fs/btrfs/super.c

   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/string.h>
  28 #include <linux/smp_lock.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 "ctree.h"
  41 #include "disk-io.h"
  42 #include "transaction.h"
  43 #include "btrfs_inode.h"
  44 #include "ioctl.h"
  45 #include "print-tree.h"
  46 #include "xattr.h"
  47 #include "volumes.h"
  48 #include "version.h"
  49 #include "export.h"
  50 #include "compression.h"
  51
  52 #define BTRFS_SUPER_MAGIC 0x9123683E
  53
  54 static struct super_operations btrfs_super_ops;
  55
  56 static void btrfs_put_super (struct super_block * sb)
  57 {
  58         struct btrfs_root *root = btrfs_sb(sb);
  59         struct btrfs_fs_info *fs = root->fs_info;
  60         int ret;
  61
  62         ret = close_ctree(root);
  63         if (ret) {
  64                 printk("close ctree returns %d\n", ret);
  65         }
  66         btrfs_sysfs_del_super(fs);
  67         sb->s_fs_info = NULL;
  68 }
  69
  70 enum {
  71         Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
  72         Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
  73         Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_err,
  74 };
  75
  76 static match_table_t tokens = {
  77         {Opt_degraded, "degraded"},
  78         {Opt_subvol, "subvol=%s"},
  79         {Opt_device, "device=%s"},
  80         {Opt_nodatasum, "nodatasum"},
  81         {Opt_nodatacow, "nodatacow"},
  82         {Opt_nobarrier, "nobarrier"},
  83         {Opt_max_extent, "max_extent=%s"},
  84         {Opt_max_inline, "max_inline=%s"},
  85         {Opt_alloc_start, "alloc_start=%s"},
  86         {Opt_thread_pool, "thread_pool=%d"},
  87         {Opt_compress, "compress"},
  88         {Opt_ssd, "ssd"},
  89         {Opt_noacl, "noacl"},
  90         {Opt_err, NULL},
  91 };
  92
  93 u64 btrfs_parse_size(char *str)
  94 {
  95         u64 res;
  96         int mult = 1;
  97         char *end;
  98         char last;
  99
 100         res = simple_strtoul(str, &end, 10);
 101
 102         last = end[0];
 103         if (isalpha(last)) {
 104                 last = tolower(last);
 105                 switch (last) {
 106                 case 'g':
 107                         mult *= 1024;
 108                 case 'm':
 109                         mult *= 1024;
 110                 case 'k':
 111                         mult *= 1024;
 112                 }
 113                 res = res * mult;
 114         }
 115         return res;
 116 }
 117
 118 /*
 119  * Regular mount options parser.  Everything that is needed only when
 120  * reading in a new superblock is parsed here.
 121  */
 122 int btrfs_parse_options(struct btrfs_root *root, char *options)
 123 {
 124         struct btrfs_fs_info *info = root->fs_info;
 125         substring_t args[MAX_OPT_ARGS];
 126         char *p, *num;
 127         int intarg;
 128
 129         if (!options)
 130                 return 0;
 131
 132         /*
 133          * strsep changes the string, duplicate it because parse_options
 134          * gets called twice
 135          */
 136         options = kstrdup(options, GFP_NOFS);
 137         if (!options)
 138                 return -ENOMEM;
 139
 140
 141         while ((p = strsep(&options, ",")) != NULL) {
 142                 int token;
 143                 if (!*p)
 144                         continue;
 145
 146                 token = match_token(p, tokens, args);
 147                 switch (token) {
 148                 case Opt_degraded:
 149                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 150                         btrfs_set_opt(info->mount_opt, DEGRADED);
 151                         break;
 152                 case Opt_subvol:
 153                 case Opt_device:
 154                         /*
 155                          * These are parsed by btrfs_parse_early_options
 156                          * and can be happily ignored here.
 157                          */
 158                         break;
 159                 case Opt_nodatasum:
 160                         printk(KERN_INFO "btrfs: setting nodatacsum\n");
 161                         btrfs_set_opt(info->mount_opt, NODATASUM);
 162                         break;
 163                 case Opt_nodatacow:
 164                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 165                         btrfs_set_opt(info->mount_opt, NODATACOW);
 166                         btrfs_set_opt(info->mount_opt, NODATASUM);
 167                         break;
 168                 case Opt_compress:
 169                         printk(KERN_INFO "btrfs: use compression\n");
 170                         btrfs_set_opt(info->mount_opt, COMPRESS);
 171                         break;
 172                 case Opt_ssd:
 173                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 174                         btrfs_set_opt(info->mount_opt, SSD);
 175                         break;
 176                 case Opt_nobarrier:
 177                         printk(KERN_INFO "btrfs: turning off barriers\n");
 178                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 179                         break;
 180                 case Opt_thread_pool:
 181                         intarg = 0;
 182                         match_int(&args[0], &intarg);
 183                         if (intarg) {
 184                                 info->thread_pool_size = intarg;
 185                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 186                                        info->thread_pool_size);
 187                         }
 188                         break;
 189                 case Opt_max_extent:
 190                         num = match_strdup(&args[0]);
 191                         if (num) {
 192                                 info->max_extent = btrfs_parse_size(num);
 193                                 kfree(num);
 194
 195                                 info->max_extent = max_t(u64,
 196                                         info->max_extent, root->sectorsize);
 197                                 printk(KERN_INFO "btrfs: max_extent at %llu\n",
 198                                        info->max_extent);
 199                         }
 200                         break;
 201                 case Opt_max_inline:
 202                         num = match_strdup(&args[0]);
 203                         if (num) {
 204                                 info->max_inline = btrfs_parse_size(num);
 205                                 kfree(num);
 206
 207                                 if (info->max_inline) {
 208                                         info->max_inline = max_t(u64,
 209                                                 info->max_inline,
 210                                                 root->sectorsize);
 211                                 }
 212                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 213                                         info->max_inline);
 214                         }
 215                         break;
 216                 case Opt_alloc_start:
 217                         num = match_strdup(&args[0]);
 218                         if (num) {
 219                                 info->alloc_start = btrfs_parse_size(num);
 220                                 kfree(num);
 221                                 printk(KERN_INFO
 222                                         "btrfs: allocations start at %llu\n",
 223                                         info->alloc_start);
 224                         }
 225                         break;
 226                 case Opt_noacl:
 227                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 228                         break;
 229                 default:
 230                         break;
 231                 }
 232         }
 233         kfree(options);
 234         return 0;
 235 }
 236
 237 /*
 238  * Parse mount options that are required early in the mount process.
 239  *
 240  * All other options will be parsed on much later in the mount process and
 241  * only when we need to allocate a new super block.
 242  */
 243 static int btrfs_parse_early_options(const char *options, int flags,
 244                 void *holder, char **subvol_name,
 245                 struct btrfs_fs_devices **fs_devices)
 246 {
 247         substring_t args[MAX_OPT_ARGS];
 248         char *opts, *p;
 249         int error = 0;
 250
 251         if (!options)
 252                 goto out;
 253
 254         /*
 255          * strsep changes the string, duplicate it because parse_options
 256          * gets called twice
 257          */
 258         opts = kstrdup(options, GFP_KERNEL);
 259         if (!opts)
 260                 return -ENOMEM;
 261
 262         while ((p = strsep(&opts, ",")) != NULL) {
 263                 int token;
 264                 if (!*p)
 265                         continue;
 266
 267                 token = match_token(p, tokens, args);
 268                 switch (token) {
 269                 case Opt_subvol:
 270                         *subvol_name = match_strdup(&args[0]);
 271                         break;
 272                 case Opt_device:
 273                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 274                                         flags, holder, fs_devices);
 275                         if (error)
 276                                 goto out_free_opts;
 277                         break;
 278                 default:
 279                         break;
 280                 }
 281         }
 282
 283  out_free_opts:
 284         kfree(opts);
 285  out:
 286         /*
 287          * If no subvolume name is specified we use the default one.  Allocate
 288          * a copy of the string "default" here so that code later in the
 289          * mount path doesn't care if it's the default volume or another one.
 290          */
 291         if (!*subvol_name) {
 292                 *subvol_name = kstrdup("default", GFP_KERNEL);
 293                 if (!*subvol_name)
 294                         return -ENOMEM;
 295         }
 296         return error;
 297 }
 298
 299 static int btrfs_fill_super(struct super_block * sb,
 300                             struct btrfs_fs_devices *fs_devices,
 301                             void * data, int silent)
 302 {
 303         struct inode * inode;
 304         struct dentry * root_dentry;
 305         struct btrfs_super_block *disk_super;
 306         struct btrfs_root *tree_root;
 307         struct btrfs_inode *bi;
 308         int err;
 309
 310         sb->s_maxbytes = MAX_LFS_FILESIZE;
 311         sb->s_magic = BTRFS_SUPER_MAGIC;
 312         sb->s_op = &btrfs_super_ops;
 313         sb->s_export_op = &btrfs_export_ops;
 314         sb->s_xattr = btrfs_xattr_handlers;
 315         sb->s_time_gran = 1;
 316         sb->s_flags |= MS_POSIXACL;
 317
 318         tree_root = open_ctree(sb, fs_devices, (char *)data);
 319
 320         if (IS_ERR(tree_root)) {
 321                 printk("btrfs: open_ctree failed\n");
 322                 return PTR_ERR(tree_root);
 323         }
 324         sb->s_fs_info = tree_root;
 325         disk_super = &tree_root->fs_info->super_copy;
 326         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 327                                   tree_root);
 328         bi = BTRFS_I(inode);
 329         bi->location.objectid = inode->i_ino;
 330         bi->location.offset = 0;
 331         bi->root = tree_root;
 332
 333         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 334
 335         if (!inode) {
 336                 err = -ENOMEM;
 337                 goto fail_close;
 338         }
 339         if (inode->i_state & I_NEW) {
 340                 btrfs_read_locked_inode(inode);
 341                 unlock_new_inode(inode);
 342         }
 343
 344         root_dentry = d_alloc_root(inode);
 345         if (!root_dentry) {
 346                 iput(inode);
 347                 err = -ENOMEM;
 348                 goto fail_close;
 349         }
 350
 351         /* this does the super kobj at the same time */
 352         err = btrfs_sysfs_add_super(tree_root->fs_info);
 353         if (err)
 354                 goto fail_close;
 355
 356         sb->s_root = root_dentry;
 357
 358         save_mount_options(sb, data);
 359         return 0;
 360
 361 fail_close:
 362         close_ctree(tree_root);
 363         return err;
 364 }
 365
 366 int btrfs_sync_fs(struct super_block *sb, int wait)
 367 {
 368         struct btrfs_trans_handle *trans;
 369         struct btrfs_root *root;
 370         int ret;
 371         root = btrfs_sb(sb);
 372
 373         sb->s_dirt = 0;
 374         if (!wait) {
 375                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 376                 return 0;
 377         }
 378
 379         btrfs_start_delalloc_inodes(root);
 380         btrfs_wait_ordered_extents(root, 0);
 381
 382         btrfs_clean_old_snapshots(root);
 383         trans = btrfs_start_transaction(root, 1);
 384         ret = btrfs_commit_transaction(trans, root);
 385         sb->s_dirt = 0;
 386         return ret;
 387 }
 388
 389 static void btrfs_write_super(struct super_block *sb)
 390 {
 391         sb->s_dirt = 0;
 392 }
 393
 394 static int btrfs_test_super(struct super_block *s, void *data)
 395 {
 396         struct btrfs_fs_devices *test_fs_devices = data;
 397         struct btrfs_root *root = btrfs_sb(s);
 398
 399         return root->fs_info->fs_devices == test_fs_devices;
 400 }
 401
 402 /*
 403  * Find a superblock for the given device / mount point.
 404  *
 405  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 406  *        for multiple device setup.  Make sure to keep it in sync.
 407  */
 408 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 409                 const char *dev_name, void *data, struct vfsmount *mnt)
 410 {
 411         char *subvol_name = NULL;
 412         struct block_device *bdev = NULL;
 413         struct super_block *s;
 414         struct dentry *root;
 415         struct btrfs_fs_devices *fs_devices = NULL;
 416         int error = 0;
 417
 418         error = btrfs_parse_early_options(data, flags, fs_type,
 419                                           &subvol_name, &fs_devices);
 420         if (error)
 421                 goto error;
 422
 423         error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
 424         if (error)
 425                 goto error_free_subvol_name;
 426
 427         error = btrfs_open_devices(fs_devices, flags, fs_type);
 428         if (error)
 429                 goto error_free_subvol_name;
 430
 431         bdev = fs_devices->latest_bdev;
 432         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 433         if (IS_ERR(s))
 434                 goto error_s;
 435
 436         if (s->s_root) {
 437                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 438                         up_write(&s->s_umount);
 439                         deactivate_super(s);
 440                         error = -EBUSY;
 441                         goto error_bdev;
 442                 }
 443
 444         } else {
 445                 char b[BDEVNAME_SIZE];
 446
 447                 s->s_flags = flags;
 448                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 449                 error = btrfs_fill_super(s, fs_devices, data,
 450                                          flags & MS_SILENT ? 1 : 0);
 451                 if (error) {
 452                         up_write(&s->s_umount);
 453                         deactivate_super(s);
 454                         goto error;
 455                 }
 456
 457                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 458                 s->s_flags |= MS_ACTIVE;
 459         }
 460
 461         if (!strcmp(subvol_name, "."))
 462                 root = dget(s->s_root);
 463         else {
 464                 mutex_lock(&s->s_root->d_inode->i_mutex);
 465                 root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name));
 466                 mutex_unlock(&s->s_root->d_inode->i_mutex);
 467                 if (IS_ERR(root)) {
 468                         up_write(&s->s_umount);
 469                         deactivate_super(s);
 470                         error = PTR_ERR(root);
 471                         goto error;
 472                 }
 473                 if (!root->d_inode) {
 474                         dput(root);
 475                         up_write(&s->s_umount);
 476                         deactivate_super(s);
 477                         error = -ENXIO;
 478                         goto error;
 479                 }
 480         }
 481
 482         mnt->mnt_sb = s;
 483         mnt->mnt_root = root;
 484
 485         kfree(subvol_name);
 486         return 0;
 487
 488 error_s:
 489         error = PTR_ERR(s);
 490 error_bdev:
 491         btrfs_close_devices(fs_devices);
 492 error_free_subvol_name:
 493         kfree(subvol_name);
 494 error:
 495         return error;
 496 }
 497
 498 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 499 {
 500         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 501         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 502         int bits = dentry->d_sb->s_blocksize_bits;
 503         __be32 *fsid = (__be32 *)root->fs_info->fsid;
 504
 505         buf->f_namelen = BTRFS_NAME_LEN;
 506         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 507         buf->f_bfree = buf->f_blocks -
 508                 (btrfs_super_bytes_used(disk_super) >> bits);
 509         buf->f_bavail = buf->f_bfree;
 510         buf->f_bsize = dentry->d_sb->s_blocksize;
 511         buf->f_type = BTRFS_SUPER_MAGIC;
 512         /* We treat it as constant endianness (it doesn't matter _which_)
 513            because we want the fsid to come out the same whether mounted
 514            on a big-endian or little-endian host */
 515         buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
 516         buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
 517         /* Mask in the root object ID too, to disambiguate subvols */
 518         buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
 519         buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
 520
 521         return 0;
 522 }
 523
 524 static struct file_system_type btrfs_fs_type = {
 525         .owner          = THIS_MODULE,
 526         .name           = "btrfs",
 527         .get_sb         = btrfs_get_sb,
 528         .kill_sb        = kill_anon_super,
 529         .fs_flags       = FS_REQUIRES_DEV,
 530 };
 531
 532 /*
 533  * used by btrfsctl to scan devices when no FS is mounted
 534  */
 535 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 536                                 unsigned long arg)
 537 {
 538         struct btrfs_ioctl_vol_args *vol;
 539         struct btrfs_fs_devices *fs_devices;
 540         int ret = 0;
 541         int len;
 542
 543         vol = kmalloc(sizeof(*vol), GFP_KERNEL);
 544         if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
 545                 ret = -EFAULT;
 546                 goto out;
 547         }
 548         len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
 549         switch (cmd) {
 550         case BTRFS_IOC_SCAN_DEV:
 551                 ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
 552                                             &btrfs_fs_type, &fs_devices);
 553                 break;
 554         }
 555 out:
 556         kfree(vol);
 557         return ret;
 558 }
 559
 560 static void btrfs_write_super_lockfs(struct super_block *sb)
 561 {
 562         struct btrfs_root *root = btrfs_sb(sb);
 563         mutex_lock(&root->fs_info->transaction_kthread_mutex);
 564         mutex_lock(&root->fs_info->cleaner_mutex);
 565 }
 566
 567 static void btrfs_unlockfs(struct super_block *sb)
 568 {
 569         struct btrfs_root *root = btrfs_sb(sb);
 570         mutex_unlock(&root->fs_info->cleaner_mutex);
 571         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 572 }
 573
 574 static struct super_operations btrfs_super_ops = {
 575         .delete_inode   = btrfs_delete_inode,
 576         .put_super      = btrfs_put_super,
 577         .write_super    = btrfs_write_super,
 578         .sync_fs        = btrfs_sync_fs,
 579         .show_options   = generic_show_options,
 580         .write_inode    = btrfs_write_inode,
 581         .dirty_inode    = btrfs_dirty_inode,
 582         .alloc_inode    = btrfs_alloc_inode,
 583         .destroy_inode  = btrfs_destroy_inode,
 584         .statfs         = btrfs_statfs,
 585         .write_super_lockfs = btrfs_write_super_lockfs,
 586         .unlockfs       = btrfs_unlockfs,
 587 };
 588
 589 static const struct file_operations btrfs_ctl_fops = {
 590         .unlocked_ioctl  = btrfs_control_ioctl,
 591         .compat_ioctl = btrfs_control_ioctl,
 592         .owner   = THIS_MODULE,
 593 };
 594
 595 static struct miscdevice btrfs_misc = {
 596         .minor          = MISC_DYNAMIC_MINOR,
 597         .name           = "btrfs-control",
 598         .fops           = &btrfs_ctl_fops
 599 };
 600
 601 static int btrfs_interface_init(void)
 602 {
 603         return misc_register(&btrfs_misc);
 604 }
 605
 606 void btrfs_interface_exit(void)
 607 {
 608         if (misc_deregister(&btrfs_misc) < 0)
 609                 printk("misc_deregister failed for control device");
 610 }
 611
 612 static int __init init_btrfs_fs(void)
 613 {
 614         int err;
 615
 616         err = btrfs_init_sysfs();
 617         if (err)
 618                 return err;
 619
 620         err = btrfs_init_cachep();
 621         if (err)
 622                 goto free_sysfs;
 623
 624         err = extent_io_init();
 625         if (err)
 626                 goto free_cachep;
 627
 628         err = extent_map_init();
 629         if (err)
 630                 goto free_extent_io;
 631
 632         err = btrfs_interface_init();
 633         if (err)
 634                 goto free_extent_map;
 635
 636         err = register_filesystem(&btrfs_fs_type);
 637         if (err)
 638                 goto unregister_ioctl;
 639
 640         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 641         return 0;
 642
 643 unregister_ioctl:
 644         btrfs_interface_exit();
 645 free_extent_map:
 646         extent_map_exit();
 647 free_extent_io:
 648         extent_io_exit();
 649 free_cachep:
 650         btrfs_destroy_cachep();
 651 free_sysfs:
 652         btrfs_exit_sysfs();
 653         return err;
 654 }
 655
 656 static void __exit exit_btrfs_fs(void)
 657 {
 658         btrfs_destroy_cachep();
 659         extent_map_exit();
 660         extent_io_exit();
 661         btrfs_interface_exit();
 662         unregister_filesystem(&btrfs_fs_type);
 663         btrfs_exit_sysfs();
 664         btrfs_cleanup_fs_uuids();
 665         btrfs_zlib_exit();
 666 }
 667
 668 module_init(init_btrfs_fs)
 669 module_exit(exit_btrfs_fs)
 670
 671 MODULE_LICENSE("GPL");