SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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_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_nodatasum, "nodatasum"},
  77         {Opt_nodatacow, "nodatacow"},
  78         {Opt_nobarrier, "nobarrier"},
  79         {Opt_max_extent, "max_extent=%s"},
  80         {Opt_max_inline, "max_inline=%s"},
  81         {Opt_alloc_start, "alloc_start=%s"},
  82         {Opt_ssd, "ssd"},
  83         {Opt_err, NULL}
  84 };
  85
  86 u64 btrfs_parse_size(char *str)
  87 {
  88         u64 res;
  89         int mult = 1;
  90         char *end;
  91         char last;
  92
  93         res = simple_strtoul(str, &end, 10);
  94
  95         last = end[0];
  96         if (isalpha(last)) {
  97                 last = tolower(last);
  98                 switch (last) {
  99                 case 'g':
 100                         mult *= 1024;
 101                 case 'm':
 102                         mult *= 1024;
 103                 case 'k':
 104                         mult *= 1024;
 105                 }
 106                 res = res * mult;
 107         }
 108         return res;
 109 }
 110
 111 int btrfs_parse_options(char *options, struct btrfs_root *root,
 112                         char **subvol_name)
 113 {
 114         char * p;
 115         struct btrfs_fs_info *info = NULL;
 116         substring_t args[MAX_OPT_ARGS];
 117
 118         if (!options)
 119                 return 1;
 120
 121         /*
 122          * strsep changes the string, duplicate it because parse_options
 123          * gets called twice
 124          */
 125         options = kstrdup(options, GFP_NOFS);
 126         if (!options)
 127                 return -ENOMEM;
 128
 129         if (root)
 130                 info = root->fs_info;
 131
 132         while ((p = strsep (&options, ",")) != NULL) {
 133                 int token;
 134                 if (!*p)
 135                         continue;
 136
 137                 token = match_token(p, tokens, args);
 138                 switch (token) {
 139                 case Opt_degraded:
 140                         if (info) {
 141                                 printk("btrfs: allowing degraded mounts\n");
 142                                 btrfs_set_opt(info->mount_opt, DEGRADED);
 143                         }
 144                         break;
 145                 case Opt_subvol:
 146                         if (subvol_name) {
 147                                 *subvol_name = match_strdup(&args[0]);
 148                         }
 149                         break;
 150                 case Opt_nodatasum:
 151                         if (info) {
 152                                 printk("btrfs: setting nodatacsum\n");
 153                                 btrfs_set_opt(info->mount_opt, NODATASUM);
 154                         }
 155                         break;
 156                 case Opt_nodatacow:
 157                         if (info) {
 158                                 printk("btrfs: setting nodatacow\n");
 159                                 btrfs_set_opt(info->mount_opt, NODATACOW);
 160                                 btrfs_set_opt(info->mount_opt, NODATASUM);
 161                         }
 162                         break;
 163                 case Opt_ssd:
 164                         if (info) {
 165                                 printk("btrfs: use ssd allocation scheme\n");
 166                                 btrfs_set_opt(info->mount_opt, SSD);
 167                         }
 168                         break;
 169                 case Opt_nobarrier:
 170                         if (info) {
 171                                 printk("btrfs: turning off barriers\n");
 172                                 btrfs_set_opt(info->mount_opt, NOBARRIER);
 173                         }
 174                         break;
 175                 case Opt_max_extent:
 176                         if (info) {
 177                                 char *num = match_strdup(&args[0]);
 178                                 if (num) {
 179                                         info->max_extent =
 180                                                 btrfs_parse_size(num);
 181                                         kfree(num);
 182
 183                                         info->max_extent = max_t(u64,
 184                                                          info->max_extent,
 185                                                          root->sectorsize);
 186                                         printk("btrfs: max_extent at %Lu\n",
 187                                                info->max_extent);
 188                                 }
 189                         }
 190                         break;
 191                 case Opt_max_inline:
 192                         if (info) {
 193                                 char *num = match_strdup(&args[0]);
 194                                 if (num) {
 195                                         info->max_inline =
 196                                                 btrfs_parse_size(num);
 197                                         kfree(num);
 198
 199                                         info->max_inline = max_t(u64,
 200                                                          info->max_inline,
 201                                                          root->sectorsize);
 202                                         printk("btrfs: max_inline at %Lu\n",
 203                                                info->max_inline);
 204                                 }
 205                         }
 206                         break;
 207                 case Opt_alloc_start:
 208                         if (info) {
 209                                 char *num = match_strdup(&args[0]);
 210                                 if (num) {
 211                                         info->alloc_start =
 212                                                 btrfs_parse_size(num);
 213                                         kfree(num);
 214                                         printk("btrfs: allocations start at "
 215                                                "%Lu\n", info->alloc_start);
 216                                 }
 217                         }
 218                         break;
 219                 default:
 220                         break;
 221                 }
 222         }
 223         kfree(options);
 224         return 1;
 225 }
 226
 227 static int btrfs_fill_super(struct super_block * sb,
 228                             struct btrfs_fs_devices *fs_devices,
 229                             void * data, int silent)
 230 {
 231         struct inode * inode;
 232         struct dentry * root_dentry;
 233         struct btrfs_super_block *disk_super;
 234         struct btrfs_root *tree_root;
 235         struct btrfs_inode *bi;
 236         int err;
 237
 238         sb->s_maxbytes = MAX_LFS_FILESIZE;
 239         sb->s_magic = BTRFS_SUPER_MAGIC;
 240         sb->s_op = &btrfs_super_ops;
 241         sb->s_xattr = btrfs_xattr_handlers;
 242         sb->s_time_gran = 1;
 243
 244         tree_root = open_ctree(sb, fs_devices, (char *)data);
 245
 246         if (IS_ERR(tree_root)) {
 247                 printk("btrfs: open_ctree failed\n");
 248                 return PTR_ERR(tree_root);
 249         }
 250         sb->s_fs_info = tree_root;
 251         disk_super = &tree_root->fs_info->super_copy;
 252         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 253                                   tree_root);
 254         bi = BTRFS_I(inode);
 255         bi->location.objectid = inode->i_ino;
 256         bi->location.offset = 0;
 257         bi->root = tree_root;
 258
 259         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 260
 261         if (!inode) {
 262                 err = -ENOMEM;
 263                 goto fail_close;
 264         }
 265         if (inode->i_state & I_NEW) {
 266                 btrfs_read_locked_inode(inode);
 267                 unlock_new_inode(inode);
 268         }
 269
 270         root_dentry = d_alloc_root(inode);
 271         if (!root_dentry) {
 272                 iput(inode);
 273                 err = -ENOMEM;
 274                 goto fail_close;
 275         }
 276
 277         /* this does the super kobj at the same time */
 278         err = btrfs_sysfs_add_super(tree_root->fs_info);
 279         if (err)
 280                 goto fail_close;
 281
 282         sb->s_root = root_dentry;
 283         btrfs_transaction_queue_work(tree_root, HZ * 30);
 284
 285 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
 286         save_mount_options(sb, data);
 287 #endif
 288
 289         return 0;
 290
 291 fail_close:
 292         close_ctree(tree_root);
 293         return err;
 294 }
 295
 296 static int btrfs_sync_fs(struct super_block *sb, int wait)
 297 {
 298         struct btrfs_trans_handle *trans;
 299         struct btrfs_root *root;
 300         int ret;
 301         root = btrfs_sb(sb);
 302
 303         sb->s_dirt = 0;
 304         if (!wait) {
 305                 filemap_flush(root->fs_info->btree_inode->i_mapping);
 306                 return 0;
 307         }
 308         btrfs_clean_old_snapshots(root);
 309         mutex_lock(&root->fs_info->fs_mutex);
 310         btrfs_defrag_dirty_roots(root->fs_info);
 311         trans = btrfs_start_transaction(root, 1);
 312         ret = btrfs_commit_transaction(trans, root);
 313         sb->s_dirt = 0;
 314         mutex_unlock(&root->fs_info->fs_mutex);
 315         return ret;
 316 }
 317
 318 static void btrfs_write_super(struct super_block *sb)
 319 {
 320         sb->s_dirt = 0;
 321 }
 322
 323 static int btrfs_test_super(struct super_block *s, void *data)
 324 {
 325         struct btrfs_fs_devices *test_fs_devices = data;
 326         struct btrfs_root *root = btrfs_sb(s);
 327
 328         return root->fs_info->fs_devices == test_fs_devices;
 329 }
 330
 331 int btrfs_get_sb_bdev(struct file_system_type *fs_type,
 332         int flags, const char *dev_name, void *data,
 333         struct vfsmount *mnt, const char *subvol)
 334 {
 335         struct block_device *bdev = NULL;
 336         struct super_block *s;
 337         struct dentry *root;
 338         struct btrfs_fs_devices *fs_devices = NULL;
 339         int error = 0;
 340
 341         error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
 342         if (error)
 343                 return error;
 344
 345         error = btrfs_open_devices(fs_devices, flags, fs_type);
 346         if (error)
 347                 return error;
 348
 349         bdev = fs_devices->latest_bdev;
 350         btrfs_lock_volumes();
 351         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 352         btrfs_unlock_volumes();
 353         if (IS_ERR(s))
 354                 goto error_s;
 355
 356         if (s->s_root) {
 357                 if ((flags ^ s->s_flags) & MS_RDONLY) {
 358                         up_write(&s->s_umount);
 359                         deactivate_super(s);
 360                         error = -EBUSY;
 361                         goto error_bdev;
 362                 }
 363
 364         } else {
 365                 char b[BDEVNAME_SIZE];
 366
 367                 s->s_flags = flags;
 368                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 369                 error = btrfs_fill_super(s, fs_devices, data,
 370                                          flags & MS_SILENT ? 1 : 0);
 371                 if (error) {
 372                         up_write(&s->s_umount);
 373                         deactivate_super(s);
 374                         goto error;
 375                 }
 376
 377                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 378                 s->s_flags |= MS_ACTIVE;
 379         }
 380
 381         if (subvol) {
 382                 root = lookup_one_len(subvol, s->s_root, strlen(subvol));
 383                 if (IS_ERR(root)) {
 384                         up_write(&s->s_umount);
 385                         deactivate_super(s);
 386                         error = PTR_ERR(root);
 387                         goto error;
 388                 }
 389                 if (!root->d_inode) {
 390                         dput(root);
 391                         up_write(&s->s_umount);
 392                         deactivate_super(s);
 393                         error = -ENXIO;
 394                         goto error;
 395                 }
 396         } else {
 397                 root = dget(s->s_root);
 398         }
 399
 400         mnt->mnt_sb = s;
 401         mnt->mnt_root = root;
 402         return 0;
 403
 404 error_s:
 405         error = PTR_ERR(s);
 406 error_bdev:
 407         btrfs_close_devices(fs_devices);
 408 error:
 409         return error;
 410 }
 411 /* end copy & paste */
 412
 413 static int btrfs_get_sb(struct file_system_type *fs_type,
 414         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
 415 {
 416         int ret;
 417         char *subvol_name = NULL;
 418
 419         btrfs_parse_options((char *)data, NULL, &subvol_name);
 420         ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, mnt,
 421                         subvol_name ? subvol_name : "default");
 422         if (subvol_name)
 423                 kfree(subvol_name);
 424         return ret;
 425 }
 426
 427 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 428 {
 429         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 430         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
 431         int bits = dentry->d_sb->s_blocksize_bits;
 432
 433         buf->f_namelen = BTRFS_NAME_LEN;
 434         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 435         buf->f_bfree = buf->f_blocks -
 436                 (btrfs_super_bytes_used(disk_super) >> bits);
 437         buf->f_bavail = buf->f_bfree;
 438         buf->f_bsize = dentry->d_sb->s_blocksize;
 439         buf->f_type = BTRFS_SUPER_MAGIC;
 440         return 0;
 441 }
 442
 443 static struct file_system_type btrfs_fs_type = {
 444         .owner          = THIS_MODULE,
 445         .name           = "btrfs",
 446         .get_sb         = btrfs_get_sb,
 447         .kill_sb        = kill_anon_super,
 448         .fs_flags       = FS_REQUIRES_DEV,
 449 };
 450
 451 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 452                                 unsigned long arg)
 453 {
 454         struct btrfs_ioctl_vol_args *vol;
 455         struct btrfs_fs_devices *fs_devices;
 456         int ret = 0;
 457         int len;
 458
 459         vol = kmalloc(sizeof(*vol), GFP_KERNEL);
 460         if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
 461                 ret = -EFAULT;
 462                 goto out;
 463         }
 464         len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
 465         switch (cmd) {
 466         case BTRFS_IOC_SCAN_DEV:
 467                 ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
 468                                             &btrfs_fs_type, &fs_devices);
 469                 break;
 470         }
 471 out:
 472         kfree(vol);
 473         return ret;
 474 }
 475
 476 static void btrfs_write_super_lockfs(struct super_block *sb)
 477 {
 478         struct btrfs_root *root = btrfs_sb(sb);
 479         btrfs_transaction_flush_work(root);
 480 }
 481
 482 static void btrfs_unlockfs(struct super_block *sb)
 483 {
 484         struct btrfs_root *root = btrfs_sb(sb);
 485         btrfs_transaction_queue_work(root, HZ * 30);
 486 }
 487
 488 static struct super_operations btrfs_super_ops = {
 489         .delete_inode   = btrfs_delete_inode,
 490         .put_super      = btrfs_put_super,
 491         .write_super    = btrfs_write_super,
 492         .sync_fs        = btrfs_sync_fs,
 493 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
 494         .read_inode     = btrfs_read_locked_inode,
 495 #else
 496         .show_options   = generic_show_options,
 497 #endif
 498         .write_inode    = btrfs_write_inode,
 499         .dirty_inode    = btrfs_dirty_inode,
 500         .alloc_inode    = btrfs_alloc_inode,
 501         .destroy_inode  = btrfs_destroy_inode,
 502         .statfs         = btrfs_statfs,
 503         .write_super_lockfs = btrfs_write_super_lockfs,
 504         .unlockfs       = btrfs_unlockfs,
 505 };
 506
 507 static const struct file_operations btrfs_ctl_fops = {
 508         .unlocked_ioctl  = btrfs_control_ioctl,
 509         .compat_ioctl = btrfs_control_ioctl,
 510         .owner   = THIS_MODULE,
 511 };
 512
 513 static struct miscdevice btrfs_misc = {
 514         .minor          = MISC_DYNAMIC_MINOR,
 515         .name           = "btrfs-control",
 516         .fops           = &btrfs_ctl_fops
 517 };
 518
 519 static int btrfs_interface_init(void)
 520 {
 521         return misc_register(&btrfs_misc);
 522 }
 523
 524 void btrfs_interface_exit(void)
 525 {
 526         if (misc_deregister(&btrfs_misc) < 0)
 527                 printk("misc_deregister failed for control device");
 528 }
 529
 530 static int __init init_btrfs_fs(void)
 531 {
 532         int err;
 533
 534         err = btrfs_init_sysfs();
 535         if (err)
 536                 return err;
 537
 538         btrfs_init_transaction_sys();
 539         err = btrfs_init_cachep();
 540         if (err)
 541                 goto free_transaction_sys;
 542
 543         err = extent_io_init();
 544         if (err)
 545                 goto free_cachep;
 546
 547         err = extent_map_init();
 548         if (err)
 549                 goto free_extent_io;
 550
 551         err = btrfs_interface_init();
 552         if (err)
 553                 goto free_extent_map;
 554         err = register_filesystem(&btrfs_fs_type);
 555         if (err)
 556                 goto unregister_ioctl;
 557         return 0;
 558
 559 unregister_ioctl:
 560         btrfs_interface_exit();
 561 free_extent_map:
 562         extent_map_exit();
 563 free_extent_io:
 564         extent_io_exit();
 565 free_cachep:
 566         btrfs_destroy_cachep();
 567 free_transaction_sys:
 568         btrfs_exit_transaction_sys();
 569         btrfs_exit_sysfs();
 570         return err;
 571 }
 572
 573 static void __exit exit_btrfs_fs(void)
 574 {
 575         btrfs_exit_transaction_sys();
 576         btrfs_destroy_cachep();
 577         extent_map_exit();
 578         extent_io_exit();
 579         btrfs_interface_exit();
 580         unregister_filesystem(&btrfs_fs_type);
 581         btrfs_exit_sysfs();
 582         btrfs_cleanup_fs_uuids();
 583 }
 584
 585 module_init(init_btrfs_fs)
 586 module_exit(exit_btrfs_fs)
 587
 588 MODULE_LICENSE("GPL");