blob: 0abf4b0a9010da69a18d8df9245851349eba3e1f [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050030#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020031#include <linux/ratelimit.h>
Stefan Behrens6463fe52013-04-19 15:08:05 +000032#include <linux/uuid.h>
Stefan Behrens803b2f52013-08-15 17:11:21 +020033#include <linux/semaphore.h>
David Sterba7e75bf32011-03-18 22:56:43 +000034#include <asm/unaligned.h>
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +000037#include "hash.h"
Chris Masone089f052007-03-16 16:20:31 -040038#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040039#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040040#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040041#include "print-tree.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080045#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010046#include "check-integrity.h"
Josef Bacik606686e2012-06-04 14:03:51 -040047#include "rcu-string.h"
Stefan Behrens8dabb742012-11-06 13:15:27 +010048#include "dev-replace.h"
David Woodhouse53b381b2013-01-29 18:40:14 -050049#include "raid56.h"
Jeff Mahoney5ac1d202013-11-01 13:06:58 -040050#include "sysfs.h"
Josef Bacikfcebe452014-05-13 17:30:47 -070051#include "qgroup.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050052
Josef Bacikde0022b2012-09-25 14:25:58 -040053#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
Chris Masond1310b22008-01-24 16:13:08 -050057static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040058static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040059static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010060static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080061 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010062static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080063static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010065static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080066static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 struct extent_io_tree *dirty_pages,
68 int mark);
69static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000071static int btrfs_cleanup_transaction(struct btrfs_root *root);
72static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040073
Chris Masond352ac62008-09-29 15:18:18 -040074/*
David Sterba97eb6b62014-07-30 00:55:42 +020075 * btrfs_end_io_wq structs are used to do processing in task context when an IO
76 * is complete. This is used during reads to verify checksums, and it is used
Chris Masond352ac62008-09-29 15:18:18 -040077 * by writes to insert metadata for new file extents after IO is complete.
78 */
David Sterba97eb6b62014-07-30 00:55:42 +020079struct btrfs_end_io_wq {
Chris Masonce9adaa2008-04-09 16:28:12 -040080 struct bio *bio;
81 bio_end_io_t *end_io;
82 void *private;
83 struct btrfs_fs_info *info;
84 int error;
David Sterbabfebd8b2014-07-30 00:25:45 +020085 enum btrfs_wq_endio_type metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040086 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040087 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040088};
Chris Mason0da54682007-11-07 21:08:01 -050089
David Sterba97eb6b62014-07-30 00:55:42 +020090static struct kmem_cache *btrfs_end_io_wq_cache;
91
92int __init btrfs_end_io_wq_init(void)
93{
94 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
95 sizeof(struct btrfs_end_io_wq),
96 0,
97 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
98 NULL);
99 if (!btrfs_end_io_wq_cache)
100 return -ENOMEM;
101 return 0;
102}
103
104void btrfs_end_io_wq_exit(void)
105{
106 if (btrfs_end_io_wq_cache)
107 kmem_cache_destroy(btrfs_end_io_wq_cache);
108}
109
Chris Masond352ac62008-09-29 15:18:18 -0400110/*
111 * async submit bios are used to offload expensive checksumming
112 * onto the worker threads. They checksum file and metadata bios
113 * just before they are sent down the IO stack.
114 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400115struct async_submit_bio {
116 struct inode *inode;
117 struct bio *bio;
118 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500119 extent_submit_bio_hook_t *submit_bio_start;
120 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400121 int rw;
122 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400123 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400124 /*
125 * bio_offset is optional, can be used if the pages in the bio
126 * can't tell us where in the file the bio should go
127 */
128 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400129 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100130 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400131};
132
Chris Mason85d4e462011-07-26 16:11:19 -0400133/*
134 * Lockdep class keys for extent_buffer->lock's in this root. For a given
135 * eb, the lockdep key is determined by the btrfs_root it belongs to and
136 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500137 *
Chris Mason85d4e462011-07-26 16:11:19 -0400138 * Different roots are used for different purposes and may nest inside each
139 * other and they require separate keysets. As lockdep keys should be
140 * static, assign keysets according to the purpose of the root as indicated
141 * by btrfs_root->objectid. This ensures that all special purpose roots
142 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500143 *
Chris Mason85d4e462011-07-26 16:11:19 -0400144 * Lock-nesting across peer nodes is always done with the immediate parent
145 * node locked thus preventing deadlock. As lockdep doesn't know this, use
146 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500147 *
Chris Mason85d4e462011-07-26 16:11:19 -0400148 * The key is set by the readpage_end_io_hook after the buffer has passed
149 * csum validation but before the pages are unlocked. It is also set by
150 * btrfs_init_new_buffer on freshly allocated blocks.
151 *
152 * We also add a check to make sure the highest level of the tree is the
153 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
154 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500155 */
156#ifdef CONFIG_DEBUG_LOCK_ALLOC
157# if BTRFS_MAX_LEVEL != 8
158# error
159# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400160
161static struct btrfs_lockdep_keyset {
162 u64 id; /* root objectid */
163 const char *name_stem; /* lock name stem */
164 char names[BTRFS_MAX_LEVEL + 1][20];
165 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
166} btrfs_lockdep_keysets[] = {
167 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
168 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
169 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
170 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
171 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
172 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000173 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400174 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
175 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
176 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200177 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
Chris Mason85d4e462011-07-26 16:11:19 -0400178 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500179};
Chris Mason85d4e462011-07-26 16:11:19 -0400180
181void __init btrfs_init_lockdep(void)
182{
183 int i, j;
184
185 /* initialize lockdep class names */
186 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
187 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
188
189 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
190 snprintf(ks->names[j], sizeof(ks->names[j]),
191 "btrfs-%s-%02d", ks->name_stem, j);
192 }
193}
194
195void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
196 int level)
197{
198 struct btrfs_lockdep_keyset *ks;
199
200 BUG_ON(level >= ARRAY_SIZE(ks->keys));
201
202 /* find the matching keyset, id 0 is the default entry */
203 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
204 if (ks->id == objectid)
205 break;
206
207 lockdep_set_class_and_name(&eb->lock,
208 &ks->keys[level], ks->names[level]);
209}
210
Chris Mason4008c042009-02-12 14:09:45 -0500211#endif
212
Chris Masond352ac62008-09-29 15:18:18 -0400213/*
214 * extents on the btree inode are pretty simple, there's one extent
215 * that covers the entire device
216 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500217static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200218 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500219 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400220{
221 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
222 struct extent_map *em;
223 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400224
Chris Mason890871b2009-09-02 16:24:52 -0400225 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500226 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400227 if (em) {
228 em->bdev =
229 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400230 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400231 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400232 }
Chris Mason890871b2009-09-02 16:24:52 -0400233 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400234
David Sterba172ddd62011-04-21 00:48:27 +0200235 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400236 if (!em) {
237 em = ERR_PTR(-ENOMEM);
238 goto out;
239 }
240 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400241 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400242 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400243 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400244 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500245
Chris Mason890871b2009-09-02 16:24:52 -0400246 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400247 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400248 if (ret == -EEXIST) {
249 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400250 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600251 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600252 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400253 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400254 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600255 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400256 }
Chris Mason890871b2009-09-02 16:24:52 -0400257 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400258
Chris Mason5f39d392007-10-15 16:14:19 -0400259out:
260 return em;
261}
262
Liu Bob0496682013-03-14 14:57:45 +0000263u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400264{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000265 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400266}
267
268void btrfs_csum_final(u32 crc, char *result)
269{
David Sterba7e75bf32011-03-18 22:56:43 +0000270 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400271}
272
Chris Masond352ac62008-09-29 15:18:18 -0400273/*
274 * compute the csum for a btree block, and either verify it or write it
275 * into the csum field of the block.
276 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400277static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
278 int verify)
279{
David Sterba6c417612011-04-13 15:41:04 +0200280 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500281 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400282 unsigned long len;
283 unsigned long cur_len;
284 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400285 char *kaddr;
286 unsigned long map_start;
287 unsigned long map_len;
288 int err;
289 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500290 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400291
292 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500293 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400294 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400295 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500296 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400297 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400298 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000299 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400300 crc, cur_len);
301 len -= cur_len;
302 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400303 }
Josef Bacik607d4322008-12-02 07:17:45 -0500304 if (csum_size > sizeof(inline_result)) {
305 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
306 if (!result)
307 return 1;
308 } else {
309 result = (char *)&inline_result;
310 }
311
Chris Mason19c00dd2007-10-15 16:19:22 -0400312 btrfs_csum_final(crc, result);
313
314 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500315 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500316 u32 val;
317 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500318 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500319
Josef Bacik607d4322008-12-02 07:17:45 -0500320 read_extent_buffer(buf, &val, 0, csum_size);
Frank Holtonefe120a2013-12-20 11:37:06 -0500321 printk_ratelimited(KERN_INFO
322 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
323 "level %d\n",
324 root->fs_info->sb->s_id, buf->start,
325 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500326 if (result != (char *)&inline_result)
327 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400328 return 1;
329 }
330 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500331 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400332 }
Josef Bacik607d4322008-12-02 07:17:45 -0500333 if (result != (char *)&inline_result)
334 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400335 return 0;
336}
337
Chris Masond352ac62008-09-29 15:18:18 -0400338/*
339 * we can't consider a given block up to date unless the transid of the
340 * block matches the transid in the parent node's pointer. This is how we
341 * detect blocks that either didn't get written at all or got written
342 * in the wrong place.
343 */
Chris Mason1259ab72008-05-12 13:39:03 -0400344static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400345 struct extent_buffer *eb, u64 parent_transid,
346 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400347{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000348 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400349 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200350 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400351
352 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
353 return 0;
354
Chris Masonb9fab912012-05-06 07:23:47 -0400355 if (atomic)
356 return -EAGAIN;
357
Josef Bacika26e8c92014-03-28 17:07:27 -0400358 if (need_lock) {
359 btrfs_tree_read_lock(eb);
360 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
361 }
362
Josef Bacik2ac55d42010-02-03 19:33:23 +0000363 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100364 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400365 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400366 btrfs_header_generation(eb) == parent_transid) {
367 ret = 0;
368 goto out;
369 }
Wang Shilong29549ae2014-07-04 17:59:06 +0800370 printk_ratelimited(KERN_INFO "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n",
371 eb->fs_info->sb->s_id, eb->start,
372 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400373 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400374
375 /*
376 * Things reading via commit roots that don't have normal protection,
377 * like send, can have a really old block in cache that may point at a
378 * block that has been free'd and re-allocated. So don't clear uptodate
379 * if we find an eb that is under IO (dirty/writeback) because we could
380 * end up reading in the stale data and then writing it back out and
381 * making everybody very sad.
382 */
383 if (!extent_buffer_under_io(eb))
384 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400385out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000386 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
387 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700388 if (need_lock)
389 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400390 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400391}
392
Chris Masond352ac62008-09-29 15:18:18 -0400393/*
David Sterba1104a882013-03-06 15:57:46 +0100394 * Return 0 if the superblock checksum type matches the checksum value of that
395 * algorithm. Pass the raw disk superblock data.
396 */
397static int btrfs_check_super_csum(char *raw_disk_sb)
398{
399 struct btrfs_super_block *disk_sb =
400 (struct btrfs_super_block *)raw_disk_sb;
401 u16 csum_type = btrfs_super_csum_type(disk_sb);
402 int ret = 0;
403
404 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
405 u32 crc = ~(u32)0;
406 const int csum_size = sizeof(crc);
407 char result[csum_size];
408
409 /*
410 * The super_block structure does not span the whole
411 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
412 * is filled with zeros and is included in the checkum.
413 */
414 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
415 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
416 btrfs_csum_final(crc, result);
417
418 if (memcmp(raw_disk_sb, result, csum_size))
419 ret = 1;
Chris Mason667e7d92013-05-07 11:00:13 -0400420
421 if (ret && btrfs_super_generation(disk_sb) < 10) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500422 printk(KERN_WARNING
423 "BTRFS: super block crcs don't match, older mkfs detected\n");
Chris Mason667e7d92013-05-07 11:00:13 -0400424 ret = 0;
425 }
David Sterba1104a882013-03-06 15:57:46 +0100426 }
427
428 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500429 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100430 csum_type);
431 ret = 1;
432 }
433
434 return ret;
435}
436
437/*
Chris Masond352ac62008-09-29 15:18:18 -0400438 * helper to read a given tree block, doing retries as required when
439 * the checksums don't match and we have alternate mirrors to try.
440 */
Chris Masonf1885912008-04-09 16:28:12 -0400441static int btree_read_extent_buffer_pages(struct btrfs_root *root,
442 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400443 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400444{
445 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400446 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400447 int ret;
448 int num_copies = 0;
449 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400450 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400451
Josef Bacika826d6d2011-03-16 13:42:43 -0400452 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400453 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
454 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200455 ret = read_extent_buffer_pages(io_tree, eb, start,
456 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400457 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600458 if (!ret) {
459 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400460 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600461 break;
462 else
463 ret = -EIO;
464 }
Chris Masond3977122009-01-05 21:25:51 -0500465
Josef Bacika826d6d2011-03-16 13:42:43 -0400466 /*
467 * This buffer's crc is fine, but its contents are corrupted, so
468 * there is no reason to read the other copies, they won't be
469 * any less wrong.
470 */
471 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400472 break;
473
Stefan Behrens5d964052012-11-05 14:59:07 +0100474 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400475 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400476 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400477 break;
Chris Mason42352982008-04-28 16:40:52 -0400478
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400479 if (!failed_mirror) {
480 failed = 1;
481 failed_mirror = eb->read_mirror;
482 }
483
Chris Masonf1885912008-04-09 16:28:12 -0400484 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400485 if (mirror_num == failed_mirror)
486 mirror_num++;
487
Chris Mason42352982008-04-28 16:40:52 -0400488 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400489 break;
Chris Masonf1885912008-04-09 16:28:12 -0400490 }
Josef Bacikea466792012-03-26 21:57:36 -0400491
Stefan Behrensc0901582012-07-10 07:30:17 -0600492 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400493 repair_eb_io_failure(root, eb, failed_mirror);
494
495 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400496}
Chris Mason19c00dd2007-10-15 16:19:22 -0400497
Chris Masond352ac62008-09-29 15:18:18 -0400498/*
Chris Masond3977122009-01-05 21:25:51 -0500499 * checksum a dirty tree block before IO. This has extra checks to make sure
500 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400501 */
Chris Masond3977122009-01-05 21:25:51 -0500502
Christoph Hellwigb2950862008-12-02 09:54:17 -0500503static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400504{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000505 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400506 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400507 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400508
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500509 eb = (struct extent_buffer *)page->private;
510 if (page != eb->pages[0])
511 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400512 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530513 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500514 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400515 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400516 return 0;
517}
518
Yan Zheng2b820322008-11-17 21:11:30 -0500519static int check_tree_block_fsid(struct btrfs_root *root,
520 struct extent_buffer *eb)
521{
522 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
523 u8 fsid[BTRFS_UUID_SIZE];
524 int ret = 1;
525
Ross Kirk0a4e5582013-09-24 10:12:38 +0100526 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500527 while (fs_devices) {
528 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
529 ret = 0;
530 break;
531 }
532 fs_devices = fs_devices->seed;
533 }
534 return ret;
535}
536
Josef Bacika826d6d2011-03-16 13:42:43 -0400537#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500538 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
539 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200540 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400541
542static noinline int check_leaf(struct btrfs_root *root,
543 struct extent_buffer *leaf)
544{
545 struct btrfs_key key;
546 struct btrfs_key leaf_key;
547 u32 nritems = btrfs_header_nritems(leaf);
548 int slot;
549
550 if (nritems == 0)
551 return 0;
552
553 /* Check the 0 item */
554 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
555 BTRFS_LEAF_DATA_SIZE(root)) {
556 CORRUPT("invalid item offset size pair", leaf, root, 0);
557 return -EIO;
558 }
559
560 /*
561 * Check to make sure each items keys are in the correct order and their
562 * offsets make sense. We only have to loop through nritems-1 because
563 * we check the current slot against the next slot, which verifies the
564 * next slot's offset+size makes sense and that the current's slot
565 * offset is correct.
566 */
567 for (slot = 0; slot < nritems - 1; slot++) {
568 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
569 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
570
571 /* Make sure the keys are in the right order */
572 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
573 CORRUPT("bad key order", leaf, root, slot);
574 return -EIO;
575 }
576
577 /*
578 * Make sure the offset and ends are right, remember that the
579 * item data starts at the end of the leaf and grows towards the
580 * front.
581 */
582 if (btrfs_item_offset_nr(leaf, slot) !=
583 btrfs_item_end_nr(leaf, slot + 1)) {
584 CORRUPT("slot offset bad", leaf, root, slot);
585 return -EIO;
586 }
587
588 /*
589 * Check to make sure that we don't point outside of the leaf,
590 * just incase all the items are consistent to eachother, but
591 * all point outside of the leaf.
592 */
593 if (btrfs_item_end_nr(leaf, slot) >
594 BTRFS_LEAF_DATA_SIZE(root)) {
595 CORRUPT("slot end outside of leaf", leaf, root, slot);
596 return -EIO;
597 }
598 }
599
600 return 0;
601}
602
Miao Xiefacc8a222013-07-25 19:22:34 +0800603static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
604 u64 phy_offset, struct page *page,
605 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400606{
Chris Masonce9adaa2008-04-09 16:28:12 -0400607 u64 found_start;
608 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400609 struct extent_buffer *eb;
610 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400611 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400612 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400613
Chris Masonce9adaa2008-04-09 16:28:12 -0400614 if (!page->private)
615 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500616
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500617 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500618
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400619 /* the pending IO might have been the only thing that kept this buffer
620 * in memory. Make sure we have a ref for all this other checks
621 */
622 extent_buffer_get(eb);
623
624 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400625 if (!reads_done)
626 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400627
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400628 eb->read_mirror = mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400629 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
630 ret = -EIO;
631 goto err;
632 }
633
Chris Masonce9adaa2008-04-09 16:28:12 -0400634 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400635 if (found_start != eb->start) {
Wang Shilong29549ae2014-07-04 17:59:06 +0800636 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400637 "%llu %llu\n",
Wang Shilong29549ae2014-07-04 17:59:06 +0800638 eb->fs_info->sb->s_id, found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400639 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400640 goto err;
641 }
Yan Zheng2b820322008-11-17 21:11:30 -0500642 if (check_tree_block_fsid(root, eb)) {
Wang Shilong29549ae2014-07-04 17:59:06 +0800643 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad fsid on block %llu\n",
644 eb->fs_info->sb->s_id, eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400645 ret = -EIO;
646 goto err;
647 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400648 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400649 if (found_level >= BTRFS_MAX_LEVEL) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500650 btrfs_info(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400651 (int)btrfs_header_level(eb));
652 ret = -EIO;
653 goto err;
654 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400655
Chris Mason85d4e462011-07-26 16:11:19 -0400656 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
657 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500658
Chris Masonce9adaa2008-04-09 16:28:12 -0400659 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400660 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400661 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400662 goto err;
663 }
664
665 /*
666 * If this is a leaf block and it is corrupt, set the corrupt bit so
667 * that we don't try and read the other copies of this block, just
668 * return -EIO.
669 */
670 if (found_level == 0 && check_leaf(root, eb)) {
671 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
672 ret = -EIO;
673 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400674
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400675 if (!ret)
676 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400677err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400678 if (reads_done &&
679 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200680 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200681
David Woodhouse53b381b2013-01-29 18:40:14 -0500682 if (ret) {
683 /*
684 * our io error hook is going to dec the io pages
685 * again, we have to make sure it has something
686 * to decrement
687 */
688 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400689 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500690 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400691 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400692out:
Chris Masonf1885912008-04-09 16:28:12 -0400693 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400694}
695
Josef Bacikea466792012-03-26 21:57:36 -0400696static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200697{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200698 struct extent_buffer *eb;
699 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
700
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500701 eb = (struct extent_buffer *)page->private;
Josef Bacikea466792012-03-26 21:57:36 -0400702 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400703 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500704 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400705 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200706 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200707 return -EIO; /* we fixed nothing */
708}
709
Chris Masonce9adaa2008-04-09 16:28:12 -0400710static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400711{
David Sterba97eb6b62014-07-30 00:55:42 +0200712 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400713 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800714 struct btrfs_workqueue *wq;
715 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400716
Chris Masonce9adaa2008-04-09 16:28:12 -0400717 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400718 end_io_wq->error = err;
Chris Masond20f7042008-12-08 16:58:54 -0500719
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200720 if (bio->bi_rw & REQ_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800721 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
722 wq = fs_info->endio_meta_write_workers;
723 func = btrfs_endio_meta_write_helper;
724 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
725 wq = fs_info->endio_freespace_worker;
726 func = btrfs_freespace_write_helper;
727 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
728 wq = fs_info->endio_raid56_workers;
729 func = btrfs_endio_raid56_helper;
730 } else {
731 wq = fs_info->endio_write_workers;
732 func = btrfs_endio_write_helper;
733 }
Chris Masond20f7042008-12-08 16:58:54 -0500734 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800735 if (unlikely(end_io_wq->metadata ==
736 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
737 wq = fs_info->endio_repair_workers;
738 func = btrfs_endio_repair_helper;
739 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800740 wq = fs_info->endio_raid56_workers;
741 func = btrfs_endio_raid56_helper;
742 } else if (end_io_wq->metadata) {
743 wq = fs_info->endio_meta_workers;
744 func = btrfs_endio_meta_helper;
745 } else {
746 wq = fs_info->endio_workers;
747 func = btrfs_endio_helper;
748 }
Chris Masond20f7042008-12-08 16:58:54 -0500749 }
Liu Bo9e0af232014-08-15 23:36:53 +0800750
751 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
752 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400753}
754
Chris Mason22c59942008-04-09 16:28:12 -0400755int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200756 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400757{
David Sterba97eb6b62014-07-30 00:55:42 +0200758 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800759
David Sterba97eb6b62014-07-30 00:55:42 +0200760 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400761 if (!end_io_wq)
762 return -ENOMEM;
763
764 end_io_wq->private = bio->bi_private;
765 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400766 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400767 end_io_wq->error = 0;
768 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400769 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400770
771 bio->bi_private = end_io_wq;
772 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400773 return 0;
774}
775
Chris Masonb64a2852008-08-20 13:39:41 -0400776unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400777{
778 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800779 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400780 info->fs_devices->open_devices);
781 return 256 * limit;
782}
783
Chris Mason4a69a412008-11-06 22:03:00 -0500784static void run_one_async_start(struct btrfs_work *work)
785{
Chris Mason4a69a412008-11-06 22:03:00 -0500786 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100787 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500788
789 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100790 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
791 async->mirror_num, async->bio_flags,
792 async->bio_offset);
793 if (ret)
794 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500795}
796
797static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400798{
799 struct btrfs_fs_info *fs_info;
800 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400801 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400802
803 async = container_of(work, struct async_submit_bio, work);
804 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400805
Chris Masonb64a2852008-08-20 13:39:41 -0400806 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400807 limit = limit * 2 / 3;
808
Josef Bacik66657b32012-08-01 15:36:24 -0400809 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400810 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400811 wake_up(&fs_info->async_submit_wait);
812
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100813 /* If an error occured we just want to clean up the bio and move on */
814 if (async->error) {
815 bio_endio(async->bio, async->error);
816 return;
817 }
818
Chris Mason4a69a412008-11-06 22:03:00 -0500819 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400820 async->mirror_num, async->bio_flags,
821 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500822}
823
824static void run_one_async_free(struct btrfs_work *work)
825{
826 struct async_submit_bio *async;
827
828 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400829 kfree(async);
830}
831
Chris Mason44b8bd72008-04-16 11:14:51 -0400832int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
833 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400834 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400835 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500836 extent_submit_bio_hook_t *submit_bio_start,
837 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400838{
839 struct async_submit_bio *async;
840
841 async = kmalloc(sizeof(*async), GFP_NOFS);
842 if (!async)
843 return -ENOMEM;
844
845 async->inode = inode;
846 async->rw = rw;
847 async->bio = bio;
848 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500849 async->submit_bio_start = submit_bio_start;
850 async->submit_bio_done = submit_bio_done;
851
Liu Bo9e0af232014-08-15 23:36:53 +0800852 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800853 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500854
Chris Masonc8b97812008-10-29 14:49:59 -0400855 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400856 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400857
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100858 async->error = 0;
859
Chris Masoncb03c742008-05-15 16:15:45 -0400860 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400861
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200862 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800863 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400864
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800865 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400866
Chris Masond3977122009-01-05 21:25:51 -0500867 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500868 atomic_read(&fs_info->nr_async_submits)) {
869 wait_event(fs_info->async_submit_wait,
870 (atomic_read(&fs_info->nr_async_submits) == 0));
871 }
872
Chris Mason44b8bd72008-04-16 11:14:51 -0400873 return 0;
874}
875
Chris Masonce3ed712008-09-23 13:14:12 -0400876static int btree_csum_one_bio(struct bio *bio)
877{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800878 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400879 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800880 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400881
Kent Overstreet2c30c712013-11-07 12:20:26 -0800882 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400883 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100884 ret = csum_dirty_buffer(root, bvec->bv_page);
885 if (ret)
886 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400887 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800888
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100889 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400890}
891
Chris Mason4a69a412008-11-06 22:03:00 -0500892static int __btree_submit_bio_start(struct inode *inode, int rw,
893 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400894 unsigned long bio_flags,
895 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400896{
Chris Mason8b712842008-06-11 16:50:36 -0400897 /*
898 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400899 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400900 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100901 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500902}
Chris Mason22c59942008-04-09 16:28:12 -0400903
Chris Mason4a69a412008-11-06 22:03:00 -0500904static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400905 int mirror_num, unsigned long bio_flags,
906 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500907{
Stefan Behrens61891922012-11-05 18:51:52 +0100908 int ret;
909
Chris Mason8b712842008-06-11 16:50:36 -0400910 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500911 * when we're called for a write, we're already in the async
912 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400913 */
Stefan Behrens61891922012-11-05 18:51:52 +0100914 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
915 if (ret)
916 bio_endio(bio, ret);
917 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400918}
919
Josef Bacikde0022b2012-09-25 14:25:58 -0400920static int check_async_write(struct inode *inode, unsigned long bio_flags)
921{
922 if (bio_flags & EXTENT_BIO_TREE_LOG)
923 return 0;
924#ifdef CONFIG_X86
925 if (cpu_has_xmm4_2)
926 return 0;
927#endif
928 return 1;
929}
930
Chris Mason44b8bd72008-04-16 11:14:51 -0400931static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400932 int mirror_num, unsigned long bio_flags,
933 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400934{
Josef Bacikde0022b2012-09-25 14:25:58 -0400935 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500936 int ret;
937
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200938 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500939 /*
940 * called for a read, do the setup so that checksum validation
941 * can happen in the async kernel threads
942 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400943 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +0200944 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -0400945 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100946 goto out_w_error;
947 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
948 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400949 } else if (!async) {
950 ret = btree_csum_one_bio(bio);
951 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100952 goto out_w_error;
953 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
954 mirror_num, 0);
955 } else {
956 /*
957 * kthread helpers are used to submit writes so that
958 * checksumming can happen in parallel across all CPUs
959 */
960 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
961 inode, rw, bio, mirror_num, 0,
962 bio_offset,
963 __btree_submit_bio_start,
964 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400965 }
Chris Masond313d7a2009-04-20 15:50:09 -0400966
Stefan Behrens61891922012-11-05 18:51:52 +0100967 if (ret) {
968out_w_error:
969 bio_endio(bio, ret);
970 }
971 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400972}
973
Jan Beulich3dd14622010-12-07 14:54:09 +0000974#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500975static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800976 struct page *newpage, struct page *page,
977 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500978{
979 /*
980 * we can't safely write a btree page from here,
981 * we haven't done the locking hook
982 */
983 if (PageDirty(page))
984 return -EAGAIN;
985 /*
986 * Buffers may be managed in a filesystem specific way.
987 * We must have no buffers or drop them.
988 */
989 if (page_has_private(page) &&
990 !try_to_release_page(page, GFP_KERNEL))
991 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800992 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500993}
Jan Beulich3dd14622010-12-07 14:54:09 +0000994#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500995
Chris Mason0da54682007-11-07 21:08:01 -0500996
997static int btree_writepages(struct address_space *mapping,
998 struct writeback_control *wbc)
999{
Miao Xiee2d84522013-01-29 10:09:20 +00001000 struct btrfs_fs_info *fs_info;
1001 int ret;
1002
Chris Masond8d5f3e2007-12-11 12:42:00 -05001003 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001004
1005 if (wbc->for_kupdate)
1006 return 0;
1007
Miao Xiee2d84522013-01-29 10:09:20 +00001008 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001009 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001010 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1011 BTRFS_DIRTY_METADATA_THRESH);
1012 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001013 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001014 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001015 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001016}
1017
Christoph Hellwigb2950862008-12-02 09:54:17 -05001018static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001019{
Chris Masond1310b22008-01-24 16:13:08 -05001020 struct extent_io_tree *tree;
1021 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001022 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001023}
1024
Chris Mason70dec802008-01-29 09:59:12 -05001025static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001026{
Chris Mason98509cfc2008-09-11 15:51:43 -04001027 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001028 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001029
David Sterbaf7a52a42013-04-26 14:56:29 +00001030 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001031}
Chris Mason123abc82007-03-14 14:14:43 -04001032
Lukas Czernerd47992f2013-05-21 23:17:23 -04001033static void btree_invalidatepage(struct page *page, unsigned int offset,
1034 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001035{
Chris Masond1310b22008-01-24 16:13:08 -05001036 struct extent_io_tree *tree;
1037 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001038 extent_invalidatepage(tree, page, offset);
1039 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001040 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001041 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1042 "page private not zero on page %llu",
1043 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001044 ClearPagePrivate(page);
1045 set_page_private(page, 0);
1046 page_cache_release(page);
1047 }
Chris Masond98237b2007-03-28 13:57:48 -04001048}
1049
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001050static int btree_set_page_dirty(struct page *page)
1051{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001052#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001053 struct extent_buffer *eb;
1054
1055 BUG_ON(!PagePrivate(page));
1056 eb = (struct extent_buffer *)page->private;
1057 BUG_ON(!eb);
1058 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1059 BUG_ON(!atomic_read(&eb->refs));
1060 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001061#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001062 return __set_page_dirty_nobuffers(page);
1063}
1064
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001065static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001066 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001067 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001068 .releasepage = btree_releasepage,
1069 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001070#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001071 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001072#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001073 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001074};
1075
Chris Masonca7a79a2008-05-12 12:59:19 -04001076int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1077 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001078{
Chris Mason5f39d392007-10-15 16:14:19 -04001079 struct extent_buffer *buf = NULL;
1080 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001081 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001082
Chris Masondb945352007-10-15 16:15:53 -04001083 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001084 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001085 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001086 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001087 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001088 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001089 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001090}
1091
Arne Jansenab0fff02011-05-23 14:25:41 +02001092int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1093 int mirror_num, struct extent_buffer **eb)
1094{
1095 struct extent_buffer *buf = NULL;
1096 struct inode *btree_inode = root->fs_info->btree_inode;
1097 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1098 int ret;
1099
1100 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1101 if (!buf)
1102 return 0;
1103
1104 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1105
1106 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1107 btree_get_extent, mirror_num);
1108 if (ret) {
1109 free_extent_buffer(buf);
1110 return ret;
1111 }
1112
1113 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1114 free_extent_buffer(buf);
1115 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001116 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001117 *eb = buf;
1118 } else {
1119 free_extent_buffer(buf);
1120 }
1121 return 0;
1122}
1123
Chris Mason0999df52008-04-01 13:48:14 -04001124struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1125 u64 bytenr, u32 blocksize)
1126{
Josef Bacikf28491e2013-12-16 13:24:27 -05001127 return find_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001128}
1129
1130struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1131 u64 bytenr, u32 blocksize)
1132{
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001133#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1134 if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
1135 return alloc_test_extent_buffer(root->fs_info, bytenr,
1136 blocksize);
1137#endif
Josef Bacikf28491e2013-12-16 13:24:27 -05001138 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001139}
1140
1141
Chris Masone02119d2008-09-05 16:13:11 -04001142int btrfs_write_tree_block(struct extent_buffer *buf)
1143{
Chris Mason727011e2010-08-06 13:21:20 -04001144 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001145 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001146}
1147
1148int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1149{
Chris Mason727011e2010-08-06 13:21:20 -04001150 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001151 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001152}
1153
Chris Masondb945352007-10-15 16:15:53 -04001154struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001155 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001156{
Chris Mason5f39d392007-10-15 16:14:19 -04001157 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001158 int ret;
1159
Chris Masondb945352007-10-15 16:15:53 -04001160 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001161 if (!buf)
1162 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001163
Chris Masonca7a79a2008-05-12 12:59:19 -04001164 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001165 if (ret) {
1166 free_extent_buffer(buf);
1167 return NULL;
1168 }
Chris Mason5f39d392007-10-15 16:14:19 -04001169 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001170
Chris Masoneb60cea2007-02-02 09:18:22 -05001171}
1172
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001173void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1174 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001175{
Miao Xiee2d84522013-01-29 10:09:20 +00001176 struct btrfs_fs_info *fs_info = root->fs_info;
1177
Chris Mason55c69072008-01-09 15:55:33 -05001178 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001179 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001180 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001181
Chris Masonb9473432009-03-13 11:00:37 -04001182 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001183 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1184 -buf->len,
1185 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001186 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1187 btrfs_set_lock_blocking(buf);
1188 clear_extent_buffer_dirty(buf);
1189 }
Chris Mason925baed2008-06-25 16:01:30 -04001190 }
Chris Mason5f39d392007-10-15 16:14:19 -04001191}
1192
Miao Xie8257b2d2014-03-06 13:38:19 +08001193static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1194{
1195 struct btrfs_subvolume_writers *writers;
1196 int ret;
1197
1198 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1199 if (!writers)
1200 return ERR_PTR(-ENOMEM);
1201
1202 ret = percpu_counter_init(&writers->counter, 0);
1203 if (ret < 0) {
1204 kfree(writers);
1205 return ERR_PTR(ret);
1206 }
1207
1208 init_waitqueue_head(&writers->wait);
1209 return writers;
1210}
1211
1212static void
1213btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1214{
1215 percpu_counter_destroy(&writers->counter);
1216 kfree(writers);
1217}
1218
David Sterba707e8a02014-06-04 19:22:26 +02001219static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1220 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
Jeff Mahoney143bede2012-03-01 14:56:26 +01001221 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001222{
Chris Masoncfaa7292007-02-21 17:04:57 -05001223 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001224 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001225 root->sectorsize = sectorsize;
1226 root->nodesize = nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001227 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001228 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001229 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001230
Chris Mason0f7d52f2007-04-09 10:42:37 -04001231 root->objectid = objectid;
1232 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001233 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001234 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001235 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001236 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001237 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001238 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001239 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001240 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001241
1242 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001243 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001244 INIT_LIST_HEAD(&root->delalloc_inodes);
1245 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001246 INIT_LIST_HEAD(&root->ordered_extents);
1247 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001248 INIT_LIST_HEAD(&root->logged_list[0]);
1249 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001250 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001251 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001252 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001253 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001254 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001255 spin_lock_init(&root->log_extents_lock[0]);
1256 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001257 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001258 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001259 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001260 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f1832009-01-21 12:54:03 -05001261 init_waitqueue_head(&root->log_writer_wait);
1262 init_waitqueue_head(&root->log_commit_wait[0]);
1263 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001264 INIT_LIST_HEAD(&root->log_ctxs[0]);
1265 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f1832009-01-21 12:54:03 -05001266 atomic_set(&root->log_commit[0], 0);
1267 atomic_set(&root->log_commit[1], 0);
1268 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001269 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001270 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001271 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001272 atomic_set(&root->will_be_snapshoted, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001273 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001274 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001275 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001276 if (fs_info)
1277 extent_io_tree_init(&root->dirty_log_pages,
1278 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001279
Chris Mason3768f362007-03-13 16:47:54 -04001280 memset(&root->root_key, 0, sizeof(root->root_key));
1281 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001282 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001283 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001284 if (fs_info)
1285 root->defrag_trans_start = fs_info->generation;
1286 else
1287 root->defrag_trans_start = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001288 init_completion(&root->kobj_unregister);
Chris Mason4d775672007-04-20 20:23:12 -04001289 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001290 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001291
Anand Jain5f3ab902012-12-07 09:28:54 +00001292 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001293}
1294
Al Virof84a8bd2011-11-17 15:57:57 -05001295static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001296{
1297 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1298 if (root)
1299 root->fs_info = fs_info;
1300 return root;
1301}
1302
Josef Bacik06ea65a2013-09-19 16:07:01 -04001303#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1304/* Should only be used by the testing infrastructure */
1305struct btrfs_root *btrfs_alloc_dummy_root(void)
1306{
1307 struct btrfs_root *root;
1308
1309 root = btrfs_alloc_root(NULL);
1310 if (!root)
1311 return ERR_PTR(-ENOMEM);
David Sterba707e8a02014-06-04 19:22:26 +02001312 __setup_root(4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001313 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001314 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001315
1316 return root;
1317}
1318#endif
1319
Arne Jansen20897f52011-09-13 12:44:20 +02001320struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1321 struct btrfs_fs_info *fs_info,
1322 u64 objectid)
1323{
1324 struct extent_buffer *leaf;
1325 struct btrfs_root *tree_root = fs_info->tree_root;
1326 struct btrfs_root *root;
1327 struct btrfs_key key;
1328 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001329 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001330
1331 root = btrfs_alloc_root(fs_info);
1332 if (!root)
1333 return ERR_PTR(-ENOMEM);
1334
David Sterba707e8a02014-06-04 19:22:26 +02001335 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1336 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001337 root->root_key.objectid = objectid;
1338 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1339 root->root_key.offset = 0;
1340
David Sterba707e8a02014-06-04 19:22:26 +02001341 leaf = btrfs_alloc_free_block(trans, root, root->nodesize,
Arne Jansen20897f52011-09-13 12:44:20 +02001342 0, objectid, NULL, 0, 0, 0);
1343 if (IS_ERR(leaf)) {
1344 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001345 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001346 goto fail;
1347 }
1348
Arne Jansen20897f52011-09-13 12:44:20 +02001349 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1350 btrfs_set_header_bytenr(leaf, leaf->start);
1351 btrfs_set_header_generation(leaf, trans->transid);
1352 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1353 btrfs_set_header_owner(leaf, objectid);
1354 root->node = leaf;
1355
Ross Kirk0a4e5582013-09-24 10:12:38 +01001356 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001357 BTRFS_FSID_SIZE);
1358 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001359 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001360 BTRFS_UUID_SIZE);
1361 btrfs_mark_buffer_dirty(leaf);
1362
1363 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001364 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001365
1366 root->root_item.flags = 0;
1367 root->root_item.byte_limit = 0;
1368 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1369 btrfs_set_root_generation(&root->root_item, trans->transid);
1370 btrfs_set_root_level(&root->root_item, 0);
1371 btrfs_set_root_refs(&root->root_item, 1);
1372 btrfs_set_root_used(&root->root_item, leaf->len);
1373 btrfs_set_root_last_snapshot(&root->root_item, 0);
1374 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001375 uuid_le_gen(&uuid);
1376 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001377 root->root_item.drop_level = 0;
1378
1379 key.objectid = objectid;
1380 key.type = BTRFS_ROOT_ITEM_KEY;
1381 key.offset = 0;
1382 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1383 if (ret)
1384 goto fail;
1385
1386 btrfs_tree_unlock(leaf);
1387
Arne Jansen20897f52011-09-13 12:44:20 +02001388 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001389
1390fail:
1391 if (leaf) {
1392 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001393 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001394 free_extent_buffer(leaf);
1395 }
1396 kfree(root);
1397
1398 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001399}
1400
Yan Zheng7237f1832009-01-21 12:54:03 -05001401static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1402 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001403{
1404 struct btrfs_root *root;
1405 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f1832009-01-21 12:54:03 -05001406 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001407
Al Viro6f07e422011-11-17 00:46:16 -05001408 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001409 if (!root)
Yan Zheng7237f1832009-01-21 12:54:03 -05001410 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001411
David Sterba707e8a02014-06-04 19:22:26 +02001412 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1413 tree_root->stripesize, root, fs_info,
1414 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001415
1416 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1417 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1418 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001419
Yan Zheng7237f1832009-01-21 12:54:03 -05001420 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001421 * DON'T set REF_COWS for log trees
1422 *
Yan Zheng7237f1832009-01-21 12:54:03 -05001423 * log trees do not get reference counted because they go away
1424 * before a real commit is actually done. They do store pointers
1425 * to file data extents, and those reference counts still get
1426 * updated (along with back refs to the log tree).
1427 */
Chris Masone02119d2008-09-05 16:13:11 -04001428
David Sterba707e8a02014-06-04 19:22:26 +02001429 leaf = btrfs_alloc_free_block(trans, root, root->nodesize, 0,
Arne Jansen66d7e7f2011-09-12 15:26:38 +02001430 BTRFS_TREE_LOG_OBJECTID, NULL,
Jan Schmidt5581a512012-05-16 17:04:52 +02001431 0, 0, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001432 if (IS_ERR(leaf)) {
1433 kfree(root);
1434 return ERR_CAST(leaf);
1435 }
Chris Masone02119d2008-09-05 16:13:11 -04001436
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001437 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1438 btrfs_set_header_bytenr(leaf, leaf->start);
1439 btrfs_set_header_generation(leaf, trans->transid);
1440 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1441 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f1832009-01-21 12:54:03 -05001442 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001443
1444 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001445 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001446 btrfs_mark_buffer_dirty(root->node);
1447 btrfs_tree_unlock(root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001448 return root;
1449}
1450
1451int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1452 struct btrfs_fs_info *fs_info)
1453{
1454 struct btrfs_root *log_root;
1455
1456 log_root = alloc_log_tree(trans, fs_info);
1457 if (IS_ERR(log_root))
1458 return PTR_ERR(log_root);
1459 WARN_ON(fs_info->log_root_tree);
1460 fs_info->log_root_tree = log_root;
1461 return 0;
1462}
1463
1464int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1465 struct btrfs_root *root)
1466{
1467 struct btrfs_root *log_root;
1468 struct btrfs_inode_item *inode_item;
1469
1470 log_root = alloc_log_tree(trans, root->fs_info);
1471 if (IS_ERR(log_root))
1472 return PTR_ERR(log_root);
1473
1474 log_root->last_trans = trans->transid;
1475 log_root->root_key.offset = root->root_key.objectid;
1476
1477 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001478 btrfs_set_stack_inode_generation(inode_item, 1);
1479 btrfs_set_stack_inode_size(inode_item, 3);
1480 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001481 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001482 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f1832009-01-21 12:54:03 -05001483
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001484 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001485
1486 WARN_ON(root->log_root);
1487 root->log_root = log_root;
1488 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001489 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001490 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001491 return 0;
1492}
1493
Stefan Behrens35a36212013-08-14 18:12:25 +02001494static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1495 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001496{
1497 struct btrfs_root *root;
1498 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001499 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001500 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001501 u32 blocksize;
Miao Xiecb517ea2013-05-15 07:48:19 +00001502 int ret;
1503
1504 path = btrfs_alloc_path();
1505 if (!path)
1506 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001507
Al Viro6f07e422011-11-17 00:46:16 -05001508 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001509 if (!root) {
1510 ret = -ENOMEM;
1511 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001512 }
1513
David Sterba707e8a02014-06-04 19:22:26 +02001514 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1515 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001516
Miao Xiecb517ea2013-05-15 07:48:19 +00001517 ret = btrfs_find_root(tree_root, key, path,
1518 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001519 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001520 if (ret > 0)
1521 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001522 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001523 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001524
Yan Zheng84234f32008-10-29 14:49:05 -04001525 generation = btrfs_root_generation(&root->root_item);
David Sterba707e8a02014-06-04 19:22:26 +02001526 blocksize = root->nodesize;
Chris Masondb945352007-10-15 16:15:53 -04001527 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001528 blocksize, generation);
Miao Xiecb517ea2013-05-15 07:48:19 +00001529 if (!root->node) {
1530 ret = -ENOMEM;
1531 goto find_fail;
1532 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1533 ret = -EIO;
1534 goto read_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001535 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001536 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001537out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001538 btrfs_free_path(path);
1539 return root;
1540
1541read_fail:
1542 free_extent_buffer(root->node);
1543find_fail:
1544 kfree(root);
1545alloc_fail:
1546 root = ERR_PTR(ret);
1547 goto out;
1548}
1549
1550struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1551 struct btrfs_key *location)
1552{
1553 struct btrfs_root *root;
1554
1555 root = btrfs_read_tree_root(tree_root, location);
1556 if (IS_ERR(root))
1557 return root;
1558
1559 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001560 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001561 btrfs_check_and_init_root_item(&root->root_item);
1562 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001563
Chris Mason5eda7b52007-06-22 14:16:25 -04001564 return root;
1565}
1566
Miao Xiecb517ea2013-05-15 07:48:19 +00001567int btrfs_init_fs_root(struct btrfs_root *root)
1568{
1569 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001570 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001571
1572 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1573 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1574 GFP_NOFS);
1575 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1576 ret = -ENOMEM;
1577 goto fail;
1578 }
1579
Miao Xie8257b2d2014-03-06 13:38:19 +08001580 writers = btrfs_alloc_subvolume_writers();
1581 if (IS_ERR(writers)) {
1582 ret = PTR_ERR(writers);
1583 goto fail;
1584 }
1585 root->subv_writers = writers;
1586
Miao Xiecb517ea2013-05-15 07:48:19 +00001587 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001588 spin_lock_init(&root->ino_cache_lock);
1589 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001590
1591 ret = get_anon_bdev(&root->anon_dev);
1592 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001593 goto free_writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001594 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001595
1596free_writers:
1597 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001598fail:
1599 kfree(root->free_ino_ctl);
1600 kfree(root->free_ino_pinned);
1601 return ret;
1602}
1603
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001604static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1605 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001606{
1607 struct btrfs_root *root;
1608
1609 spin_lock(&fs_info->fs_roots_radix_lock);
1610 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1611 (unsigned long)root_id);
1612 spin_unlock(&fs_info->fs_roots_radix_lock);
1613 return root;
1614}
1615
1616int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1617 struct btrfs_root *root)
1618{
1619 int ret;
1620
1621 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1622 if (ret)
1623 return ret;
1624
1625 spin_lock(&fs_info->fs_roots_radix_lock);
1626 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1627 (unsigned long)root->root_key.objectid,
1628 root);
1629 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001630 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001631 spin_unlock(&fs_info->fs_roots_radix_lock);
1632 radix_tree_preload_end();
1633
1634 return ret;
1635}
1636
Miao Xiec00869f2013-09-25 21:47:44 +08001637struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1638 struct btrfs_key *location,
1639 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001640{
1641 struct btrfs_root *root;
1642 int ret;
1643
Chris Masonedbd8d42007-12-21 16:27:24 -05001644 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1645 return fs_info->tree_root;
1646 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1647 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001648 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1649 return fs_info->chunk_root;
1650 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1651 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001652 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1653 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001654 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1655 return fs_info->quota_root ? fs_info->quota_root :
1656 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001657 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1658 return fs_info->uuid_root ? fs_info->uuid_root :
1659 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001660again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001661 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001662 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001663 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001664 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001665 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001666 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001667
Miao Xiecb517ea2013-05-15 07:48:19 +00001668 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001669 if (IS_ERR(root))
1670 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001671
Miao Xiec00869f2013-09-25 21:47:44 +08001672 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001673 ret = -ENOENT;
1674 goto fail;
1675 }
1676
Miao Xiecb517ea2013-05-15 07:48:19 +00001677 ret = btrfs_init_fs_root(root);
1678 if (ret)
1679 goto fail;
1680
Kelley Nielsen3f870c22013-11-04 19:37:39 -08001681 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1682 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001683 if (ret < 0)
1684 goto fail;
1685 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001686 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001687
Miao Xiecb517ea2013-05-15 07:48:19 +00001688 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001689 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001690 if (ret == -EEXIST) {
1691 free_fs_root(root);
1692 goto again;
1693 }
1694 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001695 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001696 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001697fail:
1698 free_fs_root(root);
1699 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001700}
1701
Chris Mason04160082008-03-26 10:28:07 -04001702static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1703{
1704 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1705 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001706 struct btrfs_device *device;
1707 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001708
Xiao Guangrong1f781602011-04-20 10:09:16 +00001709 rcu_read_lock();
1710 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001711 if (!device->bdev)
1712 continue;
Chris Mason04160082008-03-26 10:28:07 -04001713 bdi = blk_get_backing_dev_info(device->bdev);
1714 if (bdi && bdi_congested(bdi, bdi_bits)) {
1715 ret = 1;
1716 break;
1717 }
1718 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001719 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001720 return ret;
1721}
1722
Chris Mason04160082008-03-26 10:28:07 -04001723static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1724{
Jens Axboead081f12009-06-12 14:43:40 +02001725 int err;
1726
1727 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001728 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001729 if (err)
1730 return err;
1731
Chris Mason4575c9c2008-04-18 16:13:31 -04001732 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001733 bdi->congested_fn = btrfs_congested_fn;
1734 bdi->congested_data = info;
1735 return 0;
1736}
1737
Chris Mason8b712842008-06-11 16:50:36 -04001738/*
1739 * called by the kthread helper functions to finally call the bio end_io
1740 * functions. This is where read checksum verification actually happens
1741 */
1742static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001743{
Chris Masonce9adaa2008-04-09 16:28:12 -04001744 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001745 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001746 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001747
David Sterba97eb6b62014-07-30 00:55:42 +02001748 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001749 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001750
Chris Mason8b712842008-06-11 16:50:36 -04001751 error = end_io_wq->error;
1752 bio->bi_private = end_io_wq->private;
1753 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001754 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Kent Overstreetbc1e79a2013-12-03 13:24:08 -08001755 bio_endio_nodec(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001756}
1757
Chris Masona74a4b92008-06-25 16:01:31 -04001758static int cleaner_kthread(void *arg)
1759{
1760 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001761 int again;
Chris Masona74a4b92008-06-25 16:01:31 -04001762
1763 do {
Miao Xied0278242013-05-14 10:20:40 +00001764 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001765
Miao Xied0278242013-05-14 10:20:40 +00001766 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001767 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001768 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001769
Miao Xied0278242013-05-14 10:20:40 +00001770 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1771 goto sleep;
1772
Miao Xiedc7f3702013-05-14 10:20:42 +00001773 /*
1774 * Avoid the problem that we change the status of the fs
1775 * during the above check and trylock.
1776 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001777 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001778 mutex_unlock(&root->fs_info->cleaner_mutex);
1779 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001780 }
1781
Miao Xied0278242013-05-14 10:20:40 +00001782 btrfs_run_delayed_iputs(root);
Josef Bacik47ab2a62014-09-18 11:20:02 -04001783 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001784 again = btrfs_clean_one_deleted_snapshot(root);
1785 mutex_unlock(&root->fs_info->cleaner_mutex);
1786
1787 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001788 * The defragger has dealt with the R/O remount and umount,
1789 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001790 */
1791 btrfs_run_defrag_inodes(root->fs_info);
1792sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001793 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001794 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001795 if (!kthread_should_stop())
1796 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001797 __set_current_state(TASK_RUNNING);
1798 }
1799 } while (!kthread_should_stop());
1800 return 0;
1801}
1802
1803static int transaction_kthread(void *arg)
1804{
1805 struct btrfs_root *root = arg;
1806 struct btrfs_trans_handle *trans;
1807 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001808 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001809 unsigned long now;
1810 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001811 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001812
1813 do {
Jan Kara914b2002012-03-12 16:05:50 +01001814 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001815 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001816 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1817
Josef Bacika4abeea2011-04-11 17:25:13 -04001818 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001819 cur = root->fs_info->running_transaction;
1820 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001821 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001822 goto sleep;
1823 }
Yan Zheng31153d82008-07-28 15:32:19 -04001824
Chris Masona74a4b92008-06-25 16:01:31 -04001825 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001826 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001827 (now < cur->start_time ||
1828 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001829 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001830 delay = HZ * 5;
1831 goto sleep;
1832 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001833 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001834 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001835
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001836 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001837 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001838 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001839 if (PTR_ERR(trans) != -ENOENT)
1840 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001841 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001842 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001843 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001844 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001845 } else {
1846 btrfs_end_transaction(trans, root);
1847 }
Chris Masona74a4b92008-06-25 16:01:31 -04001848sleep:
1849 wake_up_process(root->fs_info->cleaner_kthread);
1850 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1851
Josef Bacik4e121c02013-09-27 16:32:39 -04001852 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1853 &root->fs_info->fs_state)))
1854 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001855 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001856 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001857 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001858 (!btrfs_transaction_blocked(root->fs_info) ||
1859 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001860 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001861 __set_current_state(TASK_RUNNING);
1862 }
1863 } while (!kthread_should_stop());
1864 return 0;
1865}
1866
Chris Masonaf31f5e2011-11-03 15:17:42 -04001867/*
1868 * this will find the highest generation in the array of
1869 * root backups. The index of the highest array is returned,
1870 * or -1 if we can't find anything.
1871 *
1872 * We check to make sure the array is valid by comparing the
1873 * generation of the latest root in the array with the generation
1874 * in the super block. If they don't match we pitch it.
1875 */
1876static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1877{
1878 u64 cur;
1879 int newest_index = -1;
1880 struct btrfs_root_backup *root_backup;
1881 int i;
1882
1883 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1884 root_backup = info->super_copy->super_roots + i;
1885 cur = btrfs_backup_tree_root_gen(root_backup);
1886 if (cur == newest_gen)
1887 newest_index = i;
1888 }
1889
1890 /* check to see if we actually wrapped around */
1891 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1892 root_backup = info->super_copy->super_roots;
1893 cur = btrfs_backup_tree_root_gen(root_backup);
1894 if (cur == newest_gen)
1895 newest_index = 0;
1896 }
1897 return newest_index;
1898}
1899
1900
1901/*
1902 * find the oldest backup so we know where to store new entries
1903 * in the backup array. This will set the backup_root_index
1904 * field in the fs_info struct
1905 */
1906static void find_oldest_super_backup(struct btrfs_fs_info *info,
1907 u64 newest_gen)
1908{
1909 int newest_index = -1;
1910
1911 newest_index = find_newest_super_backup(info, newest_gen);
1912 /* if there was garbage in there, just move along */
1913 if (newest_index == -1) {
1914 info->backup_root_index = 0;
1915 } else {
1916 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1917 }
1918}
1919
1920/*
1921 * copy all the root pointers into the super backup array.
1922 * this will bump the backup pointer by one when it is
1923 * done
1924 */
1925static void backup_super_roots(struct btrfs_fs_info *info)
1926{
1927 int next_backup;
1928 struct btrfs_root_backup *root_backup;
1929 int last_backup;
1930
1931 next_backup = info->backup_root_index;
1932 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1933 BTRFS_NUM_BACKUP_ROOTS;
1934
1935 /*
1936 * just overwrite the last backup if we're at the same generation
1937 * this happens only at umount
1938 */
1939 root_backup = info->super_for_commit->super_roots + last_backup;
1940 if (btrfs_backup_tree_root_gen(root_backup) ==
1941 btrfs_header_generation(info->tree_root->node))
1942 next_backup = last_backup;
1943
1944 root_backup = info->super_for_commit->super_roots + next_backup;
1945
1946 /*
1947 * make sure all of our padding and empty slots get zero filled
1948 * regardless of which ones we use today
1949 */
1950 memset(root_backup, 0, sizeof(*root_backup));
1951
1952 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1953
1954 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1955 btrfs_set_backup_tree_root_gen(root_backup,
1956 btrfs_header_generation(info->tree_root->node));
1957
1958 btrfs_set_backup_tree_root_level(root_backup,
1959 btrfs_header_level(info->tree_root->node));
1960
1961 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1962 btrfs_set_backup_chunk_root_gen(root_backup,
1963 btrfs_header_generation(info->chunk_root->node));
1964 btrfs_set_backup_chunk_root_level(root_backup,
1965 btrfs_header_level(info->chunk_root->node));
1966
1967 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1968 btrfs_set_backup_extent_root_gen(root_backup,
1969 btrfs_header_generation(info->extent_root->node));
1970 btrfs_set_backup_extent_root_level(root_backup,
1971 btrfs_header_level(info->extent_root->node));
1972
Chris Mason7c7e82a2011-11-06 18:50:56 -05001973 /*
1974 * we might commit during log recovery, which happens before we set
1975 * the fs_root. Make sure it is valid before we fill it in.
1976 */
1977 if (info->fs_root && info->fs_root->node) {
1978 btrfs_set_backup_fs_root(root_backup,
1979 info->fs_root->node->start);
1980 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001981 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001982 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001983 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001984 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001985
1986 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1987 btrfs_set_backup_dev_root_gen(root_backup,
1988 btrfs_header_generation(info->dev_root->node));
1989 btrfs_set_backup_dev_root_level(root_backup,
1990 btrfs_header_level(info->dev_root->node));
1991
1992 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1993 btrfs_set_backup_csum_root_gen(root_backup,
1994 btrfs_header_generation(info->csum_root->node));
1995 btrfs_set_backup_csum_root_level(root_backup,
1996 btrfs_header_level(info->csum_root->node));
1997
1998 btrfs_set_backup_total_bytes(root_backup,
1999 btrfs_super_total_bytes(info->super_copy));
2000 btrfs_set_backup_bytes_used(root_backup,
2001 btrfs_super_bytes_used(info->super_copy));
2002 btrfs_set_backup_num_devices(root_backup,
2003 btrfs_super_num_devices(info->super_copy));
2004
2005 /*
2006 * if we don't copy this out to the super_copy, it won't get remembered
2007 * for the next commit
2008 */
2009 memcpy(&info->super_copy->super_roots,
2010 &info->super_for_commit->super_roots,
2011 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2012}
2013
2014/*
2015 * this copies info out of the root backup array and back into
2016 * the in-memory super block. It is meant to help iterate through
2017 * the array, so you send it the number of backups you've already
2018 * tried and the last backup index you used.
2019 *
2020 * this returns -1 when it has tried all the backups
2021 */
2022static noinline int next_root_backup(struct btrfs_fs_info *info,
2023 struct btrfs_super_block *super,
2024 int *num_backups_tried, int *backup_index)
2025{
2026 struct btrfs_root_backup *root_backup;
2027 int newest = *backup_index;
2028
2029 if (*num_backups_tried == 0) {
2030 u64 gen = btrfs_super_generation(super);
2031
2032 newest = find_newest_super_backup(info, gen);
2033 if (newest == -1)
2034 return -1;
2035
2036 *backup_index = newest;
2037 *num_backups_tried = 1;
2038 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2039 /* we've tried all the backups, all done */
2040 return -1;
2041 } else {
2042 /* jump to the next oldest backup */
2043 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2044 BTRFS_NUM_BACKUP_ROOTS;
2045 *backup_index = newest;
2046 *num_backups_tried += 1;
2047 }
2048 root_backup = super->super_roots + newest;
2049
2050 btrfs_set_super_generation(super,
2051 btrfs_backup_tree_root_gen(root_backup));
2052 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2053 btrfs_set_super_root_level(super,
2054 btrfs_backup_tree_root_level(root_backup));
2055 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2056
2057 /*
2058 * fixme: the total bytes and num_devices need to match or we should
2059 * need a fsck
2060 */
2061 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2062 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2063 return 0;
2064}
2065
Liu Bo7abadb62013-03-17 02:10:31 +00002066/* helper to cleanup workers */
2067static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2068{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002069 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002070 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002071 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002072 btrfs_destroy_workqueue(fs_info->endio_workers);
2073 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2074 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002075 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002076 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002077 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2078 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2079 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002080 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002081 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002082 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002083 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002084 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002085 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002086 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002087}
2088
Rashika2e9f5952013-10-31 02:45:20 +05302089static void free_root_extent_buffers(struct btrfs_root *root)
2090{
2091 if (root) {
2092 free_extent_buffer(root->node);
2093 free_extent_buffer(root->commit_root);
2094 root->node = NULL;
2095 root->commit_root = NULL;
2096 }
2097}
2098
Chris Masonaf31f5e2011-11-03 15:17:42 -04002099/* helper to cleanup tree roots */
2100static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2101{
Rashika2e9f5952013-10-31 02:45:20 +05302102 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002103
Rashika2e9f5952013-10-31 02:45:20 +05302104 free_root_extent_buffers(info->dev_root);
2105 free_root_extent_buffers(info->extent_root);
2106 free_root_extent_buffers(info->csum_root);
2107 free_root_extent_buffers(info->quota_root);
2108 free_root_extent_buffers(info->uuid_root);
2109 if (chunk_root)
2110 free_root_extent_buffers(info->chunk_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002111}
2112
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002113void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002114{
2115 int ret;
2116 struct btrfs_root *gang[8];
2117 int i;
2118
2119 while (!list_empty(&fs_info->dead_roots)) {
2120 gang[0] = list_entry(fs_info->dead_roots.next,
2121 struct btrfs_root, root_list);
2122 list_del(&gang[0]->root_list);
2123
Miao Xie27cdeb72014-04-02 19:51:05 +08002124 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002125 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002126 } else {
2127 free_extent_buffer(gang[0]->node);
2128 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002129 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002130 }
2131 }
2132
2133 while (1) {
2134 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2135 (void **)gang, 0,
2136 ARRAY_SIZE(gang));
2137 if (!ret)
2138 break;
2139 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002140 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002141 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002142
2143 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2144 btrfs_free_log_root_tree(NULL, fs_info);
2145 btrfs_destroy_pinned_extent(fs_info->tree_root,
2146 fs_info->pinned_extents);
2147 }
Josef Bacik171f6532013-04-24 16:35:41 -04002148}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002149
Al Viroad2b2c82011-11-17 01:10:02 -05002150int open_ctree(struct super_block *sb,
2151 struct btrfs_fs_devices *fs_devices,
2152 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002153{
Chris Masondb945352007-10-15 16:15:53 -04002154 u32 sectorsize;
2155 u32 nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002156 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002157 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002158 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002159 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002160 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002161 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002162 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002163 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002164 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002165 struct btrfs_root *extent_root;
2166 struct btrfs_root *csum_root;
2167 struct btrfs_root *chunk_root;
2168 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002169 struct btrfs_root *quota_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002170 struct btrfs_root *uuid_root;
Chris Masone02119d2008-09-05 16:13:11 -04002171 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002172 int ret;
Yane58ca022008-04-01 11:21:34 -04002173 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002174 int num_backups_tried = 0;
2175 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002176 int max_active;
2177 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Stefan Behrens70f80172013-08-15 17:11:23 +02002178 bool create_uuid_tree;
2179 bool check_uuid_tree;
Chris Mason4543df72008-06-11 21:47:56 -04002180
Al Virof84a8bd2011-11-17 15:57:57 -05002181 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002182 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002183 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002184 err = -ENOMEM;
2185 goto fail;
2186 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002187
2188 ret = init_srcu_struct(&fs_info->subvol_srcu);
2189 if (ret) {
2190 err = ret;
2191 goto fail;
2192 }
2193
2194 ret = setup_bdi(fs_info, &fs_info->bdi);
2195 if (ret) {
2196 err = ret;
2197 goto fail_srcu;
2198 }
2199
Miao Xiee2d84522013-01-29 10:09:20 +00002200 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2201 if (ret) {
2202 err = ret;
2203 goto fail_bdi;
2204 }
2205 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2206 (1 + ilog2(nr_cpu_ids));
2207
Miao Xie963d6782013-01-29 10:10:51 +00002208 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2209 if (ret) {
2210 err = ret;
2211 goto fail_dirty_metadata_bytes;
2212 }
2213
Miao Xiec404e0d2014-01-30 16:46:55 +08002214 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2215 if (ret) {
2216 err = ret;
2217 goto fail_delalloc_bytes;
2218 }
2219
Yan, Zheng76dda932009-09-21 16:00:26 -04002220 fs_info->btree_inode = new_inode(sb);
2221 if (!fs_info->btree_inode) {
2222 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002223 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002224 }
2225
Chris Masona6591712011-07-19 12:04:14 -04002226 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002227
Yan, Zheng76dda932009-09-21 16:00:26 -04002228 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002229 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002230 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002231 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002232 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002233 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002234 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002235 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002236 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002237 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002238 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002239 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002240 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002241 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002242 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002243 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002244 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002245 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002246 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002247 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002248 mutex_init(&fs_info->delalloc_root_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002249 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002250
Josef Bacik58176a92007-08-29 15:47:34 -04002251 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002252 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002253 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002254 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002255 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002256 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002257 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2258 BTRFS_BLOCK_RSV_GLOBAL);
2259 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2260 BTRFS_BLOCK_RSV_DELALLOC);
2261 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2262 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2263 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2264 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2265 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002266 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002267 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002268 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002269 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002270 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002271 atomic_set(&fs_info->qgroup_op_seq, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002272 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002273 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002274 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002275 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002276 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002277 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002278 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002279 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
Josef Bacik0a2b2a82014-01-23 10:54:11 -05002280 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
Arne Jansen90519d62011-05-23 14:30:00 +02002281 /* readahead state */
2282 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2283 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002284
Chris Masonb34b0862008-12-19 15:43:22 -05002285 fs_info->thread_pool_size = min_t(unsigned long,
2286 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002287
Miao Xie199c2a92013-05-15 07:48:23 +00002288 INIT_LIST_HEAD(&fs_info->ordered_roots);
2289 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002290 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2291 GFP_NOFS);
2292 if (!fs_info->delayed_root) {
2293 err = -ENOMEM;
2294 goto fail_iput;
2295 }
2296 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002297
Arne Jansena2de7332011-03-08 14:14:00 +01002298 mutex_init(&fs_info->scrub_lock);
2299 atomic_set(&fs_info->scrubs_running, 0);
2300 atomic_set(&fs_info->scrub_pause_req, 0);
2301 atomic_set(&fs_info->scrubs_paused, 0);
2302 atomic_set(&fs_info->scrub_cancel_req, 0);
Miao Xiec404e0d2014-01-30 16:46:55 +08002303 init_waitqueue_head(&fs_info->replace_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002304 init_waitqueue_head(&fs_info->scrub_pause_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002305 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002306#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2307 fs_info->check_integrity_print_mask = 0;
2308#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002309
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002310 spin_lock_init(&fs_info->balance_lock);
2311 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002312 atomic_set(&fs_info->balance_running, 0);
2313 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002314 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002315 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002316 init_waitqueue_head(&fs_info->balance_wait_q);
Miao Xie21c7e752014-05-13 17:29:04 -07002317 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Chris Masona061fc82008-05-07 11:43:44 -04002318
Chris Mason0afbaf82008-04-18 14:17:20 -04002319 sb->s_blocksize = 4096;
2320 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002321 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002322
Yan, Zheng76dda932009-09-21 16:00:26 -04002323 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002324 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002325 /*
2326 * we set the i_size on the btree inode to the max possible int.
2327 * the real end of the address space is determined by all of
2328 * the devices in the system
2329 */
2330 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002331 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002332 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2333
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002334 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002335 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002336 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002337 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002338 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002339
Chris Masond1310b22008-01-24 16:13:08 -05002340 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2341
Chris Mason0f7d52f2007-04-09 10:42:37 -04002342 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2343 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2344 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002345 set_bit(BTRFS_INODE_DUMMY,
2346 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Filipe David Borba Manana778ba822013-10-06 22:22:33 +01002347 btrfs_insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002348
Chris Masone02119d2008-09-05 16:13:11 -04002349 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002350 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002351 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002352
Yan Zheng11833d62009-09-11 16:11:19 -04002353 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002354 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002355 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002356 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002357 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002358 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002359
Chris Masond98237b2007-03-28 13:57:48 -04002360
Chris Mason5a3f23d2009-03-31 13:27:11 -04002361 mutex_init(&fs_info->ordered_operations_mutex);
Josef Bacik9ffba8c2013-08-14 11:33:56 -04002362 mutex_init(&fs_info->ordered_extent_flush_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002363 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002364 mutex_init(&fs_info->chunk_mutex);
2365 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002366 mutex_init(&fs_info->cleaner_mutex);
2367 mutex_init(&fs_info->volume_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002368 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002369 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002370 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002371 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Stefan Behrense922e082012-11-05 17:26:40 +01002372 fs_info->dev_replace.lock_owner = 0;
2373 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2374 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2375 mutex_init(&fs_info->dev_replace.lock_management_lock);
2376 mutex_init(&fs_info->dev_replace.lock);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002377
Arne Jansen416ac512011-09-13 12:56:09 +02002378 spin_lock_init(&fs_info->qgroup_lock);
Wang Shilongf2f6ed32013-04-07 10:50:16 +00002379 mutex_init(&fs_info->qgroup_ioctl_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002380 fs_info->qgroup_tree = RB_ROOT;
Josef Bacikfcebe452014-05-13 17:30:47 -07002381 fs_info->qgroup_op_tree = RB_ROOT;
Arne Jansen416ac512011-09-13 12:56:09 +02002382 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2383 fs_info->qgroup_seq = 1;
2384 fs_info->quota_enabled = 0;
2385 fs_info->pending_quota_state = 0;
Wang Shilong1e8f9152013-05-06 11:03:27 +00002386 fs_info->qgroup_ulist = NULL;
Jan Schmidt2f232032013-04-25 16:04:51 +00002387 mutex_init(&fs_info->qgroup_rescan_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002388
Chris Masonfa9c0d792009-04-03 09:47:43 -04002389 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2390 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2391
Chris Mason7d9eb122008-07-08 14:19:17 -04002392 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002393 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002394 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002395 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002396
David Woodhouse53b381b2013-01-29 18:40:14 -05002397 ret = btrfs_alloc_stripe_hash_table(fs_info);
2398 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002399 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002400 goto fail_alloc;
2401 }
2402
David Sterba707e8a02014-06-04 19:22:26 +02002403 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002404 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002405
Chris Mason3c4bb262012-03-27 18:56:56 -04002406 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002407
2408 /*
2409 * Read super block and check the signature bytes only
2410 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002411 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002412 if (!bh) {
2413 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002414 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002415 }
Chris Mason39279cc2007-06-12 06:35:45 -04002416
David Sterba1104a882013-03-06 15:57:46 +01002417 /*
2418 * We want to check superblock checksum, the type is stored inside.
2419 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2420 */
2421 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002422 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002423 err = -EINVAL;
2424 goto fail_alloc;
2425 }
2426
2427 /*
2428 * super_copy is zeroed at allocation time and we never touch the
2429 * following bytes up to INFO_SIZE, the checksum is calculated from
2430 * the whole block of INFO_SIZE
2431 */
David Sterba6c417612011-04-13 15:41:04 +02002432 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2433 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2434 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002435 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002436
David Sterba6c417612011-04-13 15:41:04 +02002437 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002438
David Sterba1104a882013-03-06 15:57:46 +01002439 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2440 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002441 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002442 err = -EINVAL;
2443 goto fail_alloc;
2444 }
2445
David Sterba6c417612011-04-13 15:41:04 +02002446 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002447 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002448 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002449
liuboacce9522011-01-06 19:30:25 +08002450 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002451 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2452 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002453
Liu Bo75e7cb72011-03-22 10:12:20 +00002454 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002455 * run through our array of backup supers and setup
2456 * our ring pointer to the oldest one
2457 */
2458 generation = btrfs_super_generation(disk_super);
2459 find_oldest_super_backup(fs_info, generation);
2460
2461 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002462 * In the long term, we'll store the compression type in the super
2463 * block, and it'll be used for per file compression control.
2464 */
2465 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2466
Yan Zheng2b820322008-11-17 21:11:30 -05002467 ret = btrfs_parse_options(tree_root, options);
2468 if (ret) {
2469 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002470 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002471 }
Chris Masondfe25022008-05-13 13:46:40 -04002472
Josef Bacikf2b636e2008-12-02 06:36:08 -05002473 features = btrfs_super_incompat_flags(disk_super) &
2474 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2475 if (features) {
2476 printk(KERN_ERR "BTRFS: couldn't mount because of "
2477 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002478 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002479 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002480 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002481 }
2482
David Sterba707e8a02014-06-04 19:22:26 +02002483 /*
2484 * Leafsize and nodesize were always equal, this is only a sanity check.
2485 */
2486 if (le32_to_cpu(disk_super->__unused_leafsize) !=
Chris Mason727011e2010-08-06 13:21:20 -04002487 btrfs_super_nodesize(disk_super)) {
2488 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2489 "blocksizes don't match. node %d leaf %d\n",
2490 btrfs_super_nodesize(disk_super),
David Sterba707e8a02014-06-04 19:22:26 +02002491 le32_to_cpu(disk_super->__unused_leafsize));
Chris Mason727011e2010-08-06 13:21:20 -04002492 err = -EINVAL;
2493 goto fail_alloc;
2494 }
David Sterba707e8a02014-06-04 19:22:26 +02002495 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002496 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2497 "blocksize (%d) was too large\n",
David Sterba707e8a02014-06-04 19:22:26 +02002498 btrfs_super_nodesize(disk_super));
Chris Mason727011e2010-08-06 13:21:20 -04002499 err = -EINVAL;
2500 goto fail_alloc;
2501 }
2502
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002503 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002504 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002505 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002506 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002507
Josef Bacik3173a182013-03-07 14:22:04 -05002508 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
Frank Holtonefe120a2013-12-20 11:37:06 -05002509 printk(KERN_ERR "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002510
Chris Mason727011e2010-08-06 13:21:20 -04002511 /*
2512 * flag our filesystem as having big metadata blocks if
2513 * they are bigger than the page size
2514 */
David Sterba707e8a02014-06-04 19:22:26 +02002515 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002516 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002517 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002518 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2519 }
2520
Chris Masonbc3f1162012-03-29 17:02:47 -04002521 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002522 sectorsize = btrfs_super_sectorsize(disk_super);
2523 stripesize = btrfs_super_stripesize(disk_super);
David Sterba707e8a02014-06-04 19:22:26 +02002524 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002525 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002526
2527 /*
2528 * mixed block groups end up with duplicate but slightly offset
2529 * extent buffers for the same range. It leads to corruptions
2530 */
2531 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002532 (sectorsize != nodesize)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002533 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002534 "are not allowed for mixed block groups on %s\n",
2535 sb->s_id);
2536 goto fail_alloc;
2537 }
2538
Miao Xieceda0862013-04-11 10:30:16 +00002539 /*
2540 * Needn't use the lock because there is no other task which will
2541 * update the flag.
2542 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002543 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002544
Josef Bacikf2b636e2008-12-02 06:36:08 -05002545 features = btrfs_super_compat_ro_flags(disk_super) &
2546 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2547 if (!(sb->s_flags & MS_RDONLY) && features) {
2548 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2549 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002550 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002551 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002552 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002553 }
Chris Mason61d92c32009-10-02 19:11:56 -04002554
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002555 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002556
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002557 fs_info->workers =
2558 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2559 max_active, 16);
Chris Masonc8b97812008-10-29 14:49:59 -04002560
Qu Wenruoafe3d242014-02-28 10:46:07 +08002561 fs_info->delalloc_workers =
2562 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
Chris Mason771ed682008-11-06 22:02:51 -05002563
Qu Wenruoa44903a2014-02-28 10:46:09 +08002564 fs_info->flush_workers =
2565 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
Miao Xie8ccf6f192012-10-25 09:28:04 +00002566
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002567 fs_info->caching_workers =
2568 btrfs_alloc_workqueue("cache", flags, max_active, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002569
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002570 /*
2571 * a higher idle thresh on the submit workers makes it much more
Chris Mason61b49442008-07-31 15:42:53 -04002572 * likely that bios will be send down in a sane order to the
2573 * devices
2574 */
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002575 fs_info->submit_workers =
2576 btrfs_alloc_workqueue("submit", flags,
2577 min_t(u64, fs_devices->num_devices,
2578 max_active), 64);
Chris Mason53863232008-08-15 15:34:18 -04002579
Qu Wenruodc6e3202014-02-28 10:46:14 +08002580 fs_info->fixup_workers =
2581 btrfs_alloc_workqueue("fixup", flags, 1, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002582
2583 /*
2584 * endios are largely parallel and should have a very
2585 * low idle thresh
2586 */
Qu Wenruofccb5d82014-02-28 10:46:10 +08002587 fs_info->endio_workers =
2588 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2589 fs_info->endio_meta_workers =
2590 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2591 fs_info->endio_meta_write_workers =
2592 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2593 fs_info->endio_raid56_workers =
2594 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
Miao Xie8b110e32014-09-12 18:44:03 +08002595 fs_info->endio_repair_workers =
2596 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002597 fs_info->rmw_workers =
2598 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002599 fs_info->endio_write_workers =
2600 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2601 fs_info->endio_freespace_worker =
2602 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002603 fs_info->delayed_workers =
2604 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002605 fs_info->readahead_workers =
2606 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002607 fs_info->qgroup_rescan_workers =
2608 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
Chris Masona79b7d42014-05-22 16:18:52 -07002609 fs_info->extent_workers =
2610 btrfs_alloc_workqueue("extent-refs", flags,
2611 min_t(u64, fs_devices->num_devices,
2612 max_active), 8);
Chris Masonb51912c2009-02-04 09:23:24 -05002613
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002614 if (!(fs_info->workers && fs_info->delalloc_workers &&
Qu Wenruofccb5d82014-02-28 10:46:10 +08002615 fs_info->submit_workers && fs_info->flush_workers &&
2616 fs_info->endio_workers && fs_info->endio_meta_workers &&
2617 fs_info->endio_meta_write_workers &&
Miao Xie8b110e32014-09-12 18:44:03 +08002618 fs_info->endio_repair_workers &&
Qu Wenruofccb5d82014-02-28 10:46:10 +08002619 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002620 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
Qu Wenruodc6e3202014-02-28 10:46:14 +08002621 fs_info->caching_workers && fs_info->readahead_workers &&
Qu Wenruofc97fab2014-02-28 10:46:16 +08002622 fs_info->fixup_workers && fs_info->delayed_workers &&
Andrey Utkin56094ee2014-08-09 14:51:15 +03002623 fs_info->extent_workers &&
Qu Wenruofc97fab2014-02-28 10:46:16 +08002624 fs_info->qgroup_rescan_workers)) {
Ilya Dryomovfed425c2012-06-22 12:13:01 -06002625 err = -ENOMEM;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002626 goto fail_sb_buffer;
2627 }
Chris Mason4543df72008-06-11 21:47:56 -04002628
Chris Mason4575c9c2008-04-18 16:13:31 -04002629 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002630 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2631 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002632
Chris Masondb945352007-10-15 16:15:53 -04002633 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002634 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002635 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002636
2637 sb->s_blocksize = sectorsize;
2638 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002639
Qu Wenruo3cae2102013-07-16 11:19:18 +08002640 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002641 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002642 goto fail_sb_buffer;
2643 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002644
Liu Bo8d082fb2012-04-03 09:56:53 +08002645 if (sectorsize != PAGE_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002646 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
Liu Bo8d082fb2012-04-03 09:56:53 +08002647 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002648 goto fail_sb_buffer;
2649 }
2650
Chris Mason925baed2008-06-25 16:01:30 -04002651 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002652 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002653 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002654 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002655 printk(KERN_WARNING "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002656 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002657 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002658 }
Chris Mason0b86a832008-03-24 15:01:56 -04002659
David Sterba707e8a02014-06-04 19:22:26 +02002660 blocksize = tree_root->nodesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002661 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002662
David Sterba707e8a02014-06-04 19:22:26 +02002663 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2664 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002665
2666 chunk_root->node = read_tree_block(chunk_root,
2667 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002668 blocksize, generation);
Josef Bacik416bc652013-04-23 14:17:42 -04002669 if (!chunk_root->node ||
2670 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002671 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002672 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002673 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002674 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002675 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2676 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002677
Chris Masone17cade2008-04-15 15:41:47 -04002678 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002679 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002680
Chris Mason0b86a832008-03-24 15:01:56 -04002681 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002682 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002683 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002684 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002685 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002686 }
Chris Mason0b86a832008-03-24 15:01:56 -04002687
Stefan Behrens8dabb742012-11-06 13:15:27 +01002688 /*
2689 * keep the device that is marked to be the target device for the
2690 * dev_replace procedure
2691 */
2692 btrfs_close_extra_devices(fs_info, fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002693
Chris Masona6b0d5c2012-02-20 20:53:43 -05002694 if (!fs_devices->latest_bdev) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002695 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002696 sb->s_id);
2697 goto fail_tree_roots;
2698 }
2699
Chris Masonaf31f5e2011-11-03 15:17:42 -04002700retry_root_backup:
David Sterba707e8a02014-06-04 19:22:26 +02002701 blocksize = tree_root->nodesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002702 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002703
Chris Masone20d96d2007-03-22 12:13:20 -04002704 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002705 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002706 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002707 if (!tree_root->node ||
2708 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002709 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002710 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002711
2712 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002713 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002714
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002715 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2716 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002717 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002718
Miao Xiecb517ea2013-05-15 07:48:19 +00002719 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2720 location.type = BTRFS_ROOT_ITEM_KEY;
2721 location.offset = 0;
2722
2723 extent_root = btrfs_read_tree_root(tree_root, &location);
2724 if (IS_ERR(extent_root)) {
2725 ret = PTR_ERR(extent_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002726 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002727 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002728 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002729 fs_info->extent_root = extent_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002730
Miao Xiecb517ea2013-05-15 07:48:19 +00002731 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2732 dev_root = btrfs_read_tree_root(tree_root, &location);
2733 if (IS_ERR(dev_root)) {
2734 ret = PTR_ERR(dev_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002735 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002736 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002737 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002738 fs_info->dev_root = dev_root;
2739 btrfs_init_devices_late(fs_info);
Chris Mason3768f362007-03-13 16:47:54 -04002740
Miao Xiecb517ea2013-05-15 07:48:19 +00002741 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2742 csum_root = btrfs_read_tree_root(tree_root, &location);
2743 if (IS_ERR(csum_root)) {
2744 ret = PTR_ERR(csum_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002745 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002746 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002747 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002748 fs_info->csum_root = csum_root;
Chris Masond20f7042008-12-08 16:58:54 -05002749
Miao Xiecb517ea2013-05-15 07:48:19 +00002750 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2751 quota_root = btrfs_read_tree_root(tree_root, &location);
2752 if (!IS_ERR(quota_root)) {
Miao Xie27cdeb72014-04-02 19:51:05 +08002753 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002754 fs_info->quota_enabled = 1;
2755 fs_info->pending_quota_state = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002756 fs_info->quota_root = quota_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002757 }
2758
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002759 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2760 uuid_root = btrfs_read_tree_root(tree_root, &location);
2761 if (IS_ERR(uuid_root)) {
2762 ret = PTR_ERR(uuid_root);
2763 if (ret != -ENOENT)
2764 goto recovery_tree_root;
2765 create_uuid_tree = true;
Stefan Behrens70f80172013-08-15 17:11:23 +02002766 check_uuid_tree = false;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002767 } else {
Miao Xie27cdeb72014-04-02 19:51:05 +08002768 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002769 fs_info->uuid_root = uuid_root;
Stefan Behrens70f80172013-08-15 17:11:23 +02002770 create_uuid_tree = false;
2771 check_uuid_tree =
2772 generation != btrfs_super_uuid_tree_generation(disk_super);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002773 }
2774
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002775 fs_info->generation = generation;
2776 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002777
Ilya Dryomov68310a52012-06-22 12:24:12 -06002778 ret = btrfs_recover_balance(fs_info);
2779 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002780 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002781 goto fail_block_groups;
2782 }
2783
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002784 ret = btrfs_init_dev_stats(fs_info);
2785 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002786 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002787 ret);
2788 goto fail_block_groups;
2789 }
2790
Stefan Behrens8dabb742012-11-06 13:15:27 +01002791 ret = btrfs_init_dev_replace(fs_info);
2792 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002793 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002794 goto fail_block_groups;
2795 }
2796
2797 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2798
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002799 ret = btrfs_sysfs_add_one(fs_info);
2800 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002801 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002802 goto fail_block_groups;
2803 }
2804
liuboc59021f2011-03-07 02:13:14 +00002805 ret = btrfs_init_space_info(fs_info);
2806 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002807 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002808 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002809 }
2810
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002811 ret = btrfs_read_block_groups(extent_root);
2812 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002813 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002814 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002815 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002816 fs_info->num_tolerated_disk_barrier_failures =
2817 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002818 if (fs_info->fs_devices->missing_devices >
2819 fs_info->num_tolerated_disk_barrier_failures &&
2820 !(sb->s_flags & MS_RDONLY)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002821 printk(KERN_WARNING "BTRFS: "
2822 "too many missing devices, writeable mount is not allowed\n");
Anand Jain2365dd32014-01-22 11:15:51 +08002823 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002824 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002825
Chris Masona74a4b92008-06-25 16:01:31 -04002826 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2827 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002828 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08002829 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04002830
2831 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2832 tree_root,
2833 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002834 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002835 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002836
Chris Masonc2898112009-06-10 09:51:32 -04002837 if (!btrfs_test_opt(tree_root, SSD) &&
2838 !btrfs_test_opt(tree_root, NOSSD) &&
2839 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002840 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04002841 "mode\n");
2842 btrfs_set_opt(fs_info->mount_opt, SSD);
2843 }
2844
Qu Wenruo3818aea2014-01-13 13:36:06 +08002845 /* Set the real inode map cache flag */
2846 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2847 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2848
Stefan Behrens21adbd52011-11-09 13:44:05 +01002849#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2850 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2851 ret = btrfsic_mount(tree_root, fs_devices,
2852 btrfs_test_opt(tree_root,
2853 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2854 1 : 0,
2855 fs_info->check_integrity_print_mask);
2856 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05002857 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01002858 " integrity check module %s\n", sb->s_id);
2859 }
2860#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002861 ret = btrfs_read_qgroup_config(fs_info);
2862 if (ret)
2863 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002864
liuboacce9522011-01-06 19:30:25 +08002865 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002866 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002867 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002868
Chris Mason7c2ca462008-11-19 15:13:35 -05002869 if (fs_devices->rw_devices == 0) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002870 printk(KERN_WARNING "BTRFS: log replay required "
Chris Masond3977122009-01-05 21:25:51 -05002871 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002872 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002873 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002874 }
David Sterba707e8a02014-06-04 19:22:26 +02002875 blocksize = tree_root->nodesize;
Chris Masone02119d2008-09-05 16:13:11 -04002876
Al Viro6f07e422011-11-17 00:46:16 -05002877 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002878 if (!log_tree_root) {
2879 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002880 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002881 }
Chris Masone02119d2008-09-05 16:13:11 -04002882
David Sterba707e8a02014-06-04 19:22:26 +02002883 __setup_root(nodesize, sectorsize, stripesize,
Chris Masone02119d2008-09-05 16:13:11 -04002884 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2885
2886 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002887 blocksize,
2888 generation + 1);
Josef Bacik416bc652013-04-23 14:17:42 -04002889 if (!log_tree_root->node ||
2890 !extent_buffer_uptodate(log_tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002891 printk(KERN_ERR "BTRFS: failed to read log tree\n");
Josef Bacik416bc652013-04-23 14:17:42 -04002892 free_extent_buffer(log_tree_root->node);
2893 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002894 goto fail_qgroup;
Josef Bacik416bc652013-04-23 14:17:42 -04002895 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002896 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002897 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002898 if (ret) {
2899 btrfs_error(tree_root->fs_info, ret,
2900 "Failed to recover log tree");
2901 free_extent_buffer(log_tree_root->node);
2902 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002903 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002904 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002905
2906 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002907 ret = btrfs_commit_super(tree_root);
2908 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002909 goto fail_qgroup;
Yan Zhenge556ce22008-11-20 10:25:19 -05002910 }
Chris Masone02119d2008-09-05 16:13:11 -04002911 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002912
Yan, Zheng76dda932009-09-21 16:00:26 -04002913 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002914 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002915 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04002916
Chris Mason7c2ca462008-11-19 15:13:35 -05002917 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002918 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002919 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002920 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002921
Wang Shilong5f316482014-06-26 11:08:16 +08002922 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002923 ret = btrfs_recover_relocation(tree_root);
Wang Shilong5f316482014-06-26 11:08:16 +08002924 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00002925 if (ret < 0) {
2926 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05002927 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00002928 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002929 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002930 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002931 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002932
Chris Mason3de45862008-11-17 21:02:50 -05002933 location.objectid = BTRFS_FS_TREE_OBJECTID;
2934 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00002935 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05002936
Chris Mason3de45862008-11-17 21:02:50 -05002937 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002938 if (IS_ERR(fs_info->fs_root)) {
2939 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002940 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002941 }
Chris Masonc2898112009-06-10 09:51:32 -04002942
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002943 if (sb->s_flags & MS_RDONLY)
2944 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002945
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002946 down_read(&fs_info->cleanup_work_sem);
2947 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2948 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002949 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002950 close_ctree(tree_root);
2951 return ret;
2952 }
2953 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002954
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002955 ret = btrfs_resume_balance_async(fs_info);
2956 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002957 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002958 close_ctree(tree_root);
2959 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002960 }
2961
Stefan Behrens8dabb742012-11-06 13:15:27 +01002962 ret = btrfs_resume_dev_replace_async(fs_info);
2963 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002964 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01002965 close_ctree(tree_root);
2966 return ret;
2967 }
2968
Jan Schmidtb382a322013-05-28 15:47:24 +00002969 btrfs_qgroup_rescan_resume(fs_info);
2970
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002971 if (create_uuid_tree) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002972 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002973 ret = btrfs_create_uuid_tree(fs_info);
2974 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002975 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002976 ret);
2977 close_ctree(tree_root);
2978 return ret;
2979 }
Stefan Behrensf420ee12013-08-15 17:11:24 +02002980 } else if (check_uuid_tree ||
2981 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002982 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02002983 ret = btrfs_check_uuid_tree(fs_info);
2984 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002985 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02002986 ret);
2987 close_ctree(tree_root);
2988 return ret;
2989 }
2990 } else {
2991 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002992 }
2993
Josef Bacik47ab2a62014-09-18 11:20:02 -04002994 fs_info->open = 1;
2995
Al Viroad2b2c82011-11-17 01:10:02 -05002996 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002997
Arne Jansenbcef60f2011-09-13 15:23:30 +02002998fail_qgroup:
2999 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05003000fail_trans_kthread:
3001 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04003002 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003003 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04003004fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04003005 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05003006
3007 /*
3008 * make sure we're done with the btree inode before we stop our
3009 * kthreads
3010 */
3011 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05003012
Anand Jain2365dd32014-01-22 11:15:51 +08003013fail_sysfs:
3014 btrfs_sysfs_remove_one(fs_info);
3015
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003016fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003017 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003018 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003019
3020fail_tree_roots:
3021 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003022 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003023
Chris Mason39279cc2007-06-12 06:35:45 -04003024fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003025 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003026fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003027fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003028 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3029
Chris Mason4543df72008-06-11 21:47:56 -04003030 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003031fail_bio_counter:
3032 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003033fail_delalloc_bytes:
3034 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003035fail_dirty_metadata_bytes:
3036 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003037fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003038 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003039fail_srcu:
3040 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003041fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003042 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003043 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003044 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003045
3046recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003047 if (!btrfs_test_opt(tree_root, RECOVERY))
3048 goto fail_tree_roots;
3049
3050 free_root_pointers(fs_info, 0);
3051
3052 /* don't use the log in recovery mode, it won't be valid */
3053 btrfs_set_super_log_root(disk_super, 0);
3054
3055 /* we can't trust the free space cache either */
3056 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3057
3058 ret = next_root_backup(fs_info, fs_info->super_copy,
3059 &num_backups_tried, &backup_index);
3060 if (ret == -1)
3061 goto fail_block_groups;
3062 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003063}
3064
Chris Masonf2984462008-04-10 16:19:33 -04003065static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3066{
Chris Masonf2984462008-04-10 16:19:33 -04003067 if (uptodate) {
3068 set_buffer_uptodate(bh);
3069 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003070 struct btrfs_device *device = (struct btrfs_device *)
3071 bh->b_private;
3072
Frank Holtonefe120a2013-12-20 11:37:06 -05003073 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
Josef Bacik606686e2012-06-04 14:03:51 -04003074 "I/O error on %s\n",
3075 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003076 /* note, we dont' set_buffer_write_io_error because we have
3077 * our own ways of dealing with the IO errors
3078 */
Chris Masonf2984462008-04-10 16:19:33 -04003079 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003080 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003081 }
3082 unlock_buffer(bh);
3083 put_bh(bh);
3084}
3085
Yan Zhenga512bbf2008-12-08 16:46:26 -05003086struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3087{
3088 struct buffer_head *bh;
3089 struct buffer_head *latest = NULL;
3090 struct btrfs_super_block *super;
3091 int i;
3092 u64 transid = 0;
3093 u64 bytenr;
3094
3095 /* we would like to check all the supers, but that would make
3096 * a btrfs mount succeed after a mkfs from a different FS.
3097 * So, we need to add a special mount option to scan for
3098 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3099 */
3100 for (i = 0; i < 1; i++) {
3101 bytenr = btrfs_sb_offset(i);
Anand Jain8068a472013-07-26 01:29:35 +08003102 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3103 i_size_read(bdev->bd_inode))
Yan Zhenga512bbf2008-12-08 16:46:26 -05003104 break;
Anand Jain8068a472013-07-26 01:29:35 +08003105 bh = __bread(bdev, bytenr / 4096,
3106 BTRFS_SUPER_INFO_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003107 if (!bh)
3108 continue;
3109
3110 super = (struct btrfs_super_block *)bh->b_data;
3111 if (btrfs_super_bytenr(super) != bytenr ||
Qu Wenruo3cae2102013-07-16 11:19:18 +08003112 btrfs_super_magic(super) != BTRFS_MAGIC) {
Yan Zhenga512bbf2008-12-08 16:46:26 -05003113 brelse(bh);
3114 continue;
3115 }
3116
3117 if (!latest || btrfs_super_generation(super) > transid) {
3118 brelse(latest);
3119 latest = bh;
3120 transid = btrfs_super_generation(super);
3121 } else {
3122 brelse(bh);
3123 }
3124 }
3125 return latest;
3126}
3127
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003128/*
3129 * this should be called twice, once with wait == 0 and
3130 * once with wait == 1. When wait == 0 is done, all the buffer heads
3131 * we write are pinned.
3132 *
3133 * They are released when wait == 1 is done.
3134 * max_mirrors must be the same for both runs, and it indicates how
3135 * many supers on this one device should be written.
3136 *
3137 * max_mirrors == 0 means to write them all.
3138 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003139static int write_dev_supers(struct btrfs_device *device,
3140 struct btrfs_super_block *sb,
3141 int do_barriers, int wait, int max_mirrors)
3142{
3143 struct buffer_head *bh;
3144 int i;
3145 int ret;
3146 int errors = 0;
3147 u32 crc;
3148 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003149
3150 if (max_mirrors == 0)
3151 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3152
Yan Zhenga512bbf2008-12-08 16:46:26 -05003153 for (i = 0; i < max_mirrors; i++) {
3154 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003155 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3156 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003157 break;
3158
3159 if (wait) {
3160 bh = __find_get_block(device->bdev, bytenr / 4096,
3161 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003162 if (!bh) {
3163 errors++;
3164 continue;
3165 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003166 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003167 if (!buffer_uptodate(bh))
3168 errors++;
3169
3170 /* drop our reference */
3171 brelse(bh);
3172
3173 /* drop the reference from the wait == 0 run */
3174 brelse(bh);
3175 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003176 } else {
3177 btrfs_set_super_bytenr(sb, bytenr);
3178
3179 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003180 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003181 BTRFS_CSUM_SIZE, crc,
3182 BTRFS_SUPER_INFO_SIZE -
3183 BTRFS_CSUM_SIZE);
3184 btrfs_csum_final(crc, sb->csum);
3185
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003186 /*
3187 * one reference for us, and we leave it for the
3188 * caller
3189 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003190 bh = __getblk(device->bdev, bytenr / 4096,
3191 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003192 if (!bh) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003193 printk(KERN_ERR "BTRFS: couldn't get super "
Josef Bacik634554d2013-04-29 10:05:57 -04003194 "buffer head for bytenr %Lu\n", bytenr);
3195 errors++;
3196 continue;
3197 }
3198
Yan Zhenga512bbf2008-12-08 16:46:26 -05003199 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3200
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003201 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003202 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003203
3204 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003205 lock_buffer(bh);
3206 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003207 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003208 }
3209
Chris Mason387125f2011-11-18 15:07:51 -05003210 /*
3211 * we fua the first super. The others we allow
3212 * to go down lazy.
3213 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003214 if (i == 0)
3215 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3216 else
3217 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003218 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003219 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003220 }
3221 return errors < i ? 0 : -1;
3222}
3223
Chris Mason387125f2011-11-18 15:07:51 -05003224/*
3225 * endio for the write_dev_flush, this will wake anyone waiting
3226 * for the barrier when it is done
3227 */
3228static void btrfs_end_empty_barrier(struct bio *bio, int err)
3229{
3230 if (err) {
3231 if (err == -EOPNOTSUPP)
3232 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3233 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3234 }
3235 if (bio->bi_private)
3236 complete(bio->bi_private);
3237 bio_put(bio);
3238}
3239
3240/*
3241 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3242 * sent down. With wait == 1, it waits for the previous flush.
3243 *
3244 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3245 * capable
3246 */
3247static int write_dev_flush(struct btrfs_device *device, int wait)
3248{
3249 struct bio *bio;
3250 int ret = 0;
3251
3252 if (device->nobarriers)
3253 return 0;
3254
3255 if (wait) {
3256 bio = device->flush_bio;
3257 if (!bio)
3258 return 0;
3259
3260 wait_for_completion(&device->flush_wait);
3261
3262 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003263 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
Josef Bacik606686e2012-06-04 14:03:51 -04003264 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05003265 device->nobarriers = 1;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003266 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
Chris Mason387125f2011-11-18 15:07:51 -05003267 ret = -EIO;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003268 btrfs_dev_stat_inc_and_print(device,
3269 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003270 }
3271
3272 /* drop the reference from the wait == 0 run */
3273 bio_put(bio);
3274 device->flush_bio = NULL;
3275
3276 return ret;
3277 }
3278
3279 /*
3280 * one reference for us, and we leave it for the
3281 * caller
3282 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003283 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003284 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003285 if (!bio)
3286 return -ENOMEM;
3287
3288 bio->bi_end_io = btrfs_end_empty_barrier;
3289 bio->bi_bdev = device->bdev;
3290 init_completion(&device->flush_wait);
3291 bio->bi_private = &device->flush_wait;
3292 device->flush_bio = bio;
3293
3294 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003295 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003296
3297 return 0;
3298}
3299
3300/*
3301 * send an empty flush down to each device in parallel,
3302 * then wait for them
3303 */
3304static int barrier_all_devices(struct btrfs_fs_info *info)
3305{
3306 struct list_head *head;
3307 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003308 int errors_send = 0;
3309 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003310 int ret;
3311
3312 /* send down all the barriers */
3313 head = &info->fs_devices->devices;
3314 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003315 if (dev->missing)
3316 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003317 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003318 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003319 continue;
3320 }
3321 if (!dev->in_fs_metadata || !dev->writeable)
3322 continue;
3323
3324 ret = write_dev_flush(dev, 0);
3325 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003326 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003327 }
3328
3329 /* wait for all the barriers */
3330 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003331 if (dev->missing)
3332 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003333 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003334 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003335 continue;
3336 }
3337 if (!dev->in_fs_metadata || !dev->writeable)
3338 continue;
3339
3340 ret = write_dev_flush(dev, 1);
3341 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003342 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003343 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003344 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3345 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003346 return -EIO;
3347 return 0;
3348}
3349
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003350int btrfs_calc_num_tolerated_disk_barrier_failures(
3351 struct btrfs_fs_info *fs_info)
3352{
3353 struct btrfs_ioctl_space_info space;
3354 struct btrfs_space_info *sinfo;
3355 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3356 BTRFS_BLOCK_GROUP_SYSTEM,
3357 BTRFS_BLOCK_GROUP_METADATA,
3358 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3359 int num_types = 4;
3360 int i;
3361 int c;
3362 int num_tolerated_disk_barrier_failures =
3363 (int)fs_info->fs_devices->num_devices;
3364
3365 for (i = 0; i < num_types; i++) {
3366 struct btrfs_space_info *tmp;
3367
3368 sinfo = NULL;
3369 rcu_read_lock();
3370 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3371 if (tmp->flags == types[i]) {
3372 sinfo = tmp;
3373 break;
3374 }
3375 }
3376 rcu_read_unlock();
3377
3378 if (!sinfo)
3379 continue;
3380
3381 down_read(&sinfo->groups_sem);
3382 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3383 if (!list_empty(&sinfo->block_groups[c])) {
3384 u64 flags;
3385
3386 btrfs_get_block_group_info(
3387 &sinfo->block_groups[c], &space);
3388 if (space.total_bytes == 0 ||
3389 space.used_bytes == 0)
3390 continue;
3391 flags = space.flags;
3392 /*
3393 * return
3394 * 0: if dup, single or RAID0 is configured for
3395 * any of metadata, system or data, else
3396 * 1: if RAID5 is configured, or if RAID1 or
3397 * RAID10 is configured and only two mirrors
3398 * are used, else
3399 * 2: if RAID6 is configured, else
3400 * num_mirrors - 1: if RAID1 or RAID10 is
3401 * configured and more than
3402 * 2 mirrors are used.
3403 */
3404 if (num_tolerated_disk_barrier_failures > 0 &&
3405 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3406 BTRFS_BLOCK_GROUP_RAID0)) ||
3407 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3408 == 0)))
3409 num_tolerated_disk_barrier_failures = 0;
David Woodhouse53b381b2013-01-29 18:40:14 -05003410 else if (num_tolerated_disk_barrier_failures > 1) {
3411 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3412 BTRFS_BLOCK_GROUP_RAID5 |
3413 BTRFS_BLOCK_GROUP_RAID10)) {
3414 num_tolerated_disk_barrier_failures = 1;
3415 } else if (flags &
Henrik Nordvik15b0a892013-04-29 18:09:23 +00003416 BTRFS_BLOCK_GROUP_RAID6) {
David Woodhouse53b381b2013-01-29 18:40:14 -05003417 num_tolerated_disk_barrier_failures = 2;
3418 }
3419 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003420 }
3421 }
3422 up_read(&sinfo->groups_sem);
3423 }
3424
3425 return num_tolerated_disk_barrier_failures;
3426}
3427
Eric Sandeen48a3b632013-04-25 20:41:01 +00003428static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003429{
Chris Masone5e9a522009-06-10 15:17:02 -04003430 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003431 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003432 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003433 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003434 int ret;
3435 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003436 int max_errors;
3437 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003438 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003439
3440 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003441 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003442
David Sterba6c417612011-04-13 15:41:04 +02003443 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003444 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003445
Chris Mason174ba502011-05-27 10:03:58 -04003446 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003447 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003448 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003449
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003450 if (do_barriers) {
3451 ret = barrier_all_devices(root->fs_info);
3452 if (ret) {
3453 mutex_unlock(
3454 &root->fs_info->fs_devices->device_list_mutex);
3455 btrfs_error(root->fs_info, ret,
3456 "errors while submitting device barriers.");
3457 return ret;
3458 }
3459 }
Chris Mason387125f2011-11-18 15:07:51 -05003460
Xiao Guangrong1f781602011-04-20 10:09:16 +00003461 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003462 if (!dev->bdev) {
3463 total_errors++;
3464 continue;
3465 }
Yan Zheng2b820322008-11-17 21:11:30 -05003466 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003467 continue;
3468
Yan Zheng2b820322008-11-17 21:11:30 -05003469 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003470 btrfs_set_stack_device_type(dev_item, dev->type);
3471 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003472 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003473 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003474 btrfs_set_stack_device_bytes_used(dev_item,
3475 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003476 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3477 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3478 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3479 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003480 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003481
Chris Masona061fc82008-05-07 11:43:44 -04003482 flags = btrfs_super_flags(sb);
3483 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003484
Yan Zhenga512bbf2008-12-08 16:46:26 -05003485 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003486 if (ret)
3487 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003488 }
Chris Masona236aed2008-04-29 09:38:00 -04003489 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003490 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003491 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003492 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003493
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003494 /* FUA is masked off if unsupported and can't be the reason */
3495 btrfs_error(root->fs_info, -EIO,
3496 "%d errors while writing supers", total_errors);
3497 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003498 }
Chris Masonf2984462008-04-10 16:19:33 -04003499
Yan Zhenga512bbf2008-12-08 16:46:26 -05003500 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003501 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003502 if (!dev->bdev)
3503 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003504 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003505 continue;
3506
Yan Zhenga512bbf2008-12-08 16:46:26 -05003507 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3508 if (ret)
3509 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003510 }
Chris Mason174ba502011-05-27 10:03:58 -04003511 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003512 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003513 btrfs_error(root->fs_info, -EIO,
3514 "%d errors while writing supers", total_errors);
3515 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003516 }
Chris Masonf2984462008-04-10 16:19:33 -04003517 return 0;
3518}
3519
Yan Zhenga512bbf2008-12-08 16:46:26 -05003520int write_ctree_super(struct btrfs_trans_handle *trans,
3521 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003522{
Rashikaf570e7572013-10-31 16:50:42 +05303523 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003524}
3525
Miao Xiecb517ea2013-05-15 07:48:19 +00003526/* Drop a fs root from the radix tree and free it. */
3527void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3528 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003529{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003530 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003531 radix_tree_delete(&fs_info->fs_roots_radix,
3532 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003533 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003534
3535 if (btrfs_root_refs(&root->root_item) == 0)
3536 synchronize_srcu(&fs_info->subvol_srcu);
3537
Liu Bo1a4319c2014-01-13 19:53:53 +08003538 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003539 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003540
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003541 if (root->free_ino_pinned)
3542 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3543 if (root->free_ino_ctl)
3544 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003545 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003546}
3547
3548static void free_fs_root(struct btrfs_root *root)
3549{
David Sterba57cdc8d2014-02-05 02:37:48 +01003550 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003551 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003552 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3553 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003554 if (root->anon_dev)
3555 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003556 if (root->subv_writers)
3557 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003558 free_extent_buffer(root->node);
3559 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003560 kfree(root->free_ino_ctl);
3561 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003562 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003563 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003564}
3565
Miao Xiecb517ea2013-05-15 07:48:19 +00003566void btrfs_free_fs_root(struct btrfs_root *root)
3567{
3568 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003569}
3570
Yan Zhengc146afa2008-11-12 14:34:12 -05003571int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3572{
3573 u64 root_objectid = 0;
3574 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003575 int i = 0;
3576 int err = 0;
3577 unsigned int ret = 0;
3578 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003579
3580 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003581 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003582 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3583 (void **)gang, root_objectid,
3584 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003585 if (!ret) {
3586 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003587 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003588 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003589 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003590
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003591 for (i = 0; i < ret; i++) {
3592 /* Avoid to grab roots in dead_roots */
3593 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3594 gang[i] = NULL;
3595 continue;
3596 }
3597 /* grab all the search result for later use */
3598 gang[i] = btrfs_grab_fs_root(gang[i]);
3599 }
3600 srcu_read_unlock(&fs_info->subvol_srcu, index);
3601
3602 for (i = 0; i < ret; i++) {
3603 if (!gang[i])
3604 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003605 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003606 err = btrfs_orphan_cleanup(gang[i]);
3607 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003608 break;
3609 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003610 }
3611 root_objectid++;
3612 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003613
3614 /* release the uncleaned roots due to error */
3615 for (; i < ret; i++) {
3616 if (gang[i])
3617 btrfs_put_fs_root(gang[i]);
3618 }
3619 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003620}
3621
3622int btrfs_commit_super(struct btrfs_root *root)
3623{
3624 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003625
3626 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00003627 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003628 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003629 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003630
3631 /* wait until ongoing cleanup work done */
3632 down_write(&root->fs_info->cleanup_work_sem);
3633 up_write(&root->fs_info->cleanup_work_sem);
3634
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003635 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003636 if (IS_ERR(trans))
3637 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003638 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003639}
3640
David Sterba3abdbd72014-06-04 18:10:45 +02003641void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003642{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003643 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003644 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003645
Chris Masonfacda1e2007-06-08 18:11:48 -04003646 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003647 smp_mb();
3648
Stefan Behrens803b2f52013-08-15 17:11:21 +02003649 /* wait for the uuid_scan task to finish */
3650 down(&fs_info->uuid_tree_rescan_sem);
3651 /* avoid complains from lockdep et al., set sem back to initial state */
3652 up(&fs_info->uuid_tree_rescan_sem);
3653
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003654 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003655 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003656
Stefan Behrens8dabb742012-11-06 13:15:27 +01003657 btrfs_dev_replace_suspend_for_unmount(fs_info);
3658
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003659 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003660
3661 /* wait for any defraggers to finish */
3662 wait_event(fs_info->transaction_wait,
3663 (atomic_read(&fs_info->defrag_running) == 0));
3664
3665 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003666 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003667
Miao Xie21c7e752014-05-13 17:29:04 -07003668 cancel_work_sync(&fs_info->async_reclaim_work);
3669
Yan Zhengc146afa2008-11-12 14:34:12 -05003670 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003671 ret = btrfs_commit_super(root);
3672 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003673 btrfs_err(root->fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003674 }
3675
Miao Xie87533c42013-01-29 10:14:48 +00003676 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003677 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003678
Al Viroe3029d92011-11-17 00:56:18 -05003679 kthread_stop(fs_info->transaction_kthread);
3680 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003681
Yan Zhengf25784b2009-07-28 08:41:57 -04003682 fs_info->closing = 2;
3683 smp_mb();
3684
Arne Jansenbcef60f2011-09-13 15:23:30 +02003685 btrfs_free_qgroup_config(root->fs_info);
3686
Miao Xie963d6782013-01-29 10:10:51 +00003687 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003688 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003689 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003690 }
Yanbcc63ab2008-07-30 16:29:20 -04003691
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003692 btrfs_sysfs_remove_one(fs_info);
3693
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003694 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003695
Liu Bo1a4319c2014-01-13 19:53:53 +08003696 btrfs_put_block_group_cache(fs_info);
3697
Josef Bacik2b1360d2013-10-07 15:14:44 -04003698 btrfs_free_block_groups(fs_info);
3699
Wang Shilongde348ee2014-04-09 19:23:22 +08003700 /*
3701 * we must make sure there is not any read request to
3702 * submit after we stopping all workers.
3703 */
3704 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003705 btrfs_stop_all_workers(fs_info);
3706
Josef Bacik47ab2a62014-09-18 11:20:02 -04003707 fs_info->open = 0;
Josef Bacik13e6c372013-05-30 16:55:44 -04003708 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003709
Josef Bacik13e6c372013-05-30 16:55:44 -04003710 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003711
Stefan Behrens21adbd52011-11-09 13:44:05 +01003712#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3713 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3714 btrfsic_unmount(root, fs_info->fs_devices);
3715#endif
3716
Chris Masondfe25022008-05-13 13:46:40 -04003717 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003718 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003719
Miao Xiee2d84522013-01-29 10:09:20 +00003720 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003721 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003722 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003723 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003724 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003725
David Woodhouse53b381b2013-01-29 18:40:14 -05003726 btrfs_free_stripe_hash_table(fs_info);
3727
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003728 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3729 root->orphan_block_rsv = NULL;
Chris Masoneb60cea2007-02-02 09:18:22 -05003730}
3731
Chris Masonb9fab912012-05-06 07:23:47 -04003732int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3733 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003734{
Chris Mason1259ab72008-05-12 13:39:03 -04003735 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003736 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003737
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003738 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003739 if (!ret)
3740 return ret;
3741
3742 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003743 parent_transid, atomic);
3744 if (ret == -EAGAIN)
3745 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003746 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003747}
Chris Mason6702ed42007-08-07 16:15:09 -04003748
Chris Mason5f39d392007-10-15 16:14:19 -04003749int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3750{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003751 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003752}
3753
3754void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3755{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003756 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003757 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003758 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003759
Josef Bacik06ea65a2013-09-19 16:07:01 -04003760#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3761 /*
3762 * This is a fast path so only do this check if we have sanity tests
3763 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3764 * outside of the sanity tests.
3765 */
3766 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3767 return;
3768#endif
3769 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003770 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003771 if (transid != root->fs_info->generation)
3772 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003773 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003774 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003775 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003776 if (!was_dirty)
3777 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3778 buf->len,
3779 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003780#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3781 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3782 btrfs_print_leaf(root, buf);
3783 ASSERT(0);
3784 }
3785#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05003786}
3787
Liu Bob53d3f52012-11-14 14:34:34 +00003788static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3789 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003790{
Chris Mason188de642008-05-09 11:52:25 -04003791 /*
3792 * looks as though older kernels can get into trouble with
3793 * this code, they end up stuck in balance_dirty_pages forever
3794 */
Miao Xiee2d84522013-01-29 10:09:20 +00003795 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04003796
Jens Axboe6933c022009-03-17 09:36:37 +01003797 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003798 return;
3799
Liu Bob53d3f52012-11-14 14:34:34 +00003800 if (flush_delayed)
3801 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003802
Miao Xiee2d84522013-01-29 10:09:20 +00003803 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3804 BTRFS_DIRTY_METADATA_THRESH);
3805 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08003806 balance_dirty_pages_ratelimited(
3807 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08003808 }
3809 return;
3810}
3811
Liu Bob53d3f52012-11-14 14:34:34 +00003812void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08003813{
Liu Bob53d3f52012-11-14 14:34:34 +00003814 __btrfs_btree_balance_dirty(root, 1);
3815}
Miao Xie16cdcec2011-04-22 18:12:22 +08003816
Liu Bob53d3f52012-11-14 14:34:34 +00003817void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3818{
3819 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04003820}
Chris Mason6b800532007-10-15 16:17:34 -04003821
Chris Masonca7a79a2008-05-12 12:59:19 -04003822int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003823{
Chris Mason727011e2010-08-06 13:21:20 -04003824 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003825 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003826}
Chris Mason0da54682007-11-07 21:08:01 -05003827
David Sterbafcd1f0652012-03-06 00:06:18 +01003828static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003829 int read_only)
3830{
David Sterba1104a882013-03-06 15:57:46 +01003831 /*
3832 * Placeholder for checks
3833 */
David Sterbafcd1f0652012-03-06 00:06:18 +01003834 return 0;
liuboacce9522011-01-06 19:30:25 +08003835}
3836
Eric Sandeen48a3b632013-04-25 20:41:01 +00003837static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003838{
liuboacce9522011-01-06 19:30:25 +08003839 mutex_lock(&root->fs_info->cleaner_mutex);
3840 btrfs_run_delayed_iputs(root);
3841 mutex_unlock(&root->fs_info->cleaner_mutex);
3842
3843 down_write(&root->fs_info->cleanup_work_sem);
3844 up_write(&root->fs_info->cleanup_work_sem);
3845
3846 /* cleanup FS via transaction */
3847 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003848}
3849
Jeff Mahoney143bede2012-03-01 14:56:26 +01003850static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003851{
liuboacce9522011-01-06 19:30:25 +08003852 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08003853
Miao Xie199c2a92013-05-15 07:48:23 +00003854 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05003855 /*
3856 * This will just short circuit the ordered completion stuff which will
3857 * make sure the ordered extent gets properly cleaned up.
3858 */
Miao Xie199c2a92013-05-15 07:48:23 +00003859 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05003860 root_extent_list)
3861 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00003862 spin_unlock(&root->ordered_extent_lock);
3863}
3864
3865static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3866{
3867 struct btrfs_root *root;
3868 struct list_head splice;
3869
3870 INIT_LIST_HEAD(&splice);
3871
3872 spin_lock(&fs_info->ordered_root_lock);
3873 list_splice_init(&fs_info->ordered_roots, &splice);
3874 while (!list_empty(&splice)) {
3875 root = list_first_entry(&splice, struct btrfs_root,
3876 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04003877 list_move_tail(&root->ordered_root,
3878 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00003879
Liu Bo2a85d9c2014-02-10 17:07:16 +08003880 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003881 btrfs_destroy_ordered_extents(root);
3882
Liu Bo2a85d9c2014-02-10 17:07:16 +08003883 cond_resched();
3884 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003885 }
3886 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003887}
3888
Stefan Behrens35a36212013-08-14 18:12:25 +02003889static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3890 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003891{
3892 struct rb_node *node;
3893 struct btrfs_delayed_ref_root *delayed_refs;
3894 struct btrfs_delayed_ref_node *ref;
3895 int ret = 0;
3896
3897 delayed_refs = &trans->delayed_refs;
3898
3899 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003900 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003901 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05003902 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08003903 return ret;
3904 }
3905
Josef Bacikd7df2c72014-01-23 09:21:38 -05003906 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3907 struct btrfs_delayed_ref_head *head;
Josef Bacike78417d2013-06-03 16:42:36 -04003908 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08003909
Josef Bacikd7df2c72014-01-23 09:21:38 -05003910 head = rb_entry(node, struct btrfs_delayed_ref_head,
3911 href_node);
3912 if (!mutex_trylock(&head->mutex)) {
3913 atomic_inc(&head->node.refs);
3914 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003915
Josef Bacikd7df2c72014-01-23 09:21:38 -05003916 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04003917 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003918 btrfs_put_delayed_ref(&head->node);
3919 spin_lock(&delayed_refs->lock);
3920 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04003921 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05003922 spin_lock(&head->lock);
3923 while ((node = rb_first(&head->ref_root)) != NULL) {
3924 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3925 rb_node);
3926 ref->in_tree = 0;
3927 rb_erase(&ref->rb_node, &head->ref_root);
3928 atomic_dec(&delayed_refs->num_entries);
3929 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003930 }
3931 if (head->must_insert_reserved)
3932 pin_bytes = true;
3933 btrfs_free_delayed_extent_op(head->extent_op);
3934 delayed_refs->num_heads--;
3935 if (head->processing == 0)
3936 delayed_refs->num_heads_ready--;
3937 atomic_dec(&delayed_refs->num_entries);
3938 head->node.in_tree = 0;
3939 rb_erase(&head->href_node, &delayed_refs->href_root);
3940 spin_unlock(&head->lock);
3941 spin_unlock(&delayed_refs->lock);
3942 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08003943
Josef Bacikd7df2c72014-01-23 09:21:38 -05003944 if (pin_bytes)
3945 btrfs_pin_extent(root, head->node.bytenr,
3946 head->node.num_bytes, 1);
3947 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08003948 cond_resched();
3949 spin_lock(&delayed_refs->lock);
3950 }
3951
3952 spin_unlock(&delayed_refs->lock);
3953
3954 return ret;
3955}
3956
Jeff Mahoney143bede2012-03-01 14:56:26 +01003957static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003958{
3959 struct btrfs_inode *btrfs_inode;
3960 struct list_head splice;
3961
3962 INIT_LIST_HEAD(&splice);
3963
Miao Xieeb73c1b2013-05-15 07:48:22 +00003964 spin_lock(&root->delalloc_lock);
3965 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003966
3967 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00003968 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3969 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08003970
3971 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00003972 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3973 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00003974 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003975
3976 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00003977
Miao Xieeb73c1b2013-05-15 07:48:22 +00003978 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003979 }
3980
Miao Xieeb73c1b2013-05-15 07:48:22 +00003981 spin_unlock(&root->delalloc_lock);
3982}
3983
3984static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3985{
3986 struct btrfs_root *root;
3987 struct list_head splice;
3988
3989 INIT_LIST_HEAD(&splice);
3990
3991 spin_lock(&fs_info->delalloc_root_lock);
3992 list_splice_init(&fs_info->delalloc_roots, &splice);
3993 while (!list_empty(&splice)) {
3994 root = list_first_entry(&splice, struct btrfs_root,
3995 delalloc_root);
3996 list_del_init(&root->delalloc_root);
3997 root = btrfs_grab_fs_root(root);
3998 BUG_ON(!root);
3999 spin_unlock(&fs_info->delalloc_root_lock);
4000
4001 btrfs_destroy_delalloc_inodes(root);
4002 btrfs_put_fs_root(root);
4003
4004 spin_lock(&fs_info->delalloc_root_lock);
4005 }
4006 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004007}
4008
4009static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4010 struct extent_io_tree *dirty_pages,
4011 int mark)
4012{
4013 int ret;
liuboacce9522011-01-06 19:30:25 +08004014 struct extent_buffer *eb;
4015 u64 start = 0;
4016 u64 end;
liuboacce9522011-01-06 19:30:25 +08004017
4018 while (1) {
4019 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004020 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004021 if (ret)
4022 break;
4023
4024 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4025 while (start <= end) {
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004026 eb = btrfs_find_tree_block(root, start,
David Sterba707e8a02014-06-04 19:22:26 +02004027 root->nodesize);
4028 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004029 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004030 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004031 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004032
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004033 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4034 &eb->bflags))
4035 clear_extent_buffer_dirty(eb);
4036 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004037 }
4038 }
4039
4040 return ret;
4041}
4042
4043static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4044 struct extent_io_tree *pinned_extents)
4045{
4046 struct extent_io_tree *unpin;
4047 u64 start;
4048 u64 end;
4049 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004050 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004051
4052 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004053again:
liuboacce9522011-01-06 19:30:25 +08004054 while (1) {
4055 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004056 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004057 if (ret)
4058 break;
4059
4060 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00004061 if (btrfs_test_opt(root, DISCARD))
4062 ret = btrfs_error_discard_extent(root, start,
4063 end + 1 - start,
4064 NULL);
liuboacce9522011-01-06 19:30:25 +08004065
4066 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4067 btrfs_error_unpin_extent_range(root, start, end);
4068 cond_resched();
4069 }
4070
Liu Boed0eaa12012-06-14 02:23:21 -06004071 if (loop) {
4072 if (unpin == &root->fs_info->freed_extents[0])
4073 unpin = &root->fs_info->freed_extents[1];
4074 else
4075 unpin = &root->fs_info->freed_extents[0];
4076 loop = false;
4077 goto again;
4078 }
4079
liuboacce9522011-01-06 19:30:25 +08004080 return 0;
4081}
4082
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004083void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4084 struct btrfs_root *root)
4085{
4086 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004087
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004088 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004089 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004090
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004091 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004092 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004093
Miao Xie67cde342012-06-14 02:23:22 -06004094 btrfs_destroy_delayed_inodes(root);
4095 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004096
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004097 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4098 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004099 btrfs_destroy_pinned_extent(root,
4100 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004101
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004102 cur_trans->state =TRANS_STATE_COMPLETED;
4103 wake_up(&cur_trans->commit_wait);
4104
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004105 /*
4106 memset(cur_trans, 0, sizeof(*cur_trans));
4107 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4108 */
4109}
4110
Eric Sandeen48a3b632013-04-25 20:41:01 +00004111static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004112{
4113 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004114
liuboacce9522011-01-06 19:30:25 +08004115 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4116
Josef Bacika4abeea2011-04-11 17:25:13 -04004117 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004118 while (!list_empty(&root->fs_info->trans_list)) {
4119 t = list_first_entry(&root->fs_info->trans_list,
4120 struct btrfs_transaction, list);
4121 if (t->state >= TRANS_STATE_COMMIT_START) {
4122 atomic_inc(&t->use_count);
4123 spin_unlock(&root->fs_info->trans_lock);
4124 btrfs_wait_for_commit(root, t->transid);
4125 btrfs_put_transaction(t);
4126 spin_lock(&root->fs_info->trans_lock);
4127 continue;
4128 }
4129 if (t == root->fs_info->running_transaction) {
4130 t->state = TRANS_STATE_COMMIT_DOING;
4131 spin_unlock(&root->fs_info->trans_lock);
4132 /*
4133 * We wait for 0 num_writers since we don't hold a trans
4134 * handle open currently for this transaction.
4135 */
4136 wait_event(t->writer_wait,
4137 atomic_read(&t->num_writers) == 0);
4138 } else {
4139 spin_unlock(&root->fs_info->trans_lock);
4140 }
4141 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004142
Josef Bacik724e2312013-09-30 11:36:38 -04004143 spin_lock(&root->fs_info->trans_lock);
4144 if (t == root->fs_info->running_transaction)
4145 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004146 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004147 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004148
Josef Bacik724e2312013-09-30 11:36:38 -04004149 btrfs_put_transaction(t);
4150 trace_btrfs_transaction_commit(root);
4151 spin_lock(&root->fs_info->trans_lock);
4152 }
4153 spin_unlock(&root->fs_info->trans_lock);
4154 btrfs_destroy_all_ordered_extents(root->fs_info);
4155 btrfs_destroy_delayed_inodes(root);
4156 btrfs_assert_delayed_root_empty(root);
4157 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4158 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004159 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004160
4161 return 0;
4162}
4163
Chris Masond1310b22008-01-24 16:13:08 -05004164static struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004165 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004166 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004167 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004168 /* note we're sharing with inode.c for the merge bio hook */
4169 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004170};