blob: 190a1b24c902014274084427705a7ed378e4b964 [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
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
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
David Woodhouse163e7832009-04-19 13:02:41 +010029#include <linux/crc32c.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090030#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050031#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020032#include <linux/ratelimit.h>
David Sterba7e75bf32011-03-18 22:56:43 +000033#include <asm/unaligned.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050034#include "compat.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Chris Masone089f052007-03-16 16:20:31 -040037#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040038#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040039#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040040#include "print-tree.h"
Chris Mason8b712842008-06-11 16:50:36 -040041#include "async-thread.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080045#include "inode-map.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050046
Chris Masond1310b22008-01-24 16:13:08 -050047static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040048static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040049static void free_fs_root(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080050static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
51 int read_only);
52static int btrfs_destroy_ordered_operations(struct btrfs_root *root);
53static int btrfs_destroy_ordered_extents(struct btrfs_root *root);
54static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
55 struct btrfs_root *root);
56static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
57static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
58static int btrfs_destroy_marked_extents(struct btrfs_root *root,
59 struct extent_io_tree *dirty_pages,
60 int mark);
61static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
62 struct extent_io_tree *pinned_extents);
63static int btrfs_cleanup_transaction(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040064
Chris Masond352ac62008-09-29 15:18:18 -040065/*
66 * end_io_wq structs are used to do processing in task context when an IO is
67 * complete. This is used during reads to verify checksums, and it is used
68 * by writes to insert metadata for new file extents after IO is complete.
69 */
Chris Masonce9adaa2008-04-09 16:28:12 -040070struct end_io_wq {
71 struct bio *bio;
72 bio_end_io_t *end_io;
73 void *private;
74 struct btrfs_fs_info *info;
75 int error;
Chris Mason22c59942008-04-09 16:28:12 -040076 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040077 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040078 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040079};
Chris Mason0da54682007-11-07 21:08:01 -050080
Chris Masond352ac62008-09-29 15:18:18 -040081/*
82 * async submit bios are used to offload expensive checksumming
83 * onto the worker threads. They checksum file and metadata bios
84 * just before they are sent down the IO stack.
85 */
Chris Mason44b8bd72008-04-16 11:14:51 -040086struct async_submit_bio {
87 struct inode *inode;
88 struct bio *bio;
89 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -050090 extent_submit_bio_hook_t *submit_bio_start;
91 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -040092 int rw;
93 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -040094 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -040095 /*
96 * bio_offset is optional, can be used if the pages in the bio
97 * can't tell us where in the file the bio should go
98 */
99 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400100 struct btrfs_work work;
Chris Mason44b8bd72008-04-16 11:14:51 -0400101};
102
Chris Mason85d4e462011-07-26 16:11:19 -0400103/*
104 * Lockdep class keys for extent_buffer->lock's in this root. For a given
105 * eb, the lockdep key is determined by the btrfs_root it belongs to and
106 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500107 *
Chris Mason85d4e462011-07-26 16:11:19 -0400108 * Different roots are used for different purposes and may nest inside each
109 * other and they require separate keysets. As lockdep keys should be
110 * static, assign keysets according to the purpose of the root as indicated
111 * by btrfs_root->objectid. This ensures that all special purpose roots
112 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500113 *
Chris Mason85d4e462011-07-26 16:11:19 -0400114 * Lock-nesting across peer nodes is always done with the immediate parent
115 * node locked thus preventing deadlock. As lockdep doesn't know this, use
116 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500117 *
Chris Mason85d4e462011-07-26 16:11:19 -0400118 * The key is set by the readpage_end_io_hook after the buffer has passed
119 * csum validation but before the pages are unlocked. It is also set by
120 * btrfs_init_new_buffer on freshly allocated blocks.
121 *
122 * We also add a check to make sure the highest level of the tree is the
123 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
124 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500125 */
126#ifdef CONFIG_DEBUG_LOCK_ALLOC
127# if BTRFS_MAX_LEVEL != 8
128# error
129# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400130
131static struct btrfs_lockdep_keyset {
132 u64 id; /* root objectid */
133 const char *name_stem; /* lock name stem */
134 char names[BTRFS_MAX_LEVEL + 1][20];
135 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
136} btrfs_lockdep_keysets[] = {
137 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
138 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
139 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
140 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
141 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
142 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
143 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
144 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
145 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
146 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
147 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500148};
Chris Mason85d4e462011-07-26 16:11:19 -0400149
150void __init btrfs_init_lockdep(void)
151{
152 int i, j;
153
154 /* initialize lockdep class names */
155 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
156 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
157
158 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
159 snprintf(ks->names[j], sizeof(ks->names[j]),
160 "btrfs-%s-%02d", ks->name_stem, j);
161 }
162}
163
164void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
165 int level)
166{
167 struct btrfs_lockdep_keyset *ks;
168
169 BUG_ON(level >= ARRAY_SIZE(ks->keys));
170
171 /* find the matching keyset, id 0 is the default entry */
172 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
173 if (ks->id == objectid)
174 break;
175
176 lockdep_set_class_and_name(&eb->lock,
177 &ks->keys[level], ks->names[level]);
178}
179
Chris Mason4008c042009-02-12 14:09:45 -0500180#endif
181
Chris Masond352ac62008-09-29 15:18:18 -0400182/*
183 * extents on the btree inode are pretty simple, there's one extent
184 * that covers the entire device
185 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500186static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200187 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500188 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400189{
190 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
191 struct extent_map *em;
192 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400193
Chris Mason890871b2009-09-02 16:24:52 -0400194 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500195 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400196 if (em) {
197 em->bdev =
198 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400199 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400200 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400201 }
Chris Mason890871b2009-09-02 16:24:52 -0400202 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400203
David Sterba172ddd62011-04-21 00:48:27 +0200204 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400205 if (!em) {
206 em = ERR_PTR(-ENOMEM);
207 goto out;
208 }
209 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400210 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400211 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400212 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400213 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500214
Chris Mason890871b2009-09-02 16:24:52 -0400215 write_lock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400216 ret = add_extent_mapping(em_tree, em);
217 if (ret == -EEXIST) {
Chris Mason0afbaf82008-04-18 14:17:20 -0400218 u64 failed_start = em->start;
219 u64 failed_len = em->len;
220
Chris Mason5f39d392007-10-15 16:14:19 -0400221 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400222 em = lookup_extent_mapping(em_tree, start, len);
Chris Mason0afbaf82008-04-18 14:17:20 -0400223 if (em) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400224 ret = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400225 } else {
226 em = lookup_extent_mapping(em_tree, failed_start,
227 failed_len);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400228 ret = -EIO;
Chris Mason0afbaf82008-04-18 14:17:20 -0400229 }
Chris Mason5f39d392007-10-15 16:14:19 -0400230 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400231 free_extent_map(em);
232 em = NULL;
Chris Mason5f39d392007-10-15 16:14:19 -0400233 }
Chris Mason890871b2009-09-02 16:24:52 -0400234 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400235
236 if (ret)
237 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400238out:
239 return em;
240}
241
Chris Mason19c00dd2007-10-15 16:19:22 -0400242u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
243{
David Woodhouse163e7832009-04-19 13:02:41 +0100244 return crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400245}
246
247void btrfs_csum_final(u32 crc, char *result)
248{
David Sterba7e75bf32011-03-18 22:56:43 +0000249 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400250}
251
Chris Masond352ac62008-09-29 15:18:18 -0400252/*
253 * compute the csum for a btree block, and either verify it or write it
254 * into the csum field of the block.
255 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400256static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
257 int verify)
258{
David Sterba6c417612011-04-13 15:41:04 +0200259 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500260 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400261 unsigned long len;
262 unsigned long cur_len;
263 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400264 char *kaddr;
265 unsigned long map_start;
266 unsigned long map_len;
267 int err;
268 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500269 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400270
271 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500272 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400273 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400274 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500275 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400276 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400277 cur_len = min(len, map_len - (offset - map_start));
278 crc = btrfs_csum_data(root, kaddr + offset - map_start,
279 crc, cur_len);
280 len -= cur_len;
281 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400282 }
Josef Bacik607d4322008-12-02 07:17:45 -0500283 if (csum_size > sizeof(inline_result)) {
284 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
285 if (!result)
286 return 1;
287 } else {
288 result = (char *)&inline_result;
289 }
290
Chris Mason19c00dd2007-10-15 16:19:22 -0400291 btrfs_csum_final(crc, result);
292
293 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500294 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500295 u32 val;
296 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500297 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500298
Josef Bacik607d4322008-12-02 07:17:45 -0500299 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba7a36dde2011-05-06 15:33:15 +0200300 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
Chris Mason193f2842009-04-27 07:29:05 -0400301 "failed on %llu wanted %X found %X "
302 "level %d\n",
303 root->fs_info->sb->s_id,
304 (unsigned long long)buf->start, val, found,
305 btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500306 if (result != (char *)&inline_result)
307 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400308 return 1;
309 }
310 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500311 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400312 }
Josef Bacik607d4322008-12-02 07:17:45 -0500313 if (result != (char *)&inline_result)
314 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400315 return 0;
316}
317
Chris Masond352ac62008-09-29 15:18:18 -0400318/*
319 * we can't consider a given block up to date unless the transid of the
320 * block matches the transid in the parent node's pointer. This is how we
321 * detect blocks that either didn't get written at all or got written
322 * in the wrong place.
323 */
Chris Mason1259ab72008-05-12 13:39:03 -0400324static int verify_parent_transid(struct extent_io_tree *io_tree,
325 struct extent_buffer *eb, u64 parent_transid)
326{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000327 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400328 int ret;
329
330 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
331 return 0;
332
Josef Bacik2ac55d42010-02-03 19:33:23 +0000333 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
334 0, &cached_state, GFP_NOFS);
335 if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400336 btrfs_header_generation(eb) == parent_transid) {
337 ret = 0;
338 goto out;
339 }
David Sterba7a36dde2011-05-06 15:33:15 +0200340 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
Chris Mason193f2842009-04-27 07:29:05 -0400341 "found %llu\n",
342 (unsigned long long)eb->start,
343 (unsigned long long)parent_transid,
344 (unsigned long long)btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400345 ret = 1;
Josef Bacik2ac55d42010-02-03 19:33:23 +0000346 clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
Chris Mason33958dc2008-07-30 10:29:12 -0400347out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000348 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
349 &cached_state, GFP_NOFS);
Chris Mason1259ab72008-05-12 13:39:03 -0400350 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400351}
352
Chris Masond352ac62008-09-29 15:18:18 -0400353/*
354 * helper to read a given tree block, doing retries as required when
355 * the checksums don't match and we have alternate mirrors to try.
356 */
Chris Masonf1885912008-04-09 16:28:12 -0400357static int btree_read_extent_buffer_pages(struct btrfs_root *root,
358 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400359 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400360{
361 struct extent_io_tree *io_tree;
362 int ret;
363 int num_copies = 0;
364 int mirror_num = 0;
365
Josef Bacika826d6d2011-03-16 13:42:43 -0400366 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400367 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
368 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200369 ret = read_extent_buffer_pages(io_tree, eb, start,
370 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400371 btree_get_extent, mirror_num);
Chris Mason1259ab72008-05-12 13:39:03 -0400372 if (!ret &&
373 !verify_parent_transid(io_tree, eb, parent_transid))
Chris Masonf1885912008-04-09 16:28:12 -0400374 return ret;
Chris Masond3977122009-01-05 21:25:51 -0500375
Josef Bacika826d6d2011-03-16 13:42:43 -0400376 /*
377 * This buffer's crc is fine, but its contents are corrupted, so
378 * there is no reason to read the other copies, they won't be
379 * any less wrong.
380 */
381 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
382 return ret;
383
Chris Masonf1885912008-04-09 16:28:12 -0400384 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
385 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400386 if (num_copies == 1)
Chris Masonf1885912008-04-09 16:28:12 -0400387 return ret;
Chris Mason42352982008-04-28 16:40:52 -0400388
Chris Masonf1885912008-04-09 16:28:12 -0400389 mirror_num++;
Chris Mason42352982008-04-28 16:40:52 -0400390 if (mirror_num > num_copies)
Chris Masonf1885912008-04-09 16:28:12 -0400391 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400392 }
Chris Masonf1885912008-04-09 16:28:12 -0400393 return -EIO;
394}
Chris Mason19c00dd2007-10-15 16:19:22 -0400395
Chris Masond352ac62008-09-29 15:18:18 -0400396/*
Chris Masond3977122009-01-05 21:25:51 -0500397 * checksum a dirty tree block before IO. This has extra checks to make sure
398 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400399 */
Chris Masond3977122009-01-05 21:25:51 -0500400
Christoph Hellwigb2950862008-12-02 09:54:17 -0500401static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400402{
Chris Masond1310b22008-01-24 16:13:08 -0500403 struct extent_io_tree *tree;
Chris Mason35ebb932007-10-30 16:56:53 -0400404 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
Chris Mason19c00dd2007-10-15 16:19:22 -0400405 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400406 unsigned long len;
407 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400408 int ret;
409
Chris Masond1310b22008-01-24 16:13:08 -0500410 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason19c00dd2007-10-15 16:19:22 -0400411
Chris Masoneb14ab82011-02-10 12:35:00 -0500412 if (page->private == EXTENT_PAGE_PRIVATE) {
413 WARN_ON(1);
Chris Mason19c00dd2007-10-15 16:19:22 -0400414 goto out;
Chris Masoneb14ab82011-02-10 12:35:00 -0500415 }
416 if (!page->private) {
417 WARN_ON(1);
Chris Mason19c00dd2007-10-15 16:19:22 -0400418 goto out;
Chris Masoneb14ab82011-02-10 12:35:00 -0500419 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400420 len = page->private >> 2;
Chris Masond3977122009-01-05 21:25:51 -0500421 WARN_ON(len == 0);
422
David Sterbaba144192011-04-21 01:12:06 +0200423 eb = alloc_extent_buffer(tree, start, len, page);
Tsutomu Itoh91ca3382011-01-05 02:32:22 +0000424 if (eb == NULL) {
425 WARN_ON(1);
426 goto out;
427 }
Chris Masonca7a79a2008-05-12 12:59:19 -0400428 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
429 btrfs_header_generation(eb));
Chris Masonf1885912008-04-09 16:28:12 -0400430 BUG_ON(ret);
Chris Mason784b4e22010-11-21 22:20:49 -0500431 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
432
Chris Mason19c00dd2007-10-15 16:19:22 -0400433 found_start = btrfs_header_bytenr(eb);
434 if (found_start != start) {
Chris Mason55c69072008-01-09 15:55:33 -0500435 WARN_ON(1);
436 goto err;
437 }
438 if (eb->first_page != page) {
Chris Mason55c69072008-01-09 15:55:33 -0500439 WARN_ON(1);
440 goto err;
441 }
442 if (!PageUptodate(page)) {
Chris Mason55c69072008-01-09 15:55:33 -0500443 WARN_ON(1);
444 goto err;
Chris Mason19c00dd2007-10-15 16:19:22 -0400445 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400446 csum_tree_block(root, eb, 0);
Chris Mason55c69072008-01-09 15:55:33 -0500447err:
Chris Mason19c00dd2007-10-15 16:19:22 -0400448 free_extent_buffer(eb);
449out:
450 return 0;
451}
452
Yan Zheng2b820322008-11-17 21:11:30 -0500453static int check_tree_block_fsid(struct btrfs_root *root,
454 struct extent_buffer *eb)
455{
456 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
457 u8 fsid[BTRFS_UUID_SIZE];
458 int ret = 1;
459
460 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
461 BTRFS_FSID_SIZE);
462 while (fs_devices) {
463 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
464 ret = 0;
465 break;
466 }
467 fs_devices = fs_devices->seed;
468 }
469 return ret;
470}
471
Josef Bacika826d6d2011-03-16 13:42:43 -0400472#define CORRUPT(reason, eb, root, slot) \
473 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
474 "root=%llu, slot=%d\n", reason, \
475 (unsigned long long)btrfs_header_bytenr(eb), \
476 (unsigned long long)root->objectid, slot)
477
478static noinline int check_leaf(struct btrfs_root *root,
479 struct extent_buffer *leaf)
480{
481 struct btrfs_key key;
482 struct btrfs_key leaf_key;
483 u32 nritems = btrfs_header_nritems(leaf);
484 int slot;
485
486 if (nritems == 0)
487 return 0;
488
489 /* Check the 0 item */
490 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
491 BTRFS_LEAF_DATA_SIZE(root)) {
492 CORRUPT("invalid item offset size pair", leaf, root, 0);
493 return -EIO;
494 }
495
496 /*
497 * Check to make sure each items keys are in the correct order and their
498 * offsets make sense. We only have to loop through nritems-1 because
499 * we check the current slot against the next slot, which verifies the
500 * next slot's offset+size makes sense and that the current's slot
501 * offset is correct.
502 */
503 for (slot = 0; slot < nritems - 1; slot++) {
504 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
505 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
506
507 /* Make sure the keys are in the right order */
508 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
509 CORRUPT("bad key order", leaf, root, slot);
510 return -EIO;
511 }
512
513 /*
514 * Make sure the offset and ends are right, remember that the
515 * item data starts at the end of the leaf and grows towards the
516 * front.
517 */
518 if (btrfs_item_offset_nr(leaf, slot) !=
519 btrfs_item_end_nr(leaf, slot + 1)) {
520 CORRUPT("slot offset bad", leaf, root, slot);
521 return -EIO;
522 }
523
524 /*
525 * Check to make sure that we don't point outside of the leaf,
526 * just incase all the items are consistent to eachother, but
527 * all point outside of the leaf.
528 */
529 if (btrfs_item_end_nr(leaf, slot) >
530 BTRFS_LEAF_DATA_SIZE(root)) {
531 CORRUPT("slot end outside of leaf", leaf, root, slot);
532 return -EIO;
533 }
534 }
535
536 return 0;
537}
538
Christoph Hellwigb2950862008-12-02 09:54:17 -0500539static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
Chris Masonce9adaa2008-04-09 16:28:12 -0400540 struct extent_state *state)
541{
542 struct extent_io_tree *tree;
543 u64 found_start;
544 int found_level;
545 unsigned long len;
546 struct extent_buffer *eb;
547 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400548 int ret = 0;
Chris Masonce9adaa2008-04-09 16:28:12 -0400549
550 tree = &BTRFS_I(page->mapping->host)->io_tree;
551 if (page->private == EXTENT_PAGE_PRIVATE)
552 goto out;
553 if (!page->private)
554 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500555
Chris Masonce9adaa2008-04-09 16:28:12 -0400556 len = page->private >> 2;
Chris Masond3977122009-01-05 21:25:51 -0500557 WARN_ON(len == 0);
558
David Sterbaba144192011-04-21 01:12:06 +0200559 eb = alloc_extent_buffer(tree, start, len, page);
Tsutomu Itoh91ca3382011-01-05 02:32:22 +0000560 if (eb == NULL) {
561 ret = -EIO;
562 goto out;
563 }
Chris Masonf1885912008-04-09 16:28:12 -0400564
Chris Masonce9adaa2008-04-09 16:28:12 -0400565 found_start = btrfs_header_bytenr(eb);
Chris Mason23a07862008-09-12 08:57:47 -0400566 if (found_start != start) {
David Sterba7a36dde2011-05-06 15:33:15 +0200567 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400568 "%llu %llu\n",
569 (unsigned long long)found_start,
570 (unsigned long long)eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400571 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400572 goto err;
573 }
574 if (eb->first_page != page) {
Chris Masond3977122009-01-05 21:25:51 -0500575 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
576 eb->first_page->index, page->index);
Chris Masonce9adaa2008-04-09 16:28:12 -0400577 WARN_ON(1);
Chris Masonf1885912008-04-09 16:28:12 -0400578 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400579 goto err;
580 }
Yan Zheng2b820322008-11-17 21:11:30 -0500581 if (check_tree_block_fsid(root, eb)) {
David Sterba7a36dde2011-05-06 15:33:15 +0200582 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
Chris Mason193f2842009-04-27 07:29:05 -0400583 (unsigned long long)eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400584 ret = -EIO;
585 goto err;
586 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400587 found_level = btrfs_header_level(eb);
588
Chris Mason85d4e462011-07-26 16:11:19 -0400589 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
590 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500591
Chris Masonce9adaa2008-04-09 16:28:12 -0400592 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400593 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400594 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400595 goto err;
596 }
597
598 /*
599 * If this is a leaf block and it is corrupt, set the corrupt bit so
600 * that we don't try and read the other copies of this block, just
601 * return -EIO.
602 */
603 if (found_level == 0 && check_leaf(root, eb)) {
604 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
605 ret = -EIO;
606 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400607
608 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
609 end = eb->start + end - 1;
Chris Masonce9adaa2008-04-09 16:28:12 -0400610err:
Arne Jansen4bb31e92011-06-10 13:55:54 +0200611 if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
612 clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
613 btree_readahead_hook(root, eb, eb->start, ret);
614 }
615
Chris Masonce9adaa2008-04-09 16:28:12 -0400616 free_extent_buffer(eb);
617out:
Chris Masonf1885912008-04-09 16:28:12 -0400618 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400619}
620
Arne Jansen4bb31e92011-06-10 13:55:54 +0200621static int btree_io_failed_hook(struct bio *failed_bio,
622 struct page *page, u64 start, u64 end,
Jan Schmidt32240a92011-11-20 07:33:38 -0500623 int mirror_num, struct extent_state *state)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200624{
625 struct extent_io_tree *tree;
626 unsigned long len;
627 struct extent_buffer *eb;
628 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
629
630 tree = &BTRFS_I(page->mapping->host)->io_tree;
631 if (page->private == EXTENT_PAGE_PRIVATE)
632 goto out;
633 if (!page->private)
634 goto out;
635
636 len = page->private >> 2;
637 WARN_ON(len == 0);
638
639 eb = alloc_extent_buffer(tree, start, len, page);
640 if (eb == NULL)
641 goto out;
642
643 if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
644 clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
645 btree_readahead_hook(root, eb, eb->start, -EIO);
646 }
Chris Masonc674e042011-11-03 22:23:13 -0400647 free_extent_buffer(eb);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200648
649out:
650 return -EIO; /* we fixed nothing */
651}
652
Chris Masonce9adaa2008-04-09 16:28:12 -0400653static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400654{
655 struct end_io_wq *end_io_wq = bio->bi_private;
656 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400657
Chris Masonce9adaa2008-04-09 16:28:12 -0400658 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400659 end_io_wq->error = err;
Chris Mason8b712842008-06-11 16:50:36 -0400660 end_io_wq->work.func = end_workqueue_fn;
661 end_io_wq->work.flags = 0;
Chris Masond20f7042008-12-08 16:58:54 -0500662
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200663 if (bio->bi_rw & REQ_WRITE) {
Josef Bacik0cb59c92010-07-02 12:14:14 -0400664 if (end_io_wq->metadata == 1)
Chris Masoncad321a2008-12-17 14:51:42 -0500665 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
666 &end_io_wq->work);
Josef Bacik0cb59c92010-07-02 12:14:14 -0400667 else if (end_io_wq->metadata == 2)
668 btrfs_queue_worker(&fs_info->endio_freespace_worker,
669 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500670 else
671 btrfs_queue_worker(&fs_info->endio_write_workers,
672 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500673 } else {
674 if (end_io_wq->metadata)
675 btrfs_queue_worker(&fs_info->endio_meta_workers,
676 &end_io_wq->work);
677 else
678 btrfs_queue_worker(&fs_info->endio_workers,
679 &end_io_wq->work);
680 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400681}
682
Josef Bacik0cb59c92010-07-02 12:14:14 -0400683/*
684 * For the metadata arg you want
685 *
686 * 0 - if data
687 * 1 - if normal metadta
688 * 2 - if writing to the free space cache area
689 */
Chris Mason22c59942008-04-09 16:28:12 -0400690int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
691 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400692{
Chris Masonce9adaa2008-04-09 16:28:12 -0400693 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400694 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
695 if (!end_io_wq)
696 return -ENOMEM;
697
698 end_io_wq->private = bio->bi_private;
699 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400700 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400701 end_io_wq->error = 0;
702 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400703 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400704
705 bio->bi_private = end_io_wq;
706 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400707 return 0;
708}
709
Chris Masonb64a2852008-08-20 13:39:41 -0400710unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400711{
712 unsigned long limit = min_t(unsigned long,
713 info->workers.max_workers,
714 info->fs_devices->open_devices);
715 return 256 * limit;
716}
717
Chris Mason4a69a412008-11-06 22:03:00 -0500718static void run_one_async_start(struct btrfs_work *work)
719{
Chris Mason4a69a412008-11-06 22:03:00 -0500720 struct async_submit_bio *async;
721
722 async = container_of(work, struct async_submit_bio, work);
Chris Mason4a69a412008-11-06 22:03:00 -0500723 async->submit_bio_start(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400724 async->mirror_num, async->bio_flags,
725 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500726}
727
728static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400729{
730 struct btrfs_fs_info *fs_info;
731 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400732 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400733
734 async = container_of(work, struct async_submit_bio, work);
735 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400736
Chris Masonb64a2852008-08-20 13:39:41 -0400737 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400738 limit = limit * 2 / 3;
739
Chris Mason8b712842008-06-11 16:50:36 -0400740 atomic_dec(&fs_info->nr_async_submits);
Chris Mason0986fe9e2008-08-15 15:34:15 -0400741
Chris Masonb64a2852008-08-20 13:39:41 -0400742 if (atomic_read(&fs_info->nr_async_submits) < limit &&
743 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400744 wake_up(&fs_info->async_submit_wait);
745
Chris Mason4a69a412008-11-06 22:03:00 -0500746 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400747 async->mirror_num, async->bio_flags,
748 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500749}
750
751static void run_one_async_free(struct btrfs_work *work)
752{
753 struct async_submit_bio *async;
754
755 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400756 kfree(async);
757}
758
Chris Mason44b8bd72008-04-16 11:14:51 -0400759int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
760 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400761 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400762 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500763 extent_submit_bio_hook_t *submit_bio_start,
764 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400765{
766 struct async_submit_bio *async;
767
768 async = kmalloc(sizeof(*async), GFP_NOFS);
769 if (!async)
770 return -ENOMEM;
771
772 async->inode = inode;
773 async->rw = rw;
774 async->bio = bio;
775 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500776 async->submit_bio_start = submit_bio_start;
777 async->submit_bio_done = submit_bio_done;
778
779 async->work.func = run_one_async_start;
780 async->work.ordered_func = run_one_async_done;
781 async->work.ordered_free = run_one_async_free;
782
Chris Mason8b712842008-06-11 16:50:36 -0400783 async->work.flags = 0;
Chris Masonc8b97812008-10-29 14:49:59 -0400784 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400785 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400786
Chris Masoncb03c742008-05-15 16:15:45 -0400787 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400788
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200789 if (rw & REQ_SYNC)
Chris Masond313d7a2009-04-20 15:50:09 -0400790 btrfs_set_work_high_prio(&async->work);
791
Chris Mason8b712842008-06-11 16:50:36 -0400792 btrfs_queue_worker(&fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400793
Chris Masond3977122009-01-05 21:25:51 -0500794 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500795 atomic_read(&fs_info->nr_async_submits)) {
796 wait_event(fs_info->async_submit_wait,
797 (atomic_read(&fs_info->nr_async_submits) == 0));
798 }
799
Chris Mason44b8bd72008-04-16 11:14:51 -0400800 return 0;
801}
802
Chris Masonce3ed712008-09-23 13:14:12 -0400803static int btree_csum_one_bio(struct bio *bio)
804{
805 struct bio_vec *bvec = bio->bi_io_vec;
806 int bio_index = 0;
807 struct btrfs_root *root;
808
809 WARN_ON(bio->bi_vcnt <= 0);
Chris Masond3977122009-01-05 21:25:51 -0500810 while (bio_index < bio->bi_vcnt) {
Chris Masonce3ed712008-09-23 13:14:12 -0400811 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
812 csum_dirty_buffer(root, bvec->bv_page);
813 bio_index++;
814 bvec++;
815 }
816 return 0;
817}
818
Chris Mason4a69a412008-11-06 22:03:00 -0500819static int __btree_submit_bio_start(struct inode *inode, int rw,
820 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400821 unsigned long bio_flags,
822 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400823{
Chris Mason8b712842008-06-11 16:50:36 -0400824 /*
825 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400826 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400827 */
Chris Mason4a69a412008-11-06 22:03:00 -0500828 btree_csum_one_bio(bio);
829 return 0;
830}
Chris Mason22c59942008-04-09 16:28:12 -0400831
Chris Mason4a69a412008-11-06 22:03:00 -0500832static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400833 int mirror_num, unsigned long bio_flags,
834 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500835{
Chris Mason8b712842008-06-11 16:50:36 -0400836 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500837 * when we're called for a write, we're already in the async
838 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400839 */
Chris Mason8b712842008-06-11 16:50:36 -0400840 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Chris Mason0b86a832008-03-24 15:01:56 -0400841}
842
Chris Mason44b8bd72008-04-16 11:14:51 -0400843static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400844 int mirror_num, unsigned long bio_flags,
845 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400846{
Chris Masoncad321a2008-12-17 14:51:42 -0500847 int ret;
848
849 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
850 bio, 1);
851 BUG_ON(ret);
852
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200853 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500854 /*
855 * called for a read, do the setup so that checksum validation
856 * can happen in the async kernel threads
857 */
Chris Mason4a69a412008-11-06 22:03:00 -0500858 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
Chris Mason6f3577b2008-11-13 09:59:36 -0500859 mirror_num, 0);
Chris Mason44b8bd72008-04-16 11:14:51 -0400860 }
Chris Masond313d7a2009-04-20 15:50:09 -0400861
Chris Masoncad321a2008-12-17 14:51:42 -0500862 /*
863 * kthread helpers are used to submit writes so that checksumming
864 * can happen in parallel across all CPUs
865 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400866 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
Chris Masonc8b97812008-10-29 14:49:59 -0400867 inode, rw, bio, mirror_num, 0,
Chris Masoneaf25d92010-05-25 09:48:28 -0400868 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500869 __btree_submit_bio_start,
870 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400871}
872
Jan Beulich3dd14622010-12-07 14:54:09 +0000873#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500874static int btree_migratepage(struct address_space *mapping,
875 struct page *newpage, struct page *page)
876{
877 /*
878 * we can't safely write a btree page from here,
879 * we haven't done the locking hook
880 */
881 if (PageDirty(page))
882 return -EAGAIN;
883 /*
884 * Buffers may be managed in a filesystem specific way.
885 * We must have no buffers or drop them.
886 */
887 if (page_has_private(page) &&
888 !try_to_release_page(page, GFP_KERNEL))
889 return -EAGAIN;
Chris Mason784b4e22010-11-21 22:20:49 -0500890 return migrate_page(mapping, newpage, page);
891}
Jan Beulich3dd14622010-12-07 14:54:09 +0000892#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500893
Chris Mason5f39d392007-10-15 16:14:19 -0400894static int btree_writepage(struct page *page, struct writeback_control *wbc)
895{
Chris Masond1310b22008-01-24 16:13:08 -0500896 struct extent_io_tree *tree;
Chris Masonb9473432009-03-13 11:00:37 -0400897 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
898 struct extent_buffer *eb;
899 int was_dirty;
Chris Mason5443be42008-08-15 15:34:16 -0400900
Chris Masonb9473432009-03-13 11:00:37 -0400901 tree = &BTRFS_I(page->mapping->host)->io_tree;
902 if (!(current->flags & PF_MEMALLOC)) {
903 return extent_write_full_page(tree, page,
904 btree_get_extent, wbc);
Chris Mason5443be42008-08-15 15:34:16 -0400905 }
Chris Masonb9473432009-03-13 11:00:37 -0400906
907 redirty_page_for_writepage(wbc, page);
Chris Mason784b4e22010-11-21 22:20:49 -0500908 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
Chris Masonb9473432009-03-13 11:00:37 -0400909 WARN_ON(!eb);
910
911 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
912 if (!was_dirty) {
913 spin_lock(&root->fs_info->delalloc_lock);
914 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
915 spin_unlock(&root->fs_info->delalloc_lock);
916 }
917 free_extent_buffer(eb);
918
919 unlock_page(page);
920 return 0;
Chris Mason5f39d392007-10-15 16:14:19 -0400921}
Chris Mason0da54682007-11-07 21:08:01 -0500922
923static int btree_writepages(struct address_space *mapping,
924 struct writeback_control *wbc)
925{
Chris Masond1310b22008-01-24 16:13:08 -0500926 struct extent_io_tree *tree;
927 tree = &BTRFS_I(mapping->host)->io_tree;
Chris Masond8d5f3e2007-12-11 12:42:00 -0500928 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Masonb9473432009-03-13 11:00:37 -0400929 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
Chris Mason793955b2007-11-26 16:34:41 -0800930 u64 num_dirty;
Chris Mason24ab9cd2008-09-24 14:51:30 -0400931 unsigned long thresh = 32 * 1024 * 1024;
Chris Mason448d6402007-11-27 07:52:01 -0800932
933 if (wbc->for_kupdate)
934 return 0;
935
Chris Masonb9473432009-03-13 11:00:37 -0400936 /* this is a bit racy, but that's ok */
937 num_dirty = root->fs_info->dirty_metadata_bytes;
Chris Masond3977122009-01-05 21:25:51 -0500938 if (num_dirty < thresh)
Chris Mason793955b2007-11-26 16:34:41 -0800939 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800940 }
Chris Mason0da54682007-11-07 21:08:01 -0500941 return extent_writepages(tree, mapping, btree_get_extent, wbc);
942}
943
Christoph Hellwigb2950862008-12-02 09:54:17 -0500944static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -0400945{
Chris Masond1310b22008-01-24 16:13:08 -0500946 struct extent_io_tree *tree;
947 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +0200948 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400949}
950
Chris Mason70dec802008-01-29 09:59:12 -0500951static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -0400952{
Chris Masond1310b22008-01-24 16:13:08 -0500953 struct extent_io_tree *tree;
954 struct extent_map_tree *map;
Chris Mason5f39d392007-10-15 16:14:19 -0400955 int ret;
956
Chris Mason98509cf2008-09-11 15:51:43 -0400957 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -0500958 return 0;
Chris Mason98509cf2008-09-11 15:51:43 -0400959
Chris Masond1310b22008-01-24 16:13:08 -0500960 tree = &BTRFS_I(page->mapping->host)->io_tree;
961 map = &BTRFS_I(page->mapping->host)->extent_tree;
Chris Mason6af118ce2008-07-22 11:18:07 -0400962
Chris Mason7b13b7b2008-04-18 10:29:50 -0400963 ret = try_release_extent_state(map, tree, page, gfp_flags);
Chris Masond3977122009-01-05 21:25:51 -0500964 if (!ret)
Chris Mason6af118ce2008-07-22 11:18:07 -0400965 return 0;
Chris Mason6af118ce2008-07-22 11:18:07 -0400966
967 ret = try_release_extent_buffer(tree, page);
Chris Mason5f39d392007-10-15 16:14:19 -0400968 if (ret == 1) {
969 ClearPagePrivate(page);
970 set_page_private(page, 0);
971 page_cache_release(page);
972 }
Chris Mason6af118ce2008-07-22 11:18:07 -0400973
Chris Masond98237b2007-03-28 13:57:48 -0400974 return ret;
Chris Masone20d96d2007-03-22 12:13:20 -0400975}
Chris Mason123abc82007-03-14 14:14:43 -0400976
Chris Mason5f39d392007-10-15 16:14:19 -0400977static void btree_invalidatepage(struct page *page, unsigned long offset)
Chris Masond98237b2007-03-28 13:57:48 -0400978{
Chris Masond1310b22008-01-24 16:13:08 -0500979 struct extent_io_tree *tree;
980 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -0400981 extent_invalidatepage(tree, page, offset);
982 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400983 if (PagePrivate(page)) {
Chris Masond3977122009-01-05 21:25:51 -0500984 printk(KERN_WARNING "btrfs warning page private not zero "
985 "on page %llu\n", (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400986 ClearPagePrivate(page);
987 set_page_private(page, 0);
988 page_cache_release(page);
989 }
Chris Masond98237b2007-03-28 13:57:48 -0400990}
991
Alexey Dobriyan7f094102009-09-21 17:01:10 -0700992static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -0400993 .readpage = btree_readpage,
994 .writepage = btree_writepage,
Chris Mason0da54682007-11-07 21:08:01 -0500995 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -0400996 .releasepage = btree_releasepage,
997 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -0500998#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500999 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001000#endif
Chris Masond98237b2007-03-28 13:57:48 -04001001};
1002
Chris Masonca7a79a2008-05-12 12:59:19 -04001003int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1004 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001005{
Chris Mason5f39d392007-10-15 16:14:19 -04001006 struct extent_buffer *buf = NULL;
1007 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001008 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001009
Chris Masondb945352007-10-15 16:15:53 -04001010 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001011 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001012 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001013 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001014 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001015 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001016 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001017}
1018
Arne Jansenab0fff02011-05-23 14:25:41 +02001019int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1020 int mirror_num, struct extent_buffer **eb)
1021{
1022 struct extent_buffer *buf = NULL;
1023 struct inode *btree_inode = root->fs_info->btree_inode;
1024 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1025 int ret;
1026
1027 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1028 if (!buf)
1029 return 0;
1030
1031 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1032
1033 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1034 btree_get_extent, mirror_num);
1035 if (ret) {
1036 free_extent_buffer(buf);
1037 return ret;
1038 }
1039
1040 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1041 free_extent_buffer(buf);
1042 return -EIO;
1043 } else if (extent_buffer_uptodate(io_tree, buf, NULL)) {
1044 *eb = buf;
1045 } else {
1046 free_extent_buffer(buf);
1047 }
1048 return 0;
1049}
1050
Chris Mason0999df52008-04-01 13:48:14 -04001051struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1052 u64 bytenr, u32 blocksize)
1053{
1054 struct inode *btree_inode = root->fs_info->btree_inode;
1055 struct extent_buffer *eb;
1056 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
David Sterbaf09d1f62011-04-21 01:08:01 +02001057 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001058 return eb;
1059}
1060
1061struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1062 u64 bytenr, u32 blocksize)
1063{
1064 struct inode *btree_inode = root->fs_info->btree_inode;
1065 struct extent_buffer *eb;
1066
1067 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
David Sterbaba144192011-04-21 01:12:06 +02001068 bytenr, blocksize, NULL);
Chris Mason0999df52008-04-01 13:48:14 -04001069 return eb;
1070}
1071
1072
Chris Masone02119d2008-09-05 16:13:11 -04001073int btrfs_write_tree_block(struct extent_buffer *buf)
1074{
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001075 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
1076 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001077}
1078
1079int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1080{
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001081 return filemap_fdatawait_range(buf->first_page->mapping,
1082 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001083}
1084
Chris Masondb945352007-10-15 16:15:53 -04001085struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001086 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001087{
Chris Mason5f39d392007-10-15 16:14:19 -04001088 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001089 int ret;
1090
Chris Masondb945352007-10-15 16:15:53 -04001091 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001092 if (!buf)
1093 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001094
Chris Masonca7a79a2008-05-12 12:59:19 -04001095 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Masonce9adaa2008-04-09 16:28:12 -04001096
Chris Masond3977122009-01-05 21:25:51 -05001097 if (ret == 0)
Chris Masonb4ce94d2009-02-04 09:25:08 -05001098 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
Chris Mason5f39d392007-10-15 16:14:19 -04001099 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001100
Chris Masoneb60cea2007-02-02 09:18:22 -05001101}
1102
Chris Masone089f052007-03-16 16:20:31 -04001103int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
Chris Mason5f39d392007-10-15 16:14:19 -04001104 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001105{
Chris Mason5f39d392007-10-15 16:14:19 -04001106 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason55c69072008-01-09 15:55:33 -05001107 if (btrfs_header_generation(buf) ==
Chris Mason925baed2008-06-25 16:01:30 -04001108 root->fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001109 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001110
Chris Masonb9473432009-03-13 11:00:37 -04001111 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1112 spin_lock(&root->fs_info->delalloc_lock);
1113 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1114 root->fs_info->dirty_metadata_bytes -= buf->len;
1115 else
1116 WARN_ON(1);
1117 spin_unlock(&root->fs_info->delalloc_lock);
1118 }
Chris Masonb4ce94d2009-02-04 09:25:08 -05001119
Chris Masonb9473432009-03-13 11:00:37 -04001120 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1121 btrfs_set_lock_blocking(buf);
Chris Masond1310b22008-01-24 16:13:08 -05001122 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
Chris Mason55c69072008-01-09 15:55:33 -05001123 buf);
Chris Mason925baed2008-06-25 16:01:30 -04001124 }
Chris Mason5f39d392007-10-15 16:14:19 -04001125 return 0;
1126}
1127
Chris Masondb945352007-10-15 16:15:53 -04001128static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001129 u32 stripesize, struct btrfs_root *root,
Chris Mason9f5fae22007-03-20 14:38:32 -04001130 struct btrfs_fs_info *fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04001131 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001132{
Chris Masoncfaa7292007-02-21 17:04:57 -05001133 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001134 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001135 root->sectorsize = sectorsize;
1136 root->nodesize = nodesize;
1137 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001138 root->stripesize = stripesize;
Chris Mason123abc82007-03-14 14:14:43 -04001139 root->ref_cows = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001140 root->track_dirty = 0;
Yan, Zhengc71bf092009-11-12 09:34:40 +00001141 root->in_radix = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001142 root->orphan_item_inserted = 0;
1143 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001144
Chris Mason9f5fae22007-03-20 14:38:32 -04001145 root->fs_info = fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001146 root->objectid = objectid;
1147 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001148 root->highest_objectid = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001149 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001150 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001151 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001152 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001153 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001154
1155 INIT_LIST_HEAD(&root->dirty_list);
Josef Bacik7b128762008-07-24 12:17:14 -04001156 INIT_LIST_HEAD(&root->orphan_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001157 INIT_LIST_HEAD(&root->root_list);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001158 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001159 spin_lock_init(&root->inode_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001160 spin_lock_init(&root->accounting_lock);
Chris Masona2135012008-06-25 16:01:30 -04001161 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001162 mutex_init(&root->log_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001163 init_waitqueue_head(&root->log_writer_wait);
1164 init_waitqueue_head(&root->log_commit_wait[0]);
1165 init_waitqueue_head(&root->log_commit_wait[1]);
1166 atomic_set(&root->log_commit[0], 0);
1167 atomic_set(&root->log_commit[1], 0);
1168 atomic_set(&root->log_writers, 0);
1169 root->log_batch = 0;
1170 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001171 root->last_log_commit = 0;
Chris Masond0c803c2008-09-11 16:17:57 -04001172 extent_io_tree_init(&root->dirty_log_pages,
David Sterbaf993c882011-04-20 23:35:57 +02001173 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001174
Chris Mason3768f362007-03-13 16:47:54 -04001175 memset(&root->root_key, 0, sizeof(root->root_key));
1176 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001177 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001178 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Chris Mason3f157a22008-06-25 16:01:31 -04001179 root->defrag_trans_start = fs_info->generation;
Josef Bacik58176a92007-08-29 15:47:34 -04001180 init_completion(&root->kobj_unregister);
Chris Mason6702ed42007-08-07 16:15:09 -04001181 root->defrag_running = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001182 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001183 root->anon_dev = 0;
Chris Mason3768f362007-03-13 16:47:54 -04001184 return 0;
1185}
1186
Chris Masondb945352007-10-15 16:15:53 -04001187static int find_and_setup_root(struct btrfs_root *tree_root,
Chris Mason9f5fae22007-03-20 14:38:32 -04001188 struct btrfs_fs_info *fs_info,
1189 u64 objectid,
Chris Masone20d96d2007-03-22 12:13:20 -04001190 struct btrfs_root *root)
Chris Mason3768f362007-03-13 16:47:54 -04001191{
1192 int ret;
Chris Masondb945352007-10-15 16:15:53 -04001193 u32 blocksize;
Yan Zheng84234f32008-10-29 14:49:05 -04001194 u64 generation;
Chris Mason3768f362007-03-13 16:47:54 -04001195
Chris Masondb945352007-10-15 16:15:53 -04001196 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001197 tree_root->sectorsize, tree_root->stripesize,
1198 root, fs_info, objectid);
Chris Mason3768f362007-03-13 16:47:54 -04001199 ret = btrfs_find_last_root(tree_root, objectid,
1200 &root->root_item, &root->root_key);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001201 if (ret > 0)
1202 return -ENOENT;
Chris Mason3768f362007-03-13 16:47:54 -04001203 BUG_ON(ret);
1204
Yan Zheng84234f32008-10-29 14:49:05 -04001205 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001206 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
Chris Masonaf31f5e2011-11-03 15:17:42 -04001207 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001208 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001209 blocksize, generation);
Chris Mason68433b72010-12-13 14:47:58 -05001210 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1211 free_extent_buffer(root->node);
Chris Masonaf31f5e2011-11-03 15:17:42 -04001212 root->node = NULL;
Chris Mason68433b72010-12-13 14:47:58 -05001213 return -EIO;
1214 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001215 root->commit_root = btrfs_root_node(root);
Chris Masond97e63b2007-02-20 16:40:44 -05001216 return 0;
1217}
1218
Yan Zheng7237f182009-01-21 12:54:03 -05001219static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1220 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001221{
1222 struct btrfs_root *root;
1223 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001224 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001225
1226 root = kzalloc(sizeof(*root), GFP_NOFS);
1227 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001228 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001229
1230 __setup_root(tree_root->nodesize, tree_root->leafsize,
1231 tree_root->sectorsize, tree_root->stripesize,
1232 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1233
1234 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1235 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1236 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Yan Zheng7237f182009-01-21 12:54:03 -05001237 /*
1238 * log trees do not get reference counted because they go away
1239 * before a real commit is actually done. They do store pointers
1240 * to file data extents, and those reference counts still get
1241 * updated (along with back refs to the log tree).
1242 */
Chris Masone02119d2008-09-05 16:13:11 -04001243 root->ref_cows = 0;
1244
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001245 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1246 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001247 if (IS_ERR(leaf)) {
1248 kfree(root);
1249 return ERR_CAST(leaf);
1250 }
Chris Masone02119d2008-09-05 16:13:11 -04001251
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001252 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1253 btrfs_set_header_bytenr(leaf, leaf->start);
1254 btrfs_set_header_generation(leaf, trans->transid);
1255 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1256 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001257 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001258
1259 write_extent_buffer(root->node, root->fs_info->fsid,
1260 (unsigned long)btrfs_header_fsid(root->node),
1261 BTRFS_FSID_SIZE);
1262 btrfs_mark_buffer_dirty(root->node);
1263 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001264 return root;
1265}
1266
1267int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1268 struct btrfs_fs_info *fs_info)
1269{
1270 struct btrfs_root *log_root;
1271
1272 log_root = alloc_log_tree(trans, fs_info);
1273 if (IS_ERR(log_root))
1274 return PTR_ERR(log_root);
1275 WARN_ON(fs_info->log_root_tree);
1276 fs_info->log_root_tree = log_root;
1277 return 0;
1278}
1279
1280int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1281 struct btrfs_root *root)
1282{
1283 struct btrfs_root *log_root;
1284 struct btrfs_inode_item *inode_item;
1285
1286 log_root = alloc_log_tree(trans, root->fs_info);
1287 if (IS_ERR(log_root))
1288 return PTR_ERR(log_root);
1289
1290 log_root->last_trans = trans->transid;
1291 log_root->root_key.offset = root->root_key.objectid;
1292
1293 inode_item = &log_root->root_item.inode;
1294 inode_item->generation = cpu_to_le64(1);
1295 inode_item->size = cpu_to_le64(3);
1296 inode_item->nlink = cpu_to_le32(1);
1297 inode_item->nbytes = cpu_to_le64(root->leafsize);
1298 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1299
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001300 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001301
1302 WARN_ON(root->log_root);
1303 root->log_root = log_root;
1304 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001305 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001306 return 0;
1307}
1308
1309struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1310 struct btrfs_key *location)
1311{
1312 struct btrfs_root *root;
1313 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001314 struct btrfs_path *path;
Chris Mason5f39d392007-10-15 16:14:19 -04001315 struct extent_buffer *l;
Yan Zheng84234f32008-10-29 14:49:05 -04001316 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001317 u32 blocksize;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001318 int ret = 0;
1319
Chris Mason5eda7b52007-06-22 14:16:25 -04001320 root = kzalloc(sizeof(*root), GFP_NOFS);
Chris Mason0cf6c622007-06-09 09:22:25 -04001321 if (!root)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001322 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001323 if (location->offset == (u64)-1) {
Chris Masondb945352007-10-15 16:15:53 -04001324 ret = find_and_setup_root(tree_root, fs_info,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001325 location->objectid, root);
1326 if (ret) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04001327 kfree(root);
1328 return ERR_PTR(ret);
1329 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001330 goto out;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001331 }
1332
Chris Masondb945352007-10-15 16:15:53 -04001333 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001334 tree_root->sectorsize, tree_root->stripesize,
1335 root, fs_info, location->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001336
1337 path = btrfs_alloc_path();
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00001338 if (!path) {
1339 kfree(root);
1340 return ERR_PTR(-ENOMEM);
1341 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001342 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001343 if (ret == 0) {
1344 l = path->nodes[0];
1345 read_extent_buffer(l, &root->root_item,
1346 btrfs_item_ptr_offset(l, path->slots[0]),
1347 sizeof(root->root_item));
1348 memcpy(&root->root_key, location, sizeof(*location));
Chris Mason0f7d52f2007-04-09 10:42:37 -04001349 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001350 btrfs_free_path(path);
1351 if (ret) {
Tsutomu Itoh5e540f72010-12-27 06:53:10 +00001352 kfree(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001353 if (ret > 0)
1354 ret = -ENOENT;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001355 return ERR_PTR(ret);
1356 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001357
Yan Zheng84234f32008-10-29 14:49:05 -04001358 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001359 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1360 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001361 blocksize, generation);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001362 root->commit_root = btrfs_root_node(root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001363 BUG_ON(!root->node);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001364out:
Li Zefan08fe4db2011-03-28 02:01:25 +00001365 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
Chris Masone02119d2008-09-05 16:13:11 -04001366 root->ref_cows = 1;
Li Zefan08fe4db2011-03-28 02:01:25 +00001367 btrfs_check_and_init_root_item(&root->root_item);
1368 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001369
Chris Mason5eda7b52007-06-22 14:16:25 -04001370 return root;
1371}
1372
Chris Masonedbd8d42007-12-21 16:27:24 -05001373struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1374 struct btrfs_key *location)
Chris Mason5eda7b52007-06-22 14:16:25 -04001375{
1376 struct btrfs_root *root;
1377 int ret;
1378
Chris Masonedbd8d42007-12-21 16:27:24 -05001379 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1380 return fs_info->tree_root;
1381 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1382 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001383 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1384 return fs_info->chunk_root;
1385 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1386 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001387 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1388 return fs_info->csum_root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001389again:
1390 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001391 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1392 (unsigned long)location->objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001393 spin_unlock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001394 if (root)
1395 return root;
1396
Chris Masone02119d2008-09-05 16:13:11 -04001397 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001398 if (IS_ERR(root))
1399 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001400
Li Zefan581bb052011-04-20 10:06:11 +08001401 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
Li Zefan581bb052011-04-20 10:06:11 +08001402 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1403 GFP_NOFS);
David Sterba35a30d72011-06-13 15:18:23 +00001404 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1405 ret = -ENOMEM;
Li Zefan581bb052011-04-20 10:06:11 +08001406 goto fail;
David Sterba35a30d72011-06-13 15:18:23 +00001407 }
Li Zefan581bb052011-04-20 10:06:11 +08001408
1409 btrfs_init_free_ino_ctl(root);
1410 mutex_init(&root->fs_commit_mutex);
1411 spin_lock_init(&root->cache_lock);
1412 init_waitqueue_head(&root->cache_wait);
1413
Al Viro0ee5dc62011-07-07 15:44:25 -04001414 ret = get_anon_bdev(&root->anon_dev);
Chris Masonac08aed2011-06-13 11:28:50 -04001415 if (ret)
1416 goto fail;
Chris Mason3394e162008-11-17 20:42:26 -05001417
Yan, Zhengd68fc572010-05-16 10:49:58 -04001418 if (btrfs_root_refs(&root->root_item) == 0) {
1419 ret = -ENOENT;
1420 goto fail;
1421 }
1422
1423 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1424 if (ret < 0)
1425 goto fail;
1426 if (ret == 0)
1427 root->orphan_item_inserted = 1;
1428
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001429 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1430 if (ret)
1431 goto fail;
1432
1433 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04001434 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1435 (unsigned long)root->root_key.objectid,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001436 root);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001437 if (ret == 0)
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001438 root->in_radix = 1;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001439
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001440 spin_unlock(&fs_info->fs_roots_radix_lock);
1441 radix_tree_preload_end();
Chris Mason0f7d52f2007-04-09 10:42:37 -04001442 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001443 if (ret == -EEXIST) {
1444 free_fs_root(root);
1445 goto again;
1446 }
1447 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001448 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001449
1450 ret = btrfs_find_dead_roots(fs_info->tree_root,
1451 root->root_key.objectid);
1452 WARN_ON(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001453 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001454fail:
1455 free_fs_root(root);
1456 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001457}
1458
Chris Mason04160082008-03-26 10:28:07 -04001459static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1460{
1461 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1462 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001463 struct btrfs_device *device;
1464 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001465
Xiao Guangrong1f781602011-04-20 10:09:16 +00001466 rcu_read_lock();
1467 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001468 if (!device->bdev)
1469 continue;
Chris Mason04160082008-03-26 10:28:07 -04001470 bdi = blk_get_backing_dev_info(device->bdev);
1471 if (bdi && bdi_congested(bdi, bdi_bits)) {
1472 ret = 1;
1473 break;
1474 }
1475 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001476 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001477 return ret;
1478}
1479
Chris Mason38b66982008-04-22 09:22:11 -04001480/*
Jens Axboead081f12009-06-12 14:43:40 +02001481 * If this fails, caller must call bdi_destroy() to get rid of the
1482 * bdi again.
1483 */
Chris Mason04160082008-03-26 10:28:07 -04001484static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1485{
Jens Axboead081f12009-06-12 14:43:40 +02001486 int err;
1487
1488 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001489 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001490 if (err)
1491 return err;
1492
Chris Mason4575c9c2008-04-18 16:13:31 -04001493 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001494 bdi->congested_fn = btrfs_congested_fn;
1495 bdi->congested_data = info;
1496 return 0;
1497}
1498
Chris Masonce9adaa2008-04-09 16:28:12 -04001499static int bio_ready_for_csum(struct bio *bio)
1500{
1501 u64 length = 0;
1502 u64 buf_len = 0;
1503 u64 start = 0;
1504 struct page *page;
1505 struct extent_io_tree *io_tree = NULL;
Chris Masonce9adaa2008-04-09 16:28:12 -04001506 struct bio_vec *bvec;
1507 int i;
1508 int ret;
1509
1510 bio_for_each_segment(bvec, bio, i) {
1511 page = bvec->bv_page;
1512 if (page->private == EXTENT_PAGE_PRIVATE) {
1513 length += bvec->bv_len;
1514 continue;
1515 }
1516 if (!page->private) {
1517 length += bvec->bv_len;
1518 continue;
1519 }
1520 length = bvec->bv_len;
1521 buf_len = page->private >> 2;
1522 start = page_offset(page) + bvec->bv_offset;
1523 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Masonce9adaa2008-04-09 16:28:12 -04001524 }
1525 /* are we fully contained in this bio? */
1526 if (buf_len <= length)
1527 return 1;
1528
1529 ret = extent_range_uptodate(io_tree, start + length,
1530 start + buf_len - 1);
Chris Masonce9adaa2008-04-09 16:28:12 -04001531 return ret;
1532}
1533
Chris Mason8b712842008-06-11 16:50:36 -04001534/*
1535 * called by the kthread helper functions to finally call the bio end_io
1536 * functions. This is where read checksum verification actually happens
1537 */
1538static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001539{
Chris Masonce9adaa2008-04-09 16:28:12 -04001540 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001541 struct end_io_wq *end_io_wq;
1542 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001543 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001544
Chris Mason8b712842008-06-11 16:50:36 -04001545 end_io_wq = container_of(work, struct end_io_wq, work);
1546 bio = end_io_wq->bio;
1547 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001548
Chris Masoncad321a2008-12-17 14:51:42 -05001549 /* metadata bio reads are special because the whole tree block must
Chris Mason8b712842008-06-11 16:50:36 -04001550 * be checksummed at once. This makes sure the entire block is in
1551 * ram and up to date before trying to verify things. For
1552 * blocksize <= pagesize, it is basically a noop
1553 */
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +02001554 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
Chris Masoncad321a2008-12-17 14:51:42 -05001555 !bio_ready_for_csum(bio)) {
Chris Masond20f7042008-12-08 16:58:54 -05001556 btrfs_queue_worker(&fs_info->endio_meta_workers,
Chris Mason8b712842008-06-11 16:50:36 -04001557 &end_io_wq->work);
1558 return;
1559 }
1560 error = end_io_wq->error;
1561 bio->bi_private = end_io_wq->private;
1562 bio->bi_end_io = end_io_wq->end_io;
1563 kfree(end_io_wq);
Chris Mason8b712842008-06-11 16:50:36 -04001564 bio_endio(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001565}
1566
Chris Masona74a4b92008-06-25 16:01:31 -04001567static int cleaner_kthread(void *arg)
1568{
1569 struct btrfs_root *root = arg;
1570
1571 do {
Chris Masona74a4b92008-06-25 16:01:31 -04001572 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
Yan, Zheng76dda932009-09-21 16:00:26 -04001573
1574 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1575 mutex_trylock(&root->fs_info->cleaner_mutex)) {
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001576 btrfs_run_delayed_iputs(root);
Yan, Zheng76dda932009-09-21 16:00:26 -04001577 btrfs_clean_old_snapshots(root);
1578 mutex_unlock(&root->fs_info->cleaner_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001579 btrfs_run_defrag_inodes(root->fs_info);
Yan, Zheng76dda932009-09-21 16:00:26 -04001580 }
Chris Masona74a4b92008-06-25 16:01:31 -04001581
1582 if (freezing(current)) {
1583 refrigerator();
1584 } else {
Chris Masona74a4b92008-06-25 16:01:31 -04001585 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001586 if (!kthread_should_stop())
1587 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001588 __set_current_state(TASK_RUNNING);
1589 }
1590 } while (!kthread_should_stop());
1591 return 0;
1592}
1593
1594static int transaction_kthread(void *arg)
1595{
1596 struct btrfs_root *root = arg;
1597 struct btrfs_trans_handle *trans;
1598 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001599 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001600 unsigned long now;
1601 unsigned long delay;
1602 int ret;
1603
1604 do {
Chris Masona74a4b92008-06-25 16:01:31 -04001605 delay = HZ * 30;
1606 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1607 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1608
Josef Bacika4abeea2011-04-11 17:25:13 -04001609 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001610 cur = root->fs_info->running_transaction;
1611 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001612 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001613 goto sleep;
1614 }
Yan Zheng31153d82008-07-28 15:32:19 -04001615
Chris Masona74a4b92008-06-25 16:01:31 -04001616 now = get_seconds();
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001617 if (!cur->blocked &&
1618 (now < cur->start_time || now - cur->start_time < 30)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001619 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001620 delay = HZ * 5;
1621 goto sleep;
1622 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001623 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001624 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001625
Josef Bacik7a7eaa42011-04-13 12:54:33 -04001626 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00001627 BUG_ON(IS_ERR(trans));
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001628 if (transid == trans->transid) {
1629 ret = btrfs_commit_transaction(trans, root);
1630 BUG_ON(ret);
1631 } else {
1632 btrfs_end_transaction(trans, root);
1633 }
Chris Masona74a4b92008-06-25 16:01:31 -04001634sleep:
1635 wake_up_process(root->fs_info->cleaner_kthread);
1636 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1637
1638 if (freezing(current)) {
1639 refrigerator();
1640 } else {
Chris Masona74a4b92008-06-25 16:01:31 -04001641 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001642 if (!kthread_should_stop() &&
1643 !btrfs_transaction_blocked(root->fs_info))
1644 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001645 __set_current_state(TASK_RUNNING);
1646 }
1647 } while (!kthread_should_stop());
1648 return 0;
1649}
1650
Chris Masonaf31f5e2011-11-03 15:17:42 -04001651/*
1652 * this will find the highest generation in the array of
1653 * root backups. The index of the highest array is returned,
1654 * or -1 if we can't find anything.
1655 *
1656 * We check to make sure the array is valid by comparing the
1657 * generation of the latest root in the array with the generation
1658 * in the super block. If they don't match we pitch it.
1659 */
1660static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1661{
1662 u64 cur;
1663 int newest_index = -1;
1664 struct btrfs_root_backup *root_backup;
1665 int i;
1666
1667 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1668 root_backup = info->super_copy->super_roots + i;
1669 cur = btrfs_backup_tree_root_gen(root_backup);
1670 if (cur == newest_gen)
1671 newest_index = i;
1672 }
1673
1674 /* check to see if we actually wrapped around */
1675 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1676 root_backup = info->super_copy->super_roots;
1677 cur = btrfs_backup_tree_root_gen(root_backup);
1678 if (cur == newest_gen)
1679 newest_index = 0;
1680 }
1681 return newest_index;
1682}
1683
1684
1685/*
1686 * find the oldest backup so we know where to store new entries
1687 * in the backup array. This will set the backup_root_index
1688 * field in the fs_info struct
1689 */
1690static void find_oldest_super_backup(struct btrfs_fs_info *info,
1691 u64 newest_gen)
1692{
1693 int newest_index = -1;
1694
1695 newest_index = find_newest_super_backup(info, newest_gen);
1696 /* if there was garbage in there, just move along */
1697 if (newest_index == -1) {
1698 info->backup_root_index = 0;
1699 } else {
1700 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1701 }
1702}
1703
1704/*
1705 * copy all the root pointers into the super backup array.
1706 * this will bump the backup pointer by one when it is
1707 * done
1708 */
1709static void backup_super_roots(struct btrfs_fs_info *info)
1710{
1711 int next_backup;
1712 struct btrfs_root_backup *root_backup;
1713 int last_backup;
1714
1715 next_backup = info->backup_root_index;
1716 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1717 BTRFS_NUM_BACKUP_ROOTS;
1718
1719 /*
1720 * just overwrite the last backup if we're at the same generation
1721 * this happens only at umount
1722 */
1723 root_backup = info->super_for_commit->super_roots + last_backup;
1724 if (btrfs_backup_tree_root_gen(root_backup) ==
1725 btrfs_header_generation(info->tree_root->node))
1726 next_backup = last_backup;
1727
1728 root_backup = info->super_for_commit->super_roots + next_backup;
1729
1730 /*
1731 * make sure all of our padding and empty slots get zero filled
1732 * regardless of which ones we use today
1733 */
1734 memset(root_backup, 0, sizeof(*root_backup));
1735
1736 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1737
1738 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1739 btrfs_set_backup_tree_root_gen(root_backup,
1740 btrfs_header_generation(info->tree_root->node));
1741
1742 btrfs_set_backup_tree_root_level(root_backup,
1743 btrfs_header_level(info->tree_root->node));
1744
1745 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1746 btrfs_set_backup_chunk_root_gen(root_backup,
1747 btrfs_header_generation(info->chunk_root->node));
1748 btrfs_set_backup_chunk_root_level(root_backup,
1749 btrfs_header_level(info->chunk_root->node));
1750
1751 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1752 btrfs_set_backup_extent_root_gen(root_backup,
1753 btrfs_header_generation(info->extent_root->node));
1754 btrfs_set_backup_extent_root_level(root_backup,
1755 btrfs_header_level(info->extent_root->node));
1756
Chris Mason7c7e82a2011-11-06 18:50:56 -05001757 /*
1758 * we might commit during log recovery, which happens before we set
1759 * the fs_root. Make sure it is valid before we fill it in.
1760 */
1761 if (info->fs_root && info->fs_root->node) {
1762 btrfs_set_backup_fs_root(root_backup,
1763 info->fs_root->node->start);
1764 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001765 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001766 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001767 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001768 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001769
1770 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1771 btrfs_set_backup_dev_root_gen(root_backup,
1772 btrfs_header_generation(info->dev_root->node));
1773 btrfs_set_backup_dev_root_level(root_backup,
1774 btrfs_header_level(info->dev_root->node));
1775
1776 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1777 btrfs_set_backup_csum_root_gen(root_backup,
1778 btrfs_header_generation(info->csum_root->node));
1779 btrfs_set_backup_csum_root_level(root_backup,
1780 btrfs_header_level(info->csum_root->node));
1781
1782 btrfs_set_backup_total_bytes(root_backup,
1783 btrfs_super_total_bytes(info->super_copy));
1784 btrfs_set_backup_bytes_used(root_backup,
1785 btrfs_super_bytes_used(info->super_copy));
1786 btrfs_set_backup_num_devices(root_backup,
1787 btrfs_super_num_devices(info->super_copy));
1788
1789 /*
1790 * if we don't copy this out to the super_copy, it won't get remembered
1791 * for the next commit
1792 */
1793 memcpy(&info->super_copy->super_roots,
1794 &info->super_for_commit->super_roots,
1795 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1796}
1797
1798/*
1799 * this copies info out of the root backup array and back into
1800 * the in-memory super block. It is meant to help iterate through
1801 * the array, so you send it the number of backups you've already
1802 * tried and the last backup index you used.
1803 *
1804 * this returns -1 when it has tried all the backups
1805 */
1806static noinline int next_root_backup(struct btrfs_fs_info *info,
1807 struct btrfs_super_block *super,
1808 int *num_backups_tried, int *backup_index)
1809{
1810 struct btrfs_root_backup *root_backup;
1811 int newest = *backup_index;
1812
1813 if (*num_backups_tried == 0) {
1814 u64 gen = btrfs_super_generation(super);
1815
1816 newest = find_newest_super_backup(info, gen);
1817 if (newest == -1)
1818 return -1;
1819
1820 *backup_index = newest;
1821 *num_backups_tried = 1;
1822 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1823 /* we've tried all the backups, all done */
1824 return -1;
1825 } else {
1826 /* jump to the next oldest backup */
1827 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1828 BTRFS_NUM_BACKUP_ROOTS;
1829 *backup_index = newest;
1830 *num_backups_tried += 1;
1831 }
1832 root_backup = super->super_roots + newest;
1833
1834 btrfs_set_super_generation(super,
1835 btrfs_backup_tree_root_gen(root_backup));
1836 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1837 btrfs_set_super_root_level(super,
1838 btrfs_backup_tree_root_level(root_backup));
1839 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1840
1841 /*
1842 * fixme: the total bytes and num_devices need to match or we should
1843 * need a fsck
1844 */
1845 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1846 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1847 return 0;
1848}
1849
1850/* helper to cleanup tree roots */
1851static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
1852{
1853 free_extent_buffer(info->tree_root->node);
1854 free_extent_buffer(info->tree_root->commit_root);
1855 free_extent_buffer(info->dev_root->node);
1856 free_extent_buffer(info->dev_root->commit_root);
1857 free_extent_buffer(info->extent_root->node);
1858 free_extent_buffer(info->extent_root->commit_root);
1859 free_extent_buffer(info->csum_root->node);
1860 free_extent_buffer(info->csum_root->commit_root);
1861
1862 info->tree_root->node = NULL;
1863 info->tree_root->commit_root = NULL;
1864 info->dev_root->node = NULL;
1865 info->dev_root->commit_root = NULL;
1866 info->extent_root->node = NULL;
1867 info->extent_root->commit_root = NULL;
1868 info->csum_root->node = NULL;
1869 info->csum_root->commit_root = NULL;
1870
1871 if (chunk_root) {
1872 free_extent_buffer(info->chunk_root->node);
1873 free_extent_buffer(info->chunk_root->commit_root);
1874 info->chunk_root->node = NULL;
1875 info->chunk_root->commit_root = NULL;
1876 }
1877}
1878
1879
Chris Mason8a4b83c2008-03-24 15:02:07 -04001880struct btrfs_root *open_ctree(struct super_block *sb,
Chris Masondfe25022008-05-13 13:46:40 -04001881 struct btrfs_fs_devices *fs_devices,
1882 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05001883{
Chris Masondb945352007-10-15 16:15:53 -04001884 u32 sectorsize;
1885 u32 nodesize;
1886 u32 leafsize;
1887 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001888 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04001889 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001890 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05001891 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04001892 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001893 struct btrfs_super_block *disk_super;
Josef Bacik450ba0e2010-11-19 14:59:15 -05001894 struct btrfs_root *tree_root = btrfs_sb(sb);
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001895 struct btrfs_fs_info *fs_info = tree_root->fs_info;
1896 struct btrfs_root *extent_root;
1897 struct btrfs_root *csum_root;
1898 struct btrfs_root *chunk_root;
1899 struct btrfs_root *dev_root;
Chris Masone02119d2008-09-05 16:13:11 -04001900 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05001901 int ret;
Yane58ca022008-04-01 11:21:34 -04001902 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04001903 int num_backups_tried = 0;
1904 int backup_index = 0;
Chris Mason4543df72008-06-11 21:47:56 -04001905
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001906 extent_root = fs_info->extent_root =
1907 kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
1908 csum_root = fs_info->csum_root =
1909 kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
1910 chunk_root = fs_info->chunk_root =
1911 kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
1912 dev_root = fs_info->dev_root =
1913 kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
Chris Mason8790d502008-04-03 16:29:03 -04001914
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001915 if (!extent_root || !csum_root || !chunk_root || !dev_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04001916 err = -ENOMEM;
1917 goto fail;
1918 }
Yan, Zheng76dda932009-09-21 16:00:26 -04001919
1920 ret = init_srcu_struct(&fs_info->subvol_srcu);
1921 if (ret) {
1922 err = ret;
1923 goto fail;
1924 }
1925
1926 ret = setup_bdi(fs_info, &fs_info->bdi);
1927 if (ret) {
1928 err = ret;
1929 goto fail_srcu;
1930 }
1931
1932 fs_info->btree_inode = new_inode(sb);
1933 if (!fs_info->btree_inode) {
1934 err = -ENOMEM;
1935 goto fail_bdi;
1936 }
1937
Chris Masona6591712011-07-19 12:04:14 -04001938 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08001939
Yan, Zheng76dda932009-09-21 16:00:26 -04001940 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04001941 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04001942 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001943 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Chris Mason19c00dd2007-10-15 16:19:22 -04001944 INIT_LIST_HEAD(&fs_info->hashers);
Chris Masonea8c2812008-08-04 23:17:27 -04001945 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
Chris Mason5a3f23d2009-03-31 13:27:11 -04001946 INIT_LIST_HEAD(&fs_info->ordered_operations);
Yan Zheng11833d62009-09-11 16:11:19 -04001947 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Chris Mason1832a6d2007-12-21 16:27:21 -05001948 spin_lock_init(&fs_info->delalloc_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04001949 spin_lock_init(&fs_info->trans_lock);
Yan Zheng31153d82008-07-28 15:32:19 -04001950 spin_lock_init(&fs_info->ref_cache_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04001951 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001952 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04001953 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04001954 spin_lock_init(&fs_info->free_chunk_lock);
Chris Mason75857172011-06-13 20:00:16 -04001955 mutex_init(&fs_info->reloc_mutex);
Chris Mason19c00dd2007-10-15 16:19:22 -04001956
Josef Bacik58176a92007-08-29 15:47:34 -04001957 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04001958 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04001959 INIT_LIST_HEAD(&fs_info->space_info);
Chris Mason0b86a832008-03-24 15:01:56 -04001960 btrfs_mapping_init(&fs_info->mapping_tree);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001961 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1962 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1963 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1964 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1965 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
Josef Bacik6d668dd2011-11-03 22:54:25 -04001966 btrfs_init_block_rsv(&fs_info->delayed_block_rsv);
Chris Masoncb03c742008-05-15 16:15:45 -04001967 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05001968 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04001969 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04001970 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04001971 atomic_set(&fs_info->defrag_running, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04001972 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05001973 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04001974 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04001975 fs_info->defrag_inodes = RB_ROOT;
Josef Bacika4abeea2011-04-11 17:25:13 -04001976 fs_info->trans_no_join = 0;
Josef Bacik2bf64752011-09-26 17:12:22 -04001977 fs_info->free_chunk_space = 0;
Chris Masonc8b97812008-10-29 14:49:59 -04001978
Arne Jansen90519d62011-05-23 14:30:00 +02001979 /* readahead state */
1980 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
1981 spin_lock_init(&fs_info->reada_lock);
1982
Chris Masonb34b0862008-12-19 15:43:22 -05001983 fs_info->thread_pool_size = min_t(unsigned long,
1984 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04001985
Chris Mason3eaa2882008-07-24 11:57:52 -04001986 INIT_LIST_HEAD(&fs_info->ordered_extents);
1987 spin_lock_init(&fs_info->ordered_extent_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08001988 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
1989 GFP_NOFS);
1990 if (!fs_info->delayed_root) {
1991 err = -ENOMEM;
1992 goto fail_iput;
1993 }
1994 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04001995
Arne Jansena2de7332011-03-08 14:14:00 +01001996 mutex_init(&fs_info->scrub_lock);
1997 atomic_set(&fs_info->scrubs_running, 0);
1998 atomic_set(&fs_info->scrub_pause_req, 0);
1999 atomic_set(&fs_info->scrubs_paused, 0);
2000 atomic_set(&fs_info->scrub_cancel_req, 0);
2001 init_waitqueue_head(&fs_info->scrub_pause_wait);
2002 init_rwsem(&fs_info->scrub_super_lock);
2003 fs_info->scrub_workers_refcnt = 0;
Arne Jansena2de7332011-03-08 14:14:00 +01002004
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002005 spin_lock_init(&fs_info->balance_lock);
2006 mutex_init(&fs_info->balance_mutex);
2007 fs_info->balance_ctl = NULL;
2008
Chris Masona061fc82008-05-07 11:43:44 -04002009 sb->s_blocksize = 4096;
2010 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002011 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04002012
Yan, Zheng76dda932009-09-21 16:00:26 -04002013 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2014 fs_info->btree_inode->i_nlink = 1;
Chris Mason0afbaf82008-04-18 14:17:20 -04002015 /*
2016 * we set the i_size on the btree inode to the max possible int.
2017 * the real end of the address space is determined by all of
2018 * the devices in the system
2019 */
2020 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002021 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002022 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2023
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002024 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002025 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002026 fs_info->btree_inode->i_mapping);
David Sterbaa8067e02011-04-21 00:34:43 +02002027 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002028
Chris Masond1310b22008-01-24 16:13:08 -05002029 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2030
Chris Mason0f7d52f2007-04-09 10:42:37 -04002031 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2032 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2033 sizeof(struct btrfs_key));
Yan, Zheng76dda932009-09-21 16:00:26 -04002034 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
Chris Mason22b0ebd2007-03-30 08:47:31 -04002035 insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002036
Chris Masone02119d2008-09-05 16:13:11 -04002037 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002038 fs_info->block_group_cache_tree = RB_ROOT;
Chris Mason925baed2008-06-25 16:01:30 -04002039
Yan Zheng11833d62009-09-11 16:11:19 -04002040 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002041 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002042 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002043 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002044 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002045 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002046
Chris Masond98237b2007-03-28 13:57:48 -04002047
Chris Mason5a3f23d2009-03-31 13:27:11 -04002048 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002049 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002050 mutex_init(&fs_info->chunk_mutex);
2051 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002052 mutex_init(&fs_info->cleaner_mutex);
2053 mutex_init(&fs_info->volume_mutex);
Yan Zheng276e6802009-07-30 09:40:40 -04002054 init_rwsem(&fs_info->extent_commit_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002055 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002056 init_rwsem(&fs_info->subvol_sem);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002057
2058 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2059 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2060
Chris Mason7d9eb122008-07-08 14:19:17 -04002061 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002062 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002063 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002064 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002065
Chris Mason0b86a832008-03-24 15:01:56 -04002066 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002067 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002068
Yan Zhenga512bbf2008-12-08 16:46:26 -05002069 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002070 if (!bh) {
2071 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002072 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002073 }
Chris Mason39279cc2007-06-12 06:35:45 -04002074
David Sterba6c417612011-04-13 15:41:04 +02002075 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2076 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2077 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002078 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002079
David Sterba6c417612011-04-13 15:41:04 +02002080 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002081
David Sterba6c417612011-04-13 15:41:04 +02002082 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002083 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002084 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002085
liuboacce9522011-01-06 19:30:25 +08002086 /* check FS state, whether FS is broken. */
2087 fs_info->fs_state |= btrfs_super_flags(disk_super);
2088
2089 btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2090
Liu Bo75e7cb72011-03-22 10:12:20 +00002091 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002092 * run through our array of backup supers and setup
2093 * our ring pointer to the oldest one
2094 */
2095 generation = btrfs_super_generation(disk_super);
2096 find_oldest_super_backup(fs_info, generation);
2097
2098 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002099 * In the long term, we'll store the compression type in the super
2100 * block, and it'll be used for per file compression control.
2101 */
2102 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2103
Yan Zheng2b820322008-11-17 21:11:30 -05002104 ret = btrfs_parse_options(tree_root, options);
2105 if (ret) {
2106 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002107 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002108 }
Chris Masondfe25022008-05-13 13:46:40 -04002109
Josef Bacikf2b636e2008-12-02 06:36:08 -05002110 features = btrfs_super_incompat_flags(disk_super) &
2111 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2112 if (features) {
2113 printk(KERN_ERR "BTRFS: couldn't mount because of "
2114 "unsupported optional features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002115 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002116 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002117 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002118 }
2119
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002120 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002121 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
2122 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
2123 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
2124 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002125
Josef Bacikf2b636e2008-12-02 06:36:08 -05002126 features = btrfs_super_compat_ro_flags(disk_super) &
2127 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2128 if (!(sb->s_flags & MS_RDONLY) && features) {
2129 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2130 "unsupported option features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002131 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002132 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002133 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002134 }
Chris Mason61d92c32009-10-02 19:11:56 -04002135
2136 btrfs_init_workers(&fs_info->generic_worker,
2137 "genwork", 1, NULL);
2138
Chris Mason5443be42008-08-15 15:34:16 -04002139 btrfs_init_workers(&fs_info->workers, "worker",
Chris Mason61d92c32009-10-02 19:11:56 -04002140 fs_info->thread_pool_size,
2141 &fs_info->generic_worker);
Chris Masonc8b97812008-10-29 14:49:59 -04002142
Chris Mason771ed682008-11-06 22:02:51 -05002143 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
Chris Mason61d92c32009-10-02 19:11:56 -04002144 fs_info->thread_pool_size,
2145 &fs_info->generic_worker);
Chris Mason771ed682008-11-06 22:02:51 -05002146
Chris Mason5443be42008-08-15 15:34:16 -04002147 btrfs_init_workers(&fs_info->submit_workers, "submit",
Chris Masonb720d202008-08-15 15:34:14 -04002148 min_t(u64, fs_devices->num_devices,
Chris Mason61d92c32009-10-02 19:11:56 -04002149 fs_info->thread_pool_size),
2150 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002151
Josef Bacikbab39bf2011-06-30 14:42:28 -04002152 btrfs_init_workers(&fs_info->caching_workers, "cache",
2153 2, &fs_info->generic_worker);
2154
Chris Mason61b49442008-07-31 15:42:53 -04002155 /* a higher idle thresh on the submit workers makes it much more
2156 * likely that bios will be send down in a sane order to the
2157 * devices
2158 */
2159 fs_info->submit_workers.idle_thresh = 64;
Chris Mason53863232008-08-15 15:34:18 -04002160
Chris Mason771ed682008-11-06 22:02:51 -05002161 fs_info->workers.idle_thresh = 16;
Chris Mason4a69a412008-11-06 22:03:00 -05002162 fs_info->workers.ordered = 1;
Chris Mason61b49442008-07-31 15:42:53 -04002163
Chris Mason771ed682008-11-06 22:02:51 -05002164 fs_info->delalloc_workers.idle_thresh = 2;
2165 fs_info->delalloc_workers.ordered = 1;
2166
Chris Mason61d92c32009-10-02 19:11:56 -04002167 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2168 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002169 btrfs_init_workers(&fs_info->endio_workers, "endio",
Chris Mason61d92c32009-10-02 19:11:56 -04002170 fs_info->thread_pool_size,
2171 &fs_info->generic_worker);
Chris Masond20f7042008-12-08 16:58:54 -05002172 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
Chris Mason61d92c32009-10-02 19:11:56 -04002173 fs_info->thread_pool_size,
2174 &fs_info->generic_worker);
Chris Masoncad321a2008-12-17 14:51:42 -05002175 btrfs_init_workers(&fs_info->endio_meta_write_workers,
Chris Mason61d92c32009-10-02 19:11:56 -04002176 "endio-meta-write", fs_info->thread_pool_size,
2177 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002178 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
Chris Mason61d92c32009-10-02 19:11:56 -04002179 fs_info->thread_pool_size,
2180 &fs_info->generic_worker);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002181 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2182 1, &fs_info->generic_worker);
Miao Xie16cdcec2011-04-22 18:12:22 +08002183 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2184 fs_info->thread_pool_size,
2185 &fs_info->generic_worker);
Arne Jansen90519d62011-05-23 14:30:00 +02002186 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2187 fs_info->thread_pool_size,
2188 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002189
2190 /*
2191 * endios are largely parallel and should have a very
2192 * low idle thresh
2193 */
2194 fs_info->endio_workers.idle_thresh = 4;
Chris Masonb51912c2009-02-04 09:23:24 -05002195 fs_info->endio_meta_workers.idle_thresh = 4;
2196
Chris Mason90428462009-08-04 16:56:34 -04002197 fs_info->endio_write_workers.idle_thresh = 2;
2198 fs_info->endio_meta_write_workers.idle_thresh = 2;
Arne Jansen90519d62011-05-23 14:30:00 +02002199 fs_info->readahead_workers.idle_thresh = 2;
Chris Mason90428462009-08-04 16:56:34 -04002200
Josef Bacik0dc3b842011-11-18 14:37:27 -05002201 /*
2202 * btrfs_start_workers can really only fail because of ENOMEM so just
2203 * return -ENOMEM if any of these fail.
2204 */
2205 ret = btrfs_start_workers(&fs_info->workers);
2206 ret |= btrfs_start_workers(&fs_info->generic_worker);
2207 ret |= btrfs_start_workers(&fs_info->submit_workers);
2208 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2209 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2210 ret |= btrfs_start_workers(&fs_info->endio_workers);
2211 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
2212 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2213 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2214 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2215 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2216 ret |= btrfs_start_workers(&fs_info->caching_workers);
2217 ret |= btrfs_start_workers(&fs_info->readahead_workers);
2218 if (ret) {
2219 ret = -ENOMEM;
2220 goto fail_sb_buffer;
2221 }
Chris Mason4543df72008-06-11 21:47:56 -04002222
Chris Mason4575c9c2008-04-18 16:13:31 -04002223 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002224 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2225 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002226
Chris Masondb945352007-10-15 16:15:53 -04002227 nodesize = btrfs_super_nodesize(disk_super);
2228 leafsize = btrfs_super_leafsize(disk_super);
2229 sectorsize = btrfs_super_sectorsize(disk_super);
Chris Mason87ee04e2007-11-30 11:30:34 -05002230 stripesize = btrfs_super_stripesize(disk_super);
Chris Masondb945352007-10-15 16:15:53 -04002231 tree_root->nodesize = nodesize;
2232 tree_root->leafsize = leafsize;
2233 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002234 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002235
2236 sb->s_blocksize = sectorsize;
2237 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002238
Chris Mason39279cc2007-06-12 06:35:45 -04002239 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
2240 sizeof(disk_super->magic))) {
Chris Masond3977122009-01-05 21:25:51 -05002241 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002242 goto fail_sb_buffer;
2243 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002244
Chris Mason925baed2008-06-25 16:01:30 -04002245 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002246 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002247 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002248 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002249 printk(KERN_WARNING "btrfs: failed to read the system "
2250 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002251 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002252 }
Chris Mason0b86a832008-03-24 15:01:56 -04002253
2254 blocksize = btrfs_level_size(tree_root,
2255 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002256 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002257
2258 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2259 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2260
2261 chunk_root->node = read_tree_block(chunk_root,
2262 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002263 blocksize, generation);
Chris Mason0b86a832008-03-24 15:01:56 -04002264 BUG_ON(!chunk_root->node);
David Woodhouse83121942009-07-22 16:52:13 -04002265 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2266 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2267 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002268 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002269 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002270 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2271 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002272
Chris Masone17cade2008-04-15 15:41:47 -04002273 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Chris Masond3977122009-01-05 21:25:51 -05002274 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2275 BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002276
Chris Mason925baed2008-06-25 16:01:30 -04002277 mutex_lock(&fs_info->chunk_mutex);
Chris Mason0b86a832008-03-24 15:01:56 -04002278 ret = btrfs_read_chunk_tree(chunk_root);
Chris Mason925baed2008-06-25 16:01:30 -04002279 mutex_unlock(&fs_info->chunk_mutex);
Yan Zheng2b820322008-11-17 21:11:30 -05002280 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002281 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2282 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002283 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002284 }
Chris Mason0b86a832008-03-24 15:01:56 -04002285
Chris Masondfe25022008-05-13 13:46:40 -04002286 btrfs_close_extra_devices(fs_devices);
2287
Chris Masonaf31f5e2011-11-03 15:17:42 -04002288retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002289 blocksize = btrfs_level_size(tree_root,
2290 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002291 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002292
Chris Masone20d96d2007-03-22 12:13:20 -04002293 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002294 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002295 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002296 if (!tree_root->node ||
2297 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
David Woodhouse83121942009-07-22 16:52:13 -04002298 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2299 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002300
2301 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002302 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002303
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002304 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2305 tree_root->commit_root = btrfs_root_node(tree_root);
Chris Masondb945352007-10-15 16:15:53 -04002306
2307 ret = find_and_setup_root(tree_root, fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04002308 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002309 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002310 goto recovery_tree_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002311 extent_root->track_dirty = 1;
2312
2313 ret = find_and_setup_root(tree_root, fs_info,
2314 BTRFS_DEV_TREE_OBJECTID, dev_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002315 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002316 goto recovery_tree_root;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002317 dev_root->track_dirty = 1;
Chris Mason3768f362007-03-13 16:47:54 -04002318
Chris Masond20f7042008-12-08 16:58:54 -05002319 ret = find_and_setup_root(tree_root, fs_info,
2320 BTRFS_CSUM_TREE_OBJECTID, csum_root);
2321 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002322 goto recovery_tree_root;
Chris Masond20f7042008-12-08 16:58:54 -05002323
2324 csum_root->track_dirty = 1;
2325
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002326 fs_info->generation = generation;
2327 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002328
liuboc59021f2011-03-07 02:13:14 +00002329 ret = btrfs_init_space_info(fs_info);
2330 if (ret) {
2331 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2332 goto fail_block_groups;
2333 }
2334
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002335 ret = btrfs_read_block_groups(extent_root);
2336 if (ret) {
2337 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2338 goto fail_block_groups;
2339 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002340
Chris Masona74a4b92008-06-25 16:01:31 -04002341 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2342 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002343 if (IS_ERR(fs_info->cleaner_kthread))
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002344 goto fail_block_groups;
Chris Masona74a4b92008-06-25 16:01:31 -04002345
2346 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2347 tree_root,
2348 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002349 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002350 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002351
Chris Masonc2898112009-06-10 09:51:32 -04002352 if (!btrfs_test_opt(tree_root, SSD) &&
2353 !btrfs_test_opt(tree_root, NOSSD) &&
2354 !fs_info->fs_devices->rotating) {
2355 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2356 "mode\n");
2357 btrfs_set_opt(fs_info->mount_opt, SSD);
2358 }
2359
liuboacce9522011-01-06 19:30:25 +08002360 /* do not make disk changes in broken FS */
2361 if (btrfs_super_log_root(disk_super) != 0 &&
2362 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
Chris Masone02119d2008-09-05 16:13:11 -04002363 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002364
Chris Mason7c2ca462008-11-19 15:13:35 -05002365 if (fs_devices->rw_devices == 0) {
Chris Masond3977122009-01-05 21:25:51 -05002366 printk(KERN_WARNING "Btrfs log replay required "
2367 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002368 err = -EIO;
2369 goto fail_trans_kthread;
2370 }
Chris Masone02119d2008-09-05 16:13:11 -04002371 blocksize =
2372 btrfs_level_size(tree_root,
2373 btrfs_super_log_root_level(disk_super));
2374
Dan Carpenter676e4c82010-05-29 09:43:07 +00002375 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
2376 if (!log_tree_root) {
2377 err = -ENOMEM;
2378 goto fail_trans_kthread;
2379 }
Chris Masone02119d2008-09-05 16:13:11 -04002380
2381 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2382 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2383
2384 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002385 blocksize,
2386 generation + 1);
Chris Masone02119d2008-09-05 16:13:11 -04002387 ret = btrfs_recover_log_trees(log_tree_root);
2388 BUG_ON(ret);
Yan Zhenge556ce22008-11-20 10:25:19 -05002389
2390 if (sb->s_flags & MS_RDONLY) {
2391 ret = btrfs_commit_super(tree_root);
2392 BUG_ON(ret);
2393 }
Chris Masone02119d2008-09-05 16:13:11 -04002394 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002395
Yan, Zheng76dda932009-09-21 16:00:26 -04002396 ret = btrfs_find_orphan_roots(tree_root);
2397 BUG_ON(ret);
2398
Chris Mason7c2ca462008-11-19 15:13:35 -05002399 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002400 ret = btrfs_cleanup_fs_roots(fs_info);
2401 BUG_ON(ret);
2402
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002403 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002404 if (ret < 0) {
2405 printk(KERN_WARNING
2406 "btrfs: failed to recover relocation\n");
2407 err = -EINVAL;
2408 goto fail_trans_kthread;
2409 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002410 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002411
Chris Mason3de45862008-11-17 21:02:50 -05002412 location.objectid = BTRFS_FS_TREE_OBJECTID;
2413 location.type = BTRFS_ROOT_ITEM_KEY;
2414 location.offset = (u64)-1;
2415
Chris Mason3de45862008-11-17 21:02:50 -05002416 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2417 if (!fs_info->fs_root)
Chris Mason7c2ca462008-11-19 15:13:35 -05002418 goto fail_trans_kthread;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002419 if (IS_ERR(fs_info->fs_root)) {
2420 err = PTR_ERR(fs_info->fs_root);
2421 goto fail_trans_kthread;
2422 }
Chris Masonc2898112009-06-10 09:51:32 -04002423
Josef Bacike3acc2a2010-01-26 14:30:53 +00002424 if (!(sb->s_flags & MS_RDONLY)) {
2425 down_read(&fs_info->cleanup_work_sem);
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002426 err = btrfs_orphan_cleanup(fs_info->fs_root);
2427 if (!err)
2428 err = btrfs_orphan_cleanup(fs_info->tree_root);
Josef Bacike3acc2a2010-01-26 14:30:53 +00002429 up_read(&fs_info->cleanup_work_sem);
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002430 if (err) {
2431 close_ctree(tree_root);
2432 return ERR_PTR(err);
2433 }
Josef Bacike3acc2a2010-01-26 14:30:53 +00002434 }
2435
Chris Mason0f7d52f2007-04-09 10:42:37 -04002436 return tree_root;
Chris Mason39279cc2007-06-12 06:35:45 -04002437
Chris Mason7c2ca462008-11-19 15:13:35 -05002438fail_trans_kthread:
2439 kthread_stop(fs_info->transaction_kthread);
Chris Mason3f157a22008-06-25 16:01:31 -04002440fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002441 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002442
2443 /*
2444 * make sure we're done with the btree inode before we stop our
2445 * kthreads
2446 */
2447 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2448 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2449
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002450fail_block_groups:
2451 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002452
2453fail_tree_roots:
2454 free_root_pointers(fs_info, 1);
2455
Chris Mason39279cc2007-06-12 06:35:45 -04002456fail_sb_buffer:
Chris Mason61d92c32009-10-02 19:11:56 -04002457 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason306c8b62011-11-03 15:21:39 -04002458 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Mason247e7432008-07-17 12:53:51 -04002459 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002460 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002461 btrfs_stop_workers(&fs_info->workers);
2462 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002463 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002464 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002465 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002466 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002467 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002468 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04002469 btrfs_stop_workers(&fs_info->caching_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002470fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04002471fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002472 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2473
Chris Mason7c2ca462008-11-19 15:13:35 -05002474 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Mason4543df72008-06-11 21:47:56 -04002475 iput(fs_info->btree_inode);
Jens Axboead081f12009-06-12 14:43:40 +02002476fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05002477 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002478fail_srcu:
2479 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002480fail:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002481 btrfs_close_devices(fs_info->fs_devices);
David Sterba6c417612011-04-13 15:41:04 +02002482 free_fs_info(fs_info);
Chris Mason39279cc2007-06-12 06:35:45 -04002483 return ERR_PTR(err);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002484
2485recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04002486 if (!btrfs_test_opt(tree_root, RECOVERY))
2487 goto fail_tree_roots;
2488
2489 free_root_pointers(fs_info, 0);
2490
2491 /* don't use the log in recovery mode, it won't be valid */
2492 btrfs_set_super_log_root(disk_super, 0);
2493
2494 /* we can't trust the free space cache either */
2495 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2496
2497 ret = next_root_backup(fs_info, fs_info->super_copy,
2498 &num_backups_tried, &backup_index);
2499 if (ret == -1)
2500 goto fail_block_groups;
2501 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05002502}
2503
Chris Masonf2984462008-04-10 16:19:33 -04002504static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2505{
2506 char b[BDEVNAME_SIZE];
2507
2508 if (uptodate) {
2509 set_buffer_uptodate(bh);
2510 } else {
David Sterba7a36dde2011-05-06 15:33:15 +02002511 printk_ratelimited(KERN_WARNING "lost page write due to "
Chris Masonf2984462008-04-10 16:19:33 -04002512 "I/O error on %s\n",
2513 bdevname(bh->b_bdev, b));
Chris Mason1259ab72008-05-12 13:39:03 -04002514 /* note, we dont' set_buffer_write_io_error because we have
2515 * our own ways of dealing with the IO errors
2516 */
Chris Masonf2984462008-04-10 16:19:33 -04002517 clear_buffer_uptodate(bh);
2518 }
2519 unlock_buffer(bh);
2520 put_bh(bh);
2521}
2522
Yan Zhenga512bbf2008-12-08 16:46:26 -05002523struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2524{
2525 struct buffer_head *bh;
2526 struct buffer_head *latest = NULL;
2527 struct btrfs_super_block *super;
2528 int i;
2529 u64 transid = 0;
2530 u64 bytenr;
2531
2532 /* we would like to check all the supers, but that would make
2533 * a btrfs mount succeed after a mkfs from a different FS.
2534 * So, we need to add a special mount option to scan for
2535 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2536 */
2537 for (i = 0; i < 1; i++) {
2538 bytenr = btrfs_sb_offset(i);
2539 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2540 break;
2541 bh = __bread(bdev, bytenr / 4096, 4096);
2542 if (!bh)
2543 continue;
2544
2545 super = (struct btrfs_super_block *)bh->b_data;
2546 if (btrfs_super_bytenr(super) != bytenr ||
2547 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2548 sizeof(super->magic))) {
2549 brelse(bh);
2550 continue;
2551 }
2552
2553 if (!latest || btrfs_super_generation(super) > transid) {
2554 brelse(latest);
2555 latest = bh;
2556 transid = btrfs_super_generation(super);
2557 } else {
2558 brelse(bh);
2559 }
2560 }
2561 return latest;
2562}
2563
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002564/*
2565 * this should be called twice, once with wait == 0 and
2566 * once with wait == 1. When wait == 0 is done, all the buffer heads
2567 * we write are pinned.
2568 *
2569 * They are released when wait == 1 is done.
2570 * max_mirrors must be the same for both runs, and it indicates how
2571 * many supers on this one device should be written.
2572 *
2573 * max_mirrors == 0 means to write them all.
2574 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002575static int write_dev_supers(struct btrfs_device *device,
2576 struct btrfs_super_block *sb,
2577 int do_barriers, int wait, int max_mirrors)
2578{
2579 struct buffer_head *bh;
2580 int i;
2581 int ret;
2582 int errors = 0;
2583 u32 crc;
2584 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002585
2586 if (max_mirrors == 0)
2587 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2588
Yan Zhenga512bbf2008-12-08 16:46:26 -05002589 for (i = 0; i < max_mirrors; i++) {
2590 bytenr = btrfs_sb_offset(i);
2591 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2592 break;
2593
2594 if (wait) {
2595 bh = __find_get_block(device->bdev, bytenr / 4096,
2596 BTRFS_SUPER_INFO_SIZE);
2597 BUG_ON(!bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002598 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002599 if (!buffer_uptodate(bh))
2600 errors++;
2601
2602 /* drop our reference */
2603 brelse(bh);
2604
2605 /* drop the reference from the wait == 0 run */
2606 brelse(bh);
2607 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002608 } else {
2609 btrfs_set_super_bytenr(sb, bytenr);
2610
2611 crc = ~(u32)0;
2612 crc = btrfs_csum_data(NULL, (char *)sb +
2613 BTRFS_CSUM_SIZE, crc,
2614 BTRFS_SUPER_INFO_SIZE -
2615 BTRFS_CSUM_SIZE);
2616 btrfs_csum_final(crc, sb->csum);
2617
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002618 /*
2619 * one reference for us, and we leave it for the
2620 * caller
2621 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002622 bh = __getblk(device->bdev, bytenr / 4096,
2623 BTRFS_SUPER_INFO_SIZE);
2624 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2625
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002626 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002627 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002628
2629 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002630 lock_buffer(bh);
2631 bh->b_end_io = btrfs_end_buffer_write_sync;
2632 }
2633
Chris Mason387125f2011-11-18 15:07:51 -05002634 /*
2635 * we fua the first super. The others we allow
2636 * to go down lazy.
2637 */
2638 ret = submit_bh(WRITE_FUA, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002639 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05002640 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002641 }
2642 return errors < i ? 0 : -1;
2643}
2644
Chris Mason387125f2011-11-18 15:07:51 -05002645/*
2646 * endio for the write_dev_flush, this will wake anyone waiting
2647 * for the barrier when it is done
2648 */
2649static void btrfs_end_empty_barrier(struct bio *bio, int err)
2650{
2651 if (err) {
2652 if (err == -EOPNOTSUPP)
2653 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
2654 clear_bit(BIO_UPTODATE, &bio->bi_flags);
2655 }
2656 if (bio->bi_private)
2657 complete(bio->bi_private);
2658 bio_put(bio);
2659}
2660
2661/*
2662 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
2663 * sent down. With wait == 1, it waits for the previous flush.
2664 *
2665 * any device where the flush fails with eopnotsupp are flagged as not-barrier
2666 * capable
2667 */
2668static int write_dev_flush(struct btrfs_device *device, int wait)
2669{
2670 struct bio *bio;
2671 int ret = 0;
2672
2673 if (device->nobarriers)
2674 return 0;
2675
2676 if (wait) {
2677 bio = device->flush_bio;
2678 if (!bio)
2679 return 0;
2680
2681 wait_for_completion(&device->flush_wait);
2682
2683 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
2684 printk("btrfs: disabling barriers on dev %s\n",
2685 device->name);
2686 device->nobarriers = 1;
2687 }
2688 if (!bio_flagged(bio, BIO_UPTODATE)) {
2689 ret = -EIO;
2690 }
2691
2692 /* drop the reference from the wait == 0 run */
2693 bio_put(bio);
2694 device->flush_bio = NULL;
2695
2696 return ret;
2697 }
2698
2699 /*
2700 * one reference for us, and we leave it for the
2701 * caller
2702 */
2703 device->flush_bio = NULL;;
2704 bio = bio_alloc(GFP_NOFS, 0);
2705 if (!bio)
2706 return -ENOMEM;
2707
2708 bio->bi_end_io = btrfs_end_empty_barrier;
2709 bio->bi_bdev = device->bdev;
2710 init_completion(&device->flush_wait);
2711 bio->bi_private = &device->flush_wait;
2712 device->flush_bio = bio;
2713
2714 bio_get(bio);
2715 submit_bio(WRITE_FLUSH, bio);
2716
2717 return 0;
2718}
2719
2720/*
2721 * send an empty flush down to each device in parallel,
2722 * then wait for them
2723 */
2724static int barrier_all_devices(struct btrfs_fs_info *info)
2725{
2726 struct list_head *head;
2727 struct btrfs_device *dev;
2728 int errors = 0;
2729 int ret;
2730
2731 /* send down all the barriers */
2732 head = &info->fs_devices->devices;
2733 list_for_each_entry_rcu(dev, head, dev_list) {
2734 if (!dev->bdev) {
2735 errors++;
2736 continue;
2737 }
2738 if (!dev->in_fs_metadata || !dev->writeable)
2739 continue;
2740
2741 ret = write_dev_flush(dev, 0);
2742 if (ret)
2743 errors++;
2744 }
2745
2746 /* wait for all the barriers */
2747 list_for_each_entry_rcu(dev, head, dev_list) {
2748 if (!dev->bdev) {
2749 errors++;
2750 continue;
2751 }
2752 if (!dev->in_fs_metadata || !dev->writeable)
2753 continue;
2754
2755 ret = write_dev_flush(dev, 1);
2756 if (ret)
2757 errors++;
2758 }
2759 if (errors)
2760 return -EIO;
2761 return 0;
2762}
2763
Yan Zhenga512bbf2008-12-08 16:46:26 -05002764int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04002765{
Chris Masone5e9a522009-06-10 15:17:02 -04002766 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04002767 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04002768 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04002769 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04002770 int ret;
2771 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04002772 int max_errors;
2773 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04002774 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04002775
David Sterba6c417612011-04-13 15:41:04 +02002776 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Masonf2984462008-04-10 16:19:33 -04002777 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002778 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04002779
David Sterba6c417612011-04-13 15:41:04 +02002780 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04002781 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04002782
Chris Mason174ba502011-05-27 10:03:58 -04002783 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04002784 head = &root->fs_info->fs_devices->devices;
Chris Mason387125f2011-11-18 15:07:51 -05002785
2786 if (do_barriers)
2787 barrier_all_devices(root->fs_info);
2788
Xiao Guangrong1f781602011-04-20 10:09:16 +00002789 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002790 if (!dev->bdev) {
2791 total_errors++;
2792 continue;
2793 }
Yan Zheng2b820322008-11-17 21:11:30 -05002794 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002795 continue;
2796
Yan Zheng2b820322008-11-17 21:11:30 -05002797 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04002798 btrfs_set_stack_device_type(dev_item, dev->type);
2799 btrfs_set_stack_device_id(dev_item, dev->devid);
2800 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2801 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2802 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2803 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2804 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2805 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05002806 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002807
Chris Masona061fc82008-05-07 11:43:44 -04002808 flags = btrfs_super_flags(sb);
2809 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04002810
Yan Zhenga512bbf2008-12-08 16:46:26 -05002811 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04002812 if (ret)
2813 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04002814 }
Chris Masona236aed2008-04-29 09:38:00 -04002815 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05002816 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2817 total_errors);
Chris Masona236aed2008-04-29 09:38:00 -04002818 BUG();
2819 }
Chris Masonf2984462008-04-10 16:19:33 -04002820
Yan Zhenga512bbf2008-12-08 16:46:26 -05002821 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00002822 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002823 if (!dev->bdev)
2824 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05002825 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002826 continue;
2827
Yan Zhenga512bbf2008-12-08 16:46:26 -05002828 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2829 if (ret)
2830 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04002831 }
Chris Mason174ba502011-05-27 10:03:58 -04002832 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04002833 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05002834 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2835 total_errors);
Chris Masona236aed2008-04-29 09:38:00 -04002836 BUG();
2837 }
Chris Masonf2984462008-04-10 16:19:33 -04002838 return 0;
2839}
2840
Yan Zhenga512bbf2008-12-08 16:46:26 -05002841int write_ctree_super(struct btrfs_trans_handle *trans,
2842 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05002843{
Chris Masone66f7092007-04-20 13:16:02 -04002844 int ret;
Chris Mason2c90e5d2007-04-02 10:50:19 -04002845
Yan Zhenga512bbf2008-12-08 16:46:26 -05002846 ret = write_all_supers(root, max_mirrors);
Chris Mason5f39d392007-10-15 16:14:19 -04002847 return ret;
Chris Masoncfaa7292007-02-21 17:04:57 -05002848}
2849
Chris Mason5eda7b52007-06-22 14:16:25 -04002850int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04002851{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002852 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04002853 radix_tree_delete(&fs_info->fs_roots_radix,
2854 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002855 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002856
2857 if (btrfs_root_refs(&root->root_item) == 0)
2858 synchronize_srcu(&fs_info->subvol_srcu);
2859
Li Zefan581bb052011-04-20 10:06:11 +08002860 __btrfs_remove_free_space_cache(root->free_ino_pinned);
2861 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002862 free_fs_root(root);
2863 return 0;
2864}
2865
2866static void free_fs_root(struct btrfs_root *root)
2867{
Li Zefan82d59022011-04-20 10:33:24 +08002868 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002869 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Al Viro0ee5dc62011-07-07 15:44:25 -04002870 if (root->anon_dev)
2871 free_anon_bdev(root->anon_dev);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002872 free_extent_buffer(root->node);
2873 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08002874 kfree(root->free_ino_ctl);
2875 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05002876 kfree(root->name);
Chris Mason2619ba12007-04-10 16:58:11 -04002877 kfree(root);
Chris Mason2619ba12007-04-10 16:58:11 -04002878}
2879
Chris Mason35b7e472007-05-02 15:53:43 -04002880static int del_fs_roots(struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04002881{
2882 int ret;
2883 struct btrfs_root *gang[8];
2884 int i;
2885
Yan, Zheng76dda932009-09-21 16:00:26 -04002886 while (!list_empty(&fs_info->dead_roots)) {
2887 gang[0] = list_entry(fs_info->dead_roots.next,
2888 struct btrfs_root, root_list);
2889 list_del(&gang[0]->root_list);
2890
2891 if (gang[0]->in_radix) {
2892 btrfs_free_fs_root(fs_info, gang[0]);
2893 } else {
2894 free_extent_buffer(gang[0]->node);
2895 free_extent_buffer(gang[0]->commit_root);
2896 kfree(gang[0]);
2897 }
2898 }
2899
Chris Masond3977122009-01-05 21:25:51 -05002900 while (1) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04002901 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2902 (void **)gang, 0,
2903 ARRAY_SIZE(gang));
2904 if (!ret)
2905 break;
Chris Mason2619ba12007-04-10 16:58:11 -04002906 for (i = 0; i < ret; i++)
Chris Mason5eda7b52007-06-22 14:16:25 -04002907 btrfs_free_fs_root(fs_info, gang[i]);
Chris Mason0f7d52f2007-04-09 10:42:37 -04002908 }
2909 return 0;
2910}
Chris Masonb4100d62007-04-12 12:14:00 -04002911
Yan Zhengc146afa2008-11-12 14:34:12 -05002912int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
2913{
2914 u64 root_objectid = 0;
2915 struct btrfs_root *gang[8];
2916 int i;
2917 int ret;
2918
2919 while (1) {
2920 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2921 (void **)gang, root_objectid,
2922 ARRAY_SIZE(gang));
2923 if (!ret)
2924 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002925
2926 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05002927 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002928 int err;
2929
Yan Zhengc146afa2008-11-12 14:34:12 -05002930 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002931 err = btrfs_orphan_cleanup(gang[i]);
2932 if (err)
2933 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05002934 }
2935 root_objectid++;
2936 }
2937 return 0;
2938}
2939
2940int btrfs_commit_super(struct btrfs_root *root)
2941{
2942 struct btrfs_trans_handle *trans;
2943 int ret;
2944
2945 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002946 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05002947 btrfs_clean_old_snapshots(root);
2948 mutex_unlock(&root->fs_info->cleaner_mutex);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002949
2950 /* wait until ongoing cleanup work done */
2951 down_write(&root->fs_info->cleanup_work_sem);
2952 up_write(&root->fs_info->cleanup_work_sem);
2953
Josef Bacik7a7eaa42011-04-13 12:54:33 -04002954 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00002955 if (IS_ERR(trans))
2956 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05002957 ret = btrfs_commit_transaction(trans, root);
2958 BUG_ON(ret);
2959 /* run commit again to drop the original snapshot */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04002960 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00002961 if (IS_ERR(trans))
2962 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05002963 btrfs_commit_transaction(trans, root);
2964 ret = btrfs_write_and_wait_transaction(NULL, root);
2965 BUG_ON(ret);
2966
Yan Zhenga512bbf2008-12-08 16:46:26 -05002967 ret = write_ctree_super(NULL, root, 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05002968 return ret;
2969}
2970
Chris Masone20d96d2007-03-22 12:13:20 -04002971int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05002972{
Chris Mason0f7d52f2007-04-09 10:42:37 -04002973 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05002974 int ret;
Chris Masone089f052007-03-16 16:20:31 -04002975
Chris Masonfacda1e2007-06-08 18:11:48 -04002976 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04002977 smp_mb();
2978
Arne Jansena2de7332011-03-08 14:14:00 +01002979 btrfs_scrub_cancel(root);
Chris Mason4cb53002011-05-24 15:35:30 -04002980
2981 /* wait for any defraggers to finish */
2982 wait_event(fs_info->transaction_wait,
2983 (atomic_read(&fs_info->defrag_running) == 0));
2984
2985 /* clear out the rbtree of defraggable inodes */
2986 btrfs_run_defrag_inodes(root->fs_info);
2987
liuboacce9522011-01-06 19:30:25 +08002988 /*
2989 * Here come 2 situations when btrfs is broken to flip readonly:
2990 *
2991 * 1. when btrfs flips readonly somewhere else before
2992 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2993 * and btrfs will skip to write sb directly to keep
2994 * ERROR state on disk.
2995 *
2996 * 2. when btrfs flips readonly just in btrfs_commit_super,
Justin P. Mattockae0e47f2011-03-01 15:06:02 +01002997 * and in such case, btrfs cannot write sb via btrfs_commit_super,
liuboacce9522011-01-06 19:30:25 +08002998 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2999 * btrfs will cleanup all FS resources first and write sb then.
3000 */
Yan Zhengc146afa2008-11-12 14:34:12 -05003001 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003002 ret = btrfs_commit_super(root);
3003 if (ret)
3004 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3005 }
3006
3007 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
3008 ret = btrfs_error_commit_super(root);
Chris Masond3977122009-01-05 21:25:51 -05003009 if (ret)
3010 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
Yan Zhengc146afa2008-11-12 14:34:12 -05003011 }
Chris Masoned2ff2c2007-03-01 18:59:40 -05003012
Josef Bacik300e4f82011-08-29 14:06:00 -04003013 btrfs_put_block_group_cache(fs_info);
3014
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003015 kthread_stop(root->fs_info->transaction_kthread);
3016 kthread_stop(root->fs_info->cleaner_kthread);
3017
Yan Zhengf25784b2009-07-28 08:41:57 -04003018 fs_info->closing = 2;
3019 smp_mb();
3020
Chris Masonb0c68f82008-01-31 11:05:37 -05003021 if (fs_info->delalloc_bytes) {
Chris Masond3977122009-01-05 21:25:51 -05003022 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07003023 (unsigned long long)fs_info->delalloc_bytes);
Chris Masonb0c68f82008-01-31 11:05:37 -05003024 }
Yan Zheng31153d82008-07-28 15:32:19 -04003025 if (fs_info->total_ref_cache_size) {
Chris Masond3977122009-01-05 21:25:51 -05003026 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
3027 (unsigned long long)fs_info->total_ref_cache_size);
Yan Zheng31153d82008-07-28 15:32:19 -04003028 }
Yanbcc63ab2008-07-30 16:29:20 -04003029
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003030 free_extent_buffer(fs_info->extent_root->node);
3031 free_extent_buffer(fs_info->extent_root->commit_root);
3032 free_extent_buffer(fs_info->tree_root->node);
3033 free_extent_buffer(fs_info->tree_root->commit_root);
3034 free_extent_buffer(root->fs_info->chunk_root->node);
3035 free_extent_buffer(root->fs_info->chunk_root->commit_root);
3036 free_extent_buffer(root->fs_info->dev_root->node);
3037 free_extent_buffer(root->fs_info->dev_root->commit_root);
3038 free_extent_buffer(root->fs_info->csum_root->node);
3039 free_extent_buffer(root->fs_info->csum_root->commit_root);
Chris Masond20f7042008-12-08 16:58:54 -05003040
Chris Mason9078a3e2007-04-26 16:46:15 -04003041 btrfs_free_block_groups(root->fs_info);
Yan Zhengc146afa2008-11-12 14:34:12 -05003042
Chris Mason0f7d52f2007-04-09 10:42:37 -04003043 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003044
Yan Zhengc146afa2008-11-12 14:34:12 -05003045 iput(fs_info->btree_inode);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003046
Chris Mason61d92c32009-10-02 19:11:56 -04003047 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04003048 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05003049 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04003050 btrfs_stop_workers(&fs_info->workers);
3051 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05003052 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05003053 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04003054 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04003055 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04003056 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08003057 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04003058 btrfs_stop_workers(&fs_info->caching_workers);
Arne Jansen90519d62011-05-23 14:30:00 +02003059 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Masond6bfde82008-04-30 13:59:35 -04003060
Chris Masondfe25022008-05-13 13:46:40 -04003061 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003062 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003063
Chris Mason04160082008-03-26 10:28:07 -04003064 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003065 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003066
David Sterba6c417612011-04-13 15:41:04 +02003067 free_fs_info(fs_info);
Li Zefan83a4d542010-12-27 16:19:53 +08003068
Chris Masoneb60cea2007-02-02 09:18:22 -05003069 return 0;
3070}
3071
Chris Mason1259ab72008-05-12 13:39:03 -04003072int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
Chris Masonccd467d2007-06-28 15:57:36 -04003073{
Chris Mason1259ab72008-05-12 13:39:03 -04003074 int ret;
Chris Mason810191f2007-10-15 16:18:55 -04003075 struct inode *btree_inode = buf->first_page->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003076
Josef Bacik2ac55d42010-02-03 19:33:23 +00003077 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
3078 NULL);
Chris Mason1259ab72008-05-12 13:39:03 -04003079 if (!ret)
3080 return ret;
3081
3082 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
3083 parent_transid);
3084 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003085}
Chris Mason6702ed42007-08-07 16:15:09 -04003086
Chris Mason5f39d392007-10-15 16:14:19 -04003087int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3088{
Chris Mason810191f2007-10-15 16:18:55 -04003089 struct inode *btree_inode = buf->first_page->mapping->host;
Chris Masond1310b22008-01-24 16:13:08 -05003090 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
Chris Mason5f39d392007-10-15 16:14:19 -04003091 buf);
3092}
3093
3094void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3095{
Chris Mason810191f2007-10-15 16:18:55 -04003096 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
Chris Mason5f39d392007-10-15 16:14:19 -04003097 u64 transid = btrfs_header_generation(buf);
3098 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonb9473432009-03-13 11:00:37 -04003099 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003100
Chris Masonb9447ef2009-03-09 11:45:38 -04003101 btrfs_assert_tree_locked(buf);
Chris Masonccd467d2007-06-28 15:57:36 -04003102 if (transid != root->fs_info->generation) {
Chris Masond3977122009-01-05 21:25:51 -05003103 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
3104 "found %llu running %llu\n",
Chris Masondb945352007-10-15 16:15:53 -04003105 (unsigned long long)buf->start,
Chris Masond3977122009-01-05 21:25:51 -05003106 (unsigned long long)transid,
3107 (unsigned long long)root->fs_info->generation);
Chris Masonccd467d2007-06-28 15:57:36 -04003108 WARN_ON(1);
3109 }
Chris Masonb9473432009-03-13 11:00:37 -04003110 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
3111 buf);
3112 if (!was_dirty) {
3113 spin_lock(&root->fs_info->delalloc_lock);
3114 root->fs_info->dirty_metadata_bytes += buf->len;
3115 spin_unlock(&root->fs_info->delalloc_lock);
3116 }
Chris Masoneb60cea2007-02-02 09:18:22 -05003117}
3118
Chris Masond3c2fdcf2007-09-17 10:58:06 -04003119void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
Chris Mason35b7e472007-05-02 15:53:43 -04003120{
Chris Mason188de642008-05-09 11:52:25 -04003121 /*
3122 * looks as though older kernels can get into trouble with
3123 * this code, they end up stuck in balance_dirty_pages forever
3124 */
Chris Masond6bfde82008-04-30 13:59:35 -04003125 u64 num_dirty;
Chris Mason771ed682008-11-06 22:02:51 -05003126 unsigned long thresh = 32 * 1024 * 1024;
Chris Masond6bfde82008-04-30 13:59:35 -04003127
Jens Axboe6933c022009-03-17 09:36:37 +01003128 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003129 return;
3130
Miao Xie16cdcec2011-04-22 18:12:22 +08003131 btrfs_balance_delayed_items(root);
3132
3133 num_dirty = root->fs_info->dirty_metadata_bytes;
3134
3135 if (num_dirty > thresh) {
3136 balance_dirty_pages_ratelimited_nr(
3137 root->fs_info->btree_inode->i_mapping, 1);
3138 }
3139 return;
3140}
3141
3142void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
3143{
3144 /*
3145 * looks as though older kernels can get into trouble with
3146 * this code, they end up stuck in balance_dirty_pages forever
3147 */
3148 u64 num_dirty;
3149 unsigned long thresh = 32 * 1024 * 1024;
3150
3151 if (current->flags & PF_MEMALLOC)
3152 return;
3153
Chris Mason585ad2c2009-05-18 10:41:58 -04003154 num_dirty = root->fs_info->dirty_metadata_bytes;
3155
Chris Masond6bfde82008-04-30 13:59:35 -04003156 if (num_dirty > thresh) {
3157 balance_dirty_pages_ratelimited_nr(
Chris Masond7fc6402008-02-18 12:12:38 -05003158 root->fs_info->btree_inode->i_mapping, 1);
Chris Masond6bfde82008-04-30 13:59:35 -04003159 }
Chris Mason188de642008-05-09 11:52:25 -04003160 return;
Chris Mason35b7e472007-05-02 15:53:43 -04003161}
Chris Mason6b800532007-10-15 16:17:34 -04003162
Chris Masonca7a79a2008-05-12 12:59:19 -04003163int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003164{
Chris Mason810191f2007-10-15 16:18:55 -04003165 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
Chris Masonce9adaa2008-04-09 16:28:12 -04003166 int ret;
Chris Masonca7a79a2008-05-12 12:59:19 -04003167 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Masond3977122009-01-05 21:25:51 -05003168 if (ret == 0)
Chris Masonb4ce94d2009-02-04 09:25:08 -05003169 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
Chris Masonce9adaa2008-04-09 16:28:12 -04003170 return ret;
Chris Mason6b800532007-10-15 16:17:34 -04003171}
Chris Mason0da54682007-11-07 21:08:01 -05003172
Chris Mason01d658f2011-11-01 10:08:06 -04003173static int btree_lock_page_hook(struct page *page, void *data,
3174 void (*flush_fn)(void *))
Chris Mason4bef0842008-09-08 11:18:08 -04003175{
3176 struct inode *inode = page->mapping->host;
Chris Masonb9473432009-03-13 11:00:37 -04003177 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4bef0842008-09-08 11:18:08 -04003178 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3179 struct extent_buffer *eb;
3180 unsigned long len;
3181 u64 bytenr = page_offset(page);
3182
3183 if (page->private == EXTENT_PAGE_PRIVATE)
3184 goto out;
3185
3186 len = page->private >> 2;
David Sterbaf09d1f62011-04-21 01:08:01 +02003187 eb = find_extent_buffer(io_tree, bytenr, len);
Chris Mason4bef0842008-09-08 11:18:08 -04003188 if (!eb)
3189 goto out;
3190
Chris Mason01d658f2011-11-01 10:08:06 -04003191 if (!btrfs_try_tree_write_lock(eb)) {
3192 flush_fn(data);
3193 btrfs_tree_lock(eb);
3194 }
Chris Mason4bef0842008-09-08 11:18:08 -04003195 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonb9473432009-03-13 11:00:37 -04003196
3197 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3198 spin_lock(&root->fs_info->delalloc_lock);
3199 if (root->fs_info->dirty_metadata_bytes >= eb->len)
3200 root->fs_info->dirty_metadata_bytes -= eb->len;
3201 else
3202 WARN_ON(1);
3203 spin_unlock(&root->fs_info->delalloc_lock);
3204 }
3205
Chris Mason4bef0842008-09-08 11:18:08 -04003206 btrfs_tree_unlock(eb);
3207 free_extent_buffer(eb);
3208out:
Chris Mason01d658f2011-11-01 10:08:06 -04003209 if (!trylock_page(page)) {
3210 flush_fn(data);
3211 lock_page(page);
3212 }
Chris Mason4bef0842008-09-08 11:18:08 -04003213 return 0;
3214}
3215
liuboacce9522011-01-06 19:30:25 +08003216static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
3217 int read_only)
3218{
3219 if (read_only)
3220 return;
3221
3222 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
3223 printk(KERN_WARNING "warning: mount fs with errors, "
3224 "running btrfsck is recommended\n");
3225}
3226
3227int btrfs_error_commit_super(struct btrfs_root *root)
3228{
3229 int ret;
3230
3231 mutex_lock(&root->fs_info->cleaner_mutex);
3232 btrfs_run_delayed_iputs(root);
3233 mutex_unlock(&root->fs_info->cleaner_mutex);
3234
3235 down_write(&root->fs_info->cleanup_work_sem);
3236 up_write(&root->fs_info->cleanup_work_sem);
3237
3238 /* cleanup FS via transaction */
3239 btrfs_cleanup_transaction(root);
3240
3241 ret = write_ctree_super(NULL, root, 0);
3242
3243 return ret;
3244}
3245
3246static int btrfs_destroy_ordered_operations(struct btrfs_root *root)
3247{
3248 struct btrfs_inode *btrfs_inode;
3249 struct list_head splice;
3250
3251 INIT_LIST_HEAD(&splice);
3252
3253 mutex_lock(&root->fs_info->ordered_operations_mutex);
3254 spin_lock(&root->fs_info->ordered_extent_lock);
3255
3256 list_splice_init(&root->fs_info->ordered_operations, &splice);
3257 while (!list_empty(&splice)) {
3258 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3259 ordered_operations);
3260
3261 list_del_init(&btrfs_inode->ordered_operations);
3262
3263 btrfs_invalidate_inodes(btrfs_inode->root);
3264 }
3265
3266 spin_unlock(&root->fs_info->ordered_extent_lock);
3267 mutex_unlock(&root->fs_info->ordered_operations_mutex);
3268
3269 return 0;
3270}
3271
3272static int btrfs_destroy_ordered_extents(struct btrfs_root *root)
3273{
3274 struct list_head splice;
3275 struct btrfs_ordered_extent *ordered;
3276 struct inode *inode;
3277
3278 INIT_LIST_HEAD(&splice);
3279
3280 spin_lock(&root->fs_info->ordered_extent_lock);
3281
3282 list_splice_init(&root->fs_info->ordered_extents, &splice);
3283 while (!list_empty(&splice)) {
3284 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
3285 root_extent_list);
3286
3287 list_del_init(&ordered->root_extent_list);
3288 atomic_inc(&ordered->refs);
3289
3290 /* the inode may be getting freed (in sys_unlink path). */
3291 inode = igrab(ordered->inode);
3292
3293 spin_unlock(&root->fs_info->ordered_extent_lock);
3294 if (inode)
3295 iput(inode);
3296
3297 atomic_set(&ordered->refs, 1);
3298 btrfs_put_ordered_extent(ordered);
3299
3300 spin_lock(&root->fs_info->ordered_extent_lock);
3301 }
3302
3303 spin_unlock(&root->fs_info->ordered_extent_lock);
3304
3305 return 0;
3306}
3307
3308static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3309 struct btrfs_root *root)
3310{
3311 struct rb_node *node;
3312 struct btrfs_delayed_ref_root *delayed_refs;
3313 struct btrfs_delayed_ref_node *ref;
3314 int ret = 0;
3315
3316 delayed_refs = &trans->delayed_refs;
3317
3318 spin_lock(&delayed_refs->lock);
3319 if (delayed_refs->num_entries == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003320 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003321 printk(KERN_INFO "delayed_refs has NO entry\n");
3322 return ret;
3323 }
3324
3325 node = rb_first(&delayed_refs->root);
3326 while (node) {
3327 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
3328 node = rb_next(node);
3329
3330 ref->in_tree = 0;
3331 rb_erase(&ref->rb_node, &delayed_refs->root);
3332 delayed_refs->num_entries--;
3333
3334 atomic_set(&ref->refs, 1);
3335 if (btrfs_delayed_ref_is_head(ref)) {
3336 struct btrfs_delayed_ref_head *head;
3337
3338 head = btrfs_delayed_node_to_head(ref);
3339 mutex_lock(&head->mutex);
3340 kfree(head->extent_op);
3341 delayed_refs->num_heads--;
3342 if (list_empty(&head->cluster))
3343 delayed_refs->num_heads_ready--;
3344 list_del_init(&head->cluster);
3345 mutex_unlock(&head->mutex);
3346 }
3347
3348 spin_unlock(&delayed_refs->lock);
3349 btrfs_put_delayed_ref(ref);
3350
3351 cond_resched();
3352 spin_lock(&delayed_refs->lock);
3353 }
3354
3355 spin_unlock(&delayed_refs->lock);
3356
3357 return ret;
3358}
3359
3360static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
3361{
3362 struct btrfs_pending_snapshot *snapshot;
3363 struct list_head splice;
3364
3365 INIT_LIST_HEAD(&splice);
3366
3367 list_splice_init(&t->pending_snapshots, &splice);
3368
3369 while (!list_empty(&splice)) {
3370 snapshot = list_entry(splice.next,
3371 struct btrfs_pending_snapshot,
3372 list);
3373
3374 list_del_init(&snapshot->list);
3375
3376 kfree(snapshot);
3377 }
3378
3379 return 0;
3380}
3381
3382static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
3383{
3384 struct btrfs_inode *btrfs_inode;
3385 struct list_head splice;
3386
3387 INIT_LIST_HEAD(&splice);
3388
liuboacce9522011-01-06 19:30:25 +08003389 spin_lock(&root->fs_info->delalloc_lock);
David Sterba5be76752011-06-09 10:02:51 +00003390 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003391
3392 while (!list_empty(&splice)) {
3393 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3394 delalloc_inodes);
3395
3396 list_del_init(&btrfs_inode->delalloc_inodes);
3397
3398 btrfs_invalidate_inodes(btrfs_inode->root);
3399 }
3400
3401 spin_unlock(&root->fs_info->delalloc_lock);
3402
3403 return 0;
3404}
3405
3406static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3407 struct extent_io_tree *dirty_pages,
3408 int mark)
3409{
3410 int ret;
3411 struct page *page;
3412 struct inode *btree_inode = root->fs_info->btree_inode;
3413 struct extent_buffer *eb;
3414 u64 start = 0;
3415 u64 end;
3416 u64 offset;
3417 unsigned long index;
3418
3419 while (1) {
3420 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
3421 mark);
3422 if (ret)
3423 break;
3424
3425 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3426 while (start <= end) {
3427 index = start >> PAGE_CACHE_SHIFT;
3428 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
3429 page = find_get_page(btree_inode->i_mapping, index);
3430 if (!page)
3431 continue;
3432 offset = page_offset(page);
3433
3434 spin_lock(&dirty_pages->buffer_lock);
3435 eb = radix_tree_lookup(
3436 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
3437 offset >> PAGE_CACHE_SHIFT);
3438 spin_unlock(&dirty_pages->buffer_lock);
3439 if (eb) {
3440 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3441 &eb->bflags);
3442 atomic_set(&eb->refs, 1);
3443 }
3444 if (PageWriteback(page))
3445 end_page_writeback(page);
3446
3447 lock_page(page);
3448 if (PageDirty(page)) {
3449 clear_page_dirty_for_io(page);
3450 spin_lock_irq(&page->mapping->tree_lock);
3451 radix_tree_tag_clear(&page->mapping->page_tree,
3452 page_index(page),
3453 PAGECACHE_TAG_DIRTY);
3454 spin_unlock_irq(&page->mapping->tree_lock);
3455 }
3456
3457 page->mapping->a_ops->invalidatepage(page, 0);
3458 unlock_page(page);
3459 }
3460 }
3461
3462 return ret;
3463}
3464
3465static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3466 struct extent_io_tree *pinned_extents)
3467{
3468 struct extent_io_tree *unpin;
3469 u64 start;
3470 u64 end;
3471 int ret;
3472
3473 unpin = pinned_extents;
3474 while (1) {
3475 ret = find_first_extent_bit(unpin, 0, &start, &end,
3476 EXTENT_DIRTY);
3477 if (ret)
3478 break;
3479
3480 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00003481 if (btrfs_test_opt(root, DISCARD))
3482 ret = btrfs_error_discard_extent(root, start,
3483 end + 1 - start,
3484 NULL);
liuboacce9522011-01-06 19:30:25 +08003485
3486 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3487 btrfs_error_unpin_extent_range(root, start, end);
3488 cond_resched();
3489 }
3490
3491 return 0;
3492}
3493
3494static int btrfs_cleanup_transaction(struct btrfs_root *root)
3495{
3496 struct btrfs_transaction *t;
3497 LIST_HEAD(list);
3498
3499 WARN_ON(1);
3500
liuboacce9522011-01-06 19:30:25 +08003501 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3502
Josef Bacika4abeea2011-04-11 17:25:13 -04003503 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003504 list_splice_init(&root->fs_info->trans_list, &list);
Josef Bacika4abeea2011-04-11 17:25:13 -04003505 root->fs_info->trans_no_join = 1;
3506 spin_unlock(&root->fs_info->trans_lock);
3507
liuboacce9522011-01-06 19:30:25 +08003508 while (!list_empty(&list)) {
3509 t = list_entry(list.next, struct btrfs_transaction, list);
3510 if (!t)
3511 break;
3512
3513 btrfs_destroy_ordered_operations(root);
3514
3515 btrfs_destroy_ordered_extents(root);
3516
3517 btrfs_destroy_delayed_refs(t, root);
3518
3519 btrfs_block_rsv_release(root,
3520 &root->fs_info->trans_block_rsv,
3521 t->dirty_pages.dirty_bytes);
3522
3523 /* FIXME: cleanup wait for commit */
3524 t->in_commit = 1;
3525 t->blocked = 1;
3526 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3527 wake_up(&root->fs_info->transaction_blocked_wait);
3528
3529 t->blocked = 0;
3530 if (waitqueue_active(&root->fs_info->transaction_wait))
3531 wake_up(&root->fs_info->transaction_wait);
liuboacce9522011-01-06 19:30:25 +08003532
liuboacce9522011-01-06 19:30:25 +08003533 t->commit_done = 1;
3534 if (waitqueue_active(&t->commit_wait))
3535 wake_up(&t->commit_wait);
liuboacce9522011-01-06 19:30:25 +08003536
3537 btrfs_destroy_pending_snapshots(t);
3538
3539 btrfs_destroy_delalloc_inodes(root);
3540
Josef Bacika4abeea2011-04-11 17:25:13 -04003541 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003542 root->fs_info->running_transaction = NULL;
Josef Bacika4abeea2011-04-11 17:25:13 -04003543 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003544
3545 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3546 EXTENT_DIRTY);
3547
3548 btrfs_destroy_pinned_extent(root,
3549 root->fs_info->pinned_extents);
3550
Josef Bacik13c5a932011-04-11 15:45:29 -04003551 atomic_set(&t->use_count, 0);
liuboacce9522011-01-06 19:30:25 +08003552 list_del_init(&t->list);
3553 memset(t, 0, sizeof(*t));
3554 kmem_cache_free(btrfs_transaction_cachep, t);
3555 }
3556
Josef Bacika4abeea2011-04-11 17:25:13 -04003557 spin_lock(&root->fs_info->trans_lock);
3558 root->fs_info->trans_no_join = 0;
3559 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003560 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08003561
3562 return 0;
3563}
3564
Chris Masond1310b22008-01-24 16:13:08 -05003565static struct extent_io_ops btree_extent_io_ops = {
Chris Mason4bef0842008-09-08 11:18:08 -04003566 .write_cache_pages_lock_hook = btree_lock_page_hook,
Chris Masonce9adaa2008-04-09 16:28:12 -04003567 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02003568 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04003569 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04003570 /* note we're sharing with inode.c for the merge bio hook */
3571 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05003572};