blob: 0c0fa78ef4525b1a1c8999fad619fb898ed7609d [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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050030#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020031#include <linux/ratelimit.h>
Stefan Behrens6463fe52013-04-19 15:08:05 +000032#include <linux/uuid.h>
Stefan Behrens803b2f52013-08-15 17:11:21 +020033#include <linux/semaphore.h>
David Sterba7e75bf32011-03-18 22:56:43 +000034#include <asm/unaligned.h>
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +000037#include "hash.h"
Chris Masone089f052007-03-16 16:20:31 -040038#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040039#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040040#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040041#include "print-tree.h"
Chris Mason8b712842008-06-11 16:50:36 -040042#include "async-thread.h"
Chris Mason925baed2008-06-25 16:01:30 -040043#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040044#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040045#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080046#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010047#include "check-integrity.h"
Josef Bacik606686e2012-06-04 14:03:51 -040048#include "rcu-string.h"
Stefan Behrens8dabb742012-11-06 13:15:27 +010049#include "dev-replace.h"
David Woodhouse53b381b2013-01-29 18:40:14 -050050#include "raid56.h"
Jeff Mahoney5ac1d202013-11-01 13:06:58 -040051#include "sysfs.h"
Josef Bacikfcebe452014-05-13 17:30:47 -070052#include "qgroup.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050053
Josef Bacikde0022b2012-09-25 14:25:58 -040054#ifdef CONFIG_X86
55#include <asm/cpufeature.h>
56#endif
57
Chris Masond1310b22008-01-24 16:13:08 -050058static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040059static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040060static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010061static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080062 int read_only);
Josef Bacik569e0f32013-02-13 11:09:14 -050063static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
64 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010065static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080066static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
67 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010068static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080069static int btrfs_destroy_marked_extents(struct btrfs_root *root,
70 struct extent_io_tree *dirty_pages,
71 int mark);
72static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
73 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000074static int btrfs_cleanup_transaction(struct btrfs_root *root);
75static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040076
Chris Masond352ac62008-09-29 15:18:18 -040077/*
78 * end_io_wq structs are used to do processing in task context when an IO is
79 * complete. This is used during reads to verify checksums, and it is used
80 * by writes to insert metadata for new file extents after IO is complete.
81 */
Chris Masonce9adaa2008-04-09 16:28:12 -040082struct end_io_wq {
83 struct bio *bio;
84 bio_end_io_t *end_io;
85 void *private;
86 struct btrfs_fs_info *info;
87 int error;
Chris Mason22c59942008-04-09 16:28:12 -040088 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040089 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040090 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040091};
Chris Mason0da54682007-11-07 21:08:01 -050092
Chris Masond352ac62008-09-29 15:18:18 -040093/*
94 * async submit bios are used to offload expensive checksumming
95 * onto the worker threads. They checksum file and metadata bios
96 * just before they are sent down the IO stack.
97 */
Chris Mason44b8bd72008-04-16 11:14:51 -040098struct async_submit_bio {
99 struct inode *inode;
100 struct bio *bio;
101 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500102 extent_submit_bio_hook_t *submit_bio_start;
103 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400104 int rw;
105 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400106 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400107 /*
108 * bio_offset is optional, can be used if the pages in the bio
109 * can't tell us where in the file the bio should go
110 */
111 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400112 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100113 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400114};
115
Chris Mason85d4e462011-07-26 16:11:19 -0400116/*
117 * Lockdep class keys for extent_buffer->lock's in this root. For a given
118 * eb, the lockdep key is determined by the btrfs_root it belongs to and
119 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500120 *
Chris Mason85d4e462011-07-26 16:11:19 -0400121 * Different roots are used for different purposes and may nest inside each
122 * other and they require separate keysets. As lockdep keys should be
123 * static, assign keysets according to the purpose of the root as indicated
124 * by btrfs_root->objectid. This ensures that all special purpose roots
125 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500126 *
Chris Mason85d4e462011-07-26 16:11:19 -0400127 * Lock-nesting across peer nodes is always done with the immediate parent
128 * node locked thus preventing deadlock. As lockdep doesn't know this, use
129 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500130 *
Chris Mason85d4e462011-07-26 16:11:19 -0400131 * The key is set by the readpage_end_io_hook after the buffer has passed
132 * csum validation but before the pages are unlocked. It is also set by
133 * btrfs_init_new_buffer on freshly allocated blocks.
134 *
135 * We also add a check to make sure the highest level of the tree is the
136 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
137 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500138 */
139#ifdef CONFIG_DEBUG_LOCK_ALLOC
140# if BTRFS_MAX_LEVEL != 8
141# error
142# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400143
144static struct btrfs_lockdep_keyset {
145 u64 id; /* root objectid */
146 const char *name_stem; /* lock name stem */
147 char names[BTRFS_MAX_LEVEL + 1][20];
148 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
149} btrfs_lockdep_keysets[] = {
150 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
151 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
152 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
153 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
154 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
155 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000156 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400157 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
158 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
159 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200160 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
Chris Mason85d4e462011-07-26 16:11:19 -0400161 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500162};
Chris Mason85d4e462011-07-26 16:11:19 -0400163
164void __init btrfs_init_lockdep(void)
165{
166 int i, j;
167
168 /* initialize lockdep class names */
169 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
170 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
171
172 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
173 snprintf(ks->names[j], sizeof(ks->names[j]),
174 "btrfs-%s-%02d", ks->name_stem, j);
175 }
176}
177
178void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
179 int level)
180{
181 struct btrfs_lockdep_keyset *ks;
182
183 BUG_ON(level >= ARRAY_SIZE(ks->keys));
184
185 /* find the matching keyset, id 0 is the default entry */
186 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
187 if (ks->id == objectid)
188 break;
189
190 lockdep_set_class_and_name(&eb->lock,
191 &ks->keys[level], ks->names[level]);
192}
193
Chris Mason4008c042009-02-12 14:09:45 -0500194#endif
195
Chris Masond352ac62008-09-29 15:18:18 -0400196/*
197 * extents on the btree inode are pretty simple, there's one extent
198 * that covers the entire device
199 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500200static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200201 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500202 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400203{
204 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
205 struct extent_map *em;
206 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400207
Chris Mason890871b2009-09-02 16:24:52 -0400208 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500209 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400210 if (em) {
211 em->bdev =
212 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400213 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400214 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400215 }
Chris Mason890871b2009-09-02 16:24:52 -0400216 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400217
David Sterba172ddd62011-04-21 00:48:27 +0200218 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400219 if (!em) {
220 em = ERR_PTR(-ENOMEM);
221 goto out;
222 }
223 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400224 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400225 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400226 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400227 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500228
Chris Mason890871b2009-09-02 16:24:52 -0400229 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400230 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400231 if (ret == -EEXIST) {
232 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400233 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600234 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600235 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400236 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400237 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600238 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400239 }
Chris Mason890871b2009-09-02 16:24:52 -0400240 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400241
Chris Mason5f39d392007-10-15 16:14:19 -0400242out:
243 return em;
244}
245
Liu Bob0496682013-03-14 14:57:45 +0000246u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400247{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000248 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400249}
250
251void btrfs_csum_final(u32 crc, char *result)
252{
David Sterba7e75bf32011-03-18 22:56:43 +0000253 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400254}
255
Chris Masond352ac62008-09-29 15:18:18 -0400256/*
257 * compute the csum for a btree block, and either verify it or write it
258 * into the csum field of the block.
259 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400260static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
261 int verify)
262{
David Sterba6c417612011-04-13 15:41:04 +0200263 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500264 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400265 unsigned long len;
266 unsigned long cur_len;
267 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400268 char *kaddr;
269 unsigned long map_start;
270 unsigned long map_len;
271 int err;
272 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500273 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400274
275 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500276 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400277 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400278 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500279 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400280 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400281 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000282 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400283 crc, cur_len);
284 len -= cur_len;
285 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400286 }
Josef Bacik607d4322008-12-02 07:17:45 -0500287 if (csum_size > sizeof(inline_result)) {
288 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
289 if (!result)
290 return 1;
291 } else {
292 result = (char *)&inline_result;
293 }
294
Chris Mason19c00dd2007-10-15 16:19:22 -0400295 btrfs_csum_final(crc, result);
296
297 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500298 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500299 u32 val;
300 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500301 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500302
Josef Bacik607d4322008-12-02 07:17:45 -0500303 read_extent_buffer(buf, &val, 0, csum_size);
Frank Holtonefe120a2013-12-20 11:37:06 -0500304 printk_ratelimited(KERN_INFO
305 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
306 "level %d\n",
307 root->fs_info->sb->s_id, buf->start,
308 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500309 if (result != (char *)&inline_result)
310 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400311 return 1;
312 }
313 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500314 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400315 }
Josef Bacik607d4322008-12-02 07:17:45 -0500316 if (result != (char *)&inline_result)
317 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400318 return 0;
319}
320
Chris Masond352ac62008-09-29 15:18:18 -0400321/*
322 * we can't consider a given block up to date unless the transid of the
323 * block matches the transid in the parent node's pointer. This is how we
324 * detect blocks that either didn't get written at all or got written
325 * in the wrong place.
326 */
Chris Mason1259ab72008-05-12 13:39:03 -0400327static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400328 struct extent_buffer *eb, u64 parent_transid,
329 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400330{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000331 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400332 int ret;
Josef Bacika26e8c92014-03-28 17:07:27 -0400333 bool need_lock = (current->journal_info ==
334 (void *)BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400335
336 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
337 return 0;
338
Chris Masonb9fab912012-05-06 07:23:47 -0400339 if (atomic)
340 return -EAGAIN;
341
Josef Bacika26e8c92014-03-28 17:07:27 -0400342 if (need_lock) {
343 btrfs_tree_read_lock(eb);
344 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
345 }
346
Josef Bacik2ac55d42010-02-03 19:33:23 +0000347 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100348 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400349 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400350 btrfs_header_generation(eb) == parent_transid) {
351 ret = 0;
352 goto out;
353 }
David Sterba7a36dde2011-05-06 15:33:15 +0200354 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
Chris Mason193f2842009-04-27 07:29:05 -0400355 "found %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200356 eb->start, parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400357 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400358
359 /*
360 * Things reading via commit roots that don't have normal protection,
361 * like send, can have a really old block in cache that may point at a
362 * block that has been free'd and re-allocated. So don't clear uptodate
363 * if we find an eb that is under IO (dirty/writeback) because we could
364 * end up reading in the stale data and then writing it back out and
365 * making everybody very sad.
366 */
367 if (!extent_buffer_under_io(eb))
368 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400369out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000370 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
371 &cached_state, GFP_NOFS);
Josef Bacika26e8c92014-03-28 17:07:27 -0400372 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400373 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400374}
375
Chris Masond352ac62008-09-29 15:18:18 -0400376/*
David Sterba1104a882013-03-06 15:57:46 +0100377 * Return 0 if the superblock checksum type matches the checksum value of that
378 * algorithm. Pass the raw disk superblock data.
379 */
380static int btrfs_check_super_csum(char *raw_disk_sb)
381{
382 struct btrfs_super_block *disk_sb =
383 (struct btrfs_super_block *)raw_disk_sb;
384 u16 csum_type = btrfs_super_csum_type(disk_sb);
385 int ret = 0;
386
387 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
388 u32 crc = ~(u32)0;
389 const int csum_size = sizeof(crc);
390 char result[csum_size];
391
392 /*
393 * The super_block structure does not span the whole
394 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
395 * is filled with zeros and is included in the checkum.
396 */
397 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
398 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
399 btrfs_csum_final(crc, result);
400
401 if (memcmp(raw_disk_sb, result, csum_size))
402 ret = 1;
Chris Mason667e7d92013-05-07 11:00:13 -0400403
404 if (ret && btrfs_super_generation(disk_sb) < 10) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500405 printk(KERN_WARNING
406 "BTRFS: super block crcs don't match, older mkfs detected\n");
Chris Mason667e7d92013-05-07 11:00:13 -0400407 ret = 0;
408 }
David Sterba1104a882013-03-06 15:57:46 +0100409 }
410
411 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500412 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100413 csum_type);
414 ret = 1;
415 }
416
417 return ret;
418}
419
420/*
Chris Masond352ac62008-09-29 15:18:18 -0400421 * helper to read a given tree block, doing retries as required when
422 * the checksums don't match and we have alternate mirrors to try.
423 */
Chris Masonf1885912008-04-09 16:28:12 -0400424static int btree_read_extent_buffer_pages(struct btrfs_root *root,
425 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400426 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400427{
428 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400429 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400430 int ret;
431 int num_copies = 0;
432 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400433 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400434
Josef Bacika826d6d2011-03-16 13:42:43 -0400435 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400436 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
437 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200438 ret = read_extent_buffer_pages(io_tree, eb, start,
439 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400440 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600441 if (!ret) {
442 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400443 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600444 break;
445 else
446 ret = -EIO;
447 }
Chris Masond3977122009-01-05 21:25:51 -0500448
Josef Bacika826d6d2011-03-16 13:42:43 -0400449 /*
450 * This buffer's crc is fine, but its contents are corrupted, so
451 * there is no reason to read the other copies, they won't be
452 * any less wrong.
453 */
454 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400455 break;
456
Stefan Behrens5d964052012-11-05 14:59:07 +0100457 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400458 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400459 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400460 break;
Chris Mason42352982008-04-28 16:40:52 -0400461
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400462 if (!failed_mirror) {
463 failed = 1;
464 failed_mirror = eb->read_mirror;
465 }
466
Chris Masonf1885912008-04-09 16:28:12 -0400467 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400468 if (mirror_num == failed_mirror)
469 mirror_num++;
470
Chris Mason42352982008-04-28 16:40:52 -0400471 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400472 break;
Chris Masonf1885912008-04-09 16:28:12 -0400473 }
Josef Bacikea466792012-03-26 21:57:36 -0400474
Stefan Behrensc0901582012-07-10 07:30:17 -0600475 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400476 repair_eb_io_failure(root, eb, failed_mirror);
477
478 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400479}
Chris Mason19c00dd2007-10-15 16:19:22 -0400480
Chris Masond352ac62008-09-29 15:18:18 -0400481/*
Chris Masond3977122009-01-05 21:25:51 -0500482 * checksum a dirty tree block before IO. This has extra checks to make sure
483 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400484 */
Chris Masond3977122009-01-05 21:25:51 -0500485
Christoph Hellwigb2950862008-12-02 09:54:17 -0500486static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400487{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000488 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400489 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400490 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400491
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500492 eb = (struct extent_buffer *)page->private;
493 if (page != eb->pages[0])
494 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400495 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530496 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500497 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400498 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400499 return 0;
500}
501
Yan Zheng2b820322008-11-17 21:11:30 -0500502static int check_tree_block_fsid(struct btrfs_root *root,
503 struct extent_buffer *eb)
504{
505 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
506 u8 fsid[BTRFS_UUID_SIZE];
507 int ret = 1;
508
Ross Kirk0a4e5582013-09-24 10:12:38 +0100509 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500510 while (fs_devices) {
511 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
512 ret = 0;
513 break;
514 }
515 fs_devices = fs_devices->seed;
516 }
517 return ret;
518}
519
Josef Bacika826d6d2011-03-16 13:42:43 -0400520#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500521 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
522 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200523 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400524
525static noinline int check_leaf(struct btrfs_root *root,
526 struct extent_buffer *leaf)
527{
528 struct btrfs_key key;
529 struct btrfs_key leaf_key;
530 u32 nritems = btrfs_header_nritems(leaf);
531 int slot;
532
533 if (nritems == 0)
534 return 0;
535
536 /* Check the 0 item */
537 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
538 BTRFS_LEAF_DATA_SIZE(root)) {
539 CORRUPT("invalid item offset size pair", leaf, root, 0);
540 return -EIO;
541 }
542
543 /*
544 * Check to make sure each items keys are in the correct order and their
545 * offsets make sense. We only have to loop through nritems-1 because
546 * we check the current slot against the next slot, which verifies the
547 * next slot's offset+size makes sense and that the current's slot
548 * offset is correct.
549 */
550 for (slot = 0; slot < nritems - 1; slot++) {
551 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
552 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
553
554 /* Make sure the keys are in the right order */
555 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
556 CORRUPT("bad key order", leaf, root, slot);
557 return -EIO;
558 }
559
560 /*
561 * Make sure the offset and ends are right, remember that the
562 * item data starts at the end of the leaf and grows towards the
563 * front.
564 */
565 if (btrfs_item_offset_nr(leaf, slot) !=
566 btrfs_item_end_nr(leaf, slot + 1)) {
567 CORRUPT("slot offset bad", leaf, root, slot);
568 return -EIO;
569 }
570
571 /*
572 * Check to make sure that we don't point outside of the leaf,
573 * just incase all the items are consistent to eachother, but
574 * all point outside of the leaf.
575 */
576 if (btrfs_item_end_nr(leaf, slot) >
577 BTRFS_LEAF_DATA_SIZE(root)) {
578 CORRUPT("slot end outside of leaf", leaf, root, slot);
579 return -EIO;
580 }
581 }
582
583 return 0;
584}
585
Miao Xiefacc8a222013-07-25 19:22:34 +0800586static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
587 u64 phy_offset, struct page *page,
588 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400589{
Chris Masonce9adaa2008-04-09 16:28:12 -0400590 u64 found_start;
591 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400592 struct extent_buffer *eb;
593 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400594 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400595 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400596
Chris Masonce9adaa2008-04-09 16:28:12 -0400597 if (!page->private)
598 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500599
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500600 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500601
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400602 /* the pending IO might have been the only thing that kept this buffer
603 * in memory. Make sure we have a ref for all this other checks
604 */
605 extent_buffer_get(eb);
606
607 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400608 if (!reads_done)
609 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400610
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400611 eb->read_mirror = mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400612 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
613 ret = -EIO;
614 goto err;
615 }
616
Chris Masonce9adaa2008-04-09 16:28:12 -0400617 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400618 if (found_start != eb->start) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500619 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400620 "%llu %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200621 found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400622 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400623 goto err;
624 }
Yan Zheng2b820322008-11-17 21:11:30 -0500625 if (check_tree_block_fsid(root, eb)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500626 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200627 eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400628 ret = -EIO;
629 goto err;
630 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400631 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400632 if (found_level >= BTRFS_MAX_LEVEL) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500633 btrfs_info(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400634 (int)btrfs_header_level(eb));
635 ret = -EIO;
636 goto err;
637 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400638
Chris Mason85d4e462011-07-26 16:11:19 -0400639 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
640 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500641
Chris Masonce9adaa2008-04-09 16:28:12 -0400642 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400643 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400644 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400645 goto err;
646 }
647
648 /*
649 * If this is a leaf block and it is corrupt, set the corrupt bit so
650 * that we don't try and read the other copies of this block, just
651 * return -EIO.
652 */
653 if (found_level == 0 && check_leaf(root, eb)) {
654 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
655 ret = -EIO;
656 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400657
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400658 if (!ret)
659 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400660err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400661 if (reads_done &&
662 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200663 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200664
David Woodhouse53b381b2013-01-29 18:40:14 -0500665 if (ret) {
666 /*
667 * our io error hook is going to dec the io pages
668 * again, we have to make sure it has something
669 * to decrement
670 */
671 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400672 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500673 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400674 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400675out:
Chris Masonf1885912008-04-09 16:28:12 -0400676 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400677}
678
Josef Bacikea466792012-03-26 21:57:36 -0400679static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200680{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200681 struct extent_buffer *eb;
682 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
683
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500684 eb = (struct extent_buffer *)page->private;
Josef Bacikea466792012-03-26 21:57:36 -0400685 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400686 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500687 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400688 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200689 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200690 return -EIO; /* we fixed nothing */
691}
692
Chris Masonce9adaa2008-04-09 16:28:12 -0400693static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400694{
695 struct end_io_wq *end_io_wq = bio->bi_private;
696 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400697
Chris Masonce9adaa2008-04-09 16:28:12 -0400698 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400699 end_io_wq->error = err;
Qu Wenruofccb5d82014-02-28 10:46:10 +0800700 btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL);
Chris Masond20f7042008-12-08 16:58:54 -0500701
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200702 if (bio->bi_rw & REQ_WRITE) {
David Woodhouse53b381b2013-01-29 18:40:14 -0500703 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800704 btrfs_queue_work(fs_info->endio_meta_write_workers,
705 &end_io_wq->work);
David Woodhouse53b381b2013-01-29 18:40:14 -0500706 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800707 btrfs_queue_work(fs_info->endio_freespace_worker,
708 &end_io_wq->work);
David Woodhouse53b381b2013-01-29 18:40:14 -0500709 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800710 btrfs_queue_work(fs_info->endio_raid56_workers,
711 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500712 else
Qu Wenruofccb5d82014-02-28 10:46:10 +0800713 btrfs_queue_work(fs_info->endio_write_workers,
714 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500715 } else {
David Woodhouse53b381b2013-01-29 18:40:14 -0500716 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800717 btrfs_queue_work(fs_info->endio_raid56_workers,
718 &end_io_wq->work);
David Woodhouse53b381b2013-01-29 18:40:14 -0500719 else if (end_io_wq->metadata)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800720 btrfs_queue_work(fs_info->endio_meta_workers,
721 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500722 else
Qu Wenruofccb5d82014-02-28 10:46:10 +0800723 btrfs_queue_work(fs_info->endio_workers,
724 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500725 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400726}
727
Josef Bacik0cb59c92010-07-02 12:14:14 -0400728/*
729 * For the metadata arg you want
730 *
731 * 0 - if data
732 * 1 - if normal metadta
733 * 2 - if writing to the free space cache area
David Woodhouse53b381b2013-01-29 18:40:14 -0500734 * 3 - raid parity work
Josef Bacik0cb59c92010-07-02 12:14:14 -0400735 */
Chris Mason22c59942008-04-09 16:28:12 -0400736int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
737 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400738{
Chris Masonce9adaa2008-04-09 16:28:12 -0400739 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400740 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
741 if (!end_io_wq)
742 return -ENOMEM;
743
744 end_io_wq->private = bio->bi_private;
745 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400746 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400747 end_io_wq->error = 0;
748 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400749 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400750
751 bio->bi_private = end_io_wq;
752 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400753 return 0;
754}
755
Chris Masonb64a2852008-08-20 13:39:41 -0400756unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400757{
758 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800759 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400760 info->fs_devices->open_devices);
761 return 256 * limit;
762}
763
Chris Mason4a69a412008-11-06 22:03:00 -0500764static void run_one_async_start(struct btrfs_work *work)
765{
Chris Mason4a69a412008-11-06 22:03:00 -0500766 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100767 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500768
769 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100770 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
771 async->mirror_num, async->bio_flags,
772 async->bio_offset);
773 if (ret)
774 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500775}
776
777static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400778{
779 struct btrfs_fs_info *fs_info;
780 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400781 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400782
783 async = container_of(work, struct async_submit_bio, work);
784 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400785
Chris Masonb64a2852008-08-20 13:39:41 -0400786 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400787 limit = limit * 2 / 3;
788
Josef Bacik66657b32012-08-01 15:36:24 -0400789 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400790 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400791 wake_up(&fs_info->async_submit_wait);
792
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100793 /* If an error occured we just want to clean up the bio and move on */
794 if (async->error) {
795 bio_endio(async->bio, async->error);
796 return;
797 }
798
Chris Mason4a69a412008-11-06 22:03:00 -0500799 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400800 async->mirror_num, async->bio_flags,
801 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500802}
803
804static void run_one_async_free(struct btrfs_work *work)
805{
806 struct async_submit_bio *async;
807
808 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400809 kfree(async);
810}
811
Chris Mason44b8bd72008-04-16 11:14:51 -0400812int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
813 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400814 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400815 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500816 extent_submit_bio_hook_t *submit_bio_start,
817 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400818{
819 struct async_submit_bio *async;
820
821 async = kmalloc(sizeof(*async), GFP_NOFS);
822 if (!async)
823 return -ENOMEM;
824
825 async->inode = inode;
826 async->rw = rw;
827 async->bio = bio;
828 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500829 async->submit_bio_start = submit_bio_start;
830 async->submit_bio_done = submit_bio_done;
831
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800832 btrfs_init_work(&async->work, run_one_async_start,
833 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500834
Chris Masonc8b97812008-10-29 14:49:59 -0400835 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400836 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400837
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100838 async->error = 0;
839
Chris Masoncb03c742008-05-15 16:15:45 -0400840 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400841
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200842 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800843 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400844
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800845 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400846
Chris Masond3977122009-01-05 21:25:51 -0500847 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500848 atomic_read(&fs_info->nr_async_submits)) {
849 wait_event(fs_info->async_submit_wait,
850 (atomic_read(&fs_info->nr_async_submits) == 0));
851 }
852
Chris Mason44b8bd72008-04-16 11:14:51 -0400853 return 0;
854}
855
Chris Masonce3ed712008-09-23 13:14:12 -0400856static int btree_csum_one_bio(struct bio *bio)
857{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800858 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400859 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800860 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400861
Kent Overstreet2c30c712013-11-07 12:20:26 -0800862 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400863 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100864 ret = csum_dirty_buffer(root, bvec->bv_page);
865 if (ret)
866 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400867 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800868
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100869 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400870}
871
Chris Mason4a69a412008-11-06 22:03:00 -0500872static int __btree_submit_bio_start(struct inode *inode, int rw,
873 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400874 unsigned long bio_flags,
875 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400876{
Chris Mason8b712842008-06-11 16:50:36 -0400877 /*
878 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400879 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400880 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100881 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500882}
Chris Mason22c59942008-04-09 16:28:12 -0400883
Chris Mason4a69a412008-11-06 22:03:00 -0500884static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400885 int mirror_num, unsigned long bio_flags,
886 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500887{
Stefan Behrens61891922012-11-05 18:51:52 +0100888 int ret;
889
Chris Mason8b712842008-06-11 16:50:36 -0400890 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500891 * when we're called for a write, we're already in the async
892 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400893 */
Stefan Behrens61891922012-11-05 18:51:52 +0100894 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
895 if (ret)
896 bio_endio(bio, ret);
897 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400898}
899
Josef Bacikde0022b2012-09-25 14:25:58 -0400900static int check_async_write(struct inode *inode, unsigned long bio_flags)
901{
902 if (bio_flags & EXTENT_BIO_TREE_LOG)
903 return 0;
904#ifdef CONFIG_X86
905 if (cpu_has_xmm4_2)
906 return 0;
907#endif
908 return 1;
909}
910
Chris Mason44b8bd72008-04-16 11:14:51 -0400911static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400912 int mirror_num, unsigned long bio_flags,
913 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400914{
Josef Bacikde0022b2012-09-25 14:25:58 -0400915 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500916 int ret;
917
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200918 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500919 /*
920 * called for a read, do the setup so that checksum validation
921 * can happen in the async kernel threads
922 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400923 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
924 bio, 1);
Chris Mason1d4284b2012-03-28 20:31:37 -0400925 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100926 goto out_w_error;
927 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
928 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400929 } else if (!async) {
930 ret = btree_csum_one_bio(bio);
931 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100932 goto out_w_error;
933 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
934 mirror_num, 0);
935 } else {
936 /*
937 * kthread helpers are used to submit writes so that
938 * checksumming can happen in parallel across all CPUs
939 */
940 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
941 inode, rw, bio, mirror_num, 0,
942 bio_offset,
943 __btree_submit_bio_start,
944 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400945 }
Chris Masond313d7a2009-04-20 15:50:09 -0400946
Stefan Behrens61891922012-11-05 18:51:52 +0100947 if (ret) {
948out_w_error:
949 bio_endio(bio, ret);
950 }
951 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400952}
953
Jan Beulich3dd14622010-12-07 14:54:09 +0000954#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500955static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800956 struct page *newpage, struct page *page,
957 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500958{
959 /*
960 * we can't safely write a btree page from here,
961 * we haven't done the locking hook
962 */
963 if (PageDirty(page))
964 return -EAGAIN;
965 /*
966 * Buffers may be managed in a filesystem specific way.
967 * We must have no buffers or drop them.
968 */
969 if (page_has_private(page) &&
970 !try_to_release_page(page, GFP_KERNEL))
971 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800972 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500973}
Jan Beulich3dd14622010-12-07 14:54:09 +0000974#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500975
Chris Mason0da54682007-11-07 21:08:01 -0500976
977static int btree_writepages(struct address_space *mapping,
978 struct writeback_control *wbc)
979{
Miao Xiee2d84522013-01-29 10:09:20 +0000980 struct btrfs_fs_info *fs_info;
981 int ret;
982
Chris Masond8d5f3e2007-12-11 12:42:00 -0500983 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -0800984
985 if (wbc->for_kupdate)
986 return 0;
987
Miao Xiee2d84522013-01-29 10:09:20 +0000988 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -0400989 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +0000990 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
991 BTRFS_DIRTY_METADATA_THRESH);
992 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -0800993 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800994 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400995 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -0500996}
997
Christoph Hellwigb2950862008-12-02 09:54:17 -0500998static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -0400999{
Chris Masond1310b22008-01-24 16:13:08 -05001000 struct extent_io_tree *tree;
1001 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001002 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001003}
1004
Chris Mason70dec802008-01-29 09:59:12 -05001005static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001006{
Chris Mason98509cfc2008-09-11 15:51:43 -04001007 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001008 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001009
David Sterbaf7a52a42013-04-26 14:56:29 +00001010 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001011}
Chris Mason123abc82007-03-14 14:14:43 -04001012
Lukas Czernerd47992f2013-05-21 23:17:23 -04001013static void btree_invalidatepage(struct page *page, unsigned int offset,
1014 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001015{
Chris Masond1310b22008-01-24 16:13:08 -05001016 struct extent_io_tree *tree;
1017 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001018 extent_invalidatepage(tree, page, offset);
1019 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001020 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001021 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1022 "page private not zero on page %llu",
1023 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001024 ClearPagePrivate(page);
1025 set_page_private(page, 0);
1026 page_cache_release(page);
1027 }
Chris Masond98237b2007-03-28 13:57:48 -04001028}
1029
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001030static int btree_set_page_dirty(struct page *page)
1031{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001032#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001033 struct extent_buffer *eb;
1034
1035 BUG_ON(!PagePrivate(page));
1036 eb = (struct extent_buffer *)page->private;
1037 BUG_ON(!eb);
1038 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1039 BUG_ON(!atomic_read(&eb->refs));
1040 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001041#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001042 return __set_page_dirty_nobuffers(page);
1043}
1044
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001045static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001046 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001047 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001048 .releasepage = btree_releasepage,
1049 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001050#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001051 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001052#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001053 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001054};
1055
Chris Masonca7a79a2008-05-12 12:59:19 -04001056int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1057 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001058{
Chris Mason5f39d392007-10-15 16:14:19 -04001059 struct extent_buffer *buf = NULL;
1060 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001061 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001062
Chris Masondb945352007-10-15 16:15:53 -04001063 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001064 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001065 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001066 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001067 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001068 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001069 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001070}
1071
Arne Jansenab0fff02011-05-23 14:25:41 +02001072int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1073 int mirror_num, struct extent_buffer **eb)
1074{
1075 struct extent_buffer *buf = NULL;
1076 struct inode *btree_inode = root->fs_info->btree_inode;
1077 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1078 int ret;
1079
1080 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1081 if (!buf)
1082 return 0;
1083
1084 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1085
1086 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1087 btree_get_extent, mirror_num);
1088 if (ret) {
1089 free_extent_buffer(buf);
1090 return ret;
1091 }
1092
1093 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1094 free_extent_buffer(buf);
1095 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001096 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001097 *eb = buf;
1098 } else {
1099 free_extent_buffer(buf);
1100 }
1101 return 0;
1102}
1103
Chris Mason0999df52008-04-01 13:48:14 -04001104struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1105 u64 bytenr, u32 blocksize)
1106{
Josef Bacikf28491e2013-12-16 13:24:27 -05001107 return find_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001108}
1109
1110struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1111 u64 bytenr, u32 blocksize)
1112{
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001113#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1114 if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
1115 return alloc_test_extent_buffer(root->fs_info, bytenr,
1116 blocksize);
1117#endif
Josef Bacikf28491e2013-12-16 13:24:27 -05001118 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001119}
1120
1121
Chris Masone02119d2008-09-05 16:13:11 -04001122int btrfs_write_tree_block(struct extent_buffer *buf)
1123{
Chris Mason727011e2010-08-06 13:21:20 -04001124 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001125 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001126}
1127
1128int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1129{
Chris Mason727011e2010-08-06 13:21:20 -04001130 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001131 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001132}
1133
Chris Masondb945352007-10-15 16:15:53 -04001134struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001135 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001136{
Chris Mason5f39d392007-10-15 16:14:19 -04001137 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001138 int ret;
1139
Chris Masondb945352007-10-15 16:15:53 -04001140 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001141 if (!buf)
1142 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001143
Chris Masonca7a79a2008-05-12 12:59:19 -04001144 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001145 if (ret) {
1146 free_extent_buffer(buf);
1147 return NULL;
1148 }
Chris Mason5f39d392007-10-15 16:14:19 -04001149 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001150
Chris Masoneb60cea2007-02-02 09:18:22 -05001151}
1152
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001153void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1154 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001155{
Miao Xiee2d84522013-01-29 10:09:20 +00001156 struct btrfs_fs_info *fs_info = root->fs_info;
1157
Chris Mason55c69072008-01-09 15:55:33 -05001158 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001159 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001160 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001161
Chris Masonb9473432009-03-13 11:00:37 -04001162 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001163 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1164 -buf->len,
1165 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001166 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1167 btrfs_set_lock_blocking(buf);
1168 clear_extent_buffer_dirty(buf);
1169 }
Chris Mason925baed2008-06-25 16:01:30 -04001170 }
Chris Mason5f39d392007-10-15 16:14:19 -04001171}
1172
Miao Xie8257b2d2014-03-06 13:38:19 +08001173static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1174{
1175 struct btrfs_subvolume_writers *writers;
1176 int ret;
1177
1178 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1179 if (!writers)
1180 return ERR_PTR(-ENOMEM);
1181
1182 ret = percpu_counter_init(&writers->counter, 0);
1183 if (ret < 0) {
1184 kfree(writers);
1185 return ERR_PTR(ret);
1186 }
1187
1188 init_waitqueue_head(&writers->wait);
1189 return writers;
1190}
1191
1192static void
1193btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1194{
1195 percpu_counter_destroy(&writers->counter);
1196 kfree(writers);
1197}
1198
Jeff Mahoney143bede2012-03-01 14:56:26 +01001199static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1200 u32 stripesize, struct btrfs_root *root,
1201 struct btrfs_fs_info *fs_info,
1202 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001203{
Chris Masoncfaa7292007-02-21 17:04:57 -05001204 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001205 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001206 root->sectorsize = sectorsize;
1207 root->nodesize = nodesize;
1208 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001209 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001210 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001211 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001212
Chris Mason0f7d52f2007-04-09 10:42:37 -04001213 root->objectid = objectid;
1214 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001215 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001216 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001217 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001218 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001219 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001220 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001221 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001222 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001223
1224 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001225 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001226 INIT_LIST_HEAD(&root->delalloc_inodes);
1227 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001228 INIT_LIST_HEAD(&root->ordered_extents);
1229 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001230 INIT_LIST_HEAD(&root->logged_list[0]);
1231 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001232 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001233 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001234 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001235 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001236 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001237 spin_lock_init(&root->log_extents_lock[0]);
1238 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001239 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001240 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001241 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001242 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f1832009-01-21 12:54:03 -05001243 init_waitqueue_head(&root->log_writer_wait);
1244 init_waitqueue_head(&root->log_commit_wait[0]);
1245 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001246 INIT_LIST_HEAD(&root->log_ctxs[0]);
1247 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f1832009-01-21 12:54:03 -05001248 atomic_set(&root->log_commit[0], 0);
1249 atomic_set(&root->log_commit[1], 0);
1250 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001251 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001252 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001253 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001254 atomic_set(&root->will_be_snapshoted, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001255 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001256 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001257 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001258 if (fs_info)
1259 extent_io_tree_init(&root->dirty_log_pages,
1260 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001261
Chris Mason3768f362007-03-13 16:47:54 -04001262 memset(&root->root_key, 0, sizeof(root->root_key));
1263 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001264 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001265 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001266 if (fs_info)
1267 root->defrag_trans_start = fs_info->generation;
1268 else
1269 root->defrag_trans_start = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001270 init_completion(&root->kobj_unregister);
Chris Mason4d775672007-04-20 20:23:12 -04001271 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001272 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001273
Anand Jain5f3ab902012-12-07 09:28:54 +00001274 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001275}
1276
Al Virof84a8bd2011-11-17 15:57:57 -05001277static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001278{
1279 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1280 if (root)
1281 root->fs_info = fs_info;
1282 return root;
1283}
1284
Josef Bacik06ea65a2013-09-19 16:07:01 -04001285#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1286/* Should only be used by the testing infrastructure */
1287struct btrfs_root *btrfs_alloc_dummy_root(void)
1288{
1289 struct btrfs_root *root;
1290
1291 root = btrfs_alloc_root(NULL);
1292 if (!root)
1293 return ERR_PTR(-ENOMEM);
1294 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001295 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001296 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001297
1298 return root;
1299}
1300#endif
1301
Arne Jansen20897f52011-09-13 12:44:20 +02001302struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1303 struct btrfs_fs_info *fs_info,
1304 u64 objectid)
1305{
1306 struct extent_buffer *leaf;
1307 struct btrfs_root *tree_root = fs_info->tree_root;
1308 struct btrfs_root *root;
1309 struct btrfs_key key;
1310 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001311 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001312
1313 root = btrfs_alloc_root(fs_info);
1314 if (!root)
1315 return ERR_PTR(-ENOMEM);
1316
1317 __setup_root(tree_root->nodesize, tree_root->leafsize,
1318 tree_root->sectorsize, tree_root->stripesize,
1319 root, fs_info, objectid);
1320 root->root_key.objectid = objectid;
1321 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1322 root->root_key.offset = 0;
1323
1324 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1325 0, objectid, NULL, 0, 0, 0);
1326 if (IS_ERR(leaf)) {
1327 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001328 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001329 goto fail;
1330 }
1331
Arne Jansen20897f52011-09-13 12:44:20 +02001332 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1333 btrfs_set_header_bytenr(leaf, leaf->start);
1334 btrfs_set_header_generation(leaf, trans->transid);
1335 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1336 btrfs_set_header_owner(leaf, objectid);
1337 root->node = leaf;
1338
Ross Kirk0a4e5582013-09-24 10:12:38 +01001339 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001340 BTRFS_FSID_SIZE);
1341 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001342 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001343 BTRFS_UUID_SIZE);
1344 btrfs_mark_buffer_dirty(leaf);
1345
1346 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001347 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001348
1349 root->root_item.flags = 0;
1350 root->root_item.byte_limit = 0;
1351 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1352 btrfs_set_root_generation(&root->root_item, trans->transid);
1353 btrfs_set_root_level(&root->root_item, 0);
1354 btrfs_set_root_refs(&root->root_item, 1);
1355 btrfs_set_root_used(&root->root_item, leaf->len);
1356 btrfs_set_root_last_snapshot(&root->root_item, 0);
1357 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001358 uuid_le_gen(&uuid);
1359 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001360 root->root_item.drop_level = 0;
1361
1362 key.objectid = objectid;
1363 key.type = BTRFS_ROOT_ITEM_KEY;
1364 key.offset = 0;
1365 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1366 if (ret)
1367 goto fail;
1368
1369 btrfs_tree_unlock(leaf);
1370
Arne Jansen20897f52011-09-13 12:44:20 +02001371 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001372
1373fail:
1374 if (leaf) {
1375 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001376 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001377 free_extent_buffer(leaf);
1378 }
1379 kfree(root);
1380
1381 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001382}
1383
Yan Zheng7237f1832009-01-21 12:54:03 -05001384static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1385 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001386{
1387 struct btrfs_root *root;
1388 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f1832009-01-21 12:54:03 -05001389 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001390
Al Viro6f07e422011-11-17 00:46:16 -05001391 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001392 if (!root)
Yan Zheng7237f1832009-01-21 12:54:03 -05001393 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001394
1395 __setup_root(tree_root->nodesize, tree_root->leafsize,
1396 tree_root->sectorsize, tree_root->stripesize,
1397 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1398
1399 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1400 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1401 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001402
Yan Zheng7237f1832009-01-21 12:54:03 -05001403 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001404 * DON'T set REF_COWS for log trees
1405 *
Yan Zheng7237f1832009-01-21 12:54:03 -05001406 * log trees do not get reference counted because they go away
1407 * before a real commit is actually done. They do store pointers
1408 * to file data extents, and those reference counts still get
1409 * updated (along with back refs to the log tree).
1410 */
Chris Masone02119d2008-09-05 16:13:11 -04001411
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001412 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
Arne Jansen66d7e7f2011-09-12 15:26:38 +02001413 BTRFS_TREE_LOG_OBJECTID, NULL,
Jan Schmidt5581a512012-05-16 17:04:52 +02001414 0, 0, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001415 if (IS_ERR(leaf)) {
1416 kfree(root);
1417 return ERR_CAST(leaf);
1418 }
Chris Masone02119d2008-09-05 16:13:11 -04001419
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001420 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1421 btrfs_set_header_bytenr(leaf, leaf->start);
1422 btrfs_set_header_generation(leaf, trans->transid);
1423 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1424 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f1832009-01-21 12:54:03 -05001425 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001426
1427 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001428 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001429 btrfs_mark_buffer_dirty(root->node);
1430 btrfs_tree_unlock(root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001431 return root;
1432}
1433
1434int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1435 struct btrfs_fs_info *fs_info)
1436{
1437 struct btrfs_root *log_root;
1438
1439 log_root = alloc_log_tree(trans, fs_info);
1440 if (IS_ERR(log_root))
1441 return PTR_ERR(log_root);
1442 WARN_ON(fs_info->log_root_tree);
1443 fs_info->log_root_tree = log_root;
1444 return 0;
1445}
1446
1447int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1448 struct btrfs_root *root)
1449{
1450 struct btrfs_root *log_root;
1451 struct btrfs_inode_item *inode_item;
1452
1453 log_root = alloc_log_tree(trans, root->fs_info);
1454 if (IS_ERR(log_root))
1455 return PTR_ERR(log_root);
1456
1457 log_root->last_trans = trans->transid;
1458 log_root->root_key.offset = root->root_key.objectid;
1459
1460 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001461 btrfs_set_stack_inode_generation(inode_item, 1);
1462 btrfs_set_stack_inode_size(inode_item, 3);
1463 btrfs_set_stack_inode_nlink(inode_item, 1);
1464 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1465 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f1832009-01-21 12:54:03 -05001466
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001467 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001468
1469 WARN_ON(root->log_root);
1470 root->log_root = log_root;
1471 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001472 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001473 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001474 return 0;
1475}
1476
Stefan Behrens35a36212013-08-14 18:12:25 +02001477static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1478 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001479{
1480 struct btrfs_root *root;
1481 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001482 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001483 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001484 u32 blocksize;
Miao Xiecb517ea2013-05-15 07:48:19 +00001485 int ret;
1486
1487 path = btrfs_alloc_path();
1488 if (!path)
1489 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001490
Al Viro6f07e422011-11-17 00:46:16 -05001491 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001492 if (!root) {
1493 ret = -ENOMEM;
1494 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001495 }
1496
Chris Masondb945352007-10-15 16:15:53 -04001497 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001498 tree_root->sectorsize, tree_root->stripesize,
Miao Xiecb517ea2013-05-15 07:48:19 +00001499 root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001500
Miao Xiecb517ea2013-05-15 07:48:19 +00001501 ret = btrfs_find_root(tree_root, key, path,
1502 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001503 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001504 if (ret > 0)
1505 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001506 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001507 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001508
Yan Zheng84234f32008-10-29 14:49:05 -04001509 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001510 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1511 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001512 blocksize, generation);
Miao Xiecb517ea2013-05-15 07:48:19 +00001513 if (!root->node) {
1514 ret = -ENOMEM;
1515 goto find_fail;
1516 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1517 ret = -EIO;
1518 goto read_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001519 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001520 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001521out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001522 btrfs_free_path(path);
1523 return root;
1524
1525read_fail:
1526 free_extent_buffer(root->node);
1527find_fail:
1528 kfree(root);
1529alloc_fail:
1530 root = ERR_PTR(ret);
1531 goto out;
1532}
1533
1534struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1535 struct btrfs_key *location)
1536{
1537 struct btrfs_root *root;
1538
1539 root = btrfs_read_tree_root(tree_root, location);
1540 if (IS_ERR(root))
1541 return root;
1542
1543 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001544 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001545 btrfs_check_and_init_root_item(&root->root_item);
1546 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001547
Chris Mason5eda7b52007-06-22 14:16:25 -04001548 return root;
1549}
1550
Miao Xiecb517ea2013-05-15 07:48:19 +00001551int btrfs_init_fs_root(struct btrfs_root *root)
1552{
1553 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001554 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001555
1556 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1557 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1558 GFP_NOFS);
1559 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1560 ret = -ENOMEM;
1561 goto fail;
1562 }
1563
Miao Xie8257b2d2014-03-06 13:38:19 +08001564 writers = btrfs_alloc_subvolume_writers();
1565 if (IS_ERR(writers)) {
1566 ret = PTR_ERR(writers);
1567 goto fail;
1568 }
1569 root->subv_writers = writers;
1570
Miao Xiecb517ea2013-05-15 07:48:19 +00001571 btrfs_init_free_ino_ctl(root);
Miao Xiecb517ea2013-05-15 07:48:19 +00001572 spin_lock_init(&root->cache_lock);
1573 init_waitqueue_head(&root->cache_wait);
1574
1575 ret = get_anon_bdev(&root->anon_dev);
1576 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001577 goto free_writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001578 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001579
1580free_writers:
1581 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001582fail:
1583 kfree(root->free_ino_ctl);
1584 kfree(root->free_ino_pinned);
1585 return ret;
1586}
1587
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001588static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1589 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001590{
1591 struct btrfs_root *root;
1592
1593 spin_lock(&fs_info->fs_roots_radix_lock);
1594 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1595 (unsigned long)root_id);
1596 spin_unlock(&fs_info->fs_roots_radix_lock);
1597 return root;
1598}
1599
1600int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1601 struct btrfs_root *root)
1602{
1603 int ret;
1604
1605 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1606 if (ret)
1607 return ret;
1608
1609 spin_lock(&fs_info->fs_roots_radix_lock);
1610 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1611 (unsigned long)root->root_key.objectid,
1612 root);
1613 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001614 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001615 spin_unlock(&fs_info->fs_roots_radix_lock);
1616 radix_tree_preload_end();
1617
1618 return ret;
1619}
1620
Miao Xiec00869f2013-09-25 21:47:44 +08001621struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1622 struct btrfs_key *location,
1623 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001624{
1625 struct btrfs_root *root;
1626 int ret;
1627
Chris Masonedbd8d42007-12-21 16:27:24 -05001628 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1629 return fs_info->tree_root;
1630 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1631 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001632 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1633 return fs_info->chunk_root;
1634 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1635 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001636 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1637 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001638 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1639 return fs_info->quota_root ? fs_info->quota_root :
1640 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001641 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1642 return fs_info->uuid_root ? fs_info->uuid_root :
1643 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001644again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001645 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001646 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001647 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001648 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001649 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001650 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001651
Miao Xiecb517ea2013-05-15 07:48:19 +00001652 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001653 if (IS_ERR(root))
1654 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001655
Miao Xiec00869f2013-09-25 21:47:44 +08001656 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001657 ret = -ENOENT;
1658 goto fail;
1659 }
1660
Miao Xiecb517ea2013-05-15 07:48:19 +00001661 ret = btrfs_init_fs_root(root);
1662 if (ret)
1663 goto fail;
1664
Kelley Nielsen3f870c22013-11-04 19:37:39 -08001665 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1666 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001667 if (ret < 0)
1668 goto fail;
1669 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001670 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001671
Miao Xiecb517ea2013-05-15 07:48:19 +00001672 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001673 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001674 if (ret == -EEXIST) {
1675 free_fs_root(root);
1676 goto again;
1677 }
1678 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001679 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001680 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001681fail:
1682 free_fs_root(root);
1683 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001684}
1685
Chris Mason04160082008-03-26 10:28:07 -04001686static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1687{
1688 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1689 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001690 struct btrfs_device *device;
1691 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001692
Xiao Guangrong1f781602011-04-20 10:09:16 +00001693 rcu_read_lock();
1694 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001695 if (!device->bdev)
1696 continue;
Chris Mason04160082008-03-26 10:28:07 -04001697 bdi = blk_get_backing_dev_info(device->bdev);
1698 if (bdi && bdi_congested(bdi, bdi_bits)) {
1699 ret = 1;
1700 break;
1701 }
1702 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001703 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001704 return ret;
1705}
1706
Chris Mason38b66982008-04-22 09:22:11 -04001707/*
Jens Axboead081f12009-06-12 14:43:40 +02001708 * If this fails, caller must call bdi_destroy() to get rid of the
1709 * bdi again.
1710 */
Chris Mason04160082008-03-26 10:28:07 -04001711static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1712{
Jens Axboead081f12009-06-12 14:43:40 +02001713 int err;
1714
1715 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001716 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001717 if (err)
1718 return err;
1719
Chris Mason4575c9c2008-04-18 16:13:31 -04001720 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001721 bdi->congested_fn = btrfs_congested_fn;
1722 bdi->congested_data = info;
1723 return 0;
1724}
1725
Chris Mason8b712842008-06-11 16:50:36 -04001726/*
1727 * called by the kthread helper functions to finally call the bio end_io
1728 * functions. This is where read checksum verification actually happens
1729 */
1730static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001731{
Chris Masonce9adaa2008-04-09 16:28:12 -04001732 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001733 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001734 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001735
Chris Mason8b712842008-06-11 16:50:36 -04001736 end_io_wq = container_of(work, struct end_io_wq, work);
1737 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001738
Chris Mason8b712842008-06-11 16:50:36 -04001739 error = end_io_wq->error;
1740 bio->bi_private = end_io_wq->private;
1741 bio->bi_end_io = end_io_wq->end_io;
1742 kfree(end_io_wq);
Kent Overstreetbc1e79a2013-12-03 13:24:08 -08001743 bio_endio_nodec(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001744}
1745
Chris Masona74a4b92008-06-25 16:01:31 -04001746static int cleaner_kthread(void *arg)
1747{
1748 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001749 int again;
Chris Masona74a4b92008-06-25 16:01:31 -04001750
1751 do {
Miao Xied0278242013-05-14 10:20:40 +00001752 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001753
Miao Xied0278242013-05-14 10:20:40 +00001754 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001755 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001756 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001757
Miao Xied0278242013-05-14 10:20:40 +00001758 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1759 goto sleep;
1760
Miao Xiedc7f3702013-05-14 10:20:42 +00001761 /*
1762 * Avoid the problem that we change the status of the fs
1763 * during the above check and trylock.
1764 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001765 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001766 mutex_unlock(&root->fs_info->cleaner_mutex);
1767 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001768 }
1769
Miao Xied0278242013-05-14 10:20:40 +00001770 btrfs_run_delayed_iputs(root);
1771 again = btrfs_clean_one_deleted_snapshot(root);
1772 mutex_unlock(&root->fs_info->cleaner_mutex);
1773
1774 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001775 * The defragger has dealt with the R/O remount and umount,
1776 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001777 */
1778 btrfs_run_defrag_inodes(root->fs_info);
1779sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001780 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001781 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001782 if (!kthread_should_stop())
1783 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001784 __set_current_state(TASK_RUNNING);
1785 }
1786 } while (!kthread_should_stop());
1787 return 0;
1788}
1789
1790static int transaction_kthread(void *arg)
1791{
1792 struct btrfs_root *root = arg;
1793 struct btrfs_trans_handle *trans;
1794 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001795 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001796 unsigned long now;
1797 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001798 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001799
1800 do {
Jan Kara914b2002012-03-12 16:05:50 +01001801 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001802 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001803 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1804
Josef Bacika4abeea2011-04-11 17:25:13 -04001805 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001806 cur = root->fs_info->running_transaction;
1807 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001808 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001809 goto sleep;
1810 }
Yan Zheng31153d82008-07-28 15:32:19 -04001811
Chris Masona74a4b92008-06-25 16:01:31 -04001812 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001813 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001814 (now < cur->start_time ||
1815 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001816 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001817 delay = HZ * 5;
1818 goto sleep;
1819 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001820 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001821 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001822
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001823 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001824 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001825 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001826 if (PTR_ERR(trans) != -ENOENT)
1827 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001828 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001829 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001830 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001831 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001832 } else {
1833 btrfs_end_transaction(trans, root);
1834 }
Chris Masona74a4b92008-06-25 16:01:31 -04001835sleep:
1836 wake_up_process(root->fs_info->cleaner_kthread);
1837 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1838
Josef Bacik4e121c02013-09-27 16:32:39 -04001839 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1840 &root->fs_info->fs_state)))
1841 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001842 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001843 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001844 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001845 (!btrfs_transaction_blocked(root->fs_info) ||
1846 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001847 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001848 __set_current_state(TASK_RUNNING);
1849 }
1850 } while (!kthread_should_stop());
1851 return 0;
1852}
1853
Chris Masonaf31f5e2011-11-03 15:17:42 -04001854/*
1855 * this will find the highest generation in the array of
1856 * root backups. The index of the highest array is returned,
1857 * or -1 if we can't find anything.
1858 *
1859 * We check to make sure the array is valid by comparing the
1860 * generation of the latest root in the array with the generation
1861 * in the super block. If they don't match we pitch it.
1862 */
1863static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1864{
1865 u64 cur;
1866 int newest_index = -1;
1867 struct btrfs_root_backup *root_backup;
1868 int i;
1869
1870 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1871 root_backup = info->super_copy->super_roots + i;
1872 cur = btrfs_backup_tree_root_gen(root_backup);
1873 if (cur == newest_gen)
1874 newest_index = i;
1875 }
1876
1877 /* check to see if we actually wrapped around */
1878 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1879 root_backup = info->super_copy->super_roots;
1880 cur = btrfs_backup_tree_root_gen(root_backup);
1881 if (cur == newest_gen)
1882 newest_index = 0;
1883 }
1884 return newest_index;
1885}
1886
1887
1888/*
1889 * find the oldest backup so we know where to store new entries
1890 * in the backup array. This will set the backup_root_index
1891 * field in the fs_info struct
1892 */
1893static void find_oldest_super_backup(struct btrfs_fs_info *info,
1894 u64 newest_gen)
1895{
1896 int newest_index = -1;
1897
1898 newest_index = find_newest_super_backup(info, newest_gen);
1899 /* if there was garbage in there, just move along */
1900 if (newest_index == -1) {
1901 info->backup_root_index = 0;
1902 } else {
1903 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1904 }
1905}
1906
1907/*
1908 * copy all the root pointers into the super backup array.
1909 * this will bump the backup pointer by one when it is
1910 * done
1911 */
1912static void backup_super_roots(struct btrfs_fs_info *info)
1913{
1914 int next_backup;
1915 struct btrfs_root_backup *root_backup;
1916 int last_backup;
1917
1918 next_backup = info->backup_root_index;
1919 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1920 BTRFS_NUM_BACKUP_ROOTS;
1921
1922 /*
1923 * just overwrite the last backup if we're at the same generation
1924 * this happens only at umount
1925 */
1926 root_backup = info->super_for_commit->super_roots + last_backup;
1927 if (btrfs_backup_tree_root_gen(root_backup) ==
1928 btrfs_header_generation(info->tree_root->node))
1929 next_backup = last_backup;
1930
1931 root_backup = info->super_for_commit->super_roots + next_backup;
1932
1933 /*
1934 * make sure all of our padding and empty slots get zero filled
1935 * regardless of which ones we use today
1936 */
1937 memset(root_backup, 0, sizeof(*root_backup));
1938
1939 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1940
1941 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1942 btrfs_set_backup_tree_root_gen(root_backup,
1943 btrfs_header_generation(info->tree_root->node));
1944
1945 btrfs_set_backup_tree_root_level(root_backup,
1946 btrfs_header_level(info->tree_root->node));
1947
1948 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1949 btrfs_set_backup_chunk_root_gen(root_backup,
1950 btrfs_header_generation(info->chunk_root->node));
1951 btrfs_set_backup_chunk_root_level(root_backup,
1952 btrfs_header_level(info->chunk_root->node));
1953
1954 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1955 btrfs_set_backup_extent_root_gen(root_backup,
1956 btrfs_header_generation(info->extent_root->node));
1957 btrfs_set_backup_extent_root_level(root_backup,
1958 btrfs_header_level(info->extent_root->node));
1959
Chris Mason7c7e82a2011-11-06 18:50:56 -05001960 /*
1961 * we might commit during log recovery, which happens before we set
1962 * the fs_root. Make sure it is valid before we fill it in.
1963 */
1964 if (info->fs_root && info->fs_root->node) {
1965 btrfs_set_backup_fs_root(root_backup,
1966 info->fs_root->node->start);
1967 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001968 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001969 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001970 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001971 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001972
1973 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1974 btrfs_set_backup_dev_root_gen(root_backup,
1975 btrfs_header_generation(info->dev_root->node));
1976 btrfs_set_backup_dev_root_level(root_backup,
1977 btrfs_header_level(info->dev_root->node));
1978
1979 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1980 btrfs_set_backup_csum_root_gen(root_backup,
1981 btrfs_header_generation(info->csum_root->node));
1982 btrfs_set_backup_csum_root_level(root_backup,
1983 btrfs_header_level(info->csum_root->node));
1984
1985 btrfs_set_backup_total_bytes(root_backup,
1986 btrfs_super_total_bytes(info->super_copy));
1987 btrfs_set_backup_bytes_used(root_backup,
1988 btrfs_super_bytes_used(info->super_copy));
1989 btrfs_set_backup_num_devices(root_backup,
1990 btrfs_super_num_devices(info->super_copy));
1991
1992 /*
1993 * if we don't copy this out to the super_copy, it won't get remembered
1994 * for the next commit
1995 */
1996 memcpy(&info->super_copy->super_roots,
1997 &info->super_for_commit->super_roots,
1998 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1999}
2000
2001/*
2002 * this copies info out of the root backup array and back into
2003 * the in-memory super block. It is meant to help iterate through
2004 * the array, so you send it the number of backups you've already
2005 * tried and the last backup index you used.
2006 *
2007 * this returns -1 when it has tried all the backups
2008 */
2009static noinline int next_root_backup(struct btrfs_fs_info *info,
2010 struct btrfs_super_block *super,
2011 int *num_backups_tried, int *backup_index)
2012{
2013 struct btrfs_root_backup *root_backup;
2014 int newest = *backup_index;
2015
2016 if (*num_backups_tried == 0) {
2017 u64 gen = btrfs_super_generation(super);
2018
2019 newest = find_newest_super_backup(info, gen);
2020 if (newest == -1)
2021 return -1;
2022
2023 *backup_index = newest;
2024 *num_backups_tried = 1;
2025 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2026 /* we've tried all the backups, all done */
2027 return -1;
2028 } else {
2029 /* jump to the next oldest backup */
2030 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2031 BTRFS_NUM_BACKUP_ROOTS;
2032 *backup_index = newest;
2033 *num_backups_tried += 1;
2034 }
2035 root_backup = super->super_roots + newest;
2036
2037 btrfs_set_super_generation(super,
2038 btrfs_backup_tree_root_gen(root_backup));
2039 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2040 btrfs_set_super_root_level(super,
2041 btrfs_backup_tree_root_level(root_backup));
2042 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2043
2044 /*
2045 * fixme: the total bytes and num_devices need to match or we should
2046 * need a fsck
2047 */
2048 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2049 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2050 return 0;
2051}
2052
Liu Bo7abadb62013-03-17 02:10:31 +00002053/* helper to cleanup workers */
2054static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2055{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002056 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002057 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002058 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002059 btrfs_destroy_workqueue(fs_info->endio_workers);
2060 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2061 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002062 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002063 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2064 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2065 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002066 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002067 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002068 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002069 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002070 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002071 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002072}
2073
Rashika2e9f5952013-10-31 02:45:20 +05302074static void free_root_extent_buffers(struct btrfs_root *root)
2075{
2076 if (root) {
2077 free_extent_buffer(root->node);
2078 free_extent_buffer(root->commit_root);
2079 root->node = NULL;
2080 root->commit_root = NULL;
2081 }
2082}
2083
Chris Masonaf31f5e2011-11-03 15:17:42 -04002084/* helper to cleanup tree roots */
2085static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2086{
Rashika2e9f5952013-10-31 02:45:20 +05302087 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002088
Rashika2e9f5952013-10-31 02:45:20 +05302089 free_root_extent_buffers(info->dev_root);
2090 free_root_extent_buffers(info->extent_root);
2091 free_root_extent_buffers(info->csum_root);
2092 free_root_extent_buffers(info->quota_root);
2093 free_root_extent_buffers(info->uuid_root);
2094 if (chunk_root)
2095 free_root_extent_buffers(info->chunk_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002096}
2097
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002098void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002099{
2100 int ret;
2101 struct btrfs_root *gang[8];
2102 int i;
2103
2104 while (!list_empty(&fs_info->dead_roots)) {
2105 gang[0] = list_entry(fs_info->dead_roots.next,
2106 struct btrfs_root, root_list);
2107 list_del(&gang[0]->root_list);
2108
Miao Xie27cdeb72014-04-02 19:51:05 +08002109 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002110 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002111 } else {
2112 free_extent_buffer(gang[0]->node);
2113 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002114 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002115 }
2116 }
2117
2118 while (1) {
2119 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2120 (void **)gang, 0,
2121 ARRAY_SIZE(gang));
2122 if (!ret)
2123 break;
2124 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002125 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002126 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002127
2128 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2129 btrfs_free_log_root_tree(NULL, fs_info);
2130 btrfs_destroy_pinned_extent(fs_info->tree_root,
2131 fs_info->pinned_extents);
2132 }
Josef Bacik171f6532013-04-24 16:35:41 -04002133}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002134
Al Viroad2b2c82011-11-17 01:10:02 -05002135int open_ctree(struct super_block *sb,
2136 struct btrfs_fs_devices *fs_devices,
2137 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002138{
Chris Masondb945352007-10-15 16:15:53 -04002139 u32 sectorsize;
2140 u32 nodesize;
2141 u32 leafsize;
2142 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002143 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002144 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002145 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002146 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002147 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002148 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002149 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002150 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002151 struct btrfs_root *extent_root;
2152 struct btrfs_root *csum_root;
2153 struct btrfs_root *chunk_root;
2154 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002155 struct btrfs_root *quota_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002156 struct btrfs_root *uuid_root;
Chris Masone02119d2008-09-05 16:13:11 -04002157 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002158 int ret;
Yane58ca022008-04-01 11:21:34 -04002159 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002160 int num_backups_tried = 0;
2161 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002162 int max_active;
2163 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Stefan Behrens70f80172013-08-15 17:11:23 +02002164 bool create_uuid_tree;
2165 bool check_uuid_tree;
Chris Mason4543df72008-06-11 21:47:56 -04002166
Al Virof84a8bd2011-11-17 15:57:57 -05002167 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002168 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002169 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002170 err = -ENOMEM;
2171 goto fail;
2172 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002173
2174 ret = init_srcu_struct(&fs_info->subvol_srcu);
2175 if (ret) {
2176 err = ret;
2177 goto fail;
2178 }
2179
2180 ret = setup_bdi(fs_info, &fs_info->bdi);
2181 if (ret) {
2182 err = ret;
2183 goto fail_srcu;
2184 }
2185
Miao Xiee2d84522013-01-29 10:09:20 +00002186 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2187 if (ret) {
2188 err = ret;
2189 goto fail_bdi;
2190 }
2191 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2192 (1 + ilog2(nr_cpu_ids));
2193
Miao Xie963d6782013-01-29 10:10:51 +00002194 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2195 if (ret) {
2196 err = ret;
2197 goto fail_dirty_metadata_bytes;
2198 }
2199
Miao Xiec404e0d2014-01-30 16:46:55 +08002200 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2201 if (ret) {
2202 err = ret;
2203 goto fail_delalloc_bytes;
2204 }
2205
Yan, Zheng76dda932009-09-21 16:00:26 -04002206 fs_info->btree_inode = new_inode(sb);
2207 if (!fs_info->btree_inode) {
2208 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002209 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002210 }
2211
Chris Masona6591712011-07-19 12:04:14 -04002212 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002213
Yan, Zheng76dda932009-09-21 16:00:26 -04002214 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002215 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002216 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002217 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002218 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002219 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002220 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002221 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002222 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002223 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002224 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002225 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002226 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002227 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002228 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002229 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002230 spin_lock_init(&fs_info->buffer_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002231 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002232 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002233 mutex_init(&fs_info->delalloc_root_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002234 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002235
Josef Bacik58176a92007-08-29 15:47:34 -04002236 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002237 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002238 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002239 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Chris Mason0b86a832008-03-24 15:01:56 -04002240 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002241 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2242 BTRFS_BLOCK_RSV_GLOBAL);
2243 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2244 BTRFS_BLOCK_RSV_DELALLOC);
2245 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2246 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2247 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2248 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2249 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002250 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002251 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002252 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002253 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002254 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002255 atomic_set(&fs_info->qgroup_op_seq, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002256 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002257 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05002258 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002259 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002260 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002261 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002262 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002263 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
Josef Bacik0a2b2a82014-01-23 10:54:11 -05002264 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
Arne Jansen90519d62011-05-23 14:30:00 +02002265 /* readahead state */
2266 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2267 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002268
Chris Masonb34b0862008-12-19 15:43:22 -05002269 fs_info->thread_pool_size = min_t(unsigned long,
2270 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002271
Miao Xie199c2a92013-05-15 07:48:23 +00002272 INIT_LIST_HEAD(&fs_info->ordered_roots);
2273 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002274 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2275 GFP_NOFS);
2276 if (!fs_info->delayed_root) {
2277 err = -ENOMEM;
2278 goto fail_iput;
2279 }
2280 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002281
Arne Jansena2de7332011-03-08 14:14:00 +01002282 mutex_init(&fs_info->scrub_lock);
2283 atomic_set(&fs_info->scrubs_running, 0);
2284 atomic_set(&fs_info->scrub_pause_req, 0);
2285 atomic_set(&fs_info->scrubs_paused, 0);
2286 atomic_set(&fs_info->scrub_cancel_req, 0);
Miao Xiec404e0d2014-01-30 16:46:55 +08002287 init_waitqueue_head(&fs_info->replace_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002288 init_waitqueue_head(&fs_info->scrub_pause_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002289 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002290#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2291 fs_info->check_integrity_print_mask = 0;
2292#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002293
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002294 spin_lock_init(&fs_info->balance_lock);
2295 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002296 atomic_set(&fs_info->balance_running, 0);
2297 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002298 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002299 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002300 init_waitqueue_head(&fs_info->balance_wait_q);
Miao Xie21c7e752014-05-13 17:29:04 -07002301 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Chris Masona061fc82008-05-07 11:43:44 -04002302
Chris Mason0afbaf82008-04-18 14:17:20 -04002303 sb->s_blocksize = 4096;
2304 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002305 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002306
Yan, Zheng76dda932009-09-21 16:00:26 -04002307 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002308 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002309 /*
2310 * we set the i_size on the btree inode to the max possible int.
2311 * the real end of the address space is determined by all of
2312 * the devices in the system
2313 */
2314 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002315 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002316 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2317
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002318 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002319 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002320 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002321 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002322 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002323
Chris Masond1310b22008-01-24 16:13:08 -05002324 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2325
Chris Mason0f7d52f2007-04-09 10:42:37 -04002326 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2327 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2328 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002329 set_bit(BTRFS_INODE_DUMMY,
2330 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Filipe David Borba Manana778ba822013-10-06 22:22:33 +01002331 btrfs_insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002332
Chris Masone02119d2008-09-05 16:13:11 -04002333 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002334 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002335 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002336
Yan Zheng11833d62009-09-11 16:11:19 -04002337 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002338 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002339 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002340 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002341 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002342 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002343
Chris Masond98237b2007-03-28 13:57:48 -04002344
Chris Mason5a3f23d2009-03-31 13:27:11 -04002345 mutex_init(&fs_info->ordered_operations_mutex);
Josef Bacik9ffba8c2013-08-14 11:33:56 -04002346 mutex_init(&fs_info->ordered_extent_flush_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002347 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002348 mutex_init(&fs_info->chunk_mutex);
2349 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002350 mutex_init(&fs_info->cleaner_mutex);
2351 mutex_init(&fs_info->volume_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002352 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002353 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002354 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002355 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Stefan Behrense922e082012-11-05 17:26:40 +01002356 fs_info->dev_replace.lock_owner = 0;
2357 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2358 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2359 mutex_init(&fs_info->dev_replace.lock_management_lock);
2360 mutex_init(&fs_info->dev_replace.lock);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002361
Arne Jansen416ac512011-09-13 12:56:09 +02002362 spin_lock_init(&fs_info->qgroup_lock);
Wang Shilongf2f6ed32013-04-07 10:50:16 +00002363 mutex_init(&fs_info->qgroup_ioctl_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002364 fs_info->qgroup_tree = RB_ROOT;
Josef Bacikfcebe452014-05-13 17:30:47 -07002365 fs_info->qgroup_op_tree = RB_ROOT;
Arne Jansen416ac512011-09-13 12:56:09 +02002366 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2367 fs_info->qgroup_seq = 1;
2368 fs_info->quota_enabled = 0;
2369 fs_info->pending_quota_state = 0;
Wang Shilong1e8f9152013-05-06 11:03:27 +00002370 fs_info->qgroup_ulist = NULL;
Jan Schmidt2f232032013-04-25 16:04:51 +00002371 mutex_init(&fs_info->qgroup_rescan_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002372
Chris Masonfa9c0d792009-04-03 09:47:43 -04002373 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2374 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2375
Chris Mason7d9eb122008-07-08 14:19:17 -04002376 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002377 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002378 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002379 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002380
David Woodhouse53b381b2013-01-29 18:40:14 -05002381 ret = btrfs_alloc_stripe_hash_table(fs_info);
2382 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002383 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002384 goto fail_alloc;
2385 }
2386
Chris Mason0b86a832008-03-24 15:01:56 -04002387 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002388 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002389
Chris Mason3c4bb262012-03-27 18:56:56 -04002390 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002391
2392 /*
2393 * Read super block and check the signature bytes only
2394 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002395 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002396 if (!bh) {
2397 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002398 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002399 }
Chris Mason39279cc2007-06-12 06:35:45 -04002400
David Sterba1104a882013-03-06 15:57:46 +01002401 /*
2402 * We want to check superblock checksum, the type is stored inside.
2403 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2404 */
2405 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002406 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002407 err = -EINVAL;
2408 goto fail_alloc;
2409 }
2410
2411 /*
2412 * super_copy is zeroed at allocation time and we never touch the
2413 * following bytes up to INFO_SIZE, the checksum is calculated from
2414 * the whole block of INFO_SIZE
2415 */
David Sterba6c417612011-04-13 15:41:04 +02002416 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2417 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2418 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002419 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002420
David Sterba6c417612011-04-13 15:41:04 +02002421 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002422
David Sterba1104a882013-03-06 15:57:46 +01002423 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2424 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002425 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002426 err = -EINVAL;
2427 goto fail_alloc;
2428 }
2429
David Sterba6c417612011-04-13 15:41:04 +02002430 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002431 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002432 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002433
liuboacce9522011-01-06 19:30:25 +08002434 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002435 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2436 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002437
Liu Bo75e7cb72011-03-22 10:12:20 +00002438 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002439 * run through our array of backup supers and setup
2440 * our ring pointer to the oldest one
2441 */
2442 generation = btrfs_super_generation(disk_super);
2443 find_oldest_super_backup(fs_info, generation);
2444
2445 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002446 * In the long term, we'll store the compression type in the super
2447 * block, and it'll be used for per file compression control.
2448 */
2449 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2450
Yan Zheng2b820322008-11-17 21:11:30 -05002451 ret = btrfs_parse_options(tree_root, options);
2452 if (ret) {
2453 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002454 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002455 }
Chris Masondfe25022008-05-13 13:46:40 -04002456
Josef Bacikf2b636e2008-12-02 06:36:08 -05002457 features = btrfs_super_incompat_flags(disk_super) &
2458 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2459 if (features) {
2460 printk(KERN_ERR "BTRFS: couldn't mount because of "
2461 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002462 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002463 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002464 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002465 }
2466
Chris Mason727011e2010-08-06 13:21:20 -04002467 if (btrfs_super_leafsize(disk_super) !=
2468 btrfs_super_nodesize(disk_super)) {
2469 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2470 "blocksizes don't match. node %d leaf %d\n",
2471 btrfs_super_nodesize(disk_super),
2472 btrfs_super_leafsize(disk_super));
2473 err = -EINVAL;
2474 goto fail_alloc;
2475 }
2476 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2477 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2478 "blocksize (%d) was too large\n",
2479 btrfs_super_leafsize(disk_super));
2480 err = -EINVAL;
2481 goto fail_alloc;
2482 }
2483
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002484 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002485 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002486 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002487 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002488
Josef Bacik3173a182013-03-07 14:22:04 -05002489 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
Frank Holtonefe120a2013-12-20 11:37:06 -05002490 printk(KERN_ERR "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002491
Chris Mason727011e2010-08-06 13:21:20 -04002492 /*
2493 * flag our filesystem as having big metadata blocks if
2494 * they are bigger than the page size
2495 */
2496 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2497 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002498 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002499 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2500 }
2501
Chris Masonbc3f1162012-03-29 17:02:47 -04002502 nodesize = btrfs_super_nodesize(disk_super);
2503 leafsize = btrfs_super_leafsize(disk_super);
2504 sectorsize = btrfs_super_sectorsize(disk_super);
2505 stripesize = btrfs_super_stripesize(disk_super);
Miao Xiee2d84522013-01-29 10:09:20 +00002506 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002507 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002508
2509 /*
2510 * mixed block groups end up with duplicate but slightly offset
2511 * extent buffers for the same range. It leads to corruptions
2512 */
2513 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2514 (sectorsize != leafsize)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002515 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002516 "are not allowed for mixed block groups on %s\n",
2517 sb->s_id);
2518 goto fail_alloc;
2519 }
2520
Miao Xieceda0862013-04-11 10:30:16 +00002521 /*
2522 * Needn't use the lock because there is no other task which will
2523 * update the flag.
2524 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002525 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002526
Josef Bacikf2b636e2008-12-02 06:36:08 -05002527 features = btrfs_super_compat_ro_flags(disk_super) &
2528 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2529 if (!(sb->s_flags & MS_RDONLY) && features) {
2530 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2531 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002532 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002533 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002534 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002535 }
Chris Mason61d92c32009-10-02 19:11:56 -04002536
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002537 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002538
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002539 fs_info->workers =
2540 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2541 max_active, 16);
Chris Masonc8b97812008-10-29 14:49:59 -04002542
Qu Wenruoafe3d242014-02-28 10:46:07 +08002543 fs_info->delalloc_workers =
2544 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
Chris Mason771ed682008-11-06 22:02:51 -05002545
Qu Wenruoa44903a2014-02-28 10:46:09 +08002546 fs_info->flush_workers =
2547 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
Miao Xie8ccf6f192012-10-25 09:28:04 +00002548
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002549 fs_info->caching_workers =
2550 btrfs_alloc_workqueue("cache", flags, max_active, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002551
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002552 /*
2553 * a higher idle thresh on the submit workers makes it much more
Chris Mason61b49442008-07-31 15:42:53 -04002554 * likely that bios will be send down in a sane order to the
2555 * devices
2556 */
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002557 fs_info->submit_workers =
2558 btrfs_alloc_workqueue("submit", flags,
2559 min_t(u64, fs_devices->num_devices,
2560 max_active), 64);
Chris Mason53863232008-08-15 15:34:18 -04002561
Qu Wenruodc6e3202014-02-28 10:46:14 +08002562 fs_info->fixup_workers =
2563 btrfs_alloc_workqueue("fixup", flags, 1, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002564
2565 /*
2566 * endios are largely parallel and should have a very
2567 * low idle thresh
2568 */
Qu Wenruofccb5d82014-02-28 10:46:10 +08002569 fs_info->endio_workers =
2570 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2571 fs_info->endio_meta_workers =
2572 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2573 fs_info->endio_meta_write_workers =
2574 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2575 fs_info->endio_raid56_workers =
2576 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002577 fs_info->rmw_workers =
2578 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002579 fs_info->endio_write_workers =
2580 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2581 fs_info->endio_freespace_worker =
2582 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002583 fs_info->delayed_workers =
2584 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002585 fs_info->readahead_workers =
2586 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002587 fs_info->qgroup_rescan_workers =
2588 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
Chris Masonb51912c2009-02-04 09:23:24 -05002589
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002590 if (!(fs_info->workers && fs_info->delalloc_workers &&
Qu Wenruofccb5d82014-02-28 10:46:10 +08002591 fs_info->submit_workers && fs_info->flush_workers &&
2592 fs_info->endio_workers && fs_info->endio_meta_workers &&
2593 fs_info->endio_meta_write_workers &&
2594 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002595 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
Qu Wenruodc6e3202014-02-28 10:46:14 +08002596 fs_info->caching_workers && fs_info->readahead_workers &&
Qu Wenruofc97fab2014-02-28 10:46:16 +08002597 fs_info->fixup_workers && fs_info->delayed_workers &&
2598 fs_info->qgroup_rescan_workers)) {
Ilya Dryomovfed425c2012-06-22 12:13:01 -06002599 err = -ENOMEM;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002600 goto fail_sb_buffer;
2601 }
Chris Mason4543df72008-06-11 21:47:56 -04002602
Chris Mason4575c9c2008-04-18 16:13:31 -04002603 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002604 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2605 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002606
Chris Masondb945352007-10-15 16:15:53 -04002607 tree_root->nodesize = nodesize;
2608 tree_root->leafsize = leafsize;
2609 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002610 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002611
2612 sb->s_blocksize = sectorsize;
2613 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002614
Qu Wenruo3cae2102013-07-16 11:19:18 +08002615 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002616 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002617 goto fail_sb_buffer;
2618 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002619
Liu Bo8d082fb2012-04-03 09:56:53 +08002620 if (sectorsize != PAGE_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002621 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
Liu Bo8d082fb2012-04-03 09:56:53 +08002622 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002623 goto fail_sb_buffer;
2624 }
2625
Chris Mason925baed2008-06-25 16:01:30 -04002626 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002627 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002628 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002629 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002630 printk(KERN_WARNING "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002631 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002632 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002633 }
Chris Mason0b86a832008-03-24 15:01:56 -04002634
2635 blocksize = btrfs_level_size(tree_root,
2636 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002637 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002638
2639 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2640 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2641
2642 chunk_root->node = read_tree_block(chunk_root,
2643 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002644 blocksize, generation);
Josef Bacik416bc652013-04-23 14:17:42 -04002645 if (!chunk_root->node ||
2646 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002647 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002648 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002649 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002650 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002651 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2652 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002653
Chris Masone17cade2008-04-15 15:41:47 -04002654 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002655 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002656
Chris Mason0b86a832008-03-24 15:01:56 -04002657 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002658 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002659 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002660 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002661 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002662 }
Chris Mason0b86a832008-03-24 15:01:56 -04002663
Stefan Behrens8dabb742012-11-06 13:15:27 +01002664 /*
2665 * keep the device that is marked to be the target device for the
2666 * dev_replace procedure
2667 */
2668 btrfs_close_extra_devices(fs_info, fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002669
Chris Masona6b0d5c2012-02-20 20:53:43 -05002670 if (!fs_devices->latest_bdev) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002671 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002672 sb->s_id);
2673 goto fail_tree_roots;
2674 }
2675
Chris Masonaf31f5e2011-11-03 15:17:42 -04002676retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002677 blocksize = btrfs_level_size(tree_root,
2678 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002679 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002680
Chris Masone20d96d2007-03-22 12:13:20 -04002681 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002682 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002683 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002684 if (!tree_root->node ||
2685 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002686 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002687 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002688
2689 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002690 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002691
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002692 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2693 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002694 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002695
Miao Xiecb517ea2013-05-15 07:48:19 +00002696 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2697 location.type = BTRFS_ROOT_ITEM_KEY;
2698 location.offset = 0;
2699
2700 extent_root = btrfs_read_tree_root(tree_root, &location);
2701 if (IS_ERR(extent_root)) {
2702 ret = PTR_ERR(extent_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002703 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002704 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002705 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002706 fs_info->extent_root = extent_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002707
Miao Xiecb517ea2013-05-15 07:48:19 +00002708 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2709 dev_root = btrfs_read_tree_root(tree_root, &location);
2710 if (IS_ERR(dev_root)) {
2711 ret = PTR_ERR(dev_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002712 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002713 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002714 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002715 fs_info->dev_root = dev_root;
2716 btrfs_init_devices_late(fs_info);
Chris Mason3768f362007-03-13 16:47:54 -04002717
Miao Xiecb517ea2013-05-15 07:48:19 +00002718 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2719 csum_root = btrfs_read_tree_root(tree_root, &location);
2720 if (IS_ERR(csum_root)) {
2721 ret = PTR_ERR(csum_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002722 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002723 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002724 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002725 fs_info->csum_root = csum_root;
Chris Masond20f7042008-12-08 16:58:54 -05002726
Miao Xiecb517ea2013-05-15 07:48:19 +00002727 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2728 quota_root = btrfs_read_tree_root(tree_root, &location);
2729 if (!IS_ERR(quota_root)) {
Miao Xie27cdeb72014-04-02 19:51:05 +08002730 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002731 fs_info->quota_enabled = 1;
2732 fs_info->pending_quota_state = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002733 fs_info->quota_root = quota_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002734 }
2735
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002736 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2737 uuid_root = btrfs_read_tree_root(tree_root, &location);
2738 if (IS_ERR(uuid_root)) {
2739 ret = PTR_ERR(uuid_root);
2740 if (ret != -ENOENT)
2741 goto recovery_tree_root;
2742 create_uuid_tree = true;
Stefan Behrens70f80172013-08-15 17:11:23 +02002743 check_uuid_tree = false;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002744 } else {
Miao Xie27cdeb72014-04-02 19:51:05 +08002745 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002746 fs_info->uuid_root = uuid_root;
Stefan Behrens70f80172013-08-15 17:11:23 +02002747 create_uuid_tree = false;
2748 check_uuid_tree =
2749 generation != btrfs_super_uuid_tree_generation(disk_super);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002750 }
2751
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002752 fs_info->generation = generation;
2753 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002754
Ilya Dryomov68310a52012-06-22 12:24:12 -06002755 ret = btrfs_recover_balance(fs_info);
2756 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002757 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002758 goto fail_block_groups;
2759 }
2760
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002761 ret = btrfs_init_dev_stats(fs_info);
2762 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002763 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002764 ret);
2765 goto fail_block_groups;
2766 }
2767
Stefan Behrens8dabb742012-11-06 13:15:27 +01002768 ret = btrfs_init_dev_replace(fs_info);
2769 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002770 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002771 goto fail_block_groups;
2772 }
2773
2774 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2775
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002776 ret = btrfs_sysfs_add_one(fs_info);
2777 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002778 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002779 goto fail_block_groups;
2780 }
2781
liuboc59021f2011-03-07 02:13:14 +00002782 ret = btrfs_init_space_info(fs_info);
2783 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002784 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002785 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002786 }
2787
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002788 ret = btrfs_read_block_groups(extent_root);
2789 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002790 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002791 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002792 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002793 fs_info->num_tolerated_disk_barrier_failures =
2794 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002795 if (fs_info->fs_devices->missing_devices >
2796 fs_info->num_tolerated_disk_barrier_failures &&
2797 !(sb->s_flags & MS_RDONLY)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002798 printk(KERN_WARNING "BTRFS: "
2799 "too many missing devices, writeable mount is not allowed\n");
Anand Jain2365dd32014-01-22 11:15:51 +08002800 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002801 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002802
Chris Masona74a4b92008-06-25 16:01:31 -04002803 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2804 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002805 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08002806 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04002807
2808 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2809 tree_root,
2810 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002811 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002812 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002813
Chris Masonc2898112009-06-10 09:51:32 -04002814 if (!btrfs_test_opt(tree_root, SSD) &&
2815 !btrfs_test_opt(tree_root, NOSSD) &&
2816 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002817 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04002818 "mode\n");
2819 btrfs_set_opt(fs_info->mount_opt, SSD);
2820 }
2821
Qu Wenruo3818aea2014-01-13 13:36:06 +08002822 /* Set the real inode map cache flag */
2823 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2824 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2825
Stefan Behrens21adbd52011-11-09 13:44:05 +01002826#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2827 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2828 ret = btrfsic_mount(tree_root, fs_devices,
2829 btrfs_test_opt(tree_root,
2830 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2831 1 : 0,
2832 fs_info->check_integrity_print_mask);
2833 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05002834 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01002835 " integrity check module %s\n", sb->s_id);
2836 }
2837#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002838 ret = btrfs_read_qgroup_config(fs_info);
2839 if (ret)
2840 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002841
liuboacce9522011-01-06 19:30:25 +08002842 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002843 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002844 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002845
Chris Mason7c2ca462008-11-19 15:13:35 -05002846 if (fs_devices->rw_devices == 0) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002847 printk(KERN_WARNING "BTRFS: log replay required "
Chris Masond3977122009-01-05 21:25:51 -05002848 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002849 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002850 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002851 }
Chris Masone02119d2008-09-05 16:13:11 -04002852 blocksize =
2853 btrfs_level_size(tree_root,
2854 btrfs_super_log_root_level(disk_super));
2855
Al Viro6f07e422011-11-17 00:46:16 -05002856 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002857 if (!log_tree_root) {
2858 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002859 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002860 }
Chris Masone02119d2008-09-05 16:13:11 -04002861
2862 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2863 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2864
2865 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002866 blocksize,
2867 generation + 1);
Josef Bacik416bc652013-04-23 14:17:42 -04002868 if (!log_tree_root->node ||
2869 !extent_buffer_uptodate(log_tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002870 printk(KERN_ERR "BTRFS: failed to read log tree\n");
Josef Bacik416bc652013-04-23 14:17:42 -04002871 free_extent_buffer(log_tree_root->node);
2872 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002873 goto fail_qgroup;
Josef Bacik416bc652013-04-23 14:17:42 -04002874 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002875 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002876 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002877 if (ret) {
2878 btrfs_error(tree_root->fs_info, ret,
2879 "Failed to recover log tree");
2880 free_extent_buffer(log_tree_root->node);
2881 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002882 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002883 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002884
2885 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002886 ret = btrfs_commit_super(tree_root);
2887 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002888 goto fail_qgroup;
Yan Zhenge556ce22008-11-20 10:25:19 -05002889 }
Chris Masone02119d2008-09-05 16:13:11 -04002890 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002891
Yan, Zheng76dda932009-09-21 16:00:26 -04002892 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002893 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002894 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04002895
Chris Mason7c2ca462008-11-19 15:13:35 -05002896 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002897 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002898 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002899 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002900
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002901 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002902 if (ret < 0) {
2903 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05002904 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00002905 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002906 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002907 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002908 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002909
Chris Mason3de45862008-11-17 21:02:50 -05002910 location.objectid = BTRFS_FS_TREE_OBJECTID;
2911 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00002912 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05002913
Chris Mason3de45862008-11-17 21:02:50 -05002914 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002915 if (IS_ERR(fs_info->fs_root)) {
2916 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002917 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002918 }
Chris Masonc2898112009-06-10 09:51:32 -04002919
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002920 if (sb->s_flags & MS_RDONLY)
2921 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002922
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002923 down_read(&fs_info->cleanup_work_sem);
2924 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2925 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002926 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002927 close_ctree(tree_root);
2928 return ret;
2929 }
2930 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002931
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002932 ret = btrfs_resume_balance_async(fs_info);
2933 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002934 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002935 close_ctree(tree_root);
2936 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002937 }
2938
Stefan Behrens8dabb742012-11-06 13:15:27 +01002939 ret = btrfs_resume_dev_replace_async(fs_info);
2940 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002941 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01002942 close_ctree(tree_root);
2943 return ret;
2944 }
2945
Jan Schmidtb382a322013-05-28 15:47:24 +00002946 btrfs_qgroup_rescan_resume(fs_info);
2947
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002948 if (create_uuid_tree) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002949 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002950 ret = btrfs_create_uuid_tree(fs_info);
2951 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002952 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002953 ret);
2954 close_ctree(tree_root);
2955 return ret;
2956 }
Stefan Behrensf420ee12013-08-15 17:11:24 +02002957 } else if (check_uuid_tree ||
2958 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002959 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02002960 ret = btrfs_check_uuid_tree(fs_info);
2961 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002962 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02002963 ret);
2964 close_ctree(tree_root);
2965 return ret;
2966 }
2967 } else {
2968 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002969 }
2970
Al Viroad2b2c82011-11-17 01:10:02 -05002971 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002972
Arne Jansenbcef60f2011-09-13 15:23:30 +02002973fail_qgroup:
2974 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05002975fail_trans_kthread:
2976 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04002977 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002978 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04002979fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002980 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002981
2982 /*
2983 * make sure we're done with the btree inode before we stop our
2984 * kthreads
2985 */
2986 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05002987
Anand Jain2365dd32014-01-22 11:15:51 +08002988fail_sysfs:
2989 btrfs_sysfs_remove_one(fs_info);
2990
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002991fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04002992 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002993 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002994
2995fail_tree_roots:
2996 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00002997 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002998
Chris Mason39279cc2007-06-12 06:35:45 -04002999fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003000 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003001fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003002fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003003 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3004
Chris Mason4543df72008-06-11 21:47:56 -04003005 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003006fail_bio_counter:
3007 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003008fail_delalloc_bytes:
3009 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003010fail_dirty_metadata_bytes:
3011 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003012fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003013 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003014fail_srcu:
3015 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003016fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003017 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003018 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003019 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003020
3021recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003022 if (!btrfs_test_opt(tree_root, RECOVERY))
3023 goto fail_tree_roots;
3024
3025 free_root_pointers(fs_info, 0);
3026
3027 /* don't use the log in recovery mode, it won't be valid */
3028 btrfs_set_super_log_root(disk_super, 0);
3029
3030 /* we can't trust the free space cache either */
3031 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3032
3033 ret = next_root_backup(fs_info, fs_info->super_copy,
3034 &num_backups_tried, &backup_index);
3035 if (ret == -1)
3036 goto fail_block_groups;
3037 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003038}
3039
Chris Masonf2984462008-04-10 16:19:33 -04003040static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3041{
Chris Masonf2984462008-04-10 16:19:33 -04003042 if (uptodate) {
3043 set_buffer_uptodate(bh);
3044 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003045 struct btrfs_device *device = (struct btrfs_device *)
3046 bh->b_private;
3047
Frank Holtonefe120a2013-12-20 11:37:06 -05003048 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
Josef Bacik606686e2012-06-04 14:03:51 -04003049 "I/O error on %s\n",
3050 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003051 /* note, we dont' set_buffer_write_io_error because we have
3052 * our own ways of dealing with the IO errors
3053 */
Chris Masonf2984462008-04-10 16:19:33 -04003054 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003055 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003056 }
3057 unlock_buffer(bh);
3058 put_bh(bh);
3059}
3060
Yan Zhenga512bbf2008-12-08 16:46:26 -05003061struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3062{
3063 struct buffer_head *bh;
3064 struct buffer_head *latest = NULL;
3065 struct btrfs_super_block *super;
3066 int i;
3067 u64 transid = 0;
3068 u64 bytenr;
3069
3070 /* we would like to check all the supers, but that would make
3071 * a btrfs mount succeed after a mkfs from a different FS.
3072 * So, we need to add a special mount option to scan for
3073 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3074 */
3075 for (i = 0; i < 1; i++) {
3076 bytenr = btrfs_sb_offset(i);
Anand Jain8068a472013-07-26 01:29:35 +08003077 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3078 i_size_read(bdev->bd_inode))
Yan Zhenga512bbf2008-12-08 16:46:26 -05003079 break;
Anand Jain8068a472013-07-26 01:29:35 +08003080 bh = __bread(bdev, bytenr / 4096,
3081 BTRFS_SUPER_INFO_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003082 if (!bh)
3083 continue;
3084
3085 super = (struct btrfs_super_block *)bh->b_data;
3086 if (btrfs_super_bytenr(super) != bytenr ||
Qu Wenruo3cae2102013-07-16 11:19:18 +08003087 btrfs_super_magic(super) != BTRFS_MAGIC) {
Yan Zhenga512bbf2008-12-08 16:46:26 -05003088 brelse(bh);
3089 continue;
3090 }
3091
3092 if (!latest || btrfs_super_generation(super) > transid) {
3093 brelse(latest);
3094 latest = bh;
3095 transid = btrfs_super_generation(super);
3096 } else {
3097 brelse(bh);
3098 }
3099 }
3100 return latest;
3101}
3102
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003103/*
3104 * this should be called twice, once with wait == 0 and
3105 * once with wait == 1. When wait == 0 is done, all the buffer heads
3106 * we write are pinned.
3107 *
3108 * They are released when wait == 1 is done.
3109 * max_mirrors must be the same for both runs, and it indicates how
3110 * many supers on this one device should be written.
3111 *
3112 * max_mirrors == 0 means to write them all.
3113 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003114static int write_dev_supers(struct btrfs_device *device,
3115 struct btrfs_super_block *sb,
3116 int do_barriers, int wait, int max_mirrors)
3117{
3118 struct buffer_head *bh;
3119 int i;
3120 int ret;
3121 int errors = 0;
3122 u32 crc;
3123 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003124
3125 if (max_mirrors == 0)
3126 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3127
Yan Zhenga512bbf2008-12-08 16:46:26 -05003128 for (i = 0; i < max_mirrors; i++) {
3129 bytenr = btrfs_sb_offset(i);
3130 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3131 break;
3132
3133 if (wait) {
3134 bh = __find_get_block(device->bdev, bytenr / 4096,
3135 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003136 if (!bh) {
3137 errors++;
3138 continue;
3139 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003140 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003141 if (!buffer_uptodate(bh))
3142 errors++;
3143
3144 /* drop our reference */
3145 brelse(bh);
3146
3147 /* drop the reference from the wait == 0 run */
3148 brelse(bh);
3149 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003150 } else {
3151 btrfs_set_super_bytenr(sb, bytenr);
3152
3153 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003154 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003155 BTRFS_CSUM_SIZE, crc,
3156 BTRFS_SUPER_INFO_SIZE -
3157 BTRFS_CSUM_SIZE);
3158 btrfs_csum_final(crc, sb->csum);
3159
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003160 /*
3161 * one reference for us, and we leave it for the
3162 * caller
3163 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003164 bh = __getblk(device->bdev, bytenr / 4096,
3165 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003166 if (!bh) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003167 printk(KERN_ERR "BTRFS: couldn't get super "
Josef Bacik634554d2013-04-29 10:05:57 -04003168 "buffer head for bytenr %Lu\n", bytenr);
3169 errors++;
3170 continue;
3171 }
3172
Yan Zhenga512bbf2008-12-08 16:46:26 -05003173 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3174
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003175 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003176 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003177
3178 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003179 lock_buffer(bh);
3180 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003181 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003182 }
3183
Chris Mason387125f2011-11-18 15:07:51 -05003184 /*
3185 * we fua the first super. The others we allow
3186 * to go down lazy.
3187 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003188 if (i == 0)
3189 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3190 else
3191 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003192 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003193 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003194 }
3195 return errors < i ? 0 : -1;
3196}
3197
Chris Mason387125f2011-11-18 15:07:51 -05003198/*
3199 * endio for the write_dev_flush, this will wake anyone waiting
3200 * for the barrier when it is done
3201 */
3202static void btrfs_end_empty_barrier(struct bio *bio, int err)
3203{
3204 if (err) {
3205 if (err == -EOPNOTSUPP)
3206 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3207 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3208 }
3209 if (bio->bi_private)
3210 complete(bio->bi_private);
3211 bio_put(bio);
3212}
3213
3214/*
3215 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3216 * sent down. With wait == 1, it waits for the previous flush.
3217 *
3218 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3219 * capable
3220 */
3221static int write_dev_flush(struct btrfs_device *device, int wait)
3222{
3223 struct bio *bio;
3224 int ret = 0;
3225
3226 if (device->nobarriers)
3227 return 0;
3228
3229 if (wait) {
3230 bio = device->flush_bio;
3231 if (!bio)
3232 return 0;
3233
3234 wait_for_completion(&device->flush_wait);
3235
3236 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003237 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
Josef Bacik606686e2012-06-04 14:03:51 -04003238 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05003239 device->nobarriers = 1;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003240 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
Chris Mason387125f2011-11-18 15:07:51 -05003241 ret = -EIO;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003242 btrfs_dev_stat_inc_and_print(device,
3243 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003244 }
3245
3246 /* drop the reference from the wait == 0 run */
3247 bio_put(bio);
3248 device->flush_bio = NULL;
3249
3250 return ret;
3251 }
3252
3253 /*
3254 * one reference for us, and we leave it for the
3255 * caller
3256 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003257 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003258 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003259 if (!bio)
3260 return -ENOMEM;
3261
3262 bio->bi_end_io = btrfs_end_empty_barrier;
3263 bio->bi_bdev = device->bdev;
3264 init_completion(&device->flush_wait);
3265 bio->bi_private = &device->flush_wait;
3266 device->flush_bio = bio;
3267
3268 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003269 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003270
3271 return 0;
3272}
3273
3274/*
3275 * send an empty flush down to each device in parallel,
3276 * then wait for them
3277 */
3278static int barrier_all_devices(struct btrfs_fs_info *info)
3279{
3280 struct list_head *head;
3281 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003282 int errors_send = 0;
3283 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003284 int ret;
3285
3286 /* send down all the barriers */
3287 head = &info->fs_devices->devices;
3288 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003289 if (dev->missing)
3290 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003291 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003292 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003293 continue;
3294 }
3295 if (!dev->in_fs_metadata || !dev->writeable)
3296 continue;
3297
3298 ret = write_dev_flush(dev, 0);
3299 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003300 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003301 }
3302
3303 /* wait for all the barriers */
3304 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003305 if (dev->missing)
3306 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003307 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003308 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003309 continue;
3310 }
3311 if (!dev->in_fs_metadata || !dev->writeable)
3312 continue;
3313
3314 ret = write_dev_flush(dev, 1);
3315 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003316 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003317 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003318 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3319 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003320 return -EIO;
3321 return 0;
3322}
3323
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003324int btrfs_calc_num_tolerated_disk_barrier_failures(
3325 struct btrfs_fs_info *fs_info)
3326{
3327 struct btrfs_ioctl_space_info space;
3328 struct btrfs_space_info *sinfo;
3329 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3330 BTRFS_BLOCK_GROUP_SYSTEM,
3331 BTRFS_BLOCK_GROUP_METADATA,
3332 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3333 int num_types = 4;
3334 int i;
3335 int c;
3336 int num_tolerated_disk_barrier_failures =
3337 (int)fs_info->fs_devices->num_devices;
3338
3339 for (i = 0; i < num_types; i++) {
3340 struct btrfs_space_info *tmp;
3341
3342 sinfo = NULL;
3343 rcu_read_lock();
3344 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3345 if (tmp->flags == types[i]) {
3346 sinfo = tmp;
3347 break;
3348 }
3349 }
3350 rcu_read_unlock();
3351
3352 if (!sinfo)
3353 continue;
3354
3355 down_read(&sinfo->groups_sem);
3356 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3357 if (!list_empty(&sinfo->block_groups[c])) {
3358 u64 flags;
3359
3360 btrfs_get_block_group_info(
3361 &sinfo->block_groups[c], &space);
3362 if (space.total_bytes == 0 ||
3363 space.used_bytes == 0)
3364 continue;
3365 flags = space.flags;
3366 /*
3367 * return
3368 * 0: if dup, single or RAID0 is configured for
3369 * any of metadata, system or data, else
3370 * 1: if RAID5 is configured, or if RAID1 or
3371 * RAID10 is configured and only two mirrors
3372 * are used, else
3373 * 2: if RAID6 is configured, else
3374 * num_mirrors - 1: if RAID1 or RAID10 is
3375 * configured and more than
3376 * 2 mirrors are used.
3377 */
3378 if (num_tolerated_disk_barrier_failures > 0 &&
3379 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3380 BTRFS_BLOCK_GROUP_RAID0)) ||
3381 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3382 == 0)))
3383 num_tolerated_disk_barrier_failures = 0;
David Woodhouse53b381b2013-01-29 18:40:14 -05003384 else if (num_tolerated_disk_barrier_failures > 1) {
3385 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3386 BTRFS_BLOCK_GROUP_RAID5 |
3387 BTRFS_BLOCK_GROUP_RAID10)) {
3388 num_tolerated_disk_barrier_failures = 1;
3389 } else if (flags &
Henrik Nordvik15b0a892013-04-29 18:09:23 +00003390 BTRFS_BLOCK_GROUP_RAID6) {
David Woodhouse53b381b2013-01-29 18:40:14 -05003391 num_tolerated_disk_barrier_failures = 2;
3392 }
3393 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003394 }
3395 }
3396 up_read(&sinfo->groups_sem);
3397 }
3398
3399 return num_tolerated_disk_barrier_failures;
3400}
3401
Eric Sandeen48a3b632013-04-25 20:41:01 +00003402static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003403{
Chris Masone5e9a522009-06-10 15:17:02 -04003404 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003405 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003406 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003407 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003408 int ret;
3409 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003410 int max_errors;
3411 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003412 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003413
3414 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003415 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003416
David Sterba6c417612011-04-13 15:41:04 +02003417 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003418 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003419
Chris Mason174ba502011-05-27 10:03:58 -04003420 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003421 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003422 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003423
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003424 if (do_barriers) {
3425 ret = barrier_all_devices(root->fs_info);
3426 if (ret) {
3427 mutex_unlock(
3428 &root->fs_info->fs_devices->device_list_mutex);
3429 btrfs_error(root->fs_info, ret,
3430 "errors while submitting device barriers.");
3431 return ret;
3432 }
3433 }
Chris Mason387125f2011-11-18 15:07:51 -05003434
Xiao Guangrong1f781602011-04-20 10:09:16 +00003435 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003436 if (!dev->bdev) {
3437 total_errors++;
3438 continue;
3439 }
Yan Zheng2b820322008-11-17 21:11:30 -05003440 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003441 continue;
3442
Yan Zheng2b820322008-11-17 21:11:30 -05003443 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003444 btrfs_set_stack_device_type(dev_item, dev->type);
3445 btrfs_set_stack_device_id(dev_item, dev->devid);
3446 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3447 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3448 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3449 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3450 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3451 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003452 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003453
Chris Masona061fc82008-05-07 11:43:44 -04003454 flags = btrfs_super_flags(sb);
3455 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003456
Yan Zhenga512bbf2008-12-08 16:46:26 -05003457 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003458 if (ret)
3459 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003460 }
Chris Masona236aed2008-04-29 09:38:00 -04003461 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003462 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003463 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003464 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003465
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003466 /* FUA is masked off if unsupported and can't be the reason */
3467 btrfs_error(root->fs_info, -EIO,
3468 "%d errors while writing supers", total_errors);
3469 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003470 }
Chris Masonf2984462008-04-10 16:19:33 -04003471
Yan Zhenga512bbf2008-12-08 16:46:26 -05003472 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003473 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003474 if (!dev->bdev)
3475 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003476 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003477 continue;
3478
Yan Zhenga512bbf2008-12-08 16:46:26 -05003479 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3480 if (ret)
3481 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003482 }
Chris Mason174ba502011-05-27 10:03:58 -04003483 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003484 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003485 btrfs_error(root->fs_info, -EIO,
3486 "%d errors while writing supers", total_errors);
3487 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003488 }
Chris Masonf2984462008-04-10 16:19:33 -04003489 return 0;
3490}
3491
Yan Zhenga512bbf2008-12-08 16:46:26 -05003492int write_ctree_super(struct btrfs_trans_handle *trans,
3493 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003494{
Rashikaf570e7572013-10-31 16:50:42 +05303495 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003496}
3497
Miao Xiecb517ea2013-05-15 07:48:19 +00003498/* Drop a fs root from the radix tree and free it. */
3499void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3500 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003501{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003502 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003503 radix_tree_delete(&fs_info->fs_roots_radix,
3504 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003505 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003506
3507 if (btrfs_root_refs(&root->root_item) == 0)
3508 synchronize_srcu(&fs_info->subvol_srcu);
3509
Liu Bo1a4319c2014-01-13 19:53:53 +08003510 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003511 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003512
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003513 if (root->free_ino_pinned)
3514 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3515 if (root->free_ino_ctl)
3516 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003517 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003518}
3519
3520static void free_fs_root(struct btrfs_root *root)
3521{
Li Zefan82d59022011-04-20 10:33:24 +08003522 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003523 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003524 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3525 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003526 if (root->anon_dev)
3527 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003528 if (root->subv_writers)
3529 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003530 free_extent_buffer(root->node);
3531 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003532 kfree(root->free_ino_ctl);
3533 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003534 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003535 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003536}
3537
Miao Xiecb517ea2013-05-15 07:48:19 +00003538void btrfs_free_fs_root(struct btrfs_root *root)
3539{
3540 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003541}
3542
Yan Zhengc146afa2008-11-12 14:34:12 -05003543int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3544{
3545 u64 root_objectid = 0;
3546 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003547 int i = 0;
3548 int err = 0;
3549 unsigned int ret = 0;
3550 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003551
3552 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003553 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003554 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3555 (void **)gang, root_objectid,
3556 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003557 if (!ret) {
3558 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003559 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003560 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003561 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003562
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003563 for (i = 0; i < ret; i++) {
3564 /* Avoid to grab roots in dead_roots */
3565 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3566 gang[i] = NULL;
3567 continue;
3568 }
3569 /* grab all the search result for later use */
3570 gang[i] = btrfs_grab_fs_root(gang[i]);
3571 }
3572 srcu_read_unlock(&fs_info->subvol_srcu, index);
3573
3574 for (i = 0; i < ret; i++) {
3575 if (!gang[i])
3576 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003577 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003578 err = btrfs_orphan_cleanup(gang[i]);
3579 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003580 break;
3581 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003582 }
3583 root_objectid++;
3584 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003585
3586 /* release the uncleaned roots due to error */
3587 for (; i < ret; i++) {
3588 if (gang[i])
3589 btrfs_put_fs_root(gang[i]);
3590 }
3591 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003592}
3593
3594int btrfs_commit_super(struct btrfs_root *root)
3595{
3596 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003597
3598 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00003599 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003600 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003601 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003602
3603 /* wait until ongoing cleanup work done */
3604 down_write(&root->fs_info->cleanup_work_sem);
3605 up_write(&root->fs_info->cleanup_work_sem);
3606
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003607 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003608 if (IS_ERR(trans))
3609 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003610 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003611}
3612
Chris Masone20d96d2007-03-22 12:13:20 -04003613int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003614{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003615 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003616 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003617
Chris Masonfacda1e2007-06-08 18:11:48 -04003618 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003619 smp_mb();
3620
Stefan Behrens803b2f52013-08-15 17:11:21 +02003621 /* wait for the uuid_scan task to finish */
3622 down(&fs_info->uuid_tree_rescan_sem);
3623 /* avoid complains from lockdep et al., set sem back to initial state */
3624 up(&fs_info->uuid_tree_rescan_sem);
3625
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003626 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003627 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003628
Stefan Behrens8dabb742012-11-06 13:15:27 +01003629 btrfs_dev_replace_suspend_for_unmount(fs_info);
3630
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003631 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003632
3633 /* wait for any defraggers to finish */
3634 wait_event(fs_info->transaction_wait,
3635 (atomic_read(&fs_info->defrag_running) == 0));
3636
3637 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003638 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003639
Miao Xie21c7e752014-05-13 17:29:04 -07003640 cancel_work_sync(&fs_info->async_reclaim_work);
3641
Yan Zhengc146afa2008-11-12 14:34:12 -05003642 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003643 ret = btrfs_commit_super(root);
3644 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003645 btrfs_err(root->fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003646 }
3647
Miao Xie87533c42013-01-29 10:14:48 +00003648 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003649 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003650
Al Viroe3029d92011-11-17 00:56:18 -05003651 kthread_stop(fs_info->transaction_kthread);
3652 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003653
Yan Zhengf25784b2009-07-28 08:41:57 -04003654 fs_info->closing = 2;
3655 smp_mb();
3656
Arne Jansenbcef60f2011-09-13 15:23:30 +02003657 btrfs_free_qgroup_config(root->fs_info);
3658
Miao Xie963d6782013-01-29 10:10:51 +00003659 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003660 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003661 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003662 }
Yanbcc63ab2008-07-30 16:29:20 -04003663
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003664 btrfs_sysfs_remove_one(fs_info);
3665
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003666 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003667
Liu Bo1a4319c2014-01-13 19:53:53 +08003668 btrfs_put_block_group_cache(fs_info);
3669
Josef Bacik2b1360d2013-10-07 15:14:44 -04003670 btrfs_free_block_groups(fs_info);
3671
Wang Shilongde348ee2014-04-09 19:23:22 +08003672 /*
3673 * we must make sure there is not any read request to
3674 * submit after we stopping all workers.
3675 */
3676 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003677 btrfs_stop_all_workers(fs_info);
3678
Josef Bacik13e6c372013-05-30 16:55:44 -04003679 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003680
Josef Bacik13e6c372013-05-30 16:55:44 -04003681 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003682
Stefan Behrens21adbd52011-11-09 13:44:05 +01003683#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3684 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3685 btrfsic_unmount(root, fs_info->fs_devices);
3686#endif
3687
Chris Masondfe25022008-05-13 13:46:40 -04003688 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003689 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003690
Miao Xiee2d84522013-01-29 10:09:20 +00003691 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003692 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003693 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003694 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003695 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003696
David Woodhouse53b381b2013-01-29 18:40:14 -05003697 btrfs_free_stripe_hash_table(fs_info);
3698
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003699 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3700 root->orphan_block_rsv = NULL;
3701
Chris Masoneb60cea2007-02-02 09:18:22 -05003702 return 0;
3703}
3704
Chris Masonb9fab912012-05-06 07:23:47 -04003705int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3706 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003707{
Chris Mason1259ab72008-05-12 13:39:03 -04003708 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003709 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003710
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003711 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003712 if (!ret)
3713 return ret;
3714
3715 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003716 parent_transid, atomic);
3717 if (ret == -EAGAIN)
3718 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003719 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003720}
Chris Mason6702ed42007-08-07 16:15:09 -04003721
Chris Mason5f39d392007-10-15 16:14:19 -04003722int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3723{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003724 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003725}
3726
3727void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3728{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003729 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003730 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003731 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003732
Josef Bacik06ea65a2013-09-19 16:07:01 -04003733#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3734 /*
3735 * This is a fast path so only do this check if we have sanity tests
3736 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3737 * outside of the sanity tests.
3738 */
3739 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3740 return;
3741#endif
3742 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003743 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003744 if (transid != root->fs_info->generation)
3745 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003746 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003747 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003748 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003749 if (!was_dirty)
3750 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3751 buf->len,
3752 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003753#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3754 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3755 btrfs_print_leaf(root, buf);
3756 ASSERT(0);
3757 }
3758#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05003759}
3760
Liu Bob53d3f52012-11-14 14:34:34 +00003761static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3762 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003763{
Chris Mason188de642008-05-09 11:52:25 -04003764 /*
3765 * looks as though older kernels can get into trouble with
3766 * this code, they end up stuck in balance_dirty_pages forever
3767 */
Miao Xiee2d84522013-01-29 10:09:20 +00003768 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04003769
Jens Axboe6933c022009-03-17 09:36:37 +01003770 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003771 return;
3772
Liu Bob53d3f52012-11-14 14:34:34 +00003773 if (flush_delayed)
3774 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003775
Miao Xiee2d84522013-01-29 10:09:20 +00003776 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3777 BTRFS_DIRTY_METADATA_THRESH);
3778 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08003779 balance_dirty_pages_ratelimited(
3780 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08003781 }
3782 return;
3783}
3784
Liu Bob53d3f52012-11-14 14:34:34 +00003785void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08003786{
Liu Bob53d3f52012-11-14 14:34:34 +00003787 __btrfs_btree_balance_dirty(root, 1);
3788}
Miao Xie16cdcec2011-04-22 18:12:22 +08003789
Liu Bob53d3f52012-11-14 14:34:34 +00003790void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3791{
3792 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04003793}
Chris Mason6b800532007-10-15 16:17:34 -04003794
Chris Masonca7a79a2008-05-12 12:59:19 -04003795int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003796{
Chris Mason727011e2010-08-06 13:21:20 -04003797 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003798 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003799}
Chris Mason0da54682007-11-07 21:08:01 -05003800
David Sterbafcd1f0652012-03-06 00:06:18 +01003801static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003802 int read_only)
3803{
David Sterba1104a882013-03-06 15:57:46 +01003804 /*
3805 * Placeholder for checks
3806 */
David Sterbafcd1f0652012-03-06 00:06:18 +01003807 return 0;
liuboacce9522011-01-06 19:30:25 +08003808}
3809
Eric Sandeen48a3b632013-04-25 20:41:01 +00003810static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003811{
liuboacce9522011-01-06 19:30:25 +08003812 mutex_lock(&root->fs_info->cleaner_mutex);
3813 btrfs_run_delayed_iputs(root);
3814 mutex_unlock(&root->fs_info->cleaner_mutex);
3815
3816 down_write(&root->fs_info->cleanup_work_sem);
3817 up_write(&root->fs_info->cleanup_work_sem);
3818
3819 /* cleanup FS via transaction */
3820 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003821}
3822
Josef Bacik569e0f32013-02-13 11:09:14 -05003823static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3824 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003825{
3826 struct btrfs_inode *btrfs_inode;
3827 struct list_head splice;
3828
3829 INIT_LIST_HEAD(&splice);
3830
3831 mutex_lock(&root->fs_info->ordered_operations_mutex);
Miao Xie199c2a92013-05-15 07:48:23 +00003832 spin_lock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003833
Josef Bacik569e0f32013-02-13 11:09:14 -05003834 list_splice_init(&t->ordered_operations, &splice);
liuboacce9522011-01-06 19:30:25 +08003835 while (!list_empty(&splice)) {
3836 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3837 ordered_operations);
3838
3839 list_del_init(&btrfs_inode->ordered_operations);
Miao Xie199c2a92013-05-15 07:48:23 +00003840 spin_unlock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003841
3842 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00003843
Miao Xie199c2a92013-05-15 07:48:23 +00003844 spin_lock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003845 }
3846
Miao Xie199c2a92013-05-15 07:48:23 +00003847 spin_unlock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003848 mutex_unlock(&root->fs_info->ordered_operations_mutex);
liuboacce9522011-01-06 19:30:25 +08003849}
3850
Jeff Mahoney143bede2012-03-01 14:56:26 +01003851static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003852{
liuboacce9522011-01-06 19:30:25 +08003853 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08003854
Miao Xie199c2a92013-05-15 07:48:23 +00003855 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05003856 /*
3857 * This will just short circuit the ordered completion stuff which will
3858 * make sure the ordered extent gets properly cleaned up.
3859 */
Miao Xie199c2a92013-05-15 07:48:23 +00003860 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05003861 root_extent_list)
3862 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00003863 spin_unlock(&root->ordered_extent_lock);
3864}
3865
3866static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3867{
3868 struct btrfs_root *root;
3869 struct list_head splice;
3870
3871 INIT_LIST_HEAD(&splice);
3872
3873 spin_lock(&fs_info->ordered_root_lock);
3874 list_splice_init(&fs_info->ordered_roots, &splice);
3875 while (!list_empty(&splice)) {
3876 root = list_first_entry(&splice, struct btrfs_root,
3877 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04003878 list_move_tail(&root->ordered_root,
3879 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00003880
Liu Bo2a85d9c2014-02-10 17:07:16 +08003881 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003882 btrfs_destroy_ordered_extents(root);
3883
Liu Bo2a85d9c2014-02-10 17:07:16 +08003884 cond_resched();
3885 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003886 }
3887 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003888}
3889
Stefan Behrens35a36212013-08-14 18:12:25 +02003890static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3891 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003892{
3893 struct rb_node *node;
3894 struct btrfs_delayed_ref_root *delayed_refs;
3895 struct btrfs_delayed_ref_node *ref;
3896 int ret = 0;
3897
3898 delayed_refs = &trans->delayed_refs;
3899
3900 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003901 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003902 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05003903 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08003904 return ret;
3905 }
3906
Josef Bacikd7df2c72014-01-23 09:21:38 -05003907 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3908 struct btrfs_delayed_ref_head *head;
Josef Bacike78417d2013-06-03 16:42:36 -04003909 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08003910
Josef Bacikd7df2c72014-01-23 09:21:38 -05003911 head = rb_entry(node, struct btrfs_delayed_ref_head,
3912 href_node);
3913 if (!mutex_trylock(&head->mutex)) {
3914 atomic_inc(&head->node.refs);
3915 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003916
Josef Bacikd7df2c72014-01-23 09:21:38 -05003917 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04003918 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003919 btrfs_put_delayed_ref(&head->node);
3920 spin_lock(&delayed_refs->lock);
3921 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04003922 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05003923 spin_lock(&head->lock);
3924 while ((node = rb_first(&head->ref_root)) != NULL) {
3925 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3926 rb_node);
3927 ref->in_tree = 0;
3928 rb_erase(&ref->rb_node, &head->ref_root);
3929 atomic_dec(&delayed_refs->num_entries);
3930 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003931 }
3932 if (head->must_insert_reserved)
3933 pin_bytes = true;
3934 btrfs_free_delayed_extent_op(head->extent_op);
3935 delayed_refs->num_heads--;
3936 if (head->processing == 0)
3937 delayed_refs->num_heads_ready--;
3938 atomic_dec(&delayed_refs->num_entries);
3939 head->node.in_tree = 0;
3940 rb_erase(&head->href_node, &delayed_refs->href_root);
3941 spin_unlock(&head->lock);
3942 spin_unlock(&delayed_refs->lock);
3943 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08003944
Josef Bacikd7df2c72014-01-23 09:21:38 -05003945 if (pin_bytes)
3946 btrfs_pin_extent(root, head->node.bytenr,
3947 head->node.num_bytes, 1);
3948 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08003949 cond_resched();
3950 spin_lock(&delayed_refs->lock);
3951 }
3952
3953 spin_unlock(&delayed_refs->lock);
3954
3955 return ret;
3956}
3957
Jeff Mahoney143bede2012-03-01 14:56:26 +01003958static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003959{
3960 struct btrfs_inode *btrfs_inode;
3961 struct list_head splice;
3962
3963 INIT_LIST_HEAD(&splice);
3964
Miao Xieeb73c1b2013-05-15 07:48:22 +00003965 spin_lock(&root->delalloc_lock);
3966 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003967
3968 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00003969 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3970 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08003971
3972 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00003973 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3974 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00003975 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003976
3977 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00003978
Miao Xieeb73c1b2013-05-15 07:48:22 +00003979 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003980 }
3981
Miao Xieeb73c1b2013-05-15 07:48:22 +00003982 spin_unlock(&root->delalloc_lock);
3983}
3984
3985static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3986{
3987 struct btrfs_root *root;
3988 struct list_head splice;
3989
3990 INIT_LIST_HEAD(&splice);
3991
3992 spin_lock(&fs_info->delalloc_root_lock);
3993 list_splice_init(&fs_info->delalloc_roots, &splice);
3994 while (!list_empty(&splice)) {
3995 root = list_first_entry(&splice, struct btrfs_root,
3996 delalloc_root);
3997 list_del_init(&root->delalloc_root);
3998 root = btrfs_grab_fs_root(root);
3999 BUG_ON(!root);
4000 spin_unlock(&fs_info->delalloc_root_lock);
4001
4002 btrfs_destroy_delalloc_inodes(root);
4003 btrfs_put_fs_root(root);
4004
4005 spin_lock(&fs_info->delalloc_root_lock);
4006 }
4007 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004008}
4009
4010static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4011 struct extent_io_tree *dirty_pages,
4012 int mark)
4013{
4014 int ret;
liuboacce9522011-01-06 19:30:25 +08004015 struct extent_buffer *eb;
4016 u64 start = 0;
4017 u64 end;
liuboacce9522011-01-06 19:30:25 +08004018
4019 while (1) {
4020 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004021 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004022 if (ret)
4023 break;
4024
4025 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4026 while (start <= end) {
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004027 eb = btrfs_find_tree_block(root, start,
4028 root->leafsize);
Josef Bacik69a85bd2013-05-08 13:30:11 -04004029 start += root->leafsize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004030 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004031 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004032 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004033
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004034 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4035 &eb->bflags))
4036 clear_extent_buffer_dirty(eb);
4037 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004038 }
4039 }
4040
4041 return ret;
4042}
4043
4044static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4045 struct extent_io_tree *pinned_extents)
4046{
4047 struct extent_io_tree *unpin;
4048 u64 start;
4049 u64 end;
4050 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004051 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004052
4053 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004054again:
liuboacce9522011-01-06 19:30:25 +08004055 while (1) {
4056 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004057 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004058 if (ret)
4059 break;
4060
4061 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00004062 if (btrfs_test_opt(root, DISCARD))
4063 ret = btrfs_error_discard_extent(root, start,
4064 end + 1 - start,
4065 NULL);
liuboacce9522011-01-06 19:30:25 +08004066
4067 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4068 btrfs_error_unpin_extent_range(root, start, end);
4069 cond_resched();
4070 }
4071
Liu Boed0eaa12012-06-14 02:23:21 -06004072 if (loop) {
4073 if (unpin == &root->fs_info->freed_extents[0])
4074 unpin = &root->fs_info->freed_extents[1];
4075 else
4076 unpin = &root->fs_info->freed_extents[0];
4077 loop = false;
4078 goto again;
4079 }
4080
liuboacce9522011-01-06 19:30:25 +08004081 return 0;
4082}
4083
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004084void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4085 struct btrfs_root *root)
4086{
Josef Bacik724e2312013-09-30 11:36:38 -04004087 btrfs_destroy_ordered_operations(cur_trans, root);
4088
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004089 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004090
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004091 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004092 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004093
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004094 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004095 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004096
Miao Xie67cde342012-06-14 02:23:22 -06004097 btrfs_destroy_delayed_inodes(root);
4098 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004099
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004100 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4101 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004102 btrfs_destroy_pinned_extent(root,
4103 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004104
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004105 cur_trans->state =TRANS_STATE_COMPLETED;
4106 wake_up(&cur_trans->commit_wait);
4107
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004108 /*
4109 memset(cur_trans, 0, sizeof(*cur_trans));
4110 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4111 */
4112}
4113
Eric Sandeen48a3b632013-04-25 20:41:01 +00004114static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004115{
4116 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004117
liuboacce9522011-01-06 19:30:25 +08004118 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4119
Josef Bacika4abeea2011-04-11 17:25:13 -04004120 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004121 while (!list_empty(&root->fs_info->trans_list)) {
4122 t = list_first_entry(&root->fs_info->trans_list,
4123 struct btrfs_transaction, list);
4124 if (t->state >= TRANS_STATE_COMMIT_START) {
4125 atomic_inc(&t->use_count);
4126 spin_unlock(&root->fs_info->trans_lock);
4127 btrfs_wait_for_commit(root, t->transid);
4128 btrfs_put_transaction(t);
4129 spin_lock(&root->fs_info->trans_lock);
4130 continue;
4131 }
4132 if (t == root->fs_info->running_transaction) {
4133 t->state = TRANS_STATE_COMMIT_DOING;
4134 spin_unlock(&root->fs_info->trans_lock);
4135 /*
4136 * We wait for 0 num_writers since we don't hold a trans
4137 * handle open currently for this transaction.
4138 */
4139 wait_event(t->writer_wait,
4140 atomic_read(&t->num_writers) == 0);
4141 } else {
4142 spin_unlock(&root->fs_info->trans_lock);
4143 }
4144 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004145
Josef Bacik724e2312013-09-30 11:36:38 -04004146 spin_lock(&root->fs_info->trans_lock);
4147 if (t == root->fs_info->running_transaction)
4148 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004149 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004150 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004151
Josef Bacik724e2312013-09-30 11:36:38 -04004152 btrfs_put_transaction(t);
4153 trace_btrfs_transaction_commit(root);
4154 spin_lock(&root->fs_info->trans_lock);
4155 }
4156 spin_unlock(&root->fs_info->trans_lock);
4157 btrfs_destroy_all_ordered_extents(root->fs_info);
4158 btrfs_destroy_delayed_inodes(root);
4159 btrfs_assert_delayed_root_empty(root);
4160 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4161 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004162 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004163
4164 return 0;
4165}
4166
Chris Masond1310b22008-01-24 16:13:08 -05004167static struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004168 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004169 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004170 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004171 /* note we're sharing with inode.c for the merge bio hook */
4172 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004173};