blob: a8f652dc940bd85148dad48c11d2d893aefe32ba [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
David Woodhouse163e7832009-04-19 13:02:41 +010029#include <linux/crc32c.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090030#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050031#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020032#include <linux/ratelimit.h>
David Sterba7e75bf32011-03-18 22:56:43 +000033#include <asm/unaligned.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050034#include "compat.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Chris Masone089f052007-03-16 16:20:31 -040037#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040038#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040039#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040040#include "print-tree.h"
Chris Mason8b712842008-06-11 16:50:36 -040041#include "async-thread.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080045#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010046#include "check-integrity.h"
Josef Bacik606686e2012-06-04 14:03:51 -040047#include "rcu-string.h"
Stefan Behrens8dabb742012-11-06 13:15:27 +010048#include "dev-replace.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050049
Josef Bacikde0022b2012-09-25 14:25:58 -040050#ifdef CONFIG_X86
51#include <asm/cpufeature.h>
52#endif
53
Chris Masond1310b22008-01-24 16:13:08 -050054static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040055static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040056static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010057static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080058 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010059static void btrfs_destroy_ordered_operations(struct btrfs_root *root);
60static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080061static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
62 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010063static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
64static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080065static int btrfs_destroy_marked_extents(struct btrfs_root *root,
66 struct extent_io_tree *dirty_pages,
67 int mark);
68static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
69 struct extent_io_tree *pinned_extents);
Chris Masonce9adaa2008-04-09 16:28:12 -040070
Chris Masond352ac62008-09-29 15:18:18 -040071/*
72 * end_io_wq structs are used to do processing in task context when an IO is
73 * complete. This is used during reads to verify checksums, and it is used
74 * by writes to insert metadata for new file extents after IO is complete.
75 */
Chris Masonce9adaa2008-04-09 16:28:12 -040076struct end_io_wq {
77 struct bio *bio;
78 bio_end_io_t *end_io;
79 void *private;
80 struct btrfs_fs_info *info;
81 int error;
Chris Mason22c59942008-04-09 16:28:12 -040082 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040083 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040084 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040085};
Chris Mason0da54682007-11-07 21:08:01 -050086
Chris Masond352ac62008-09-29 15:18:18 -040087/*
88 * async submit bios are used to offload expensive checksumming
89 * onto the worker threads. They checksum file and metadata bios
90 * just before they are sent down the IO stack.
91 */
Chris Mason44b8bd72008-04-16 11:14:51 -040092struct async_submit_bio {
93 struct inode *inode;
94 struct bio *bio;
95 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -050096 extent_submit_bio_hook_t *submit_bio_start;
97 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -040098 int rw;
99 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400100 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400101 /*
102 * bio_offset is optional, can be used if the pages in the bio
103 * can't tell us where in the file the bio should go
104 */
105 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400106 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100107 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400108};
109
Chris Mason85d4e462011-07-26 16:11:19 -0400110/*
111 * Lockdep class keys for extent_buffer->lock's in this root. For a given
112 * eb, the lockdep key is determined by the btrfs_root it belongs to and
113 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500114 *
Chris Mason85d4e462011-07-26 16:11:19 -0400115 * Different roots are used for different purposes and may nest inside each
116 * other and they require separate keysets. As lockdep keys should be
117 * static, assign keysets according to the purpose of the root as indicated
118 * by btrfs_root->objectid. This ensures that all special purpose roots
119 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500120 *
Chris Mason85d4e462011-07-26 16:11:19 -0400121 * Lock-nesting across peer nodes is always done with the immediate parent
122 * node locked thus preventing deadlock. As lockdep doesn't know this, use
123 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500124 *
Chris Mason85d4e462011-07-26 16:11:19 -0400125 * The key is set by the readpage_end_io_hook after the buffer has passed
126 * csum validation but before the pages are unlocked. It is also set by
127 * btrfs_init_new_buffer on freshly allocated blocks.
128 *
129 * We also add a check to make sure the highest level of the tree is the
130 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
131 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500132 */
133#ifdef CONFIG_DEBUG_LOCK_ALLOC
134# if BTRFS_MAX_LEVEL != 8
135# error
136# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400137
138static struct btrfs_lockdep_keyset {
139 u64 id; /* root objectid */
140 const char *name_stem; /* lock name stem */
141 char names[BTRFS_MAX_LEVEL + 1][20];
142 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
143} btrfs_lockdep_keysets[] = {
144 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
145 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
146 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
147 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
148 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
149 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
150 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
151 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
152 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
153 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
154 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500155};
Chris Mason85d4e462011-07-26 16:11:19 -0400156
157void __init btrfs_init_lockdep(void)
158{
159 int i, j;
160
161 /* initialize lockdep class names */
162 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
163 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
164
165 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
166 snprintf(ks->names[j], sizeof(ks->names[j]),
167 "btrfs-%s-%02d", ks->name_stem, j);
168 }
169}
170
171void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
172 int level)
173{
174 struct btrfs_lockdep_keyset *ks;
175
176 BUG_ON(level >= ARRAY_SIZE(ks->keys));
177
178 /* find the matching keyset, id 0 is the default entry */
179 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
180 if (ks->id == objectid)
181 break;
182
183 lockdep_set_class_and_name(&eb->lock,
184 &ks->keys[level], ks->names[level]);
185}
186
Chris Mason4008c042009-02-12 14:09:45 -0500187#endif
188
Chris Masond352ac62008-09-29 15:18:18 -0400189/*
190 * extents on the btree inode are pretty simple, there's one extent
191 * that covers the entire device
192 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500193static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200194 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500195 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400196{
197 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
198 struct extent_map *em;
199 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400200
Chris Mason890871b2009-09-02 16:24:52 -0400201 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500202 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400203 if (em) {
204 em->bdev =
205 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400206 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400207 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400208 }
Chris Mason890871b2009-09-02 16:24:52 -0400209 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400210
David Sterba172ddd62011-04-21 00:48:27 +0200211 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400212 if (!em) {
213 em = ERR_PTR(-ENOMEM);
214 goto out;
215 }
216 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400217 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400218 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400219 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400220 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500221
Chris Mason890871b2009-09-02 16:24:52 -0400222 write_lock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400223 ret = add_extent_mapping(em_tree, em);
224 if (ret == -EEXIST) {
225 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400226 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600227 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600228 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400229 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400230 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600231 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400232 }
Chris Mason890871b2009-09-02 16:24:52 -0400233 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400234
Chris Mason5f39d392007-10-15 16:14:19 -0400235out:
236 return em;
237}
238
Chris Mason19c00dd2007-10-15 16:19:22 -0400239u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
240{
David Woodhouse163e7832009-04-19 13:02:41 +0100241 return crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400242}
243
244void btrfs_csum_final(u32 crc, char *result)
245{
David Sterba7e75bf32011-03-18 22:56:43 +0000246 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400247}
248
Chris Masond352ac62008-09-29 15:18:18 -0400249/*
250 * compute the csum for a btree block, and either verify it or write it
251 * into the csum field of the block.
252 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400253static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
254 int verify)
255{
David Sterba6c417612011-04-13 15:41:04 +0200256 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500257 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400258 unsigned long len;
259 unsigned long cur_len;
260 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400261 char *kaddr;
262 unsigned long map_start;
263 unsigned long map_len;
264 int err;
265 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500266 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400267
268 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500269 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400270 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400271 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500272 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400273 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400274 cur_len = min(len, map_len - (offset - map_start));
275 crc = btrfs_csum_data(root, kaddr + offset - map_start,
276 crc, cur_len);
277 len -= cur_len;
278 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400279 }
Josef Bacik607d4322008-12-02 07:17:45 -0500280 if (csum_size > sizeof(inline_result)) {
281 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
282 if (!result)
283 return 1;
284 } else {
285 result = (char *)&inline_result;
286 }
287
Chris Mason19c00dd2007-10-15 16:19:22 -0400288 btrfs_csum_final(crc, result);
289
290 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500291 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500292 u32 val;
293 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500294 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500295
Josef Bacik607d4322008-12-02 07:17:45 -0500296 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba7a36dde2011-05-06 15:33:15 +0200297 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
Chris Mason193f2842009-04-27 07:29:05 -0400298 "failed on %llu wanted %X found %X "
299 "level %d\n",
300 root->fs_info->sb->s_id,
301 (unsigned long long)buf->start, val, found,
302 btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500303 if (result != (char *)&inline_result)
304 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400305 return 1;
306 }
307 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500308 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400309 }
Josef Bacik607d4322008-12-02 07:17:45 -0500310 if (result != (char *)&inline_result)
311 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400312 return 0;
313}
314
Chris Masond352ac62008-09-29 15:18:18 -0400315/*
316 * we can't consider a given block up to date unless the transid of the
317 * block matches the transid in the parent node's pointer. This is how we
318 * detect blocks that either didn't get written at all or got written
319 * in the wrong place.
320 */
Chris Mason1259ab72008-05-12 13:39:03 -0400321static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400322 struct extent_buffer *eb, u64 parent_transid,
323 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400324{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000325 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400326 int ret;
327
328 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
329 return 0;
330
Chris Masonb9fab912012-05-06 07:23:47 -0400331 if (atomic)
332 return -EAGAIN;
333
Josef Bacik2ac55d42010-02-03 19:33:23 +0000334 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100335 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400336 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400337 btrfs_header_generation(eb) == parent_transid) {
338 ret = 0;
339 goto out;
340 }
David Sterba7a36dde2011-05-06 15:33:15 +0200341 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
Chris Mason193f2842009-04-27 07:29:05 -0400342 "found %llu\n",
343 (unsigned long long)eb->start,
344 (unsigned long long)parent_transid,
345 (unsigned long long)btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400346 ret = 1;
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400347 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400348out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000349 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
350 &cached_state, GFP_NOFS);
Chris Mason1259ab72008-05-12 13:39:03 -0400351 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400352}
353
Chris Masond352ac62008-09-29 15:18:18 -0400354/*
355 * helper to read a given tree block, doing retries as required when
356 * the checksums don't match and we have alternate mirrors to try.
357 */
Chris Masonf1885912008-04-09 16:28:12 -0400358static int btree_read_extent_buffer_pages(struct btrfs_root *root,
359 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400360 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400361{
362 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400363 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400364 int ret;
365 int num_copies = 0;
366 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400367 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400368
Josef Bacika826d6d2011-03-16 13:42:43 -0400369 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400370 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
371 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200372 ret = read_extent_buffer_pages(io_tree, eb, start,
373 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400374 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600375 if (!ret) {
376 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400377 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600378 break;
379 else
380 ret = -EIO;
381 }
Chris Masond3977122009-01-05 21:25:51 -0500382
Josef Bacika826d6d2011-03-16 13:42:43 -0400383 /*
384 * This buffer's crc is fine, but its contents are corrupted, so
385 * there is no reason to read the other copies, they won't be
386 * any less wrong.
387 */
388 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400389 break;
390
Stefan Behrens5d964052012-11-05 14:59:07 +0100391 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400392 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400393 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400394 break;
Chris Mason42352982008-04-28 16:40:52 -0400395
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400396 if (!failed_mirror) {
397 failed = 1;
398 failed_mirror = eb->read_mirror;
399 }
400
Chris Masonf1885912008-04-09 16:28:12 -0400401 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400402 if (mirror_num == failed_mirror)
403 mirror_num++;
404
Chris Mason42352982008-04-28 16:40:52 -0400405 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400406 break;
Chris Masonf1885912008-04-09 16:28:12 -0400407 }
Josef Bacikea466792012-03-26 21:57:36 -0400408
Stefan Behrensc0901582012-07-10 07:30:17 -0600409 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400410 repair_eb_io_failure(root, eb, failed_mirror);
411
412 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400413}
Chris Mason19c00dd2007-10-15 16:19:22 -0400414
Chris Masond352ac62008-09-29 15:18:18 -0400415/*
Chris Masond3977122009-01-05 21:25:51 -0500416 * checksum a dirty tree block before IO. This has extra checks to make sure
417 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400418 */
Chris Masond3977122009-01-05 21:25:51 -0500419
Christoph Hellwigb2950862008-12-02 09:54:17 -0500420static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400421{
Chris Masond1310b22008-01-24 16:13:08 -0500422 struct extent_io_tree *tree;
Chris Mason35ebb932007-10-30 16:56:53 -0400423 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
Chris Mason19c00dd2007-10-15 16:19:22 -0400424 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400425 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400426
Chris Masond1310b22008-01-24 16:13:08 -0500427 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason19c00dd2007-10-15 16:19:22 -0400428
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500429 eb = (struct extent_buffer *)page->private;
430 if (page != eb->pages[0])
431 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400432 found_start = btrfs_header_bytenr(eb);
433 if (found_start != start) {
Chris Mason55c69072008-01-09 15:55:33 -0500434 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500435 return 0;
Chris Mason55c69072008-01-09 15:55:33 -0500436 }
Chris Mason55c69072008-01-09 15:55:33 -0500437 if (!PageUptodate(page)) {
Chris Mason55c69072008-01-09 15:55:33 -0500438 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500439 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400440 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400441 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400442 return 0;
443}
444
Yan Zheng2b820322008-11-17 21:11:30 -0500445static int check_tree_block_fsid(struct btrfs_root *root,
446 struct extent_buffer *eb)
447{
448 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
449 u8 fsid[BTRFS_UUID_SIZE];
450 int ret = 1;
451
452 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
453 BTRFS_FSID_SIZE);
454 while (fs_devices) {
455 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
456 ret = 0;
457 break;
458 }
459 fs_devices = fs_devices->seed;
460 }
461 return ret;
462}
463
Josef Bacika826d6d2011-03-16 13:42:43 -0400464#define CORRUPT(reason, eb, root, slot) \
465 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
466 "root=%llu, slot=%d\n", reason, \
467 (unsigned long long)btrfs_header_bytenr(eb), \
468 (unsigned long long)root->objectid, slot)
469
470static noinline int check_leaf(struct btrfs_root *root,
471 struct extent_buffer *leaf)
472{
473 struct btrfs_key key;
474 struct btrfs_key leaf_key;
475 u32 nritems = btrfs_header_nritems(leaf);
476 int slot;
477
478 if (nritems == 0)
479 return 0;
480
481 /* Check the 0 item */
482 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
483 BTRFS_LEAF_DATA_SIZE(root)) {
484 CORRUPT("invalid item offset size pair", leaf, root, 0);
485 return -EIO;
486 }
487
488 /*
489 * Check to make sure each items keys are in the correct order and their
490 * offsets make sense. We only have to loop through nritems-1 because
491 * we check the current slot against the next slot, which verifies the
492 * next slot's offset+size makes sense and that the current's slot
493 * offset is correct.
494 */
495 for (slot = 0; slot < nritems - 1; slot++) {
496 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
497 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
498
499 /* Make sure the keys are in the right order */
500 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
501 CORRUPT("bad key order", leaf, root, slot);
502 return -EIO;
503 }
504
505 /*
506 * Make sure the offset and ends are right, remember that the
507 * item data starts at the end of the leaf and grows towards the
508 * front.
509 */
510 if (btrfs_item_offset_nr(leaf, slot) !=
511 btrfs_item_end_nr(leaf, slot + 1)) {
512 CORRUPT("slot offset bad", leaf, root, slot);
513 return -EIO;
514 }
515
516 /*
517 * Check to make sure that we don't point outside of the leaf,
518 * just incase all the items are consistent to eachother, but
519 * all point outside of the leaf.
520 */
521 if (btrfs_item_end_nr(leaf, slot) >
522 BTRFS_LEAF_DATA_SIZE(root)) {
523 CORRUPT("slot end outside of leaf", leaf, root, slot);
524 return -EIO;
525 }
526 }
527
528 return 0;
529}
530
Chris Mason727011e2010-08-06 13:21:20 -0400531struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree,
532 struct page *page, int max_walk)
533{
534 struct extent_buffer *eb;
535 u64 start = page_offset(page);
536 u64 target = start;
537 u64 min_start;
538
539 if (start < max_walk)
540 min_start = 0;
541 else
542 min_start = start - max_walk;
543
544 while (start >= min_start) {
545 eb = find_extent_buffer(tree, start, 0);
546 if (eb) {
547 /*
548 * we found an extent buffer and it contains our page
549 * horray!
550 */
551 if (eb->start <= target &&
552 eb->start + eb->len > target)
553 return eb;
554
555 /* we found an extent buffer that wasn't for us */
556 free_extent_buffer(eb);
557 return NULL;
558 }
559 if (start == 0)
560 break;
561 start -= PAGE_CACHE_SIZE;
562 }
563 return NULL;
564}
565
Christoph Hellwigb2950862008-12-02 09:54:17 -0500566static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400567 struct extent_state *state, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400568{
569 struct extent_io_tree *tree;
570 u64 found_start;
571 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400572 struct extent_buffer *eb;
573 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400574 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400575 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400576
Chris Masonce9adaa2008-04-09 16:28:12 -0400577 if (!page->private)
578 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500579
Chris Mason727011e2010-08-06 13:21:20 -0400580 tree = &BTRFS_I(page->mapping->host)->io_tree;
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500581 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500582
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400583 /* the pending IO might have been the only thing that kept this buffer
584 * in memory. Make sure we have a ref for all this other checks
585 */
586 extent_buffer_get(eb);
587
588 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400589 if (!reads_done)
590 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400591
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400592 eb->read_mirror = mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400593 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
594 ret = -EIO;
595 goto err;
596 }
597
Chris Masonce9adaa2008-04-09 16:28:12 -0400598 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400599 if (found_start != eb->start) {
David Sterba7a36dde2011-05-06 15:33:15 +0200600 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400601 "%llu %llu\n",
602 (unsigned long long)found_start,
603 (unsigned long long)eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400604 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400605 goto err;
606 }
Yan Zheng2b820322008-11-17 21:11:30 -0500607 if (check_tree_block_fsid(root, eb)) {
David Sterba7a36dde2011-05-06 15:33:15 +0200608 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
Chris Mason193f2842009-04-27 07:29:05 -0400609 (unsigned long long)eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400610 ret = -EIO;
611 goto err;
612 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400613 found_level = btrfs_header_level(eb);
614
Chris Mason85d4e462011-07-26 16:11:19 -0400615 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
616 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500617
Chris Masonce9adaa2008-04-09 16:28:12 -0400618 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400619 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400620 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400621 goto err;
622 }
623
624 /*
625 * If this is a leaf block and it is corrupt, set the corrupt bit so
626 * that we don't try and read the other copies of this block, just
627 * return -EIO.
628 */
629 if (found_level == 0 && check_leaf(root, eb)) {
630 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
631 ret = -EIO;
632 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400633
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400634 if (!ret)
635 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400636err:
Arne Jansen4bb31e92011-06-10 13:55:54 +0200637 if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
638 clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
639 btree_readahead_hook(root, eb, eb->start, ret);
640 }
641
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400642 if (ret)
643 clear_extent_buffer_uptodate(eb);
644 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400645out:
Chris Masonf1885912008-04-09 16:28:12 -0400646 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400647}
648
Josef Bacikea466792012-03-26 21:57:36 -0400649static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200650{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200651 struct extent_buffer *eb;
652 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
653
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500654 eb = (struct extent_buffer *)page->private;
Josef Bacikea466792012-03-26 21:57:36 -0400655 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400656 eb->read_mirror = failed_mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400657 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200658 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200659 return -EIO; /* we fixed nothing */
660}
661
Chris Masonce9adaa2008-04-09 16:28:12 -0400662static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400663{
664 struct end_io_wq *end_io_wq = bio->bi_private;
665 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400666
Chris Masonce9adaa2008-04-09 16:28:12 -0400667 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400668 end_io_wq->error = err;
Chris Mason8b712842008-06-11 16:50:36 -0400669 end_io_wq->work.func = end_workqueue_fn;
670 end_io_wq->work.flags = 0;
Chris Masond20f7042008-12-08 16:58:54 -0500671
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200672 if (bio->bi_rw & REQ_WRITE) {
Josef Bacik0cb59c92010-07-02 12:14:14 -0400673 if (end_io_wq->metadata == 1)
Chris Masoncad321a2008-12-17 14:51:42 -0500674 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
675 &end_io_wq->work);
Josef Bacik0cb59c92010-07-02 12:14:14 -0400676 else if (end_io_wq->metadata == 2)
677 btrfs_queue_worker(&fs_info->endio_freespace_worker,
678 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500679 else
680 btrfs_queue_worker(&fs_info->endio_write_workers,
681 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500682 } else {
683 if (end_io_wq->metadata)
684 btrfs_queue_worker(&fs_info->endio_meta_workers,
685 &end_io_wq->work);
686 else
687 btrfs_queue_worker(&fs_info->endio_workers,
688 &end_io_wq->work);
689 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400690}
691
Josef Bacik0cb59c92010-07-02 12:14:14 -0400692/*
693 * For the metadata arg you want
694 *
695 * 0 - if data
696 * 1 - if normal metadta
697 * 2 - if writing to the free space cache area
698 */
Chris Mason22c59942008-04-09 16:28:12 -0400699int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
700 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400701{
Chris Masonce9adaa2008-04-09 16:28:12 -0400702 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400703 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
704 if (!end_io_wq)
705 return -ENOMEM;
706
707 end_io_wq->private = bio->bi_private;
708 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400709 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400710 end_io_wq->error = 0;
711 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400712 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400713
714 bio->bi_private = end_io_wq;
715 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400716 return 0;
717}
718
Chris Masonb64a2852008-08-20 13:39:41 -0400719unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400720{
721 unsigned long limit = min_t(unsigned long,
722 info->workers.max_workers,
723 info->fs_devices->open_devices);
724 return 256 * limit;
725}
726
Chris Mason4a69a412008-11-06 22:03:00 -0500727static void run_one_async_start(struct btrfs_work *work)
728{
Chris Mason4a69a412008-11-06 22:03:00 -0500729 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100730 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500731
732 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100733 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
734 async->mirror_num, async->bio_flags,
735 async->bio_offset);
736 if (ret)
737 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500738}
739
740static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400741{
742 struct btrfs_fs_info *fs_info;
743 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400744 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400745
746 async = container_of(work, struct async_submit_bio, work);
747 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400748
Chris Masonb64a2852008-08-20 13:39:41 -0400749 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400750 limit = limit * 2 / 3;
751
Josef Bacik66657b32012-08-01 15:36:24 -0400752 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400753 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400754 wake_up(&fs_info->async_submit_wait);
755
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100756 /* If an error occured we just want to clean up the bio and move on */
757 if (async->error) {
758 bio_endio(async->bio, async->error);
759 return;
760 }
761
Chris Mason4a69a412008-11-06 22:03:00 -0500762 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400763 async->mirror_num, async->bio_flags,
764 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500765}
766
767static void run_one_async_free(struct btrfs_work *work)
768{
769 struct async_submit_bio *async;
770
771 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400772 kfree(async);
773}
774
Chris Mason44b8bd72008-04-16 11:14:51 -0400775int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
776 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400777 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400778 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500779 extent_submit_bio_hook_t *submit_bio_start,
780 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400781{
782 struct async_submit_bio *async;
783
784 async = kmalloc(sizeof(*async), GFP_NOFS);
785 if (!async)
786 return -ENOMEM;
787
788 async->inode = inode;
789 async->rw = rw;
790 async->bio = bio;
791 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500792 async->submit_bio_start = submit_bio_start;
793 async->submit_bio_done = submit_bio_done;
794
795 async->work.func = run_one_async_start;
796 async->work.ordered_func = run_one_async_done;
797 async->work.ordered_free = run_one_async_free;
798
Chris Mason8b712842008-06-11 16:50:36 -0400799 async->work.flags = 0;
Chris Masonc8b97812008-10-29 14:49:59 -0400800 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400801 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400802
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100803 async->error = 0;
804
Chris Masoncb03c742008-05-15 16:15:45 -0400805 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400806
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200807 if (rw & REQ_SYNC)
Chris Masond313d7a2009-04-20 15:50:09 -0400808 btrfs_set_work_high_prio(&async->work);
809
Chris Mason8b712842008-06-11 16:50:36 -0400810 btrfs_queue_worker(&fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400811
Chris Masond3977122009-01-05 21:25:51 -0500812 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500813 atomic_read(&fs_info->nr_async_submits)) {
814 wait_event(fs_info->async_submit_wait,
815 (atomic_read(&fs_info->nr_async_submits) == 0));
816 }
817
Chris Mason44b8bd72008-04-16 11:14:51 -0400818 return 0;
819}
820
Chris Masonce3ed712008-09-23 13:14:12 -0400821static int btree_csum_one_bio(struct bio *bio)
822{
823 struct bio_vec *bvec = bio->bi_io_vec;
824 int bio_index = 0;
825 struct btrfs_root *root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100826 int ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400827
828 WARN_ON(bio->bi_vcnt <= 0);
Chris Masond3977122009-01-05 21:25:51 -0500829 while (bio_index < bio->bi_vcnt) {
Chris Masonce3ed712008-09-23 13:14:12 -0400830 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100831 ret = csum_dirty_buffer(root, bvec->bv_page);
832 if (ret)
833 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400834 bio_index++;
835 bvec++;
836 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100837 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400838}
839
Chris Mason4a69a412008-11-06 22:03:00 -0500840static int __btree_submit_bio_start(struct inode *inode, int rw,
841 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400842 unsigned long bio_flags,
843 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400844{
Chris Mason8b712842008-06-11 16:50:36 -0400845 /*
846 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400847 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400848 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100849 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500850}
Chris Mason22c59942008-04-09 16:28:12 -0400851
Chris Mason4a69a412008-11-06 22:03:00 -0500852static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400853 int mirror_num, unsigned long bio_flags,
854 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500855{
Stefan Behrens61891922012-11-05 18:51:52 +0100856 int ret;
857
Chris Mason8b712842008-06-11 16:50:36 -0400858 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500859 * when we're called for a write, we're already in the async
860 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400861 */
Stefan Behrens61891922012-11-05 18:51:52 +0100862 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
863 if (ret)
864 bio_endio(bio, ret);
865 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400866}
867
Josef Bacikde0022b2012-09-25 14:25:58 -0400868static int check_async_write(struct inode *inode, unsigned long bio_flags)
869{
870 if (bio_flags & EXTENT_BIO_TREE_LOG)
871 return 0;
872#ifdef CONFIG_X86
873 if (cpu_has_xmm4_2)
874 return 0;
875#endif
876 return 1;
877}
878
Chris Mason44b8bd72008-04-16 11:14:51 -0400879static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400880 int mirror_num, unsigned long bio_flags,
881 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400882{
Josef Bacikde0022b2012-09-25 14:25:58 -0400883 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500884 int ret;
885
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200886 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500887 /*
888 * called for a read, do the setup so that checksum validation
889 * can happen in the async kernel threads
890 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400891 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
892 bio, 1);
Chris Mason1d4284b2012-03-28 20:31:37 -0400893 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100894 goto out_w_error;
895 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
896 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400897 } else if (!async) {
898 ret = btree_csum_one_bio(bio);
899 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100900 goto out_w_error;
901 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
902 mirror_num, 0);
903 } else {
904 /*
905 * kthread helpers are used to submit writes so that
906 * checksumming can happen in parallel across all CPUs
907 */
908 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
909 inode, rw, bio, mirror_num, 0,
910 bio_offset,
911 __btree_submit_bio_start,
912 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400913 }
Chris Masond313d7a2009-04-20 15:50:09 -0400914
Stefan Behrens61891922012-11-05 18:51:52 +0100915 if (ret) {
916out_w_error:
917 bio_endio(bio, ret);
918 }
919 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400920}
921
Jan Beulich3dd14622010-12-07 14:54:09 +0000922#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500923static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800924 struct page *newpage, struct page *page,
925 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500926{
927 /*
928 * we can't safely write a btree page from here,
929 * we haven't done the locking hook
930 */
931 if (PageDirty(page))
932 return -EAGAIN;
933 /*
934 * Buffers may be managed in a filesystem specific way.
935 * We must have no buffers or drop them.
936 */
937 if (page_has_private(page) &&
938 !try_to_release_page(page, GFP_KERNEL))
939 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800940 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500941}
Jan Beulich3dd14622010-12-07 14:54:09 +0000942#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500943
Chris Mason0da54682007-11-07 21:08:01 -0500944
945static int btree_writepages(struct address_space *mapping,
946 struct writeback_control *wbc)
947{
Chris Masond1310b22008-01-24 16:13:08 -0500948 struct extent_io_tree *tree;
949 tree = &BTRFS_I(mapping->host)->io_tree;
Chris Masond8d5f3e2007-12-11 12:42:00 -0500950 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Masonb9473432009-03-13 11:00:37 -0400951 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
Chris Mason793955b2007-11-26 16:34:41 -0800952 u64 num_dirty;
Chris Mason24ab9cd2008-09-24 14:51:30 -0400953 unsigned long thresh = 32 * 1024 * 1024;
Chris Mason448d6402007-11-27 07:52:01 -0800954
955 if (wbc->for_kupdate)
956 return 0;
957
Chris Masonb9473432009-03-13 11:00:37 -0400958 /* this is a bit racy, but that's ok */
959 num_dirty = root->fs_info->dirty_metadata_bytes;
Chris Masond3977122009-01-05 21:25:51 -0500960 if (num_dirty < thresh)
Chris Mason793955b2007-11-26 16:34:41 -0800961 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800962 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400963 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -0500964}
965
Christoph Hellwigb2950862008-12-02 09:54:17 -0500966static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -0400967{
Chris Masond1310b22008-01-24 16:13:08 -0500968 struct extent_io_tree *tree;
969 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +0200970 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400971}
972
Chris Mason70dec802008-01-29 09:59:12 -0500973static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -0400974{
Chris Mason98509cf2008-09-11 15:51:43 -0400975 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -0500976 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -0500977 /*
978 * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing
979 * slab allocation from alloc_extent_state down the callchain where
980 * it'd hit a BUG_ON as those flags are not allowed.
981 */
982 gfp_flags &= ~GFP_SLAB_BUG_MASK;
983
Josef Bacik3083ee22012-03-09 16:01:49 -0500984 return try_release_extent_buffer(page, gfp_flags);
Chris Masone20d96d2007-03-22 12:13:20 -0400985}
Chris Mason123abc82007-03-14 14:14:43 -0400986
Chris Mason5f39d392007-10-15 16:14:19 -0400987static void btree_invalidatepage(struct page *page, unsigned long offset)
Chris Masond98237b2007-03-28 13:57:48 -0400988{
Chris Masond1310b22008-01-24 16:13:08 -0500989 struct extent_io_tree *tree;
990 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -0400991 extent_invalidatepage(tree, page, offset);
992 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400993 if (PagePrivate(page)) {
Chris Masond3977122009-01-05 21:25:51 -0500994 printk(KERN_WARNING "btrfs warning page private not zero "
995 "on page %llu\n", (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400996 ClearPagePrivate(page);
997 set_page_private(page, 0);
998 page_cache_release(page);
999 }
Chris Masond98237b2007-03-28 13:57:48 -04001000}
1001
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001002static int btree_set_page_dirty(struct page *page)
1003{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001004#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001005 struct extent_buffer *eb;
1006
1007 BUG_ON(!PagePrivate(page));
1008 eb = (struct extent_buffer *)page->private;
1009 BUG_ON(!eb);
1010 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1011 BUG_ON(!atomic_read(&eb->refs));
1012 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001013#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001014 return __set_page_dirty_nobuffers(page);
1015}
1016
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001017static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001018 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001019 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001020 .releasepage = btree_releasepage,
1021 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001022#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001023 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001024#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001025 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001026};
1027
Chris Masonca7a79a2008-05-12 12:59:19 -04001028int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1029 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001030{
Chris Mason5f39d392007-10-15 16:14:19 -04001031 struct extent_buffer *buf = NULL;
1032 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001033 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001034
Chris Masondb945352007-10-15 16:15:53 -04001035 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001036 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001037 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001038 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001039 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001040 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001041 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001042}
1043
Arne Jansenab0fff02011-05-23 14:25:41 +02001044int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1045 int mirror_num, struct extent_buffer **eb)
1046{
1047 struct extent_buffer *buf = NULL;
1048 struct inode *btree_inode = root->fs_info->btree_inode;
1049 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1050 int ret;
1051
1052 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1053 if (!buf)
1054 return 0;
1055
1056 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1057
1058 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1059 btree_get_extent, mirror_num);
1060 if (ret) {
1061 free_extent_buffer(buf);
1062 return ret;
1063 }
1064
1065 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1066 free_extent_buffer(buf);
1067 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001068 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001069 *eb = buf;
1070 } else {
1071 free_extent_buffer(buf);
1072 }
1073 return 0;
1074}
1075
Chris Mason0999df52008-04-01 13:48:14 -04001076struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1077 u64 bytenr, u32 blocksize)
1078{
1079 struct inode *btree_inode = root->fs_info->btree_inode;
1080 struct extent_buffer *eb;
1081 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
David Sterbaf09d1f62011-04-21 01:08:01 +02001082 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001083 return eb;
1084}
1085
1086struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1087 u64 bytenr, u32 blocksize)
1088{
1089 struct inode *btree_inode = root->fs_info->btree_inode;
1090 struct extent_buffer *eb;
1091
1092 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
Chris Mason727011e2010-08-06 13:21:20 -04001093 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001094 return eb;
1095}
1096
1097
Chris Masone02119d2008-09-05 16:13:11 -04001098int btrfs_write_tree_block(struct extent_buffer *buf)
1099{
Chris Mason727011e2010-08-06 13:21:20 -04001100 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001101 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001102}
1103
1104int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1105{
Chris Mason727011e2010-08-06 13:21:20 -04001106 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001107 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001108}
1109
Chris Masondb945352007-10-15 16:15:53 -04001110struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001111 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001112{
Chris Mason5f39d392007-10-15 16:14:19 -04001113 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001114 int ret;
1115
Chris Masondb945352007-10-15 16:15:53 -04001116 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001117 if (!buf)
1118 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001119
Chris Masonca7a79a2008-05-12 12:59:19 -04001120 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason5f39d392007-10-15 16:14:19 -04001121 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001122
Chris Masoneb60cea2007-02-02 09:18:22 -05001123}
1124
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001125void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1126 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001127{
Chris Mason55c69072008-01-09 15:55:33 -05001128 if (btrfs_header_generation(buf) ==
Chris Mason925baed2008-06-25 16:01:30 -04001129 root->fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001130 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001131
Chris Masonb9473432009-03-13 11:00:37 -04001132 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1133 spin_lock(&root->fs_info->delalloc_lock);
1134 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1135 root->fs_info->dirty_metadata_bytes -= buf->len;
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001136 else {
1137 spin_unlock(&root->fs_info->delalloc_lock);
1138 btrfs_panic(root->fs_info, -EOVERFLOW,
1139 "Can't clear %lu bytes from "
Joe Perches533574c2012-07-30 14:40:13 -07001140 " dirty_mdatadata_bytes (%llu)",
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001141 buf->len,
1142 root->fs_info->dirty_metadata_bytes);
1143 }
Chris Masonb9473432009-03-13 11:00:37 -04001144 spin_unlock(&root->fs_info->delalloc_lock);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001145
Josef Baciked7b63e2012-10-15 13:33:54 -04001146 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1147 btrfs_set_lock_blocking(buf);
1148 clear_extent_buffer_dirty(buf);
1149 }
Chris Mason925baed2008-06-25 16:01:30 -04001150 }
Chris Mason5f39d392007-10-15 16:14:19 -04001151}
1152
Jeff Mahoney143bede2012-03-01 14:56:26 +01001153static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1154 u32 stripesize, struct btrfs_root *root,
1155 struct btrfs_fs_info *fs_info,
1156 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001157{
Chris Masoncfaa7292007-02-21 17:04:57 -05001158 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001159 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001160 root->sectorsize = sectorsize;
1161 root->nodesize = nodesize;
1162 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001163 root->stripesize = stripesize;
Chris Mason123abc82007-03-14 14:14:43 -04001164 root->ref_cows = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001165 root->track_dirty = 0;
Yan, Zhengc71bf092009-11-12 09:34:40 +00001166 root->in_radix = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001167 root->orphan_item_inserted = 0;
1168 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001169
Chris Mason0f7d52f2007-04-09 10:42:37 -04001170 root->objectid = objectid;
1171 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001172 root->highest_objectid = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001173 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001174 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001175 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001176 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001177 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001178
1179 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001180 INIT_LIST_HEAD(&root->root_list);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001181 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001182 spin_lock_init(&root->inode_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001183 spin_lock_init(&root->accounting_lock);
Chris Masona2135012008-06-25 16:01:30 -04001184 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001185 mutex_init(&root->log_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001186 init_waitqueue_head(&root->log_writer_wait);
1187 init_waitqueue_head(&root->log_commit_wait[0]);
1188 init_waitqueue_head(&root->log_commit_wait[1]);
1189 atomic_set(&root->log_commit[0], 0);
1190 atomic_set(&root->log_commit[1], 0);
1191 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001192 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001193 atomic_set(&root->orphan_inodes, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001194 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001195 root->last_log_commit = 0;
Chris Masond0c803c2008-09-11 16:17:57 -04001196 extent_io_tree_init(&root->dirty_log_pages,
David Sterbaf993c882011-04-20 23:35:57 +02001197 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001198
Chris Mason3768f362007-03-13 16:47:54 -04001199 memset(&root->root_key, 0, sizeof(root->root_key));
1200 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001201 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001202 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Chris Mason3f157a22008-06-25 16:01:31 -04001203 root->defrag_trans_start = fs_info->generation;
Josef Bacik58176a92007-08-29 15:47:34 -04001204 init_completion(&root->kobj_unregister);
Chris Mason6702ed42007-08-07 16:15:09 -04001205 root->defrag_running = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001206 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001207 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001208
Anand Jain5f3ab902012-12-07 09:28:54 +00001209 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001210}
1211
Jeff Mahoney200a5c12011-10-03 23:22:43 -04001212static int __must_check find_and_setup_root(struct btrfs_root *tree_root,
1213 struct btrfs_fs_info *fs_info,
1214 u64 objectid,
1215 struct btrfs_root *root)
Chris Mason3768f362007-03-13 16:47:54 -04001216{
1217 int ret;
Chris Masondb945352007-10-15 16:15:53 -04001218 u32 blocksize;
Yan Zheng84234f32008-10-29 14:49:05 -04001219 u64 generation;
Chris Mason3768f362007-03-13 16:47:54 -04001220
Chris Masondb945352007-10-15 16:15:53 -04001221 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001222 tree_root->sectorsize, tree_root->stripesize,
1223 root, fs_info, objectid);
Chris Mason3768f362007-03-13 16:47:54 -04001224 ret = btrfs_find_last_root(tree_root, objectid,
1225 &root->root_item, &root->root_key);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001226 if (ret > 0)
1227 return -ENOENT;
Jeff Mahoney200a5c12011-10-03 23:22:43 -04001228 else if (ret < 0)
1229 return ret;
Chris Mason3768f362007-03-13 16:47:54 -04001230
Yan Zheng84234f32008-10-29 14:49:05 -04001231 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001232 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
Chris Masonaf31f5e2011-11-03 15:17:42 -04001233 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001234 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001235 blocksize, generation);
Chris Masonb9fab912012-05-06 07:23:47 -04001236 if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) {
Chris Mason68433b72010-12-13 14:47:58 -05001237 free_extent_buffer(root->node);
Chris Masonaf31f5e2011-11-03 15:17:42 -04001238 root->node = NULL;
Chris Mason68433b72010-12-13 14:47:58 -05001239 return -EIO;
1240 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001241 root->commit_root = btrfs_root_node(root);
Chris Masond97e63b2007-02-20 16:40:44 -05001242 return 0;
1243}
1244
Al Virof84a8bd2011-11-17 15:57:57 -05001245static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001246{
1247 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1248 if (root)
1249 root->fs_info = fs_info;
1250 return root;
1251}
1252
Arne Jansen20897f52011-09-13 12:44:20 +02001253struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1254 struct btrfs_fs_info *fs_info,
1255 u64 objectid)
1256{
1257 struct extent_buffer *leaf;
1258 struct btrfs_root *tree_root = fs_info->tree_root;
1259 struct btrfs_root *root;
1260 struct btrfs_key key;
1261 int ret = 0;
1262 u64 bytenr;
1263
1264 root = btrfs_alloc_root(fs_info);
1265 if (!root)
1266 return ERR_PTR(-ENOMEM);
1267
1268 __setup_root(tree_root->nodesize, tree_root->leafsize,
1269 tree_root->sectorsize, tree_root->stripesize,
1270 root, fs_info, objectid);
1271 root->root_key.objectid = objectid;
1272 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1273 root->root_key.offset = 0;
1274
1275 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1276 0, objectid, NULL, 0, 0, 0);
1277 if (IS_ERR(leaf)) {
1278 ret = PTR_ERR(leaf);
1279 goto fail;
1280 }
1281
1282 bytenr = leaf->start;
1283 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1284 btrfs_set_header_bytenr(leaf, leaf->start);
1285 btrfs_set_header_generation(leaf, trans->transid);
1286 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1287 btrfs_set_header_owner(leaf, objectid);
1288 root->node = leaf;
1289
1290 write_extent_buffer(leaf, fs_info->fsid,
1291 (unsigned long)btrfs_header_fsid(leaf),
1292 BTRFS_FSID_SIZE);
1293 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1294 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
1295 BTRFS_UUID_SIZE);
1296 btrfs_mark_buffer_dirty(leaf);
1297
1298 root->commit_root = btrfs_root_node(root);
1299 root->track_dirty = 1;
1300
1301
1302 root->root_item.flags = 0;
1303 root->root_item.byte_limit = 0;
1304 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1305 btrfs_set_root_generation(&root->root_item, trans->transid);
1306 btrfs_set_root_level(&root->root_item, 0);
1307 btrfs_set_root_refs(&root->root_item, 1);
1308 btrfs_set_root_used(&root->root_item, leaf->len);
1309 btrfs_set_root_last_snapshot(&root->root_item, 0);
1310 btrfs_set_root_dirid(&root->root_item, 0);
1311 root->root_item.drop_level = 0;
1312
1313 key.objectid = objectid;
1314 key.type = BTRFS_ROOT_ITEM_KEY;
1315 key.offset = 0;
1316 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1317 if (ret)
1318 goto fail;
1319
1320 btrfs_tree_unlock(leaf);
1321
1322fail:
1323 if (ret)
1324 return ERR_PTR(ret);
1325
1326 return root;
1327}
1328
Yan Zheng7237f182009-01-21 12:54:03 -05001329static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1330 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001331{
1332 struct btrfs_root *root;
1333 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001334 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001335
Al Viro6f07e422011-11-17 00:46:16 -05001336 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001337 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001338 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001339
1340 __setup_root(tree_root->nodesize, tree_root->leafsize,
1341 tree_root->sectorsize, tree_root->stripesize,
1342 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1343
1344 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1345 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1346 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Yan Zheng7237f182009-01-21 12:54:03 -05001347 /*
1348 * log trees do not get reference counted because they go away
1349 * before a real commit is actually done. They do store pointers
1350 * to file data extents, and those reference counts still get
1351 * updated (along with back refs to the log tree).
1352 */
Chris Masone02119d2008-09-05 16:13:11 -04001353 root->ref_cows = 0;
1354
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001355 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
Arne Jansen66d7e7f2011-09-12 15:26:38 +02001356 BTRFS_TREE_LOG_OBJECTID, NULL,
Jan Schmidt5581a512012-05-16 17:04:52 +02001357 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001358 if (IS_ERR(leaf)) {
1359 kfree(root);
1360 return ERR_CAST(leaf);
1361 }
Chris Masone02119d2008-09-05 16:13:11 -04001362
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001363 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1364 btrfs_set_header_bytenr(leaf, leaf->start);
1365 btrfs_set_header_generation(leaf, trans->transid);
1366 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1367 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001368 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001369
1370 write_extent_buffer(root->node, root->fs_info->fsid,
1371 (unsigned long)btrfs_header_fsid(root->node),
1372 BTRFS_FSID_SIZE);
1373 btrfs_mark_buffer_dirty(root->node);
1374 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001375 return root;
1376}
1377
1378int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1379 struct btrfs_fs_info *fs_info)
1380{
1381 struct btrfs_root *log_root;
1382
1383 log_root = alloc_log_tree(trans, fs_info);
1384 if (IS_ERR(log_root))
1385 return PTR_ERR(log_root);
1386 WARN_ON(fs_info->log_root_tree);
1387 fs_info->log_root_tree = log_root;
1388 return 0;
1389}
1390
1391int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1392 struct btrfs_root *root)
1393{
1394 struct btrfs_root *log_root;
1395 struct btrfs_inode_item *inode_item;
1396
1397 log_root = alloc_log_tree(trans, root->fs_info);
1398 if (IS_ERR(log_root))
1399 return PTR_ERR(log_root);
1400
1401 log_root->last_trans = trans->transid;
1402 log_root->root_key.offset = root->root_key.objectid;
1403
1404 inode_item = &log_root->root_item.inode;
1405 inode_item->generation = cpu_to_le64(1);
1406 inode_item->size = cpu_to_le64(3);
1407 inode_item->nlink = cpu_to_le32(1);
1408 inode_item->nbytes = cpu_to_le64(root->leafsize);
1409 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1410
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001411 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001412
1413 WARN_ON(root->log_root);
1414 root->log_root = log_root;
1415 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001416 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001417 return 0;
1418}
1419
1420struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1421 struct btrfs_key *location)
1422{
1423 struct btrfs_root *root;
1424 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001425 struct btrfs_path *path;
Chris Mason5f39d392007-10-15 16:14:19 -04001426 struct extent_buffer *l;
Yan Zheng84234f32008-10-29 14:49:05 -04001427 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001428 u32 blocksize;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001429 int ret = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001430 int slot;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001431
Al Viro6f07e422011-11-17 00:46:16 -05001432 root = btrfs_alloc_root(fs_info);
Chris Mason0cf6c622007-06-09 09:22:25 -04001433 if (!root)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001434 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001435 if (location->offset == (u64)-1) {
Chris Masondb945352007-10-15 16:15:53 -04001436 ret = find_and_setup_root(tree_root, fs_info,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001437 location->objectid, root);
1438 if (ret) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04001439 kfree(root);
1440 return ERR_PTR(ret);
1441 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001442 goto out;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001443 }
1444
Chris Masondb945352007-10-15 16:15:53 -04001445 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001446 tree_root->sectorsize, tree_root->stripesize,
1447 root, fs_info, location->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001448
1449 path = btrfs_alloc_path();
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00001450 if (!path) {
1451 kfree(root);
1452 return ERR_PTR(-ENOMEM);
1453 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001454 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001455 if (ret == 0) {
1456 l = path->nodes[0];
Alexander Block8ea05e32012-07-25 17:35:53 +02001457 slot = path->slots[0];
1458 btrfs_read_root_item(tree_root, l, slot, &root->root_item);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001459 memcpy(&root->root_key, location, sizeof(*location));
Chris Mason0f7d52f2007-04-09 10:42:37 -04001460 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001461 btrfs_free_path(path);
1462 if (ret) {
Tsutomu Itoh5e540f72010-12-27 06:53:10 +00001463 kfree(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001464 if (ret > 0)
1465 ret = -ENOENT;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001466 return ERR_PTR(ret);
1467 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001468
Yan Zheng84234f32008-10-29 14:49:05 -04001469 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001470 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1471 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001472 blocksize, generation);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001473 root->commit_root = btrfs_root_node(root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001474 BUG_ON(!root->node); /* -ENOMEM */
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001475out:
Li Zefan08fe4db2011-03-28 02:01:25 +00001476 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
Chris Masone02119d2008-09-05 16:13:11 -04001477 root->ref_cows = 1;
Li Zefan08fe4db2011-03-28 02:01:25 +00001478 btrfs_check_and_init_root_item(&root->root_item);
1479 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001480
Chris Mason5eda7b52007-06-22 14:16:25 -04001481 return root;
1482}
1483
Chris Masonedbd8d42007-12-21 16:27:24 -05001484struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1485 struct btrfs_key *location)
Chris Mason5eda7b52007-06-22 14:16:25 -04001486{
1487 struct btrfs_root *root;
1488 int ret;
1489
Chris Masonedbd8d42007-12-21 16:27:24 -05001490 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1491 return fs_info->tree_root;
1492 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1493 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001494 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1495 return fs_info->chunk_root;
1496 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1497 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001498 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1499 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001500 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1501 return fs_info->quota_root ? fs_info->quota_root :
1502 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001503again:
1504 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001505 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1506 (unsigned long)location->objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001507 spin_unlock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001508 if (root)
1509 return root;
1510
Chris Masone02119d2008-09-05 16:13:11 -04001511 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001512 if (IS_ERR(root))
1513 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001514
Li Zefan581bb052011-04-20 10:06:11 +08001515 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
Li Zefan581bb052011-04-20 10:06:11 +08001516 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1517 GFP_NOFS);
David Sterba35a30d72011-06-13 15:18:23 +00001518 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1519 ret = -ENOMEM;
Li Zefan581bb052011-04-20 10:06:11 +08001520 goto fail;
David Sterba35a30d72011-06-13 15:18:23 +00001521 }
Li Zefan581bb052011-04-20 10:06:11 +08001522
1523 btrfs_init_free_ino_ctl(root);
1524 mutex_init(&root->fs_commit_mutex);
1525 spin_lock_init(&root->cache_lock);
1526 init_waitqueue_head(&root->cache_wait);
1527
Al Viro0ee5dc62011-07-07 15:44:25 -04001528 ret = get_anon_bdev(&root->anon_dev);
Chris Masonac08aed2011-06-13 11:28:50 -04001529 if (ret)
1530 goto fail;
Chris Mason3394e162008-11-17 20:42:26 -05001531
Yan, Zhengd68fc572010-05-16 10:49:58 -04001532 if (btrfs_root_refs(&root->root_item) == 0) {
1533 ret = -ENOENT;
1534 goto fail;
1535 }
1536
1537 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1538 if (ret < 0)
1539 goto fail;
1540 if (ret == 0)
1541 root->orphan_item_inserted = 1;
1542
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001543 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1544 if (ret)
1545 goto fail;
1546
1547 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04001548 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1549 (unsigned long)root->root_key.objectid,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001550 root);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001551 if (ret == 0)
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001552 root->in_radix = 1;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001553
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001554 spin_unlock(&fs_info->fs_roots_radix_lock);
1555 radix_tree_preload_end();
Chris Mason0f7d52f2007-04-09 10:42:37 -04001556 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001557 if (ret == -EEXIST) {
1558 free_fs_root(root);
1559 goto again;
1560 }
1561 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001562 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001563
1564 ret = btrfs_find_dead_roots(fs_info->tree_root,
1565 root->root_key.objectid);
1566 WARN_ON(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001567 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001568fail:
1569 free_fs_root(root);
1570 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001571}
1572
Chris Mason04160082008-03-26 10:28:07 -04001573static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1574{
1575 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1576 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001577 struct btrfs_device *device;
1578 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001579
Xiao Guangrong1f781602011-04-20 10:09:16 +00001580 rcu_read_lock();
1581 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001582 if (!device->bdev)
1583 continue;
Chris Mason04160082008-03-26 10:28:07 -04001584 bdi = blk_get_backing_dev_info(device->bdev);
1585 if (bdi && bdi_congested(bdi, bdi_bits)) {
1586 ret = 1;
1587 break;
1588 }
1589 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001590 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001591 return ret;
1592}
1593
Chris Mason38b66982008-04-22 09:22:11 -04001594/*
Jens Axboead081f12009-06-12 14:43:40 +02001595 * If this fails, caller must call bdi_destroy() to get rid of the
1596 * bdi again.
1597 */
Chris Mason04160082008-03-26 10:28:07 -04001598static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1599{
Jens Axboead081f12009-06-12 14:43:40 +02001600 int err;
1601
1602 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001603 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001604 if (err)
1605 return err;
1606
Chris Mason4575c9c2008-04-18 16:13:31 -04001607 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001608 bdi->congested_fn = btrfs_congested_fn;
1609 bdi->congested_data = info;
1610 return 0;
1611}
1612
Chris Mason8b712842008-06-11 16:50:36 -04001613/*
1614 * called by the kthread helper functions to finally call the bio end_io
1615 * functions. This is where read checksum verification actually happens
1616 */
1617static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001618{
Chris Masonce9adaa2008-04-09 16:28:12 -04001619 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001620 struct end_io_wq *end_io_wq;
1621 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001622 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001623
Chris Mason8b712842008-06-11 16:50:36 -04001624 end_io_wq = container_of(work, struct end_io_wq, work);
1625 bio = end_io_wq->bio;
1626 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001627
Chris Mason8b712842008-06-11 16:50:36 -04001628 error = end_io_wq->error;
1629 bio->bi_private = end_io_wq->private;
1630 bio->bi_end_io = end_io_wq->end_io;
1631 kfree(end_io_wq);
Chris Mason8b712842008-06-11 16:50:36 -04001632 bio_endio(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001633}
1634
Chris Masona74a4b92008-06-25 16:01:31 -04001635static int cleaner_kthread(void *arg)
1636{
1637 struct btrfs_root *root = arg;
1638
1639 do {
Yan, Zheng76dda932009-09-21 16:00:26 -04001640 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1641 mutex_trylock(&root->fs_info->cleaner_mutex)) {
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001642 btrfs_run_delayed_iputs(root);
Yan, Zheng76dda932009-09-21 16:00:26 -04001643 btrfs_clean_old_snapshots(root);
1644 mutex_unlock(&root->fs_info->cleaner_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001645 btrfs_run_defrag_inodes(root->fs_info);
Yan, Zheng76dda932009-09-21 16:00:26 -04001646 }
Chris Masona74a4b92008-06-25 16:01:31 -04001647
Tejun Heoa0acae02011-11-21 12:32:22 -08001648 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001649 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001650 if (!kthread_should_stop())
1651 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001652 __set_current_state(TASK_RUNNING);
1653 }
1654 } while (!kthread_should_stop());
1655 return 0;
1656}
1657
1658static int transaction_kthread(void *arg)
1659{
1660 struct btrfs_root *root = arg;
1661 struct btrfs_trans_handle *trans;
1662 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001663 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001664 unsigned long now;
1665 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001666 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001667
1668 do {
Jan Kara914b2002012-03-12 16:05:50 +01001669 cannot_commit = false;
Chris Masona74a4b92008-06-25 16:01:31 -04001670 delay = HZ * 30;
Chris Masona74a4b92008-06-25 16:01:31 -04001671 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1672
Josef Bacika4abeea2011-04-11 17:25:13 -04001673 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001674 cur = root->fs_info->running_transaction;
1675 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001676 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001677 goto sleep;
1678 }
Yan Zheng31153d82008-07-28 15:32:19 -04001679
Chris Masona74a4b92008-06-25 16:01:31 -04001680 now = get_seconds();
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001681 if (!cur->blocked &&
1682 (now < cur->start_time || now - cur->start_time < 30)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001683 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001684 delay = HZ * 5;
1685 goto sleep;
1686 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001687 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001688 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001689
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001690 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001691 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001692 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001693 if (PTR_ERR(trans) != -ENOENT)
1694 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001695 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001696 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001697 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001698 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001699 } else {
1700 btrfs_end_transaction(trans, root);
1701 }
Chris Masona74a4b92008-06-25 16:01:31 -04001702sleep:
1703 wake_up_process(root->fs_info->cleaner_kthread);
1704 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1705
Tejun Heoa0acae02011-11-21 12:32:22 -08001706 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001707 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001708 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001709 (!btrfs_transaction_blocked(root->fs_info) ||
1710 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001711 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001712 __set_current_state(TASK_RUNNING);
1713 }
1714 } while (!kthread_should_stop());
1715 return 0;
1716}
1717
Chris Masonaf31f5e2011-11-03 15:17:42 -04001718/*
1719 * this will find the highest generation in the array of
1720 * root backups. The index of the highest array is returned,
1721 * or -1 if we can't find anything.
1722 *
1723 * We check to make sure the array is valid by comparing the
1724 * generation of the latest root in the array with the generation
1725 * in the super block. If they don't match we pitch it.
1726 */
1727static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1728{
1729 u64 cur;
1730 int newest_index = -1;
1731 struct btrfs_root_backup *root_backup;
1732 int i;
1733
1734 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1735 root_backup = info->super_copy->super_roots + i;
1736 cur = btrfs_backup_tree_root_gen(root_backup);
1737 if (cur == newest_gen)
1738 newest_index = i;
1739 }
1740
1741 /* check to see if we actually wrapped around */
1742 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1743 root_backup = info->super_copy->super_roots;
1744 cur = btrfs_backup_tree_root_gen(root_backup);
1745 if (cur == newest_gen)
1746 newest_index = 0;
1747 }
1748 return newest_index;
1749}
1750
1751
1752/*
1753 * find the oldest backup so we know where to store new entries
1754 * in the backup array. This will set the backup_root_index
1755 * field in the fs_info struct
1756 */
1757static void find_oldest_super_backup(struct btrfs_fs_info *info,
1758 u64 newest_gen)
1759{
1760 int newest_index = -1;
1761
1762 newest_index = find_newest_super_backup(info, newest_gen);
1763 /* if there was garbage in there, just move along */
1764 if (newest_index == -1) {
1765 info->backup_root_index = 0;
1766 } else {
1767 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1768 }
1769}
1770
1771/*
1772 * copy all the root pointers into the super backup array.
1773 * this will bump the backup pointer by one when it is
1774 * done
1775 */
1776static void backup_super_roots(struct btrfs_fs_info *info)
1777{
1778 int next_backup;
1779 struct btrfs_root_backup *root_backup;
1780 int last_backup;
1781
1782 next_backup = info->backup_root_index;
1783 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1784 BTRFS_NUM_BACKUP_ROOTS;
1785
1786 /*
1787 * just overwrite the last backup if we're at the same generation
1788 * this happens only at umount
1789 */
1790 root_backup = info->super_for_commit->super_roots + last_backup;
1791 if (btrfs_backup_tree_root_gen(root_backup) ==
1792 btrfs_header_generation(info->tree_root->node))
1793 next_backup = last_backup;
1794
1795 root_backup = info->super_for_commit->super_roots + next_backup;
1796
1797 /*
1798 * make sure all of our padding and empty slots get zero filled
1799 * regardless of which ones we use today
1800 */
1801 memset(root_backup, 0, sizeof(*root_backup));
1802
1803 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1804
1805 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1806 btrfs_set_backup_tree_root_gen(root_backup,
1807 btrfs_header_generation(info->tree_root->node));
1808
1809 btrfs_set_backup_tree_root_level(root_backup,
1810 btrfs_header_level(info->tree_root->node));
1811
1812 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1813 btrfs_set_backup_chunk_root_gen(root_backup,
1814 btrfs_header_generation(info->chunk_root->node));
1815 btrfs_set_backup_chunk_root_level(root_backup,
1816 btrfs_header_level(info->chunk_root->node));
1817
1818 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1819 btrfs_set_backup_extent_root_gen(root_backup,
1820 btrfs_header_generation(info->extent_root->node));
1821 btrfs_set_backup_extent_root_level(root_backup,
1822 btrfs_header_level(info->extent_root->node));
1823
Chris Mason7c7e82a2011-11-06 18:50:56 -05001824 /*
1825 * we might commit during log recovery, which happens before we set
1826 * the fs_root. Make sure it is valid before we fill it in.
1827 */
1828 if (info->fs_root && info->fs_root->node) {
1829 btrfs_set_backup_fs_root(root_backup,
1830 info->fs_root->node->start);
1831 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001832 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001833 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001834 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001835 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001836
1837 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1838 btrfs_set_backup_dev_root_gen(root_backup,
1839 btrfs_header_generation(info->dev_root->node));
1840 btrfs_set_backup_dev_root_level(root_backup,
1841 btrfs_header_level(info->dev_root->node));
1842
1843 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1844 btrfs_set_backup_csum_root_gen(root_backup,
1845 btrfs_header_generation(info->csum_root->node));
1846 btrfs_set_backup_csum_root_level(root_backup,
1847 btrfs_header_level(info->csum_root->node));
1848
1849 btrfs_set_backup_total_bytes(root_backup,
1850 btrfs_super_total_bytes(info->super_copy));
1851 btrfs_set_backup_bytes_used(root_backup,
1852 btrfs_super_bytes_used(info->super_copy));
1853 btrfs_set_backup_num_devices(root_backup,
1854 btrfs_super_num_devices(info->super_copy));
1855
1856 /*
1857 * if we don't copy this out to the super_copy, it won't get remembered
1858 * for the next commit
1859 */
1860 memcpy(&info->super_copy->super_roots,
1861 &info->super_for_commit->super_roots,
1862 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1863}
1864
1865/*
1866 * this copies info out of the root backup array and back into
1867 * the in-memory super block. It is meant to help iterate through
1868 * the array, so you send it the number of backups you've already
1869 * tried and the last backup index you used.
1870 *
1871 * this returns -1 when it has tried all the backups
1872 */
1873static noinline int next_root_backup(struct btrfs_fs_info *info,
1874 struct btrfs_super_block *super,
1875 int *num_backups_tried, int *backup_index)
1876{
1877 struct btrfs_root_backup *root_backup;
1878 int newest = *backup_index;
1879
1880 if (*num_backups_tried == 0) {
1881 u64 gen = btrfs_super_generation(super);
1882
1883 newest = find_newest_super_backup(info, gen);
1884 if (newest == -1)
1885 return -1;
1886
1887 *backup_index = newest;
1888 *num_backups_tried = 1;
1889 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1890 /* we've tried all the backups, all done */
1891 return -1;
1892 } else {
1893 /* jump to the next oldest backup */
1894 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1895 BTRFS_NUM_BACKUP_ROOTS;
1896 *backup_index = newest;
1897 *num_backups_tried += 1;
1898 }
1899 root_backup = super->super_roots + newest;
1900
1901 btrfs_set_super_generation(super,
1902 btrfs_backup_tree_root_gen(root_backup));
1903 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1904 btrfs_set_super_root_level(super,
1905 btrfs_backup_tree_root_level(root_backup));
1906 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1907
1908 /*
1909 * fixme: the total bytes and num_devices need to match or we should
1910 * need a fsck
1911 */
1912 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1913 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1914 return 0;
1915}
1916
1917/* helper to cleanup tree roots */
1918static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
1919{
1920 free_extent_buffer(info->tree_root->node);
1921 free_extent_buffer(info->tree_root->commit_root);
1922 free_extent_buffer(info->dev_root->node);
1923 free_extent_buffer(info->dev_root->commit_root);
1924 free_extent_buffer(info->extent_root->node);
1925 free_extent_buffer(info->extent_root->commit_root);
1926 free_extent_buffer(info->csum_root->node);
1927 free_extent_buffer(info->csum_root->commit_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02001928 if (info->quota_root) {
1929 free_extent_buffer(info->quota_root->node);
1930 free_extent_buffer(info->quota_root->commit_root);
1931 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001932
1933 info->tree_root->node = NULL;
1934 info->tree_root->commit_root = NULL;
1935 info->dev_root->node = NULL;
1936 info->dev_root->commit_root = NULL;
1937 info->extent_root->node = NULL;
1938 info->extent_root->commit_root = NULL;
1939 info->csum_root->node = NULL;
1940 info->csum_root->commit_root = NULL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001941 if (info->quota_root) {
1942 info->quota_root->node = NULL;
1943 info->quota_root->commit_root = NULL;
1944 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001945
1946 if (chunk_root) {
1947 free_extent_buffer(info->chunk_root->node);
1948 free_extent_buffer(info->chunk_root->commit_root);
1949 info->chunk_root->node = NULL;
1950 info->chunk_root->commit_root = NULL;
1951 }
1952}
1953
1954
Al Viroad2b2c82011-11-17 01:10:02 -05001955int open_ctree(struct super_block *sb,
1956 struct btrfs_fs_devices *fs_devices,
1957 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05001958{
Chris Masondb945352007-10-15 16:15:53 -04001959 u32 sectorsize;
1960 u32 nodesize;
1961 u32 leafsize;
1962 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001963 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04001964 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001965 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05001966 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04001967 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001968 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05001969 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05001970 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001971 struct btrfs_root *extent_root;
1972 struct btrfs_root *csum_root;
1973 struct btrfs_root *chunk_root;
1974 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001975 struct btrfs_root *quota_root;
Chris Masone02119d2008-09-05 16:13:11 -04001976 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05001977 int ret;
Yane58ca022008-04-01 11:21:34 -04001978 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04001979 int num_backups_tried = 0;
1980 int backup_index = 0;
Chris Mason4543df72008-06-11 21:47:56 -04001981
Al Virof84a8bd2011-11-17 15:57:57 -05001982 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05001983 extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info);
1984 csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info);
1985 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
1986 dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02001987 quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info);
Chris Mason8790d502008-04-03 16:29:03 -04001988
Al Virof84a8bd2011-11-17 15:57:57 -05001989 if (!tree_root || !extent_root || !csum_root ||
Arne Jansenbcef60f2011-09-13 15:23:30 +02001990 !chunk_root || !dev_root || !quota_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04001991 err = -ENOMEM;
1992 goto fail;
1993 }
Yan, Zheng76dda932009-09-21 16:00:26 -04001994
1995 ret = init_srcu_struct(&fs_info->subvol_srcu);
1996 if (ret) {
1997 err = ret;
1998 goto fail;
1999 }
2000
2001 ret = setup_bdi(fs_info, &fs_info->bdi);
2002 if (ret) {
2003 err = ret;
2004 goto fail_srcu;
2005 }
2006
2007 fs_info->btree_inode = new_inode(sb);
2008 if (!fs_info->btree_inode) {
2009 err = -ENOMEM;
2010 goto fail_bdi;
2011 }
2012
Chris Masona6591712011-07-19 12:04:14 -04002013 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002014
Yan, Zheng76dda932009-09-21 16:00:26 -04002015 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002016 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002017 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002018 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Chris Masonea8c2812008-08-04 23:17:27 -04002019 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
Chris Mason5a3f23d2009-03-31 13:27:11 -04002020 INIT_LIST_HEAD(&fs_info->ordered_operations);
Yan Zheng11833d62009-09-11 16:11:19 -04002021 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Chris Mason1832a6d2007-12-21 16:27:21 -05002022 spin_lock_init(&fs_info->delalloc_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002023 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002024 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002025 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002026 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002027 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002028 spin_lock_init(&fs_info->tree_mod_seq_lock);
2029 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002030 mutex_init(&fs_info->reloc_mutex);
Chris Mason19c00dd2007-10-15 16:19:22 -04002031
Josef Bacik58176a92007-08-29 15:47:34 -04002032 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002033 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002034 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002035 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Chris Mason0b86a832008-03-24 15:01:56 -04002036 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002037 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2038 BTRFS_BLOCK_RSV_GLOBAL);
2039 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2040 BTRFS_BLOCK_RSV_DELALLOC);
2041 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2042 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2043 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2044 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2045 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002046 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002047 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002048 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002049 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002050 atomic_set(&fs_info->defrag_running, 0);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002051 atomic_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002052 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05002053 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002054 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002055 fs_info->defrag_inodes = RB_ROOT;
Josef Bacika4abeea2011-04-11 17:25:13 -04002056 fs_info->trans_no_join = 0;
Josef Bacik2bf64752011-09-26 17:12:22 -04002057 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002058 fs_info->tree_mod_log = RB_ROOT;
Chris Masonc8b97812008-10-29 14:49:59 -04002059
Arne Jansen90519d62011-05-23 14:30:00 +02002060 /* readahead state */
2061 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2062 spin_lock_init(&fs_info->reada_lock);
Chris Mason2c90e5d2007-04-02 10:50:19 -04002063
Chris Masonb34b0862008-12-19 15:43:22 -05002064 fs_info->thread_pool_size = min_t(unsigned long,
2065 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002066
Chris Mason3eaa2882008-07-24 11:57:52 -04002067 INIT_LIST_HEAD(&fs_info->ordered_extents);
2068 spin_lock_init(&fs_info->ordered_extent_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002069 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2070 GFP_NOFS);
2071 if (!fs_info->delayed_root) {
2072 err = -ENOMEM;
2073 goto fail_iput;
2074 }
2075 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002076
Arne Jansena2de7332011-03-08 14:14:00 +01002077 mutex_init(&fs_info->scrub_lock);
2078 atomic_set(&fs_info->scrubs_running, 0);
2079 atomic_set(&fs_info->scrub_pause_req, 0);
2080 atomic_set(&fs_info->scrubs_paused, 0);
2081 atomic_set(&fs_info->scrub_cancel_req, 0);
2082 init_waitqueue_head(&fs_info->scrub_pause_wait);
2083 init_rwsem(&fs_info->scrub_super_lock);
2084 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002085#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2086 fs_info->check_integrity_print_mask = 0;
2087#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002088
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002089 spin_lock_init(&fs_info->balance_lock);
2090 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002091 atomic_set(&fs_info->balance_running, 0);
2092 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002093 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002094 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002095 init_waitqueue_head(&fs_info->balance_wait_q);
Chris Masona061fc82008-05-07 11:43:44 -04002096
Chris Mason0afbaf82008-04-18 14:17:20 -04002097 sb->s_blocksize = 4096;
2098 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002099 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002100
Yan, Zheng76dda932009-09-21 16:00:26 -04002101 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002102 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002103 /*
2104 * we set the i_size on the btree inode to the max possible int.
2105 * the real end of the address space is determined by all of
2106 * the devices in the system
2107 */
2108 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002109 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002110 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2111
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002112 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002113 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002114 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002115 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002116 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002117
Chris Masond1310b22008-01-24 16:13:08 -05002118 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2119
Chris Mason0f7d52f2007-04-09 10:42:37 -04002120 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2121 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2122 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002123 set_bit(BTRFS_INODE_DUMMY,
2124 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Chris Mason22b0ebd2007-03-30 08:47:31 -04002125 insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002126
Chris Masone02119d2008-09-05 16:13:11 -04002127 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002128 fs_info->block_group_cache_tree = RB_ROOT;
Chris Mason925baed2008-06-25 16:01:30 -04002129
Yan Zheng11833d62009-09-11 16:11:19 -04002130 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002131 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002132 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002133 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002134 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002135 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002136
Chris Masond98237b2007-03-28 13:57:48 -04002137
Chris Mason5a3f23d2009-03-31 13:27:11 -04002138 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002139 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002140 mutex_init(&fs_info->chunk_mutex);
2141 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002142 mutex_init(&fs_info->cleaner_mutex);
2143 mutex_init(&fs_info->volume_mutex);
Yan Zheng276e6802009-07-30 09:40:40 -04002144 init_rwsem(&fs_info->extent_commit_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002145 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002146 init_rwsem(&fs_info->subvol_sem);
Stefan Behrense922e082012-11-05 17:26:40 +01002147 fs_info->dev_replace.lock_owner = 0;
2148 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2149 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2150 mutex_init(&fs_info->dev_replace.lock_management_lock);
2151 mutex_init(&fs_info->dev_replace.lock);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002152
Arne Jansen416ac512011-09-13 12:56:09 +02002153 spin_lock_init(&fs_info->qgroup_lock);
2154 fs_info->qgroup_tree = RB_ROOT;
2155 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2156 fs_info->qgroup_seq = 1;
2157 fs_info->quota_enabled = 0;
2158 fs_info->pending_quota_state = 0;
2159
Chris Masonfa9c0d792009-04-03 09:47:43 -04002160 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2161 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2162
Chris Mason7d9eb122008-07-08 14:19:17 -04002163 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002164 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002165 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002166 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002167
Chris Mason0b86a832008-03-24 15:01:56 -04002168 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002169 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002170
Chris Mason3c4bb262012-03-27 18:56:56 -04002171 invalidate_bdev(fs_devices->latest_bdev);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002172 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002173 if (!bh) {
2174 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002175 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002176 }
Chris Mason39279cc2007-06-12 06:35:45 -04002177
David Sterba6c417612011-04-13 15:41:04 +02002178 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2179 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2180 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002181 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002182
David Sterba6c417612011-04-13 15:41:04 +02002183 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002184
David Sterba6c417612011-04-13 15:41:04 +02002185 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002186 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002187 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002188
liuboacce9522011-01-06 19:30:25 +08002189 /* check FS state, whether FS is broken. */
2190 fs_info->fs_state |= btrfs_super_flags(disk_super);
2191
David Sterbafcd1f0652012-03-06 00:06:18 +01002192 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2193 if (ret) {
2194 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2195 err = ret;
2196 goto fail_alloc;
2197 }
liuboacce9522011-01-06 19:30:25 +08002198
Liu Bo75e7cb72011-03-22 10:12:20 +00002199 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002200 * run through our array of backup supers and setup
2201 * our ring pointer to the oldest one
2202 */
2203 generation = btrfs_super_generation(disk_super);
2204 find_oldest_super_backup(fs_info, generation);
2205
2206 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002207 * In the long term, we'll store the compression type in the super
2208 * block, and it'll be used for per file compression control.
2209 */
2210 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2211
Yan Zheng2b820322008-11-17 21:11:30 -05002212 ret = btrfs_parse_options(tree_root, options);
2213 if (ret) {
2214 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002215 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002216 }
Chris Masondfe25022008-05-13 13:46:40 -04002217
Josef Bacikf2b636e2008-12-02 06:36:08 -05002218 features = btrfs_super_incompat_flags(disk_super) &
2219 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2220 if (features) {
2221 printk(KERN_ERR "BTRFS: couldn't mount because of "
2222 "unsupported optional features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002223 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002224 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002225 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002226 }
2227
Chris Mason727011e2010-08-06 13:21:20 -04002228 if (btrfs_super_leafsize(disk_super) !=
2229 btrfs_super_nodesize(disk_super)) {
2230 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2231 "blocksizes don't match. node %d leaf %d\n",
2232 btrfs_super_nodesize(disk_super),
2233 btrfs_super_leafsize(disk_super));
2234 err = -EINVAL;
2235 goto fail_alloc;
2236 }
2237 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2238 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2239 "blocksize (%d) was too large\n",
2240 btrfs_super_leafsize(disk_super));
2241 err = -EINVAL;
2242 goto fail_alloc;
2243 }
2244
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002245 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002246 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002247 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002248 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002249
2250 /*
2251 * flag our filesystem as having big metadata blocks if
2252 * they are bigger than the page size
2253 */
2254 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2255 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2256 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2257 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2258 }
2259
Chris Masonbc3f1162012-03-29 17:02:47 -04002260 nodesize = btrfs_super_nodesize(disk_super);
2261 leafsize = btrfs_super_leafsize(disk_super);
2262 sectorsize = btrfs_super_sectorsize(disk_super);
2263 stripesize = btrfs_super_stripesize(disk_super);
2264
2265 /*
2266 * mixed block groups end up with duplicate but slightly offset
2267 * extent buffers for the same range. It leads to corruptions
2268 */
2269 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2270 (sectorsize != leafsize)) {
2271 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2272 "are not allowed for mixed block groups on %s\n",
2273 sb->s_id);
2274 goto fail_alloc;
2275 }
2276
Li Zefana6fa6fa2010-10-25 15:12:26 +08002277 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002278
Josef Bacikf2b636e2008-12-02 06:36:08 -05002279 features = btrfs_super_compat_ro_flags(disk_super) &
2280 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2281 if (!(sb->s_flags & MS_RDONLY) && features) {
2282 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2283 "unsupported option features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002284 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002285 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002286 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002287 }
Chris Mason61d92c32009-10-02 19:11:56 -04002288
2289 btrfs_init_workers(&fs_info->generic_worker,
2290 "genwork", 1, NULL);
2291
Chris Mason5443be42008-08-15 15:34:16 -04002292 btrfs_init_workers(&fs_info->workers, "worker",
Chris Mason61d92c32009-10-02 19:11:56 -04002293 fs_info->thread_pool_size,
2294 &fs_info->generic_worker);
Chris Masonc8b97812008-10-29 14:49:59 -04002295
Chris Mason771ed682008-11-06 22:02:51 -05002296 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
Chris Mason61d92c32009-10-02 19:11:56 -04002297 fs_info->thread_pool_size,
2298 &fs_info->generic_worker);
Chris Mason771ed682008-11-06 22:02:51 -05002299
Miao Xie8ccf6f192012-10-25 09:28:04 +00002300 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
2301 fs_info->thread_pool_size,
2302 &fs_info->generic_worker);
2303
Chris Mason5443be42008-08-15 15:34:16 -04002304 btrfs_init_workers(&fs_info->submit_workers, "submit",
Chris Masonb720d202008-08-15 15:34:14 -04002305 min_t(u64, fs_devices->num_devices,
Chris Mason61d92c32009-10-02 19:11:56 -04002306 fs_info->thread_pool_size),
2307 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002308
Josef Bacikbab39bf2011-06-30 14:42:28 -04002309 btrfs_init_workers(&fs_info->caching_workers, "cache",
2310 2, &fs_info->generic_worker);
2311
Chris Mason61b49442008-07-31 15:42:53 -04002312 /* a higher idle thresh on the submit workers makes it much more
2313 * likely that bios will be send down in a sane order to the
2314 * devices
2315 */
2316 fs_info->submit_workers.idle_thresh = 64;
Chris Mason53863232008-08-15 15:34:18 -04002317
Chris Mason771ed682008-11-06 22:02:51 -05002318 fs_info->workers.idle_thresh = 16;
Chris Mason4a69a412008-11-06 22:03:00 -05002319 fs_info->workers.ordered = 1;
Chris Mason61b49442008-07-31 15:42:53 -04002320
Chris Mason771ed682008-11-06 22:02:51 -05002321 fs_info->delalloc_workers.idle_thresh = 2;
2322 fs_info->delalloc_workers.ordered = 1;
2323
Chris Mason61d92c32009-10-02 19:11:56 -04002324 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2325 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002326 btrfs_init_workers(&fs_info->endio_workers, "endio",
Chris Mason61d92c32009-10-02 19:11:56 -04002327 fs_info->thread_pool_size,
2328 &fs_info->generic_worker);
Chris Masond20f7042008-12-08 16:58:54 -05002329 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
Chris Mason61d92c32009-10-02 19:11:56 -04002330 fs_info->thread_pool_size,
2331 &fs_info->generic_worker);
Chris Masoncad321a2008-12-17 14:51:42 -05002332 btrfs_init_workers(&fs_info->endio_meta_write_workers,
Chris Mason61d92c32009-10-02 19:11:56 -04002333 "endio-meta-write", fs_info->thread_pool_size,
2334 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002335 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
Chris Mason61d92c32009-10-02 19:11:56 -04002336 fs_info->thread_pool_size,
2337 &fs_info->generic_worker);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002338 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2339 1, &fs_info->generic_worker);
Miao Xie16cdcec2011-04-22 18:12:22 +08002340 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2341 fs_info->thread_pool_size,
2342 &fs_info->generic_worker);
Arne Jansen90519d62011-05-23 14:30:00 +02002343 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2344 fs_info->thread_pool_size,
2345 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002346
2347 /*
2348 * endios are largely parallel and should have a very
2349 * low idle thresh
2350 */
2351 fs_info->endio_workers.idle_thresh = 4;
Chris Masonb51912c2009-02-04 09:23:24 -05002352 fs_info->endio_meta_workers.idle_thresh = 4;
2353
Chris Mason90428462009-08-04 16:56:34 -04002354 fs_info->endio_write_workers.idle_thresh = 2;
2355 fs_info->endio_meta_write_workers.idle_thresh = 2;
Arne Jansen90519d62011-05-23 14:30:00 +02002356 fs_info->readahead_workers.idle_thresh = 2;
Chris Mason90428462009-08-04 16:56:34 -04002357
Josef Bacik0dc3b842011-11-18 14:37:27 -05002358 /*
2359 * btrfs_start_workers can really only fail because of ENOMEM so just
2360 * return -ENOMEM if any of these fail.
2361 */
2362 ret = btrfs_start_workers(&fs_info->workers);
2363 ret |= btrfs_start_workers(&fs_info->generic_worker);
2364 ret |= btrfs_start_workers(&fs_info->submit_workers);
2365 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2366 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2367 ret |= btrfs_start_workers(&fs_info->endio_workers);
2368 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
2369 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2370 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2371 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2372 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2373 ret |= btrfs_start_workers(&fs_info->caching_workers);
2374 ret |= btrfs_start_workers(&fs_info->readahead_workers);
Miao Xie8ccf6f192012-10-25 09:28:04 +00002375 ret |= btrfs_start_workers(&fs_info->flush_workers);
Josef Bacik0dc3b842011-11-18 14:37:27 -05002376 if (ret) {
Ilya Dryomovfed425c2012-06-22 12:13:01 -06002377 err = -ENOMEM;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002378 goto fail_sb_buffer;
2379 }
Chris Mason4543df72008-06-11 21:47:56 -04002380
Chris Mason4575c9c2008-04-18 16:13:31 -04002381 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002382 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2383 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002384
Chris Masondb945352007-10-15 16:15:53 -04002385 tree_root->nodesize = nodesize;
2386 tree_root->leafsize = leafsize;
2387 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002388 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002389
2390 sb->s_blocksize = sectorsize;
2391 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002392
Chris Mason39279cc2007-06-12 06:35:45 -04002393 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
2394 sizeof(disk_super->magic))) {
Chris Masond3977122009-01-05 21:25:51 -05002395 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002396 goto fail_sb_buffer;
2397 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002398
Liu Bo8d082fb2012-04-03 09:56:53 +08002399 if (sectorsize != PAGE_SIZE) {
2400 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2401 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002402 goto fail_sb_buffer;
2403 }
2404
Chris Mason925baed2008-06-25 16:01:30 -04002405 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002406 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002407 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002408 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002409 printk(KERN_WARNING "btrfs: failed to read the system "
2410 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002411 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002412 }
Chris Mason0b86a832008-03-24 15:01:56 -04002413
2414 blocksize = btrfs_level_size(tree_root,
2415 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002416 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002417
2418 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2419 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2420
2421 chunk_root->node = read_tree_block(chunk_root,
2422 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002423 blocksize, generation);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002424 BUG_ON(!chunk_root->node); /* -ENOMEM */
David Woodhouse83121942009-07-22 16:52:13 -04002425 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2426 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2427 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002428 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002429 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002430 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2431 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002432
Chris Masone17cade2008-04-15 15:41:47 -04002433 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Chris Masond3977122009-01-05 21:25:51 -05002434 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2435 BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002436
Chris Mason0b86a832008-03-24 15:01:56 -04002437 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002438 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002439 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2440 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002441 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002442 }
Chris Mason0b86a832008-03-24 15:01:56 -04002443
Stefan Behrens8dabb742012-11-06 13:15:27 +01002444 /*
2445 * keep the device that is marked to be the target device for the
2446 * dev_replace procedure
2447 */
2448 btrfs_close_extra_devices(fs_info, fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002449
Chris Masona6b0d5c2012-02-20 20:53:43 -05002450 if (!fs_devices->latest_bdev) {
2451 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2452 sb->s_id);
2453 goto fail_tree_roots;
2454 }
2455
Chris Masonaf31f5e2011-11-03 15:17:42 -04002456retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002457 blocksize = btrfs_level_size(tree_root,
2458 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002459 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002460
Chris Masone20d96d2007-03-22 12:13:20 -04002461 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002462 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002463 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002464 if (!tree_root->node ||
2465 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
David Woodhouse83121942009-07-22 16:52:13 -04002466 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2467 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002468
2469 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002470 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002471
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002472 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2473 tree_root->commit_root = btrfs_root_node(tree_root);
Chris Masondb945352007-10-15 16:15:53 -04002474
2475 ret = find_and_setup_root(tree_root, fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04002476 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002477 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002478 goto recovery_tree_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002479 extent_root->track_dirty = 1;
2480
2481 ret = find_and_setup_root(tree_root, fs_info,
2482 BTRFS_DEV_TREE_OBJECTID, dev_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002483 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002484 goto recovery_tree_root;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002485 dev_root->track_dirty = 1;
Chris Mason3768f362007-03-13 16:47:54 -04002486
Chris Masond20f7042008-12-08 16:58:54 -05002487 ret = find_and_setup_root(tree_root, fs_info,
2488 BTRFS_CSUM_TREE_OBJECTID, csum_root);
2489 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002490 goto recovery_tree_root;
Chris Masond20f7042008-12-08 16:58:54 -05002491 csum_root->track_dirty = 1;
2492
Arne Jansenbcef60f2011-09-13 15:23:30 +02002493 ret = find_and_setup_root(tree_root, fs_info,
2494 BTRFS_QUOTA_TREE_OBJECTID, quota_root);
2495 if (ret) {
2496 kfree(quota_root);
2497 quota_root = fs_info->quota_root = NULL;
2498 } else {
2499 quota_root->track_dirty = 1;
2500 fs_info->quota_enabled = 1;
2501 fs_info->pending_quota_state = 1;
2502 }
2503
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002504 fs_info->generation = generation;
2505 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002506
Ilya Dryomov68310a52012-06-22 12:24:12 -06002507 ret = btrfs_recover_balance(fs_info);
2508 if (ret) {
2509 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2510 goto fail_block_groups;
2511 }
2512
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002513 ret = btrfs_init_dev_stats(fs_info);
2514 if (ret) {
2515 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2516 ret);
2517 goto fail_block_groups;
2518 }
2519
Stefan Behrens8dabb742012-11-06 13:15:27 +01002520 ret = btrfs_init_dev_replace(fs_info);
2521 if (ret) {
2522 pr_err("btrfs: failed to init dev_replace: %d\n", ret);
2523 goto fail_block_groups;
2524 }
2525
2526 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2527
liuboc59021f2011-03-07 02:13:14 +00002528 ret = btrfs_init_space_info(fs_info);
2529 if (ret) {
2530 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2531 goto fail_block_groups;
2532 }
2533
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002534 ret = btrfs_read_block_groups(extent_root);
2535 if (ret) {
2536 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2537 goto fail_block_groups;
2538 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002539 fs_info->num_tolerated_disk_barrier_failures =
2540 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002541 if (fs_info->fs_devices->missing_devices >
2542 fs_info->num_tolerated_disk_barrier_failures &&
2543 !(sb->s_flags & MS_RDONLY)) {
2544 printk(KERN_WARNING
2545 "Btrfs: too many missing devices, writeable mount is not allowed\n");
2546 goto fail_block_groups;
2547 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002548
Chris Masona74a4b92008-06-25 16:01:31 -04002549 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2550 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002551 if (IS_ERR(fs_info->cleaner_kthread))
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002552 goto fail_block_groups;
Chris Masona74a4b92008-06-25 16:01:31 -04002553
2554 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2555 tree_root,
2556 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002557 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002558 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002559
Chris Masonc2898112009-06-10 09:51:32 -04002560 if (!btrfs_test_opt(tree_root, SSD) &&
2561 !btrfs_test_opt(tree_root, NOSSD) &&
2562 !fs_info->fs_devices->rotating) {
2563 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2564 "mode\n");
2565 btrfs_set_opt(fs_info->mount_opt, SSD);
2566 }
2567
Stefan Behrens21adbd52011-11-09 13:44:05 +01002568#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2569 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2570 ret = btrfsic_mount(tree_root, fs_devices,
2571 btrfs_test_opt(tree_root,
2572 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2573 1 : 0,
2574 fs_info->check_integrity_print_mask);
2575 if (ret)
2576 printk(KERN_WARNING "btrfs: failed to initialize"
2577 " integrity check module %s\n", sb->s_id);
2578 }
2579#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002580 ret = btrfs_read_qgroup_config(fs_info);
2581 if (ret)
2582 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002583
liuboacce9522011-01-06 19:30:25 +08002584 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002585 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002586 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002587
Chris Mason7c2ca462008-11-19 15:13:35 -05002588 if (fs_devices->rw_devices == 0) {
Chris Masond3977122009-01-05 21:25:51 -05002589 printk(KERN_WARNING "Btrfs log replay required "
2590 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002591 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002592 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002593 }
Chris Masone02119d2008-09-05 16:13:11 -04002594 blocksize =
2595 btrfs_level_size(tree_root,
2596 btrfs_super_log_root_level(disk_super));
2597
Al Viro6f07e422011-11-17 00:46:16 -05002598 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002599 if (!log_tree_root) {
2600 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002601 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002602 }
Chris Masone02119d2008-09-05 16:13:11 -04002603
2604 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2605 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2606
2607 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002608 blocksize,
2609 generation + 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002610 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002611 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002612 if (ret) {
2613 btrfs_error(tree_root->fs_info, ret,
2614 "Failed to recover log tree");
2615 free_extent_buffer(log_tree_root->node);
2616 kfree(log_tree_root);
2617 goto fail_trans_kthread;
2618 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002619
2620 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002621 ret = btrfs_commit_super(tree_root);
2622 if (ret)
2623 goto fail_trans_kthread;
Yan Zhenge556ce22008-11-20 10:25:19 -05002624 }
Chris Masone02119d2008-09-05 16:13:11 -04002625 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002626
Yan, Zheng76dda932009-09-21 16:00:26 -04002627 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002628 if (ret)
2629 goto fail_trans_kthread;
Yan, Zheng76dda932009-09-21 16:00:26 -04002630
Chris Mason7c2ca462008-11-19 15:13:35 -05002631 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002632 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002633 if (ret)
2634 goto fail_trans_kthread;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002635
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002636 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002637 if (ret < 0) {
2638 printk(KERN_WARNING
2639 "btrfs: failed to recover relocation\n");
2640 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002641 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002642 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002643 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002644
Chris Mason3de45862008-11-17 21:02:50 -05002645 location.objectid = BTRFS_FS_TREE_OBJECTID;
2646 location.type = BTRFS_ROOT_ITEM_KEY;
2647 location.offset = (u64)-1;
2648
Chris Mason3de45862008-11-17 21:02:50 -05002649 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2650 if (!fs_info->fs_root)
Arne Jansenbcef60f2011-09-13 15:23:30 +02002651 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002652 if (IS_ERR(fs_info->fs_root)) {
2653 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002654 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002655 }
Chris Masonc2898112009-06-10 09:51:32 -04002656
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002657 if (sb->s_flags & MS_RDONLY)
2658 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002659
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002660 down_read(&fs_info->cleanup_work_sem);
2661 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2662 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002663 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002664 close_ctree(tree_root);
2665 return ret;
2666 }
2667 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002668
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002669 ret = btrfs_resume_balance_async(fs_info);
2670 if (ret) {
2671 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2672 close_ctree(tree_root);
2673 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002674 }
2675
Stefan Behrens8dabb742012-11-06 13:15:27 +01002676 ret = btrfs_resume_dev_replace_async(fs_info);
2677 if (ret) {
2678 pr_warn("btrfs: failed to resume dev_replace\n");
2679 close_ctree(tree_root);
2680 return ret;
2681 }
2682
Al Viroad2b2c82011-11-17 01:10:02 -05002683 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002684
Arne Jansenbcef60f2011-09-13 15:23:30 +02002685fail_qgroup:
2686 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05002687fail_trans_kthread:
2688 kthread_stop(fs_info->transaction_kthread);
Chris Mason3f157a22008-06-25 16:01:31 -04002689fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002690 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002691
2692 /*
2693 * make sure we're done with the btree inode before we stop our
2694 * kthreads
2695 */
2696 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2697 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2698
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002699fail_block_groups:
2700 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002701
2702fail_tree_roots:
2703 free_root_pointers(fs_info, 1);
2704
Chris Mason39279cc2007-06-12 06:35:45 -04002705fail_sb_buffer:
Chris Mason61d92c32009-10-02 19:11:56 -04002706 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason306c8b62011-11-03 15:21:39 -04002707 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Mason247e7432008-07-17 12:53:51 -04002708 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002709 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002710 btrfs_stop_workers(&fs_info->workers);
2711 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002712 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002713 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002714 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002715 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002716 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002717 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04002718 btrfs_stop_workers(&fs_info->caching_workers);
Miao Xie8ccf6f192012-10-25 09:28:04 +00002719 btrfs_stop_workers(&fs_info->flush_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002720fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04002721fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002722 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2723
Chris Mason7c2ca462008-11-19 15:13:35 -05002724 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Mason4543df72008-06-11 21:47:56 -04002725 iput(fs_info->btree_inode);
Jens Axboead081f12009-06-12 14:43:40 +02002726fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05002727 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002728fail_srcu:
2729 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002730fail:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002731 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05002732 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002733
2734recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04002735 if (!btrfs_test_opt(tree_root, RECOVERY))
2736 goto fail_tree_roots;
2737
2738 free_root_pointers(fs_info, 0);
2739
2740 /* don't use the log in recovery mode, it won't be valid */
2741 btrfs_set_super_log_root(disk_super, 0);
2742
2743 /* we can't trust the free space cache either */
2744 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2745
2746 ret = next_root_backup(fs_info, fs_info->super_copy,
2747 &num_backups_tried, &backup_index);
2748 if (ret == -1)
2749 goto fail_block_groups;
2750 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05002751}
2752
Chris Masonf2984462008-04-10 16:19:33 -04002753static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2754{
Chris Masonf2984462008-04-10 16:19:33 -04002755 if (uptodate) {
2756 set_buffer_uptodate(bh);
2757 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02002758 struct btrfs_device *device = (struct btrfs_device *)
2759 bh->b_private;
2760
Josef Bacik606686e2012-06-04 14:03:51 -04002761 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
2762 "I/O error on %s\n",
2763 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04002764 /* note, we dont' set_buffer_write_io_error because we have
2765 * our own ways of dealing with the IO errors
2766 */
Chris Masonf2984462008-04-10 16:19:33 -04002767 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02002768 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04002769 }
2770 unlock_buffer(bh);
2771 put_bh(bh);
2772}
2773
Yan Zhenga512bbf2008-12-08 16:46:26 -05002774struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2775{
2776 struct buffer_head *bh;
2777 struct buffer_head *latest = NULL;
2778 struct btrfs_super_block *super;
2779 int i;
2780 u64 transid = 0;
2781 u64 bytenr;
2782
2783 /* we would like to check all the supers, but that would make
2784 * a btrfs mount succeed after a mkfs from a different FS.
2785 * So, we need to add a special mount option to scan for
2786 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2787 */
2788 for (i = 0; i < 1; i++) {
2789 bytenr = btrfs_sb_offset(i);
2790 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2791 break;
2792 bh = __bread(bdev, bytenr / 4096, 4096);
2793 if (!bh)
2794 continue;
2795
2796 super = (struct btrfs_super_block *)bh->b_data;
2797 if (btrfs_super_bytenr(super) != bytenr ||
2798 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2799 sizeof(super->magic))) {
2800 brelse(bh);
2801 continue;
2802 }
2803
2804 if (!latest || btrfs_super_generation(super) > transid) {
2805 brelse(latest);
2806 latest = bh;
2807 transid = btrfs_super_generation(super);
2808 } else {
2809 brelse(bh);
2810 }
2811 }
2812 return latest;
2813}
2814
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002815/*
2816 * this should be called twice, once with wait == 0 and
2817 * once with wait == 1. When wait == 0 is done, all the buffer heads
2818 * we write are pinned.
2819 *
2820 * They are released when wait == 1 is done.
2821 * max_mirrors must be the same for both runs, and it indicates how
2822 * many supers on this one device should be written.
2823 *
2824 * max_mirrors == 0 means to write them all.
2825 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002826static int write_dev_supers(struct btrfs_device *device,
2827 struct btrfs_super_block *sb,
2828 int do_barriers, int wait, int max_mirrors)
2829{
2830 struct buffer_head *bh;
2831 int i;
2832 int ret;
2833 int errors = 0;
2834 u32 crc;
2835 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002836
2837 if (max_mirrors == 0)
2838 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2839
Yan Zhenga512bbf2008-12-08 16:46:26 -05002840 for (i = 0; i < max_mirrors; i++) {
2841 bytenr = btrfs_sb_offset(i);
2842 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2843 break;
2844
2845 if (wait) {
2846 bh = __find_get_block(device->bdev, bytenr / 4096,
2847 BTRFS_SUPER_INFO_SIZE);
2848 BUG_ON(!bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002849 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002850 if (!buffer_uptodate(bh))
2851 errors++;
2852
2853 /* drop our reference */
2854 brelse(bh);
2855
2856 /* drop the reference from the wait == 0 run */
2857 brelse(bh);
2858 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002859 } else {
2860 btrfs_set_super_bytenr(sb, bytenr);
2861
2862 crc = ~(u32)0;
2863 crc = btrfs_csum_data(NULL, (char *)sb +
2864 BTRFS_CSUM_SIZE, crc,
2865 BTRFS_SUPER_INFO_SIZE -
2866 BTRFS_CSUM_SIZE);
2867 btrfs_csum_final(crc, sb->csum);
2868
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002869 /*
2870 * one reference for us, and we leave it for the
2871 * caller
2872 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002873 bh = __getblk(device->bdev, bytenr / 4096,
2874 BTRFS_SUPER_INFO_SIZE);
2875 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2876
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002877 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002878 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002879
2880 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002881 lock_buffer(bh);
2882 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02002883 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002884 }
2885
Chris Mason387125f2011-11-18 15:07:51 -05002886 /*
2887 * we fua the first super. The others we allow
2888 * to go down lazy.
2889 */
Stefan Behrens21adbd52011-11-09 13:44:05 +01002890 ret = btrfsic_submit_bh(WRITE_FUA, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002891 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05002892 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002893 }
2894 return errors < i ? 0 : -1;
2895}
2896
Chris Mason387125f2011-11-18 15:07:51 -05002897/*
2898 * endio for the write_dev_flush, this will wake anyone waiting
2899 * for the barrier when it is done
2900 */
2901static void btrfs_end_empty_barrier(struct bio *bio, int err)
2902{
2903 if (err) {
2904 if (err == -EOPNOTSUPP)
2905 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
2906 clear_bit(BIO_UPTODATE, &bio->bi_flags);
2907 }
2908 if (bio->bi_private)
2909 complete(bio->bi_private);
2910 bio_put(bio);
2911}
2912
2913/*
2914 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
2915 * sent down. With wait == 1, it waits for the previous flush.
2916 *
2917 * any device where the flush fails with eopnotsupp are flagged as not-barrier
2918 * capable
2919 */
2920static int write_dev_flush(struct btrfs_device *device, int wait)
2921{
2922 struct bio *bio;
2923 int ret = 0;
2924
2925 if (device->nobarriers)
2926 return 0;
2927
2928 if (wait) {
2929 bio = device->flush_bio;
2930 if (!bio)
2931 return 0;
2932
2933 wait_for_completion(&device->flush_wait);
2934
2935 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Josef Bacik606686e2012-06-04 14:03:51 -04002936 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
2937 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05002938 device->nobarriers = 1;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002939 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
Chris Mason387125f2011-11-18 15:07:51 -05002940 ret = -EIO;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002941 btrfs_dev_stat_inc_and_print(device,
2942 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05002943 }
2944
2945 /* drop the reference from the wait == 0 run */
2946 bio_put(bio);
2947 device->flush_bio = NULL;
2948
2949 return ret;
2950 }
2951
2952 /*
2953 * one reference for us, and we leave it for the
2954 * caller
2955 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01002956 device->flush_bio = NULL;
Chris Mason387125f2011-11-18 15:07:51 -05002957 bio = bio_alloc(GFP_NOFS, 0);
2958 if (!bio)
2959 return -ENOMEM;
2960
2961 bio->bi_end_io = btrfs_end_empty_barrier;
2962 bio->bi_bdev = device->bdev;
2963 init_completion(&device->flush_wait);
2964 bio->bi_private = &device->flush_wait;
2965 device->flush_bio = bio;
2966
2967 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002968 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05002969
2970 return 0;
2971}
2972
2973/*
2974 * send an empty flush down to each device in parallel,
2975 * then wait for them
2976 */
2977static int barrier_all_devices(struct btrfs_fs_info *info)
2978{
2979 struct list_head *head;
2980 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002981 int errors_send = 0;
2982 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05002983 int ret;
2984
2985 /* send down all the barriers */
2986 head = &info->fs_devices->devices;
2987 list_for_each_entry_rcu(dev, head, dev_list) {
2988 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002989 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05002990 continue;
2991 }
2992 if (!dev->in_fs_metadata || !dev->writeable)
2993 continue;
2994
2995 ret = write_dev_flush(dev, 0);
2996 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002997 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05002998 }
2999
3000 /* wait for all the barriers */
3001 list_for_each_entry_rcu(dev, head, dev_list) {
3002 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003003 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003004 continue;
3005 }
3006 if (!dev->in_fs_metadata || !dev->writeable)
3007 continue;
3008
3009 ret = write_dev_flush(dev, 1);
3010 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003011 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003012 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003013 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3014 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003015 return -EIO;
3016 return 0;
3017}
3018
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003019int btrfs_calc_num_tolerated_disk_barrier_failures(
3020 struct btrfs_fs_info *fs_info)
3021{
3022 struct btrfs_ioctl_space_info space;
3023 struct btrfs_space_info *sinfo;
3024 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3025 BTRFS_BLOCK_GROUP_SYSTEM,
3026 BTRFS_BLOCK_GROUP_METADATA,
3027 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3028 int num_types = 4;
3029 int i;
3030 int c;
3031 int num_tolerated_disk_barrier_failures =
3032 (int)fs_info->fs_devices->num_devices;
3033
3034 for (i = 0; i < num_types; i++) {
3035 struct btrfs_space_info *tmp;
3036
3037 sinfo = NULL;
3038 rcu_read_lock();
3039 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3040 if (tmp->flags == types[i]) {
3041 sinfo = tmp;
3042 break;
3043 }
3044 }
3045 rcu_read_unlock();
3046
3047 if (!sinfo)
3048 continue;
3049
3050 down_read(&sinfo->groups_sem);
3051 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3052 if (!list_empty(&sinfo->block_groups[c])) {
3053 u64 flags;
3054
3055 btrfs_get_block_group_info(
3056 &sinfo->block_groups[c], &space);
3057 if (space.total_bytes == 0 ||
3058 space.used_bytes == 0)
3059 continue;
3060 flags = space.flags;
3061 /*
3062 * return
3063 * 0: if dup, single or RAID0 is configured for
3064 * any of metadata, system or data, else
3065 * 1: if RAID5 is configured, or if RAID1 or
3066 * RAID10 is configured and only two mirrors
3067 * are used, else
3068 * 2: if RAID6 is configured, else
3069 * num_mirrors - 1: if RAID1 or RAID10 is
3070 * configured and more than
3071 * 2 mirrors are used.
3072 */
3073 if (num_tolerated_disk_barrier_failures > 0 &&
3074 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3075 BTRFS_BLOCK_GROUP_RAID0)) ||
3076 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3077 == 0)))
3078 num_tolerated_disk_barrier_failures = 0;
3079 else if (num_tolerated_disk_barrier_failures > 1
3080 &&
3081 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3082 BTRFS_BLOCK_GROUP_RAID10)))
3083 num_tolerated_disk_barrier_failures = 1;
3084 }
3085 }
3086 up_read(&sinfo->groups_sem);
3087 }
3088
3089 return num_tolerated_disk_barrier_failures;
3090}
3091
Yan Zhenga512bbf2008-12-08 16:46:26 -05003092int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003093{
Chris Masone5e9a522009-06-10 15:17:02 -04003094 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003095 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003096 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003097 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003098 int ret;
3099 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003100 int max_errors;
3101 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003102 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003103
David Sterba6c417612011-04-13 15:41:04 +02003104 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Masonf2984462008-04-10 16:19:33 -04003105 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003106 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003107
David Sterba6c417612011-04-13 15:41:04 +02003108 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003109 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003110
Chris Mason174ba502011-05-27 10:03:58 -04003111 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003112 head = &root->fs_info->fs_devices->devices;
Chris Mason387125f2011-11-18 15:07:51 -05003113
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003114 if (do_barriers) {
3115 ret = barrier_all_devices(root->fs_info);
3116 if (ret) {
3117 mutex_unlock(
3118 &root->fs_info->fs_devices->device_list_mutex);
3119 btrfs_error(root->fs_info, ret,
3120 "errors while submitting device barriers.");
3121 return ret;
3122 }
3123 }
Chris Mason387125f2011-11-18 15:07:51 -05003124
Xiao Guangrong1f781602011-04-20 10:09:16 +00003125 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003126 if (!dev->bdev) {
3127 total_errors++;
3128 continue;
3129 }
Yan Zheng2b820322008-11-17 21:11:30 -05003130 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003131 continue;
3132
Yan Zheng2b820322008-11-17 21:11:30 -05003133 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003134 btrfs_set_stack_device_type(dev_item, dev->type);
3135 btrfs_set_stack_device_id(dev_item, dev->devid);
3136 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3137 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3138 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3139 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3140 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3141 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003142 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003143
Chris Masona061fc82008-05-07 11:43:44 -04003144 flags = btrfs_super_flags(sb);
3145 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003146
Yan Zhenga512bbf2008-12-08 16:46:26 -05003147 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003148 if (ret)
3149 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003150 }
Chris Masona236aed2008-04-29 09:38:00 -04003151 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05003152 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3153 total_errors);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003154
3155 /* This shouldn't happen. FUA is masked off if unsupported */
Chris Masona236aed2008-04-29 09:38:00 -04003156 BUG();
3157 }
Chris Masonf2984462008-04-10 16:19:33 -04003158
Yan Zhenga512bbf2008-12-08 16:46:26 -05003159 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003160 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003161 if (!dev->bdev)
3162 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003163 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003164 continue;
3165
Yan Zhenga512bbf2008-12-08 16:46:26 -05003166 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3167 if (ret)
3168 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003169 }
Chris Mason174ba502011-05-27 10:03:58 -04003170 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003171 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003172 btrfs_error(root->fs_info, -EIO,
3173 "%d errors while writing supers", total_errors);
3174 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003175 }
Chris Masonf2984462008-04-10 16:19:33 -04003176 return 0;
3177}
3178
Yan Zhenga512bbf2008-12-08 16:46:26 -05003179int write_ctree_super(struct btrfs_trans_handle *trans,
3180 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003181{
Chris Masone66f7092007-04-20 13:16:02 -04003182 int ret;
Chris Mason2c90e5d2007-04-02 10:50:19 -04003183
Yan Zhenga512bbf2008-12-08 16:46:26 -05003184 ret = write_all_supers(root, max_mirrors);
Chris Mason5f39d392007-10-15 16:14:19 -04003185 return ret;
Chris Masoncfaa7292007-02-21 17:04:57 -05003186}
3187
Jeff Mahoney143bede2012-03-01 14:56:26 +01003188void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003189{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003190 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003191 radix_tree_delete(&fs_info->fs_roots_radix,
3192 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003193 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003194
3195 if (btrfs_root_refs(&root->root_item) == 0)
3196 synchronize_srcu(&fs_info->subvol_srcu);
3197
Li Zefan581bb052011-04-20 10:06:11 +08003198 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3199 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003200 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003201}
3202
3203static void free_fs_root(struct btrfs_root *root)
3204{
Li Zefan82d59022011-04-20 10:33:24 +08003205 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003206 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Al Viro0ee5dc62011-07-07 15:44:25 -04003207 if (root->anon_dev)
3208 free_anon_bdev(root->anon_dev);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003209 free_extent_buffer(root->node);
3210 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003211 kfree(root->free_ino_ctl);
3212 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003213 kfree(root->name);
Chris Mason2619ba12007-04-10 16:58:11 -04003214 kfree(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003215}
3216
Jeff Mahoney143bede2012-03-01 14:56:26 +01003217static void del_fs_roots(struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04003218{
3219 int ret;
3220 struct btrfs_root *gang[8];
3221 int i;
3222
Yan, Zheng76dda932009-09-21 16:00:26 -04003223 while (!list_empty(&fs_info->dead_roots)) {
3224 gang[0] = list_entry(fs_info->dead_roots.next,
3225 struct btrfs_root, root_list);
3226 list_del(&gang[0]->root_list);
3227
3228 if (gang[0]->in_radix) {
3229 btrfs_free_fs_root(fs_info, gang[0]);
3230 } else {
3231 free_extent_buffer(gang[0]->node);
3232 free_extent_buffer(gang[0]->commit_root);
3233 kfree(gang[0]);
3234 }
3235 }
3236
Chris Masond3977122009-01-05 21:25:51 -05003237 while (1) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04003238 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3239 (void **)gang, 0,
3240 ARRAY_SIZE(gang));
3241 if (!ret)
3242 break;
Chris Mason2619ba12007-04-10 16:58:11 -04003243 for (i = 0; i < ret; i++)
Chris Mason5eda7b52007-06-22 14:16:25 -04003244 btrfs_free_fs_root(fs_info, gang[i]);
Chris Mason0f7d52f2007-04-09 10:42:37 -04003245 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04003246}
Chris Masonb4100d62007-04-12 12:14:00 -04003247
Yan Zhengc146afa2008-11-12 14:34:12 -05003248int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3249{
3250 u64 root_objectid = 0;
3251 struct btrfs_root *gang[8];
3252 int i;
3253 int ret;
3254
3255 while (1) {
3256 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3257 (void **)gang, root_objectid,
3258 ARRAY_SIZE(gang));
3259 if (!ret)
3260 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003261
3262 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05003263 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003264 int err;
3265
Yan Zhengc146afa2008-11-12 14:34:12 -05003266 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003267 err = btrfs_orphan_cleanup(gang[i]);
3268 if (err)
3269 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003270 }
3271 root_objectid++;
3272 }
3273 return 0;
3274}
3275
3276int btrfs_commit_super(struct btrfs_root *root)
3277{
3278 struct btrfs_trans_handle *trans;
3279 int ret;
3280
3281 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003282 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003283 btrfs_clean_old_snapshots(root);
3284 mutex_unlock(&root->fs_info->cleaner_mutex);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003285
3286 /* wait until ongoing cleanup work done */
3287 down_write(&root->fs_info->cleanup_work_sem);
3288 up_write(&root->fs_info->cleanup_work_sem);
3289
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003290 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003291 if (IS_ERR(trans))
3292 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05003293 ret = btrfs_commit_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003294 if (ret)
3295 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003296 /* run commit again to drop the original snapshot */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003297 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003298 if (IS_ERR(trans))
3299 return PTR_ERR(trans);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003300 ret = btrfs_commit_transaction(trans, root);
3301 if (ret)
3302 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003303 ret = btrfs_write_and_wait_transaction(NULL, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003304 if (ret) {
3305 btrfs_error(root->fs_info, ret,
3306 "Failed to sync btree inode to disk.");
3307 return ret;
3308 }
Yan Zhengc146afa2008-11-12 14:34:12 -05003309
Yan Zhenga512bbf2008-12-08 16:46:26 -05003310 ret = write_ctree_super(NULL, root, 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05003311 return ret;
3312}
3313
Chris Masone20d96d2007-03-22 12:13:20 -04003314int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003315{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003316 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003317 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003318
Chris Masonfacda1e2007-06-08 18:11:48 -04003319 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003320 smp_mb();
3321
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003322 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003323 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003324
Stefan Behrens8dabb742012-11-06 13:15:27 +01003325 btrfs_dev_replace_suspend_for_unmount(fs_info);
3326
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003327 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003328
3329 /* wait for any defraggers to finish */
3330 wait_event(fs_info->transaction_wait,
3331 (atomic_read(&fs_info->defrag_running) == 0));
3332
3333 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003334 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003335
Yan Zhengc146afa2008-11-12 14:34:12 -05003336 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003337 ret = btrfs_commit_super(root);
3338 if (ret)
3339 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3340 }
3341
Stefan Behrens68ce9682012-08-01 05:45:52 -06003342 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
3343 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003344
Josef Bacik300e4f82011-08-29 14:06:00 -04003345 btrfs_put_block_group_cache(fs_info);
3346
Al Viroe3029d92011-11-17 00:56:18 -05003347 kthread_stop(fs_info->transaction_kthread);
3348 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003349
Yan Zhengf25784b2009-07-28 08:41:57 -04003350 fs_info->closing = 2;
3351 smp_mb();
3352
Arne Jansenbcef60f2011-09-13 15:23:30 +02003353 btrfs_free_qgroup_config(root->fs_info);
3354
Chris Masonb0c68f82008-01-31 11:05:37 -05003355 if (fs_info->delalloc_bytes) {
Chris Masond3977122009-01-05 21:25:51 -05003356 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07003357 (unsigned long long)fs_info->delalloc_bytes);
Chris Masonb0c68f82008-01-31 11:05:37 -05003358 }
Yanbcc63ab2008-07-30 16:29:20 -04003359
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003360 free_extent_buffer(fs_info->extent_root->node);
3361 free_extent_buffer(fs_info->extent_root->commit_root);
3362 free_extent_buffer(fs_info->tree_root->node);
3363 free_extent_buffer(fs_info->tree_root->commit_root);
Al Viroe3029d92011-11-17 00:56:18 -05003364 free_extent_buffer(fs_info->chunk_root->node);
3365 free_extent_buffer(fs_info->chunk_root->commit_root);
3366 free_extent_buffer(fs_info->dev_root->node);
3367 free_extent_buffer(fs_info->dev_root->commit_root);
3368 free_extent_buffer(fs_info->csum_root->node);
3369 free_extent_buffer(fs_info->csum_root->commit_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003370 if (fs_info->quota_root) {
3371 free_extent_buffer(fs_info->quota_root->node);
3372 free_extent_buffer(fs_info->quota_root->commit_root);
3373 }
Chris Masond20f7042008-12-08 16:58:54 -05003374
Al Viroe3029d92011-11-17 00:56:18 -05003375 btrfs_free_block_groups(fs_info);
Yan Zhengc146afa2008-11-12 14:34:12 -05003376
Chris Mason0f7d52f2007-04-09 10:42:37 -04003377 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003378
Yan Zhengc146afa2008-11-12 14:34:12 -05003379 iput(fs_info->btree_inode);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003380
Chris Mason61d92c32009-10-02 19:11:56 -04003381 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04003382 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05003383 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04003384 btrfs_stop_workers(&fs_info->workers);
3385 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05003386 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05003387 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04003388 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04003389 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04003390 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08003391 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04003392 btrfs_stop_workers(&fs_info->caching_workers);
Arne Jansen90519d62011-05-23 14:30:00 +02003393 btrfs_stop_workers(&fs_info->readahead_workers);
Miao Xie8ccf6f192012-10-25 09:28:04 +00003394 btrfs_stop_workers(&fs_info->flush_workers);
Chris Masond6bfde82008-04-30 13:59:35 -04003395
Stefan Behrens21adbd52011-11-09 13:44:05 +01003396#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3397 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3398 btrfsic_unmount(root, fs_info->fs_devices);
3399#endif
3400
Chris Masondfe25022008-05-13 13:46:40 -04003401 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003402 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003403
Chris Mason04160082008-03-26 10:28:07 -04003404 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003405 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003406
Chris Masoneb60cea2007-02-02 09:18:22 -05003407 return 0;
3408}
3409
Chris Masonb9fab912012-05-06 07:23:47 -04003410int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3411 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003412{
Chris Mason1259ab72008-05-12 13:39:03 -04003413 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003414 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003415
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003416 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003417 if (!ret)
3418 return ret;
3419
3420 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003421 parent_transid, atomic);
3422 if (ret == -EAGAIN)
3423 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003424 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003425}
Chris Mason6702ed42007-08-07 16:15:09 -04003426
Chris Mason5f39d392007-10-15 16:14:19 -04003427int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3428{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003429 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003430}
3431
3432void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3433{
Chris Mason727011e2010-08-06 13:21:20 -04003434 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Mason5f39d392007-10-15 16:14:19 -04003435 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003436 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003437
Chris Masonb9447ef2009-03-09 11:45:38 -04003438 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003439 if (transid != root->fs_info->generation)
3440 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003441 "found %llu running %llu\n",
Chris Masondb945352007-10-15 16:15:53 -04003442 (unsigned long long)buf->start,
Chris Masond3977122009-01-05 21:25:51 -05003443 (unsigned long long)transid,
3444 (unsigned long long)root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003445 was_dirty = set_extent_buffer_dirty(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003446 if (!was_dirty) {
3447 spin_lock(&root->fs_info->delalloc_lock);
3448 root->fs_info->dirty_metadata_bytes += buf->len;
3449 spin_unlock(&root->fs_info->delalloc_lock);
3450 }
Chris Masoneb60cea2007-02-02 09:18:22 -05003451}
3452
Liu Bob53d3f52012-11-14 14:34:34 +00003453static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3454 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003455{
Chris Mason188de642008-05-09 11:52:25 -04003456 /*
3457 * looks as though older kernels can get into trouble with
3458 * this code, they end up stuck in balance_dirty_pages forever
3459 */
Chris Masond6bfde82008-04-30 13:59:35 -04003460 u64 num_dirty;
Chris Mason771ed682008-11-06 22:02:51 -05003461 unsigned long thresh = 32 * 1024 * 1024;
Chris Masond6bfde82008-04-30 13:59:35 -04003462
Jens Axboe6933c022009-03-17 09:36:37 +01003463 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003464 return;
3465
Liu Bob53d3f52012-11-14 14:34:34 +00003466 if (flush_delayed)
3467 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003468
3469 num_dirty = root->fs_info->dirty_metadata_bytes;
3470
3471 if (num_dirty > thresh) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08003472 balance_dirty_pages_ratelimited(
3473 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08003474 }
3475 return;
3476}
3477
Liu Bob53d3f52012-11-14 14:34:34 +00003478void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08003479{
Liu Bob53d3f52012-11-14 14:34:34 +00003480 __btrfs_btree_balance_dirty(root, 1);
3481}
Miao Xie16cdcec2011-04-22 18:12:22 +08003482
Liu Bob53d3f52012-11-14 14:34:34 +00003483void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3484{
3485 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04003486}
Chris Mason6b800532007-10-15 16:17:34 -04003487
Chris Masonca7a79a2008-05-12 12:59:19 -04003488int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003489{
Chris Mason727011e2010-08-06 13:21:20 -04003490 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003491 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003492}
Chris Mason0da54682007-11-07 21:08:01 -05003493
David Sterbafcd1f0652012-03-06 00:06:18 +01003494static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003495 int read_only)
3496{
David Sterbafcd1f0652012-03-06 00:06:18 +01003497 if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) {
3498 printk(KERN_ERR "btrfs: unsupported checksum algorithm\n");
3499 return -EINVAL;
3500 }
liuboacce9522011-01-06 19:30:25 +08003501
David Sterbafcd1f0652012-03-06 00:06:18 +01003502 if (read_only)
3503 return 0;
3504
David Sterbafcd1f0652012-03-06 00:06:18 +01003505 return 0;
liuboacce9522011-01-06 19:30:25 +08003506}
3507
Stefan Behrens68ce9682012-08-01 05:45:52 -06003508void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003509{
liuboacce9522011-01-06 19:30:25 +08003510 mutex_lock(&root->fs_info->cleaner_mutex);
3511 btrfs_run_delayed_iputs(root);
3512 mutex_unlock(&root->fs_info->cleaner_mutex);
3513
3514 down_write(&root->fs_info->cleanup_work_sem);
3515 up_write(&root->fs_info->cleanup_work_sem);
3516
3517 /* cleanup FS via transaction */
3518 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003519}
3520
Jeff Mahoney143bede2012-03-01 14:56:26 +01003521static void btrfs_destroy_ordered_operations(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003522{
3523 struct btrfs_inode *btrfs_inode;
3524 struct list_head splice;
3525
3526 INIT_LIST_HEAD(&splice);
3527
3528 mutex_lock(&root->fs_info->ordered_operations_mutex);
3529 spin_lock(&root->fs_info->ordered_extent_lock);
3530
3531 list_splice_init(&root->fs_info->ordered_operations, &splice);
3532 while (!list_empty(&splice)) {
3533 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3534 ordered_operations);
3535
3536 list_del_init(&btrfs_inode->ordered_operations);
3537
3538 btrfs_invalidate_inodes(btrfs_inode->root);
3539 }
3540
3541 spin_unlock(&root->fs_info->ordered_extent_lock);
3542 mutex_unlock(&root->fs_info->ordered_operations_mutex);
liuboacce9522011-01-06 19:30:25 +08003543}
3544
Jeff Mahoney143bede2012-03-01 14:56:26 +01003545static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003546{
3547 struct list_head splice;
3548 struct btrfs_ordered_extent *ordered;
3549 struct inode *inode;
3550
3551 INIT_LIST_HEAD(&splice);
3552
3553 spin_lock(&root->fs_info->ordered_extent_lock);
3554
3555 list_splice_init(&root->fs_info->ordered_extents, &splice);
3556 while (!list_empty(&splice)) {
3557 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
3558 root_extent_list);
3559
3560 list_del_init(&ordered->root_extent_list);
3561 atomic_inc(&ordered->refs);
3562
3563 /* the inode may be getting freed (in sys_unlink path). */
3564 inode = igrab(ordered->inode);
3565
3566 spin_unlock(&root->fs_info->ordered_extent_lock);
3567 if (inode)
3568 iput(inode);
3569
3570 atomic_set(&ordered->refs, 1);
3571 btrfs_put_ordered_extent(ordered);
3572
3573 spin_lock(&root->fs_info->ordered_extent_lock);
3574 }
3575
3576 spin_unlock(&root->fs_info->ordered_extent_lock);
liuboacce9522011-01-06 19:30:25 +08003577}
3578
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003579int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3580 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003581{
3582 struct rb_node *node;
3583 struct btrfs_delayed_ref_root *delayed_refs;
3584 struct btrfs_delayed_ref_node *ref;
3585 int ret = 0;
3586
3587 delayed_refs = &trans->delayed_refs;
3588
3589 spin_lock(&delayed_refs->lock);
3590 if (delayed_refs->num_entries == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003591 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003592 printk(KERN_INFO "delayed_refs has NO entry\n");
3593 return ret;
3594 }
3595
Josef Bacikb939d1a2012-05-31 11:06:33 -04003596 while ((node = rb_first(&delayed_refs->root)) != NULL) {
liuboacce9522011-01-06 19:30:25 +08003597 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
liuboacce9522011-01-06 19:30:25 +08003598
3599 atomic_set(&ref->refs, 1);
3600 if (btrfs_delayed_ref_is_head(ref)) {
3601 struct btrfs_delayed_ref_head *head;
3602
3603 head = btrfs_delayed_node_to_head(ref);
Josef Bacikb939d1a2012-05-31 11:06:33 -04003604 if (!mutex_trylock(&head->mutex)) {
3605 atomic_inc(&ref->refs);
3606 spin_unlock(&delayed_refs->lock);
3607
3608 /* Need to wait for the delayed ref to run */
3609 mutex_lock(&head->mutex);
3610 mutex_unlock(&head->mutex);
3611 btrfs_put_delayed_ref(ref);
3612
Josef Bacike18fca72012-06-18 07:23:18 -06003613 spin_lock(&delayed_refs->lock);
Josef Bacikb939d1a2012-05-31 11:06:33 -04003614 continue;
3615 }
3616
liuboacce9522011-01-06 19:30:25 +08003617 kfree(head->extent_op);
3618 delayed_refs->num_heads--;
3619 if (list_empty(&head->cluster))
3620 delayed_refs->num_heads_ready--;
3621 list_del_init(&head->cluster);
liuboacce9522011-01-06 19:30:25 +08003622 }
Josef Bacikb939d1a2012-05-31 11:06:33 -04003623 ref->in_tree = 0;
3624 rb_erase(&ref->rb_node, &delayed_refs->root);
3625 delayed_refs->num_entries--;
3626
liuboacce9522011-01-06 19:30:25 +08003627 spin_unlock(&delayed_refs->lock);
3628 btrfs_put_delayed_ref(ref);
3629
3630 cond_resched();
3631 spin_lock(&delayed_refs->lock);
3632 }
3633
3634 spin_unlock(&delayed_refs->lock);
3635
3636 return ret;
3637}
3638
Jeff Mahoney143bede2012-03-01 14:56:26 +01003639static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
liuboacce9522011-01-06 19:30:25 +08003640{
3641 struct btrfs_pending_snapshot *snapshot;
3642 struct list_head splice;
3643
3644 INIT_LIST_HEAD(&splice);
3645
3646 list_splice_init(&t->pending_snapshots, &splice);
3647
3648 while (!list_empty(&splice)) {
3649 snapshot = list_entry(splice.next,
3650 struct btrfs_pending_snapshot,
3651 list);
3652
3653 list_del_init(&snapshot->list);
3654
3655 kfree(snapshot);
3656 }
liuboacce9522011-01-06 19:30:25 +08003657}
3658
Jeff Mahoney143bede2012-03-01 14:56:26 +01003659static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003660{
3661 struct btrfs_inode *btrfs_inode;
3662 struct list_head splice;
3663
3664 INIT_LIST_HEAD(&splice);
3665
liuboacce9522011-01-06 19:30:25 +08003666 spin_lock(&root->fs_info->delalloc_lock);
David Sterba5be76752011-06-09 10:02:51 +00003667 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003668
3669 while (!list_empty(&splice)) {
3670 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3671 delalloc_inodes);
3672
3673 list_del_init(&btrfs_inode->delalloc_inodes);
3674
3675 btrfs_invalidate_inodes(btrfs_inode->root);
3676 }
3677
3678 spin_unlock(&root->fs_info->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003679}
3680
3681static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3682 struct extent_io_tree *dirty_pages,
3683 int mark)
3684{
3685 int ret;
3686 struct page *page;
3687 struct inode *btree_inode = root->fs_info->btree_inode;
3688 struct extent_buffer *eb;
3689 u64 start = 0;
3690 u64 end;
3691 u64 offset;
3692 unsigned long index;
3693
3694 while (1) {
3695 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04003696 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08003697 if (ret)
3698 break;
3699
3700 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3701 while (start <= end) {
3702 index = start >> PAGE_CACHE_SHIFT;
3703 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
3704 page = find_get_page(btree_inode->i_mapping, index);
3705 if (!page)
3706 continue;
3707 offset = page_offset(page);
3708
3709 spin_lock(&dirty_pages->buffer_lock);
3710 eb = radix_tree_lookup(
3711 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
3712 offset >> PAGE_CACHE_SHIFT);
3713 spin_unlock(&dirty_pages->buffer_lock);
Josef Bacikee670f0a2012-05-31 15:54:30 -04003714 if (eb)
liuboacce9522011-01-06 19:30:25 +08003715 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3716 &eb->bflags);
liuboacce9522011-01-06 19:30:25 +08003717 if (PageWriteback(page))
3718 end_page_writeback(page);
3719
3720 lock_page(page);
3721 if (PageDirty(page)) {
3722 clear_page_dirty_for_io(page);
3723 spin_lock_irq(&page->mapping->tree_lock);
3724 radix_tree_tag_clear(&page->mapping->page_tree,
3725 page_index(page),
3726 PAGECACHE_TAG_DIRTY);
3727 spin_unlock_irq(&page->mapping->tree_lock);
3728 }
3729
liuboacce9522011-01-06 19:30:25 +08003730 unlock_page(page);
Josef Bacikee670f0a2012-05-31 15:54:30 -04003731 page_cache_release(page);
liuboacce9522011-01-06 19:30:25 +08003732 }
3733 }
3734
3735 return ret;
3736}
3737
3738static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3739 struct extent_io_tree *pinned_extents)
3740{
3741 struct extent_io_tree *unpin;
3742 u64 start;
3743 u64 end;
3744 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06003745 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08003746
3747 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06003748again:
liuboacce9522011-01-06 19:30:25 +08003749 while (1) {
3750 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04003751 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08003752 if (ret)
3753 break;
3754
3755 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00003756 if (btrfs_test_opt(root, DISCARD))
3757 ret = btrfs_error_discard_extent(root, start,
3758 end + 1 - start,
3759 NULL);
liuboacce9522011-01-06 19:30:25 +08003760
3761 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3762 btrfs_error_unpin_extent_range(root, start, end);
3763 cond_resched();
3764 }
3765
Liu Boed0eaa12012-06-14 02:23:21 -06003766 if (loop) {
3767 if (unpin == &root->fs_info->freed_extents[0])
3768 unpin = &root->fs_info->freed_extents[1];
3769 else
3770 unpin = &root->fs_info->freed_extents[0];
3771 loop = false;
3772 goto again;
3773 }
3774
liuboacce9522011-01-06 19:30:25 +08003775 return 0;
3776}
3777
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003778void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3779 struct btrfs_root *root)
3780{
3781 btrfs_destroy_delayed_refs(cur_trans, root);
3782 btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
3783 cur_trans->dirty_pages.dirty_bytes);
3784
3785 /* FIXME: cleanup wait for commit */
3786 cur_trans->in_commit = 1;
3787 cur_trans->blocked = 1;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003788 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003789
3790 cur_trans->blocked = 0;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003791 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003792
3793 cur_trans->commit_done = 1;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003794 wake_up(&cur_trans->commit_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003795
Miao Xie67cde342012-06-14 02:23:22 -06003796 btrfs_destroy_delayed_inodes(root);
3797 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003798
3799 btrfs_destroy_pending_snapshots(cur_trans);
3800
3801 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
3802 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06003803 btrfs_destroy_pinned_extent(root,
3804 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003805
3806 /*
3807 memset(cur_trans, 0, sizeof(*cur_trans));
3808 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3809 */
3810}
3811
3812int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003813{
3814 struct btrfs_transaction *t;
3815 LIST_HEAD(list);
3816
liuboacce9522011-01-06 19:30:25 +08003817 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3818
Josef Bacika4abeea2011-04-11 17:25:13 -04003819 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003820 list_splice_init(&root->fs_info->trans_list, &list);
Josef Bacika4abeea2011-04-11 17:25:13 -04003821 root->fs_info->trans_no_join = 1;
3822 spin_unlock(&root->fs_info->trans_lock);
3823
liuboacce9522011-01-06 19:30:25 +08003824 while (!list_empty(&list)) {
3825 t = list_entry(list.next, struct btrfs_transaction, list);
3826 if (!t)
3827 break;
3828
3829 btrfs_destroy_ordered_operations(root);
3830
3831 btrfs_destroy_ordered_extents(root);
3832
3833 btrfs_destroy_delayed_refs(t, root);
3834
3835 btrfs_block_rsv_release(root,
3836 &root->fs_info->trans_block_rsv,
3837 t->dirty_pages.dirty_bytes);
3838
3839 /* FIXME: cleanup wait for commit */
3840 t->in_commit = 1;
3841 t->blocked = 1;
Josef Bacik66657b32012-08-01 15:36:24 -04003842 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003843 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3844 wake_up(&root->fs_info->transaction_blocked_wait);
3845
3846 t->blocked = 0;
Josef Bacik66657b32012-08-01 15:36:24 -04003847 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003848 if (waitqueue_active(&root->fs_info->transaction_wait))
3849 wake_up(&root->fs_info->transaction_wait);
liuboacce9522011-01-06 19:30:25 +08003850
liuboacce9522011-01-06 19:30:25 +08003851 t->commit_done = 1;
Josef Bacik66657b32012-08-01 15:36:24 -04003852 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003853 if (waitqueue_active(&t->commit_wait))
3854 wake_up(&t->commit_wait);
liuboacce9522011-01-06 19:30:25 +08003855
Miao Xie67cde342012-06-14 02:23:22 -06003856 btrfs_destroy_delayed_inodes(root);
3857 btrfs_assert_delayed_root_empty(root);
3858
liuboacce9522011-01-06 19:30:25 +08003859 btrfs_destroy_pending_snapshots(t);
3860
3861 btrfs_destroy_delalloc_inodes(root);
3862
Josef Bacika4abeea2011-04-11 17:25:13 -04003863 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003864 root->fs_info->running_transaction = NULL;
Josef Bacika4abeea2011-04-11 17:25:13 -04003865 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003866
3867 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3868 EXTENT_DIRTY);
3869
3870 btrfs_destroy_pinned_extent(root,
3871 root->fs_info->pinned_extents);
3872
Josef Bacik13c5a932011-04-11 15:45:29 -04003873 atomic_set(&t->use_count, 0);
liuboacce9522011-01-06 19:30:25 +08003874 list_del_init(&t->list);
3875 memset(t, 0, sizeof(*t));
3876 kmem_cache_free(btrfs_transaction_cachep, t);
3877 }
3878
Josef Bacika4abeea2011-04-11 17:25:13 -04003879 spin_lock(&root->fs_info->trans_lock);
3880 root->fs_info->trans_no_join = 0;
3881 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003882 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08003883
3884 return 0;
3885}
3886
Chris Masond1310b22008-01-24 16:13:08 -05003887static struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04003888 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02003889 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04003890 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04003891 /* note we're sharing with inode.c for the merge bio hook */
3892 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05003893};