blob: 4e47849d7427247dd179e8d4f320c5fcda687809 [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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090028#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050029#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020030#include <linux/ratelimit.h>
Stefan Behrens6463fe52013-04-19 15:08:05 +000031#include <linux/uuid.h>
Stefan Behrens803b2f52013-08-15 17:11:21 +020032#include <linux/semaphore.h>
David Sterba7e75bf32011-03-18 22:56:43 +000033#include <asm/unaligned.h>
Chris Masoneb60cea2007-02-02 09:18:22 -050034#include "ctree.h"
35#include "disk-io.h"
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +000036#include "hash.h"
Chris Masone089f052007-03-16 16:20:31 -040037#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040038#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040039#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040040#include "print-tree.h"
Chris Mason925baed2008-06-25 16:01:30 -040041#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040042#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040043#include "free-space-cache.h"
Omar Sandoval70f6d822015-09-29 20:50:38 -070044#include "free-space-tree.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"
Anand Jainebb87652016-03-10 17:26:59 +080052#include "compression.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050053
Josef Bacikde0022b2012-09-25 14:25:58 -040054#ifdef CONFIG_X86
55#include <asm/cpufeature.h>
56#endif
57
Qu Wenruo319e4d02015-12-15 09:14:36 +080058#define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\
59 BTRFS_HEADER_FLAG_RELOC |\
60 BTRFS_SUPER_FLAG_ERROR |\
61 BTRFS_SUPER_FLAG_SEEDING |\
62 BTRFS_SUPER_FLAG_METADUMP)
63
David Sterbae8c9f182015-01-02 18:23:10 +010064static const struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040065static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040066static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010067static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080068 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010069static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080070static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
71 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010072static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080073static int btrfs_destroy_marked_extents(struct btrfs_root *root,
74 struct extent_io_tree *dirty_pages,
75 int mark);
76static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
77 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000078static int btrfs_cleanup_transaction(struct btrfs_root *root);
79static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040080
Chris Masond352ac62008-09-29 15:18:18 -040081/*
David Sterba97eb6b62014-07-30 00:55:42 +020082 * btrfs_end_io_wq structs are used to do processing in task context when an IO
83 * is complete. This is used during reads to verify checksums, and it is used
Chris Masond352ac62008-09-29 15:18:18 -040084 * by writes to insert metadata for new file extents after IO is complete.
85 */
David Sterba97eb6b62014-07-30 00:55:42 +020086struct btrfs_end_io_wq {
Chris Masonce9adaa2008-04-09 16:28:12 -040087 struct bio *bio;
88 bio_end_io_t *end_io;
89 void *private;
90 struct btrfs_fs_info *info;
91 int error;
David Sterbabfebd8b2014-07-30 00:25:45 +020092 enum btrfs_wq_endio_type metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040093 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040094 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040095};
Chris Mason0da54682007-11-07 21:08:01 -050096
David Sterba97eb6b62014-07-30 00:55:42 +020097static struct kmem_cache *btrfs_end_io_wq_cache;
98
99int __init btrfs_end_io_wq_init(void)
100{
101 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
102 sizeof(struct btrfs_end_io_wq),
103 0,
104 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
105 NULL);
106 if (!btrfs_end_io_wq_cache)
107 return -ENOMEM;
108 return 0;
109}
110
111void btrfs_end_io_wq_exit(void)
112{
Kinglong Mee5598e902016-01-29 21:36:35 +0800113 kmem_cache_destroy(btrfs_end_io_wq_cache);
David Sterba97eb6b62014-07-30 00:55:42 +0200114}
115
Chris Masond352ac62008-09-29 15:18:18 -0400116/*
117 * async submit bios are used to offload expensive checksumming
118 * onto the worker threads. They checksum file and metadata bios
119 * just before they are sent down the IO stack.
120 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400121struct async_submit_bio {
122 struct inode *inode;
123 struct bio *bio;
124 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500125 extent_submit_bio_hook_t *submit_bio_start;
126 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400127 int rw;
128 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400129 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400130 /*
131 * bio_offset is optional, can be used if the pages in the bio
132 * can't tell us where in the file the bio should go
133 */
134 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400135 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100136 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400137};
138
Chris Mason85d4e462011-07-26 16:11:19 -0400139/*
140 * Lockdep class keys for extent_buffer->lock's in this root. For a given
141 * eb, the lockdep key is determined by the btrfs_root it belongs to and
142 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500143 *
Chris Mason85d4e462011-07-26 16:11:19 -0400144 * Different roots are used for different purposes and may nest inside each
145 * other and they require separate keysets. As lockdep keys should be
146 * static, assign keysets according to the purpose of the root as indicated
147 * by btrfs_root->objectid. This ensures that all special purpose roots
148 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500149 *
Chris Mason85d4e462011-07-26 16:11:19 -0400150 * Lock-nesting across peer nodes is always done with the immediate parent
151 * node locked thus preventing deadlock. As lockdep doesn't know this, use
152 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500153 *
Chris Mason85d4e462011-07-26 16:11:19 -0400154 * The key is set by the readpage_end_io_hook after the buffer has passed
155 * csum validation but before the pages are unlocked. It is also set by
156 * btrfs_init_new_buffer on freshly allocated blocks.
157 *
158 * We also add a check to make sure the highest level of the tree is the
159 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
160 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500161 */
162#ifdef CONFIG_DEBUG_LOCK_ALLOC
163# if BTRFS_MAX_LEVEL != 8
164# error
165# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400166
167static struct btrfs_lockdep_keyset {
168 u64 id; /* root objectid */
169 const char *name_stem; /* lock name stem */
170 char names[BTRFS_MAX_LEVEL + 1][20];
171 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
172} btrfs_lockdep_keysets[] = {
173 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
174 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
175 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
176 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
177 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
178 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000179 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400180 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
181 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
182 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200183 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
David Sterba6b20e0a2016-01-25 16:30:22 +0100184 { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" },
Chris Mason85d4e462011-07-26 16:11:19 -0400185 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500186};
Chris Mason85d4e462011-07-26 16:11:19 -0400187
188void __init btrfs_init_lockdep(void)
189{
190 int i, j;
191
192 /* initialize lockdep class names */
193 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
194 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
195
196 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
197 snprintf(ks->names[j], sizeof(ks->names[j]),
198 "btrfs-%s-%02d", ks->name_stem, j);
199 }
200}
201
202void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
203 int level)
204{
205 struct btrfs_lockdep_keyset *ks;
206
207 BUG_ON(level >= ARRAY_SIZE(ks->keys));
208
209 /* find the matching keyset, id 0 is the default entry */
210 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
211 if (ks->id == objectid)
212 break;
213
214 lockdep_set_class_and_name(&eb->lock,
215 &ks->keys[level], ks->names[level]);
216}
217
Chris Mason4008c042009-02-12 14:09:45 -0500218#endif
219
Chris Masond352ac62008-09-29 15:18:18 -0400220/*
221 * extents on the btree inode are pretty simple, there's one extent
222 * that covers the entire device
223 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500224static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200225 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500226 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400227{
228 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
229 struct extent_map *em;
230 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400231
Chris Mason890871b2009-09-02 16:24:52 -0400232 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500233 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400234 if (em) {
235 em->bdev =
236 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400237 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400238 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400239 }
Chris Mason890871b2009-09-02 16:24:52 -0400240 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400241
David Sterba172ddd62011-04-21 00:48:27 +0200242 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400243 if (!em) {
244 em = ERR_PTR(-ENOMEM);
245 goto out;
246 }
247 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400248 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400249 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400250 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400251 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500252
Chris Mason890871b2009-09-02 16:24:52 -0400253 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400254 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400255 if (ret == -EEXIST) {
256 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400257 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600258 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600259 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400260 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400261 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600262 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400263 }
Chris Mason890871b2009-09-02 16:24:52 -0400264 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400265
Chris Mason5f39d392007-10-15 16:14:19 -0400266out:
267 return em;
268}
269
Liu Bob0496682013-03-14 14:57:45 +0000270u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400271{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000272 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400273}
274
275void btrfs_csum_final(u32 crc, char *result)
276{
David Sterba7e75bf32011-03-18 22:56:43 +0000277 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400278}
279
Chris Masond352ac62008-09-29 15:18:18 -0400280/*
281 * compute the csum for a btree block, and either verify it or write it
282 * into the csum field of the block.
283 */
Daniel Dressler01d58472014-11-21 17:15:07 +0900284static int csum_tree_block(struct btrfs_fs_info *fs_info,
285 struct extent_buffer *buf,
Chris Mason19c00dd2007-10-15 16:19:22 -0400286 int verify)
287{
Daniel Dressler01d58472014-11-21 17:15:07 +0900288 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500289 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400290 unsigned long len;
291 unsigned long cur_len;
292 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400293 char *kaddr;
294 unsigned long map_start;
295 unsigned long map_len;
296 int err;
297 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500298 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400299
300 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500301 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400302 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400303 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500304 if (err)
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200305 return err;
Chris Mason19c00dd2007-10-15 16:19:22 -0400306 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000307 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400308 crc, cur_len);
309 len -= cur_len;
310 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400311 }
Josef Bacik607d4322008-12-02 07:17:45 -0500312 if (csum_size > sizeof(inline_result)) {
David Sterba31e818f2015-02-20 18:00:26 +0100313 result = kzalloc(csum_size, GFP_NOFS);
Josef Bacik607d4322008-12-02 07:17:45 -0500314 if (!result)
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200315 return -ENOMEM;
Josef Bacik607d4322008-12-02 07:17:45 -0500316 } else {
317 result = (char *)&inline_result;
318 }
319
Chris Mason19c00dd2007-10-15 16:19:22 -0400320 btrfs_csum_final(crc, result);
321
322 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500323 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500324 u32 val;
325 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500326 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500327
Josef Bacik607d4322008-12-02 07:17:45 -0500328 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba94647322015-10-08 11:01:36 +0200329 btrfs_warn_rl(fs_info,
330 "%s checksum verify failed on %llu wanted %X found %X "
331 "level %d",
Daniel Dressler01d58472014-11-21 17:15:07 +0900332 fs_info->sb->s_id, buf->start,
Frank Holtonefe120a2013-12-20 11:37:06 -0500333 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500334 if (result != (char *)&inline_result)
335 kfree(result);
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200336 return -EUCLEAN;
Chris Mason19c00dd2007-10-15 16:19:22 -0400337 }
338 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500339 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400340 }
Josef Bacik607d4322008-12-02 07:17:45 -0500341 if (result != (char *)&inline_result)
342 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400343 return 0;
344}
345
Chris Masond352ac62008-09-29 15:18:18 -0400346/*
347 * we can't consider a given block up to date unless the transid of the
348 * block matches the transid in the parent node's pointer. This is how we
349 * detect blocks that either didn't get written at all or got written
350 * in the wrong place.
351 */
Chris Mason1259ab72008-05-12 13:39:03 -0400352static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400353 struct extent_buffer *eb, u64 parent_transid,
354 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400355{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000356 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400357 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200358 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400359
360 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
361 return 0;
362
Chris Masonb9fab912012-05-06 07:23:47 -0400363 if (atomic)
364 return -EAGAIN;
365
Josef Bacika26e8c92014-03-28 17:07:27 -0400366 if (need_lock) {
367 btrfs_tree_read_lock(eb);
368 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
369 }
370
Josef Bacik2ac55d42010-02-03 19:33:23 +0000371 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
David Sterbaff13db42015-12-03 14:30:40 +0100372 &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400373 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400374 btrfs_header_generation(eb) == parent_transid) {
375 ret = 0;
376 goto out;
377 }
David Sterba94647322015-10-08 11:01:36 +0200378 btrfs_err_rl(eb->fs_info,
379 "parent transid verify failed on %llu wanted %llu found %llu",
380 eb->start,
Wang Shilong29549ae2014-07-04 17:59:06 +0800381 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400382 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400383
384 /*
385 * Things reading via commit roots that don't have normal protection,
386 * like send, can have a really old block in cache that may point at a
387 * block that has been free'd and re-allocated. So don't clear uptodate
388 * if we find an eb that is under IO (dirty/writeback) because we could
389 * end up reading in the stale data and then writing it back out and
390 * making everybody very sad.
391 */
392 if (!extent_buffer_under_io(eb))
393 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400394out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000395 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
396 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700397 if (need_lock)
398 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400399 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400400}
401
Chris Masond352ac62008-09-29 15:18:18 -0400402/*
David Sterba1104a882013-03-06 15:57:46 +0100403 * Return 0 if the superblock checksum type matches the checksum value of that
404 * algorithm. Pass the raw disk superblock data.
405 */
406static int btrfs_check_super_csum(char *raw_disk_sb)
407{
408 struct btrfs_super_block *disk_sb =
409 (struct btrfs_super_block *)raw_disk_sb;
410 u16 csum_type = btrfs_super_csum_type(disk_sb);
411 int ret = 0;
412
413 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
414 u32 crc = ~(u32)0;
415 const int csum_size = sizeof(crc);
416 char result[csum_size];
417
418 /*
419 * The super_block structure does not span the whole
420 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
421 * is filled with zeros and is included in the checkum.
422 */
423 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
424 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
425 btrfs_csum_final(crc, result);
426
427 if (memcmp(raw_disk_sb, result, csum_size))
428 ret = 1;
429 }
430
431 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500432 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100433 csum_type);
434 ret = 1;
435 }
436
437 return ret;
438}
439
440/*
Chris Masond352ac62008-09-29 15:18:18 -0400441 * helper to read a given tree block, doing retries as required when
442 * the checksums don't match and we have alternate mirrors to try.
443 */
Chris Masonf1885912008-04-09 16:28:12 -0400444static int btree_read_extent_buffer_pages(struct btrfs_root *root,
445 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400446 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400447{
448 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400449 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400450 int ret;
451 int num_copies = 0;
452 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400453 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400454
Josef Bacika826d6d2011-03-16 13:42:43 -0400455 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400456 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
457 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200458 ret = read_extent_buffer_pages(io_tree, eb, start,
459 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400460 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600461 if (!ret) {
462 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400463 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600464 break;
465 else
466 ret = -EIO;
467 }
Chris Masond3977122009-01-05 21:25:51 -0500468
Josef Bacika826d6d2011-03-16 13:42:43 -0400469 /*
470 * This buffer's crc is fine, but its contents are corrupted, so
471 * there is no reason to read the other copies, they won't be
472 * any less wrong.
473 */
474 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400475 break;
476
Stefan Behrens5d964052012-11-05 14:59:07 +0100477 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400478 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400479 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400480 break;
Chris Mason42352982008-04-28 16:40:52 -0400481
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400482 if (!failed_mirror) {
483 failed = 1;
484 failed_mirror = eb->read_mirror;
485 }
486
Chris Masonf1885912008-04-09 16:28:12 -0400487 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400488 if (mirror_num == failed_mirror)
489 mirror_num++;
490
Chris Mason42352982008-04-28 16:40:52 -0400491 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400492 break;
Chris Masonf1885912008-04-09 16:28:12 -0400493 }
Josef Bacikea466792012-03-26 21:57:36 -0400494
Stefan Behrensc0901582012-07-10 07:30:17 -0600495 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400496 repair_eb_io_failure(root, eb, failed_mirror);
497
498 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400499}
Chris Mason19c00dd2007-10-15 16:19:22 -0400500
Chris Masond352ac62008-09-29 15:18:18 -0400501/*
Chris Masond3977122009-01-05 21:25:51 -0500502 * checksum a dirty tree block before IO. This has extra checks to make sure
503 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400504 */
Chris Masond3977122009-01-05 21:25:51 -0500505
Daniel Dressler01d58472014-11-21 17:15:07 +0900506static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400507{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000508 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400509 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400510 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400511
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500512 eb = (struct extent_buffer *)page->private;
513 if (page != eb->pages[0])
514 return 0;
Alex Lyakas0f805532016-03-10 13:10:15 +0200515
Chris Mason19c00dd2007-10-15 16:19:22 -0400516 found_start = btrfs_header_bytenr(eb);
Alex Lyakas0f805532016-03-10 13:10:15 +0200517 /*
518 * Please do not consolidate these warnings into a single if.
519 * It is useful to know what went wrong.
520 */
521 if (WARN_ON(found_start != start))
522 return -EUCLEAN;
523 if (WARN_ON(!PageUptodate(page)))
524 return -EUCLEAN;
525
526 ASSERT(memcmp_extent_buffer(eb, fs_info->fsid,
527 btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0);
528
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200529 return csum_tree_block(fs_info, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400530}
531
Daniel Dressler01d58472014-11-21 17:15:07 +0900532static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
Yan Zheng2b820322008-11-17 21:11:30 -0500533 struct extent_buffer *eb)
534{
Daniel Dressler01d58472014-11-21 17:15:07 +0900535 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
Yan Zheng2b820322008-11-17 21:11:30 -0500536 u8 fsid[BTRFS_UUID_SIZE];
537 int ret = 1;
538
Ross Kirk0a4e5582013-09-24 10:12:38 +0100539 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500540 while (fs_devices) {
541 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
542 ret = 0;
543 break;
544 }
545 fs_devices = fs_devices->seed;
546 }
547 return ret;
548}
549
Josef Bacika826d6d2011-03-16 13:42:43 -0400550#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500551 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
552 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200553 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400554
555static noinline int check_leaf(struct btrfs_root *root,
556 struct extent_buffer *leaf)
557{
558 struct btrfs_key key;
559 struct btrfs_key leaf_key;
560 u32 nritems = btrfs_header_nritems(leaf);
561 int slot;
562
563 if (nritems == 0)
564 return 0;
565
566 /* Check the 0 item */
567 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
568 BTRFS_LEAF_DATA_SIZE(root)) {
569 CORRUPT("invalid item offset size pair", leaf, root, 0);
570 return -EIO;
571 }
572
573 /*
574 * Check to make sure each items keys are in the correct order and their
575 * offsets make sense. We only have to loop through nritems-1 because
576 * we check the current slot against the next slot, which verifies the
577 * next slot's offset+size makes sense and that the current's slot
578 * offset is correct.
579 */
580 for (slot = 0; slot < nritems - 1; slot++) {
581 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
582 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
583
584 /* Make sure the keys are in the right order */
585 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
586 CORRUPT("bad key order", leaf, root, slot);
587 return -EIO;
588 }
589
590 /*
591 * Make sure the offset and ends are right, remember that the
592 * item data starts at the end of the leaf and grows towards the
593 * front.
594 */
595 if (btrfs_item_offset_nr(leaf, slot) !=
596 btrfs_item_end_nr(leaf, slot + 1)) {
597 CORRUPT("slot offset bad", leaf, root, slot);
598 return -EIO;
599 }
600
601 /*
602 * Check to make sure that we don't point outside of the leaf,
603 * just incase all the items are consistent to eachother, but
604 * all point outside of the leaf.
605 */
606 if (btrfs_item_end_nr(leaf, slot) >
607 BTRFS_LEAF_DATA_SIZE(root)) {
608 CORRUPT("slot end outside of leaf", leaf, root, slot);
609 return -EIO;
610 }
611 }
612
613 return 0;
614}
615
Miao Xiefacc8a222013-07-25 19:22:34 +0800616static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
617 u64 phy_offset, struct page *page,
618 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400619{
Chris Masonce9adaa2008-04-09 16:28:12 -0400620 u64 found_start;
621 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400622 struct extent_buffer *eb;
623 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Zhao Lei02873e42015-12-31 22:46:45 +0800624 struct btrfs_fs_info *fs_info = root->fs_info;
Chris Masonf1885912008-04-09 16:28:12 -0400625 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400626 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400627
Chris Masonce9adaa2008-04-09 16:28:12 -0400628 if (!page->private)
629 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500630
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500631 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500632
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400633 /* the pending IO might have been the only thing that kept this buffer
634 * in memory. Make sure we have a ref for all this other checks
635 */
636 extent_buffer_get(eb);
637
638 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400639 if (!reads_done)
640 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400641
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400642 eb->read_mirror = mirror;
Filipe Manana656f30d2014-09-26 12:25:56 +0100643 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
Josef Bacikea466792012-03-26 21:57:36 -0400644 ret = -EIO;
645 goto err;
646 }
647
Chris Masonce9adaa2008-04-09 16:28:12 -0400648 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400649 if (found_start != eb->start) {
Zhao Lei02873e42015-12-31 22:46:45 +0800650 btrfs_err_rl(fs_info, "bad tree block start %llu %llu",
651 found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400652 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400653 goto err;
654 }
Zhao Lei02873e42015-12-31 22:46:45 +0800655 if (check_tree_block_fsid(fs_info, eb)) {
656 btrfs_err_rl(fs_info, "bad fsid on block %llu",
657 eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400658 ret = -EIO;
659 goto err;
660 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400661 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400662 if (found_level >= BTRFS_MAX_LEVEL) {
Zhao Lei02873e42015-12-31 22:46:45 +0800663 btrfs_err(fs_info, "bad tree block level %d",
664 (int)btrfs_header_level(eb));
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400665 ret = -EIO;
666 goto err;
667 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400668
Chris Mason85d4e462011-07-26 16:11:19 -0400669 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
670 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500671
Zhao Lei02873e42015-12-31 22:46:45 +0800672 ret = csum_tree_block(fs_info, eb, 1);
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200673 if (ret)
Josef Bacika826d6d2011-03-16 13:42:43 -0400674 goto err;
Josef Bacika826d6d2011-03-16 13:42:43 -0400675
676 /*
677 * If this is a leaf block and it is corrupt, set the corrupt bit so
678 * that we don't try and read the other copies of this block, just
679 * return -EIO.
680 */
681 if (found_level == 0 && check_leaf(root, eb)) {
682 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
683 ret = -EIO;
684 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400685
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400686 if (!ret)
687 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400688err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400689 if (reads_done &&
690 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Zhao Lei02873e42015-12-31 22:46:45 +0800691 btree_readahead_hook(fs_info, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200692
David Woodhouse53b381b2013-01-29 18:40:14 -0500693 if (ret) {
694 /*
695 * our io error hook is going to dec the io pages
696 * again, we have to make sure it has something
697 * to decrement
698 */
699 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400700 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500701 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400702 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400703out:
Chris Masonf1885912008-04-09 16:28:12 -0400704 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400705}
706
Josef Bacikea466792012-03-26 21:57:36 -0400707static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200708{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200709 struct extent_buffer *eb;
Arne Jansen4bb31e92011-06-10 13:55:54 +0200710
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500711 eb = (struct extent_buffer *)page->private;
Filipe Manana656f30d2014-09-26 12:25:56 +0100712 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400713 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500714 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400715 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Zhao Lei02873e42015-12-31 22:46:45 +0800716 btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200717 return -EIO; /* we fixed nothing */
718}
719
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200720static void end_workqueue_bio(struct bio *bio)
Chris Masonce9adaa2008-04-09 16:28:12 -0400721{
David Sterba97eb6b62014-07-30 00:55:42 +0200722 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400723 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800724 struct btrfs_workqueue *wq;
725 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400726
Chris Masonce9adaa2008-04-09 16:28:12 -0400727 fs_info = end_io_wq->info;
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200728 end_io_wq->error = bio->bi_error;
Chris Masond20f7042008-12-08 16:58:54 -0500729
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200730 if (bio->bi_rw & REQ_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800731 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
732 wq = fs_info->endio_meta_write_workers;
733 func = btrfs_endio_meta_write_helper;
734 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
735 wq = fs_info->endio_freespace_worker;
736 func = btrfs_freespace_write_helper;
737 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
738 wq = fs_info->endio_raid56_workers;
739 func = btrfs_endio_raid56_helper;
740 } else {
741 wq = fs_info->endio_write_workers;
742 func = btrfs_endio_write_helper;
743 }
Chris Masond20f7042008-12-08 16:58:54 -0500744 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800745 if (unlikely(end_io_wq->metadata ==
746 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
747 wq = fs_info->endio_repair_workers;
748 func = btrfs_endio_repair_helper;
749 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800750 wq = fs_info->endio_raid56_workers;
751 func = btrfs_endio_raid56_helper;
752 } else if (end_io_wq->metadata) {
753 wq = fs_info->endio_meta_workers;
754 func = btrfs_endio_meta_helper;
755 } else {
756 wq = fs_info->endio_workers;
757 func = btrfs_endio_helper;
758 }
Chris Masond20f7042008-12-08 16:58:54 -0500759 }
Liu Bo9e0af232014-08-15 23:36:53 +0800760
761 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
762 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400763}
764
Chris Mason22c59942008-04-09 16:28:12 -0400765int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200766 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400767{
David Sterba97eb6b62014-07-30 00:55:42 +0200768 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800769
David Sterba97eb6b62014-07-30 00:55:42 +0200770 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400771 if (!end_io_wq)
772 return -ENOMEM;
773
774 end_io_wq->private = bio->bi_private;
775 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400776 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400777 end_io_wq->error = 0;
778 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400779 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400780
781 bio->bi_private = end_io_wq;
782 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400783 return 0;
784}
785
Chris Masonb64a2852008-08-20 13:39:41 -0400786unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400787{
788 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800789 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400790 info->fs_devices->open_devices);
791 return 256 * limit;
792}
793
Chris Mason4a69a412008-11-06 22:03:00 -0500794static void run_one_async_start(struct btrfs_work *work)
795{
Chris Mason4a69a412008-11-06 22:03:00 -0500796 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100797 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500798
799 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100800 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
801 async->mirror_num, async->bio_flags,
802 async->bio_offset);
803 if (ret)
804 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500805}
806
807static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400808{
809 struct btrfs_fs_info *fs_info;
810 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400811 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400812
813 async = container_of(work, struct async_submit_bio, work);
814 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400815
Chris Masonb64a2852008-08-20 13:39:41 -0400816 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400817 limit = limit * 2 / 3;
818
David Sterbaee863952015-02-16 19:41:40 +0100819 /*
820 * atomic_dec_return implies a barrier for waitqueue_active
821 */
Josef Bacik66657b32012-08-01 15:36:24 -0400822 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400823 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400824 wake_up(&fs_info->async_submit_wait);
825
Adam Buchbinderbb7ab3b2016-03-04 11:23:12 -0800826 /* If an error occurred we just want to clean up the bio and move on */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100827 if (async->error) {
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200828 async->bio->bi_error = async->error;
829 bio_endio(async->bio);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100830 return;
831 }
832
Chris Mason4a69a412008-11-06 22:03:00 -0500833 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400834 async->mirror_num, async->bio_flags,
835 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500836}
837
838static void run_one_async_free(struct btrfs_work *work)
839{
840 struct async_submit_bio *async;
841
842 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400843 kfree(async);
844}
845
Chris Mason44b8bd72008-04-16 11:14:51 -0400846int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
847 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400848 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400849 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500850 extent_submit_bio_hook_t *submit_bio_start,
851 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400852{
853 struct async_submit_bio *async;
854
855 async = kmalloc(sizeof(*async), GFP_NOFS);
856 if (!async)
857 return -ENOMEM;
858
859 async->inode = inode;
860 async->rw = rw;
861 async->bio = bio;
862 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500863 async->submit_bio_start = submit_bio_start;
864 async->submit_bio_done = submit_bio_done;
865
Liu Bo9e0af232014-08-15 23:36:53 +0800866 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800867 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500868
Chris Masonc8b97812008-10-29 14:49:59 -0400869 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400870 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400871
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100872 async->error = 0;
873
Chris Masoncb03c742008-05-15 16:15:45 -0400874 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400875
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200876 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800877 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400878
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800879 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400880
Chris Masond3977122009-01-05 21:25:51 -0500881 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500882 atomic_read(&fs_info->nr_async_submits)) {
883 wait_event(fs_info->async_submit_wait,
884 (atomic_read(&fs_info->nr_async_submits) == 0));
885 }
886
Chris Mason44b8bd72008-04-16 11:14:51 -0400887 return 0;
888}
889
Chris Masonce3ed712008-09-23 13:14:12 -0400890static int btree_csum_one_bio(struct bio *bio)
891{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800892 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400893 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800894 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400895
Kent Overstreet2c30c712013-11-07 12:20:26 -0800896 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400897 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Daniel Dressler01d58472014-11-21 17:15:07 +0900898 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100899 if (ret)
900 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400901 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800902
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100903 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400904}
905
Chris Mason4a69a412008-11-06 22:03:00 -0500906static int __btree_submit_bio_start(struct inode *inode, int rw,
907 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400908 unsigned long bio_flags,
909 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400910{
Chris Mason8b712842008-06-11 16:50:36 -0400911 /*
912 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400913 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400914 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100915 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500916}
Chris Mason22c59942008-04-09 16:28:12 -0400917
Chris Mason4a69a412008-11-06 22:03:00 -0500918static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400919 int mirror_num, unsigned long bio_flags,
920 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500921{
Stefan Behrens61891922012-11-05 18:51:52 +0100922 int ret;
923
Chris Mason8b712842008-06-11 16:50:36 -0400924 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500925 * when we're called for a write, we're already in the async
926 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400927 */
Stefan Behrens61891922012-11-05 18:51:52 +0100928 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200929 if (ret) {
930 bio->bi_error = ret;
931 bio_endio(bio);
932 }
Stefan Behrens61891922012-11-05 18:51:52 +0100933 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400934}
935
Josef Bacikde0022b2012-09-25 14:25:58 -0400936static int check_async_write(struct inode *inode, unsigned long bio_flags)
937{
938 if (bio_flags & EXTENT_BIO_TREE_LOG)
939 return 0;
940#ifdef CONFIG_X86
Borislav Petkovbc696ca2016-01-26 22:12:05 +0100941 if (static_cpu_has(X86_FEATURE_XMM4_2))
Josef Bacikde0022b2012-09-25 14:25:58 -0400942 return 0;
943#endif
944 return 1;
945}
946
Chris Mason44b8bd72008-04-16 11:14:51 -0400947static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400948 int mirror_num, unsigned long bio_flags,
949 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400950{
Josef Bacikde0022b2012-09-25 14:25:58 -0400951 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500952 int ret;
953
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200954 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500955 /*
956 * called for a read, do the setup so that checksum validation
957 * can happen in the async kernel threads
958 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400959 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +0200960 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -0400961 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100962 goto out_w_error;
963 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
964 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400965 } else if (!async) {
966 ret = btree_csum_one_bio(bio);
967 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100968 goto out_w_error;
969 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
970 mirror_num, 0);
971 } else {
972 /*
973 * kthread helpers are used to submit writes so that
974 * checksumming can happen in parallel across all CPUs
975 */
976 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
977 inode, rw, bio, mirror_num, 0,
978 bio_offset,
979 __btree_submit_bio_start,
980 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400981 }
Chris Masond313d7a2009-04-20 15:50:09 -0400982
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200983 if (ret)
984 goto out_w_error;
985 return 0;
986
Stefan Behrens61891922012-11-05 18:51:52 +0100987out_w_error:
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200988 bio->bi_error = ret;
989 bio_endio(bio);
Stefan Behrens61891922012-11-05 18:51:52 +0100990 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400991}
992
Jan Beulich3dd14622010-12-07 14:54:09 +0000993#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500994static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800995 struct page *newpage, struct page *page,
996 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500997{
998 /*
999 * we can't safely write a btree page from here,
1000 * we haven't done the locking hook
1001 */
1002 if (PageDirty(page))
1003 return -EAGAIN;
1004 /*
1005 * Buffers may be managed in a filesystem specific way.
1006 * We must have no buffers or drop them.
1007 */
1008 if (page_has_private(page) &&
1009 !try_to_release_page(page, GFP_KERNEL))
1010 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -08001011 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -05001012}
Jan Beulich3dd14622010-12-07 14:54:09 +00001013#endif
Chris Mason784b4e22010-11-21 22:20:49 -05001014
Chris Mason0da54682007-11-07 21:08:01 -05001015
1016static int btree_writepages(struct address_space *mapping,
1017 struct writeback_control *wbc)
1018{
Miao Xiee2d84522013-01-29 10:09:20 +00001019 struct btrfs_fs_info *fs_info;
1020 int ret;
1021
Chris Masond8d5f3e2007-12-11 12:42:00 -05001022 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001023
1024 if (wbc->for_kupdate)
1025 return 0;
1026
Miao Xiee2d84522013-01-29 10:09:20 +00001027 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001028 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001029 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1030 BTRFS_DIRTY_METADATA_THRESH);
1031 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001032 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001033 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001034 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001035}
1036
Christoph Hellwigb2950862008-12-02 09:54:17 -05001037static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001038{
Chris Masond1310b22008-01-24 16:13:08 -05001039 struct extent_io_tree *tree;
1040 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001041 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001042}
1043
Chris Mason70dec802008-01-29 09:59:12 -05001044static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001045{
Chris Mason98509cf2008-09-11 15:51:43 -04001046 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001047 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001048
David Sterbaf7a52a42013-04-26 14:56:29 +00001049 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001050}
Chris Mason123abc82007-03-14 14:14:43 -04001051
Lukas Czernerd47992f2013-05-21 23:17:23 -04001052static void btree_invalidatepage(struct page *page, unsigned int offset,
1053 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001054{
Chris Masond1310b22008-01-24 16:13:08 -05001055 struct extent_io_tree *tree;
1056 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001057 extent_invalidatepage(tree, page, offset);
1058 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001059 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001060 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1061 "page private not zero on page %llu",
1062 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001063 ClearPagePrivate(page);
1064 set_page_private(page, 0);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001065 put_page(page);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001066 }
Chris Masond98237b2007-03-28 13:57:48 -04001067}
1068
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001069static int btree_set_page_dirty(struct page *page)
1070{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001071#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001072 struct extent_buffer *eb;
1073
1074 BUG_ON(!PagePrivate(page));
1075 eb = (struct extent_buffer *)page->private;
1076 BUG_ON(!eb);
1077 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1078 BUG_ON(!atomic_read(&eb->refs));
1079 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001080#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001081 return __set_page_dirty_nobuffers(page);
1082}
1083
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001084static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001085 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001086 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001087 .releasepage = btree_releasepage,
1088 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001089#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001090 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001091#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001092 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001093};
1094
David Sterbad3e46fe2014-06-15 02:04:19 +02001095void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
Chris Mason090d1872007-05-01 08:53:32 -04001096{
Chris Mason5f39d392007-10-15 16:14:19 -04001097 struct extent_buffer *buf = NULL;
1098 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason090d1872007-05-01 08:53:32 -04001099
David Sterbaa83fffb2014-06-15 02:39:54 +02001100 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001101 if (!buf)
David Sterba6197d862014-06-15 00:49:36 +02001102 return;
Chris Masond1310b22008-01-24 16:13:08 -05001103 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001104 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001105 free_extent_buffer(buf);
Chris Mason090d1872007-05-01 08:53:32 -04001106}
1107
David Sterbac0dcaa42014-06-15 02:24:21 +02001108int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
Arne Jansenab0fff02011-05-23 14:25:41 +02001109 int mirror_num, struct extent_buffer **eb)
1110{
1111 struct extent_buffer *buf = NULL;
1112 struct inode *btree_inode = root->fs_info->btree_inode;
1113 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1114 int ret;
1115
David Sterbaa83fffb2014-06-15 02:39:54 +02001116 buf = btrfs_find_create_tree_block(root, bytenr);
Arne Jansenab0fff02011-05-23 14:25:41 +02001117 if (!buf)
1118 return 0;
1119
1120 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1121
1122 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1123 btree_get_extent, mirror_num);
1124 if (ret) {
1125 free_extent_buffer(buf);
1126 return ret;
1127 }
1128
1129 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1130 free_extent_buffer(buf);
1131 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001132 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001133 *eb = buf;
1134 } else {
1135 free_extent_buffer(buf);
1136 }
1137 return 0;
1138}
1139
Daniel Dressler01d58472014-11-21 17:15:07 +09001140struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
David Sterba0308af42014-06-15 01:43:40 +02001141 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001142{
Daniel Dressler01d58472014-11-21 17:15:07 +09001143 return find_extent_buffer(fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001144}
1145
1146struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
David Sterbaa83fffb2014-06-15 02:39:54 +02001147 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001148{
David Sterbafccb84c2014-09-29 23:53:21 +02001149 if (btrfs_test_is_dummy_root(root))
David Sterbace3e6982014-06-15 03:00:04 +02001150 return alloc_test_extent_buffer(root->fs_info, bytenr);
1151 return alloc_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001152}
1153
1154
Chris Masone02119d2008-09-05 16:13:11 -04001155int btrfs_write_tree_block(struct extent_buffer *buf)
1156{
Chris Mason727011e2010-08-06 13:21:20 -04001157 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001158 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001159}
1160
1161int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1162{
Chris Mason727011e2010-08-06 13:21:20 -04001163 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001164 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001165}
1166
Chris Masondb945352007-10-15 16:15:53 -04001167struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
David Sterbace86cd52014-06-15 01:07:32 +02001168 u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001169{
Chris Mason5f39d392007-10-15 16:14:19 -04001170 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001171 int ret;
1172
David Sterbaa83fffb2014-06-15 02:39:54 +02001173 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001174 if (!buf)
Liu Bo64c043d2015-05-25 17:30:15 +08001175 return ERR_PTR(-ENOMEM);
Chris Masone4204de2008-01-08 15:46:27 -05001176
Chris Masonca7a79a2008-05-12 12:59:19 -04001177 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001178 if (ret) {
1179 free_extent_buffer(buf);
Liu Bo64c043d2015-05-25 17:30:15 +08001180 return ERR_PTR(ret);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001181 }
Chris Mason5f39d392007-10-15 16:14:19 -04001182 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001183
Chris Masoneb60cea2007-02-02 09:18:22 -05001184}
1185
Daniel Dressler01d58472014-11-21 17:15:07 +09001186void clean_tree_block(struct btrfs_trans_handle *trans,
1187 struct btrfs_fs_info *fs_info,
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001188 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001189{
Chris Mason55c69072008-01-09 15:55:33 -05001190 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001191 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001192 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001193
Chris Masonb9473432009-03-13 11:00:37 -04001194 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001195 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1196 -buf->len,
1197 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001198 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1199 btrfs_set_lock_blocking(buf);
1200 clear_extent_buffer_dirty(buf);
1201 }
Chris Mason925baed2008-06-25 16:01:30 -04001202 }
Chris Mason5f39d392007-10-15 16:14:19 -04001203}
1204
Miao Xie8257b2d2014-03-06 13:38:19 +08001205static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1206{
1207 struct btrfs_subvolume_writers *writers;
1208 int ret;
1209
1210 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1211 if (!writers)
1212 return ERR_PTR(-ENOMEM);
1213
Tejun Heo908c7f12014-09-08 09:51:29 +09001214 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
Miao Xie8257b2d2014-03-06 13:38:19 +08001215 if (ret < 0) {
1216 kfree(writers);
1217 return ERR_PTR(ret);
1218 }
1219
1220 init_waitqueue_head(&writers->wait);
1221 return writers;
1222}
1223
1224static void
1225btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1226{
1227 percpu_counter_destroy(&writers->counter);
1228 kfree(writers);
1229}
1230
David Sterba707e8a02014-06-04 19:22:26 +02001231static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1232 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
Jeff Mahoney143bede2012-03-01 14:56:26 +01001233 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001234{
Chris Masoncfaa7292007-02-21 17:04:57 -05001235 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001236 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001237 root->sectorsize = sectorsize;
1238 root->nodesize = nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001239 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001240 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001241 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001242
Chris Mason0f7d52f2007-04-09 10:42:37 -04001243 root->objectid = objectid;
1244 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001245 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001246 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001247 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001248 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001249 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001250 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001251 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001252 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001253
1254 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001255 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001256 INIT_LIST_HEAD(&root->delalloc_inodes);
1257 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001258 INIT_LIST_HEAD(&root->ordered_extents);
1259 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001260 INIT_LIST_HEAD(&root->logged_list[0]);
1261 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001262 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001263 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001264 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001265 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001266 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001267 spin_lock_init(&root->log_extents_lock[0]);
1268 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001269 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001270 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001271 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001272 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001273 init_waitqueue_head(&root->log_writer_wait);
1274 init_waitqueue_head(&root->log_commit_wait[0]);
1275 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001276 INIT_LIST_HEAD(&root->log_ctxs[0]);
1277 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f182009-01-21 12:54:03 -05001278 atomic_set(&root->log_commit[0], 0);
1279 atomic_set(&root->log_commit[1], 0);
1280 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001281 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001282 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001283 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001284 atomic_set(&root->will_be_snapshoted, 0);
Qu Wenruo55eeaf02015-09-08 17:08:38 +08001285 atomic_set(&root->qgroup_meta_rsv, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001286 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001287 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001288 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001289 if (fs_info)
1290 extent_io_tree_init(&root->dirty_log_pages,
1291 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001292
Chris Mason3768f362007-03-13 16:47:54 -04001293 memset(&root->root_key, 0, sizeof(root->root_key));
1294 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001295 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001296 if (fs_info)
1297 root->defrag_trans_start = fs_info->generation;
1298 else
1299 root->defrag_trans_start = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001300 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001301 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001302
Anand Jain5f3ab902012-12-07 09:28:54 +00001303 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001304}
1305
David Sterba74e4d822016-02-11 11:01:55 +01001306static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info,
1307 gfp_t flags)
Al Viro6f07e422011-11-17 00:46:16 -05001308{
David Sterba74e4d822016-02-11 11:01:55 +01001309 struct btrfs_root *root = kzalloc(sizeof(*root), flags);
Al Viro6f07e422011-11-17 00:46:16 -05001310 if (root)
1311 root->fs_info = fs_info;
1312 return root;
1313}
1314
Josef Bacik06ea65a2013-09-19 16:07:01 -04001315#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1316/* Should only be used by the testing infrastructure */
1317struct btrfs_root *btrfs_alloc_dummy_root(void)
1318{
1319 struct btrfs_root *root;
1320
David Sterba74e4d822016-02-11 11:01:55 +01001321 root = btrfs_alloc_root(NULL, GFP_KERNEL);
Josef Bacik06ea65a2013-09-19 16:07:01 -04001322 if (!root)
1323 return ERR_PTR(-ENOMEM);
David Sterba707e8a02014-06-04 19:22:26 +02001324 __setup_root(4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001325 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001326 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001327
1328 return root;
1329}
1330#endif
1331
Arne Jansen20897f52011-09-13 12:44:20 +02001332struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1333 struct btrfs_fs_info *fs_info,
1334 u64 objectid)
1335{
1336 struct extent_buffer *leaf;
1337 struct btrfs_root *tree_root = fs_info->tree_root;
1338 struct btrfs_root *root;
1339 struct btrfs_key key;
1340 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001341 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001342
David Sterba74e4d822016-02-11 11:01:55 +01001343 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Arne Jansen20897f52011-09-13 12:44:20 +02001344 if (!root)
1345 return ERR_PTR(-ENOMEM);
1346
David Sterba707e8a02014-06-04 19:22:26 +02001347 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1348 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001349 root->root_key.objectid = objectid;
1350 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1351 root->root_key.offset = 0;
1352
David Sterba4d75f8a2014-06-15 01:54:12 +02001353 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Arne Jansen20897f52011-09-13 12:44:20 +02001354 if (IS_ERR(leaf)) {
1355 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001356 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001357 goto fail;
1358 }
1359
Arne Jansen20897f52011-09-13 12:44:20 +02001360 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1361 btrfs_set_header_bytenr(leaf, leaf->start);
1362 btrfs_set_header_generation(leaf, trans->transid);
1363 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1364 btrfs_set_header_owner(leaf, objectid);
1365 root->node = leaf;
1366
Ross Kirk0a4e5582013-09-24 10:12:38 +01001367 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001368 BTRFS_FSID_SIZE);
1369 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001370 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001371 BTRFS_UUID_SIZE);
1372 btrfs_mark_buffer_dirty(leaf);
1373
1374 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001375 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001376
1377 root->root_item.flags = 0;
1378 root->root_item.byte_limit = 0;
1379 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1380 btrfs_set_root_generation(&root->root_item, trans->transid);
1381 btrfs_set_root_level(&root->root_item, 0);
1382 btrfs_set_root_refs(&root->root_item, 1);
1383 btrfs_set_root_used(&root->root_item, leaf->len);
1384 btrfs_set_root_last_snapshot(&root->root_item, 0);
1385 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001386 uuid_le_gen(&uuid);
1387 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001388 root->root_item.drop_level = 0;
1389
1390 key.objectid = objectid;
1391 key.type = BTRFS_ROOT_ITEM_KEY;
1392 key.offset = 0;
1393 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1394 if (ret)
1395 goto fail;
1396
1397 btrfs_tree_unlock(leaf);
1398
Arne Jansen20897f52011-09-13 12:44:20 +02001399 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001400
1401fail:
1402 if (leaf) {
1403 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001404 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001405 free_extent_buffer(leaf);
1406 }
1407 kfree(root);
1408
1409 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001410}
1411
Yan Zheng7237f182009-01-21 12:54:03 -05001412static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1413 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001414{
1415 struct btrfs_root *root;
1416 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001417 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001418
David Sterba74e4d822016-02-11 11:01:55 +01001419 root = btrfs_alloc_root(fs_info, GFP_NOFS);
Chris Masone02119d2008-09-05 16:13:11 -04001420 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001421 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001422
David Sterba707e8a02014-06-04 19:22:26 +02001423 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1424 tree_root->stripesize, root, fs_info,
1425 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001426
1427 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1428 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1429 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001430
Yan Zheng7237f182009-01-21 12:54:03 -05001431 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001432 * DON'T set REF_COWS for log trees
1433 *
Yan Zheng7237f182009-01-21 12:54:03 -05001434 * log trees do not get reference counted because they go away
1435 * before a real commit is actually done. They do store pointers
1436 * to file data extents, and those reference counts still get
1437 * updated (along with back refs to the log tree).
1438 */
Chris Masone02119d2008-09-05 16:13:11 -04001439
David Sterba4d75f8a2014-06-15 01:54:12 +02001440 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1441 NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001442 if (IS_ERR(leaf)) {
1443 kfree(root);
1444 return ERR_CAST(leaf);
1445 }
Chris Masone02119d2008-09-05 16:13:11 -04001446
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001447 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1448 btrfs_set_header_bytenr(leaf, leaf->start);
1449 btrfs_set_header_generation(leaf, trans->transid);
1450 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1451 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001452 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001453
1454 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001455 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001456 btrfs_mark_buffer_dirty(root->node);
1457 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001458 return root;
1459}
1460
1461int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1462 struct btrfs_fs_info *fs_info)
1463{
1464 struct btrfs_root *log_root;
1465
1466 log_root = alloc_log_tree(trans, fs_info);
1467 if (IS_ERR(log_root))
1468 return PTR_ERR(log_root);
1469 WARN_ON(fs_info->log_root_tree);
1470 fs_info->log_root_tree = log_root;
1471 return 0;
1472}
1473
1474int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1475 struct btrfs_root *root)
1476{
1477 struct btrfs_root *log_root;
1478 struct btrfs_inode_item *inode_item;
1479
1480 log_root = alloc_log_tree(trans, root->fs_info);
1481 if (IS_ERR(log_root))
1482 return PTR_ERR(log_root);
1483
1484 log_root->last_trans = trans->transid;
1485 log_root->root_key.offset = root->root_key.objectid;
1486
1487 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001488 btrfs_set_stack_inode_generation(inode_item, 1);
1489 btrfs_set_stack_inode_size(inode_item, 3);
1490 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001491 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001492 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f182009-01-21 12:54:03 -05001493
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001494 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001495
1496 WARN_ON(root->log_root);
1497 root->log_root = log_root;
1498 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001499 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001500 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001501 return 0;
1502}
1503
Stefan Behrens35a36212013-08-14 18:12:25 +02001504static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1505 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001506{
1507 struct btrfs_root *root;
1508 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001509 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001510 u64 generation;
Miao Xiecb517ea2013-05-15 07:48:19 +00001511 int ret;
1512
1513 path = btrfs_alloc_path();
1514 if (!path)
1515 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001516
David Sterba74e4d822016-02-11 11:01:55 +01001517 root = btrfs_alloc_root(fs_info, GFP_NOFS);
Miao Xiecb517ea2013-05-15 07:48:19 +00001518 if (!root) {
1519 ret = -ENOMEM;
1520 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001521 }
1522
David Sterba707e8a02014-06-04 19:22:26 +02001523 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1524 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001525
Miao Xiecb517ea2013-05-15 07:48:19 +00001526 ret = btrfs_find_root(tree_root, key, path,
1527 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001528 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001529 if (ret > 0)
1530 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001531 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001532 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001533
Yan Zheng84234f32008-10-29 14:49:05 -04001534 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001535 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
David Sterbace86cd52014-06-15 01:07:32 +02001536 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08001537 if (IS_ERR(root->node)) {
1538 ret = PTR_ERR(root->node);
Miao Xiecb517ea2013-05-15 07:48:19 +00001539 goto find_fail;
1540 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1541 ret = -EIO;
Liu Bo64c043d2015-05-25 17:30:15 +08001542 free_extent_buffer(root->node);
1543 goto find_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001544 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001545 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001546out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001547 btrfs_free_path(path);
1548 return root;
1549
Miao Xiecb517ea2013-05-15 07:48:19 +00001550find_fail:
1551 kfree(root);
1552alloc_fail:
1553 root = ERR_PTR(ret);
1554 goto out;
1555}
1556
1557struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1558 struct btrfs_key *location)
1559{
1560 struct btrfs_root *root;
1561
1562 root = btrfs_read_tree_root(tree_root, location);
1563 if (IS_ERR(root))
1564 return root;
1565
1566 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001567 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001568 btrfs_check_and_init_root_item(&root->root_item);
1569 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001570
Chris Mason5eda7b52007-06-22 14:16:25 -04001571 return root;
1572}
1573
Miao Xiecb517ea2013-05-15 07:48:19 +00001574int btrfs_init_fs_root(struct btrfs_root *root)
1575{
1576 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001577 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001578
1579 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1580 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1581 GFP_NOFS);
1582 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1583 ret = -ENOMEM;
1584 goto fail;
1585 }
1586
Miao Xie8257b2d2014-03-06 13:38:19 +08001587 writers = btrfs_alloc_subvolume_writers();
1588 if (IS_ERR(writers)) {
1589 ret = PTR_ERR(writers);
1590 goto fail;
1591 }
1592 root->subv_writers = writers;
1593
Miao Xiecb517ea2013-05-15 07:48:19 +00001594 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001595 spin_lock_init(&root->ino_cache_lock);
1596 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001597
1598 ret = get_anon_bdev(&root->anon_dev);
1599 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001600 goto free_writers;
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301601
1602 mutex_lock(&root->objectid_mutex);
1603 ret = btrfs_find_highest_objectid(root,
1604 &root->highest_objectid);
1605 if (ret) {
1606 mutex_unlock(&root->objectid_mutex);
1607 goto free_root_dev;
1608 }
1609
1610 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1611
1612 mutex_unlock(&root->objectid_mutex);
1613
Miao Xiecb517ea2013-05-15 07:48:19 +00001614 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001615
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301616free_root_dev:
1617 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08001618free_writers:
1619 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001620fail:
1621 kfree(root->free_ino_ctl);
1622 kfree(root->free_ino_pinned);
1623 return ret;
1624}
1625
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001626static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1627 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001628{
1629 struct btrfs_root *root;
1630
1631 spin_lock(&fs_info->fs_roots_radix_lock);
1632 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1633 (unsigned long)root_id);
1634 spin_unlock(&fs_info->fs_roots_radix_lock);
1635 return root;
1636}
1637
1638int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1639 struct btrfs_root *root)
1640{
1641 int ret;
1642
1643 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1644 if (ret)
1645 return ret;
1646
1647 spin_lock(&fs_info->fs_roots_radix_lock);
1648 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1649 (unsigned long)root->root_key.objectid,
1650 root);
1651 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001652 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001653 spin_unlock(&fs_info->fs_roots_radix_lock);
1654 radix_tree_preload_end();
1655
1656 return ret;
1657}
1658
Miao Xiec00869f2013-09-25 21:47:44 +08001659struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1660 struct btrfs_key *location,
1661 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001662{
1663 struct btrfs_root *root;
David Sterba381cf652015-01-02 18:45:16 +01001664 struct btrfs_path *path;
David Sterba1d4c08e2015-01-02 19:36:14 +01001665 struct btrfs_key key;
Chris Mason5eda7b52007-06-22 14:16:25 -04001666 int ret;
1667
Chris Masonedbd8d42007-12-21 16:27:24 -05001668 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1669 return fs_info->tree_root;
1670 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1671 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001672 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1673 return fs_info->chunk_root;
1674 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1675 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001676 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1677 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001678 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1679 return fs_info->quota_root ? fs_info->quota_root :
1680 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001681 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1682 return fs_info->uuid_root ? fs_info->uuid_root :
1683 ERR_PTR(-ENOENT);
Omar Sandoval70f6d822015-09-29 20:50:38 -07001684 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1685 return fs_info->free_space_root ? fs_info->free_space_root :
1686 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001687again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001688 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001689 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001690 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001691 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001692 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001693 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001694
Miao Xiecb517ea2013-05-15 07:48:19 +00001695 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001696 if (IS_ERR(root))
1697 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001698
Miao Xiec00869f2013-09-25 21:47:44 +08001699 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001700 ret = -ENOENT;
1701 goto fail;
1702 }
1703
Miao Xiecb517ea2013-05-15 07:48:19 +00001704 ret = btrfs_init_fs_root(root);
1705 if (ret)
1706 goto fail;
1707
David Sterba381cf652015-01-02 18:45:16 +01001708 path = btrfs_alloc_path();
1709 if (!path) {
1710 ret = -ENOMEM;
1711 goto fail;
1712 }
David Sterba1d4c08e2015-01-02 19:36:14 +01001713 key.objectid = BTRFS_ORPHAN_OBJECTID;
1714 key.type = BTRFS_ORPHAN_ITEM_KEY;
1715 key.offset = location->objectid;
1716
1717 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
David Sterba381cf652015-01-02 18:45:16 +01001718 btrfs_free_path(path);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001719 if (ret < 0)
1720 goto fail;
1721 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001722 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001723
Miao Xiecb517ea2013-05-15 07:48:19 +00001724 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001725 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001726 if (ret == -EEXIST) {
1727 free_fs_root(root);
1728 goto again;
1729 }
1730 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001731 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001732 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001733fail:
1734 free_fs_root(root);
1735 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001736}
1737
Chris Mason04160082008-03-26 10:28:07 -04001738static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1739{
1740 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1741 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001742 struct btrfs_device *device;
1743 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001744
Xiao Guangrong1f781602011-04-20 10:09:16 +00001745 rcu_read_lock();
1746 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001747 if (!device->bdev)
1748 continue;
Chris Mason04160082008-03-26 10:28:07 -04001749 bdi = blk_get_backing_dev_info(device->bdev);
Tejun Heoff9ea322014-09-08 08:03:56 +09001750 if (bdi_congested(bdi, bdi_bits)) {
Chris Mason04160082008-03-26 10:28:07 -04001751 ret = 1;
1752 break;
1753 }
1754 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001755 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001756 return ret;
1757}
1758
Chris Mason04160082008-03-26 10:28:07 -04001759static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1760{
Jens Axboead081f12009-06-12 14:43:40 +02001761 int err;
1762
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001763 err = bdi_setup_and_register(bdi, "btrfs");
Jens Axboead081f12009-06-12 14:43:40 +02001764 if (err)
1765 return err;
1766
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001767 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
Chris Mason04160082008-03-26 10:28:07 -04001768 bdi->congested_fn = btrfs_congested_fn;
1769 bdi->congested_data = info;
Chris Masonda2f0f72015-07-02 13:57:22 -07001770 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
Chris Mason04160082008-03-26 10:28:07 -04001771 return 0;
1772}
1773
Chris Mason8b712842008-06-11 16:50:36 -04001774/*
1775 * called by the kthread helper functions to finally call the bio end_io
1776 * functions. This is where read checksum verification actually happens
1777 */
1778static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001779{
Chris Masonce9adaa2008-04-09 16:28:12 -04001780 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001781 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001782
David Sterba97eb6b62014-07-30 00:55:42 +02001783 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001784 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001785
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001786 bio->bi_error = end_io_wq->error;
Chris Mason8b712842008-06-11 16:50:36 -04001787 bio->bi_private = end_io_wq->private;
1788 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001789 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001790 bio_endio(bio);
Chris Mason44b8bd72008-04-16 11:14:51 -04001791}
1792
Chris Masona74a4b92008-06-25 16:01:31 -04001793static int cleaner_kthread(void *arg)
1794{
1795 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001796 int again;
Filipe Mananada288d22015-06-13 06:55:31 +01001797 struct btrfs_trans_handle *trans;
Chris Masona74a4b92008-06-25 16:01:31 -04001798
1799 do {
Miao Xied0278242013-05-14 10:20:40 +00001800 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001801
Miao Xied0278242013-05-14 10:20:40 +00001802 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001803 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001804 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001805
Miao Xied0278242013-05-14 10:20:40 +00001806 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1807 goto sleep;
1808
Miao Xiedc7f3702013-05-14 10:20:42 +00001809 /*
1810 * Avoid the problem that we change the status of the fs
1811 * during the above check and trylock.
1812 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001813 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001814 mutex_unlock(&root->fs_info->cleaner_mutex);
1815 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001816 }
1817
Filipe Mananac2d6cb12016-01-15 11:05:12 +00001818 mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex);
Miao Xied0278242013-05-14 10:20:40 +00001819 btrfs_run_delayed_iputs(root);
Filipe Mananac2d6cb12016-01-15 11:05:12 +00001820 mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex);
1821
Miao Xied0278242013-05-14 10:20:40 +00001822 again = btrfs_clean_one_deleted_snapshot(root);
1823 mutex_unlock(&root->fs_info->cleaner_mutex);
1824
1825 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001826 * The defragger has dealt with the R/O remount and umount,
1827 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001828 */
1829 btrfs_run_defrag_inodes(root->fs_info);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01001830
1831 /*
1832 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1833 * with relocation (btrfs_relocate_chunk) and relocation
1834 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1835 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1836 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1837 * unused block groups.
1838 */
1839 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001840sleep:
Jiri Kosina838fe182016-03-15 11:28:54 +01001841 if (!again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001842 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001843 if (!kthread_should_stop())
1844 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001845 __set_current_state(TASK_RUNNING);
1846 }
1847 } while (!kthread_should_stop());
Filipe Mananada288d22015-06-13 06:55:31 +01001848
1849 /*
1850 * Transaction kthread is stopped before us and wakes us up.
1851 * However we might have started a new transaction and COWed some
1852 * tree blocks when deleting unused block groups for example. So
1853 * make sure we commit the transaction we started to have a clean
1854 * shutdown when evicting the btree inode - if it has dirty pages
1855 * when we do the final iput() on it, eviction will trigger a
1856 * writeback for it which will fail with null pointer dereferences
1857 * since work queues and other resources were already released and
1858 * destroyed by the time the iput/eviction/writeback is made.
1859 */
1860 trans = btrfs_attach_transaction(root);
1861 if (IS_ERR(trans)) {
1862 if (PTR_ERR(trans) != -ENOENT)
1863 btrfs_err(root->fs_info,
1864 "cleaner transaction attach returned %ld",
1865 PTR_ERR(trans));
1866 } else {
1867 int ret;
1868
1869 ret = btrfs_commit_transaction(trans, root);
1870 if (ret)
1871 btrfs_err(root->fs_info,
1872 "cleaner open transaction commit returned %d",
1873 ret);
1874 }
1875
Chris Masona74a4b92008-06-25 16:01:31 -04001876 return 0;
1877}
1878
1879static int transaction_kthread(void *arg)
1880{
1881 struct btrfs_root *root = arg;
1882 struct btrfs_trans_handle *trans;
1883 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001884 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001885 unsigned long now;
1886 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001887 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001888
1889 do {
Jan Kara914b2002012-03-12 16:05:50 +01001890 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001891 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001892 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1893
Josef Bacika4abeea2011-04-11 17:25:13 -04001894 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001895 cur = root->fs_info->running_transaction;
1896 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001897 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001898 goto sleep;
1899 }
Yan Zheng31153d82008-07-28 15:32:19 -04001900
Chris Masona74a4b92008-06-25 16:01:31 -04001901 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001902 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001903 (now < cur->start_time ||
1904 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001905 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001906 delay = HZ * 5;
1907 goto sleep;
1908 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001909 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001910 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001911
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001912 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001913 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001914 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001915 if (PTR_ERR(trans) != -ENOENT)
1916 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001917 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001918 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001919 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001920 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001921 } else {
1922 btrfs_end_transaction(trans, root);
1923 }
Chris Masona74a4b92008-06-25 16:01:31 -04001924sleep:
1925 wake_up_process(root->fs_info->cleaner_kthread);
1926 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1927
Josef Bacik4e121c02013-09-27 16:32:39 -04001928 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1929 &root->fs_info->fs_state)))
1930 btrfs_cleanup_transaction(root);
Jiri Kosinace63f892016-03-15 11:28:59 +01001931 set_current_state(TASK_INTERRUPTIBLE);
1932 if (!kthread_should_stop() &&
1933 (!btrfs_transaction_blocked(root->fs_info) ||
1934 cannot_commit))
1935 schedule_timeout(delay);
1936 __set_current_state(TASK_RUNNING);
Chris Masona74a4b92008-06-25 16:01:31 -04001937 } while (!kthread_should_stop());
1938 return 0;
1939}
1940
Chris Masonaf31f5e2011-11-03 15:17:42 -04001941/*
1942 * this will find the highest generation in the array of
1943 * root backups. The index of the highest array is returned,
1944 * or -1 if we can't find anything.
1945 *
1946 * We check to make sure the array is valid by comparing the
1947 * generation of the latest root in the array with the generation
1948 * in the super block. If they don't match we pitch it.
1949 */
1950static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1951{
1952 u64 cur;
1953 int newest_index = -1;
1954 struct btrfs_root_backup *root_backup;
1955 int i;
1956
1957 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1958 root_backup = info->super_copy->super_roots + i;
1959 cur = btrfs_backup_tree_root_gen(root_backup);
1960 if (cur == newest_gen)
1961 newest_index = i;
1962 }
1963
1964 /* check to see if we actually wrapped around */
1965 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1966 root_backup = info->super_copy->super_roots;
1967 cur = btrfs_backup_tree_root_gen(root_backup);
1968 if (cur == newest_gen)
1969 newest_index = 0;
1970 }
1971 return newest_index;
1972}
1973
1974
1975/*
1976 * find the oldest backup so we know where to store new entries
1977 * in the backup array. This will set the backup_root_index
1978 * field in the fs_info struct
1979 */
1980static void find_oldest_super_backup(struct btrfs_fs_info *info,
1981 u64 newest_gen)
1982{
1983 int newest_index = -1;
1984
1985 newest_index = find_newest_super_backup(info, newest_gen);
1986 /* if there was garbage in there, just move along */
1987 if (newest_index == -1) {
1988 info->backup_root_index = 0;
1989 } else {
1990 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1991 }
1992}
1993
1994/*
1995 * copy all the root pointers into the super backup array.
1996 * this will bump the backup pointer by one when it is
1997 * done
1998 */
1999static void backup_super_roots(struct btrfs_fs_info *info)
2000{
2001 int next_backup;
2002 struct btrfs_root_backup *root_backup;
2003 int last_backup;
2004
2005 next_backup = info->backup_root_index;
2006 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
2007 BTRFS_NUM_BACKUP_ROOTS;
2008
2009 /*
2010 * just overwrite the last backup if we're at the same generation
2011 * this happens only at umount
2012 */
2013 root_backup = info->super_for_commit->super_roots + last_backup;
2014 if (btrfs_backup_tree_root_gen(root_backup) ==
2015 btrfs_header_generation(info->tree_root->node))
2016 next_backup = last_backup;
2017
2018 root_backup = info->super_for_commit->super_roots + next_backup;
2019
2020 /*
2021 * make sure all of our padding and empty slots get zero filled
2022 * regardless of which ones we use today
2023 */
2024 memset(root_backup, 0, sizeof(*root_backup));
2025
2026 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
2027
2028 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
2029 btrfs_set_backup_tree_root_gen(root_backup,
2030 btrfs_header_generation(info->tree_root->node));
2031
2032 btrfs_set_backup_tree_root_level(root_backup,
2033 btrfs_header_level(info->tree_root->node));
2034
2035 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
2036 btrfs_set_backup_chunk_root_gen(root_backup,
2037 btrfs_header_generation(info->chunk_root->node));
2038 btrfs_set_backup_chunk_root_level(root_backup,
2039 btrfs_header_level(info->chunk_root->node));
2040
2041 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2042 btrfs_set_backup_extent_root_gen(root_backup,
2043 btrfs_header_generation(info->extent_root->node));
2044 btrfs_set_backup_extent_root_level(root_backup,
2045 btrfs_header_level(info->extent_root->node));
2046
Chris Mason7c7e82a2011-11-06 18:50:56 -05002047 /*
2048 * we might commit during log recovery, which happens before we set
2049 * the fs_root. Make sure it is valid before we fill it in.
2050 */
2051 if (info->fs_root && info->fs_root->node) {
2052 btrfs_set_backup_fs_root(root_backup,
2053 info->fs_root->node->start);
2054 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002055 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002056 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002057 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002058 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002059
2060 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2061 btrfs_set_backup_dev_root_gen(root_backup,
2062 btrfs_header_generation(info->dev_root->node));
2063 btrfs_set_backup_dev_root_level(root_backup,
2064 btrfs_header_level(info->dev_root->node));
2065
2066 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2067 btrfs_set_backup_csum_root_gen(root_backup,
2068 btrfs_header_generation(info->csum_root->node));
2069 btrfs_set_backup_csum_root_level(root_backup,
2070 btrfs_header_level(info->csum_root->node));
2071
2072 btrfs_set_backup_total_bytes(root_backup,
2073 btrfs_super_total_bytes(info->super_copy));
2074 btrfs_set_backup_bytes_used(root_backup,
2075 btrfs_super_bytes_used(info->super_copy));
2076 btrfs_set_backup_num_devices(root_backup,
2077 btrfs_super_num_devices(info->super_copy));
2078
2079 /*
2080 * if we don't copy this out to the super_copy, it won't get remembered
2081 * for the next commit
2082 */
2083 memcpy(&info->super_copy->super_roots,
2084 &info->super_for_commit->super_roots,
2085 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2086}
2087
2088/*
2089 * this copies info out of the root backup array and back into
2090 * the in-memory super block. It is meant to help iterate through
2091 * the array, so you send it the number of backups you've already
2092 * tried and the last backup index you used.
2093 *
2094 * this returns -1 when it has tried all the backups
2095 */
2096static noinline int next_root_backup(struct btrfs_fs_info *info,
2097 struct btrfs_super_block *super,
2098 int *num_backups_tried, int *backup_index)
2099{
2100 struct btrfs_root_backup *root_backup;
2101 int newest = *backup_index;
2102
2103 if (*num_backups_tried == 0) {
2104 u64 gen = btrfs_super_generation(super);
2105
2106 newest = find_newest_super_backup(info, gen);
2107 if (newest == -1)
2108 return -1;
2109
2110 *backup_index = newest;
2111 *num_backups_tried = 1;
2112 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2113 /* we've tried all the backups, all done */
2114 return -1;
2115 } else {
2116 /* jump to the next oldest backup */
2117 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2118 BTRFS_NUM_BACKUP_ROOTS;
2119 *backup_index = newest;
2120 *num_backups_tried += 1;
2121 }
2122 root_backup = super->super_roots + newest;
2123
2124 btrfs_set_super_generation(super,
2125 btrfs_backup_tree_root_gen(root_backup));
2126 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2127 btrfs_set_super_root_level(super,
2128 btrfs_backup_tree_root_level(root_backup));
2129 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2130
2131 /*
2132 * fixme: the total bytes and num_devices need to match or we should
2133 * need a fsck
2134 */
2135 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2136 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2137 return 0;
2138}
2139
Liu Bo7abadb62013-03-17 02:10:31 +00002140/* helper to cleanup workers */
2141static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2142{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002143 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002144 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002145 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002146 btrfs_destroy_workqueue(fs_info->endio_workers);
2147 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2148 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002149 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002150 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002151 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2152 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2153 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002154 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002155 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002156 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002157 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002158 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002159 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002160 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002161}
2162
Rashika2e9f5952013-10-31 02:45:20 +05302163static void free_root_extent_buffers(struct btrfs_root *root)
2164{
2165 if (root) {
2166 free_extent_buffer(root->node);
2167 free_extent_buffer(root->commit_root);
2168 root->node = NULL;
2169 root->commit_root = NULL;
2170 }
2171}
2172
Chris Masonaf31f5e2011-11-03 15:17:42 -04002173/* helper to cleanup tree roots */
2174static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2175{
Rashika2e9f5952013-10-31 02:45:20 +05302176 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002177
Rashika2e9f5952013-10-31 02:45:20 +05302178 free_root_extent_buffers(info->dev_root);
2179 free_root_extent_buffers(info->extent_root);
2180 free_root_extent_buffers(info->csum_root);
2181 free_root_extent_buffers(info->quota_root);
2182 free_root_extent_buffers(info->uuid_root);
2183 if (chunk_root)
2184 free_root_extent_buffers(info->chunk_root);
Omar Sandoval70f6d822015-09-29 20:50:38 -07002185 free_root_extent_buffers(info->free_space_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002186}
2187
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002188void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002189{
2190 int ret;
2191 struct btrfs_root *gang[8];
2192 int i;
2193
2194 while (!list_empty(&fs_info->dead_roots)) {
2195 gang[0] = list_entry(fs_info->dead_roots.next,
2196 struct btrfs_root, root_list);
2197 list_del(&gang[0]->root_list);
2198
Miao Xie27cdeb72014-04-02 19:51:05 +08002199 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002200 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002201 } else {
2202 free_extent_buffer(gang[0]->node);
2203 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002204 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002205 }
2206 }
2207
2208 while (1) {
2209 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2210 (void **)gang, 0,
2211 ARRAY_SIZE(gang));
2212 if (!ret)
2213 break;
2214 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002215 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002216 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002217
2218 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2219 btrfs_free_log_root_tree(NULL, fs_info);
2220 btrfs_destroy_pinned_extent(fs_info->tree_root,
2221 fs_info->pinned_extents);
2222 }
Josef Bacik171f6532013-04-24 16:35:41 -04002223}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002224
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002225static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2226{
2227 mutex_init(&fs_info->scrub_lock);
2228 atomic_set(&fs_info->scrubs_running, 0);
2229 atomic_set(&fs_info->scrub_pause_req, 0);
2230 atomic_set(&fs_info->scrubs_paused, 0);
2231 atomic_set(&fs_info->scrub_cancel_req, 0);
2232 init_waitqueue_head(&fs_info->scrub_pause_wait);
2233 fs_info->scrub_workers_refcnt = 0;
2234}
2235
Eric Sandeen779a65a2014-08-01 18:12:39 -05002236static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2237{
2238 spin_lock_init(&fs_info->balance_lock);
2239 mutex_init(&fs_info->balance_mutex);
2240 atomic_set(&fs_info->balance_running, 0);
2241 atomic_set(&fs_info->balance_pause_req, 0);
2242 atomic_set(&fs_info->balance_cancel_req, 0);
2243 fs_info->balance_ctl = NULL;
2244 init_waitqueue_head(&fs_info->balance_wait_q);
2245}
2246
Eric Sandeenf37938e2014-08-01 18:12:40 -05002247static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2248 struct btrfs_root *tree_root)
2249{
2250 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2251 set_nlink(fs_info->btree_inode, 1);
2252 /*
2253 * we set the i_size on the btree inode to the max possible int.
2254 * the real end of the address space is determined by all of
2255 * the devices in the system
2256 */
2257 fs_info->btree_inode->i_size = OFFSET_MAX;
2258 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Eric Sandeenf37938e2014-08-01 18:12:40 -05002259
2260 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2261 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2262 fs_info->btree_inode->i_mapping);
2263 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2264 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2265
2266 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2267
2268 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2269 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2270 sizeof(struct btrfs_key));
2271 set_bit(BTRFS_INODE_DUMMY,
2272 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2273 btrfs_insert_inode_hash(fs_info->btree_inode);
2274}
2275
Eric Sandeenad618362014-08-01 18:12:41 -05002276static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2277{
2278 fs_info->dev_replace.lock_owner = 0;
2279 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2280 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
Liu Bo73beece2015-07-17 16:49:19 +08002281 rwlock_init(&fs_info->dev_replace.lock);
2282 atomic_set(&fs_info->dev_replace.read_locks, 0);
2283 atomic_set(&fs_info->dev_replace.blocking_readers, 0);
Eric Sandeenad618362014-08-01 18:12:41 -05002284 init_waitqueue_head(&fs_info->replace_wait);
Liu Bo73beece2015-07-17 16:49:19 +08002285 init_waitqueue_head(&fs_info->dev_replace.read_lock_wq);
Eric Sandeenad618362014-08-01 18:12:41 -05002286}
2287
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002288static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2289{
2290 spin_lock_init(&fs_info->qgroup_lock);
2291 mutex_init(&fs_info->qgroup_ioctl_lock);
2292 fs_info->qgroup_tree = RB_ROOT;
2293 fs_info->qgroup_op_tree = RB_ROOT;
2294 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2295 fs_info->qgroup_seq = 1;
2296 fs_info->quota_enabled = 0;
2297 fs_info->pending_quota_state = 0;
2298 fs_info->qgroup_ulist = NULL;
2299 mutex_init(&fs_info->qgroup_rescan_lock);
2300}
2301
Eric Sandeen2a458192015-02-16 16:29:26 +01002302static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2303 struct btrfs_fs_devices *fs_devices)
2304{
2305 int max_active = fs_info->thread_pool_size;
David Sterba6f011052015-02-16 18:34:01 +01002306 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Eric Sandeen2a458192015-02-16 16:29:26 +01002307
2308 fs_info->workers =
2309 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2310 max_active, 16);
2311
2312 fs_info->delalloc_workers =
2313 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
2314
2315 fs_info->flush_workers =
2316 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
2317
2318 fs_info->caching_workers =
2319 btrfs_alloc_workqueue("cache", flags, max_active, 0);
2320
2321 /*
2322 * a higher idle thresh on the submit workers makes it much more
2323 * likely that bios will be send down in a sane order to the
2324 * devices
2325 */
2326 fs_info->submit_workers =
2327 btrfs_alloc_workqueue("submit", flags,
2328 min_t(u64, fs_devices->num_devices,
2329 max_active), 64);
2330
2331 fs_info->fixup_workers =
2332 btrfs_alloc_workqueue("fixup", flags, 1, 0);
2333
2334 /*
2335 * endios are largely parallel and should have a very
2336 * low idle thresh
2337 */
2338 fs_info->endio_workers =
2339 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2340 fs_info->endio_meta_workers =
2341 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2342 fs_info->endio_meta_write_workers =
2343 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2344 fs_info->endio_raid56_workers =
2345 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
2346 fs_info->endio_repair_workers =
2347 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
2348 fs_info->rmw_workers =
2349 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
2350 fs_info->endio_write_workers =
2351 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2352 fs_info->endio_freespace_worker =
2353 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
2354 fs_info->delayed_workers =
2355 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
2356 fs_info->readahead_workers =
2357 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
2358 fs_info->qgroup_rescan_workers =
2359 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
2360 fs_info->extent_workers =
2361 btrfs_alloc_workqueue("extent-refs", flags,
2362 min_t(u64, fs_devices->num_devices,
2363 max_active), 8);
2364
2365 if (!(fs_info->workers && fs_info->delalloc_workers &&
2366 fs_info->submit_workers && fs_info->flush_workers &&
2367 fs_info->endio_workers && fs_info->endio_meta_workers &&
2368 fs_info->endio_meta_write_workers &&
2369 fs_info->endio_repair_workers &&
2370 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2371 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2372 fs_info->caching_workers && fs_info->readahead_workers &&
2373 fs_info->fixup_workers && fs_info->delayed_workers &&
2374 fs_info->extent_workers &&
2375 fs_info->qgroup_rescan_workers)) {
2376 return -ENOMEM;
2377 }
2378
2379 return 0;
2380}
2381
Eric Sandeen63443bf2014-08-01 18:12:46 -05002382static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2383 struct btrfs_fs_devices *fs_devices)
2384{
2385 int ret;
2386 struct btrfs_root *tree_root = fs_info->tree_root;
2387 struct btrfs_root *log_tree_root;
2388 struct btrfs_super_block *disk_super = fs_info->super_copy;
2389 u64 bytenr = btrfs_super_log_root(disk_super);
2390
2391 if (fs_devices->rw_devices == 0) {
David Sterbaf14d1042015-10-08 11:37:06 +02002392 btrfs_warn(fs_info, "log replay required on RO media");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002393 return -EIO;
2394 }
2395
David Sterba74e4d822016-02-11 11:01:55 +01002396 log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Eric Sandeen63443bf2014-08-01 18:12:46 -05002397 if (!log_tree_root)
2398 return -ENOMEM;
2399
2400 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2401 tree_root->stripesize, log_tree_root, fs_info,
2402 BTRFS_TREE_LOG_OBJECTID);
2403
2404 log_tree_root->node = read_tree_block(tree_root, bytenr,
2405 fs_info->generation + 1);
Liu Bo64c043d2015-05-25 17:30:15 +08002406 if (IS_ERR(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002407 btrfs_warn(fs_info, "failed to read log tree");
Liu Bo0eeff232015-06-11 14:16:44 +08002408 ret = PTR_ERR(log_tree_root->node);
Liu Bo64c043d2015-05-25 17:30:15 +08002409 kfree(log_tree_root);
Liu Bo0eeff232015-06-11 14:16:44 +08002410 return ret;
Liu Bo64c043d2015-05-25 17:30:15 +08002411 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002412 btrfs_err(fs_info, "failed to read log tree");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002413 free_extent_buffer(log_tree_root->node);
2414 kfree(log_tree_root);
2415 return -EIO;
2416 }
2417 /* returns with log_tree_root freed on success */
2418 ret = btrfs_recover_log_trees(log_tree_root);
2419 if (ret) {
Anand Jaina4553fe2015-09-25 14:43:01 +08002420 btrfs_std_error(tree_root->fs_info, ret,
Eric Sandeen63443bf2014-08-01 18:12:46 -05002421 "Failed to recover log tree");
2422 free_extent_buffer(log_tree_root->node);
2423 kfree(log_tree_root);
2424 return ret;
2425 }
2426
2427 if (fs_info->sb->s_flags & MS_RDONLY) {
2428 ret = btrfs_commit_super(tree_root);
2429 if (ret)
2430 return ret;
2431 }
2432
2433 return 0;
2434}
2435
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002436static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2437 struct btrfs_root *tree_root)
2438{
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002439 struct btrfs_root *root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002440 struct btrfs_key location;
2441 int ret;
2442
2443 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2444 location.type = BTRFS_ROOT_ITEM_KEY;
2445 location.offset = 0;
2446
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002447 root = btrfs_read_tree_root(tree_root, &location);
2448 if (IS_ERR(root))
2449 return PTR_ERR(root);
2450 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2451 fs_info->extent_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002452
2453 location.objectid = BTRFS_DEV_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002454 root = btrfs_read_tree_root(tree_root, &location);
2455 if (IS_ERR(root))
2456 return PTR_ERR(root);
2457 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2458 fs_info->dev_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002459 btrfs_init_devices_late(fs_info);
2460
2461 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002462 root = btrfs_read_tree_root(tree_root, &location);
2463 if (IS_ERR(root))
2464 return PTR_ERR(root);
2465 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2466 fs_info->csum_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002467
2468 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002469 root = btrfs_read_tree_root(tree_root, &location);
2470 if (!IS_ERR(root)) {
2471 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002472 fs_info->quota_enabled = 1;
2473 fs_info->pending_quota_state = 1;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002474 fs_info->quota_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002475 }
2476
2477 location.objectid = BTRFS_UUID_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002478 root = btrfs_read_tree_root(tree_root, &location);
2479 if (IS_ERR(root)) {
2480 ret = PTR_ERR(root);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002481 if (ret != -ENOENT)
2482 return ret;
2483 } else {
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002484 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2485 fs_info->uuid_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002486 }
2487
Omar Sandoval70f6d822015-09-29 20:50:38 -07002488 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2489 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2490 root = btrfs_read_tree_root(tree_root, &location);
2491 if (IS_ERR(root))
2492 return PTR_ERR(root);
2493 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2494 fs_info->free_space_root = root;
2495 }
2496
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002497 return 0;
2498}
2499
Al Viroad2b2c82011-11-17 01:10:02 -05002500int open_ctree(struct super_block *sb,
2501 struct btrfs_fs_devices *fs_devices,
2502 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002503{
Chris Masondb945352007-10-15 16:15:53 -04002504 u32 sectorsize;
2505 u32 nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002506 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002507 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002508 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002509 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002510 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002511 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002512 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002513 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002514 struct btrfs_root *chunk_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002515 int ret;
Yane58ca022008-04-01 11:21:34 -04002516 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002517 int num_backups_tried = 0;
2518 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002519 int max_active;
Chris Mason4543df72008-06-11 21:47:56 -04002520
David Sterba74e4d822016-02-11 11:01:55 +01002521 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
2522 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Miao Xiecb517ea2013-05-15 07:48:19 +00002523 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002524 err = -ENOMEM;
2525 goto fail;
2526 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002527
2528 ret = init_srcu_struct(&fs_info->subvol_srcu);
2529 if (ret) {
2530 err = ret;
2531 goto fail;
2532 }
2533
2534 ret = setup_bdi(fs_info, &fs_info->bdi);
2535 if (ret) {
2536 err = ret;
2537 goto fail_srcu;
2538 }
2539
Tejun Heo908c7f12014-09-08 09:51:29 +09002540 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
Miao Xiee2d84522013-01-29 10:09:20 +00002541 if (ret) {
2542 err = ret;
2543 goto fail_bdi;
2544 }
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002545 fs_info->dirty_metadata_batch = PAGE_SIZE *
Miao Xiee2d84522013-01-29 10:09:20 +00002546 (1 + ilog2(nr_cpu_ids));
2547
Tejun Heo908c7f12014-09-08 09:51:29 +09002548 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
Miao Xie963d6782013-01-29 10:10:51 +00002549 if (ret) {
2550 err = ret;
2551 goto fail_dirty_metadata_bytes;
2552 }
2553
Tejun Heo908c7f12014-09-08 09:51:29 +09002554 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
Miao Xiec404e0d2014-01-30 16:46:55 +08002555 if (ret) {
2556 err = ret;
2557 goto fail_delalloc_bytes;
2558 }
2559
Yan, Zheng76dda932009-09-21 16:00:26 -04002560 fs_info->btree_inode = new_inode(sb);
2561 if (!fs_info->btree_inode) {
2562 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002563 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002564 }
2565
Chris Masona6591712011-07-19 12:04:14 -04002566 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002567
Yan, Zheng76dda932009-09-21 16:00:26 -04002568 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002569 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002570 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002571 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002572 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002573 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002574 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002575 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002576 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002577 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002578 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002579 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002580 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002581 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002582 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002583 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002584 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002585 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002586 rwlock_init(&fs_info->tree_mod_log_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002587 mutex_init(&fs_info->unused_bg_unpin_mutex);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01002588 mutex_init(&fs_info->delete_unused_bgs_mutex);
Chris Mason75857172011-06-13 20:00:16 -04002589 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002590 mutex_init(&fs_info->delalloc_root_mutex);
Filipe Mananac2d6cb12016-01-15 11:05:12 +00002591 mutex_init(&fs_info->cleaner_delayed_iput_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002592 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002593
Chris Mason0b86a832008-03-24 15:01:56 -04002594 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002595 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002596 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002597 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002598 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002599 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2600 BTRFS_BLOCK_RSV_GLOBAL);
2601 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2602 BTRFS_BLOCK_RSV_DELALLOC);
2603 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2604 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2605 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2606 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2607 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002608 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002609 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002610 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002611 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002612 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002613 atomic_set(&fs_info->qgroup_op_seq, 0);
Zhao Lei2fefd552016-01-07 18:38:48 +08002614 atomic_set(&fs_info->reada_works_cnt, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002615 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002616 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002617 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002618 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002619 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002620 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002621 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002622 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
David Sterbaf8c269d2015-01-16 17:21:12 +01002623 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
Arne Jansen90519d62011-05-23 14:30:00 +02002624 /* readahead state */
Mel Gormand0164ad2015-11-06 16:28:21 -08002625 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
Arne Jansen90519d62011-05-23 14:30:00 +02002626 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002627
Chris Masonb34b0862008-12-19 15:43:22 -05002628 fs_info->thread_pool_size = min_t(unsigned long,
2629 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002630
Miao Xie199c2a92013-05-15 07:48:23 +00002631 INIT_LIST_HEAD(&fs_info->ordered_roots);
2632 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002633 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
David Sterba74e4d822016-02-11 11:01:55 +01002634 GFP_KERNEL);
Miao Xie16cdcec2011-04-22 18:12:22 +08002635 if (!fs_info->delayed_root) {
2636 err = -ENOMEM;
2637 goto fail_iput;
2638 }
2639 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002640
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002641 btrfs_init_scrub(fs_info);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002642#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2643 fs_info->check_integrity_print_mask = 0;
2644#endif
Eric Sandeen779a65a2014-08-01 18:12:39 -05002645 btrfs_init_balance(fs_info);
Miao Xie21c7e752014-05-13 17:29:04 -07002646 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Arne Jansena2de7332011-03-08 14:14:00 +01002647
Chris Masona061fc82008-05-07 11:43:44 -04002648 sb->s_blocksize = 4096;
2649 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002650 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04002651
Eric Sandeenf37938e2014-08-01 18:12:40 -05002652 btrfs_init_btree_inode(fs_info, tree_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002653
Chris Masone02119d2008-09-05 16:13:11 -04002654 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002655 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002656 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002657
Yan Zheng11833d62009-09-11 16:11:19 -04002658 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002659 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002660 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002661 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002662 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002663 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002664
Chris Masond98237b2007-03-28 13:57:48 -04002665
Chris Mason5a3f23d2009-03-31 13:27:11 -04002666 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002667 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002668 mutex_init(&fs_info->chunk_mutex);
2669 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002670 mutex_init(&fs_info->cleaner_mutex);
2671 mutex_init(&fs_info->volume_mutex);
Chris Mason1bbc6212015-04-06 12:46:08 -07002672 mutex_init(&fs_info->ro_block_group_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002673 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002674 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002675 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002676 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002677
Eric Sandeenad618362014-08-01 18:12:41 -05002678 btrfs_init_dev_replace_locks(fs_info);
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002679 btrfs_init_qgroup(fs_info);
Arne Jansen416ac512011-09-13 12:56:09 +02002680
Chris Masonfa9c0d792009-04-03 09:47:43 -04002681 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2682 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2683
Chris Mason7d9eb122008-07-08 14:19:17 -04002684 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002685 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002686 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002687 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002688
Filipe Manana04216822014-11-27 21:14:15 +00002689 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2690
David Woodhouse53b381b2013-01-29 18:40:14 -05002691 ret = btrfs_alloc_stripe_hash_table(fs_info);
2692 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002693 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002694 goto fail_alloc;
2695 }
2696
David Sterba707e8a02014-06-04 19:22:26 +02002697 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002698 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002699
Chris Mason3c4bb262012-03-27 18:56:56 -04002700 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002701
2702 /*
2703 * Read super block and check the signature bytes only
2704 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002705 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Anand Jain92fc03f2015-08-14 18:32:51 +08002706 if (IS_ERR(bh)) {
2707 err = PTR_ERR(bh);
Miao Xie16cdcec2011-04-22 18:12:22 +08002708 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002709 }
Chris Mason39279cc2007-06-12 06:35:45 -04002710
David Sterba1104a882013-03-06 15:57:46 +01002711 /*
2712 * We want to check superblock checksum, the type is stored inside.
2713 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2714 */
2715 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002716 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002717 err = -EINVAL;
Anand Jainb2acddd2015-10-07 17:23:23 +08002718 brelse(bh);
David Sterba1104a882013-03-06 15:57:46 +01002719 goto fail_alloc;
2720 }
2721
2722 /*
2723 * super_copy is zeroed at allocation time and we never touch the
2724 * following bytes up to INFO_SIZE, the checksum is calculated from
2725 * the whole block of INFO_SIZE
2726 */
David Sterba6c417612011-04-13 15:41:04 +02002727 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2728 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2729 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002730 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002731
David Sterba6c417612011-04-13 15:41:04 +02002732 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002733
David Sterba1104a882013-03-06 15:57:46 +01002734 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2735 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002736 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002737 err = -EINVAL;
2738 goto fail_alloc;
2739 }
2740
David Sterba6c417612011-04-13 15:41:04 +02002741 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002742 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002743 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002744
liuboacce9522011-01-06 19:30:25 +08002745 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002746 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2747 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002748
Liu Bo75e7cb72011-03-22 10:12:20 +00002749 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002750 * run through our array of backup supers and setup
2751 * our ring pointer to the oldest one
2752 */
2753 generation = btrfs_super_generation(disk_super);
2754 find_oldest_super_backup(fs_info, generation);
2755
2756 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002757 * In the long term, we'll store the compression type in the super
2758 * block, and it'll be used for per file compression control.
2759 */
2760 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2761
Qu Wenruo96da0912016-01-19 10:23:03 +08002762 ret = btrfs_parse_options(tree_root, options, sb->s_flags);
Yan Zheng2b820322008-11-17 21:11:30 -05002763 if (ret) {
2764 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002765 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002766 }
Chris Masondfe25022008-05-13 13:46:40 -04002767
Josef Bacikf2b636e2008-12-02 06:36:08 -05002768 features = btrfs_super_incompat_flags(disk_super) &
2769 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2770 if (features) {
2771 printk(KERN_ERR "BTRFS: couldn't mount because of "
2772 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002773 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002774 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002775 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002776 }
2777
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002778 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002779 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002780 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002781 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002782
Josef Bacik3173a182013-03-07 14:22:04 -05002783 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
David Sterba5efa0492014-12-19 18:38:47 +01002784 printk(KERN_INFO "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002785
Chris Mason727011e2010-08-06 13:21:20 -04002786 /*
2787 * flag our filesystem as having big metadata blocks if
2788 * they are bigger than the page size
2789 */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002790 if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002791 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002792 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002793 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2794 }
2795
Chris Masonbc3f1162012-03-29 17:02:47 -04002796 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002797 sectorsize = btrfs_super_sectorsize(disk_super);
2798 stripesize = btrfs_super_stripesize(disk_super);
David Sterba707e8a02014-06-04 19:22:26 +02002799 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002800 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002801
2802 /*
2803 * mixed block groups end up with duplicate but slightly offset
2804 * extent buffers for the same range. It leads to corruptions
2805 */
2806 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002807 (sectorsize != nodesize)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002808 printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002809 "are not allowed for mixed block groups on %s\n",
2810 sb->s_id);
2811 goto fail_alloc;
2812 }
2813
Miao Xieceda0862013-04-11 10:30:16 +00002814 /*
2815 * Needn't use the lock because there is no other task which will
2816 * update the flag.
2817 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002818 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002819
Josef Bacikf2b636e2008-12-02 06:36:08 -05002820 features = btrfs_super_compat_ro_flags(disk_super) &
2821 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2822 if (!(sb->s_flags & MS_RDONLY) && features) {
2823 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2824 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002825 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002826 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002827 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002828 }
Chris Mason61d92c32009-10-02 19:11:56 -04002829
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002830 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002831
Eric Sandeen2a458192015-02-16 16:29:26 +01002832 ret = btrfs_init_workqueues(fs_info, fs_devices);
2833 if (ret) {
2834 err = ret;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002835 goto fail_sb_buffer;
2836 }
Chris Mason4543df72008-06-11 21:47:56 -04002837
Chris Mason4575c9c2008-04-18 16:13:31 -04002838 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002839 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002840 SZ_4M / PAGE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002841
Chris Masondb945352007-10-15 16:15:53 -04002842 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002843 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002844 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002845
2846 sb->s_blocksize = sectorsize;
2847 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002848
Chris Mason925baed2008-06-25 16:01:30 -04002849 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002850 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002851 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002852 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002853 printk(KERN_ERR "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002854 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002855 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002856 }
Chris Mason0b86a832008-03-24 15:01:56 -04002857
Yan Zheng84234f32008-10-29 14:49:05 -04002858 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002859
David Sterba707e8a02014-06-04 19:22:26 +02002860 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2861 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002862
2863 chunk_root->node = read_tree_block(chunk_root,
2864 btrfs_super_chunk_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002865 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002866 if (IS_ERR(chunk_root->node) ||
2867 !extent_buffer_uptodate(chunk_root->node)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002868 printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002869 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302870 if (!IS_ERR(chunk_root->node))
2871 free_extent_buffer(chunk_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002872 chunk_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002873 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002874 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002875 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2876 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002877
Chris Masone17cade2008-04-15 15:41:47 -04002878 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002879 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002880
Chris Mason0b86a832008-03-24 15:01:56 -04002881 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002882 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002883 printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002884 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002885 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002886 }
Chris Mason0b86a832008-03-24 15:01:56 -04002887
Stefan Behrens8dabb742012-11-06 13:15:27 +01002888 /*
2889 * keep the device that is marked to be the target device for the
2890 * dev_replace procedure
2891 */
Eric Sandeen9eaed212014-08-01 18:12:35 -05002892 btrfs_close_extra_devices(fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002893
Chris Masona6b0d5c2012-02-20 20:53:43 -05002894 if (!fs_devices->latest_bdev) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002895 printk(KERN_ERR "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002896 sb->s_id);
2897 goto fail_tree_roots;
2898 }
2899
Chris Masonaf31f5e2011-11-03 15:17:42 -04002900retry_root_backup:
Yan Zheng84234f32008-10-29 14:49:05 -04002901 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002902
Chris Masone20d96d2007-03-22 12:13:20 -04002903 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002904 btrfs_super_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002905 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002906 if (IS_ERR(tree_root->node) ||
2907 !extent_buffer_uptodate(tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002908 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002909 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302910 if (!IS_ERR(tree_root->node))
2911 free_extent_buffer(tree_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002912 tree_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002913 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002914 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002915
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002916 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2917 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002918 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002919
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05302920 mutex_lock(&tree_root->objectid_mutex);
2921 ret = btrfs_find_highest_objectid(tree_root,
2922 &tree_root->highest_objectid);
2923 if (ret) {
2924 mutex_unlock(&tree_root->objectid_mutex);
2925 goto recovery_tree_root;
2926 }
2927
2928 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
2929
2930 mutex_unlock(&tree_root->objectid_mutex);
2931
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002932 ret = btrfs_read_roots(fs_info, tree_root);
2933 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002934 goto recovery_tree_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002935
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002936 fs_info->generation = generation;
2937 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002938
Ilya Dryomov68310a52012-06-22 12:24:12 -06002939 ret = btrfs_recover_balance(fs_info);
2940 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002941 printk(KERN_ERR "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002942 goto fail_block_groups;
2943 }
2944
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002945 ret = btrfs_init_dev_stats(fs_info);
2946 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002947 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002948 ret);
2949 goto fail_block_groups;
2950 }
2951
Stefan Behrens8dabb742012-11-06 13:15:27 +01002952 ret = btrfs_init_dev_replace(fs_info);
2953 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002954 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002955 goto fail_block_groups;
2956 }
2957
Eric Sandeen9eaed212014-08-01 18:12:35 -05002958 btrfs_close_extra_devices(fs_devices, 1);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002959
Anand Jainb7c35e82015-03-10 06:38:38 +08002960 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
2961 if (ret) {
2962 pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret);
2963 goto fail_block_groups;
2964 }
2965
2966 ret = btrfs_sysfs_add_device(fs_devices);
2967 if (ret) {
2968 pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret);
2969 goto fail_fsdev_sysfs;
2970 }
2971
Anand Jain96f31362015-08-14 18:32:46 +08002972 ret = btrfs_sysfs_add_mounted(fs_info);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002973 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002974 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08002975 goto fail_fsdev_sysfs;
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002976 }
2977
liuboc59021f2011-03-07 02:13:14 +00002978 ret = btrfs_init_space_info(fs_info);
2979 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002980 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002981 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002982 }
2983
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002984 ret = btrfs_read_block_groups(fs_info->extent_root);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002985 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002986 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002987 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002988 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002989 fs_info->num_tolerated_disk_barrier_failures =
2990 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002991 if (fs_info->fs_devices->missing_devices >
2992 fs_info->num_tolerated_disk_barrier_failures &&
2993 !(sb->s_flags & MS_RDONLY)) {
Zhao Lei78fa1772015-07-20 17:54:50 +08002994 pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n",
2995 fs_info->fs_devices->missing_devices,
2996 fs_info->num_tolerated_disk_barrier_failures);
Anand Jain2365dd32014-01-22 11:15:51 +08002997 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002998 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002999
Chris Masona74a4b92008-06-25 16:01:31 -04003000 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
3001 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05003002 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08003003 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04003004
3005 fs_info->transaction_kthread = kthread_run(transaction_kthread,
3006 tree_root,
3007 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05003008 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04003009 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04003010
Chris Masonc2898112009-06-10 09:51:32 -04003011 if (!btrfs_test_opt(tree_root, SSD) &&
3012 !btrfs_test_opt(tree_root, NOSSD) &&
3013 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003014 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04003015 "mode\n");
3016 btrfs_set_opt(fs_info->mount_opt, SSD);
3017 }
3018
David Sterba572d9ab2014-02-05 15:26:17 +01003019 /*
3020 * Mount does not set all options immediatelly, we can do it now and do
3021 * not have to wait for transaction commit
3022 */
3023 btrfs_apply_pending_changes(fs_info);
Qu Wenruo3818aea2014-01-13 13:36:06 +08003024
Stefan Behrens21adbd52011-11-09 13:44:05 +01003025#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3026 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
3027 ret = btrfsic_mount(tree_root, fs_devices,
3028 btrfs_test_opt(tree_root,
3029 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3030 1 : 0,
3031 fs_info->check_integrity_print_mask);
3032 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003033 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01003034 " integrity check module %s\n", sb->s_id);
3035 }
3036#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02003037 ret = btrfs_read_qgroup_config(fs_info);
3038 if (ret)
3039 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01003040
Qu Wenruo96da0912016-01-19 10:23:03 +08003041 /* do not make disk changes in broken FS or nologreplay is given */
3042 if (btrfs_super_log_root(disk_super) != 0 &&
3043 !btrfs_test_opt(tree_root, NOLOGREPLAY)) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003044 ret = btrfs_replay_log(fs_info, fs_devices);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003045 if (ret) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003046 err = ret;
Wang Shilong28c16cb2014-04-23 19:33:35 +08003047 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003048 }
Chris Masone02119d2008-09-05 16:13:11 -04003049 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003050
Yan, Zheng76dda932009-09-21 16:00:26 -04003051 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003052 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003053 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04003054
Chris Mason7c2ca462008-11-19 15:13:35 -05003055 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04003056 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06003057 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003058 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04003059
Wang Shilong5f316482014-06-26 11:08:16 +08003060 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003061 ret = btrfs_recover_relocation(tree_root);
Wang Shilong5f316482014-06-26 11:08:16 +08003062 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00003063 if (ret < 0) {
3064 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05003065 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00003066 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02003067 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00003068 }
Chris Mason7c2ca462008-11-19 15:13:35 -05003069 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003070
Chris Mason3de45862008-11-17 21:02:50 -05003071 location.objectid = BTRFS_FS_TREE_OBJECTID;
3072 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00003073 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05003074
Chris Mason3de45862008-11-17 21:02:50 -05003075 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003076 if (IS_ERR(fs_info->fs_root)) {
3077 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003078 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003079 }
Chris Masonc2898112009-06-10 09:51:32 -04003080
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003081 if (sb->s_flags & MS_RDONLY)
3082 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02003083
Chris Mason511711a2015-12-30 07:52:35 -08003084 if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) &&
3085 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3086 pr_info("BTRFS: creating free space tree\n");
3087 ret = btrfs_create_free_space_tree(fs_info);
3088 if (ret) {
3089 pr_warn("BTRFS: failed to create free space tree %d\n",
3090 ret);
3091 close_ctree(tree_root);
3092 return ret;
3093 }
3094 }
3095
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003096 down_read(&fs_info->cleanup_work_sem);
3097 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3098 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00003099 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003100 close_ctree(tree_root);
3101 return ret;
3102 }
3103 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02003104
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003105 ret = btrfs_resume_balance_async(fs_info);
3106 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003107 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003108 close_ctree(tree_root);
3109 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00003110 }
3111
Stefan Behrens8dabb742012-11-06 13:15:27 +01003112 ret = btrfs_resume_dev_replace_async(fs_info);
3113 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003114 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01003115 close_ctree(tree_root);
3116 return ret;
3117 }
3118
Jan Schmidtb382a322013-05-28 15:47:24 +00003119 btrfs_qgroup_rescan_resume(fs_info);
3120
Omar Sandoval70f6d822015-09-29 20:50:38 -07003121 if (btrfs_test_opt(tree_root, CLEAR_CACHE) &&
3122 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3123 pr_info("BTRFS: clearing free space tree\n");
3124 ret = btrfs_clear_free_space_tree(fs_info);
3125 if (ret) {
3126 pr_warn("BTRFS: failed to clear free space tree %d\n",
3127 ret);
3128 close_ctree(tree_root);
3129 return ret;
3130 }
3131 }
3132
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003133 if (!fs_info->uuid_root) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003134 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003135 ret = btrfs_create_uuid_tree(fs_info);
3136 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003137 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003138 ret);
3139 close_ctree(tree_root);
3140 return ret;
3141 }
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003142 } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) ||
3143 fs_info->generation !=
3144 btrfs_super_uuid_tree_generation(disk_super)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003145 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02003146 ret = btrfs_check_uuid_tree(fs_info);
3147 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003148 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02003149 ret);
3150 close_ctree(tree_root);
3151 return ret;
3152 }
3153 } else {
3154 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003155 }
3156
Josef Bacik47ab2a62014-09-18 11:20:02 -04003157 fs_info->open = 1;
3158
Qu Wenruo8dcddfa2016-01-19 10:23:02 +08003159 /*
3160 * backuproot only affect mount behavior, and if open_ctree succeeded,
3161 * no need to keep the flag
3162 */
3163 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
3164
Al Viroad2b2c82011-11-17 01:10:02 -05003165 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04003166
Arne Jansenbcef60f2011-09-13 15:23:30 +02003167fail_qgroup:
3168 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05003169fail_trans_kthread:
3170 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04003171 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003172 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04003173fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04003174 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05003175
3176 /*
3177 * make sure we're done with the btree inode before we stop our
3178 * kthreads
3179 */
3180 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05003181
Anand Jain2365dd32014-01-22 11:15:51 +08003182fail_sysfs:
Anand Jain6618a592015-08-14 18:32:47 +08003183 btrfs_sysfs_remove_mounted(fs_info);
Anand Jain2365dd32014-01-22 11:15:51 +08003184
Anand Jainb7c35e82015-03-10 06:38:38 +08003185fail_fsdev_sysfs:
3186 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3187
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003188fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003189 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003190 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003191
3192fail_tree_roots:
3193 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003194 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003195
Chris Mason39279cc2007-06-12 06:35:45 -04003196fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003197 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003198fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003199fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003200 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3201
Chris Mason4543df72008-06-11 21:47:56 -04003202 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003203fail_bio_counter:
3204 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003205fail_delalloc_bytes:
3206 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003207fail_dirty_metadata_bytes:
3208 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003209fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003210 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003211fail_srcu:
3212 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003213fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003214 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003215 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003216 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003217
3218recovery_tree_root:
Qu Wenruo8dcddfa2016-01-19 10:23:02 +08003219 if (!btrfs_test_opt(tree_root, USEBACKUPROOT))
Chris Masonaf31f5e2011-11-03 15:17:42 -04003220 goto fail_tree_roots;
3221
3222 free_root_pointers(fs_info, 0);
3223
3224 /* don't use the log in recovery mode, it won't be valid */
3225 btrfs_set_super_log_root(disk_super, 0);
3226
3227 /* we can't trust the free space cache either */
3228 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3229
3230 ret = next_root_backup(fs_info, fs_info->super_copy,
3231 &num_backups_tried, &backup_index);
3232 if (ret == -1)
3233 goto fail_block_groups;
3234 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003235}
3236
Chris Masonf2984462008-04-10 16:19:33 -04003237static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3238{
Chris Masonf2984462008-04-10 16:19:33 -04003239 if (uptodate) {
3240 set_buffer_uptodate(bh);
3241 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003242 struct btrfs_device *device = (struct btrfs_device *)
3243 bh->b_private;
3244
David Sterbab14af3b2015-10-08 10:43:10 +02003245 btrfs_warn_rl_in_rcu(device->dev_root->fs_info,
3246 "lost page write due to IO error on %s",
Josef Bacik606686e2012-06-04 14:03:51 -04003247 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003248 /* note, we dont' set_buffer_write_io_error because we have
3249 * our own ways of dealing with the IO errors
3250 */
Chris Masonf2984462008-04-10 16:19:33 -04003251 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003252 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003253 }
3254 unlock_buffer(bh);
3255 put_bh(bh);
3256}
3257
Anand Jain29c36d72015-08-14 18:32:58 +08003258int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3259 struct buffer_head **bh_ret)
3260{
3261 struct buffer_head *bh;
3262 struct btrfs_super_block *super;
3263 u64 bytenr;
3264
3265 bytenr = btrfs_sb_offset(copy_num);
3266 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3267 return -EINVAL;
3268
3269 bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE);
3270 /*
3271 * If we fail to read from the underlying devices, as of now
3272 * the best option we have is to mark it EIO.
3273 */
3274 if (!bh)
3275 return -EIO;
3276
3277 super = (struct btrfs_super_block *)bh->b_data;
3278 if (btrfs_super_bytenr(super) != bytenr ||
3279 btrfs_super_magic(super) != BTRFS_MAGIC) {
3280 brelse(bh);
3281 return -EINVAL;
3282 }
3283
3284 *bh_ret = bh;
3285 return 0;
3286}
3287
3288
Yan Zhenga512bbf2008-12-08 16:46:26 -05003289struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3290{
3291 struct buffer_head *bh;
3292 struct buffer_head *latest = NULL;
3293 struct btrfs_super_block *super;
3294 int i;
3295 u64 transid = 0;
Anand Jain92fc03f2015-08-14 18:32:51 +08003296 int ret = -EINVAL;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003297
3298 /* we would like to check all the supers, but that would make
3299 * a btrfs mount succeed after a mkfs from a different FS.
3300 * So, we need to add a special mount option to scan for
3301 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3302 */
3303 for (i = 0; i < 1; i++) {
Anand Jain29c36d72015-08-14 18:32:58 +08003304 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3305 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003306 continue;
3307
3308 super = (struct btrfs_super_block *)bh->b_data;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003309
3310 if (!latest || btrfs_super_generation(super) > transid) {
3311 brelse(latest);
3312 latest = bh;
3313 transid = btrfs_super_generation(super);
3314 } else {
3315 brelse(bh);
3316 }
3317 }
Anand Jain92fc03f2015-08-14 18:32:51 +08003318
3319 if (!latest)
3320 return ERR_PTR(ret);
3321
Yan Zhenga512bbf2008-12-08 16:46:26 -05003322 return latest;
3323}
3324
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003325/*
3326 * this should be called twice, once with wait == 0 and
3327 * once with wait == 1. When wait == 0 is done, all the buffer heads
3328 * we write are pinned.
3329 *
3330 * They are released when wait == 1 is done.
3331 * max_mirrors must be the same for both runs, and it indicates how
3332 * many supers on this one device should be written.
3333 *
3334 * max_mirrors == 0 means to write them all.
3335 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003336static int write_dev_supers(struct btrfs_device *device,
3337 struct btrfs_super_block *sb,
3338 int do_barriers, int wait, int max_mirrors)
3339{
3340 struct buffer_head *bh;
3341 int i;
3342 int ret;
3343 int errors = 0;
3344 u32 crc;
3345 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003346
3347 if (max_mirrors == 0)
3348 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3349
Yan Zhenga512bbf2008-12-08 16:46:26 -05003350 for (i = 0; i < max_mirrors; i++) {
3351 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003352 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3353 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003354 break;
3355
3356 if (wait) {
3357 bh = __find_get_block(device->bdev, bytenr / 4096,
3358 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003359 if (!bh) {
3360 errors++;
3361 continue;
3362 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003363 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003364 if (!buffer_uptodate(bh))
3365 errors++;
3366
3367 /* drop our reference */
3368 brelse(bh);
3369
3370 /* drop the reference from the wait == 0 run */
3371 brelse(bh);
3372 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003373 } else {
3374 btrfs_set_super_bytenr(sb, bytenr);
3375
3376 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003377 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003378 BTRFS_CSUM_SIZE, crc,
3379 BTRFS_SUPER_INFO_SIZE -
3380 BTRFS_CSUM_SIZE);
3381 btrfs_csum_final(crc, sb->csum);
3382
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003383 /*
3384 * one reference for us, and we leave it for the
3385 * caller
3386 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003387 bh = __getblk(device->bdev, bytenr / 4096,
3388 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003389 if (!bh) {
David Sterbaf14d1042015-10-08 11:37:06 +02003390 btrfs_err(device->dev_root->fs_info,
3391 "couldn't get super buffer head for bytenr %llu",
3392 bytenr);
Josef Bacik634554d2013-04-29 10:05:57 -04003393 errors++;
3394 continue;
3395 }
3396
Yan Zhenga512bbf2008-12-08 16:46:26 -05003397 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3398
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003399 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003400 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003401
3402 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003403 lock_buffer(bh);
3404 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003405 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003406 }
3407
Chris Mason387125f2011-11-18 15:07:51 -05003408 /*
3409 * we fua the first super. The others we allow
3410 * to go down lazy.
3411 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003412 if (i == 0)
3413 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3414 else
3415 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003416 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003417 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003418 }
3419 return errors < i ? 0 : -1;
3420}
3421
Chris Mason387125f2011-11-18 15:07:51 -05003422/*
3423 * endio for the write_dev_flush, this will wake anyone waiting
3424 * for the barrier when it is done
3425 */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003426static void btrfs_end_empty_barrier(struct bio *bio)
Chris Mason387125f2011-11-18 15:07:51 -05003427{
Chris Mason387125f2011-11-18 15:07:51 -05003428 if (bio->bi_private)
3429 complete(bio->bi_private);
3430 bio_put(bio);
3431}
3432
3433/*
3434 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3435 * sent down. With wait == 1, it waits for the previous flush.
3436 *
3437 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3438 * capable
3439 */
3440static int write_dev_flush(struct btrfs_device *device, int wait)
3441{
3442 struct bio *bio;
3443 int ret = 0;
3444
3445 if (device->nobarriers)
3446 return 0;
3447
3448 if (wait) {
3449 bio = device->flush_bio;
3450 if (!bio)
3451 return 0;
3452
3453 wait_for_completion(&device->flush_wait);
3454
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003455 if (bio->bi_error) {
3456 ret = bio->bi_error;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003457 btrfs_dev_stat_inc_and_print(device,
3458 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003459 }
3460
3461 /* drop the reference from the wait == 0 run */
3462 bio_put(bio);
3463 device->flush_bio = NULL;
3464
3465 return ret;
3466 }
3467
3468 /*
3469 * one reference for us, and we leave it for the
3470 * caller
3471 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003472 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003473 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003474 if (!bio)
3475 return -ENOMEM;
3476
3477 bio->bi_end_io = btrfs_end_empty_barrier;
3478 bio->bi_bdev = device->bdev;
3479 init_completion(&device->flush_wait);
3480 bio->bi_private = &device->flush_wait;
3481 device->flush_bio = bio;
3482
3483 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003484 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003485
3486 return 0;
3487}
3488
3489/*
3490 * send an empty flush down to each device in parallel,
3491 * then wait for them
3492 */
3493static int barrier_all_devices(struct btrfs_fs_info *info)
3494{
3495 struct list_head *head;
3496 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003497 int errors_send = 0;
3498 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003499 int ret;
3500
3501 /* send down all the barriers */
3502 head = &info->fs_devices->devices;
3503 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003504 if (dev->missing)
3505 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003506 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003507 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003508 continue;
3509 }
3510 if (!dev->in_fs_metadata || !dev->writeable)
3511 continue;
3512
3513 ret = write_dev_flush(dev, 0);
3514 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003515 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003516 }
3517
3518 /* wait for all the barriers */
3519 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003520 if (dev->missing)
3521 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003522 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003523 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003524 continue;
3525 }
3526 if (!dev->in_fs_metadata || !dev->writeable)
3527 continue;
3528
3529 ret = write_dev_flush(dev, 1);
3530 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003531 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003532 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003533 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3534 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003535 return -EIO;
3536 return 0;
3537}
3538
Zhao Lei943c6e92015-08-19 15:54:15 +08003539int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3540{
Zhao Lei8789f4f2015-09-15 21:08:07 +08003541 int raid_type;
3542 int min_tolerated = INT_MAX;
Zhao Lei943c6e92015-08-19 15:54:15 +08003543
Zhao Lei8789f4f2015-09-15 21:08:07 +08003544 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3545 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3546 min_tolerated = min(min_tolerated,
3547 btrfs_raid_array[BTRFS_RAID_SINGLE].
3548 tolerated_failures);
Zhao Lei943c6e92015-08-19 15:54:15 +08003549
Zhao Lei8789f4f2015-09-15 21:08:07 +08003550 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3551 if (raid_type == BTRFS_RAID_SINGLE)
3552 continue;
3553 if (!(flags & btrfs_raid_group[raid_type]))
3554 continue;
3555 min_tolerated = min(min_tolerated,
3556 btrfs_raid_array[raid_type].
3557 tolerated_failures);
3558 }
Zhao Lei943c6e92015-08-19 15:54:15 +08003559
Zhao Lei8789f4f2015-09-15 21:08:07 +08003560 if (min_tolerated == INT_MAX) {
3561 pr_warn("BTRFS: unknown raid flag: %llu\n", flags);
3562 min_tolerated = 0;
3563 }
3564
3565 return min_tolerated;
Zhao Lei943c6e92015-08-19 15:54:15 +08003566}
3567
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003568int btrfs_calc_num_tolerated_disk_barrier_failures(
3569 struct btrfs_fs_info *fs_info)
3570{
3571 struct btrfs_ioctl_space_info space;
3572 struct btrfs_space_info *sinfo;
3573 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3574 BTRFS_BLOCK_GROUP_SYSTEM,
3575 BTRFS_BLOCK_GROUP_METADATA,
3576 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003577 int i;
3578 int c;
3579 int num_tolerated_disk_barrier_failures =
3580 (int)fs_info->fs_devices->num_devices;
3581
Zhao Lei2c458042015-07-16 15:00:46 +08003582 for (i = 0; i < ARRAY_SIZE(types); i++) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003583 struct btrfs_space_info *tmp;
3584
3585 sinfo = NULL;
3586 rcu_read_lock();
3587 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3588 if (tmp->flags == types[i]) {
3589 sinfo = tmp;
3590 break;
3591 }
3592 }
3593 rcu_read_unlock();
3594
3595 if (!sinfo)
3596 continue;
3597
3598 down_read(&sinfo->groups_sem);
3599 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
Zhao Lei2c458042015-07-16 15:00:46 +08003600 u64 flags;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003601
Zhao Lei2c458042015-07-16 15:00:46 +08003602 if (list_empty(&sinfo->block_groups[c]))
3603 continue;
3604
3605 btrfs_get_block_group_info(&sinfo->block_groups[c],
3606 &space);
3607 if (space.total_bytes == 0 || space.used_bytes == 0)
3608 continue;
3609 flags = space.flags;
Zhao Lei943c6e92015-08-19 15:54:15 +08003610
3611 num_tolerated_disk_barrier_failures = min(
3612 num_tolerated_disk_barrier_failures,
3613 btrfs_get_num_tolerated_disk_barrier_failures(
3614 flags));
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003615 }
3616 up_read(&sinfo->groups_sem);
3617 }
3618
3619 return num_tolerated_disk_barrier_failures;
3620}
3621
Eric Sandeen48a3b632013-04-25 20:41:01 +00003622static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003623{
Chris Masone5e9a522009-06-10 15:17:02 -04003624 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003625 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003626 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003627 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003628 int ret;
3629 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003630 int max_errors;
3631 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003632 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003633
3634 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003635 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003636
David Sterba6c417612011-04-13 15:41:04 +02003637 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003638 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003639
Chris Mason174ba502011-05-27 10:03:58 -04003640 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003641 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003642 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003643
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003644 if (do_barriers) {
3645 ret = barrier_all_devices(root->fs_info);
3646 if (ret) {
3647 mutex_unlock(
3648 &root->fs_info->fs_devices->device_list_mutex);
Anand Jaina4553fe2015-09-25 14:43:01 +08003649 btrfs_std_error(root->fs_info, ret,
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003650 "errors while submitting device barriers.");
3651 return ret;
3652 }
3653 }
Chris Mason387125f2011-11-18 15:07:51 -05003654
Xiao Guangrong1f781602011-04-20 10:09:16 +00003655 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003656 if (!dev->bdev) {
3657 total_errors++;
3658 continue;
3659 }
Yan Zheng2b820322008-11-17 21:11:30 -05003660 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003661 continue;
3662
Yan Zheng2b820322008-11-17 21:11:30 -05003663 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003664 btrfs_set_stack_device_type(dev_item, dev->type);
3665 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003666 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003667 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003668 btrfs_set_stack_device_bytes_used(dev_item,
3669 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003670 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3671 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3672 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3673 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003674 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003675
Chris Masona061fc82008-05-07 11:43:44 -04003676 flags = btrfs_super_flags(sb);
3677 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003678
Yan Zhenga512bbf2008-12-08 16:46:26 -05003679 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003680 if (ret)
3681 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003682 }
Chris Masona236aed2008-04-29 09:38:00 -04003683 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003684 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003685 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003686 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003687
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003688 /* FUA is masked off if unsupported and can't be the reason */
Anand Jaina4553fe2015-09-25 14:43:01 +08003689 btrfs_std_error(root->fs_info, -EIO,
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003690 "%d errors while writing supers", total_errors);
3691 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003692 }
Chris Masonf2984462008-04-10 16:19:33 -04003693
Yan Zhenga512bbf2008-12-08 16:46:26 -05003694 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003695 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003696 if (!dev->bdev)
3697 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003698 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003699 continue;
3700
Yan Zhenga512bbf2008-12-08 16:46:26 -05003701 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3702 if (ret)
3703 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003704 }
Chris Mason174ba502011-05-27 10:03:58 -04003705 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003706 if (total_errors > max_errors) {
Anand Jaina4553fe2015-09-25 14:43:01 +08003707 btrfs_std_error(root->fs_info, -EIO,
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003708 "%d errors while writing supers", total_errors);
3709 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003710 }
Chris Masonf2984462008-04-10 16:19:33 -04003711 return 0;
3712}
3713
Yan Zhenga512bbf2008-12-08 16:46:26 -05003714int write_ctree_super(struct btrfs_trans_handle *trans,
3715 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003716{
Rashikaf570e7572013-10-31 16:50:42 +05303717 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003718}
3719
Miao Xiecb517ea2013-05-15 07:48:19 +00003720/* Drop a fs root from the radix tree and free it. */
3721void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3722 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003723{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003724 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003725 radix_tree_delete(&fs_info->fs_roots_radix,
3726 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003727 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003728
3729 if (btrfs_root_refs(&root->root_item) == 0)
3730 synchronize_srcu(&fs_info->subvol_srcu);
3731
Liu Bo1a4319c2014-01-13 19:53:53 +08003732 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003733 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003734
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003735 if (root->free_ino_pinned)
3736 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3737 if (root->free_ino_ctl)
3738 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003739 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003740}
3741
3742static void free_fs_root(struct btrfs_root *root)
3743{
David Sterba57cdc8d2014-02-05 02:37:48 +01003744 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003745 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003746 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3747 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003748 if (root->anon_dev)
3749 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003750 if (root->subv_writers)
3751 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003752 free_extent_buffer(root->node);
3753 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003754 kfree(root->free_ino_ctl);
3755 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003756 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003757 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003758}
3759
Miao Xiecb517ea2013-05-15 07:48:19 +00003760void btrfs_free_fs_root(struct btrfs_root *root)
3761{
3762 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003763}
3764
Yan Zhengc146afa2008-11-12 14:34:12 -05003765int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3766{
3767 u64 root_objectid = 0;
3768 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003769 int i = 0;
3770 int err = 0;
3771 unsigned int ret = 0;
3772 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003773
3774 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003775 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003776 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3777 (void **)gang, root_objectid,
3778 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003779 if (!ret) {
3780 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003781 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003782 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003783 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003784
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003785 for (i = 0; i < ret; i++) {
3786 /* Avoid to grab roots in dead_roots */
3787 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3788 gang[i] = NULL;
3789 continue;
3790 }
3791 /* grab all the search result for later use */
3792 gang[i] = btrfs_grab_fs_root(gang[i]);
3793 }
3794 srcu_read_unlock(&fs_info->subvol_srcu, index);
3795
3796 for (i = 0; i < ret; i++) {
3797 if (!gang[i])
3798 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003799 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003800 err = btrfs_orphan_cleanup(gang[i]);
3801 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003802 break;
3803 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003804 }
3805 root_objectid++;
3806 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003807
3808 /* release the uncleaned roots due to error */
3809 for (; i < ret; i++) {
3810 if (gang[i])
3811 btrfs_put_fs_root(gang[i]);
3812 }
3813 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003814}
3815
3816int btrfs_commit_super(struct btrfs_root *root)
3817{
3818 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003819
3820 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003821 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003822 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003823 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003824
3825 /* wait until ongoing cleanup work done */
3826 down_write(&root->fs_info->cleanup_work_sem);
3827 up_write(&root->fs_info->cleanup_work_sem);
3828
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003829 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003830 if (IS_ERR(trans))
3831 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003832 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003833}
3834
David Sterba3abdbd72014-06-04 18:10:45 +02003835void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003836{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003837 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003838 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003839
Chris Masonfacda1e2007-06-08 18:11:48 -04003840 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003841 smp_mb();
3842
Justin Maggard7343dd62015-11-04 15:56:16 -08003843 /* wait for the qgroup rescan worker to stop */
3844 btrfs_qgroup_wait_for_completion(fs_info);
3845
Stefan Behrens803b2f52013-08-15 17:11:21 +02003846 /* wait for the uuid_scan task to finish */
3847 down(&fs_info->uuid_tree_rescan_sem);
3848 /* avoid complains from lockdep et al., set sem back to initial state */
3849 up(&fs_info->uuid_tree_rescan_sem);
3850
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003851 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003852 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003853
Stefan Behrens8dabb742012-11-06 13:15:27 +01003854 btrfs_dev_replace_suspend_for_unmount(fs_info);
3855
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003856 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003857
3858 /* wait for any defraggers to finish */
3859 wait_event(fs_info->transaction_wait,
3860 (atomic_read(&fs_info->defrag_running) == 0));
3861
3862 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003863 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003864
Miao Xie21c7e752014-05-13 17:29:04 -07003865 cancel_work_sync(&fs_info->async_reclaim_work);
3866
Yan Zhengc146afa2008-11-12 14:34:12 -05003867 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003868 /*
3869 * If the cleaner thread is stopped and there are
3870 * block groups queued for removal, the deletion will be
3871 * skipped when we quit the cleaner thread.
3872 */
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003873 btrfs_delete_unused_bgs(root->fs_info);
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003874
liuboacce9522011-01-06 19:30:25 +08003875 ret = btrfs_commit_super(root);
3876 if (ret)
Eric Sandeen04892342014-08-01 18:12:36 -05003877 btrfs_err(fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003878 }
3879
Miao Xie87533c42013-01-29 10:14:48 +00003880 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003881 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003882
Al Viroe3029d92011-11-17 00:56:18 -05003883 kthread_stop(fs_info->transaction_kthread);
3884 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003885
Yan Zhengf25784b2009-07-28 08:41:57 -04003886 fs_info->closing = 2;
3887 smp_mb();
3888
Eric Sandeen04892342014-08-01 18:12:36 -05003889 btrfs_free_qgroup_config(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003890
Miao Xie963d6782013-01-29 10:10:51 +00003891 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Eric Sandeen04892342014-08-01 18:12:36 -05003892 btrfs_info(fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003893 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003894 }
Yanbcc63ab2008-07-30 16:29:20 -04003895
Anand Jain6618a592015-08-14 18:32:47 +08003896 btrfs_sysfs_remove_mounted(fs_info);
Anand Jainb7c35e82015-03-10 06:38:38 +08003897 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003898
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003899 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003900
Liu Bo1a4319c2014-01-13 19:53:53 +08003901 btrfs_put_block_group_cache(fs_info);
3902
Josef Bacik2b1360d2013-10-07 15:14:44 -04003903 btrfs_free_block_groups(fs_info);
3904
Wang Shilongde348ee2014-04-09 19:23:22 +08003905 /*
3906 * we must make sure there is not any read request to
3907 * submit after we stopping all workers.
3908 */
3909 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003910 btrfs_stop_all_workers(fs_info);
3911
Josef Bacik47ab2a62014-09-18 11:20:02 -04003912 fs_info->open = 0;
Josef Bacik13e6c372013-05-30 16:55:44 -04003913 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003914
Josef Bacik13e6c372013-05-30 16:55:44 -04003915 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003916
Stefan Behrens21adbd52011-11-09 13:44:05 +01003917#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3918 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3919 btrfsic_unmount(root, fs_info->fs_devices);
3920#endif
3921
Chris Masondfe25022008-05-13 13:46:40 -04003922 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003923 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003924
Miao Xiee2d84522013-01-29 10:09:20 +00003925 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003926 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003927 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003928 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003929 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003930
David Woodhouse53b381b2013-01-29 18:40:14 -05003931 btrfs_free_stripe_hash_table(fs_info);
3932
Chris Masoncdfb0802015-04-06 18:17:00 -07003933 __btrfs_free_block_rsv(root->orphan_block_rsv);
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003934 root->orphan_block_rsv = NULL;
Filipe Manana04216822014-11-27 21:14:15 +00003935
3936 lock_chunks(root);
3937 while (!list_empty(&fs_info->pinned_chunks)) {
3938 struct extent_map *em;
3939
3940 em = list_first_entry(&fs_info->pinned_chunks,
3941 struct extent_map, list);
3942 list_del_init(&em->list);
3943 free_extent_map(em);
3944 }
3945 unlock_chunks(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003946}
3947
Chris Masonb9fab912012-05-06 07:23:47 -04003948int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3949 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003950{
Chris Mason1259ab72008-05-12 13:39:03 -04003951 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003952 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003953
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003954 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003955 if (!ret)
3956 return ret;
3957
3958 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003959 parent_transid, atomic);
3960 if (ret == -EAGAIN)
3961 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003962 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003963}
Chris Mason6702ed42007-08-07 16:15:09 -04003964
Chris Mason5f39d392007-10-15 16:14:19 -04003965void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3966{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003967 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003968 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003969 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003970
Josef Bacik06ea65a2013-09-19 16:07:01 -04003971#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3972 /*
3973 * This is a fast path so only do this check if we have sanity tests
3974 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3975 * outside of the sanity tests.
3976 */
3977 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3978 return;
3979#endif
3980 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003981 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003982 if (transid != root->fs_info->generation)
3983 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003984 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003985 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003986 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003987 if (!was_dirty)
3988 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3989 buf->len,
3990 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003991#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3992 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3993 btrfs_print_leaf(root, buf);
3994 ASSERT(0);
3995 }
3996#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05003997}
3998
Liu Bob53d3f52012-11-14 14:34:34 +00003999static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
4000 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04004001{
Chris Mason188de642008-05-09 11:52:25 -04004002 /*
4003 * looks as though older kernels can get into trouble with
4004 * this code, they end up stuck in balance_dirty_pages forever
4005 */
Miao Xiee2d84522013-01-29 10:09:20 +00004006 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04004007
Jens Axboe6933c022009-03-17 09:36:37 +01004008 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04004009 return;
4010
Liu Bob53d3f52012-11-14 14:34:34 +00004011 if (flush_delayed)
4012 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08004013
Miao Xiee2d84522013-01-29 10:09:20 +00004014 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
4015 BTRFS_DIRTY_METADATA_THRESH);
4016 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08004017 balance_dirty_pages_ratelimited(
4018 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08004019 }
Miao Xie16cdcec2011-04-22 18:12:22 +08004020}
4021
Liu Bob53d3f52012-11-14 14:34:34 +00004022void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08004023{
Liu Bob53d3f52012-11-14 14:34:34 +00004024 __btrfs_btree_balance_dirty(root, 1);
4025}
Miao Xie16cdcec2011-04-22 18:12:22 +08004026
Liu Bob53d3f52012-11-14 14:34:34 +00004027void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
4028{
4029 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04004030}
Chris Mason6b800532007-10-15 16:17:34 -04004031
Chris Masonca7a79a2008-05-12 12:59:19 -04004032int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04004033{
Chris Mason727011e2010-08-06 13:21:20 -04004034 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04004035 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04004036}
Chris Mason0da54682007-11-07 21:08:01 -05004037
David Sterbafcd1f0652012-03-06 00:06:18 +01004038static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08004039 int read_only)
4040{
David Sterbac9260932014-09-30 19:16:47 +02004041 struct btrfs_super_block *sb = fs_info->super_copy;
Qu Wenruo319e4d02015-12-15 09:14:36 +08004042 u64 nodesize = btrfs_super_nodesize(sb);
4043 u64 sectorsize = btrfs_super_sectorsize(sb);
David Sterbac9260932014-09-30 19:16:47 +02004044 int ret = 0;
4045
Qu Wenruo319e4d02015-12-15 09:14:36 +08004046 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
4047 printk(KERN_ERR "BTRFS: no valid FS found\n");
4048 ret = -EINVAL;
4049 }
4050 if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP)
4051 printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n",
4052 btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP);
David Sterba21e76262014-10-27 13:52:21 +01004053 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
4054 printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n",
4055 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004056 ret = -EINVAL;
4057 }
David Sterba21e76262014-10-27 13:52:21 +01004058 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
4059 printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n",
4060 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004061 ret = -EINVAL;
4062 }
David Sterba21e76262014-10-27 13:52:21 +01004063 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
4064 printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n",
4065 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004066 ret = -EINVAL;
4067 }
4068
David Sterba1104a882013-03-06 15:57:46 +01004069 /*
Qu Wenruo319e4d02015-12-15 09:14:36 +08004070 * Check sectorsize and nodesize first, other check will need it.
4071 * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here.
David Sterba1104a882013-03-06 15:57:46 +01004072 */
Qu Wenruo319e4d02015-12-15 09:14:36 +08004073 if (!is_power_of_2(sectorsize) || sectorsize < 4096 ||
4074 sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) {
4075 printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize);
David Sterba75d6ad32014-10-31 17:18:08 +01004076 ret = -EINVAL;
4077 }
Qu Wenruo319e4d02015-12-15 09:14:36 +08004078 /* Only PAGE SIZE is supported yet */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03004079 if (sectorsize != PAGE_SIZE) {
Qu Wenruo319e4d02015-12-15 09:14:36 +08004080 printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n",
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03004081 sectorsize, PAGE_SIZE);
Qu Wenruo319e4d02015-12-15 09:14:36 +08004082 ret = -EINVAL;
4083 }
4084 if (!is_power_of_2(nodesize) || nodesize < sectorsize ||
4085 nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
4086 printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize);
4087 ret = -EINVAL;
4088 }
4089 if (nodesize != le32_to_cpu(sb->__unused_leafsize)) {
4090 printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n",
4091 le32_to_cpu(sb->__unused_leafsize),
4092 nodesize);
4093 ret = -EINVAL;
4094 }
4095
4096 /* Root alignment check */
4097 if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) {
4098 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
4099 btrfs_super_root(sb));
4100 ret = -EINVAL;
4101 }
4102 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) {
4103 printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n",
4104 btrfs_super_chunk_root(sb));
4105 ret = -EINVAL;
4106 }
4107 if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) {
4108 printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
4109 btrfs_super_log_root(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004110 ret = -EINVAL;
4111 }
4112
David Sterbac9260932014-09-30 19:16:47 +02004113 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
4114 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
4115 fs_info->fsid, sb->dev_item.fsid);
4116 ret = -EINVAL;
4117 }
4118
4119 /*
4120 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4121 * done later
4122 */
David Sterba21e76262014-10-27 13:52:21 +01004123 if (btrfs_super_num_devices(sb) > (1UL << 31))
David Sterbac9260932014-09-30 19:16:47 +02004124 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004125 btrfs_super_num_devices(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004126 if (btrfs_super_num_devices(sb) == 0) {
4127 printk(KERN_ERR "BTRFS: number of devices is 0\n");
4128 ret = -EINVAL;
4129 }
David Sterbac9260932014-09-30 19:16:47 +02004130
David Sterba21e76262014-10-27 13:52:21 +01004131 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
David Sterbac9260932014-09-30 19:16:47 +02004132 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
David Sterba21e76262014-10-27 13:52:21 +01004133 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
David Sterbac9260932014-09-30 19:16:47 +02004134 ret = -EINVAL;
4135 }
4136
4137 /*
David Sterbace7fca52014-10-31 18:42:05 +01004138 * Obvious sys_chunk_array corruptions, it must hold at least one key
4139 * and one chunk
4140 */
4141 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
4142 printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
4143 btrfs_super_sys_array_size(sb),
4144 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
4145 ret = -EINVAL;
4146 }
4147 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4148 + sizeof(struct btrfs_chunk)) {
Fabian Frederickd2207122015-02-14 13:10:22 +01004149 printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n",
David Sterbace7fca52014-10-31 18:42:05 +01004150 btrfs_super_sys_array_size(sb),
4151 sizeof(struct btrfs_disk_key)
4152 + sizeof(struct btrfs_chunk));
4153 ret = -EINVAL;
4154 }
4155
4156 /*
David Sterbac9260932014-09-30 19:16:47 +02004157 * The generation is a global counter, we'll trust it more than the others
4158 * but it's still possible that it's the one that's wrong.
4159 */
David Sterba21e76262014-10-27 13:52:21 +01004160 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
David Sterbac9260932014-09-30 19:16:47 +02004161 printk(KERN_WARNING
4162 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004163 btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb));
4164 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4165 && btrfs_super_cache_generation(sb) != (u64)-1)
David Sterbac9260932014-09-30 19:16:47 +02004166 printk(KERN_WARNING
4167 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004168 btrfs_super_generation(sb), btrfs_super_cache_generation(sb));
David Sterbac9260932014-09-30 19:16:47 +02004169
4170 return ret;
liuboacce9522011-01-06 19:30:25 +08004171}
4172
Eric Sandeen48a3b632013-04-25 20:41:01 +00004173static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004174{
liuboacce9522011-01-06 19:30:25 +08004175 mutex_lock(&root->fs_info->cleaner_mutex);
4176 btrfs_run_delayed_iputs(root);
4177 mutex_unlock(&root->fs_info->cleaner_mutex);
4178
4179 down_write(&root->fs_info->cleanup_work_sem);
4180 up_write(&root->fs_info->cleanup_work_sem);
4181
4182 /* cleanup FS via transaction */
4183 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08004184}
4185
Jeff Mahoney143bede2012-03-01 14:56:26 +01004186static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004187{
liuboacce9522011-01-06 19:30:25 +08004188 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08004189
Miao Xie199c2a92013-05-15 07:48:23 +00004190 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05004191 /*
4192 * This will just short circuit the ordered completion stuff which will
4193 * make sure the ordered extent gets properly cleaned up.
4194 */
Miao Xie199c2a92013-05-15 07:48:23 +00004195 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05004196 root_extent_list)
4197 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00004198 spin_unlock(&root->ordered_extent_lock);
4199}
4200
4201static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4202{
4203 struct btrfs_root *root;
4204 struct list_head splice;
4205
4206 INIT_LIST_HEAD(&splice);
4207
4208 spin_lock(&fs_info->ordered_root_lock);
4209 list_splice_init(&fs_info->ordered_roots, &splice);
4210 while (!list_empty(&splice)) {
4211 root = list_first_entry(&splice, struct btrfs_root,
4212 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04004213 list_move_tail(&root->ordered_root,
4214 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00004215
Liu Bo2a85d9c2014-02-10 17:07:16 +08004216 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004217 btrfs_destroy_ordered_extents(root);
4218
Liu Bo2a85d9c2014-02-10 17:07:16 +08004219 cond_resched();
4220 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004221 }
4222 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08004223}
4224
Stefan Behrens35a36212013-08-14 18:12:25 +02004225static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4226 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004227{
4228 struct rb_node *node;
4229 struct btrfs_delayed_ref_root *delayed_refs;
4230 struct btrfs_delayed_ref_node *ref;
4231 int ret = 0;
4232
4233 delayed_refs = &trans->delayed_refs;
4234
4235 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004236 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04004237 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05004238 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08004239 return ret;
4240 }
4241
Josef Bacikd7df2c72014-01-23 09:21:38 -05004242 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4243 struct btrfs_delayed_ref_head *head;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004244 struct btrfs_delayed_ref_node *tmp;
Josef Bacike78417d2013-06-03 16:42:36 -04004245 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08004246
Josef Bacikd7df2c72014-01-23 09:21:38 -05004247 head = rb_entry(node, struct btrfs_delayed_ref_head,
4248 href_node);
4249 if (!mutex_trylock(&head->mutex)) {
4250 atomic_inc(&head->node.refs);
4251 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08004252
Josef Bacikd7df2c72014-01-23 09:21:38 -05004253 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04004254 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004255 btrfs_put_delayed_ref(&head->node);
4256 spin_lock(&delayed_refs->lock);
4257 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04004258 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05004259 spin_lock(&head->lock);
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004260 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4261 list) {
Josef Bacikd7df2c72014-01-23 09:21:38 -05004262 ref->in_tree = 0;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004263 list_del(&ref->list);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004264 atomic_dec(&delayed_refs->num_entries);
4265 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004266 }
4267 if (head->must_insert_reserved)
4268 pin_bytes = true;
4269 btrfs_free_delayed_extent_op(head->extent_op);
4270 delayed_refs->num_heads--;
4271 if (head->processing == 0)
4272 delayed_refs->num_heads_ready--;
4273 atomic_dec(&delayed_refs->num_entries);
4274 head->node.in_tree = 0;
4275 rb_erase(&head->href_node, &delayed_refs->href_root);
4276 spin_unlock(&head->lock);
4277 spin_unlock(&delayed_refs->lock);
4278 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08004279
Josef Bacikd7df2c72014-01-23 09:21:38 -05004280 if (pin_bytes)
4281 btrfs_pin_extent(root, head->node.bytenr,
4282 head->node.num_bytes, 1);
4283 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08004284 cond_resched();
4285 spin_lock(&delayed_refs->lock);
4286 }
4287
4288 spin_unlock(&delayed_refs->lock);
4289
4290 return ret;
4291}
4292
Jeff Mahoney143bede2012-03-01 14:56:26 +01004293static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004294{
4295 struct btrfs_inode *btrfs_inode;
4296 struct list_head splice;
4297
4298 INIT_LIST_HEAD(&splice);
4299
Miao Xieeb73c1b2013-05-15 07:48:22 +00004300 spin_lock(&root->delalloc_lock);
4301 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08004302
4303 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00004304 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4305 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08004306
4307 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00004308 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4309 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00004310 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004311
4312 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00004313
Miao Xieeb73c1b2013-05-15 07:48:22 +00004314 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004315 }
4316
Miao Xieeb73c1b2013-05-15 07:48:22 +00004317 spin_unlock(&root->delalloc_lock);
4318}
4319
4320static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4321{
4322 struct btrfs_root *root;
4323 struct list_head splice;
4324
4325 INIT_LIST_HEAD(&splice);
4326
4327 spin_lock(&fs_info->delalloc_root_lock);
4328 list_splice_init(&fs_info->delalloc_roots, &splice);
4329 while (!list_empty(&splice)) {
4330 root = list_first_entry(&splice, struct btrfs_root,
4331 delalloc_root);
4332 list_del_init(&root->delalloc_root);
4333 root = btrfs_grab_fs_root(root);
4334 BUG_ON(!root);
4335 spin_unlock(&fs_info->delalloc_root_lock);
4336
4337 btrfs_destroy_delalloc_inodes(root);
4338 btrfs_put_fs_root(root);
4339
4340 spin_lock(&fs_info->delalloc_root_lock);
4341 }
4342 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004343}
4344
4345static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4346 struct extent_io_tree *dirty_pages,
4347 int mark)
4348{
4349 int ret;
liuboacce9522011-01-06 19:30:25 +08004350 struct extent_buffer *eb;
4351 u64 start = 0;
4352 u64 end;
liuboacce9522011-01-06 19:30:25 +08004353
4354 while (1) {
4355 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004356 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004357 if (ret)
4358 break;
4359
4360 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4361 while (start <= end) {
Daniel Dressler01d58472014-11-21 17:15:07 +09004362 eb = btrfs_find_tree_block(root->fs_info, start);
David Sterba707e8a02014-06-04 19:22:26 +02004363 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004364 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004365 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004366 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004367
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004368 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4369 &eb->bflags))
4370 clear_extent_buffer_dirty(eb);
4371 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004372 }
4373 }
4374
4375 return ret;
4376}
4377
4378static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4379 struct extent_io_tree *pinned_extents)
4380{
4381 struct extent_io_tree *unpin;
4382 u64 start;
4383 u64 end;
4384 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004385 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004386
4387 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004388again:
liuboacce9522011-01-06 19:30:25 +08004389 while (1) {
4390 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004391 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004392 if (ret)
4393 break;
4394
liuboacce9522011-01-06 19:30:25 +08004395 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4396 btrfs_error_unpin_extent_range(root, start, end);
4397 cond_resched();
4398 }
4399
Liu Boed0eaa12012-06-14 02:23:21 -06004400 if (loop) {
4401 if (unpin == &root->fs_info->freed_extents[0])
4402 unpin = &root->fs_info->freed_extents[1];
4403 else
4404 unpin = &root->fs_info->freed_extents[0];
4405 loop = false;
4406 goto again;
4407 }
4408
liuboacce9522011-01-06 19:30:25 +08004409 return 0;
4410}
4411
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004412void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4413 struct btrfs_root *root)
4414{
4415 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004416
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004417 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004418 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004419
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004420 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004421 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004422
Miao Xie67cde342012-06-14 02:23:22 -06004423 btrfs_destroy_delayed_inodes(root);
4424 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004425
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004426 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4427 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004428 btrfs_destroy_pinned_extent(root,
4429 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004430
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004431 cur_trans->state =TRANS_STATE_COMPLETED;
4432 wake_up(&cur_trans->commit_wait);
4433
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004434 /*
4435 memset(cur_trans, 0, sizeof(*cur_trans));
4436 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4437 */
4438}
4439
Eric Sandeen48a3b632013-04-25 20:41:01 +00004440static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004441{
4442 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004443
liuboacce9522011-01-06 19:30:25 +08004444 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4445
Josef Bacika4abeea2011-04-11 17:25:13 -04004446 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004447 while (!list_empty(&root->fs_info->trans_list)) {
4448 t = list_first_entry(&root->fs_info->trans_list,
4449 struct btrfs_transaction, list);
4450 if (t->state >= TRANS_STATE_COMMIT_START) {
4451 atomic_inc(&t->use_count);
4452 spin_unlock(&root->fs_info->trans_lock);
4453 btrfs_wait_for_commit(root, t->transid);
4454 btrfs_put_transaction(t);
4455 spin_lock(&root->fs_info->trans_lock);
4456 continue;
4457 }
4458 if (t == root->fs_info->running_transaction) {
4459 t->state = TRANS_STATE_COMMIT_DOING;
4460 spin_unlock(&root->fs_info->trans_lock);
4461 /*
4462 * We wait for 0 num_writers since we don't hold a trans
4463 * handle open currently for this transaction.
4464 */
4465 wait_event(t->writer_wait,
4466 atomic_read(&t->num_writers) == 0);
4467 } else {
4468 spin_unlock(&root->fs_info->trans_lock);
4469 }
4470 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004471
Josef Bacik724e2312013-09-30 11:36:38 -04004472 spin_lock(&root->fs_info->trans_lock);
4473 if (t == root->fs_info->running_transaction)
4474 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004475 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004476 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004477
Josef Bacik724e2312013-09-30 11:36:38 -04004478 btrfs_put_transaction(t);
4479 trace_btrfs_transaction_commit(root);
4480 spin_lock(&root->fs_info->trans_lock);
4481 }
4482 spin_unlock(&root->fs_info->trans_lock);
4483 btrfs_destroy_all_ordered_extents(root->fs_info);
4484 btrfs_destroy_delayed_inodes(root);
4485 btrfs_assert_delayed_root_empty(root);
4486 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4487 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004488 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004489
4490 return 0;
4491}
4492
David Sterbae8c9f182015-01-02 18:23:10 +01004493static const struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004494 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004495 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004496 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004497 /* note we're sharing with inode.c for the merge bio hook */
4498 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004499};