blob: 29c69e60d3b000de3e373c9b33742800771f0109 [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 Mason98509cf2008-09-11 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 Mason9ad6b7bc2008-04-18 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 Mason9ad6b7bc2008-04-18 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 Masonb9447ef82009-03-09 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 "
1119 " dirty_mdatadata_bytes (%lu)",
1120 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 Zheng7237f182009-01-21 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);
Josef Bacik8a35d952012-05-23 14:26:42 -04001171 atomic_set(&root->orphan_inodes, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001172 root->log_batch = 0;
1173 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 Zheng7237f182009-01-21 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 {
Chris Masona74a4b92008-06-25 16:01:31 -04001619 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
Yan, Zheng76dda932009-09-21 16:00:26 -04001620
1621 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1622 mutex_trylock(&root->fs_info->cleaner_mutex)) {
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001623 btrfs_run_delayed_iputs(root);
Yan, Zheng76dda932009-09-21 16:00:26 -04001624 btrfs_clean_old_snapshots(root);
1625 mutex_unlock(&root->fs_info->cleaner_mutex);
Chris Mason4cb53002011-05-24 15:35:30 -04001626 btrfs_run_defrag_inodes(root->fs_info);
Yan, Zheng76dda932009-09-21 16:00:26 -04001627 }
Chris Masona74a4b92008-06-25 16:01:31 -04001628
Tejun Heoa0acae02011-11-21 12:32:22 -08001629 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001630 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001631 if (!kthread_should_stop())
1632 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001633 __set_current_state(TASK_RUNNING);
1634 }
1635 } while (!kthread_should_stop());
1636 return 0;
1637}
1638
1639static int transaction_kthread(void *arg)
1640{
1641 struct btrfs_root *root = arg;
1642 struct btrfs_trans_handle *trans;
1643 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001644 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001645 unsigned long now;
1646 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001647 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001648
1649 do {
Jan Kara914b2002012-03-12 16:05:50 +01001650 cannot_commit = false;
Chris Masona74a4b92008-06-25 16:01:31 -04001651 delay = HZ * 30;
1652 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1653 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1654
Josef Bacika4abeea2011-04-11 17:25:13 -04001655 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001656 cur = root->fs_info->running_transaction;
1657 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001658 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001659 goto sleep;
1660 }
Yan Zheng31153d82008-07-28 15:32:19 -04001661
Chris Masona74a4b92008-06-25 16:01:31 -04001662 now = get_seconds();
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001663 if (!cur->blocked &&
1664 (now < cur->start_time || now - cur->start_time < 30)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001665 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001666 delay = HZ * 5;
1667 goto sleep;
1668 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001669 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001670 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001671
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001672 /* If the file system is aborted, this will always fail. */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04001673 trans = btrfs_join_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001674 if (IS_ERR(trans)) {
1675 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001676 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001677 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001678 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001679 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001680 } else {
1681 btrfs_end_transaction(trans, root);
1682 }
Chris Masona74a4b92008-06-25 16:01:31 -04001683sleep:
1684 wake_up_process(root->fs_info->cleaner_kthread);
1685 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1686
Tejun Heoa0acae02011-11-21 12:32:22 -08001687 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001688 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001689 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001690 (!btrfs_transaction_blocked(root->fs_info) ||
1691 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001692 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001693 __set_current_state(TASK_RUNNING);
1694 }
1695 } while (!kthread_should_stop());
1696 return 0;
1697}
1698
Chris Masonaf31f5e2011-11-03 15:17:42 -04001699/*
1700 * this will find the highest generation in the array of
1701 * root backups. The index of the highest array is returned,
1702 * or -1 if we can't find anything.
1703 *
1704 * We check to make sure the array is valid by comparing the
1705 * generation of the latest root in the array with the generation
1706 * in the super block. If they don't match we pitch it.
1707 */
1708static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1709{
1710 u64 cur;
1711 int newest_index = -1;
1712 struct btrfs_root_backup *root_backup;
1713 int i;
1714
1715 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1716 root_backup = info->super_copy->super_roots + i;
1717 cur = btrfs_backup_tree_root_gen(root_backup);
1718 if (cur == newest_gen)
1719 newest_index = i;
1720 }
1721
1722 /* check to see if we actually wrapped around */
1723 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1724 root_backup = info->super_copy->super_roots;
1725 cur = btrfs_backup_tree_root_gen(root_backup);
1726 if (cur == newest_gen)
1727 newest_index = 0;
1728 }
1729 return newest_index;
1730}
1731
1732
1733/*
1734 * find the oldest backup so we know where to store new entries
1735 * in the backup array. This will set the backup_root_index
1736 * field in the fs_info struct
1737 */
1738static void find_oldest_super_backup(struct btrfs_fs_info *info,
1739 u64 newest_gen)
1740{
1741 int newest_index = -1;
1742
1743 newest_index = find_newest_super_backup(info, newest_gen);
1744 /* if there was garbage in there, just move along */
1745 if (newest_index == -1) {
1746 info->backup_root_index = 0;
1747 } else {
1748 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1749 }
1750}
1751
1752/*
1753 * copy all the root pointers into the super backup array.
1754 * this will bump the backup pointer by one when it is
1755 * done
1756 */
1757static void backup_super_roots(struct btrfs_fs_info *info)
1758{
1759 int next_backup;
1760 struct btrfs_root_backup *root_backup;
1761 int last_backup;
1762
1763 next_backup = info->backup_root_index;
1764 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1765 BTRFS_NUM_BACKUP_ROOTS;
1766
1767 /*
1768 * just overwrite the last backup if we're at the same generation
1769 * this happens only at umount
1770 */
1771 root_backup = info->super_for_commit->super_roots + last_backup;
1772 if (btrfs_backup_tree_root_gen(root_backup) ==
1773 btrfs_header_generation(info->tree_root->node))
1774 next_backup = last_backup;
1775
1776 root_backup = info->super_for_commit->super_roots + next_backup;
1777
1778 /*
1779 * make sure all of our padding and empty slots get zero filled
1780 * regardless of which ones we use today
1781 */
1782 memset(root_backup, 0, sizeof(*root_backup));
1783
1784 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1785
1786 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1787 btrfs_set_backup_tree_root_gen(root_backup,
1788 btrfs_header_generation(info->tree_root->node));
1789
1790 btrfs_set_backup_tree_root_level(root_backup,
1791 btrfs_header_level(info->tree_root->node));
1792
1793 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1794 btrfs_set_backup_chunk_root_gen(root_backup,
1795 btrfs_header_generation(info->chunk_root->node));
1796 btrfs_set_backup_chunk_root_level(root_backup,
1797 btrfs_header_level(info->chunk_root->node));
1798
1799 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1800 btrfs_set_backup_extent_root_gen(root_backup,
1801 btrfs_header_generation(info->extent_root->node));
1802 btrfs_set_backup_extent_root_level(root_backup,
1803 btrfs_header_level(info->extent_root->node));
1804
Chris Mason7c7e82a2011-11-06 18:50:56 -05001805 /*
1806 * we might commit during log recovery, which happens before we set
1807 * the fs_root. Make sure it is valid before we fill it in.
1808 */
1809 if (info->fs_root && info->fs_root->node) {
1810 btrfs_set_backup_fs_root(root_backup,
1811 info->fs_root->node->start);
1812 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001813 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001814 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001815 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001816 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001817
1818 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1819 btrfs_set_backup_dev_root_gen(root_backup,
1820 btrfs_header_generation(info->dev_root->node));
1821 btrfs_set_backup_dev_root_level(root_backup,
1822 btrfs_header_level(info->dev_root->node));
1823
1824 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1825 btrfs_set_backup_csum_root_gen(root_backup,
1826 btrfs_header_generation(info->csum_root->node));
1827 btrfs_set_backup_csum_root_level(root_backup,
1828 btrfs_header_level(info->csum_root->node));
1829
1830 btrfs_set_backup_total_bytes(root_backup,
1831 btrfs_super_total_bytes(info->super_copy));
1832 btrfs_set_backup_bytes_used(root_backup,
1833 btrfs_super_bytes_used(info->super_copy));
1834 btrfs_set_backup_num_devices(root_backup,
1835 btrfs_super_num_devices(info->super_copy));
1836
1837 /*
1838 * if we don't copy this out to the super_copy, it won't get remembered
1839 * for the next commit
1840 */
1841 memcpy(&info->super_copy->super_roots,
1842 &info->super_for_commit->super_roots,
1843 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1844}
1845
1846/*
1847 * this copies info out of the root backup array and back into
1848 * the in-memory super block. It is meant to help iterate through
1849 * the array, so you send it the number of backups you've already
1850 * tried and the last backup index you used.
1851 *
1852 * this returns -1 when it has tried all the backups
1853 */
1854static noinline int next_root_backup(struct btrfs_fs_info *info,
1855 struct btrfs_super_block *super,
1856 int *num_backups_tried, int *backup_index)
1857{
1858 struct btrfs_root_backup *root_backup;
1859 int newest = *backup_index;
1860
1861 if (*num_backups_tried == 0) {
1862 u64 gen = btrfs_super_generation(super);
1863
1864 newest = find_newest_super_backup(info, gen);
1865 if (newest == -1)
1866 return -1;
1867
1868 *backup_index = newest;
1869 *num_backups_tried = 1;
1870 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1871 /* we've tried all the backups, all done */
1872 return -1;
1873 } else {
1874 /* jump to the next oldest backup */
1875 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1876 BTRFS_NUM_BACKUP_ROOTS;
1877 *backup_index = newest;
1878 *num_backups_tried += 1;
1879 }
1880 root_backup = super->super_roots + newest;
1881
1882 btrfs_set_super_generation(super,
1883 btrfs_backup_tree_root_gen(root_backup));
1884 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1885 btrfs_set_super_root_level(super,
1886 btrfs_backup_tree_root_level(root_backup));
1887 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1888
1889 /*
1890 * fixme: the total bytes and num_devices need to match or we should
1891 * need a fsck
1892 */
1893 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1894 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1895 return 0;
1896}
1897
1898/* helper to cleanup tree roots */
1899static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
1900{
1901 free_extent_buffer(info->tree_root->node);
1902 free_extent_buffer(info->tree_root->commit_root);
1903 free_extent_buffer(info->dev_root->node);
1904 free_extent_buffer(info->dev_root->commit_root);
1905 free_extent_buffer(info->extent_root->node);
1906 free_extent_buffer(info->extent_root->commit_root);
1907 free_extent_buffer(info->csum_root->node);
1908 free_extent_buffer(info->csum_root->commit_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02001909 if (info->quota_root) {
1910 free_extent_buffer(info->quota_root->node);
1911 free_extent_buffer(info->quota_root->commit_root);
1912 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001913
1914 info->tree_root->node = NULL;
1915 info->tree_root->commit_root = NULL;
1916 info->dev_root->node = NULL;
1917 info->dev_root->commit_root = NULL;
1918 info->extent_root->node = NULL;
1919 info->extent_root->commit_root = NULL;
1920 info->csum_root->node = NULL;
1921 info->csum_root->commit_root = NULL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001922 if (info->quota_root) {
1923 info->quota_root->node = NULL;
1924 info->quota_root->commit_root = NULL;
1925 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001926
1927 if (chunk_root) {
1928 free_extent_buffer(info->chunk_root->node);
1929 free_extent_buffer(info->chunk_root->commit_root);
1930 info->chunk_root->node = NULL;
1931 info->chunk_root->commit_root = NULL;
1932 }
1933}
1934
1935
Al Viroad2b2c82011-11-17 01:10:02 -05001936int open_ctree(struct super_block *sb,
1937 struct btrfs_fs_devices *fs_devices,
1938 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05001939{
Chris Masondb945352007-10-15 16:15:53 -04001940 u32 sectorsize;
1941 u32 nodesize;
1942 u32 leafsize;
1943 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001944 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04001945 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05001946 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05001947 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04001948 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001949 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05001950 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05001951 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02001952 struct btrfs_root *extent_root;
1953 struct btrfs_root *csum_root;
1954 struct btrfs_root *chunk_root;
1955 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001956 struct btrfs_root *quota_root;
Chris Masone02119d2008-09-05 16:13:11 -04001957 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05001958 int ret;
Yane58ca022008-04-01 11:21:34 -04001959 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04001960 int num_backups_tried = 0;
1961 int backup_index = 0;
Chris Mason4543df72008-06-11 21:47:56 -04001962
Al Virof84a8bd2011-11-17 15:57:57 -05001963 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05001964 extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info);
1965 csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info);
1966 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
1967 dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02001968 quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info);
Chris Mason8790d502008-04-03 16:29:03 -04001969
Al Virof84a8bd2011-11-17 15:57:57 -05001970 if (!tree_root || !extent_root || !csum_root ||
Arne Jansenbcef60f2011-09-13 15:23:30 +02001971 !chunk_root || !dev_root || !quota_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04001972 err = -ENOMEM;
1973 goto fail;
1974 }
Yan, Zheng76dda932009-09-21 16:00:26 -04001975
1976 ret = init_srcu_struct(&fs_info->subvol_srcu);
1977 if (ret) {
1978 err = ret;
1979 goto fail;
1980 }
1981
1982 ret = setup_bdi(fs_info, &fs_info->bdi);
1983 if (ret) {
1984 err = ret;
1985 goto fail_srcu;
1986 }
1987
1988 fs_info->btree_inode = new_inode(sb);
1989 if (!fs_info->btree_inode) {
1990 err = -ENOMEM;
1991 goto fail_bdi;
1992 }
1993
Chris Masona6591712011-07-19 12:04:14 -04001994 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08001995
Yan, Zheng76dda932009-09-21 16:00:26 -04001996 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04001997 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04001998 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00001999 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Chris Mason19c00dd2007-10-15 16:19:22 -04002000 INIT_LIST_HEAD(&fs_info->hashers);
Chris Masonea8c2812008-08-04 23:17:27 -04002001 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
Chris Mason5a3f23d2009-03-31 13:27:11 -04002002 INIT_LIST_HEAD(&fs_info->ordered_operations);
Yan Zheng11833d62009-09-11 16:11:19 -04002003 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Chris Mason1832a6d2007-12-21 16:27:21 -05002004 spin_lock_init(&fs_info->delalloc_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002005 spin_lock_init(&fs_info->trans_lock);
Yan Zheng31153d82008-07-28 15:32:19 -04002006 spin_lock_init(&fs_info->ref_cache_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002007 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002008 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002009 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002010 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002011 spin_lock_init(&fs_info->tree_mod_seq_lock);
2012 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002013 mutex_init(&fs_info->reloc_mutex);
Chris Mason19c00dd2007-10-15 16:19:22 -04002014
Josef Bacik58176a92007-08-29 15:47:34 -04002015 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002016 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002017 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002018 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Chris Mason0b86a832008-03-24 15:01:56 -04002019 btrfs_mapping_init(&fs_info->mapping_tree);
Yan, Zhengf0486c62010-05-16 10:46:25 -04002020 btrfs_init_block_rsv(&fs_info->global_block_rsv);
2021 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
2022 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
2023 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
2024 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
Josef Bacik6d668dd2011-11-03 22:54:25 -04002025 btrfs_init_block_rsv(&fs_info->delayed_block_rsv);
Chris Masoncb03c742008-05-15 16:15:45 -04002026 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002027 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002028 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002029 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002030 atomic_set(&fs_info->defrag_running, 0);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002031 atomic_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002032 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05002033 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002034 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002035 fs_info->defrag_inodes = RB_ROOT;
Josef Bacika4abeea2011-04-11 17:25:13 -04002036 fs_info->trans_no_join = 0;
Josef Bacik2bf64752011-09-26 17:12:22 -04002037 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002038 fs_info->tree_mod_log = RB_ROOT;
Chris Masonc8b97812008-10-29 14:49:59 -04002039
Arne Jansen90519d62011-05-23 14:30:00 +02002040 /* readahead state */
2041 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2042 spin_lock_init(&fs_info->reada_lock);
Chris Mason2c90e5d2007-04-02 10:50:19 -04002043
Chris Masonb34b0862008-12-19 15:43:22 -05002044 fs_info->thread_pool_size = min_t(unsigned long,
2045 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002046
Chris Mason3eaa2882008-07-24 11:57:52 -04002047 INIT_LIST_HEAD(&fs_info->ordered_extents);
2048 spin_lock_init(&fs_info->ordered_extent_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002049 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2050 GFP_NOFS);
2051 if (!fs_info->delayed_root) {
2052 err = -ENOMEM;
2053 goto fail_iput;
2054 }
2055 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002056
Arne Jansena2de7332011-03-08 14:14:00 +01002057 mutex_init(&fs_info->scrub_lock);
2058 atomic_set(&fs_info->scrubs_running, 0);
2059 atomic_set(&fs_info->scrub_pause_req, 0);
2060 atomic_set(&fs_info->scrubs_paused, 0);
2061 atomic_set(&fs_info->scrub_cancel_req, 0);
2062 init_waitqueue_head(&fs_info->scrub_pause_wait);
2063 init_rwsem(&fs_info->scrub_super_lock);
2064 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002065#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2066 fs_info->check_integrity_print_mask = 0;
2067#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002068
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002069 spin_lock_init(&fs_info->balance_lock);
2070 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002071 atomic_set(&fs_info->balance_running, 0);
2072 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002073 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002074 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002075 init_waitqueue_head(&fs_info->balance_wait_q);
Chris Masona061fc82008-05-07 11:43:44 -04002076
Chris Mason0afbaf82008-04-18 14:17:20 -04002077 sb->s_blocksize = 4096;
2078 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002079 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002080
Yan, Zheng76dda932009-09-21 16:00:26 -04002081 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002082 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002083 /*
2084 * we set the i_size on the btree inode to the max possible int.
2085 * the real end of the address space is determined by all of
2086 * the devices in the system
2087 */
2088 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002089 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002090 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2091
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002092 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002093 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002094 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002095 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002096 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002097
Chris Masond1310b22008-01-24 16:13:08 -05002098 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2099
Chris Mason0f7d52f2007-04-09 10:42:37 -04002100 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2101 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2102 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002103 set_bit(BTRFS_INODE_DUMMY,
2104 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Chris Mason22b0ebd2007-03-30 08:47:31 -04002105 insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002106
Chris Masone02119d2008-09-05 16:13:11 -04002107 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002108 fs_info->block_group_cache_tree = RB_ROOT;
Chris Mason925baed2008-06-25 16:01:30 -04002109
Yan Zheng11833d62009-09-11 16:11:19 -04002110 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002111 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002112 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002113 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002114 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002115 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002116
Chris Masond98237b2007-03-28 13:57:48 -04002117
Chris Mason5a3f23d2009-03-31 13:27:11 -04002118 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002119 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002120 mutex_init(&fs_info->chunk_mutex);
2121 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002122 mutex_init(&fs_info->cleaner_mutex);
2123 mutex_init(&fs_info->volume_mutex);
Yan Zheng276e6802009-07-30 09:40:40 -04002124 init_rwsem(&fs_info->extent_commit_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002125 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002126 init_rwsem(&fs_info->subvol_sem);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002127
Arne Jansen416ac512011-09-13 12:56:09 +02002128 spin_lock_init(&fs_info->qgroup_lock);
2129 fs_info->qgroup_tree = RB_ROOT;
2130 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2131 fs_info->qgroup_seq = 1;
2132 fs_info->quota_enabled = 0;
2133 fs_info->pending_quota_state = 0;
2134
Chris Masonfa9c0d792009-04-03 09:47:43 -04002135 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2136 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2137
Chris Mason7d9eb122008-07-08 14:19:17 -04002138 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002139 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002140 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002141 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002142
Chris Mason0b86a832008-03-24 15:01:56 -04002143 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002144 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002145
Chris Mason3c4bb262012-03-27 18:56:56 -04002146 invalidate_bdev(fs_devices->latest_bdev);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002147 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002148 if (!bh) {
2149 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002150 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002151 }
Chris Mason39279cc2007-06-12 06:35:45 -04002152
David Sterba6c417612011-04-13 15:41:04 +02002153 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2154 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2155 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002156 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002157
David Sterba6c417612011-04-13 15:41:04 +02002158 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002159
David Sterba6c417612011-04-13 15:41:04 +02002160 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002161 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002162 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002163
liuboacce9522011-01-06 19:30:25 +08002164 /* check FS state, whether FS is broken. */
2165 fs_info->fs_state |= btrfs_super_flags(disk_super);
2166
David Sterbafcd1f0652012-03-06 00:06:18 +01002167 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2168 if (ret) {
2169 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2170 err = ret;
2171 goto fail_alloc;
2172 }
liuboacce9522011-01-06 19:30:25 +08002173
Liu Bo75e7cb72011-03-22 10:12:20 +00002174 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002175 * run through our array of backup supers and setup
2176 * our ring pointer to the oldest one
2177 */
2178 generation = btrfs_super_generation(disk_super);
2179 find_oldest_super_backup(fs_info, generation);
2180
2181 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002182 * In the long term, we'll store the compression type in the super
2183 * block, and it'll be used for per file compression control.
2184 */
2185 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2186
Yan Zheng2b820322008-11-17 21:11:30 -05002187 ret = btrfs_parse_options(tree_root, options);
2188 if (ret) {
2189 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002190 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002191 }
Chris Masondfe25022008-05-13 13:46:40 -04002192
Josef Bacikf2b636e2008-12-02 06:36:08 -05002193 features = btrfs_super_incompat_flags(disk_super) &
2194 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2195 if (features) {
2196 printk(KERN_ERR "BTRFS: couldn't mount because of "
2197 "unsupported optional features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002198 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002199 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002200 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002201 }
2202
Chris Mason727011e2010-08-06 13:21:20 -04002203 if (btrfs_super_leafsize(disk_super) !=
2204 btrfs_super_nodesize(disk_super)) {
2205 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2206 "blocksizes don't match. node %d leaf %d\n",
2207 btrfs_super_nodesize(disk_super),
2208 btrfs_super_leafsize(disk_super));
2209 err = -EINVAL;
2210 goto fail_alloc;
2211 }
2212 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2213 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2214 "blocksize (%d) was too large\n",
2215 btrfs_super_leafsize(disk_super));
2216 err = -EINVAL;
2217 goto fail_alloc;
2218 }
2219
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002220 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002221 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002222 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002223 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002224
2225 /*
2226 * flag our filesystem as having big metadata blocks if
2227 * they are bigger than the page size
2228 */
2229 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2230 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2231 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2232 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2233 }
2234
Chris Masonbc3f1162012-03-29 17:02:47 -04002235 nodesize = btrfs_super_nodesize(disk_super);
2236 leafsize = btrfs_super_leafsize(disk_super);
2237 sectorsize = btrfs_super_sectorsize(disk_super);
2238 stripesize = btrfs_super_stripesize(disk_super);
2239
2240 /*
2241 * mixed block groups end up with duplicate but slightly offset
2242 * extent buffers for the same range. It leads to corruptions
2243 */
2244 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2245 (sectorsize != leafsize)) {
2246 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2247 "are not allowed for mixed block groups on %s\n",
2248 sb->s_id);
2249 goto fail_alloc;
2250 }
2251
Li Zefana6fa6fa2010-10-25 15:12:26 +08002252 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002253
Josef Bacikf2b636e2008-12-02 06:36:08 -05002254 features = btrfs_super_compat_ro_flags(disk_super) &
2255 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2256 if (!(sb->s_flags & MS_RDONLY) && features) {
2257 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2258 "unsupported option features (%Lx).\n",
Joel Becker21380932009-04-21 12:38:29 -07002259 (unsigned long long)features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002260 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002261 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002262 }
Chris Mason61d92c32009-10-02 19:11:56 -04002263
2264 btrfs_init_workers(&fs_info->generic_worker,
2265 "genwork", 1, NULL);
2266
Chris Mason5443be42008-08-15 15:34:16 -04002267 btrfs_init_workers(&fs_info->workers, "worker",
Chris Mason61d92c32009-10-02 19:11:56 -04002268 fs_info->thread_pool_size,
2269 &fs_info->generic_worker);
Chris Masonc8b97812008-10-29 14:49:59 -04002270
Chris Mason771ed682008-11-06 22:02:51 -05002271 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
Chris Mason61d92c32009-10-02 19:11:56 -04002272 fs_info->thread_pool_size,
2273 &fs_info->generic_worker);
Chris Mason771ed682008-11-06 22:02:51 -05002274
Chris Mason5443be42008-08-15 15:34:16 -04002275 btrfs_init_workers(&fs_info->submit_workers, "submit",
Chris Masonb720d202008-08-15 15:34:14 -04002276 min_t(u64, fs_devices->num_devices,
Chris Mason61d92c32009-10-02 19:11:56 -04002277 fs_info->thread_pool_size),
2278 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002279
Josef Bacikbab39bf2011-06-30 14:42:28 -04002280 btrfs_init_workers(&fs_info->caching_workers, "cache",
2281 2, &fs_info->generic_worker);
2282
Chris Mason61b49442008-07-31 15:42:53 -04002283 /* a higher idle thresh on the submit workers makes it much more
2284 * likely that bios will be send down in a sane order to the
2285 * devices
2286 */
2287 fs_info->submit_workers.idle_thresh = 64;
Chris Mason53863232008-08-15 15:34:18 -04002288
Chris Mason771ed682008-11-06 22:02:51 -05002289 fs_info->workers.idle_thresh = 16;
Chris Mason4a69a412008-11-06 22:03:00 -05002290 fs_info->workers.ordered = 1;
Chris Mason61b49442008-07-31 15:42:53 -04002291
Chris Mason771ed682008-11-06 22:02:51 -05002292 fs_info->delalloc_workers.idle_thresh = 2;
2293 fs_info->delalloc_workers.ordered = 1;
2294
Chris Mason61d92c32009-10-02 19:11:56 -04002295 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2296 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002297 btrfs_init_workers(&fs_info->endio_workers, "endio",
Chris Mason61d92c32009-10-02 19:11:56 -04002298 fs_info->thread_pool_size,
2299 &fs_info->generic_worker);
Chris Masond20f7042008-12-08 16:58:54 -05002300 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
Chris Mason61d92c32009-10-02 19:11:56 -04002301 fs_info->thread_pool_size,
2302 &fs_info->generic_worker);
Chris Masoncad321a2008-12-17 14:51:42 -05002303 btrfs_init_workers(&fs_info->endio_meta_write_workers,
Chris Mason61d92c32009-10-02 19:11:56 -04002304 "endio-meta-write", fs_info->thread_pool_size,
2305 &fs_info->generic_worker);
Chris Mason5443be42008-08-15 15:34:16 -04002306 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
Chris Mason61d92c32009-10-02 19:11:56 -04002307 fs_info->thread_pool_size,
2308 &fs_info->generic_worker);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002309 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2310 1, &fs_info->generic_worker);
Miao Xie16cdcec2011-04-22 18:12:22 +08002311 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2312 fs_info->thread_pool_size,
2313 &fs_info->generic_worker);
Arne Jansen90519d62011-05-23 14:30:00 +02002314 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2315 fs_info->thread_pool_size,
2316 &fs_info->generic_worker);
Chris Mason61b49442008-07-31 15:42:53 -04002317
2318 /*
2319 * endios are largely parallel and should have a very
2320 * low idle thresh
2321 */
2322 fs_info->endio_workers.idle_thresh = 4;
Chris Masonb51912c2009-02-04 09:23:24 -05002323 fs_info->endio_meta_workers.idle_thresh = 4;
2324
Chris Mason90428462009-08-04 16:56:34 -04002325 fs_info->endio_write_workers.idle_thresh = 2;
2326 fs_info->endio_meta_write_workers.idle_thresh = 2;
Arne Jansen90519d62011-05-23 14:30:00 +02002327 fs_info->readahead_workers.idle_thresh = 2;
Chris Mason90428462009-08-04 16:56:34 -04002328
Josef Bacik0dc3b842011-11-18 14:37:27 -05002329 /*
2330 * btrfs_start_workers can really only fail because of ENOMEM so just
2331 * return -ENOMEM if any of these fail.
2332 */
2333 ret = btrfs_start_workers(&fs_info->workers);
2334 ret |= btrfs_start_workers(&fs_info->generic_worker);
2335 ret |= btrfs_start_workers(&fs_info->submit_workers);
2336 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2337 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2338 ret |= btrfs_start_workers(&fs_info->endio_workers);
2339 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
2340 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2341 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2342 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2343 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2344 ret |= btrfs_start_workers(&fs_info->caching_workers);
2345 ret |= btrfs_start_workers(&fs_info->readahead_workers);
2346 if (ret) {
Ilya Dryomovfed425c2012-06-22 12:13:01 -06002347 err = -ENOMEM;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002348 goto fail_sb_buffer;
2349 }
Chris Mason4543df72008-06-11 21:47:56 -04002350
Chris Mason4575c9c2008-04-18 16:13:31 -04002351 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002352 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2353 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002354
Chris Masondb945352007-10-15 16:15:53 -04002355 tree_root->nodesize = nodesize;
2356 tree_root->leafsize = leafsize;
2357 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002358 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002359
2360 sb->s_blocksize = sectorsize;
2361 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002362
Chris Mason39279cc2007-06-12 06:35:45 -04002363 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
2364 sizeof(disk_super->magic))) {
Chris Masond3977122009-01-05 21:25:51 -05002365 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002366 goto fail_sb_buffer;
2367 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002368
Liu Bo8d082fb2012-04-03 09:56:53 +08002369 if (sectorsize != PAGE_SIZE) {
2370 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2371 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002372 goto fail_sb_buffer;
2373 }
2374
Chris Mason925baed2008-06-25 16:01:30 -04002375 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002376 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002377 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002378 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002379 printk(KERN_WARNING "btrfs: failed to read the system "
2380 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002381 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002382 }
Chris Mason0b86a832008-03-24 15:01:56 -04002383
2384 blocksize = btrfs_level_size(tree_root,
2385 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002386 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002387
2388 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2389 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2390
2391 chunk_root->node = read_tree_block(chunk_root,
2392 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002393 blocksize, generation);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002394 BUG_ON(!chunk_root->node); /* -ENOMEM */
David Woodhouse83121942009-07-22 16:52:13 -04002395 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2396 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2397 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002398 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002399 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002400 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2401 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002402
Chris Masone17cade2008-04-15 15:41:47 -04002403 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Chris Masond3977122009-01-05 21:25:51 -05002404 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2405 BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002406
Chris Mason0b86a832008-03-24 15:01:56 -04002407 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002408 if (ret) {
Chris Masond3977122009-01-05 21:25:51 -05002409 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2410 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002411 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002412 }
Chris Mason0b86a832008-03-24 15:01:56 -04002413
Chris Masondfe25022008-05-13 13:46:40 -04002414 btrfs_close_extra_devices(fs_devices);
2415
Chris Masona6b0d5c2012-02-20 20:53:43 -05002416 if (!fs_devices->latest_bdev) {
2417 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2418 sb->s_id);
2419 goto fail_tree_roots;
2420 }
2421
Chris Masonaf31f5e2011-11-03 15:17:42 -04002422retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002423 blocksize = btrfs_level_size(tree_root,
2424 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002425 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002426
Chris Masone20d96d2007-03-22 12:13:20 -04002427 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002428 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002429 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002430 if (!tree_root->node ||
2431 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
David Woodhouse83121942009-07-22 16:52:13 -04002432 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2433 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002434
2435 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002436 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002437
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002438 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2439 tree_root->commit_root = btrfs_root_node(tree_root);
Chris Masondb945352007-10-15 16:15:53 -04002440
2441 ret = find_and_setup_root(tree_root, fs_info,
Chris Masone20d96d2007-03-22 12:13:20 -04002442 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002443 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002444 goto recovery_tree_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002445 extent_root->track_dirty = 1;
2446
2447 ret = find_and_setup_root(tree_root, fs_info,
2448 BTRFS_DEV_TREE_OBJECTID, dev_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002449 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002450 goto recovery_tree_root;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002451 dev_root->track_dirty = 1;
Chris Mason3768f362007-03-13 16:47:54 -04002452
Chris Masond20f7042008-12-08 16:58:54 -05002453 ret = find_and_setup_root(tree_root, fs_info,
2454 BTRFS_CSUM_TREE_OBJECTID, csum_root);
2455 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002456 goto recovery_tree_root;
Chris Masond20f7042008-12-08 16:58:54 -05002457 csum_root->track_dirty = 1;
2458
Arne Jansenbcef60f2011-09-13 15:23:30 +02002459 ret = find_and_setup_root(tree_root, fs_info,
2460 BTRFS_QUOTA_TREE_OBJECTID, quota_root);
2461 if (ret) {
2462 kfree(quota_root);
2463 quota_root = fs_info->quota_root = NULL;
2464 } else {
2465 quota_root->track_dirty = 1;
2466 fs_info->quota_enabled = 1;
2467 fs_info->pending_quota_state = 1;
2468 }
2469
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002470 fs_info->generation = generation;
2471 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002472
Ilya Dryomov68310a52012-06-22 12:24:12 -06002473 ret = btrfs_recover_balance(fs_info);
2474 if (ret) {
2475 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2476 goto fail_block_groups;
2477 }
2478
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002479 ret = btrfs_init_dev_stats(fs_info);
2480 if (ret) {
2481 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2482 ret);
2483 goto fail_block_groups;
2484 }
2485
liuboc59021f2011-03-07 02:13:14 +00002486 ret = btrfs_init_space_info(fs_info);
2487 if (ret) {
2488 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2489 goto fail_block_groups;
2490 }
2491
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002492 ret = btrfs_read_block_groups(extent_root);
2493 if (ret) {
2494 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2495 goto fail_block_groups;
2496 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002497
Chris Masona74a4b92008-06-25 16:01:31 -04002498 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2499 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002500 if (IS_ERR(fs_info->cleaner_kthread))
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002501 goto fail_block_groups;
Chris Masona74a4b92008-06-25 16:01:31 -04002502
2503 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2504 tree_root,
2505 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002506 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002507 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002508
Chris Masonc2898112009-06-10 09:51:32 -04002509 if (!btrfs_test_opt(tree_root, SSD) &&
2510 !btrfs_test_opt(tree_root, NOSSD) &&
2511 !fs_info->fs_devices->rotating) {
2512 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2513 "mode\n");
2514 btrfs_set_opt(fs_info->mount_opt, SSD);
2515 }
2516
Stefan Behrens21adbd52011-11-09 13:44:05 +01002517#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2518 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2519 ret = btrfsic_mount(tree_root, fs_devices,
2520 btrfs_test_opt(tree_root,
2521 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2522 1 : 0,
2523 fs_info->check_integrity_print_mask);
2524 if (ret)
2525 printk(KERN_WARNING "btrfs: failed to initialize"
2526 " integrity check module %s\n", sb->s_id);
2527 }
2528#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002529 ret = btrfs_read_qgroup_config(fs_info);
2530 if (ret)
2531 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002532
liuboacce9522011-01-06 19:30:25 +08002533 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002534 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002535 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002536
Chris Mason7c2ca462008-11-19 15:13:35 -05002537 if (fs_devices->rw_devices == 0) {
Chris Masond3977122009-01-05 21:25:51 -05002538 printk(KERN_WARNING "Btrfs log replay required "
2539 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002540 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002541 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002542 }
Chris Masone02119d2008-09-05 16:13:11 -04002543 blocksize =
2544 btrfs_level_size(tree_root,
2545 btrfs_super_log_root_level(disk_super));
2546
Al Viro6f07e422011-11-17 00:46:16 -05002547 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002548 if (!log_tree_root) {
2549 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002550 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002551 }
Chris Masone02119d2008-09-05 16:13:11 -04002552
2553 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2554 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2555
2556 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002557 blocksize,
2558 generation + 1);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002559 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002560 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002561 if (ret) {
2562 btrfs_error(tree_root->fs_info, ret,
2563 "Failed to recover log tree");
2564 free_extent_buffer(log_tree_root->node);
2565 kfree(log_tree_root);
2566 goto fail_trans_kthread;
2567 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002568
2569 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002570 ret = btrfs_commit_super(tree_root);
2571 if (ret)
2572 goto fail_trans_kthread;
Yan Zhenge556ce22008-11-20 10:25:19 -05002573 }
Chris Masone02119d2008-09-05 16:13:11 -04002574 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002575
Yan, Zheng76dda932009-09-21 16:00:26 -04002576 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002577 if (ret)
2578 goto fail_trans_kthread;
Yan, Zheng76dda932009-09-21 16:00:26 -04002579
Chris Mason7c2ca462008-11-19 15:13:35 -05002580 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002581 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002582 if (ret)
2583 goto fail_trans_kthread;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002584
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002585 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002586 if (ret < 0) {
2587 printk(KERN_WARNING
2588 "btrfs: failed to recover relocation\n");
2589 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002590 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002591 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002592 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002593
Chris Mason3de45862008-11-17 21:02:50 -05002594 location.objectid = BTRFS_FS_TREE_OBJECTID;
2595 location.type = BTRFS_ROOT_ITEM_KEY;
2596 location.offset = (u64)-1;
2597
Chris Mason3de45862008-11-17 21:02:50 -05002598 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2599 if (!fs_info->fs_root)
Arne Jansenbcef60f2011-09-13 15:23:30 +02002600 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002601 if (IS_ERR(fs_info->fs_root)) {
2602 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002603 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002604 }
Chris Masonc2898112009-06-10 09:51:32 -04002605
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002606 if (sb->s_flags & MS_RDONLY)
2607 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002608
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002609 down_read(&fs_info->cleanup_work_sem);
2610 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2611 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002612 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002613 close_ctree(tree_root);
2614 return ret;
2615 }
2616 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002617
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002618 ret = btrfs_resume_balance_async(fs_info);
2619 if (ret) {
2620 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2621 close_ctree(tree_root);
2622 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002623 }
2624
Al Viroad2b2c82011-11-17 01:10:02 -05002625 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002626
Arne Jansenbcef60f2011-09-13 15:23:30 +02002627fail_qgroup:
2628 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05002629fail_trans_kthread:
2630 kthread_stop(fs_info->transaction_kthread);
Chris Mason3f157a22008-06-25 16:01:31 -04002631fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002632 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002633
2634 /*
2635 * make sure we're done with the btree inode before we stop our
2636 * kthreads
2637 */
2638 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2639 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2640
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002641fail_block_groups:
2642 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002643
2644fail_tree_roots:
2645 free_root_pointers(fs_info, 1);
2646
Chris Mason39279cc2007-06-12 06:35:45 -04002647fail_sb_buffer:
Chris Mason61d92c32009-10-02 19:11:56 -04002648 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason306c8b62011-11-03 15:21:39 -04002649 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Mason247e7432008-07-17 12:53:51 -04002650 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05002651 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04002652 btrfs_stop_workers(&fs_info->workers);
2653 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05002654 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05002655 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04002656 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04002657 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04002658 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002659 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04002660 btrfs_stop_workers(&fs_info->caching_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08002661fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04002662fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002663 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2664
Chris Mason7c2ca462008-11-19 15:13:35 -05002665 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Mason4543df72008-06-11 21:47:56 -04002666 iput(fs_info->btree_inode);
Jens Axboead081f12009-06-12 14:43:40 +02002667fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05002668 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04002669fail_srcu:
2670 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05002671fail:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002672 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05002673 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002674
2675recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04002676 if (!btrfs_test_opt(tree_root, RECOVERY))
2677 goto fail_tree_roots;
2678
2679 free_root_pointers(fs_info, 0);
2680
2681 /* don't use the log in recovery mode, it won't be valid */
2682 btrfs_set_super_log_root(disk_super, 0);
2683
2684 /* we can't trust the free space cache either */
2685 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2686
2687 ret = next_root_backup(fs_info, fs_info->super_copy,
2688 &num_backups_tried, &backup_index);
2689 if (ret == -1)
2690 goto fail_block_groups;
2691 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05002692}
2693
Chris Masonf2984462008-04-10 16:19:33 -04002694static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2695{
Chris Masonf2984462008-04-10 16:19:33 -04002696 if (uptodate) {
2697 set_buffer_uptodate(bh);
2698 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02002699 struct btrfs_device *device = (struct btrfs_device *)
2700 bh->b_private;
2701
Josef Bacik606686e2012-06-04 14:03:51 -04002702 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
2703 "I/O error on %s\n",
2704 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04002705 /* note, we dont' set_buffer_write_io_error because we have
2706 * our own ways of dealing with the IO errors
2707 */
Chris Masonf2984462008-04-10 16:19:33 -04002708 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02002709 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04002710 }
2711 unlock_buffer(bh);
2712 put_bh(bh);
2713}
2714
Yan Zhenga512bbf2008-12-08 16:46:26 -05002715struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2716{
2717 struct buffer_head *bh;
2718 struct buffer_head *latest = NULL;
2719 struct btrfs_super_block *super;
2720 int i;
2721 u64 transid = 0;
2722 u64 bytenr;
2723
2724 /* we would like to check all the supers, but that would make
2725 * a btrfs mount succeed after a mkfs from a different FS.
2726 * So, we need to add a special mount option to scan for
2727 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2728 */
2729 for (i = 0; i < 1; i++) {
2730 bytenr = btrfs_sb_offset(i);
2731 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2732 break;
2733 bh = __bread(bdev, bytenr / 4096, 4096);
2734 if (!bh)
2735 continue;
2736
2737 super = (struct btrfs_super_block *)bh->b_data;
2738 if (btrfs_super_bytenr(super) != bytenr ||
2739 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2740 sizeof(super->magic))) {
2741 brelse(bh);
2742 continue;
2743 }
2744
2745 if (!latest || btrfs_super_generation(super) > transid) {
2746 brelse(latest);
2747 latest = bh;
2748 transid = btrfs_super_generation(super);
2749 } else {
2750 brelse(bh);
2751 }
2752 }
2753 return latest;
2754}
2755
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002756/*
2757 * this should be called twice, once with wait == 0 and
2758 * once with wait == 1. When wait == 0 is done, all the buffer heads
2759 * we write are pinned.
2760 *
2761 * They are released when wait == 1 is done.
2762 * max_mirrors must be the same for both runs, and it indicates how
2763 * many supers on this one device should be written.
2764 *
2765 * max_mirrors == 0 means to write them all.
2766 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002767static int write_dev_supers(struct btrfs_device *device,
2768 struct btrfs_super_block *sb,
2769 int do_barriers, int wait, int max_mirrors)
2770{
2771 struct buffer_head *bh;
2772 int i;
2773 int ret;
2774 int errors = 0;
2775 u32 crc;
2776 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002777
2778 if (max_mirrors == 0)
2779 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2780
Yan Zhenga512bbf2008-12-08 16:46:26 -05002781 for (i = 0; i < max_mirrors; i++) {
2782 bytenr = btrfs_sb_offset(i);
2783 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2784 break;
2785
2786 if (wait) {
2787 bh = __find_get_block(device->bdev, bytenr / 4096,
2788 BTRFS_SUPER_INFO_SIZE);
2789 BUG_ON(!bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002790 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002791 if (!buffer_uptodate(bh))
2792 errors++;
2793
2794 /* drop our reference */
2795 brelse(bh);
2796
2797 /* drop the reference from the wait == 0 run */
2798 brelse(bh);
2799 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002800 } else {
2801 btrfs_set_super_bytenr(sb, bytenr);
2802
2803 crc = ~(u32)0;
2804 crc = btrfs_csum_data(NULL, (char *)sb +
2805 BTRFS_CSUM_SIZE, crc,
2806 BTRFS_SUPER_INFO_SIZE -
2807 BTRFS_CSUM_SIZE);
2808 btrfs_csum_final(crc, sb->csum);
2809
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002810 /*
2811 * one reference for us, and we leave it for the
2812 * caller
2813 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002814 bh = __getblk(device->bdev, bytenr / 4096,
2815 BTRFS_SUPER_INFO_SIZE);
2816 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2817
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002818 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002819 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002820
2821 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05002822 lock_buffer(bh);
2823 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02002824 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002825 }
2826
Chris Mason387125f2011-11-18 15:07:51 -05002827 /*
2828 * we fua the first super. The others we allow
2829 * to go down lazy.
2830 */
Stefan Behrens21adbd52011-11-09 13:44:05 +01002831 ret = btrfsic_submit_bh(WRITE_FUA, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04002832 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05002833 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05002834 }
2835 return errors < i ? 0 : -1;
2836}
2837
Chris Mason387125f2011-11-18 15:07:51 -05002838/*
2839 * endio for the write_dev_flush, this will wake anyone waiting
2840 * for the barrier when it is done
2841 */
2842static void btrfs_end_empty_barrier(struct bio *bio, int err)
2843{
2844 if (err) {
2845 if (err == -EOPNOTSUPP)
2846 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
2847 clear_bit(BIO_UPTODATE, &bio->bi_flags);
2848 }
2849 if (bio->bi_private)
2850 complete(bio->bi_private);
2851 bio_put(bio);
2852}
2853
2854/*
2855 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
2856 * sent down. With wait == 1, it waits for the previous flush.
2857 *
2858 * any device where the flush fails with eopnotsupp are flagged as not-barrier
2859 * capable
2860 */
2861static int write_dev_flush(struct btrfs_device *device, int wait)
2862{
2863 struct bio *bio;
2864 int ret = 0;
2865
2866 if (device->nobarriers)
2867 return 0;
2868
2869 if (wait) {
2870 bio = device->flush_bio;
2871 if (!bio)
2872 return 0;
2873
2874 wait_for_completion(&device->flush_wait);
2875
2876 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Josef Bacik606686e2012-06-04 14:03:51 -04002877 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
2878 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05002879 device->nobarriers = 1;
2880 }
2881 if (!bio_flagged(bio, BIO_UPTODATE)) {
2882 ret = -EIO;
Stefan Behrens442a4f62012-05-25 16:06:08 +02002883 if (!bio_flagged(bio, BIO_EOPNOTSUPP))
2884 btrfs_dev_stat_inc_and_print(device,
2885 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05002886 }
2887
2888 /* drop the reference from the wait == 0 run */
2889 bio_put(bio);
2890 device->flush_bio = NULL;
2891
2892 return ret;
2893 }
2894
2895 /*
2896 * one reference for us, and we leave it for the
2897 * caller
2898 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01002899 device->flush_bio = NULL;
Chris Mason387125f2011-11-18 15:07:51 -05002900 bio = bio_alloc(GFP_NOFS, 0);
2901 if (!bio)
2902 return -ENOMEM;
2903
2904 bio->bi_end_io = btrfs_end_empty_barrier;
2905 bio->bi_bdev = device->bdev;
2906 init_completion(&device->flush_wait);
2907 bio->bi_private = &device->flush_wait;
2908 device->flush_bio = bio;
2909
2910 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002911 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05002912
2913 return 0;
2914}
2915
2916/*
2917 * send an empty flush down to each device in parallel,
2918 * then wait for them
2919 */
2920static int barrier_all_devices(struct btrfs_fs_info *info)
2921{
2922 struct list_head *head;
2923 struct btrfs_device *dev;
2924 int errors = 0;
2925 int ret;
2926
2927 /* send down all the barriers */
2928 head = &info->fs_devices->devices;
2929 list_for_each_entry_rcu(dev, head, dev_list) {
2930 if (!dev->bdev) {
2931 errors++;
2932 continue;
2933 }
2934 if (!dev->in_fs_metadata || !dev->writeable)
2935 continue;
2936
2937 ret = write_dev_flush(dev, 0);
2938 if (ret)
2939 errors++;
2940 }
2941
2942 /* wait for all the barriers */
2943 list_for_each_entry_rcu(dev, head, dev_list) {
2944 if (!dev->bdev) {
2945 errors++;
2946 continue;
2947 }
2948 if (!dev->in_fs_metadata || !dev->writeable)
2949 continue;
2950
2951 ret = write_dev_flush(dev, 1);
2952 if (ret)
2953 errors++;
2954 }
2955 if (errors)
2956 return -EIO;
2957 return 0;
2958}
2959
Yan Zhenga512bbf2008-12-08 16:46:26 -05002960int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04002961{
Chris Masone5e9a522009-06-10 15:17:02 -04002962 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04002963 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04002964 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04002965 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04002966 int ret;
2967 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04002968 int max_errors;
2969 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04002970 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04002971
David Sterba6c417612011-04-13 15:41:04 +02002972 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Masonf2984462008-04-10 16:19:33 -04002973 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002974 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04002975
David Sterba6c417612011-04-13 15:41:04 +02002976 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04002977 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04002978
Chris Mason174ba502011-05-27 10:03:58 -04002979 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04002980 head = &root->fs_info->fs_devices->devices;
Chris Mason387125f2011-11-18 15:07:51 -05002981
2982 if (do_barriers)
2983 barrier_all_devices(root->fs_info);
2984
Xiao Guangrong1f781602011-04-20 10:09:16 +00002985 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04002986 if (!dev->bdev) {
2987 total_errors++;
2988 continue;
2989 }
Yan Zheng2b820322008-11-17 21:11:30 -05002990 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04002991 continue;
2992
Yan Zheng2b820322008-11-17 21:11:30 -05002993 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04002994 btrfs_set_stack_device_type(dev_item, dev->type);
2995 btrfs_set_stack_device_id(dev_item, dev->devid);
2996 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2997 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2998 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2999 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3000 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3001 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003002 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003003
Chris Masona061fc82008-05-07 11:43:44 -04003004 flags = btrfs_super_flags(sb);
3005 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003006
Yan Zhenga512bbf2008-12-08 16:46:26 -05003007 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003008 if (ret)
3009 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003010 }
Chris Masona236aed2008-04-29 09:38:00 -04003011 if (total_errors > max_errors) {
Chris Masond3977122009-01-05 21:25:51 -05003012 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3013 total_errors);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003014
3015 /* This shouldn't happen. FUA is masked off if unsupported */
Chris Masona236aed2008-04-29 09:38:00 -04003016 BUG();
3017 }
Chris Masonf2984462008-04-10 16:19:33 -04003018
Yan Zhenga512bbf2008-12-08 16:46:26 -05003019 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003020 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003021 if (!dev->bdev)
3022 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003023 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003024 continue;
3025
Yan Zhenga512bbf2008-12-08 16:46:26 -05003026 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3027 if (ret)
3028 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003029 }
Chris Mason174ba502011-05-27 10:03:58 -04003030 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003031 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003032 btrfs_error(root->fs_info, -EIO,
3033 "%d errors while writing supers", total_errors);
3034 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003035 }
Chris Masonf2984462008-04-10 16:19:33 -04003036 return 0;
3037}
3038
Yan Zhenga512bbf2008-12-08 16:46:26 -05003039int write_ctree_super(struct btrfs_trans_handle *trans,
3040 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003041{
Chris Masone66f7092007-04-20 13:16:02 -04003042 int ret;
Chris Mason2c90e5d2007-04-02 10:50:19 -04003043
Yan Zhenga512bbf2008-12-08 16:46:26 -05003044 ret = write_all_supers(root, max_mirrors);
Chris Mason5f39d392007-10-15 16:14:19 -04003045 return ret;
Chris Masoncfaa7292007-02-21 17:04:57 -05003046}
3047
Jeff Mahoney143bede2012-03-01 14:56:26 +01003048void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003049{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003050 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003051 radix_tree_delete(&fs_info->fs_roots_radix,
3052 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003053 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003054
3055 if (btrfs_root_refs(&root->root_item) == 0)
3056 synchronize_srcu(&fs_info->subvol_srcu);
3057
Li Zefan581bb052011-04-20 10:06:11 +08003058 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3059 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003060 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003061}
3062
3063static void free_fs_root(struct btrfs_root *root)
3064{
Li Zefan82d59022011-04-20 10:33:24 +08003065 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003066 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Al Viro0ee5dc62011-07-07 15:44:25 -04003067 if (root->anon_dev)
3068 free_anon_bdev(root->anon_dev);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003069 free_extent_buffer(root->node);
3070 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003071 kfree(root->free_ino_ctl);
3072 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003073 kfree(root->name);
Chris Mason2619ba12007-04-10 16:58:11 -04003074 kfree(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003075}
3076
Jeff Mahoney143bede2012-03-01 14:56:26 +01003077static void del_fs_roots(struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04003078{
3079 int ret;
3080 struct btrfs_root *gang[8];
3081 int i;
3082
Yan, Zheng76dda932009-09-21 16:00:26 -04003083 while (!list_empty(&fs_info->dead_roots)) {
3084 gang[0] = list_entry(fs_info->dead_roots.next,
3085 struct btrfs_root, root_list);
3086 list_del(&gang[0]->root_list);
3087
3088 if (gang[0]->in_radix) {
3089 btrfs_free_fs_root(fs_info, gang[0]);
3090 } else {
3091 free_extent_buffer(gang[0]->node);
3092 free_extent_buffer(gang[0]->commit_root);
3093 kfree(gang[0]);
3094 }
3095 }
3096
Chris Masond3977122009-01-05 21:25:51 -05003097 while (1) {
Chris Mason0f7d52f2007-04-09 10:42:37 -04003098 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3099 (void **)gang, 0,
3100 ARRAY_SIZE(gang));
3101 if (!ret)
3102 break;
Chris Mason2619ba12007-04-10 16:58:11 -04003103 for (i = 0; i < ret; i++)
Chris Mason5eda7b52007-06-22 14:16:25 -04003104 btrfs_free_fs_root(fs_info, gang[i]);
Chris Mason0f7d52f2007-04-09 10:42:37 -04003105 }
Chris Mason0f7d52f2007-04-09 10:42:37 -04003106}
Chris Masonb4100d62007-04-12 12:14:00 -04003107
Yan Zhengc146afa2008-11-12 14:34:12 -05003108int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3109{
3110 u64 root_objectid = 0;
3111 struct btrfs_root *gang[8];
3112 int i;
3113 int ret;
3114
3115 while (1) {
3116 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3117 (void **)gang, root_objectid,
3118 ARRAY_SIZE(gang));
3119 if (!ret)
3120 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003121
3122 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05003123 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003124 int err;
3125
Yan Zhengc146afa2008-11-12 14:34:12 -05003126 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003127 err = btrfs_orphan_cleanup(gang[i]);
3128 if (err)
3129 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003130 }
3131 root_objectid++;
3132 }
3133 return 0;
3134}
3135
3136int btrfs_commit_super(struct btrfs_root *root)
3137{
3138 struct btrfs_trans_handle *trans;
3139 int ret;
3140
3141 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003142 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003143 btrfs_clean_old_snapshots(root);
3144 mutex_unlock(&root->fs_info->cleaner_mutex);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003145
3146 /* wait until ongoing cleanup work done */
3147 down_write(&root->fs_info->cleanup_work_sem);
3148 up_write(&root->fs_info->cleanup_work_sem);
3149
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003150 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003151 if (IS_ERR(trans))
3152 return PTR_ERR(trans);
Yan Zhengc146afa2008-11-12 14:34:12 -05003153 ret = btrfs_commit_transaction(trans, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003154 if (ret)
3155 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003156 /* run commit again to drop the original snapshot */
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003157 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003158 if (IS_ERR(trans))
3159 return PTR_ERR(trans);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003160 ret = btrfs_commit_transaction(trans, root);
3161 if (ret)
3162 return ret;
Yan Zhengc146afa2008-11-12 14:34:12 -05003163 ret = btrfs_write_and_wait_transaction(NULL, root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003164 if (ret) {
3165 btrfs_error(root->fs_info, ret,
3166 "Failed to sync btree inode to disk.");
3167 return ret;
3168 }
Yan Zhengc146afa2008-11-12 14:34:12 -05003169
Yan Zhenga512bbf2008-12-08 16:46:26 -05003170 ret = write_ctree_super(NULL, root, 0);
Yan Zhengc146afa2008-11-12 14:34:12 -05003171 return ret;
3172}
3173
Chris Masone20d96d2007-03-22 12:13:20 -04003174int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003175{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003176 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003177 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003178
Chris Masonfacda1e2007-06-08 18:11:48 -04003179 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003180 smp_mb();
3181
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003182 /* pause restriper - we want to resume on mount */
3183 btrfs_pause_balance(root->fs_info);
3184
Arne Jansena2de7332011-03-08 14:14:00 +01003185 btrfs_scrub_cancel(root);
Chris Mason4cb53002011-05-24 15:35:30 -04003186
3187 /* wait for any defraggers to finish */
3188 wait_event(fs_info->transaction_wait,
3189 (atomic_read(&fs_info->defrag_running) == 0));
3190
3191 /* clear out the rbtree of defraggable inodes */
Al Viroe3029d92011-11-17 00:56:18 -05003192 btrfs_run_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003193
Yan Zhengc146afa2008-11-12 14:34:12 -05003194 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003195 ret = btrfs_commit_super(root);
3196 if (ret)
3197 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3198 }
3199
Stefan Behrens68ce9682012-08-01 05:45:52 -06003200 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
3201 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003202
Josef Bacik300e4f82011-08-29 14:06:00 -04003203 btrfs_put_block_group_cache(fs_info);
3204
Al Viroe3029d92011-11-17 00:56:18 -05003205 kthread_stop(fs_info->transaction_kthread);
3206 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003207
Yan Zhengf25784b2009-07-28 08:41:57 -04003208 fs_info->closing = 2;
3209 smp_mb();
3210
Arne Jansenbcef60f2011-09-13 15:23:30 +02003211 btrfs_free_qgroup_config(root->fs_info);
3212
Chris Masonb0c68f82008-01-31 11:05:37 -05003213 if (fs_info->delalloc_bytes) {
Chris Masond3977122009-01-05 21:25:51 -05003214 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
Joel Becker21380932009-04-21 12:38:29 -07003215 (unsigned long long)fs_info->delalloc_bytes);
Chris Masonb0c68f82008-01-31 11:05:37 -05003216 }
Yan Zheng31153d82008-07-28 15:32:19 -04003217 if (fs_info->total_ref_cache_size) {
Chris Masond3977122009-01-05 21:25:51 -05003218 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
3219 (unsigned long long)fs_info->total_ref_cache_size);
Yan Zheng31153d82008-07-28 15:32:19 -04003220 }
Yanbcc63ab2008-07-30 16:29:20 -04003221
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003222 free_extent_buffer(fs_info->extent_root->node);
3223 free_extent_buffer(fs_info->extent_root->commit_root);
3224 free_extent_buffer(fs_info->tree_root->node);
3225 free_extent_buffer(fs_info->tree_root->commit_root);
Al Viroe3029d92011-11-17 00:56:18 -05003226 free_extent_buffer(fs_info->chunk_root->node);
3227 free_extent_buffer(fs_info->chunk_root->commit_root);
3228 free_extent_buffer(fs_info->dev_root->node);
3229 free_extent_buffer(fs_info->dev_root->commit_root);
3230 free_extent_buffer(fs_info->csum_root->node);
3231 free_extent_buffer(fs_info->csum_root->commit_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003232 if (fs_info->quota_root) {
3233 free_extent_buffer(fs_info->quota_root->node);
3234 free_extent_buffer(fs_info->quota_root->commit_root);
3235 }
Chris Masond20f7042008-12-08 16:58:54 -05003236
Al Viroe3029d92011-11-17 00:56:18 -05003237 btrfs_free_block_groups(fs_info);
Yan Zhengc146afa2008-11-12 14:34:12 -05003238
Chris Mason0f7d52f2007-04-09 10:42:37 -04003239 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003240
Yan Zhengc146afa2008-11-12 14:34:12 -05003241 iput(fs_info->btree_inode);
Chris Mason9ad6b7bc2008-04-18 16:11:30 -04003242
Chris Mason61d92c32009-10-02 19:11:56 -04003243 btrfs_stop_workers(&fs_info->generic_worker);
Chris Mason247e7432008-07-17 12:53:51 -04003244 btrfs_stop_workers(&fs_info->fixup_workers);
Chris Mason771ed682008-11-06 22:02:51 -05003245 btrfs_stop_workers(&fs_info->delalloc_workers);
Chris Mason8b712842008-06-11 16:50:36 -04003246 btrfs_stop_workers(&fs_info->workers);
3247 btrfs_stop_workers(&fs_info->endio_workers);
Chris Masond20f7042008-12-08 16:58:54 -05003248 btrfs_stop_workers(&fs_info->endio_meta_workers);
Chris Masoncad321a2008-12-17 14:51:42 -05003249 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
Chris Masone6dcd2d2008-07-17 12:53:50 -04003250 btrfs_stop_workers(&fs_info->endio_write_workers);
Josef Bacik0cb59c92010-07-02 12:14:14 -04003251 btrfs_stop_workers(&fs_info->endio_freespace_worker);
Chris Mason1cc127b2008-06-12 14:46:17 -04003252 btrfs_stop_workers(&fs_info->submit_workers);
Miao Xie16cdcec2011-04-22 18:12:22 +08003253 btrfs_stop_workers(&fs_info->delayed_workers);
Josef Bacikbab39bf2011-06-30 14:42:28 -04003254 btrfs_stop_workers(&fs_info->caching_workers);
Arne Jansen90519d62011-05-23 14:30:00 +02003255 btrfs_stop_workers(&fs_info->readahead_workers);
Chris Masond6bfde82008-04-30 13:59:35 -04003256
Stefan Behrens21adbd52011-11-09 13:44:05 +01003257#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3258 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3259 btrfsic_unmount(root, fs_info->fs_devices);
3260#endif
3261
Chris Masondfe25022008-05-13 13:46:40 -04003262 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003263 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003264
Chris Mason04160082008-03-26 10:28:07 -04003265 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003266 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003267
Chris Masoneb60cea2007-02-02 09:18:22 -05003268 return 0;
3269}
3270
Chris Masonb9fab912012-05-06 07:23:47 -04003271int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3272 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003273{
Chris Mason1259ab72008-05-12 13:39:03 -04003274 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003275 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003276
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003277 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003278 if (!ret)
3279 return ret;
3280
3281 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003282 parent_transid, atomic);
3283 if (ret == -EAGAIN)
3284 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003285 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003286}
Chris Mason6702ed42007-08-07 16:15:09 -04003287
Chris Mason5f39d392007-10-15 16:14:19 -04003288int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3289{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003290 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003291}
3292
3293void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3294{
Chris Mason727011e2010-08-06 13:21:20 -04003295 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Mason5f39d392007-10-15 16:14:19 -04003296 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003297 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003298
Chris Masonb9447ef82009-03-09 11:45:38 -04003299 btrfs_assert_tree_locked(buf);
Chris Masonccd467d2007-06-28 15:57:36 -04003300 if (transid != root->fs_info->generation) {
Chris Masond3977122009-01-05 21:25:51 -05003301 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
3302 "found %llu running %llu\n",
Chris Masondb945352007-10-15 16:15:53 -04003303 (unsigned long long)buf->start,
Chris Masond3977122009-01-05 21:25:51 -05003304 (unsigned long long)transid,
3305 (unsigned long long)root->fs_info->generation);
Chris Masonccd467d2007-06-28 15:57:36 -04003306 WARN_ON(1);
3307 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003308 was_dirty = set_extent_buffer_dirty(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003309 if (!was_dirty) {
3310 spin_lock(&root->fs_info->delalloc_lock);
3311 root->fs_info->dirty_metadata_bytes += buf->len;
3312 spin_unlock(&root->fs_info->delalloc_lock);
3313 }
Chris Masoneb60cea2007-02-02 09:18:22 -05003314}
3315
Chris Masond3c2fdcf2007-09-17 10:58:06 -04003316void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
Chris Mason35b7e472007-05-02 15:53:43 -04003317{
Chris Mason188de642008-05-09 11:52:25 -04003318 /*
3319 * looks as though older kernels can get into trouble with
3320 * this code, they end up stuck in balance_dirty_pages forever
3321 */
Chris Masond6bfde82008-04-30 13:59:35 -04003322 u64 num_dirty;
Chris Mason771ed682008-11-06 22:02:51 -05003323 unsigned long thresh = 32 * 1024 * 1024;
Chris Masond6bfde82008-04-30 13:59:35 -04003324
Jens Axboe6933c022009-03-17 09:36:37 +01003325 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003326 return;
3327
Miao Xie16cdcec2011-04-22 18:12:22 +08003328 btrfs_balance_delayed_items(root);
3329
3330 num_dirty = root->fs_info->dirty_metadata_bytes;
3331
3332 if (num_dirty > thresh) {
3333 balance_dirty_pages_ratelimited_nr(
3334 root->fs_info->btree_inode->i_mapping, 1);
3335 }
3336 return;
3337}
3338
3339void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
3340{
3341 /*
3342 * looks as though older kernels can get into trouble with
3343 * this code, they end up stuck in balance_dirty_pages forever
3344 */
3345 u64 num_dirty;
3346 unsigned long thresh = 32 * 1024 * 1024;
3347
3348 if (current->flags & PF_MEMALLOC)
3349 return;
3350
Chris Mason585ad2c2009-05-18 10:41:58 -04003351 num_dirty = root->fs_info->dirty_metadata_bytes;
3352
Chris Masond6bfde82008-04-30 13:59:35 -04003353 if (num_dirty > thresh) {
3354 balance_dirty_pages_ratelimited_nr(
Chris Masond7fc6402008-02-18 12:12:38 -05003355 root->fs_info->btree_inode->i_mapping, 1);
Chris Masond6bfde82008-04-30 13:59:35 -04003356 }
Chris Mason188de642008-05-09 11:52:25 -04003357 return;
Chris Mason35b7e472007-05-02 15:53:43 -04003358}
Chris Mason6b800532007-10-15 16:17:34 -04003359
Chris Masonca7a79a2008-05-12 12:59:19 -04003360int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003361{
Chris Mason727011e2010-08-06 13:21:20 -04003362 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003363 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003364}
Chris Mason0da54682007-11-07 21:08:01 -05003365
Arne Jansen20897f52011-09-13 12:44:20 +02003366int btree_lock_page_hook(struct page *page, void *data,
Chris Mason01d658f2011-11-01 10:08:06 -04003367 void (*flush_fn)(void *))
Chris Mason4bef0842008-09-08 11:18:08 -04003368{
3369 struct inode *inode = page->mapping->host;
Chris Masonb9473432009-03-13 11:00:37 -04003370 struct btrfs_root *root = BTRFS_I(inode)->root;
Chris Mason4bef0842008-09-08 11:18:08 -04003371 struct extent_buffer *eb;
Chris Mason4bef0842008-09-08 11:18:08 -04003372
Josef Bacik4f2de97a2012-03-07 16:20:05 -05003373 /*
3374 * We culled this eb but the page is still hanging out on the mapping,
3375 * carry on.
3376 */
3377 if (!PagePrivate(page))
Chris Mason4bef0842008-09-08 11:18:08 -04003378 goto out;
3379
Josef Bacik4f2de97a2012-03-07 16:20:05 -05003380 eb = (struct extent_buffer *)page->private;
3381 if (!eb) {
3382 WARN_ON(1);
3383 goto out;
3384 }
3385 if (page != eb->pages[0])
Chris Mason4bef0842008-09-08 11:18:08 -04003386 goto out;
3387
Chris Mason01d658f2011-11-01 10:08:06 -04003388 if (!btrfs_try_tree_write_lock(eb)) {
3389 flush_fn(data);
3390 btrfs_tree_lock(eb);
3391 }
Chris Mason4bef0842008-09-08 11:18:08 -04003392 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonb9473432009-03-13 11:00:37 -04003393
3394 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3395 spin_lock(&root->fs_info->delalloc_lock);
3396 if (root->fs_info->dirty_metadata_bytes >= eb->len)
3397 root->fs_info->dirty_metadata_bytes -= eb->len;
3398 else
3399 WARN_ON(1);
3400 spin_unlock(&root->fs_info->delalloc_lock);
3401 }
3402
Chris Mason4bef0842008-09-08 11:18:08 -04003403 btrfs_tree_unlock(eb);
Chris Mason4bef0842008-09-08 11:18:08 -04003404out:
Chris Mason01d658f2011-11-01 10:08:06 -04003405 if (!trylock_page(page)) {
3406 flush_fn(data);
3407 lock_page(page);
3408 }
Chris Mason4bef0842008-09-08 11:18:08 -04003409 return 0;
3410}
3411
David Sterbafcd1f0652012-03-06 00:06:18 +01003412static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003413 int read_only)
3414{
David Sterbafcd1f0652012-03-06 00:06:18 +01003415 if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) {
3416 printk(KERN_ERR "btrfs: unsupported checksum algorithm\n");
3417 return -EINVAL;
3418 }
liuboacce9522011-01-06 19:30:25 +08003419
David Sterbafcd1f0652012-03-06 00:06:18 +01003420 if (read_only)
3421 return 0;
3422
David Sterbafcd1f0652012-03-06 00:06:18 +01003423 return 0;
liuboacce9522011-01-06 19:30:25 +08003424}
3425
Stefan Behrens68ce9682012-08-01 05:45:52 -06003426void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003427{
liuboacce9522011-01-06 19:30:25 +08003428 mutex_lock(&root->fs_info->cleaner_mutex);
3429 btrfs_run_delayed_iputs(root);
3430 mutex_unlock(&root->fs_info->cleaner_mutex);
3431
3432 down_write(&root->fs_info->cleanup_work_sem);
3433 up_write(&root->fs_info->cleanup_work_sem);
3434
3435 /* cleanup FS via transaction */
3436 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003437}
3438
Jeff Mahoney143bede2012-03-01 14:56:26 +01003439static void btrfs_destroy_ordered_operations(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003440{
3441 struct btrfs_inode *btrfs_inode;
3442 struct list_head splice;
3443
3444 INIT_LIST_HEAD(&splice);
3445
3446 mutex_lock(&root->fs_info->ordered_operations_mutex);
3447 spin_lock(&root->fs_info->ordered_extent_lock);
3448
3449 list_splice_init(&root->fs_info->ordered_operations, &splice);
3450 while (!list_empty(&splice)) {
3451 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3452 ordered_operations);
3453
3454 list_del_init(&btrfs_inode->ordered_operations);
3455
3456 btrfs_invalidate_inodes(btrfs_inode->root);
3457 }
3458
3459 spin_unlock(&root->fs_info->ordered_extent_lock);
3460 mutex_unlock(&root->fs_info->ordered_operations_mutex);
liuboacce9522011-01-06 19:30:25 +08003461}
3462
Jeff Mahoney143bede2012-03-01 14:56:26 +01003463static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003464{
3465 struct list_head splice;
3466 struct btrfs_ordered_extent *ordered;
3467 struct inode *inode;
3468
3469 INIT_LIST_HEAD(&splice);
3470
3471 spin_lock(&root->fs_info->ordered_extent_lock);
3472
3473 list_splice_init(&root->fs_info->ordered_extents, &splice);
3474 while (!list_empty(&splice)) {
3475 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
3476 root_extent_list);
3477
3478 list_del_init(&ordered->root_extent_list);
3479 atomic_inc(&ordered->refs);
3480
3481 /* the inode may be getting freed (in sys_unlink path). */
3482 inode = igrab(ordered->inode);
3483
3484 spin_unlock(&root->fs_info->ordered_extent_lock);
3485 if (inode)
3486 iput(inode);
3487
3488 atomic_set(&ordered->refs, 1);
3489 btrfs_put_ordered_extent(ordered);
3490
3491 spin_lock(&root->fs_info->ordered_extent_lock);
3492 }
3493
3494 spin_unlock(&root->fs_info->ordered_extent_lock);
liuboacce9522011-01-06 19:30:25 +08003495}
3496
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003497int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3498 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003499{
3500 struct rb_node *node;
3501 struct btrfs_delayed_ref_root *delayed_refs;
3502 struct btrfs_delayed_ref_node *ref;
3503 int ret = 0;
3504
3505 delayed_refs = &trans->delayed_refs;
3506
3507 spin_lock(&delayed_refs->lock);
3508 if (delayed_refs->num_entries == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003509 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003510 printk(KERN_INFO "delayed_refs has NO entry\n");
3511 return ret;
3512 }
3513
Josef Bacikb939d1a2012-05-31 11:06:33 -04003514 while ((node = rb_first(&delayed_refs->root)) != NULL) {
liuboacce9522011-01-06 19:30:25 +08003515 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
liuboacce9522011-01-06 19:30:25 +08003516
3517 atomic_set(&ref->refs, 1);
3518 if (btrfs_delayed_ref_is_head(ref)) {
3519 struct btrfs_delayed_ref_head *head;
3520
3521 head = btrfs_delayed_node_to_head(ref);
Josef Bacikb939d1a2012-05-31 11:06:33 -04003522 if (!mutex_trylock(&head->mutex)) {
3523 atomic_inc(&ref->refs);
3524 spin_unlock(&delayed_refs->lock);
3525
3526 /* Need to wait for the delayed ref to run */
3527 mutex_lock(&head->mutex);
3528 mutex_unlock(&head->mutex);
3529 btrfs_put_delayed_ref(ref);
3530
Josef Bacike18fca72012-06-18 07:23:18 -06003531 spin_lock(&delayed_refs->lock);
Josef Bacikb939d1a2012-05-31 11:06:33 -04003532 continue;
3533 }
3534
liuboacce9522011-01-06 19:30:25 +08003535 kfree(head->extent_op);
3536 delayed_refs->num_heads--;
3537 if (list_empty(&head->cluster))
3538 delayed_refs->num_heads_ready--;
3539 list_del_init(&head->cluster);
liuboacce9522011-01-06 19:30:25 +08003540 }
Josef Bacikb939d1a2012-05-31 11:06:33 -04003541 ref->in_tree = 0;
3542 rb_erase(&ref->rb_node, &delayed_refs->root);
3543 delayed_refs->num_entries--;
3544
liuboacce9522011-01-06 19:30:25 +08003545 spin_unlock(&delayed_refs->lock);
3546 btrfs_put_delayed_ref(ref);
3547
3548 cond_resched();
3549 spin_lock(&delayed_refs->lock);
3550 }
3551
3552 spin_unlock(&delayed_refs->lock);
3553
3554 return ret;
3555}
3556
Jeff Mahoney143bede2012-03-01 14:56:26 +01003557static void btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
liuboacce9522011-01-06 19:30:25 +08003558{
3559 struct btrfs_pending_snapshot *snapshot;
3560 struct list_head splice;
3561
3562 INIT_LIST_HEAD(&splice);
3563
3564 list_splice_init(&t->pending_snapshots, &splice);
3565
3566 while (!list_empty(&splice)) {
3567 snapshot = list_entry(splice.next,
3568 struct btrfs_pending_snapshot,
3569 list);
3570
3571 list_del_init(&snapshot->list);
3572
3573 kfree(snapshot);
3574 }
liuboacce9522011-01-06 19:30:25 +08003575}
3576
Jeff Mahoney143bede2012-03-01 14:56:26 +01003577static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003578{
3579 struct btrfs_inode *btrfs_inode;
3580 struct list_head splice;
3581
3582 INIT_LIST_HEAD(&splice);
3583
liuboacce9522011-01-06 19:30:25 +08003584 spin_lock(&root->fs_info->delalloc_lock);
David Sterba5be76752011-06-09 10:02:51 +00003585 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003586
3587 while (!list_empty(&splice)) {
3588 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3589 delalloc_inodes);
3590
3591 list_del_init(&btrfs_inode->delalloc_inodes);
3592
3593 btrfs_invalidate_inodes(btrfs_inode->root);
3594 }
3595
3596 spin_unlock(&root->fs_info->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003597}
3598
3599static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3600 struct extent_io_tree *dirty_pages,
3601 int mark)
3602{
3603 int ret;
3604 struct page *page;
3605 struct inode *btree_inode = root->fs_info->btree_inode;
3606 struct extent_buffer *eb;
3607 u64 start = 0;
3608 u64 end;
3609 u64 offset;
3610 unsigned long index;
3611
3612 while (1) {
3613 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
3614 mark);
3615 if (ret)
3616 break;
3617
3618 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3619 while (start <= end) {
3620 index = start >> PAGE_CACHE_SHIFT;
3621 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
3622 page = find_get_page(btree_inode->i_mapping, index);
3623 if (!page)
3624 continue;
3625 offset = page_offset(page);
3626
3627 spin_lock(&dirty_pages->buffer_lock);
3628 eb = radix_tree_lookup(
3629 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
3630 offset >> PAGE_CACHE_SHIFT);
3631 spin_unlock(&dirty_pages->buffer_lock);
Josef Bacikee670f0a2012-05-31 15:54:30 -04003632 if (eb)
liuboacce9522011-01-06 19:30:25 +08003633 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3634 &eb->bflags);
liuboacce9522011-01-06 19:30:25 +08003635 if (PageWriteback(page))
3636 end_page_writeback(page);
3637
3638 lock_page(page);
3639 if (PageDirty(page)) {
3640 clear_page_dirty_for_io(page);
3641 spin_lock_irq(&page->mapping->tree_lock);
3642 radix_tree_tag_clear(&page->mapping->page_tree,
3643 page_index(page),
3644 PAGECACHE_TAG_DIRTY);
3645 spin_unlock_irq(&page->mapping->tree_lock);
3646 }
3647
liuboacce9522011-01-06 19:30:25 +08003648 unlock_page(page);
Josef Bacikee670f0a2012-05-31 15:54:30 -04003649 page_cache_release(page);
liuboacce9522011-01-06 19:30:25 +08003650 }
3651 }
3652
3653 return ret;
3654}
3655
3656static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3657 struct extent_io_tree *pinned_extents)
3658{
3659 struct extent_io_tree *unpin;
3660 u64 start;
3661 u64 end;
3662 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06003663 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08003664
3665 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06003666again:
liuboacce9522011-01-06 19:30:25 +08003667 while (1) {
3668 ret = find_first_extent_bit(unpin, 0, &start, &end,
3669 EXTENT_DIRTY);
3670 if (ret)
3671 break;
3672
3673 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00003674 if (btrfs_test_opt(root, DISCARD))
3675 ret = btrfs_error_discard_extent(root, start,
3676 end + 1 - start,
3677 NULL);
liuboacce9522011-01-06 19:30:25 +08003678
3679 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3680 btrfs_error_unpin_extent_range(root, start, end);
3681 cond_resched();
3682 }
3683
Liu Boed0eaa12012-06-14 02:23:21 -06003684 if (loop) {
3685 if (unpin == &root->fs_info->freed_extents[0])
3686 unpin = &root->fs_info->freed_extents[1];
3687 else
3688 unpin = &root->fs_info->freed_extents[0];
3689 loop = false;
3690 goto again;
3691 }
3692
liuboacce9522011-01-06 19:30:25 +08003693 return 0;
3694}
3695
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003696void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3697 struct btrfs_root *root)
3698{
3699 btrfs_destroy_delayed_refs(cur_trans, root);
3700 btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
3701 cur_trans->dirty_pages.dirty_bytes);
3702
3703 /* FIXME: cleanup wait for commit */
3704 cur_trans->in_commit = 1;
3705 cur_trans->blocked = 1;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003706 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003707
3708 cur_trans->blocked = 0;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003709 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003710
3711 cur_trans->commit_done = 1;
Josef Bacikd7096fc2012-05-31 15:49:57 -04003712 wake_up(&cur_trans->commit_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003713
Miao Xie67cde342012-06-14 02:23:22 -06003714 btrfs_destroy_delayed_inodes(root);
3715 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003716
3717 btrfs_destroy_pending_snapshots(cur_trans);
3718
3719 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
3720 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06003721 btrfs_destroy_pinned_extent(root,
3722 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01003723
3724 /*
3725 memset(cur_trans, 0, sizeof(*cur_trans));
3726 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3727 */
3728}
3729
3730int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003731{
3732 struct btrfs_transaction *t;
3733 LIST_HEAD(list);
3734
liuboacce9522011-01-06 19:30:25 +08003735 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3736
Josef Bacika4abeea2011-04-11 17:25:13 -04003737 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003738 list_splice_init(&root->fs_info->trans_list, &list);
Josef Bacika4abeea2011-04-11 17:25:13 -04003739 root->fs_info->trans_no_join = 1;
3740 spin_unlock(&root->fs_info->trans_lock);
3741
liuboacce9522011-01-06 19:30:25 +08003742 while (!list_empty(&list)) {
3743 t = list_entry(list.next, struct btrfs_transaction, list);
3744 if (!t)
3745 break;
3746
3747 btrfs_destroy_ordered_operations(root);
3748
3749 btrfs_destroy_ordered_extents(root);
3750
3751 btrfs_destroy_delayed_refs(t, root);
3752
3753 btrfs_block_rsv_release(root,
3754 &root->fs_info->trans_block_rsv,
3755 t->dirty_pages.dirty_bytes);
3756
3757 /* FIXME: cleanup wait for commit */
3758 t->in_commit = 1;
3759 t->blocked = 1;
Josef Bacik66657b32012-08-01 15:36:24 -04003760 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003761 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3762 wake_up(&root->fs_info->transaction_blocked_wait);
3763
3764 t->blocked = 0;
Josef Bacik66657b32012-08-01 15:36:24 -04003765 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003766 if (waitqueue_active(&root->fs_info->transaction_wait))
3767 wake_up(&root->fs_info->transaction_wait);
liuboacce9522011-01-06 19:30:25 +08003768
liuboacce9522011-01-06 19:30:25 +08003769 t->commit_done = 1;
Josef Bacik66657b32012-08-01 15:36:24 -04003770 smp_mb();
liuboacce9522011-01-06 19:30:25 +08003771 if (waitqueue_active(&t->commit_wait))
3772 wake_up(&t->commit_wait);
liuboacce9522011-01-06 19:30:25 +08003773
Miao Xie67cde342012-06-14 02:23:22 -06003774 btrfs_destroy_delayed_inodes(root);
3775 btrfs_assert_delayed_root_empty(root);
3776
liuboacce9522011-01-06 19:30:25 +08003777 btrfs_destroy_pending_snapshots(t);
3778
3779 btrfs_destroy_delalloc_inodes(root);
3780
Josef Bacika4abeea2011-04-11 17:25:13 -04003781 spin_lock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003782 root->fs_info->running_transaction = NULL;
Josef Bacika4abeea2011-04-11 17:25:13 -04003783 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003784
3785 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3786 EXTENT_DIRTY);
3787
3788 btrfs_destroy_pinned_extent(root,
3789 root->fs_info->pinned_extents);
3790
Josef Bacik13c5a932011-04-11 15:45:29 -04003791 atomic_set(&t->use_count, 0);
liuboacce9522011-01-06 19:30:25 +08003792 list_del_init(&t->list);
3793 memset(t, 0, sizeof(*t));
3794 kmem_cache_free(btrfs_transaction_cachep, t);
3795 }
3796
Josef Bacika4abeea2011-04-11 17:25:13 -04003797 spin_lock(&root->fs_info->trans_lock);
3798 root->fs_info->trans_no_join = 0;
3799 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08003800 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08003801
3802 return 0;
3803}
3804
Chris Masond1310b22008-01-24 16:13:08 -05003805static struct extent_io_ops btree_extent_io_ops = {
Chris Mason4bef0842008-09-08 11:18:08 -04003806 .write_cache_pages_lock_hook = btree_lock_page_hook,
Chris Masonce9adaa2008-04-09 16:28:12 -04003807 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02003808 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04003809 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04003810 /* note we're sharing with inode.c for the merge bio hook */
3811 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05003812};