blob: 7b55eee15a51265e3d670cafa57c2772b8f0614a [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
David Woodhouse163e7832009-04-19 13:02:41 +010029#include <linux/crc32c.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090030#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050031#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020032#include <linux/ratelimit.h>
David Sterba7e75bf32011-03-18 22:56:43 +000033#include <asm/unaligned.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050034#include "compat.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Chris Masone089f052007-03-16 16:20:31 -040037#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040038#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040039#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040040#include "print-tree.h"
Chris Mason8b712842008-06-11 16:50:36 -040041#include "async-thread.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080045#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010046#include "check-integrity.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050047
Chris Masond1310b22008-01-24 16:13:08 -050048static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040049static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040050static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010051static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080052 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010053static void btrfs_destroy_ordered_operations(struct btrfs_root *root);
54static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080055static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
56 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010057static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
58static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080059static int btrfs_destroy_marked_extents(struct btrfs_root *root,
60 struct extent_io_tree *dirty_pages,
61 int mark);
62static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
63 struct extent_io_tree *pinned_extents);
Chris Masonce9adaa2008-04-09 16:28:12 -040064
Chris Masond352ac62008-09-29 15:18:18 -040065/*
66 * end_io_wq structs are used to do processing in task context when an IO is
67 * complete. This is used during reads to verify checksums, and it is used
68 * by writes to insert metadata for new file extents after IO is complete.
69 */
Chris Masonce9adaa2008-04-09 16:28:12 -040070struct end_io_wq {
71 struct bio *bio;
72 bio_end_io_t *end_io;
73 void *private;
74 struct btrfs_fs_info *info;
75 int error;
Chris Mason22c59942008-04-09 16:28:12 -040076 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040077 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040078 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040079};
Chris Mason0da54682007-11-07 21:08:01 -050080
Chris Masond352ac62008-09-29 15:18:18 -040081/*
82 * async submit bios are used to offload expensive checksumming
83 * onto the worker threads. They checksum file and metadata bios
84 * just before they are sent down the IO stack.
85 */
Chris Mason44b8bd72008-04-16 11:14:51 -040086struct async_submit_bio {
87 struct inode *inode;
88 struct bio *bio;
89 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -050090 extent_submit_bio_hook_t *submit_bio_start;
91 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -040092 int rw;
93 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -040094 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -040095 /*
96 * bio_offset is optional, can be used if the pages in the bio
97 * can't tell us where in the file the bio should go
98 */
99 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400100 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100101 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400102};
103
Chris Mason85d4e462011-07-26 16:11:19 -0400104/*
105 * Lockdep class keys for extent_buffer->lock's in this root. For a given
106 * eb, the lockdep key is determined by the btrfs_root it belongs to and
107 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500108 *
Chris Mason85d4e462011-07-26 16:11:19 -0400109 * Different roots are used for different purposes and may nest inside each
110 * other and they require separate keysets. As lockdep keys should be
111 * static, assign keysets according to the purpose of the root as indicated
112 * by btrfs_root->objectid. This ensures that all special purpose roots
113 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500114 *
Chris Mason85d4e462011-07-26 16:11:19 -0400115 * Lock-nesting across peer nodes is always done with the immediate parent
116 * node locked thus preventing deadlock. As lockdep doesn't know this, use
117 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500118 *
Chris Mason85d4e462011-07-26 16:11:19 -0400119 * The key is set by the readpage_end_io_hook after the buffer has passed
120 * csum validation but before the pages are unlocked. It is also set by
121 * btrfs_init_new_buffer on freshly allocated blocks.
122 *
123 * We also add a check to make sure the highest level of the tree is the
124 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
125 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500126 */
127#ifdef CONFIG_DEBUG_LOCK_ALLOC
128# if BTRFS_MAX_LEVEL != 8
129# error
130# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400131
132static struct btrfs_lockdep_keyset {
133 u64 id; /* root objectid */
134 const char *name_stem; /* lock name stem */
135 char names[BTRFS_MAX_LEVEL + 1][20];
136 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
137} btrfs_lockdep_keysets[] = {
138 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
139 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
140 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
141 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
142 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
143 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
144 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
145 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
146 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
147 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
148 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500149};
Chris Mason85d4e462011-07-26 16:11:19 -0400150
151void __init btrfs_init_lockdep(void)
152{
153 int i, j;
154
155 /* initialize lockdep class names */
156 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
157 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
158
159 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
160 snprintf(ks->names[j], sizeof(ks->names[j]),
161 "btrfs-%s-%02d", ks->name_stem, j);
162 }
163}
164
165void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
166 int level)
167{
168 struct btrfs_lockdep_keyset *ks;
169
170 BUG_ON(level >= ARRAY_SIZE(ks->keys));
171
172 /* find the matching keyset, id 0 is the default entry */
173 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
174 if (ks->id == objectid)
175 break;
176
177 lockdep_set_class_and_name(&eb->lock,
178 &ks->keys[level], ks->names[level]);
179}
180
Chris Mason4008c042009-02-12 14:09:45 -0500181#endif
182
Chris Masond352ac62008-09-29 15:18:18 -0400183/*
184 * extents on the btree inode are pretty simple, there's one extent
185 * that covers the entire device
186 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500187static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200188 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500189 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400190{
191 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
192 struct extent_map *em;
193 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400194
Chris Mason890871b2009-09-02 16:24:52 -0400195 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500196 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400197 if (em) {
198 em->bdev =
199 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400200 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400201 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400202 }
Chris Mason890871b2009-09-02 16:24:52 -0400203 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400204
David Sterba172ddd62011-04-21 00:48:27 +0200205 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400206 if (!em) {
207 em = ERR_PTR(-ENOMEM);
208 goto out;
209 }
210 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400211 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400212 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400213 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400214 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500215
Chris Mason890871b2009-09-02 16:24:52 -0400216 write_lock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400217 ret = add_extent_mapping(em_tree, em);
218 if (ret == -EEXIST) {
Chris Mason0afbaf82008-04-18 14:17:20 -0400219 u64 failed_start = em->start;
220 u64 failed_len = em->len;
221
Chris Mason5f39d392007-10-15 16:14:19 -0400222 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400223 em = lookup_extent_mapping(em_tree, start, len);
Chris Mason0afbaf82008-04-18 14:17:20 -0400224 if (em) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400225 ret = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400226 } else {
227 em = lookup_extent_mapping(em_tree, failed_start,
228 failed_len);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400229 ret = -EIO;
Chris Mason0afbaf82008-04-18 14:17:20 -0400230 }
Chris Mason5f39d392007-10-15 16:14:19 -0400231 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400232 free_extent_map(em);
233 em = NULL;
Chris Mason5f39d392007-10-15 16:14:19 -0400234 }
Chris Mason890871b2009-09-02 16:24:52 -0400235 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400236
237 if (ret)
238 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400239out:
240 return em;
241}
242
Chris Mason19c00dd2007-10-15 16:19:22 -0400243u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
244{
David Woodhouse163e7832009-04-19 13:02:41 +0100245 return crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400246}
247
248void btrfs_csum_final(u32 crc, char *result)
249{
David Sterba7e75bf32011-03-18 22:56:43 +0000250 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400251}
252
Chris Masond352ac62008-09-29 15:18:18 -0400253/*
254 * compute the csum for a btree block, and either verify it or write it
255 * into the csum field of the block.
256 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400257static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
258 int verify)
259{
David Sterba6c417612011-04-13 15:41:04 +0200260 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500261 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400262 unsigned long len;
263 unsigned long cur_len;
264 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400265 char *kaddr;
266 unsigned long map_start;
267 unsigned long map_len;
268 int err;
269 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500270 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400271
272 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500273 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400274 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400275 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500276 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400277 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400278 cur_len = min(len, map_len - (offset - map_start));
279 crc = btrfs_csum_data(root, kaddr + offset - map_start,
280 crc, cur_len);
281 len -= cur_len;
282 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400283 }
Josef Bacik607d4322008-12-02 07:17:45 -0500284 if (csum_size > sizeof(inline_result)) {
285 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
286 if (!result)
287 return 1;
288 } else {
289 result = (char *)&inline_result;
290 }
291
Chris Mason19c00dd2007-10-15 16:19:22 -0400292 btrfs_csum_final(crc, result);
293
294 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500295 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500296 u32 val;
297 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500298 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500299
Josef Bacik607d4322008-12-02 07:17:45 -0500300 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba7a36dde2011-05-06 15:33:15 +0200301 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
Chris Mason193f2842009-04-27 07:29:05 -0400302 "failed on %llu wanted %X found %X "
303 "level %d\n",
304 root->fs_info->sb->s_id,
305 (unsigned long long)buf->start, val, found,
306 btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500307 if (result != (char *)&inline_result)
308 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400309 return 1;
310 }
311 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500312 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400313 }
Josef Bacik607d4322008-12-02 07:17:45 -0500314 if (result != (char *)&inline_result)
315 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400316 return 0;
317}
318
Chris Masond352ac62008-09-29 15:18:18 -0400319/*
320 * we can't consider a given block up to date unless the transid of the
321 * block matches the transid in the parent node's pointer. This is how we
322 * detect blocks that either didn't get written at all or got written
323 * in the wrong place.
324 */
Chris Mason1259ab72008-05-12 13:39:03 -0400325static int verify_parent_transid(struct extent_io_tree *io_tree,
326 struct extent_buffer *eb, u64 parent_transid)
327{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000328 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400329 int ret;
330
331 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
332 return 0;
333
Josef Bacik2ac55d42010-02-03 19:33:23 +0000334 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100335 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400336 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400337 btrfs_header_generation(eb) == parent_transid) {
338 ret = 0;
339 goto out;
340 }
David Sterba7a36dde2011-05-06 15:33:15 +0200341 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
Chris Mason193f2842009-04-27 07:29:05 -0400342 "found %llu\n",
343 (unsigned long long)eb->start,
344 (unsigned long long)parent_transid,
345 (unsigned long long)btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400346 ret = 1;
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400347 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400348out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000349 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
350 &cached_state, GFP_NOFS);
Chris Mason1259ab72008-05-12 13:39:03 -0400351 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400352}
353
Chris Masond352ac62008-09-29 15:18:18 -0400354/*
355 * helper to read a given tree block, doing retries as required when
356 * the checksums don't match and we have alternate mirrors to try.
357 */
Chris Masonf1885912008-04-09 16:28:12 -0400358static int btree_read_extent_buffer_pages(struct btrfs_root *root,
359 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400360 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400361{
362 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400363 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400364 int ret;
365 int num_copies = 0;
366 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400367 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400368
Josef Bacika826d6d2011-03-16 13:42:43 -0400369 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400370 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
371 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200372 ret = read_extent_buffer_pages(io_tree, eb, start,
373 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400374 btree_get_extent, mirror_num);
Chris Mason727011e2010-08-06 13:21:20 -0400375 if (!ret && !verify_parent_transid(io_tree, eb, parent_transid))
Josef Bacikea466792012-03-26 21:57:36 -0400376 break;
Chris Masond3977122009-01-05 21:25:51 -0500377
Josef Bacika826d6d2011-03-16 13:42:43 -0400378 /*
379 * This buffer's crc is fine, but its contents are corrupted, so
380 * there is no reason to read the other copies, they won't be
381 * any less wrong.
382 */
383 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400384 break;
385
386 if (!failed_mirror) {
387 failed = 1;
388 printk(KERN_ERR "failed mirror was %d\n", eb->failed_mirror);
389 failed_mirror = eb->failed_mirror;
390 }
Josef Bacika826d6d2011-03-16 13:42:43 -0400391
Chris Masonf1885912008-04-09 16:28:12 -0400392 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
393 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400394 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400395 break;
Chris Mason42352982008-04-28 16:40:52 -0400396
Chris Masonf1885912008-04-09 16:28:12 -0400397 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400398 if (mirror_num == failed_mirror)
399 mirror_num++;
400
Chris Mason42352982008-04-28 16:40:52 -0400401 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400402 break;
Chris Masonf1885912008-04-09 16:28:12 -0400403 }
Josef Bacikea466792012-03-26 21:57:36 -0400404
405 if (failed && !ret)
406 repair_eb_io_failure(root, eb, failed_mirror);
407
408 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400409}
Chris Mason19c00dd2007-10-15 16:19:22 -0400410
Chris Masond352ac62008-09-29 15:18:18 -0400411/*
Chris Masond3977122009-01-05 21:25:51 -0500412 * checksum a dirty tree block before IO. This has extra checks to make sure
413 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400414 */
Chris Masond3977122009-01-05 21:25:51 -0500415
Christoph Hellwigb2950862008-12-02 09:54:17 -0500416static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400417{
Chris Masond1310b22008-01-24 16:13:08 -0500418 struct extent_io_tree *tree;
Chris Mason35ebb932007-10-30 16:56:53 -0400419 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
Chris Mason19c00dd2007-10-15 16:19:22 -0400420 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400421 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400422
Chris Masond1310b22008-01-24 16:13:08 -0500423 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason19c00dd2007-10-15 16:19:22 -0400424
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500425 eb = (struct extent_buffer *)page->private;
426 if (page != eb->pages[0])
427 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400428 found_start = btrfs_header_bytenr(eb);
429 if (found_start != start) {
Chris Mason55c69072008-01-09 15:55:33 -0500430 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500431 return 0;
Chris Mason55c69072008-01-09 15:55:33 -0500432 }
Chris Mason727011e2010-08-06 13:21:20 -0400433 if (eb->pages[0] != page) {
Chris Mason55c69072008-01-09 15:55:33 -0500434 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500435 return 0;
Chris Mason55c69072008-01-09 15:55:33 -0500436 }
437 if (!PageUptodate(page)) {
Chris Mason55c69072008-01-09 15:55:33 -0500438 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500439 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400440 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400441 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400442 return 0;
443}
444
Yan Zheng2b820322008-11-17 21:11:30 -0500445static int check_tree_block_fsid(struct btrfs_root *root,
446 struct extent_buffer *eb)
447{
448 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
449 u8 fsid[BTRFS_UUID_SIZE];
450 int ret = 1;
451
452 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
453 BTRFS_FSID_SIZE);
454 while (fs_devices) {
455 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
456 ret = 0;
457 break;
458 }
459 fs_devices = fs_devices->seed;
460 }
461 return ret;
462}
463
Josef Bacika826d6d2011-03-16 13:42:43 -0400464#define CORRUPT(reason, eb, root, slot) \
465 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
466 "root=%llu, slot=%d\n", reason, \
467 (unsigned long long)btrfs_header_bytenr(eb), \
468 (unsigned long long)root->objectid, slot)
469
470static noinline int check_leaf(struct btrfs_root *root,
471 struct extent_buffer *leaf)
472{
473 struct btrfs_key key;
474 struct btrfs_key leaf_key;
475 u32 nritems = btrfs_header_nritems(leaf);
476 int slot;
477
478 if (nritems == 0)
479 return 0;
480
481 /* Check the 0 item */
482 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
483 BTRFS_LEAF_DATA_SIZE(root)) {
484 CORRUPT("invalid item offset size pair", leaf, root, 0);
485 return -EIO;
486 }
487
488 /*
489 * Check to make sure each items keys are in the correct order and their
490 * offsets make sense. We only have to loop through nritems-1 because
491 * we check the current slot against the next slot, which verifies the
492 * next slot's offset+size makes sense and that the current's slot
493 * offset is correct.
494 */
495 for (slot = 0; slot < nritems - 1; slot++) {
496 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
497 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
498
499 /* Make sure the keys are in the right order */
500 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
501 CORRUPT("bad key order", leaf, root, slot);
502 return -EIO;
503 }
504
505 /*
506 * Make sure the offset and ends are right, remember that the
507 * item data starts at the end of the leaf and grows towards the
508 * front.
509 */
510 if (btrfs_item_offset_nr(leaf, slot) !=
511 btrfs_item_end_nr(leaf, slot + 1)) {
512 CORRUPT("slot offset bad", leaf, root, slot);
513 return -EIO;
514 }
515
516 /*
517 * Check to make sure that we don't point outside of the leaf,
518 * just incase all the items are consistent to eachother, but
519 * all point outside of the leaf.
520 */
521 if (btrfs_item_end_nr(leaf, slot) >
522 BTRFS_LEAF_DATA_SIZE(root)) {
523 CORRUPT("slot end outside of leaf", leaf, root, slot);
524 return -EIO;
525 }
526 }
527
528 return 0;
529}
530
Chris Mason727011e2010-08-06 13:21:20 -0400531struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree,
532 struct page *page, int max_walk)
533{
534 struct extent_buffer *eb;
535 u64 start = page_offset(page);
536 u64 target = start;
537 u64 min_start;
538
539 if (start < max_walk)
540 min_start = 0;
541 else
542 min_start = start - max_walk;
543
544 while (start >= min_start) {
545 eb = find_extent_buffer(tree, start, 0);
546 if (eb) {
547 /*
548 * we found an extent buffer and it contains our page
549 * horray!
550 */
551 if (eb->start <= target &&
552 eb->start + eb->len > target)
553 return eb;
554
555 /* we found an extent buffer that wasn't for us */
556 free_extent_buffer(eb);
557 return NULL;
558 }
559 if (start == 0)
560 break;
561 start -= PAGE_CACHE_SIZE;
562 }
563 return NULL;
564}
565
Christoph Hellwigb2950862008-12-02 09:54:17 -0500566static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
Chris Masonce9adaa2008-04-09 16:28:12 -0400567 struct extent_state *state)
568{
569 struct extent_io_tree *tree;
570 u64 found_start;
571 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400572 struct extent_buffer *eb;
573 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400574 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400575 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400576
Chris Masonce9adaa2008-04-09 16:28:12 -0400577 if (!page->private)
578 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500579
Chris Mason727011e2010-08-06 13:21:20 -0400580 tree = &BTRFS_I(page->mapping->host)->io_tree;
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500581 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500582
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400583 /* the pending IO might have been the only thing that kept this buffer
584 * in memory. Make sure we have a ref for all this other checks
585 */
586 extent_buffer_get(eb);
587
588 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400589 if (!reads_done)
590 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400591
Josef Bacikea466792012-03-26 21:57:36 -0400592 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
593 ret = -EIO;
594 goto err;
595 }
596
Chris Masonce9adaa2008-04-09 16:28:12 -0400597 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400598 if (found_start != eb->start) {
David Sterba7a36dde2011-05-06 15:33:15 +0200599 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400600 "%llu %llu\n",
601 (unsigned long long)found_start,
602 (unsigned long long)eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400603 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400604 goto err;
605 }
Yan Zheng2b820322008-11-17 21:11:30 -0500606 if (check_tree_block_fsid(root, eb)) {
David Sterba7a36dde2011-05-06 15:33:15 +0200607 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
Chris Mason193f2842009-04-27 07:29:05 -0400608 (unsigned long long)eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400609 ret = -EIO;
610 goto err;
611 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400612 found_level = btrfs_header_level(eb);
613
Chris Mason85d4e462011-07-26 16:11:19 -0400614 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
615 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500616
Chris Masonce9adaa2008-04-09 16:28:12 -0400617 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400618 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400619 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400620 goto err;
621 }
622
623 /*
624 * If this is a leaf block and it is corrupt, set the corrupt bit so
625 * that we don't try and read the other copies of this block, just
626 * return -EIO.
627 */
628 if (found_level == 0 && check_leaf(root, eb)) {
629 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
630 ret = -EIO;
631 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400632
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400633 if (!ret)
634 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400635err:
Arne Jansen4bb31e92011-06-10 13:55:54 +0200636 if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
637 clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
638 btree_readahead_hook(root, eb, eb->start, ret);
639 }
640
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400641 if (ret)
642 clear_extent_buffer_uptodate(eb);
643 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400644out:
Chris Masonf1885912008-04-09 16:28:12 -0400645 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400646}
647
Josef Bacikea466792012-03-26 21:57:36 -0400648static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200649{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200650 struct extent_buffer *eb;
651 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
652
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500653 eb = (struct extent_buffer *)page->private;
Josef Bacikea466792012-03-26 21:57:36 -0400654 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
655 eb->failed_mirror = failed_mirror;
656 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200657 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200658 return -EIO; /* we fixed nothing */
659}
660
Chris Masonce9adaa2008-04-09 16:28:12 -0400661static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400662{
663 struct end_io_wq *end_io_wq = bio->bi_private;
664 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400665
Chris Masonce9adaa2008-04-09 16:28:12 -0400666 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400667 end_io_wq->error = err;
Chris Mason8b712842008-06-11 16:50:36 -0400668 end_io_wq->work.func = end_workqueue_fn;
669 end_io_wq->work.flags = 0;
Chris Masond20f7042008-12-08 16:58:54 -0500670
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200671 if (bio->bi_rw & REQ_WRITE) {
Josef Bacik0cb59c92010-07-02 12:14:14 -0400672 if (end_io_wq->metadata == 1)
Chris Masoncad321a2008-12-17 14:51:42 -0500673 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
674 &end_io_wq->work);
Josef Bacik0cb59c92010-07-02 12:14:14 -0400675 else if (end_io_wq->metadata == 2)
676 btrfs_queue_worker(&fs_info->endio_freespace_worker,
677 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500678 else
679 btrfs_queue_worker(&fs_info->endio_write_workers,
680 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500681 } else {
682 if (end_io_wq->metadata)
683 btrfs_queue_worker(&fs_info->endio_meta_workers,
684 &end_io_wq->work);
685 else
686 btrfs_queue_worker(&fs_info->endio_workers,
687 &end_io_wq->work);
688 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400689}
690
Josef Bacik0cb59c92010-07-02 12:14:14 -0400691/*
692 * For the metadata arg you want
693 *
694 * 0 - if data
695 * 1 - if normal metadta
696 * 2 - if writing to the free space cache area
697 */
Chris Mason22c59942008-04-09 16:28:12 -0400698int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
699 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400700{
Chris Masonce9adaa2008-04-09 16:28:12 -0400701 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400702 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
703 if (!end_io_wq)
704 return -ENOMEM;
705
706 end_io_wq->private = bio->bi_private;
707 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400708 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400709 end_io_wq->error = 0;
710 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400711 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400712
713 bio->bi_private = end_io_wq;
714 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400715 return 0;
716}
717
Chris Masonb64a2852008-08-20 13:39:41 -0400718unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400719{
720 unsigned long limit = min_t(unsigned long,
721 info->workers.max_workers,
722 info->fs_devices->open_devices);
723 return 256 * limit;
724}
725
Chris Mason4a69a412008-11-06 22:03:00 -0500726static void run_one_async_start(struct btrfs_work *work)
727{
Chris Mason4a69a412008-11-06 22:03:00 -0500728 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100729 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500730
731 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100732 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
733 async->mirror_num, async->bio_flags,
734 async->bio_offset);
735 if (ret)
736 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500737}
738
739static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400740{
741 struct btrfs_fs_info *fs_info;
742 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400743 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400744
745 async = container_of(work, struct async_submit_bio, work);
746 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400747
Chris Masonb64a2852008-08-20 13:39:41 -0400748 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400749 limit = limit * 2 / 3;
750
Chris Mason8b712842008-06-11 16:50:36 -0400751 atomic_dec(&fs_info->nr_async_submits);
Chris Mason0986fe9e2008-08-15 15:34:15 -0400752
Chris Masonb64a2852008-08-20 13:39:41 -0400753 if (atomic_read(&fs_info->nr_async_submits) < limit &&
754 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400755 wake_up(&fs_info->async_submit_wait);
756
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100757 /* If an error occured we just want to clean up the bio and move on */
758 if (async->error) {
759 bio_endio(async->bio, async->error);
760 return;
761 }
762
Chris Mason4a69a412008-11-06 22:03:00 -0500763 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400764 async->mirror_num, async->bio_flags,
765 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500766}
767
768static void run_one_async_free(struct btrfs_work *work)
769{
770 struct async_submit_bio *async;
771
772 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400773 kfree(async);
774}
775
Chris Mason44b8bd72008-04-16 11:14:51 -0400776int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
777 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400778 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400779 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500780 extent_submit_bio_hook_t *submit_bio_start,
781 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400782{
783 struct async_submit_bio *async;
784
785 async = kmalloc(sizeof(*async), GFP_NOFS);
786 if (!async)
787 return -ENOMEM;
788
789 async->inode = inode;
790 async->rw = rw;
791 async->bio = bio;
792 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500793 async->submit_bio_start = submit_bio_start;
794 async->submit_bio_done = submit_bio_done;
795
796 async->work.func = run_one_async_start;
797 async->work.ordered_func = run_one_async_done;
798 async->work.ordered_free = run_one_async_free;
799
Chris Mason8b712842008-06-11 16:50:36 -0400800 async->work.flags = 0;
Chris Masonc8b97812008-10-29 14:49:59 -0400801 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400802 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400803
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100804 async->error = 0;
805
Chris Masoncb03c742008-05-15 16:15:45 -0400806 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400807
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200808 if (rw & REQ_SYNC)
Chris Masond313d7a2009-04-20 15:50:09 -0400809 btrfs_set_work_high_prio(&async->work);
810
Chris Mason8b712842008-06-11 16:50:36 -0400811 btrfs_queue_worker(&fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400812
Chris Masond3977122009-01-05 21:25:51 -0500813 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500814 atomic_read(&fs_info->nr_async_submits)) {
815 wait_event(fs_info->async_submit_wait,
816 (atomic_read(&fs_info->nr_async_submits) == 0));
817 }
818
Chris Mason44b8bd72008-04-16 11:14:51 -0400819 return 0;
820}
821
Chris Masonce3ed712008-09-23 13:14:12 -0400822static int btree_csum_one_bio(struct bio *bio)
823{
824 struct bio_vec *bvec = bio->bi_io_vec;
825 int bio_index = 0;
826 struct btrfs_root *root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100827 int ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400828
829 WARN_ON(bio->bi_vcnt <= 0);
Chris Masond3977122009-01-05 21:25:51 -0500830 while (bio_index < bio->bi_vcnt) {
Chris Masonce3ed712008-09-23 13:14:12 -0400831 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100832 ret = csum_dirty_buffer(root, bvec->bv_page);
833 if (ret)
834 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400835 bio_index++;
836 bvec++;
837 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100838 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400839}
840
Chris Mason4a69a412008-11-06 22:03:00 -0500841static int __btree_submit_bio_start(struct inode *inode, int rw,
842 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400843 unsigned long bio_flags,
844 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400845{
Chris Mason8b712842008-06-11 16:50:36 -0400846 /*
847 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400848 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400849 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100850 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500851}
Chris Mason22c59942008-04-09 16:28:12 -0400852
Chris Mason4a69a412008-11-06 22:03:00 -0500853static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400854 int mirror_num, unsigned long bio_flags,
855 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500856{
Chris Mason8b712842008-06-11 16:50:36 -0400857 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500858 * when we're called for a write, we're already in the async
859 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400860 */
Chris Mason8b712842008-06-11 16:50:36 -0400861 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Chris Mason0b86a832008-03-24 15:01:56 -0400862}
863
Chris Mason44b8bd72008-04-16 11:14:51 -0400864static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400865 int mirror_num, unsigned long bio_flags,
866 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400867{
Chris Masoncad321a2008-12-17 14:51:42 -0500868 int ret;
869
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200870 if (!(rw & REQ_WRITE)) {
Chris Masonf3f266a2012-03-23 10:22:46 -0400871
Chris Mason4a69a412008-11-06 22:03:00 -0500872 /*
873 * called for a read, do the setup so that checksum validation
874 * can happen in the async kernel threads
875 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400876 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
877 bio, 1);
Chris Mason1d4284b2012-03-28 20:31:37 -0400878 if (ret)
879 return ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500880 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
Chris Mason6f3577b2008-11-13 09:59:36 -0500881 mirror_num, 0);
Chris Mason44b8bd72008-04-16 11:14:51 -0400882 }
Chris Masond313d7a2009-04-20 15:50:09 -0400883
Chris Masoncad321a2008-12-17 14:51:42 -0500884 /*
885 * kthread helpers are used to submit writes so that checksumming
886 * can happen in parallel across all CPUs
887 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400888 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
Chris Masonc8b97812008-10-29 14:49:59 -0400889 inode, rw, bio, mirror_num, 0,
Chris Masoneaf25d92010-05-25 09:48:28 -0400890 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500891 __btree_submit_bio_start,
892 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400893}
894
Jan Beulich3dd14622010-12-07 14:54:09 +0000895#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500896static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800897 struct page *newpage, struct page *page,
898 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500899{
900 /*
901 * we can't safely write a btree page from here,
902 * we haven't done the locking hook
903 */
904 if (PageDirty(page))
905 return -EAGAIN;
906 /*
907 * Buffers may be managed in a filesystem specific way.
908 * We must have no buffers or drop them.
909 */
910 if (page_has_private(page) &&
911 !try_to_release_page(page, GFP_KERNEL))
912 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800913 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500914}
Jan Beulich3dd14622010-12-07 14:54:09 +0000915#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500916
Chris Mason0da54682007-11-07 21:08:01 -0500917
918static int btree_writepages(struct address_space *mapping,
919 struct writeback_control *wbc)
920{
Chris Masond1310b22008-01-24 16:13:08 -0500921 struct extent_io_tree *tree;
922 tree = &BTRFS_I(mapping->host)->io_tree;
Chris Masond8d5f3e2007-12-11 12:42:00 -0500923 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Masonb9473432009-03-13 11:00:37 -0400924 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
Chris Mason793955b2007-11-26 16:34:41 -0800925 u64 num_dirty;
Chris Mason24ab9cd2008-09-24 14:51:30 -0400926 unsigned long thresh = 32 * 1024 * 1024;
Chris Mason448d6402007-11-27 07:52:01 -0800927
928 if (wbc->for_kupdate)
929 return 0;
930
Chris Masonb9473432009-03-13 11:00:37 -0400931 /* this is a bit racy, but that's ok */
932 num_dirty = root->fs_info->dirty_metadata_bytes;
Chris Masond3977122009-01-05 21:25:51 -0500933 if (num_dirty < thresh)
Chris Mason793955b2007-11-26 16:34:41 -0800934 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800935 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400936 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -0500937}
938
Christoph Hellwigb2950862008-12-02 09:54:17 -0500939static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -0400940{
Chris Masond1310b22008-01-24 16:13:08 -0500941 struct extent_io_tree *tree;
942 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +0200943 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400944}
945
Chris Mason70dec802008-01-29 09:59:12 -0500946static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -0400947{
Chris Mason98509cfc2008-09-11 15:51:43 -0400948 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -0500949 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -0500950 /*
951 * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing
952 * slab allocation from alloc_extent_state down the callchain where
953 * it'd hit a BUG_ON as those flags are not allowed.
954 */
955 gfp_flags &= ~GFP_SLAB_BUG_MASK;
956
Josef Bacik3083ee22012-03-09 16:01:49 -0500957 return try_release_extent_buffer(page, gfp_flags);
Chris Masone20d96d2007-03-22 12:13:20 -0400958}
Chris Mason123abc82007-03-14 14:14:43 -0400959
Chris Mason5f39d392007-10-15 16:14:19 -0400960static void btree_invalidatepage(struct page *page, unsigned long offset)
Chris Masond98237b2007-03-28 13:57:48 -0400961{
Chris Masond1310b22008-01-24 16:13:08 -0500962 struct extent_io_tree *tree;
963 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -0400964 extent_invalidatepage(tree, page, offset);
965 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400966 if (PagePrivate(page)) {
Chris Masond3977122009-01-05 21:25:51 -0500967 printk(KERN_WARNING "btrfs warning page private not zero "
968 "on page %llu\n", (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400969 ClearPagePrivate(page);
970 set_page_private(page, 0);
971 page_cache_release(page);
972 }
Chris Masond98237b2007-03-28 13:57:48 -0400973}
974
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400975static int btree_set_page_dirty(struct page *page)
976{
977 struct extent_buffer *eb;
978
979 BUG_ON(!PagePrivate(page));
980 eb = (struct extent_buffer *)page->private;
981 BUG_ON(!eb);
982 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
983 BUG_ON(!atomic_read(&eb->refs));
984 btrfs_assert_tree_locked(eb);
985 return __set_page_dirty_nobuffers(page);
986}
987
Alexey Dobriyan7f094102009-09-21 17:01:10 -0700988static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -0400989 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -0500990 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -0400991 .releasepage = btree_releasepage,
992 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -0500993#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500994 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -0500995#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400996 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -0400997};
998
Chris Masonca7a79a2008-05-12 12:59:19 -0400999int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1000 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001001{
Chris Mason5f39d392007-10-15 16:14:19 -04001002 struct extent_buffer *buf = NULL;
1003 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001004 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001005
Chris Masondb945352007-10-15 16:15:53 -04001006 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001007 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001008 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001009 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001010 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001011 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001012 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001013}
1014
Arne Jansenab0fff02011-05-23 14:25:41 +02001015int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1016 int mirror_num, struct extent_buffer **eb)
1017{
1018 struct extent_buffer *buf = NULL;
1019 struct inode *btree_inode = root->fs_info->btree_inode;
1020 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1021 int ret;
1022
1023 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1024 if (!buf)
1025 return 0;
1026
1027 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1028
1029 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1030 btree_get_extent, mirror_num);
1031 if (ret) {
1032 free_extent_buffer(buf);
1033 return ret;
1034 }
1035
1036 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1037 free_extent_buffer(buf);
1038 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001039 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001040 *eb = buf;
1041 } else {
1042 free_extent_buffer(buf);
1043 }
1044 return 0;
1045}
1046
Chris Mason0999df52008-04-01 13:48:14 -04001047struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1048 u64 bytenr, u32 blocksize)
1049{
1050 struct inode *btree_inode = root->fs_info->btree_inode;
1051 struct extent_buffer *eb;
1052 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
David Sterbaf09d1f62011-04-21 01:08:01 +02001053 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001054 return eb;
1055}
1056
1057struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1058 u64 bytenr, u32 blocksize)
1059{
1060 struct inode *btree_inode = root->fs_info->btree_inode;
1061 struct extent_buffer *eb;
1062
1063 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
Chris Mason727011e2010-08-06 13:21:20 -04001064 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001065 return eb;
1066}
1067
1068
Chris Masone02119d2008-09-05 16:13:11 -04001069int btrfs_write_tree_block(struct extent_buffer *buf)
1070{
Chris Mason727011e2010-08-06 13:21:20 -04001071 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001072 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001073}
1074
1075int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1076{
Chris Mason727011e2010-08-06 13:21:20 -04001077 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001078 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001079}
1080
Chris Masondb945352007-10-15 16:15:53 -04001081struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001082 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001083{
Chris Mason5f39d392007-10-15 16:14:19 -04001084 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001085 int ret;
1086
Chris Masondb945352007-10-15 16:15:53 -04001087 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001088 if (!buf)
1089 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001090
Chris Masonca7a79a2008-05-12 12:59:19 -04001091 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason5f39d392007-10-15 16:14:19 -04001092 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001093
Chris Masoneb60cea2007-02-02 09:18:22 -05001094}
1095
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001096void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1097 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001098{
Chris Mason55c69072008-01-09 15:55:33 -05001099 if (btrfs_header_generation(buf) ==
Chris Mason925baed2008-06-25 16:01:30 -04001100 root->fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001101 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001102
Chris Masonb9473432009-03-13 11:00:37 -04001103 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1104 spin_lock(&root->fs_info->delalloc_lock);
1105 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1106 root->fs_info->dirty_metadata_bytes -= buf->len;
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001107 else {
1108 spin_unlock(&root->fs_info->delalloc_lock);
1109 btrfs_panic(root->fs_info, -EOVERFLOW,
1110 "Can't clear %lu bytes from "
1111 " dirty_mdatadata_bytes (%lu)",
1112 buf->len,
1113 root->fs_info->dirty_metadata_bytes);
1114 }
Chris Masonb9473432009-03-13 11:00:37 -04001115 spin_unlock(&root->fs_info->delalloc_lock);
1116 }
Chris Masonb4ce94d2009-02-04 09:25:08 -05001117
Chris Masonb9473432009-03-13 11:00:37 -04001118 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1119 btrfs_set_lock_blocking(buf);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001120 clear_extent_buffer_dirty(buf);
Chris Mason925baed2008-06-25 16:01:30 -04001121 }
Chris Mason5f39d392007-10-15 16:14:19 -04001122}
1123
Jeff Mahoney143bede2012-03-01 14:56:26 +01001124static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1125 u32 stripesize, struct btrfs_root *root,
1126 struct btrfs_fs_info *fs_info,
1127 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001128{
Chris Masoncfaa7292007-02-21 17:04:57 -05001129 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001130 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001131 root->sectorsize = sectorsize;
1132 root->nodesize = nodesize;
1133 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001134 root->stripesize = stripesize;
Chris Mason123abc82007-03-14 14:14:43 -04001135 root->ref_cows = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001136 root->track_dirty = 0;
Yan, Zhengc71bf092009-11-12 09:34:40 +00001137 root->in_radix = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001138 root->orphan_item_inserted = 0;
1139 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001140
Chris Mason0f7d52f2007-04-09 10:42:37 -04001141 root->objectid = objectid;
1142 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001143 root->highest_objectid = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001144 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001145 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001146 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001147 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001148 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001149
1150 INIT_LIST_HEAD(&root->dirty_list);
Josef Bacik7b128762008-07-24 12:17:14 -04001151 INIT_LIST_HEAD(&root->orphan_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001152 INIT_LIST_HEAD(&root->root_list);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001153 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001154 spin_lock_init(&root->inode_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001155 spin_lock_init(&root->accounting_lock);
Chris Masona2135012008-06-25 16:01:30 -04001156 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001157 mutex_init(&root->log_mutex);
Yan Zheng7237f1832009-01-21 12:54:03 -05001158 init_waitqueue_head(&root->log_writer_wait);
1159 init_waitqueue_head(&root->log_commit_wait[0]);
1160 init_waitqueue_head(&root->log_commit_wait[1]);
1161 atomic_set(&root->log_commit[0], 0);
1162 atomic_set(&root->log_commit[1], 0);
1163 atomic_set(&root->log_writers, 0);
1164 root->log_batch = 0;
1165 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001166 root->last_log_commit = 0;
Chris Masond0c803c2008-09-11 16:17:57 -04001167 extent_io_tree_init(&root->dirty_log_pages,
David Sterbaf993c882011-04-20 23:35:57 +02001168 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001169
Chris Mason3768f362007-03-13 16:47:54 -04001170 memset(&root->root_key, 0, sizeof(root->root_key));
1171 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001172 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001173 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Chris Mason3f157a22008-06-25 16:01:31 -04001174 root->defrag_trans_start = fs_info->generation;
Josef Bacik58176a92007-08-29 15:47:34 -04001175 init_completion(&root->kobj_unregister);
Chris Mason6702ed42007-08-07 16:15:09 -04001176 root->defrag_running = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001177 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001178 root->anon_dev = 0;
Chris Mason3768f362007-03-13 16:47:54 -04001179}
1180
Jeff Mahoney200a5c12011-10-03 23:22:43 -04001181static int __must_check find_and_setup_root(struct btrfs_root *tree_root,
1182 struct btrfs_fs_info *fs_info,
1183 u64 objectid,
1184 struct btrfs_root *root)
Chris Mason3768f362007-03-13 16:47:54 -04001185{
1186 int ret;
Chris Masondb945352007-10-15 16:15:53 -04001187 u32 blocksize;
Yan Zheng84234f32008-10-29 14:49:05 -04001188 u64 generation;
Chris Mason3768f362007-03-13 16:47:54 -04001189
Chris Masondb945352007-10-15 16:15:53 -04001190 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001191 tree_root->sectorsize, tree_root->stripesize,
1192 root, fs_info, objectid);
Chris Mason3768f362007-03-13 16:47:54 -04001193 ret = btrfs_find_last_root(tree_root, objectid,
1194 &root->root_item, &root->root_key);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001195 if (ret > 0)
1196 return -ENOENT;
Jeff Mahoney200a5c12011-10-03 23:22:43 -04001197 else if (ret < 0)
1198 return ret;
Chris Mason3768f362007-03-13 16:47:54 -04001199
Yan Zheng84234f32008-10-29 14:49:05 -04001200 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001201 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
Chris Masonaf31f5e2011-11-03 15:17:42 -04001202 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001203 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001204 blocksize, generation);
Chris Mason68433b72010-12-13 14:47:58 -05001205 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1206 free_extent_buffer(root->node);
Chris Masonaf31f5e2011-11-03 15:17:42 -04001207 root->node = NULL;
Chris Mason68433b72010-12-13 14:47:58 -05001208 return -EIO;
1209 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001210 root->commit_root = btrfs_root_node(root);
Chris Masond97e63b2007-02-20 16:40:44 -05001211 return 0;
1212}
1213
Al Virof84a8bd2011-11-17 15:57:57 -05001214static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001215{
1216 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1217 if (root)
1218 root->fs_info = fs_info;
1219 return root;
1220}
1221
Yan Zheng7237f1832009-01-21 12:54:03 -05001222static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1223 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001224{
1225 struct btrfs_root *root;
1226 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f1832009-01-21 12:54:03 -05001227 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001228
Al Viro6f07e422011-11-17 00:46:16 -05001229 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001230 if (!root)
Yan Zheng7237f1832009-01-21 12:54:03 -05001231 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001232
1233 __setup_root(tree_root->nodesize, tree_root->leafsize,
1234 tree_root->sectorsize, tree_root->stripesize,
1235 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1236
1237 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1238 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1239 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Yan Zheng7237f1832009-01-21 12:54:03 -05001240 /*
1241 * log trees do not get reference counted because they go away
1242 * before a real commit is actually done. They do store pointers
1243 * to file data extents, and those reference counts still get
1244 * updated (along with back refs to the log tree).
1245 */
Chris Masone02119d2008-09-05 16:13:11 -04001246 root->ref_cows = 0;
1247
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001248 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
Arne Jansen66d7e7f2011-09-12 15:26:38 +02001249 BTRFS_TREE_LOG_OBJECTID, NULL,
1250 0, 0, 0, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001251 if (IS_ERR(leaf)) {
1252 kfree(root);
1253 return ERR_CAST(leaf);
1254 }
Chris Masone02119d2008-09-05 16:13:11 -04001255
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001256 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1257 btrfs_set_header_bytenr(leaf, leaf->start);
1258 btrfs_set_header_generation(leaf, trans->transid);
1259 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1260 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f1832009-01-21 12:54:03 -05001261 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001262
1263 write_extent_buffer(root->node, root->fs_info->fsid,
1264 (unsigned long)btrfs_header_fsid(root->node),
1265 BTRFS_FSID_SIZE);
1266 btrfs_mark_buffer_dirty(root->node);
1267 btrfs_tree_unlock(root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001268 return root;
1269}
1270
1271int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1272 struct btrfs_fs_info *fs_info)
1273{
1274 struct btrfs_root *log_root;
1275
1276 log_root = alloc_log_tree(trans, fs_info);
1277 if (IS_ERR(log_root))
1278 return PTR_ERR(log_root);
1279 WARN_ON(fs_info->log_root_tree);
1280 fs_info->log_root_tree = log_root;
1281 return 0;
1282}
1283
1284int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1285 struct btrfs_root *root)
1286{
1287 struct btrfs_root *log_root;
1288 struct btrfs_inode_item *inode_item;
1289
1290 log_root = alloc_log_tree(trans, root->fs_info);
1291 if (IS_ERR(log_root))
1292 return PTR_ERR(log_root);
1293
1294 log_root->last_trans = trans->transid;
1295 log_root->root_key.offset = root->root_key.objectid;
1296
1297 inode_item = &log_root->root_item.inode;
1298 inode_item->generation = cpu_to_le64(1);
1299 inode_item->size = cpu_to_le64(3);
1300 inode_item->nlink = cpu_to_le32(1);
1301 inode_item->nbytes = cpu_to_le64(root->leafsize);
1302 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1303
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001304 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001305
1306 WARN_ON(root->log_root);
1307 root->log_root = log_root;
1308 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001309 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001310 return 0;
1311}
1312
1313struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1314 struct btrfs_key *location)
1315{
1316 struct btrfs_root *root;
1317 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001318 struct btrfs_path *path;
Chris Mason5f39d392007-10-15 16:14:19 -04001319 struct extent_buffer *l;
Yan Zheng84234f32008-10-29 14:49:05 -04001320 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001321 u32 blocksize;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001322 int ret = 0;
1323
Al Viro6f07e422011-11-17 00:46:16 -05001324 root = btrfs_alloc_root(fs_info);
Chris Mason0cf6c622007-06-09 09:22:25 -04001325 if (!root)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001326 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001327 if (location->offset == (u64)-1) {
Chris Masondb945352007-10-15 16:15:53 -04001328 ret = find_and_setup_root(tree_root, fs_info,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001329 location->objectid, root);
1330 if (ret) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04001331 kfree(root);
1332 return ERR_PTR(ret);
1333 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001334 goto out;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001335 }
1336
Chris Masondb945352007-10-15 16:15:53 -04001337 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001338 tree_root->sectorsize, tree_root->stripesize,
1339 root, fs_info, location->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001340
1341 path = btrfs_alloc_path();
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00001342 if (!path) {
1343 kfree(root);
1344 return ERR_PTR(-ENOMEM);
1345 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001346 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001347 if (ret == 0) {
1348 l = path->nodes[0];
1349 read_extent_buffer(l, &root->root_item,
1350 btrfs_item_ptr_offset(l, path->slots[0]),
1351 sizeof(root->root_item));
1352 memcpy(&root->root_key, location, sizeof(*location));
Chris Mason0f7d52f2007-04-09 10:42:37 -04001353 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001354 btrfs_free_path(path);
1355 if (ret) {
Tsutomu Itoh5e540f72010-12-27 06:53:10 +00001356 kfree(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001357 if (ret > 0)
1358 ret = -ENOENT;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001359 return ERR_PTR(ret);
1360 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001361
Yan Zheng84234f32008-10-29 14:49:05 -04001362 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001363 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1364 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001365 blocksize, generation);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001366 root->commit_root = btrfs_root_node(root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001367 BUG_ON(!root->node); /* -ENOMEM */
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001368out:
Li Zefan08fe4db2011-03-28 02:01:25 +00001369 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
Chris Masone02119d2008-09-05 16:13:11 -04001370 root->ref_cows = 1;
Li Zefan08fe4db2011-03-28 02:01:25 +00001371 btrfs_check_and_init_root_item(&root->root_item);
1372 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001373
Chris Mason5eda7b52007-06-22 14:16:25 -04001374 return root;
1375}
1376
Chris Masonedbd8d42007-12-21 16:27:24 -05001377struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1378 struct btrfs_key *location)
Chris Mason5eda7b52007-06-22 14:16:25 -04001379{
1380 struct btrfs_root *root;
1381 int ret;
1382
Chris Masonedbd8d42007-12-21 16:27:24 -05001383 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1384 return fs_info->tree_root;
1385 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1386 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001387 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1388 return fs_info->chunk_root;
1389 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1390 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001391 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1392 return fs_info->csum_root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001393again:
1394 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001395 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1396 (unsigned long)location->objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001397 spin_unlock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001398 if (root)
1399 return root;
1400
Chris Masone02119d2008-09-05 16:13:11 -04001401 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001402 if (IS_ERR(root))
1403 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001404
Li Zefan581bb052011-04-20 10:06:11 +08001405 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
Li Zefan581bb052011-04-20 10:06:11 +08001406 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1407 GFP_NOFS);
David Sterba35a30d72011-06-13 15:18:23 +00001408 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1409 ret = -ENOMEM;
Li Zefan581bb052011-04-20 10:06:11 +08001410 goto fail;
David Sterba35a30d72011-06-13 15:18:23 +00001411 }
Li Zefan581bb052011-04-20 10:06:11 +08001412
1413 btrfs_init_free_ino_ctl(root);
1414 mutex_init(&root->fs_commit_mutex);
1415 spin_lock_init(&root->cache_lock);
1416 init_waitqueue_head(&root->cache_wait);
1417
Al Viro0ee5dc62011-07-07 15:44:25 -04001418 ret = get_anon_bdev(&root->anon_dev);
Chris Masonac08aed2011-06-13 11:28:50 -04001419 if (ret)
1420 goto fail;
Chris Mason3394e162008-11-17 20:42:26 -05001421
Yan, Zhengd68fc572010-05-16 10:49:58 -04001422 if (btrfs_root_refs(&root->root_item) == 0) {
1423 ret = -ENOENT;
1424 goto fail;
1425 }
1426
1427 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1428 if (ret < 0)
1429 goto fail;
1430 if (ret == 0)
1431 root->orphan_item_inserted = 1;
1432
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001433 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1434 if (ret)
1435 goto fail;
1436
1437 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04001438 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1439 (unsigned long)root->root_key.objectid,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001440 root);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001441 if (ret == 0)
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001442 root->in_radix = 1;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001443
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001444 spin_unlock(&fs_info->fs_roots_radix_lock);
1445 radix_tree_preload_end();
Chris Mason0f7d52f2007-04-09 10:42:37 -04001446 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001447 if (ret == -EEXIST) {
1448 free_fs_root(root);
1449 goto again;
1450 }
1451 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001452 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001453
1454 ret = btrfs_find_dead_roots(fs_info->tree_root,
1455 root->root_key.objectid);
1456 WARN_ON(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001457 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001458fail:
1459 free_fs_root(root);
1460 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001461}
1462
Chris Mason04160082008-03-26 10:28:07 -04001463static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1464{
1465 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1466 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001467 struct btrfs_device *device;
1468 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001469
Xiao Guangrong1f781602011-04-20 10:09:16 +00001470 rcu_read_lock();
1471 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001472 if (!device->bdev)
1473 continue;
Chris Mason04160082008-03-26 10:28:07 -04001474 bdi = blk_get_backing_dev_info(device->bdev);
1475 if (bdi && bdi_congested(bdi, bdi_bits)) {
1476 ret = 1;
1477 break;
1478 }
1479 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001480 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001481 return ret;
1482}
1483
Chris Mason38b66982008-04-22 09:22:11 -04001484/*
Jens Axboead081f12009-06-12 14:43:40 +02001485 * If this fails, caller must call bdi_destroy() to get rid of the
1486 * bdi again.
1487 */
Chris Mason04160082008-03-26 10:28:07 -04001488static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1489{
Jens Axboead081f12009-06-12 14:43:40 +02001490 int err;
1491
1492 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001493 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001494 if (err)
1495 return err;
1496
Chris Mason4575c9c2008-04-18 16:13:31 -04001497 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001498 bdi->congested_fn = btrfs_congested_fn;
1499 bdi->congested_data = info;
1500 return 0;
1501}
1502
Chris Mason8b712842008-06-11 16:50:36 -04001503/*
1504 * called by the kthread helper functions to finally call the bio end_io
1505 * functions. This is where read checksum verification actually happens
1506 */
1507static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001508{
Chris Masonce9adaa2008-04-09 16:28:12 -04001509 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001510 struct end_io_wq *end_io_wq;
1511 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001512 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001513
Chris Mason8b712842008-06-11 16:50:36 -04001514 end_io_wq = container_of(work, struct end_io_wq, work);
1515 bio = end_io_wq->bio;
1516 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001517
Chris Mason8b712842008-06-11 16:50:36 -04001518 error = end_io_wq->error;
1519 bio->bi_private = end_io_wq->private;
1520 bio->bi_end_io = end_io_wq->end_io;
1521 kfree(end_io_wq);
Chris Mason8b712842008-06-11 16:50:36 -04001522 bio_endio(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001523}
1524
Chris Masona74a4b92008-06-25 16:01:31 -04001525static int cleaner_kthread(void *arg)
1526{
1527 struct btrfs_root *root = arg;
1528
1529 do {
Chris Masona74a4b92008-06-25 16:01:31 -04001530 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
Yan, Zheng76dda932009-09-21 16:00:26 -04001531
1532 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1533 mutex_trylock(&root->fs_info->cleaner_mutex)) {
Yan, Zheng24bbcf042009-11-12 09:36:34 +00001534 btrfs_run_delayed_iputs(root);
Yan, Zheng76dda932009-09-21 16:00:26 -04001535 btrfs_clean_old_snapshots(root);
1536 mutex_unlock(&root->fs_info->cleaner_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001537 btrfs_run_defrag_inodes(root->fs_info);
Yan, Zheng76dda932009-09-21 16:00:26 -04001538 }
Chris Masona74a4b92008-06-25 16:01:31 -04001539
Tejun Heoa0acae02011-11-21 12:32:22 -08001540 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001541 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001542 if (!kthread_should_stop())
1543 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001544 __set_current_state(TASK_RUNNING);
1545 }
1546 } while (!kthread_should_stop());
1547 return 0;
1548}
1549
1550static int transaction_kthread(void *arg)
1551{
1552 struct btrfs_root *root = arg;
1553 struct btrfs_trans_handle *trans;
1554 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001555 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001556 unsigned long now;
1557 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001558 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001559
1560 do {
Jan Kara914b2002012-03-12 16:05:50 +01001561 cannot_commit = false;
Chris Masona74a4b92008-06-25 16:01:31 -04001562 delay = HZ * 30;
1563 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1564 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1565
Josef Bacika4abeea2011-04-11 17:25:13 -04001566 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001567 cur = root->fs_info->running_transaction;
1568 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001569 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001570 goto sleep;
1571 }
Yan Zheng31153d82008-07-28 15:32:19 -04001572
Chris Masona74a4b92008-06-25 16:01:31 -04001573 now = get_seconds();
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001574 if (!cur->blocked &&
1575 (now < cur->start_time || now - cur->start_time < 30)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001576 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001577 delay = HZ * 5;
1578 goto sleep;
1579 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001580 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001581 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001582
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001583 /* If the file system is aborted, this will always fail. */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04001584 trans = btrfs_join_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001585 if (IS_ERR(trans)) {
1586 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001587 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001588 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001589 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001590 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001591 } else {
1592 btrfs_end_transaction(trans, root);
1593 }
Chris Masona74a4b92008-06-25 16:01:31 -04001594sleep:
1595 wake_up_process(root->fs_info->cleaner_kthread);
1596 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1597
Tejun Heoa0acae02011-11-21 12:32:22 -08001598 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001599 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001600 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001601 (!btrfs_transaction_blocked(root->fs_info) ||
1602 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001603 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001604 __set_current_state(TASK_RUNNING);
1605 }
1606 } while (!kthread_should_stop());
1607 return 0;
1608}
1609
Chris Masonaf31f5e2011-11-03 15:17:42 -04001610/*
1611 * this will find the highest generation in the array of
1612 * root backups. The index of the highest array is returned,
1613 * or -1 if we can't find anything.
1614 *
1615 * We check to make sure the array is valid by comparing the
1616 * generation of the latest root in the array with the generation
1617 * in the super block. If they don't match we pitch it.
1618 */
1619static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1620{
1621 u64 cur;
1622 int newest_index = -1;
1623 struct btrfs_root_backup *root_backup;
1624 int i;
1625
1626 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1627 root_backup = info->super_copy->super_roots + i;
1628 cur = btrfs_backup_tree_root_gen(root_backup);
1629 if (cur == newest_gen)
1630 newest_index = i;
1631 }
1632
1633 /* check to see if we actually wrapped around */
1634 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1635 root_backup = info->super_copy->super_roots;
1636 cur = btrfs_backup_tree_root_gen(root_backup);
1637 if (cur == newest_gen)
1638 newest_index = 0;
1639 }
1640 return newest_index;
1641}
1642
1643
1644/*
1645 * find the oldest backup so we know where to store new entries
1646 * in the backup array. This will set the backup_root_index
1647 * field in the fs_info struct
1648 */
1649static void find_oldest_super_backup(struct btrfs_fs_info *info,
1650 u64 newest_gen)
1651{
1652 int newest_index = -1;
1653
1654 newest_index = find_newest_super_backup(info, newest_gen);
1655 /* if there was garbage in there, just move along */
1656 if (newest_index == -1) {
1657 info->backup_root_index = 0;
1658 } else {
1659 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1660 }
1661}
1662
1663/*
1664 * copy all the root pointers into the super backup array.
1665 * this will bump the backup pointer by one when it is
1666 * done
1667 */
1668static void backup_super_roots(struct btrfs_fs_info *info)
1669{
1670 int next_backup;
1671 struct btrfs_root_backup *root_backup;
1672 int last_backup;
1673
1674 next_backup = info->backup_root_index;
1675 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1676 BTRFS_NUM_BACKUP_ROOTS;
1677
1678 /*
1679 * just overwrite the last backup if we're at the same generation
1680 * this happens only at umount
1681 */
1682 root_backup = info->super_for_commit->super_roots + last_backup;
1683 if (btrfs_backup_tree_root_gen(root_backup) ==
1684 btrfs_header_generation(info->tree_root->node))
1685 next_backup = last_backup;
1686
1687 root_backup = info->super_for_commit->super_roots + next_backup;
1688
1689 /*
1690 * make sure all of our padding and empty slots get zero filled
1691 * regardless of which ones we use today
1692 */
1693 memset(root_backup, 0, sizeof(*root_backup));
1694
1695 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1696
1697 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1698 btrfs_set_backup_tree_root_gen(root_backup,
1699 btrfs_header_generation(info->tree_root->node));
1700
1701 btrfs_set_backup_tree_root_level(root_backup,
1702 btrfs_header_level(info->tree_root->node));
1703
1704 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1705 btrfs_set_backup_chunk_root_gen(root_backup,
1706 btrfs_header_generation(info->chunk_root->node));
1707 btrfs_set_backup_chunk_root_level(root_backup,
1708 btrfs_header_level(info->chunk_root->node));
1709
1710 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1711 btrfs_set_backup_extent_root_gen(root_backup,
1712 btrfs_header_generation(info->extent_root->node));
1713 btrfs_set_backup_extent_root_level(root_backup,
1714 btrfs_header_level(info->extent_root->node));
1715
Chris Mason7c7e82a2011-11-06 18:50:56 -05001716 /*
1717 * we might commit during log recovery, which happens before we set
1718 * the fs_root. Make sure it is valid before we fill it in.
1719 */
1720 if (info->fs_root && info->fs_root->node) {
1721 btrfs_set_backup_fs_root(root_backup,
1722 info->fs_root->node->start);
1723 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001724 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001725 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001726 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001727 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001728
1729 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1730 btrfs_set_backup_dev_root_gen(root_backup,
1731 btrfs_header_generation(info->dev_root->node));
1732 btrfs_set_backup_dev_root_level(root_backup,
1733 btrfs_header_level(info->dev_root->node));
1734
1735 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1736 btrfs_set_backup_csum_root_gen(root_backup,
1737 btrfs_header_generation(info->csum_root->node));
1738 btrfs_set_backup_csum_root_level(root_backup,
1739 btrfs_header_level(info->csum_root->node));
1740
1741 btrfs_set_backup_total_bytes(root_backup,
1742 btrfs_super_total_bytes(info->super_copy));
1743 btrfs_set_backup_bytes_used(root_backup,
1744 btrfs_super_bytes_used(info->super_copy));
1745 btrfs_set_backup_num_devices(root_backup,
1746 btrfs_super_num_devices(info->super_copy));
1747
1748 /*
1749 * if we don't copy this out to the super_copy, it won't get remembered
1750 * for the next commit
1751 */
1752 memcpy(&info->super_copy->super_roots,
1753 &info->super_for_commit->super_roots,
1754 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1755}
1756
1757/*
1758 * this copies info out of the root backup array and back into
1759 * the in-memory super block. It is meant to help iterate through
1760 * the array, so you send it the number of backups you've already
1761 * tried and the last backup index you used.
1762 *
1763 * this returns -1 when it has tried all the backups
1764 */
1765static noinline int next_root_backup(struct btrfs_fs_info *info,
1766 struct btrfs_super_block *super,
1767 int *num_backups_tried, int *backup_index)
1768{
1769 struct btrfs_root_backup *root_backup;
1770 int newest = *backup_index;
1771
1772 if (*num_backups_tried == 0) {
1773 u64 gen = btrfs_super_generation(super);
1774
1775 newest = find_newest_super_backup(info, gen);
1776 if (newest == -1)
1777 return -1;
1778
1779 *backup_index = newest;
1780 *num_backups_tried = 1;
1781 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1782 /* we've tried all the backups, all done */
1783 return -1;
1784 } else {
1785 /* jump to the next oldest backup */
1786 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1787 BTRFS_NUM_BACKUP_ROOTS;
1788 *backup_index = newest;
1789 *num_backups_tried += 1;
1790 }
1791 root_backup = super->super_roots + newest;
1792
1793 btrfs_set_super_generation(super,
1794 btrfs_backup_tree_root_gen(root_backup));
1795 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1796 btrfs_set_super_root_level(super,
1797 btrfs_backup_tree_root_level(root_backup));
1798 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1799
1800 /*
1801 * fixme: the total bytes and num_devices need to match or we should
1802 * need a fsck
1803 */
1804 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1805 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1806 return 0;
1807}
1808
1809/* helper to cleanup tree roots */
1810static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
1811{
1812 free_extent_buffer(info->tree_root->node);
1813 free_extent_buffer(info->tree_root->commit_root);
1814 free_extent_buffer(info->dev_root->node);
1815 free_extent_buffer(info->dev_root->commit_root);
1816 free_extent_buffer(info->extent_root->node);
1817 free_extent_buffer(info->extent_root->commit_root);
1818 free_extent_buffer(info->csum_root->node);
1819 free_extent_buffer(info->csum_root->commit_root);
1820
1821 info->tree_root->node = NULL;
1822 info->tree_root->commit_root = NULL;
1823 info->dev_root->node = NULL;
1824 info->dev_root->commit_root = NULL;
1825 info->extent_root->node = NULL;
1826 info->extent_root->commit_root = NULL;
1827 info->csum_root->node = NULL;
1828 info->csum_root->commit_root = NULL;
1829
1830 if (chunk_root) {
1831 free_extent_buffer(info->chunk_root->node);
1832 free_extent_buffer(info->chunk_root->commit_root);
1833 info->chunk_root->node = NULL;
1834 info->chunk_root->commit_root = NULL;
1835 }
1836}
1837
1838
Al Viroad2b2c82011-11-17 01:10:02 -05001839int open_ctree(struct super_block *sb,
1840 struct btrfs_fs_devices *fs_devices,
1841 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05001842{
Chris Masondb945352007-10-15 16:15:53 -04001843 u32 sectorsize;
1844 u32 nodesize;
1845 u32 leafsize;
1846 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001847 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04001848 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001849 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05001850 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04001851 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001852 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05001853 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05001854 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001855 struct btrfs_root *extent_root;
1856 struct btrfs_root *csum_root;
1857 struct btrfs_root *chunk_root;
1858 struct btrfs_root *dev_root;
Chris Masone02119d2008-09-05 16:13:11 -04001859 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05001860 int ret;
Yane58ca022008-04-01 11:21:34 -04001861 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04001862 int num_backups_tried = 0;
1863 int backup_index = 0;
Chris Mason4543df72008-06-11 21:47:56 -04001864
Al Virof84a8bd2011-11-17 15:57:57 -05001865 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05001866 extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info);
1867 csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info);
1868 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
1869 dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info);
Chris Mason8790d502008-04-03 16:29:03 -04001870
Al Virof84a8bd2011-11-17 15:57:57 -05001871 if (!tree_root || !extent_root || !csum_root ||
1872 !chunk_root || !dev_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04001873 err = -ENOMEM;
1874 goto fail;
1875 }
Yan, Zheng76dda932009-09-21 16:00:26 -04001876
1877 ret = init_srcu_struct(&fs_info->subvol_srcu);
1878 if (ret) {
1879 err = ret;
1880 goto fail;
1881 }
1882
1883 ret = setup_bdi(fs_info, &fs_info->bdi);
1884 if (ret) {
1885 err = ret;
1886 goto fail_srcu;
1887 }
1888
1889 fs_info->btree_inode = new_inode(sb);
1890 if (!fs_info->btree_inode) {
1891 err = -ENOMEM;
1892 goto fail_bdi;
1893 }
1894
Chris Masona6591712011-07-19 12:04:14 -04001895 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08001896
Yan, Zheng76dda932009-09-21 16:00:26 -04001897 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04001898 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04001899 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00001900 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Chris Mason19c00dd2007-10-15 16:19:22 -04001901 INIT_LIST_HEAD(&fs_info->hashers);
Chris Masonea8c2812008-08-04 23:17:27 -04001902 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
Chris Mason5a3f23d2009-03-31 13:27:11 -04001903 INIT_LIST_HEAD(&fs_info->ordered_operations);
Yan Zheng11833d62009-09-11 16:11:19 -04001904 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Chris Mason1832a6d2007-12-21 16:27:21 -05001905 spin_lock_init(&fs_info->delalloc_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04001906 spin_lock_init(&fs_info->trans_lock);
Yan Zheng31153d82008-07-28 15:32:19 -04001907 spin_lock_init(&fs_info->ref_cache_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04001908 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00001909 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04001910 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04001911 spin_lock_init(&fs_info->free_chunk_lock);
Chris Mason75857172011-06-13 20:00:16 -04001912 mutex_init(&fs_info->reloc_mutex);
Chris Mason19c00dd2007-10-15 16:19:22 -04001913
Josef Bacik58176a92007-08-29 15:47:34 -04001914 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04001915 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04001916 INIT_LIST_HEAD(&fs_info->space_info);
Chris Mason0b86a832008-03-24 15:01:56 -04001917 btrfs_mapping_init(&fs_info->mapping_tree);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001918 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1919 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1920 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1921 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1922 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
Josef Bacik6d668dd2011-11-03 22:54:25 -04001923 btrfs_init_block_rsv(&fs_info->delayed_block_rsv);
Chris Masoncb03c742008-05-15 16:15:45 -04001924 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05001925 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04001926 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04001927 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04001928 atomic_set(&fs_info->defrag_running, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04001929 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05001930 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04001931 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04001932 fs_info->defrag_inodes = RB_ROOT;
Josef Bacika4abeea2011-04-11 17:25:13 -04001933 fs_info->trans_no_join = 0;
Josef Bacik2bf64752011-09-26 17:12:22 -04001934 fs_info->free_chunk_space = 0;
Chris Masonc8b97812008-10-29 14:49:59 -04001935
Arne Jansen90519d62011-05-23 14:30:00 +02001936 /* readahead state */
1937 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
1938 spin_lock_init(&fs_info->reada_lock);
Chris Mason2c90e5d2007-04-02 10:50:19 -04001939
Chris Masonb34b0862008-12-19 15:43:22 -05001940 fs_info->thread_pool_size = min_t(unsigned long,
1941 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04001942
Chris Mason3eaa2882008-07-24 11:57:52 -04001943 INIT_LIST_HEAD(&fs_info->ordered_extents);
1944 spin_lock_init(&fs_info->ordered_extent_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08001945 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
1946 GFP_NOFS);
1947 if (!fs_info->delayed_root) {
1948 err = -ENOMEM;
1949 goto fail_iput;
1950 }
1951 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04001952
Arne Jansena2de7332011-03-08 14:14:00 +01001953 mutex_init(&fs_info->scrub_lock);
1954 atomic_set(&fs_info->scrubs_running, 0);
1955 atomic_set(&fs_info->scrub_pause_req, 0);
1956 atomic_set(&fs_info->scrubs_paused, 0);
1957 atomic_set(&fs_info->scrub_cancel_req, 0);
1958 init_waitqueue_head(&fs_info->scrub_pause_wait);
1959 init_rwsem(&fs_info->scrub_super_lock);
1960 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01001961#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1962 fs_info->check_integrity_print_mask = 0;
1963#endif
Arne Jansena2de7332011-03-08 14:14:00 +01001964
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001965 spin_lock_init(&fs_info->balance_lock);
1966 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02001967 atomic_set(&fs_info->balance_running, 0);
1968 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02001969 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02001970 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02001971 init_waitqueue_head(&fs_info->balance_wait_q);
Chris Masona061fc82008-05-07 11:43:44 -04001972
Chris Mason0afbaf82008-04-18 14:17:20 -04001973 sb->s_blocksize = 4096;
1974 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02001975 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04001976
Yan, Zheng76dda932009-09-21 16:00:26 -04001977 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02001978 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04001979 /*
1980 * we set the i_size on the btree inode to the max possible int.
1981 * the real end of the address space is determined by all of
1982 * the devices in the system
1983 */
1984 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04001985 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04001986 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1987
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001988 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05001989 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02001990 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001991 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02001992 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05001993
Chris Masond1310b22008-01-24 16:13:08 -05001994 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1995
Chris Mason0f7d52f2007-04-09 10:42:37 -04001996 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1997 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1998 sizeof(struct btrfs_key));
Yan, Zheng76dda932009-09-21 16:00:26 -04001999 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
Chris Mason22b0ebd2007-03-30 08:47:31 -04002000 insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002001
Chris Masone02119d2008-09-05 16:13:11 -04002002 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002003 fs_info->block_group_cache_tree = RB_ROOT;
Chris Mason925baed2008-06-25 16:01:30 -04002004
Yan Zheng11833d62009-09-11 16:11:19 -04002005 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002006 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002007 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002008 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002009 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002010 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002011
Chris Masond98237b2007-03-28 13:57:48 -04002012
Chris Mason5a3f23d2009-03-31 13:27:11 -04002013 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002014 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002015 mutex_init(&fs_info->chunk_mutex);
2016 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002017 mutex_init(&fs_info->cleaner_mutex);
2018 mutex_init(&fs_info->volume_mutex);
Yan Zheng276e6802009-07-30 09:40:40 -04002019 init_rwsem(&fs_info->extent_commit_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002020 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002021 init_rwsem(&fs_info->subvol_sem);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002022
2023 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2024 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2025
Chris Mason7d9eb122008-07-08 14:19:17 -04002026 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002027 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002028 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002029 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002030
Chris Mason0b86a832008-03-24 15:01:56 -04002031 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002032 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002033
Yan Zhenga512bbf2008-12-08 16:46:26 -05002034 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002035 if (!bh) {
2036 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002037 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002038 }
Chris Mason39279cc2007-06-12 06:35:45 -04002039
David Sterba6c417612011-04-13 15:41:04 +02002040 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2041 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2042 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002043 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002044
David Sterba6c417612011-04-13 15:41:04 +02002045 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002046
David Sterba6c417612011-04-13 15:41:04 +02002047 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002048 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002049 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002050
liuboacce9522011-01-06 19:30:25 +08002051 /* check FS state, whether FS is broken. */
2052 fs_info->fs_state |= btrfs_super_flags(disk_super);
2053
David Sterbafcd1f0652012-03-06 00:06:18 +01002054 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2055 if (ret) {
2056 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2057 err = ret;
2058 goto fail_alloc;
2059 }
liuboacce9522011-01-06 19:30:25 +08002060
Liu Bo75e7cb72011-03-22 10:12:20 +00002061 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002062 * run through our array of backup supers and setup
2063 * our ring pointer to the oldest one
2064 */
2065 generation = btrfs_super_generation(disk_super);
2066 find_oldest_super_backup(fs_info, generation);
2067
2068 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002069 * In the long term, we'll store the compression type in the super
2070 * block, and it'll be used for per file compression control.
2071 */
2072 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2073
Yan Zheng2b820322008-11-17 21:11:30 -05002074 ret = btrfs_parse_options(tree_root, options);
2075 if (ret) {
2076 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002077 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002078 }
Chris Masondfe25022008-05-13 13:46:40 -04002079
Josef Bacikf2b636e2008-12-02 06:36:08 -05002080 features = btrfs_super_incompat_flags(disk_super) &
2081 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2082 if (features) {
2083 printk(KERN_ERR "BTRFS: couldn't mount because of "
2084 "unsupported optional features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002085 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002086 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002087 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002088 }
2089
Chris Mason727011e2010-08-06 13:21:20 -04002090 if (btrfs_super_leafsize(disk_super) !=
2091 btrfs_super_nodesize(disk_super)) {
2092 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2093 "blocksizes don't match. node %d leaf %d\n",
2094 btrfs_super_nodesize(disk_super),
2095 btrfs_super_leafsize(disk_super));
2096 err = -EINVAL;
2097 goto fail_alloc;
2098 }
2099 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2100 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2101 "blocksize (%d) was too large\n",
2102 btrfs_super_leafsize(disk_super));
2103 err = -EINVAL;
2104 goto fail_alloc;
2105 }
2106
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002107 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002108 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
2109 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
2110 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002111
2112 /*
2113 * flag our filesystem as having big metadata blocks if
2114 * they are bigger than the page size
2115 */
2116 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2117 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2118 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2119 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2120 }
2121
Li Zefana6fa6fa2010-10-25 15:12:26 +08002122 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002123
Josef Bacikf2b636e2008-12-02 06:36:08 -05002124 features = btrfs_super_compat_ro_flags(disk_super) &
2125 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2126 if (!(sb->s_flags & MS_RDONLY) && features) {
2127 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2128 "unsupported option features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002129 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002130 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002131 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002132 }
Chris Mason61d92c32009-10-02 19:11:56 -04002133
2134 btrfs_init_workers(&fs_info->generic_worker,
2135 "genwork", 1, NULL);
2136
Chris Mason5443be42008-08-15 15:34:16 -04002137 btrfs_init_workers(&fs_info->workers, "worker",
Chris Mason61d92c32009-10-02 19:11:56 -04002138 fs_info->thread_pool_size,
2139 &fs_info->generic_worker);
Chris Masonc8b97812008-10-29 14:49:59 -04002140
Chris Mason771ed682008-11-06 22:02:51 -05002141 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
Chris Mason61d92c32009-10-02 19:11:56 -04002142 fs_info->thread_pool_size,
2143 &fs_info->generic_worker);
Chris Mason771ed682008-11-06 22:02:51 -05002144
Chris Mason5443be42008-08-15 15:34:16 -04002145 btrfs_init_workers(&fs_info->submit_workers, "submit",
Chris Masonb720d202008-08-15 15:34:14 -04002146 min_t(u64, fs_devices->num_devices,
Chris Mason61d92c32009-10-02 19:11:56 -04002147 fs_info->thread_pool_size),
2148 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002149
Josef Bacikbab39bf2011-06-30 14:42:28 -04002150 btrfs_init_workers(&fs_info->caching_workers, "cache",
2151 2, &fs_info->generic_worker);
2152
Chris Mason61b49442008-07-31 15:42:53 -04002153 /* a higher idle thresh on the submit workers makes it much more
2154 * likely that bios will be send down in a sane order to the
2155 * devices
2156 */
2157 fs_info->submit_workers.idle_thresh = 64;
Chris Mason53863232008-08-15 15:34:18 -04002158
Chris Mason771ed682008-11-06 22:02:51 -05002159 fs_info->workers.idle_thresh = 16;
Chris Mason4a69a412008-11-06 22:03:00 -05002160 fs_info->workers.ordered = 1;
Chris Mason61b49442008-07-31 15:42:53 -04002161
Chris Mason771ed682008-11-06 22:02:51 -05002162 fs_info->delalloc_workers.idle_thresh = 2;
2163 fs_info->delalloc_workers.ordered = 1;
2164
Chris Mason61d92c32009-10-02 19:11:56 -04002165 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2166 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002167 btrfs_init_workers(&fs_info->endio_workers, "endio",
Chris Mason61d92c32009-10-02 19:11:56 -04002168 fs_info->thread_pool_size,
2169 &fs_info->generic_worker);
Chris Masond20f7042008-12-08 16:58:54 -05002170 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
Chris Mason61d92c32009-10-02 19:11:56 -04002171 fs_info->thread_pool_size,
2172 &fs_info->generic_worker);
Chris Masoncad321a2008-12-17 14:51:42 -05002173 btrfs_init_workers(&fs_info->endio_meta_write_workers,
Chris Mason61d92c32009-10-02 19:11:56 -04002174 "endio-meta-write", fs_info->thread_pool_size,
2175 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002176 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
Chris Mason61d92c32009-10-02 19:11:56 -04002177 fs_info->thread_pool_size,
2178 &fs_info->generic_worker);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002179 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2180 1, &fs_info->generic_worker);
Miao Xie16cdcec2011-04-22 18:12:22 +08002181 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2182 fs_info->thread_pool_size,
2183 &fs_info->generic_worker);
Arne Jansen90519d62011-05-23 14:30:00 +02002184 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2185 fs_info->thread_pool_size,
2186 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002187
2188 /*
2189 * endios are largely parallel and should have a very
2190 * low idle thresh
2191 */
2192 fs_info->endio_workers.idle_thresh = 4;
Chris Masonb51912c2009-02-04 09:23:24 -05002193 fs_info->endio_meta_workers.idle_thresh = 4;
2194
Chris Mason90428462009-08-04 16:56:34 -04002195 fs_info->endio_write_workers.idle_thresh = 2;
2196 fs_info->endio_meta_write_workers.idle_thresh = 2;
Arne Jansen90519d62011-05-23 14:30:00 +02002197 fs_info->readahead_workers.idle_thresh = 2;
Chris Mason90428462009-08-04 16:56:34 -04002198
Josef Bacik0dc3b842011-11-18 14:37:27 -05002199 /*
2200 * btrfs_start_workers can really only fail because of ENOMEM so just
2201 * return -ENOMEM if any of these fail.
2202 */
2203 ret = btrfs_start_workers(&fs_info->workers);
2204 ret |= btrfs_start_workers(&fs_info->generic_worker);
2205 ret |= btrfs_start_workers(&fs_info->submit_workers);
2206 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2207 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2208 ret |= btrfs_start_workers(&fs_info->endio_workers);
2209 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
2210 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2211 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2212 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2213 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2214 ret |= btrfs_start_workers(&fs_info->caching_workers);
2215 ret |= btrfs_start_workers(&fs_info->readahead_workers);
2216 if (ret) {
2217 ret = -ENOMEM;
2218 goto fail_sb_buffer;
2219 }
Chris Mason4543df72008-06-11 21:47:56 -04002220
Chris Mason4575c9c2008-04-18 16:13:31 -04002221 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002222 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2223 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002224
Chris Masondb945352007-10-15 16:15:53 -04002225 nodesize = btrfs_super_nodesize(disk_super);
2226 leafsize = btrfs_super_leafsize(disk_super);
2227 sectorsize = btrfs_super_sectorsize(disk_super);
Chris Mason87ee04e2007-11-30 11:30:34 -05002228 stripesize = btrfs_super_stripesize(disk_super);
Chris Masondb945352007-10-15 16:15:53 -04002229 tree_root->nodesize = nodesize;
2230 tree_root->leafsize = leafsize;
2231 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002232 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002233
2234 sb->s_blocksize = sectorsize;
2235 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002236
Chris Mason39279cc2007-06-12 06:35:45 -04002237 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
2238 sizeof(disk_super->magic))) {
Chris Masond3977122009-01-05 21:25:51 -05002239 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002240 goto fail_sb_buffer;
2241 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002242
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002243 if (sectorsize < PAGE_SIZE) {
2244 printk(KERN_WARNING "btrfs: Incompatible sector size "
2245 "found on %s\n", sb->s_id);
2246 goto fail_sb_buffer;
2247 }
2248
David Sterba65139ed2012-02-17 12:26:09 +01002249 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2250 (leafsize != nodesize || sectorsize != nodesize)) {
2251 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2252 "are not allowed for mixed block groups on %s\n",
2253 sb->s_id);
2254 goto fail_sb_buffer;
2255 }
2256
Chris Mason925baed2008-06-25 16:01:30 -04002257 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002258 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002259 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002260 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002261 printk(KERN_WARNING "btrfs: failed to read the system "
2262 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002263 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002264 }
Chris Mason0b86a832008-03-24 15:01:56 -04002265
2266 blocksize = btrfs_level_size(tree_root,
2267 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002268 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002269
2270 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2271 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2272
2273 chunk_root->node = read_tree_block(chunk_root,
2274 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002275 blocksize, generation);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002276 BUG_ON(!chunk_root->node); /* -ENOMEM */
David Woodhouse83121942009-07-22 16:52:13 -04002277 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2278 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2279 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002280 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002281 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002282 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2283 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002284
Chris Masone17cade2008-04-15 15:41:47 -04002285 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Chris Masond3977122009-01-05 21:25:51 -05002286 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2287 BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002288
Chris Mason0b86a832008-03-24 15:01:56 -04002289 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002290 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002291 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2292 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002293 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002294 }
Chris Mason0b86a832008-03-24 15:01:56 -04002295
Chris Masondfe25022008-05-13 13:46:40 -04002296 btrfs_close_extra_devices(fs_devices);
2297
Chris Masona6b0d5c2012-02-20 20:53:43 -05002298 if (!fs_devices->latest_bdev) {
2299 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2300 sb->s_id);
2301 goto fail_tree_roots;
2302 }
2303
Chris Masonaf31f5e2011-11-03 15:17:42 -04002304retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002305 blocksize = btrfs_level_size(tree_root,
2306 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002307 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002308
Chris Masone20d96d2007-03-22 12:13:20 -04002309 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002310 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002311 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002312 if (!tree_root->node ||
2313 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
David Woodhouse83121942009-07-22 16:52:13 -04002314 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2315 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002316
2317 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002318 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002319
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002320 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2321 tree_root->commit_root = btrfs_root_node(tree_root);
Chris Masondb945352007-10-15 16:15:53 -04002322
2323 ret = find_and_setup_root(tree_root, fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04002324 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002325 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002326 goto recovery_tree_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002327 extent_root->track_dirty = 1;
2328
2329 ret = find_and_setup_root(tree_root, fs_info,
2330 BTRFS_DEV_TREE_OBJECTID, dev_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002331 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002332 goto recovery_tree_root;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002333 dev_root->track_dirty = 1;
Chris Mason3768f362007-03-13 16:47:54 -04002334
Chris Masond20f7042008-12-08 16:58:54 -05002335 ret = find_and_setup_root(tree_root, fs_info,
2336 BTRFS_CSUM_TREE_OBJECTID, csum_root);
2337 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002338 goto recovery_tree_root;
Chris Masond20f7042008-12-08 16:58:54 -05002339
2340 csum_root->track_dirty = 1;
2341
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002342 fs_info->generation = generation;
2343 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002344
liuboc59021f2011-03-07 02:13:14 +00002345 ret = btrfs_init_space_info(fs_info);
2346 if (ret) {
2347 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2348 goto fail_block_groups;
2349 }
2350
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002351 ret = btrfs_read_block_groups(extent_root);
2352 if (ret) {
2353 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2354 goto fail_block_groups;
2355 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002356
Chris Masona74a4b92008-06-25 16:01:31 -04002357 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2358 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002359 if (IS_ERR(fs_info->cleaner_kthread))
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002360 goto fail_block_groups;
Chris Masona74a4b92008-06-25 16:01:31 -04002361
2362 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2363 tree_root,
2364 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002365 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002366 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002367
Chris Masonc2898112009-06-10 09:51:32 -04002368 if (!btrfs_test_opt(tree_root, SSD) &&
2369 !btrfs_test_opt(tree_root, NOSSD) &&
2370 !fs_info->fs_devices->rotating) {
2371 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2372 "mode\n");
2373 btrfs_set_opt(fs_info->mount_opt, SSD);
2374 }
2375
Stefan Behrens21adbd52011-11-09 13:44:05 +01002376#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2377 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2378 ret = btrfsic_mount(tree_root, fs_devices,
2379 btrfs_test_opt(tree_root,
2380 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2381 1 : 0,
2382 fs_info->check_integrity_print_mask);
2383 if (ret)
2384 printk(KERN_WARNING "btrfs: failed to initialize"
2385 " integrity check module %s\n", sb->s_id);
2386 }
2387#endif
2388
liuboacce9522011-01-06 19:30:25 +08002389 /* do not make disk changes in broken FS */
2390 if (btrfs_super_log_root(disk_super) != 0 &&
2391 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
Chris Masone02119d2008-09-05 16:13:11 -04002392 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002393
Chris Mason7c2ca462008-11-19 15:13:35 -05002394 if (fs_devices->rw_devices == 0) {
Chris Masond3977122009-01-05 21:25:51 -05002395 printk(KERN_WARNING "Btrfs log replay required "
2396 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002397 err = -EIO;
2398 goto fail_trans_kthread;
2399 }
Chris Masone02119d2008-09-05 16:13:11 -04002400 blocksize =
2401 btrfs_level_size(tree_root,
2402 btrfs_super_log_root_level(disk_super));
2403
Al Viro6f07e422011-11-17 00:46:16 -05002404 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002405 if (!log_tree_root) {
2406 err = -ENOMEM;
2407 goto fail_trans_kthread;
2408 }
Chris Masone02119d2008-09-05 16:13:11 -04002409
2410 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2411 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2412
2413 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002414 blocksize,
2415 generation + 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002416 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002417 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002418 if (ret) {
2419 btrfs_error(tree_root->fs_info, ret,
2420 "Failed to recover log tree");
2421 free_extent_buffer(log_tree_root->node);
2422 kfree(log_tree_root);
2423 goto fail_trans_kthread;
2424 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002425
2426 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002427 ret = btrfs_commit_super(tree_root);
2428 if (ret)
2429 goto fail_trans_kthread;
Yan Zhenge556ce22008-11-20 10:25:19 -05002430 }
Chris Masone02119d2008-09-05 16:13:11 -04002431 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002432
Yan, Zheng76dda932009-09-21 16:00:26 -04002433 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002434 if (ret)
2435 goto fail_trans_kthread;
Yan, Zheng76dda932009-09-21 16:00:26 -04002436
Chris Mason7c2ca462008-11-19 15:13:35 -05002437 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002438 ret = btrfs_cleanup_fs_roots(fs_info);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002439 if (ret) {
2440 }
Yan, Zhengd68fc572010-05-16 10:49:58 -04002441
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002442 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002443 if (ret < 0) {
2444 printk(KERN_WARNING
2445 "btrfs: failed to recover relocation\n");
2446 err = -EINVAL;
2447 goto fail_trans_kthread;
2448 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002449 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002450
Chris Mason3de45862008-11-17 21:02:50 -05002451 location.objectid = BTRFS_FS_TREE_OBJECTID;
2452 location.type = BTRFS_ROOT_ITEM_KEY;
2453 location.offset = (u64)-1;
2454
Chris Mason3de45862008-11-17 21:02:50 -05002455 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2456 if (!fs_info->fs_root)
Chris Mason7c2ca462008-11-19 15:13:35 -05002457 goto fail_trans_kthread;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002458 if (IS_ERR(fs_info->fs_root)) {
2459 err = PTR_ERR(fs_info->fs_root);
2460 goto fail_trans_kthread;
2461 }
Chris Masonc2898112009-06-10 09:51:32 -04002462
Josef Bacike3acc2a2010-01-26 14:30:53 +00002463 if (!(sb->s_flags & MS_RDONLY)) {
2464 down_read(&fs_info->cleanup_work_sem);
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002465 err = btrfs_orphan_cleanup(fs_info->fs_root);
2466 if (!err)
2467 err = btrfs_orphan_cleanup(fs_info->tree_root);
Josef Bacike3acc2a2010-01-26 14:30:53 +00002468 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002469
2470 if (!err)
2471 err = btrfs_recover_balance(fs_info->tree_root);
2472
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002473 if (err) {
2474 close_ctree(tree_root);
Al Viroad2b2c82011-11-17 01:10:02 -05002475 return err;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002476 }
Josef Bacike3acc2a2010-01-26 14:30:53 +00002477 }
2478
Al Viroad2b2c82011-11-17 01:10:02 -05002479 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002480
Chris Mason7c2ca462008-11-19 15:13:35 -05002481fail_trans_kthread:
2482 kthread_stop(fs_info->transaction_kthread);
Chris Mason3f157a22008-06-25 16:01:31 -04002483fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002484 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002485
2486 /*
2487 * make sure we're done with the btree inode before we stop our
2488 * kthreads
2489 */
2490 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2491 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2492
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002493fail_block_groups:
2494 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002495
2496fail_tree_roots:
2497 free_root_pointers(fs_info, 1);
2498
Chris Mason39279cc2007-06-12 06:35:45 -04002499fail_sb_buffer:
Chris Mason61d92c32009-10-02 19:11:56 -04002500 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason306c8b62011-11-03 15:21:39 -04002501 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Mason247e7432008-07-17 12:53:51 -04002502 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002503 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002504 btrfs_stop_workers(&fs_info->workers);
2505 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002506 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002507 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002508 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002509 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002510 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002511 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04002512 btrfs_stop_workers(&fs_info->caching_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002513fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04002514fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002515 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2516
Chris Mason7c2ca462008-11-19 15:13:35 -05002517 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Mason4543df72008-06-11 21:47:56 -04002518 iput(fs_info->btree_inode);
Jens Axboead081f12009-06-12 14:43:40 +02002519fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05002520 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002521fail_srcu:
2522 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002523fail:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002524 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05002525 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002526
2527recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04002528 if (!btrfs_test_opt(tree_root, RECOVERY))
2529 goto fail_tree_roots;
2530
2531 free_root_pointers(fs_info, 0);
2532
2533 /* don't use the log in recovery mode, it won't be valid */
2534 btrfs_set_super_log_root(disk_super, 0);
2535
2536 /* we can't trust the free space cache either */
2537 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2538
2539 ret = next_root_backup(fs_info, fs_info->super_copy,
2540 &num_backups_tried, &backup_index);
2541 if (ret == -1)
2542 goto fail_block_groups;
2543 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05002544}
2545
Chris Masonf2984462008-04-10 16:19:33 -04002546static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2547{
2548 char b[BDEVNAME_SIZE];
2549
2550 if (uptodate) {
2551 set_buffer_uptodate(bh);
2552 } else {
David Sterba7a36dde2011-05-06 15:33:15 +02002553 printk_ratelimited(KERN_WARNING "lost page write due to "
Chris Masonf2984462008-04-10 16:19:33 -04002554 "I/O error on %s\n",
2555 bdevname(bh->b_bdev, b));
Chris Mason1259ab72008-05-12 13:39:03 -04002556 /* note, we dont' set_buffer_write_io_error because we have
2557 * our own ways of dealing with the IO errors
2558 */
Chris Masonf2984462008-04-10 16:19:33 -04002559 clear_buffer_uptodate(bh);
2560 }
2561 unlock_buffer(bh);
2562 put_bh(bh);
2563}
2564
Yan Zhenga512bbf2008-12-08 16:46:26 -05002565struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2566{
2567 struct buffer_head *bh;
2568 struct buffer_head *latest = NULL;
2569 struct btrfs_super_block *super;
2570 int i;
2571 u64 transid = 0;
2572 u64 bytenr;
2573
2574 /* we would like to check all the supers, but that would make
2575 * a btrfs mount succeed after a mkfs from a different FS.
2576 * So, we need to add a special mount option to scan for
2577 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2578 */
2579 for (i = 0; i < 1; i++) {
2580 bytenr = btrfs_sb_offset(i);
2581 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2582 break;
2583 bh = __bread(bdev, bytenr / 4096, 4096);
2584 if (!bh)
2585 continue;
2586
2587 super = (struct btrfs_super_block *)bh->b_data;
2588 if (btrfs_super_bytenr(super) != bytenr ||
2589 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2590 sizeof(super->magic))) {
2591 brelse(bh);
2592 continue;
2593 }
2594
2595 if (!latest || btrfs_super_generation(super) > transid) {
2596 brelse(latest);
2597 latest = bh;
2598 transid = btrfs_super_generation(super);
2599 } else {
2600 brelse(bh);
2601 }
2602 }
2603 return latest;
2604}
2605
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002606/*
2607 * this should be called twice, once with wait == 0 and
2608 * once with wait == 1. When wait == 0 is done, all the buffer heads
2609 * we write are pinned.
2610 *
2611 * They are released when wait == 1 is done.
2612 * max_mirrors must be the same for both runs, and it indicates how
2613 * many supers on this one device should be written.
2614 *
2615 * max_mirrors == 0 means to write them all.
2616 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002617static int write_dev_supers(struct btrfs_device *device,
2618 struct btrfs_super_block *sb,
2619 int do_barriers, int wait, int max_mirrors)
2620{
2621 struct buffer_head *bh;
2622 int i;
2623 int ret;
2624 int errors = 0;
2625 u32 crc;
2626 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002627
2628 if (max_mirrors == 0)
2629 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2630
Yan Zhenga512bbf2008-12-08 16:46:26 -05002631 for (i = 0; i < max_mirrors; i++) {
2632 bytenr = btrfs_sb_offset(i);
2633 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2634 break;
2635
2636 if (wait) {
2637 bh = __find_get_block(device->bdev, bytenr / 4096,
2638 BTRFS_SUPER_INFO_SIZE);
2639 BUG_ON(!bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002640 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002641 if (!buffer_uptodate(bh))
2642 errors++;
2643
2644 /* drop our reference */
2645 brelse(bh);
2646
2647 /* drop the reference from the wait == 0 run */
2648 brelse(bh);
2649 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002650 } else {
2651 btrfs_set_super_bytenr(sb, bytenr);
2652
2653 crc = ~(u32)0;
2654 crc = btrfs_csum_data(NULL, (char *)sb +
2655 BTRFS_CSUM_SIZE, crc,
2656 BTRFS_SUPER_INFO_SIZE -
2657 BTRFS_CSUM_SIZE);
2658 btrfs_csum_final(crc, sb->csum);
2659
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002660 /*
2661 * one reference for us, and we leave it for the
2662 * caller
2663 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002664 bh = __getblk(device->bdev, bytenr / 4096,
2665 BTRFS_SUPER_INFO_SIZE);
2666 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2667
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002668 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002669 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002670
2671 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002672 lock_buffer(bh);
2673 bh->b_end_io = btrfs_end_buffer_write_sync;
2674 }
2675
Chris Mason387125f2011-11-18 15:07:51 -05002676 /*
2677 * we fua the first super. The others we allow
2678 * to go down lazy.
2679 */
Stefan Behrens21adbd52011-11-09 13:44:05 +01002680 ret = btrfsic_submit_bh(WRITE_FUA, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002681 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05002682 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002683 }
2684 return errors < i ? 0 : -1;
2685}
2686
Chris Mason387125f2011-11-18 15:07:51 -05002687/*
2688 * endio for the write_dev_flush, this will wake anyone waiting
2689 * for the barrier when it is done
2690 */
2691static void btrfs_end_empty_barrier(struct bio *bio, int err)
2692{
2693 if (err) {
2694 if (err == -EOPNOTSUPP)
2695 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
2696 clear_bit(BIO_UPTODATE, &bio->bi_flags);
2697 }
2698 if (bio->bi_private)
2699 complete(bio->bi_private);
2700 bio_put(bio);
2701}
2702
2703/*
2704 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
2705 * sent down. With wait == 1, it waits for the previous flush.
2706 *
2707 * any device where the flush fails with eopnotsupp are flagged as not-barrier
2708 * capable
2709 */
2710static int write_dev_flush(struct btrfs_device *device, int wait)
2711{
2712 struct bio *bio;
2713 int ret = 0;
2714
2715 if (device->nobarriers)
2716 return 0;
2717
2718 if (wait) {
2719 bio = device->flush_bio;
2720 if (!bio)
2721 return 0;
2722
2723 wait_for_completion(&device->flush_wait);
2724
2725 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
2726 printk("btrfs: disabling barriers on dev %s\n",
2727 device->name);
2728 device->nobarriers = 1;
2729 }
2730 if (!bio_flagged(bio, BIO_UPTODATE)) {
2731 ret = -EIO;
2732 }
2733
2734 /* drop the reference from the wait == 0 run */
2735 bio_put(bio);
2736 device->flush_bio = NULL;
2737
2738 return ret;
2739 }
2740
2741 /*
2742 * one reference for us, and we leave it for the
2743 * caller
2744 */
2745 device->flush_bio = NULL;;
2746 bio = bio_alloc(GFP_NOFS, 0);
2747 if (!bio)
2748 return -ENOMEM;
2749
2750 bio->bi_end_io = btrfs_end_empty_barrier;
2751 bio->bi_bdev = device->bdev;
2752 init_completion(&device->flush_wait);
2753 bio->bi_private = &device->flush_wait;
2754 device->flush_bio = bio;
2755
2756 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002757 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05002758
2759 return 0;
2760}
2761
2762/*
2763 * send an empty flush down to each device in parallel,
2764 * then wait for them
2765 */
2766static int barrier_all_devices(struct btrfs_fs_info *info)
2767{
2768 struct list_head *head;
2769 struct btrfs_device *dev;
2770 int errors = 0;
2771 int ret;
2772
2773 /* send down all the barriers */
2774 head = &info->fs_devices->devices;
2775 list_for_each_entry_rcu(dev, head, dev_list) {
2776 if (!dev->bdev) {
2777 errors++;
2778 continue;
2779 }
2780 if (!dev->in_fs_metadata || !dev->writeable)
2781 continue;
2782
2783 ret = write_dev_flush(dev, 0);
2784 if (ret)
2785 errors++;
2786 }
2787
2788 /* wait for all the barriers */
2789 list_for_each_entry_rcu(dev, head, dev_list) {
2790 if (!dev->bdev) {
2791 errors++;
2792 continue;
2793 }
2794 if (!dev->in_fs_metadata || !dev->writeable)
2795 continue;
2796
2797 ret = write_dev_flush(dev, 1);
2798 if (ret)
2799 errors++;
2800 }
2801 if (errors)
2802 return -EIO;
2803 return 0;
2804}
2805
Yan Zhenga512bbf2008-12-08 16:46:26 -05002806int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04002807{
Chris Masone5e9a522009-06-10 15:17:02 -04002808 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04002809 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04002810 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04002811 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04002812 int ret;
2813 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04002814 int max_errors;
2815 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04002816 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04002817
David Sterba6c417612011-04-13 15:41:04 +02002818 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Masonf2984462008-04-10 16:19:33 -04002819 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002820 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04002821
David Sterba6c417612011-04-13 15:41:04 +02002822 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04002823 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04002824
Chris Mason174ba502011-05-27 10:03:58 -04002825 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04002826 head = &root->fs_info->fs_devices->devices;
Chris Mason387125f2011-11-18 15:07:51 -05002827
2828 if (do_barriers)
2829 barrier_all_devices(root->fs_info);
2830
Xiao Guangrong1f781602011-04-20 10:09:16 +00002831 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002832 if (!dev->bdev) {
2833 total_errors++;
2834 continue;
2835 }
Yan Zheng2b820322008-11-17 21:11:30 -05002836 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002837 continue;
2838
Yan Zheng2b820322008-11-17 21:11:30 -05002839 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04002840 btrfs_set_stack_device_type(dev_item, dev->type);
2841 btrfs_set_stack_device_id(dev_item, dev->devid);
2842 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2843 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2844 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2845 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2846 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2847 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05002848 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002849
Chris Masona061fc82008-05-07 11:43:44 -04002850 flags = btrfs_super_flags(sb);
2851 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04002852
Yan Zhenga512bbf2008-12-08 16:46:26 -05002853 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04002854 if (ret)
2855 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04002856 }
Chris Masona236aed2008-04-29 09:38:00 -04002857 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05002858 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2859 total_errors);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002860
2861 /* This shouldn't happen. FUA is masked off if unsupported */
Chris Masona236aed2008-04-29 09:38:00 -04002862 BUG();
2863 }
Chris Masonf2984462008-04-10 16:19:33 -04002864
Yan Zhenga512bbf2008-12-08 16:46:26 -05002865 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00002866 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002867 if (!dev->bdev)
2868 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05002869 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002870 continue;
2871
Yan Zhenga512bbf2008-12-08 16:46:26 -05002872 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2873 if (ret)
2874 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04002875 }
Chris Mason174ba502011-05-27 10:03:58 -04002876 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04002877 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002878 btrfs_error(root->fs_info, -EIO,
2879 "%d errors while writing supers", total_errors);
2880 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04002881 }
Chris Masonf2984462008-04-10 16:19:33 -04002882 return 0;
2883}
2884
Yan Zhenga512bbf2008-12-08 16:46:26 -05002885int write_ctree_super(struct btrfs_trans_handle *trans,
2886 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05002887{
Chris Masone66f7092007-04-20 13:16:02 -04002888 int ret;
Chris Mason2c90e5d2007-04-02 10:50:19 -04002889
Yan Zhenga512bbf2008-12-08 16:46:26 -05002890 ret = write_all_supers(root, max_mirrors);
Chris Mason5f39d392007-10-15 16:14:19 -04002891 return ret;
Chris Masoncfaa7292007-02-21 17:04:57 -05002892}
2893
Jeff Mahoney49b25e02012-03-01 17:24:58 +01002894/* Kill all outstanding I/O */
2895void btrfs_abort_devices(struct btrfs_root *root)
2896{
2897 struct list_head *head;
2898 struct btrfs_device *dev;
2899 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
2900 head = &root->fs_info->fs_devices->devices;
2901 list_for_each_entry_rcu(dev, head, dev_list) {
2902 blk_abort_queue(dev->bdev->bd_disk->queue);
2903 }
2904 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
2905}
2906
Jeff Mahoney143bede2012-03-01 14:56:26 +01002907void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04002908{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002909 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04002910 radix_tree_delete(&fs_info->fs_roots_radix,
2911 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002912 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002913
2914 if (btrfs_root_refs(&root->root_item) == 0)
2915 synchronize_srcu(&fs_info->subvol_srcu);
2916
Li Zefan581bb052011-04-20 10:06:11 +08002917 __btrfs_remove_free_space_cache(root->free_ino_pinned);
2918 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002919 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002920}
2921
2922static void free_fs_root(struct btrfs_root *root)
2923{
Li Zefan82d59022011-04-20 10:33:24 +08002924 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002925 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Al Viro0ee5dc62011-07-07 15:44:25 -04002926 if (root->anon_dev)
2927 free_anon_bdev(root->anon_dev);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04002928 free_extent_buffer(root->node);
2929 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08002930 kfree(root->free_ino_ctl);
2931 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05002932 kfree(root->name);
Chris Mason2619ba12007-04-10 16:58:11 -04002933 kfree(root);
Chris Mason2619ba12007-04-10 16:58:11 -04002934}
2935
Jeff Mahoney143bede2012-03-01 14:56:26 +01002936static void del_fs_roots(struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04002937{
2938 int ret;
2939 struct btrfs_root *gang[8];
2940 int i;
2941
Yan, Zheng76dda932009-09-21 16:00:26 -04002942 while (!list_empty(&fs_info->dead_roots)) {
2943 gang[0] = list_entry(fs_info->dead_roots.next,
2944 struct btrfs_root, root_list);
2945 list_del(&gang[0]->root_list);
2946
2947 if (gang[0]->in_radix) {
2948 btrfs_free_fs_root(fs_info, gang[0]);
2949 } else {
2950 free_extent_buffer(gang[0]->node);
2951 free_extent_buffer(gang[0]->commit_root);
2952 kfree(gang[0]);
2953 }
2954 }
2955
Chris Masond3977122009-01-05 21:25:51 -05002956 while (1) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04002957 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2958 (void **)gang, 0,
2959 ARRAY_SIZE(gang));
2960 if (!ret)
2961 break;
Chris Mason2619ba12007-04-10 16:58:11 -04002962 for (i = 0; i < ret; i++)
Chris Mason5eda7b52007-06-22 14:16:25 -04002963 btrfs_free_fs_root(fs_info, gang[i]);
Chris Mason0f7d52f2007-04-09 10:42:37 -04002964 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04002965}
Chris Masonb4100d62007-04-12 12:14:00 -04002966
Yan Zhengc146afa2008-11-12 14:34:12 -05002967int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
2968{
2969 u64 root_objectid = 0;
2970 struct btrfs_root *gang[8];
2971 int i;
2972 int ret;
2973
2974 while (1) {
2975 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2976 (void **)gang, root_objectid,
2977 ARRAY_SIZE(gang));
2978 if (!ret)
2979 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002980
2981 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05002982 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002983 int err;
2984
Yan Zhengc146afa2008-11-12 14:34:12 -05002985 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05002986 err = btrfs_orphan_cleanup(gang[i]);
2987 if (err)
2988 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05002989 }
2990 root_objectid++;
2991 }
2992 return 0;
2993}
2994
2995int btrfs_commit_super(struct btrfs_root *root)
2996{
2997 struct btrfs_trans_handle *trans;
2998 int ret;
2999
3000 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00003001 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003002 btrfs_clean_old_snapshots(root);
3003 mutex_unlock(&root->fs_info->cleaner_mutex);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003004
3005 /* wait until ongoing cleanup work done */
3006 down_write(&root->fs_info->cleanup_work_sem);
3007 up_write(&root->fs_info->cleanup_work_sem);
3008
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003009 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003010 if (IS_ERR(trans))
3011 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05003012 ret = btrfs_commit_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003013 if (ret)
3014 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003015 /* run commit again to drop the original snapshot */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003016 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003017 if (IS_ERR(trans))
3018 return PTR_ERR(trans);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003019 ret = btrfs_commit_transaction(trans, root);
3020 if (ret)
3021 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003022 ret = btrfs_write_and_wait_transaction(NULL, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003023 if (ret) {
3024 btrfs_error(root->fs_info, ret,
3025 "Failed to sync btree inode to disk.");
3026 return ret;
3027 }
Yan Zhengc146afa2008-11-12 14:34:12 -05003028
Yan Zhenga512bbf2008-12-08 16:46:26 -05003029 ret = write_ctree_super(NULL, root, 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05003030 return ret;
3031}
3032
Chris Masone20d96d2007-03-22 12:13:20 -04003033int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003034{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003035 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003036 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003037
Chris Masonfacda1e2007-06-08 18:11:48 -04003038 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003039 smp_mb();
3040
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003041 /* pause restriper - we want to resume on mount */
3042 btrfs_pause_balance(root->fs_info);
3043
Arne Jansena2de7332011-03-08 14:14:00 +01003044 btrfs_scrub_cancel(root);
Chris Mason4cb53002011-05-24 15:35:30 -04003045
3046 /* wait for any defraggers to finish */
3047 wait_event(fs_info->transaction_wait,
3048 (atomic_read(&fs_info->defrag_running) == 0));
3049
3050 /* clear out the rbtree of defraggable inodes */
Al Viroe3029d92011-11-17 00:56:18 -05003051 btrfs_run_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003052
liuboacce9522011-01-06 19:30:25 +08003053 /*
3054 * Here come 2 situations when btrfs is broken to flip readonly:
3055 *
3056 * 1. when btrfs flips readonly somewhere else before
3057 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
3058 * and btrfs will skip to write sb directly to keep
3059 * ERROR state on disk.
3060 *
3061 * 2. when btrfs flips readonly just in btrfs_commit_super,
Justin P. Mattockae0e47f2011-03-01 15:06:02 +01003062 * and in such case, btrfs cannot write sb via btrfs_commit_super,
liuboacce9522011-01-06 19:30:25 +08003063 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
3064 * btrfs will cleanup all FS resources first and write sb then.
3065 */
Yan Zhengc146afa2008-11-12 14:34:12 -05003066 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003067 ret = btrfs_commit_super(root);
3068 if (ret)
3069 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3070 }
3071
3072 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
3073 ret = btrfs_error_commit_super(root);
Chris Masond3977122009-01-05 21:25:51 -05003074 if (ret)
3075 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
Yan Zhengc146afa2008-11-12 14:34:12 -05003076 }
Chris Masoned2ff2c2007-03-01 18:59:40 -05003077
Josef Bacik300e4f82011-08-29 14:06:00 -04003078 btrfs_put_block_group_cache(fs_info);
3079
Al Viroe3029d92011-11-17 00:56:18 -05003080 kthread_stop(fs_info->transaction_kthread);
3081 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003082
Yan Zhengf25784b2009-07-28 08:41:57 -04003083 fs_info->closing = 2;
3084 smp_mb();
3085
Chris Masonb0c68f82008-01-31 11:05:37 -05003086 if (fs_info->delalloc_bytes) {
Chris Masond3977122009-01-05 21:25:51 -05003087 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07003088 (unsigned long long)fs_info->delalloc_bytes);
Chris Masonb0c68f82008-01-31 11:05:37 -05003089 }
Yan Zheng31153d82008-07-28 15:32:19 -04003090 if (fs_info->total_ref_cache_size) {
Chris Masond3977122009-01-05 21:25:51 -05003091 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
3092 (unsigned long long)fs_info->total_ref_cache_size);
Yan Zheng31153d82008-07-28 15:32:19 -04003093 }
Yanbcc63ab2008-07-30 16:29:20 -04003094
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003095 free_extent_buffer(fs_info->extent_root->node);
3096 free_extent_buffer(fs_info->extent_root->commit_root);
3097 free_extent_buffer(fs_info->tree_root->node);
3098 free_extent_buffer(fs_info->tree_root->commit_root);
Al Viroe3029d92011-11-17 00:56:18 -05003099 free_extent_buffer(fs_info->chunk_root->node);
3100 free_extent_buffer(fs_info->chunk_root->commit_root);
3101 free_extent_buffer(fs_info->dev_root->node);
3102 free_extent_buffer(fs_info->dev_root->commit_root);
3103 free_extent_buffer(fs_info->csum_root->node);
3104 free_extent_buffer(fs_info->csum_root->commit_root);
Chris Masond20f7042008-12-08 16:58:54 -05003105
Al Viroe3029d92011-11-17 00:56:18 -05003106 btrfs_free_block_groups(fs_info);
Yan Zhengc146afa2008-11-12 14:34:12 -05003107
Chris Mason0f7d52f2007-04-09 10:42:37 -04003108 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003109
Yan Zhengc146afa2008-11-12 14:34:12 -05003110 iput(fs_info->btree_inode);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003111
Chris Mason61d92c32009-10-02 19:11:56 -04003112 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04003113 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05003114 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04003115 btrfs_stop_workers(&fs_info->workers);
3116 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05003117 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05003118 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04003119 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04003120 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04003121 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08003122 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04003123 btrfs_stop_workers(&fs_info->caching_workers);
Arne Jansen90519d62011-05-23 14:30:00 +02003124 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Masond6bfde82008-04-30 13:59:35 -04003125
Stefan Behrens21adbd52011-11-09 13:44:05 +01003126#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3127 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3128 btrfsic_unmount(root, fs_info->fs_devices);
3129#endif
3130
Chris Masondfe25022008-05-13 13:46:40 -04003131 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003132 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003133
Chris Mason04160082008-03-26 10:28:07 -04003134 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003135 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003136
Chris Masoneb60cea2007-02-02 09:18:22 -05003137 return 0;
3138}
3139
Chris Mason1259ab72008-05-12 13:39:03 -04003140int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
Chris Masonccd467d2007-06-28 15:57:36 -04003141{
Chris Mason1259ab72008-05-12 13:39:03 -04003142 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003143 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003144
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003145 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003146 if (!ret)
3147 return ret;
3148
3149 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
3150 parent_transid);
3151 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003152}
Chris Mason6702ed42007-08-07 16:15:09 -04003153
Chris Mason5f39d392007-10-15 16:14:19 -04003154int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3155{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003156 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003157}
3158
3159void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3160{
Chris Mason727011e2010-08-06 13:21:20 -04003161 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Mason5f39d392007-10-15 16:14:19 -04003162 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003163 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003164
Chris Masonb9447ef2009-03-09 11:45:38 -04003165 btrfs_assert_tree_locked(buf);
Chris Masonccd467d2007-06-28 15:57:36 -04003166 if (transid != root->fs_info->generation) {
Chris Masond3977122009-01-05 21:25:51 -05003167 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
3168 "found %llu running %llu\n",
Chris Masondb945352007-10-15 16:15:53 -04003169 (unsigned long long)buf->start,
Chris Masond3977122009-01-05 21:25:51 -05003170 (unsigned long long)transid,
3171 (unsigned long long)root->fs_info->generation);
Chris Masonccd467d2007-06-28 15:57:36 -04003172 WARN_ON(1);
3173 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003174 was_dirty = set_extent_buffer_dirty(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003175 if (!was_dirty) {
3176 spin_lock(&root->fs_info->delalloc_lock);
3177 root->fs_info->dirty_metadata_bytes += buf->len;
3178 spin_unlock(&root->fs_info->delalloc_lock);
3179 }
Chris Masoneb60cea2007-02-02 09:18:22 -05003180}
3181
Chris Masond3c2fdcf2007-09-17 10:58:06 -04003182void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
Chris Mason35b7e472007-05-02 15:53:43 -04003183{
Chris Mason188de642008-05-09 11:52:25 -04003184 /*
3185 * looks as though older kernels can get into trouble with
3186 * this code, they end up stuck in balance_dirty_pages forever
3187 */
Chris Masond6bfde82008-04-30 13:59:35 -04003188 u64 num_dirty;
Chris Mason771ed682008-11-06 22:02:51 -05003189 unsigned long thresh = 32 * 1024 * 1024;
Chris Masond6bfde82008-04-30 13:59:35 -04003190
Jens Axboe6933c022009-03-17 09:36:37 +01003191 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003192 return;
3193
Miao Xie16cdcec2011-04-22 18:12:22 +08003194 btrfs_balance_delayed_items(root);
3195
3196 num_dirty = root->fs_info->dirty_metadata_bytes;
3197
3198 if (num_dirty > thresh) {
3199 balance_dirty_pages_ratelimited_nr(
3200 root->fs_info->btree_inode->i_mapping, 1);
3201 }
3202 return;
3203}
3204
3205void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
3206{
3207 /*
3208 * looks as though older kernels can get into trouble with
3209 * this code, they end up stuck in balance_dirty_pages forever
3210 */
3211 u64 num_dirty;
3212 unsigned long thresh = 32 * 1024 * 1024;
3213
3214 if (current->flags & PF_MEMALLOC)
3215 return;
3216
Chris Mason585ad2c2009-05-18 10:41:58 -04003217 num_dirty = root->fs_info->dirty_metadata_bytes;
3218
Chris Masond6bfde82008-04-30 13:59:35 -04003219 if (num_dirty > thresh) {
3220 balance_dirty_pages_ratelimited_nr(
Chris Masond7fc6402008-02-18 12:12:38 -05003221 root->fs_info->btree_inode->i_mapping, 1);
Chris Masond6bfde82008-04-30 13:59:35 -04003222 }
Chris Mason188de642008-05-09 11:52:25 -04003223 return;
Chris Mason35b7e472007-05-02 15:53:43 -04003224}
Chris Mason6b800532007-10-15 16:17:34 -04003225
Chris Masonca7a79a2008-05-12 12:59:19 -04003226int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003227{
Chris Mason727011e2010-08-06 13:21:20 -04003228 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003229 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003230}
Chris Mason0da54682007-11-07 21:08:01 -05003231
Chris Mason01d658f2011-11-01 10:08:06 -04003232static int btree_lock_page_hook(struct page *page, void *data,
3233 void (*flush_fn)(void *))
Chris Mason4bef0842008-09-08 11:18:08 -04003234{
3235 struct inode *inode = page->mapping->host;
Chris Masonb9473432009-03-13 11:00:37 -04003236 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4bef0842008-09-08 11:18:08 -04003237 struct extent_buffer *eb;
Chris Mason4bef0842008-09-08 11:18:08 -04003238
Josef Bacik4f2de97a2012-03-07 16:20:05 -05003239 /*
3240 * We culled this eb but the page is still hanging out on the mapping,
3241 * carry on.
3242 */
3243 if (!PagePrivate(page))
Chris Mason4bef0842008-09-08 11:18:08 -04003244 goto out;
3245
Josef Bacik4f2de97a2012-03-07 16:20:05 -05003246 eb = (struct extent_buffer *)page->private;
3247 if (!eb) {
3248 WARN_ON(1);
3249 goto out;
3250 }
3251 if (page != eb->pages[0])
Chris Mason4bef0842008-09-08 11:18:08 -04003252 goto out;
3253
Chris Mason01d658f2011-11-01 10:08:06 -04003254 if (!btrfs_try_tree_write_lock(eb)) {
3255 flush_fn(data);
3256 btrfs_tree_lock(eb);
3257 }
Chris Mason4bef0842008-09-08 11:18:08 -04003258 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonb9473432009-03-13 11:00:37 -04003259
3260 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3261 spin_lock(&root->fs_info->delalloc_lock);
3262 if (root->fs_info->dirty_metadata_bytes >= eb->len)
3263 root->fs_info->dirty_metadata_bytes -= eb->len;
3264 else
3265 WARN_ON(1);
3266 spin_unlock(&root->fs_info->delalloc_lock);
3267 }
3268
Chris Mason4bef0842008-09-08 11:18:08 -04003269 btrfs_tree_unlock(eb);
Chris Mason4bef0842008-09-08 11:18:08 -04003270out:
Chris Mason01d658f2011-11-01 10:08:06 -04003271 if (!trylock_page(page)) {
3272 flush_fn(data);
3273 lock_page(page);
3274 }
Chris Mason4bef0842008-09-08 11:18:08 -04003275 return 0;
3276}
3277
David Sterbafcd1f0652012-03-06 00:06:18 +01003278static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003279 int read_only)
3280{
David Sterbafcd1f0652012-03-06 00:06:18 +01003281 if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) {
3282 printk(KERN_ERR "btrfs: unsupported checksum algorithm\n");
3283 return -EINVAL;
3284 }
liuboacce9522011-01-06 19:30:25 +08003285
David Sterbafcd1f0652012-03-06 00:06:18 +01003286 if (read_only)
3287 return 0;
3288
3289 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
liuboacce9522011-01-06 19:30:25 +08003290 printk(KERN_WARNING "warning: mount fs with errors, "
3291 "running btrfsck is recommended\n");
David Sterbafcd1f0652012-03-06 00:06:18 +01003292 }
3293
3294 return 0;
liuboacce9522011-01-06 19:30:25 +08003295}
3296
3297int btrfs_error_commit_super(struct btrfs_root *root)
3298{
3299 int ret;
3300
3301 mutex_lock(&root->fs_info->cleaner_mutex);
3302 btrfs_run_delayed_iputs(root);
3303 mutex_unlock(&root->fs_info->cleaner_mutex);
3304
3305 down_write(&root->fs_info->cleanup_work_sem);
3306 up_write(&root->fs_info->cleanup_work_sem);
3307
3308 /* cleanup FS via transaction */
3309 btrfs_cleanup_transaction(root);
3310
3311 ret = write_ctree_super(NULL, root, 0);
3312
3313 return ret;
3314}
3315
Jeff Mahoney143bede2012-03-01 14:56:26 +01003316static void btrfs_destroy_ordered_operations(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003317{
3318 struct btrfs_inode *btrfs_inode;
3319 struct list_head splice;
3320
3321 INIT_LIST_HEAD(&splice);
3322
3323 mutex_lock(&root->fs_info->ordered_operations_mutex);
3324 spin_lock(&root->fs_info->ordered_extent_lock);
3325
3326 list_splice_init(&root->fs_info->ordered_operations, &splice);
3327 while (!list_empty(&splice)) {
3328 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3329 ordered_operations);
3330
3331 list_del_init(&btrfs_inode->ordered_operations);
3332
3333 btrfs_invalidate_inodes(btrfs_inode->root);
3334 }
3335
3336 spin_unlock(&root->fs_info->ordered_extent_lock);
3337 mutex_unlock(&root->fs_info->ordered_operations_mutex);
liuboacce9522011-01-06 19:30:25 +08003338}
3339
Jeff Mahoney143bede2012-03-01 14:56:26 +01003340static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003341{
3342 struct list_head splice;
3343 struct btrfs_ordered_extent *ordered;
3344 struct inode *inode;
3345
3346 INIT_LIST_HEAD(&splice);
3347
3348 spin_lock(&root->fs_info->ordered_extent_lock);
3349
3350 list_splice_init(&root->fs_info->ordered_extents, &splice);
3351 while (!list_empty(&splice)) {
3352 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
3353 root_extent_list);
3354
3355 list_del_init(&ordered->root_extent_list);
3356 atomic_inc(&ordered->refs);
3357
3358 /* the inode may be getting freed (in sys_unlink path). */
3359 inode = igrab(ordered->inode);
3360
3361 spin_unlock(&root->fs_info->ordered_extent_lock);
3362 if (inode)
3363 iput(inode);
3364
3365 atomic_set(&ordered->refs, 1);
3366 btrfs_put_ordered_extent(ordered);
3367
3368 spin_lock(&root->fs_info->ordered_extent_lock);
3369 }
3370
3371 spin_unlock(&root->fs_info->ordered_extent_lock);
liuboacce9522011-01-06 19:30:25 +08003372}
3373
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003374int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3375 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003376{
3377 struct rb_node *node;
3378 struct btrfs_delayed_ref_root *delayed_refs;
3379 struct btrfs_delayed_ref_node *ref;
3380 int ret = 0;
3381
3382 delayed_refs = &trans->delayed_refs;
3383
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003384again:
liuboacce9522011-01-06 19:30:25 +08003385 spin_lock(&delayed_refs->lock);
3386 if (delayed_refs->num_entries == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003387 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003388 printk(KERN_INFO "delayed_refs has NO entry\n");
3389 return ret;
3390 }
3391
3392 node = rb_first(&delayed_refs->root);
3393 while (node) {
3394 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
3395 node = rb_next(node);
3396
3397 ref->in_tree = 0;
3398 rb_erase(&ref->rb_node, &delayed_refs->root);
3399 delayed_refs->num_entries--;
3400
3401 atomic_set(&ref->refs, 1);
3402 if (btrfs_delayed_ref_is_head(ref)) {
3403 struct btrfs_delayed_ref_head *head;
3404
3405 head = btrfs_delayed_node_to_head(ref);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003406 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003407 mutex_lock(&head->mutex);
3408 kfree(head->extent_op);
3409 delayed_refs->num_heads--;
3410 if (list_empty(&head->cluster))
3411 delayed_refs->num_heads_ready--;
3412 list_del_init(&head->cluster);
3413 mutex_unlock(&head->mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003414 btrfs_put_delayed_ref(ref);
3415 goto again;
liuboacce9522011-01-06 19:30:25 +08003416 }
liuboacce9522011-01-06 19:30:25 +08003417 spin_unlock(&delayed_refs->lock);
3418 btrfs_put_delayed_ref(ref);
3419
3420 cond_resched();
3421 spin_lock(&delayed_refs->lock);
3422 }
3423
3424 spin_unlock(&delayed_refs->lock);
3425
3426 return ret;
3427}
3428
Jeff Mahoney143bede2012-03-01 14:56:26 +01003429static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
liuboacce9522011-01-06 19:30:25 +08003430{
3431 struct btrfs_pending_snapshot *snapshot;
3432 struct list_head splice;
3433
3434 INIT_LIST_HEAD(&splice);
3435
3436 list_splice_init(&t->pending_snapshots, &splice);
3437
3438 while (!list_empty(&splice)) {
3439 snapshot = list_entry(splice.next,
3440 struct btrfs_pending_snapshot,
3441 list);
3442
3443 list_del_init(&snapshot->list);
3444
3445 kfree(snapshot);
3446 }
liuboacce9522011-01-06 19:30:25 +08003447}
3448
Jeff Mahoney143bede2012-03-01 14:56:26 +01003449static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003450{
3451 struct btrfs_inode *btrfs_inode;
3452 struct list_head splice;
3453
3454 INIT_LIST_HEAD(&splice);
3455
liuboacce9522011-01-06 19:30:25 +08003456 spin_lock(&root->fs_info->delalloc_lock);
David Sterba5be76752011-06-09 10:02:51 +00003457 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003458
3459 while (!list_empty(&splice)) {
3460 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3461 delalloc_inodes);
3462
3463 list_del_init(&btrfs_inode->delalloc_inodes);
3464
3465 btrfs_invalidate_inodes(btrfs_inode->root);
3466 }
3467
3468 spin_unlock(&root->fs_info->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003469}
3470
3471static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3472 struct extent_io_tree *dirty_pages,
3473 int mark)
3474{
3475 int ret;
3476 struct page *page;
3477 struct inode *btree_inode = root->fs_info->btree_inode;
3478 struct extent_buffer *eb;
3479 u64 start = 0;
3480 u64 end;
3481 u64 offset;
3482 unsigned long index;
3483
3484 while (1) {
3485 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
3486 mark);
3487 if (ret)
3488 break;
3489
3490 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3491 while (start <= end) {
3492 index = start >> PAGE_CACHE_SHIFT;
3493 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
3494 page = find_get_page(btree_inode->i_mapping, index);
3495 if (!page)
3496 continue;
3497 offset = page_offset(page);
3498
3499 spin_lock(&dirty_pages->buffer_lock);
3500 eb = radix_tree_lookup(
3501 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
3502 offset >> PAGE_CACHE_SHIFT);
3503 spin_unlock(&dirty_pages->buffer_lock);
3504 if (eb) {
3505 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3506 &eb->bflags);
3507 atomic_set(&eb->refs, 1);
3508 }
3509 if (PageWriteback(page))
3510 end_page_writeback(page);
3511
3512 lock_page(page);
3513 if (PageDirty(page)) {
3514 clear_page_dirty_for_io(page);
3515 spin_lock_irq(&page->mapping->tree_lock);
3516 radix_tree_tag_clear(&page->mapping->page_tree,
3517 page_index(page),
3518 PAGECACHE_TAG_DIRTY);
3519 spin_unlock_irq(&page->mapping->tree_lock);
3520 }
3521
3522 page->mapping->a_ops->invalidatepage(page, 0);
3523 unlock_page(page);
3524 }
3525 }
3526
3527 return ret;
3528}
3529
3530static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3531 struct extent_io_tree *pinned_extents)
3532{
3533 struct extent_io_tree *unpin;
3534 u64 start;
3535 u64 end;
3536 int ret;
3537
3538 unpin = pinned_extents;
3539 while (1) {
3540 ret = find_first_extent_bit(unpin, 0, &start, &end,
3541 EXTENT_DIRTY);
3542 if (ret)
3543 break;
3544
3545 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00003546 if (btrfs_test_opt(root, DISCARD))
3547 ret = btrfs_error_discard_extent(root, start,
3548 end + 1 - start,
3549 NULL);
liuboacce9522011-01-06 19:30:25 +08003550
3551 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3552 btrfs_error_unpin_extent_range(root, start, end);
3553 cond_resched();
3554 }
3555
3556 return 0;
3557}
3558
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003559void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3560 struct btrfs_root *root)
3561{
3562 btrfs_destroy_delayed_refs(cur_trans, root);
3563 btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
3564 cur_trans->dirty_pages.dirty_bytes);
3565
3566 /* FIXME: cleanup wait for commit */
3567 cur_trans->in_commit = 1;
3568 cur_trans->blocked = 1;
3569 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3570 wake_up(&root->fs_info->transaction_blocked_wait);
3571
3572 cur_trans->blocked = 0;
3573 if (waitqueue_active(&root->fs_info->transaction_wait))
3574 wake_up(&root->fs_info->transaction_wait);
3575
3576 cur_trans->commit_done = 1;
3577 if (waitqueue_active(&cur_trans->commit_wait))
3578 wake_up(&cur_trans->commit_wait);
3579
3580 btrfs_destroy_pending_snapshots(cur_trans);
3581
3582 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
3583 EXTENT_DIRTY);
3584
3585 /*
3586 memset(cur_trans, 0, sizeof(*cur_trans));
3587 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3588 */
3589}
3590
3591int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003592{
3593 struct btrfs_transaction *t;
3594 LIST_HEAD(list);
3595
liuboacce9522011-01-06 19:30:25 +08003596 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3597
Josef Bacika4abeea2011-04-11 17:25:13 -04003598 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003599 list_splice_init(&root->fs_info->trans_list, &list);
Josef Bacika4abeea2011-04-11 17:25:13 -04003600 root->fs_info->trans_no_join = 1;
3601 spin_unlock(&root->fs_info->trans_lock);
3602
liuboacce9522011-01-06 19:30:25 +08003603 while (!list_empty(&list)) {
3604 t = list_entry(list.next, struct btrfs_transaction, list);
3605 if (!t)
3606 break;
3607
3608 btrfs_destroy_ordered_operations(root);
3609
3610 btrfs_destroy_ordered_extents(root);
3611
3612 btrfs_destroy_delayed_refs(t, root);
3613
3614 btrfs_block_rsv_release(root,
3615 &root->fs_info->trans_block_rsv,
3616 t->dirty_pages.dirty_bytes);
3617
3618 /* FIXME: cleanup wait for commit */
3619 t->in_commit = 1;
3620 t->blocked = 1;
3621 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3622 wake_up(&root->fs_info->transaction_blocked_wait);
3623
3624 t->blocked = 0;
3625 if (waitqueue_active(&root->fs_info->transaction_wait))
3626 wake_up(&root->fs_info->transaction_wait);
liuboacce9522011-01-06 19:30:25 +08003627
liuboacce9522011-01-06 19:30:25 +08003628 t->commit_done = 1;
3629 if (waitqueue_active(&t->commit_wait))
3630 wake_up(&t->commit_wait);
liuboacce9522011-01-06 19:30:25 +08003631
3632 btrfs_destroy_pending_snapshots(t);
3633
3634 btrfs_destroy_delalloc_inodes(root);
3635
Josef Bacika4abeea2011-04-11 17:25:13 -04003636 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003637 root->fs_info->running_transaction = NULL;
Josef Bacika4abeea2011-04-11 17:25:13 -04003638 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003639
3640 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3641 EXTENT_DIRTY);
3642
3643 btrfs_destroy_pinned_extent(root,
3644 root->fs_info->pinned_extents);
3645
Josef Bacik13c5a932011-04-11 15:45:29 -04003646 atomic_set(&t->use_count, 0);
liuboacce9522011-01-06 19:30:25 +08003647 list_del_init(&t->list);
3648 memset(t, 0, sizeof(*t));
3649 kmem_cache_free(btrfs_transaction_cachep, t);
3650 }
3651
Josef Bacika4abeea2011-04-11 17:25:13 -04003652 spin_lock(&root->fs_info->trans_lock);
3653 root->fs_info->trans_no_join = 0;
3654 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003655 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08003656
3657 return 0;
3658}
3659
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003660static int btree_writepage_io_failed_hook(struct bio *bio, struct page *page,
3661 u64 start, u64 end,
3662 struct extent_state *state)
3663{
3664 struct super_block *sb = page->mapping->host->i_sb;
3665 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
3666 btrfs_error(fs_info, -EIO,
3667 "Error occured while writing out btree at %llu", start);
3668 return -EIO;
3669}
3670
Chris Masond1310b22008-01-24 16:13:08 -05003671static struct extent_io_ops btree_extent_io_ops = {
Chris Mason4bef0842008-09-08 11:18:08 -04003672 .write_cache_pages_lock_hook = btree_lock_page_hook,
Chris Masonce9adaa2008-04-09 16:28:12 -04003673 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02003674 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04003675 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04003676 /* note we're sharing with inode.c for the merge bio hook */
3677 .merge_bio_hook = btrfs_merge_bio_hook,
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003678 .writepage_io_failed_hook = btree_writepage_io_failed_hook,
Chris Mason0da54682007-11-07 21:08:01 -05003679};