + long_opts, NULL)) >= 0) { xtime = NULL; xsign = NULL; if (neg_opt) @@ -565,13 +1122,8 @@ static int lfs_find(int argc, char **argv) } if (!isoption && pathstart == -1) pathstart = optind - 1; - if (isoption && pathstart != -1 && pathend == -1) { + if (isoption && pathstart != -1 && pathend == -1) pathend = optind - 2; - if ((c == 1 && strcmp(optarg, "!") == 0) || - c == 'P' || c == 'p' || c == 'O' || - c == 'q' || c == 'r' || c == 'v') - pathend = optind - 1; - } switch (c) { case 0: /* Long options. */ @@ -583,41 +1135,59 @@ static int lfs_find(int argc, char **argv) if (strcmp(optarg, "!") == 0) neg_opt = 2; break; - case 'A': - xtime = ¶m.atime; - xsign = ¶m.asign; - param.exclude_atime = !!neg_opt; - case 'C': - if (c == 'C') { - xtime = ¶m.ctime; - xsign = ¶m.csign; - param.exclude_ctime = !!neg_opt; - } - case 'M': - if (c == 'M') { - xtime = ¶m.mtime; - xsign = ¶m.msign; - param.exclude_mtime = !!neg_opt; + case 'A': + xtime = ¶m.fp_atime; + xsign = ¶m.fp_asign; + param.fp_exclude_atime = !!neg_opt; + /* no break, this falls through to 'C' for ctime */ + case 'C': + if (c == 'C') { + xtime = ¶m.fp_ctime; + xsign = ¶m.fp_csign; + param.fp_exclude_ctime = !!neg_opt; + } + /* no break, this falls through to 'M' for mtime */ + case 'M': + if (c == 'M') { + xtime = ¶m.fp_mtime; + xsign = ¶m.fp_msign; + param.fp_exclude_mtime = !!neg_opt; + } + rc = set_time(&t, xtime, optarg); + if (rc == INT_MAX) { + ret = -1; + goto err; + } + if (rc) + *xsign = rc; + break; + case 'c': + if (optarg[0] == '+') { + param.stripecount_sign = -1; + optarg++; + } else if (optarg[0] == '-') { + param.stripecount_sign = 1; + optarg++; } - new_fashion = 1; - ret = set_time(&t, xtime, optarg); - if (ret == INT_MAX) { + + param.stripecount = strtoul(optarg, &endptr, 0); + if (*endptr != '\0') { + fprintf(stderr,"error: bad stripe_count '%s'\n", + optarg); ret = -1; goto err; } - if (ret) - *xsign = ret; - break; - case 'D': - new_fashion = 1; - param.maxdepth = strtol(optarg, 0, 0); + param.check_stripecount = 1; + param.exclude_stripecount = !!neg_opt; break; - case 'g': - case 'G': - new_fashion = 1; - ret = name2id(¶m.gid, optarg, GROUP); - if (ret) { - param.gid = strtoul(optarg, &endptr, 10); + case 'D': + param.fp_max_depth = strtol(optarg, 0, 0); + break; + case 'g': + case 'G': + rc = name2id(¶m.fp_gid, optarg, GROUP); + if (rc) { + param.fp_gid = strtoul(optarg, &endptr, 10); if (*endptr != '\0') { fprintf(stderr, "Group/GID: %s cannot " "be found.\n", optarg); @@ -625,15 +1195,21 @@ static int lfs_find(int argc, char **argv) goto err; } } - param.exclude_gid = !!neg_opt; - param.check_gid = 1; + param.fp_exclude_gid = !!neg_opt; + param.fp_check_gid = 1; break; + case 'L': + ret = name2layout(¶m.layout, optarg); + if (ret) + goto err; + param.exclude_layout = !!neg_opt; + param.check_layout = 1; + break; case 'u': case 'U': - new_fashion = 1; - ret = name2id(¶m.uid, optarg, USER); - if (ret) { - param.uid = strtoul(optarg, &endptr, 10); + rc = name2id(¶m.fp_uid, optarg, USER); + if (rc) { + param.fp_uid = strtoul(optarg, &endptr, 10); if (*endptr != '\0') { fprintf(stderr, "User/UID: %s cannot " "be found.\n", optarg); @@ -641,11 +1217,10 @@ static int lfs_find(int argc, char **argv) goto err; } } - param.exclude_uid = !!neg_opt; - param.check_uid = 1; + param.fp_exclude_uid = !!neg_opt; + param.fp_check_uid = 1; break; case FIND_POOL_OPT: - new_fashion = 1; if (strlen(optarg) > LOV_MAXPOOLNAME) { fprintf(stderr, "Pool name %s is too long" @@ -662,11 +1237,11 @@ static int lfs_find(int argc, char **argv) param.check_pool = 1; break; case 'n': - new_fashion = 1; param.pattern = (char *)optarg; param.exclude_pattern = !!neg_opt; break; case 'm': + case 'i': case 'O': { char *buf, *token, *next, *p; int len = 1; @@ -688,40 +1263,54 @@ static int lfs_find(int argc, char **argv) token++; } } - if (c == 'O') { - param.exclude_obd = !!neg_opt; - param.num_alloc_obds += len; - tmp = realloc(param.obduuid, - param.num_alloc_obds * - sizeof(*param.obduuid)); - if (tmp == NULL) - GOTO(err_free, ret = -ENOMEM); - param.obduuid = tmp; - } else { + if (c == 'm') { param.exclude_mdt = !!neg_opt; param.num_alloc_mdts += len; tmp = realloc(param.mdtuuid, param.num_alloc_mdts * sizeof(*param.mdtuuid)); - if (tmp == NULL) - GOTO(err_free, ret = -ENOMEM); + if (tmp == NULL) { + ret = -ENOMEM; + goto err_free; + } + param.mdtuuid = tmp; + } else { + param.exclude_obd = !!neg_opt; + param.num_alloc_obds += len; + tmp = realloc(param.obduuid, + param.num_alloc_obds * + sizeof(*param.obduuid)); + if (tmp == NULL) { + ret = -ENOMEM; + goto err_free; + } + + param.obduuid = tmp; } for (token = buf; token && *token; token = next) { - char *uuid; - if (c == 'O') - uuid = - param.obduuid[param.num_obds++].uuid; - else - uuid = - param.mdtuuid[param.num_mdts++].uuid; + struct obd_uuid *puuid; + if (c == 'm') { + puuid = + ¶m.mdtuuid[param.num_mdts++]; + } else { + puuid = + ¶m.obduuid[param.num_obds++]; + } p = strchr(token, ','); next = 0; if (p) { *p = 0; next = p+1; } - strcpy((char *)uuid, token); + + if (strlen(token) > sizeof(puuid->uuid) - 1) { + ret = -E2BIG; + goto err_free; + } + + strncpy(puuid->uuid, token, + sizeof(puuid->uuid)); } err_free: if (buf) @@ -729,68 +1318,80 @@ err_free: break; } case 'p': - new_fashion = 1; param.zeroend = 1; break; case 'P': break; - case 'q': - new_fashion = 0; - param.quiet++; - param.verbose = 0; - break; - case 'r': - new_fashion = 0; - param.recursive = 1; - break; - case 't': - param.exclude_type = !!neg_opt; - switch(optarg[0]) { - case 'b': param.type = S_IFBLK; break; - case 'c': param.type = S_IFCHR; break; - case 'd': param.type = S_IFDIR; break; - case 'f': param.type = S_IFREG; break; - case 'l': param.type = S_IFLNK; break; - case 'p': param.type = S_IFIFO; break; - case 's': param.type = S_IFSOCK; break; -#ifdef S_IFDOOR /* Solaris only */ - case 'D': param.type = S_IFDOOR; break; -#endif - default: fprintf(stderr, "error: %s: bad type '%s'\n", - argv[0], optarg); - ret = CMD_HELP; - goto err; - }; - break; - case 's': - if (optarg[0] == '+') - param.size_sign = -1; - else if (optarg[0] == '-') - param.size_sign = +1; - - if (param.size_sign) - optarg++; - ret = parse_size(optarg, ¶m.size, - ¶m.size_units, 0); - if (ret) { - fprintf(stderr,"error: bad size '%s'\n", - optarg); - goto err; - } - param.check_size = 1; - param.exclude_size = !!neg_opt; - break; - case 'v': - new_fashion = 0; - param.verbose++; - param.quiet = 0; - break; - case '?': - ret = CMD_HELP; - goto err; + case 's': + if (optarg[0] == '+') { + param.size_sign = -1; + optarg++; + } else if (optarg[0] == '-') { + param.size_sign = 1; + optarg++; + } + + ret = llapi_parse_size(optarg, ¶m.size, + ¶m.size_units, 0); + if (ret) { + fprintf(stderr, "error: bad file size '%s'\n", + optarg); + goto err; + } + param.check_size = 1; + param.exclude_size = !!neg_opt; + break; + case 'S': + if (optarg[0] == '+') { + param.stripesize_sign = -1; + optarg++; + } else if (optarg[0] == '-') { + param.stripesize_sign = 1; + optarg++; + } + + ret = llapi_parse_size(optarg, ¶m.stripesize, + ¶m.stripesize_units, 0); + if (ret) { + fprintf(stderr, "error: bad stripe_size '%s'\n", + optarg); + goto err; + } + param.check_stripesize = 1; + param.exclude_stripesize = !!neg_opt; + break; + case 't': + param.fp_exclude_type = !!neg_opt; + switch (optarg[0]) { + case 'b': + param.fp_type = S_IFBLK; + break; + case 'c': + param.fp_type = S_IFCHR; + break; + case 'd': + param.fp_type = S_IFDIR; + break; + case 'f': + param.fp_type = S_IFREG; + break; + case 'l': + param.fp_type = S_IFLNK; + break; + case 'p': + param.fp_type = S_IFIFO; + break; + case 's': + param.fp_type = S_IFSOCK; + break; + default: + fprintf(stderr, "error: %s: bad type '%s'\n", + argv[0], optarg); + ret = CMD_HELP; + goto err; + }; + break; default: - fprintf(stderr, "error: %s: option '%s' unrecognized\n", - argv[0], argv[optind - 1]); ret = CMD_HELP; goto err; }; @@ -806,25 +1407,11 @@ err_free: pathend = argc; } - if (new_fashion) { - param.quiet = 1; - } else { - static int deprecated_warning; - if (!deprecated_warning) { - fprintf(stderr, "lfs find: -q, -r, -v options " - "deprecated. Use 'lfs getstripe' instead.\n"); - deprecated_warning = 1; - } - if (!param.recursive && param.maxdepth == -1) - param.maxdepth = 1; - } - - do { - if (new_fashion) - ret = llapi_find(argv[pathstart], ¶m); - else - ret = llapi_getstripe(argv[pathstart], ¶m); - } while (++pathstart < pathend && !ret); + do { + rc = llapi_find(argv[pathstart], ¶m); + if (rc != 0 && ret == 0) + ret = rc; + } while (++pathstart < pathend); if (ret) fprintf(stderr, "error: %s failed for %s.\n", @@ -839,120 +1426,176 @@ err: return ret; } -static int lfs_getstripe(int argc, char **argv) +static int lfs_getstripe_internal(int argc, char **argv, + struct find_param *param) { - struct option long_opts[] = { - {"count", 0, 0, 'c'}, - {"directory", 0, 0, 'd'}, - {"generation", 0, 0, 'g'}, - {"index", 0, 0, 'i'}, - {"mdt", 0, 0, 'M'}, - {"offset", 0, 0, 'o'}, - {"obd", 1, 0, 'O'}, - {"pool", 0, 0, 'p'}, - {"quiet", 0, 0, 'q'}, - {"recursive", 0, 0, 'r'}, - {"raw", 0, 0, 'R'}, - {"size", 0, 0, 's'}, - {"verbose", 0, 0, 'v'}, - {0, 0, 0, 0} - }; - int c, rc; - struct find_param param = { 0 }; - - param.maxdepth = 1; - optind = 0; - while ((c = getopt_long(argc, argv, "cdghiMoO:pqrRsv", - long_opts, NULL)) != -1) { - switch (c) { - case 'O': - if (param.obduuid) { - fprintf(stderr, - "error: %s: only one obduuid allowed", - argv[0]); - return CMD_HELP; - } - param.obduuid = (struct obd_uuid *)optarg; - break; - case 'q': - param.quiet++; - break; - case 'd': - param.maxdepth = 0; - break; - case 'r': - param.recursive = 1; - break; - case 'v': - param.verbose = VERBOSE_ALL | VERBOSE_DETAIL; - break; - case 'c': - if (!(param.verbose & VERBOSE_DETAIL)) { - param.verbose |= VERBOSE_COUNT; - param.maxdepth = 0; - } - break; - case 's': - if (!(param.verbose & VERBOSE_DETAIL)) { - param.verbose |= VERBOSE_SIZE; - param.maxdepth = 0; - } - break; - case 'i': - case 'o': - if (!(param.verbose & VERBOSE_DETAIL)) { - param.verbose |= VERBOSE_OFFSET; - param.maxdepth = 0; - } - break; - case 'p': - if (!(param.verbose & VERBOSE_DETAIL)) { - param.verbose |= VERBOSE_POOL; - param.maxdepth = 0; - } - break; - case 'g': - if (!(param.verbose & VERBOSE_DETAIL)) { - param.verbose |= VERBOSE_GENERATION; - param.maxdepth = 0; - } - break; - case 'M': - if (!(param.verbose & VERBOSE_DETAIL)) - param.maxdepth = 0; - param.get_mdt_index = 1; - break; - case 'R': - param.raw = 1; - break; - case '?': - return CMD_HELP; - default: - fprintf(stderr, "error: %s: option '%s' unrecognized\n", - argv[0], argv[optind - 1]); - return CMD_HELP; - } - } - - if (optind >= argc) - return CMD_HELP; - - if (param.recursive) - param.maxdepth = -1; - - if (!param.verbose) - param.verbose = VERBOSE_ALL; - if (param.quiet) - param.verbose = VERBOSE_OBJID; - - do { - rc = llapi_getstripe(argv[optind], ¶m); - } while (++optind < argc && !rc); - - if (rc) - fprintf(stderr, "error: %s failed for %s.\n", - argv[0], argv[optind - 1]); - return rc; + struct option long_opts[] = { +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "stripe-count", but was explicitly + * made "stripe-count" for consistency with other options, + * and to separate it from "mdt-count" when DNE arrives. */ + {"count", no_argument, 0, 'c'}, +#endif + {"stripe-count", no_argument, 0, 'c'}, + {"stripe_count", no_argument, 0, 'c'}, + {"directory", no_argument, 0, 'd'}, + {"default", no_argument, 0, 'D'}, + {"generation", no_argument, 0, 'g'}, +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "stripe-index", but was explicitly + * made "stripe-index" for consistency with other options, + * and to separate it from "mdt-index" when DNE arrives. */ + {"index", no_argument, 0, 'i'}, +#endif + {"stripe-index", no_argument, 0, 'i'}, + {"stripe_index", no_argument, 0, 'i'}, + {"layout", no_argument, 0, 'L'}, + {"mdt-index", no_argument, 0, 'M'}, + {"mdt_index", no_argument, 0, 'M'}, +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "stripe-index", but was confusing + * with "file offset" (which will eventually be needed for + * with different layouts by offset), so deprecate it. */ + {"offset", no_argument, 0, 'o'}, +#endif + {"obd", required_argument, 0, 'O'}, + {"ost", required_argument, 0, 'O'}, + {"pool", no_argument, 0, 'p'}, + {"quiet", no_argument, 0, 'q'}, + {"recursive", no_argument, 0, 'r'}, + {"raw", no_argument, 0, 'R'}, +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + /* This formerly implied "--stripe-size", but was confusing + * with "lfs find --size|-s", which means "file size", so use + * the consistent "--stripe-size|-S" for all commands. */ + {"size", no_argument, 0, 's'}, +#endif + {"stripe-size", no_argument, 0, 'S'}, + {"stripe_size", no_argument, 0, 'S'}, + {"verbose", no_argument, 0, 'v'}, + {0, 0, 0, 0} + }; + int c, rc; + + param->fp_max_depth = 1; + optind = 0; + while ((c = getopt_long(argc, argv, "cdDghiLMoO:pqrRsSv", + long_opts, NULL)) != -1) { + switch (c) { + case 'O': + if (param->obduuid) { + fprintf(stderr, + "error: %s: only one obduuid allowed", + argv[0]); + return CMD_HELP; + } + param->obduuid = (struct obd_uuid *)optarg; + break; + case 'q': + param->quiet++; + break; + case 'd': + param->fp_max_depth = 0; + break; + case 'D': + param->get_default_lmv = 1; + break; + case 'r': + param->recursive = 1; + break; + case 'v': + param->verbose = VERBOSE_ALL | VERBOSE_DETAIL; + break; + case 'c': +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0) + if (strcmp(argv[optind - 1], "--count") == 0) + fprintf(stderr, "warning: '--count' deprecated," + " use '--stripe-count' instead\n"); +#endif + if (!(param->verbose & VERBOSE_DETAIL)) { + param->verbose |= VERBOSE_COUNT; + param->fp_max_depth = 0; + } + break; +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + case 's': +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0) + fprintf(stderr, "warning: '--size|-s' deprecated, " + "use '--stripe-size|-S' instead\n"); +#endif +#endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) */ + case 'S': + if (!(param->verbose & VERBOSE_DETAIL)) { + param->verbose |= VERBOSE_SIZE; + param->fp_max_depth = 0; + } + break; +#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 9, 53, 0) + case 'o': + fprintf(stderr, "warning: '--offset|-o' deprecated, " + "use '--stripe-index|-i' instead\n"); +#endif + case 'i': +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 6, 53, 0) + if (strcmp(argv[optind - 1], "--index") == 0) + fprintf(stderr, "warning: '--index' deprecated" + ", use '--stripe-index' instead\n"); +#endif + if (!(param->verbose & VERBOSE_DETAIL)) { + param->verbose |= VERBOSE_OFFSET; + param->fp_max_depth = 0; + } + break; + case 'p': + if (!(param->verbose & VERBOSE_DETAIL)) { + param->verbose |= VERBOSE_POOL; + param->fp_max_depth = 0; + } + break; + case 'g': + if (!(param->verbose & VERBOSE_DETAIL)) { + param->verbose |= VERBOSE_GENERATION; + param->fp_max_depth = 0; + } + break; + case 'L': + if (!(param->verbose & VERBOSE_DETAIL)) { + param->verbose |= VERBOSE_LAYOUT; + param->fp_max_depth = 0; + } + break; + case 'M': + if (!(param->verbose & VERBOSE_DETAIL)) + param->fp_max_depth = 0; + param->verbose |= VERBOSE_MDTINDEX; + break; + case 'R': + param->raw = 1; + break; + default: + return CMD_HELP; + } + } + + if (optind >= argc) + return CMD_HELP; + + if (param->recursive) + param->fp_max_depth = -1; + + if (!param->verbose) + param->verbose = VERBOSE_ALL; + if (param->quiet) + param->verbose = VERBOSE_OBJID; + + do { + rc = llapi_getstripe(argv[optind], param); + } while (++optind < argc && !rc); + + if (rc) + fprintf(stderr, "error: %s failed for %s.\n", + argv[0], argv[optind - 1]); + return rc; } static int lfs_tgts(int argc, char **argv) @@ -993,6 +1636,241 @@ static int lfs_tgts(int argc, char **argv) return rc; } +static int lfs_getstripe(int argc, char **argv) +{ + struct find_param param = { 0 }; + return lfs_getstripe_internal(argc, argv, ¶m); +} + +/* functions */ +static int lfs_getdirstripe(int argc, char **argv) +{ + struct find_param param = { 0 }; + + param.get_lmv = 1; + return lfs_getstripe_internal(argc, argv, ¶m); +} + +/* functions */ +static int lfs_setdirstripe(int argc, char **argv) +{ + char *dname; + int result; + unsigned int stripe_offset = -1; + unsigned int stripe_count = 1; + enum lmv_hash_type hash_type; + char *end; + int c; + char *stripe_offset_opt = NULL; + char *stripe_count_opt = NULL; + char *stripe_hash_opt = NULL; + char *mode_opt = NULL; + int default_stripe = 0; + mode_t mode = S_IRWXU | S_IRWXG | S_IRWXO; + mode_t previous_mode = 0; + + struct option long_opts[] = { + {"count", required_argument, 0, 'c'}, + {"index", required_argument, 0, 'i'}, + {"mode", required_argument, 0, 'm'}, + {"hash-type", required_argument, 0, 't'}, + {"default_stripe", no_argument, 0, 'D'}, + {0, 0, 0, 0} + }; + + optind = 0; + + while ((c = getopt_long(argc, argv, "c:Di:m:t:", long_opts, + NULL)) >= 0) { + switch (c) { + case 0: + /* Long options. */ + break; + case 'c': + stripe_count_opt = optarg; + break; + case 'D': + default_stripe = 1; + break; + case 'i': + stripe_offset_opt = optarg; + break; + case 'm': + mode_opt = optarg; + break; + case 't': + stripe_hash_opt = optarg; + break; + default: + fprintf(stderr, "error: %s: option '%s' " + "unrecognized\n", + argv[0], argv[optind - 1]); + return CMD_HELP; + } + } + + if (optind == argc) { + fprintf(stderr, "error: %s: missing dirname\n", + argv[0]); + return CMD_HELP; + } + + if (stripe_offset_opt == NULL && stripe_count_opt == NULL) { + fprintf(stderr, "error: %s: missing stripe offset and count.\n", + argv[0]); + return CMD_HELP; + } + + if (stripe_offset_opt != NULL) { + /* get the stripe offset */ + stripe_offset = strtoul(stripe_offset_opt, &end, 0); + if (*end != '\0') { + fprintf(stderr, "error: %s: bad stripe offset '%s'\n", + argv[0], stripe_offset_opt); + return CMD_HELP; + } + } + + if (mode_opt != NULL) { + mode = strtoul(mode_opt, &end, 8); + if (*end != '\0') { + fprintf(stderr, "error: %s: bad mode '%s'\n", + argv[0], mode_opt); + return CMD_HELP; + } + previous_mode = umask(0); + } + + if (stripe_hash_opt == NULL || + strcmp(stripe_hash_opt, LMV_HASH_NAME_FNV_1A_64) == 0) { + hash_type = LMV_HASH_TYPE_FNV_1A_64; + } else if (strcmp(stripe_hash_opt, LMV_HASH_NAME_ALL_CHARS) == 0) { + hash_type = LMV_HASH_TYPE_ALL_CHARS; + } else { + fprintf(stderr, "error: %s: bad stripe hash type '%s'\n", + argv[0], stripe_hash_opt); + return CMD_HELP; + } + + /* get the stripe count */ + if (stripe_count_opt != NULL) { + stripe_count = strtoul(stripe_count_opt, &end, 0); + if (*end != '\0') { + fprintf(stderr, "error: %s: bad stripe count '%s'\n", + argv[0], stripe_count_opt); + return CMD_HELP; + } + } + + dname = argv[optind]; + do { + if (default_stripe == 1) { + result = llapi_dir_set_default_lmv_stripe(dname, + stripe_offset, stripe_count, + hash_type, NULL); + } else { + result = llapi_dir_create_pool(dname, mode, + stripe_offset, + stripe_count, hash_type, + NULL); + } + + if (result) { + fprintf(stderr, "error: %s: create stripe dir '%s' " + "failed\n", argv[0], dname); + break; + } + dname = argv[++optind]; + } while (dname != NULL); + + if (mode_opt != NULL) + umask(previous_mode); + + return result; +} + +/* functions */ +static int lfs_rmentry(int argc, char **argv) +{ + char *dname; + int index; + int result = 0; + + if (argc <= 1) { + fprintf(stderr, "error: %s: missing dirname\n", + argv[0]); + return CMD_HELP; + } + + index = 1; + dname = argv[index]; + while (dname != NULL) { + result = llapi_direntry_remove(dname); + if (result) { + fprintf(stderr, "error: %s: remove dir entry '%s' " + "failed\n", argv[0], dname); + break; + } + dname = argv[++index]; + } + return result; +} + +static int lfs_mv(int argc, char **argv) +{ + struct find_param param = { + .fp_max_depth = -1, + .mdtindex = -1, + }; + char *end; + int c; + int rc = 0; + struct option long_opts[] = { + {"--mdt-index", required_argument, 0, 'M'}, + {"verbose", no_argument, 0, 'v'}, + {0, 0, 0, 0} + }; + + while ((c = getopt_long(argc, argv, "M:v", long_opts, NULL)) != -1) { + switch (c) { + case 'M': { + param.mdtindex = strtoul(optarg, &end, 0); + if (*end != '\0') { + fprintf(stderr, "%s: invalid MDT index'%s'\n", + argv[0], optarg); + return CMD_HELP; + } + break; + } + case 'v': { + param.verbose = VERBOSE_DETAIL; + break; + } + default: + fprintf(stderr, "error: %s: unrecognized option '%s'\n", + argv[0], argv[optind - 1]); + return CMD_HELP; + } + } + + if (param.mdtindex == -1) { + fprintf(stderr, "%s MDT index must be indicated\n", argv[0]); + return CMD_HELP; + } + + if (optind >= argc) { + fprintf(stderr, "%s missing operand path\n", argv[0]); + return CMD_HELP; + } + + param.migrate = 1; + rc = llapi_mv(argv[optind], ¶m); + if (rc != 0) + fprintf(stderr, "cannot migrate '%s' to MDT%04x: %s\n", + argv[optind], param.mdtindex, strerror(-rc)); + return rc; +} + static int lfs_osts(int argc, char **argv) { return lfs_tgts(argc, argv); @@ -1027,7 +1905,10 @@ static int showdf(char *mntdir, struct obd_statfs *stat, double ratio = 0; char *suffix = "KMGTPEZY"; /* Note if we have >2^64 bytes/fs these buffers will need to be grown */ - char tbuf[20], ubuf[20], abuf[20], rbuf[20]; + char tbuf[3 * sizeof(__u64)]; + char ubuf[3 * sizeof(__u64)]; + char abuf[3 * sizeof(__u64)]; + char rbuf[3 * sizeof(__u64)]; if (!uuid || !stat) return -EINVAL; @@ -1105,17 +1986,20 @@ struct ll_stat_type { char *st_name; }; -static int mntdf(char *mntdir, char *fsname, char *pool, int ishow, int cooked) +static int mntdf(char *mntdir, char *fsname, char *pool, int ishow, + int cooked, int lazy) { - struct obd_statfs stat_buf, sum = { .os_bsize = 1 }; - struct obd_uuid uuid_buf; - char *poolname = NULL; - struct ll_stat_type types[] = { { LL_STATFS_LMV, "MDT" }, - { LL_STATFS_LOV, "OST" }, - { 0, NULL } }; - struct ll_stat_type *tp; - __u32 index; - int rc; + struct obd_statfs stat_buf, sum = { .os_bsize = 1 }; + struct obd_uuid uuid_buf; + char *poolname = NULL; + struct ll_stat_type types[] = { { LL_STATFS_LMV, "MDT" }, + { LL_STATFS_LOV, "OST" }, + { 0, NULL } }; + struct ll_stat_type *tp; + __u64 ost_ffree = 0; + __u32 index; + __u32 type; + int rc; if (pool) { poolname = strchr(pool, '.'); @@ -1142,11 +2026,15 @@ static int mntdf(char *mntdir, char *fsname, char *pool, int ishow, int cooked) for (index = 0; ; index++) { memset(&stat_buf, 0, sizeof(struct obd_statfs)); memset(&uuid_buf, 0, sizeof(struct obd_uuid)); - rc = llapi_obd_statfs(mntdir, tp->st_op, index, + type = lazy ? tp->st_op | LL_STATFS_NODELAY : tp->st_op; + rc = llapi_obd_statfs(mntdir, type, index, &stat_buf, &uuid_buf); if (rc == -ENODEV) break; + if (rc == -EAGAIN) + continue; + if (poolname && tp->st_op == LL_STATFS_LOV && llapi_search_ost(fsname, poolname, obd_uuid2str(&uuid_buf)) != 1) @@ -1159,61 +2047,75 @@ static int mntdf(char *mntdir, char *fsname, char *pool, int ishow, int cooked) * it in so that we can print an error message. */ if (uuid_buf.uuid[0] == '\0') sprintf(uuid_buf.uuid, "%s%04x", - tp->st_name, index); - showdf(mntdir,&stat_buf,obd_uuid2str(&uuid_buf), - ishow, cooked, tp->st_name, index, rc); + tp->st_name, index); + showdf(mntdir, &stat_buf, obd_uuid2str(&uuid_buf), + ishow, cooked, tp->st_name, index, rc); if (rc == 0) { if (tp->st_op == LL_STATFS_LMV) { sum.os_ffree += stat_buf.os_ffree; sum.os_files += stat_buf.os_files; } else /* if (tp->st_op == LL_STATFS_LOV) */ { - sum.os_blocks += stat_buf.os_blocks * - stat_buf.os_bsize; - sum.os_bfree += stat_buf.os_bfree * - stat_buf.os_bsize; - sum.os_bavail += stat_buf.os_bavail * - stat_buf.os_bsize; - } - } else if (rc == -EINVAL || rc == -EFAULT) { - break; - } - } - } - - printf("\n"); - showdf(mntdir, &sum, "filesystem summary:", ishow, cooked, NULL, 0,0); - printf("\n"); - return 0; + sum.os_blocks += stat_buf.os_blocks * + stat_buf.os_bsize; + sum.os_bfree += stat_buf.os_bfree * + stat_buf.os_bsize; + sum.os_bavail += stat_buf.os_bavail * + stat_buf.os_bsize; + ost_ffree += stat_buf.os_ffree; + } + } else if (rc == -EINVAL || rc == -EFAULT) { + break; + } + } + } + + /* If we don't have as many objects free on the OST as inodes + * on the MDS, we reduce the total number of inodes to + * compensate, so that the "inodes in use" number is correct. + * Matches ll_statfs_internal() so the results are consistent. */ + if (ost_ffree < sum.os_ffree) { + sum.os_files = (sum.os_files - sum.os_ffree) + ost_ffree; + sum.os_ffree = ost_ffree; + } + printf("\n"); + showdf(mntdir, &sum, "filesystem summary:", ishow, cooked, NULL, 0, 0); + printf("\n"); + return 0; } static int lfs_df(int argc, char **argv) { char mntdir[PATH_MAX] = {'\0'}, path[PATH_MAX] = {'\0'}; int ishow = 0, cooked = 0; + int lazy = 0; int c, rc = 0, index = 0; char fsname[PATH_MAX] = "", *pool_name = NULL; struct option long_opts[] = { {"pool", required_argument, 0, 'p'}, + {"lazy", 0, 0, 'l'}, {0, 0, 0, 0} }; - optind = 0; - while ((c = getopt_long(argc, argv, "hip:", long_opts, NULL)) != -1) { - switch (c) { - case 'i': - ishow = 1; - break; - case 'h': - cooked = 1; - break; - case 'p': - pool_name = optarg; - break; - default: - return CMD_HELP; - } - } + optind = 0; + while ((c = getopt_long(argc, argv, "hilp:", long_opts, NULL)) != -1) { + switch (c) { + case 'i': + ishow = 1; + break; + case 'h': + cooked = 1; + break; + case 'l': + lazy = 1; + break; + case 'p': + pool_name = optarg; + break; + default: + return CMD_HELP; + } + } if (optind < argc && !realpath(argv[optind], path)) { rc = -errno; fprintf(stderr, "error: invalid path '%s': %s\n", @@ -1226,7 +2128,7 @@ static int lfs_df(int argc, char **argv) if (mntdir[0] == '\0') continue; - rc = mntdf(mntdir, fsname, pool_name, ishow, cooked); + rc = mntdf(mntdir, fsname, pool_name, ishow, cooked, lazy); if (rc || path[0] != '\0') break; fsname[0] = '\0'; /* avoid matching in next loop */ @@ -1311,9 +2213,7 @@ static int lfs_check(int argc, char **argv) return rc; } - rc = llapi_target_iterate(num_types, obd_types, - mntdir, llapi_ping_target); - + rc = llapi_target_check(num_types, obd_types, mntdir); if (rc) fprintf(stderr, "error: %s: %s status failed\n", argv[0],argv[1]); @@ -1322,31 +2222,6 @@ static int lfs_check(int argc, char **argv) } -static int lfs_catinfo(int argc, char **argv) -{ - char mntdir[PATH_MAX] = {'\0'}; - int rc; - - if (argc < 2 || (!strcmp(argv[1],"config") && argc < 3)) - return CMD_HELP; - - if (strcmp(argv[1], "config") && strcmp(argv[1], "deletions")) - return CMD_HELP; - - rc = llapi_search_mounts(NULL, 0, mntdir, NULL); - if (rc == 0 && mntdir[0] != '\0') { - if (argc == 3) - rc = llapi_catinfo(mntdir, argv[1], argv[2]); - else - rc = llapi_catinfo(mntdir, argv[1], NULL); - } else { - fprintf(stderr, "no lustre_lite mounted.\n"); - rc = -1; - } - - return rc; -} - static int lfs_join(int argc, char **argv) { fprintf(stderr, "join two lustre files into one.\n" @@ -1355,32 +2230,6 @@ static int lfs_join(int argc, char **argv) } #ifdef HAVE_SYS_QUOTA_H -static int lfs_quotachown(int argc, char **argv) -{ - - int c,rc; - int flag = 0; - - optind = 0; - while ((c = getopt(argc, argv, "i")) != -1) { - switch (c) { - case 'i': - flag++; - break; - default: - fprintf(stderr, "error: %s: option '-%c' " - "unrecognized\n", argv[0], c); - return CMD_HELP; - } - } - if (optind == argc) - return CMD_HELP; - rc = llapi_quotachown(argv[optind], flag); - if(rc) - fprintf(stderr,"error: change file owner/group failed.\n"); - return rc; -} - static int lfs_quotacheck(int argc, char **argv) { int c, check_type = 0; @@ -1419,7 +2268,14 @@ static int lfs_quotacheck(int argc, char **argv) mnt = argv[optind]; rc = llapi_quotacheck(mnt, check_type); - if (rc) { + if (rc == -EOPNOTSUPP) { + fprintf(stderr, "error: quotacheck not supported by the quota " + "master.\nPlease note that quotacheck is deprecated as " + "of lustre 2.4.0 since space accounting is always " + "enabled.\nFilesystems not formatted with 2.4 utils or " + "beyond can be upgraded with tunefs.lustre --quota.\n"); + return rc; + } else if (rc) { fprintf(stderr, "quotacheck failed: %s\n", strerror(-rc)); return rc; } @@ -1491,7 +2347,16 @@ static int lfs_quotaon(int argc, char **argv) rc = llapi_quotactl(mnt, &qctl); if (rc) { - if (rc == -EALREADY) { + if (rc == -EOPNOTSUPP) { + fprintf(stderr, "error: quotaon not supported by the " + "quota master.\nPlease note that quotaon/off is" + " deprecated as of lustre 2.4.0.\nQuota " + "enforcement should now be enabled on the MGS " + "via:\nmgs# lctl conf_param ${FSNAME}.quota." + "=\n(ost for block quota, mdt " + "for inode quota, u for user and g for group" + "\n"); + } else if (rc == -EALREADY) { rc = 0; } else if (rc == -ENOENT) { fprintf(stderr, "error: cannot find quota database, " @@ -1547,7 +2412,14 @@ static int lfs_quotaoff(int argc, char **argv) rc = llapi_quotactl(mnt, &qctl); if (rc) { - if (rc == -EALREADY) { + if (rc == -EOPNOTSUPP) { + fprintf(stderr, "error: quotaoff not supported by the " + "quota master.\nPlease note that quotaon/off is" + " deprecated as of lustre 2.4.0.\nQuota " + "enforcement can be disabled on the MGS via:\n" + "mgs# lctl conf_param ${FSNAME}.quota." + "=\"\"\n"); + } else if (rc == -EALREADY) { rc = 0; } else { if (*obd_type) @@ -1561,54 +2433,6 @@ static int lfs_quotaoff(int argc, char **argv) return rc; } -static int lfs_quotainv(int argc, char **argv) -{ - int c; - char *mnt; - struct if_quotactl qctl; - int rc; - - memset(&qctl, 0, sizeof(qctl)); - qctl.qc_cmd = LUSTRE_Q_INVALIDATE; - - optind = 0; - while ((c = getopt(argc, argv, "fgu")) != -1) { - switch (c) { - case 'u': - qctl.qc_type |= 0x01; - break; - case 'g': - qctl.qc_type |= 0x02; - break; - case 'f': - qctl.qc_cmd = LUSTRE_Q_FINVALIDATE; - break; - default: - fprintf(stderr, "error: %s: option '-%c' " - "unrecognized\n", argv[0], c); - return CMD_HELP; - } - } - - if (qctl.qc_type) - qctl.qc_type--; - else /* by default, invalidate quota for both user & group */ - qctl.qc_type = 0x02; - - if (argc == optind) - return CMD_HELP; - - mnt = argv[optind]; - - rc = llapi_quotactl(mnt, &qctl); - if (rc) { - fprintf(stderr, "quotainv failed: %s\n", strerror(-rc)); - return rc; - } - - return 0; -} - #define ARG2INT(nr, str, msg) \ do { \ char *endp; \ @@ -1678,17 +2502,17 @@ error: return ULONG_MAX; } -#define ARG2ULL(nr, str, def_units) \ -do { \ - unsigned long long limit, units = def_units; \ - int rc; \ - \ - rc = parse_size(str, &limit, &units, 1); \ - if (rc < 0) { \ - fprintf(stderr, "error: bad limit value %s\n", str); \ - return CMD_HELP; \ - } \ - nr = limit; \ +#define ARG2ULL(nr, str, def_units) \ +do { \ + unsigned long long limit, units = def_units; \ + int rc; \ + \ + rc = llapi_parse_size(str, &limit, &units, 1); \ + if (rc < 0) { \ + fprintf(stderr, "error: bad limit value %s\n", str); \ + return CMD_HELP; \ + } \ + nr = limit; \ } while (0) static inline int has_times_option(int argc, char **argv) @@ -1838,19 +2662,43 @@ int lfs_setquota(int argc, char **argv) ARG2ULL(dqb->dqb_bsoftlimit, optarg, 1024); dqb->dqb_bsoftlimit >>= 10; limit_mask |= BSLIMIT; + if (dqb->dqb_bsoftlimit && + dqb->dqb_bsoftlimit <= 1024) /* <= 1M? */ + fprintf(stderr, "warning: block softlimit is " + "smaller than the miminal qunit size, " + "please see the help of setquota or " + "Lustre manual for details.\n"); break; case 'B': ARG2ULL(dqb->dqb_bhardlimit, optarg, 1024); dqb->dqb_bhardlimit >>= 10; limit_mask |= BHLIMIT; + if (dqb->dqb_bhardlimit && + dqb->dqb_bhardlimit <= 1024) /* <= 1M? */ + fprintf(stderr, "warning: block hardlimit is " + "smaller than the miminal qunit size, " + "please see the help of setquota or " + "Lustre manual for details.\n"); break; case 'i': ARG2ULL(dqb->dqb_isoftlimit, optarg, 1); limit_mask |= ISLIMIT; + if (dqb->dqb_isoftlimit && + dqb->dqb_isoftlimit <= 1024) /* <= 1K inodes? */ + fprintf(stderr, "warning: inode softlimit is " + "smaller than the miminal qunit size, " + "please see the help of setquota or " + "Lustre manual for details.\n"); break; case 'I': ARG2ULL(dqb->dqb_ihardlimit, optarg, 1); limit_mask |= IHLIMIT; + if (dqb->dqb_ihardlimit && + dqb->dqb_ihardlimit <= 1024) /* <= 1K inodes? */ + fprintf(stderr, "warning: inode hardlimit is " + "smaller than the miminal qunit size, " + "please see the help of setquota or " + "Lustre manual for details.\n"); break; default: /* getopt prints error message for us when opterr != 0 */ return CMD_HELP; @@ -1990,29 +2838,50 @@ static void diff2str(time_t seconds, char *buf, time_t now) __sec2str(seconds - now, buf); } -static void print_quota_title(char *name, struct if_quotactl *qctl) +static void print_quota_title(char *name, struct if_quotactl *qctl, + bool human_readable) +{ + printf("Disk quotas for %s %s (%cid %u):\n", + type2name(qctl->qc_type), name, + *type2name(qctl->qc_type), qctl->qc_id); + printf("%15s%8s %7s%8s%8s%8s %7s%8s%8s\n", + "Filesystem", human_readable ? "used" : "kbytes", + "quota", "limit", "grace", + "files", "quota", "limit", "grace"); +} + +static void kbytes2str(__u64 num, char *buf, bool h) { - printf("Disk quotas for %s %s (%cid %u):\n", - type2name(qctl->qc_type), name, - *type2name(qctl->qc_type), qctl->qc_id); - printf("%15s%8s %7s%8s%8s%8s %7s%8s%8s\n", - "Filesystem", - "kbytes", "quota", "limit", "grace", - "files", "quota", "limit", "grace"); + if (!h) { + sprintf(buf, LPU64, num); + } else { + if (num >> 30) + sprintf(buf, "%5.4gT", (double)num / (1 << 30)); + else if (num >> 20) + sprintf(buf, "%5.4gG", (double)num / (1 << 20)); + else if (num >> 10) + sprintf(buf, "%5.4gM", (double)num / (1 << 10)); + else + sprintf(buf, LPU64"%s", num, "k"); + } } -static void print_quota(char *mnt, struct if_quotactl *qctl, int type, int rc) +static void print_quota(char *mnt, struct if_quotactl *qctl, int type, + int rc, bool h) { time_t now; time(&now); if (qctl->qc_cmd == LUSTRE_Q_GETQUOTA || qctl->qc_cmd == Q_GETOQUOTA) { - int bover = 0, iover = 0; - struct obd_dqblk *dqb = &qctl->qc_dqblk; + int bover = 0, iover = 0; + struct obd_dqblk *dqb = &qctl->qc_dqblk; + char numbuf[3][32]; + char timebuf[40]; + char strbuf[32]; if (dqb->dqb_bhardlimit && - toqb(dqb->dqb_curspace) >= dqb->dqb_bhardlimit) { + lustre_stoqb(dqb->dqb_curspace) >= dqb->dqb_bhardlimit) { bover = 1; } else if (dqb->dqb_bsoftlimit && dqb->dqb_btime) { if (dqb->dqb_btime > now) { @@ -2026,68 +2895,67 @@ static void print_quota(char *mnt, struct if_quotactl *qctl, int type, int rc) dqb->dqb_curinodes >= dqb->dqb_ihardlimit) { iover = 1; } else if (dqb->dqb_isoftlimit && dqb->dqb_itime) { - if (dqb->dqb_btime > now) { - iover = 2; - } else { - iover = 3; - } + if (dqb->dqb_itime > now) { + iover = 2; + } else { + iover = 3; + } } -#if 0 /* XXX: always print quotas even when no usages */ - if (dqb->dqb_curspace || dqb->dqb_curinodes) -#endif - { - char numbuf[3][32]; - char timebuf[40]; - if (strlen(mnt) > 15) - printf("%s\n%15s", mnt, ""); - else - printf("%15s", mnt); + if (strlen(mnt) > 15) + printf("%s\n%15s", mnt, ""); + else + printf("%15s", mnt); + + if (bover) + diff2str(dqb->dqb_btime, timebuf, now); + + kbytes2str(lustre_stoqb(dqb->dqb_curspace), strbuf, h); + if (rc == -EREMOTEIO) + sprintf(numbuf[0], "%s*", strbuf); + else + sprintf(numbuf[0], (dqb->dqb_valid & QIF_SPACE) ? + "%s" : "[%s]", strbuf); + + kbytes2str(dqb->dqb_bsoftlimit, strbuf, h); + if (type == QC_GENERAL) + sprintf(numbuf[1], (dqb->dqb_valid & QIF_BLIMITS) ? + "%s" : "[%s]", strbuf); + else + sprintf(numbuf[1], "%s", "-"); + + kbytes2str(dqb->dqb_bhardlimit, strbuf, h); + sprintf(numbuf[2], (dqb->dqb_valid & QIF_BLIMITS) ? + "%s" : "[%s]", strbuf); + + printf(" %7s%c %6s %7s %7s", + numbuf[0], bover ? '*' : ' ', numbuf[1], + numbuf[2], bover > 1 ? timebuf : "-"); + + if (iover) + diff2str(dqb->dqb_itime, timebuf, now); + + sprintf(numbuf[0], (dqb->dqb_valid & QIF_INODES) ? + LPU64 : "["LPU64"]", dqb->dqb_curinodes); + + if (type == QC_GENERAL) + sprintf(numbuf[1], (dqb->dqb_valid & QIF_ILIMITS) ? + LPU64 : "["LPU64"]", dqb->dqb_isoftlimit); + else + sprintf(numbuf[1], "%s", "-"); + + sprintf(numbuf[2], (dqb->dqb_valid & QIF_ILIMITS) ? + LPU64 : "["LPU64"]", dqb->dqb_ihardlimit); + + if (type != QC_OSTIDX) + printf(" %7s%c %6s %7s %7s", + numbuf[0], iover ? '*' : ' ', numbuf[1], + numbuf[2], iover > 1 ? timebuf : "-"); + else + printf(" %7s %7s %7s %7s", "-", "-", "-", "-"); + printf("\n"); - if (bover) - diff2str(dqb->dqb_btime, timebuf, now); - if (rc == -EREMOTEIO) - sprintf(numbuf[0], LPU64"*", - toqb(dqb->dqb_curspace)); - else - sprintf(numbuf[0], - (dqb->dqb_valid & QIF_SPACE) ? - LPU64 : "["LPU64"]", - toqb(dqb->dqb_curspace)); - if (type == QC_GENERAL) - sprintf(numbuf[1], (dqb->dqb_valid & QIF_BLIMITS) - ? LPU64 : "["LPU64"]", - dqb->dqb_bsoftlimit); - else - sprintf(numbuf[1], "%s", "-"); - sprintf(numbuf[2], (dqb->dqb_valid & QIF_BLIMITS) - ? LPU64 : "["LPU64"]", dqb->dqb_bhardlimit); - printf(" %7s%c %6s %7s %7s", - numbuf[0], bover ? '*' : ' ', numbuf[1], - numbuf[2], bover > 1 ? timebuf : "-"); - - if (iover) - diff2str(dqb->dqb_itime, timebuf, now); - - sprintf(numbuf[0], (dqb->dqb_valid & QIF_INODES) ? - LPU64 : "["LPU64"]", dqb->dqb_curinodes); - if (type == QC_GENERAL) - sprintf(numbuf[1], (dqb->dqb_valid & QIF_ILIMITS) - ? LPU64 : "["LPU64"]", - dqb->dqb_isoftlimit); - else - sprintf(numbuf[1], "%s", "-"); - sprintf(numbuf[2], (dqb->dqb_valid & QIF_ILIMITS) ? - LPU64 : "["LPU64"]", dqb->dqb_ihardlimit); - if (type != QC_OSTIDX) - printf(" %7s%c %6s %7s %7s", - numbuf[0], iover ? '*' : ' ', numbuf[1], - numbuf[2], iover > 1 ? timebuf : "-"); - else - printf(" %7s %7s %7s %7s", "-", "-", "-", "-"); - printf("\n"); - } } else if (qctl->qc_cmd == LUSTRE_Q_GETINFO || qctl->qc_cmd == Q_GETOINFO) { char bgtimebuf[40]; @@ -2100,7 +2968,8 @@ static void print_quota(char *mnt, struct if_quotactl *qctl, int type, int rc) } } -static int print_obd_quota(char *mnt, struct if_quotactl *qctl, int is_mdt) +static int print_obd_quota(char *mnt, struct if_quotactl *qctl, int is_mdt, + bool h, __u64 *total) { int rc = 0, rc1 = 0, count = 0; __u32 valid = qctl->qc_valid; @@ -2129,29 +2998,33 @@ static int print_obd_quota(char *mnt, struct if_quotactl *qctl, int is_mdt) continue; } - print_quota(obd_uuid2str(&qctl->obd_uuid), qctl, qctl->qc_valid, 0); - } - + print_quota(obd_uuid2str(&qctl->obd_uuid), qctl, + qctl->qc_valid, 0, h); + *total += is_mdt ? qctl->qc_dqblk.dqb_ihardlimit : + qctl->qc_dqblk.dqb_bhardlimit; + } out: - qctl->qc_valid = valid; - return rc ? : rc1; + qctl->qc_valid = valid; + return rc ? : rc1; } static int lfs_quota(int argc, char **argv) { - int c; - char *mnt, *name = NULL; - struct if_quotactl qctl = { .qc_cmd = LUSTRE_Q_GETQUOTA, - .qc_type = UGQUOTA }; - char *obd_type = (char *)qctl.obd_type; - char *obd_uuid = (char *)qctl.obd_uuid.uuid; - int rc, rc1 = 0, rc2 = 0, rc3 = 0, - verbose = 0, pass = 0, quiet = 0, inacc; - char *endptr; - __u32 valid = QC_GENERAL, idx = 0; - - optind = 0; - while ((c = getopt(argc, argv, "gi:I:o:qtuv")) != -1) { + int c; + char *mnt, *name = NULL; + struct if_quotactl qctl = { .qc_cmd = LUSTRE_Q_GETQUOTA, + .qc_type = UGQUOTA }; + char *obd_type = (char *)qctl.obd_type; + char *obd_uuid = (char *)qctl.obd_uuid.uuid; + int rc, rc1 = 0, rc2 = 0, rc3 = 0, + verbose = 0, pass = 0, quiet = 0, inacc; + char *endptr; + __u32 valid = QC_GENERAL, idx = 0; + __u64 total_ialloc = 0, total_balloc = 0; + bool human_readable = false; + + optind = 0; + while ((c = getopt(argc, argv, "gi:I:o:qtuvh")) != -1) { switch (c) { case 'u': if (qctl.qc_type != UGQUOTA) { @@ -2172,7 +3045,7 @@ static int lfs_quota(int argc, char **argv) break; case 'o': valid = qctl.qc_valid = QC_UUID; - strncpy(obd_uuid, optarg, sizeof(qctl.obd_uuid)); + strlcpy(obd_uuid, optarg, sizeof(qctl.obd_uuid)); break; case 'i': valid = qctl.qc_valid = QC_MDTIDX; @@ -2188,6 +3061,9 @@ static int lfs_quota(int argc, char **argv) case 'q': quiet = 1; break; + case 'h': + human_readable = true; + break; default: fprintf(stderr, "error: %s: option '-%c' " "unrecognized\n", argv[0], c); @@ -2258,8 +3134,8 @@ ug_output: } } - if (qctl.qc_cmd == LUSTRE_Q_GETQUOTA && !quiet) - print_quota_title(name, &qctl); + if (qctl.qc_cmd == LUSTRE_Q_GETQUOTA && !quiet) + print_quota_title(name, &qctl, human_readable); if (rc1 && *obd_type) fprintf(stderr, "%s %s ", obd_type, obd_uuid); @@ -2267,15 +3143,24 @@ ug_output: if (qctl.qc_valid != QC_GENERAL) mnt = ""; - inacc = (qctl.qc_cmd == LUSTRE_Q_GETQUOTA) && - ((qctl.qc_dqblk.dqb_valid&(QIF_LIMITS|QIF_USAGE))!=(QIF_LIMITS|QIF_USAGE)); + inacc = (qctl.qc_cmd == LUSTRE_Q_GETQUOTA) && + ((qctl.qc_dqblk.dqb_valid & (QIF_LIMITS|QIF_USAGE)) != + (QIF_LIMITS|QIF_USAGE)); - print_quota(mnt, &qctl, QC_GENERAL, rc1); + print_quota(mnt, &qctl, QC_GENERAL, rc1, human_readable); - if (qctl.qc_valid == QC_GENERAL && qctl.qc_cmd != LUSTRE_Q_GETINFO && verbose) { - rc2 = print_obd_quota(mnt, &qctl, 1); - rc3 = print_obd_quota(mnt, &qctl, 0); - } + if (qctl.qc_valid == QC_GENERAL && qctl.qc_cmd != LUSTRE_Q_GETINFO && + verbose) { + char strbuf[32]; + + rc2 = print_obd_quota(mnt, &qctl, 1, human_readable, + &total_ialloc); + rc3 = print_obd_quota(mnt, &qctl, 0, human_readable, + &total_balloc); + kbytes2str(total_balloc, strbuf, human_readable); + printf("Total allocated inode limit: "LPU64", total " + "allocated block limit: %s\n", total_ialloc, strbuf); + } if (rc1 || rc2 || rc3 || inacc) printf("Some errors happened when getting quota info. " @@ -2312,10 +3197,10 @@ static int flushctx_ioctl(char *mp) static int lfs_flushctx(int argc, char **argv) { - int kdestroy = 0, c; - FILE *proc; - char procline[PATH_MAX], *line; - int rc = 0; + int kdestroy = 0, c; + char mntdir[PATH_MAX] = {'\0'}; + int index = 0; + int rc = 0; optind = 0; while ((c = getopt(argc, argv, "k")) != -1) { @@ -2331,45 +3216,24 @@ static int lfs_flushctx(int argc, char **argv) } if (kdestroy) { - int rc; if ((rc = system("kdestroy > /dev/null")) != 0) { rc = WEXITSTATUS(rc); fprintf(stderr, "error destroying tickets: %d, continuing\n", rc); } } - if (optind >= argc) { - /* flush for all mounted lustre fs. */ - proc = fopen("/proc/mounts", "r"); - if (!proc) { - fprintf(stderr, "error: %s: can't open /proc/mounts\n", - argv[0]); - return -1; - } - - while ((line = fgets(procline, PATH_MAX, proc)) != NULL) { - char dev[PATH_MAX]; - char mp[PATH_MAX]; - char fs[PATH_MAX]; - - if (sscanf(line, "%s %s %s", dev, mp, fs) != 3) { - fprintf(stderr, "%s: unexpected format in " - "/proc/mounts\n", - argv[0]); - return -1; - } + if (optind >= argc) { + /* flush for all mounted lustre fs. */ + while (!llapi_search_mounts(NULL, index++, mntdir, NULL)) { + /* Check if we have a mount point */ + if (mntdir[0] == '\0') + continue; - if (strcmp(fs, "lustre") != 0) - continue; - /* we use '@' to determine it's a client. are there - * any other better way? - */ - if (strchr(dev, '@') == NULL) - continue; + if (flushctx_ioctl(mntdir)) + rc = -1; - if (flushctx_ioctl(mp)) - rc = -1; - } + mntdir[0] = '\0'; /* avoid matching in next loop */ + } } else { /* flush fs as specified */ while (optind < argc) { @@ -2377,7 +3241,6 @@ static int lfs_flushctx(int argc, char **argv) rc = -1; } } - return rc; } @@ -2418,7 +3281,7 @@ static int lfs_ls(int argc, char **argv) static int lfs_changelog(int argc, char **argv) { void *changelog_priv; - struct changelog_rec *rec; + struct changelog_rec *rec; long long startrec = 0, endrec = 0; char *mdd; struct option long_opts[] = { @@ -2452,15 +3315,16 @@ static int lfs_changelog(int argc, char **argv) if (argc > optind) endrec = strtoll(argv[optind++], NULL, 10); - rc = llapi_changelog_start(&changelog_priv, - CHANGELOG_FLAG_BLOCK | - (follow ? CHANGELOG_FLAG_FOLLOW : 0), - mdd, startrec); - if (rc < 0) { - fprintf(stderr, "Can't start changelog: %s\n", - strerror(errno = -rc)); - return rc; - } + rc = llapi_changelog_start(&changelog_priv, + CHANGELOG_FLAG_BLOCK | + CHANGELOG_FLAG_JOBID | + (follow ? CHANGELOG_FLAG_FOLLOW : 0), + mdd, startrec); + if (rc < 0) { + fprintf(stderr, "Can't start changelog: %s\n", + strerror(errno = -rc)); + return rc; + } while ((rc = llapi_changelog_recv(changelog_priv, &rec)) == 0) { time_t secs; @@ -2475,23 +3339,42 @@ static int lfs_changelog(int argc, char **argv) continue; } - secs = rec->cr_time >> 30; - gmtime_r(&secs, &ts); - printf(LPU64" %02d%-5s %02d:%02d:%02d.%06d %04d.%02d.%02d " - "0x%x t="DFID, rec->cr_index, rec->cr_type, - changelog_type2str(rec->cr_type), - ts.tm_hour, ts.tm_min, ts.tm_sec, - (int)(rec->cr_time & ((1<<30) - 1)), - ts.tm_year+1900, ts.tm_mon+1, ts.tm_mday, - rec->cr_flags & CLF_FLAGMASK, PFID(&rec->cr_tfid)); - if (rec->cr_namelen) - /* namespace rec includes parent and filename */ - printf(" p="DFID" %.*s\n", PFID(&rec->cr_pfid), - rec->cr_namelen, rec->cr_name); - else - printf("\n"); - - llapi_changelog_free(&rec); + secs = rec->cr_time >> 30; + gmtime_r(&secs, &ts); + printf(LPU64" %02d%-5s %02d:%02d:%02d.%06d %04d.%02d.%02d " + "0x%x t="DFID, rec->cr_index, rec->cr_type, + changelog_type2str(rec->cr_type), + ts.tm_hour, ts.tm_min, ts.tm_sec, + (int)(rec->cr_time & ((1<<30) - 1)), + ts.tm_year + 1900, ts.tm_mon + 1, ts.tm_mday, + rec->cr_flags & CLF_FLAGMASK, PFID(&rec->cr_tfid)); + + if (rec->cr_flags & CLF_JOBID) { + struct changelog_ext_jobid *jid = + changelog_rec_jobid(rec); + + if (jid->cr_jobid[0] != '\0') + printf(" j=%s", jid->cr_jobid); + } + + if (rec->cr_namelen) + printf(" p="DFID" %.*s", PFID(&rec->cr_pfid), + rec->cr_namelen, changelog_rec_name(rec)); + + if (rec->cr_flags & CLF_RENAME) { + struct changelog_ext_rename *rnm = + changelog_rec_rename(rec); + + if (!fid_is_zero(&rnm->cr_sfid)) + printf(" s="DFID" sp="DFID" %.*s", + PFID(&rnm->cr_sfid), + PFID(&rnm->cr_spfid), + (int)changelog_rec_snamelen(rec), + changelog_rec_sname(rec)); + } + printf("\n"); + + llapi_changelog_free(&rec); } llapi_changelog_fini(&changelog_priv); @@ -2533,7 +3416,7 @@ static int lfs_fid2path(int argc, char **argv) int linkno = -1; int lnktmp; int printcur = 0; - int rc; + int rc = 0; optind = 0; @@ -2557,68 +3440,644 @@ static int lfs_fid2path(int argc, char **argv) return CMD_HELP; } } - device = argv[optind++]; - fid = argv[optind++]; - if (optind != argc) - return CMD_HELP; - path = calloc(1, PATH_MAX); + if (argc < 3) + return CMD_HELP; + + device = argv[optind++]; + path = calloc(1, PATH_MAX); + if (path == NULL) { + fprintf(stderr, "error: Not enough memory\n"); + return -errno; + } + + rc = 0; + while (optind < argc) { + fid = argv[optind++]; + + lnktmp = (linkno >= 0) ? linkno : 0; + while (1) { + int oldtmp = lnktmp; + long long rectmp = recno; + int rc2; + rc2 = llapi_fid2path(device, fid, path, PATH_MAX, + &rectmp, &lnktmp); + if (rc2 < 0) { + fprintf(stderr, "%s: error on FID %s: %s\n", + argv[0], fid, strerror(errno = -rc2)); + if (rc == 0) + rc = rc2; + break; + } + + if (printcur) + fprintf(stdout, "%lld ", rectmp); + if (device[0] == '/') { + fprintf(stdout, "%s", device); + if (device[strlen(device) - 1] != '/') + fprintf(stdout, "/"); + } else if (path[0] == '\0') { + fprintf(stdout, "/"); + } + fprintf(stdout, "%s\n", path); + + if (linkno >= 0) + /* specified linkno */ + break; + if (oldtmp == lnktmp) + /* no more links */ + break; + } + } + + free(path); + return rc; +} - lnktmp = (linkno >= 0) ? linkno : 0; - while (1) { - int oldtmp = lnktmp; - long long rectmp = recno; - rc = llapi_fid2path(device, fid, path, PATH_MAX, &rectmp, - &lnktmp); - if (rc < 0) { - fprintf(stderr, "%s error: %s\n", argv[0], - strerror(errno = -rc)); - break; - } +static int lfs_path2fid(int argc, char **argv) +{ + struct option long_opts[] = { + {"parents", no_argument, 0, 'p'}, + {0, 0, 0, 0} + }; + char **path; + const char short_opts[] = "p"; + const char *sep = ""; + lustre_fid fid; + int rc = 0; + bool show_parents = false; + + optind = 0; + while ((rc = getopt_long(argc, argv, short_opts, + long_opts, NULL)) != -1) { + switch (rc) { + case 'p': + show_parents = true; + break; + default: + fprintf(stderr, "error: %s: option '%s' unrecognized\n", + argv[0], argv[optind - 1]); + return CMD_HELP; + } + } + + if (optind > argc - 1) + return CMD_HELP; + else if (optind < argc - 1) + sep = ": "; + + rc = 0; + for (path = argv + optind; *path != NULL; path++) { + int err = 0; + if (!show_parents) { + err = llapi_path2fid(*path, &fid); + if (!err) + printf("%s%s"DFID"\n", + *sep != '\0' ? *path : "", sep, + PFID(&fid)); + } else { + char name[NAME_MAX + 1]; + unsigned int linkno = 0; + + while ((err = llapi_path2parent(*path, linkno, &fid, + name, sizeof(name))) == 0) { + if (*sep != '\0' && linkno == 0) + printf("%s%s", *path, sep); + + printf("%s"DFID"/%s", linkno != 0 ? "\t" : "", + PFID(&fid), name); + linkno++; + } + + /* err == -ENODATA is end-of-loop */ + if (linkno > 0 && err == -ENODATA) { + printf("\n"); + err = 0; + } + } + + if (err) { + fprintf(stderr, "%s: can't get %sfid for %s: %s\n", + argv[0], show_parents ? "parent " : "", *path, + strerror(-err)); + if (rc == 0) { + rc = err; + errno = -err; + } + } + } + + return rc; +} - if (printcur) - fprintf(stdout, "%lld ", rectmp); - if (device[0] == '/') { - fprintf(stdout, "%s", device); - if (device[strlen(device) - 1] != '/') - fprintf(stdout, "/"); - } else if (path[0] == '\0') { - fprintf(stdout, "/"); - } - fprintf(stdout, "%s\n", path); +static int lfs_data_version(int argc, char **argv) +{ + char *path; + __u64 data_version; + int fd; + int rc; + int c; + int data_version_flags = LL_DV_RD_FLUSH; /* Read by default */ + + if (argc < 2) + return CMD_HELP; + + optind = 0; + while ((c = getopt(argc, argv, "nrw")) != -1) { + switch (c) { + case 'n': + data_version_flags = 0; + break; + case 'r': + data_version_flags |= LL_DV_RD_FLUSH; + break; + case 'w': + data_version_flags |= LL_DV_WR_FLUSH; + break; + default: + return CMD_HELP; + } + } + if (optind == argc) + return CMD_HELP; + + path = argv[optind]; + fd = open(path, O_RDONLY); + if (fd < 0) + err(errno, "cannot open file %s", path); + + rc = llapi_get_data_version(fd, &data_version, data_version_flags); + if (rc < 0) + err(errno, "cannot get version for %s", path); + else + printf(LPU64 "\n", data_version); + + close(fd); + return rc; +} - if (linkno >= 0) - /* specified linkno */ - break; - if (oldtmp == lnktmp) - /* no more links */ - break; - } +static int lfs_hsm_state(int argc, char **argv) +{ + int rc; + int i = 1; + char *path; + struct hsm_user_state hus; + + if (argc < 2) + return CMD_HELP; + + do { + path = argv[i]; + + rc = llapi_hsm_state_get(path, &hus); + if (rc) { + fprintf(stderr, "can't get hsm state for %s: %s\n", + path, strerror(errno = -rc)); + return rc; + } + + /* Display path name and status flags */ + printf("%s: (0x%08x)", path, hus.hus_states); + + if (hus.hus_states & HS_RELEASED) + printf(" released"); + if (hus.hus_states & HS_EXISTS) + printf(" exists"); + if (hus.hus_states & HS_DIRTY) + printf(" dirty"); + if (hus.hus_states & HS_ARCHIVED) + printf(" archived"); + /* Display user-settable flags */ + if (hus.hus_states & HS_NORELEASE) + printf(" never_release"); + if (hus.hus_states & HS_NOARCHIVE) + printf(" never_archive"); + if (hus.hus_states & HS_LOST) + printf(" lost_from_hsm"); + + if (hus.hus_archive_id != 0) + printf(", archive_id:%d", hus.hus_archive_id); + printf("\n"); + + } while (++i < argc); + + return 0; +} - free(path); - return rc; +#define LFS_HSM_SET 0 +#define LFS_HSM_CLEAR 1 + +/** + * Generic function to set or clear HSM flags. + * Used by hsm_set and hsm_clear. + * + * @mode if LFS_HSM_SET, set the flags, if LFS_HSM_CLEAR, clear the flags. + */ +static int lfs_hsm_change_flags(int argc, char **argv, int mode) +{ + struct option long_opts[] = { + {"lost", 0, 0, 'l'}, + {"norelease", 0, 0, 'r'}, + {"noarchive", 0, 0, 'a'}, + {"archived", 0, 0, 'A'}, + {"dirty", 0, 0, 'd'}, + {"exists", 0, 0, 'e'}, + {0, 0, 0, 0} + }; + char short_opts[] = "lraAde"; + __u64 mask = 0; + int c, rc; + char *path; + + if (argc < 3) + return CMD_HELP; + + optind = 0; + while ((c = getopt_long(argc, argv, short_opts, + long_opts, NULL)) != -1) { + switch (c) { + case 'l': + mask |= HS_LOST; + break; + case 'a': + mask |= HS_NOARCHIVE; + break; + case 'A': + mask |= HS_ARCHIVED; + break; + case 'r': + mask |= HS_NORELEASE; + break; + case 'd': + mask |= HS_DIRTY; + break; + case 'e': + mask |= HS_EXISTS; + break; + case '?': + return CMD_HELP; + default: + fprintf(stderr, "error: %s: option '%s' unrecognized\n", + argv[0], argv[optind - 1]); + return CMD_HELP; + } + } + + /* User should have specified a flag */ + if (mask == 0) + return CMD_HELP; + + while (optind < argc) { + + path = argv[optind]; + + /* If mode == 0, this means we apply the mask. */ + if (mode == LFS_HSM_SET) + rc = llapi_hsm_state_set(path, mask, 0, 0); + else + rc = llapi_hsm_state_set(path, 0, mask, 0); + + if (rc != 0) { + fprintf(stderr, "Can't change hsm flags for %s: %s\n", + path, strerror(errno = -rc)); + return rc; + } + optind++; + } + + return 0; } -static int lfs_path2fid(int argc, char **argv) +static int lfs_hsm_action(int argc, char **argv) { - char *path; - lustre_fid fid; - int rc; + int rc; + int i = 1; + char *path; + struct hsm_current_action hca; + struct hsm_extent he; + enum hsm_user_action hua; + enum hsm_progress_states hps; + + if (argc < 2) + return CMD_HELP; + + do { + path = argv[i]; + + rc = llapi_hsm_current_action(path, &hca); + if (rc) { + fprintf(stderr, "can't get hsm action for %s: %s\n", + path, strerror(errno = -rc)); + return rc; + } + he = hca.hca_location; + hua = hca.hca_action; + hps = hca.hca_state; + + printf("%s: %s", path, hsm_user_action2name(hua)); + + /* Skip file without action */ + if (hca.hca_action == HUA_NONE) { + printf("\n"); + continue; + } + + printf(" %s ", hsm_progress_state2name(hps)); + + if ((hps == HPS_RUNNING) && + (hua == HUA_ARCHIVE || hua == HUA_RESTORE)) + printf("("LPX64 " bytes moved)\n", he.length); + else if ((he.offset + he.length) == LUSTRE_EOF) + printf("(from "LPX64 " to EOF)\n", he.offset); + else + printf("(from "LPX64 " to "LPX64")\n", + he.offset, he.offset + he.length); + + } while (++i < argc); + + return 0; +} - if (argc != 2) - return CMD_HELP; +static int lfs_hsm_set(int argc, char **argv) +{ + return lfs_hsm_change_flags(argc, argv, LFS_HSM_SET); +} - path = argv[1]; - rc = llapi_path2fid(path, &fid); - if (rc) { - fprintf(stderr, "can't get fid for %s: %s\n", path, - strerror(errno = -rc)); - return rc; - } +static int lfs_hsm_clear(int argc, char **argv) +{ + return lfs_hsm_change_flags(argc, argv, LFS_HSM_CLEAR); +} + +/** + * Check file state and return its fid, to be used by lfs_hsm_request(). + * + * \param[in] file Path to file to check + * \param[in,out] fid Pointer to allocated lu_fid struct. + * \param[in,out] last_dev Pointer to last device id used. + * + * \return 0 on success. + */ +static int lfs_hsm_prepare_file(char *file, struct lu_fid *fid, + dev_t *last_dev) +{ + struct stat st; + int rc; + + rc = lstat(file, &st); + if (rc) { + fprintf(stderr, "Cannot stat %s: %s\n", file, strerror(errno)); + return -errno; + } + /* Checking for regular file as archiving as posix copytool + * rejects archiving files other than regular files + */ + if (!S_ISREG(st.st_mode)) { + fprintf(stderr, "error: \"%s\" is not a regular file\n", file); + return CMD_HELP; + } + /* A request should be ... */ + if (*last_dev != st.st_dev && *last_dev != 0) { + fprintf(stderr, "All files should be " + "on the same filesystem: %s\n", file); + return -EINVAL; + } + *last_dev = st.st_dev; + + rc = llapi_path2fid(file, fid); + if (rc) { + fprintf(stderr, "Cannot read FID of %s: %s\n", + file, strerror(-rc)); + return rc; + } + return 0; +} - printf(DFID"\n", PFID(&fid)); +static int lfs_hsm_request(int argc, char **argv, int action) +{ + struct option long_opts[] = { + {"filelist", 1, 0, 'l'}, + {"data", 1, 0, 'D'}, + {"archive", 1, 0, 'a'}, + {0, 0, 0, 0} + }; + dev_t last_dev = 0; + char short_opts[] = "l:D:a:"; + struct hsm_user_request *hur, *oldhur; + int c, i; + size_t len; + int nbfile; + char *line = NULL; + char *filelist = NULL; + char fullpath[PATH_MAX]; + char *opaque = NULL; + int opaque_len = 0; + int archive_id = 0; + FILE *fp; + int nbfile_alloc = 0; + char some_file[PATH_MAX+1] = ""; + int rc; + + if (argc < 2) + return CMD_HELP; + + optind = 0; + while ((c = getopt_long(argc, argv, short_opts, + long_opts, NULL)) != -1) { + switch (c) { + case 'l': + filelist = optarg; + break; + case 'D': + opaque = optarg; + break; + case 'a': + if (action != HUA_ARCHIVE) { + fprintf(stderr, + "error: -a is supported only " + "when archiving\n"); + return CMD_HELP; + } + archive_id = atoi(optarg); + break; + case '?': + return CMD_HELP; + default: + fprintf(stderr, "error: %s: option '%s' unrecognized\n", + argv[0], argv[optind - 1]); + return CMD_HELP; + } + } + + /* All remaining args are files, so we have at least nbfile */ + nbfile = argc - optind; + + if ((nbfile == 0) && (filelist == NULL)) + return CMD_HELP; + + if (opaque != NULL) + opaque_len = strlen(opaque); + + /* Alloc the request structure with enough place to store all files + * from command line. */ + hur = llapi_hsm_user_request_alloc(nbfile, opaque_len); + if (hur == NULL) { + fprintf(stderr, "Cannot create the request: %s\n", + strerror(errno)); + return errno; + } + nbfile_alloc = nbfile; + + hur->hur_request.hr_action = action; + hur->hur_request.hr_archive_id = archive_id; + hur->hur_request.hr_flags = 0; + + /* All remaining args are files, add them */ + if (nbfile != 0) { + if (strlen(argv[optind]) > sizeof(some_file)-1) { + free(hur); + return -E2BIG; + } + strncpy(some_file, argv[optind], sizeof(some_file)); + } + + for (i = 0; i < nbfile; i++) { + hur->hur_user_item[i].hui_extent.length = -1; + rc = lfs_hsm_prepare_file(argv[optind + i], + &hur->hur_user_item[i].hui_fid, + &last_dev); + hur->hur_request.hr_itemcount++; + if (rc) + goto out_free; + } + + /* from here stop using nb_file, use hur->hur_request.hr_itemcount */ + + /* If a filelist was specified, read the filelist from it. */ + if (filelist != NULL) { + fp = fopen(filelist, "r"); + if (fp == NULL) { + fprintf(stderr, "Cannot read the file list %s: %s\n", + filelist, strerror(errno)); + rc = -errno; + goto out_free; + } + + while ((rc = getline(&line, &len, fp)) != -1) { + struct hsm_user_item *hui; + + /* If allocated buffer was too small, gets something + * bigger */ + if (nbfile_alloc <= hur->hur_request.hr_itemcount) { + ssize_t size; + nbfile_alloc = nbfile_alloc * 2 + 1; + oldhur = hur; + hur = llapi_hsm_user_request_alloc(nbfile_alloc, + opaque_len); + if (hur == NULL) { + fprintf(stderr, "hsm: cannot allocate " + "the request: %s\n", + strerror(errno)); + hur = oldhur; + rc = -errno; + fclose(fp); + goto out_free; + } + size = hur_len(oldhur); + if (size < 0) { + fprintf(stderr, "hsm: cannot allocate " + "%u files + %u bytes data\n", + oldhur->hur_request.hr_itemcount, + oldhur->hur_request.hr_data_len); + free(hur); + hur = oldhur; + rc = -E2BIG; + fclose(fp); + goto out_free; + } + memcpy(hur, oldhur, size); + free(oldhur); + } + + /* Chop CR */ + if (line[strlen(line) - 1] == '\n') + line[strlen(line) - 1] = '\0'; + + hui = + &hur->hur_user_item[hur->hur_request.hr_itemcount]; + hui->hui_extent.length = -1; + rc = lfs_hsm_prepare_file(line, &hui->hui_fid, + &last_dev); + hur->hur_request.hr_itemcount++; + if (rc) { + fclose(fp); + goto out_free; + } + + if ((some_file[0] == '\0') && + (strlen(line) < sizeof(some_file))) + strcpy(some_file, line); + } + + rc = fclose(fp); + if (line) + free(line); + } + + /* If a --data was used, add it to the request */ + hur->hur_request.hr_data_len = opaque_len; + if (opaque != NULL) + memcpy(hur_data(hur), opaque, opaque_len); + + /* Send the HSM request */ + if (realpath(some_file, fullpath) == NULL) { + fprintf(stderr, "Could not find path '%s': %s\n", + some_file, strerror(errno)); + } + rc = llapi_hsm_request(fullpath, hur); + if (rc) { + fprintf(stderr, "Cannot send HSM request (use of %s): %s\n", + some_file, strerror(-rc)); + goto out_free; + } + +out_free: + free(hur); + return rc; +} - return 0; +static int lfs_hsm_archive(int argc, char **argv) +{ + return lfs_hsm_request(argc, argv, HUA_ARCHIVE); +} + +static int lfs_hsm_restore(int argc, char **argv) +{ + return lfs_hsm_request(argc, argv, HUA_RESTORE); +} + +static int lfs_hsm_release(int argc, char **argv) +{ + return lfs_hsm_request(argc, argv, HUA_RELEASE); +} + +static int lfs_hsm_remove(int argc, char **argv) +{ + return lfs_hsm_request(argc, argv, HUA_REMOVE); +} + +static int lfs_hsm_cancel(int argc, char **argv) +{ + return lfs_hsm_request(argc, argv, HUA_CANCEL); +} + +static int lfs_swap_layouts(int argc, char **argv) +{ + if (argc != 3) + return CMD_HELP; + + return llapi_swap_layouts(argv[1], argv[2], 0, 0, + SWAP_LAYOUTS_KEEP_MTIME | + SWAP_LAYOUTS_KEEP_ATIME); } int main(int argc, char **argv) @@ -2627,13 +4086,7 @@ int main(int argc, char **argv) setlinebuf(stdout); - ptl_initialize(argc, argv); - if (obd_initialize(argc, argv) < 0) - exit(2); - if (dbg_initialize(argc, argv) < 0) - exit(3); - - Parser_init("lfs > ", cmdlist); + Parser_init("lfs > ", cmdlist); if (argc > 1) { rc = Parser_execarg(argc - 1, argv + 1, cmdlist); @@ -2641,7 +4094,10 @@ int main(int argc, char **argv) rc = Parser_commands(); } - obd_finalize(argc, argv); return rc < 0 ? -rc : rc; } +#ifdef _LUSTRE_IDL_H_ +/* Everything we need here should be included by lustreapi.h. */ +# error "lfs should not depend on lustre_idl.h" +#endif /* _LUSTRE_IDL_H_ */