linux/fs/btrfs/super.c
<<
>>
Prefs
   1/*
   2 * Copyright (C) 2007 Oracle.  All rights reserved.
   3 *
   4 * This program is free software; you can redistribute it and/or
   5 * modify it under the terms of the GNU General Public
   6 * License v2 as published by the Free Software Foundation.
   7 *
   8 * This program is distributed in the hope that it will be useful,
   9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11 * General Public License for more details.
  12 *
  13 * You should have received a copy of the GNU General Public
  14 * License along with this program; if not, write to the
  15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16 * Boston, MA 021110-1307, USA.
  17 */
  18
  19#include <linux/blkdev.h>
  20#include <linux/module.h>
  21#include <linux/buffer_head.h>
  22#include <linux/fs.h>
  23#include <linux/pagemap.h>
  24#include <linux/highmem.h>
  25#include <linux/time.h>
  26#include <linux/init.h>
  27#include <linux/seq_file.h>
  28#include <linux/string.h>
  29#include <linux/backing-dev.h>
  30#include <linux/mount.h>
  31#include <linux/mpage.h>
  32#include <linux/swap.h>
  33#include <linux/writeback.h>
  34#include <linux/statfs.h>
  35#include <linux/compat.h>
  36#include <linux/parser.h>
  37#include <linux/ctype.h>
  38#include <linux/namei.h>
  39#include <linux/miscdevice.h>
  40#include <linux/magic.h>
  41#include <linux/slab.h>
  42#include <linux/cleancache.h>
  43#include <linux/ratelimit.h>
  44#include "compat.h"
  45#include "delayed-inode.h"
  46#include "ctree.h"
  47#include "disk-io.h"
  48#include "transaction.h"
  49#include "btrfs_inode.h"
  50#include "ioctl.h"
  51#include "print-tree.h"
  52#include "xattr.h"
  53#include "volumes.h"
  54#include "version.h"
  55#include "export.h"
  56#include "compression.h"
  57#include "rcu-string.h"
  58
  59#define CREATE_TRACE_POINTS
  60#include <trace/events/btrfs.h>
  61
  62static const struct super_operations btrfs_super_ops;
  63static struct file_system_type btrfs_fs_type;
  64
  65static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
  66                                      char nbuf[16])
  67{
  68        char *errstr = NULL;
  69
  70        switch (errno) {
  71        case -EIO:
  72                errstr = "IO failure";
  73                break;
  74        case -ENOMEM:
  75                errstr = "Out of memory";
  76                break;
  77        case -EROFS:
  78                errstr = "Readonly filesystem";
  79                break;
  80        case -EEXIST:
  81                errstr = "Object already exists";
  82                break;
  83        default:
  84                if (nbuf) {
  85                        if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  86                                errstr = nbuf;
  87                }
  88                break;
  89        }
  90
  91        return errstr;
  92}
  93
  94static void __save_error_info(struct btrfs_fs_info *fs_info)
  95{
  96        /*
  97         * today we only save the error info into ram.  Long term we'll
  98         * also send it down to the disk
  99         */
 100        fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR;
 101}
 102
 103static void save_error_info(struct btrfs_fs_info *fs_info)
 104{
 105        __save_error_info(fs_info);
 106}
 107
 108/* btrfs handle error by forcing the filesystem readonly */
 109static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
 110{
 111        struct super_block *sb = fs_info->sb;
 112
 113        if (sb->s_flags & MS_RDONLY)
 114                return;
 115
 116        if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
 117                sb->s_flags |= MS_RDONLY;
 118                printk(KERN_INFO "btrfs is forced readonly\n");
 119                __btrfs_scrub_cancel(fs_info);
 120//              WARN_ON(1);
 121        }
 122}
 123
 124#ifdef CONFIG_PRINTK
 125/*
 126 * __btrfs_std_error decodes expected errors from the caller and
 127 * invokes the approciate error response.
 128 */
 129void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 130                       unsigned int line, int errno, const char *fmt, ...)
 131{
 132        struct super_block *sb = fs_info->sb;
 133        char nbuf[16];
 134        const char *errstr;
 135        va_list args;
 136        va_start(args, fmt);
 137
 138        /*
 139         * Special case: if the error is EROFS, and we're already
 140         * under MS_RDONLY, then it is safe here.
 141         */
 142        if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 143                return;
 144
 145        errstr = btrfs_decode_error(fs_info, errno, nbuf);
 146        if (fmt) {
 147                struct va_format vaf = {
 148                        .fmt = fmt,
 149                        .va = &args,
 150                };
 151
 152                printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s (%pV)\n",
 153                        sb->s_id, function, line, errstr, &vaf);
 154        } else {
 155                printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n",
 156                        sb->s_id, function, line, errstr);
 157        }
 158
 159        /* Don't go through full error handling during mount */
 160        if (sb->s_flags & MS_BORN) {
 161                save_error_info(fs_info);
 162                btrfs_handle_error(fs_info);
 163        }
 164        va_end(args);
 165}
 166
 167static const char * const logtypes[] = {
 168        "emergency",
 169        "alert",
 170        "critical",
 171        "error",
 172        "warning",
 173        "notice",
 174        "info",
 175        "debug",
 176};
 177
 178void btrfs_printk(struct btrfs_fs_info *fs_info, const char *fmt, ...)
 179{
 180        struct super_block *sb = fs_info->sb;
 181        char lvl[4];
 182        struct va_format vaf;
 183        va_list args;
 184        const char *type = logtypes[4];
 185        int kern_level;
 186
 187        va_start(args, fmt);
 188
 189        kern_level = printk_get_level(fmt);
 190        if (kern_level) {
 191                size_t size = printk_skip_level(fmt) - fmt;
 192                memcpy(lvl, fmt,  size);
 193                lvl[size] = '\0';
 194                fmt += size;
 195                type = logtypes[kern_level - '0'];
 196        } else
 197                *lvl = '\0';
 198
 199        vaf.fmt = fmt;
 200        vaf.va = &args;
 201
 202        printk("%sBTRFS %s (device %s): %pV", lvl, type, sb->s_id, &vaf);
 203
 204        va_end(args);
 205}
 206
 207#else
 208
 209void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
 210                       unsigned int line, int errno, const char *fmt, ...)
 211{
 212        struct super_block *sb = fs_info->sb;
 213
 214        /*
 215         * Special case: if the error is EROFS, and we're already
 216         * under MS_RDONLY, then it is safe here.
 217         */
 218        if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
 219                return;
 220
 221        /* Don't go through full error handling during mount */
 222        if (sb->s_flags & MS_BORN) {
 223                save_error_info(fs_info);
 224                btrfs_handle_error(fs_info);
 225        }
 226}
 227#endif
 228
 229/*
 230 * We only mark the transaction aborted and then set the file system read-only.
 231 * This will prevent new transactions from starting or trying to join this
 232 * one.
 233 *
 234 * This means that error recovery at the call site is limited to freeing
 235 * any local memory allocations and passing the error code up without
 236 * further cleanup. The transaction should complete as it normally would
 237 * in the call path but will return -EIO.
 238 *
 239 * We'll complete the cleanup in btrfs_end_transaction and
 240 * btrfs_commit_transaction.
 241 */
 242void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
 243                               struct btrfs_root *root, const char *function,
 244                               unsigned int line, int errno)
 245{
 246        WARN_ONCE(1, KERN_DEBUG "btrfs: Transaction aborted");
 247        trans->aborted = errno;
 248        /* Nothing used. The other threads that have joined this
 249         * transaction may be able to continue. */
 250        if (!trans->blocks_used) {
 251                btrfs_printk(root->fs_info, "Aborting unused transaction.\n");
 252                return;
 253        }
 254        trans->transaction->aborted = errno;
 255        __btrfs_std_error(root->fs_info, function, line, errno, NULL);
 256}
 257/*
 258 * __btrfs_panic decodes unexpected, fatal errors from the caller,
 259 * issues an alert, and either panics or BUGs, depending on mount options.
 260 */
 261void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
 262                   unsigned int line, int errno, const char *fmt, ...)
 263{
 264        char nbuf[16];
 265        char *s_id = "<unknown>";
 266        const char *errstr;
 267        struct va_format vaf = { .fmt = fmt };
 268        va_list args;
 269
 270        if (fs_info)
 271                s_id = fs_info->sb->s_id;
 272
 273        va_start(args, fmt);
 274        vaf.va = &args;
 275
 276        errstr = btrfs_decode_error(fs_info, errno, nbuf);
 277        if (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)
 278                panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
 279                        s_id, function, line, &vaf, errstr);
 280
 281        printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (%s)\n",
 282               s_id, function, line, &vaf, errstr);
 283        va_end(args);
 284        /* Caller calls BUG() */
 285}
 286
 287static void btrfs_put_super(struct super_block *sb)
 288{
 289        (void)close_ctree(btrfs_sb(sb)->tree_root);
 290        /* FIXME: need to fix VFS to return error? */
 291        /* AV: return it _where_?  ->put_super() can be triggered by any number
 292         * of async events, up to and including delivery of SIGKILL to the
 293         * last process that kept it busy.  Or segfault in the aforementioned
 294         * process...  Whom would you report that to?
 295         */
 296}
 297
 298enum {
 299        Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 300        Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 301        Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
 302        Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
 303        Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 304        Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 305        Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
 306        Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
 307        Opt_check_integrity, Opt_check_integrity_including_extent_data,
 308        Opt_check_integrity_print_mask, Opt_fatal_errors,
 309        Opt_err,
 310};
 311
 312static match_table_t tokens = {
 313        {Opt_degraded, "degraded"},
 314        {Opt_subvol, "subvol=%s"},
 315        {Opt_subvolid, "subvolid=%d"},
 316        {Opt_device, "device=%s"},
 317        {Opt_nodatasum, "nodatasum"},
 318        {Opt_nodatacow, "nodatacow"},
 319        {Opt_nobarrier, "nobarrier"},
 320        {Opt_max_inline, "max_inline=%s"},
 321        {Opt_alloc_start, "alloc_start=%s"},
 322        {Opt_thread_pool, "thread_pool=%d"},
 323        {Opt_compress, "compress"},
 324        {Opt_compress_type, "compress=%s"},
 325        {Opt_compress_force, "compress-force"},
 326        {Opt_compress_force_type, "compress-force=%s"},
 327        {Opt_ssd, "ssd"},
 328        {Opt_ssd_spread, "ssd_spread"},
 329        {Opt_nossd, "nossd"},
 330        {Opt_noacl, "noacl"},
 331        {Opt_notreelog, "notreelog"},
 332        {Opt_flushoncommit, "flushoncommit"},
 333        {Opt_ratio, "metadata_ratio=%d"},
 334        {Opt_discard, "discard"},
 335        {Opt_space_cache, "space_cache"},
 336        {Opt_clear_cache, "clear_cache"},
 337        {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 338        {Opt_enospc_debug, "enospc_debug"},
 339        {Opt_subvolrootid, "subvolrootid=%d"},
 340        {Opt_defrag, "autodefrag"},
 341        {Opt_inode_cache, "inode_cache"},
 342        {Opt_no_space_cache, "nospace_cache"},
 343        {Opt_recovery, "recovery"},
 344        {Opt_skip_balance, "skip_balance"},
 345        {Opt_check_integrity, "check_int"},
 346        {Opt_check_integrity_including_extent_data, "check_int_data"},
 347        {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
 348        {Opt_fatal_errors, "fatal_errors=%s"},
 349        {Opt_err, NULL},
 350};
 351
 352/*
 353 * Regular mount options parser.  Everything that is needed only when
 354 * reading in a new superblock is parsed here.
 355 * XXX JDM: This needs to be cleaned up for remount.
 356 */
 357int btrfs_parse_options(struct btrfs_root *root, char *options)
 358{
 359        struct btrfs_fs_info *info = root->fs_info;
 360        substring_t args[MAX_OPT_ARGS];
 361        char *p, *num, *orig = NULL;
 362        u64 cache_gen;
 363        int intarg;
 364        int ret = 0;
 365        char *compress_type;
 366        bool compress_force = false;
 367
 368        cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
 369        if (cache_gen)
 370                btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 371
 372        if (!options)
 373                goto out;
 374
 375        /*
 376         * strsep changes the string, duplicate it because parse_options
 377         * gets called twice
 378         */
 379        options = kstrdup(options, GFP_NOFS);
 380        if (!options)
 381                return -ENOMEM;
 382
 383        orig = options;
 384
 385        while ((p = strsep(&options, ",")) != NULL) {
 386                int token;
 387                if (!*p)
 388                        continue;
 389
 390                token = match_token(p, tokens, args);
 391                switch (token) {
 392                case Opt_degraded:
 393                        printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 394                        btrfs_set_opt(info->mount_opt, DEGRADED);
 395                        break;
 396                case Opt_subvol:
 397                case Opt_subvolid:
 398                case Opt_subvolrootid:
 399                case Opt_device:
 400                        /*
 401                         * These are parsed by btrfs_parse_early_options
 402                         * and can be happily ignored here.
 403                         */
 404                        break;
 405                case Opt_nodatasum:
 406                        printk(KERN_INFO "btrfs: setting nodatasum\n");
 407                        btrfs_set_opt(info->mount_opt, NODATASUM);
 408                        break;
 409                case Opt_nodatacow:
 410                        printk(KERN_INFO "btrfs: setting nodatacow\n");
 411                        btrfs_set_opt(info->mount_opt, NODATACOW);
 412                        btrfs_set_opt(info->mount_opt, NODATASUM);
 413                        break;
 414                case Opt_compress_force:
 415                case Opt_compress_force_type:
 416                        compress_force = true;
 417                case Opt_compress:
 418                case Opt_compress_type:
 419                        if (token == Opt_compress ||
 420                            token == Opt_compress_force ||
 421                            strcmp(args[0].from, "zlib") == 0) {
 422                                compress_type = "zlib";
 423                                info->compress_type = BTRFS_COMPRESS_ZLIB;
 424                                btrfs_set_opt(info->mount_opt, COMPRESS);
 425                        } else if (strcmp(args[0].from, "lzo") == 0) {
 426                                compress_type = "lzo";
 427                                info->compress_type = BTRFS_COMPRESS_LZO;
 428                                btrfs_set_opt(info->mount_opt, COMPRESS);
 429                                btrfs_set_fs_incompat(info, COMPRESS_LZO);
 430                        } else if (strncmp(args[0].from, "no", 2) == 0) {
 431                                compress_type = "no";
 432                                info->compress_type = BTRFS_COMPRESS_NONE;
 433                                btrfs_clear_opt(info->mount_opt, COMPRESS);
 434                                btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 435                                compress_force = false;
 436                        } else {
 437                                ret = -EINVAL;
 438                                goto out;
 439                        }
 440
 441                        if (compress_force) {
 442                                btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
 443                                pr_info("btrfs: force %s compression\n",
 444                                        compress_type);
 445                        } else
 446                                pr_info("btrfs: use %s compression\n",
 447                                        compress_type);
 448                        break;
 449                case Opt_ssd:
 450                        printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 451                        btrfs_set_opt(info->mount_opt, SSD);
 452                        break;
 453                case Opt_ssd_spread:
 454                        printk(KERN_INFO "btrfs: use spread ssd "
 455                               "allocation scheme\n");
 456                        btrfs_set_opt(info->mount_opt, SSD);
 457                        btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 458                        break;
 459                case Opt_nossd:
 460                        printk(KERN_INFO "btrfs: not using ssd allocation "
 461                               "scheme\n");
 462                        btrfs_set_opt(info->mount_opt, NOSSD);
 463                        btrfs_clear_opt(info->mount_opt, SSD);
 464                        btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 465                        break;
 466                case Opt_nobarrier:
 467                        printk(KERN_INFO "btrfs: turning off barriers\n");
 468                        btrfs_set_opt(info->mount_opt, NOBARRIER);
 469                        break;
 470                case Opt_thread_pool:
 471                        intarg = 0;
 472                        match_int(&args[0], &intarg);
 473                        if (intarg)
 474                                info->thread_pool_size = intarg;
 475                        break;
 476                case Opt_max_inline:
 477                        num = match_strdup(&args[0]);
 478                        if (num) {
 479                                info->max_inline = memparse(num, NULL);
 480                                kfree(num);
 481
 482                                if (info->max_inline) {
 483                                        info->max_inline = max_t(u64,
 484                                                info->max_inline,
 485                                                root->sectorsize);
 486                                }
 487                                printk(KERN_INFO "btrfs: max_inline at %llu\n",
 488                                        (unsigned long long)info->max_inline);
 489                        }
 490                        break;
 491                case Opt_alloc_start:
 492                        num = match_strdup(&args[0]);
 493                        if (num) {
 494                                info->alloc_start = memparse(num, NULL);
 495                                kfree(num);
 496                                printk(KERN_INFO
 497                                        "btrfs: allocations start at %llu\n",
 498                                        (unsigned long long)info->alloc_start);
 499                        }
 500                        break;
 501                case Opt_noacl:
 502                        root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 503                        break;
 504                case Opt_notreelog:
 505                        printk(KERN_INFO "btrfs: disabling tree log\n");
 506                        btrfs_set_opt(info->mount_opt, NOTREELOG);
 507                        break;
 508                case Opt_flushoncommit:
 509                        printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
 510                        btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
 511                        break;
 512                case Opt_ratio:
 513                        intarg = 0;
 514                        match_int(&args[0], &intarg);
 515                        if (intarg) {
 516                                info->metadata_ratio = intarg;
 517                                printk(KERN_INFO "btrfs: metadata ratio %d\n",
 518                                       info->metadata_ratio);
 519                        }
 520                        break;
 521                case Opt_discard:
 522                        btrfs_set_opt(info->mount_opt, DISCARD);
 523                        break;
 524                case Opt_space_cache:
 525                        btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 526                        break;
 527                case Opt_no_space_cache:
 528                        printk(KERN_INFO "btrfs: disabling disk space caching\n");
 529                        btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
 530                        break;
 531                case Opt_inode_cache:
 532                        printk(KERN_INFO "btrfs: enabling inode map caching\n");
 533                        btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
 534                        break;
 535                case Opt_clear_cache:
 536                        printk(KERN_INFO "btrfs: force clearing of disk cache\n");
 537                        btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
 538                        break;
 539                case Opt_user_subvol_rm_allowed:
 540                        btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 541                        break;
 542                case Opt_enospc_debug:
 543                        btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
 544                        break;
 545                case Opt_defrag:
 546                        printk(KERN_INFO "btrfs: enabling auto defrag");
 547                        btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 548                        break;
 549                case Opt_recovery:
 550                        printk(KERN_INFO "btrfs: enabling auto recovery");
 551                        btrfs_set_opt(info->mount_opt, RECOVERY);
 552                        break;
 553                case Opt_skip_balance:
 554                        btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
 555                        break;
 556#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
 557                case Opt_check_integrity_including_extent_data:
 558                        printk(KERN_INFO "btrfs: enabling check integrity"
 559                               " including extent data\n");
 560                        btrfs_set_opt(info->mount_opt,
 561                                      CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
 562                        btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 563                        break;
 564                case Opt_check_integrity:
 565                        printk(KERN_INFO "btrfs: enabling check integrity\n");
 566                        btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
 567                        break;
 568                case Opt_check_integrity_print_mask:
 569                        intarg = 0;
 570                        match_int(&args[0], &intarg);
 571                        if (intarg) {
 572                                info->check_integrity_print_mask = intarg;
 573                                printk(KERN_INFO "btrfs:"
 574                                       " check_integrity_print_mask 0x%x\n",
 575                                       info->check_integrity_print_mask);
 576                        }
 577                        break;
 578#else
 579                case Opt_check_integrity_including_extent_data:
 580                case Opt_check_integrity:
 581                case Opt_check_integrity_print_mask:
 582                        printk(KERN_ERR "btrfs: support for check_integrity*"
 583                               " not compiled in!\n");
 584                        ret = -EINVAL;
 585                        goto out;
 586#endif
 587                case Opt_fatal_errors:
 588                        if (strcmp(args[0].from, "panic") == 0)
 589                                btrfs_set_opt(info->mount_opt,
 590                                              PANIC_ON_FATAL_ERROR);
 591                        else if (strcmp(args[0].from, "bug") == 0)
 592                                btrfs_clear_opt(info->mount_opt,
 593                                              PANIC_ON_FATAL_ERROR);
 594                        else {
 595                                ret = -EINVAL;
 596                                goto out;
 597                        }
 598                        break;
 599                case Opt_err:
 600                        printk(KERN_INFO "btrfs: unrecognized mount option "
 601                               "'%s'\n", p);
 602                        ret = -EINVAL;
 603                        goto out;
 604                default:
 605                        break;
 606                }
 607        }
 608out:
 609        if (!ret && btrfs_test_opt(root, SPACE_CACHE))
 610                printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 611        kfree(orig);
 612        return ret;
 613}
 614
 615/*
 616 * Parse mount options that are required early in the mount process.
 617 *
 618 * All other options will be parsed on much later in the mount process and
 619 * only when we need to allocate a new super block.
 620 */
 621static int btrfs_parse_early_options(const char *options, fmode_t flags,
 622                void *holder, char **subvol_name, u64 *subvol_objectid,
 623                u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
 624{
 625        substring_t args[MAX_OPT_ARGS];
 626        char *device_name, *opts, *orig, *p;
 627        int error = 0;
 628        int intarg;
 629
 630        if (!options)
 631                return 0;
 632
 633        /*
 634         * strsep changes the string, duplicate it because parse_options
 635         * gets called twice
 636         */
 637        opts = kstrdup(options, GFP_KERNEL);
 638        if (!opts)
 639                return -ENOMEM;
 640        orig = opts;
 641
 642        while ((p = strsep(&opts, ",")) != NULL) {
 643                int token;
 644                if (!*p)
 645                        continue;
 646
 647                token = match_token(p, tokens, args);
 648                switch (token) {
 649                case Opt_subvol:
 650                        kfree(*subvol_name);
 651                        *subvol_name = match_strdup(&args[0]);
 652                        break;
 653                case Opt_subvolid:
 654                        intarg = 0;
 655                        error = match_int(&args[0], &intarg);
 656                        if (!error) {
 657                                /* we want the original fs_tree */
 658                                if (!intarg)
 659                                        *subvol_objectid =
 660                                                BTRFS_FS_TREE_OBJECTID;
 661                                else
 662                                        *subvol_objectid = intarg;
 663                        }
 664                        break;
 665                case Opt_subvolrootid:
 666                        intarg = 0;
 667                        error = match_int(&args[0], &intarg);
 668                        if (!error) {
 669                                /* we want the original fs_tree */
 670                                if (!intarg)
 671                                        *subvol_rootid =
 672                                                BTRFS_FS_TREE_OBJECTID;
 673                                else
 674                                        *subvol_rootid = intarg;
 675                        }
 676                        break;
 677                case Opt_device:
 678                        device_name = match_strdup(&args[0]);
 679                        if (!device_name) {
 680                                error = -ENOMEM;
 681                                goto out;
 682                        }
 683                        error = btrfs_scan_one_device(device_name,
 684                                        flags, holder, fs_devices);
 685                        kfree(device_name);
 686                        if (error)
 687                                goto out;
 688                        break;
 689                default:
 690                        break;
 691                }
 692        }
 693
 694out:
 695        kfree(orig);
 696        return error;
 697}
 698
 699static struct dentry *get_default_root(struct super_block *sb,
 700                                       u64 subvol_objectid)
 701{
 702        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 703        struct btrfs_root *root = fs_info->tree_root;
 704        struct btrfs_root *new_root;
 705        struct btrfs_dir_item *di;
 706        struct btrfs_path *path;
 707        struct btrfs_key location;
 708        struct inode *inode;
 709        u64 dir_id;
 710        int new = 0;
 711
 712        /*
 713         * We have a specific subvol we want to mount, just setup location and
 714         * go look up the root.
 715         */
 716        if (subvol_objectid) {
 717                location.objectid = subvol_objectid;
 718                location.type = BTRFS_ROOT_ITEM_KEY;
 719                location.offset = (u64)-1;
 720                goto find_root;
 721        }
 722
 723        path = btrfs_alloc_path();
 724        if (!path)
 725                return ERR_PTR(-ENOMEM);
 726        path->leave_spinning = 1;
 727
 728        /*
 729         * Find the "default" dir item which points to the root item that we
 730         * will mount by default if we haven't been given a specific subvolume
 731         * to mount.
 732         */
 733        dir_id = btrfs_super_root_dir(fs_info->super_copy);
 734        di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 735        if (IS_ERR(di)) {
 736                btrfs_free_path(path);
 737                return ERR_CAST(di);
 738        }
 739        if (!di) {
 740                /*
 741                 * Ok the default dir item isn't there.  This is weird since
 742                 * it's always been there, but don't freak out, just try and
 743                 * mount to root most subvolume.
 744                 */
 745                btrfs_free_path(path);
 746                dir_id = BTRFS_FIRST_FREE_OBJECTID;
 747                new_root = fs_info->fs_root;
 748                goto setup_root;
 749        }
 750
 751        btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
 752        btrfs_free_path(path);
 753
 754find_root:
 755        new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 756        if (IS_ERR(new_root))
 757                return ERR_CAST(new_root);
 758
 759        if (btrfs_root_refs(&new_root->root_item) == 0)
 760                return ERR_PTR(-ENOENT);
 761
 762        dir_id = btrfs_root_dirid(&new_root->root_item);
 763setup_root:
 764        location.objectid = dir_id;
 765        location.type = BTRFS_INODE_ITEM_KEY;
 766        location.offset = 0;
 767
 768        inode = btrfs_iget(sb, &location, new_root, &new);
 769        if (IS_ERR(inode))
 770                return ERR_CAST(inode);
 771
 772        /*
 773         * If we're just mounting the root most subvol put the inode and return
 774         * a reference to the dentry.  We will have already gotten a reference
 775         * to the inode in btrfs_fill_super so we're good to go.
 776         */
 777        if (!new && sb->s_root->d_inode == inode) {
 778                iput(inode);
 779                return dget(sb->s_root);
 780        }
 781
 782        return d_obtain_alias(inode);
 783}
 784
 785static int btrfs_fill_super(struct super_block *sb,
 786                            struct btrfs_fs_devices *fs_devices,
 787                            void *data, int silent)
 788{
 789        struct inode *inode;
 790        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 791        struct btrfs_key key;
 792        int err;
 793
 794        sb->s_maxbytes = MAX_LFS_FILESIZE;
 795        sb->s_magic = BTRFS_SUPER_MAGIC;
 796        sb->s_op = &btrfs_super_ops;
 797        sb->s_d_op = &btrfs_dentry_operations;
 798        sb->s_export_op = &btrfs_export_ops;
 799        sb->s_xattr = btrfs_xattr_handlers;
 800        sb->s_time_gran = 1;
 801#ifdef CONFIG_BTRFS_FS_POSIX_ACL
 802        sb->s_flags |= MS_POSIXACL;
 803#endif
 804        sb->s_flags |= MS_I_VERSION;
 805        err = open_ctree(sb, fs_devices, (char *)data);
 806        if (err) {
 807                printk("btrfs: open_ctree failed\n");
 808                return err;
 809        }
 810
 811        key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 812        key.type = BTRFS_INODE_ITEM_KEY;
 813        key.offset = 0;
 814        inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 815        if (IS_ERR(inode)) {
 816                err = PTR_ERR(inode);
 817                goto fail_close;
 818        }
 819
 820        sb->s_root = d_make_root(inode);
 821        if (!sb->s_root) {
 822                err = -ENOMEM;
 823                goto fail_close;
 824        }
 825
 826        save_mount_options(sb, data);
 827        cleancache_init_fs(sb);
 828        sb->s_flags |= MS_ACTIVE;
 829        return 0;
 830
 831fail_close:
 832        close_ctree(fs_info->tree_root);
 833        return err;
 834}
 835
 836int btrfs_sync_fs(struct super_block *sb, int wait)
 837{
 838        struct btrfs_trans_handle *trans;
 839        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 840        struct btrfs_root *root = fs_info->tree_root;
 841
 842        trace_btrfs_sync_fs(wait);
 843
 844        if (!wait) {
 845                filemap_flush(fs_info->btree_inode->i_mapping);
 846                return 0;
 847        }
 848
 849        btrfs_wait_ordered_extents(root, 0, 0);
 850
 851        spin_lock(&fs_info->trans_lock);
 852        if (!fs_info->running_transaction) {
 853                spin_unlock(&fs_info->trans_lock);
 854                return 0;
 855        }
 856        spin_unlock(&fs_info->trans_lock);
 857
 858        trans = btrfs_join_transaction(root);
 859        if (IS_ERR(trans))
 860                return PTR_ERR(trans);
 861        return btrfs_commit_transaction(trans, root);
 862}
 863
 864static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 865{
 866        struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
 867        struct btrfs_root *root = info->tree_root;
 868        char *compress_type;
 869
 870        if (btrfs_test_opt(root, DEGRADED))
 871                seq_puts(seq, ",degraded");
 872        if (btrfs_test_opt(root, NODATASUM))
 873                seq_puts(seq, ",nodatasum");
 874        if (btrfs_test_opt(root, NODATACOW))
 875                seq_puts(seq, ",nodatacow");
 876        if (btrfs_test_opt(root, NOBARRIER))
 877                seq_puts(seq, ",nobarrier");
 878        if (info->max_inline != 8192 * 1024)
 879                seq_printf(seq, ",max_inline=%llu",
 880                           (unsigned long long)info->max_inline);
 881        if (info->alloc_start != 0)
 882                seq_printf(seq, ",alloc_start=%llu",
 883                           (unsigned long long)info->alloc_start);
 884        if (info->thread_pool_size !=  min_t(unsigned long,
 885                                             num_online_cpus() + 2, 8))
 886                seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 887        if (btrfs_test_opt(root, COMPRESS)) {
 888                if (info->compress_type == BTRFS_COMPRESS_ZLIB)
 889                        compress_type = "zlib";
 890                else
 891                        compress_type = "lzo";
 892                if (btrfs_test_opt(root, FORCE_COMPRESS))
 893                        seq_printf(seq, ",compress-force=%s", compress_type);
 894                else
 895                        seq_printf(seq, ",compress=%s", compress_type);
 896        }
 897        if (btrfs_test_opt(root, NOSSD))
 898                seq_puts(seq, ",nossd");
 899        if (btrfs_test_opt(root, SSD_SPREAD))
 900                seq_puts(seq, ",ssd_spread");
 901        else if (btrfs_test_opt(root, SSD))
 902                seq_puts(seq, ",ssd");
 903        if (btrfs_test_opt(root, NOTREELOG))
 904                seq_puts(seq, ",notreelog");
 905        if (btrfs_test_opt(root, FLUSHONCOMMIT))
 906                seq_puts(seq, ",flushoncommit");
 907        if (btrfs_test_opt(root, DISCARD))
 908                seq_puts(seq, ",discard");
 909        if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 910                seq_puts(seq, ",noacl");
 911        if (btrfs_test_opt(root, SPACE_CACHE))
 912                seq_puts(seq, ",space_cache");
 913        else
 914                seq_puts(seq, ",nospace_cache");
 915        if (btrfs_test_opt(root, CLEAR_CACHE))
 916                seq_puts(seq, ",clear_cache");
 917        if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
 918                seq_puts(seq, ",user_subvol_rm_allowed");
 919        if (btrfs_test_opt(root, ENOSPC_DEBUG))
 920                seq_puts(seq, ",enospc_debug");
 921        if (btrfs_test_opt(root, AUTO_DEFRAG))
 922                seq_puts(seq, ",autodefrag");
 923        if (btrfs_test_opt(root, INODE_MAP_CACHE))
 924                seq_puts(seq, ",inode_cache");
 925        if (btrfs_test_opt(root, SKIP_BALANCE))
 926                seq_puts(seq, ",skip_balance");
 927        if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
 928                seq_puts(seq, ",fatal_errors=panic");
 929        return 0;
 930}
 931
 932static int btrfs_test_super(struct super_block *s, void *data)
 933{
 934        struct btrfs_fs_info *p = data;
 935        struct btrfs_fs_info *fs_info = btrfs_sb(s);
 936
 937        return fs_info->fs_devices == p->fs_devices;
 938}
 939
 940static int btrfs_set_super(struct super_block *s, void *data)
 941{
 942        int err = set_anon_super(s, data);
 943        if (!err)
 944                s->s_fs_info = data;
 945        return err;
 946}
 947
 948/*
 949 * subvolumes are identified by ino 256
 950 */
 951static inline int is_subvolume_inode(struct inode *inode)
 952{
 953        if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
 954                return 1;
 955        return 0;
 956}
 957
 958/*
 959 * This will strip out the subvol=%s argument for an argument string and add
 960 * subvolid=0 to make sure we get the actual tree root for path walking to the
 961 * subvol we want.
 962 */
 963static char *setup_root_args(char *args)
 964{
 965        unsigned len = strlen(args) + 2 + 1;
 966        char *src, *dst, *buf;
 967
 968        /*
 969         * We need the same args as before, but with this substitution:
 970         * s!subvol=[^,]+!subvolid=0!
 971         *
 972         * Since the replacement string is up to 2 bytes longer than the
 973         * original, allocate strlen(args) + 2 + 1 bytes.
 974         */
 975
 976        src = strstr(args, "subvol=");
 977        /* This shouldn't happen, but just in case.. */
 978        if (!src)
 979                return NULL;
 980
 981        buf = dst = kmalloc(len, GFP_NOFS);
 982        if (!buf)
 983                return NULL;
 984
 985        /*
 986         * If the subvol= arg is not at the start of the string,
 987         * copy whatever precedes it into buf.
 988         */
 989        if (src != args) {
 990                *src++ = '\0';
 991                strcpy(buf, args);
 992                dst += strlen(args);
 993        }
 994
 995        strcpy(dst, "subvolid=0");
 996        dst += strlen("subvolid=0");
 997
 998        /*
 999         * If there is a "," after the original subvol=... string,
1000         * copy that suffix into our buffer.  Otherwise, we're done.
1001         */
1002        src = strchr(src, ',');
1003        if (src)
1004                strcpy(dst, src);
1005
1006        return buf;
1007}
1008
1009static struct dentry *mount_subvol(const char *subvol_name, int flags,
1010                                   const char *device_name, char *data)
1011{
1012        struct dentry *root;
1013        struct vfsmount *mnt;
1014        char *newargs;
1015
1016        newargs = setup_root_args(data);
1017        if (!newargs)
1018                return ERR_PTR(-ENOMEM);
1019        mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1020                             newargs);
1021        kfree(newargs);
1022        if (IS_ERR(mnt))
1023                return ERR_CAST(mnt);
1024
1025        root = mount_subtree(mnt, subvol_name);
1026
1027        if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1028                struct super_block *s = root->d_sb;
1029                dput(root);
1030                root = ERR_PTR(-EINVAL);
1031                deactivate_locked_super(s);
1032                printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
1033                                subvol_name);
1034        }
1035
1036        return root;
1037}
1038
1039/*
1040 * Find a superblock for the given device / mount point.
1041 *
1042 * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1043 *        for multiple device setup.  Make sure to keep it in sync.
1044 */
1045static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1046                const char *device_name, void *data)
1047{
1048        struct block_device *bdev = NULL;
1049        struct super_block *s;
1050        struct dentry *root;
1051        struct btrfs_fs_devices *fs_devices = NULL;
1052        struct btrfs_fs_info *fs_info = NULL;
1053        fmode_t mode = FMODE_READ;
1054        char *subvol_name = NULL;
1055        u64 subvol_objectid = 0;
1056        u64 subvol_rootid = 0;
1057        int error = 0;
1058
1059        if (!(flags & MS_RDONLY))
1060                mode |= FMODE_WRITE;
1061
1062        error = btrfs_parse_early_options(data, mode, fs_type,
1063                                          &subvol_name, &subvol_objectid,
1064                                          &subvol_rootid, &fs_devices);
1065        if (error) {
1066                kfree(subvol_name);
1067                return ERR_PTR(error);
1068        }
1069
1070        if (subvol_name) {
1071                root = mount_subvol(subvol_name, flags, device_name, data);
1072                kfree(subvol_name);
1073                return root;
1074        }
1075
1076        error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1077        if (error)
1078                return ERR_PTR(error);
1079
1080        /*
1081         * Setup a dummy root and fs_info for test/set super.  This is because
1082         * we don't actually fill this stuff out until open_ctree, but we need
1083         * it for searching for existing supers, so this lets us do that and
1084         * then open_ctree will properly initialize everything later.
1085         */
1086        fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1087        if (!fs_info)
1088                return ERR_PTR(-ENOMEM);
1089
1090        fs_info->fs_devices = fs_devices;
1091
1092        fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1093        fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1094        if (!fs_info->super_copy || !fs_info->super_for_commit) {
1095                error = -ENOMEM;
1096                goto error_fs_info;
1097        }
1098
1099        error = btrfs_open_devices(fs_devices, mode, fs_type);
1100        if (error)
1101                goto error_fs_info;
1102
1103        if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1104                error = -EACCES;
1105                goto error_close_devices;
1106        }
1107
1108        bdev = fs_devices->latest_bdev;
1109        s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1110                 fs_info);
1111        if (IS_ERR(s)) {
1112                error = PTR_ERR(s);
1113                goto error_close_devices;
1114        }
1115
1116        if (s->s_root) {
1117                btrfs_close_devices(fs_devices);
1118                free_fs_info(fs_info);
1119                if ((flags ^ s->s_flags) & MS_RDONLY)
1120                        error = -EBUSY;
1121        } else {
1122                char b[BDEVNAME_SIZE];
1123
1124                strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1125                btrfs_sb(s)->bdev_holder = fs_type;
1126                error = btrfs_fill_super(s, fs_devices, data,
1127                                         flags & MS_SILENT ? 1 : 0);
1128        }
1129
1130        root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1131        if (IS_ERR(root))
1132                deactivate_locked_super(s);
1133
1134        return root;
1135
1136error_close_devices:
1137        btrfs_close_devices(fs_devices);
1138error_fs_info:
1139        free_fs_info(fs_info);
1140        return ERR_PTR(error);
1141}
1142
1143static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit)
1144{
1145        spin_lock_irq(&workers->lock);
1146        workers->max_workers = new_limit;
1147        spin_unlock_irq(&workers->lock);
1148}
1149
1150static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1151                                     int new_pool_size, int old_pool_size)
1152{
1153        if (new_pool_size == old_pool_size)
1154                return;
1155
1156        fs_info->thread_pool_size = new_pool_size;
1157
1158        printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n",
1159               old_pool_size, new_pool_size);
1160
1161        btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size);
1162        btrfs_set_max_workers(&fs_info->workers, new_pool_size);
1163        btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size);
1164        btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size);
1165        btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size);
1166        btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size);
1167        btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size);
1168        btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size);
1169        btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size);
1170        btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size);
1171        btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size);
1172        btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size);
1173        btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size);
1174        btrfs_set_max_workers(&fs_info->scrub_workers, new_pool_size);
1175}
1176
1177static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1178{
1179        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1180        struct btrfs_root *root = fs_info->tree_root;
1181        unsigned old_flags = sb->s_flags;
1182        unsigned long old_opts = fs_info->mount_opt;
1183        unsigned long old_compress_type = fs_info->compress_type;
1184        u64 old_max_inline = fs_info->max_inline;
1185        u64 old_alloc_start = fs_info->alloc_start;
1186        int old_thread_pool_size = fs_info->thread_pool_size;
1187        unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1188        int ret;
1189
1190        ret = btrfs_parse_options(root, data);
1191        if (ret) {
1192                ret = -EINVAL;
1193                goto restore;
1194        }
1195
1196        btrfs_resize_thread_pool(fs_info,
1197                fs_info->thread_pool_size, old_thread_pool_size);
1198
1199        if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1200                return 0;
1201
1202        if (*flags & MS_RDONLY) {
1203                sb->s_flags |= MS_RDONLY;
1204
1205                ret = btrfs_commit_super(root);
1206                if (ret)
1207                        goto restore;
1208        } else {
1209                if (fs_info->fs_devices->rw_devices == 0) {
1210                        ret = -EACCES;
1211                        goto restore;
1212                }
1213
1214                if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1215                        ret = -EINVAL;
1216                        goto restore;
1217                }
1218
1219                ret = btrfs_cleanup_fs_roots(fs_info);
1220                if (ret)
1221                        goto restore;
1222
1223                /* recover relocation */
1224                ret = btrfs_recover_relocation(root);
1225                if (ret)
1226                        goto restore;
1227
1228                ret = btrfs_resume_balance_async(fs_info);
1229                if (ret)
1230                        goto restore;
1231
1232                sb->s_flags &= ~MS_RDONLY;
1233        }
1234
1235        return 0;
1236
1237restore:
1238        /* We've hit an error - don't reset MS_RDONLY */
1239        if (sb->s_flags & MS_RDONLY)
1240                old_flags |= MS_RDONLY;
1241        sb->s_flags = old_flags;
1242        fs_info->mount_opt = old_opts;
1243        fs_info->compress_type = old_compress_type;
1244        fs_info->max_inline = old_max_inline;
1245        fs_info->alloc_start = old_alloc_start;
1246        btrfs_resize_thread_pool(fs_info,
1247                old_thread_pool_size, fs_info->thread_pool_size);
1248        fs_info->metadata_ratio = old_metadata_ratio;
1249        return ret;
1250}
1251
1252/* Used to sort the devices by max_avail(descending sort) */
1253static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1254                                       const void *dev_info2)
1255{
1256        if (((struct btrfs_device_info *)dev_info1)->max_avail >
1257            ((struct btrfs_device_info *)dev_info2)->max_avail)
1258                return -1;
1259        else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1260                 ((struct btrfs_device_info *)dev_info2)->max_avail)
1261                return 1;
1262        else
1263        return 0;
1264}
1265
1266/*
1267 * sort the devices by max_avail, in which max free extent size of each device
1268 * is stored.(Descending Sort)
1269 */
1270static inline void btrfs_descending_sort_devices(
1271                                        struct btrfs_device_info *devices,
1272                                        size_t nr_devices)
1273{
1274        sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1275             btrfs_cmp_device_free_bytes, NULL);
1276}
1277
1278/*
1279 * The helper to calc the free space on the devices that can be used to store
1280 * file data.
1281 */
1282static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1283{
1284        struct btrfs_fs_info *fs_info = root->fs_info;
1285        struct btrfs_device_info *devices_info;
1286        struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1287        struct btrfs_device *device;
1288        u64 skip_space;
1289        u64 type;
1290        u64 avail_space;
1291        u64 used_space;
1292        u64 min_stripe_size;
1293        int min_stripes = 1, num_stripes = 1;
1294        int i = 0, nr_devices;
1295        int ret;
1296
1297        nr_devices = fs_info->fs_devices->open_devices;
1298        BUG_ON(!nr_devices);
1299
1300        devices_info = kmalloc(sizeof(*devices_info) * nr_devices,
1301                               GFP_NOFS);
1302        if (!devices_info)
1303                return -ENOMEM;
1304
1305        /* calc min stripe number for data space alloction */
1306        type = btrfs_get_alloc_profile(root, 1);
1307        if (type & BTRFS_BLOCK_GROUP_RAID0) {
1308                min_stripes = 2;
1309                num_stripes = nr_devices;
1310        } else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1311                min_stripes = 2;
1312                num_stripes = 2;
1313        } else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1314                min_stripes = 4;
1315                num_stripes = 4;
1316        }
1317
1318        if (type & BTRFS_BLOCK_GROUP_DUP)
1319                min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1320        else
1321                min_stripe_size = BTRFS_STRIPE_LEN;
1322
1323        list_for_each_entry(device, &fs_devices->devices, dev_list) {
1324                if (!device->in_fs_metadata || !device->bdev)
1325                        continue;
1326
1327                avail_space = device->total_bytes - device->bytes_used;
1328
1329                /* align with stripe_len */
1330                do_div(avail_space, BTRFS_STRIPE_LEN);
1331                avail_space *= BTRFS_STRIPE_LEN;
1332
1333                /*
1334                 * In order to avoid overwritting the superblock on the drive,
1335                 * btrfs starts at an offset of at least 1MB when doing chunk
1336                 * allocation.
1337                 */
1338                skip_space = 1024 * 1024;
1339
1340                /* user can set the offset in fs_info->alloc_start. */
1341                if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1342                    device->total_bytes)
1343                        skip_space = max(fs_info->alloc_start, skip_space);
1344
1345                /*
1346                 * btrfs can not use the free space in [0, skip_space - 1],
1347                 * we must subtract it from the total. In order to implement
1348                 * it, we account the used space in this range first.
1349                 */
1350                ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1351                                                     &used_space);
1352                if (ret) {
1353                        kfree(devices_info);
1354                        return ret;
1355                }
1356
1357                /* calc the free space in [0, skip_space - 1] */
1358                skip_space -= used_space;
1359
1360                /*
1361                 * we can use the free space in [0, skip_space - 1], subtract
1362                 * it from the total.
1363                 */
1364                if (avail_space && avail_space >= skip_space)
1365                        avail_space -= skip_space;
1366                else
1367                        avail_space = 0;
1368
1369                if (avail_space < min_stripe_size)
1370                        continue;
1371
1372                devices_info[i].dev = device;
1373                devices_info[i].max_avail = avail_space;
1374
1375                i++;
1376        }
1377
1378        nr_devices = i;
1379
1380        btrfs_descending_sort_devices(devices_info, nr_devices);
1381
1382        i = nr_devices - 1;
1383        avail_space = 0;
1384        while (nr_devices >= min_stripes) {
1385                if (num_stripes > nr_devices)
1386                        num_stripes = nr_devices;
1387
1388                if (devices_info[i].max_avail >= min_stripe_size) {
1389                        int j;
1390                        u64 alloc_size;
1391
1392                        avail_space += devices_info[i].max_avail * num_stripes;
1393                        alloc_size = devices_info[i].max_avail;
1394                        for (j = i + 1 - num_stripes; j <= i; j++)
1395                                devices_info[j].max_avail -= alloc_size;
1396                }
1397                i--;
1398                nr_devices--;
1399        }
1400
1401        kfree(devices_info);
1402        *free_bytes = avail_space;
1403        return 0;
1404}
1405
1406static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1407{
1408        struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1409        struct btrfs_super_block *disk_super = fs_info->super_copy;
1410        struct list_head *head = &fs_info->space_info;
1411        struct btrfs_space_info *found;
1412        u64 total_used = 0;
1413        u64 total_free_data = 0;
1414        int bits = dentry->d_sb->s_blocksize_bits;
1415        __be32 *fsid = (__be32 *)fs_info->fsid;
1416        int ret;
1417
1418        /* holding chunk_muext to avoid allocating new chunks */
1419        mutex_lock(&fs_info->chunk_mutex);
1420        rcu_read_lock();
1421        list_for_each_entry_rcu(found, head, list) {
1422                if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1423                        total_free_data += found->disk_total - found->disk_used;
1424                        total_free_data -=
1425                                btrfs_account_ro_block_groups_free_space(found);
1426                }
1427
1428                total_used += found->disk_used;
1429        }
1430        rcu_read_unlock();
1431
1432        buf->f_namelen = BTRFS_NAME_LEN;
1433        buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
1434        buf->f_bfree = buf->f_blocks - (total_used >> bits);
1435        buf->f_bsize = dentry->d_sb->s_blocksize;
1436        buf->f_type = BTRFS_SUPER_MAGIC;
1437        buf->f_bavail = total_free_data;
1438        ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1439        if (ret) {
1440                mutex_unlock(&fs_info->chunk_mutex);
1441                return ret;
1442        }
1443        buf->f_bavail += total_free_data;
1444        buf->f_bavail = buf->f_bavail >> bits;
1445        mutex_unlock(&fs_info->chunk_mutex);
1446
1447        /* We treat it as constant endianness (it doesn't matter _which_)
1448           because we want the fsid to come out the same whether mounted
1449           on a big-endian or little-endian host */
1450        buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1451        buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1452        /* Mask in the root object ID too, to disambiguate subvols */
1453        buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1454        buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1455
1456        return 0;
1457}
1458
1459static void btrfs_kill_super(struct super_block *sb)
1460{
1461        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1462        kill_anon_super(sb);
1463        free_fs_info(fs_info);
1464}
1465
1466static struct file_system_type btrfs_fs_type = {
1467        .owner          = THIS_MODULE,
1468        .name           = "btrfs",
1469        .mount          = btrfs_mount,
1470        .kill_sb        = btrfs_kill_super,
1471        .fs_flags       = FS_REQUIRES_DEV,
1472};
1473
1474/*
1475 * used by btrfsctl to scan devices when no FS is mounted
1476 */
1477static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1478                                unsigned long arg)
1479{
1480        struct btrfs_ioctl_vol_args *vol;
1481        struct btrfs_fs_devices *fs_devices;
1482        int ret = -ENOTTY;
1483
1484        if (!capable(CAP_SYS_ADMIN))
1485                return -EPERM;
1486
1487        vol = memdup_user((void __user *)arg, sizeof(*vol));
1488        if (IS_ERR(vol))
1489                return PTR_ERR(vol);
1490
1491        switch (cmd) {
1492        case BTRFS_IOC_SCAN_DEV:
1493                ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1494                                            &btrfs_fs_type, &fs_devices);
1495                break;
1496        case BTRFS_IOC_DEVICES_READY:
1497                ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1498                                            &btrfs_fs_type, &fs_devices);
1499                if (ret)
1500                        break;
1501                ret = !(fs_devices->num_devices == fs_devices->total_devices);
1502                break;
1503        }
1504
1505        kfree(vol);
1506        return ret;
1507}
1508
1509static int btrfs_freeze(struct super_block *sb)
1510{
1511        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1512        mutex_lock(&fs_info->transaction_kthread_mutex);
1513        mutex_lock(&fs_info->cleaner_mutex);
1514        return 0;
1515}
1516
1517static int btrfs_unfreeze(struct super_block *sb)
1518{
1519        struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1520        mutex_unlock(&fs_info->cleaner_mutex);
1521        mutex_unlock(&fs_info->transaction_kthread_mutex);
1522        return 0;
1523}
1524
1525static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1526{
1527        struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1528        struct btrfs_fs_devices *cur_devices;
1529        struct btrfs_device *dev, *first_dev = NULL;
1530        struct list_head *head;
1531        struct rcu_string *name;
1532
1533        mutex_lock(&fs_info->fs_devices->device_list_mutex);
1534        cur_devices = fs_info->fs_devices;
1535        while (cur_devices) {
1536                head = &cur_devices->devices;
1537                list_for_each_entry(dev, head, dev_list) {
1538                        if (dev->missing)
1539                                continue;
1540                        if (!first_dev || dev->devid < first_dev->devid)
1541                                first_dev = dev;
1542                }
1543                cur_devices = cur_devices->seed;
1544        }
1545
1546        if (first_dev) {
1547                rcu_read_lock();
1548                name = rcu_dereference(first_dev->name);
1549                seq_escape(m, name->str, " \t\n\\");
1550                rcu_read_unlock();
1551        } else {
1552                WARN_ON(1);
1553        }
1554        mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1555        return 0;
1556}
1557
1558static const struct super_operations btrfs_super_ops = {
1559        .drop_inode     = btrfs_drop_inode,
1560        .evict_inode    = btrfs_evict_inode,
1561        .put_super      = btrfs_put_super,
1562        .sync_fs        = btrfs_sync_fs,
1563        .show_options   = btrfs_show_options,
1564        .show_devname   = btrfs_show_devname,
1565        .write_inode    = btrfs_write_inode,
1566        .alloc_inode    = btrfs_alloc_inode,
1567        .destroy_inode  = btrfs_destroy_inode,
1568        .statfs         = btrfs_statfs,
1569        .remount_fs     = btrfs_remount,
1570        .freeze_fs      = btrfs_freeze,
1571        .unfreeze_fs    = btrfs_unfreeze,
1572};
1573
1574static const struct file_operations btrfs_ctl_fops = {
1575        .unlocked_ioctl  = btrfs_control_ioctl,
1576        .compat_ioctl = btrfs_control_ioctl,
1577        .owner   = THIS_MODULE,
1578        .llseek = noop_llseek,
1579};
1580
1581static struct miscdevice btrfs_misc = {
1582        .minor          = BTRFS_MINOR,
1583        .name           = "btrfs-control",
1584        .fops           = &btrfs_ctl_fops
1585};
1586
1587MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1588MODULE_ALIAS("devname:btrfs-control");
1589
1590static int btrfs_interface_init(void)
1591{
1592        return misc_register(&btrfs_misc);
1593}
1594
1595static void btrfs_interface_exit(void)
1596{
1597        if (misc_deregister(&btrfs_misc) < 0)
1598                printk(KERN_INFO "misc_deregister failed for control device");
1599}
1600
1601static int __init init_btrfs_fs(void)
1602{
1603        int err;
1604
1605        err = btrfs_init_sysfs();
1606        if (err)
1607                return err;
1608
1609        btrfs_init_compress();
1610
1611        err = btrfs_init_cachep();
1612        if (err)
1613                goto free_compress;
1614
1615        err = extent_io_init();
1616        if (err)
1617                goto free_cachep;
1618
1619        err = extent_map_init();
1620        if (err)
1621                goto free_extent_io;
1622
1623        err = btrfs_delayed_inode_init();
1624        if (err)
1625                goto free_extent_map;
1626
1627        err = btrfs_interface_init();
1628        if (err)
1629                goto free_delayed_inode;
1630
1631        err = register_filesystem(&btrfs_fs_type);
1632        if (err)
1633                goto unregister_ioctl;
1634
1635        btrfs_init_lockdep();
1636
1637        printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
1638        return 0;
1639
1640unregister_ioctl:
1641        btrfs_interface_exit();
1642free_delayed_inode:
1643        btrfs_delayed_inode_exit();
1644free_extent_map:
1645        extent_map_exit();
1646free_extent_io:
1647        extent_io_exit();
1648free_cachep:
1649        btrfs_destroy_cachep();
1650free_compress:
1651        btrfs_exit_compress();
1652        btrfs_exit_sysfs();
1653        return err;
1654}
1655
1656static void __exit exit_btrfs_fs(void)
1657{
1658        btrfs_destroy_cachep();
1659        btrfs_delayed_inode_exit();
1660        extent_map_exit();
1661        extent_io_exit();
1662        btrfs_interface_exit();
1663        unregister_filesystem(&btrfs_fs_type);
1664        btrfs_exit_sysfs();
1665        btrfs_cleanup_fs_uuids();
1666        btrfs_exit_compress();
1667}
1668
1669module_init(init_btrfs_fs)
1670module_exit(exit_btrfs_fs)
1671
1672MODULE_LICENSE("GPL");
1673
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.