blob: 8054f7ccf46a49c371b5cf524acee513c16f6160 [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"
Chris Masoneb60cea2007-02-02 09:18:22 -050048
Chris Masond1310b22008-01-24 16:13:08 -050049static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040050static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040051static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010052static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080053 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010054static void btrfs_destroy_ordered_operations(struct btrfs_root *root);
55static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080056static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
57 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010058static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
59static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080060static int btrfs_destroy_marked_extents(struct btrfs_root *root,
61 struct extent_io_tree *dirty_pages,
62 int mark);
63static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
64 struct extent_io_tree *pinned_extents);
Chris Masonce9adaa2008-04-09 16:28:12 -040065
Chris Masond352ac62008-09-29 15:18:18 -040066/*
67 * end_io_wq structs are used to do processing in task context when an IO is
68 * complete. This is used during reads to verify checksums, and it is used
69 * by writes to insert metadata for new file extents after IO is complete.
70 */
Chris Masonce9adaa2008-04-09 16:28:12 -040071struct end_io_wq {
72 struct bio *bio;
73 bio_end_io_t *end_io;
74 void *private;
75 struct btrfs_fs_info *info;
76 int error;
Chris Mason22c59942008-04-09 16:28:12 -040077 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040078 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040079 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040080};
Chris Mason0da54682007-11-07 21:08:01 -050081
Chris Masond352ac62008-09-29 15:18:18 -040082/*
83 * async submit bios are used to offload expensive checksumming
84 * onto the worker threads. They checksum file and metadata bios
85 * just before they are sent down the IO stack.
86 */
Chris Mason44b8bd72008-04-16 11:14:51 -040087struct async_submit_bio {
88 struct inode *inode;
89 struct bio *bio;
90 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -050091 extent_submit_bio_hook_t *submit_bio_start;
92 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -040093 int rw;
94 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -040095 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -040096 /*
97 * bio_offset is optional, can be used if the pages in the bio
98 * can't tell us where in the file the bio should go
99 */
100 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400101 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100102 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400103};
104
Chris Mason85d4e462011-07-26 16:11:19 -0400105/*
106 * Lockdep class keys for extent_buffer->lock's in this root. For a given
107 * eb, the lockdep key is determined by the btrfs_root it belongs to and
108 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500109 *
Chris Mason85d4e462011-07-26 16:11:19 -0400110 * Different roots are used for different purposes and may nest inside each
111 * other and they require separate keysets. As lockdep keys should be
112 * static, assign keysets according to the purpose of the root as indicated
113 * by btrfs_root->objectid. This ensures that all special purpose roots
114 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500115 *
Chris Mason85d4e462011-07-26 16:11:19 -0400116 * Lock-nesting across peer nodes is always done with the immediate parent
117 * node locked thus preventing deadlock. As lockdep doesn't know this, use
118 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500119 *
Chris Mason85d4e462011-07-26 16:11:19 -0400120 * The key is set by the readpage_end_io_hook after the buffer has passed
121 * csum validation but before the pages are unlocked. It is also set by
122 * btrfs_init_new_buffer on freshly allocated blocks.
123 *
124 * We also add a check to make sure the highest level of the tree is the
125 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
126 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500127 */
128#ifdef CONFIG_DEBUG_LOCK_ALLOC
129# if BTRFS_MAX_LEVEL != 8
130# error
131# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400132
133static struct btrfs_lockdep_keyset {
134 u64 id; /* root objectid */
135 const char *name_stem; /* lock name stem */
136 char names[BTRFS_MAX_LEVEL + 1][20];
137 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
138} btrfs_lockdep_keysets[] = {
139 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
140 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
141 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
142 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
143 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
144 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
145 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
146 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
147 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
148 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
149 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500150};
Chris Mason85d4e462011-07-26 16:11:19 -0400151
152void __init btrfs_init_lockdep(void)
153{
154 int i, j;
155
156 /* initialize lockdep class names */
157 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
158 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
159
160 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
161 snprintf(ks->names[j], sizeof(ks->names[j]),
162 "btrfs-%s-%02d", ks->name_stem, j);
163 }
164}
165
166void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
167 int level)
168{
169 struct btrfs_lockdep_keyset *ks;
170
171 BUG_ON(level >= ARRAY_SIZE(ks->keys));
172
173 /* find the matching keyset, id 0 is the default entry */
174 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
175 if (ks->id == objectid)
176 break;
177
178 lockdep_set_class_and_name(&eb->lock,
179 &ks->keys[level], ks->names[level]);
180}
181
Chris Mason4008c042009-02-12 14:09:45 -0500182#endif
183
Chris Masond352ac62008-09-29 15:18:18 -0400184/*
185 * extents on the btree inode are pretty simple, there's one extent
186 * that covers the entire device
187 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500188static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200189 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500190 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400191{
192 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
193 struct extent_map *em;
194 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400195
Chris Mason890871b2009-09-02 16:24:52 -0400196 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500197 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400198 if (em) {
199 em->bdev =
200 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400201 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400202 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400203 }
Chris Mason890871b2009-09-02 16:24:52 -0400204 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400205
David Sterba172ddd62011-04-21 00:48:27 +0200206 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400207 if (!em) {
208 em = ERR_PTR(-ENOMEM);
209 goto out;
210 }
211 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400212 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400213 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400214 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400215 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500216
Chris Mason890871b2009-09-02 16:24:52 -0400217 write_lock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400218 ret = add_extent_mapping(em_tree, em);
219 if (ret == -EEXIST) {
Chris Mason0afbaf82008-04-18 14:17:20 -0400220 u64 failed_start = em->start;
221 u64 failed_len = em->len;
222
Chris Mason5f39d392007-10-15 16:14:19 -0400223 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400224 em = lookup_extent_mapping(em_tree, start, len);
Chris Mason0afbaf82008-04-18 14:17:20 -0400225 if (em) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400226 ret = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400227 } else {
228 em = lookup_extent_mapping(em_tree, failed_start,
229 failed_len);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400230 ret = -EIO;
Chris Mason0afbaf82008-04-18 14:17:20 -0400231 }
Chris Mason5f39d392007-10-15 16:14:19 -0400232 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400233 free_extent_map(em);
234 em = NULL;
Chris Mason5f39d392007-10-15 16:14:19 -0400235 }
Chris Mason890871b2009-09-02 16:24:52 -0400236 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400237
238 if (ret)
239 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400240out:
241 return em;
242}
243
Chris Mason19c00dd2007-10-15 16:19:22 -0400244u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
245{
David Woodhouse163e7832009-04-19 13:02:41 +0100246 return crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400247}
248
249void btrfs_csum_final(u32 crc, char *result)
250{
David Sterba7e75bf32011-03-18 22:56:43 +0000251 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400252}
253
Chris Masond352ac62008-09-29 15:18:18 -0400254/*
255 * compute the csum for a btree block, and either verify it or write it
256 * into the csum field of the block.
257 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400258static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
259 int verify)
260{
David Sterba6c417612011-04-13 15:41:04 +0200261 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500262 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400263 unsigned long len;
264 unsigned long cur_len;
265 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400266 char *kaddr;
267 unsigned long map_start;
268 unsigned long map_len;
269 int err;
270 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500271 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400272
273 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500274 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400275 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400276 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500277 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400278 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400279 cur_len = min(len, map_len - (offset - map_start));
280 crc = btrfs_csum_data(root, kaddr + offset - map_start,
281 crc, cur_len);
282 len -= cur_len;
283 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400284 }
Josef Bacik607d4322008-12-02 07:17:45 -0500285 if (csum_size > sizeof(inline_result)) {
286 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
287 if (!result)
288 return 1;
289 } else {
290 result = (char *)&inline_result;
291 }
292
Chris Mason19c00dd2007-10-15 16:19:22 -0400293 btrfs_csum_final(crc, result);
294
295 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500296 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500297 u32 val;
298 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500299 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500300
Josef Bacik607d4322008-12-02 07:17:45 -0500301 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba7a36dde2011-05-06 15:33:15 +0200302 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
Chris Mason193f2842009-04-27 07:29:05 -0400303 "failed on %llu wanted %X found %X "
304 "level %d\n",
305 root->fs_info->sb->s_id,
306 (unsigned long long)buf->start, val, found,
307 btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500308 if (result != (char *)&inline_result)
309 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400310 return 1;
311 }
312 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500313 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400314 }
Josef Bacik607d4322008-12-02 07:17:45 -0500315 if (result != (char *)&inline_result)
316 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400317 return 0;
318}
319
Chris Masond352ac62008-09-29 15:18:18 -0400320/*
321 * we can't consider a given block up to date unless the transid of the
322 * block matches the transid in the parent node's pointer. This is how we
323 * detect blocks that either didn't get written at all or got written
324 * in the wrong place.
325 */
Chris Mason1259ab72008-05-12 13:39:03 -0400326static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400327 struct extent_buffer *eb, u64 parent_transid,
328 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400329{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000330 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400331 int ret;
332
333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 return 0;
335
Chris Masonb9fab912012-05-06 07:23:47 -0400336 if (atomic)
337 return -EAGAIN;
338
Josef Bacik2ac55d42010-02-03 19:33:23 +0000339 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100340 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400341 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400342 btrfs_header_generation(eb) == parent_transid) {
343 ret = 0;
344 goto out;
345 }
David Sterba7a36dde2011-05-06 15:33:15 +0200346 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
Chris Mason193f2842009-04-27 07:29:05 -0400347 "found %llu\n",
348 (unsigned long long)eb->start,
349 (unsigned long long)parent_transid,
350 (unsigned long long)btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400351 ret = 1;
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400352 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400353out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000354 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
355 &cached_state, GFP_NOFS);
Chris Mason1259ab72008-05-12 13:39:03 -0400356 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400357}
358
Chris Masond352ac62008-09-29 15:18:18 -0400359/*
360 * helper to read a given tree block, doing retries as required when
361 * the checksums don't match and we have alternate mirrors to try.
362 */
Chris Masonf1885912008-04-09 16:28:12 -0400363static int btree_read_extent_buffer_pages(struct btrfs_root *root,
364 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400365 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400366{
367 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400368 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400369 int ret;
370 int num_copies = 0;
371 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400372 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400373
Josef Bacika826d6d2011-03-16 13:42:43 -0400374 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400375 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
376 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200377 ret = read_extent_buffer_pages(io_tree, eb, start,
378 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400379 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600380 if (!ret) {
381 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400382 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600383 break;
384 else
385 ret = -EIO;
386 }
Chris Masond3977122009-01-05 21:25:51 -0500387
Josef Bacika826d6d2011-03-16 13:42:43 -0400388 /*
389 * This buffer's crc is fine, but its contents are corrupted, so
390 * there is no reason to read the other copies, they won't be
391 * any less wrong.
392 */
393 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400394 break;
395
Chris Masonf1885912008-04-09 16:28:12 -0400396 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
397 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400398 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400399 break;
Chris Mason42352982008-04-28 16:40:52 -0400400
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400401 if (!failed_mirror) {
402 failed = 1;
403 failed_mirror = eb->read_mirror;
404 }
405
Chris Masonf1885912008-04-09 16:28:12 -0400406 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400407 if (mirror_num == failed_mirror)
408 mirror_num++;
409
Chris Mason42352982008-04-28 16:40:52 -0400410 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400411 break;
Chris Masonf1885912008-04-09 16:28:12 -0400412 }
Josef Bacikea466792012-03-26 21:57:36 -0400413
Stefan Behrensc0901582012-07-10 07:30:17 -0600414 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400415 repair_eb_io_failure(root, eb, failed_mirror);
416
417 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400418}
Chris Mason19c00dd2007-10-15 16:19:22 -0400419
Chris Masond352ac62008-09-29 15:18:18 -0400420/*
Chris Masond3977122009-01-05 21:25:51 -0500421 * checksum a dirty tree block before IO. This has extra checks to make sure
422 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400423 */
Chris Masond3977122009-01-05 21:25:51 -0500424
Christoph Hellwigb2950862008-12-02 09:54:17 -0500425static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400426{
Chris Masond1310b22008-01-24 16:13:08 -0500427 struct extent_io_tree *tree;
Chris Mason35ebb932007-10-30 16:56:53 -0400428 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
Chris Mason19c00dd2007-10-15 16:19:22 -0400429 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400430 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400431
Chris Masond1310b22008-01-24 16:13:08 -0500432 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason19c00dd2007-10-15 16:19:22 -0400433
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500434 eb = (struct extent_buffer *)page->private;
435 if (page != eb->pages[0])
436 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400437 found_start = btrfs_header_bytenr(eb);
438 if (found_start != start) {
Chris Mason55c69072008-01-09 15:55:33 -0500439 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500440 return 0;
Chris Mason55c69072008-01-09 15:55:33 -0500441 }
Chris Mason727011e2010-08-06 13:21:20 -0400442 if (eb->pages[0] != page) {
Chris Mason55c69072008-01-09 15:55:33 -0500443 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500444 return 0;
Chris Mason55c69072008-01-09 15:55:33 -0500445 }
446 if (!PageUptodate(page)) {
Chris Mason55c69072008-01-09 15:55:33 -0500447 WARN_ON(1);
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500448 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400449 }
Chris Mason19c00dd2007-10-15 16:19:22 -0400450 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400451 return 0;
452}
453
Yan Zheng2b820322008-11-17 21:11:30 -0500454static int check_tree_block_fsid(struct btrfs_root *root,
455 struct extent_buffer *eb)
456{
457 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
458 u8 fsid[BTRFS_UUID_SIZE];
459 int ret = 1;
460
461 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
462 BTRFS_FSID_SIZE);
463 while (fs_devices) {
464 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
465 ret = 0;
466 break;
467 }
468 fs_devices = fs_devices->seed;
469 }
470 return ret;
471}
472
Josef Bacika826d6d2011-03-16 13:42:43 -0400473#define CORRUPT(reason, eb, root, slot) \
474 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
475 "root=%llu, slot=%d\n", reason, \
476 (unsigned long long)btrfs_header_bytenr(eb), \
477 (unsigned long long)root->objectid, slot)
478
479static noinline int check_leaf(struct btrfs_root *root,
480 struct extent_buffer *leaf)
481{
482 struct btrfs_key key;
483 struct btrfs_key leaf_key;
484 u32 nritems = btrfs_header_nritems(leaf);
485 int slot;
486
487 if (nritems == 0)
488 return 0;
489
490 /* Check the 0 item */
491 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
492 BTRFS_LEAF_DATA_SIZE(root)) {
493 CORRUPT("invalid item offset size pair", leaf, root, 0);
494 return -EIO;
495 }
496
497 /*
498 * Check to make sure each items keys are in the correct order and their
499 * offsets make sense. We only have to loop through nritems-1 because
500 * we check the current slot against the next slot, which verifies the
501 * next slot's offset+size makes sense and that the current's slot
502 * offset is correct.
503 */
504 for (slot = 0; slot < nritems - 1; slot++) {
505 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
506 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
507
508 /* Make sure the keys are in the right order */
509 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
510 CORRUPT("bad key order", leaf, root, slot);
511 return -EIO;
512 }
513
514 /*
515 * Make sure the offset and ends are right, remember that the
516 * item data starts at the end of the leaf and grows towards the
517 * front.
518 */
519 if (btrfs_item_offset_nr(leaf, slot) !=
520 btrfs_item_end_nr(leaf, slot + 1)) {
521 CORRUPT("slot offset bad", leaf, root, slot);
522 return -EIO;
523 }
524
525 /*
526 * Check to make sure that we don't point outside of the leaf,
527 * just incase all the items are consistent to eachother, but
528 * all point outside of the leaf.
529 */
530 if (btrfs_item_end_nr(leaf, slot) >
531 BTRFS_LEAF_DATA_SIZE(root)) {
532 CORRUPT("slot end outside of leaf", leaf, root, slot);
533 return -EIO;
534 }
535 }
536
537 return 0;
538}
539
Chris Mason727011e2010-08-06 13:21:20 -0400540struct extent_buffer *find_eb_for_page(struct extent_io_tree *tree,
541 struct page *page, int max_walk)
542{
543 struct extent_buffer *eb;
544 u64 start = page_offset(page);
545 u64 target = start;
546 u64 min_start;
547
548 if (start < max_walk)
549 min_start = 0;
550 else
551 min_start = start - max_walk;
552
553 while (start >= min_start) {
554 eb = find_extent_buffer(tree, start, 0);
555 if (eb) {
556 /*
557 * we found an extent buffer and it contains our page
558 * horray!
559 */
560 if (eb->start <= target &&
561 eb->start + eb->len > target)
562 return eb;
563
564 /* we found an extent buffer that wasn't for us */
565 free_extent_buffer(eb);
566 return NULL;
567 }
568 if (start == 0)
569 break;
570 start -= PAGE_CACHE_SIZE;
571 }
572 return NULL;
573}
574
Christoph Hellwigb2950862008-12-02 09:54:17 -0500575static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400576 struct extent_state *state, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400577{
578 struct extent_io_tree *tree;
579 u64 found_start;
580 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400581 struct extent_buffer *eb;
582 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400583 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400584 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400585
Chris Masonce9adaa2008-04-09 16:28:12 -0400586 if (!page->private)
587 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500588
Chris Mason727011e2010-08-06 13:21:20 -0400589 tree = &BTRFS_I(page->mapping->host)->io_tree;
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500590 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500591
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400592 /* the pending IO might have been the only thing that kept this buffer
593 * in memory. Make sure we have a ref for all this other checks
594 */
595 extent_buffer_get(eb);
596
597 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400598 if (!reads_done)
599 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400600
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400601 eb->read_mirror = mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400602 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
603 ret = -EIO;
604 goto err;
605 }
606
Chris Masonce9adaa2008-04-09 16:28:12 -0400607 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400608 if (found_start != eb->start) {
David Sterba7a36dde2011-05-06 15:33:15 +0200609 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400610 "%llu %llu\n",
611 (unsigned long long)found_start,
612 (unsigned long long)eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400613 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400614 goto err;
615 }
Yan Zheng2b820322008-11-17 21:11:30 -0500616 if (check_tree_block_fsid(root, eb)) {
David Sterba7a36dde2011-05-06 15:33:15 +0200617 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
Chris Mason193f2842009-04-27 07:29:05 -0400618 (unsigned long long)eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400619 ret = -EIO;
620 goto err;
621 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400622 found_level = btrfs_header_level(eb);
623
Chris Mason85d4e462011-07-26 16:11:19 -0400624 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
625 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500626
Chris Masonce9adaa2008-04-09 16:28:12 -0400627 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400628 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400629 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400630 goto err;
631 }
632
633 /*
634 * If this is a leaf block and it is corrupt, set the corrupt bit so
635 * that we don't try and read the other copies of this block, just
636 * return -EIO.
637 */
638 if (found_level == 0 && check_leaf(root, eb)) {
639 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
640 ret = -EIO;
641 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400642
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400643 if (!ret)
644 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400645err:
Arne Jansen4bb31e92011-06-10 13:55:54 +0200646 if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
647 clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
648 btree_readahead_hook(root, eb, eb->start, ret);
649 }
650
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400651 if (ret)
652 clear_extent_buffer_uptodate(eb);
653 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400654out:
Chris Masonf1885912008-04-09 16:28:12 -0400655 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400656}
657
Josef Bacikea466792012-03-26 21:57:36 -0400658static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200659{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200660 struct extent_buffer *eb;
661 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
662
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500663 eb = (struct extent_buffer *)page->private;
Josef Bacikea466792012-03-26 21:57:36 -0400664 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400665 eb->read_mirror = failed_mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400666 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200667 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200668 return -EIO; /* we fixed nothing */
669}
670
Chris Masonce9adaa2008-04-09 16:28:12 -0400671static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400672{
673 struct end_io_wq *end_io_wq = bio->bi_private;
674 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400675
Chris Masonce9adaa2008-04-09 16:28:12 -0400676 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400677 end_io_wq->error = err;
Chris Mason8b712842008-06-11 16:50:36 -0400678 end_io_wq->work.func = end_workqueue_fn;
679 end_io_wq->work.flags = 0;
Chris Masond20f7042008-12-08 16:58:54 -0500680
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200681 if (bio->bi_rw & REQ_WRITE) {
Josef Bacik0cb59c92010-07-02 12:14:14 -0400682 if (end_io_wq->metadata == 1)
Chris Masoncad321a2008-12-17 14:51:42 -0500683 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
684 &end_io_wq->work);
Josef Bacik0cb59c92010-07-02 12:14:14 -0400685 else if (end_io_wq->metadata == 2)
686 btrfs_queue_worker(&fs_info->endio_freespace_worker,
687 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500688 else
689 btrfs_queue_worker(&fs_info->endio_write_workers,
690 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500691 } else {
692 if (end_io_wq->metadata)
693 btrfs_queue_worker(&fs_info->endio_meta_workers,
694 &end_io_wq->work);
695 else
696 btrfs_queue_worker(&fs_info->endio_workers,
697 &end_io_wq->work);
698 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400699}
700
Josef Bacik0cb59c92010-07-02 12:14:14 -0400701/*
702 * For the metadata arg you want
703 *
704 * 0 - if data
705 * 1 - if normal metadta
706 * 2 - if writing to the free space cache area
707 */
Chris Mason22c59942008-04-09 16:28:12 -0400708int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
709 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400710{
Chris Masonce9adaa2008-04-09 16:28:12 -0400711 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400712 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
713 if (!end_io_wq)
714 return -ENOMEM;
715
716 end_io_wq->private = bio->bi_private;
717 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400718 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400719 end_io_wq->error = 0;
720 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400721 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400722
723 bio->bi_private = end_io_wq;
724 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400725 return 0;
726}
727
Chris Masonb64a2852008-08-20 13:39:41 -0400728unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400729{
730 unsigned long limit = min_t(unsigned long,
731 info->workers.max_workers,
732 info->fs_devices->open_devices);
733 return 256 * limit;
734}
735
Chris Mason4a69a412008-11-06 22:03:00 -0500736static void run_one_async_start(struct btrfs_work *work)
737{
Chris Mason4a69a412008-11-06 22:03:00 -0500738 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100739 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500740
741 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100742 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
743 async->mirror_num, async->bio_flags,
744 async->bio_offset);
745 if (ret)
746 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500747}
748
749static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400750{
751 struct btrfs_fs_info *fs_info;
752 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400753 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400754
755 async = container_of(work, struct async_submit_bio, work);
756 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400757
Chris Masonb64a2852008-08-20 13:39:41 -0400758 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400759 limit = limit * 2 / 3;
760
Josef Bacik66657b32012-08-01 15:36:24 -0400761 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400762 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400763 wake_up(&fs_info->async_submit_wait);
764
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100765 /* If an error occured we just want to clean up the bio and move on */
766 if (async->error) {
767 bio_endio(async->bio, async->error);
768 return;
769 }
770
Chris Mason4a69a412008-11-06 22:03:00 -0500771 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400772 async->mirror_num, async->bio_flags,
773 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500774}
775
776static void run_one_async_free(struct btrfs_work *work)
777{
778 struct async_submit_bio *async;
779
780 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400781 kfree(async);
782}
783
Chris Mason44b8bd72008-04-16 11:14:51 -0400784int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
785 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400786 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400787 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500788 extent_submit_bio_hook_t *submit_bio_start,
789 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400790{
791 struct async_submit_bio *async;
792
793 async = kmalloc(sizeof(*async), GFP_NOFS);
794 if (!async)
795 return -ENOMEM;
796
797 async->inode = inode;
798 async->rw = rw;
799 async->bio = bio;
800 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500801 async->submit_bio_start = submit_bio_start;
802 async->submit_bio_done = submit_bio_done;
803
804 async->work.func = run_one_async_start;
805 async->work.ordered_func = run_one_async_done;
806 async->work.ordered_free = run_one_async_free;
807
Chris Mason8b712842008-06-11 16:50:36 -0400808 async->work.flags = 0;
Chris Masonc8b97812008-10-29 14:49:59 -0400809 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400810 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400811
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100812 async->error = 0;
813
Chris Masoncb03c742008-05-15 16:15:45 -0400814 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400815
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200816 if (rw & REQ_SYNC)
Chris Masond313d7a2009-04-20 15:50:09 -0400817 btrfs_set_work_high_prio(&async->work);
818
Chris Mason8b712842008-06-11 16:50:36 -0400819 btrfs_queue_worker(&fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400820
Chris Masond3977122009-01-05 21:25:51 -0500821 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500822 atomic_read(&fs_info->nr_async_submits)) {
823 wait_event(fs_info->async_submit_wait,
824 (atomic_read(&fs_info->nr_async_submits) == 0));
825 }
826
Chris Mason44b8bd72008-04-16 11:14:51 -0400827 return 0;
828}
829
Chris Masonce3ed712008-09-23 13:14:12 -0400830static int btree_csum_one_bio(struct bio *bio)
831{
832 struct bio_vec *bvec = bio->bi_io_vec;
833 int bio_index = 0;
834 struct btrfs_root *root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100835 int ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400836
837 WARN_ON(bio->bi_vcnt <= 0);
Chris Masond3977122009-01-05 21:25:51 -0500838 while (bio_index < bio->bi_vcnt) {
Chris Masonce3ed712008-09-23 13:14:12 -0400839 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100840 ret = csum_dirty_buffer(root, bvec->bv_page);
841 if (ret)
842 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400843 bio_index++;
844 bvec++;
845 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100846 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400847}
848
Chris Mason4a69a412008-11-06 22:03:00 -0500849static int __btree_submit_bio_start(struct inode *inode, int rw,
850 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400851 unsigned long bio_flags,
852 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400853{
Chris Mason8b712842008-06-11 16:50:36 -0400854 /*
855 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400856 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400857 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100858 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500859}
Chris Mason22c59942008-04-09 16:28:12 -0400860
Chris Mason4a69a412008-11-06 22:03:00 -0500861static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400862 int mirror_num, unsigned long bio_flags,
863 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500864{
Chris Mason8b712842008-06-11 16:50:36 -0400865 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500866 * when we're called for a write, we're already in the async
867 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400868 */
Chris Mason8b712842008-06-11 16:50:36 -0400869 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Chris Mason0b86a832008-03-24 15:01:56 -0400870}
871
Chris Mason44b8bd72008-04-16 11:14:51 -0400872static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400873 int mirror_num, unsigned long bio_flags,
874 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400875{
Chris Masoncad321a2008-12-17 14:51:42 -0500876 int ret;
877
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200878 if (!(rw & REQ_WRITE)) {
Chris Masonf3f266a2012-03-23 10:22:46 -0400879
Chris Mason4a69a412008-11-06 22:03:00 -0500880 /*
881 * called for a read, do the setup so that checksum validation
882 * can happen in the async kernel threads
883 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400884 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
885 bio, 1);
Chris Mason1d4284b2012-03-28 20:31:37 -0400886 if (ret)
887 return ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500888 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
Chris Mason6f3577b2008-11-13 09:59:36 -0500889 mirror_num, 0);
Chris Mason44b8bd72008-04-16 11:14:51 -0400890 }
Chris Masond313d7a2009-04-20 15:50:09 -0400891
Chris Masoncad321a2008-12-17 14:51:42 -0500892 /*
893 * kthread helpers are used to submit writes so that checksumming
894 * can happen in parallel across all CPUs
895 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400896 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
Chris Masonc8b97812008-10-29 14:49:59 -0400897 inode, rw, bio, mirror_num, 0,
Chris Masoneaf25d92010-05-25 09:48:28 -0400898 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500899 __btree_submit_bio_start,
900 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400901}
902
Jan Beulich3dd14622010-12-07 14:54:09 +0000903#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500904static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800905 struct page *newpage, struct page *page,
906 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500907{
908 /*
909 * we can't safely write a btree page from here,
910 * we haven't done the locking hook
911 */
912 if (PageDirty(page))
913 return -EAGAIN;
914 /*
915 * Buffers may be managed in a filesystem specific way.
916 * We must have no buffers or drop them.
917 */
918 if (page_has_private(page) &&
919 !try_to_release_page(page, GFP_KERNEL))
920 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800921 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500922}
Jan Beulich3dd14622010-12-07 14:54:09 +0000923#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500924
Chris Mason0da54682007-11-07 21:08:01 -0500925
926static int btree_writepages(struct address_space *mapping,
927 struct writeback_control *wbc)
928{
Chris Masond1310b22008-01-24 16:13:08 -0500929 struct extent_io_tree *tree;
930 tree = &BTRFS_I(mapping->host)->io_tree;
Chris Masond8d5f3e2007-12-11 12:42:00 -0500931 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Masonb9473432009-03-13 11:00:37 -0400932 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
Chris Mason793955b2007-11-26 16:34:41 -0800933 u64 num_dirty;
Chris Mason24ab9cd2008-09-24 14:51:30 -0400934 unsigned long thresh = 32 * 1024 * 1024;
Chris Mason448d6402007-11-27 07:52:01 -0800935
936 if (wbc->for_kupdate)
937 return 0;
938
Chris Masonb9473432009-03-13 11:00:37 -0400939 /* this is a bit racy, but that's ok */
940 num_dirty = root->fs_info->dirty_metadata_bytes;
Chris Masond3977122009-01-05 21:25:51 -0500941 if (num_dirty < thresh)
Chris Mason793955b2007-11-26 16:34:41 -0800942 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800943 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400944 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -0500945}
946
Christoph Hellwigb2950862008-12-02 09:54:17 -0500947static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -0400948{
Chris Masond1310b22008-01-24 16:13:08 -0500949 struct extent_io_tree *tree;
950 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +0200951 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400952}
953
Chris Mason70dec802008-01-29 09:59:12 -0500954static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -0400955{
Chris Mason98509cfc2008-09-11 15:51:43 -0400956 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -0500957 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -0500958 /*
959 * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing
960 * slab allocation from alloc_extent_state down the callchain where
961 * it'd hit a BUG_ON as those flags are not allowed.
962 */
963 gfp_flags &= ~GFP_SLAB_BUG_MASK;
964
Josef Bacik3083ee22012-03-09 16:01:49 -0500965 return try_release_extent_buffer(page, gfp_flags);
Chris Masone20d96d2007-03-22 12:13:20 -0400966}
Chris Mason123abc82007-03-14 14:14:43 -0400967
Chris Mason5f39d392007-10-15 16:14:19 -0400968static void btree_invalidatepage(struct page *page, unsigned long offset)
Chris Masond98237b2007-03-28 13:57:48 -0400969{
Chris Masond1310b22008-01-24 16:13:08 -0500970 struct extent_io_tree *tree;
971 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -0400972 extent_invalidatepage(tree, page, offset);
973 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400974 if (PagePrivate(page)) {
Chris Masond3977122009-01-05 21:25:51 -0500975 printk(KERN_WARNING "btrfs warning page private not zero "
976 "on page %llu\n", (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -0400977 ClearPagePrivate(page);
978 set_page_private(page, 0);
979 page_cache_release(page);
980 }
Chris Masond98237b2007-03-28 13:57:48 -0400981}
982
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400983static int btree_set_page_dirty(struct page *page)
984{
985 struct extent_buffer *eb;
986
987 BUG_ON(!PagePrivate(page));
988 eb = (struct extent_buffer *)page->private;
989 BUG_ON(!eb);
990 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
991 BUG_ON(!atomic_read(&eb->refs));
992 btrfs_assert_tree_locked(eb);
993 return __set_page_dirty_nobuffers(page);
994}
995
Alexey Dobriyan7f094102009-09-21 17:01:10 -0700996static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -0400997 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -0500998 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -0400999 .releasepage = btree_releasepage,
1000 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001001#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001002 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001003#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001004 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001005};
1006
Chris Masonca7a79a2008-05-12 12:59:19 -04001007int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1008 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001009{
Chris Mason5f39d392007-10-15 16:14:19 -04001010 struct extent_buffer *buf = NULL;
1011 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001012 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001013
Chris Masondb945352007-10-15 16:15:53 -04001014 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001015 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001016 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001017 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001018 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001019 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001020 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001021}
1022
Arne Jansenab0fff02011-05-23 14:25:41 +02001023int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1024 int mirror_num, struct extent_buffer **eb)
1025{
1026 struct extent_buffer *buf = NULL;
1027 struct inode *btree_inode = root->fs_info->btree_inode;
1028 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1029 int ret;
1030
1031 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1032 if (!buf)
1033 return 0;
1034
1035 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1036
1037 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1038 btree_get_extent, mirror_num);
1039 if (ret) {
1040 free_extent_buffer(buf);
1041 return ret;
1042 }
1043
1044 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1045 free_extent_buffer(buf);
1046 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001047 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001048 *eb = buf;
1049 } else {
1050 free_extent_buffer(buf);
1051 }
1052 return 0;
1053}
1054
Chris Mason0999df52008-04-01 13:48:14 -04001055struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1056 u64 bytenr, u32 blocksize)
1057{
1058 struct inode *btree_inode = root->fs_info->btree_inode;
1059 struct extent_buffer *eb;
1060 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
David Sterbaf09d1f62011-04-21 01:08:01 +02001061 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001062 return eb;
1063}
1064
1065struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1066 u64 bytenr, u32 blocksize)
1067{
1068 struct inode *btree_inode = root->fs_info->btree_inode;
1069 struct extent_buffer *eb;
1070
1071 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
Chris Mason727011e2010-08-06 13:21:20 -04001072 bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001073 return eb;
1074}
1075
1076
Chris Masone02119d2008-09-05 16:13:11 -04001077int btrfs_write_tree_block(struct extent_buffer *buf)
1078{
Chris Mason727011e2010-08-06 13:21:20 -04001079 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001080 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001081}
1082
1083int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1084{
Chris Mason727011e2010-08-06 13:21:20 -04001085 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001086 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001087}
1088
Chris Masondb945352007-10-15 16:15:53 -04001089struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001090 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001091{
Chris Mason5f39d392007-10-15 16:14:19 -04001092 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001093 int ret;
1094
Chris Masondb945352007-10-15 16:15:53 -04001095 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001096 if (!buf)
1097 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001098
Chris Masonca7a79a2008-05-12 12:59:19 -04001099 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason5f39d392007-10-15 16:14:19 -04001100 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001101
Chris Masoneb60cea2007-02-02 09:18:22 -05001102}
1103
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001104void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1105 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001106{
Chris Mason55c69072008-01-09 15:55:33 -05001107 if (btrfs_header_generation(buf) ==
Chris Mason925baed2008-06-25 16:01:30 -04001108 root->fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001109 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001110
Chris Masonb9473432009-03-13 11:00:37 -04001111 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1112 spin_lock(&root->fs_info->delalloc_lock);
1113 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1114 root->fs_info->dirty_metadata_bytes -= buf->len;
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001115 else {
1116 spin_unlock(&root->fs_info->delalloc_lock);
1117 btrfs_panic(root->fs_info, -EOVERFLOW,
1118 "Can't clear %lu bytes from "
Joe Perches533574c2012-07-30 14:40:13 -07001119 " dirty_mdatadata_bytes (%llu)",
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001120 buf->len,
1121 root->fs_info->dirty_metadata_bytes);
1122 }
Chris Masonb9473432009-03-13 11:00:37 -04001123 spin_unlock(&root->fs_info->delalloc_lock);
1124 }
Chris Masonb4ce94d2009-02-04 09:25:08 -05001125
Chris Masonb9473432009-03-13 11:00:37 -04001126 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1127 btrfs_set_lock_blocking(buf);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001128 clear_extent_buffer_dirty(buf);
Chris Mason925baed2008-06-25 16:01:30 -04001129 }
Chris Mason5f39d392007-10-15 16:14:19 -04001130}
1131
Jeff Mahoney143bede2012-03-01 14:56:26 +01001132static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1133 u32 stripesize, struct btrfs_root *root,
1134 struct btrfs_fs_info *fs_info,
1135 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001136{
Chris Masoncfaa7292007-02-21 17:04:57 -05001137 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001138 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001139 root->sectorsize = sectorsize;
1140 root->nodesize = nodesize;
1141 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001142 root->stripesize = stripesize;
Chris Mason123abc82007-03-14 14:14:43 -04001143 root->ref_cows = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001144 root->track_dirty = 0;
Yan, Zhengc71bf092009-11-12 09:34:40 +00001145 root->in_radix = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001146 root->orphan_item_inserted = 0;
1147 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001148
Chris Mason0f7d52f2007-04-09 10:42:37 -04001149 root->objectid = objectid;
1150 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001151 root->highest_objectid = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001152 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001153 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001154 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001155 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001156 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001157
1158 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001159 INIT_LIST_HEAD(&root->root_list);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001160 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001161 spin_lock_init(&root->inode_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001162 spin_lock_init(&root->accounting_lock);
Chris Masona2135012008-06-25 16:01:30 -04001163 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001164 mutex_init(&root->log_mutex);
Yan Zheng7237f1832009-01-21 12:54:03 -05001165 init_waitqueue_head(&root->log_writer_wait);
1166 init_waitqueue_head(&root->log_commit_wait[0]);
1167 init_waitqueue_head(&root->log_commit_wait[1]);
1168 atomic_set(&root->log_commit[0], 0);
1169 atomic_set(&root->log_commit[1], 0);
1170 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001171 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001172 atomic_set(&root->orphan_inodes, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001173 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001174 root->last_log_commit = 0;
Chris Masond0c803c2008-09-11 16:17:57 -04001175 extent_io_tree_init(&root->dirty_log_pages,
David Sterbaf993c882011-04-20 23:35:57 +02001176 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001177
Chris Mason3768f362007-03-13 16:47:54 -04001178 memset(&root->root_key, 0, sizeof(root->root_key));
1179 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001180 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001181 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Chris Mason3f157a22008-06-25 16:01:31 -04001182 root->defrag_trans_start = fs_info->generation;
Josef Bacik58176a92007-08-29 15:47:34 -04001183 init_completion(&root->kobj_unregister);
Chris Mason6702ed42007-08-07 16:15:09 -04001184 root->defrag_running = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001185 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001186 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001187
1188 spin_lock_init(&root->root_times_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001189}
1190
Jeff Mahoney200a5c12011-10-03 23:22:43 -04001191static int __must_check find_and_setup_root(struct btrfs_root *tree_root,
1192 struct btrfs_fs_info *fs_info,
1193 u64 objectid,
1194 struct btrfs_root *root)
Chris Mason3768f362007-03-13 16:47:54 -04001195{
1196 int ret;
Chris Masondb945352007-10-15 16:15:53 -04001197 u32 blocksize;
Yan Zheng84234f32008-10-29 14:49:05 -04001198 u64 generation;
Chris Mason3768f362007-03-13 16:47:54 -04001199
Chris Masondb945352007-10-15 16:15:53 -04001200 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001201 tree_root->sectorsize, tree_root->stripesize,
1202 root, fs_info, objectid);
Chris Mason3768f362007-03-13 16:47:54 -04001203 ret = btrfs_find_last_root(tree_root, objectid,
1204 &root->root_item, &root->root_key);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001205 if (ret > 0)
1206 return -ENOENT;
Jeff Mahoney200a5c12011-10-03 23:22:43 -04001207 else if (ret < 0)
1208 return ret;
Chris Mason3768f362007-03-13 16:47:54 -04001209
Yan Zheng84234f32008-10-29 14:49:05 -04001210 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001211 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
Chris Masonaf31f5e2011-11-03 15:17:42 -04001212 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001213 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001214 blocksize, generation);
Chris Masonb9fab912012-05-06 07:23:47 -04001215 if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) {
Chris Mason68433b72010-12-13 14:47:58 -05001216 free_extent_buffer(root->node);
Chris Masonaf31f5e2011-11-03 15:17:42 -04001217 root->node = NULL;
Chris Mason68433b72010-12-13 14:47:58 -05001218 return -EIO;
1219 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001220 root->commit_root = btrfs_root_node(root);
Chris Masond97e63b2007-02-20 16:40:44 -05001221 return 0;
1222}
1223
Al Virof84a8bd2011-11-17 15:57:57 -05001224static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001225{
1226 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1227 if (root)
1228 root->fs_info = fs_info;
1229 return root;
1230}
1231
Arne Jansen20897f52011-09-13 12:44:20 +02001232struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1233 struct btrfs_fs_info *fs_info,
1234 u64 objectid)
1235{
1236 struct extent_buffer *leaf;
1237 struct btrfs_root *tree_root = fs_info->tree_root;
1238 struct btrfs_root *root;
1239 struct btrfs_key key;
1240 int ret = 0;
1241 u64 bytenr;
1242
1243 root = btrfs_alloc_root(fs_info);
1244 if (!root)
1245 return ERR_PTR(-ENOMEM);
1246
1247 __setup_root(tree_root->nodesize, tree_root->leafsize,
1248 tree_root->sectorsize, tree_root->stripesize,
1249 root, fs_info, objectid);
1250 root->root_key.objectid = objectid;
1251 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1252 root->root_key.offset = 0;
1253
1254 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1255 0, objectid, NULL, 0, 0, 0);
1256 if (IS_ERR(leaf)) {
1257 ret = PTR_ERR(leaf);
1258 goto fail;
1259 }
1260
1261 bytenr = leaf->start;
1262 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1263 btrfs_set_header_bytenr(leaf, leaf->start);
1264 btrfs_set_header_generation(leaf, trans->transid);
1265 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1266 btrfs_set_header_owner(leaf, objectid);
1267 root->node = leaf;
1268
1269 write_extent_buffer(leaf, fs_info->fsid,
1270 (unsigned long)btrfs_header_fsid(leaf),
1271 BTRFS_FSID_SIZE);
1272 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1273 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
1274 BTRFS_UUID_SIZE);
1275 btrfs_mark_buffer_dirty(leaf);
1276
1277 root->commit_root = btrfs_root_node(root);
1278 root->track_dirty = 1;
1279
1280
1281 root->root_item.flags = 0;
1282 root->root_item.byte_limit = 0;
1283 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1284 btrfs_set_root_generation(&root->root_item, trans->transid);
1285 btrfs_set_root_level(&root->root_item, 0);
1286 btrfs_set_root_refs(&root->root_item, 1);
1287 btrfs_set_root_used(&root->root_item, leaf->len);
1288 btrfs_set_root_last_snapshot(&root->root_item, 0);
1289 btrfs_set_root_dirid(&root->root_item, 0);
1290 root->root_item.drop_level = 0;
1291
1292 key.objectid = objectid;
1293 key.type = BTRFS_ROOT_ITEM_KEY;
1294 key.offset = 0;
1295 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1296 if (ret)
1297 goto fail;
1298
1299 btrfs_tree_unlock(leaf);
1300
1301fail:
1302 if (ret)
1303 return ERR_PTR(ret);
1304
1305 return root;
1306}
1307
Yan Zheng7237f1832009-01-21 12:54:03 -05001308static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1309 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001310{
1311 struct btrfs_root *root;
1312 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f1832009-01-21 12:54:03 -05001313 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001314
Al Viro6f07e422011-11-17 00:46:16 -05001315 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001316 if (!root)
Yan Zheng7237f1832009-01-21 12:54:03 -05001317 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001318
1319 __setup_root(tree_root->nodesize, tree_root->leafsize,
1320 tree_root->sectorsize, tree_root->stripesize,
1321 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1322
1323 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1324 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1325 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Yan Zheng7237f1832009-01-21 12:54:03 -05001326 /*
1327 * log trees do not get reference counted because they go away
1328 * before a real commit is actually done. They do store pointers
1329 * to file data extents, and those reference counts still get
1330 * updated (along with back refs to the log tree).
1331 */
Chris Masone02119d2008-09-05 16:13:11 -04001332 root->ref_cows = 0;
1333
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001334 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
Arne Jansen66d7e7f2011-09-12 15:26:38 +02001335 BTRFS_TREE_LOG_OBJECTID, NULL,
Jan Schmidt5581a512012-05-16 17:04:52 +02001336 0, 0, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001337 if (IS_ERR(leaf)) {
1338 kfree(root);
1339 return ERR_CAST(leaf);
1340 }
Chris Masone02119d2008-09-05 16:13:11 -04001341
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001342 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1343 btrfs_set_header_bytenr(leaf, leaf->start);
1344 btrfs_set_header_generation(leaf, trans->transid);
1345 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1346 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f1832009-01-21 12:54:03 -05001347 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001348
1349 write_extent_buffer(root->node, root->fs_info->fsid,
1350 (unsigned long)btrfs_header_fsid(root->node),
1351 BTRFS_FSID_SIZE);
1352 btrfs_mark_buffer_dirty(root->node);
1353 btrfs_tree_unlock(root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001354 return root;
1355}
1356
1357int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1358 struct btrfs_fs_info *fs_info)
1359{
1360 struct btrfs_root *log_root;
1361
1362 log_root = alloc_log_tree(trans, fs_info);
1363 if (IS_ERR(log_root))
1364 return PTR_ERR(log_root);
1365 WARN_ON(fs_info->log_root_tree);
1366 fs_info->log_root_tree = log_root;
1367 return 0;
1368}
1369
1370int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1371 struct btrfs_root *root)
1372{
1373 struct btrfs_root *log_root;
1374 struct btrfs_inode_item *inode_item;
1375
1376 log_root = alloc_log_tree(trans, root->fs_info);
1377 if (IS_ERR(log_root))
1378 return PTR_ERR(log_root);
1379
1380 log_root->last_trans = trans->transid;
1381 log_root->root_key.offset = root->root_key.objectid;
1382
1383 inode_item = &log_root->root_item.inode;
1384 inode_item->generation = cpu_to_le64(1);
1385 inode_item->size = cpu_to_le64(3);
1386 inode_item->nlink = cpu_to_le32(1);
1387 inode_item->nbytes = cpu_to_le64(root->leafsize);
1388 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1389
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001390 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001391
1392 WARN_ON(root->log_root);
1393 root->log_root = log_root;
1394 root->log_transid = 0;
Chris Mason257c62e2009-10-13 13:21:08 -04001395 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001396 return 0;
1397}
1398
1399struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1400 struct btrfs_key *location)
1401{
1402 struct btrfs_root *root;
1403 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001404 struct btrfs_path *path;
Chris Mason5f39d392007-10-15 16:14:19 -04001405 struct extent_buffer *l;
Yan Zheng84234f32008-10-29 14:49:05 -04001406 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001407 u32 blocksize;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001408 int ret = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001409 int slot;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001410
Al Viro6f07e422011-11-17 00:46:16 -05001411 root = btrfs_alloc_root(fs_info);
Chris Mason0cf6c622007-06-09 09:22:25 -04001412 if (!root)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001413 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001414 if (location->offset == (u64)-1) {
Chris Masondb945352007-10-15 16:15:53 -04001415 ret = find_and_setup_root(tree_root, fs_info,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001416 location->objectid, root);
1417 if (ret) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04001418 kfree(root);
1419 return ERR_PTR(ret);
1420 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001421 goto out;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001422 }
1423
Chris Masondb945352007-10-15 16:15:53 -04001424 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001425 tree_root->sectorsize, tree_root->stripesize,
1426 root, fs_info, location->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001427
1428 path = btrfs_alloc_path();
Tsutomu Itohdb5b4932011-03-23 08:14:16 +00001429 if (!path) {
1430 kfree(root);
1431 return ERR_PTR(-ENOMEM);
1432 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001433 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001434 if (ret == 0) {
1435 l = path->nodes[0];
Alexander Block8ea05e32012-07-25 17:35:53 +02001436 slot = path->slots[0];
1437 btrfs_read_root_item(tree_root, l, slot, &root->root_item);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001438 memcpy(&root->root_key, location, sizeof(*location));
Chris Mason0f7d52f2007-04-09 10:42:37 -04001439 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04001440 btrfs_free_path(path);
1441 if (ret) {
Tsutomu Itoh5e540f72010-12-27 06:53:10 +00001442 kfree(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001443 if (ret > 0)
1444 ret = -ENOENT;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001445 return ERR_PTR(ret);
1446 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001447
Yan Zheng84234f32008-10-29 14:49:05 -04001448 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001449 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1450 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001451 blocksize, generation);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001452 root->commit_root = btrfs_root_node(root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001453 BUG_ON(!root->node); /* -ENOMEM */
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001454out:
Li Zefan08fe4db2011-03-28 02:01:25 +00001455 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
Chris Masone02119d2008-09-05 16:13:11 -04001456 root->ref_cows = 1;
Li Zefan08fe4db2011-03-28 02:01:25 +00001457 btrfs_check_and_init_root_item(&root->root_item);
1458 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001459
Chris Mason5eda7b52007-06-22 14:16:25 -04001460 return root;
1461}
1462
Chris Masonedbd8d42007-12-21 16:27:24 -05001463struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1464 struct btrfs_key *location)
Chris Mason5eda7b52007-06-22 14:16:25 -04001465{
1466 struct btrfs_root *root;
1467 int ret;
1468
Chris Masonedbd8d42007-12-21 16:27:24 -05001469 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1470 return fs_info->tree_root;
1471 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1472 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001473 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1474 return fs_info->chunk_root;
1475 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1476 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001477 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1478 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001479 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1480 return fs_info->quota_root ? fs_info->quota_root :
1481 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001482again:
1483 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001484 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1485 (unsigned long)location->objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001486 spin_unlock(&fs_info->fs_roots_radix_lock);
Chris Mason5eda7b52007-06-22 14:16:25 -04001487 if (root)
1488 return root;
1489
Chris Masone02119d2008-09-05 16:13:11 -04001490 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001491 if (IS_ERR(root))
1492 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001493
Li Zefan581bb052011-04-20 10:06:11 +08001494 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
Li Zefan581bb052011-04-20 10:06:11 +08001495 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1496 GFP_NOFS);
David Sterba35a30d72011-06-13 15:18:23 +00001497 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1498 ret = -ENOMEM;
Li Zefan581bb052011-04-20 10:06:11 +08001499 goto fail;
David Sterba35a30d72011-06-13 15:18:23 +00001500 }
Li Zefan581bb052011-04-20 10:06:11 +08001501
1502 btrfs_init_free_ino_ctl(root);
1503 mutex_init(&root->fs_commit_mutex);
1504 spin_lock_init(&root->cache_lock);
1505 init_waitqueue_head(&root->cache_wait);
1506
Al Viro0ee5dc62011-07-07 15:44:25 -04001507 ret = get_anon_bdev(&root->anon_dev);
Chris Masonac08aed2011-06-13 11:28:50 -04001508 if (ret)
1509 goto fail;
Chris Mason3394e162008-11-17 20:42:26 -05001510
Yan, Zhengd68fc572010-05-16 10:49:58 -04001511 if (btrfs_root_refs(&root->root_item) == 0) {
1512 ret = -ENOENT;
1513 goto fail;
1514 }
1515
1516 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1517 if (ret < 0)
1518 goto fail;
1519 if (ret == 0)
1520 root->orphan_item_inserted = 1;
1521
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001522 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1523 if (ret)
1524 goto fail;
1525
1526 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04001527 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1528 (unsigned long)root->root_key.objectid,
Chris Mason0f7d52f2007-04-09 10:42:37 -04001529 root);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001530 if (ret == 0)
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001531 root->in_radix = 1;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001532
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001533 spin_unlock(&fs_info->fs_roots_radix_lock);
1534 radix_tree_preload_end();
Chris Mason0f7d52f2007-04-09 10:42:37 -04001535 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001536 if (ret == -EEXIST) {
1537 free_fs_root(root);
1538 goto again;
1539 }
1540 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001541 }
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001542
1543 ret = btrfs_find_dead_roots(fs_info->tree_root,
1544 root->root_key.objectid);
1545 WARN_ON(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001546 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001547fail:
1548 free_fs_root(root);
1549 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001550}
1551
Chris Mason04160082008-03-26 10:28:07 -04001552static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1553{
1554 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1555 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001556 struct btrfs_device *device;
1557 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001558
Xiao Guangrong1f781602011-04-20 10:09:16 +00001559 rcu_read_lock();
1560 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001561 if (!device->bdev)
1562 continue;
Chris Mason04160082008-03-26 10:28:07 -04001563 bdi = blk_get_backing_dev_info(device->bdev);
1564 if (bdi && bdi_congested(bdi, bdi_bits)) {
1565 ret = 1;
1566 break;
1567 }
1568 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001569 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001570 return ret;
1571}
1572
Chris Mason38b66982008-04-22 09:22:11 -04001573/*
Jens Axboead081f12009-06-12 14:43:40 +02001574 * If this fails, caller must call bdi_destroy() to get rid of the
1575 * bdi again.
1576 */
Chris Mason04160082008-03-26 10:28:07 -04001577static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1578{
Jens Axboead081f12009-06-12 14:43:40 +02001579 int err;
1580
1581 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001582 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001583 if (err)
1584 return err;
1585
Chris Mason4575c9c2008-04-18 16:13:31 -04001586 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001587 bdi->congested_fn = btrfs_congested_fn;
1588 bdi->congested_data = info;
1589 return 0;
1590}
1591
Chris Mason8b712842008-06-11 16:50:36 -04001592/*
1593 * called by the kthread helper functions to finally call the bio end_io
1594 * functions. This is where read checksum verification actually happens
1595 */
1596static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001597{
Chris Masonce9adaa2008-04-09 16:28:12 -04001598 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001599 struct end_io_wq *end_io_wq;
1600 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001601 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001602
Chris Mason8b712842008-06-11 16:50:36 -04001603 end_io_wq = container_of(work, struct end_io_wq, work);
1604 bio = end_io_wq->bio;
1605 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -04001606
Chris Mason8b712842008-06-11 16:50:36 -04001607 error = end_io_wq->error;
1608 bio->bi_private = end_io_wq->private;
1609 bio->bi_end_io = end_io_wq->end_io;
1610 kfree(end_io_wq);
Chris Mason8b712842008-06-11 16:50:36 -04001611 bio_endio(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001612}
1613
Chris Masona74a4b92008-06-25 16:01:31 -04001614static int cleaner_kthread(void *arg)
1615{
1616 struct btrfs_root *root = arg;
1617
1618 do {
Yan, Zheng76dda932009-09-21 16:00:26 -04001619 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1620 mutex_trylock(&root->fs_info->cleaner_mutex)) {
Yan, Zheng24bbcf042009-11-12 09:36:34 +00001621 btrfs_run_delayed_iputs(root);
Yan, Zheng76dda932009-09-21 16:00:26 -04001622 btrfs_clean_old_snapshots(root);
1623 mutex_unlock(&root->fs_info->cleaner_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001624 btrfs_run_defrag_inodes(root->fs_info);
Yan, Zheng76dda932009-09-21 16:00:26 -04001625 }
Chris Masona74a4b92008-06-25 16:01:31 -04001626
Tejun Heoa0acae02011-11-21 12:32:22 -08001627 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001628 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001629 if (!kthread_should_stop())
1630 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001631 __set_current_state(TASK_RUNNING);
1632 }
1633 } while (!kthread_should_stop());
1634 return 0;
1635}
1636
1637static int transaction_kthread(void *arg)
1638{
1639 struct btrfs_root *root = arg;
1640 struct btrfs_trans_handle *trans;
1641 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001642 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001643 unsigned long now;
1644 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001645 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001646
1647 do {
Jan Kara914b2002012-03-12 16:05:50 +01001648 cannot_commit = false;
Chris Masona74a4b92008-06-25 16:01:31 -04001649 delay = HZ * 30;
Chris Masona74a4b92008-06-25 16:01:31 -04001650 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1651
Josef Bacika4abeea2011-04-11 17:25:13 -04001652 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001653 cur = root->fs_info->running_transaction;
1654 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001655 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001656 goto sleep;
1657 }
Yan Zheng31153d82008-07-28 15:32:19 -04001658
Chris Masona74a4b92008-06-25 16:01:31 -04001659 now = get_seconds();
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001660 if (!cur->blocked &&
1661 (now < cur->start_time || now - cur->start_time < 30)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001662 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001663 delay = HZ * 5;
1664 goto sleep;
1665 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001666 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001667 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001668
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001669 /* If the file system is aborted, this will always fail. */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04001670 trans = btrfs_join_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001671 if (IS_ERR(trans)) {
1672 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001673 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001674 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001675 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001676 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001677 } else {
1678 btrfs_end_transaction(trans, root);
1679 }
Chris Masona74a4b92008-06-25 16:01:31 -04001680sleep:
1681 wake_up_process(root->fs_info->cleaner_kthread);
1682 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1683
Tejun Heoa0acae02011-11-21 12:32:22 -08001684 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001685 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001686 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001687 (!btrfs_transaction_blocked(root->fs_info) ||
1688 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001689 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001690 __set_current_state(TASK_RUNNING);
1691 }
1692 } while (!kthread_should_stop());
1693 return 0;
1694}
1695
Chris Masonaf31f5e2011-11-03 15:17:42 -04001696/*
1697 * this will find the highest generation in the array of
1698 * root backups. The index of the highest array is returned,
1699 * or -1 if we can't find anything.
1700 *
1701 * We check to make sure the array is valid by comparing the
1702 * generation of the latest root in the array with the generation
1703 * in the super block. If they don't match we pitch it.
1704 */
1705static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1706{
1707 u64 cur;
1708 int newest_index = -1;
1709 struct btrfs_root_backup *root_backup;
1710 int i;
1711
1712 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1713 root_backup = info->super_copy->super_roots + i;
1714 cur = btrfs_backup_tree_root_gen(root_backup);
1715 if (cur == newest_gen)
1716 newest_index = i;
1717 }
1718
1719 /* check to see if we actually wrapped around */
1720 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1721 root_backup = info->super_copy->super_roots;
1722 cur = btrfs_backup_tree_root_gen(root_backup);
1723 if (cur == newest_gen)
1724 newest_index = 0;
1725 }
1726 return newest_index;
1727}
1728
1729
1730/*
1731 * find the oldest backup so we know where to store new entries
1732 * in the backup array. This will set the backup_root_index
1733 * field in the fs_info struct
1734 */
1735static void find_oldest_super_backup(struct btrfs_fs_info *info,
1736 u64 newest_gen)
1737{
1738 int newest_index = -1;
1739
1740 newest_index = find_newest_super_backup(info, newest_gen);
1741 /* if there was garbage in there, just move along */
1742 if (newest_index == -1) {
1743 info->backup_root_index = 0;
1744 } else {
1745 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1746 }
1747}
1748
1749/*
1750 * copy all the root pointers into the super backup array.
1751 * this will bump the backup pointer by one when it is
1752 * done
1753 */
1754static void backup_super_roots(struct btrfs_fs_info *info)
1755{
1756 int next_backup;
1757 struct btrfs_root_backup *root_backup;
1758 int last_backup;
1759
1760 next_backup = info->backup_root_index;
1761 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1762 BTRFS_NUM_BACKUP_ROOTS;
1763
1764 /*
1765 * just overwrite the last backup if we're at the same generation
1766 * this happens only at umount
1767 */
1768 root_backup = info->super_for_commit->super_roots + last_backup;
1769 if (btrfs_backup_tree_root_gen(root_backup) ==
1770 btrfs_header_generation(info->tree_root->node))
1771 next_backup = last_backup;
1772
1773 root_backup = info->super_for_commit->super_roots + next_backup;
1774
1775 /*
1776 * make sure all of our padding and empty slots get zero filled
1777 * regardless of which ones we use today
1778 */
1779 memset(root_backup, 0, sizeof(*root_backup));
1780
1781 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1782
1783 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1784 btrfs_set_backup_tree_root_gen(root_backup,
1785 btrfs_header_generation(info->tree_root->node));
1786
1787 btrfs_set_backup_tree_root_level(root_backup,
1788 btrfs_header_level(info->tree_root->node));
1789
1790 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1791 btrfs_set_backup_chunk_root_gen(root_backup,
1792 btrfs_header_generation(info->chunk_root->node));
1793 btrfs_set_backup_chunk_root_level(root_backup,
1794 btrfs_header_level(info->chunk_root->node));
1795
1796 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1797 btrfs_set_backup_extent_root_gen(root_backup,
1798 btrfs_header_generation(info->extent_root->node));
1799 btrfs_set_backup_extent_root_level(root_backup,
1800 btrfs_header_level(info->extent_root->node));
1801
Chris Mason7c7e82a2011-11-06 18:50:56 -05001802 /*
1803 * we might commit during log recovery, which happens before we set
1804 * the fs_root. Make sure it is valid before we fill it in.
1805 */
1806 if (info->fs_root && info->fs_root->node) {
1807 btrfs_set_backup_fs_root(root_backup,
1808 info->fs_root->node->start);
1809 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001810 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001811 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001812 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001813 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001814
1815 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1816 btrfs_set_backup_dev_root_gen(root_backup,
1817 btrfs_header_generation(info->dev_root->node));
1818 btrfs_set_backup_dev_root_level(root_backup,
1819 btrfs_header_level(info->dev_root->node));
1820
1821 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1822 btrfs_set_backup_csum_root_gen(root_backup,
1823 btrfs_header_generation(info->csum_root->node));
1824 btrfs_set_backup_csum_root_level(root_backup,
1825 btrfs_header_level(info->csum_root->node));
1826
1827 btrfs_set_backup_total_bytes(root_backup,
1828 btrfs_super_total_bytes(info->super_copy));
1829 btrfs_set_backup_bytes_used(root_backup,
1830 btrfs_super_bytes_used(info->super_copy));
1831 btrfs_set_backup_num_devices(root_backup,
1832 btrfs_super_num_devices(info->super_copy));
1833
1834 /*
1835 * if we don't copy this out to the super_copy, it won't get remembered
1836 * for the next commit
1837 */
1838 memcpy(&info->super_copy->super_roots,
1839 &info->super_for_commit->super_roots,
1840 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1841}
1842
1843/*
1844 * this copies info out of the root backup array and back into
1845 * the in-memory super block. It is meant to help iterate through
1846 * the array, so you send it the number of backups you've already
1847 * tried and the last backup index you used.
1848 *
1849 * this returns -1 when it has tried all the backups
1850 */
1851static noinline int next_root_backup(struct btrfs_fs_info *info,
1852 struct btrfs_super_block *super,
1853 int *num_backups_tried, int *backup_index)
1854{
1855 struct btrfs_root_backup *root_backup;
1856 int newest = *backup_index;
1857
1858 if (*num_backups_tried == 0) {
1859 u64 gen = btrfs_super_generation(super);
1860
1861 newest = find_newest_super_backup(info, gen);
1862 if (newest == -1)
1863 return -1;
1864
1865 *backup_index = newest;
1866 *num_backups_tried = 1;
1867 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1868 /* we've tried all the backups, all done */
1869 return -1;
1870 } else {
1871 /* jump to the next oldest backup */
1872 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1873 BTRFS_NUM_BACKUP_ROOTS;
1874 *backup_index = newest;
1875 *num_backups_tried += 1;
1876 }
1877 root_backup = super->super_roots + newest;
1878
1879 btrfs_set_super_generation(super,
1880 btrfs_backup_tree_root_gen(root_backup));
1881 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1882 btrfs_set_super_root_level(super,
1883 btrfs_backup_tree_root_level(root_backup));
1884 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1885
1886 /*
1887 * fixme: the total bytes and num_devices need to match or we should
1888 * need a fsck
1889 */
1890 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1891 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1892 return 0;
1893}
1894
1895/* helper to cleanup tree roots */
1896static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
1897{
1898 free_extent_buffer(info->tree_root->node);
1899 free_extent_buffer(info->tree_root->commit_root);
1900 free_extent_buffer(info->dev_root->node);
1901 free_extent_buffer(info->dev_root->commit_root);
1902 free_extent_buffer(info->extent_root->node);
1903 free_extent_buffer(info->extent_root->commit_root);
1904 free_extent_buffer(info->csum_root->node);
1905 free_extent_buffer(info->csum_root->commit_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02001906 if (info->quota_root) {
1907 free_extent_buffer(info->quota_root->node);
1908 free_extent_buffer(info->quota_root->commit_root);
1909 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001910
1911 info->tree_root->node = NULL;
1912 info->tree_root->commit_root = NULL;
1913 info->dev_root->node = NULL;
1914 info->dev_root->commit_root = NULL;
1915 info->extent_root->node = NULL;
1916 info->extent_root->commit_root = NULL;
1917 info->csum_root->node = NULL;
1918 info->csum_root->commit_root = NULL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001919 if (info->quota_root) {
1920 info->quota_root->node = NULL;
1921 info->quota_root->commit_root = NULL;
1922 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001923
1924 if (chunk_root) {
1925 free_extent_buffer(info->chunk_root->node);
1926 free_extent_buffer(info->chunk_root->commit_root);
1927 info->chunk_root->node = NULL;
1928 info->chunk_root->commit_root = NULL;
1929 }
1930}
1931
1932
Al Viroad2b2c82011-11-17 01:10:02 -05001933int open_ctree(struct super_block *sb,
1934 struct btrfs_fs_devices *fs_devices,
1935 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05001936{
Chris Masondb945352007-10-15 16:15:53 -04001937 u32 sectorsize;
1938 u32 nodesize;
1939 u32 leafsize;
1940 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001941 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04001942 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001943 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05001944 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04001945 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001946 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05001947 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05001948 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001949 struct btrfs_root *extent_root;
1950 struct btrfs_root *csum_root;
1951 struct btrfs_root *chunk_root;
1952 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001953 struct btrfs_root *quota_root;
Chris Masone02119d2008-09-05 16:13:11 -04001954 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05001955 int ret;
Yane58ca022008-04-01 11:21:34 -04001956 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04001957 int num_backups_tried = 0;
1958 int backup_index = 0;
Chris Mason4543df72008-06-11 21:47:56 -04001959
Al Virof84a8bd2011-11-17 15:57:57 -05001960 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05001961 extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info);
1962 csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info);
1963 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
1964 dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02001965 quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info);
Chris Mason8790d502008-04-03 16:29:03 -04001966
Al Virof84a8bd2011-11-17 15:57:57 -05001967 if (!tree_root || !extent_root || !csum_root ||
Arne Jansenbcef60f2011-09-13 15:23:30 +02001968 !chunk_root || !dev_root || !quota_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04001969 err = -ENOMEM;
1970 goto fail;
1971 }
Yan, Zheng76dda932009-09-21 16:00:26 -04001972
1973 ret = init_srcu_struct(&fs_info->subvol_srcu);
1974 if (ret) {
1975 err = ret;
1976 goto fail;
1977 }
1978
1979 ret = setup_bdi(fs_info, &fs_info->bdi);
1980 if (ret) {
1981 err = ret;
1982 goto fail_srcu;
1983 }
1984
1985 fs_info->btree_inode = new_inode(sb);
1986 if (!fs_info->btree_inode) {
1987 err = -ENOMEM;
1988 goto fail_bdi;
1989 }
1990
Chris Masona6591712011-07-19 12:04:14 -04001991 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08001992
Yan, Zheng76dda932009-09-21 16:00:26 -04001993 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04001994 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04001995 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00001996 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Chris Mason19c00dd2007-10-15 16:19:22 -04001997 INIT_LIST_HEAD(&fs_info->hashers);
Chris Masonea8c2812008-08-04 23:17:27 -04001998 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
Chris Mason5a3f23d2009-03-31 13:27:11 -04001999 INIT_LIST_HEAD(&fs_info->ordered_operations);
Yan Zheng11833d62009-09-11 16:11:19 -04002000 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Chris Mason1832a6d2007-12-21 16:27:21 -05002001 spin_lock_init(&fs_info->delalloc_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002002 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002003 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002004 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002005 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002006 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002007 spin_lock_init(&fs_info->tree_mod_seq_lock);
2008 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002009 mutex_init(&fs_info->reloc_mutex);
Chris Mason19c00dd2007-10-15 16:19:22 -04002010
Josef Bacik58176a92007-08-29 15:47:34 -04002011 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002012 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002013 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002014 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Chris Mason0b86a832008-03-24 15:01:56 -04002015 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002016 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2017 BTRFS_BLOCK_RSV_GLOBAL);
2018 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2019 BTRFS_BLOCK_RSV_DELALLOC);
2020 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2021 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2022 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2023 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2024 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002025 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002026 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002027 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002028 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002029 atomic_set(&fs_info->defrag_running, 0);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002030 atomic_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002031 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05002032 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002033 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002034 fs_info->defrag_inodes = RB_ROOT;
Josef Bacika4abeea2011-04-11 17:25:13 -04002035 fs_info->trans_no_join = 0;
Josef Bacik2bf64752011-09-26 17:12:22 -04002036 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002037 fs_info->tree_mod_log = RB_ROOT;
Chris Masonc8b97812008-10-29 14:49:59 -04002038
Arne Jansen90519d62011-05-23 14:30:00 +02002039 /* readahead state */
2040 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2041 spin_lock_init(&fs_info->reada_lock);
Chris Mason2c90e5d2007-04-02 10:50:19 -04002042
Chris Masonb34b0862008-12-19 15:43:22 -05002043 fs_info->thread_pool_size = min_t(unsigned long,
2044 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002045
Chris Mason3eaa2882008-07-24 11:57:52 -04002046 INIT_LIST_HEAD(&fs_info->ordered_extents);
2047 spin_lock_init(&fs_info->ordered_extent_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002048 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2049 GFP_NOFS);
2050 if (!fs_info->delayed_root) {
2051 err = -ENOMEM;
2052 goto fail_iput;
2053 }
2054 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002055
Arne Jansena2de7332011-03-08 14:14:00 +01002056 mutex_init(&fs_info->scrub_lock);
2057 atomic_set(&fs_info->scrubs_running, 0);
2058 atomic_set(&fs_info->scrub_pause_req, 0);
2059 atomic_set(&fs_info->scrubs_paused, 0);
2060 atomic_set(&fs_info->scrub_cancel_req, 0);
2061 init_waitqueue_head(&fs_info->scrub_pause_wait);
2062 init_rwsem(&fs_info->scrub_super_lock);
2063 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002064#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2065 fs_info->check_integrity_print_mask = 0;
2066#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002067
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002068 spin_lock_init(&fs_info->balance_lock);
2069 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002070 atomic_set(&fs_info->balance_running, 0);
2071 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002072 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002073 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002074 init_waitqueue_head(&fs_info->balance_wait_q);
Chris Masona061fc82008-05-07 11:43:44 -04002075
Chris Mason0afbaf82008-04-18 14:17:20 -04002076 sb->s_blocksize = 4096;
2077 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002078 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002079
Yan, Zheng76dda932009-09-21 16:00:26 -04002080 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002081 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002082 /*
2083 * we set the i_size on the btree inode to the max possible int.
2084 * the real end of the address space is determined by all of
2085 * the devices in the system
2086 */
2087 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002088 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002089 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2090
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002091 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002092 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002093 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002094 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002095 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002096
Chris Masond1310b22008-01-24 16:13:08 -05002097 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2098
Chris Mason0f7d52f2007-04-09 10:42:37 -04002099 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2100 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2101 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002102 set_bit(BTRFS_INODE_DUMMY,
2103 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Chris Mason22b0ebd2007-03-30 08:47:31 -04002104 insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002105
Chris Masone02119d2008-09-05 16:13:11 -04002106 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002107 fs_info->block_group_cache_tree = RB_ROOT;
Chris Mason925baed2008-06-25 16:01:30 -04002108
Yan Zheng11833d62009-09-11 16:11:19 -04002109 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002110 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002111 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002112 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002113 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002114 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002115
Chris Masond98237b2007-03-28 13:57:48 -04002116
Chris Mason5a3f23d2009-03-31 13:27:11 -04002117 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002118 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002119 mutex_init(&fs_info->chunk_mutex);
2120 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002121 mutex_init(&fs_info->cleaner_mutex);
2122 mutex_init(&fs_info->volume_mutex);
Yan Zheng276e6802009-07-30 09:40:40 -04002123 init_rwsem(&fs_info->extent_commit_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002124 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002125 init_rwsem(&fs_info->subvol_sem);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002126
Arne Jansen416ac512011-09-13 12:56:09 +02002127 spin_lock_init(&fs_info->qgroup_lock);
2128 fs_info->qgroup_tree = RB_ROOT;
2129 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2130 fs_info->qgroup_seq = 1;
2131 fs_info->quota_enabled = 0;
2132 fs_info->pending_quota_state = 0;
2133
Chris Masonfa9c0d792009-04-03 09:47:43 -04002134 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2135 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2136
Chris Mason7d9eb122008-07-08 14:19:17 -04002137 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002138 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002139 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002140 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002141
Chris Mason0b86a832008-03-24 15:01:56 -04002142 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002143 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002144
Chris Mason3c4bb262012-03-27 18:56:56 -04002145 invalidate_bdev(fs_devices->latest_bdev);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002146 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002147 if (!bh) {
2148 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002149 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002150 }
Chris Mason39279cc2007-06-12 06:35:45 -04002151
David Sterba6c417612011-04-13 15:41:04 +02002152 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2153 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2154 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002155 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002156
David Sterba6c417612011-04-13 15:41:04 +02002157 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002158
David Sterba6c417612011-04-13 15:41:04 +02002159 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002160 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002161 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002162
liuboacce9522011-01-06 19:30:25 +08002163 /* check FS state, whether FS is broken. */
2164 fs_info->fs_state |= btrfs_super_flags(disk_super);
2165
David Sterbafcd1f0652012-03-06 00:06:18 +01002166 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2167 if (ret) {
2168 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2169 err = ret;
2170 goto fail_alloc;
2171 }
liuboacce9522011-01-06 19:30:25 +08002172
Liu Bo75e7cb72011-03-22 10:12:20 +00002173 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002174 * run through our array of backup supers and setup
2175 * our ring pointer to the oldest one
2176 */
2177 generation = btrfs_super_generation(disk_super);
2178 find_oldest_super_backup(fs_info, generation);
2179
2180 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002181 * In the long term, we'll store the compression type in the super
2182 * block, and it'll be used for per file compression control.
2183 */
2184 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2185
Yan Zheng2b820322008-11-17 21:11:30 -05002186 ret = btrfs_parse_options(tree_root, options);
2187 if (ret) {
2188 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002189 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002190 }
Chris Masondfe25022008-05-13 13:46:40 -04002191
Josef Bacikf2b636e2008-12-02 06:36:08 -05002192 features = btrfs_super_incompat_flags(disk_super) &
2193 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2194 if (features) {
2195 printk(KERN_ERR "BTRFS: couldn't mount because of "
2196 "unsupported optional features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002197 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002198 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002199 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002200 }
2201
Chris Mason727011e2010-08-06 13:21:20 -04002202 if (btrfs_super_leafsize(disk_super) !=
2203 btrfs_super_nodesize(disk_super)) {
2204 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2205 "blocksizes don't match. node %d leaf %d\n",
2206 btrfs_super_nodesize(disk_super),
2207 btrfs_super_leafsize(disk_super));
2208 err = -EINVAL;
2209 goto fail_alloc;
2210 }
2211 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2212 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2213 "blocksize (%d) was too large\n",
2214 btrfs_super_leafsize(disk_super));
2215 err = -EINVAL;
2216 goto fail_alloc;
2217 }
2218
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002219 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002220 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002221 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002222 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002223
2224 /*
2225 * flag our filesystem as having big metadata blocks if
2226 * they are bigger than the page size
2227 */
2228 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2229 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2230 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2231 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2232 }
2233
Chris Masonbc3f1162012-03-29 17:02:47 -04002234 nodesize = btrfs_super_nodesize(disk_super);
2235 leafsize = btrfs_super_leafsize(disk_super);
2236 sectorsize = btrfs_super_sectorsize(disk_super);
2237 stripesize = btrfs_super_stripesize(disk_super);
2238
2239 /*
2240 * mixed block groups end up with duplicate but slightly offset
2241 * extent buffers for the same range. It leads to corruptions
2242 */
2243 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2244 (sectorsize != leafsize)) {
2245 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2246 "are not allowed for mixed block groups on %s\n",
2247 sb->s_id);
2248 goto fail_alloc;
2249 }
2250
Li Zefana6fa6fa2010-10-25 15:12:26 +08002251 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002252
Josef Bacikf2b636e2008-12-02 06:36:08 -05002253 features = btrfs_super_compat_ro_flags(disk_super) &
2254 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2255 if (!(sb->s_flags & MS_RDONLY) && features) {
2256 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2257 "unsupported option features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002258 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002259 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002260 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002261 }
Chris Mason61d92c32009-10-02 19:11:56 -04002262
2263 btrfs_init_workers(&fs_info->generic_worker,
2264 "genwork", 1, NULL);
2265
Chris Mason5443be42008-08-15 15:34:16 -04002266 btrfs_init_workers(&fs_info->workers, "worker",
Chris Mason61d92c32009-10-02 19:11:56 -04002267 fs_info->thread_pool_size,
2268 &fs_info->generic_worker);
Chris Masonc8b97812008-10-29 14:49:59 -04002269
Chris Mason771ed682008-11-06 22:02:51 -05002270 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
Chris Mason61d92c32009-10-02 19:11:56 -04002271 fs_info->thread_pool_size,
2272 &fs_info->generic_worker);
Chris Mason771ed682008-11-06 22:02:51 -05002273
Chris Mason5443be42008-08-15 15:34:16 -04002274 btrfs_init_workers(&fs_info->submit_workers, "submit",
Chris Masonb720d202008-08-15 15:34:14 -04002275 min_t(u64, fs_devices->num_devices,
Chris Mason61d92c32009-10-02 19:11:56 -04002276 fs_info->thread_pool_size),
2277 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002278
Josef Bacikbab39bf2011-06-30 14:42:28 -04002279 btrfs_init_workers(&fs_info->caching_workers, "cache",
2280 2, &fs_info->generic_worker);
2281
Chris Mason61b49442008-07-31 15:42:53 -04002282 /* a higher idle thresh on the submit workers makes it much more
2283 * likely that bios will be send down in a sane order to the
2284 * devices
2285 */
2286 fs_info->submit_workers.idle_thresh = 64;
Chris Mason53863232008-08-15 15:34:18 -04002287
Chris Mason771ed682008-11-06 22:02:51 -05002288 fs_info->workers.idle_thresh = 16;
Chris Mason4a69a412008-11-06 22:03:00 -05002289 fs_info->workers.ordered = 1;
Chris Mason61b49442008-07-31 15:42:53 -04002290
Chris Mason771ed682008-11-06 22:02:51 -05002291 fs_info->delalloc_workers.idle_thresh = 2;
2292 fs_info->delalloc_workers.ordered = 1;
2293
Chris Mason61d92c32009-10-02 19:11:56 -04002294 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2295 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002296 btrfs_init_workers(&fs_info->endio_workers, "endio",
Chris Mason61d92c32009-10-02 19:11:56 -04002297 fs_info->thread_pool_size,
2298 &fs_info->generic_worker);
Chris Masond20f7042008-12-08 16:58:54 -05002299 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
Chris Mason61d92c32009-10-02 19:11:56 -04002300 fs_info->thread_pool_size,
2301 &fs_info->generic_worker);
Chris Masoncad321a2008-12-17 14:51:42 -05002302 btrfs_init_workers(&fs_info->endio_meta_write_workers,
Chris Mason61d92c32009-10-02 19:11:56 -04002303 "endio-meta-write", fs_info->thread_pool_size,
2304 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002305 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
Chris Mason61d92c32009-10-02 19:11:56 -04002306 fs_info->thread_pool_size,
2307 &fs_info->generic_worker);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002308 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2309 1, &fs_info->generic_worker);
Miao Xie16cdcec2011-04-22 18:12:22 +08002310 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2311 fs_info->thread_pool_size,
2312 &fs_info->generic_worker);
Arne Jansen90519d62011-05-23 14:30:00 +02002313 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2314 fs_info->thread_pool_size,
2315 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002316
2317 /*
2318 * endios are largely parallel and should have a very
2319 * low idle thresh
2320 */
2321 fs_info->endio_workers.idle_thresh = 4;
Chris Masonb51912c2009-02-04 09:23:24 -05002322 fs_info->endio_meta_workers.idle_thresh = 4;
2323
Chris Mason90428462009-08-04 16:56:34 -04002324 fs_info->endio_write_workers.idle_thresh = 2;
2325 fs_info->endio_meta_write_workers.idle_thresh = 2;
Arne Jansen90519d62011-05-23 14:30:00 +02002326 fs_info->readahead_workers.idle_thresh = 2;
Chris Mason90428462009-08-04 16:56:34 -04002327
Josef Bacik0dc3b842011-11-18 14:37:27 -05002328 /*
2329 * btrfs_start_workers can really only fail because of ENOMEM so just
2330 * return -ENOMEM if any of these fail.
2331 */
2332 ret = btrfs_start_workers(&fs_info->workers);
2333 ret |= btrfs_start_workers(&fs_info->generic_worker);
2334 ret |= btrfs_start_workers(&fs_info->submit_workers);
2335 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2336 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2337 ret |= btrfs_start_workers(&fs_info->endio_workers);
2338 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
2339 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2340 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2341 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2342 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2343 ret |= btrfs_start_workers(&fs_info->caching_workers);
2344 ret |= btrfs_start_workers(&fs_info->readahead_workers);
2345 if (ret) {
Ilya Dryomovfed425c2012-06-22 12:13:01 -06002346 err = -ENOMEM;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002347 goto fail_sb_buffer;
2348 }
Chris Mason4543df72008-06-11 21:47:56 -04002349
Chris Mason4575c9c2008-04-18 16:13:31 -04002350 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002351 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2352 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002353
Chris Masondb945352007-10-15 16:15:53 -04002354 tree_root->nodesize = nodesize;
2355 tree_root->leafsize = leafsize;
2356 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002357 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002358
2359 sb->s_blocksize = sectorsize;
2360 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002361
Chris Mason39279cc2007-06-12 06:35:45 -04002362 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
2363 sizeof(disk_super->magic))) {
Chris Masond3977122009-01-05 21:25:51 -05002364 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002365 goto fail_sb_buffer;
2366 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002367
Liu Bo8d082fb2012-04-03 09:56:53 +08002368 if (sectorsize != PAGE_SIZE) {
2369 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2370 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002371 goto fail_sb_buffer;
2372 }
2373
Chris Mason925baed2008-06-25 16:01:30 -04002374 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002375 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002376 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002377 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002378 printk(KERN_WARNING "btrfs: failed to read the system "
2379 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002380 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002381 }
Chris Mason0b86a832008-03-24 15:01:56 -04002382
2383 blocksize = btrfs_level_size(tree_root,
2384 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002385 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002386
2387 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2388 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2389
2390 chunk_root->node = read_tree_block(chunk_root,
2391 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002392 blocksize, generation);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002393 BUG_ON(!chunk_root->node); /* -ENOMEM */
David Woodhouse83121942009-07-22 16:52:13 -04002394 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2395 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2396 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002397 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002398 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002399 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2400 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002401
Chris Masone17cade2008-04-15 15:41:47 -04002402 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Chris Masond3977122009-01-05 21:25:51 -05002403 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2404 BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002405
Chris Mason0b86a832008-03-24 15:01:56 -04002406 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002407 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002408 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2409 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002410 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002411 }
Chris Mason0b86a832008-03-24 15:01:56 -04002412
Chris Masondfe25022008-05-13 13:46:40 -04002413 btrfs_close_extra_devices(fs_devices);
2414
Chris Masona6b0d5c2012-02-20 20:53:43 -05002415 if (!fs_devices->latest_bdev) {
2416 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2417 sb->s_id);
2418 goto fail_tree_roots;
2419 }
2420
Chris Masonaf31f5e2011-11-03 15:17:42 -04002421retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002422 blocksize = btrfs_level_size(tree_root,
2423 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002424 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002425
Chris Masone20d96d2007-03-22 12:13:20 -04002426 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002427 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002428 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002429 if (!tree_root->node ||
2430 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
David Woodhouse83121942009-07-22 16:52:13 -04002431 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2432 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002433
2434 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002435 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002436
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002437 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2438 tree_root->commit_root = btrfs_root_node(tree_root);
Chris Masondb945352007-10-15 16:15:53 -04002439
2440 ret = find_and_setup_root(tree_root, fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04002441 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002442 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002443 goto recovery_tree_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002444 extent_root->track_dirty = 1;
2445
2446 ret = find_and_setup_root(tree_root, fs_info,
2447 BTRFS_DEV_TREE_OBJECTID, dev_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002448 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002449 goto recovery_tree_root;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002450 dev_root->track_dirty = 1;
Chris Mason3768f362007-03-13 16:47:54 -04002451
Chris Masond20f7042008-12-08 16:58:54 -05002452 ret = find_and_setup_root(tree_root, fs_info,
2453 BTRFS_CSUM_TREE_OBJECTID, csum_root);
2454 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002455 goto recovery_tree_root;
Chris Masond20f7042008-12-08 16:58:54 -05002456 csum_root->track_dirty = 1;
2457
Arne Jansenbcef60f2011-09-13 15:23:30 +02002458 ret = find_and_setup_root(tree_root, fs_info,
2459 BTRFS_QUOTA_TREE_OBJECTID, quota_root);
2460 if (ret) {
2461 kfree(quota_root);
2462 quota_root = fs_info->quota_root = NULL;
2463 } else {
2464 quota_root->track_dirty = 1;
2465 fs_info->quota_enabled = 1;
2466 fs_info->pending_quota_state = 1;
2467 }
2468
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002469 fs_info->generation = generation;
2470 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002471
Ilya Dryomov68310a52012-06-22 12:24:12 -06002472 ret = btrfs_recover_balance(fs_info);
2473 if (ret) {
2474 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2475 goto fail_block_groups;
2476 }
2477
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002478 ret = btrfs_init_dev_stats(fs_info);
2479 if (ret) {
2480 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2481 ret);
2482 goto fail_block_groups;
2483 }
2484
liuboc59021f2011-03-07 02:13:14 +00002485 ret = btrfs_init_space_info(fs_info);
2486 if (ret) {
2487 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2488 goto fail_block_groups;
2489 }
2490
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002491 ret = btrfs_read_block_groups(extent_root);
2492 if (ret) {
2493 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2494 goto fail_block_groups;
2495 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002496
Chris Masona74a4b92008-06-25 16:01:31 -04002497 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2498 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002499 if (IS_ERR(fs_info->cleaner_kthread))
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002500 goto fail_block_groups;
Chris Masona74a4b92008-06-25 16:01:31 -04002501
2502 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2503 tree_root,
2504 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002505 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002506 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002507
Chris Masonc2898112009-06-10 09:51:32 -04002508 if (!btrfs_test_opt(tree_root, SSD) &&
2509 !btrfs_test_opt(tree_root, NOSSD) &&
2510 !fs_info->fs_devices->rotating) {
2511 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2512 "mode\n");
2513 btrfs_set_opt(fs_info->mount_opt, SSD);
2514 }
2515
Stefan Behrens21adbd52011-11-09 13:44:05 +01002516#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2517 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2518 ret = btrfsic_mount(tree_root, fs_devices,
2519 btrfs_test_opt(tree_root,
2520 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2521 1 : 0,
2522 fs_info->check_integrity_print_mask);
2523 if (ret)
2524 printk(KERN_WARNING "btrfs: failed to initialize"
2525 " integrity check module %s\n", sb->s_id);
2526 }
2527#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002528 ret = btrfs_read_qgroup_config(fs_info);
2529 if (ret)
2530 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002531
liuboacce9522011-01-06 19:30:25 +08002532 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002533 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002534 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002535
Chris Mason7c2ca462008-11-19 15:13:35 -05002536 if (fs_devices->rw_devices == 0) {
Chris Masond3977122009-01-05 21:25:51 -05002537 printk(KERN_WARNING "Btrfs log replay required "
2538 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002539 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002540 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002541 }
Chris Masone02119d2008-09-05 16:13:11 -04002542 blocksize =
2543 btrfs_level_size(tree_root,
2544 btrfs_super_log_root_level(disk_super));
2545
Al Viro6f07e422011-11-17 00:46:16 -05002546 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002547 if (!log_tree_root) {
2548 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002549 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002550 }
Chris Masone02119d2008-09-05 16:13:11 -04002551
2552 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2553 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2554
2555 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002556 blocksize,
2557 generation + 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002558 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002559 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002560 if (ret) {
2561 btrfs_error(tree_root->fs_info, ret,
2562 "Failed to recover log tree");
2563 free_extent_buffer(log_tree_root->node);
2564 kfree(log_tree_root);
2565 goto fail_trans_kthread;
2566 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002567
2568 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002569 ret = btrfs_commit_super(tree_root);
2570 if (ret)
2571 goto fail_trans_kthread;
Yan Zhenge556ce22008-11-20 10:25:19 -05002572 }
Chris Masone02119d2008-09-05 16:13:11 -04002573 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002574
Yan, Zheng76dda932009-09-21 16:00:26 -04002575 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002576 if (ret)
2577 goto fail_trans_kthread;
Yan, Zheng76dda932009-09-21 16:00:26 -04002578
Chris Mason7c2ca462008-11-19 15:13:35 -05002579 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002580 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002581 if (ret)
2582 goto fail_trans_kthread;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002583
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002584 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002585 if (ret < 0) {
2586 printk(KERN_WARNING
2587 "btrfs: failed to recover relocation\n");
2588 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002589 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002590 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002591 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002592
Chris Mason3de45862008-11-17 21:02:50 -05002593 location.objectid = BTRFS_FS_TREE_OBJECTID;
2594 location.type = BTRFS_ROOT_ITEM_KEY;
2595 location.offset = (u64)-1;
2596
Chris Mason3de45862008-11-17 21:02:50 -05002597 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2598 if (!fs_info->fs_root)
Arne Jansenbcef60f2011-09-13 15:23:30 +02002599 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002600 if (IS_ERR(fs_info->fs_root)) {
2601 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002602 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002603 }
Chris Masonc2898112009-06-10 09:51:32 -04002604
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002605 if (sb->s_flags & MS_RDONLY)
2606 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002607
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002608 down_read(&fs_info->cleanup_work_sem);
2609 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2610 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002611 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002612 close_ctree(tree_root);
2613 return ret;
2614 }
2615 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002616
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002617 ret = btrfs_resume_balance_async(fs_info);
2618 if (ret) {
2619 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2620 close_ctree(tree_root);
2621 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002622 }
2623
Al Viroad2b2c82011-11-17 01:10:02 -05002624 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002625
Arne Jansenbcef60f2011-09-13 15:23:30 +02002626fail_qgroup:
2627 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05002628fail_trans_kthread:
2629 kthread_stop(fs_info->transaction_kthread);
Chris Mason3f157a22008-06-25 16:01:31 -04002630fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002631 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002632
2633 /*
2634 * make sure we're done with the btree inode before we stop our
2635 * kthreads
2636 */
2637 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2638 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2639
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002640fail_block_groups:
2641 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002642
2643fail_tree_roots:
2644 free_root_pointers(fs_info, 1);
2645
Chris Mason39279cc2007-06-12 06:35:45 -04002646fail_sb_buffer:
Chris Mason61d92c32009-10-02 19:11:56 -04002647 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason306c8b62011-11-03 15:21:39 -04002648 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Mason247e7432008-07-17 12:53:51 -04002649 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002650 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002651 btrfs_stop_workers(&fs_info->workers);
2652 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002653 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002654 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002655 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002656 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002657 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002658 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04002659 btrfs_stop_workers(&fs_info->caching_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002660fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04002661fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002662 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2663
Chris Mason7c2ca462008-11-19 15:13:35 -05002664 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Mason4543df72008-06-11 21:47:56 -04002665 iput(fs_info->btree_inode);
Jens Axboead081f12009-06-12 14:43:40 +02002666fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05002667 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002668fail_srcu:
2669 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002670fail:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002671 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05002672 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002673
2674recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04002675 if (!btrfs_test_opt(tree_root, RECOVERY))
2676 goto fail_tree_roots;
2677
2678 free_root_pointers(fs_info, 0);
2679
2680 /* don't use the log in recovery mode, it won't be valid */
2681 btrfs_set_super_log_root(disk_super, 0);
2682
2683 /* we can't trust the free space cache either */
2684 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2685
2686 ret = next_root_backup(fs_info, fs_info->super_copy,
2687 &num_backups_tried, &backup_index);
2688 if (ret == -1)
2689 goto fail_block_groups;
2690 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05002691}
2692
Chris Masonf2984462008-04-10 16:19:33 -04002693static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2694{
Chris Masonf2984462008-04-10 16:19:33 -04002695 if (uptodate) {
2696 set_buffer_uptodate(bh);
2697 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02002698 struct btrfs_device *device = (struct btrfs_device *)
2699 bh->b_private;
2700
Josef Bacik606686e2012-06-04 14:03:51 -04002701 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
2702 "I/O error on %s\n",
2703 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04002704 /* note, we dont' set_buffer_write_io_error because we have
2705 * our own ways of dealing with the IO errors
2706 */
Chris Masonf2984462008-04-10 16:19:33 -04002707 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02002708 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04002709 }
2710 unlock_buffer(bh);
2711 put_bh(bh);
2712}
2713
Yan Zhenga512bbf2008-12-08 16:46:26 -05002714struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2715{
2716 struct buffer_head *bh;
2717 struct buffer_head *latest = NULL;
2718 struct btrfs_super_block *super;
2719 int i;
2720 u64 transid = 0;
2721 u64 bytenr;
2722
2723 /* we would like to check all the supers, but that would make
2724 * a btrfs mount succeed after a mkfs from a different FS.
2725 * So, we need to add a special mount option to scan for
2726 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2727 */
2728 for (i = 0; i < 1; i++) {
2729 bytenr = btrfs_sb_offset(i);
2730 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2731 break;
2732 bh = __bread(bdev, bytenr / 4096, 4096);
2733 if (!bh)
2734 continue;
2735
2736 super = (struct btrfs_super_block *)bh->b_data;
2737 if (btrfs_super_bytenr(super) != bytenr ||
2738 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2739 sizeof(super->magic))) {
2740 brelse(bh);
2741 continue;
2742 }
2743
2744 if (!latest || btrfs_super_generation(super) > transid) {
2745 brelse(latest);
2746 latest = bh;
2747 transid = btrfs_super_generation(super);
2748 } else {
2749 brelse(bh);
2750 }
2751 }
2752 return latest;
2753}
2754
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002755/*
2756 * this should be called twice, once with wait == 0 and
2757 * once with wait == 1. When wait == 0 is done, all the buffer heads
2758 * we write are pinned.
2759 *
2760 * They are released when wait == 1 is done.
2761 * max_mirrors must be the same for both runs, and it indicates how
2762 * many supers on this one device should be written.
2763 *
2764 * max_mirrors == 0 means to write them all.
2765 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002766static int write_dev_supers(struct btrfs_device *device,
2767 struct btrfs_super_block *sb,
2768 int do_barriers, int wait, int max_mirrors)
2769{
2770 struct buffer_head *bh;
2771 int i;
2772 int ret;
2773 int errors = 0;
2774 u32 crc;
2775 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002776
2777 if (max_mirrors == 0)
2778 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2779
Yan Zhenga512bbf2008-12-08 16:46:26 -05002780 for (i = 0; i < max_mirrors; i++) {
2781 bytenr = btrfs_sb_offset(i);
2782 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2783 break;
2784
2785 if (wait) {
2786 bh = __find_get_block(device->bdev, bytenr / 4096,
2787 BTRFS_SUPER_INFO_SIZE);
2788 BUG_ON(!bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002789 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002790 if (!buffer_uptodate(bh))
2791 errors++;
2792
2793 /* drop our reference */
2794 brelse(bh);
2795
2796 /* drop the reference from the wait == 0 run */
2797 brelse(bh);
2798 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002799 } else {
2800 btrfs_set_super_bytenr(sb, bytenr);
2801
2802 crc = ~(u32)0;
2803 crc = btrfs_csum_data(NULL, (char *)sb +
2804 BTRFS_CSUM_SIZE, crc,
2805 BTRFS_SUPER_INFO_SIZE -
2806 BTRFS_CSUM_SIZE);
2807 btrfs_csum_final(crc, sb->csum);
2808
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002809 /*
2810 * one reference for us, and we leave it for the
2811 * caller
2812 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002813 bh = __getblk(device->bdev, bytenr / 4096,
2814 BTRFS_SUPER_INFO_SIZE);
2815 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2816
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002817 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002818 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002819
2820 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002821 lock_buffer(bh);
2822 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02002823 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002824 }
2825
Chris Mason387125f2011-11-18 15:07:51 -05002826 /*
2827 * we fua the first super. The others we allow
2828 * to go down lazy.
2829 */
Stefan Behrens21adbd52011-11-09 13:44:05 +01002830 ret = btrfsic_submit_bh(WRITE_FUA, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002831 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05002832 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002833 }
2834 return errors < i ? 0 : -1;
2835}
2836
Chris Mason387125f2011-11-18 15:07:51 -05002837/*
2838 * endio for the write_dev_flush, this will wake anyone waiting
2839 * for the barrier when it is done
2840 */
2841static void btrfs_end_empty_barrier(struct bio *bio, int err)
2842{
2843 if (err) {
2844 if (err == -EOPNOTSUPP)
2845 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
2846 clear_bit(BIO_UPTODATE, &bio->bi_flags);
2847 }
2848 if (bio->bi_private)
2849 complete(bio->bi_private);
2850 bio_put(bio);
2851}
2852
2853/*
2854 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
2855 * sent down. With wait == 1, it waits for the previous flush.
2856 *
2857 * any device where the flush fails with eopnotsupp are flagged as not-barrier
2858 * capable
2859 */
2860static int write_dev_flush(struct btrfs_device *device, int wait)
2861{
2862 struct bio *bio;
2863 int ret = 0;
2864
2865 if (device->nobarriers)
2866 return 0;
2867
2868 if (wait) {
2869 bio = device->flush_bio;
2870 if (!bio)
2871 return 0;
2872
2873 wait_for_completion(&device->flush_wait);
2874
2875 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Josef Bacik606686e2012-06-04 14:03:51 -04002876 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
2877 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05002878 device->nobarriers = 1;
2879 }
2880 if (!bio_flagged(bio, BIO_UPTODATE)) {
2881 ret = -EIO;
Stefan Behrens442a4f62012-05-25 16:06:08 +02002882 if (!bio_flagged(bio, BIO_EOPNOTSUPP))
2883 btrfs_dev_stat_inc_and_print(device,
2884 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05002885 }
2886
2887 /* drop the reference from the wait == 0 run */
2888 bio_put(bio);
2889 device->flush_bio = NULL;
2890
2891 return ret;
2892 }
2893
2894 /*
2895 * one reference for us, and we leave it for the
2896 * caller
2897 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01002898 device->flush_bio = NULL;
Chris Mason387125f2011-11-18 15:07:51 -05002899 bio = bio_alloc(GFP_NOFS, 0);
2900 if (!bio)
2901 return -ENOMEM;
2902
2903 bio->bi_end_io = btrfs_end_empty_barrier;
2904 bio->bi_bdev = device->bdev;
2905 init_completion(&device->flush_wait);
2906 bio->bi_private = &device->flush_wait;
2907 device->flush_bio = bio;
2908
2909 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002910 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05002911
2912 return 0;
2913}
2914
2915/*
2916 * send an empty flush down to each device in parallel,
2917 * then wait for them
2918 */
2919static int barrier_all_devices(struct btrfs_fs_info *info)
2920{
2921 struct list_head *head;
2922 struct btrfs_device *dev;
2923 int errors = 0;
2924 int ret;
2925
2926 /* send down all the barriers */
2927 head = &info->fs_devices->devices;
2928 list_for_each_entry_rcu(dev, head, dev_list) {
2929 if (!dev->bdev) {
2930 errors++;
2931 continue;
2932 }
2933 if (!dev->in_fs_metadata || !dev->writeable)
2934 continue;
2935
2936 ret = write_dev_flush(dev, 0);
2937 if (ret)
2938 errors++;
2939 }
2940
2941 /* wait for all the barriers */
2942 list_for_each_entry_rcu(dev, head, dev_list) {
2943 if (!dev->bdev) {
2944 errors++;
2945 continue;
2946 }
2947 if (!dev->in_fs_metadata || !dev->writeable)
2948 continue;
2949
2950 ret = write_dev_flush(dev, 1);
2951 if (ret)
2952 errors++;
2953 }
2954 if (errors)
2955 return -EIO;
2956 return 0;
2957}
2958
Yan Zhenga512bbf2008-12-08 16:46:26 -05002959int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04002960{
Chris Masone5e9a522009-06-10 15:17:02 -04002961 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04002962 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04002963 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04002964 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04002965 int ret;
2966 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04002967 int max_errors;
2968 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04002969 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04002970
David Sterba6c417612011-04-13 15:41:04 +02002971 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Masonf2984462008-04-10 16:19:33 -04002972 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002973 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04002974
David Sterba6c417612011-04-13 15:41:04 +02002975 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04002976 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04002977
Chris Mason174ba502011-05-27 10:03:58 -04002978 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04002979 head = &root->fs_info->fs_devices->devices;
Chris Mason387125f2011-11-18 15:07:51 -05002980
2981 if (do_barriers)
2982 barrier_all_devices(root->fs_info);
2983
Xiao Guangrong1f781602011-04-20 10:09:16 +00002984 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002985 if (!dev->bdev) {
2986 total_errors++;
2987 continue;
2988 }
Yan Zheng2b820322008-11-17 21:11:30 -05002989 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002990 continue;
2991
Yan Zheng2b820322008-11-17 21:11:30 -05002992 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04002993 btrfs_set_stack_device_type(dev_item, dev->type);
2994 btrfs_set_stack_device_id(dev_item, dev->devid);
2995 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2996 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2997 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2998 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2999 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3000 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003001 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003002
Chris Masona061fc82008-05-07 11:43:44 -04003003 flags = btrfs_super_flags(sb);
3004 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003005
Yan Zhenga512bbf2008-12-08 16:46:26 -05003006 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003007 if (ret)
3008 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003009 }
Chris Masona236aed2008-04-29 09:38:00 -04003010 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05003011 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3012 total_errors);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003013
3014 /* This shouldn't happen. FUA is masked off if unsupported */
Chris Masona236aed2008-04-29 09:38:00 -04003015 BUG();
3016 }
Chris Masonf2984462008-04-10 16:19:33 -04003017
Yan Zhenga512bbf2008-12-08 16:46:26 -05003018 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003019 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003020 if (!dev->bdev)
3021 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003022 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003023 continue;
3024
Yan Zhenga512bbf2008-12-08 16:46:26 -05003025 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3026 if (ret)
3027 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003028 }
Chris Mason174ba502011-05-27 10:03:58 -04003029 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003030 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003031 btrfs_error(root->fs_info, -EIO,
3032 "%d errors while writing supers", total_errors);
3033 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003034 }
Chris Masonf2984462008-04-10 16:19:33 -04003035 return 0;
3036}
3037
Yan Zhenga512bbf2008-12-08 16:46:26 -05003038int write_ctree_super(struct btrfs_trans_handle *trans,
3039 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003040{
Chris Masone66f7092007-04-20 13:16:02 -04003041 int ret;
Chris Mason2c90e5d2007-04-02 10:50:19 -04003042
Yan Zhenga512bbf2008-12-08 16:46:26 -05003043 ret = write_all_supers(root, max_mirrors);
Chris Mason5f39d392007-10-15 16:14:19 -04003044 return ret;
Chris Masoncfaa7292007-02-21 17:04:57 -05003045}
3046
Jeff Mahoney143bede2012-03-01 14:56:26 +01003047void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003048{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003049 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003050 radix_tree_delete(&fs_info->fs_roots_radix,
3051 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003052 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003053
3054 if (btrfs_root_refs(&root->root_item) == 0)
3055 synchronize_srcu(&fs_info->subvol_srcu);
3056
Li Zefan581bb052011-04-20 10:06:11 +08003057 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3058 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003059 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003060}
3061
3062static void free_fs_root(struct btrfs_root *root)
3063{
Li Zefan82d59022011-04-20 10:33:24 +08003064 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003065 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Al Viro0ee5dc62011-07-07 15:44:25 -04003066 if (root->anon_dev)
3067 free_anon_bdev(root->anon_dev);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003068 free_extent_buffer(root->node);
3069 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003070 kfree(root->free_ino_ctl);
3071 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003072 kfree(root->name);
Chris Mason2619ba12007-04-10 16:58:11 -04003073 kfree(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003074}
3075
Jeff Mahoney143bede2012-03-01 14:56:26 +01003076static void del_fs_roots(struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04003077{
3078 int ret;
3079 struct btrfs_root *gang[8];
3080 int i;
3081
Yan, Zheng76dda932009-09-21 16:00:26 -04003082 while (!list_empty(&fs_info->dead_roots)) {
3083 gang[0] = list_entry(fs_info->dead_roots.next,
3084 struct btrfs_root, root_list);
3085 list_del(&gang[0]->root_list);
3086
3087 if (gang[0]->in_radix) {
3088 btrfs_free_fs_root(fs_info, gang[0]);
3089 } else {
3090 free_extent_buffer(gang[0]->node);
3091 free_extent_buffer(gang[0]->commit_root);
3092 kfree(gang[0]);
3093 }
3094 }
3095
Chris Masond3977122009-01-05 21:25:51 -05003096 while (1) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04003097 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3098 (void **)gang, 0,
3099 ARRAY_SIZE(gang));
3100 if (!ret)
3101 break;
Chris Mason2619ba12007-04-10 16:58:11 -04003102 for (i = 0; i < ret; i++)
Chris Mason5eda7b52007-06-22 14:16:25 -04003103 btrfs_free_fs_root(fs_info, gang[i]);
Chris Mason0f7d52f2007-04-09 10:42:37 -04003104 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04003105}
Chris Masonb4100d62007-04-12 12:14:00 -04003106
Yan Zhengc146afa2008-11-12 14:34:12 -05003107int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3108{
3109 u64 root_objectid = 0;
3110 struct btrfs_root *gang[8];
3111 int i;
3112 int ret;
3113
3114 while (1) {
3115 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3116 (void **)gang, root_objectid,
3117 ARRAY_SIZE(gang));
3118 if (!ret)
3119 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003120
3121 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05003122 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003123 int err;
3124
Yan Zhengc146afa2008-11-12 14:34:12 -05003125 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003126 err = btrfs_orphan_cleanup(gang[i]);
3127 if (err)
3128 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003129 }
3130 root_objectid++;
3131 }
3132 return 0;
3133}
3134
3135int btrfs_commit_super(struct btrfs_root *root)
3136{
3137 struct btrfs_trans_handle *trans;
3138 int ret;
3139
3140 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00003141 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003142 btrfs_clean_old_snapshots(root);
3143 mutex_unlock(&root->fs_info->cleaner_mutex);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003144
3145 /* wait until ongoing cleanup work done */
3146 down_write(&root->fs_info->cleanup_work_sem);
3147 up_write(&root->fs_info->cleanup_work_sem);
3148
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003149 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003150 if (IS_ERR(trans))
3151 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05003152 ret = btrfs_commit_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003153 if (ret)
3154 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003155 /* run commit again to drop the original snapshot */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003156 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003157 if (IS_ERR(trans))
3158 return PTR_ERR(trans);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003159 ret = btrfs_commit_transaction(trans, root);
3160 if (ret)
3161 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003162 ret = btrfs_write_and_wait_transaction(NULL, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003163 if (ret) {
3164 btrfs_error(root->fs_info, ret,
3165 "Failed to sync btree inode to disk.");
3166 return ret;
3167 }
Yan Zhengc146afa2008-11-12 14:34:12 -05003168
Yan Zhenga512bbf2008-12-08 16:46:26 -05003169 ret = write_ctree_super(NULL, root, 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05003170 return ret;
3171}
3172
Chris Masone20d96d2007-03-22 12:13:20 -04003173int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003174{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003175 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003176 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003177
Chris Masonfacda1e2007-06-08 18:11:48 -04003178 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003179 smp_mb();
3180
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003181 /* pause restriper - we want to resume on mount */
3182 btrfs_pause_balance(root->fs_info);
3183
Arne Jansena2de7332011-03-08 14:14:00 +01003184 btrfs_scrub_cancel(root);
Chris Mason4cb53002011-05-24 15:35:30 -04003185
3186 /* wait for any defraggers to finish */
3187 wait_event(fs_info->transaction_wait,
3188 (atomic_read(&fs_info->defrag_running) == 0));
3189
3190 /* clear out the rbtree of defraggable inodes */
Al Viroe3029d92011-11-17 00:56:18 -05003191 btrfs_run_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003192
Yan Zhengc146afa2008-11-12 14:34:12 -05003193 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003194 ret = btrfs_commit_super(root);
3195 if (ret)
3196 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3197 }
3198
Stefan Behrens68ce9682012-08-01 05:45:52 -06003199 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
3200 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003201
Josef Bacik300e4f82011-08-29 14:06:00 -04003202 btrfs_put_block_group_cache(fs_info);
3203
Al Viroe3029d92011-11-17 00:56:18 -05003204 kthread_stop(fs_info->transaction_kthread);
3205 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003206
Yan Zhengf25784b2009-07-28 08:41:57 -04003207 fs_info->closing = 2;
3208 smp_mb();
3209
Arne Jansenbcef60f2011-09-13 15:23:30 +02003210 btrfs_free_qgroup_config(root->fs_info);
3211
Chris Masonb0c68f82008-01-31 11:05:37 -05003212 if (fs_info->delalloc_bytes) {
Chris Masond3977122009-01-05 21:25:51 -05003213 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07003214 (unsigned long long)fs_info->delalloc_bytes);
Chris Masonb0c68f82008-01-31 11:05:37 -05003215 }
Yanbcc63ab2008-07-30 16:29:20 -04003216
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003217 free_extent_buffer(fs_info->extent_root->node);
3218 free_extent_buffer(fs_info->extent_root->commit_root);
3219 free_extent_buffer(fs_info->tree_root->node);
3220 free_extent_buffer(fs_info->tree_root->commit_root);
Al Viroe3029d92011-11-17 00:56:18 -05003221 free_extent_buffer(fs_info->chunk_root->node);
3222 free_extent_buffer(fs_info->chunk_root->commit_root);
3223 free_extent_buffer(fs_info->dev_root->node);
3224 free_extent_buffer(fs_info->dev_root->commit_root);
3225 free_extent_buffer(fs_info->csum_root->node);
3226 free_extent_buffer(fs_info->csum_root->commit_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003227 if (fs_info->quota_root) {
3228 free_extent_buffer(fs_info->quota_root->node);
3229 free_extent_buffer(fs_info->quota_root->commit_root);
3230 }
Chris Masond20f7042008-12-08 16:58:54 -05003231
Al Viroe3029d92011-11-17 00:56:18 -05003232 btrfs_free_block_groups(fs_info);
Yan Zhengc146afa2008-11-12 14:34:12 -05003233
Chris Mason0f7d52f2007-04-09 10:42:37 -04003234 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003235
Yan Zhengc146afa2008-11-12 14:34:12 -05003236 iput(fs_info->btree_inode);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003237
Chris Mason61d92c32009-10-02 19:11:56 -04003238 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04003239 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05003240 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04003241 btrfs_stop_workers(&fs_info->workers);
3242 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05003243 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05003244 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04003245 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04003246 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04003247 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08003248 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04003249 btrfs_stop_workers(&fs_info->caching_workers);
Arne Jansen90519d62011-05-23 14:30:00 +02003250 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Masond6bfde82008-04-30 13:59:35 -04003251
Stefan Behrens21adbd52011-11-09 13:44:05 +01003252#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3253 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3254 btrfsic_unmount(root, fs_info->fs_devices);
3255#endif
3256
Chris Masondfe25022008-05-13 13:46:40 -04003257 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003258 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003259
Chris Mason04160082008-03-26 10:28:07 -04003260 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003261 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003262
Chris Masoneb60cea2007-02-02 09:18:22 -05003263 return 0;
3264}
3265
Chris Masonb9fab912012-05-06 07:23:47 -04003266int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3267 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003268{
Chris Mason1259ab72008-05-12 13:39:03 -04003269 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003270 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003271
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003272 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003273 if (!ret)
3274 return ret;
3275
3276 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003277 parent_transid, atomic);
3278 if (ret == -EAGAIN)
3279 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003280 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003281}
Chris Mason6702ed42007-08-07 16:15:09 -04003282
Chris Mason5f39d392007-10-15 16:14:19 -04003283int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3284{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003285 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003286}
3287
3288void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3289{
Chris Mason727011e2010-08-06 13:21:20 -04003290 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Mason5f39d392007-10-15 16:14:19 -04003291 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003292 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003293
Chris Masonb9447ef2009-03-09 11:45:38 -04003294 btrfs_assert_tree_locked(buf);
Chris Masonccd467d2007-06-28 15:57:36 -04003295 if (transid != root->fs_info->generation) {
Chris Masond3977122009-01-05 21:25:51 -05003296 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
3297 "found %llu running %llu\n",
Chris Masondb945352007-10-15 16:15:53 -04003298 (unsigned long long)buf->start,
Chris Masond3977122009-01-05 21:25:51 -05003299 (unsigned long long)transid,
3300 (unsigned long long)root->fs_info->generation);
Chris Masonccd467d2007-06-28 15:57:36 -04003301 WARN_ON(1);
3302 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003303 was_dirty = set_extent_buffer_dirty(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003304 if (!was_dirty) {
3305 spin_lock(&root->fs_info->delalloc_lock);
3306 root->fs_info->dirty_metadata_bytes += buf->len;
3307 spin_unlock(&root->fs_info->delalloc_lock);
3308 }
Chris Masoneb60cea2007-02-02 09:18:22 -05003309}
3310
Chris Masond3c2fdcf2007-09-17 10:58:06 -04003311void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
Chris Mason35b7e472007-05-02 15:53:43 -04003312{
Chris Mason188de642008-05-09 11:52:25 -04003313 /*
3314 * looks as though older kernels can get into trouble with
3315 * this code, they end up stuck in balance_dirty_pages forever
3316 */
Chris Masond6bfde82008-04-30 13:59:35 -04003317 u64 num_dirty;
Chris Mason771ed682008-11-06 22:02:51 -05003318 unsigned long thresh = 32 * 1024 * 1024;
Chris Masond6bfde82008-04-30 13:59:35 -04003319
Jens Axboe6933c022009-03-17 09:36:37 +01003320 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003321 return;
3322
Miao Xie16cdcec2011-04-22 18:12:22 +08003323 btrfs_balance_delayed_items(root);
3324
3325 num_dirty = root->fs_info->dirty_metadata_bytes;
3326
3327 if (num_dirty > thresh) {
3328 balance_dirty_pages_ratelimited_nr(
3329 root->fs_info->btree_inode->i_mapping, 1);
3330 }
3331 return;
3332}
3333
3334void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
3335{
3336 /*
3337 * looks as though older kernels can get into trouble with
3338 * this code, they end up stuck in balance_dirty_pages forever
3339 */
3340 u64 num_dirty;
3341 unsigned long thresh = 32 * 1024 * 1024;
3342
3343 if (current->flags & PF_MEMALLOC)
3344 return;
3345
Chris Mason585ad2c2009-05-18 10:41:58 -04003346 num_dirty = root->fs_info->dirty_metadata_bytes;
3347
Chris Masond6bfde82008-04-30 13:59:35 -04003348 if (num_dirty > thresh) {
3349 balance_dirty_pages_ratelimited_nr(
Chris Masond7fc6402008-02-18 12:12:38 -05003350 root->fs_info->btree_inode->i_mapping, 1);
Chris Masond6bfde82008-04-30 13:59:35 -04003351 }
Chris Mason188de642008-05-09 11:52:25 -04003352 return;
Chris Mason35b7e472007-05-02 15:53:43 -04003353}
Chris Mason6b800532007-10-15 16:17:34 -04003354
Chris Masonca7a79a2008-05-12 12:59:19 -04003355int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003356{
Chris Mason727011e2010-08-06 13:21:20 -04003357 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003358 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003359}
Chris Mason0da54682007-11-07 21:08:01 -05003360
Arne Jansen20897f52011-09-13 12:44:20 +02003361int btree_lock_page_hook(struct page *page, void *data,
Chris Mason01d658f2011-11-01 10:08:06 -04003362 void (*flush_fn)(void *))
Chris Mason4bef0842008-09-08 11:18:08 -04003363{
3364 struct inode *inode = page->mapping->host;
Chris Masonb9473432009-03-13 11:00:37 -04003365 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4bef0842008-09-08 11:18:08 -04003366 struct extent_buffer *eb;
Chris Mason4bef0842008-09-08 11:18:08 -04003367
Josef Bacik4f2de97a2012-03-07 16:20:05 -05003368 /*
3369 * We culled this eb but the page is still hanging out on the mapping,
3370 * carry on.
3371 */
3372 if (!PagePrivate(page))
Chris Mason4bef0842008-09-08 11:18:08 -04003373 goto out;
3374
Josef Bacik4f2de97a2012-03-07 16:20:05 -05003375 eb = (struct extent_buffer *)page->private;
3376 if (!eb) {
3377 WARN_ON(1);
3378 goto out;
3379 }
3380 if (page != eb->pages[0])
Chris Mason4bef0842008-09-08 11:18:08 -04003381 goto out;
3382
Chris Mason01d658f2011-11-01 10:08:06 -04003383 if (!btrfs_try_tree_write_lock(eb)) {
3384 flush_fn(data);
3385 btrfs_tree_lock(eb);
3386 }
Chris Mason4bef0842008-09-08 11:18:08 -04003387 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonb9473432009-03-13 11:00:37 -04003388
3389 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3390 spin_lock(&root->fs_info->delalloc_lock);
3391 if (root->fs_info->dirty_metadata_bytes >= eb->len)
3392 root->fs_info->dirty_metadata_bytes -= eb->len;
3393 else
3394 WARN_ON(1);
3395 spin_unlock(&root->fs_info->delalloc_lock);
3396 }
3397
Chris Mason4bef0842008-09-08 11:18:08 -04003398 btrfs_tree_unlock(eb);
Chris Mason4bef0842008-09-08 11:18:08 -04003399out:
Chris Mason01d658f2011-11-01 10:08:06 -04003400 if (!trylock_page(page)) {
3401 flush_fn(data);
3402 lock_page(page);
3403 }
Chris Mason4bef0842008-09-08 11:18:08 -04003404 return 0;
3405}
3406
David Sterbafcd1f0652012-03-06 00:06:18 +01003407static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003408 int read_only)
3409{
David Sterbafcd1f0652012-03-06 00:06:18 +01003410 if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) {
3411 printk(KERN_ERR "btrfs: unsupported checksum algorithm\n");
3412 return -EINVAL;
3413 }
liuboacce9522011-01-06 19:30:25 +08003414
David Sterbafcd1f0652012-03-06 00:06:18 +01003415 if (read_only)
3416 return 0;
3417
David Sterbafcd1f0652012-03-06 00:06:18 +01003418 return 0;
liuboacce9522011-01-06 19:30:25 +08003419}
3420
Stefan Behrens68ce9682012-08-01 05:45:52 -06003421void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003422{
liuboacce9522011-01-06 19:30:25 +08003423 mutex_lock(&root->fs_info->cleaner_mutex);
3424 btrfs_run_delayed_iputs(root);
3425 mutex_unlock(&root->fs_info->cleaner_mutex);
3426
3427 down_write(&root->fs_info->cleanup_work_sem);
3428 up_write(&root->fs_info->cleanup_work_sem);
3429
3430 /* cleanup FS via transaction */
3431 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003432}
3433
Jeff Mahoney143bede2012-03-01 14:56:26 +01003434static void btrfs_destroy_ordered_operations(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003435{
3436 struct btrfs_inode *btrfs_inode;
3437 struct list_head splice;
3438
3439 INIT_LIST_HEAD(&splice);
3440
3441 mutex_lock(&root->fs_info->ordered_operations_mutex);
3442 spin_lock(&root->fs_info->ordered_extent_lock);
3443
3444 list_splice_init(&root->fs_info->ordered_operations, &splice);
3445 while (!list_empty(&splice)) {
3446 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3447 ordered_operations);
3448
3449 list_del_init(&btrfs_inode->ordered_operations);
3450
3451 btrfs_invalidate_inodes(btrfs_inode->root);
3452 }
3453
3454 spin_unlock(&root->fs_info->ordered_extent_lock);
3455 mutex_unlock(&root->fs_info->ordered_operations_mutex);
liuboacce9522011-01-06 19:30:25 +08003456}
3457
Jeff Mahoney143bede2012-03-01 14:56:26 +01003458static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003459{
3460 struct list_head splice;
3461 struct btrfs_ordered_extent *ordered;
3462 struct inode *inode;
3463
3464 INIT_LIST_HEAD(&splice);
3465
3466 spin_lock(&root->fs_info->ordered_extent_lock);
3467
3468 list_splice_init(&root->fs_info->ordered_extents, &splice);
3469 while (!list_empty(&splice)) {
3470 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
3471 root_extent_list);
3472
3473 list_del_init(&ordered->root_extent_list);
3474 atomic_inc(&ordered->refs);
3475
3476 /* the inode may be getting freed (in sys_unlink path). */
3477 inode = igrab(ordered->inode);
3478
3479 spin_unlock(&root->fs_info->ordered_extent_lock);
3480 if (inode)
3481 iput(inode);
3482
3483 atomic_set(&ordered->refs, 1);
3484 btrfs_put_ordered_extent(ordered);
3485
3486 spin_lock(&root->fs_info->ordered_extent_lock);
3487 }
3488
3489 spin_unlock(&root->fs_info->ordered_extent_lock);
liuboacce9522011-01-06 19:30:25 +08003490}
3491
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003492int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3493 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003494{
3495 struct rb_node *node;
3496 struct btrfs_delayed_ref_root *delayed_refs;
3497 struct btrfs_delayed_ref_node *ref;
3498 int ret = 0;
3499
3500 delayed_refs = &trans->delayed_refs;
3501
3502 spin_lock(&delayed_refs->lock);
3503 if (delayed_refs->num_entries == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003504 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003505 printk(KERN_INFO "delayed_refs has NO entry\n");
3506 return ret;
3507 }
3508
Josef Bacikb939d1a2012-05-31 11:06:33 -04003509 while ((node = rb_first(&delayed_refs->root)) != NULL) {
liuboacce9522011-01-06 19:30:25 +08003510 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
liuboacce9522011-01-06 19:30:25 +08003511
3512 atomic_set(&ref->refs, 1);
3513 if (btrfs_delayed_ref_is_head(ref)) {
3514 struct btrfs_delayed_ref_head *head;
3515
3516 head = btrfs_delayed_node_to_head(ref);
Josef Bacikb939d1a2012-05-31 11:06:33 -04003517 if (!mutex_trylock(&head->mutex)) {
3518 atomic_inc(&ref->refs);
3519 spin_unlock(&delayed_refs->lock);
3520
3521 /* Need to wait for the delayed ref to run */
3522 mutex_lock(&head->mutex);
3523 mutex_unlock(&head->mutex);
3524 btrfs_put_delayed_ref(ref);
3525
Josef Bacike18fca72012-06-18 07:23:18 -06003526 spin_lock(&delayed_refs->lock);
Josef Bacikb939d1a2012-05-31 11:06:33 -04003527 continue;
3528 }
3529
liuboacce9522011-01-06 19:30:25 +08003530 kfree(head->extent_op);
3531 delayed_refs->num_heads--;
3532 if (list_empty(&head->cluster))
3533 delayed_refs->num_heads_ready--;
3534 list_del_init(&head->cluster);
liuboacce9522011-01-06 19:30:25 +08003535 }
Josef Bacikb939d1a2012-05-31 11:06:33 -04003536 ref->in_tree = 0;
3537 rb_erase(&ref->rb_node, &delayed_refs->root);
3538 delayed_refs->num_entries--;
3539
liuboacce9522011-01-06 19:30:25 +08003540 spin_unlock(&delayed_refs->lock);
3541 btrfs_put_delayed_ref(ref);
3542
3543 cond_resched();
3544 spin_lock(&delayed_refs->lock);
3545 }
3546
3547 spin_unlock(&delayed_refs->lock);
3548
3549 return ret;
3550}
3551
Jeff Mahoney143bede2012-03-01 14:56:26 +01003552static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
liuboacce9522011-01-06 19:30:25 +08003553{
3554 struct btrfs_pending_snapshot *snapshot;
3555 struct list_head splice;
3556
3557 INIT_LIST_HEAD(&splice);
3558
3559 list_splice_init(&t->pending_snapshots, &splice);
3560
3561 while (!list_empty(&splice)) {
3562 snapshot = list_entry(splice.next,
3563 struct btrfs_pending_snapshot,
3564 list);
3565
3566 list_del_init(&snapshot->list);
3567
3568 kfree(snapshot);
3569 }
liuboacce9522011-01-06 19:30:25 +08003570}
3571
Jeff Mahoney143bede2012-03-01 14:56:26 +01003572static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003573{
3574 struct btrfs_inode *btrfs_inode;
3575 struct list_head splice;
3576
3577 INIT_LIST_HEAD(&splice);
3578
liuboacce9522011-01-06 19:30:25 +08003579 spin_lock(&root->fs_info->delalloc_lock);
David Sterba5be76752011-06-09 10:02:51 +00003580 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003581
3582 while (!list_empty(&splice)) {
3583 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3584 delalloc_inodes);
3585
3586 list_del_init(&btrfs_inode->delalloc_inodes);
3587
3588 btrfs_invalidate_inodes(btrfs_inode->root);
3589 }
3590
3591 spin_unlock(&root->fs_info->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003592}
3593
3594static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3595 struct extent_io_tree *dirty_pages,
3596 int mark)
3597{
3598 int ret;
3599 struct page *page;
3600 struct inode *btree_inode = root->fs_info->btree_inode;
3601 struct extent_buffer *eb;
3602 u64 start = 0;
3603 u64 end;
3604 u64 offset;
3605 unsigned long index;
3606
3607 while (1) {
3608 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
3609 mark);
3610 if (ret)
3611 break;
3612
3613 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3614 while (start <= end) {
3615 index = start >> PAGE_CACHE_SHIFT;
3616 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
3617 page = find_get_page(btree_inode->i_mapping, index);
3618 if (!page)
3619 continue;
3620 offset = page_offset(page);
3621
3622 spin_lock(&dirty_pages->buffer_lock);
3623 eb = radix_tree_lookup(
3624 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
3625 offset >> PAGE_CACHE_SHIFT);
3626 spin_unlock(&dirty_pages->buffer_lock);
Josef Bacikee670f0a2012-05-31 15:54:30 -04003627 if (eb)
liuboacce9522011-01-06 19:30:25 +08003628 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3629 &eb->bflags);
liuboacce9522011-01-06 19:30:25 +08003630 if (PageWriteback(page))
3631 end_page_writeback(page);
3632
3633 lock_page(page);
3634 if (PageDirty(page)) {
3635 clear_page_dirty_for_io(page);
3636 spin_lock_irq(&page->mapping->tree_lock);
3637 radix_tree_tag_clear(&page->mapping->page_tree,
3638 page_index(page),
3639 PAGECACHE_TAG_DIRTY);
3640 spin_unlock_irq(&page->mapping->tree_lock);
3641 }
3642
liuboacce9522011-01-06 19:30:25 +08003643 unlock_page(page);
Josef Bacikee670f0a2012-05-31 15:54:30 -04003644 page_cache_release(page);
liuboacce9522011-01-06 19:30:25 +08003645 }
3646 }
3647
3648 return ret;
3649}
3650
3651static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3652 struct extent_io_tree *pinned_extents)
3653{
3654 struct extent_io_tree *unpin;
3655 u64 start;
3656 u64 end;
3657 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06003658 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08003659
3660 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06003661again:
liuboacce9522011-01-06 19:30:25 +08003662 while (1) {
3663 ret = find_first_extent_bit(unpin, 0, &start, &end,
3664 EXTENT_DIRTY);
3665 if (ret)
3666 break;
3667
3668 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00003669 if (btrfs_test_opt(root, DISCARD))
3670 ret = btrfs_error_discard_extent(root, start,
3671 end + 1 - start,
3672 NULL);
liuboacce9522011-01-06 19:30:25 +08003673
3674 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3675 btrfs_error_unpin_extent_range(root, start, end);
3676 cond_resched();
3677 }
3678
Liu Boed0eaa12012-06-14 02:23:21 -06003679 if (loop) {
3680 if (unpin == &root->fs_info->freed_extents[0])
3681 unpin = &root->fs_info->freed_extents[1];
3682 else
3683 unpin = &root->fs_info->freed_extents[0];
3684 loop = false;
3685 goto again;
3686 }
3687
liuboacce9522011-01-06 19:30:25 +08003688 return 0;
3689}
3690
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003691void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3692 struct btrfs_root *root)
3693{
3694 btrfs_destroy_delayed_refs(cur_trans, root);
3695 btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
3696 cur_trans->dirty_pages.dirty_bytes);
3697
3698 /* FIXME: cleanup wait for commit */
3699 cur_trans->in_commit = 1;
3700 cur_trans->blocked = 1;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003701 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003702
3703 cur_trans->blocked = 0;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003704 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003705
3706 cur_trans->commit_done = 1;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003707 wake_up(&cur_trans->commit_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003708
Miao Xie67cde342012-06-14 02:23:22 -06003709 btrfs_destroy_delayed_inodes(root);
3710 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003711
3712 btrfs_destroy_pending_snapshots(cur_trans);
3713
3714 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
3715 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06003716 btrfs_destroy_pinned_extent(root,
3717 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003718
3719 /*
3720 memset(cur_trans, 0, sizeof(*cur_trans));
3721 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3722 */
3723}
3724
3725int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003726{
3727 struct btrfs_transaction *t;
3728 LIST_HEAD(list);
3729
liuboacce9522011-01-06 19:30:25 +08003730 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3731
Josef Bacika4abeea2011-04-11 17:25:13 -04003732 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003733 list_splice_init(&root->fs_info->trans_list, &list);
Josef Bacika4abeea2011-04-11 17:25:13 -04003734 root->fs_info->trans_no_join = 1;
3735 spin_unlock(&root->fs_info->trans_lock);
3736
liuboacce9522011-01-06 19:30:25 +08003737 while (!list_empty(&list)) {
3738 t = list_entry(list.next, struct btrfs_transaction, list);
3739 if (!t)
3740 break;
3741
3742 btrfs_destroy_ordered_operations(root);
3743
3744 btrfs_destroy_ordered_extents(root);
3745
3746 btrfs_destroy_delayed_refs(t, root);
3747
3748 btrfs_block_rsv_release(root,
3749 &root->fs_info->trans_block_rsv,
3750 t->dirty_pages.dirty_bytes);
3751
3752 /* FIXME: cleanup wait for commit */
3753 t->in_commit = 1;
3754 t->blocked = 1;
Josef Bacik66657b32012-08-01 15:36:24 -04003755 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003756 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3757 wake_up(&root->fs_info->transaction_blocked_wait);
3758
3759 t->blocked = 0;
Josef Bacik66657b32012-08-01 15:36:24 -04003760 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003761 if (waitqueue_active(&root->fs_info->transaction_wait))
3762 wake_up(&root->fs_info->transaction_wait);
liuboacce9522011-01-06 19:30:25 +08003763
liuboacce9522011-01-06 19:30:25 +08003764 t->commit_done = 1;
Josef Bacik66657b32012-08-01 15:36:24 -04003765 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003766 if (waitqueue_active(&t->commit_wait))
3767 wake_up(&t->commit_wait);
liuboacce9522011-01-06 19:30:25 +08003768
Miao Xie67cde342012-06-14 02:23:22 -06003769 btrfs_destroy_delayed_inodes(root);
3770 btrfs_assert_delayed_root_empty(root);
3771
liuboacce9522011-01-06 19:30:25 +08003772 btrfs_destroy_pending_snapshots(t);
3773
3774 btrfs_destroy_delalloc_inodes(root);
3775
Josef Bacika4abeea2011-04-11 17:25:13 -04003776 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003777 root->fs_info->running_transaction = NULL;
Josef Bacika4abeea2011-04-11 17:25:13 -04003778 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003779
3780 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3781 EXTENT_DIRTY);
3782
3783 btrfs_destroy_pinned_extent(root,
3784 root->fs_info->pinned_extents);
3785
Josef Bacik13c5a932011-04-11 15:45:29 -04003786 atomic_set(&t->use_count, 0);
liuboacce9522011-01-06 19:30:25 +08003787 list_del_init(&t->list);
3788 memset(t, 0, sizeof(*t));
3789 kmem_cache_free(btrfs_transaction_cachep, t);
3790 }
3791
Josef Bacika4abeea2011-04-11 17:25:13 -04003792 spin_lock(&root->fs_info->trans_lock);
3793 root->fs_info->trans_no_join = 0;
3794 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003795 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08003796
3797 return 0;
3798}
3799
Chris Masond1310b22008-01-24 16:13:08 -05003800static struct extent_io_ops btree_extent_io_ops = {
Chris Mason4bef0842008-09-08 11:18:08 -04003801 .write_cache_pages_lock_hook = btree_lock_page_hook,
Chris Masonce9adaa2008-04-09 16:28:12 -04003802 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02003803 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04003804 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04003805 /* note we're sharing with inode.c for the merge bio hook */
3806 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05003807};