Btrfs: Misc 2.6.25 updates
[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 "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_max_inline, 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_max_inline, "max_inline=%s"},
77         {Opt_alloc_start, "alloc_start=%s"},
78         {Opt_ssd, "ssd"},
79         {Opt_err, NULL}
80 };
81
82 u64 btrfs_parse_size(char *str)
83 {
84         u64 res;
85         int mult = 1;
86         char *end;
87         char last;
88
89         res = simple_strtoul(str, &end, 10);
90
91         last = end[0];
92         if (isalpha(last)) {
93                 last = tolower(last);
94                 switch (last) {
95                 case 'g':
96                         mult *= 1024;
97                 case 'm':
98                         mult *= 1024;
99                 case 'k':
100                         mult *= 1024;
101                 }
102                 res = res * mult;
103         }
104         return res;
105 }
106
107 static int parse_options (char * options,
108                           struct btrfs_root *root,
109                           char **subvol_name)
110 {
111         char * p;
112         struct btrfs_fs_info *info = NULL;
113         substring_t args[MAX_OPT_ARGS];
114
115         if (!options)
116                 return 1;
117
118         /*
119          * strsep changes the string, duplicate it because parse_options
120          * gets called twice
121          */
122         options = kstrdup(options, GFP_NOFS);
123         if (!options)
124                 return -ENOMEM;
125
126         if (root)
127                 info = root->fs_info;
128
129         while ((p = strsep (&options, ",")) != NULL) {
130                 int token;
131                 if (!*p)
132                         continue;
133
134                 token = match_token(p, tokens, args);
135                 switch (token) {
136                 case Opt_subvol:
137                         if (subvol_name) {
138                                 *subvol_name = match_strdup(&args[0]);
139                         }
140                         break;
141                 case Opt_nodatasum:
142                         if (info) {
143                                 printk("btrfs: setting nodatacsum\n");
144                                 btrfs_set_opt(info->mount_opt, NODATASUM);
145                         }
146                         break;
147                 case Opt_nodatacow:
148                         if (info) {
149                                 printk("btrfs: setting nodatacow\n");
150                                 btrfs_set_opt(info->mount_opt, NODATACOW);
151                                 btrfs_set_opt(info->mount_opt, NODATASUM);
152                         }
153                         break;
154                 case Opt_ssd:
155                         if (info) {
156                                 printk("btrfs: use ssd allocation scheme\n");
157                                 btrfs_set_opt(info->mount_opt, SSD);
158                         }
159                         break;
160                 case Opt_nobarrier:
161                         if (info) {
162                                 printk("btrfs: turning off barriers\n");
163                                 btrfs_set_opt(info->mount_opt, NOBARRIER);
164                         }
165                         break;
166                 case Opt_max_extent:
167                         if (info) {
168                                 char *num = match_strdup(&args[0]);
169                                 if (num) {
170                                         info->max_extent =
171                                                 btrfs_parse_size(num);
172                                         kfree(num);
173
174                                         info->max_extent = max_t(u64,
175                                                          info->max_extent,
176                                                          root->sectorsize);
177                                         printk("btrfs: max_extent at %Lu\n",
178                                                info->max_extent);
179                                 }
180                         }
181                         break;
182                 case Opt_max_inline:
183                         if (info) {
184                                 char *num = match_strdup(&args[0]);
185                                 if (num) {
186                                         info->max_inline =
187                                                 btrfs_parse_size(num);
188                                         kfree(num);
189
190                                         info->max_inline = max_t(u64,
191                                                          info->max_inline,
192                                                          root->sectorsize);
193                                         printk("btrfs: max_inline at %Lu\n",
194                                                info->max_inline);
195                                 }
196                         }
197                         break;
198                 case Opt_alloc_start:
199                         if (info) {
200                                 char *num = match_strdup(&args[0]);
201                                 if (num) {
202                                         info->alloc_start =
203                                                 btrfs_parse_size(num);
204                                         kfree(num);
205                                         printk("btrfs: allocations start at "
206                                                "%Lu\n", info->alloc_start);
207                                 }
208                         }
209                         break;
210                 default:
211                         break;
212                 }
213         }
214         kfree(options);
215         return 1;
216 }
217
218 static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
219 {
220         struct inode * inode;
221         struct dentry * root_dentry;
222         struct btrfs_super_block *disk_super;
223         struct btrfs_root *tree_root;
224         struct btrfs_inode *bi;
225         int err;
226
227         sb->s_maxbytes = MAX_LFS_FILESIZE;
228         sb->s_magic = BTRFS_SUPER_MAGIC;
229         sb->s_op = &btrfs_super_ops;
230         sb->s_xattr = btrfs_xattr_handlers;
231         sb->s_time_gran = 1;
232
233         tree_root = open_ctree(sb);
234
235         if (!tree_root || IS_ERR(tree_root)) {
236                 printk("btrfs: open_ctree failed\n");
237                 return -EIO;
238         }
239         sb->s_fs_info = tree_root;
240         disk_super = &tree_root->fs_info->super_copy;
241         inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
242                                   tree_root);
243         bi = BTRFS_I(inode);
244         bi->location.objectid = inode->i_ino;
245         bi->location.offset = 0;
246         bi->root = tree_root;
247
248         btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
249
250         if (!inode) {
251                 err = -ENOMEM;
252                 goto fail_close;
253         }
254         if (inode->i_state & I_NEW) {
255                 btrfs_read_locked_inode(inode);
256                 unlock_new_inode(inode);
257         }
258
259         root_dentry = d_alloc_root(inode);
260         if (!root_dentry) {
261                 iput(inode);
262                 err = -ENOMEM;
263                 goto fail_close;
264         }
265
266         parse_options((char *)data, tree_root, NULL);
267
268         /* this does the super kobj at the same time */
269         err = btrfs_sysfs_add_super(tree_root->fs_info);
270         if (err)
271                 goto fail_close;
272
273         sb->s_root = root_dentry;
274         btrfs_transaction_queue_work(tree_root, HZ * 30);
275
276 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
277         save_mount_options(sb, data);
278 #endif
279
280         return 0;
281
282 fail_close:
283         close_ctree(tree_root);
284         return err;
285 }
286
287 static int btrfs_sync_fs(struct super_block *sb, int wait)
288 {
289         struct btrfs_trans_handle *trans;
290         struct btrfs_root *root;
291         int ret;
292         root = btrfs_sb(sb);
293
294         sb->s_dirt = 0;
295         if (!wait) {
296                 filemap_flush(root->fs_info->btree_inode->i_mapping);
297                 return 0;
298         }
299         btrfs_clean_old_snapshots(root);
300         mutex_lock(&root->fs_info->fs_mutex);
301         btrfs_defrag_dirty_roots(root->fs_info);
302         trans = btrfs_start_transaction(root, 1);
303         ret = btrfs_commit_transaction(trans, root);
304         sb->s_dirt = 0;
305         mutex_unlock(&root->fs_info->fs_mutex);
306         return ret;
307 }
308
309 static void btrfs_write_super(struct super_block *sb)
310 {
311         sb->s_dirt = 0;
312 }
313
314 /*
315  * This is almost a copy of get_sb_bdev in fs/super.c.
316  * We need the local copy to allow direct mounting of
317  * subvolumes, but this could be easily integrated back
318  * into the generic version.  --hch
319  */
320
321 /* start copy & paste */
322 static int set_bdev_super(struct super_block *s, void *data)
323 {
324         s->s_bdev = data;
325         s->s_dev = s->s_bdev->bd_dev;
326         return 0;
327 }
328
329 static int test_bdev_super(struct super_block *s, void *data)
330 {
331         return (void *)s->s_bdev == data;
332 }
333
334 int btrfs_get_sb_bdev(struct file_system_type *fs_type,
335         int flags, const char *dev_name, void *data,
336         int (*fill_super)(struct super_block *, void *, int),
337         struct vfsmount *mnt, const char *subvol)
338 {
339         struct block_device *bdev = NULL;
340         struct super_block *s;
341         struct dentry *root;
342         int error = 0;
343
344         bdev = open_bdev_excl(dev_name, flags, fs_type);
345         if (IS_ERR(bdev))
346                 return PTR_ERR(bdev);
347
348         /*
349          * once the super is inserted into the list by sget, s_umount
350          * will protect the lockfs code from trying to start a snapshot
351          * while we are mounting
352          */
353         down(&bdev->bd_mount_sem);
354         s = sget(fs_type, test_bdev_super, set_bdev_super, bdev);
355         up(&bdev->bd_mount_sem);
356         if (IS_ERR(s))
357                 goto error_s;
358
359         if (s->s_root) {
360                 if ((flags ^ s->s_flags) & MS_RDONLY) {
361                         up_write(&s->s_umount);
362                         deactivate_super(s);
363                         error = -EBUSY;
364                         goto error_bdev;
365                 }
366
367                 close_bdev_excl(bdev);
368         } else {
369                 char b[BDEVNAME_SIZE];
370
371                 s->s_flags = flags;
372                 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
373                 sb_set_blocksize(s, block_size(bdev));
374                 error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
375                 if (error) {
376                         up_write(&s->s_umount);
377                         deactivate_super(s);
378                         goto error;
379                 }
380
381                 s->s_flags |= MS_ACTIVE;
382         }
383
384         if (subvol) {
385                 root = lookup_one_len(subvol, s->s_root, strlen(subvol));
386                 if (IS_ERR(root)) {
387                         up_write(&s->s_umount);
388                         deactivate_super(s);
389                         error = PTR_ERR(root);
390                         goto error;
391                 }
392                 if (!root->d_inode) {
393                         dput(root);
394                         up_write(&s->s_umount);
395                         deactivate_super(s);
396                         error = -ENXIO;
397                         goto error;
398                 }
399         } else {
400                 root = dget(s->s_root);
401         }
402
403         mnt->mnt_sb = s;
404         mnt->mnt_root = root;
405         return 0;
406
407 error_s:
408         error = PTR_ERR(s);
409 error_bdev:
410         close_bdev_excl(bdev);
411 error:
412         return error;
413 }
414 /* end copy & paste */
415
416 static int btrfs_get_sb(struct file_system_type *fs_type,
417         int flags, const char *dev_name, void *data, struct vfsmount *mnt)
418 {
419         int ret;
420         char *subvol_name = NULL;
421
422         parse_options((char *)data, NULL, &subvol_name);
423         ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data,
424                         btrfs_fill_super, mnt,
425                         subvol_name ? subvol_name : "default");
426         if (subvol_name)
427                 kfree(subvol_name);
428         return ret;
429 }
430
431 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
432 {
433         struct btrfs_root *root = btrfs_sb(dentry->d_sb);
434         struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
435         int bits = dentry->d_sb->s_blocksize_bits;
436
437         buf->f_namelen = BTRFS_NAME_LEN;
438         buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
439         buf->f_bfree = buf->f_blocks -
440                 (btrfs_super_bytes_used(disk_super) >> bits);
441         buf->f_bavail = buf->f_bfree;
442         buf->f_bsize = dentry->d_sb->s_blocksize;
443         buf->f_type = BTRFS_SUPER_MAGIC;
444         return 0;
445 }
446
447 static struct file_system_type btrfs_fs_type = {
448         .owner          = THIS_MODULE,
449         .name           = "btrfs",
450         .get_sb         = btrfs_get_sb,
451         .kill_sb        = kill_block_super,
452         .fs_flags       = FS_REQUIRES_DEV,
453 };
454 static void btrfs_write_super_lockfs(struct super_block *sb)
455 {
456         struct btrfs_root *root = btrfs_sb(sb);
457         btrfs_transaction_flush_work(root);
458 }
459
460 static void btrfs_unlockfs(struct super_block *sb)
461 {
462         struct btrfs_root *root = btrfs_sb(sb);
463         btrfs_transaction_queue_work(root, HZ * 30);
464 }
465
466 static struct super_operations btrfs_super_ops = {
467         .delete_inode   = btrfs_delete_inode,
468         .put_inode      = btrfs_put_inode,
469         .put_super      = btrfs_put_super,
470         .write_super    = btrfs_write_super,
471         .sync_fs        = btrfs_sync_fs,
472 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
473         .read_inode     = btrfs_read_locked_inode,
474 #else
475         .show_options   = generic_show_options,
476 #endif
477         .write_inode    = btrfs_write_inode,
478         .dirty_inode    = btrfs_dirty_inode,
479         .alloc_inode    = btrfs_alloc_inode,
480         .destroy_inode  = btrfs_destroy_inode,
481         .statfs         = btrfs_statfs,
482         .write_super_lockfs = btrfs_write_super_lockfs,
483         .unlockfs       = btrfs_unlockfs,
484 };
485 static int __init init_btrfs_fs(void)
486 {
487         int err;
488
489         err = btrfs_init_sysfs();
490         if (err)
491                 return err;
492
493         btrfs_init_transaction_sys();
494         err = btrfs_init_cachep();
495         if (err)
496                 goto free_transaction_sys;
497
498         err = extent_io_init();
499         if (err)
500                 goto free_cachep;
501
502         err = extent_map_init();
503         if (err)
504                 goto free_extent_io;
505
506         err = register_filesystem(&btrfs_fs_type);
507         if (err)
508                 goto free_extent_map;
509         return 0;
510
511 free_extent_map:
512         extent_map_exit();
513 free_extent_io:
514         extent_io_exit();
515 free_cachep:
516         btrfs_destroy_cachep();
517 free_transaction_sys:
518         btrfs_exit_transaction_sys();
519         btrfs_exit_sysfs();
520         return err;
521 }
522
523 static void __exit exit_btrfs_fs(void)
524 {
525         btrfs_exit_transaction_sys();
526         btrfs_destroy_cachep();
527         extent_map_exit();
528         extent_io_exit();
529         unregister_filesystem(&btrfs_fs_type);
530         btrfs_exit_sysfs();
531 }
532
533 module_init(init_btrfs_fs)
534 module_exit(exit_btrfs_fs)
535
536 MODULE_LICENSE("GPL");