16cbd5570SChris Mason /* 26cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 36cbd5570SChris Mason * 46cbd5570SChris Mason * This program is free software; you can redistribute it and/or 56cbd5570SChris Mason * modify it under the terms of the GNU General Public 66cbd5570SChris Mason * License v2 as published by the Free Software Foundation. 76cbd5570SChris Mason * 86cbd5570SChris Mason * This program is distributed in the hope that it will be useful, 96cbd5570SChris Mason * but WITHOUT ANY WARRANTY; without even the implied warranty of 106cbd5570SChris Mason * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 116cbd5570SChris Mason * General Public License for more details. 126cbd5570SChris Mason * 136cbd5570SChris Mason * You should have received a copy of the GNU General Public 146cbd5570SChris Mason * License along with this program; if not, write to the 156cbd5570SChris Mason * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 166cbd5570SChris Mason * Boston, MA 021110-1307, USA. 176cbd5570SChris Mason */ 186cbd5570SChris Mason 194b82d6e4SYan #include <linux/blkdev.h> 202e635a27SChris Mason #include <linux/module.h> 21e20d96d6SChris Mason #include <linux/buffer_head.h> 222e635a27SChris Mason #include <linux/fs.h> 232e635a27SChris Mason #include <linux/pagemap.h> 242e635a27SChris Mason #include <linux/highmem.h> 252e635a27SChris Mason #include <linux/time.h> 262e635a27SChris Mason #include <linux/init.h> 27a9572a15SEric Paris #include <linux/seq_file.h> 282e635a27SChris Mason #include <linux/string.h> 292e635a27SChris Mason #include <linux/backing-dev.h> 304b82d6e4SYan #include <linux/mount.h> 31dee26a9fSChris Mason #include <linux/mpage.h> 3275dfe396SChris Mason #include <linux/swap.h> 3375dfe396SChris Mason #include <linux/writeback.h> 348fd17795SChris Mason #include <linux/statfs.h> 3508607c1bSChris Mason #include <linux/compat.h> 3695e05289SChris Mason #include <linux/parser.h> 37c59f8951SChris Mason #include <linux/ctype.h> 386da6abaeSChris Mason #include <linux/namei.h> 39a9218f6bSChris Mason #include <linux/miscdevice.h> 401bcbf313SQinghuang Feng #include <linux/magic.h> 415a0e3ad6STejun Heo #include <linux/slab.h> 4290a887c9SDan Magenheimer #include <linux/cleancache.h> 4322c44fe6SJosef Bacik #include <linux/ratelimit.h> 4455e301fdSFilipe Brandenburger #include <linux/btrfs.h> 454b4e25f2SChris Mason #include "compat.h" 4616cdcec7SMiao Xie #include "delayed-inode.h" 472e635a27SChris Mason #include "ctree.h" 48e20d96d6SChris Mason #include "disk-io.h" 49d5719762SChris Mason #include "transaction.h" 502c90e5d6SChris Mason #include "btrfs_inode.h" 513a686375SChris Mason #include "print-tree.h" 525103e947SJosef Bacik #include "xattr.h" 538a4b83ccSChris Mason #include "volumes.h" 54be6e8dc0SBalaji Rao #include "export.h" 55c8b97818SChris Mason #include "compression.h" 569c5085c1SJosef Bacik #include "rcu-string.h" 578dabb742SStefan Behrens #include "dev-replace.h" 5874255aa0SJosef Bacik #include "free-space-cache.h" 592e635a27SChris Mason 601abe9b8aSliubo #define CREATE_TRACE_POINTS 611abe9b8aSliubo #include <trace/events/btrfs.h> 621abe9b8aSliubo 63b87221deSAlexey Dobriyan static const struct super_operations btrfs_super_ops; 64830c4adbSJosef Bacik static struct file_system_type btrfs_fs_type; 65e20d96d6SChris Mason 6608748810SDavid Sterba static const char *btrfs_decode_error(int errno) 67acce952bSliubo { 6808748810SDavid Sterba char *errstr = "unknown"; 69acce952bSliubo 70acce952bSliubo switch (errno) { 71acce952bSliubo case -EIO: 72acce952bSliubo errstr = "IO failure"; 73acce952bSliubo break; 74acce952bSliubo case -ENOMEM: 75acce952bSliubo errstr = "Out of memory"; 76acce952bSliubo break; 77acce952bSliubo case -EROFS: 78acce952bSliubo errstr = "Readonly filesystem"; 79acce952bSliubo break; 808c342930SJeff Mahoney case -EEXIST: 818c342930SJeff Mahoney errstr = "Object already exists"; 828c342930SJeff Mahoney break; 8394ef7280SDavid Sterba case -ENOSPC: 8494ef7280SDavid Sterba errstr = "No space left"; 8594ef7280SDavid Sterba break; 8694ef7280SDavid Sterba case -ENOENT: 8794ef7280SDavid Sterba errstr = "No such entry"; 8894ef7280SDavid Sterba break; 89acce952bSliubo } 90acce952bSliubo 91acce952bSliubo return errstr; 92acce952bSliubo } 93acce952bSliubo 94bbece8a3SDavid Sterba static void save_error_info(struct btrfs_fs_info *fs_info) 95acce952bSliubo { 96acce952bSliubo /* 97acce952bSliubo * today we only save the error info into ram. Long term we'll 98acce952bSliubo * also send it down to the disk 99acce952bSliubo */ 10087533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 101acce952bSliubo } 102acce952bSliubo 103acce952bSliubo /* btrfs handle error by forcing the filesystem readonly */ 104acce952bSliubo static void btrfs_handle_error(struct btrfs_fs_info *fs_info) 105acce952bSliubo { 106acce952bSliubo struct super_block *sb = fs_info->sb; 107acce952bSliubo 108acce952bSliubo if (sb->s_flags & MS_RDONLY) 109acce952bSliubo return; 110acce952bSliubo 11187533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 112acce952bSliubo sb->s_flags |= MS_RDONLY; 113c2cf52ebSSimon Kirby btrfs_info(fs_info, "forced readonly"); 1141acd6831SStefan Behrens /* 1151acd6831SStefan Behrens * Note that a running device replace operation is not 1161acd6831SStefan Behrens * canceled here although there is no way to update 1171acd6831SStefan Behrens * the progress. It would add the risk of a deadlock, 1181acd6831SStefan Behrens * therefore the canceling is ommited. The only penalty 1191acd6831SStefan Behrens * is that some I/O remains active until the procedure 1201acd6831SStefan Behrens * completes. The next time when the filesystem is 1211acd6831SStefan Behrens * mounted writeable again, the device replace 1221acd6831SStefan Behrens * operation continues. 1231acd6831SStefan Behrens */ 124acce952bSliubo } 125acce952bSliubo } 126acce952bSliubo 127533574c6SJoe Perches #ifdef CONFIG_PRINTK 128acce952bSliubo /* 129acce952bSliubo * __btrfs_std_error decodes expected errors from the caller and 130acce952bSliubo * invokes the approciate error response. 131acce952bSliubo */ 132acce952bSliubo void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, 1334da35113SJeff Mahoney unsigned int line, int errno, const char *fmt, ...) 134acce952bSliubo { 135acce952bSliubo struct super_block *sb = fs_info->sb; 136acce952bSliubo const char *errstr; 137acce952bSliubo 138acce952bSliubo /* 139acce952bSliubo * Special case: if the error is EROFS, and we're already 140acce952bSliubo * under MS_RDONLY, then it is safe here. 141acce952bSliubo */ 142acce952bSliubo if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) 143acce952bSliubo return; 144acce952bSliubo 14508748810SDavid Sterba errstr = btrfs_decode_error(errno); 1464da35113SJeff Mahoney if (fmt) { 14737252a66SEric Sandeen struct va_format vaf; 14837252a66SEric Sandeen va_list args; 14937252a66SEric Sandeen 15037252a66SEric Sandeen va_start(args, fmt); 15137252a66SEric Sandeen vaf.fmt = fmt; 15237252a66SEric Sandeen vaf.va = &args; 1534da35113SJeff Mahoney 15408748810SDavid Sterba printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s (%pV)\n", 15508748810SDavid Sterba sb->s_id, function, line, errno, errstr, &vaf); 15637252a66SEric Sandeen va_end(args); 1574da35113SJeff Mahoney } else { 15808748810SDavid Sterba printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s\n", 15908748810SDavid Sterba sb->s_id, function, line, errno, errstr); 1604da35113SJeff Mahoney } 161acce952bSliubo 1624da35113SJeff Mahoney /* Don't go through full error handling during mount */ 1634da35113SJeff Mahoney save_error_info(fs_info); 164cf79ffb5SJosef Bacik if (sb->s_flags & MS_BORN) 165acce952bSliubo btrfs_handle_error(fs_info); 166acce952bSliubo } 1674da35113SJeff Mahoney 168533574c6SJoe Perches static const char * const logtypes[] = { 1694da35113SJeff Mahoney "emergency", 1704da35113SJeff Mahoney "alert", 1714da35113SJeff Mahoney "critical", 1724da35113SJeff Mahoney "error", 1734da35113SJeff Mahoney "warning", 1744da35113SJeff Mahoney "notice", 1754da35113SJeff Mahoney "info", 1764da35113SJeff Mahoney "debug", 1774da35113SJeff Mahoney }; 1784da35113SJeff Mahoney 179c2cf52ebSSimon Kirby void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...) 1804da35113SJeff Mahoney { 1814da35113SJeff Mahoney struct super_block *sb = fs_info->sb; 1824da35113SJeff Mahoney char lvl[4]; 1834da35113SJeff Mahoney struct va_format vaf; 1844da35113SJeff Mahoney va_list args; 1854da35113SJeff Mahoney const char *type = logtypes[4]; 186533574c6SJoe Perches int kern_level; 1874da35113SJeff Mahoney 1884da35113SJeff Mahoney va_start(args, fmt); 1894da35113SJeff Mahoney 190533574c6SJoe Perches kern_level = printk_get_level(fmt); 191533574c6SJoe Perches if (kern_level) { 192533574c6SJoe Perches size_t size = printk_skip_level(fmt) - fmt; 193533574c6SJoe Perches memcpy(lvl, fmt, size); 194533574c6SJoe Perches lvl[size] = '\0'; 195533574c6SJoe Perches fmt += size; 196533574c6SJoe Perches type = logtypes[kern_level - '0']; 1974da35113SJeff Mahoney } else 1984da35113SJeff Mahoney *lvl = '\0'; 1994da35113SJeff Mahoney 2004da35113SJeff Mahoney vaf.fmt = fmt; 2014da35113SJeff Mahoney vaf.va = &args; 202533574c6SJoe Perches 203c2cf52ebSSimon Kirby printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf); 204533574c6SJoe Perches 205533574c6SJoe Perches va_end(args); 2064da35113SJeff Mahoney } 207acce952bSliubo 208533574c6SJoe Perches #else 209533574c6SJoe Perches 210533574c6SJoe Perches void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, 211533574c6SJoe Perches unsigned int line, int errno, const char *fmt, ...) 212533574c6SJoe Perches { 213533574c6SJoe Perches struct super_block *sb = fs_info->sb; 214533574c6SJoe Perches 215533574c6SJoe Perches /* 216533574c6SJoe Perches * Special case: if the error is EROFS, and we're already 217533574c6SJoe Perches * under MS_RDONLY, then it is safe here. 218533574c6SJoe Perches */ 219533574c6SJoe Perches if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) 220533574c6SJoe Perches return; 221533574c6SJoe Perches 222533574c6SJoe Perches /* Don't go through full error handling during mount */ 223533574c6SJoe Perches if (sb->s_flags & MS_BORN) { 224533574c6SJoe Perches save_error_info(fs_info); 225533574c6SJoe Perches btrfs_handle_error(fs_info); 226533574c6SJoe Perches } 227533574c6SJoe Perches } 228533574c6SJoe Perches #endif 229533574c6SJoe Perches 2308c342930SJeff Mahoney /* 23149b25e05SJeff Mahoney * We only mark the transaction aborted and then set the file system read-only. 23249b25e05SJeff Mahoney * This will prevent new transactions from starting or trying to join this 23349b25e05SJeff Mahoney * one. 23449b25e05SJeff Mahoney * 23549b25e05SJeff Mahoney * This means that error recovery at the call site is limited to freeing 23649b25e05SJeff Mahoney * any local memory allocations and passing the error code up without 23749b25e05SJeff Mahoney * further cleanup. The transaction should complete as it normally would 23849b25e05SJeff Mahoney * in the call path but will return -EIO. 23949b25e05SJeff Mahoney * 24049b25e05SJeff Mahoney * We'll complete the cleanup in btrfs_end_transaction and 24149b25e05SJeff Mahoney * btrfs_commit_transaction. 24249b25e05SJeff Mahoney */ 24349b25e05SJeff Mahoney void __btrfs_abort_transaction(struct btrfs_trans_handle *trans, 24449b25e05SJeff Mahoney struct btrfs_root *root, const char *function, 24549b25e05SJeff Mahoney unsigned int line, int errno) 24649b25e05SJeff Mahoney { 24708748810SDavid Sterba /* 24808748810SDavid Sterba * Report first abort since mount 24908748810SDavid Sterba */ 25008748810SDavid Sterba if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED, 25108748810SDavid Sterba &root->fs_info->fs_state)) { 25208748810SDavid Sterba WARN(1, KERN_DEBUG "btrfs: Transaction aborted (error %d)\n", 25308748810SDavid Sterba errno); 25408748810SDavid Sterba } 25549b25e05SJeff Mahoney trans->aborted = errno; 25649b25e05SJeff Mahoney /* Nothing used. The other threads that have joined this 25749b25e05SJeff Mahoney * transaction may be able to continue. */ 25849b25e05SJeff Mahoney if (!trans->blocks_used) { 25969ce977aSMiao Xie const char *errstr; 26069ce977aSMiao Xie 26108748810SDavid Sterba errstr = btrfs_decode_error(errno); 262c2cf52ebSSimon Kirby btrfs_warn(root->fs_info, 263c2cf52ebSSimon Kirby "%s:%d: Aborting unused transaction(%s).", 26469ce977aSMiao Xie function, line, errstr); 26549b25e05SJeff Mahoney return; 26649b25e05SJeff Mahoney } 2678d25a086SMiao Xie ACCESS_ONCE(trans->transaction->aborted) = errno; 268501407aaSJosef Bacik /* Wake up anybody who may be waiting on this transaction */ 269501407aaSJosef Bacik wake_up(&root->fs_info->transaction_wait); 270501407aaSJosef Bacik wake_up(&root->fs_info->transaction_blocked_wait); 27149b25e05SJeff Mahoney __btrfs_std_error(root->fs_info, function, line, errno, NULL); 27249b25e05SJeff Mahoney } 27349b25e05SJeff Mahoney /* 2748c342930SJeff Mahoney * __btrfs_panic decodes unexpected, fatal errors from the caller, 2758c342930SJeff Mahoney * issues an alert, and either panics or BUGs, depending on mount options. 2768c342930SJeff Mahoney */ 2778c342930SJeff Mahoney void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function, 2788c342930SJeff Mahoney unsigned int line, int errno, const char *fmt, ...) 2798c342930SJeff Mahoney { 2808c342930SJeff Mahoney char *s_id = "<unknown>"; 2818c342930SJeff Mahoney const char *errstr; 2828c342930SJeff Mahoney struct va_format vaf = { .fmt = fmt }; 2838c342930SJeff Mahoney va_list args; 2848c342930SJeff Mahoney 2858c342930SJeff Mahoney if (fs_info) 2868c342930SJeff Mahoney s_id = fs_info->sb->s_id; 2878c342930SJeff Mahoney 2888c342930SJeff Mahoney va_start(args, fmt); 2898c342930SJeff Mahoney vaf.va = &args; 2908c342930SJeff Mahoney 29108748810SDavid Sterba errstr = btrfs_decode_error(errno); 292aa43a17cSEric Sandeen if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)) 29308748810SDavid Sterba panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n", 29408748810SDavid Sterba s_id, function, line, &vaf, errno, errstr); 2958c342930SJeff Mahoney 29608748810SDavid Sterba printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n", 29708748810SDavid Sterba s_id, function, line, &vaf, errno, errstr); 2988c342930SJeff Mahoney va_end(args); 2998c342930SJeff Mahoney /* Caller calls BUG() */ 3008c342930SJeff Mahoney } 301e20d96d6SChris Mason 302e20d96d6SChris Mason static void btrfs_put_super(struct super_block *sb) 303e20d96d6SChris Mason { 304815745cfSAl Viro (void)close_ctree(btrfs_sb(sb)->tree_root); 305aea52e19SAl Viro /* FIXME: need to fix VFS to return error? */ 306aea52e19SAl Viro /* AV: return it _where_? ->put_super() can be triggered by any number 307aea52e19SAl Viro * of async events, up to and including delivery of SIGKILL to the 308aea52e19SAl Viro * last process that kept it busy. Or segfault in the aforementioned 309aea52e19SAl Viro * process... Whom would you report that to? 310aea52e19SAl Viro */ 311e20d96d6SChris Mason } 3122e635a27SChris Mason 31395e05289SChris Mason enum { 31473f73415SJosef Bacik Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum, 315287a0ab9SJosef Bacik Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, 316287a0ab9SJosef Bacik Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress, 317261507a0SLi Zefan Opt_compress_type, Opt_compress_force, Opt_compress_force_type, 318261507a0SLi Zefan Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, 31991435650SChris Mason Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, 3209555c6c1SIlya Dryomov Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache, 3219555c6c1SIlya Dryomov Opt_no_space_cache, Opt_recovery, Opt_skip_balance, 32221adbd5cSStefan Behrens Opt_check_integrity, Opt_check_integrity_including_extent_data, 3238c342930SJeff Mahoney Opt_check_integrity_print_mask, Opt_fatal_errors, 324*8b87dc17SDavid Sterba Opt_commit_interval, 3259555c6c1SIlya Dryomov Opt_err, 32695e05289SChris Mason }; 32795e05289SChris Mason 32895e05289SChris Mason static match_table_t tokens = { 329dfe25020SChris Mason {Opt_degraded, "degraded"}, 33095e05289SChris Mason {Opt_subvol, "subvol=%s"}, 3311493381fSWang Shilong {Opt_subvolid, "subvolid=%s"}, 33243e570b0SChristoph Hellwig {Opt_device, "device=%s"}, 333b6cda9bcSChris Mason {Opt_nodatasum, "nodatasum"}, 334be20aa9dSChris Mason {Opt_nodatacow, "nodatacow"}, 33521ad10cfSChris Mason {Opt_nobarrier, "nobarrier"}, 3366f568d35SChris Mason {Opt_max_inline, "max_inline=%s"}, 3378f662a76SChris Mason {Opt_alloc_start, "alloc_start=%s"}, 3384543df7eSChris Mason {Opt_thread_pool, "thread_pool=%d"}, 339c8b97818SChris Mason {Opt_compress, "compress"}, 340261507a0SLi Zefan {Opt_compress_type, "compress=%s"}, 341a555f810SChris Mason {Opt_compress_force, "compress-force"}, 342261507a0SLi Zefan {Opt_compress_force_type, "compress-force=%s"}, 343e18e4809SChris Mason {Opt_ssd, "ssd"}, 344451d7585SChris Mason {Opt_ssd_spread, "ssd_spread"}, 3453b30c22fSChris Mason {Opt_nossd, "nossd"}, 34633268eafSJosef Bacik {Opt_noacl, "noacl"}, 3473a5e1404SSage Weil {Opt_notreelog, "notreelog"}, 348dccae999SSage Weil {Opt_flushoncommit, "flushoncommit"}, 34997e728d4SJosef Bacik {Opt_ratio, "metadata_ratio=%d"}, 350e244a0aeSChristoph Hellwig {Opt_discard, "discard"}, 3510af3d00bSJosef Bacik {Opt_space_cache, "space_cache"}, 35288c2ba3bSJosef Bacik {Opt_clear_cache, "clear_cache"}, 3534260f7c7SSage Weil {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, 35491435650SChris Mason {Opt_enospc_debug, "enospc_debug"}, 355e15d0542SXin Zhong {Opt_subvolrootid, "subvolrootid=%d"}, 3564cb5300bSChris Mason {Opt_defrag, "autodefrag"}, 3574b9465cbSChris Mason {Opt_inode_cache, "inode_cache"}, 3588965593eSDavid Sterba {Opt_no_space_cache, "nospace_cache"}, 359af31f5e5SChris Mason {Opt_recovery, "recovery"}, 3609555c6c1SIlya Dryomov {Opt_skip_balance, "skip_balance"}, 36121adbd5cSStefan Behrens {Opt_check_integrity, "check_int"}, 36221adbd5cSStefan Behrens {Opt_check_integrity_including_extent_data, "check_int_data"}, 36321adbd5cSStefan Behrens {Opt_check_integrity_print_mask, "check_int_print_mask=%d"}, 3648c342930SJeff Mahoney {Opt_fatal_errors, "fatal_errors=%s"}, 365*8b87dc17SDavid Sterba {Opt_commit_interval, "commit=%d"}, 36633268eafSJosef Bacik {Opt_err, NULL}, 36795e05289SChris Mason }; 36895e05289SChris Mason 369edf24abeSChristoph Hellwig /* 370edf24abeSChristoph Hellwig * Regular mount options parser. Everything that is needed only when 371edf24abeSChristoph Hellwig * reading in a new superblock is parsed here. 37249b25e05SJeff Mahoney * XXX JDM: This needs to be cleaned up for remount. 373edf24abeSChristoph Hellwig */ 374edf24abeSChristoph Hellwig int btrfs_parse_options(struct btrfs_root *root, char *options) 37595e05289SChris Mason { 376edf24abeSChristoph Hellwig struct btrfs_fs_info *info = root->fs_info; 37795e05289SChris Mason substring_t args[MAX_OPT_ARGS]; 37873bc1876SJosef Bacik char *p, *num, *orig = NULL; 37973bc1876SJosef Bacik u64 cache_gen; 3804543df7eSChris Mason int intarg; 381a7a3f7caSSage Weil int ret = 0; 382261507a0SLi Zefan char *compress_type; 383261507a0SLi Zefan bool compress_force = false; 384b6cda9bcSChris Mason 3856c41761fSDavid Sterba cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy); 38673bc1876SJosef Bacik if (cache_gen) 38773bc1876SJosef Bacik btrfs_set_opt(info->mount_opt, SPACE_CACHE); 38873bc1876SJosef Bacik 38995e05289SChris Mason if (!options) 39073bc1876SJosef Bacik goto out; 39195e05289SChris Mason 392be20aa9dSChris Mason /* 393be20aa9dSChris Mason * strsep changes the string, duplicate it because parse_options 394be20aa9dSChris Mason * gets called twice 395be20aa9dSChris Mason */ 396be20aa9dSChris Mason options = kstrdup(options, GFP_NOFS); 397be20aa9dSChris Mason if (!options) 398be20aa9dSChris Mason return -ENOMEM; 399be20aa9dSChris Mason 400da495eccSJosef Bacik orig = options; 401be20aa9dSChris Mason 40295e05289SChris Mason while ((p = strsep(&options, ",")) != NULL) { 40395e05289SChris Mason int token; 40495e05289SChris Mason if (!*p) 40595e05289SChris Mason continue; 40695e05289SChris Mason 40795e05289SChris Mason token = match_token(p, tokens, args); 40895e05289SChris Mason switch (token) { 409dfe25020SChris Mason case Opt_degraded: 410edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: allowing degraded mounts\n"); 411dfe25020SChris Mason btrfs_set_opt(info->mount_opt, DEGRADED); 412dfe25020SChris Mason break; 41395e05289SChris Mason case Opt_subvol: 41473f73415SJosef Bacik case Opt_subvolid: 415e15d0542SXin Zhong case Opt_subvolrootid: 41643e570b0SChristoph Hellwig case Opt_device: 417edf24abeSChristoph Hellwig /* 41843e570b0SChristoph Hellwig * These are parsed by btrfs_parse_early_options 419edf24abeSChristoph Hellwig * and can be happily ignored here. 420edf24abeSChristoph Hellwig */ 42195e05289SChris Mason break; 422b6cda9bcSChris Mason case Opt_nodatasum: 423067c28adSChris Mason printk(KERN_INFO "btrfs: setting nodatasum\n"); 424b6cda9bcSChris Mason btrfs_set_opt(info->mount_opt, NODATASUM); 425be20aa9dSChris Mason break; 426be20aa9dSChris Mason case Opt_nodatacow: 427bedb2ccaSAndrei Popa if (!btrfs_test_opt(root, COMPRESS) || 428bedb2ccaSAndrei Popa !btrfs_test_opt(root, FORCE_COMPRESS)) { 429bedb2ccaSAndrei Popa printk(KERN_INFO "btrfs: setting nodatacow, compression disabled\n"); 430bedb2ccaSAndrei Popa } else { 431edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: setting nodatacow\n"); 432bedb2ccaSAndrei Popa } 433bedb2ccaSAndrei Popa info->compress_type = BTRFS_COMPRESS_NONE; 434bedb2ccaSAndrei Popa btrfs_clear_opt(info->mount_opt, COMPRESS); 435bedb2ccaSAndrei Popa btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 436be20aa9dSChris Mason btrfs_set_opt(info->mount_opt, NODATACOW); 437be20aa9dSChris Mason btrfs_set_opt(info->mount_opt, NODATASUM); 438b6cda9bcSChris Mason break; 439a555f810SChris Mason case Opt_compress_force: 440261507a0SLi Zefan case Opt_compress_force_type: 441261507a0SLi Zefan compress_force = true; 4421c697d4aSEric Sandeen /* Fallthrough */ 443261507a0SLi Zefan case Opt_compress: 444261507a0SLi Zefan case Opt_compress_type: 445261507a0SLi Zefan if (token == Opt_compress || 446261507a0SLi Zefan token == Opt_compress_force || 447261507a0SLi Zefan strcmp(args[0].from, "zlib") == 0) { 448261507a0SLi Zefan compress_type = "zlib"; 449261507a0SLi Zefan info->compress_type = BTRFS_COMPRESS_ZLIB; 450063849eaSArnd Hannemann btrfs_set_opt(info->mount_opt, COMPRESS); 451bedb2ccaSAndrei Popa btrfs_clear_opt(info->mount_opt, NODATACOW); 452bedb2ccaSAndrei Popa btrfs_clear_opt(info->mount_opt, NODATASUM); 453a6fa6faeSLi Zefan } else if (strcmp(args[0].from, "lzo") == 0) { 454a6fa6faeSLi Zefan compress_type = "lzo"; 455a6fa6faeSLi Zefan info->compress_type = BTRFS_COMPRESS_LZO; 456063849eaSArnd Hannemann btrfs_set_opt(info->mount_opt, COMPRESS); 457bedb2ccaSAndrei Popa btrfs_clear_opt(info->mount_opt, NODATACOW); 458bedb2ccaSAndrei Popa btrfs_clear_opt(info->mount_opt, NODATASUM); 4592b0ce2c2SMitch Harder btrfs_set_fs_incompat(info, COMPRESS_LZO); 460063849eaSArnd Hannemann } else if (strncmp(args[0].from, "no", 2) == 0) { 461063849eaSArnd Hannemann compress_type = "no"; 462063849eaSArnd Hannemann info->compress_type = BTRFS_COMPRESS_NONE; 463063849eaSArnd Hannemann btrfs_clear_opt(info->mount_opt, COMPRESS); 464063849eaSArnd Hannemann btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 465063849eaSArnd Hannemann compress_force = false; 466261507a0SLi Zefan } else { 467261507a0SLi Zefan ret = -EINVAL; 468261507a0SLi Zefan goto out; 469261507a0SLi Zefan } 470261507a0SLi Zefan 471261507a0SLi Zefan if (compress_force) { 472261507a0SLi Zefan btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); 473261507a0SLi Zefan pr_info("btrfs: force %s compression\n", 474261507a0SLi Zefan compress_type); 475261507a0SLi Zefan } else 476261507a0SLi Zefan pr_info("btrfs: use %s compression\n", 477261507a0SLi Zefan compress_type); 478a555f810SChris Mason break; 479e18e4809SChris Mason case Opt_ssd: 480edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); 481e18e4809SChris Mason btrfs_set_opt(info->mount_opt, SSD); 482e18e4809SChris Mason break; 483451d7585SChris Mason case Opt_ssd_spread: 484451d7585SChris Mason printk(KERN_INFO "btrfs: use spread ssd " 485451d7585SChris Mason "allocation scheme\n"); 486451d7585SChris Mason btrfs_set_opt(info->mount_opt, SSD); 487451d7585SChris Mason btrfs_set_opt(info->mount_opt, SSD_SPREAD); 488451d7585SChris Mason break; 4893b30c22fSChris Mason case Opt_nossd: 490451d7585SChris Mason printk(KERN_INFO "btrfs: not using ssd allocation " 491451d7585SChris Mason "scheme\n"); 492c289811cSChris Mason btrfs_set_opt(info->mount_opt, NOSSD); 4933b30c22fSChris Mason btrfs_clear_opt(info->mount_opt, SSD); 494451d7585SChris Mason btrfs_clear_opt(info->mount_opt, SSD_SPREAD); 4953b30c22fSChris Mason break; 49621ad10cfSChris Mason case Opt_nobarrier: 497edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: turning off barriers\n"); 49821ad10cfSChris Mason btrfs_set_opt(info->mount_opt, NOBARRIER); 49921ad10cfSChris Mason break; 5004543df7eSChris Mason case Opt_thread_pool: 5012c334e87SWang Shilong ret = match_int(&args[0], &intarg); 5022c334e87SWang Shilong if (ret) { 5032c334e87SWang Shilong goto out; 5042c334e87SWang Shilong } else if (intarg > 0) { 5054543df7eSChris Mason info->thread_pool_size = intarg; 5062c334e87SWang Shilong } else { 5072c334e87SWang Shilong ret = -EINVAL; 5082c334e87SWang Shilong goto out; 5092c334e87SWang Shilong } 5104543df7eSChris Mason break; 5116f568d35SChris Mason case Opt_max_inline: 512edf24abeSChristoph Hellwig num = match_strdup(&args[0]); 5136f568d35SChris Mason if (num) { 51491748467SAkinobu Mita info->max_inline = memparse(num, NULL); 5156f568d35SChris Mason kfree(num); 5166f568d35SChris Mason 51715ada040SChris Mason if (info->max_inline) { 5186f568d35SChris Mason info->max_inline = max_t(u64, 51915ada040SChris Mason info->max_inline, 52015ada040SChris Mason root->sectorsize); 52115ada040SChris Mason } 522edf24abeSChristoph Hellwig printk(KERN_INFO "btrfs: max_inline at %llu\n", 52321380931SJoel Becker (unsigned long long)info->max_inline); 5242c334e87SWang Shilong } else { 5252c334e87SWang Shilong ret = -ENOMEM; 5262c334e87SWang Shilong goto out; 5276f568d35SChris Mason } 5286f568d35SChris Mason break; 5298f662a76SChris Mason case Opt_alloc_start: 530edf24abeSChristoph Hellwig num = match_strdup(&args[0]); 5318f662a76SChris Mason if (num) { 532c018daecSMiao Xie mutex_lock(&info->chunk_mutex); 53391748467SAkinobu Mita info->alloc_start = memparse(num, NULL); 534c018daecSMiao Xie mutex_unlock(&info->chunk_mutex); 5358f662a76SChris Mason kfree(num); 536edf24abeSChristoph Hellwig printk(KERN_INFO 537edf24abeSChristoph Hellwig "btrfs: allocations start at %llu\n", 53821380931SJoel Becker (unsigned long long)info->alloc_start); 5392c334e87SWang Shilong } else { 5402c334e87SWang Shilong ret = -ENOMEM; 5412c334e87SWang Shilong goto out; 5428f662a76SChris Mason } 5438f662a76SChris Mason break; 54433268eafSJosef Bacik case Opt_noacl: 54533268eafSJosef Bacik root->fs_info->sb->s_flags &= ~MS_POSIXACL; 54633268eafSJosef Bacik break; 5473a5e1404SSage Weil case Opt_notreelog: 5483a5e1404SSage Weil printk(KERN_INFO "btrfs: disabling tree log\n"); 5493a5e1404SSage Weil btrfs_set_opt(info->mount_opt, NOTREELOG); 5503a5e1404SSage Weil break; 551dccae999SSage Weil case Opt_flushoncommit: 552dccae999SSage Weil printk(KERN_INFO "btrfs: turning on flush-on-commit\n"); 553dccae999SSage Weil btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); 554dccae999SSage Weil break; 55597e728d4SJosef Bacik case Opt_ratio: 5562c334e87SWang Shilong ret = match_int(&args[0], &intarg); 5572c334e87SWang Shilong if (ret) { 5582c334e87SWang Shilong goto out; 5592c334e87SWang Shilong } else if (intarg >= 0) { 56097e728d4SJosef Bacik info->metadata_ratio = intarg; 56197e728d4SJosef Bacik printk(KERN_INFO "btrfs: metadata ratio %d\n", 56297e728d4SJosef Bacik info->metadata_ratio); 5632c334e87SWang Shilong } else { 5642c334e87SWang Shilong ret = -EINVAL; 5652c334e87SWang Shilong goto out; 56697e728d4SJosef Bacik } 56797e728d4SJosef Bacik break; 568e244a0aeSChristoph Hellwig case Opt_discard: 569e244a0aeSChristoph Hellwig btrfs_set_opt(info->mount_opt, DISCARD); 570e244a0aeSChristoph Hellwig break; 5710af3d00bSJosef Bacik case Opt_space_cache: 5720af3d00bSJosef Bacik btrfs_set_opt(info->mount_opt, SPACE_CACHE); 5730de90876SJosef Bacik break; 57473bc1876SJosef Bacik case Opt_no_space_cache: 57573bc1876SJosef Bacik printk(KERN_INFO "btrfs: disabling disk space caching\n"); 57673bc1876SJosef Bacik btrfs_clear_opt(info->mount_opt, SPACE_CACHE); 57773bc1876SJosef Bacik break; 5784b9465cbSChris Mason case Opt_inode_cache: 5794b9465cbSChris Mason printk(KERN_INFO "btrfs: enabling inode map caching\n"); 5804b9465cbSChris Mason btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); 5814b9465cbSChris Mason break; 58288c2ba3bSJosef Bacik case Opt_clear_cache: 58388c2ba3bSJosef Bacik printk(KERN_INFO "btrfs: force clearing of disk cache\n"); 58488c2ba3bSJosef Bacik btrfs_set_opt(info->mount_opt, CLEAR_CACHE); 5850af3d00bSJosef Bacik break; 5864260f7c7SSage Weil case Opt_user_subvol_rm_allowed: 5874260f7c7SSage Weil btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); 5884260f7c7SSage Weil break; 58991435650SChris Mason case Opt_enospc_debug: 59091435650SChris Mason btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG); 59191435650SChris Mason break; 5924cb5300bSChris Mason case Opt_defrag: 59348940662SDaniel J Blueman printk(KERN_INFO "btrfs: enabling auto defrag\n"); 5944cb5300bSChris Mason btrfs_set_opt(info->mount_opt, AUTO_DEFRAG); 5954cb5300bSChris Mason break; 596af31f5e5SChris Mason case Opt_recovery: 59748940662SDaniel J Blueman printk(KERN_INFO "btrfs: enabling auto recovery\n"); 598af31f5e5SChris Mason btrfs_set_opt(info->mount_opt, RECOVERY); 599af31f5e5SChris Mason break; 6009555c6c1SIlya Dryomov case Opt_skip_balance: 6019555c6c1SIlya Dryomov btrfs_set_opt(info->mount_opt, SKIP_BALANCE); 6029555c6c1SIlya Dryomov break; 60321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 60421adbd5cSStefan Behrens case Opt_check_integrity_including_extent_data: 60521adbd5cSStefan Behrens printk(KERN_INFO "btrfs: enabling check integrity" 60621adbd5cSStefan Behrens " including extent data\n"); 60721adbd5cSStefan Behrens btrfs_set_opt(info->mount_opt, 60821adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA); 60921adbd5cSStefan Behrens btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); 61021adbd5cSStefan Behrens break; 61121adbd5cSStefan Behrens case Opt_check_integrity: 61221adbd5cSStefan Behrens printk(KERN_INFO "btrfs: enabling check integrity\n"); 61321adbd5cSStefan Behrens btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); 61421adbd5cSStefan Behrens break; 61521adbd5cSStefan Behrens case Opt_check_integrity_print_mask: 6162c334e87SWang Shilong ret = match_int(&args[0], &intarg); 6172c334e87SWang Shilong if (ret) { 6182c334e87SWang Shilong goto out; 6192c334e87SWang Shilong } else if (intarg >= 0) { 62021adbd5cSStefan Behrens info->check_integrity_print_mask = intarg; 62121adbd5cSStefan Behrens printk(KERN_INFO "btrfs:" 62221adbd5cSStefan Behrens " check_integrity_print_mask 0x%x\n", 62321adbd5cSStefan Behrens info->check_integrity_print_mask); 6242c334e87SWang Shilong } else { 6252c334e87SWang Shilong ret = -EINVAL; 6262c334e87SWang Shilong goto out; 62721adbd5cSStefan Behrens } 62821adbd5cSStefan Behrens break; 62921adbd5cSStefan Behrens #else 63021adbd5cSStefan Behrens case Opt_check_integrity_including_extent_data: 63121adbd5cSStefan Behrens case Opt_check_integrity: 63221adbd5cSStefan Behrens case Opt_check_integrity_print_mask: 63321adbd5cSStefan Behrens printk(KERN_ERR "btrfs: support for check_integrity*" 63421adbd5cSStefan Behrens " not compiled in!\n"); 63521adbd5cSStefan Behrens ret = -EINVAL; 63621adbd5cSStefan Behrens goto out; 63721adbd5cSStefan Behrens #endif 6388c342930SJeff Mahoney case Opt_fatal_errors: 6398c342930SJeff Mahoney if (strcmp(args[0].from, "panic") == 0) 6408c342930SJeff Mahoney btrfs_set_opt(info->mount_opt, 6418c342930SJeff Mahoney PANIC_ON_FATAL_ERROR); 6428c342930SJeff Mahoney else if (strcmp(args[0].from, "bug") == 0) 6438c342930SJeff Mahoney btrfs_clear_opt(info->mount_opt, 6448c342930SJeff Mahoney PANIC_ON_FATAL_ERROR); 6458c342930SJeff Mahoney else { 6468c342930SJeff Mahoney ret = -EINVAL; 6478c342930SJeff Mahoney goto out; 6488c342930SJeff Mahoney } 6498c342930SJeff Mahoney break; 650*8b87dc17SDavid Sterba case Opt_commit_interval: 651*8b87dc17SDavid Sterba intarg = 0; 652*8b87dc17SDavid Sterba ret = match_int(&args[0], &intarg); 653*8b87dc17SDavid Sterba if (ret < 0) { 654*8b87dc17SDavid Sterba printk(KERN_ERR 655*8b87dc17SDavid Sterba "btrfs: invalid commit interval\n"); 656*8b87dc17SDavid Sterba ret = -EINVAL; 657*8b87dc17SDavid Sterba goto out; 658*8b87dc17SDavid Sterba } 659*8b87dc17SDavid Sterba if (intarg > 0) { 660*8b87dc17SDavid Sterba if (intarg > 300) { 661*8b87dc17SDavid Sterba printk(KERN_WARNING 662*8b87dc17SDavid Sterba "btrfs: excessive commit interval %d\n", 663*8b87dc17SDavid Sterba intarg); 664*8b87dc17SDavid Sterba } 665*8b87dc17SDavid Sterba info->commit_interval = intarg; 666*8b87dc17SDavid Sterba } else { 667*8b87dc17SDavid Sterba printk(KERN_INFO 668*8b87dc17SDavid Sterba "btrfs: using default commit interval %ds\n", 669*8b87dc17SDavid Sterba BTRFS_DEFAULT_COMMIT_INTERVAL); 670*8b87dc17SDavid Sterba info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 671*8b87dc17SDavid Sterba } 672*8b87dc17SDavid Sterba break; 673a7a3f7caSSage Weil case Opt_err: 674a7a3f7caSSage Weil printk(KERN_INFO "btrfs: unrecognized mount option " 675a7a3f7caSSage Weil "'%s'\n", p); 676a7a3f7caSSage Weil ret = -EINVAL; 677a7a3f7caSSage Weil goto out; 67895e05289SChris Mason default: 679be20aa9dSChris Mason break; 68095e05289SChris Mason } 68195e05289SChris Mason } 682a7a3f7caSSage Weil out: 68373bc1876SJosef Bacik if (!ret && btrfs_test_opt(root, SPACE_CACHE)) 68473bc1876SJosef Bacik printk(KERN_INFO "btrfs: disk space caching is enabled\n"); 685da495eccSJosef Bacik kfree(orig); 686a7a3f7caSSage Weil return ret; 687edf24abeSChristoph Hellwig } 688edf24abeSChristoph Hellwig 689edf24abeSChristoph Hellwig /* 690edf24abeSChristoph Hellwig * Parse mount options that are required early in the mount process. 691edf24abeSChristoph Hellwig * 692edf24abeSChristoph Hellwig * All other options will be parsed on much later in the mount process and 693edf24abeSChristoph Hellwig * only when we need to allocate a new super block. 694edf24abeSChristoph Hellwig */ 69597288f2cSChristoph Hellwig static int btrfs_parse_early_options(const char *options, fmode_t flags, 69673f73415SJosef Bacik void *holder, char **subvol_name, u64 *subvol_objectid, 6975e2a4b25SDavid Sterba struct btrfs_fs_devices **fs_devices) 698edf24abeSChristoph Hellwig { 699edf24abeSChristoph Hellwig substring_t args[MAX_OPT_ARGS]; 70083c8c9bdSJeff Liu char *device_name, *opts, *orig, *p; 7011493381fSWang Shilong char *num = NULL; 702edf24abeSChristoph Hellwig int error = 0; 703edf24abeSChristoph Hellwig 704edf24abeSChristoph Hellwig if (!options) 705830c4adbSJosef Bacik return 0; 706edf24abeSChristoph Hellwig 707edf24abeSChristoph Hellwig /* 708edf24abeSChristoph Hellwig * strsep changes the string, duplicate it because parse_options 709edf24abeSChristoph Hellwig * gets called twice 710edf24abeSChristoph Hellwig */ 711edf24abeSChristoph Hellwig opts = kstrdup(options, GFP_KERNEL); 712edf24abeSChristoph Hellwig if (!opts) 713edf24abeSChristoph Hellwig return -ENOMEM; 7143f3d0bc0STero Roponen orig = opts; 715edf24abeSChristoph Hellwig 716edf24abeSChristoph Hellwig while ((p = strsep(&opts, ",")) != NULL) { 717edf24abeSChristoph Hellwig int token; 718edf24abeSChristoph Hellwig if (!*p) 719edf24abeSChristoph Hellwig continue; 720edf24abeSChristoph Hellwig 721edf24abeSChristoph Hellwig token = match_token(p, tokens, args); 722edf24abeSChristoph Hellwig switch (token) { 723edf24abeSChristoph Hellwig case Opt_subvol: 724a90e8b6fSIlya Dryomov kfree(*subvol_name); 725edf24abeSChristoph Hellwig *subvol_name = match_strdup(&args[0]); 7262c334e87SWang Shilong if (!*subvol_name) { 7272c334e87SWang Shilong error = -ENOMEM; 7282c334e87SWang Shilong goto out; 7292c334e87SWang Shilong } 730edf24abeSChristoph Hellwig break; 73173f73415SJosef Bacik case Opt_subvolid: 7321493381fSWang Shilong num = match_strdup(&args[0]); 7331493381fSWang Shilong if (num) { 7341493381fSWang Shilong *subvol_objectid = memparse(num, NULL); 7351493381fSWang Shilong kfree(num); 7364849f01dSJosef Bacik /* we want the original fs_tree */ 7371493381fSWang Shilong if (!*subvol_objectid) 7384849f01dSJosef Bacik *subvol_objectid = 7394849f01dSJosef Bacik BTRFS_FS_TREE_OBJECTID; 7402c334e87SWang Shilong } else { 7412c334e87SWang Shilong error = -EINVAL; 7422c334e87SWang Shilong goto out; 7434849f01dSJosef Bacik } 74473f73415SJosef Bacik break; 745e15d0542SXin Zhong case Opt_subvolrootid: 7465e2a4b25SDavid Sterba printk(KERN_WARNING 7475e2a4b25SDavid Sterba "btrfs: 'subvolrootid' mount option is deprecated and has no effect\n"); 748e15d0542SXin Zhong break; 74943e570b0SChristoph Hellwig case Opt_device: 75083c8c9bdSJeff Liu device_name = match_strdup(&args[0]); 75183c8c9bdSJeff Liu if (!device_name) { 75283c8c9bdSJeff Liu error = -ENOMEM; 75383c8c9bdSJeff Liu goto out; 75483c8c9bdSJeff Liu } 75583c8c9bdSJeff Liu error = btrfs_scan_one_device(device_name, 75643e570b0SChristoph Hellwig flags, holder, fs_devices); 75783c8c9bdSJeff Liu kfree(device_name); 75843e570b0SChristoph Hellwig if (error) 759830c4adbSJosef Bacik goto out; 76043e570b0SChristoph Hellwig break; 761edf24abeSChristoph Hellwig default: 762edf24abeSChristoph Hellwig break; 763edf24abeSChristoph Hellwig } 764edf24abeSChristoph Hellwig } 765edf24abeSChristoph Hellwig 766edf24abeSChristoph Hellwig out: 767830c4adbSJosef Bacik kfree(orig); 768edf24abeSChristoph Hellwig return error; 76995e05289SChris Mason } 77095e05289SChris Mason 77173f73415SJosef Bacik static struct dentry *get_default_root(struct super_block *sb, 77273f73415SJosef Bacik u64 subvol_objectid) 77373f73415SJosef Bacik { 774815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 775815745cfSAl Viro struct btrfs_root *root = fs_info->tree_root; 77673f73415SJosef Bacik struct btrfs_root *new_root; 77773f73415SJosef Bacik struct btrfs_dir_item *di; 77873f73415SJosef Bacik struct btrfs_path *path; 77973f73415SJosef Bacik struct btrfs_key location; 78073f73415SJosef Bacik struct inode *inode; 78173f73415SJosef Bacik u64 dir_id; 78273f73415SJosef Bacik int new = 0; 78373f73415SJosef Bacik 78473f73415SJosef Bacik /* 78573f73415SJosef Bacik * We have a specific subvol we want to mount, just setup location and 78673f73415SJosef Bacik * go look up the root. 78773f73415SJosef Bacik */ 78873f73415SJosef Bacik if (subvol_objectid) { 78973f73415SJosef Bacik location.objectid = subvol_objectid; 79073f73415SJosef Bacik location.type = BTRFS_ROOT_ITEM_KEY; 79173f73415SJosef Bacik location.offset = (u64)-1; 79273f73415SJosef Bacik goto find_root; 79373f73415SJosef Bacik } 79473f73415SJosef Bacik 79573f73415SJosef Bacik path = btrfs_alloc_path(); 79673f73415SJosef Bacik if (!path) 79773f73415SJosef Bacik return ERR_PTR(-ENOMEM); 79873f73415SJosef Bacik path->leave_spinning = 1; 79973f73415SJosef Bacik 80073f73415SJosef Bacik /* 80173f73415SJosef Bacik * Find the "default" dir item which points to the root item that we 80273f73415SJosef Bacik * will mount by default if we haven't been given a specific subvolume 80373f73415SJosef Bacik * to mount. 80473f73415SJosef Bacik */ 805815745cfSAl Viro dir_id = btrfs_super_root_dir(fs_info->super_copy); 80673f73415SJosef Bacik di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); 807b0839166SJulia Lawall if (IS_ERR(di)) { 808b0839166SJulia Lawall btrfs_free_path(path); 809fb4f6f91SDan Carpenter return ERR_CAST(di); 810b0839166SJulia Lawall } 81173f73415SJosef Bacik if (!di) { 81273f73415SJosef Bacik /* 81373f73415SJosef Bacik * Ok the default dir item isn't there. This is weird since 81473f73415SJosef Bacik * it's always been there, but don't freak out, just try and 81573f73415SJosef Bacik * mount to root most subvolume. 81673f73415SJosef Bacik */ 81773f73415SJosef Bacik btrfs_free_path(path); 81873f73415SJosef Bacik dir_id = BTRFS_FIRST_FREE_OBJECTID; 819815745cfSAl Viro new_root = fs_info->fs_root; 82073f73415SJosef Bacik goto setup_root; 82173f73415SJosef Bacik } 82273f73415SJosef Bacik 82373f73415SJosef Bacik btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 82473f73415SJosef Bacik btrfs_free_path(path); 82573f73415SJosef Bacik 82673f73415SJosef Bacik find_root: 827815745cfSAl Viro new_root = btrfs_read_fs_root_no_name(fs_info, &location); 82873f73415SJosef Bacik if (IS_ERR(new_root)) 829d0b678cbSJulia Lawall return ERR_CAST(new_root); 83073f73415SJosef Bacik 83173f73415SJosef Bacik dir_id = btrfs_root_dirid(&new_root->root_item); 83273f73415SJosef Bacik setup_root: 83373f73415SJosef Bacik location.objectid = dir_id; 83473f73415SJosef Bacik location.type = BTRFS_INODE_ITEM_KEY; 83573f73415SJosef Bacik location.offset = 0; 83673f73415SJosef Bacik 83773f73415SJosef Bacik inode = btrfs_iget(sb, &location, new_root, &new); 8384cbd1149SDan Carpenter if (IS_ERR(inode)) 8394cbd1149SDan Carpenter return ERR_CAST(inode); 84073f73415SJosef Bacik 84173f73415SJosef Bacik /* 84273f73415SJosef Bacik * If we're just mounting the root most subvol put the inode and return 84373f73415SJosef Bacik * a reference to the dentry. We will have already gotten a reference 84473f73415SJosef Bacik * to the inode in btrfs_fill_super so we're good to go. 84573f73415SJosef Bacik */ 84673f73415SJosef Bacik if (!new && sb->s_root->d_inode == inode) { 84773f73415SJosef Bacik iput(inode); 84873f73415SJosef Bacik return dget(sb->s_root); 84973f73415SJosef Bacik } 85073f73415SJosef Bacik 851ba5b8958SJosef Bacik return d_obtain_alias(inode); 85273f73415SJosef Bacik } 85373f73415SJosef Bacik 8548a4b83ccSChris Mason static int btrfs_fill_super(struct super_block *sb, 8558a4b83ccSChris Mason struct btrfs_fs_devices *fs_devices, 8568a4b83ccSChris Mason void *data, int silent) 8572e635a27SChris Mason { 8582e635a27SChris Mason struct inode *inode; 859815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 8605d4f98a2SYan Zheng struct btrfs_key key; 86139279cc3SChris Mason int err; 8622e635a27SChris Mason 8632e635a27SChris Mason sb->s_maxbytes = MAX_LFS_FILESIZE; 8642e635a27SChris Mason sb->s_magic = BTRFS_SUPER_MAGIC; 865e20d96d6SChris Mason sb->s_op = &btrfs_super_ops; 866af53d29aSAl Viro sb->s_d_op = &btrfs_dentry_operations; 867be6e8dc0SBalaji Rao sb->s_export_op = &btrfs_export_ops; 8685103e947SJosef Bacik sb->s_xattr = btrfs_xattr_handlers; 8692e635a27SChris Mason sb->s_time_gran = 1; 8700eda294dSChris Mason #ifdef CONFIG_BTRFS_FS_POSIX_ACL 87133268eafSJosef Bacik sb->s_flags |= MS_POSIXACL; 87249cf6f45SChris Ball #endif 8730c4d2d95SJosef Bacik sb->s_flags |= MS_I_VERSION; 874ad2b2c80SAl Viro err = open_ctree(sb, fs_devices, (char *)data); 875ad2b2c80SAl Viro if (err) { 876e20d96d6SChris Mason printk("btrfs: open_ctree failed\n"); 877ad2b2c80SAl Viro return err; 878e20d96d6SChris Mason } 879b888db2bSChris Mason 8805d4f98a2SYan Zheng key.objectid = BTRFS_FIRST_FREE_OBJECTID; 8815d4f98a2SYan Zheng key.type = BTRFS_INODE_ITEM_KEY; 8825d4f98a2SYan Zheng key.offset = 0; 88398c7089cSAl Viro inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL); 8845d4f98a2SYan Zheng if (IS_ERR(inode)) { 8855d4f98a2SYan Zheng err = PTR_ERR(inode); 88639279cc3SChris Mason goto fail_close; 88739279cc3SChris Mason } 8882e635a27SChris Mason 88948fde701SAl Viro sb->s_root = d_make_root(inode); 89048fde701SAl Viro if (!sb->s_root) { 89139279cc3SChris Mason err = -ENOMEM; 89239279cc3SChris Mason goto fail_close; 8932e635a27SChris Mason } 89458176a96SJosef Bacik 8956885f308SChris Mason save_mount_options(sb, data); 89690a887c9SDan Magenheimer cleancache_init_fs(sb); 89759553edfSAl Viro sb->s_flags |= MS_ACTIVE; 8982e635a27SChris Mason return 0; 8992e635a27SChris Mason 90039279cc3SChris Mason fail_close: 901815745cfSAl Viro close_ctree(fs_info->tree_root); 902d5719762SChris Mason return err; 903d5719762SChris Mason } 904d5719762SChris Mason 9056bf13c0cSSage Weil int btrfs_sync_fs(struct super_block *sb, int wait) 906d5719762SChris Mason { 907d5719762SChris Mason struct btrfs_trans_handle *trans; 908815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 909815745cfSAl Viro struct btrfs_root *root = fs_info->tree_root; 910df2ce34cSChris Mason 9111abe9b8aSliubo trace_btrfs_sync_fs(wait); 9121abe9b8aSliubo 913d561c025SChris Mason if (!wait) { 914815745cfSAl Viro filemap_flush(fs_info->btree_inode->i_mapping); 915df2ce34cSChris Mason return 0; 916d561c025SChris Mason } 917771ed689SChris Mason 918c73e2936SJosef Bacik btrfs_wait_all_ordered_extents(fs_info, 1); 919771ed689SChris Mason 920d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 92160376ce4SJosef Bacik if (IS_ERR(trans)) { 922354aa0fbSMiao Xie /* no transaction, don't bother */ 923354aa0fbSMiao Xie if (PTR_ERR(trans) == -ENOENT) 924bd7de2c9SJosef Bacik return 0; 92598d5dc13STsutomu Itoh return PTR_ERR(trans); 92660376ce4SJosef Bacik } 927bd7de2c9SJosef Bacik return btrfs_commit_transaction(trans, root); 928d5719762SChris Mason } 929d5719762SChris Mason 93034c80b1dSAl Viro static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) 931a9572a15SEric Paris { 932815745cfSAl Viro struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); 933815745cfSAl Viro struct btrfs_root *root = info->tree_root; 934200da64eSTsutomu Itoh char *compress_type; 935a9572a15SEric Paris 936a9572a15SEric Paris if (btrfs_test_opt(root, DEGRADED)) 937a9572a15SEric Paris seq_puts(seq, ",degraded"); 938a9572a15SEric Paris if (btrfs_test_opt(root, NODATASUM)) 939a9572a15SEric Paris seq_puts(seq, ",nodatasum"); 940a9572a15SEric Paris if (btrfs_test_opt(root, NODATACOW)) 941a9572a15SEric Paris seq_puts(seq, ",nodatacow"); 942a9572a15SEric Paris if (btrfs_test_opt(root, NOBARRIER)) 943a9572a15SEric Paris seq_puts(seq, ",nobarrier"); 944a9572a15SEric Paris if (info->max_inline != 8192 * 1024) 94521380931SJoel Becker seq_printf(seq, ",max_inline=%llu", 94621380931SJoel Becker (unsigned long long)info->max_inline); 947a9572a15SEric Paris if (info->alloc_start != 0) 94821380931SJoel Becker seq_printf(seq, ",alloc_start=%llu", 94921380931SJoel Becker (unsigned long long)info->alloc_start); 950a9572a15SEric Paris if (info->thread_pool_size != min_t(unsigned long, 951a9572a15SEric Paris num_online_cpus() + 2, 8)) 952a9572a15SEric Paris seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); 953200da64eSTsutomu Itoh if (btrfs_test_opt(root, COMPRESS)) { 954200da64eSTsutomu Itoh if (info->compress_type == BTRFS_COMPRESS_ZLIB) 955200da64eSTsutomu Itoh compress_type = "zlib"; 956200da64eSTsutomu Itoh else 957200da64eSTsutomu Itoh compress_type = "lzo"; 958200da64eSTsutomu Itoh if (btrfs_test_opt(root, FORCE_COMPRESS)) 959200da64eSTsutomu Itoh seq_printf(seq, ",compress-force=%s", compress_type); 960200da64eSTsutomu Itoh else 961200da64eSTsutomu Itoh seq_printf(seq, ",compress=%s", compress_type); 962200da64eSTsutomu Itoh } 963c289811cSChris Mason if (btrfs_test_opt(root, NOSSD)) 964c289811cSChris Mason seq_puts(seq, ",nossd"); 965451d7585SChris Mason if (btrfs_test_opt(root, SSD_SPREAD)) 966451d7585SChris Mason seq_puts(seq, ",ssd_spread"); 967451d7585SChris Mason else if (btrfs_test_opt(root, SSD)) 968a9572a15SEric Paris seq_puts(seq, ",ssd"); 9693a5e1404SSage Weil if (btrfs_test_opt(root, NOTREELOG)) 9706b65c5c6SSage Weil seq_puts(seq, ",notreelog"); 971dccae999SSage Weil if (btrfs_test_opt(root, FLUSHONCOMMIT)) 9726b65c5c6SSage Weil seq_puts(seq, ",flushoncommit"); 97320a5239aSMatthew Wilcox if (btrfs_test_opt(root, DISCARD)) 97420a5239aSMatthew Wilcox seq_puts(seq, ",discard"); 975a9572a15SEric Paris if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) 976a9572a15SEric Paris seq_puts(seq, ",noacl"); 977200da64eSTsutomu Itoh if (btrfs_test_opt(root, SPACE_CACHE)) 978200da64eSTsutomu Itoh seq_puts(seq, ",space_cache"); 97973bc1876SJosef Bacik else 9808965593eSDavid Sterba seq_puts(seq, ",nospace_cache"); 981200da64eSTsutomu Itoh if (btrfs_test_opt(root, CLEAR_CACHE)) 982200da64eSTsutomu Itoh seq_puts(seq, ",clear_cache"); 983200da64eSTsutomu Itoh if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 984200da64eSTsutomu Itoh seq_puts(seq, ",user_subvol_rm_allowed"); 9850942caa3SDavid Sterba if (btrfs_test_opt(root, ENOSPC_DEBUG)) 9860942caa3SDavid Sterba seq_puts(seq, ",enospc_debug"); 9870942caa3SDavid Sterba if (btrfs_test_opt(root, AUTO_DEFRAG)) 9880942caa3SDavid Sterba seq_puts(seq, ",autodefrag"); 9890942caa3SDavid Sterba if (btrfs_test_opt(root, INODE_MAP_CACHE)) 9900942caa3SDavid Sterba seq_puts(seq, ",inode_cache"); 9919555c6c1SIlya Dryomov if (btrfs_test_opt(root, SKIP_BALANCE)) 9929555c6c1SIlya Dryomov seq_puts(seq, ",skip_balance"); 9938507d216SWang Shilong if (btrfs_test_opt(root, RECOVERY)) 9948507d216SWang Shilong seq_puts(seq, ",recovery"); 9958507d216SWang Shilong #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 9968507d216SWang Shilong if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA)) 9978507d216SWang Shilong seq_puts(seq, ",check_int_data"); 9988507d216SWang Shilong else if (btrfs_test_opt(root, CHECK_INTEGRITY)) 9998507d216SWang Shilong seq_puts(seq, ",check_int"); 10008507d216SWang Shilong if (info->check_integrity_print_mask) 10018507d216SWang Shilong seq_printf(seq, ",check_int_print_mask=%d", 10028507d216SWang Shilong info->check_integrity_print_mask); 10038507d216SWang Shilong #endif 10048507d216SWang Shilong if (info->metadata_ratio) 10058507d216SWang Shilong seq_printf(seq, ",metadata_ratio=%d", 10068507d216SWang Shilong info->metadata_ratio); 10078c342930SJeff Mahoney if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR)) 10088c342930SJeff Mahoney seq_puts(seq, ",fatal_errors=panic"); 1009*8b87dc17SDavid Sterba if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) 1010*8b87dc17SDavid Sterba seq_printf(seq, ",commit=%d", info->commit_interval); 1011a9572a15SEric Paris return 0; 1012a9572a15SEric Paris } 1013a9572a15SEric Paris 1014a061fc8dSChris Mason static int btrfs_test_super(struct super_block *s, void *data) 10152e635a27SChris Mason { 1016815745cfSAl Viro struct btrfs_fs_info *p = data; 1017815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(s); 10184b82d6e4SYan 1019815745cfSAl Viro return fs_info->fs_devices == p->fs_devices; 10204b82d6e4SYan } 10214b82d6e4SYan 1022450ba0eaSJosef Bacik static int btrfs_set_super(struct super_block *s, void *data) 1023450ba0eaSJosef Bacik { 10246de1d09dSAl Viro int err = set_anon_super(s, data); 10256de1d09dSAl Viro if (!err) 1026450ba0eaSJosef Bacik s->s_fs_info = data; 10276de1d09dSAl Viro return err; 1028450ba0eaSJosef Bacik } 1029450ba0eaSJosef Bacik 1030830c4adbSJosef Bacik /* 1031f9d9ef62SDavid Sterba * subvolumes are identified by ino 256 1032f9d9ef62SDavid Sterba */ 1033f9d9ef62SDavid Sterba static inline int is_subvolume_inode(struct inode *inode) 1034f9d9ef62SDavid Sterba { 1035f9d9ef62SDavid Sterba if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) 1036f9d9ef62SDavid Sterba return 1; 1037f9d9ef62SDavid Sterba return 0; 1038f9d9ef62SDavid Sterba } 1039f9d9ef62SDavid Sterba 1040f9d9ef62SDavid Sterba /* 1041830c4adbSJosef Bacik * This will strip out the subvol=%s argument for an argument string and add 1042830c4adbSJosef Bacik * subvolid=0 to make sure we get the actual tree root for path walking to the 1043830c4adbSJosef Bacik * subvol we want. 1044830c4adbSJosef Bacik */ 1045830c4adbSJosef Bacik static char *setup_root_args(char *args) 1046830c4adbSJosef Bacik { 1047f60d16a8SJim Meyering unsigned len = strlen(args) + 2 + 1; 1048f60d16a8SJim Meyering char *src, *dst, *buf; 1049830c4adbSJosef Bacik 1050830c4adbSJosef Bacik /* 1051f60d16a8SJim Meyering * We need the same args as before, but with this substitution: 1052f60d16a8SJim Meyering * s!subvol=[^,]+!subvolid=0! 1053830c4adbSJosef Bacik * 1054f60d16a8SJim Meyering * Since the replacement string is up to 2 bytes longer than the 1055f60d16a8SJim Meyering * original, allocate strlen(args) + 2 + 1 bytes. 1056830c4adbSJosef Bacik */ 1057830c4adbSJosef Bacik 1058f60d16a8SJim Meyering src = strstr(args, "subvol="); 1059830c4adbSJosef Bacik /* This shouldn't happen, but just in case.. */ 1060f60d16a8SJim Meyering if (!src) 1061830c4adbSJosef Bacik return NULL; 1062f60d16a8SJim Meyering 1063f60d16a8SJim Meyering buf = dst = kmalloc(len, GFP_NOFS); 1064f60d16a8SJim Meyering if (!buf) 1065f60d16a8SJim Meyering return NULL; 1066830c4adbSJosef Bacik 1067830c4adbSJosef Bacik /* 1068f60d16a8SJim Meyering * If the subvol= arg is not at the start of the string, 1069f60d16a8SJim Meyering * copy whatever precedes it into buf. 1070830c4adbSJosef Bacik */ 1071f60d16a8SJim Meyering if (src != args) { 1072f60d16a8SJim Meyering *src++ = '\0'; 1073f60d16a8SJim Meyering strcpy(buf, args); 1074f60d16a8SJim Meyering dst += strlen(args); 1075830c4adbSJosef Bacik } 1076830c4adbSJosef Bacik 1077f60d16a8SJim Meyering strcpy(dst, "subvolid=0"); 1078f60d16a8SJim Meyering dst += strlen("subvolid=0"); 1079830c4adbSJosef Bacik 1080830c4adbSJosef Bacik /* 1081f60d16a8SJim Meyering * If there is a "," after the original subvol=... string, 1082f60d16a8SJim Meyering * copy that suffix into our buffer. Otherwise, we're done. 1083830c4adbSJosef Bacik */ 1084f60d16a8SJim Meyering src = strchr(src, ','); 1085f60d16a8SJim Meyering if (src) 1086f60d16a8SJim Meyering strcpy(dst, src); 1087830c4adbSJosef Bacik 1088f60d16a8SJim Meyering return buf; 1089830c4adbSJosef Bacik } 1090830c4adbSJosef Bacik 1091830c4adbSJosef Bacik static struct dentry *mount_subvol(const char *subvol_name, int flags, 1092830c4adbSJosef Bacik const char *device_name, char *data) 1093830c4adbSJosef Bacik { 1094830c4adbSJosef Bacik struct dentry *root; 1095830c4adbSJosef Bacik struct vfsmount *mnt; 1096830c4adbSJosef Bacik char *newargs; 1097830c4adbSJosef Bacik 1098830c4adbSJosef Bacik newargs = setup_root_args(data); 1099830c4adbSJosef Bacik if (!newargs) 1100830c4adbSJosef Bacik return ERR_PTR(-ENOMEM); 1101830c4adbSJosef Bacik mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, 1102830c4adbSJosef Bacik newargs); 1103830c4adbSJosef Bacik kfree(newargs); 1104830c4adbSJosef Bacik if (IS_ERR(mnt)) 1105830c4adbSJosef Bacik return ERR_CAST(mnt); 1106830c4adbSJosef Bacik 1107ea441d11SAl Viro root = mount_subtree(mnt, subvol_name); 1108830c4adbSJosef Bacik 1109ea441d11SAl Viro if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) { 1110ea441d11SAl Viro struct super_block *s = root->d_sb; 1111ea441d11SAl Viro dput(root); 1112ea441d11SAl Viro root = ERR_PTR(-EINVAL); 1113ea441d11SAl Viro deactivate_locked_super(s); 1114f9d9ef62SDavid Sterba printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n", 1115f9d9ef62SDavid Sterba subvol_name); 1116f9d9ef62SDavid Sterba } 1117f9d9ef62SDavid Sterba 1118830c4adbSJosef Bacik return root; 1119830c4adbSJosef Bacik } 1120450ba0eaSJosef Bacik 1121edf24abeSChristoph Hellwig /* 1122edf24abeSChristoph Hellwig * Find a superblock for the given device / mount point. 1123edf24abeSChristoph Hellwig * 1124edf24abeSChristoph Hellwig * Note: This is based on get_sb_bdev from fs/super.c with a few additions 1125edf24abeSChristoph Hellwig * for multiple device setup. Make sure to keep it in sync. 1126edf24abeSChristoph Hellwig */ 1127061dbc6bSAl Viro static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, 1128306e16ceSDavid Sterba const char *device_name, void *data) 11294b82d6e4SYan { 11304b82d6e4SYan struct block_device *bdev = NULL; 11314b82d6e4SYan struct super_block *s; 11324b82d6e4SYan struct dentry *root; 11338a4b83ccSChris Mason struct btrfs_fs_devices *fs_devices = NULL; 1134450ba0eaSJosef Bacik struct btrfs_fs_info *fs_info = NULL; 113597288f2cSChristoph Hellwig fmode_t mode = FMODE_READ; 113673f73415SJosef Bacik char *subvol_name = NULL; 113773f73415SJosef Bacik u64 subvol_objectid = 0; 11384b82d6e4SYan int error = 0; 11394b82d6e4SYan 114097288f2cSChristoph Hellwig if (!(flags & MS_RDONLY)) 114197288f2cSChristoph Hellwig mode |= FMODE_WRITE; 114297288f2cSChristoph Hellwig 114397288f2cSChristoph Hellwig error = btrfs_parse_early_options(data, mode, fs_type, 114473f73415SJosef Bacik &subvol_name, &subvol_objectid, 11455e2a4b25SDavid Sterba &fs_devices); 1146f23c8af8SIlya Dryomov if (error) { 1147f23c8af8SIlya Dryomov kfree(subvol_name); 1148061dbc6bSAl Viro return ERR_PTR(error); 1149f23c8af8SIlya Dryomov } 1150edf24abeSChristoph Hellwig 1151830c4adbSJosef Bacik if (subvol_name) { 1152830c4adbSJosef Bacik root = mount_subvol(subvol_name, flags, device_name, data); 1153830c4adbSJosef Bacik kfree(subvol_name); 1154830c4adbSJosef Bacik return root; 1155830c4adbSJosef Bacik } 1156830c4adbSJosef Bacik 1157306e16ceSDavid Sterba error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices); 11588a4b83ccSChris Mason if (error) 1159830c4adbSJosef Bacik return ERR_PTR(error); 11604b82d6e4SYan 1161450ba0eaSJosef Bacik /* 1162450ba0eaSJosef Bacik * Setup a dummy root and fs_info for test/set super. This is because 1163450ba0eaSJosef Bacik * we don't actually fill this stuff out until open_ctree, but we need 1164450ba0eaSJosef Bacik * it for searching for existing supers, so this lets us do that and 1165450ba0eaSJosef Bacik * then open_ctree will properly initialize everything later. 1166450ba0eaSJosef Bacik */ 1167450ba0eaSJosef Bacik fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); 116804d21a24SIlya Dryomov if (!fs_info) 116904d21a24SIlya Dryomov return ERR_PTR(-ENOMEM); 117004d21a24SIlya Dryomov 1171450ba0eaSJosef Bacik fs_info->fs_devices = fs_devices; 1172450ba0eaSJosef Bacik 11736c41761fSDavid Sterba fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); 11746c41761fSDavid Sterba fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); 11756c41761fSDavid Sterba if (!fs_info->super_copy || !fs_info->super_for_commit) { 11766c41761fSDavid Sterba error = -ENOMEM; 117704d21a24SIlya Dryomov goto error_fs_info; 117804d21a24SIlya Dryomov } 117904d21a24SIlya Dryomov 118004d21a24SIlya Dryomov error = btrfs_open_devices(fs_devices, mode, fs_type); 118104d21a24SIlya Dryomov if (error) 118204d21a24SIlya Dryomov goto error_fs_info; 118304d21a24SIlya Dryomov 118404d21a24SIlya Dryomov if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { 118504d21a24SIlya Dryomov error = -EACCES; 11866c41761fSDavid Sterba goto error_close_devices; 11876c41761fSDavid Sterba } 11886c41761fSDavid Sterba 1189dfe25020SChris Mason bdev = fs_devices->latest_bdev; 11909249e17fSDavid Howells s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC, 11919249e17fSDavid Howells fs_info); 1192830c4adbSJosef Bacik if (IS_ERR(s)) { 1193830c4adbSJosef Bacik error = PTR_ERR(s); 1194830c4adbSJosef Bacik goto error_close_devices; 1195830c4adbSJosef Bacik } 11964b82d6e4SYan 11974b82d6e4SYan if (s->s_root) { 11982b82032cSYan Zheng btrfs_close_devices(fs_devices); 11996c41761fSDavid Sterba free_fs_info(fs_info); 120059553edfSAl Viro if ((flags ^ s->s_flags) & MS_RDONLY) 120159553edfSAl Viro error = -EBUSY; 12024b82d6e4SYan } else { 12034b82d6e4SYan char b[BDEVNAME_SIZE]; 12044b82d6e4SYan 12054b82d6e4SYan strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); 1206815745cfSAl Viro btrfs_sb(s)->bdev_holder = fs_type; 12078a4b83ccSChris Mason error = btrfs_fill_super(s, fs_devices, data, 12088a4b83ccSChris Mason flags & MS_SILENT ? 1 : 0); 12094b82d6e4SYan } 12104b82d6e4SYan 121159553edfSAl Viro root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error); 121259553edfSAl Viro if (IS_ERR(root)) 1213e15d0542SXin Zhong deactivate_locked_super(s); 12144b82d6e4SYan 1215061dbc6bSAl Viro return root; 12164b82d6e4SYan 1217c146afadSYan Zheng error_close_devices: 12188a4b83ccSChris Mason btrfs_close_devices(fs_devices); 121904d21a24SIlya Dryomov error_fs_info: 12206c41761fSDavid Sterba free_fs_info(fs_info); 1221061dbc6bSAl Viro return ERR_PTR(error); 12224b82d6e4SYan } 12232e635a27SChris Mason 12240d2450abSSergei Trofimovich static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit) 12250d2450abSSergei Trofimovich { 12260d2450abSSergei Trofimovich spin_lock_irq(&workers->lock); 12270d2450abSSergei Trofimovich workers->max_workers = new_limit; 12280d2450abSSergei Trofimovich spin_unlock_irq(&workers->lock); 12290d2450abSSergei Trofimovich } 12300d2450abSSergei Trofimovich 12310d2450abSSergei Trofimovich static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, 12320d2450abSSergei Trofimovich int new_pool_size, int old_pool_size) 12330d2450abSSergei Trofimovich { 12340d2450abSSergei Trofimovich if (new_pool_size == old_pool_size) 12350d2450abSSergei Trofimovich return; 12360d2450abSSergei Trofimovich 12370d2450abSSergei Trofimovich fs_info->thread_pool_size = new_pool_size; 12380d2450abSSergei Trofimovich 12390d2450abSSergei Trofimovich printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n", 12400d2450abSSergei Trofimovich old_pool_size, new_pool_size); 12410d2450abSSergei Trofimovich 12420d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size); 12430d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->workers, new_pool_size); 12440d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size); 12450d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size); 12460d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size); 12470d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size); 12480d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size); 12490d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size); 12500d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size); 12510d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size); 12520d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size); 12530d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size); 12540d2450abSSergei Trofimovich btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size); 1255ff023aacSStefan Behrens btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers, 1256ff023aacSStefan Behrens new_pool_size); 12570d2450abSSergei Trofimovich } 12580d2450abSSergei Trofimovich 1259f42a34b2SMiao Xie static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info) 1260dc81cdc5SMiao Xie { 1261dc81cdc5SMiao Xie set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1262f42a34b2SMiao Xie } 1263dc81cdc5SMiao Xie 1264f42a34b2SMiao Xie static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, 1265f42a34b2SMiao Xie unsigned long old_opts, int flags) 1266f42a34b2SMiao Xie { 1267dc81cdc5SMiao Xie if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1268dc81cdc5SMiao Xie (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1269dc81cdc5SMiao Xie (flags & MS_RDONLY))) { 1270dc81cdc5SMiao Xie /* wait for any defraggers to finish */ 1271dc81cdc5SMiao Xie wait_event(fs_info->transaction_wait, 1272dc81cdc5SMiao Xie (atomic_read(&fs_info->defrag_running) == 0)); 1273dc81cdc5SMiao Xie if (flags & MS_RDONLY) 1274dc81cdc5SMiao Xie sync_filesystem(fs_info->sb); 1275dc81cdc5SMiao Xie } 1276dc81cdc5SMiao Xie } 1277dc81cdc5SMiao Xie 1278dc81cdc5SMiao Xie static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, 1279dc81cdc5SMiao Xie unsigned long old_opts) 1280dc81cdc5SMiao Xie { 1281dc81cdc5SMiao Xie /* 1282dc81cdc5SMiao Xie * We need cleanup all defragable inodes if the autodefragment is 1283dc81cdc5SMiao Xie * close or the fs is R/O. 1284dc81cdc5SMiao Xie */ 1285dc81cdc5SMiao Xie if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1286dc81cdc5SMiao Xie (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1287dc81cdc5SMiao Xie (fs_info->sb->s_flags & MS_RDONLY))) { 1288dc81cdc5SMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 1289dc81cdc5SMiao Xie } 1290dc81cdc5SMiao Xie 1291dc81cdc5SMiao Xie clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1292dc81cdc5SMiao Xie } 1293dc81cdc5SMiao Xie 1294c146afadSYan Zheng static int btrfs_remount(struct super_block *sb, int *flags, char *data) 1295c146afadSYan Zheng { 1296815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1297815745cfSAl Viro struct btrfs_root *root = fs_info->tree_root; 129849b25e05SJeff Mahoney unsigned old_flags = sb->s_flags; 129949b25e05SJeff Mahoney unsigned long old_opts = fs_info->mount_opt; 130049b25e05SJeff Mahoney unsigned long old_compress_type = fs_info->compress_type; 130149b25e05SJeff Mahoney u64 old_max_inline = fs_info->max_inline; 130249b25e05SJeff Mahoney u64 old_alloc_start = fs_info->alloc_start; 130349b25e05SJeff Mahoney int old_thread_pool_size = fs_info->thread_pool_size; 130449b25e05SJeff Mahoney unsigned int old_metadata_ratio = fs_info->metadata_ratio; 1305c146afadSYan Zheng int ret; 1306c146afadSYan Zheng 1307f42a34b2SMiao Xie btrfs_remount_prepare(fs_info); 1308dc81cdc5SMiao Xie 1309b288052eSChris Mason ret = btrfs_parse_options(root, data); 131049b25e05SJeff Mahoney if (ret) { 131149b25e05SJeff Mahoney ret = -EINVAL; 131249b25e05SJeff Mahoney goto restore; 131349b25e05SJeff Mahoney } 1314b288052eSChris Mason 1315f42a34b2SMiao Xie btrfs_remount_begin(fs_info, old_opts, *flags); 13160d2450abSSergei Trofimovich btrfs_resize_thread_pool(fs_info, 13170d2450abSSergei Trofimovich fs_info->thread_pool_size, old_thread_pool_size); 13180d2450abSSergei Trofimovich 1319c146afadSYan Zheng if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) 1320dc81cdc5SMiao Xie goto out; 1321c146afadSYan Zheng 1322c146afadSYan Zheng if (*flags & MS_RDONLY) { 13238dabb742SStefan Behrens /* 13248dabb742SStefan Behrens * this also happens on 'umount -rf' or on shutdown, when 13258dabb742SStefan Behrens * the filesystem is busy. 13268dabb742SStefan Behrens */ 1327c146afadSYan Zheng sb->s_flags |= MS_RDONLY; 1328c146afadSYan Zheng 13298dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 13308dabb742SStefan Behrens btrfs_scrub_cancel(fs_info); 1331061594efSMiao Xie btrfs_pause_balance(fs_info); 13328dabb742SStefan Behrens 1333c146afadSYan Zheng ret = btrfs_commit_super(root); 133449b25e05SJeff Mahoney if (ret) 133549b25e05SJeff Mahoney goto restore; 1336c146afadSYan Zheng } else { 13378a3db184SSergei Trofimovich if (fs_info->fs_devices->rw_devices == 0) { 133849b25e05SJeff Mahoney ret = -EACCES; 133949b25e05SJeff Mahoney goto restore; 13408a3db184SSergei Trofimovich } 13412b82032cSYan Zheng 1342292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 1343292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 1344292fd7fcSStefan Behrens !(*flags & MS_RDONLY)) { 1345292fd7fcSStefan Behrens printk(KERN_WARNING 1346292fd7fcSStefan Behrens "Btrfs: too many missing devices, writeable remount is not allowed\n"); 1347292fd7fcSStefan Behrens ret = -EACCES; 1348292fd7fcSStefan Behrens goto restore; 1349292fd7fcSStefan Behrens } 1350292fd7fcSStefan Behrens 13518a3db184SSergei Trofimovich if (btrfs_super_log_root(fs_info->super_copy) != 0) { 135249b25e05SJeff Mahoney ret = -EINVAL; 135349b25e05SJeff Mahoney goto restore; 13548a3db184SSergei Trofimovich } 1355c146afadSYan Zheng 1356815745cfSAl Viro ret = btrfs_cleanup_fs_roots(fs_info); 135749b25e05SJeff Mahoney if (ret) 135849b25e05SJeff Mahoney goto restore; 1359c146afadSYan Zheng 1360d68fc57bSYan, Zheng /* recover relocation */ 1361d68fc57bSYan, Zheng ret = btrfs_recover_relocation(root); 136249b25e05SJeff Mahoney if (ret) 136349b25e05SJeff Mahoney goto restore; 1364c146afadSYan Zheng 13652b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 13662b6ba629SIlya Dryomov if (ret) 13672b6ba629SIlya Dryomov goto restore; 13682b6ba629SIlya Dryomov 13698dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 13708dabb742SStefan Behrens if (ret) { 13718dabb742SStefan Behrens pr_warn("btrfs: failed to resume dev_replace\n"); 13728dabb742SStefan Behrens goto restore; 13738dabb742SStefan Behrens } 1374c146afadSYan Zheng sb->s_flags &= ~MS_RDONLY; 1375c146afadSYan Zheng } 1376dc81cdc5SMiao Xie out: 1377dc81cdc5SMiao Xie btrfs_remount_cleanup(fs_info, old_opts); 1378c146afadSYan Zheng return 0; 137949b25e05SJeff Mahoney 138049b25e05SJeff Mahoney restore: 138149b25e05SJeff Mahoney /* We've hit an error - don't reset MS_RDONLY */ 138249b25e05SJeff Mahoney if (sb->s_flags & MS_RDONLY) 138349b25e05SJeff Mahoney old_flags |= MS_RDONLY; 138449b25e05SJeff Mahoney sb->s_flags = old_flags; 138549b25e05SJeff Mahoney fs_info->mount_opt = old_opts; 138649b25e05SJeff Mahoney fs_info->compress_type = old_compress_type; 138749b25e05SJeff Mahoney fs_info->max_inline = old_max_inline; 1388c018daecSMiao Xie mutex_lock(&fs_info->chunk_mutex); 138949b25e05SJeff Mahoney fs_info->alloc_start = old_alloc_start; 1390c018daecSMiao Xie mutex_unlock(&fs_info->chunk_mutex); 13910d2450abSSergei Trofimovich btrfs_resize_thread_pool(fs_info, 13920d2450abSSergei Trofimovich old_thread_pool_size, fs_info->thread_pool_size); 139349b25e05SJeff Mahoney fs_info->metadata_ratio = old_metadata_ratio; 1394dc81cdc5SMiao Xie btrfs_remount_cleanup(fs_info, old_opts); 139549b25e05SJeff Mahoney return ret; 1396c146afadSYan Zheng } 1397c146afadSYan Zheng 1398bcd53741SArne Jansen /* Used to sort the devices by max_avail(descending sort) */ 1399bcd53741SArne Jansen static int btrfs_cmp_device_free_bytes(const void *dev_info1, 1400bcd53741SArne Jansen const void *dev_info2) 1401bcd53741SArne Jansen { 1402bcd53741SArne Jansen if (((struct btrfs_device_info *)dev_info1)->max_avail > 1403bcd53741SArne Jansen ((struct btrfs_device_info *)dev_info2)->max_avail) 1404bcd53741SArne Jansen return -1; 1405bcd53741SArne Jansen else if (((struct btrfs_device_info *)dev_info1)->max_avail < 1406bcd53741SArne Jansen ((struct btrfs_device_info *)dev_info2)->max_avail) 1407bcd53741SArne Jansen return 1; 1408bcd53741SArne Jansen else 1409bcd53741SArne Jansen return 0; 1410bcd53741SArne Jansen } 1411bcd53741SArne Jansen 1412bcd53741SArne Jansen /* 1413bcd53741SArne Jansen * sort the devices by max_avail, in which max free extent size of each device 1414bcd53741SArne Jansen * is stored.(Descending Sort) 1415bcd53741SArne Jansen */ 1416bcd53741SArne Jansen static inline void btrfs_descending_sort_devices( 1417bcd53741SArne Jansen struct btrfs_device_info *devices, 1418bcd53741SArne Jansen size_t nr_devices) 1419bcd53741SArne Jansen { 1420bcd53741SArne Jansen sort(devices, nr_devices, sizeof(struct btrfs_device_info), 1421bcd53741SArne Jansen btrfs_cmp_device_free_bytes, NULL); 1422bcd53741SArne Jansen } 1423bcd53741SArne Jansen 14246d07bcecSMiao Xie /* 14256d07bcecSMiao Xie * The helper to calc the free space on the devices that can be used to store 14266d07bcecSMiao Xie * file data. 14276d07bcecSMiao Xie */ 14286d07bcecSMiao Xie static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) 14296d07bcecSMiao Xie { 14306d07bcecSMiao Xie struct btrfs_fs_info *fs_info = root->fs_info; 14316d07bcecSMiao Xie struct btrfs_device_info *devices_info; 14326d07bcecSMiao Xie struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 14336d07bcecSMiao Xie struct btrfs_device *device; 14346d07bcecSMiao Xie u64 skip_space; 14356d07bcecSMiao Xie u64 type; 14366d07bcecSMiao Xie u64 avail_space; 14376d07bcecSMiao Xie u64 used_space; 14386d07bcecSMiao Xie u64 min_stripe_size; 143939fb26c3SMiao Xie int min_stripes = 1, num_stripes = 1; 14406d07bcecSMiao Xie int i = 0, nr_devices; 14416d07bcecSMiao Xie int ret; 14426d07bcecSMiao Xie 1443b772a86eSLi Zefan nr_devices = fs_info->fs_devices->open_devices; 14446d07bcecSMiao Xie BUG_ON(!nr_devices); 14456d07bcecSMiao Xie 14466d07bcecSMiao Xie devices_info = kmalloc(sizeof(*devices_info) * nr_devices, 14476d07bcecSMiao Xie GFP_NOFS); 14486d07bcecSMiao Xie if (!devices_info) 14496d07bcecSMiao Xie return -ENOMEM; 14506d07bcecSMiao Xie 14516d07bcecSMiao Xie /* calc min stripe number for data space alloction */ 14526d07bcecSMiao Xie type = btrfs_get_alloc_profile(root, 1); 145339fb26c3SMiao Xie if (type & BTRFS_BLOCK_GROUP_RAID0) { 14546d07bcecSMiao Xie min_stripes = 2; 145539fb26c3SMiao Xie num_stripes = nr_devices; 145639fb26c3SMiao Xie } else if (type & BTRFS_BLOCK_GROUP_RAID1) { 14576d07bcecSMiao Xie min_stripes = 2; 145839fb26c3SMiao Xie num_stripes = 2; 145939fb26c3SMiao Xie } else if (type & BTRFS_BLOCK_GROUP_RAID10) { 14606d07bcecSMiao Xie min_stripes = 4; 146139fb26c3SMiao Xie num_stripes = 4; 146239fb26c3SMiao Xie } 14636d07bcecSMiao Xie 14646d07bcecSMiao Xie if (type & BTRFS_BLOCK_GROUP_DUP) 14656d07bcecSMiao Xie min_stripe_size = 2 * BTRFS_STRIPE_LEN; 14666d07bcecSMiao Xie else 14676d07bcecSMiao Xie min_stripe_size = BTRFS_STRIPE_LEN; 14686d07bcecSMiao Xie 1469b772a86eSLi Zefan list_for_each_entry(device, &fs_devices->devices, dev_list) { 147063a212abSStefan Behrens if (!device->in_fs_metadata || !device->bdev || 147163a212abSStefan Behrens device->is_tgtdev_for_dev_replace) 14726d07bcecSMiao Xie continue; 14736d07bcecSMiao Xie 14746d07bcecSMiao Xie avail_space = device->total_bytes - device->bytes_used; 14756d07bcecSMiao Xie 14766d07bcecSMiao Xie /* align with stripe_len */ 14776d07bcecSMiao Xie do_div(avail_space, BTRFS_STRIPE_LEN); 14786d07bcecSMiao Xie avail_space *= BTRFS_STRIPE_LEN; 14796d07bcecSMiao Xie 14806d07bcecSMiao Xie /* 14816d07bcecSMiao Xie * In order to avoid overwritting the superblock on the drive, 14826d07bcecSMiao Xie * btrfs starts at an offset of at least 1MB when doing chunk 14836d07bcecSMiao Xie * allocation. 14846d07bcecSMiao Xie */ 14856d07bcecSMiao Xie skip_space = 1024 * 1024; 14866d07bcecSMiao Xie 14876d07bcecSMiao Xie /* user can set the offset in fs_info->alloc_start. */ 14886d07bcecSMiao Xie if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= 14896d07bcecSMiao Xie device->total_bytes) 14906d07bcecSMiao Xie skip_space = max(fs_info->alloc_start, skip_space); 14916d07bcecSMiao Xie 14926d07bcecSMiao Xie /* 14936d07bcecSMiao Xie * btrfs can not use the free space in [0, skip_space - 1], 14946d07bcecSMiao Xie * we must subtract it from the total. In order to implement 14956d07bcecSMiao Xie * it, we account the used space in this range first. 14966d07bcecSMiao Xie */ 14976d07bcecSMiao Xie ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, 14986d07bcecSMiao Xie &used_space); 14996d07bcecSMiao Xie if (ret) { 15006d07bcecSMiao Xie kfree(devices_info); 15016d07bcecSMiao Xie return ret; 15026d07bcecSMiao Xie } 15036d07bcecSMiao Xie 15046d07bcecSMiao Xie /* calc the free space in [0, skip_space - 1] */ 15056d07bcecSMiao Xie skip_space -= used_space; 15066d07bcecSMiao Xie 15076d07bcecSMiao Xie /* 15086d07bcecSMiao Xie * we can use the free space in [0, skip_space - 1], subtract 15096d07bcecSMiao Xie * it from the total. 15106d07bcecSMiao Xie */ 15116d07bcecSMiao Xie if (avail_space && avail_space >= skip_space) 15126d07bcecSMiao Xie avail_space -= skip_space; 15136d07bcecSMiao Xie else 15146d07bcecSMiao Xie avail_space = 0; 15156d07bcecSMiao Xie 15166d07bcecSMiao Xie if (avail_space < min_stripe_size) 15176d07bcecSMiao Xie continue; 15186d07bcecSMiao Xie 15196d07bcecSMiao Xie devices_info[i].dev = device; 15206d07bcecSMiao Xie devices_info[i].max_avail = avail_space; 15216d07bcecSMiao Xie 15226d07bcecSMiao Xie i++; 15236d07bcecSMiao Xie } 15246d07bcecSMiao Xie 15256d07bcecSMiao Xie nr_devices = i; 15266d07bcecSMiao Xie 15276d07bcecSMiao Xie btrfs_descending_sort_devices(devices_info, nr_devices); 15286d07bcecSMiao Xie 15296d07bcecSMiao Xie i = nr_devices - 1; 15306d07bcecSMiao Xie avail_space = 0; 15316d07bcecSMiao Xie while (nr_devices >= min_stripes) { 153239fb26c3SMiao Xie if (num_stripes > nr_devices) 153339fb26c3SMiao Xie num_stripes = nr_devices; 153439fb26c3SMiao Xie 15356d07bcecSMiao Xie if (devices_info[i].max_avail >= min_stripe_size) { 15366d07bcecSMiao Xie int j; 15376d07bcecSMiao Xie u64 alloc_size; 15386d07bcecSMiao Xie 153939fb26c3SMiao Xie avail_space += devices_info[i].max_avail * num_stripes; 15406d07bcecSMiao Xie alloc_size = devices_info[i].max_avail; 154139fb26c3SMiao Xie for (j = i + 1 - num_stripes; j <= i; j++) 15426d07bcecSMiao Xie devices_info[j].max_avail -= alloc_size; 15436d07bcecSMiao Xie } 15446d07bcecSMiao Xie i--; 15456d07bcecSMiao Xie nr_devices--; 15466d07bcecSMiao Xie } 15476d07bcecSMiao Xie 15486d07bcecSMiao Xie kfree(devices_info); 15496d07bcecSMiao Xie *free_bytes = avail_space; 15506d07bcecSMiao Xie return 0; 15516d07bcecSMiao Xie } 15526d07bcecSMiao Xie 15538fd17795SChris Mason static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 15548fd17795SChris Mason { 1555815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); 1556815745cfSAl Viro struct btrfs_super_block *disk_super = fs_info->super_copy; 1557815745cfSAl Viro struct list_head *head = &fs_info->space_info; 1558bd4d1088SJosef Bacik struct btrfs_space_info *found; 1559bd4d1088SJosef Bacik u64 total_used = 0; 15606d07bcecSMiao Xie u64 total_free_data = 0; 1561db94535dSChris Mason int bits = dentry->d_sb->s_blocksize_bits; 1562815745cfSAl Viro __be32 *fsid = (__be32 *)fs_info->fsid; 15636d07bcecSMiao Xie int ret; 15648fd17795SChris Mason 15656d07bcecSMiao Xie /* holding chunk_muext to avoid allocating new chunks */ 1566815745cfSAl Viro mutex_lock(&fs_info->chunk_mutex); 1567bd4d1088SJosef Bacik rcu_read_lock(); 156889a55897SJosef Bacik list_for_each_entry_rcu(found, head, list) { 15696d07bcecSMiao Xie if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 15706d07bcecSMiao Xie total_free_data += found->disk_total - found->disk_used; 15716d07bcecSMiao Xie total_free_data -= 15726d07bcecSMiao Xie btrfs_account_ro_block_groups_free_space(found); 15736d07bcecSMiao Xie } 15746d07bcecSMiao Xie 1575b742bb82SYan, Zheng total_used += found->disk_used; 157689a55897SJosef Bacik } 1577bd4d1088SJosef Bacik rcu_read_unlock(); 1578bd4d1088SJosef Bacik 15798fd17795SChris Mason buf->f_namelen = BTRFS_NAME_LEN; 1580db94535dSChris Mason buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; 1581bd4d1088SJosef Bacik buf->f_bfree = buf->f_blocks - (total_used >> bits); 15828fd17795SChris Mason buf->f_bsize = dentry->d_sb->s_blocksize; 15838fd17795SChris Mason buf->f_type = BTRFS_SUPER_MAGIC; 15846d07bcecSMiao Xie buf->f_bavail = total_free_data; 1585815745cfSAl Viro ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data); 15866d07bcecSMiao Xie if (ret) { 1587815745cfSAl Viro mutex_unlock(&fs_info->chunk_mutex); 15886d07bcecSMiao Xie return ret; 15896d07bcecSMiao Xie } 15906d07bcecSMiao Xie buf->f_bavail += total_free_data; 15916d07bcecSMiao Xie buf->f_bavail = buf->f_bavail >> bits; 1592815745cfSAl Viro mutex_unlock(&fs_info->chunk_mutex); 1593d397712bSChris Mason 15949d03632eSDavid Woodhouse /* We treat it as constant endianness (it doesn't matter _which_) 15959d03632eSDavid Woodhouse because we want the fsid to come out the same whether mounted 15969d03632eSDavid Woodhouse on a big-endian or little-endian host */ 15979d03632eSDavid Woodhouse buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 15989d03632eSDavid Woodhouse buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 159932d48fa1SDavid Woodhouse /* Mask in the root object ID too, to disambiguate subvols */ 160032d48fa1SDavid Woodhouse buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; 160132d48fa1SDavid Woodhouse buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; 160232d48fa1SDavid Woodhouse 16038fd17795SChris Mason return 0; 16048fd17795SChris Mason } 1605b5133862SChris Mason 1606aea52e19SAl Viro static void btrfs_kill_super(struct super_block *sb) 1607aea52e19SAl Viro { 1608815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1609aea52e19SAl Viro kill_anon_super(sb); 1610aea52e19SAl Viro free_fs_info(fs_info); 1611aea52e19SAl Viro } 1612aea52e19SAl Viro 16132e635a27SChris Mason static struct file_system_type btrfs_fs_type = { 16142e635a27SChris Mason .owner = THIS_MODULE, 16152e635a27SChris Mason .name = "btrfs", 1616061dbc6bSAl Viro .mount = btrfs_mount, 1617aea52e19SAl Viro .kill_sb = btrfs_kill_super, 16182e635a27SChris Mason .fs_flags = FS_REQUIRES_DEV, 16192e635a27SChris Mason }; 16207f78e035SEric W. Biederman MODULE_ALIAS_FS("btrfs"); 1621a9218f6bSChris Mason 1622d352ac68SChris Mason /* 1623d352ac68SChris Mason * used by btrfsctl to scan devices when no FS is mounted 1624d352ac68SChris Mason */ 16258a4b83ccSChris Mason static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 16268a4b83ccSChris Mason unsigned long arg) 16278a4b83ccSChris Mason { 16288a4b83ccSChris Mason struct btrfs_ioctl_vol_args *vol; 16298a4b83ccSChris Mason struct btrfs_fs_devices *fs_devices; 1630c071fcfdSChris Mason int ret = -ENOTTY; 16318a4b83ccSChris Mason 1632e441d54dSChris Mason if (!capable(CAP_SYS_ADMIN)) 1633e441d54dSChris Mason return -EPERM; 1634e441d54dSChris Mason 1635dae7b665SLi Zefan vol = memdup_user((void __user *)arg, sizeof(*vol)); 1636dae7b665SLi Zefan if (IS_ERR(vol)) 1637dae7b665SLi Zefan return PTR_ERR(vol); 1638c071fcfdSChris Mason 16398a4b83ccSChris Mason switch (cmd) { 16408a4b83ccSChris Mason case BTRFS_IOC_SCAN_DEV: 164197288f2cSChristoph Hellwig ret = btrfs_scan_one_device(vol->name, FMODE_READ, 16428a4b83ccSChris Mason &btrfs_fs_type, &fs_devices); 16438a4b83ccSChris Mason break; 164402db0844SJosef Bacik case BTRFS_IOC_DEVICES_READY: 164502db0844SJosef Bacik ret = btrfs_scan_one_device(vol->name, FMODE_READ, 164602db0844SJosef Bacik &btrfs_fs_type, &fs_devices); 164702db0844SJosef Bacik if (ret) 164802db0844SJosef Bacik break; 164902db0844SJosef Bacik ret = !(fs_devices->num_devices == fs_devices->total_devices); 165002db0844SJosef Bacik break; 16518a4b83ccSChris Mason } 1652dae7b665SLi Zefan 16538a4b83ccSChris Mason kfree(vol); 1654f819d837SLinda Knippers return ret; 16558a4b83ccSChris Mason } 16568a4b83ccSChris Mason 16570176260fSLinus Torvalds static int btrfs_freeze(struct super_block *sb) 1658ed0dab6bSYan { 1659354aa0fbSMiao Xie struct btrfs_trans_handle *trans; 1660354aa0fbSMiao Xie struct btrfs_root *root = btrfs_sb(sb)->tree_root; 1661354aa0fbSMiao Xie 1662d4edf39bSMiao Xie trans = btrfs_attach_transaction_barrier(root); 1663354aa0fbSMiao Xie if (IS_ERR(trans)) { 1664354aa0fbSMiao Xie /* no transaction, don't bother */ 1665354aa0fbSMiao Xie if (PTR_ERR(trans) == -ENOENT) 16660176260fSLinus Torvalds return 0; 1667354aa0fbSMiao Xie return PTR_ERR(trans); 1668354aa0fbSMiao Xie } 1669354aa0fbSMiao Xie return btrfs_commit_transaction(trans, root); 1670ed0dab6bSYan } 1671ed0dab6bSYan 16720176260fSLinus Torvalds static int btrfs_unfreeze(struct super_block *sb) 1673ed0dab6bSYan { 16740176260fSLinus Torvalds return 0; 1675ed0dab6bSYan } 16762e635a27SChris Mason 16779c5085c1SJosef Bacik static int btrfs_show_devname(struct seq_file *m, struct dentry *root) 16789c5085c1SJosef Bacik { 16799c5085c1SJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); 16809c5085c1SJosef Bacik struct btrfs_fs_devices *cur_devices; 16819c5085c1SJosef Bacik struct btrfs_device *dev, *first_dev = NULL; 16829c5085c1SJosef Bacik struct list_head *head; 16839c5085c1SJosef Bacik struct rcu_string *name; 16849c5085c1SJosef Bacik 16859c5085c1SJosef Bacik mutex_lock(&fs_info->fs_devices->device_list_mutex); 16869c5085c1SJosef Bacik cur_devices = fs_info->fs_devices; 16879c5085c1SJosef Bacik while (cur_devices) { 16889c5085c1SJosef Bacik head = &cur_devices->devices; 16899c5085c1SJosef Bacik list_for_each_entry(dev, head, dev_list) { 1690aa9ddcd4SJosef Bacik if (dev->missing) 1691aa9ddcd4SJosef Bacik continue; 16929c5085c1SJosef Bacik if (!first_dev || dev->devid < first_dev->devid) 16939c5085c1SJosef Bacik first_dev = dev; 16949c5085c1SJosef Bacik } 16959c5085c1SJosef Bacik cur_devices = cur_devices->seed; 16969c5085c1SJosef Bacik } 16979c5085c1SJosef Bacik 16989c5085c1SJosef Bacik if (first_dev) { 16999c5085c1SJosef Bacik rcu_read_lock(); 17009c5085c1SJosef Bacik name = rcu_dereference(first_dev->name); 17019c5085c1SJosef Bacik seq_escape(m, name->str, " \t\n\\"); 17029c5085c1SJosef Bacik rcu_read_unlock(); 17039c5085c1SJosef Bacik } else { 17049c5085c1SJosef Bacik WARN_ON(1); 17059c5085c1SJosef Bacik } 17069c5085c1SJosef Bacik mutex_unlock(&fs_info->fs_devices->device_list_mutex); 17079c5085c1SJosef Bacik return 0; 17089c5085c1SJosef Bacik } 17099c5085c1SJosef Bacik 1710b87221deSAlexey Dobriyan static const struct super_operations btrfs_super_ops = { 171176dda93cSYan, Zheng .drop_inode = btrfs_drop_inode, 1712bd555975SAl Viro .evict_inode = btrfs_evict_inode, 1713e20d96d6SChris Mason .put_super = btrfs_put_super, 1714d5719762SChris Mason .sync_fs = btrfs_sync_fs, 1715a9572a15SEric Paris .show_options = btrfs_show_options, 17169c5085c1SJosef Bacik .show_devname = btrfs_show_devname, 17174730a4bcSChris Mason .write_inode = btrfs_write_inode, 17182c90e5d6SChris Mason .alloc_inode = btrfs_alloc_inode, 17192c90e5d6SChris Mason .destroy_inode = btrfs_destroy_inode, 17208fd17795SChris Mason .statfs = btrfs_statfs, 1721c146afadSYan Zheng .remount_fs = btrfs_remount, 17220176260fSLinus Torvalds .freeze_fs = btrfs_freeze, 17230176260fSLinus Torvalds .unfreeze_fs = btrfs_unfreeze, 1724e20d96d6SChris Mason }; 1725a9218f6bSChris Mason 1726a9218f6bSChris Mason static const struct file_operations btrfs_ctl_fops = { 1727a9218f6bSChris Mason .unlocked_ioctl = btrfs_control_ioctl, 1728a9218f6bSChris Mason .compat_ioctl = btrfs_control_ioctl, 1729a9218f6bSChris Mason .owner = THIS_MODULE, 17306038f373SArnd Bergmann .llseek = noop_llseek, 1731a9218f6bSChris Mason }; 1732a9218f6bSChris Mason 1733a9218f6bSChris Mason static struct miscdevice btrfs_misc = { 1734578454ffSKay Sievers .minor = BTRFS_MINOR, 1735a9218f6bSChris Mason .name = "btrfs-control", 1736a9218f6bSChris Mason .fops = &btrfs_ctl_fops 1737a9218f6bSChris Mason }; 1738a9218f6bSChris Mason 1739578454ffSKay Sievers MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 1740578454ffSKay Sievers MODULE_ALIAS("devname:btrfs-control"); 1741578454ffSKay Sievers 1742a9218f6bSChris Mason static int btrfs_interface_init(void) 1743a9218f6bSChris Mason { 1744a9218f6bSChris Mason return misc_register(&btrfs_misc); 1745a9218f6bSChris Mason } 1746a9218f6bSChris Mason 1747b2950863SChristoph Hellwig static void btrfs_interface_exit(void) 1748a9218f6bSChris Mason { 1749a9218f6bSChris Mason if (misc_deregister(&btrfs_misc) < 0) 175048940662SDaniel J Blueman printk(KERN_INFO "btrfs: misc_deregister failed for control device\n"); 1751a9218f6bSChris Mason } 1752a9218f6bSChris Mason 175385965600SDavid Sterba static void btrfs_print_info(void) 175485965600SDavid Sterba { 175585965600SDavid Sterba printk(KERN_INFO "Btrfs loaded" 175685965600SDavid Sterba #ifdef CONFIG_BTRFS_DEBUG 175785965600SDavid Sterba ", debug=on" 175885965600SDavid Sterba #endif 175985965600SDavid Sterba #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 176085965600SDavid Sterba ", integrity-checker=on" 176185965600SDavid Sterba #endif 176285965600SDavid Sterba "\n"); 176385965600SDavid Sterba } 176485965600SDavid Sterba 17652e635a27SChris Mason static int __init init_btrfs_fs(void) 17662e635a27SChris Mason { 17672c90e5d6SChris Mason int err; 176858176a96SJosef Bacik 176958176a96SJosef Bacik err = btrfs_init_sysfs(); 177058176a96SJosef Bacik if (err) 177158176a96SJosef Bacik return err; 177258176a96SJosef Bacik 1773143bede5SJeff Mahoney btrfs_init_compress(); 1774d1310b2eSChris Mason 1775261507a0SLi Zefan err = btrfs_init_cachep(); 1776261507a0SLi Zefan if (err) 1777261507a0SLi Zefan goto free_compress; 1778261507a0SLi Zefan 1779d1310b2eSChris Mason err = extent_io_init(); 17802f4cbe64SWyatt Banks if (err) 17812f4cbe64SWyatt Banks goto free_cachep; 17822f4cbe64SWyatt Banks 1783d1310b2eSChris Mason err = extent_map_init(); 1784d1310b2eSChris Mason if (err) 1785d1310b2eSChris Mason goto free_extent_io; 1786d1310b2eSChris Mason 17876352b91dSMiao Xie err = ordered_data_init(); 17882f4cbe64SWyatt Banks if (err) 17892f4cbe64SWyatt Banks goto free_extent_map; 1790c8b97818SChris Mason 17916352b91dSMiao Xie err = btrfs_delayed_inode_init(); 17926352b91dSMiao Xie if (err) 17936352b91dSMiao Xie goto free_ordered_data; 17946352b91dSMiao Xie 17959247f317SMiao Xie err = btrfs_auto_defrag_init(); 179616cdcec7SMiao Xie if (err) 179716cdcec7SMiao Xie goto free_delayed_inode; 179816cdcec7SMiao Xie 179978a6184aSMiao Xie err = btrfs_delayed_ref_init(); 18009247f317SMiao Xie if (err) 18019247f317SMiao Xie goto free_auto_defrag; 18029247f317SMiao Xie 180378a6184aSMiao Xie err = btrfs_interface_init(); 180478a6184aSMiao Xie if (err) 180578a6184aSMiao Xie goto free_delayed_ref; 180678a6184aSMiao Xie 1807a9218f6bSChris Mason err = register_filesystem(&btrfs_fs_type); 1808a9218f6bSChris Mason if (err) 1809a9218f6bSChris Mason goto unregister_ioctl; 1810b3c3da71SChris Mason 1811e565d4b9SJan Schmidt btrfs_init_lockdep(); 1812e565d4b9SJan Schmidt 181385965600SDavid Sterba btrfs_print_info(); 181474255aa0SJosef Bacik btrfs_test_free_space_cache(); 181574255aa0SJosef Bacik 18162f4cbe64SWyatt Banks return 0; 18172f4cbe64SWyatt Banks 1818a9218f6bSChris Mason unregister_ioctl: 1819a9218f6bSChris Mason btrfs_interface_exit(); 182078a6184aSMiao Xie free_delayed_ref: 182178a6184aSMiao Xie btrfs_delayed_ref_exit(); 18229247f317SMiao Xie free_auto_defrag: 18239247f317SMiao Xie btrfs_auto_defrag_exit(); 182416cdcec7SMiao Xie free_delayed_inode: 182516cdcec7SMiao Xie btrfs_delayed_inode_exit(); 18266352b91dSMiao Xie free_ordered_data: 18276352b91dSMiao Xie ordered_data_exit(); 18282f4cbe64SWyatt Banks free_extent_map: 18292f4cbe64SWyatt Banks extent_map_exit(); 1830d1310b2eSChris Mason free_extent_io: 1831d1310b2eSChris Mason extent_io_exit(); 18322f4cbe64SWyatt Banks free_cachep: 18332f4cbe64SWyatt Banks btrfs_destroy_cachep(); 1834261507a0SLi Zefan free_compress: 1835261507a0SLi Zefan btrfs_exit_compress(); 18362f4cbe64SWyatt Banks btrfs_exit_sysfs(); 18372c90e5d6SChris Mason return err; 18382e635a27SChris Mason } 18392e635a27SChris Mason 18402e635a27SChris Mason static void __exit exit_btrfs_fs(void) 18412e635a27SChris Mason { 184239279cc3SChris Mason btrfs_destroy_cachep(); 184378a6184aSMiao Xie btrfs_delayed_ref_exit(); 18449247f317SMiao Xie btrfs_auto_defrag_exit(); 184516cdcec7SMiao Xie btrfs_delayed_inode_exit(); 18466352b91dSMiao Xie ordered_data_exit(); 1847a52d9a80SChris Mason extent_map_exit(); 1848d1310b2eSChris Mason extent_io_exit(); 1849a9218f6bSChris Mason btrfs_interface_exit(); 18502e635a27SChris Mason unregister_filesystem(&btrfs_fs_type); 185158176a96SJosef Bacik btrfs_exit_sysfs(); 18528a4b83ccSChris Mason btrfs_cleanup_fs_uuids(); 1853261507a0SLi Zefan btrfs_exit_compress(); 18542e635a27SChris Mason } 18552e635a27SChris Mason 18562e635a27SChris Mason module_init(init_btrfs_fs) 18572e635a27SChris Mason module_exit(exit_btrfs_fs) 18582e635a27SChris Mason 18592e635a27SChris Mason MODULE_LICENSE("GPL"); 1860