Btrfs: Pass down the expected generation number when reading tree blocks
[linux-2.6] / 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_subvol, Opt_nodatasum, Opt_nodatacow, Opt_max_extent,
69         Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, Opt_err,
70 };
71
72 static match_table_t tokens = {
73         {Opt_subvol, "subvol=%s"},
74         {Opt_nodatasum, "nodatasum"},
75         {Opt_nodatacow, "nodatacow"},
76         {Opt_nobarrier, "nobarrier"},
77         {Opt_max_extent, "max_extent=%s"},
78         {Opt_max_inline, "max_inline=%s"},
79         {Opt_alloc_start, "alloc_start=%s"},
80         {Opt_ssd, "ssd"},
81         {Opt_err, NULL}
82 };
83
84 u64 btrfs_parse_size(char *str)
85 {
86         u64 res;
87         int mult = 1;
88         char *end;
89         char last;
90
91         res = simple_strtoul(str, &end, 10);
92
93         last = end[0];
94         if (isalpha(last)) {
95                 last = tolower(last);
96                 switch (last) {
97                 case 'g':
98                         mult *= 1024;
99                 case 'm':
100                         mult *= 1024;
101                 case 'k':
102                         mult *= 1024;
103                 }
104                 res = res * mult;
105         }
106         return res;
107 }
108
109 static int parse_options (char * options,
110                           struct btrfs_root *root,
111                           char **subvol_name)
112 {
113         char * p;
114         struct btrfs_fs_info *info = NULL;
115         substring_t args[MAX_OPT_ARGS];
116
117         if (!options)
118                 return 1;
119
120         /*
121          * strsep changes the string, duplicate it because parse_options
122          * gets called twice
123          */
124         options = kstrdup(options, GFP_NOFS);
125         if (!options)
126                 return -ENOMEM;
127
128         if (root)
129                 info = root->fs_info;
130
131         while ((p = strsep (&options, ",")) != NULL) {
132                 int token;
133                 if (!*p)
134                         continue;
135
136                 token = match_token(p, tokens, args);
137                 switch (token) {
138                 case Opt_subvol:
139                         if (subvol_name) {
140                                 *subvol_name = match_strdup(&args[0]);
141                         }
142                         break;
143                 case Opt_nodatasum:
144                         if (info) {
145                                 printk("btrfs: setting nodatacsum\n");
146                                 btrfs_set_opt(info->mount_opt, NODATASUM);
147                         }
148                         break;
149                 case Opt_nodatacow:
150                         if (info) {
151                                 printk("btrfs: setting nodatacow\n");
152                                 btrfs_set_opt(info->mount_opt, NODATACOW);
153                                 btrfs_set_opt(info->mount_opt, NODATASUM);
154                         }
155                         break;
156                 case Opt_ssd:
157                         if (info) {
158                                 printk("btrfs: use ssd allocation scheme\n");
159                                 btrfs_set_opt(info->mount_opt, SSD);
160                         }
161                         break;
162                 case Opt_nobarrier:
163                         if (info) {
164                                 printk("btrfs: turning off barriers\n");
165                                 btrfs_set_opt(info->mount_opt, NOBARRIER);
166                         }
167                         break;
168                 case Opt_max_extent:
169                         if (info) {
170                                 char *num = match_strdup(&args[0]);
171                                 if (num) {
172                                         info->max_extent =
173                                                 btrfs_parse_size(num);
174                                         kfree(num);
175
176                                         info->max_extent = max_t(u64,
177                                                          info->max_extent,
178                                                          root->sectorsize);
179                                         printk("btrfs: max_extent at %Lu\n",
180                                                info->max_extent);
181                                 }
182                         }
183                         break;
184                 case Opt_max_inline:
185                         if (info) {
186                                 char *num = match_strdup(&args[0]);
187                                 if (num) {
188                                         info->max_inline =
189                                                 btrfs_parse_size(num);
190                                         kfree(num);
191
192                                         info->max_inline = max_t(u64,
193                                                          info->max_inline,
194                                                          root->sectorsize);
195                                         printk("btrfs: max_inline at %Lu\n",
196                                                info->max_inline);
197                                 }
198                         }
199                         break;
200                 case Opt_alloc_start:
201                         if (info) {
202                                 char *num = match_strdup(&args[0]);
203                                 if (num) {
204                                         info->alloc_start =
205                                                 btrfs_parse_size(num);
206                                         kfree(num);
207                                         printk("btrfs: allocations start at "
208                                                "%Lu\n", info->alloc_start);
209                                 }
210                         }
211                         break;
212                 default:
213                         break;
214                 }
215         }
216         kfree(options);
217         return 1;
218 }
219
220 static int btrfs_fill_super(struct super_block * sb,
221                             struct btrfs_fs_devices *fs_devices,
222                             void * data, int silent)
223 {
224         struct inode * inode;
225         struct dentry * root_dentry;
226         struct btrfs_super_block *disk_super;
227         struct btrfs_root *tree_root;
228         struct btrfs_inode *bi;
229         int err;
230
231         sb->s_maxbytes = MAX_LFS_FILESIZE;
232         sb->s_magic = BTRFS_SUPER_MAGIC;
233         sb->s_op = &btrfs_super_ops;
234         sb->s_xattr = btrfs_xattr_handlers;
235         sb->s_time_gran = 1;
236
237         tree_root = open_ctree(sb, fs_devices);
238
239         if (IS_ERR(tree_root)) {
240                 printk("btrfs: open_ctree failed\n");
241                 return PTR_ERR(tree_root);
242         }
243         sb->s_fs_info = tree_root;
244         disk_super = &tree_root->fs_info->super_copy;
245         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
246                                   tree_root);
247         bi = BTRFS_I(inode);
248         bi->location.objectid = inode->i_ino;
249         bi->location.offset = 0;
250         bi->root = tree_root;
251
252         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
253
254         if (!inode) {
255                 err = -ENOMEM;
256                 goto fail_close;
257         }
258         if (inode->i_state & I_NEW) {
259                 btrfs_read_locked_inode(inode);
260                 unlock_new_inode(inode);
261         }
262
263         root_dentry = d_alloc_root(inode);
264         if (!root_dentry) {
265                 iput(inode);
266                 err = -ENOMEM;
267                 goto fail_close;
268         }
269
270         parse_options((char *)data, tree_root, NULL);
271
272         /* this does the super kobj at the same time */
273         err = btrfs_sysfs_add_super(tree_root->fs_info);
274         if (err)
275                 goto fail_close;
276
277         sb->s_root = root_dentry;
278         btrfs_transaction_queue_work(tree_root, HZ * 30);
279
280 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
281         save_mount_options(sb, data);
282 #endif
283
284         return 0;
285
286 fail_close:
287         close_ctree(tree_root);
288         return err;
289 }
290
291 static int btrfs_sync_fs(struct super_block *sb, int wait)
292 {
293         struct btrfs_trans_handle *trans;
294         struct btrfs_root *root;
295         int ret;
296         root = btrfs_sb(sb);
297
298         sb->s_dirt = 0;
299         if (!wait) {
300                 filemap_flush(root->fs_info->btree_inode->i_mapping);
301                 return 0;
302         }
303         btrfs_clean_old_snapshots(root);
304         mutex_lock(&root->fs_info->fs_mutex);
305         btrfs_defrag_dirty_roots(root->fs_info);
306         trans = btrfs_start_transaction(root, 1);
307         ret = btrfs_commit_transaction(trans, root);
308         sb->s_dirt = 0;
309         mutex_unlock(&root->fs_info->fs_mutex);
310         return ret;
311 }
312
313 static void btrfs_write_super(struct super_block *sb)
314 {
315         sb->s_dirt = 0;
316 }
317
318 static int btrfs_test_super(struct super_block *s, void *data)
319 {
320         struct btrfs_fs_devices *test_fs_devices = data;
321         struct btrfs_root *root = btrfs_sb(s);
322
323         return root->fs_info->fs_devices == test_fs_devices;
324 }
325
326 int btrfs_get_sb_bdev(struct file_system_type *fs_type,
327         int flags, const char *dev_name, void *data,
328         struct vfsmount *mnt, const char *subvol)
329 {
330         struct block_device *bdev = NULL;
331         struct super_block *s;
332         struct dentry *root;
333         struct btrfs_fs_devices *fs_devices = NULL;
334         int error = 0;
335
336         error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
337         if (error)
338                 return error;
339
340         error = btrfs_open_devices(fs_devices, flags, fs_type);
341         if (error)
342                 return error;
343
344         bdev = fs_devices->lowest_bdev;
345         btrfs_lock_volumes();
346         s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
347         btrfs_unlock_volumes();
348         if (IS_ERR(s))
349                 goto error_s;
350
351         if (s->s_root) {
352                 if ((flags ^ s->s_flags) & MS_RDONLY) {
353                         up_write(&s->s_umount);
354                         deactivate_super(s);
355                         error = -EBUSY;
356                         goto error_bdev;
357                 }
358
359         } else {
360                 char b[BDEVNAME_SIZE];
361
362                 s->s_flags = flags;
363                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
364                 error = btrfs_fill_super(s, fs_devices, data,
365                                          flags & MS_SILENT ? 1 : 0);
366                 if (error) {
367                         up_write(&s->s_umount);
368                         deactivate_super(s);
369                         goto error;
370                 }
371
372                 btrfs_sb(s)->fs_info->bdev_holder = fs_type;
373                 s->s_flags |= MS_ACTIVE;
374         }
375
376         if (subvol) {
377                 root = lookup_one_len(subvol, s->s_root, strlen(subvol));
378                 if (IS_ERR(root)) {
379                         up_write(&s->s_umount);
380                         deactivate_super(s);
381                         error = PTR_ERR(root);
382                         goto error;
383                 }
384                 if (!root->d_inode) {
385                         dput(root);
386                         up_write(&s->s_umount);
387                         deactivate_super(s);
388                         error = -ENXIO;
389                         goto error;
390                 }
391         } else {
392                 root = dget(s->s_root);
393         }
394
395         mnt->mnt_sb = s;
396         mnt->mnt_root = root;
397         return 0;
398
399 error_s:
400         error = PTR_ERR(s);
401 error_bdev:
402         btrfs_close_devices(fs_devices);
403 error:
404         return error;
405 }
406 /* end copy & paste */
407
408 static int btrfs_get_sb(struct file_system_type *fs_type,
409         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
410 {
411         int ret;
412         char *subvol_name = NULL;
413
414         parse_options((char *)data, NULL, &subvol_name);
415         ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, mnt,
416                         subvol_name ? subvol_name : "default");
417         if (subvol_name)
418                 kfree(subvol_name);
419         return ret;
420 }
421
422 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
423 {
424         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
425         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
426         int bits = dentry->d_sb->s_blocksize_bits;
427
428         buf->f_namelen = BTRFS_NAME_LEN;
429         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
430         buf->f_bfree = buf->f_blocks -
431                 (btrfs_super_bytes_used(disk_super) >> bits);
432         buf->f_bavail = buf->f_bfree;
433         buf->f_bsize = dentry->d_sb->s_blocksize;
434         buf->f_type = BTRFS_SUPER_MAGIC;
435         return 0;
436 }
437
438 static struct file_system_type btrfs_fs_type = {
439         .owner          = THIS_MODULE,
440         .name           = "btrfs",
441         .get_sb         = btrfs_get_sb,
442         .kill_sb        = kill_anon_super,
443         .fs_flags       = FS_REQUIRES_DEV,
444 };
445
446 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
447                                 unsigned long arg)
448 {
449         struct btrfs_ioctl_vol_args *vol;
450         struct btrfs_fs_devices *fs_devices;
451         int ret;
452         int len;
453
454         vol = kmalloc(sizeof(*vol), GFP_KERNEL);
455         if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
456                 ret = -EFAULT;
457                 goto out;
458         }
459         len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
460         switch (cmd) {
461         case BTRFS_IOC_SCAN_DEV:
462                 ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
463                                             &btrfs_fs_type, &fs_devices);
464                 break;
465         }
466 out:
467         kfree(vol);
468         return 0;
469 }
470
471 static void btrfs_write_super_lockfs(struct super_block *sb)
472 {
473         struct btrfs_root *root = btrfs_sb(sb);
474         btrfs_transaction_flush_work(root);
475 }
476
477 static void btrfs_unlockfs(struct super_block *sb)
478 {
479         struct btrfs_root *root = btrfs_sb(sb);
480         btrfs_transaction_queue_work(root, HZ * 30);
481 }
482
483 static struct super_operations btrfs_super_ops = {
484         .delete_inode   = btrfs_delete_inode,
485         .put_inode      = btrfs_put_inode,
486         .put_super      = btrfs_put_super,
487         .write_super    = btrfs_write_super,
488         .sync_fs        = btrfs_sync_fs,
489 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
490         .read_inode     = btrfs_read_locked_inode,
491 #else
492         .show_options   = generic_show_options,
493 #endif
494         .write_inode    = btrfs_write_inode,
495         .dirty_inode    = btrfs_dirty_inode,
496         .alloc_inode    = btrfs_alloc_inode,
497         .destroy_inode  = btrfs_destroy_inode,
498         .statfs         = btrfs_statfs,
499         .write_super_lockfs = btrfs_write_super_lockfs,
500         .unlockfs       = btrfs_unlockfs,
501 };
502
503 static const struct file_operations btrfs_ctl_fops = {
504         .unlocked_ioctl  = btrfs_control_ioctl,
505         .compat_ioctl = btrfs_control_ioctl,
506         .owner   = THIS_MODULE,
507 };
508
509 static struct miscdevice btrfs_misc = {
510         .minor          = MISC_DYNAMIC_MINOR,
511         .name           = "btrfs-control",
512         .fops           = &btrfs_ctl_fops
513 };
514
515 static int btrfs_interface_init(void)
516 {
517         return misc_register(&btrfs_misc);
518 }
519
520 void btrfs_interface_exit(void)
521 {
522         if (misc_deregister(&btrfs_misc) < 0)
523                 printk("misc_deregister failed for control device");
524 }
525
526 static int __init init_btrfs_fs(void)
527 {
528         int err;
529
530         err = btrfs_init_sysfs();
531         if (err)
532                 return err;
533
534         btrfs_init_transaction_sys();
535         err = btrfs_init_cachep();
536         if (err)
537                 goto free_transaction_sys;
538
539         err = extent_io_init();
540         if (err)
541                 goto free_cachep;
542
543         err = extent_map_init();
544         if (err)
545                 goto free_extent_io;
546
547         err = btrfs_interface_init();
548         if (err)
549                 goto free_extent_map;
550         err = register_filesystem(&btrfs_fs_type);
551         if (err)
552                 goto unregister_ioctl;
553         return 0;
554
555 unregister_ioctl:
556         btrfs_interface_exit();
557 free_extent_map:
558         extent_map_exit();
559 free_extent_io:
560         extent_io_exit();
561 free_cachep:
562         btrfs_destroy_cachep();
563 free_transaction_sys:
564         btrfs_exit_transaction_sys();
565         btrfs_exit_sysfs();
566         return err;
567 }
568
569 static void __exit exit_btrfs_fs(void)
570 {
571         btrfs_exit_transaction_sys();
572         btrfs_destroy_cachep();
573         extent_map_exit();
574         extent_io_exit();
575         btrfs_interface_exit();
576         unregister_filesystem(&btrfs_fs_type);
577         btrfs_exit_sysfs();
578         btrfs_cleanup_fs_uuids();
579 }
580
581 module_init(init_btrfs_fs)
582 module_exit(exit_btrfs_fs)
583
584 MODULE_LICENSE("GPL");