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