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_thread_pool, 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_thread_pool, "thread_pool=%d"},
  84         {Opt_ssd, "ssd"},
  85         {Opt_err, NULL}
  86 };
  87
  88 u64 btrfs_parse_size(char *str)
  89 {
  90         u64 res;
  91         int mult = 1;
  92         char *end;
  93         char last;
  94
  95         res = simple_strtoul(str, &end, 10);
  96
  97         last = end[0];
  98         if (isalpha(last)) {
  99                 last = tolower(last);
 100                 switch (last) {
 101                 case 'g':
 102                         mult *= 1024;
 103                 case 'm':
 104                         mult *= 1024;
 105                 case 'k':
 106                         mult *= 1024;
 107                 }
 108                 res = res * mult;
 109         }
 110         return res;
 111 }
 112
 113 /*
 114  * Regular mount options parser.  Everything that is needed only when
 115  * reading in a new superblock is parsed here.
 116  */
 117 int btrfs_parse_options(struct btrfs_root *root, char *options)
 118 {
 119         struct btrfs_fs_info *info = root->fs_info;
 120         substring_t args[MAX_OPT_ARGS];
 121         char *p, *num;
 122         int intarg;
 123
 124         if (!options)
 125                 return 0;
 126
 127         /*
 128          * strsep changes the string, duplicate it because parse_options
 129          * gets called twice
 130          */
 131         options = kstrdup(options, GFP_NOFS);
 132         if (!options)
 133                 return -ENOMEM;
 134
 135
 136         while ((p = strsep(&options, ",")) != NULL) {
 137                 int token;
 138                 if (!*p)
 139                         continue;
 140
 141                 token = match_token(p, tokens, args);
 142                 switch (token) {
 143                 case Opt_degraded:
 144                         printk(KERN_INFO "btrfs: allowing degraded mounts\n");
 145                         btrfs_set_opt(info->mount_opt, DEGRADED);
 146                         break;
 147                 case Opt_subvol:
 148                 case Opt_device:
 149                         /*
 150                          * These are parsed by btrfs_parse_early_options
 151                          * and can be happily ignored here.
 152                          */
 153                         break;
 154                 case Opt_nodatasum:
 155                         printk(KERN_INFO "btrfs: setting nodatacsum\n");
 156                         btrfs_set_opt(info->mount_opt, NODATASUM);
 157                         break;
 158                 case Opt_nodatacow:
 159                         printk(KERN_INFO "btrfs: setting nodatacow\n");
 160                         btrfs_set_opt(info->mount_opt, NODATACOW);
 161                         btrfs_set_opt(info->mount_opt, NODATASUM);
 162                         break;
 163                 case Opt_ssd:
 164                         printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 165                         btrfs_set_opt(info->mount_opt, SSD);
 166                         break;
 167                 case Opt_nobarrier:
 168                         printk(KERN_INFO "btrfs: turning off barriers\n");
 169                         btrfs_set_opt(info->mount_opt, NOBARRIER);
 170                         break;
 171                 case Opt_thread_pool:
 172                         intarg = 0;
 173                         match_int(&args[0], &intarg);
 174                         if (intarg) {
 175                                 info->thread_pool_size = intarg;
 176                                 printk(KERN_INFO "btrfs: thread pool %d\n",
 177                                        info->thread_pool_size);
 178                         }
 179                         break;
 180                 case Opt_max_extent:
 181                         num = match_strdup(&args[0]);
 182                         if (num) {
 183                                 info->max_extent = btrfs_parse_size(num);
 184                                 kfree(num);
 185
 186                                 info->max_extent = max_t(u64,
 187                                         info->max_extent, root->sectorsize);
 188                                 printk(KERN_INFO "btrfs: max_extent at %llu\n",
 189                                        info->max_extent);
 190                         }
 191                         break;
 192                 case Opt_max_inline:
 193                         num = match_strdup(&args[0]);
 194                         if (num) {
 195                                 info->max_inline = btrfs_parse_size(num);
 196                                 kfree(num);
 197
 198                                 if (info->max_inline) {
 199                                         info->max_inline = max_t(u64,
 200                                                 info->max_inline,
 201                                                 root->sectorsize);
 202                                 }
 203                                 printk(KERN_INFO "btrfs: max_inline at %llu\n",
 204                                         info->max_inline);
 205                         }
 206                         break;
 207                 case Opt_alloc_start:
 208                         num = match_strdup(&args[0]);
 209                         if (num) {
 210                                 info->alloc_start = btrfs_parse_size(num);
 211                                 kfree(num);
 212                                 printk(KERN_INFO
 213                                         "btrfs: allocations start at %llu\n",
 214                                         info->alloc_start);
 215                         }
 216                         break;
 217                 default:
 218                         break;
 219                 }
 220         }
 221         kfree(options);
 222         return 0;
 223 }
 224
 225 /*
 226  * Parse mount options that are required early in the mount process.
 227  *
 228  * All other options will be parsed on much later in the mount process and
 229  * only when we need to allocate a new super block.
 230  */
 231 static int btrfs_parse_early_options(const char *options, int flags,
 232                 void *holder, char **subvol_name,
 233                 struct btrfs_fs_devices **fs_devices)
 234 {
 235         substring_t args[MAX_OPT_ARGS];
 236         char *opts, *p;
 237         int error = 0;
 238
 239         if (!options)
 240                 goto out;
 241
 242         /*
 243          * strsep changes the string, duplicate it because parse_options
 244          * gets called twice
 245          */
 246         opts = kstrdup(options, GFP_KERNEL);
 247         if (!opts)
 248                 return -ENOMEM;
 249
 250         while ((p = strsep(&opts, ",")) != NULL) {
 251                 int token;
 252                 if (!*p)
 253                         continue;
 254
 255                 token = match_token(p, tokens, args);
 256                 switch (token) {
 257                 case Opt_subvol:
 258                         *subvol_name = match_strdup(&args[0]);
 259                         break;
 260                 case Opt_device:
 261                         error = btrfs_scan_one_device(match_strdup(&args[0]),
 262                                         flags, holder, fs_devices);
 263                         if (error)
 264                                 goto out_free_opts;
 265                         break;
 266                 default:
 267                         break;
 268                 }
 269         }
 270
 271  out_free_opts:
 272         kfree(opts);
 273  out:
 274         /*
 275          * If no subvolume name is specified we use the default one.  Allocate
 276          * a copy of the string "default" here so that code later in the
 277          * mount path doesn't care if it's the default volume or another one.
 278          */
 279         if (!*subvol_name) {
 280                 *subvol_name = kstrdup("default", GFP_KERNEL);
 281                 if (!*subvol_name)
 282                         return -ENOMEM;
 283         }
 284         return error;
 285 }
 286
 287 static int btrfs_fill_super(struct super_block * sb,
 288                             struct btrfs_fs_devices *fs_devices,
 289                             void * data, int silent)
 290 {
 291         struct inode * inode;
 292         struct dentry * root_dentry;
 293         struct btrfs_super_block *disk_super;
 294         struct btrfs_root *tree_root;
 295         struct btrfs_inode *bi;
 296         int err;
 297
 298         sb->s_maxbytes = MAX_LFS_FILESIZE;
 299         sb->s_magic = BTRFS_SUPER_MAGIC;
 300         sb->s_op = &btrfs_super_ops;
 301         sb->s_xattr = btrfs_xattr_handlers;
 302         sb->s_time_gran = 1;
 303
 304         tree_root = open_ctree(sb, fs_devices, (char *)data);
 305
 306         if (IS_ERR(tree_root)) {
 307                 printk("btrfs: open_ctree failed\n");
 308                 return PTR_ERR(tree_root);
 309         }
 310         sb->s_fs_info = tree_root;
 311         disk_super = &tree_root->fs_info->super_copy;
 312         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 313                                   tree_root);
 314         bi = BTRFS_I(inode);
 315         bi->location.objectid = inode->i_ino;
 316         bi->location.offset = 0;
 317         bi->root = tree_root;
 318
 319         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 320
 321         if (!inode) {
 322                 err = -ENOMEM;
 323                 goto fail_close;
 324         }
 325         if (inode->i_state & I_NEW) {
 326                 btrfs_read_locked_inode(inode);
 327                 unlock_new_inode(inode);
 328         }
 329
 330         root_dentry = d_alloc_root(inode);
 331         if (!root_dentry) {
 332                 iput(inode);
 333                 err = -ENOMEM;
 334                 goto fail_close;
 335         }
 336
 337         /* this does the super kobj at the same time */
 338         err = btrfs_sysfs_add_super(tree_root->fs_info);
 339         if (err)
 340                 goto fail_close;
 341
 342         sb->s_root = root_dentry;
 343         btrfs_transaction_queue_work(tree_root, HZ * 30);
 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         btrfs_defrag_dirty_roots(root->fs_info);
 370         trans = btrfs_start_transaction(root, 1);
 371         ret = btrfs_commit_transaction(trans, root);
 372         sb->s_dirt = 0;
 373         return ret;
 374 }
 375
 376 static void btrfs_write_super(struct super_block *sb)
 377 {
 378         sb->s_dirt = 0;
 379 }
 380
 381 static int btrfs_test_super(struct super_block *s, void *data)
 382 {
 383         struct btrfs_fs_devices *test_fs_devices = data;
 384         struct btrfs_root *root = btrfs_sb(s);
 385
 386         return root->fs_info->fs_devices == test_fs_devices;
 387 }
 388
 389 /*
 390  * Find a superblock for the given device / mount point.
 391  *
 392  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
 393  *        for multiple device setup.  Make sure to keep it in sync.
 394  */
 395 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 396                 const char *dev_name, void *data, struct vfsmount *mnt)
 397 {
 398         char *subvol_name = NULL;
 399         struct block_device *bdev = NULL;
 400         struct super_block *s;
 401         struct dentry *root;
 402         struct btrfs_fs_devices *fs_devices = NULL;
 403         int error = 0;
 404
 405         error = btrfs_parse_early_options(data, flags, fs_type,
 406                                           &subvol_name, &fs_devices);
 407         if (error)
 408                 goto error;
 409
 410         error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
 411         if (error)
 412                 goto error_free_subvol_name;
 413
 414         error = btrfs_open_devices(fs_devices, flags, fs_type);
 415         if (error)
 416                 goto error_free_subvol_name;
 417
 418         bdev = fs_devices->latest_bdev;
 419         btrfs_lock_volumes();
 420         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 421         btrfs_unlock_volumes();
 422         if (IS_ERR(s))
 423                 goto error_s;
 424
 425         if (s->s_root) {
 426                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 427                         up_write(&s->s_umount);
 428                         deactivate_super(s);
 429                         error = -EBUSY;
 430                         goto error_bdev;
 431                 }
 432
 433         } else {
 434                 char b[BDEVNAME_SIZE];
 435
 436                 s->s_flags = flags;
 437                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 438                 error = btrfs_fill_super(s, fs_devices, data,
 439                                          flags & MS_SILENT ? 1 : 0);
 440                 if (error) {
 441                         up_write(&s->s_umount);
 442                         deactivate_super(s);
 443                         goto error;
 444                 }
 445
 446                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 447                 s->s_flags |= MS_ACTIVE;
 448         }
 449
 450         root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name));
 451         if (IS_ERR(root)) {
 452                 up_write(&s->s_umount);
 453                 deactivate_super(s);
 454                 error = PTR_ERR(root);
 455                 goto error;
 456         }
 457         if (!root->d_inode) {
 458                 dput(root);
 459                 up_write(&s->s_umount);
 460                 deactivate_super(s);
 461                 error = -ENXIO;
 462                 goto error;
 463         }
 464
 465         mnt->mnt_sb = s;
 466         mnt->mnt_root = root;
 467
 468         kfree(subvol_name);
 469         return 0;
 470
 471 error_s:
 472         error = PTR_ERR(s);
 473 error_bdev:
 474         btrfs_close_devices(fs_devices);
 475 error_free_subvol_name:
 476         kfree(subvol_name);
 477 error:
 478         return error;
 479 }
 480
 481 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 482 {
 483         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 484         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 485         int bits = dentry->d_sb->s_blocksize_bits;
 486
 487         buf->f_namelen = BTRFS_NAME_LEN;
 488         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 489         buf->f_bfree = buf->f_blocks -
 490                 (btrfs_super_bytes_used(disk_super) >> bits);
 491         buf->f_bavail = buf->f_bfree;
 492         buf->f_bsize = dentry->d_sb->s_blocksize;
 493         buf->f_type = BTRFS_SUPER_MAGIC;
 494         return 0;
 495 }
 496
 497 static struct file_system_type btrfs_fs_type = {
 498         .owner          = THIS_MODULE,
 499         .name           = "btrfs",
 500         .get_sb         = btrfs_get_sb,
 501         .kill_sb        = kill_anon_super,
 502         .fs_flags       = FS_REQUIRES_DEV,
 503 };
 504
 505 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 506                                 unsigned long arg)
 507 {
 508         struct btrfs_ioctl_vol_args *vol;
 509         struct btrfs_fs_devices *fs_devices;
 510         int ret = 0;
 511         int len;
 512
 513         vol = kmalloc(sizeof(*vol), GFP_KERNEL);
 514         if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
 515                 ret = -EFAULT;
 516                 goto out;
 517         }
 518         len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
 519         switch (cmd) {
 520         case BTRFS_IOC_SCAN_DEV:
 521                 ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
 522                                             &btrfs_fs_type, &fs_devices);
 523                 break;
 524         }
 525 out:
 526         kfree(vol);
 527         return ret;
 528 }
 529
 530 static void btrfs_write_super_lockfs(struct super_block *sb)
 531 {
 532         struct btrfs_root *root = btrfs_sb(sb);
 533         btrfs_transaction_flush_work(root);
 534 }
 535
 536 static void btrfs_unlockfs(struct super_block *sb)
 537 {
 538         struct btrfs_root *root = btrfs_sb(sb);
 539         btrfs_transaction_queue_work(root, HZ * 30);
 540 }
 541
 542 static struct super_operations btrfs_super_ops = {
 543         .delete_inode   = btrfs_delete_inode,
 544         .put_super      = btrfs_put_super,
 545         .write_super    = btrfs_write_super,
 546         .sync_fs        = btrfs_sync_fs,
 547 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
 548         .read_inode     = btrfs_read_locked_inode,
 549 #else
 550         .show_options   = generic_show_options,
 551 #endif
 552         .write_inode    = btrfs_write_inode,
 553         .dirty_inode    = btrfs_dirty_inode,
 554         .alloc_inode    = btrfs_alloc_inode,
 555         .destroy_inode  = btrfs_destroy_inode,
 556         .statfs         = btrfs_statfs,
 557         .write_super_lockfs = btrfs_write_super_lockfs,
 558         .unlockfs       = btrfs_unlockfs,
 559 };
 560
 561 static const struct file_operations btrfs_ctl_fops = {
 562         .unlocked_ioctl  = btrfs_control_ioctl,
 563         .compat_ioctl = btrfs_control_ioctl,
 564         .owner   = THIS_MODULE,
 565 };
 566
 567 static struct miscdevice btrfs_misc = {
 568         .minor          = MISC_DYNAMIC_MINOR,
 569         .name           = "btrfs-control",
 570         .fops           = &btrfs_ctl_fops
 571 };
 572
 573 static int btrfs_interface_init(void)
 574 {
 575         return misc_register(&btrfs_misc);
 576 }
 577
 578 void btrfs_interface_exit(void)
 579 {
 580         if (misc_deregister(&btrfs_misc) < 0)
 581                 printk("misc_deregister failed for control device");
 582 }
 583
 584 static int __init init_btrfs_fs(void)
 585 {
 586         int err;
 587
 588         err = btrfs_init_sysfs();
 589         if (err)
 590                 return err;
 591
 592         btrfs_init_transaction_sys();
 593         err = btrfs_init_cachep();
 594         if (err)
 595                 goto free_transaction_sys;
 596
 597         err = extent_io_init();
 598         if (err)
 599                 goto free_cachep;
 600
 601         err = extent_map_init();
 602         if (err)
 603                 goto free_extent_io;
 604
 605         err = btrfs_interface_init();
 606         if (err)
 607                 goto free_extent_map;
 608         err = register_filesystem(&btrfs_fs_type);
 609         if (err)
 610                 goto unregister_ioctl;
 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_transaction_sys:
 622         btrfs_exit_transaction_sys();
 623         btrfs_exit_sysfs();
 624         return err;
 625 }
 626
 627 static void __exit exit_btrfs_fs(void)
 628 {
 629         btrfs_exit_transaction_sys();
 630         btrfs_destroy_cachep();
 631         extent_map_exit();
 632         extent_io_exit();
 633         btrfs_interface_exit();
 634         unregister_filesystem(&btrfs_fs_type);
 635         btrfs_exit_sysfs();
 636         btrfs_cleanup_fs_uuids();
 637 }
 638
 639 module_init(init_btrfs_fs)
 640 module_exit(exit_btrfs_fs)
 641
 642 MODULE_LICENSE("GPL");