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 "ctree.h"
  40 #include "disk-io.h"
  41 #include "transaction.h"
  42 #include "btrfs_inode.h"
  43 #include "ioctl.h"
  44 #include "print-tree.h"
  45 #include "xattr.h"
  46
  47 #define BTRFS_SUPER_MAGIC 0x9123683E
  48
  49 static struct super_operations btrfs_super_ops;
  50
  51 static void btrfs_put_super (struct super_block * sb)
  52 {
  53         struct btrfs_root *root = btrfs_sb(sb);
  54         struct btrfs_fs_info *fs = root->fs_info;
  55         int ret;
  56
  57         ret = close_ctree(root);
  58         if (ret) {
  59                 printk("close ctree returns %d\n", ret);
  60         }
  61         btrfs_sysfs_del_super(fs);
  62         sb->s_fs_info = NULL;
  63 }
  64
  65 enum {
  66         Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_max_extent,
  67         Opt_alloc_start, Opt_nobarrier, Opt_ssd, Opt_err,
  68 };
  69
  70 static match_table_t tokens = {
  71         {Opt_subvol, "subvol=%s"},
  72         {Opt_nodatasum, "nodatasum"},
  73         {Opt_nodatacow, "nodatacow"},
  74         {Opt_nobarrier, "nobarrier"},
  75         {Opt_max_extent, "max_extent=%s"},
  76         {Opt_alloc_start, "alloc_start=%s"},
  77         {Opt_ssd, "ssd"},
  78         {Opt_err, NULL}
  79 };
  80
  81 u64 btrfs_parse_size(char *str)
  82 {
  83         u64 res;
  84         int mult = 1;
  85         char *end;
  86         char last;
  87
  88         res = simple_strtoul(str, &end, 10);
  89
  90         last = end[0];
  91         if (isalpha(last)) {
  92                 last = tolower(last);
  93                 switch (last) {
  94                 case 'g':
  95                         mult *= 1024;
  96                 case 'm':
  97                         mult *= 1024;
  98                 case 'k':
  99                         mult *= 1024;
 100                 }
 101                 res = res * mult;
 102         }
 103         return res;
 104 }
 105
 106 static int parse_options (char * options,
 107                           struct btrfs_root *root,
 108                           char **subvol_name)
 109 {
 110         char * p;
 111         struct btrfs_fs_info *info = NULL;
 112         substring_t args[MAX_OPT_ARGS];
 113
 114         if (!options)
 115                 return 1;
 116
 117         /*
 118          * strsep changes the string, duplicate it because parse_options
 119          * gets called twice
 120          */
 121         options = kstrdup(options, GFP_NOFS);
 122         if (!options)
 123                 return -ENOMEM;
 124
 125         if (root)
 126                 info = root->fs_info;
 127
 128         while ((p = strsep (&options, ",")) != NULL) {
 129                 int token;
 130                 if (!*p)
 131                         continue;
 132
 133                 token = match_token(p, tokens, args);
 134                 switch (token) {
 135                 case Opt_subvol:
 136                         if (subvol_name) {
 137                                 *subvol_name = match_strdup(&args[0]);
 138                         }
 139                         break;
 140                 case Opt_nodatasum:
 141                         if (info) {
 142                                 printk("btrfs: setting nodatacsum\n");
 143                                 btrfs_set_opt(info->mount_opt, NODATASUM);
 144                         }
 145                         break;
 146                 case Opt_nodatacow:
 147                         if (info) {
 148                                 printk("btrfs: setting nodatacow\n");
 149                                 btrfs_set_opt(info->mount_opt, NODATACOW);
 150                                 btrfs_set_opt(info->mount_opt, NODATASUM);
 151                         }
 152                         break;
 153                 case Opt_ssd:
 154                         if (info) {
 155                                 printk("btrfs: use ssd allocation scheme\n");
 156                                 btrfs_set_opt(info->mount_opt, SSD);
 157                         }
 158                         break;
 159                 case Opt_nobarrier:
 160                         if (info) {
 161                                 printk("btrfs: turning off barriers\n");
 162                                 btrfs_set_opt(info->mount_opt, NOBARRIER);
 163                         }
 164                         break;
 165                 case Opt_max_extent:
 166                         if (info) {
 167                                 char *num = match_strdup(&args[0]);
 168                                 if (num) {
 169                                         info->max_extent =
 170                                                 btrfs_parse_size(num);
 171                                         kfree(num);
 172
 173                                         info->max_extent = max_t(u64,
 174                                                          info->max_extent,
 175                                                          root->sectorsize);
 176                                         printk("btrfs: max_extent at %Lu\n",
 177                                                info->max_extent);
 178                                 }
 179                         }
 180                         break;
 181                 case Opt_alloc_start:
 182                         if (info) {
 183                                 char *num = match_strdup(&args[0]);
 184                                 if (num) {
 185                                         info->alloc_start =
 186                                                 btrfs_parse_size(num);
 187                                         kfree(num);
 188                                         printk("btrfs: allocations start at "
 189                                                "%Lu\n", info->alloc_start);
 190                                 }
 191                         }
 192                         break;
 193                 default:
 194                         break;
 195                 }
 196         }
 197         kfree(options);
 198         return 1;
 199 }
 200
 201 static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 202 {
 203         struct inode * inode;
 204         struct dentry * root_dentry;
 205         struct btrfs_super_block *disk_super;
 206         struct btrfs_root *tree_root;
 207         struct btrfs_inode *bi;
 208         int err;
 209
 210         sb->s_maxbytes = MAX_LFS_FILESIZE;
 211         sb->s_magic = BTRFS_SUPER_MAGIC;
 212         sb->s_op = &btrfs_super_ops;
 213         sb->s_xattr = btrfs_xattr_handlers;
 214         sb->s_time_gran = 1;
 215
 216         tree_root = open_ctree(sb);
 217
 218         if (!tree_root || IS_ERR(tree_root)) {
 219                 printk("btrfs: open_ctree failed\n");
 220                 return -EIO;
 221         }
 222         sb->s_fs_info = tree_root;
 223         disk_super = &tree_root->fs_info->super_copy;
 224         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 225                                   tree_root);
 226         bi = BTRFS_I(inode);
 227         bi->location.objectid = inode->i_ino;
 228         bi->location.offset = 0;
 229         bi->root = tree_root;
 230
 231         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 232
 233         if (!inode) {
 234                 err = -ENOMEM;
 235                 goto fail_close;
 236         }
 237         if (inode->i_state & I_NEW) {
 238                 btrfs_read_locked_inode(inode);
 239                 unlock_new_inode(inode);
 240         }
 241
 242         root_dentry = d_alloc_root(inode);
 243         if (!root_dentry) {
 244                 iput(inode);
 245                 err = -ENOMEM;
 246                 goto fail_close;
 247         }
 248
 249         parse_options((char *)data, tree_root, NULL);
 250
 251         /* this does the super kobj at the same time */
 252         err = btrfs_sysfs_add_super(tree_root->fs_info);
 253         if (err)
 254                 goto fail_close;
 255
 256         sb->s_root = root_dentry;
 257         btrfs_transaction_queue_work(tree_root, HZ * 30);
 258         return 0;
 259
 260 fail_close:
 261         close_ctree(tree_root);
 262         return err;
 263 }
 264
 265 static int btrfs_sync_fs(struct super_block *sb, int wait)
 266 {
 267         struct btrfs_trans_handle *trans;
 268         struct btrfs_root *root;
 269         int ret;
 270         root = btrfs_sb(sb);
 271
 272         sb->s_dirt = 0;
 273         if (!wait) {
 274                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 275                 return 0;
 276         }
 277         btrfs_clean_old_snapshots(root);
 278         mutex_lock(&root->fs_info->fs_mutex);
 279         btrfs_defrag_dirty_roots(root->fs_info);
 280         trans = btrfs_start_transaction(root, 1);
 281         ret = btrfs_commit_transaction(trans, root);
 282         sb->s_dirt = 0;
 283         mutex_unlock(&root->fs_info->fs_mutex);
 284         return ret;
 285 }
 286
 287 static void btrfs_write_super(struct super_block *sb)
 288 {
 289         sb->s_dirt = 0;
 290 }
 291
 292 /*
 293  * This is almost a copy of get_sb_bdev in fs/super.c.
 294  * We need the local copy to allow direct mounting of
 295  * subvolumes, but this could be easily integrated back
 296  * into the generic version.  --hch
 297  */
 298
 299 /* start copy & paste */
 300 static int set_bdev_super(struct super_block *s, void *data)
 301 {
 302         s->s_bdev = data;
 303         s->s_dev = s->s_bdev->bd_dev;
 304         return 0;
 305 }
 306
 307 static int test_bdev_super(struct super_block *s, void *data)
 308 {
 309         return (void *)s->s_bdev == data;
 310 }
 311
 312 int btrfs_get_sb_bdev(struct file_system_type *fs_type,
 313         int flags, const char *dev_name, void *data,
 314         int (*fill_super)(struct super_block *, void *, int),
 315         struct vfsmount *mnt, const char *subvol)
 316 {
 317         struct block_device *bdev = NULL;
 318         struct super_block *s;
 319         struct dentry *root;
 320         int error = 0;
 321
 322         bdev = open_bdev_excl(dev_name, flags, fs_type);
 323         if (IS_ERR(bdev))
 324                 return PTR_ERR(bdev);
 325
 326         /*
 327          * once the super is inserted into the list by sget, s_umount
 328          * will protect the lockfs code from trying to start a snapshot
 329          * while we are mounting
 330          */
 331         down(&bdev->bd_mount_sem);
 332         s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
 333         up(&bdev->bd_mount_sem);
 334         if (IS_ERR(s))
 335                 goto error_s;
 336
 337         if (s->s_root) {
 338                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 339                         up_write(&s->s_umount);
 340                         deactivate_super(s);
 341                         error = -EBUSY;
 342                         goto error_bdev;
 343                 }
 344
 345                 close_bdev_excl(bdev);
 346         } else {
 347                 char b[BDEVNAME_SIZE];
 348
 349                 s->s_flags = flags;
 350                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 351                 sb_set_blocksize(s, block_size(bdev));
 352                 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
 353                 if (error) {
 354                         up_write(&s->s_umount);
 355                         deactivate_super(s);
 356                         goto error;
 357                 }
 358
 359                 s->s_flags |= MS_ACTIVE;
 360         }
 361
 362         if (subvol) {
 363                 root = lookup_one_len(subvol, s->s_root, strlen(subvol));
 364                 if (IS_ERR(root)) {
 365                         up_write(&s->s_umount);
 366                         deactivate_super(s);
 367                         error = PTR_ERR(root);
 368                         goto error;
 369                 }
 370                 if (!root->d_inode) {
 371                         dput(root);
 372                         up_write(&s->s_umount);
 373                         deactivate_super(s);
 374                         error = -ENXIO;
 375                         goto error;
 376                 }
 377         } else {
 378                 root = dget(s->s_root);
 379         }
 380
 381         mnt->mnt_sb = s;
 382         mnt->mnt_root = root;
 383         return 0;
 384
 385 error_s:
 386         error = PTR_ERR(s);
 387 error_bdev:
 388         close_bdev_excl(bdev);
 389 error:
 390         return error;
 391 }
 392 /* end copy & paste */
 393
 394 static int btrfs_get_sb(struct file_system_type *fs_type,
 395         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
 396 {
 397         int ret;
 398         char *subvol_name = NULL;
 399
 400         parse_options((char *)data, NULL, &subvol_name);
 401         ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data,
 402                         btrfs_fill_super, mnt,
 403                         subvol_name ? subvol_name : "default");
 404         if (subvol_name)
 405                 kfree(subvol_name);
 406         return ret;
 407 }
 408
 409 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 410 {
 411         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 412         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 413         int bits = dentry->d_sb->s_blocksize_bits;
 414
 415         buf->f_namelen = BTRFS_NAME_LEN;
 416         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 417         buf->f_bfree = buf->f_blocks -
 418                 (btrfs_super_bytes_used(disk_super) >> bits);
 419         buf->f_bavail = buf->f_bfree;
 420         buf->f_bsize = dentry->d_sb->s_blocksize;
 421         buf->f_type = BTRFS_SUPER_MAGIC;
 422         return 0;
 423 }
 424
 425 static struct file_system_type btrfs_fs_type = {
 426         .owner          = THIS_MODULE,
 427         .name           = "btrfs",
 428         .get_sb         = btrfs_get_sb,
 429         .kill_sb        = kill_block_super,
 430         .fs_flags       = FS_REQUIRES_DEV,
 431 };
 432 static void btrfs_write_super_lockfs(struct super_block *sb)
 433 {
 434         struct btrfs_root *root = btrfs_sb(sb);
 435         btrfs_transaction_flush_work(root);
 436 }
 437
 438 static void btrfs_unlockfs(struct super_block *sb)
 439 {
 440         struct btrfs_root *root = btrfs_sb(sb);
 441         btrfs_transaction_queue_work(root, HZ * 30);
 442 }
 443
 444 static struct super_operations btrfs_super_ops = {
 445         .delete_inode   = btrfs_delete_inode,
 446         .put_inode      = btrfs_put_inode,
 447         .put_super      = btrfs_put_super,
 448         .read_inode     = btrfs_read_locked_inode,
 449         .write_super    = btrfs_write_super,
 450         .sync_fs        = btrfs_sync_fs,
 451         .write_inode    = btrfs_write_inode,
 452         .dirty_inode    = btrfs_dirty_inode,
 453         .alloc_inode    = btrfs_alloc_inode,
 454         .destroy_inode  = btrfs_destroy_inode,
 455         .statfs         = btrfs_statfs,
 456         .write_super_lockfs = btrfs_write_super_lockfs,
 457         .unlockfs       = btrfs_unlockfs,
 458 };
 459 static int __init init_btrfs_fs(void)
 460 {
 461         int err;
 462
 463         err = btrfs_init_sysfs();
 464         if (err)
 465                 return err;
 466
 467         btrfs_init_transaction_sys();
 468         err = btrfs_init_cachep();
 469         if (err)
 470                 goto free_transaction_sys;
 471
 472         err = extent_io_init();
 473         if (err)
 474                 goto free_cachep;
 475
 476         err = extent_map_init();
 477         if (err)
 478                 goto free_extent_io;
 479
 480         err = register_filesystem(&btrfs_fs_type);
 481         if (err)
 482                 goto free_extent_map;
 483         return 0;
 484
 485 free_extent_map:
 486         extent_map_exit();
 487 free_extent_io:
 488         extent_io_exit();
 489 free_cachep:
 490         btrfs_destroy_cachep();
 491 free_transaction_sys:
 492         btrfs_exit_transaction_sys();
 493         btrfs_exit_sysfs();
 494         return err;
 495 }
 496
 497 static void __exit exit_btrfs_fs(void)
 498 {
 499         btrfs_exit_transaction_sys();
 500         btrfs_destroy_cachep();
 501         extent_map_exit();
 502         extent_io_exit();
 503         unregister_filesystem(&btrfs_fs_type);
 504         btrfs_exit_sysfs();
 505 }
 506
 507 module_init(init_btrfs_fs)
 508 module_exit(exit_btrfs_fs)
 509
 510 MODULE_LICENSE("GPL");