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