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
  50 #define BTRFS_SUPER_MAGIC 0x9123683E
  51
  52 static struct super_operations btrfs_super_ops;
  53
  54 static void btrfs_put_super (struct super_block * sb)
  55 {
  56         struct btrfs_root *root = btrfs_sb(sb);
  57         struct btrfs_fs_info *fs = root->fs_info;
  58         int ret;
  59
  60         ret = close_ctree(root);
  61         if (ret) {
  62                 printk("close ctree returns %d\n", ret);
  63         }
  64         btrfs_sysfs_del_super(fs);
  65         sb->s_fs_info = NULL;
  66 }
  67
  68 enum {
  69         Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
  70         Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
  71         Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_err,
  72 };
  73
  74 static match_table_t tokens = {
  75         {Opt_degraded, "degraded"},
  76         {Opt_subvol, "subvol=%s"},
  77         {Opt_device, "device=%s"},
  78         {Opt_nodatasum, "nodatasum"},
  79         {Opt_nodatacow, "nodatacow"},
  80         {Opt_nobarrier, "nobarrier"},
  81         {Opt_max_extent, "max_extent=%s"},
  82         {Opt_max_inline, "max_inline=%s"},
  83         {Opt_alloc_start, "alloc_start=%s"},
  84         {Opt_thread_pool, "thread_pool=%d"},
  85         {Opt_ssd, "ssd"},
  86         {Opt_noacl, "noacl"},
  87         {Opt_err, NULL},
  88 };
  89
  90 u64 btrfs_parse_size(char *str)
  91 {
  92         u64 res;
  93         int mult = 1;
  94         char *end;
  95         char last;
  96
  97         res = simple_strtoul(str, &end, 10);
  98
  99         last = end[0];
 100         if (isalpha(last)) {
 101                 last = tolower(last);
 102                 switch (last) {
 103                 case 'g':
 104                         mult *= 1024;
 105                 case 'm':
 106                         mult *= 1024;
 107                 case 'k':
 108                         mult *= 1024;
 109                 }
 110                 res = res * mult;
 111         }
 112         return res;
 113 }
 114
 115 /*
 116  * Regular mount options parser.  Everything that is needed only when
 117  * reading in a new superblock is parsed here.
 118  */
 119 int btrfs_parse_options(struct btrfs_root *root, char *options)
 120 {
 121         struct btrfs_fs_info *info = root->fs_info;
 122         substring_t args[MAX_OPT_ARGS];
 123         char *p, *num;
 124         int intarg;
 125
 126         if (!options)
 127                 return 0;
 128
 129         /*
 130          * strsep changes the string, duplicate it because parse_options
 131          * gets called twice
 132          */
 133         options = kstrdup(options, GFP_NOFS);
 134         if (!options)
 135                 return -ENOMEM;
 136
 137
 138         while ((p = strsep(&options, ",")) != NULL) {
 139                 int token;
 140                 if (!*p)
 141                         continue;
 142
 143                 token = match_token(p, tokens, args);
 144                 switch (token) {
 145                 case Opt_degraded:
 146                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 147                         btrfs_set_opt(info->mount_opt, DEGRADED);
 148                         break;
 149                 case Opt_subvol:
 150                 case Opt_device:
 151                         /*
 152                          * These are parsed by btrfs_parse_early_options
 153                          * and can be happily ignored here.
 154                          */
 155                         break;
 156                 case Opt_nodatasum:
 157                         printk(KERN_INFO "btrfs: setting nodatacsum\n");
 158                         btrfs_set_opt(info->mount_opt, NODATASUM);
 159                         break;
 160                 case Opt_nodatacow:
 161                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 162                         btrfs_set_opt(info->mount_opt, NODATACOW);
 163                         btrfs_set_opt(info->mount_opt, NODATASUM);
 164                         break;
 165                 case Opt_ssd:
 166                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 167                         btrfs_set_opt(info->mount_opt, SSD);
 168                         break;
 169                 case Opt_nobarrier:
 170                         printk(KERN_INFO "btrfs: turning off barriers\n");
 171                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 172                         break;
 173                 case Opt_thread_pool:
 174                         intarg = 0;
 175                         match_int(&args[0], &intarg);
 176                         if (intarg) {
 177                                 info->thread_pool_size = intarg;
 178                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 179                                        info->thread_pool_size);
 180                         }
 181                         break;
 182                 case Opt_max_extent:
 183                         num = match_strdup(&args[0]);
 184                         if (num) {
 185                                 info->max_extent = btrfs_parse_size(num);
 186                                 kfree(num);
 187
 188                                 info->max_extent = max_t(u64,
 189                                         info->max_extent, root->sectorsize);
 190                                 printk(KERN_INFO "btrfs: max_extent at %llu\n",
 191                                        info->max_extent);
 192                         }
 193                         break;
 194                 case Opt_max_inline:
 195                         num = match_strdup(&args[0]);
 196                         if (num) {
 197                                 info->max_inline = btrfs_parse_size(num);
 198                                 kfree(num);
 199
 200                                 if (info->max_inline) {
 201                                         info->max_inline = max_t(u64,
 202                                                 info->max_inline,
 203                                                 root->sectorsize);
 204                                 }
 205                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 206                                         info->max_inline);
 207                         }
 208                         break;
 209                 case Opt_alloc_start:
 210                         num = match_strdup(&args[0]);
 211                         if (num) {
 212                                 info->alloc_start = btrfs_parse_size(num);
 213                                 kfree(num);
 214                                 printk(KERN_INFO
 215                                         "btrfs: allocations start at %llu\n",
 216                                         info->alloc_start);
 217                         }
 218                         break;
 219                 case Opt_noacl:
 220                         root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 221                         break;
 222                 default:
 223                         break;
 224                 }
 225         }
 226         kfree(options);
 227         return 0;
 228 }
 229
 230 /*
 231  * Parse mount options that are required early in the mount process.
 232  *
 233  * All other options will be parsed on much later in the mount process and
 234  * only when we need to allocate a new super block.
 235  */
 236 static int btrfs_parse_early_options(const char *options, int flags,
 237                 void *holder, char **subvol_name,
 238                 struct btrfs_fs_devices **fs_devices)
 239 {
 240         substring_t args[MAX_OPT_ARGS];
 241         char *opts, *p;
 242         int error = 0;
 243
 244         if (!options)
 245                 goto out;
 246
 247         /*
 248          * strsep changes the string, duplicate it because parse_options
 249          * gets called twice
 250          */
 251         opts = kstrdup(options, GFP_KERNEL);
 252         if (!opts)
 253                 return -ENOMEM;
 254
 255         while ((p = strsep(&opts, ",")) != NULL) {
 256                 int token;
 257                 if (!*p)
 258                         continue;
 259
 260                 token = match_token(p, tokens, args);
 261                 switch (token) {
 262                 case Opt_subvol:
 263                         *subvol_name = match_strdup(&args[0]);
 264                         break;
 265                 case Opt_device:
 266                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 267                                         flags, holder, fs_devices);
 268                         if (error)
 269                                 goto out_free_opts;
 270                         break;
 271                 default:
 272                         break;
 273                 }
 274         }
 275
 276  out_free_opts:
 277         kfree(opts);
 278  out:
 279         /*
 280          * If no subvolume name is specified we use the default one.  Allocate
 281          * a copy of the string "default" here so that code later in the
 282          * mount path doesn't care if it's the default volume or another one.
 283          */
 284         if (!*subvol_name) {
 285                 *subvol_name = kstrdup("default", GFP_KERNEL);
 286                 if (!*subvol_name)
 287                         return -ENOMEM;
 288         }
 289         return error;
 290 }
 291
 292 static int btrfs_fill_super(struct super_block * sb,
 293                             struct btrfs_fs_devices *fs_devices,
 294                             void * data, int silent)
 295 {
 296         struct inode * inode;
 297         struct dentry * root_dentry;
 298         struct btrfs_super_block *disk_super;
 299         struct btrfs_root *tree_root;
 300         struct btrfs_inode *bi;
 301         int err;
 302
 303         sb->s_maxbytes = MAX_LFS_FILESIZE;
 304         sb->s_magic = BTRFS_SUPER_MAGIC;
 305         sb->s_op = &btrfs_super_ops;
 306         sb->s_xattr = btrfs_xattr_handlers;
 307         sb->s_time_gran = 1;
 308         sb->s_flags |= MS_POSIXACL;
 309
 310         tree_root = open_ctree(sb, fs_devices, (char *)data);
 311
 312         if (IS_ERR(tree_root)) {
 313                 printk("btrfs: open_ctree failed\n");
 314                 return PTR_ERR(tree_root);
 315         }
 316         sb->s_fs_info = tree_root;
 317         disk_super = &tree_root->fs_info->super_copy;
 318         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 319                                   tree_root);
 320         bi = BTRFS_I(inode);
 321         bi->location.objectid = inode->i_ino;
 322         bi->location.offset = 0;
 323         bi->root = tree_root;
 324
 325         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 326
 327         if (!inode) {
 328                 err = -ENOMEM;
 329                 goto fail_close;
 330         }
 331         if (inode->i_state & I_NEW) {
 332                 btrfs_read_locked_inode(inode);
 333                 unlock_new_inode(inode);
 334         }
 335
 336         root_dentry = d_alloc_root(inode);
 337         if (!root_dentry) {
 338                 iput(inode);
 339                 err = -ENOMEM;
 340                 goto fail_close;
 341         }
 342
 343         /* this does the super kobj at the same time */
 344         err = btrfs_sysfs_add_super(tree_root->fs_info);
 345         if (err)
 346                 goto fail_close;
 347
 348         sb->s_root = root_dentry;
 349
 350 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
 351         save_mount_options(sb, data);
 352 #endif
 353
 354         return 0;
 355
 356 fail_close:
 357         close_ctree(tree_root);
 358         return err;
 359 }
 360
 361 int btrfs_sync_fs(struct super_block *sb, int wait)
 362 {
 363         struct btrfs_trans_handle *trans;
 364         struct btrfs_root *root;
 365         int ret;
 366         root = btrfs_sb(sb);
 367
 368         sb->s_dirt = 0;
 369         if (!wait) {
 370                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 371                 return 0;
 372         }
 373         btrfs_clean_old_snapshots(root);
 374         trans = btrfs_start_transaction(root, 1);
 375         ret = btrfs_commit_transaction(trans, root);
 376         sb->s_dirt = 0;
 377         return ret;
 378 }
 379
 380 static void btrfs_write_super(struct super_block *sb)
 381 {
 382         sb->s_dirt = 0;
 383 }
 384
 385 static int btrfs_test_super(struct super_block *s, void *data)
 386 {
 387         struct btrfs_fs_devices *test_fs_devices = data;
 388         struct btrfs_root *root = btrfs_sb(s);
 389
 390         return root->fs_info->fs_devices == test_fs_devices;
 391 }
 392
 393 /*
 394  * Find a superblock for the given device / mount point.
 395  *
 396  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 397  *        for multiple device setup.  Make sure to keep it in sync.
 398  */
 399 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 400                 const char *dev_name, void *data, struct vfsmount *mnt)
 401 {
 402         char *subvol_name = NULL;
 403         struct block_device *bdev = NULL;
 404         struct super_block *s;
 405         struct dentry *root;
 406         struct btrfs_fs_devices *fs_devices = NULL;
 407         int error = 0;
 408
 409         error = btrfs_parse_early_options(data, flags, fs_type,
 410                                           &subvol_name, &fs_devices);
 411         if (error)
 412                 goto error;
 413
 414         error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
 415         if (error)
 416                 goto error_free_subvol_name;
 417
 418         error = btrfs_open_devices(fs_devices, flags, fs_type);
 419         if (error)
 420                 goto error_free_subvol_name;
 421
 422         bdev = fs_devices->latest_bdev;
 423         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 424         if (IS_ERR(s))
 425                 goto error_s;
 426
 427         if (s->s_root) {
 428                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 429                         up_write(&s->s_umount);
 430                         deactivate_super(s);
 431                         error = -EBUSY;
 432                         goto error_bdev;
 433                 }
 434
 435         } else {
 436                 char b[BDEVNAME_SIZE];
 437
 438                 s->s_flags = flags;
 439                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 440                 error = btrfs_fill_super(s, fs_devices, data,
 441                                          flags & MS_SILENT ? 1 : 0);
 442                 if (error) {
 443                         up_write(&s->s_umount);
 444                         deactivate_super(s);
 445                         goto error;
 446                 }
 447
 448                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 449                 s->s_flags |= MS_ACTIVE;
 450         }
 451
 452         mutex_lock(&s->s_root->d_inode->i_mutex);
 453         root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name));
 454         mutex_unlock(&s->s_root->d_inode->i_mutex);
 455         if (IS_ERR(root)) {
 456                 up_write(&s->s_umount);
 457                 deactivate_super(s);
 458                 error = PTR_ERR(root);
 459                 goto error;
 460         }
 461         if (!root->d_inode) {
 462                 dput(root);
 463                 up_write(&s->s_umount);
 464                 deactivate_super(s);
 465                 error = -ENXIO;
 466                 goto error;
 467         }
 468
 469         mnt->mnt_sb = s;
 470         mnt->mnt_root = root;
 471
 472         kfree(subvol_name);
 473         return 0;
 474
 475 error_s:
 476         error = PTR_ERR(s);
 477 error_bdev:
 478         btrfs_close_devices(fs_devices);
 479 error_free_subvol_name:
 480         kfree(subvol_name);
 481 error:
 482         return error;
 483 }
 484
 485 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 486 {
 487         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 488         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 489         int bits = dentry->d_sb->s_blocksize_bits;
 490
 491         buf->f_namelen = BTRFS_NAME_LEN;
 492         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 493         buf->f_bfree = buf->f_blocks -
 494                 (btrfs_super_bytes_used(disk_super) >> bits);
 495         buf->f_bavail = buf->f_bfree;
 496         buf->f_bsize = dentry->d_sb->s_blocksize;
 497         buf->f_type = BTRFS_SUPER_MAGIC;
 498         return 0;
 499 }
 500
 501 static struct file_system_type btrfs_fs_type = {
 502         .owner          = THIS_MODULE,
 503         .name           = "btrfs",
 504         .get_sb         = btrfs_get_sb,
 505         .kill_sb        = kill_anon_super,
 506         .fs_flags       = FS_REQUIRES_DEV,
 507 };
 508
 509 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 510                                 unsigned long arg)
 511 {
 512         struct btrfs_ioctl_vol_args *vol;
 513         struct btrfs_fs_devices *fs_devices;
 514         int ret = 0;
 515         int len;
 516
 517         vol = kmalloc(sizeof(*vol), GFP_KERNEL);
 518         if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
 519                 ret = -EFAULT;
 520                 goto out;
 521         }
 522         len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
 523         switch (cmd) {
 524         case BTRFS_IOC_SCAN_DEV:
 525                 ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
 526                                             &btrfs_fs_type, &fs_devices);
 527                 break;
 528         }
 529 out:
 530         kfree(vol);
 531         return ret;
 532 }
 533
 534 static void btrfs_write_super_lockfs(struct super_block *sb)
 535 {
 536         struct btrfs_root *root = btrfs_sb(sb);
 537         mutex_lock(&root->fs_info->transaction_kthread_mutex);
 538         mutex_lock(&root->fs_info->cleaner_mutex);
 539 }
 540
 541 static void btrfs_unlockfs(struct super_block *sb)
 542 {
 543         struct btrfs_root *root = btrfs_sb(sb);
 544         mutex_unlock(&root->fs_info->cleaner_mutex);
 545         mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 546 }
 547
 548 static struct super_operations btrfs_super_ops = {
 549         .delete_inode   = btrfs_delete_inode,
 550         .put_super      = btrfs_put_super,
 551         .write_super    = btrfs_write_super,
 552         .sync_fs        = btrfs_sync_fs,
 553 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
 554         .read_inode     = btrfs_read_locked_inode,
 555 #else
 556         .show_options   = generic_show_options,
 557 #endif
 558         .write_inode    = btrfs_write_inode,
 559         .dirty_inode    = btrfs_dirty_inode,
 560         .alloc_inode    = btrfs_alloc_inode,
 561         .destroy_inode  = btrfs_destroy_inode,
 562         .statfs         = btrfs_statfs,
 563         .write_super_lockfs = btrfs_write_super_lockfs,
 564         .unlockfs       = btrfs_unlockfs,
 565 };
 566
 567 static const struct file_operations btrfs_ctl_fops = {
 568         .unlocked_ioctl  = btrfs_control_ioctl,
 569         .compat_ioctl = btrfs_control_ioctl,
 570         .owner   = THIS_MODULE,
 571 };
 572
 573 static struct miscdevice btrfs_misc = {
 574         .minor          = MISC_DYNAMIC_MINOR,
 575         .name           = "btrfs-control",
 576         .fops           = &btrfs_ctl_fops
 577 };
 578
 579 static int btrfs_interface_init(void)
 580 {
 581         return misc_register(&btrfs_misc);
 582 }
 583
 584 void btrfs_interface_exit(void)
 585 {
 586         if (misc_deregister(&btrfs_misc) < 0)
 587                 printk("misc_deregister failed for control device");
 588 }
 589
 590 static int __init init_btrfs_fs(void)
 591 {
 592         int err;
 593
 594         err = btrfs_init_sysfs();
 595         if (err)
 596                 return err;
 597
 598         err = btrfs_init_cachep();
 599         if (err)
 600                 goto free_sysfs;
 601
 602         err = extent_io_init();
 603         if (err)
 604                 goto free_cachep;
 605
 606         err = extent_map_init();
 607         if (err)
 608                 goto free_extent_io;
 609
 610         err = btrfs_interface_init();
 611         if (err)
 612                 goto free_extent_map;
 613         err = register_filesystem(&btrfs_fs_type);
 614         if (err)
 615                 goto unregister_ioctl;
 616
 617         printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 618         return 0;
 619
 620 unregister_ioctl:
 621         btrfs_interface_exit();
 622 free_extent_map:
 623         extent_map_exit();
 624 free_extent_io:
 625         extent_io_exit();
 626 free_cachep:
 627         btrfs_destroy_cachep();
 628 free_sysfs:
 629         btrfs_exit_sysfs();
 630         return err;
 631 }
 632
 633 static void __exit exit_btrfs_fs(void)
 634 {
 635         btrfs_destroy_cachep();
 636         extent_map_exit();
 637         extent_io_exit();
 638         btrfs_interface_exit();
 639         unregister_filesystem(&btrfs_fs_type);
 640         btrfs_exit_sysfs();
 641         btrfs_cleanup_fs_uuids();
 642 }
 643
 644 module_init(init_btrfs_fs)
 645 module_exit(exit_btrfs_fs)
 646
 647 MODULE_LICENSE("GPL");