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
  49 #define BTRFS_SUPER_MAGIC 0x9123683E
  50
  51 static struct super_operations btrfs_super_ops;
  52
  53 static void btrfs_put_super (struct super_block * sb)
  54 {
  55         struct btrfs_root *root = btrfs_sb(sb);
  56         struct btrfs_fs_info *fs = root->fs_info;
  57         int ret;
  58
  59         ret = close_ctree(root);
  60         if (ret) {
  61                 printk("close ctree returns %d\n", ret);
  62         }
  63         btrfs_sysfs_del_super(fs);
  64         sb->s_fs_info = NULL;
  65 }
  66
  67 enum {
  68         Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
  69         Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
  70         Opt_ssd, Opt_err,
  71 };
  72
  73 static match_table_t tokens = {
  74         {Opt_degraded, "degraded"},
  75         {Opt_subvol, "subvol=%s"},
  76         {Opt_device, "device=%s"},
  77         {Opt_nodatasum, "nodatasum"},
  78         {Opt_nodatacow, "nodatacow"},
  79         {Opt_nobarrier, "nobarrier"},
  80         {Opt_max_extent, "max_extent=%s"},
  81         {Opt_max_inline, "max_inline=%s"},
  82         {Opt_alloc_start, "alloc_start=%s"},
  83         {Opt_ssd, "ssd"},
  84         {Opt_err, NULL}
  85 };
  86
  87 u64 btrfs_parse_size(char *str)
  88 {
  89         u64 res;
  90         int mult = 1;
  91         char *end;
  92         char last;
  93
  94         res = simple_strtoul(str, &end, 10);
  95
  96         last = end[0];
  97         if (isalpha(last)) {
  98                 last = tolower(last);
  99                 switch (last) {
 100                 case 'g':
 101                         mult *= 1024;
 102                 case 'm':
 103                         mult *= 1024;
 104                 case 'k':
 105                         mult *= 1024;
 106                 }
 107                 res = res * mult;
 108         }
 109         return res;
 110 }
 111
 112 /*
 113  * Regular mount options parser.  Everything that is needed only when
 114  * reading in a new superblock is parsed here.
 115  */
 116 int btrfs_parse_options(struct btrfs_root *root, char *options)
 117 {
 118         struct btrfs_fs_info *info = root->fs_info;
 119         substring_t args[MAX_OPT_ARGS];
 120         char *p, *num;
 121
 122         if (!options)
 123                 return 0;
 124
 125         /*
 126          * strsep changes the string, duplicate it because parse_options
 127          * gets called twice
 128          */
 129         options = kstrdup(options, GFP_NOFS);
 130         if (!options)
 131                 return -ENOMEM;
 132
 133
 134         while ((p = strsep(&options, ",")) != NULL) {
 135                 int token;
 136                 if (!*p)
 137                         continue;
 138
 139                 token = match_token(p, tokens, args);
 140                 switch (token) {
 141                 case Opt_degraded:
 142                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 143                         btrfs_set_opt(info->mount_opt, DEGRADED);
 144                         break;
 145                 case Opt_subvol:
 146                 case Opt_device:
 147                         /*
 148                          * These are parsed by btrfs_parse_early_options
 149                          * and can be happily ignored here.
 150                          */
 151                         break;
 152                 case Opt_nodatasum:
 153                         printk(KERN_INFO "btrfs: setting nodatacsum\n");
 154                         btrfs_set_opt(info->mount_opt, NODATASUM);
 155                         break;
 156                 case Opt_nodatacow:
 157                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 158                         btrfs_set_opt(info->mount_opt, NODATACOW);
 159                         btrfs_set_opt(info->mount_opt, NODATASUM);
 160                         break;
 161                 case Opt_ssd:
 162                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 163                         btrfs_set_opt(info->mount_opt, SSD);
 164                         break;
 165                 case Opt_nobarrier:
 166                         printk(KERN_INFO "btrfs: turning off barriers\n");
 167                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 168                         break;
 169                 case Opt_max_extent:
 170                         num = match_strdup(&args[0]);
 171                         if (num) {
 172                                 info->max_extent = btrfs_parse_size(num);
 173                                 kfree(num);
 174
 175                                 info->max_extent = max_t(u64,
 176                                         info->max_extent, root->sectorsize);
 177                                 printk(KERN_INFO "btrfs: max_extent at %llu\n",
 178                                        info->max_extent);
 179                         }
 180                         break;
 181                 case Opt_max_inline:
 182                         num = match_strdup(&args[0]);
 183                         if (num) {
 184                                 info->max_inline = btrfs_parse_size(num);
 185                                 kfree(num);
 186
 187                                 info->max_inline = max_t(u64,
 188                                         info->max_inline, root->sectorsize);
 189                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 190                                         info->max_inline);
 191                         }
 192                         break;
 193                 case Opt_alloc_start:
 194                         num = match_strdup(&args[0]);
 195                         if (num) {
 196                                 info->alloc_start = btrfs_parse_size(num);
 197                                 kfree(num);
 198                                 printk(KERN_INFO
 199                                         "btrfs: allocations start at %llu\n",
 200                                         info->alloc_start);
 201                         }
 202                         break;
 203                 default:
 204                         break;
 205                 }
 206         }
 207         kfree(options);
 208         return 0;
 209 }
 210
 211 /*
 212  * Parse mount options that are required early in the mount process.
 213  *
 214  * All other options will be parsed on much later in the mount process and
 215  * only when we need to allocate a new super block.
 216  */
 217 static int btrfs_parse_early_options(const char *options, int flags,
 218                 void *holder, char **subvol_name,
 219                 struct btrfs_fs_devices **fs_devices)
 220 {
 221         substring_t args[MAX_OPT_ARGS];
 222         char *opts, *p;
 223         int error = 0;
 224
 225         if (!options)
 226                 goto out;
 227
 228         /*
 229          * strsep changes the string, duplicate it because parse_options
 230          * gets called twice
 231          */
 232         opts = kstrdup(options, GFP_KERNEL);
 233         if (!opts)
 234                 return -ENOMEM;
 235
 236         while ((p = strsep(&opts, ",")) != NULL) {
 237                 int token;
 238                 if (!*p)
 239                         continue;
 240
 241                 token = match_token(p, tokens, args);
 242                 switch (token) {
 243                 case Opt_subvol:
 244                         *subvol_name = match_strdup(&args[0]);
 245                         break;
 246                 case Opt_device:
 247                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 248                                         flags, holder, fs_devices);
 249                         if (error)
 250                                 goto out_free_opts;
 251                         break;
 252                 default:
 253                         break;
 254                 }
 255         }
 256
 257  out_free_opts:
 258         kfree(opts);
 259  out:
 260         /*
 261          * If no subvolume name is specified we use the default one.  Allocate
 262          * a copy of the string "default" here so that code later in the
 263          * mount path doesn't care if it's the default volume or another one.
 264          */
 265         if (!*subvol_name) {
 266                 *subvol_name = kstrdup("default", GFP_KERNEL);
 267                 if (!*subvol_name)
 268                         return -ENOMEM;
 269         }
 270         return error;
 271 }
 272
 273 static int btrfs_fill_super(struct super_block * sb,
 274                             struct btrfs_fs_devices *fs_devices,
 275                             void * data, int silent)
 276 {
 277         struct inode * inode;
 278         struct dentry * root_dentry;
 279         struct btrfs_super_block *disk_super;
 280         struct btrfs_root *tree_root;
 281         struct btrfs_inode *bi;
 282         int err;
 283
 284         sb->s_maxbytes = MAX_LFS_FILESIZE;
 285         sb->s_magic = BTRFS_SUPER_MAGIC;
 286         sb->s_op = &btrfs_super_ops;
 287         sb->s_xattr = btrfs_xattr_handlers;
 288         sb->s_time_gran = 1;
 289
 290         tree_root = open_ctree(sb, fs_devices, (char *)data);
 291
 292         if (IS_ERR(tree_root)) {
 293                 printk("btrfs: open_ctree failed\n");
 294                 return PTR_ERR(tree_root);
 295         }
 296         sb->s_fs_info = tree_root;
 297         disk_super = &tree_root->fs_info->super_copy;
 298         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 299                                   tree_root);
 300         bi = BTRFS_I(inode);
 301         bi->location.objectid = inode->i_ino;
 302         bi->location.offset = 0;
 303         bi->root = tree_root;
 304
 305         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 306
 307         if (!inode) {
 308                 err = -ENOMEM;
 309                 goto fail_close;
 310         }
 311         if (inode->i_state & I_NEW) {
 312                 btrfs_read_locked_inode(inode);
 313                 unlock_new_inode(inode);
 314         }
 315
 316         root_dentry = d_alloc_root(inode);
 317         if (!root_dentry) {
 318                 iput(inode);
 319                 err = -ENOMEM;
 320                 goto fail_close;
 321         }
 322
 323         /* this does the super kobj at the same time */
 324         err = btrfs_sysfs_add_super(tree_root->fs_info);
 325         if (err)
 326                 goto fail_close;
 327
 328         sb->s_root = root_dentry;
 329         btrfs_transaction_queue_work(tree_root, HZ * 30);
 330
 331 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
 332         save_mount_options(sb, data);
 333 #endif
 334
 335         return 0;
 336
 337 fail_close:
 338         close_ctree(tree_root);
 339         return err;
 340 }
 341
 342 int btrfs_sync_fs(struct super_block *sb, int wait)
 343 {
 344         struct btrfs_trans_handle *trans;
 345         struct btrfs_root *root;
 346         int ret;
 347         root = btrfs_sb(sb);
 348
 349         sb->s_dirt = 0;
 350         if (!wait) {
 351                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 352                 return 0;
 353         }
 354         btrfs_clean_old_snapshots(root);
 355         mutex_lock(&root->fs_info->fs_mutex);
 356         btrfs_defrag_dirty_roots(root->fs_info);
 357         trans = btrfs_start_transaction(root, 1);
 358         ret = btrfs_commit_transaction(trans, root);
 359         sb->s_dirt = 0;
 360         mutex_unlock(&root->fs_info->fs_mutex);
 361         return ret;
 362 }
 363
 364 static void btrfs_write_super(struct super_block *sb)
 365 {
 366         sb->s_dirt = 0;
 367 }
 368
 369 static int btrfs_test_super(struct super_block *s, void *data)
 370 {
 371         struct btrfs_fs_devices *test_fs_devices = data;
 372         struct btrfs_root *root = btrfs_sb(s);
 373
 374         return root->fs_info->fs_devices == test_fs_devices;
 375 }
 376
 377 /*
 378  * Find a superblock for the given device / mount point.
 379  *
 380  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 381  *        for multiple device setup.  Make sure to keep it in sync.
 382  */
 383 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 384                 const char *dev_name, void *data, struct vfsmount *mnt)
 385 {
 386         char *subvol_name = NULL;
 387         struct block_device *bdev = NULL;
 388         struct super_block *s;
 389         struct dentry *root;
 390         struct btrfs_fs_devices *fs_devices = NULL;
 391         int error = 0;
 392
 393         error = btrfs_parse_early_options(data, flags, fs_type,
 394                                           &subvol_name, &fs_devices);
 395         if (error)
 396                 goto error;
 397
 398         error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
 399         if (error)
 400                 goto error_free_subvol_name;
 401
 402         error = btrfs_open_devices(fs_devices, flags, fs_type);
 403         if (error)
 404                 goto error_free_subvol_name;
 405
 406         bdev = fs_devices->latest_bdev;
 407         btrfs_lock_volumes();
 408         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 409         btrfs_unlock_volumes();
 410         if (IS_ERR(s))
 411                 goto error_s;
 412
 413         if (s->s_root) {
 414                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 415                         up_write(&s->s_umount);
 416                         deactivate_super(s);
 417                         error = -EBUSY;
 418                         goto error_bdev;
 419                 }
 420
 421         } else {
 422                 char b[BDEVNAME_SIZE];
 423
 424                 s->s_flags = flags;
 425                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 426                 error = btrfs_fill_super(s, fs_devices, data,
 427                                          flags & MS_SILENT ? 1 : 0);
 428                 if (error) {
 429                         up_write(&s->s_umount);
 430                         deactivate_super(s);
 431                         goto error;
 432                 }
 433
 434                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 435                 s->s_flags |= MS_ACTIVE;
 436         }
 437
 438         root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name));
 439         if (IS_ERR(root)) {
 440                 up_write(&s->s_umount);
 441                 deactivate_super(s);
 442                 error = PTR_ERR(root);
 443                 goto error;
 444         }
 445         if (!root->d_inode) {
 446                 dput(root);
 447                 up_write(&s->s_umount);
 448                 deactivate_super(s);
 449                 error = -ENXIO;
 450                 goto error;
 451         }
 452
 453         mnt->mnt_sb = s;
 454         mnt->mnt_root = root;
 455
 456         kfree(subvol_name);
 457         return 0;
 458
 459 error_s:
 460         error = PTR_ERR(s);
 461 error_bdev:
 462         btrfs_close_devices(fs_devices);
 463 error_free_subvol_name:
 464         kfree(subvol_name);
 465 error:
 466         return error;
 467 }
 468
 469 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 470 {
 471         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 472         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 473         int bits = dentry->d_sb->s_blocksize_bits;
 474
 475         buf->f_namelen = BTRFS_NAME_LEN;
 476         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 477         buf->f_bfree = buf->f_blocks -
 478                 (btrfs_super_bytes_used(disk_super) >> bits);
 479         buf->f_bavail = buf->f_bfree;
 480         buf->f_bsize = dentry->d_sb->s_blocksize;
 481         buf->f_type = BTRFS_SUPER_MAGIC;
 482         return 0;
 483 }
 484
 485 static struct file_system_type btrfs_fs_type = {
 486         .owner          = THIS_MODULE,
 487         .name           = "btrfs",
 488         .get_sb         = btrfs_get_sb,
 489         .kill_sb        = kill_anon_super,
 490         .fs_flags       = FS_REQUIRES_DEV,
 491 };
 492
 493 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 494                                 unsigned long arg)
 495 {
 496         struct btrfs_ioctl_vol_args *vol;
 497         struct btrfs_fs_devices *fs_devices;
 498         int ret = 0;
 499         int len;
 500
 501         vol = kmalloc(sizeof(*vol), GFP_KERNEL);
 502         if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
 503                 ret = -EFAULT;
 504                 goto out;
 505         }
 506         len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
 507         switch (cmd) {
 508         case BTRFS_IOC_SCAN_DEV:
 509                 ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
 510                                             &btrfs_fs_type, &fs_devices);
 511                 break;
 512         }
 513 out:
 514         kfree(vol);
 515         return ret;
 516 }
 517
 518 static void btrfs_write_super_lockfs(struct super_block *sb)
 519 {
 520         struct btrfs_root *root = btrfs_sb(sb);
 521         btrfs_transaction_flush_work(root);
 522 }
 523
 524 static void btrfs_unlockfs(struct super_block *sb)
 525 {
 526         struct btrfs_root *root = btrfs_sb(sb);
 527         btrfs_transaction_queue_work(root, HZ * 30);
 528 }
 529
 530 static struct super_operations btrfs_super_ops = {
 531         .delete_inode   = btrfs_delete_inode,
 532         .put_super      = btrfs_put_super,
 533         .write_super    = btrfs_write_super,
 534         .sync_fs        = btrfs_sync_fs,
 535 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
 536         .read_inode     = btrfs_read_locked_inode,
 537 #else
 538         .show_options   = generic_show_options,
 539 #endif
 540         .write_inode    = btrfs_write_inode,
 541         .dirty_inode    = btrfs_dirty_inode,
 542         .alloc_inode    = btrfs_alloc_inode,
 543         .destroy_inode  = btrfs_destroy_inode,
 544         .statfs         = btrfs_statfs,
 545         .write_super_lockfs = btrfs_write_super_lockfs,
 546         .unlockfs       = btrfs_unlockfs,
 547 };
 548
 549 static const struct file_operations btrfs_ctl_fops = {
 550         .unlocked_ioctl  = btrfs_control_ioctl,
 551         .compat_ioctl = btrfs_control_ioctl,
 552         .owner   = THIS_MODULE,
 553 };
 554
 555 static struct miscdevice btrfs_misc = {
 556         .minor          = MISC_DYNAMIC_MINOR,
 557         .name           = "btrfs-control",
 558         .fops           = &btrfs_ctl_fops
 559 };
 560
 561 static int btrfs_interface_init(void)
 562 {
 563         return misc_register(&btrfs_misc);
 564 }
 565
 566 void btrfs_interface_exit(void)
 567 {
 568         if (misc_deregister(&btrfs_misc) < 0)
 569                 printk("misc_deregister failed for control device");
 570 }
 571
 572 static int __init init_btrfs_fs(void)
 573 {
 574         int err;
 575
 576         err = btrfs_init_sysfs();
 577         if (err)
 578                 return err;
 579
 580         btrfs_init_transaction_sys();
 581         err = btrfs_init_cachep();
 582         if (err)
 583                 goto free_transaction_sys;
 584
 585         err = extent_io_init();
 586         if (err)
 587                 goto free_cachep;
 588
 589         err = extent_map_init();
 590         if (err)
 591                 goto free_extent_io;
 592
 593         err = btrfs_interface_init();
 594         if (err)
 595                 goto free_extent_map;
 596         err = register_filesystem(&btrfs_fs_type);
 597         if (err)
 598                 goto unregister_ioctl;
 599         return 0;
 600
 601 unregister_ioctl:
 602         btrfs_interface_exit();
 603 free_extent_map:
 604         extent_map_exit();
 605 free_extent_io:
 606         extent_io_exit();
 607 free_cachep:
 608         btrfs_destroy_cachep();
 609 free_transaction_sys:
 610         btrfs_exit_transaction_sys();
 611         btrfs_exit_sysfs();
 612         return err;
 613 }
 614
 615 static void __exit exit_btrfs_fs(void)
 616 {
 617         btrfs_exit_transaction_sys();
 618         btrfs_destroy_cachep();
 619         extent_map_exit();
 620         extent_io_exit();
 621         btrfs_interface_exit();
 622         unregister_filesystem(&btrfs_fs_type);
 623         btrfs_exit_sysfs();
 624         btrfs_cleanup_fs_uuids();
 625 }
 626
 627 module_init(init_btrfs_fs)
 628 module_exit(exit_btrfs_fs)
 629
 630 MODULE_LICENSE("GPL");