]> git.karo-electronics.de Git - mv-sheeva.git/blob - fs/btrfs/disk-io.c
8242933a1d8b411bc2ed7e9dec030d04ec9ac263
[mv-sheeva.git] / fs / btrfs / disk-io.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/fs.h>
20 #include <linux/blkdev.h>
21 #include <linux/crc32c.h>
22 #include <linux/scatterlist.h>
23 #include <linux/swap.h>
24 #include <linux/radix-tree.h>
25 #include <linux/writeback.h>
26 #include <linux/buffer_head.h> // for block_sync_page
27 #include "ctree.h"
28 #include "disk-io.h"
29 #include "transaction.h"
30 #include "btrfs_inode.h"
31
32 #if 0
33 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
34 {
35         if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
36                 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
37                        (unsigned long long)extent_buffer_blocknr(buf),
38                        (unsigned long long)btrfs_header_blocknr(buf));
39                 return 1;
40         }
41         return 0;
42 }
43 #endif
44
45 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
46                                             u64 blocknr)
47 {
48         struct inode *btree_inode = root->fs_info->btree_inode;
49         return find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
50                                    blocknr * root->sectorsize,
51                                    root->sectorsize, GFP_NOFS);
52 }
53
54 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
55                                                  u64 blocknr)
56 {
57         struct inode *btree_inode = root->fs_info->btree_inode;
58         return alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
59                                    blocknr * root->sectorsize,
60                                    root->sectorsize, GFP_NOFS);
61 }
62
63 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
64                                     size_t page_offset, u64 start, u64 end,
65                                     int create)
66 {
67         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
68         struct extent_map *em;
69         int ret;
70
71 again:
72         em = lookup_extent_mapping(em_tree, start, end);
73         if (em) {
74                 goto out;
75         }
76         em = alloc_extent_map(GFP_NOFS);
77         if (!em) {
78                 em = ERR_PTR(-ENOMEM);
79                 goto out;
80         }
81         em->start = 0;
82         em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1;
83         em->block_start = 0;
84         em->block_end = em->end;
85         em->bdev = inode->i_sb->s_bdev;
86         ret = add_extent_mapping(em_tree, em);
87         if (ret == -EEXIST) {
88                 free_extent_map(em);
89                 em = NULL;
90                 goto again;
91         } else if (ret) {
92                 em = ERR_PTR(ret);
93         }
94 out:
95         return em;
96 }
97
98 static int btree_writepage(struct page *page, struct writeback_control *wbc)
99 {
100         struct extent_map_tree *tree;
101         tree = &BTRFS_I(page->mapping->host)->extent_tree;
102         return extent_write_full_page(tree, page, btree_get_extent, wbc);
103 }
104 int btree_readpage(struct file *file, struct page *page)
105 {
106         struct extent_map_tree *tree;
107         tree = &BTRFS_I(page->mapping->host)->extent_tree;
108         return extent_read_full_page(tree, page, btree_get_extent);
109 }
110
111 static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
112 {
113         struct extent_map_tree *tree;
114         int ret;
115
116         BUG_ON(page->private != 1);
117         tree = &BTRFS_I(page->mapping->host)->extent_tree;
118         ret = try_release_extent_mapping(tree, page);
119         if (ret == 1) {
120                 ClearPagePrivate(page);
121                 set_page_private(page, 0);
122                 page_cache_release(page);
123         }
124         return ret;
125 }
126
127 static void btree_invalidatepage(struct page *page, unsigned long offset)
128 {
129         struct extent_map_tree *tree;
130         tree = &BTRFS_I(page->mapping->host)->extent_tree;
131         extent_invalidatepage(tree, page, offset);
132         btree_releasepage(page, GFP_NOFS);
133 }
134
135 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
136                     char *result)
137 {
138         return 0;
139 #if 0
140         u32 crc;
141         crc = crc32c(0, data, len);
142         memcpy(result, &crc, BTRFS_CRC32_SIZE);
143         return 0;
144 #endif
145 }
146
147 #if 0
148 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
149                            int verify)
150 {
151         return 0;
152         char result[BTRFS_CRC32_SIZE];
153         int ret;
154         struct btrfs_node *node;
155
156         ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
157                               bh->b_size - BTRFS_CSUM_SIZE, result);
158         if (ret)
159                 return ret;
160         if (verify) {
161                 if (memcmp(bh->b_data, result, BTRFS_CRC32_SIZE)) {
162                         printk("btrfs: %s checksum verify failed on %llu\n",
163                                root->fs_info->sb->s_id,
164                                (unsigned long long)bh_blocknr(bh));
165                         return 1;
166                 }
167         } else {
168                 node = btrfs_buffer_node(bh);
169                 memcpy(node->header.csum, result, BTRFS_CRC32_SIZE);
170         }
171         return 0;
172 }
173 #endif
174
175 #if 0
176 static int btree_writepage(struct page *page, struct writeback_control *wbc)
177 {
178         struct buffer_head *bh;
179         struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
180         struct buffer_head *head;
181         if (!page_has_buffers(page)) {
182                 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
183                                         (1 << BH_Dirty)|(1 << BH_Uptodate));
184         }
185         head = page_buffers(page);
186         bh = head;
187         do {
188                 if (buffer_dirty(bh))
189                         csum_tree_block(root, bh, 0);
190                 bh = bh->b_this_page;
191         } while (bh != head);
192         return block_write_full_page(page, btree_get_block, wbc);
193 }
194 #endif
195
196 static struct address_space_operations btree_aops = {
197         .readpage       = btree_readpage,
198         .writepage      = btree_writepage,
199         .releasepage    = btree_releasepage,
200         .invalidatepage = btree_invalidatepage,
201         .sync_page      = block_sync_page,
202 };
203
204 int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
205 {
206         struct extent_buffer *buf = NULL;
207         struct inode *btree_inode = root->fs_info->btree_inode;
208         int ret = 0;
209
210         buf = btrfs_find_create_tree_block(root, blocknr);
211         if (!buf)
212                 return 0;
213         read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
214                                  buf, 0);
215         free_extent_buffer(buf);
216         return ret;
217 }
218
219 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
220 {
221         struct extent_buffer *buf = NULL;
222         struct inode *btree_inode = root->fs_info->btree_inode;
223
224         buf = btrfs_find_create_tree_block(root, blocknr);
225         if (!buf)
226                 return NULL;
227         read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
228                                  buf, 1);
229         return buf;
230 }
231
232 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
233                      struct extent_buffer *buf)
234 {
235         struct inode *btree_inode = root->fs_info->btree_inode;
236         clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
237         return 0;
238 }
239
240 int wait_on_tree_block_writeback(struct btrfs_root *root,
241                                  struct extent_buffer *buf)
242 {
243         struct inode *btree_inode = root->fs_info->btree_inode;
244         wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree,
245                                         buf);
246         return 0;
247 }
248
249 int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
250 {
251         struct inode *btree_inode = root->fs_info->btree_inode;
252         set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
253         return 0;
254 }
255
256 static int __setup_root(int blocksize,
257                         struct btrfs_root *root,
258                         struct btrfs_fs_info *fs_info,
259                         u64 objectid)
260 {
261         root->node = NULL;
262         root->inode = NULL;
263         root->commit_root = NULL;
264         root->sectorsize = blocksize;
265         root->nodesize = blocksize;
266         root->leafsize = blocksize;
267         root->ref_cows = 0;
268         root->fs_info = fs_info;
269         root->objectid = objectid;
270         root->last_trans = 0;
271         root->highest_inode = 0;
272         root->last_inode_alloc = 0;
273         root->name = NULL;
274         memset(&root->root_key, 0, sizeof(root->root_key));
275         memset(&root->root_item, 0, sizeof(root->root_item));
276         memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
277         memset(&root->root_kobj, 0, sizeof(root->root_kobj));
278         init_completion(&root->kobj_unregister);
279         init_rwsem(&root->snap_sem);
280         root->defrag_running = 0;
281         root->defrag_level = 0;
282         root->root_key.objectid = objectid;
283         return 0;
284 }
285
286 static int find_and_setup_root(int blocksize,
287                                struct btrfs_root *tree_root,
288                                struct btrfs_fs_info *fs_info,
289                                u64 objectid,
290                                struct btrfs_root *root)
291 {
292         int ret;
293
294         __setup_root(blocksize, root, fs_info, objectid);
295         ret = btrfs_find_last_root(tree_root, objectid,
296                                    &root->root_item, &root->root_key);
297         BUG_ON(ret);
298
299         root->node = read_tree_block(root,
300                                      btrfs_root_blocknr(&root->root_item));
301         BUG_ON(!root->node);
302         return 0;
303 }
304
305 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
306                                                struct btrfs_key *location)
307 {
308         struct btrfs_root *root;
309         struct btrfs_root *tree_root = fs_info->tree_root;
310         struct btrfs_path *path;
311         struct extent_buffer *l;
312         u64 highest_inode;
313         int ret = 0;
314
315         root = kzalloc(sizeof(*root), GFP_NOFS);
316         if (!root)
317                 return ERR_PTR(-ENOMEM);
318         if (location->offset == (u64)-1) {
319                 ret = find_and_setup_root(fs_info->sb->s_blocksize,
320                                           fs_info->tree_root, fs_info,
321                                           location->objectid, root);
322                 if (ret) {
323                         kfree(root);
324                         return ERR_PTR(ret);
325                 }
326                 goto insert;
327         }
328
329         __setup_root(fs_info->sb->s_blocksize, root, fs_info,
330                      location->objectid);
331
332         path = btrfs_alloc_path();
333         BUG_ON(!path);
334         ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
335         if (ret != 0) {
336                 if (ret > 0)
337                         ret = -ENOENT;
338                 goto out;
339         }
340         l = path->nodes[0];
341         read_extent_buffer(l, &root->root_item,
342                btrfs_item_ptr_offset(l, path->slots[0]),
343                sizeof(root->root_item));
344         ret = 0;
345 out:
346         btrfs_release_path(root, path);
347         btrfs_free_path(path);
348         if (ret) {
349                 kfree(root);
350                 return ERR_PTR(ret);
351         }
352         root->node = read_tree_block(root,
353                                      btrfs_root_blocknr(&root->root_item));
354         BUG_ON(!root->node);
355 insert:
356         root->ref_cows = 1;
357         ret = btrfs_find_highest_inode(root, &highest_inode);
358         if (ret == 0) {
359                 root->highest_inode = highest_inode;
360                 root->last_inode_alloc = highest_inode;
361         }
362         return root;
363 }
364
365 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
366                                       struct btrfs_key *location,
367                                       const char *name, int namelen)
368 {
369         struct btrfs_root *root;
370         int ret;
371
372         root = radix_tree_lookup(&fs_info->fs_roots_radix,
373                                  (unsigned long)location->objectid);
374         if (root)
375                 return root;
376
377         root = btrfs_read_fs_root_no_radix(fs_info, location);
378         if (IS_ERR(root))
379                 return root;
380         ret = radix_tree_insert(&fs_info->fs_roots_radix,
381                                 (unsigned long)root->root_key.objectid,
382                                 root);
383         if (ret) {
384                 free_extent_buffer(root->node);
385                 kfree(root);
386                 return ERR_PTR(ret);
387         }
388
389         ret = btrfs_set_root_name(root, name, namelen);
390         if (ret) {
391                 free_extent_buffer(root->node);
392                 kfree(root);
393                 return ERR_PTR(ret);
394         }
395
396         ret = btrfs_sysfs_add_root(root);
397         if (ret) {
398                 free_extent_buffer(root->node);
399                 kfree(root->name);
400                 kfree(root);
401                 return ERR_PTR(ret);
402         }
403
404         ret = btrfs_find_dead_roots(fs_info->tree_root,
405                                     root->root_key.objectid, root);
406         BUG_ON(ret);
407
408         return root;
409 }
410
411 struct btrfs_root *open_ctree(struct super_block *sb)
412 {
413         struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
414                                                  GFP_NOFS);
415         struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
416                                                GFP_NOFS);
417         struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
418                                                 GFP_NOFS);
419         int ret;
420         int err = -EIO;
421         struct btrfs_super_block *disk_super;
422
423         if (!extent_root || !tree_root || !fs_info) {
424                 err = -ENOMEM;
425                 goto fail;
426         }
427         init_bit_radix(&fs_info->pinned_radix);
428         init_bit_radix(&fs_info->pending_del_radix);
429         init_bit_radix(&fs_info->extent_map_radix);
430         init_bit_radix(&fs_info->extent_ins_radix);
431         INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
432         INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
433         INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
434         INIT_LIST_HEAD(&fs_info->trans_list);
435         INIT_LIST_HEAD(&fs_info->dead_roots);
436         memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
437         init_completion(&fs_info->kobj_unregister);
438         sb_set_blocksize(sb, 4096);
439         fs_info->running_transaction = NULL;
440         fs_info->last_trans_committed = 0;
441         fs_info->tree_root = tree_root;
442         fs_info->extent_root = extent_root;
443         fs_info->sb = sb;
444         fs_info->btree_inode = new_inode(sb);
445         fs_info->btree_inode->i_ino = 1;
446         fs_info->btree_inode->i_nlink = 1;
447         fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
448         fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
449         extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
450                              fs_info->btree_inode->i_mapping,
451                              GFP_NOFS);
452         fs_info->do_barriers = 1;
453         fs_info->closing = 0;
454
455         INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
456         BTRFS_I(fs_info->btree_inode)->root = tree_root;
457         memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
458                sizeof(struct btrfs_key));
459         insert_inode_hash(fs_info->btree_inode);
460         mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
461
462         mutex_init(&fs_info->trans_mutex);
463         mutex_init(&fs_info->fs_mutex);
464
465         __setup_root(sb->s_blocksize, tree_root,
466                      fs_info, BTRFS_ROOT_TREE_OBJECTID);
467
468         fs_info->sb_buffer = read_tree_block(tree_root,
469                                              BTRFS_SUPER_INFO_OFFSET /
470                                              sb->s_blocksize);
471
472         if (!fs_info->sb_buffer)
473                 goto fail_iput;
474
475         read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
476                            sizeof(fs_info->super_copy));
477
478         read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
479                            (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
480                            BTRFS_FSID_SIZE);
481         disk_super = &fs_info->super_copy;
482         if (!btrfs_super_root(disk_super))
483                 goto fail_sb_buffer;
484
485         i_size_write(fs_info->btree_inode,
486                      btrfs_super_total_blocks(disk_super) <<
487                      fs_info->btree_inode->i_blkbits);
488
489
490         if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
491                     sizeof(disk_super->magic))) {
492                 printk("btrfs: valid FS not found on %s\n", sb->s_id);
493                 goto fail_sb_buffer;
494         }
495         tree_root->node = read_tree_block(tree_root,
496                                           btrfs_super_root(disk_super));
497         if (!tree_root->node)
498                 goto fail_sb_buffer;
499
500         mutex_lock(&fs_info->fs_mutex);
501         ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
502                                   BTRFS_EXTENT_TREE_OBJECTID, extent_root);
503         if (ret) {
504                 mutex_unlock(&fs_info->fs_mutex);
505                 goto fail_tree_root;
506         }
507
508         btrfs_read_block_groups(extent_root);
509
510         fs_info->generation = btrfs_super_generation(disk_super) + 1;
511         mutex_unlock(&fs_info->fs_mutex);
512         return tree_root;
513
514 fail_tree_root:
515         free_extent_buffer(tree_root->node);
516 fail_sb_buffer:
517         free_extent_buffer(fs_info->sb_buffer);
518 fail_iput:
519         iput(fs_info->btree_inode);
520 fail:
521         kfree(extent_root);
522         kfree(tree_root);
523         kfree(fs_info);
524         return ERR_PTR(err);
525 }
526
527 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
528                       *root)
529 {
530         int ret;
531         struct extent_buffer *super = root->fs_info->sb_buffer;
532         struct inode *btree_inode = root->fs_info->btree_inode;
533
534         set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super);
535         ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
536                                      super->start, super->len);
537         return ret;
538 }
539
540 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
541 {
542         radix_tree_delete(&fs_info->fs_roots_radix,
543                           (unsigned long)root->root_key.objectid);
544         btrfs_sysfs_del_root(root);
545         if (root->inode)
546                 iput(root->inode);
547         if (root->node)
548                 free_extent_buffer(root->node);
549         if (root->commit_root)
550                 free_extent_buffer(root->commit_root);
551         if (root->name)
552                 kfree(root->name);
553         kfree(root);
554         return 0;
555 }
556
557 static int del_fs_roots(struct btrfs_fs_info *fs_info)
558 {
559         int ret;
560         struct btrfs_root *gang[8];
561         int i;
562
563         while(1) {
564                 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
565                                              (void **)gang, 0,
566                                              ARRAY_SIZE(gang));
567                 if (!ret)
568                         break;
569                 for (i = 0; i < ret; i++)
570                         btrfs_free_fs_root(fs_info, gang[i]);
571         }
572         return 0;
573 }
574
575 int close_ctree(struct btrfs_root *root)
576 {
577         int ret;
578         struct btrfs_trans_handle *trans;
579         struct btrfs_fs_info *fs_info = root->fs_info;
580
581         fs_info->closing = 1;
582         btrfs_transaction_flush_work(root);
583         mutex_lock(&fs_info->fs_mutex);
584         btrfs_defrag_dirty_roots(root->fs_info);
585         trans = btrfs_start_transaction(root, 1);
586         ret = btrfs_commit_transaction(trans, root);
587         /* run commit again to  drop the original snapshot */
588         trans = btrfs_start_transaction(root, 1);
589         btrfs_commit_transaction(trans, root);
590         ret = btrfs_write_and_wait_transaction(NULL, root);
591         BUG_ON(ret);
592         write_ctree_super(NULL, root);
593         mutex_unlock(&fs_info->fs_mutex);
594
595         if (fs_info->extent_root->node)
596                 free_extent_buffer(fs_info->extent_root->node);
597         if (fs_info->tree_root->node)
598                 free_extent_buffer(fs_info->tree_root->node);
599         free_extent_buffer(fs_info->sb_buffer);
600         truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
601         iput(fs_info->btree_inode);
602
603         btrfs_free_block_groups(root->fs_info);
604         del_fs_roots(fs_info);
605         kfree(fs_info->extent_root);
606         kfree(fs_info->tree_root);
607         return 0;
608 }
609
610 int btrfs_buffer_uptodate(struct extent_buffer *buf)
611 {
612         struct inode *btree_inode = buf->first_page->mapping->host;
613         return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf);
614 }
615
616 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
617 {
618         struct inode *btree_inode = buf->first_page->mapping->host;
619         return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree,
620                                           buf);
621 }
622
623 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
624 {
625         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
626         u64 transid = btrfs_header_generation(buf);
627         struct inode *btree_inode = root->fs_info->btree_inode;
628
629         if (transid != root->fs_info->generation) {
630                 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
631                         (unsigned long long)extent_buffer_blocknr(buf),
632                         transid, root->fs_info->generation);
633                 WARN_ON(1);
634         }
635         set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
636 }
637
638 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
639 {
640         balance_dirty_pages_ratelimited_nr(
641                         root->fs_info->btree_inode->i_mapping, nr);
642 }