blob: deba3d9c88534850f4481dc833f2cf87081fd830 [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 Sterba7e75bf32011-03-18 22:56:43 +000032#include <asm/unaligned.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050033#include "compat.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050034#include "ctree.h"
35#include "disk-io.h"
Chris Masone089f052007-03-16 16:20:31 -040036#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040037#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040038#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040039#include "print-tree.h"
Chris Mason8b712842008-06-11 16:50:36 -040040#include "async-thread.h"
Chris Mason925baed2008-06-25 16:01:30 -040041#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040042#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040043#include "free-space-cache.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050044
Chris Masond1310b22008-01-24 16:13:08 -050045static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040046static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040047static void free_fs_root(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080048static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
49 int read_only);
50static int btrfs_destroy_ordered_operations(struct btrfs_root *root);
51static int btrfs_destroy_ordered_extents(struct btrfs_root *root);
52static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
53 struct btrfs_root *root);
54static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
55static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
56static int btrfs_destroy_marked_extents(struct btrfs_root *root,
57 struct extent_io_tree *dirty_pages,
58 int mark);
59static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
60 struct extent_io_tree *pinned_extents);
61static int btrfs_cleanup_transaction(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040062
Chris Masond352ac62008-09-29 15:18:18 -040063/*
64 * end_io_wq structs are used to do processing in task context when an IO is
65 * complete. This is used during reads to verify checksums, and it is used
66 * by writes to insert metadata for new file extents after IO is complete.
67 */
Chris Masonce9adaa2008-04-09 16:28:12 -040068struct end_io_wq {
69 struct bio *bio;
70 bio_end_io_t *end_io;
71 void *private;
72 struct btrfs_fs_info *info;
73 int error;
Chris Mason22c59942008-04-09 16:28:12 -040074 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040075 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040076 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040077};
Chris Mason0da54682007-11-07 21:08:01 -050078
Chris Masond352ac62008-09-29 15:18:18 -040079/*
80 * async submit bios are used to offload expensive checksumming
81 * onto the worker threads. They checksum file and metadata bios
82 * just before they are sent down the IO stack.
83 */
Chris Mason44b8bd72008-04-16 11:14:51 -040084struct async_submit_bio {
85 struct inode *inode;
86 struct bio *bio;
87 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -050088 extent_submit_bio_hook_t *submit_bio_start;
89 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -040090 int rw;
91 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -040092 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -040093 /*
94 * bio_offset is optional, can be used if the pages in the bio
95 * can't tell us where in the file the bio should go
96 */
97 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -040098 struct btrfs_work work;
Chris Mason44b8bd72008-04-16 11:14:51 -040099};
100
Chris Mason4008c042009-02-12 14:09:45 -0500101/* These are used to set the lockdep class on the extent buffer locks.
102 * The class is set by the readpage_end_io_hook after the buffer has
103 * passed csum validation but before the pages are unlocked.
104 *
105 * The lockdep class is also set by btrfs_init_new_buffer on freshly
106 * allocated blocks.
107 *
108 * The class is based on the level in the tree block, which allows lockdep
109 * to know that lower nodes nest inside the locks of higher nodes.
110 *
111 * We also add a check to make sure the highest level of the tree is
112 * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this
113 * code needs update as well.
114 */
115#ifdef CONFIG_DEBUG_LOCK_ALLOC
116# if BTRFS_MAX_LEVEL != 8
117# error
118# endif
119static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
120static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
121 /* leaf */
122 "btrfs-extent-00",
123 "btrfs-extent-01",
124 "btrfs-extent-02",
125 "btrfs-extent-03",
126 "btrfs-extent-04",
127 "btrfs-extent-05",
128 "btrfs-extent-06",
129 "btrfs-extent-07",
130 /* highest possible level */
131 "btrfs-extent-08",
132};
133#endif
134
Chris Masond352ac62008-09-29 15:18:18 -0400135/*
136 * extents on the btree inode are pretty simple, there's one extent
137 * that covers the entire device
138 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500139static struct extent_map *btree_get_extent(struct inode *inode,
140 struct page *page, size_t page_offset, u64 start, u64 len,
141 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400142{
143 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
144 struct extent_map *em;
145 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400146
Chris Mason890871b2009-09-02 16:24:52 -0400147 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500148 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400149 if (em) {
150 em->bdev =
151 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400152 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400153 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400154 }
Chris Mason890871b2009-09-02 16:24:52 -0400155 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400156
Chris Mason5f39d392007-10-15 16:14:19 -0400157 em = alloc_extent_map(GFP_NOFS);
158 if (!em) {
159 em = ERR_PTR(-ENOMEM);
160 goto out;
161 }
162 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400163 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400164 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400165 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400166 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500167
Chris Mason890871b2009-09-02 16:24:52 -0400168 write_lock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400169 ret = add_extent_mapping(em_tree, em);
170 if (ret == -EEXIST) {
Chris Mason0afbaf82008-04-18 14:17:20 -0400171 u64 failed_start = em->start;
172 u64 failed_len = em->len;
173
Chris Mason5f39d392007-10-15 16:14:19 -0400174 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400175 em = lookup_extent_mapping(em_tree, start, len);
Chris Mason0afbaf82008-04-18 14:17:20 -0400176 if (em) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400177 ret = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400178 } else {
179 em = lookup_extent_mapping(em_tree, failed_start,
180 failed_len);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400181 ret = -EIO;
Chris Mason0afbaf82008-04-18 14:17:20 -0400182 }
Chris Mason5f39d392007-10-15 16:14:19 -0400183 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400184 free_extent_map(em);
185 em = NULL;
Chris Mason5f39d392007-10-15 16:14:19 -0400186 }
Chris Mason890871b2009-09-02 16:24:52 -0400187 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400188
189 if (ret)
190 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400191out:
192 return em;
193}
194
Chris Mason19c00dd2007-10-15 16:19:22 -0400195u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
196{
David Woodhouse163e7832009-04-19 13:02:41 +0100197 return crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400198}
199
200void btrfs_csum_final(u32 crc, char *result)
201{
David Sterba7e75bf32011-03-18 22:56:43 +0000202 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400203}
204
Chris Masond352ac62008-09-29 15:18:18 -0400205/*
206 * compute the csum for a btree block, and either verify it or write it
207 * into the csum field of the block.
208 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400209static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
210 int verify)
211{
Josef Bacik607d4322008-12-02 07:17:45 -0500212 u16 csum_size =
213 btrfs_super_csum_size(&root->fs_info->super_copy);
214 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400215 unsigned long len;
216 unsigned long cur_len;
217 unsigned long offset = BTRFS_CSUM_SIZE;
218 char *map_token = NULL;
219 char *kaddr;
220 unsigned long map_start;
221 unsigned long map_len;
222 int err;
223 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500224 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400225
226 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500227 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400228 err = map_private_extent_buffer(buf, offset, 32,
229 &map_token, &kaddr,
230 &map_start, &map_len, KM_USER0);
Chris Masond3977122009-01-05 21:25:51 -0500231 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400232 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400233 cur_len = min(len, map_len - (offset - map_start));
234 crc = btrfs_csum_data(root, kaddr + offset - map_start,
235 crc, cur_len);
236 len -= cur_len;
237 offset += cur_len;
238 unmap_extent_buffer(buf, map_token, KM_USER0);
239 }
Josef Bacik607d4322008-12-02 07:17:45 -0500240 if (csum_size > sizeof(inline_result)) {
241 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
242 if (!result)
243 return 1;
244 } else {
245 result = (char *)&inline_result;
246 }
247
Chris Mason19c00dd2007-10-15 16:19:22 -0400248 btrfs_csum_final(crc, result);
249
250 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500251 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500252 u32 val;
253 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500254 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500255
Josef Bacik607d4322008-12-02 07:17:45 -0500256 read_extent_buffer(buf, &val, 0, csum_size);
Chris Mason193f2842009-04-27 07:29:05 -0400257 if (printk_ratelimit()) {
258 printk(KERN_INFO "btrfs: %s checksum verify "
259 "failed on %llu wanted %X found %X "
260 "level %d\n",
261 root->fs_info->sb->s_id,
262 (unsigned long long)buf->start, val, found,
263 btrfs_header_level(buf));
264 }
Josef Bacik607d4322008-12-02 07:17:45 -0500265 if (result != (char *)&inline_result)
266 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400267 return 1;
268 }
269 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500270 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400271 }
Josef Bacik607d4322008-12-02 07:17:45 -0500272 if (result != (char *)&inline_result)
273 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400274 return 0;
275}
276
Chris Masond352ac62008-09-29 15:18:18 -0400277/*
278 * we can't consider a given block up to date unless the transid of the
279 * block matches the transid in the parent node's pointer. This is how we
280 * detect blocks that either didn't get written at all or got written
281 * in the wrong place.
282 */
Chris Mason1259ab72008-05-12 13:39:03 -0400283static int verify_parent_transid(struct extent_io_tree *io_tree,
284 struct extent_buffer *eb, u64 parent_transid)
285{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000286 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400287 int ret;
288
289 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
290 return 0;
291
Josef Bacik2ac55d42010-02-03 19:33:23 +0000292 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
293 0, &cached_state, GFP_NOFS);
294 if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400295 btrfs_header_generation(eb) == parent_transid) {
296 ret = 0;
297 goto out;
298 }
Chris Mason193f2842009-04-27 07:29:05 -0400299 if (printk_ratelimit()) {
300 printk("parent transid verify failed on %llu wanted %llu "
301 "found %llu\n",
302 (unsigned long long)eb->start,
303 (unsigned long long)parent_transid,
304 (unsigned long long)btrfs_header_generation(eb));
305 }
Chris Mason1259ab72008-05-12 13:39:03 -0400306 ret = 1;
Josef Bacik2ac55d42010-02-03 19:33:23 +0000307 clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
Chris Mason33958dc2008-07-30 10:29:12 -0400308out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000309 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
310 &cached_state, GFP_NOFS);
Chris Mason1259ab72008-05-12 13:39:03 -0400311 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400312}
313
Chris Masond352ac62008-09-29 15:18:18 -0400314/*
315 * helper to read a given tree block, doing retries as required when
316 * the checksums don't match and we have alternate mirrors to try.
317 */
Chris Masonf1885912008-04-09 16:28:12 -0400318static int btree_read_extent_buffer_pages(struct btrfs_root *root,
319 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400320 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400321{
322 struct extent_io_tree *io_tree;
323 int ret;
324 int num_copies = 0;
325 int mirror_num = 0;
326
Josef Bacika826d6d2011-03-16 13:42:43 -0400327 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400328 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
329 while (1) {
330 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
331 btree_get_extent, mirror_num);
Chris Mason1259ab72008-05-12 13:39:03 -0400332 if (!ret &&
333 !verify_parent_transid(io_tree, eb, parent_transid))
Chris Masonf1885912008-04-09 16:28:12 -0400334 return ret;
Chris Masond3977122009-01-05 21:25:51 -0500335
Josef Bacika826d6d2011-03-16 13:42:43 -0400336 /*
337 * This buffer's crc is fine, but its contents are corrupted, so
338 * there is no reason to read the other copies, they won't be
339 * any less wrong.
340 */
341 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
342 return ret;
343
Chris Masonf1885912008-04-09 16:28:12 -0400344 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
345 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400346 if (num_copies == 1)
Chris Masonf1885912008-04-09 16:28:12 -0400347 return ret;
Chris Mason42352982008-04-28 16:40:52 -0400348
Chris Masonf1885912008-04-09 16:28:12 -0400349 mirror_num++;
Chris Mason42352982008-04-28 16:40:52 -0400350 if (mirror_num > num_copies)
Chris Masonf1885912008-04-09 16:28:12 -0400351 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400352 }
Chris Masonf1885912008-04-09 16:28:12 -0400353 return -EIO;
354}
Chris Mason19c00dd2007-10-15 16:19:22 -0400355
Chris Masond352ac62008-09-29 15:18:18 -0400356/*
Chris Masond3977122009-01-05 21:25:51 -0500357 * checksum a dirty tree block before IO. This has extra checks to make sure
358 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400359 */
Chris Masond3977122009-01-05 21:25:51 -0500360
Christoph Hellwigb2950862008-12-02 09:54:17 -0500361static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400362{
Chris Masond1310b22008-01-24 16:13:08 -0500363 struct extent_io_tree *tree;
Chris Mason35ebb932007-10-30 16:56:53 -0400364 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
Chris Mason19c00dd2007-10-15 16:19:22 -0400365 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400366 unsigned long len;
367 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400368 int ret;
369
Chris Masond1310b22008-01-24 16:13:08 -0500370 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason19c00dd2007-10-15 16:19:22 -0400371
Chris Masoneb14ab82011-02-10 12:35:00 -0500372 if (page->private == EXTENT_PAGE_PRIVATE) {
373 WARN_ON(1);
Chris Mason19c00dd2007-10-15 16:19:22 -0400374 goto out;
Chris Masoneb14ab82011-02-10 12:35:00 -0500375 }
376 if (!page->private) {
377 WARN_ON(1);
Chris Mason19c00dd2007-10-15 16:19:22 -0400378 goto out;
Chris Masoneb14ab82011-02-10 12:35:00 -0500379 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400380 len = page->private >> 2;
Chris Masond3977122009-01-05 21:25:51 -0500381 WARN_ON(len == 0);
382
Chris Mason19c00dd2007-10-15 16:19:22 -0400383 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
Tsutomu Itoh91ca3382011-01-05 02:32:22 +0000384 if (eb == NULL) {
385 WARN_ON(1);
386 goto out;
387 }
Chris Masonca7a79a2008-05-12 12:59:19 -0400388 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
389 btrfs_header_generation(eb));
Chris Masonf1885912008-04-09 16:28:12 -0400390 BUG_ON(ret);
Chris Mason784b4e22010-11-21 22:20:49 -0500391 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
392
Chris Mason19c00dd2007-10-15 16:19:22 -0400393 found_start = btrfs_header_bytenr(eb);
394 if (found_start != start) {
Chris Mason55c69072008-01-09 15:55:33 -0500395 WARN_ON(1);
396 goto err;
397 }
398 if (eb->first_page != page) {
Chris Mason55c69072008-01-09 15:55:33 -0500399 WARN_ON(1);
400 goto err;
401 }
402 if (!PageUptodate(page)) {
Chris Mason55c69072008-01-09 15:55:33 -0500403 WARN_ON(1);
404 goto err;
Chris Mason19c00dd2007-10-15 16:19:22 -0400405 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400406 csum_tree_block(root, eb, 0);
Chris Mason55c69072008-01-09 15:55:33 -0500407err:
Chris Mason19c00dd2007-10-15 16:19:22 -0400408 free_extent_buffer(eb);
409out:
410 return 0;
411}
412
Yan Zheng2b820322008-11-17 21:11:30 -0500413static int check_tree_block_fsid(struct btrfs_root *root,
414 struct extent_buffer *eb)
415{
416 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
417 u8 fsid[BTRFS_UUID_SIZE];
418 int ret = 1;
419
420 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
421 BTRFS_FSID_SIZE);
422 while (fs_devices) {
423 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
424 ret = 0;
425 break;
426 }
427 fs_devices = fs_devices->seed;
428 }
429 return ret;
430}
431
Josef Bacika826d6d2011-03-16 13:42:43 -0400432#define CORRUPT(reason, eb, root, slot) \
433 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
434 "root=%llu, slot=%d\n", reason, \
435 (unsigned long long)btrfs_header_bytenr(eb), \
436 (unsigned long long)root->objectid, slot)
437
438static noinline int check_leaf(struct btrfs_root *root,
439 struct extent_buffer *leaf)
440{
441 struct btrfs_key key;
442 struct btrfs_key leaf_key;
443 u32 nritems = btrfs_header_nritems(leaf);
444 int slot;
445
446 if (nritems == 0)
447 return 0;
448
449 /* Check the 0 item */
450 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
451 BTRFS_LEAF_DATA_SIZE(root)) {
452 CORRUPT("invalid item offset size pair", leaf, root, 0);
453 return -EIO;
454 }
455
456 /*
457 * Check to make sure each items keys are in the correct order and their
458 * offsets make sense. We only have to loop through nritems-1 because
459 * we check the current slot against the next slot, which verifies the
460 * next slot's offset+size makes sense and that the current's slot
461 * offset is correct.
462 */
463 for (slot = 0; slot < nritems - 1; slot++) {
464 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
465 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
466
467 /* Make sure the keys are in the right order */
468 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
469 CORRUPT("bad key order", leaf, root, slot);
470 return -EIO;
471 }
472
473 /*
474 * Make sure the offset and ends are right, remember that the
475 * item data starts at the end of the leaf and grows towards the
476 * front.
477 */
478 if (btrfs_item_offset_nr(leaf, slot) !=
479 btrfs_item_end_nr(leaf, slot + 1)) {
480 CORRUPT("slot offset bad", leaf, root, slot);
481 return -EIO;
482 }
483
484 /*
485 * Check to make sure that we don't point outside of the leaf,
486 * just incase all the items are consistent to eachother, but
487 * all point outside of the leaf.
488 */
489 if (btrfs_item_end_nr(leaf, slot) >
490 BTRFS_LEAF_DATA_SIZE(root)) {
491 CORRUPT("slot end outside of leaf", leaf, root, slot);
492 return -EIO;
493 }
494 }
495
496 return 0;
497}
498
Chris Mason4008c042009-02-12 14:09:45 -0500499#ifdef CONFIG_DEBUG_LOCK_ALLOC
500void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
501{
502 lockdep_set_class_and_name(&eb->lock,
503 &btrfs_eb_class[level],
504 btrfs_eb_name[level]);
505}
506#endif
507
Christoph Hellwigb2950862008-12-02 09:54:17 -0500508static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
Chris Masonce9adaa2008-04-09 16:28:12 -0400509 struct extent_state *state)
510{
511 struct extent_io_tree *tree;
512 u64 found_start;
513 int found_level;
514 unsigned long len;
515 struct extent_buffer *eb;
516 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400517 int ret = 0;
Chris Masonce9adaa2008-04-09 16:28:12 -0400518
519 tree = &BTRFS_I(page->mapping->host)->io_tree;
520 if (page->private == EXTENT_PAGE_PRIVATE)
521 goto out;
522 if (!page->private)
523 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500524
Chris Masonce9adaa2008-04-09 16:28:12 -0400525 len = page->private >> 2;
Chris Masond3977122009-01-05 21:25:51 -0500526 WARN_ON(len == 0);
527
Chris Masonce9adaa2008-04-09 16:28:12 -0400528 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
Tsutomu Itoh91ca3382011-01-05 02:32:22 +0000529 if (eb == NULL) {
530 ret = -EIO;
531 goto out;
532 }
Chris Masonf1885912008-04-09 16:28:12 -0400533
Chris Masonce9adaa2008-04-09 16:28:12 -0400534 found_start = btrfs_header_bytenr(eb);
Chris Mason23a07862008-09-12 08:57:47 -0400535 if (found_start != start) {
Chris Mason193f2842009-04-27 07:29:05 -0400536 if (printk_ratelimit()) {
537 printk(KERN_INFO "btrfs bad tree block start "
538 "%llu %llu\n",
539 (unsigned long long)found_start,
540 (unsigned long long)eb->start);
541 }
Chris Masonf1885912008-04-09 16:28:12 -0400542 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400543 goto err;
544 }
545 if (eb->first_page != page) {
Chris Masond3977122009-01-05 21:25:51 -0500546 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
547 eb->first_page->index, page->index);
Chris Masonce9adaa2008-04-09 16:28:12 -0400548 WARN_ON(1);
Chris Masonf1885912008-04-09 16:28:12 -0400549 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400550 goto err;
551 }
Yan Zheng2b820322008-11-17 21:11:30 -0500552 if (check_tree_block_fsid(root, eb)) {
Chris Mason193f2842009-04-27 07:29:05 -0400553 if (printk_ratelimit()) {
554 printk(KERN_INFO "btrfs bad fsid on block %llu\n",
555 (unsigned long long)eb->start);
556 }
Chris Mason1259ab72008-05-12 13:39:03 -0400557 ret = -EIO;
558 goto err;
559 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400560 found_level = btrfs_header_level(eb);
561
Chris Mason4008c042009-02-12 14:09:45 -0500562 btrfs_set_buffer_lockdep_class(eb, found_level);
563
Chris Masonce9adaa2008-04-09 16:28:12 -0400564 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400565 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400566 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400567 goto err;
568 }
569
570 /*
571 * If this is a leaf block and it is corrupt, set the corrupt bit so
572 * that we don't try and read the other copies of this block, just
573 * return -EIO.
574 */
575 if (found_level == 0 && check_leaf(root, eb)) {
576 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
577 ret = -EIO;
578 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400579
580 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
581 end = eb->start + end - 1;
Chris Masonce9adaa2008-04-09 16:28:12 -0400582err:
583 free_extent_buffer(eb);
584out:
Chris Masonf1885912008-04-09 16:28:12 -0400585 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400586}
587
Chris Masonce9adaa2008-04-09 16:28:12 -0400588static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400589{
590 struct end_io_wq *end_io_wq = bio->bi_private;
591 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400592
Chris Masonce9adaa2008-04-09 16:28:12 -0400593 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400594 end_io_wq->error = err;
Chris Mason8b712842008-06-11 16:50:36 -0400595 end_io_wq->work.func = end_workqueue_fn;
596 end_io_wq->work.flags = 0;
Chris Masond20f7042008-12-08 16:58:54 -0500597
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200598 if (bio->bi_rw & REQ_WRITE) {
Josef Bacik0cb59c92010-07-02 12:14:14 -0400599 if (end_io_wq->metadata == 1)
Chris Masoncad321a2008-12-17 14:51:42 -0500600 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
601 &end_io_wq->work);
Josef Bacik0cb59c92010-07-02 12:14:14 -0400602 else if (end_io_wq->metadata == 2)
603 btrfs_queue_worker(&fs_info->endio_freespace_worker,
604 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500605 else
606 btrfs_queue_worker(&fs_info->endio_write_workers,
607 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500608 } else {
609 if (end_io_wq->metadata)
610 btrfs_queue_worker(&fs_info->endio_meta_workers,
611 &end_io_wq->work);
612 else
613 btrfs_queue_worker(&fs_info->endio_workers,
614 &end_io_wq->work);
615 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400616}
617
Josef Bacik0cb59c92010-07-02 12:14:14 -0400618/*
619 * For the metadata arg you want
620 *
621 * 0 - if data
622 * 1 - if normal metadta
623 * 2 - if writing to the free space cache area
624 */
Chris Mason22c59942008-04-09 16:28:12 -0400625int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
626 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400627{
Chris Masonce9adaa2008-04-09 16:28:12 -0400628 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400629 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
630 if (!end_io_wq)
631 return -ENOMEM;
632
633 end_io_wq->private = bio->bi_private;
634 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400635 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400636 end_io_wq->error = 0;
637 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400638 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400639
640 bio->bi_private = end_io_wq;
641 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400642 return 0;
643}
644
Chris Masonb64a2852008-08-20 13:39:41 -0400645unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400646{
647 unsigned long limit = min_t(unsigned long,
648 info->workers.max_workers,
649 info->fs_devices->open_devices);
650 return 256 * limit;
651}
652
Chris Mason777e6bd2008-08-15 15:34:15 -0400653int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
Chris Mason0986fe9e2008-08-15 15:34:15 -0400654{
Chris Masonb64a2852008-08-20 13:39:41 -0400655 return atomic_read(&info->nr_async_bios) >
656 btrfs_async_submit_limit(info);
Chris Mason0986fe9e2008-08-15 15:34:15 -0400657}
658
Chris Mason4a69a412008-11-06 22:03:00 -0500659static void run_one_async_start(struct btrfs_work *work)
660{
Chris Mason4a69a412008-11-06 22:03:00 -0500661 struct async_submit_bio *async;
662
663 async = container_of(work, struct async_submit_bio, work);
Chris Mason4a69a412008-11-06 22:03:00 -0500664 async->submit_bio_start(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400665 async->mirror_num, async->bio_flags,
666 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500667}
668
669static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400670{
671 struct btrfs_fs_info *fs_info;
672 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400673 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400674
675 async = container_of(work, struct async_submit_bio, work);
676 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400677
Chris Masonb64a2852008-08-20 13:39:41 -0400678 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400679 limit = limit * 2 / 3;
680
Chris Mason8b712842008-06-11 16:50:36 -0400681 atomic_dec(&fs_info->nr_async_submits);
Chris Mason0986fe9e2008-08-15 15:34:15 -0400682
Chris Masonb64a2852008-08-20 13:39:41 -0400683 if (atomic_read(&fs_info->nr_async_submits) < limit &&
684 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400685 wake_up(&fs_info->async_submit_wait);
686
Chris Mason4a69a412008-11-06 22:03:00 -0500687 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400688 async->mirror_num, async->bio_flags,
689 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500690}
691
692static void run_one_async_free(struct btrfs_work *work)
693{
694 struct async_submit_bio *async;
695
696 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400697 kfree(async);
698}
699
Chris Mason44b8bd72008-04-16 11:14:51 -0400700int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
701 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400702 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400703 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500704 extent_submit_bio_hook_t *submit_bio_start,
705 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400706{
707 struct async_submit_bio *async;
708
709 async = kmalloc(sizeof(*async), GFP_NOFS);
710 if (!async)
711 return -ENOMEM;
712
713 async->inode = inode;
714 async->rw = rw;
715 async->bio = bio;
716 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500717 async->submit_bio_start = submit_bio_start;
718 async->submit_bio_done = submit_bio_done;
719
720 async->work.func = run_one_async_start;
721 async->work.ordered_func = run_one_async_done;
722 async->work.ordered_free = run_one_async_free;
723
Chris Mason8b712842008-06-11 16:50:36 -0400724 async->work.flags = 0;
Chris Masonc8b97812008-10-29 14:49:59 -0400725 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400726 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400727
Chris Masoncb03c742008-05-15 16:15:45 -0400728 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400729
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200730 if (rw & REQ_SYNC)
Chris Masond313d7a2009-04-20 15:50:09 -0400731 btrfs_set_work_high_prio(&async->work);
732
Chris Mason8b712842008-06-11 16:50:36 -0400733 btrfs_queue_worker(&fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400734
Chris Masond3977122009-01-05 21:25:51 -0500735 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500736 atomic_read(&fs_info->nr_async_submits)) {
737 wait_event(fs_info->async_submit_wait,
738 (atomic_read(&fs_info->nr_async_submits) == 0));
739 }
740
Chris Mason44b8bd72008-04-16 11:14:51 -0400741 return 0;
742}
743
Chris Masonce3ed712008-09-23 13:14:12 -0400744static int btree_csum_one_bio(struct bio *bio)
745{
746 struct bio_vec *bvec = bio->bi_io_vec;
747 int bio_index = 0;
748 struct btrfs_root *root;
749
750 WARN_ON(bio->bi_vcnt <= 0);
Chris Masond3977122009-01-05 21:25:51 -0500751 while (bio_index < bio->bi_vcnt) {
Chris Masonce3ed712008-09-23 13:14:12 -0400752 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
753 csum_dirty_buffer(root, bvec->bv_page);
754 bio_index++;
755 bvec++;
756 }
757 return 0;
758}
759
Chris Mason4a69a412008-11-06 22:03:00 -0500760static int __btree_submit_bio_start(struct inode *inode, int rw,
761 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400762 unsigned long bio_flags,
763 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400764{
Chris Mason8b712842008-06-11 16:50:36 -0400765 /*
766 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400767 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400768 */
Chris Mason4a69a412008-11-06 22:03:00 -0500769 btree_csum_one_bio(bio);
770 return 0;
771}
Chris Mason22c59942008-04-09 16:28:12 -0400772
Chris Mason4a69a412008-11-06 22:03:00 -0500773static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400774 int mirror_num, unsigned long bio_flags,
775 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500776{
Chris Mason8b712842008-06-11 16:50:36 -0400777 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500778 * when we're called for a write, we're already in the async
779 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400780 */
Chris Mason8b712842008-06-11 16:50:36 -0400781 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Chris Mason0b86a832008-03-24 15:01:56 -0400782}
783
Chris Mason44b8bd72008-04-16 11:14:51 -0400784static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400785 int mirror_num, unsigned long bio_flags,
786 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400787{
Chris Masoncad321a2008-12-17 14:51:42 -0500788 int ret;
789
790 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
791 bio, 1);
792 BUG_ON(ret);
793
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200794 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500795 /*
796 * called for a read, do the setup so that checksum validation
797 * can happen in the async kernel threads
798 */
Chris Mason4a69a412008-11-06 22:03:00 -0500799 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
Chris Mason6f3577b2008-11-13 09:59:36 -0500800 mirror_num, 0);
Chris Mason44b8bd72008-04-16 11:14:51 -0400801 }
Chris Masond313d7a2009-04-20 15:50:09 -0400802
Chris Masoncad321a2008-12-17 14:51:42 -0500803 /*
804 * kthread helpers are used to submit writes so that checksumming
805 * can happen in parallel across all CPUs
806 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400807 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
Chris Masonc8b97812008-10-29 14:49:59 -0400808 inode, rw, bio, mirror_num, 0,
Chris Masoneaf25d92010-05-25 09:48:28 -0400809 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500810 __btree_submit_bio_start,
811 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400812}
813
Jan Beulich3dd14622010-12-07 14:54:09 +0000814#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500815static int btree_migratepage(struct address_space *mapping,
816 struct page *newpage, struct page *page)
817{
818 /*
819 * we can't safely write a btree page from here,
820 * we haven't done the locking hook
821 */
822 if (PageDirty(page))
823 return -EAGAIN;
824 /*
825 * Buffers may be managed in a filesystem specific way.
826 * We must have no buffers or drop them.
827 */
828 if (page_has_private(page) &&
829 !try_to_release_page(page, GFP_KERNEL))
830 return -EAGAIN;
Chris Mason784b4e22010-11-21 22:20:49 -0500831 return migrate_page(mapping, newpage, page);
832}
Jan Beulich3dd14622010-12-07 14:54:09 +0000833#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500834
Chris Mason5f39d392007-10-15 16:14:19 -0400835static int btree_writepage(struct page *page, struct writeback_control *wbc)
836{
Chris Masond1310b22008-01-24 16:13:08 -0500837 struct extent_io_tree *tree;
Chris Masonb9473432009-03-13 11:00:37 -0400838 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
839 struct extent_buffer *eb;
840 int was_dirty;
Chris Mason5443be42008-08-15 15:34:16 -0400841
Chris Masonb9473432009-03-13 11:00:37 -0400842 tree = &BTRFS_I(page->mapping->host)->io_tree;
843 if (!(current->flags & PF_MEMALLOC)) {
844 return extent_write_full_page(tree, page,
845 btree_get_extent, wbc);
Chris Mason5443be42008-08-15 15:34:16 -0400846 }
Chris Masonb9473432009-03-13 11:00:37 -0400847
848 redirty_page_for_writepage(wbc, page);
Chris Mason784b4e22010-11-21 22:20:49 -0500849 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
Chris Masonb9473432009-03-13 11:00:37 -0400850 WARN_ON(!eb);
851
852 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
853 if (!was_dirty) {
854 spin_lock(&root->fs_info->delalloc_lock);
855 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
856 spin_unlock(&root->fs_info->delalloc_lock);
857 }
858 free_extent_buffer(eb);
859
860 unlock_page(page);
861 return 0;
Chris Mason5f39d392007-10-15 16:14:19 -0400862}
Chris Mason0da54682007-11-07 21:08:01 -0500863
864static int btree_writepages(struct address_space *mapping,
865 struct writeback_control *wbc)
866{
Chris Masond1310b22008-01-24 16:13:08 -0500867 struct extent_io_tree *tree;
868 tree = &BTRFS_I(mapping->host)->io_tree;
Chris Masond8d5f3e2007-12-11 12:42:00 -0500869 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Masonb9473432009-03-13 11:00:37 -0400870 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
Chris Mason793955b2007-11-26 16:34:41 -0800871 u64 num_dirty;
Chris Mason24ab9cd2008-09-24 14:51:30 -0400872 unsigned long thresh = 32 * 1024 * 1024;
Chris Mason448d6402007-11-27 07:52:01 -0800873
874 if (wbc->for_kupdate)
875 return 0;
876
Chris Masonb9473432009-03-13 11:00:37 -0400877 /* this is a bit racy, but that's ok */
878 num_dirty = root->fs_info->dirty_metadata_bytes;
Chris Masond3977122009-01-05 21:25:51 -0500879 if (num_dirty < thresh)
Chris Mason793955b2007-11-26 16:34:41 -0800880 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800881 }
Chris Mason0da54682007-11-07 21:08:01 -0500882 return extent_writepages(tree, mapping, btree_get_extent, wbc);
883}
884
Christoph Hellwigb2950862008-12-02 09:54:17 -0500885static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -0400886{
Chris Masond1310b22008-01-24 16:13:08 -0500887 struct extent_io_tree *tree;
888 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -0400889 return extent_read_full_page(tree, page, btree_get_extent);
890}
891
Chris Mason70dec802008-01-29 09:59:12 -0500892static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -0400893{
Chris Masond1310b22008-01-24 16:13:08 -0500894 struct extent_io_tree *tree;
895 struct extent_map_tree *map;
Chris Mason5f39d392007-10-15 16:14:19 -0400896 int ret;
897
Chris Mason98509cf2008-09-11 15:51:43 -0400898 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -0500899 return 0;
Chris Mason98509cf2008-09-11 15:51:43 -0400900
Chris Masond1310b22008-01-24 16:13:08 -0500901 tree = &BTRFS_I(page->mapping->host)->io_tree;
902 map = &BTRFS_I(page->mapping->host)->extent_tree;
Chris Mason6af118ce2008-07-22 11:18:07 -0400903
Chris Mason7b13b7b2008-04-18 10:29:50 -0400904 ret = try_release_extent_state(map, tree, page, gfp_flags);
Chris Masond3977122009-01-05 21:25:51 -0500905 if (!ret)
Chris Mason6af118ce2008-07-22 11:18:07 -0400906 return 0;
Chris Mason6af118ce2008-07-22 11:18:07 -0400907
908 ret = try_release_extent_buffer(tree, page);
Chris Mason5f39d392007-10-15 16:14:19 -0400909 if (ret == 1) {
910 ClearPagePrivate(page);
911 set_page_private(page, 0);
912 page_cache_release(page);
913 }
Chris Mason6af118ce2008-07-22 11:18:07 -0400914
Chris Masond98237b2007-03-28 13:57:48 -0400915 return ret;
Chris Masone20d96d2007-03-22 12:13:20 -0400916}
Chris Mason123abc82007-03-14 14:14:43 -0400917
Chris Mason5f39d392007-10-15 16:14:19 -0400918static void btree_invalidatepage(struct page *page, unsigned long offset)
Chris Masond98237b2007-03-28 13:57:48 -0400919{
Chris Masond1310b22008-01-24 16:13:08 -0500920 struct extent_io_tree *tree;
921 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -0400922 extent_invalidatepage(tree, page, offset);
923 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7bc2008-04-18 16:11:30 -0400924 if (PagePrivate(page)) {
Chris Masond3977122009-01-05 21:25:51 -0500925 printk(KERN_WARNING "btrfs warning page private not zero "
926 "on page %llu\n", (unsigned long long)page_offset(page));
Chris Mason9ad6b7bc2008-04-18 16:11:30 -0400927 ClearPagePrivate(page);
928 set_page_private(page, 0);
929 page_cache_release(page);
930 }
Chris Masond98237b2007-03-28 13:57:48 -0400931}
932
Alexey Dobriyan7f094102009-09-21 17:01:10 -0700933static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -0400934 .readpage = btree_readpage,
935 .writepage = btree_writepage,
Chris Mason0da54682007-11-07 21:08:01 -0500936 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -0400937 .releasepage = btree_releasepage,
938 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -0500939#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500940 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -0500941#endif
Chris Masond98237b2007-03-28 13:57:48 -0400942};
943
Chris Masonca7a79a2008-05-12 12:59:19 -0400944int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
945 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -0400946{
Chris Mason5f39d392007-10-15 16:14:19 -0400947 struct extent_buffer *buf = NULL;
948 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -0400949 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -0400950
Chris Masondb945352007-10-15 16:15:53 -0400951 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -0400952 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -0400953 return 0;
Chris Masond1310b22008-01-24 16:13:08 -0500954 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Chris Masonf1885912008-04-09 16:28:12 -0400955 buf, 0, 0, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400956 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -0400957 return ret;
Chris Mason090d1872007-05-01 08:53:32 -0400958}
959
Chris Mason0999df52008-04-01 13:48:14 -0400960struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
961 u64 bytenr, u32 blocksize)
962{
963 struct inode *btree_inode = root->fs_info->btree_inode;
964 struct extent_buffer *eb;
965 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
966 bytenr, blocksize, GFP_NOFS);
967 return eb;
968}
969
970struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
971 u64 bytenr, u32 blocksize)
972{
973 struct inode *btree_inode = root->fs_info->btree_inode;
974 struct extent_buffer *eb;
975
976 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
977 bytenr, blocksize, NULL, GFP_NOFS);
978 return eb;
979}
980
981
Chris Masone02119d2008-09-05 16:13:11 -0400982int btrfs_write_tree_block(struct extent_buffer *buf)
983{
Christoph Hellwig8aa38c32009-10-01 12:58:30 -0400984 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
985 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -0400986}
987
988int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
989{
Christoph Hellwig8aa38c32009-10-01 12:58:30 -0400990 return filemap_fdatawait_range(buf->first_page->mapping,
991 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -0400992}
993
Chris Masondb945352007-10-15 16:15:53 -0400994struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -0400995 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -0400996{
Chris Mason5f39d392007-10-15 16:14:19 -0400997 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400998 int ret;
999
Chris Masondb945352007-10-15 16:15:53 -04001000 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001001 if (!buf)
1002 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001003
Chris Masonca7a79a2008-05-12 12:59:19 -04001004 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Masonce9adaa2008-04-09 16:28:12 -04001005
Chris Masond3977122009-01-05 21:25:51 -05001006 if (ret == 0)
Chris Masonb4ce94d2009-02-04 09:25:08 -05001007 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
Chris Mason5f39d392007-10-15 16:14:19 -04001008 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001009
Chris Masoneb60cea2007-02-02 09:18:22 -05001010}
1011
Chris Masone089f052007-03-16 16:20:31 -04001012int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
Chris Mason5f39d392007-10-15 16:14:19 -04001013 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001014{
Chris Mason5f39d392007-10-15 16:14:19 -04001015 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason55c69072008-01-09 15:55:33 -05001016 if (btrfs_header_generation(buf) ==
Chris Mason925baed2008-06-25 16:01:30 -04001017 root->fs_info->running_transaction->transid) {
Chris Masonb9447ef82009-03-09 11:45:38 -04001018 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001019
Chris Masonb9473432009-03-13 11:00:37 -04001020 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1021 spin_lock(&root->fs_info->delalloc_lock);
1022 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1023 root->fs_info->dirty_metadata_bytes -= buf->len;
1024 else
1025 WARN_ON(1);
1026 spin_unlock(&root->fs_info->delalloc_lock);
1027 }
Chris Masonb4ce94d2009-02-04 09:25:08 -05001028
Chris Masonb9473432009-03-13 11:00:37 -04001029 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1030 btrfs_set_lock_blocking(buf);
Chris Masond1310b22008-01-24 16:13:08 -05001031 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
Chris Mason55c69072008-01-09 15:55:33 -05001032 buf);
Chris Mason925baed2008-06-25 16:01:30 -04001033 }
Chris Mason5f39d392007-10-15 16:14:19 -04001034 return 0;
1035}
1036
Chris Masondb945352007-10-15 16:15:53 -04001037static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001038 u32 stripesize, struct btrfs_root *root,
Chris Mason9f5fae22007-03-20 14:38:32 -04001039 struct btrfs_fs_info *fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04001040 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001041{
Chris Masoncfaa7292007-02-21 17:04:57 -05001042 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001043 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001044 root->sectorsize = sectorsize;
1045 root->nodesize = nodesize;
1046 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001047 root->stripesize = stripesize;
Chris Mason123abc82007-03-14 14:14:43 -04001048 root->ref_cows = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001049 root->track_dirty = 0;
Yan, Zhengc71bf092009-11-12 09:34:40 +00001050 root->in_radix = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001051 root->orphan_item_inserted = 0;
1052 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001053
Chris Mason9f5fae22007-03-20 14:38:32 -04001054 root->fs_info = fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001055 root->objectid = objectid;
1056 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001057 root->highest_objectid = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001058 root->name = NULL;
Chris Mason4313b392008-01-03 09:08:48 -05001059 root->in_sysfs = 0;
Eric Paris6bef4d32010-02-23 19:43:04 +00001060 root->inode_tree = RB_ROOT;
Yan, Zhengf0486c62010-05-16 10:46:25 -04001061 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001062 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001063
1064 INIT_LIST_HEAD(&root->dirty_list);
Josef Bacik7b128762008-07-24 12:17:14 -04001065 INIT_LIST_HEAD(&root->orphan_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001066 INIT_LIST_HEAD(&root->root_list);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001067 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001068 spin_lock_init(&root->inode_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001069 spin_lock_init(&root->accounting_lock);
Chris Masona2135012008-06-25 16:01:30 -04001070 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001071 mutex_init(&root->log_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001072 init_waitqueue_head(&root->log_writer_wait);
1073 init_waitqueue_head(&root->log_commit_wait[0]);
1074 init_waitqueue_head(&root->log_commit_wait[1]);
1075 atomic_set(&root->log_commit[0], 0);
1076 atomic_set(&root->log_commit[1], 0);
1077 atomic_set(&root->log_writers, 0);
1078 root->log_batch = 0;
1079 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001080 root->last_log_commit = 0;
Chris Masond0c803c2008-09-11 16:17:57 -04001081 extent_io_tree_init(&root->dirty_log_pages,
1082 fs_info->btree_inode->i_mapping, GFP_NOFS);
Chris Mason017e5362008-07-28 15:32:51 -04001083
Chris Mason3768f362007-03-13 16:47:54 -04001084 memset(&root->root_key, 0, sizeof(root->root_key));
1085 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001086 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001087 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Chris Mason3f157a22008-06-25 16:01:31 -04001088 root->defrag_trans_start = fs_info->generation;
Josef Bacik58176a92007-08-29 15:47:34 -04001089 init_completion(&root->kobj_unregister);
Chris Mason6702ed42007-08-07 16:15:09 -04001090 root->defrag_running = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001091 root->root_key.objectid = objectid;
Chris Mason3394e162008-11-17 20:42:26 -05001092 root->anon_super.s_root = NULL;
1093 root->anon_super.s_dev = 0;
1094 INIT_LIST_HEAD(&root->anon_super.s_list);
1095 INIT_LIST_HEAD(&root->anon_super.s_instances);
1096 init_rwsem(&root->anon_super.s_umount);
1097
Chris Mason3768f362007-03-13 16:47:54 -04001098 return 0;
1099}
1100
Chris Masondb945352007-10-15 16:15:53 -04001101static int find_and_setup_root(struct btrfs_root *tree_root,
Chris Mason9f5fae22007-03-20 14:38:32 -04001102 struct btrfs_fs_info *fs_info,
1103 u64 objectid,
Chris Masone20d96d2007-03-22 12:13:20 -04001104 struct btrfs_root *root)
Chris Mason3768f362007-03-13 16:47:54 -04001105{
1106 int ret;
Chris Masondb945352007-10-15 16:15:53 -04001107 u32 blocksize;
Yan Zheng84234f32008-10-29 14:49:05 -04001108 u64 generation;
Chris Mason3768f362007-03-13 16:47:54 -04001109
Chris Masondb945352007-10-15 16:15:53 -04001110 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001111 tree_root->sectorsize, tree_root->stripesize,
1112 root, fs_info, objectid);
Chris Mason3768f362007-03-13 16:47:54 -04001113 ret = btrfs_find_last_root(tree_root, objectid,
1114 &root->root_item, &root->root_key);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001115 if (ret > 0)
1116 return -ENOENT;
Chris Mason3768f362007-03-13 16:47:54 -04001117 BUG_ON(ret);
1118
Yan Zheng84234f32008-10-29 14:49:05 -04001119 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001120 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1121 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001122 blocksize, generation);
Chris Mason68433b72010-12-13 14:47:58 -05001123 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1124 free_extent_buffer(root->node);
1125 return -EIO;
1126 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001127 root->commit_root = btrfs_root_node(root);
Chris Masond97e63b2007-02-20 16:40:44 -05001128 return 0;
1129}
1130
Yan Zheng7237f182009-01-21 12:54:03 -05001131static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1132 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001133{
1134 struct btrfs_root *root;
1135 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001136 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001137
1138 root = kzalloc(sizeof(*root), GFP_NOFS);
1139 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001140 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001141
1142 __setup_root(tree_root->nodesize, tree_root->leafsize,
1143 tree_root->sectorsize, tree_root->stripesize,
1144 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1145
1146 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1147 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1148 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Yan Zheng7237f182009-01-21 12:54:03 -05001149 /*
1150 * log trees do not get reference counted because they go away
1151 * before a real commit is actually done. They do store pointers
1152 * to file data extents, and those reference counts still get
1153 * updated (along with back refs to the log tree).
1154 */
Chris Masone02119d2008-09-05 16:13:11 -04001155 root->ref_cows = 0;
1156
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001157 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1158 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001159 if (IS_ERR(leaf)) {
1160 kfree(root);
1161 return ERR_CAST(leaf);
1162 }
Chris Masone02119d2008-09-05 16:13:11 -04001163
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001164 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1165 btrfs_set_header_bytenr(leaf, leaf->start);
1166 btrfs_set_header_generation(leaf, trans->transid);
1167 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1168 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001169 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001170
1171 write_extent_buffer(root->node, root->fs_info->fsid,
1172 (unsigned long)btrfs_header_fsid(root->node),
1173 BTRFS_FSID_SIZE);
1174 btrfs_mark_buffer_dirty(root->node);
1175 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001176 return root;
1177}
1178
1179int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1180 struct btrfs_fs_info *fs_info)
1181{
1182 struct btrfs_root *log_root;
1183
1184 log_root = alloc_log_tree(trans, fs_info);
1185 if (IS_ERR(log_root))
1186 return PTR_ERR(log_root);
1187 WARN_ON(fs_info->log_root_tree);
1188 fs_info->log_root_tree = log_root;
1189 return 0;
1190}
1191
1192int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1193 struct btrfs_root *root)
1194{
1195 struct btrfs_root *log_root;
1196 struct btrfs_inode_item *inode_item;
1197
1198 log_root = alloc_log_tree(trans, root->fs_info);
1199 if (IS_ERR(log_root))
1200 return PTR_ERR(log_root);
1201
1202 log_root->last_trans = trans->transid;
1203 log_root->root_key.offset = root->root_key.objectid;
1204
1205 inode_item = &log_root->root_item.inode;
1206 inode_item->generation = cpu_to_le64(1);
1207 inode_item->size = cpu_to_le64(3);
1208 inode_item->nlink = cpu_to_le32(1);
1209 inode_item->nbytes = cpu_to_le64(root->leafsize);
1210 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1211
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001212 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001213
1214 WARN_ON(root->log_root);
1215 root->log_root = log_root;
1216 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001217 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001218 return 0;
1219}
1220
1221struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1222 struct btrfs_key *location)
1223{
1224 struct btrfs_root *root;
1225 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001226 struct btrfs_path *path;
Chris Mason5f39d392007-10-15 16:14:19 -04001227 struct extent_buffer *l;
Yan Zheng84234f32008-10-29 14:49:05 -04001228 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001229 u32 blocksize;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001230 int ret = 0;
1231
Chris Mason5eda7b52007-06-22 14:16:25 -04001232 root = kzalloc(sizeof(*root), GFP_NOFS);
Chris Mason0cf6c622007-06-09 09:22:25 -04001233 if (!root)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001234 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001235 if (location->offset == (u64)-1) {
Chris Masondb945352007-10-15 16:15:53 -04001236 ret = find_and_setup_root(tree_root, fs_info,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001237 location->objectid, root);
1238 if (ret) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04001239 kfree(root);
1240 return ERR_PTR(ret);
1241 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001242 goto out;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001243 }
1244
Chris Masondb945352007-10-15 16:15:53 -04001245 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001246 tree_root->sectorsize, tree_root->stripesize,
1247 root, fs_info, location->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001248
1249 path = btrfs_alloc_path();
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00001250 if (!path) {
1251 kfree(root);
1252 return ERR_PTR(-ENOMEM);
1253 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001254 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001255 if (ret == 0) {
1256 l = path->nodes[0];
1257 read_extent_buffer(l, &root->root_item,
1258 btrfs_item_ptr_offset(l, path->slots[0]),
1259 sizeof(root->root_item));
1260 memcpy(&root->root_key, location, sizeof(*location));
Chris Mason0f7d52f2007-04-09 10:42:37 -04001261 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001262 btrfs_free_path(path);
1263 if (ret) {
Tsutomu Itoh5e540f72010-12-27 06:53:10 +00001264 kfree(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001265 if (ret > 0)
1266 ret = -ENOENT;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001267 return ERR_PTR(ret);
1268 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001269
Yan Zheng84234f32008-10-29 14:49:05 -04001270 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001271 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1272 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001273 blocksize, generation);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001274 root->commit_root = btrfs_root_node(root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001275 BUG_ON(!root->node);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001276out:
Li Zefan08fe4db2011-03-28 02:01:25 +00001277 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
Chris Masone02119d2008-09-05 16:13:11 -04001278 root->ref_cows = 1;
Li Zefan08fe4db2011-03-28 02:01:25 +00001279 btrfs_check_and_init_root_item(&root->root_item);
1280 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001281
Chris Mason5eda7b52007-06-22 14:16:25 -04001282 return root;
1283}
1284
Chris Masondc17ff82008-01-08 15:46:30 -05001285struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1286 u64 root_objectid)
1287{
1288 struct btrfs_root *root;
1289
1290 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
1291 return fs_info->tree_root;
1292 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
1293 return fs_info->extent_root;
1294
1295 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1296 (unsigned long)root_objectid);
1297 return root;
1298}
1299
Chris Masonedbd8d42007-12-21 16:27:24 -05001300struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1301 struct btrfs_key *location)
Chris Mason5eda7b52007-06-22 14:16:25 -04001302{
1303 struct btrfs_root *root;
1304 int ret;
1305
Chris Masonedbd8d42007-12-21 16:27:24 -05001306 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1307 return fs_info->tree_root;
1308 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1309 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001310 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1311 return fs_info->chunk_root;
1312 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1313 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001314 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1315 return fs_info->csum_root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001316again:
1317 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001318 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1319 (unsigned long)location->objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001320 spin_unlock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001321 if (root)
1322 return root;
1323
Chris Masone02119d2008-09-05 16:13:11 -04001324 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001325 if (IS_ERR(root))
1326 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001327
1328 set_anon_super(&root->anon_super, NULL);
1329
Yan, Zhengd68fc572010-05-16 10:49:58 -04001330 if (btrfs_root_refs(&root->root_item) == 0) {
1331 ret = -ENOENT;
1332 goto fail;
1333 }
1334
1335 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1336 if (ret < 0)
1337 goto fail;
1338 if (ret == 0)
1339 root->orphan_item_inserted = 1;
1340
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001341 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1342 if (ret)
1343 goto fail;
1344
1345 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04001346 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1347 (unsigned long)root->root_key.objectid,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001348 root);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001349 if (ret == 0)
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001350 root->in_radix = 1;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001351
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001352 spin_unlock(&fs_info->fs_roots_radix_lock);
1353 radix_tree_preload_end();
Chris Mason0f7d52f2007-04-09 10:42:37 -04001354 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001355 if (ret == -EEXIST) {
1356 free_fs_root(root);
1357 goto again;
1358 }
1359 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001360 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001361
1362 ret = btrfs_find_dead_roots(fs_info->tree_root,
1363 root->root_key.objectid);
1364 WARN_ON(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001365 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001366fail:
1367 free_fs_root(root);
1368 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001369}
1370
1371struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
1372 struct btrfs_key *location,
1373 const char *name, int namelen)
1374{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001375 return btrfs_read_fs_root_no_name(fs_info, location);
1376#if 0
Chris Masonedbd8d42007-12-21 16:27:24 -05001377 struct btrfs_root *root;
1378 int ret;
1379
1380 root = btrfs_read_fs_root_no_name(fs_info, location);
1381 if (!root)
1382 return NULL;
Josef Bacik58176a92007-08-29 15:47:34 -04001383
Chris Mason4313b392008-01-03 09:08:48 -05001384 if (root->in_sysfs)
1385 return root;
1386
Josef Bacik58176a92007-08-29 15:47:34 -04001387 ret = btrfs_set_root_name(root, name, namelen);
1388 if (ret) {
Chris Mason5f39d392007-10-15 16:14:19 -04001389 free_extent_buffer(root->node);
Josef Bacik58176a92007-08-29 15:47:34 -04001390 kfree(root);
1391 return ERR_PTR(ret);
1392 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001393
Josef Bacik58176a92007-08-29 15:47:34 -04001394 ret = btrfs_sysfs_add_root(root);
1395 if (ret) {
Chris Mason5f39d392007-10-15 16:14:19 -04001396 free_extent_buffer(root->node);
Josef Bacik58176a92007-08-29 15:47:34 -04001397 kfree(root->name);
1398 kfree(root);
1399 return ERR_PTR(ret);
1400 }
Chris Mason4313b392008-01-03 09:08:48 -05001401 root->in_sysfs = 1;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001402 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001403#endif
Chris Mason0f7d52f2007-04-09 10:42:37 -04001404}
Chris Mason04160082008-03-26 10:28:07 -04001405
1406static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1407{
1408 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1409 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001410 struct btrfs_device *device;
1411 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001412
Xiao Guangrong1f781602011-04-20 10:09:16 +00001413 rcu_read_lock();
1414 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001415 if (!device->bdev)
1416 continue;
Chris Mason04160082008-03-26 10:28:07 -04001417 bdi = blk_get_backing_dev_info(device->bdev);
1418 if (bdi && bdi_congested(bdi, bdi_bits)) {
1419 ret = 1;
1420 break;
1421 }
1422 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001423 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001424 return ret;
1425}
1426
Chris Mason38b66982008-04-22 09:22:11 -04001427/*
Jens Axboead081f12009-06-12 14:43:40 +02001428 * If this fails, caller must call bdi_destroy() to get rid of the
1429 * bdi again.
1430 */
Chris Mason04160082008-03-26 10:28:07 -04001431static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1432{
Jens Axboead081f12009-06-12 14:43:40 +02001433 int err;
1434
1435 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001436 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001437 if (err)
1438 return err;
1439
Chris Mason4575c9c2008-04-18 16:13:31 -04001440 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001441 bdi->congested_fn = btrfs_congested_fn;
1442 bdi->congested_data = info;
1443 return 0;
1444}
1445
Chris Masonce9adaa2008-04-09 16:28:12 -04001446static int bio_ready_for_csum(struct bio *bio)
1447{
1448 u64 length = 0;
1449 u64 buf_len = 0;
1450 u64 start = 0;
1451 struct page *page;
1452 struct extent_io_tree *io_tree = NULL;
Chris Masonce9adaa2008-04-09 16:28:12 -04001453 struct bio_vec *bvec;
1454 int i;
1455 int ret;
1456
1457 bio_for_each_segment(bvec, bio, i) {
1458 page = bvec->bv_page;
1459 if (page->private == EXTENT_PAGE_PRIVATE) {
1460 length += bvec->bv_len;
1461 continue;
1462 }
1463 if (!page->private) {
1464 length += bvec->bv_len;
1465 continue;
1466 }
1467 length = bvec->bv_len;
1468 buf_len = page->private >> 2;
1469 start = page_offset(page) + bvec->bv_offset;
1470 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Masonce9adaa2008-04-09 16:28:12 -04001471 }
1472 /* are we fully contained in this bio? */
1473 if (buf_len <= length)
1474 return 1;
1475
1476 ret = extent_range_uptodate(io_tree, start + length,
1477 start + buf_len - 1);
Chris Masonce9adaa2008-04-09 16:28:12 -04001478 return ret;
1479}
1480
Chris Mason8b712842008-06-11 16:50:36 -04001481/*
1482 * called by the kthread helper functions to finally call the bio end_io
1483 * functions. This is where read checksum verification actually happens
1484 */
1485static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001486{
Chris Masonce9adaa2008-04-09 16:28:12 -04001487 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001488 struct end_io_wq *end_io_wq;
1489 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001490 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001491
Chris Mason8b712842008-06-11 16:50:36 -04001492 end_io_wq = container_of(work, struct end_io_wq, work);
1493 bio = end_io_wq->bio;
1494 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001495
Chris Masoncad321a2008-12-17 14:51:42 -05001496 /* metadata bio reads are special because the whole tree block must
Chris Mason8b712842008-06-11 16:50:36 -04001497 * be checksummed at once. This makes sure the entire block is in
1498 * ram and up to date before trying to verify things. For
1499 * blocksize <= pagesize, it is basically a noop
1500 */
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +02001501 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
Chris Masoncad321a2008-12-17 14:51:42 -05001502 !bio_ready_for_csum(bio)) {
Chris Masond20f7042008-12-08 16:58:54 -05001503 btrfs_queue_worker(&fs_info->endio_meta_workers,
Chris Mason8b712842008-06-11 16:50:36 -04001504 &end_io_wq->work);
1505 return;
1506 }
1507 error = end_io_wq->error;
1508 bio->bi_private = end_io_wq->private;
1509 bio->bi_end_io = end_io_wq->end_io;
1510 kfree(end_io_wq);
Chris Mason8b712842008-06-11 16:50:36 -04001511 bio_endio(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001512}
1513
Chris Masona74a4b92008-06-25 16:01:31 -04001514static int cleaner_kthread(void *arg)
1515{
1516 struct btrfs_root *root = arg;
1517
1518 do {
Chris Masona74a4b92008-06-25 16:01:31 -04001519 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
Yan, Zheng76dda932009-09-21 16:00:26 -04001520
1521 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1522 mutex_trylock(&root->fs_info->cleaner_mutex)) {
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001523 btrfs_run_delayed_iputs(root);
Yan, Zheng76dda932009-09-21 16:00:26 -04001524 btrfs_clean_old_snapshots(root);
1525 mutex_unlock(&root->fs_info->cleaner_mutex);
1526 }
Chris Masona74a4b92008-06-25 16:01:31 -04001527
1528 if (freezing(current)) {
1529 refrigerator();
1530 } else {
Chris Masona74a4b92008-06-25 16:01:31 -04001531 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001532 if (!kthread_should_stop())
1533 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001534 __set_current_state(TASK_RUNNING);
1535 }
1536 } while (!kthread_should_stop());
1537 return 0;
1538}
1539
1540static int transaction_kthread(void *arg)
1541{
1542 struct btrfs_root *root = arg;
1543 struct btrfs_trans_handle *trans;
1544 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001545 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001546 unsigned long now;
1547 unsigned long delay;
1548 int ret;
1549
1550 do {
Chris Masona74a4b92008-06-25 16:01:31 -04001551 delay = HZ * 30;
1552 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1553 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1554
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001555 spin_lock(&root->fs_info->new_trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001556 cur = root->fs_info->running_transaction;
1557 if (!cur) {
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001558 spin_unlock(&root->fs_info->new_trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001559 goto sleep;
1560 }
Yan Zheng31153d82008-07-28 15:32:19 -04001561
Chris Masona74a4b92008-06-25 16:01:31 -04001562 now = get_seconds();
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001563 if (!cur->blocked &&
1564 (now < cur->start_time || now - cur->start_time < 30)) {
1565 spin_unlock(&root->fs_info->new_trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001566 delay = HZ * 5;
1567 goto sleep;
1568 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001569 transid = cur->transid;
1570 spin_unlock(&root->fs_info->new_trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001571
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001572 trans = btrfs_join_transaction(root, 1);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00001573 BUG_ON(IS_ERR(trans));
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001574 if (transid == trans->transid) {
1575 ret = btrfs_commit_transaction(trans, root);
1576 BUG_ON(ret);
1577 } else {
1578 btrfs_end_transaction(trans, root);
1579 }
Chris Masona74a4b92008-06-25 16:01:31 -04001580sleep:
1581 wake_up_process(root->fs_info->cleaner_kthread);
1582 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1583
1584 if (freezing(current)) {
1585 refrigerator();
1586 } else {
Chris Masona74a4b92008-06-25 16:01:31 -04001587 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001588 if (!kthread_should_stop() &&
1589 !btrfs_transaction_blocked(root->fs_info))
1590 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001591 __set_current_state(TASK_RUNNING);
1592 }
1593 } while (!kthread_should_stop());
1594 return 0;
1595}
1596
Chris Mason8a4b83c2008-03-24 15:02:07 -04001597struct btrfs_root *open_ctree(struct super_block *sb,
Chris Masondfe25022008-05-13 13:46:40 -04001598 struct btrfs_fs_devices *fs_devices,
1599 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05001600{
Chris Masondb945352007-10-15 16:15:53 -04001601 u32 sectorsize;
1602 u32 nodesize;
1603 u32 leafsize;
1604 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001605 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04001606 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001607 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05001608 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04001609 struct buffer_head *bh;
Chris Masone02119d2008-09-05 16:13:11 -04001610 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
Chris Masone20d96d2007-03-22 12:13:20 -04001611 GFP_NOFS);
Chris Masond20f7042008-12-08 16:58:54 -05001612 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1613 GFP_NOFS);
Josef Bacik450ba0e2010-11-19 14:59:15 -05001614 struct btrfs_root *tree_root = btrfs_sb(sb);
1615 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Masone02119d2008-09-05 16:13:11 -04001616 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
Chris Mason0b86a832008-03-24 15:01:56 -04001617 GFP_NOFS);
Chris Masone02119d2008-09-05 16:13:11 -04001618 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
Chris Mason0b86a832008-03-24 15:01:56 -04001619 GFP_NOFS);
Chris Masone02119d2008-09-05 16:13:11 -04001620 struct btrfs_root *log_tree_root;
1621
Chris Masoneb60cea2007-02-02 09:18:22 -05001622 int ret;
Yane58ca022008-04-01 11:21:34 -04001623 int err = -EINVAL;
Chris Mason4543df72008-06-11 21:47:56 -04001624
Chris Mason2c90e5d2007-04-02 10:50:19 -04001625 struct btrfs_super_block *disk_super;
Chris Mason8790d502008-04-03 16:29:03 -04001626
Jim Meyering0463bb42008-10-01 19:09:04 -04001627 if (!extent_root || !tree_root || !fs_info ||
Chris Masond20f7042008-12-08 16:58:54 -05001628 !chunk_root || !dev_root || !csum_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04001629 err = -ENOMEM;
1630 goto fail;
1631 }
Yan, Zheng76dda932009-09-21 16:00:26 -04001632
1633 ret = init_srcu_struct(&fs_info->subvol_srcu);
1634 if (ret) {
1635 err = ret;
1636 goto fail;
1637 }
1638
1639 ret = setup_bdi(fs_info, &fs_info->bdi);
1640 if (ret) {
1641 err = ret;
1642 goto fail_srcu;
1643 }
1644
1645 fs_info->btree_inode = new_inode(sb);
1646 if (!fs_info->btree_inode) {
1647 err = -ENOMEM;
1648 goto fail_bdi;
1649 }
1650
Miao Xie1561ded2011-03-27 16:07:36 +08001651 fs_info->btree_inode->i_mapping->flags &= ~__GFP_FS;
1652
Yan, Zheng76dda932009-09-21 16:00:26 -04001653 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04001654 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04001655 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001656 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Chris Mason19c00dd2007-10-15 16:19:22 -04001657 INIT_LIST_HEAD(&fs_info->hashers);
Chris Masonea8c2812008-08-04 23:17:27 -04001658 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
Chris Mason5a3f23d2009-03-31 13:27:11 -04001659 INIT_LIST_HEAD(&fs_info->ordered_operations);
Yan Zheng11833d62009-09-11 16:11:19 -04001660 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Chris Mason1832a6d2007-12-21 16:27:21 -05001661 spin_lock_init(&fs_info->delalloc_lock);
Chris Masoncee36a02008-01-15 08:40:48 -05001662 spin_lock_init(&fs_info->new_trans_lock);
Yan Zheng31153d82008-07-28 15:32:19 -04001663 spin_lock_init(&fs_info->ref_cache_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04001664 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001665 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04001666
Josef Bacik58176a92007-08-29 15:47:34 -04001667 init_completion(&fs_info->kobj_unregister);
Chris Mason9f5fae22007-03-20 14:38:32 -04001668 fs_info->tree_root = tree_root;
1669 fs_info->extent_root = extent_root;
Chris Masond20f7042008-12-08 16:58:54 -05001670 fs_info->csum_root = csum_root;
Chris Mason0b86a832008-03-24 15:01:56 -04001671 fs_info->chunk_root = chunk_root;
1672 fs_info->dev_root = dev_root;
Chris Mason8a4b83c2008-03-24 15:02:07 -04001673 fs_info->fs_devices = fs_devices;
Chris Mason0b86a832008-03-24 15:01:56 -04001674 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04001675 INIT_LIST_HEAD(&fs_info->space_info);
Chris Mason0b86a832008-03-24 15:01:56 -04001676 btrfs_mapping_init(&fs_info->mapping_tree);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001677 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1678 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1679 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1680 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1681 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
1682 INIT_LIST_HEAD(&fs_info->durable_block_rsv_list);
1683 mutex_init(&fs_info->durable_block_rsv_mutex);
Chris Masoncb03c742008-05-15 16:15:45 -04001684 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05001685 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04001686 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04001687 atomic_set(&fs_info->nr_async_bios, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04001688 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05001689 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04001690 fs_info->metadata_ratio = 0;
Chris Masonc8b97812008-10-29 14:49:59 -04001691
Chris Masonb34b0862008-12-19 15:43:22 -05001692 fs_info->thread_pool_size = min_t(unsigned long,
1693 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04001694
Chris Mason3eaa2882008-07-24 11:57:52 -04001695 INIT_LIST_HEAD(&fs_info->ordered_extents);
1696 spin_lock_init(&fs_info->ordered_extent_lock);
1697
Chris Masona061fc82008-05-07 11:43:44 -04001698 sb->s_blocksize = 4096;
1699 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02001700 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04001701
Yan, Zheng76dda932009-09-21 16:00:26 -04001702 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
1703 fs_info->btree_inode->i_nlink = 1;
Chris Mason0afbaf82008-04-18 14:17:20 -04001704 /*
1705 * we set the i_size on the btree inode to the max possible int.
1706 * the real end of the address space is determined by all of
1707 * the devices in the system
1708 */
1709 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04001710 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04001711 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1712
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001713 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05001714 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
Chris Mason5f39d392007-10-15 16:14:19 -04001715 fs_info->btree_inode->i_mapping,
1716 GFP_NOFS);
Chris Masond1310b22008-01-24 16:13:08 -05001717 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1718 GFP_NOFS);
Chris Mason0da54682007-11-07 21:08:01 -05001719
Chris Masond1310b22008-01-24 16:13:08 -05001720 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1721
Chris Mason0f7d52f2007-04-09 10:42:37 -04001722 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1723 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1724 sizeof(struct btrfs_key));
Yan, Zheng76dda932009-09-21 16:00:26 -04001725 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
Chris Mason22b0ebd2007-03-30 08:47:31 -04001726 insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04001727
Chris Masone02119d2008-09-05 16:13:11 -04001728 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00001729 fs_info->block_group_cache_tree = RB_ROOT;
Chris Mason925baed2008-06-25 16:01:30 -04001730
Yan Zheng11833d62009-09-11 16:11:19 -04001731 extent_io_tree_init(&fs_info->freed_extents[0],
Chris Mason7d9eb122008-07-08 14:19:17 -04001732 fs_info->btree_inode->i_mapping, GFP_NOFS);
Yan Zheng11833d62009-09-11 16:11:19 -04001733 extent_io_tree_init(&fs_info->freed_extents[1],
1734 fs_info->btree_inode->i_mapping, GFP_NOFS);
1735 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04001736 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04001737
Chris Masond98237b2007-03-28 13:57:48 -04001738
Chris Mason509659c2007-05-10 12:36:17 -04001739 mutex_init(&fs_info->trans_mutex);
Chris Mason5a3f23d2009-03-31 13:27:11 -04001740 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04001741 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04001742 mutex_init(&fs_info->chunk_mutex);
1743 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04001744 mutex_init(&fs_info->cleaner_mutex);
1745 mutex_init(&fs_info->volume_mutex);
Yan Zheng276e6802009-07-30 09:40:40 -04001746 init_rwsem(&fs_info->extent_commit_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00001747 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04001748 init_rwsem(&fs_info->subvol_sem);
Chris Masonfa9c0d792009-04-03 09:47:43 -04001749
1750 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
1751 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
1752
Chris Mason3768f362007-03-13 16:47:54 -04001753 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason19c00dd2007-10-15 16:19:22 -04001754 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04001755 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04001756 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05001757
Chris Mason0b86a832008-03-24 15:01:56 -04001758 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04001759 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04001760
Yan Zhenga512bbf2008-12-08 16:46:26 -05001761 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00001762 if (!bh) {
1763 err = -EINVAL;
Chris Mason39279cc2007-06-12 06:35:45 -04001764 goto fail_iput;
Dave Young20b45072011-01-08 10:09:13 +00001765 }
Chris Mason39279cc2007-06-12 06:35:45 -04001766
Chris Masona061fc82008-05-07 11:43:44 -04001767 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
Yan Zheng2d69a0f2009-01-05 15:43:42 -05001768 memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
1769 sizeof(fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04001770 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04001771
Chris Masona061fc82008-05-07 11:43:44 -04001772 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04001773
Chris Mason5f39d392007-10-15 16:14:19 -04001774 disk_super = &fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001775 if (!btrfs_super_root(disk_super))
Josef Bacikc6e2bac2008-12-02 06:36:10 -05001776 goto fail_iput;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001777
liuboacce9522011-01-06 19:30:25 +08001778 /* check FS state, whether FS is broken. */
1779 fs_info->fs_state |= btrfs_super_flags(disk_super);
1780
1781 btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
1782
Liu Bo75e7cb72011-03-22 10:12:20 +00001783 /*
1784 * In the long term, we'll store the compression type in the super
1785 * block, and it'll be used for per file compression control.
1786 */
1787 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
1788
Yan Zheng2b820322008-11-17 21:11:30 -05001789 ret = btrfs_parse_options(tree_root, options);
1790 if (ret) {
1791 err = ret;
Josef Bacikc6e2bac2008-12-02 06:36:10 -05001792 goto fail_iput;
Yan Zheng2b820322008-11-17 21:11:30 -05001793 }
Chris Masondfe25022008-05-13 13:46:40 -04001794
Josef Bacikf2b636e2008-12-02 06:36:08 -05001795 features = btrfs_super_incompat_flags(disk_super) &
1796 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1797 if (features) {
1798 printk(KERN_ERR "BTRFS: couldn't mount because of "
1799 "unsupported optional features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07001800 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05001801 err = -EINVAL;
Josef Bacikc6e2bac2008-12-02 06:36:10 -05001802 goto fail_iput;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001803 }
1804
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001805 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08001806 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
1807 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
1808 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1809 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001810
Josef Bacikf2b636e2008-12-02 06:36:08 -05001811 features = btrfs_super_compat_ro_flags(disk_super) &
1812 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1813 if (!(sb->s_flags & MS_RDONLY) && features) {
1814 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1815 "unsupported option features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07001816 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05001817 err = -EINVAL;
Josef Bacikc6e2bac2008-12-02 06:36:10 -05001818 goto fail_iput;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001819 }
Chris Mason61d92c32009-10-02 19:11:56 -04001820
1821 btrfs_init_workers(&fs_info->generic_worker,
1822 "genwork", 1, NULL);
1823
Chris Mason5443be42008-08-15 15:34:16 -04001824 btrfs_init_workers(&fs_info->workers, "worker",
Chris Mason61d92c32009-10-02 19:11:56 -04001825 fs_info->thread_pool_size,
1826 &fs_info->generic_worker);
Chris Masonc8b97812008-10-29 14:49:59 -04001827
Chris Mason771ed682008-11-06 22:02:51 -05001828 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
Chris Mason61d92c32009-10-02 19:11:56 -04001829 fs_info->thread_pool_size,
1830 &fs_info->generic_worker);
Chris Mason771ed682008-11-06 22:02:51 -05001831
Chris Mason5443be42008-08-15 15:34:16 -04001832 btrfs_init_workers(&fs_info->submit_workers, "submit",
Chris Masonb720d202008-08-15 15:34:14 -04001833 min_t(u64, fs_devices->num_devices,
Chris Mason61d92c32009-10-02 19:11:56 -04001834 fs_info->thread_pool_size),
1835 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04001836
1837 /* a higher idle thresh on the submit workers makes it much more
1838 * likely that bios will be send down in a sane order to the
1839 * devices
1840 */
1841 fs_info->submit_workers.idle_thresh = 64;
Chris Mason53863232008-08-15 15:34:18 -04001842
Chris Mason771ed682008-11-06 22:02:51 -05001843 fs_info->workers.idle_thresh = 16;
Chris Mason4a69a412008-11-06 22:03:00 -05001844 fs_info->workers.ordered = 1;
Chris Mason61b49442008-07-31 15:42:53 -04001845
Chris Mason771ed682008-11-06 22:02:51 -05001846 fs_info->delalloc_workers.idle_thresh = 2;
1847 fs_info->delalloc_workers.ordered = 1;
1848
Chris Mason61d92c32009-10-02 19:11:56 -04001849 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1850 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04001851 btrfs_init_workers(&fs_info->endio_workers, "endio",
Chris Mason61d92c32009-10-02 19:11:56 -04001852 fs_info->thread_pool_size,
1853 &fs_info->generic_worker);
Chris Masond20f7042008-12-08 16:58:54 -05001854 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
Chris Mason61d92c32009-10-02 19:11:56 -04001855 fs_info->thread_pool_size,
1856 &fs_info->generic_worker);
Chris Masoncad321a2008-12-17 14:51:42 -05001857 btrfs_init_workers(&fs_info->endio_meta_write_workers,
Chris Mason61d92c32009-10-02 19:11:56 -04001858 "endio-meta-write", fs_info->thread_pool_size,
1859 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04001860 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
Chris Mason61d92c32009-10-02 19:11:56 -04001861 fs_info->thread_pool_size,
1862 &fs_info->generic_worker);
Josef Bacik0cb59c92010-07-02 12:14:14 -04001863 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
1864 1, &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04001865
1866 /*
1867 * endios are largely parallel and should have a very
1868 * low idle thresh
1869 */
1870 fs_info->endio_workers.idle_thresh = 4;
Chris Masonb51912c2009-02-04 09:23:24 -05001871 fs_info->endio_meta_workers.idle_thresh = 4;
1872
Chris Mason90428462009-08-04 16:56:34 -04001873 fs_info->endio_write_workers.idle_thresh = 2;
1874 fs_info->endio_meta_write_workers.idle_thresh = 2;
1875
Chris Mason4543df72008-06-11 21:47:56 -04001876 btrfs_start_workers(&fs_info->workers, 1);
Chris Mason61d92c32009-10-02 19:11:56 -04001877 btrfs_start_workers(&fs_info->generic_worker, 1);
Chris Mason1cc127b2008-06-12 14:46:17 -04001878 btrfs_start_workers(&fs_info->submit_workers, 1);
Chris Mason771ed682008-11-06 22:02:51 -05001879 btrfs_start_workers(&fs_info->delalloc_workers, 1);
Chris Mason247e7432008-07-17 12:53:51 -04001880 btrfs_start_workers(&fs_info->fixup_workers, 1);
Chris Mason90428462009-08-04 16:56:34 -04001881 btrfs_start_workers(&fs_info->endio_workers, 1);
1882 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1883 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1884 btrfs_start_workers(&fs_info->endio_write_workers, 1);
Josef Bacik0cb59c92010-07-02 12:14:14 -04001885 btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
Chris Mason4543df72008-06-11 21:47:56 -04001886
Chris Mason4575c9c2008-04-18 16:13:31 -04001887 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04001888 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1889 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04001890
Chris Masondb945352007-10-15 16:15:53 -04001891 nodesize = btrfs_super_nodesize(disk_super);
1892 leafsize = btrfs_super_leafsize(disk_super);
1893 sectorsize = btrfs_super_sectorsize(disk_super);
Chris Mason87ee04e2007-11-30 11:30:34 -05001894 stripesize = btrfs_super_stripesize(disk_super);
Chris Masondb945352007-10-15 16:15:53 -04001895 tree_root->nodesize = nodesize;
1896 tree_root->leafsize = leafsize;
1897 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001898 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04001899
1900 sb->s_blocksize = sectorsize;
1901 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04001902
Chris Mason39279cc2007-06-12 06:35:45 -04001903 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1904 sizeof(disk_super->magic))) {
Chris Masond3977122009-01-05 21:25:51 -05001905 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04001906 goto fail_sb_buffer;
1907 }
Chris Mason19c00dd2007-10-15 16:19:22 -04001908
Chris Mason925baed2008-06-25 16:01:30 -04001909 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05001910 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04001911 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04001912 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05001913 printk(KERN_WARNING "btrfs: failed to read the system "
1914 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001915 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04001916 }
Chris Mason0b86a832008-03-24 15:01:56 -04001917
1918 blocksize = btrfs_level_size(tree_root,
1919 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04001920 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04001921
1922 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1923 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1924
1925 chunk_root->node = read_tree_block(chunk_root,
1926 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04001927 blocksize, generation);
Chris Mason0b86a832008-03-24 15:01:56 -04001928 BUG_ON(!chunk_root->node);
David Woodhouse83121942009-07-22 16:52:13 -04001929 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
1930 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
1931 sb->s_id);
1932 goto fail_chunk_root;
1933 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001934 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
1935 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04001936
Chris Masone17cade2008-04-15 15:41:47 -04001937 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Chris Masond3977122009-01-05 21:25:51 -05001938 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1939 BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04001940
Chris Mason925baed2008-06-25 16:01:30 -04001941 mutex_lock(&fs_info->chunk_mutex);
Chris Mason0b86a832008-03-24 15:01:56 -04001942 ret = btrfs_read_chunk_tree(chunk_root);
Chris Mason925baed2008-06-25 16:01:30 -04001943 mutex_unlock(&fs_info->chunk_mutex);
Yan Zheng2b820322008-11-17 21:11:30 -05001944 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05001945 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1946 sb->s_id);
Yan Zheng2b820322008-11-17 21:11:30 -05001947 goto fail_chunk_root;
1948 }
Chris Mason0b86a832008-03-24 15:01:56 -04001949
Chris Masondfe25022008-05-13 13:46:40 -04001950 btrfs_close_extra_devices(fs_devices);
1951
Chris Masondb945352007-10-15 16:15:53 -04001952 blocksize = btrfs_level_size(tree_root,
1953 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04001954 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04001955
Chris Masone20d96d2007-03-22 12:13:20 -04001956 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04001957 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04001958 blocksize, generation);
Chris Mason39279cc2007-06-12 06:35:45 -04001959 if (!tree_root->node)
Yan Zheng2b820322008-11-17 21:11:30 -05001960 goto fail_chunk_root;
David Woodhouse83121942009-07-22 16:52:13 -04001961 if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
1962 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
1963 sb->s_id);
1964 goto fail_tree_root;
1965 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001966 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
1967 tree_root->commit_root = btrfs_root_node(tree_root);
Chris Masondb945352007-10-15 16:15:53 -04001968
1969 ret = find_and_setup_root(tree_root, fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04001970 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
Chris Mason0b86a832008-03-24 15:01:56 -04001971 if (ret)
Chris Mason39279cc2007-06-12 06:35:45 -04001972 goto fail_tree_root;
Chris Mason0b86a832008-03-24 15:01:56 -04001973 extent_root->track_dirty = 1;
1974
1975 ret = find_and_setup_root(tree_root, fs_info,
1976 BTRFS_DEV_TREE_OBJECTID, dev_root);
Chris Mason0b86a832008-03-24 15:01:56 -04001977 if (ret)
1978 goto fail_extent_root;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001979 dev_root->track_dirty = 1;
Chris Mason3768f362007-03-13 16:47:54 -04001980
Chris Masond20f7042008-12-08 16:58:54 -05001981 ret = find_and_setup_root(tree_root, fs_info,
1982 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1983 if (ret)
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001984 goto fail_dev_root;
Chris Masond20f7042008-12-08 16:58:54 -05001985
1986 csum_root->track_dirty = 1;
1987
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001988 fs_info->generation = generation;
1989 fs_info->last_trans_committed = generation;
1990 fs_info->data_alloc_profile = (u64)-1;
1991 fs_info->metadata_alloc_profile = (u64)-1;
1992 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1993
liuboc59021f2011-03-07 02:13:14 +00001994 ret = btrfs_init_space_info(fs_info);
1995 if (ret) {
1996 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
1997 goto fail_block_groups;
1998 }
1999
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002000 ret = btrfs_read_block_groups(extent_root);
2001 if (ret) {
2002 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2003 goto fail_block_groups;
2004 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002005
Chris Masona74a4b92008-06-25 16:01:31 -04002006 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2007 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002008 if (IS_ERR(fs_info->cleaner_kthread))
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002009 goto fail_block_groups;
Chris Masona74a4b92008-06-25 16:01:31 -04002010
2011 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2012 tree_root,
2013 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002014 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002015 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002016
Chris Masonc2898112009-06-10 09:51:32 -04002017 if (!btrfs_test_opt(tree_root, SSD) &&
2018 !btrfs_test_opt(tree_root, NOSSD) &&
2019 !fs_info->fs_devices->rotating) {
2020 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2021 "mode\n");
2022 btrfs_set_opt(fs_info->mount_opt, SSD);
2023 }
2024
liuboacce9522011-01-06 19:30:25 +08002025 /* do not make disk changes in broken FS */
2026 if (btrfs_super_log_root(disk_super) != 0 &&
2027 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
Chris Masone02119d2008-09-05 16:13:11 -04002028 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002029
Chris Mason7c2ca462008-11-19 15:13:35 -05002030 if (fs_devices->rw_devices == 0) {
Chris Masond3977122009-01-05 21:25:51 -05002031 printk(KERN_WARNING "Btrfs log replay required "
2032 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002033 err = -EIO;
2034 goto fail_trans_kthread;
2035 }
Chris Masone02119d2008-09-05 16:13:11 -04002036 blocksize =
2037 btrfs_level_size(tree_root,
2038 btrfs_super_log_root_level(disk_super));
2039
Dan Carpenter676e4c82010-05-29 09:43:07 +00002040 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
2041 if (!log_tree_root) {
2042 err = -ENOMEM;
2043 goto fail_trans_kthread;
2044 }
Chris Masone02119d2008-09-05 16:13:11 -04002045
2046 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2047 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2048
2049 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002050 blocksize,
2051 generation + 1);
Chris Masone02119d2008-09-05 16:13:11 -04002052 ret = btrfs_recover_log_trees(log_tree_root);
2053 BUG_ON(ret);
Yan Zhenge556ce22008-11-20 10:25:19 -05002054
2055 if (sb->s_flags & MS_RDONLY) {
2056 ret = btrfs_commit_super(tree_root);
2057 BUG_ON(ret);
2058 }
Chris Masone02119d2008-09-05 16:13:11 -04002059 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002060
Yan, Zheng76dda932009-09-21 16:00:26 -04002061 ret = btrfs_find_orphan_roots(tree_root);
2062 BUG_ON(ret);
2063
Chris Mason7c2ca462008-11-19 15:13:35 -05002064 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002065 ret = btrfs_cleanup_fs_roots(fs_info);
2066 BUG_ON(ret);
2067
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002068 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002069 if (ret < 0) {
2070 printk(KERN_WARNING
2071 "btrfs: failed to recover relocation\n");
2072 err = -EINVAL;
2073 goto fail_trans_kthread;
2074 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002075 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002076
Chris Mason3de45862008-11-17 21:02:50 -05002077 location.objectid = BTRFS_FS_TREE_OBJECTID;
2078 location.type = BTRFS_ROOT_ITEM_KEY;
2079 location.offset = (u64)-1;
2080
Chris Mason3de45862008-11-17 21:02:50 -05002081 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2082 if (!fs_info->fs_root)
Chris Mason7c2ca462008-11-19 15:13:35 -05002083 goto fail_trans_kthread;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002084 if (IS_ERR(fs_info->fs_root)) {
2085 err = PTR_ERR(fs_info->fs_root);
2086 goto fail_trans_kthread;
2087 }
Chris Masonc2898112009-06-10 09:51:32 -04002088
Josef Bacike3acc2a2010-01-26 14:30:53 +00002089 if (!(sb->s_flags & MS_RDONLY)) {
2090 down_read(&fs_info->cleanup_work_sem);
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002091 err = btrfs_orphan_cleanup(fs_info->fs_root);
2092 if (!err)
2093 err = btrfs_orphan_cleanup(fs_info->tree_root);
Josef Bacike3acc2a2010-01-26 14:30:53 +00002094 up_read(&fs_info->cleanup_work_sem);
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002095 if (err) {
2096 close_ctree(tree_root);
2097 return ERR_PTR(err);
2098 }
Josef Bacike3acc2a2010-01-26 14:30:53 +00002099 }
2100
Chris Mason0f7d52f2007-04-09 10:42:37 -04002101 return tree_root;
Chris Mason39279cc2007-06-12 06:35:45 -04002102
Chris Mason7c2ca462008-11-19 15:13:35 -05002103fail_trans_kthread:
2104 kthread_stop(fs_info->transaction_kthread);
Chris Mason3f157a22008-06-25 16:01:31 -04002105fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002106 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002107
2108 /*
2109 * make sure we're done with the btree inode before we stop our
2110 * kthreads
2111 */
2112 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2113 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2114
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002115fail_block_groups:
2116 btrfs_free_block_groups(fs_info);
Chris Masond20f7042008-12-08 16:58:54 -05002117 free_extent_buffer(csum_root->node);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002118 free_extent_buffer(csum_root->commit_root);
2119fail_dev_root:
2120 free_extent_buffer(dev_root->node);
2121 free_extent_buffer(dev_root->commit_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002122fail_extent_root:
2123 free_extent_buffer(extent_root->node);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002124 free_extent_buffer(extent_root->commit_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002125fail_tree_root:
Chris Mason5f39d392007-10-15 16:14:19 -04002126 free_extent_buffer(tree_root->node);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002127 free_extent_buffer(tree_root->commit_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002128fail_chunk_root:
2129 free_extent_buffer(chunk_root->node);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002130 free_extent_buffer(chunk_root->commit_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002131fail_sb_buffer:
Chris Mason61d92c32009-10-02 19:11:56 -04002132 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04002133 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002134 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002135 btrfs_stop_workers(&fs_info->workers);
2136 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002137 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002138 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002139 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002140 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002141 btrfs_stop_workers(&fs_info->submit_workers);
Chris Mason4543df72008-06-11 21:47:56 -04002142fail_iput:
Chris Mason7c2ca462008-11-19 15:13:35 -05002143 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Mason4543df72008-06-11 21:47:56 -04002144 iput(fs_info->btree_inode);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002145
Chris Masondfe25022008-05-13 13:46:40 -04002146 btrfs_close_devices(fs_info->fs_devices);
Chris Mason84eed902008-04-25 09:04:37 -04002147 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Jens Axboead081f12009-06-12 14:43:40 +02002148fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05002149 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002150fail_srcu:
2151 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002152fail:
Chris Mason39279cc2007-06-12 06:35:45 -04002153 kfree(extent_root);
2154 kfree(tree_root);
2155 kfree(fs_info);
Jim Meyering83afeac2008-10-01 19:09:51 -04002156 kfree(chunk_root);
2157 kfree(dev_root);
Chris Masond20f7042008-12-08 16:58:54 -05002158 kfree(csum_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002159 return ERR_PTR(err);
Chris Masoneb60cea2007-02-02 09:18:22 -05002160}
2161
Chris Masonf2984462008-04-10 16:19:33 -04002162static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2163{
2164 char b[BDEVNAME_SIZE];
2165
2166 if (uptodate) {
2167 set_buffer_uptodate(bh);
2168 } else {
Christoph Hellwigc3b9a622010-08-18 05:29:12 -04002169 if (printk_ratelimit()) {
Chris Masonf2984462008-04-10 16:19:33 -04002170 printk(KERN_WARNING "lost page write due to "
2171 "I/O error on %s\n",
2172 bdevname(bh->b_bdev, b));
2173 }
Chris Mason1259ab72008-05-12 13:39:03 -04002174 /* note, we dont' set_buffer_write_io_error because we have
2175 * our own ways of dealing with the IO errors
2176 */
Chris Masonf2984462008-04-10 16:19:33 -04002177 clear_buffer_uptodate(bh);
2178 }
2179 unlock_buffer(bh);
2180 put_bh(bh);
2181}
2182
Yan Zhenga512bbf2008-12-08 16:46:26 -05002183struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2184{
2185 struct buffer_head *bh;
2186 struct buffer_head *latest = NULL;
2187 struct btrfs_super_block *super;
2188 int i;
2189 u64 transid = 0;
2190 u64 bytenr;
2191
2192 /* we would like to check all the supers, but that would make
2193 * a btrfs mount succeed after a mkfs from a different FS.
2194 * So, we need to add a special mount option to scan for
2195 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2196 */
2197 for (i = 0; i < 1; i++) {
2198 bytenr = btrfs_sb_offset(i);
2199 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2200 break;
2201 bh = __bread(bdev, bytenr / 4096, 4096);
2202 if (!bh)
2203 continue;
2204
2205 super = (struct btrfs_super_block *)bh->b_data;
2206 if (btrfs_super_bytenr(super) != bytenr ||
2207 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2208 sizeof(super->magic))) {
2209 brelse(bh);
2210 continue;
2211 }
2212
2213 if (!latest || btrfs_super_generation(super) > transid) {
2214 brelse(latest);
2215 latest = bh;
2216 transid = btrfs_super_generation(super);
2217 } else {
2218 brelse(bh);
2219 }
2220 }
2221 return latest;
2222}
2223
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002224/*
2225 * this should be called twice, once with wait == 0 and
2226 * once with wait == 1. When wait == 0 is done, all the buffer heads
2227 * we write are pinned.
2228 *
2229 * They are released when wait == 1 is done.
2230 * max_mirrors must be the same for both runs, and it indicates how
2231 * many supers on this one device should be written.
2232 *
2233 * max_mirrors == 0 means to write them all.
2234 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002235static int write_dev_supers(struct btrfs_device *device,
2236 struct btrfs_super_block *sb,
2237 int do_barriers, int wait, int max_mirrors)
2238{
2239 struct buffer_head *bh;
2240 int i;
2241 int ret;
2242 int errors = 0;
2243 u32 crc;
2244 u64 bytenr;
2245 int last_barrier = 0;
2246
2247 if (max_mirrors == 0)
2248 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2249
2250 /* make sure only the last submit_bh does a barrier */
2251 if (do_barriers) {
2252 for (i = 0; i < max_mirrors; i++) {
2253 bytenr = btrfs_sb_offset(i);
2254 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2255 device->total_bytes)
2256 break;
2257 last_barrier = i;
2258 }
2259 }
2260
2261 for (i = 0; i < max_mirrors; i++) {
2262 bytenr = btrfs_sb_offset(i);
2263 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2264 break;
2265
2266 if (wait) {
2267 bh = __find_get_block(device->bdev, bytenr / 4096,
2268 BTRFS_SUPER_INFO_SIZE);
2269 BUG_ON(!bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002270 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002271 if (!buffer_uptodate(bh))
2272 errors++;
2273
2274 /* drop our reference */
2275 brelse(bh);
2276
2277 /* drop the reference from the wait == 0 run */
2278 brelse(bh);
2279 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002280 } else {
2281 btrfs_set_super_bytenr(sb, bytenr);
2282
2283 crc = ~(u32)0;
2284 crc = btrfs_csum_data(NULL, (char *)sb +
2285 BTRFS_CSUM_SIZE, crc,
2286 BTRFS_SUPER_INFO_SIZE -
2287 BTRFS_CSUM_SIZE);
2288 btrfs_csum_final(crc, sb->csum);
2289
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002290 /*
2291 * one reference for us, and we leave it for the
2292 * caller
2293 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002294 bh = __getblk(device->bdev, bytenr / 4096,
2295 BTRFS_SUPER_INFO_SIZE);
2296 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2297
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002298 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002299 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002300
2301 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002302 lock_buffer(bh);
2303 bh->b_end_io = btrfs_end_buffer_write_sync;
2304 }
2305
Christoph Hellwigc3b9a622010-08-18 05:29:12 -04002306 if (i == last_barrier && do_barriers)
2307 ret = submit_bh(WRITE_FLUSH_FUA, bh);
2308 else
Chris Masonffbd5172009-04-20 15:50:09 -04002309 ret = submit_bh(WRITE_SYNC, bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002310
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002311 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05002312 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002313 }
2314 return errors < i ? 0 : -1;
2315}
2316
2317int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04002318{
Chris Masone5e9a522009-06-10 15:17:02 -04002319 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04002320 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04002321 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04002322 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04002323 int ret;
2324 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04002325 int max_errors;
2326 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04002327 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04002328
Chris Masona236aed2008-04-29 09:38:00 -04002329 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
Chris Masonf2984462008-04-10 16:19:33 -04002330 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2331
Chris Masona061fc82008-05-07 11:43:44 -04002332 sb = &root->fs_info->super_for_commit;
2333 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04002334
Xiao Guangrong1f781602011-04-20 10:09:16 +00002335 rcu_read_lock();
Chris Masone5e9a522009-06-10 15:17:02 -04002336 head = &root->fs_info->fs_devices->devices;
Xiao Guangrong1f781602011-04-20 10:09:16 +00002337 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002338 if (!dev->bdev) {
2339 total_errors++;
2340 continue;
2341 }
Yan Zheng2b820322008-11-17 21:11:30 -05002342 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002343 continue;
2344
Yan Zheng2b820322008-11-17 21:11:30 -05002345 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04002346 btrfs_set_stack_device_type(dev_item, dev->type);
2347 btrfs_set_stack_device_id(dev_item, dev->devid);
2348 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2349 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2350 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2351 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2352 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2353 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05002354 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002355
Chris Masona061fc82008-05-07 11:43:44 -04002356 flags = btrfs_super_flags(sb);
2357 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04002358
Yan Zhenga512bbf2008-12-08 16:46:26 -05002359 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04002360 if (ret)
2361 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04002362 }
Chris Masona236aed2008-04-29 09:38:00 -04002363 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05002364 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2365 total_errors);
Chris Masona236aed2008-04-29 09:38:00 -04002366 BUG();
2367 }
Chris Masonf2984462008-04-10 16:19:33 -04002368
Yan Zhenga512bbf2008-12-08 16:46:26 -05002369 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00002370 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002371 if (!dev->bdev)
2372 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05002373 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002374 continue;
2375
Yan Zhenga512bbf2008-12-08 16:46:26 -05002376 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2377 if (ret)
2378 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04002379 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00002380 rcu_read_unlock();
Chris Masona236aed2008-04-29 09:38:00 -04002381 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05002382 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2383 total_errors);
Chris Masona236aed2008-04-29 09:38:00 -04002384 BUG();
2385 }
Chris Masonf2984462008-04-10 16:19:33 -04002386 return 0;
2387}
2388
Yan Zhenga512bbf2008-12-08 16:46:26 -05002389int write_ctree_super(struct btrfs_trans_handle *trans,
2390 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05002391{
Chris Masone66f7092007-04-20 13:16:02 -04002392 int ret;
Chris Mason2c90e5d2007-04-02 10:50:19 -04002393
Yan Zhenga512bbf2008-12-08 16:46:26 -05002394 ret = write_all_supers(root, max_mirrors);
Chris Mason5f39d392007-10-15 16:14:19 -04002395 return ret;
Chris Masoncfaa7292007-02-21 17:04:57 -05002396}
2397
Chris Mason5eda7b52007-06-22 14:16:25 -04002398int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04002399{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002400 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04002401 radix_tree_delete(&fs_info->fs_roots_radix,
2402 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002403 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002404
2405 if (btrfs_root_refs(&root->root_item) == 0)
2406 synchronize_srcu(&fs_info->subvol_srcu);
2407
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002408 free_fs_root(root);
2409 return 0;
2410}
2411
2412static void free_fs_root(struct btrfs_root *root)
2413{
2414 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Chris Mason3394e162008-11-17 20:42:26 -05002415 if (root->anon_super.s_dev) {
2416 down_write(&root->anon_super.s_umount);
2417 kill_anon_super(&root->anon_super);
2418 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002419 free_extent_buffer(root->node);
2420 free_extent_buffer(root->commit_root);
Chris Masond3977122009-01-05 21:25:51 -05002421 kfree(root->name);
Chris Mason2619ba12007-04-10 16:58:11 -04002422 kfree(root);
Chris Mason2619ba12007-04-10 16:58:11 -04002423}
2424
Chris Mason35b7e472007-05-02 15:53:43 -04002425static int del_fs_roots(struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04002426{
2427 int ret;
2428 struct btrfs_root *gang[8];
2429 int i;
2430
Yan, Zheng76dda932009-09-21 16:00:26 -04002431 while (!list_empty(&fs_info->dead_roots)) {
2432 gang[0] = list_entry(fs_info->dead_roots.next,
2433 struct btrfs_root, root_list);
2434 list_del(&gang[0]->root_list);
2435
2436 if (gang[0]->in_radix) {
2437 btrfs_free_fs_root(fs_info, gang[0]);
2438 } else {
2439 free_extent_buffer(gang[0]->node);
2440 free_extent_buffer(gang[0]->commit_root);
2441 kfree(gang[0]);
2442 }
2443 }
2444
Chris Masond3977122009-01-05 21:25:51 -05002445 while (1) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04002446 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2447 (void **)gang, 0,
2448 ARRAY_SIZE(gang));
2449 if (!ret)
2450 break;
Chris Mason2619ba12007-04-10 16:58:11 -04002451 for (i = 0; i < ret; i++)
Chris Mason5eda7b52007-06-22 14:16:25 -04002452 btrfs_free_fs_root(fs_info, gang[i]);
Chris Mason0f7d52f2007-04-09 10:42:37 -04002453 }
2454 return 0;
2455}
Chris Masonb4100d62007-04-12 12:14:00 -04002456
Yan Zhengc146afa2008-11-12 14:34:12 -05002457int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
2458{
2459 u64 root_objectid = 0;
2460 struct btrfs_root *gang[8];
2461 int i;
2462 int ret;
2463
2464 while (1) {
2465 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2466 (void **)gang, root_objectid,
2467 ARRAY_SIZE(gang));
2468 if (!ret)
2469 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002470
2471 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05002472 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002473 int err;
2474
Yan Zhengc146afa2008-11-12 14:34:12 -05002475 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002476 err = btrfs_orphan_cleanup(gang[i]);
2477 if (err)
2478 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05002479 }
2480 root_objectid++;
2481 }
2482 return 0;
2483}
2484
2485int btrfs_commit_super(struct btrfs_root *root)
2486{
2487 struct btrfs_trans_handle *trans;
2488 int ret;
2489
2490 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002491 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05002492 btrfs_clean_old_snapshots(root);
2493 mutex_unlock(&root->fs_info->cleaner_mutex);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002494
2495 /* wait until ongoing cleanup work done */
2496 down_write(&root->fs_info->cleanup_work_sem);
2497 up_write(&root->fs_info->cleanup_work_sem);
2498
Yan, Zhenga22285a2010-05-16 10:48:46 -04002499 trans = btrfs_join_transaction(root, 1);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00002500 if (IS_ERR(trans))
2501 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05002502 ret = btrfs_commit_transaction(trans, root);
2503 BUG_ON(ret);
2504 /* run commit again to drop the original snapshot */
Yan, Zhenga22285a2010-05-16 10:48:46 -04002505 trans = btrfs_join_transaction(root, 1);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00002506 if (IS_ERR(trans))
2507 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05002508 btrfs_commit_transaction(trans, root);
2509 ret = btrfs_write_and_wait_transaction(NULL, root);
2510 BUG_ON(ret);
2511
Yan Zhenga512bbf2008-12-08 16:46:26 -05002512 ret = write_ctree_super(NULL, root, 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05002513 return ret;
2514}
2515
Chris Masone20d96d2007-03-22 12:13:20 -04002516int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05002517{
Chris Mason0f7d52f2007-04-09 10:42:37 -04002518 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05002519 int ret;
Chris Masone089f052007-03-16 16:20:31 -04002520
Chris Masonfacda1e2007-06-08 18:11:48 -04002521 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04002522 smp_mb();
2523
Josef Bacik0af3d002010-06-21 14:48:16 -04002524 btrfs_put_block_group_cache(fs_info);
liuboacce9522011-01-06 19:30:25 +08002525
2526 /*
2527 * Here come 2 situations when btrfs is broken to flip readonly:
2528 *
2529 * 1. when btrfs flips readonly somewhere else before
2530 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2531 * and btrfs will skip to write sb directly to keep
2532 * ERROR state on disk.
2533 *
2534 * 2. when btrfs flips readonly just in btrfs_commit_super,
Justin P. Mattockae0e47f2011-03-01 15:06:02 +01002535 * and in such case, btrfs cannot write sb via btrfs_commit_super,
liuboacce9522011-01-06 19:30:25 +08002536 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2537 * btrfs will cleanup all FS resources first and write sb then.
2538 */
Yan Zhengc146afa2008-11-12 14:34:12 -05002539 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08002540 ret = btrfs_commit_super(root);
2541 if (ret)
2542 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
2543 }
2544
2545 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
2546 ret = btrfs_error_commit_super(root);
Chris Masond3977122009-01-05 21:25:51 -05002547 if (ret)
2548 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
Yan Zhengc146afa2008-11-12 14:34:12 -05002549 }
Chris Masoned2ff2c2007-03-01 18:59:40 -05002550
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002551 kthread_stop(root->fs_info->transaction_kthread);
2552 kthread_stop(root->fs_info->cleaner_kthread);
2553
Yan Zhengf25784b2009-07-28 08:41:57 -04002554 fs_info->closing = 2;
2555 smp_mb();
2556
Chris Masonb0c68f82008-01-31 11:05:37 -05002557 if (fs_info->delalloc_bytes) {
Chris Masond3977122009-01-05 21:25:51 -05002558 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07002559 (unsigned long long)fs_info->delalloc_bytes);
Chris Masonb0c68f82008-01-31 11:05:37 -05002560 }
Yan Zheng31153d82008-07-28 15:32:19 -04002561 if (fs_info->total_ref_cache_size) {
Chris Masond3977122009-01-05 21:25:51 -05002562 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2563 (unsigned long long)fs_info->total_ref_cache_size);
Yan Zheng31153d82008-07-28 15:32:19 -04002564 }
Yanbcc63ab2008-07-30 16:29:20 -04002565
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002566 free_extent_buffer(fs_info->extent_root->node);
2567 free_extent_buffer(fs_info->extent_root->commit_root);
2568 free_extent_buffer(fs_info->tree_root->node);
2569 free_extent_buffer(fs_info->tree_root->commit_root);
2570 free_extent_buffer(root->fs_info->chunk_root->node);
2571 free_extent_buffer(root->fs_info->chunk_root->commit_root);
2572 free_extent_buffer(root->fs_info->dev_root->node);
2573 free_extent_buffer(root->fs_info->dev_root->commit_root);
2574 free_extent_buffer(root->fs_info->csum_root->node);
2575 free_extent_buffer(root->fs_info->csum_root->commit_root);
Chris Masond20f7042008-12-08 16:58:54 -05002576
Chris Mason9078a3e2007-04-26 16:46:15 -04002577 btrfs_free_block_groups(root->fs_info);
Yan Zhengc146afa2008-11-12 14:34:12 -05002578
Chris Mason0f7d52f2007-04-09 10:42:37 -04002579 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05002580
Yan Zhengc146afa2008-11-12 14:34:12 -05002581 iput(fs_info->btree_inode);
Chris Mason9ad6b7bc2008-04-18 16:11:30 -04002582
Chris Mason61d92c32009-10-02 19:11:56 -04002583 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04002584 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002585 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002586 btrfs_stop_workers(&fs_info->workers);
2587 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002588 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002589 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002590 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002591 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002592 btrfs_stop_workers(&fs_info->submit_workers);
Chris Masond6bfde82008-04-30 13:59:35 -04002593
Chris Masondfe25022008-05-13 13:46:40 -04002594 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04002595 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04002596
Chris Mason04160082008-03-26 10:28:07 -04002597 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002598 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04002599
Chris Mason0f7d52f2007-04-09 10:42:37 -04002600 kfree(fs_info->extent_root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04002601 kfree(fs_info->tree_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002602 kfree(fs_info->chunk_root);
2603 kfree(fs_info->dev_root);
Chris Masond20f7042008-12-08 16:58:54 -05002604 kfree(fs_info->csum_root);
Li Zefan83a4d542010-12-27 16:19:53 +08002605 kfree(fs_info);
2606
Chris Masoneb60cea2007-02-02 09:18:22 -05002607 return 0;
2608}
2609
Chris Mason1259ab72008-05-12 13:39:03 -04002610int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
Chris Masonccd467d2007-06-28 15:57:36 -04002611{
Chris Mason1259ab72008-05-12 13:39:03 -04002612 int ret;
Chris Mason810191f2007-10-15 16:18:55 -04002613 struct inode *btree_inode = buf->first_page->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04002614
Josef Bacik2ac55d42010-02-03 19:33:23 +00002615 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
2616 NULL);
Chris Mason1259ab72008-05-12 13:39:03 -04002617 if (!ret)
2618 return ret;
2619
2620 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2621 parent_transid);
2622 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04002623}
Chris Mason6702ed42007-08-07 16:15:09 -04002624
Chris Mason5f39d392007-10-15 16:14:19 -04002625int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
2626{
Chris Mason810191f2007-10-15 16:18:55 -04002627 struct inode *btree_inode = buf->first_page->mapping->host;
Chris Masond1310b22008-01-24 16:13:08 -05002628 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
Chris Mason5f39d392007-10-15 16:14:19 -04002629 buf);
2630}
2631
2632void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2633{
Chris Mason810191f2007-10-15 16:18:55 -04002634 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
Chris Mason5f39d392007-10-15 16:14:19 -04002635 u64 transid = btrfs_header_generation(buf);
2636 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonb9473432009-03-13 11:00:37 -04002637 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05002638
Chris Masonb9447ef82009-03-09 11:45:38 -04002639 btrfs_assert_tree_locked(buf);
Chris Masonccd467d2007-06-28 15:57:36 -04002640 if (transid != root->fs_info->generation) {
Chris Masond3977122009-01-05 21:25:51 -05002641 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2642 "found %llu running %llu\n",
Chris Masondb945352007-10-15 16:15:53 -04002643 (unsigned long long)buf->start,
Chris Masond3977122009-01-05 21:25:51 -05002644 (unsigned long long)transid,
2645 (unsigned long long)root->fs_info->generation);
Chris Masonccd467d2007-06-28 15:57:36 -04002646 WARN_ON(1);
2647 }
Chris Masonb9473432009-03-13 11:00:37 -04002648 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2649 buf);
2650 if (!was_dirty) {
2651 spin_lock(&root->fs_info->delalloc_lock);
2652 root->fs_info->dirty_metadata_bytes += buf->len;
2653 spin_unlock(&root->fs_info->delalloc_lock);
2654 }
Chris Masoneb60cea2007-02-02 09:18:22 -05002655}
2656
Chris Masond3c2fdcf2007-09-17 10:58:06 -04002657void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
Chris Mason35b7e472007-05-02 15:53:43 -04002658{
Chris Mason188de642008-05-09 11:52:25 -04002659 /*
2660 * looks as though older kernels can get into trouble with
2661 * this code, they end up stuck in balance_dirty_pages forever
2662 */
Chris Masond6bfde82008-04-30 13:59:35 -04002663 u64 num_dirty;
Chris Mason771ed682008-11-06 22:02:51 -05002664 unsigned long thresh = 32 * 1024 * 1024;
Chris Masond6bfde82008-04-30 13:59:35 -04002665
Jens Axboe6933c022009-03-17 09:36:37 +01002666 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04002667 return;
2668
Chris Mason585ad2c2009-05-18 10:41:58 -04002669 num_dirty = root->fs_info->dirty_metadata_bytes;
2670
Chris Masond6bfde82008-04-30 13:59:35 -04002671 if (num_dirty > thresh) {
2672 balance_dirty_pages_ratelimited_nr(
Chris Masond7fc6402008-02-18 12:12:38 -05002673 root->fs_info->btree_inode->i_mapping, 1);
Chris Masond6bfde82008-04-30 13:59:35 -04002674 }
Chris Mason188de642008-05-09 11:52:25 -04002675 return;
Chris Mason35b7e472007-05-02 15:53:43 -04002676}
Chris Mason6b800532007-10-15 16:17:34 -04002677
Chris Masonca7a79a2008-05-12 12:59:19 -04002678int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04002679{
Chris Mason810191f2007-10-15 16:18:55 -04002680 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
Chris Masonce9adaa2008-04-09 16:28:12 -04002681 int ret;
Chris Masonca7a79a2008-05-12 12:59:19 -04002682 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Masond3977122009-01-05 21:25:51 -05002683 if (ret == 0)
Chris Masonb4ce94d2009-02-04 09:25:08 -05002684 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
Chris Masonce9adaa2008-04-09 16:28:12 -04002685 return ret;
Chris Mason6b800532007-10-15 16:17:34 -04002686}
Chris Mason0da54682007-11-07 21:08:01 -05002687
Chris Mason4bef0842008-09-08 11:18:08 -04002688int btree_lock_page_hook(struct page *page)
2689{
2690 struct inode *inode = page->mapping->host;
Chris Masonb9473432009-03-13 11:00:37 -04002691 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4bef0842008-09-08 11:18:08 -04002692 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2693 struct extent_buffer *eb;
2694 unsigned long len;
2695 u64 bytenr = page_offset(page);
2696
2697 if (page->private == EXTENT_PAGE_PRIVATE)
2698 goto out;
2699
2700 len = page->private >> 2;
2701 eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
2702 if (!eb)
2703 goto out;
2704
2705 btrfs_tree_lock(eb);
Chris Mason4bef0842008-09-08 11:18:08 -04002706 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonb9473432009-03-13 11:00:37 -04002707
2708 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2709 spin_lock(&root->fs_info->delalloc_lock);
2710 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2711 root->fs_info->dirty_metadata_bytes -= eb->len;
2712 else
2713 WARN_ON(1);
2714 spin_unlock(&root->fs_info->delalloc_lock);
2715 }
2716
Chris Mason4bef0842008-09-08 11:18:08 -04002717 btrfs_tree_unlock(eb);
2718 free_extent_buffer(eb);
2719out:
2720 lock_page(page);
2721 return 0;
2722}
2723
liuboacce9522011-01-06 19:30:25 +08002724static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
2725 int read_only)
2726{
2727 if (read_only)
2728 return;
2729
2730 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
2731 printk(KERN_WARNING "warning: mount fs with errors, "
2732 "running btrfsck is recommended\n");
2733}
2734
2735int btrfs_error_commit_super(struct btrfs_root *root)
2736{
2737 int ret;
2738
2739 mutex_lock(&root->fs_info->cleaner_mutex);
2740 btrfs_run_delayed_iputs(root);
2741 mutex_unlock(&root->fs_info->cleaner_mutex);
2742
2743 down_write(&root->fs_info->cleanup_work_sem);
2744 up_write(&root->fs_info->cleanup_work_sem);
2745
2746 /* cleanup FS via transaction */
2747 btrfs_cleanup_transaction(root);
2748
2749 ret = write_ctree_super(NULL, root, 0);
2750
2751 return ret;
2752}
2753
2754static int btrfs_destroy_ordered_operations(struct btrfs_root *root)
2755{
2756 struct btrfs_inode *btrfs_inode;
2757 struct list_head splice;
2758
2759 INIT_LIST_HEAD(&splice);
2760
2761 mutex_lock(&root->fs_info->ordered_operations_mutex);
2762 spin_lock(&root->fs_info->ordered_extent_lock);
2763
2764 list_splice_init(&root->fs_info->ordered_operations, &splice);
2765 while (!list_empty(&splice)) {
2766 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2767 ordered_operations);
2768
2769 list_del_init(&btrfs_inode->ordered_operations);
2770
2771 btrfs_invalidate_inodes(btrfs_inode->root);
2772 }
2773
2774 spin_unlock(&root->fs_info->ordered_extent_lock);
2775 mutex_unlock(&root->fs_info->ordered_operations_mutex);
2776
2777 return 0;
2778}
2779
2780static int btrfs_destroy_ordered_extents(struct btrfs_root *root)
2781{
2782 struct list_head splice;
2783 struct btrfs_ordered_extent *ordered;
2784 struct inode *inode;
2785
2786 INIT_LIST_HEAD(&splice);
2787
2788 spin_lock(&root->fs_info->ordered_extent_lock);
2789
2790 list_splice_init(&root->fs_info->ordered_extents, &splice);
2791 while (!list_empty(&splice)) {
2792 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
2793 root_extent_list);
2794
2795 list_del_init(&ordered->root_extent_list);
2796 atomic_inc(&ordered->refs);
2797
2798 /* the inode may be getting freed (in sys_unlink path). */
2799 inode = igrab(ordered->inode);
2800
2801 spin_unlock(&root->fs_info->ordered_extent_lock);
2802 if (inode)
2803 iput(inode);
2804
2805 atomic_set(&ordered->refs, 1);
2806 btrfs_put_ordered_extent(ordered);
2807
2808 spin_lock(&root->fs_info->ordered_extent_lock);
2809 }
2810
2811 spin_unlock(&root->fs_info->ordered_extent_lock);
2812
2813 return 0;
2814}
2815
2816static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2817 struct btrfs_root *root)
2818{
2819 struct rb_node *node;
2820 struct btrfs_delayed_ref_root *delayed_refs;
2821 struct btrfs_delayed_ref_node *ref;
2822 int ret = 0;
2823
2824 delayed_refs = &trans->delayed_refs;
2825
2826 spin_lock(&delayed_refs->lock);
2827 if (delayed_refs->num_entries == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04002828 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08002829 printk(KERN_INFO "delayed_refs has NO entry\n");
2830 return ret;
2831 }
2832
2833 node = rb_first(&delayed_refs->root);
2834 while (node) {
2835 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2836 node = rb_next(node);
2837
2838 ref->in_tree = 0;
2839 rb_erase(&ref->rb_node, &delayed_refs->root);
2840 delayed_refs->num_entries--;
2841
2842 atomic_set(&ref->refs, 1);
2843 if (btrfs_delayed_ref_is_head(ref)) {
2844 struct btrfs_delayed_ref_head *head;
2845
2846 head = btrfs_delayed_node_to_head(ref);
2847 mutex_lock(&head->mutex);
2848 kfree(head->extent_op);
2849 delayed_refs->num_heads--;
2850 if (list_empty(&head->cluster))
2851 delayed_refs->num_heads_ready--;
2852 list_del_init(&head->cluster);
2853 mutex_unlock(&head->mutex);
2854 }
2855
2856 spin_unlock(&delayed_refs->lock);
2857 btrfs_put_delayed_ref(ref);
2858
2859 cond_resched();
2860 spin_lock(&delayed_refs->lock);
2861 }
2862
2863 spin_unlock(&delayed_refs->lock);
2864
2865 return ret;
2866}
2867
2868static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
2869{
2870 struct btrfs_pending_snapshot *snapshot;
2871 struct list_head splice;
2872
2873 INIT_LIST_HEAD(&splice);
2874
2875 list_splice_init(&t->pending_snapshots, &splice);
2876
2877 while (!list_empty(&splice)) {
2878 snapshot = list_entry(splice.next,
2879 struct btrfs_pending_snapshot,
2880 list);
2881
2882 list_del_init(&snapshot->list);
2883
2884 kfree(snapshot);
2885 }
2886
2887 return 0;
2888}
2889
2890static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
2891{
2892 struct btrfs_inode *btrfs_inode;
2893 struct list_head splice;
2894
2895 INIT_LIST_HEAD(&splice);
2896
2897 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
2898
2899 spin_lock(&root->fs_info->delalloc_lock);
2900
2901 while (!list_empty(&splice)) {
2902 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2903 delalloc_inodes);
2904
2905 list_del_init(&btrfs_inode->delalloc_inodes);
2906
2907 btrfs_invalidate_inodes(btrfs_inode->root);
2908 }
2909
2910 spin_unlock(&root->fs_info->delalloc_lock);
2911
2912 return 0;
2913}
2914
2915static int btrfs_destroy_marked_extents(struct btrfs_root *root,
2916 struct extent_io_tree *dirty_pages,
2917 int mark)
2918{
2919 int ret;
2920 struct page *page;
2921 struct inode *btree_inode = root->fs_info->btree_inode;
2922 struct extent_buffer *eb;
2923 u64 start = 0;
2924 u64 end;
2925 u64 offset;
2926 unsigned long index;
2927
2928 while (1) {
2929 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
2930 mark);
2931 if (ret)
2932 break;
2933
2934 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
2935 while (start <= end) {
2936 index = start >> PAGE_CACHE_SHIFT;
2937 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
2938 page = find_get_page(btree_inode->i_mapping, index);
2939 if (!page)
2940 continue;
2941 offset = page_offset(page);
2942
2943 spin_lock(&dirty_pages->buffer_lock);
2944 eb = radix_tree_lookup(
2945 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
2946 offset >> PAGE_CACHE_SHIFT);
2947 spin_unlock(&dirty_pages->buffer_lock);
2948 if (eb) {
2949 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
2950 &eb->bflags);
2951 atomic_set(&eb->refs, 1);
2952 }
2953 if (PageWriteback(page))
2954 end_page_writeback(page);
2955
2956 lock_page(page);
2957 if (PageDirty(page)) {
2958 clear_page_dirty_for_io(page);
2959 spin_lock_irq(&page->mapping->tree_lock);
2960 radix_tree_tag_clear(&page->mapping->page_tree,
2961 page_index(page),
2962 PAGECACHE_TAG_DIRTY);
2963 spin_unlock_irq(&page->mapping->tree_lock);
2964 }
2965
2966 page->mapping->a_ops->invalidatepage(page, 0);
2967 unlock_page(page);
2968 }
2969 }
2970
2971 return ret;
2972}
2973
2974static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
2975 struct extent_io_tree *pinned_extents)
2976{
2977 struct extent_io_tree *unpin;
2978 u64 start;
2979 u64 end;
2980 int ret;
2981
2982 unpin = pinned_extents;
2983 while (1) {
2984 ret = find_first_extent_bit(unpin, 0, &start, &end,
2985 EXTENT_DIRTY);
2986 if (ret)
2987 break;
2988
2989 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00002990 if (btrfs_test_opt(root, DISCARD))
2991 ret = btrfs_error_discard_extent(root, start,
2992 end + 1 - start,
2993 NULL);
liuboacce9522011-01-06 19:30:25 +08002994
2995 clear_extent_dirty(unpin, start, end, GFP_NOFS);
2996 btrfs_error_unpin_extent_range(root, start, end);
2997 cond_resched();
2998 }
2999
3000 return 0;
3001}
3002
3003static int btrfs_cleanup_transaction(struct btrfs_root *root)
3004{
3005 struct btrfs_transaction *t;
3006 LIST_HEAD(list);
3007
3008 WARN_ON(1);
3009
3010 mutex_lock(&root->fs_info->trans_mutex);
3011 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3012
3013 list_splice_init(&root->fs_info->trans_list, &list);
3014 while (!list_empty(&list)) {
3015 t = list_entry(list.next, struct btrfs_transaction, list);
3016 if (!t)
3017 break;
3018
3019 btrfs_destroy_ordered_operations(root);
3020
3021 btrfs_destroy_ordered_extents(root);
3022
3023 btrfs_destroy_delayed_refs(t, root);
3024
3025 btrfs_block_rsv_release(root,
3026 &root->fs_info->trans_block_rsv,
3027 t->dirty_pages.dirty_bytes);
3028
3029 /* FIXME: cleanup wait for commit */
3030 t->in_commit = 1;
3031 t->blocked = 1;
3032 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3033 wake_up(&root->fs_info->transaction_blocked_wait);
3034
3035 t->blocked = 0;
3036 if (waitqueue_active(&root->fs_info->transaction_wait))
3037 wake_up(&root->fs_info->transaction_wait);
3038 mutex_unlock(&root->fs_info->trans_mutex);
3039
3040 mutex_lock(&root->fs_info->trans_mutex);
3041 t->commit_done = 1;
3042 if (waitqueue_active(&t->commit_wait))
3043 wake_up(&t->commit_wait);
3044 mutex_unlock(&root->fs_info->trans_mutex);
3045
3046 mutex_lock(&root->fs_info->trans_mutex);
3047
3048 btrfs_destroy_pending_snapshots(t);
3049
3050 btrfs_destroy_delalloc_inodes(root);
3051
3052 spin_lock(&root->fs_info->new_trans_lock);
3053 root->fs_info->running_transaction = NULL;
3054 spin_unlock(&root->fs_info->new_trans_lock);
3055
3056 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3057 EXTENT_DIRTY);
3058
3059 btrfs_destroy_pinned_extent(root,
3060 root->fs_info->pinned_extents);
3061
Josef Bacik13c5a932011-04-11 15:45:29 -04003062 atomic_set(&t->use_count, 0);
liuboacce9522011-01-06 19:30:25 +08003063 list_del_init(&t->list);
3064 memset(t, 0, sizeof(*t));
3065 kmem_cache_free(btrfs_transaction_cachep, t);
3066 }
3067
3068 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
3069 mutex_unlock(&root->fs_info->trans_mutex);
3070
3071 return 0;
3072}
3073
Chris Masond1310b22008-01-24 16:13:08 -05003074static struct extent_io_ops btree_extent_io_ops = {
Chris Mason4bef0842008-09-08 11:18:08 -04003075 .write_cache_pages_lock_hook = btree_lock_page_hook,
Chris Masonce9adaa2008-04-09 16:28:12 -04003076 .readpage_end_io_hook = btree_readpage_end_io_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04003077 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04003078 /* note we're sharing with inode.c for the merge bio hook */
3079 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05003080};