blob: 1cd325765aaa8cd49be5100250c00e69561a0126 [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,
Nikolay Borisovfba4b692016-06-23 21:17:08 +0300104 SLAB_MEM_SPREAD,
David Sterba97eb6b62014-07-30 00:55:42 +0200105 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 mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400128 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400129 /*
130 * bio_offset is optional, can be used if the pages in the bio
131 * can't tell us where in the file the bio should go
132 */
133 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400134 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100135 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400136};
137
Chris Mason85d4e462011-07-26 16:11:19 -0400138/*
139 * Lockdep class keys for extent_buffer->lock's in this root. For a given
140 * eb, the lockdep key is determined by the btrfs_root it belongs to and
141 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500142 *
Chris Mason85d4e462011-07-26 16:11:19 -0400143 * Different roots are used for different purposes and may nest inside each
144 * other and they require separate keysets. As lockdep keys should be
145 * static, assign keysets according to the purpose of the root as indicated
146 * by btrfs_root->objectid. This ensures that all special purpose roots
147 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500148 *
Chris Mason85d4e462011-07-26 16:11:19 -0400149 * Lock-nesting across peer nodes is always done with the immediate parent
150 * node locked thus preventing deadlock. As lockdep doesn't know this, use
151 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500152 *
Chris Mason85d4e462011-07-26 16:11:19 -0400153 * The key is set by the readpage_end_io_hook after the buffer has passed
154 * csum validation but before the pages are unlocked. It is also set by
155 * btrfs_init_new_buffer on freshly allocated blocks.
156 *
157 * We also add a check to make sure the highest level of the tree is the
158 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
159 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500160 */
161#ifdef CONFIG_DEBUG_LOCK_ALLOC
162# if BTRFS_MAX_LEVEL != 8
163# error
164# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400165
166static struct btrfs_lockdep_keyset {
167 u64 id; /* root objectid */
168 const char *name_stem; /* lock name stem */
169 char names[BTRFS_MAX_LEVEL + 1][20];
170 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
171} btrfs_lockdep_keysets[] = {
172 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
173 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
174 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
175 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
176 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
177 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000178 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400179 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
180 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
181 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200182 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
David Sterba6b20e0a2016-01-25 16:30:22 +0100183 { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" },
Chris Mason85d4e462011-07-26 16:11:19 -0400184 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500185};
Chris Mason85d4e462011-07-26 16:11:19 -0400186
187void __init btrfs_init_lockdep(void)
188{
189 int i, j;
190
191 /* initialize lockdep class names */
192 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
193 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
194
195 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
196 snprintf(ks->names[j], sizeof(ks->names[j]),
197 "btrfs-%s-%02d", ks->name_stem, j);
198 }
199}
200
201void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
202 int level)
203{
204 struct btrfs_lockdep_keyset *ks;
205
206 BUG_ON(level >= ARRAY_SIZE(ks->keys));
207
208 /* find the matching keyset, id 0 is the default entry */
209 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
210 if (ks->id == objectid)
211 break;
212
213 lockdep_set_class_and_name(&eb->lock,
214 &ks->keys[level], ks->names[level]);
215}
216
Chris Mason4008c042009-02-12 14:09:45 -0500217#endif
218
Chris Masond352ac62008-09-29 15:18:18 -0400219/*
220 * extents on the btree inode are pretty simple, there's one extent
221 * that covers the entire device
222 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500223static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200224 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500225 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400226{
227 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
228 struct extent_map *em;
229 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400230
Chris Mason890871b2009-09-02 16:24:52 -0400231 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500232 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400233 if (em) {
234 em->bdev =
235 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400236 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400237 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400238 }
Chris Mason890871b2009-09-02 16:24:52 -0400239 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400240
David Sterba172ddd62011-04-21 00:48:27 +0200241 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400242 if (!em) {
243 em = ERR_PTR(-ENOMEM);
244 goto out;
245 }
246 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400247 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400248 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400249 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400250 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500251
Chris Mason890871b2009-09-02 16:24:52 -0400252 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400253 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400254 if (ret == -EEXIST) {
255 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400256 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600257 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600258 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400259 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400260 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600261 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400262 }
Chris Mason890871b2009-09-02 16:24:52 -0400263 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400264
Chris Mason5f39d392007-10-15 16:14:19 -0400265out:
266 return em;
267}
268
Liu Bob0496682013-03-14 14:57:45 +0000269u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400270{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000271 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400272}
273
274void btrfs_csum_final(u32 crc, char *result)
275{
David Sterba7e75bf32011-03-18 22:56:43 +0000276 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400277}
278
Chris Masond352ac62008-09-29 15:18:18 -0400279/*
280 * compute the csum for a btree block, and either verify it or write it
281 * into the csum field of the block.
282 */
Daniel Dressler01d58472014-11-21 17:15:07 +0900283static int csum_tree_block(struct btrfs_fs_info *fs_info,
284 struct extent_buffer *buf,
Chris Mason19c00dd2007-10-15 16:19:22 -0400285 int verify)
286{
Daniel Dressler01d58472014-11-21 17:15:07 +0900287 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500288 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400289 unsigned long len;
290 unsigned long cur_len;
291 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400292 char *kaddr;
293 unsigned long map_start;
294 unsigned long map_len;
295 int err;
296 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500297 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400298
299 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500300 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400301 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400302 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500303 if (err)
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200304 return err;
Chris Mason19c00dd2007-10-15 16:19:22 -0400305 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000306 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400307 crc, cur_len);
308 len -= cur_len;
309 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400310 }
Josef Bacik607d4322008-12-02 07:17:45 -0500311 if (csum_size > sizeof(inline_result)) {
David Sterba31e818f2015-02-20 18:00:26 +0100312 result = kzalloc(csum_size, GFP_NOFS);
Josef Bacik607d4322008-12-02 07:17:45 -0500313 if (!result)
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200314 return -ENOMEM;
Josef Bacik607d4322008-12-02 07:17:45 -0500315 } else {
316 result = (char *)&inline_result;
317 }
318
Chris Mason19c00dd2007-10-15 16:19:22 -0400319 btrfs_csum_final(crc, result);
320
321 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500322 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500323 u32 val;
324 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500325 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500326
Josef Bacik607d4322008-12-02 07:17:45 -0500327 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba94647322015-10-08 11:01:36 +0200328 btrfs_warn_rl(fs_info,
Jeff Mahoney5d163e02016-09-20 10:05:00 -0400329 "%s checksum verify failed on %llu wanted %X found %X level %d",
Daniel Dressler01d58472014-11-21 17:15:07 +0900330 fs_info->sb->s_id, buf->start,
Frank Holtonefe120a2013-12-20 11:37:06 -0500331 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500332 if (result != (char *)&inline_result)
333 kfree(result);
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200334 return -EUCLEAN;
Chris Mason19c00dd2007-10-15 16:19:22 -0400335 }
336 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500337 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400338 }
Josef Bacik607d4322008-12-02 07:17:45 -0500339 if (result != (char *)&inline_result)
340 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400341 return 0;
342}
343
Chris Masond352ac62008-09-29 15:18:18 -0400344/*
345 * we can't consider a given block up to date unless the transid of the
346 * block matches the transid in the parent node's pointer. This is how we
347 * detect blocks that either didn't get written at all or got written
348 * in the wrong place.
349 */
Chris Mason1259ab72008-05-12 13:39:03 -0400350static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400351 struct extent_buffer *eb, u64 parent_transid,
352 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400353{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000354 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400355 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200356 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400357
358 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
359 return 0;
360
Chris Masonb9fab912012-05-06 07:23:47 -0400361 if (atomic)
362 return -EAGAIN;
363
Josef Bacika26e8c92014-03-28 17:07:27 -0400364 if (need_lock) {
365 btrfs_tree_read_lock(eb);
366 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
367 }
368
Josef Bacik2ac55d42010-02-03 19:33:23 +0000369 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
David Sterbaff13db42015-12-03 14:30:40 +0100370 &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400371 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400372 btrfs_header_generation(eb) == parent_transid) {
373 ret = 0;
374 goto out;
375 }
David Sterba94647322015-10-08 11:01:36 +0200376 btrfs_err_rl(eb->fs_info,
377 "parent transid verify failed on %llu wanted %llu found %llu",
378 eb->start,
Wang Shilong29549ae2014-07-04 17:59:06 +0800379 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400380 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400381
382 /*
383 * Things reading via commit roots that don't have normal protection,
384 * like send, can have a really old block in cache that may point at a
Nicholas D Steeves01327612016-05-19 21:18:45 -0400385 * block that has been freed and re-allocated. So don't clear uptodate
Josef Bacika26e8c92014-03-28 17:07:27 -0400386 * if we find an eb that is under IO (dirty/writeback) because we could
387 * end up reading in the stale data and then writing it back out and
388 * making everybody very sad.
389 */
390 if (!extent_buffer_under_io(eb))
391 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400392out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000393 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
394 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700395 if (need_lock)
396 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400397 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400398}
399
Chris Masond352ac62008-09-29 15:18:18 -0400400/*
David Sterba1104a882013-03-06 15:57:46 +0100401 * Return 0 if the superblock checksum type matches the checksum value of that
402 * algorithm. Pass the raw disk superblock data.
403 */
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -0400404static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info,
405 char *raw_disk_sb)
David Sterba1104a882013-03-06 15:57:46 +0100406{
407 struct btrfs_super_block *disk_sb =
408 (struct btrfs_super_block *)raw_disk_sb;
409 u16 csum_type = btrfs_super_csum_type(disk_sb);
410 int ret = 0;
411
412 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
413 u32 crc = ~(u32)0;
414 const int csum_size = sizeof(crc);
415 char result[csum_size];
416
417 /*
418 * The super_block structure does not span the whole
419 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
Nicholas D Steeves01327612016-05-19 21:18:45 -0400420 * is filled with zeros and is included in the checksum.
David Sterba1104a882013-03-06 15:57:46 +0100421 */
422 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
423 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
424 btrfs_csum_final(crc, result);
425
426 if (memcmp(raw_disk_sb, result, csum_size))
427 ret = 1;
428 }
429
430 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -0400431 btrfs_err(fs_info, "unsupported checksum algorithm %u",
David Sterba1104a882013-03-06 15:57:46 +0100432 csum_type);
433 ret = 1;
434 }
435
436 return ret;
437}
438
439/*
Chris Masond352ac62008-09-29 15:18:18 -0400440 * helper to read a given tree block, doing retries as required when
441 * the checksums don't match and we have alternate mirrors to try.
442 */
Chris Masonf1885912008-04-09 16:28:12 -0400443static int btree_read_extent_buffer_pages(struct btrfs_root *root,
444 struct extent_buffer *eb,
Josef Bacik8436ea912016-09-02 15:40:03 -0400445 u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400446{
447 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400448 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400449 int ret;
450 int num_copies = 0;
451 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400452 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400453
Josef Bacika826d6d2011-03-16 13:42:43 -0400454 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400455 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
456 while (1) {
Josef Bacik8436ea912016-09-02 15:40:03 -0400457 ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400458 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600459 if (!ret) {
460 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400461 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600462 break;
463 else
464 ret = -EIO;
465 }
Chris Masond3977122009-01-05 21:25:51 -0500466
Josef Bacika826d6d2011-03-16 13:42:43 -0400467 /*
468 * This buffer's crc is fine, but its contents are corrupted, so
469 * there is no reason to read the other copies, they won't be
470 * any less wrong.
471 */
472 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400473 break;
474
Stefan Behrens5d964052012-11-05 14:59:07 +0100475 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400476 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400477 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400478 break;
Chris Mason42352982008-04-28 16:40:52 -0400479
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400480 if (!failed_mirror) {
481 failed = 1;
482 failed_mirror = eb->read_mirror;
483 }
484
Chris Masonf1885912008-04-09 16:28:12 -0400485 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400486 if (mirror_num == failed_mirror)
487 mirror_num++;
488
Chris Mason42352982008-04-28 16:40:52 -0400489 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400490 break;
Chris Masonf1885912008-04-09 16:28:12 -0400491 }
Josef Bacikea466792012-03-26 21:57:36 -0400492
Stefan Behrensc0901582012-07-10 07:30:17 -0600493 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400494 repair_eb_io_failure(root, eb, failed_mirror);
495
496 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400497}
Chris Mason19c00dd2007-10-15 16:19:22 -0400498
Chris Masond352ac62008-09-29 15:18:18 -0400499/*
Chris Masond3977122009-01-05 21:25:51 -0500500 * checksum a dirty tree block before IO. This has extra checks to make sure
501 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400502 */
Chris Masond3977122009-01-05 21:25:51 -0500503
Daniel Dressler01d58472014-11-21 17:15:07 +0900504static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400505{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000506 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400507 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400508 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400509
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500510 eb = (struct extent_buffer *)page->private;
511 if (page != eb->pages[0])
512 return 0;
Alex Lyakas0f805532016-03-10 13:10:15 +0200513
Chris Mason19c00dd2007-10-15 16:19:22 -0400514 found_start = btrfs_header_bytenr(eb);
Alex Lyakas0f805532016-03-10 13:10:15 +0200515 /*
516 * Please do not consolidate these warnings into a single if.
517 * It is useful to know what went wrong.
518 */
519 if (WARN_ON(found_start != start))
520 return -EUCLEAN;
521 if (WARN_ON(!PageUptodate(page)))
522 return -EUCLEAN;
523
524 ASSERT(memcmp_extent_buffer(eb, fs_info->fsid,
525 btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0);
526
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200527 return csum_tree_block(fs_info, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400528}
529
Daniel Dressler01d58472014-11-21 17:15:07 +0900530static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
Yan Zheng2b820322008-11-17 21:11:30 -0500531 struct extent_buffer *eb)
532{
Daniel Dressler01d58472014-11-21 17:15:07 +0900533 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
Yan Zheng2b820322008-11-17 21:11:30 -0500534 u8 fsid[BTRFS_UUID_SIZE];
535 int ret = 1;
536
Ross Kirk0a4e5582013-09-24 10:12:38 +0100537 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500538 while (fs_devices) {
539 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
540 ret = 0;
541 break;
542 }
543 fs_devices = fs_devices->seed;
544 }
545 return ret;
546}
547
Josef Bacika826d6d2011-03-16 13:42:43 -0400548#define CORRUPT(reason, eb, root, slot) \
Liu Bo6b722c12016-09-14 17:23:24 -0700549 btrfs_crit(root->fs_info, "corrupt %s, %s: block=%llu," \
550 " root=%llu, slot=%d", \
551 btrfs_header_level(eb) == 0 ? "leaf" : "node",\
552 reason, btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400553
554static noinline int check_leaf(struct btrfs_root *root,
555 struct extent_buffer *leaf)
556{
557 struct btrfs_key key;
558 struct btrfs_key leaf_key;
559 u32 nritems = btrfs_header_nritems(leaf);
560 int slot;
561
Filipe Mananaa1e0e042016-11-23 16:21:18 +0000562 /*
563 * Extent buffers from a relocation tree have a owner field that
564 * corresponds to the subvolume tree they are based on. So just from an
565 * extent buffer alone we can not find out what is the id of the
566 * corresponding subvolume tree, so we can not figure out if the extent
567 * buffer corresponds to the root of the relocation tree or not. So skip
568 * this check for relocation trees.
569 */
570 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
Liu Bo1ba98d02016-08-23 15:22:58 -0700571 struct btrfs_root *check_root;
572
573 key.objectid = btrfs_header_owner(leaf);
574 key.type = BTRFS_ROOT_ITEM_KEY;
575 key.offset = (u64)-1;
576
577 check_root = btrfs_get_fs_root(root->fs_info, &key, false);
578 /*
579 * The only reason we also check NULL here is that during
580 * open_ctree() some roots has not yet been set up.
581 */
582 if (!IS_ERR_OR_NULL(check_root)) {
Liu Bo7d470f02016-09-02 12:35:34 -0700583 struct extent_buffer *eb;
584
585 eb = btrfs_root_node(check_root);
Liu Bo1ba98d02016-08-23 15:22:58 -0700586 /* if leaf is the root, then it's fine */
Liu Bo7d470f02016-09-02 12:35:34 -0700587 if (leaf != eb) {
Liu Bo1ba98d02016-08-23 15:22:58 -0700588 CORRUPT("non-root leaf's nritems is 0",
Liu Bo7d470f02016-09-02 12:35:34 -0700589 leaf, check_root, 0);
590 free_extent_buffer(eb);
Liu Bo1ba98d02016-08-23 15:22:58 -0700591 return -EIO;
592 }
Liu Bo7d470f02016-09-02 12:35:34 -0700593 free_extent_buffer(eb);
Liu Bo1ba98d02016-08-23 15:22:58 -0700594 }
Josef Bacika826d6d2011-03-16 13:42:43 -0400595 return 0;
Liu Bo1ba98d02016-08-23 15:22:58 -0700596 }
Josef Bacika826d6d2011-03-16 13:42:43 -0400597
Filipe Mananaa1e0e042016-11-23 16:21:18 +0000598 if (nritems == 0)
599 return 0;
600
Josef Bacika826d6d2011-03-16 13:42:43 -0400601 /* Check the 0 item */
602 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
603 BTRFS_LEAF_DATA_SIZE(root)) {
604 CORRUPT("invalid item offset size pair", leaf, root, 0);
605 return -EIO;
606 }
607
608 /*
609 * Check to make sure each items keys are in the correct order and their
610 * offsets make sense. We only have to loop through nritems-1 because
611 * we check the current slot against the next slot, which verifies the
612 * next slot's offset+size makes sense and that the current's slot
613 * offset is correct.
614 */
615 for (slot = 0; slot < nritems - 1; slot++) {
616 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
617 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
618
619 /* Make sure the keys are in the right order */
620 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
621 CORRUPT("bad key order", leaf, root, slot);
622 return -EIO;
623 }
624
625 /*
626 * Make sure the offset and ends are right, remember that the
627 * item data starts at the end of the leaf and grows towards the
628 * front.
629 */
630 if (btrfs_item_offset_nr(leaf, slot) !=
631 btrfs_item_end_nr(leaf, slot + 1)) {
632 CORRUPT("slot offset bad", leaf, root, slot);
633 return -EIO;
634 }
635
636 /*
637 * Check to make sure that we don't point outside of the leaf,
Nicholas D Steeves01327612016-05-19 21:18:45 -0400638 * just in case all the items are consistent to each other, but
Josef Bacika826d6d2011-03-16 13:42:43 -0400639 * all point outside of the leaf.
640 */
641 if (btrfs_item_end_nr(leaf, slot) >
642 BTRFS_LEAF_DATA_SIZE(root)) {
643 CORRUPT("slot end outside of leaf", leaf, root, slot);
644 return -EIO;
645 }
646 }
647
648 return 0;
649}
650
Liu Bo053ab70f2016-08-23 17:37:45 -0700651static int check_node(struct btrfs_root *root, struct extent_buffer *node)
652{
653 unsigned long nr = btrfs_header_nritems(node);
Liu Bo6b722c12016-09-14 17:23:24 -0700654 struct btrfs_key key, next_key;
655 int slot;
656 u64 bytenr;
657 int ret = 0;
Liu Bo053ab70f2016-08-23 17:37:45 -0700658
659 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) {
660 btrfs_crit(root->fs_info,
661 "corrupt node: block %llu root %llu nritems %lu",
662 node->start, root->objectid, nr);
663 return -EIO;
664 }
Liu Bo6b722c12016-09-14 17:23:24 -0700665
666 for (slot = 0; slot < nr - 1; slot++) {
667 bytenr = btrfs_node_blockptr(node, slot);
668 btrfs_node_key_to_cpu(node, &key, slot);
669 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
670
671 if (!bytenr) {
672 CORRUPT("invalid item slot", node, root, slot);
673 ret = -EIO;
674 goto out;
675 }
676
677 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
678 CORRUPT("bad key order", node, root, slot);
679 ret = -EIO;
680 goto out;
681 }
682 }
683out:
684 return ret;
Liu Bo053ab70f2016-08-23 17:37:45 -0700685}
686
Miao Xiefacc8a222013-07-25 19:22:34 +0800687static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
688 u64 phy_offset, struct page *page,
689 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400690{
Chris Masonce9adaa2008-04-09 16:28:12 -0400691 u64 found_start;
692 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400693 struct extent_buffer *eb;
694 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Zhao Lei02873e42015-12-31 22:46:45 +0800695 struct btrfs_fs_info *fs_info = root->fs_info;
Chris Masonf1885912008-04-09 16:28:12 -0400696 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400697 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400698
Chris Masonce9adaa2008-04-09 16:28:12 -0400699 if (!page->private)
700 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500701
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500702 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500703
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400704 /* the pending IO might have been the only thing that kept this buffer
705 * in memory. Make sure we have a ref for all this other checks
706 */
707 extent_buffer_get(eb);
708
709 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400710 if (!reads_done)
711 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400712
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400713 eb->read_mirror = mirror;
Filipe Manana656f30d2014-09-26 12:25:56 +0100714 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
Josef Bacikea466792012-03-26 21:57:36 -0400715 ret = -EIO;
716 goto err;
717 }
718
Chris Masonce9adaa2008-04-09 16:28:12 -0400719 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400720 if (found_start != eb->start) {
Zhao Lei02873e42015-12-31 22:46:45 +0800721 btrfs_err_rl(fs_info, "bad tree block start %llu %llu",
722 found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400723 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400724 goto err;
725 }
Zhao Lei02873e42015-12-31 22:46:45 +0800726 if (check_tree_block_fsid(fs_info, eb)) {
727 btrfs_err_rl(fs_info, "bad fsid on block %llu",
728 eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400729 ret = -EIO;
730 goto err;
731 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400732 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400733 if (found_level >= BTRFS_MAX_LEVEL) {
Zhao Lei02873e42015-12-31 22:46:45 +0800734 btrfs_err(fs_info, "bad tree block level %d",
735 (int)btrfs_header_level(eb));
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400736 ret = -EIO;
737 goto err;
738 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400739
Chris Mason85d4e462011-07-26 16:11:19 -0400740 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
741 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500742
Zhao Lei02873e42015-12-31 22:46:45 +0800743 ret = csum_tree_block(fs_info, eb, 1);
Alex Lyakas8bd98f02016-03-10 13:09:46 +0200744 if (ret)
Josef Bacika826d6d2011-03-16 13:42:43 -0400745 goto err;
Josef Bacika826d6d2011-03-16 13:42:43 -0400746
747 /*
748 * If this is a leaf block and it is corrupt, set the corrupt bit so
749 * that we don't try and read the other copies of this block, just
750 * return -EIO.
751 */
752 if (found_level == 0 && check_leaf(root, eb)) {
753 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
754 ret = -EIO;
755 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400756
Liu Bo053ab70f2016-08-23 17:37:45 -0700757 if (found_level > 0 && check_node(root, eb))
758 ret = -EIO;
759
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400760 if (!ret)
761 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400762err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400763 if (reads_done &&
764 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Zhao Lei02873e42015-12-31 22:46:45 +0800765 btree_readahead_hook(fs_info, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200766
David Woodhouse53b381b2013-01-29 18:40:14 -0500767 if (ret) {
768 /*
769 * our io error hook is going to dec the io pages
770 * again, we have to make sure it has something
771 * to decrement
772 */
773 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400774 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500775 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400776 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400777out:
Chris Masonf1885912008-04-09 16:28:12 -0400778 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400779}
780
Josef Bacikea466792012-03-26 21:57:36 -0400781static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200782{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200783 struct extent_buffer *eb;
Arne Jansen4bb31e92011-06-10 13:55:54 +0200784
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500785 eb = (struct extent_buffer *)page->private;
Filipe Manana656f30d2014-09-26 12:25:56 +0100786 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400787 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500788 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400789 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Zhao Lei02873e42015-12-31 22:46:45 +0800790 btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200791 return -EIO; /* we fixed nothing */
792}
793
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200794static void end_workqueue_bio(struct bio *bio)
Chris Masonce9adaa2008-04-09 16:28:12 -0400795{
David Sterba97eb6b62014-07-30 00:55:42 +0200796 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400797 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800798 struct btrfs_workqueue *wq;
799 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400800
Chris Masonce9adaa2008-04-09 16:28:12 -0400801 fs_info = end_io_wq->info;
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200802 end_io_wq->error = bio->bi_error;
Chris Masond20f7042008-12-08 16:58:54 -0500803
Mike Christie37226b22016-06-05 14:31:52 -0500804 if (bio_op(bio) == REQ_OP_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800805 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
806 wq = fs_info->endio_meta_write_workers;
807 func = btrfs_endio_meta_write_helper;
808 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
809 wq = fs_info->endio_freespace_worker;
810 func = btrfs_freespace_write_helper;
811 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
812 wq = fs_info->endio_raid56_workers;
813 func = btrfs_endio_raid56_helper;
814 } else {
815 wq = fs_info->endio_write_workers;
816 func = btrfs_endio_write_helper;
817 }
Chris Masond20f7042008-12-08 16:58:54 -0500818 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800819 if (unlikely(end_io_wq->metadata ==
820 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
821 wq = fs_info->endio_repair_workers;
822 func = btrfs_endio_repair_helper;
823 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800824 wq = fs_info->endio_raid56_workers;
825 func = btrfs_endio_raid56_helper;
826 } else if (end_io_wq->metadata) {
827 wq = fs_info->endio_meta_workers;
828 func = btrfs_endio_meta_helper;
829 } else {
830 wq = fs_info->endio_workers;
831 func = btrfs_endio_helper;
832 }
Chris Masond20f7042008-12-08 16:58:54 -0500833 }
Liu Bo9e0af232014-08-15 23:36:53 +0800834
835 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
836 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400837}
838
Chris Mason22c59942008-04-09 16:28:12 -0400839int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200840 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400841{
David Sterba97eb6b62014-07-30 00:55:42 +0200842 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800843
David Sterba97eb6b62014-07-30 00:55:42 +0200844 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400845 if (!end_io_wq)
846 return -ENOMEM;
847
848 end_io_wq->private = bio->bi_private;
849 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400850 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400851 end_io_wq->error = 0;
852 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400853 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400854
855 bio->bi_private = end_io_wq;
856 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400857 return 0;
858}
859
Chris Masonb64a2852008-08-20 13:39:41 -0400860unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400861{
862 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800863 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400864 info->fs_devices->open_devices);
865 return 256 * limit;
866}
867
Chris Mason4a69a412008-11-06 22:03:00 -0500868static void run_one_async_start(struct btrfs_work *work)
869{
Chris Mason4a69a412008-11-06 22:03:00 -0500870 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100871 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500872
873 async = container_of(work, struct async_submit_bio, work);
Mike Christie81a75f62016-06-05 14:31:54 -0500874 ret = async->submit_bio_start(async->inode, async->bio,
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100875 async->mirror_num, async->bio_flags,
876 async->bio_offset);
877 if (ret)
878 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500879}
880
881static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400882{
883 struct btrfs_fs_info *fs_info;
884 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400885 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400886
887 async = container_of(work, struct async_submit_bio, work);
888 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400889
Chris Masonb64a2852008-08-20 13:39:41 -0400890 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400891 limit = limit * 2 / 3;
892
David Sterbaee863952015-02-16 19:41:40 +0100893 /*
894 * atomic_dec_return implies a barrier for waitqueue_active
895 */
Josef Bacik66657b32012-08-01 15:36:24 -0400896 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400897 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400898 wake_up(&fs_info->async_submit_wait);
899
Adam Buchbinderbb7ab3b2016-03-04 11:23:12 -0800900 /* If an error occurred we just want to clean up the bio and move on */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100901 if (async->error) {
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200902 async->bio->bi_error = async->error;
903 bio_endio(async->bio);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100904 return;
905 }
906
Mike Christie81a75f62016-06-05 14:31:54 -0500907 async->submit_bio_done(async->inode, async->bio, async->mirror_num,
908 async->bio_flags, async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500909}
910
911static void run_one_async_free(struct btrfs_work *work)
912{
913 struct async_submit_bio *async;
914
915 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400916 kfree(async);
917}
918
Chris Mason44b8bd72008-04-16 11:14:51 -0400919int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
Mike Christie81a75f62016-06-05 14:31:54 -0500920 struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400921 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400922 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500923 extent_submit_bio_hook_t *submit_bio_start,
924 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400925{
926 struct async_submit_bio *async;
927
928 async = kmalloc(sizeof(*async), GFP_NOFS);
929 if (!async)
930 return -ENOMEM;
931
932 async->inode = inode;
Chris Mason44b8bd72008-04-16 11:14:51 -0400933 async->bio = bio;
934 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500935 async->submit_bio_start = submit_bio_start;
936 async->submit_bio_done = submit_bio_done;
937
Liu Bo9e0af232014-08-15 23:36:53 +0800938 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800939 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500940
Chris Masonc8b97812008-10-29 14:49:59 -0400941 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400942 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400943
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100944 async->error = 0;
945
Chris Masoncb03c742008-05-15 16:15:45 -0400946 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400947
Jens Axboe1eff9d32016-08-05 15:35:16 -0600948 if (bio->bi_opf & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800949 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400950
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800951 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400952
Chris Masond3977122009-01-05 21:25:51 -0500953 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500954 atomic_read(&fs_info->nr_async_submits)) {
955 wait_event(fs_info->async_submit_wait,
956 (atomic_read(&fs_info->nr_async_submits) == 0));
957 }
958
Chris Mason44b8bd72008-04-16 11:14:51 -0400959 return 0;
960}
961
Chris Masonce3ed712008-09-23 13:14:12 -0400962static int btree_csum_one_bio(struct bio *bio)
963{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800964 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400965 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800966 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400967
Kent Overstreet2c30c712013-11-07 12:20:26 -0800968 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400969 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Daniel Dressler01d58472014-11-21 17:15:07 +0900970 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100971 if (ret)
972 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400973 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800974
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100975 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400976}
977
Mike Christie81a75f62016-06-05 14:31:54 -0500978static int __btree_submit_bio_start(struct inode *inode, struct bio *bio,
979 int mirror_num, unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400980 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400981{
Chris Mason8b712842008-06-11 16:50:36 -0400982 /*
983 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400984 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400985 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100986 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500987}
Chris Mason22c59942008-04-09 16:28:12 -0400988
Mike Christie81a75f62016-06-05 14:31:54 -0500989static int __btree_submit_bio_done(struct inode *inode, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400990 int mirror_num, unsigned long bio_flags,
991 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500992{
Stefan Behrens61891922012-11-05 18:51:52 +0100993 int ret;
994
Chris Mason8b712842008-06-11 16:50:36 -0400995 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500996 * when we're called for a write, we're already in the async
997 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400998 */
Mike Christie81a75f62016-06-05 14:31:54 -0500999 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 1);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001000 if (ret) {
1001 bio->bi_error = ret;
1002 bio_endio(bio);
1003 }
Stefan Behrens61891922012-11-05 18:51:52 +01001004 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -04001005}
1006
Josef Bacikde0022b2012-09-25 14:25:58 -04001007static int check_async_write(struct inode *inode, unsigned long bio_flags)
1008{
1009 if (bio_flags & EXTENT_BIO_TREE_LOG)
1010 return 0;
1011#ifdef CONFIG_X86
Borislav Petkovbc696ca2016-01-26 22:12:05 +01001012 if (static_cpu_has(X86_FEATURE_XMM4_2))
Josef Bacikde0022b2012-09-25 14:25:58 -04001013 return 0;
1014#endif
1015 return 1;
1016}
1017
Mike Christie81a75f62016-06-05 14:31:54 -05001018static int btree_submit_bio_hook(struct inode *inode, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -04001019 int mirror_num, unsigned long bio_flags,
1020 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -04001021{
Josef Bacikde0022b2012-09-25 14:25:58 -04001022 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -05001023 int ret;
1024
Mike Christie37226b22016-06-05 14:31:52 -05001025 if (bio_op(bio) != REQ_OP_WRITE) {
Chris Mason4a69a412008-11-06 22:03:00 -05001026 /*
1027 * called for a read, do the setup so that checksum validation
1028 * can happen in the async kernel threads
1029 */
Chris Masonf3f266a2012-03-23 10:22:46 -04001030 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +02001031 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -04001032 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +01001033 goto out_w_error;
Mike Christie81a75f62016-06-05 14:31:54 -05001034 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -04001035 } else if (!async) {
1036 ret = btree_csum_one_bio(bio);
1037 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +01001038 goto out_w_error;
Mike Christie81a75f62016-06-05 14:31:54 -05001039 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0);
Stefan Behrens61891922012-11-05 18:51:52 +01001040 } else {
1041 /*
1042 * kthread helpers are used to submit writes so that
1043 * checksumming can happen in parallel across all CPUs
1044 */
1045 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
Mike Christie81a75f62016-06-05 14:31:54 -05001046 inode, bio, mirror_num, 0,
Stefan Behrens61891922012-11-05 18:51:52 +01001047 bio_offset,
1048 __btree_submit_bio_start,
1049 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -04001050 }
Chris Masond313d7a2009-04-20 15:50:09 -04001051
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001052 if (ret)
1053 goto out_w_error;
1054 return 0;
1055
Stefan Behrens61891922012-11-05 18:51:52 +01001056out_w_error:
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001057 bio->bi_error = ret;
1058 bio_endio(bio);
Stefan Behrens61891922012-11-05 18:51:52 +01001059 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -04001060}
1061
Jan Beulich3dd14622010-12-07 14:54:09 +00001062#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001063static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -08001064 struct page *newpage, struct page *page,
1065 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -05001066{
1067 /*
1068 * we can't safely write a btree page from here,
1069 * we haven't done the locking hook
1070 */
1071 if (PageDirty(page))
1072 return -EAGAIN;
1073 /*
1074 * Buffers may be managed in a filesystem specific way.
1075 * We must have no buffers or drop them.
1076 */
1077 if (page_has_private(page) &&
1078 !try_to_release_page(page, GFP_KERNEL))
1079 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -08001080 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -05001081}
Jan Beulich3dd14622010-12-07 14:54:09 +00001082#endif
Chris Mason784b4e22010-11-21 22:20:49 -05001083
Chris Mason0da54682007-11-07 21:08:01 -05001084
1085static int btree_writepages(struct address_space *mapping,
1086 struct writeback_control *wbc)
1087{
Miao Xiee2d84522013-01-29 10:09:20 +00001088 struct btrfs_fs_info *fs_info;
1089 int ret;
1090
Chris Masond8d5f3e2007-12-11 12:42:00 -05001091 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001092
1093 if (wbc->for_kupdate)
1094 return 0;
1095
Miao Xiee2d84522013-01-29 10:09:20 +00001096 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001097 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001098 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1099 BTRFS_DIRTY_METADATA_THRESH);
1100 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001101 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001102 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001103 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001104}
1105
Christoph Hellwigb2950862008-12-02 09:54:17 -05001106static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001107{
Chris Masond1310b22008-01-24 16:13:08 -05001108 struct extent_io_tree *tree;
1109 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001110 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001111}
1112
Chris Mason70dec802008-01-29 09:59:12 -05001113static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001114{
Chris Mason98509cf2008-09-11 15:51:43 -04001115 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001116 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001117
David Sterbaf7a52a42013-04-26 14:56:29 +00001118 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001119}
Chris Mason123abc82007-03-14 14:14:43 -04001120
Lukas Czernerd47992f2013-05-21 23:17:23 -04001121static void btree_invalidatepage(struct page *page, unsigned int offset,
1122 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001123{
Chris Masond1310b22008-01-24 16:13:08 -05001124 struct extent_io_tree *tree;
1125 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001126 extent_invalidatepage(tree, page, offset);
1127 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001128 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001129 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1130 "page private not zero on page %llu",
1131 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001132 ClearPagePrivate(page);
1133 set_page_private(page, 0);
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001134 put_page(page);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001135 }
Chris Masond98237b2007-03-28 13:57:48 -04001136}
1137
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001138static int btree_set_page_dirty(struct page *page)
1139{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001140#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001141 struct extent_buffer *eb;
1142
1143 BUG_ON(!PagePrivate(page));
1144 eb = (struct extent_buffer *)page->private;
1145 BUG_ON(!eb);
1146 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1147 BUG_ON(!atomic_read(&eb->refs));
1148 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001149#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001150 return __set_page_dirty_nobuffers(page);
1151}
1152
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001153static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001154 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001155 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001156 .releasepage = btree_releasepage,
1157 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001158#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001159 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001160#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001161 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001162};
1163
David Sterbad3e46fe2014-06-15 02:04:19 +02001164void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
Chris Mason090d1872007-05-01 08:53:32 -04001165{
Chris Mason5f39d392007-10-15 16:14:19 -04001166 struct extent_buffer *buf = NULL;
1167 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason090d1872007-05-01 08:53:32 -04001168
David Sterbaa83fffb2014-06-15 02:39:54 +02001169 buf = btrfs_find_create_tree_block(root, bytenr);
Liu Boc871b0f2016-06-06 12:01:23 -07001170 if (IS_ERR(buf))
David Sterba6197d862014-06-15 00:49:36 +02001171 return;
Chris Masond1310b22008-01-24 16:13:08 -05001172 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Josef Bacik8436ea912016-09-02 15:40:03 -04001173 buf, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001174 free_extent_buffer(buf);
Chris Mason090d1872007-05-01 08:53:32 -04001175}
1176
David Sterbac0dcaa42014-06-15 02:24:21 +02001177int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
Arne Jansenab0fff02011-05-23 14:25:41 +02001178 int mirror_num, struct extent_buffer **eb)
1179{
1180 struct extent_buffer *buf = NULL;
1181 struct inode *btree_inode = root->fs_info->btree_inode;
1182 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1183 int ret;
1184
David Sterbaa83fffb2014-06-15 02:39:54 +02001185 buf = btrfs_find_create_tree_block(root, bytenr);
Liu Boc871b0f2016-06-06 12:01:23 -07001186 if (IS_ERR(buf))
Arne Jansenab0fff02011-05-23 14:25:41 +02001187 return 0;
1188
1189 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1190
Josef Bacik8436ea912016-09-02 15:40:03 -04001191 ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK,
Arne Jansenab0fff02011-05-23 14:25:41 +02001192 btree_get_extent, mirror_num);
1193 if (ret) {
1194 free_extent_buffer(buf);
1195 return ret;
1196 }
1197
1198 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1199 free_extent_buffer(buf);
1200 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001201 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001202 *eb = buf;
1203 } else {
1204 free_extent_buffer(buf);
1205 }
1206 return 0;
1207}
1208
Daniel Dressler01d58472014-11-21 17:15:07 +09001209struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
David Sterba0308af42014-06-15 01:43:40 +02001210 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001211{
Daniel Dressler01d58472014-11-21 17:15:07 +09001212 return find_extent_buffer(fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001213}
1214
1215struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
David Sterbaa83fffb2014-06-15 02:39:54 +02001216 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001217{
Jeff Mahoneyf5ee5c92016-06-21 09:52:41 -04001218 if (btrfs_is_testing(root->fs_info))
Feifei Xub9ef22d2016-06-01 19:18:25 +08001219 return alloc_test_extent_buffer(root->fs_info, bytenr,
1220 root->nodesize);
David Sterbace3e6982014-06-15 03:00:04 +02001221 return alloc_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001222}
1223
1224
Chris Masone02119d2008-09-05 16:13:11 -04001225int btrfs_write_tree_block(struct extent_buffer *buf)
1226{
Chris Mason727011e2010-08-06 13:21:20 -04001227 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001228 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001229}
1230
1231int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1232{
Chris Mason727011e2010-08-06 13:21:20 -04001233 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001234 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001235}
1236
Chris Masondb945352007-10-15 16:15:53 -04001237struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
David Sterbace86cd52014-06-15 01:07:32 +02001238 u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001239{
Chris Mason5f39d392007-10-15 16:14:19 -04001240 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001241 int ret;
1242
David Sterbaa83fffb2014-06-15 02:39:54 +02001243 buf = btrfs_find_create_tree_block(root, bytenr);
Liu Boc871b0f2016-06-06 12:01:23 -07001244 if (IS_ERR(buf))
1245 return buf;
Chris Masone4204de2008-01-08 15:46:27 -05001246
Josef Bacik8436ea912016-09-02 15:40:03 -04001247 ret = btree_read_extent_buffer_pages(root, buf, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001248 if (ret) {
1249 free_extent_buffer(buf);
Liu Bo64c043d2015-05-25 17:30:15 +08001250 return ERR_PTR(ret);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001251 }
Chris Mason5f39d392007-10-15 16:14:19 -04001252 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001253
Chris Masoneb60cea2007-02-02 09:18:22 -05001254}
1255
Daniel Dressler01d58472014-11-21 17:15:07 +09001256void clean_tree_block(struct btrfs_trans_handle *trans,
1257 struct btrfs_fs_info *fs_info,
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001258 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001259{
Chris Mason55c69072008-01-09 15:55:33 -05001260 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001261 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001262 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001263
Chris Masonb9473432009-03-13 11:00:37 -04001264 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001265 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1266 -buf->len,
1267 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001268 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1269 btrfs_set_lock_blocking(buf);
1270 clear_extent_buffer_dirty(buf);
1271 }
Chris Mason925baed2008-06-25 16:01:30 -04001272 }
Chris Mason5f39d392007-10-15 16:14:19 -04001273}
1274
Miao Xie8257b2d2014-03-06 13:38:19 +08001275static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1276{
1277 struct btrfs_subvolume_writers *writers;
1278 int ret;
1279
1280 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1281 if (!writers)
1282 return ERR_PTR(-ENOMEM);
1283
Tejun Heo908c7f12014-09-08 09:51:29 +09001284 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
Miao Xie8257b2d2014-03-06 13:38:19 +08001285 if (ret < 0) {
1286 kfree(writers);
1287 return ERR_PTR(ret);
1288 }
1289
1290 init_waitqueue_head(&writers->wait);
1291 return writers;
1292}
1293
1294static void
1295btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1296{
1297 percpu_counter_destroy(&writers->counter);
1298 kfree(writers);
1299}
1300
David Sterba707e8a02014-06-04 19:22:26 +02001301static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1302 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
Jeff Mahoney143bede2012-03-01 14:56:26 +01001303 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001304{
Jeff Mahoney7c0260e2016-06-20 14:14:09 -04001305 bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
Chris Masoncfaa7292007-02-21 17:04:57 -05001306 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001307 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001308 root->sectorsize = sectorsize;
1309 root->nodesize = nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001310 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001311 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001312 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001313
Chris Mason0f7d52f2007-04-09 10:42:37 -04001314 root->objectid = objectid;
1315 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001316 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001317 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001318 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001319 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001320 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001321 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001322 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001323 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001324
1325 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001326 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001327 INIT_LIST_HEAD(&root->delalloc_inodes);
1328 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001329 INIT_LIST_HEAD(&root->ordered_extents);
1330 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001331 INIT_LIST_HEAD(&root->logged_list[0]);
1332 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001333 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001334 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001335 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001336 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001337 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001338 spin_lock_init(&root->log_extents_lock[0]);
1339 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001340 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001341 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001342 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001343 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001344 init_waitqueue_head(&root->log_writer_wait);
1345 init_waitqueue_head(&root->log_commit_wait[0]);
1346 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001347 INIT_LIST_HEAD(&root->log_ctxs[0]);
1348 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f182009-01-21 12:54:03 -05001349 atomic_set(&root->log_commit[0], 0);
1350 atomic_set(&root->log_commit[1], 0);
1351 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001352 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001353 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001354 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001355 atomic_set(&root->will_be_snapshoted, 0);
Qu Wenruo55eeaf02015-09-08 17:08:38 +08001356 atomic_set(&root->qgroup_meta_rsv, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001357 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001358 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001359 root->last_log_commit = 0;
Jeff Mahoney7c0260e2016-06-20 14:14:09 -04001360 if (!dummy)
Josef Bacik06ea65a2013-09-19 16:07:01 -04001361 extent_io_tree_init(&root->dirty_log_pages,
1362 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001363
Chris Mason3768f362007-03-13 16:47:54 -04001364 memset(&root->root_key, 0, sizeof(root->root_key));
1365 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001366 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Jeff Mahoney7c0260e2016-06-20 14:14:09 -04001367 if (!dummy)
Josef Bacik06ea65a2013-09-19 16:07:01 -04001368 root->defrag_trans_start = fs_info->generation;
1369 else
1370 root->defrag_trans_start = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001371 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001372 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001373
Anand Jain5f3ab902012-12-07 09:28:54 +00001374 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001375}
1376
David Sterba74e4d822016-02-11 11:01:55 +01001377static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info,
1378 gfp_t flags)
Al Viro6f07e422011-11-17 00:46:16 -05001379{
David Sterba74e4d822016-02-11 11:01:55 +01001380 struct btrfs_root *root = kzalloc(sizeof(*root), flags);
Al Viro6f07e422011-11-17 00:46:16 -05001381 if (root)
1382 root->fs_info = fs_info;
1383 return root;
1384}
1385
Josef Bacik06ea65a2013-09-19 16:07:01 -04001386#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1387/* Should only be used by the testing infrastructure */
Jeff Mahoney7c0260e2016-06-20 14:14:09 -04001388struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info,
1389 u32 sectorsize, u32 nodesize)
Josef Bacik06ea65a2013-09-19 16:07:01 -04001390{
1391 struct btrfs_root *root;
1392
Jeff Mahoney7c0260e2016-06-20 14:14:09 -04001393 if (!fs_info)
1394 return ERR_PTR(-EINVAL);
1395
1396 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Josef Bacik06ea65a2013-09-19 16:07:01 -04001397 if (!root)
1398 return ERR_PTR(-ENOMEM);
Feifei Xub9ef22d2016-06-01 19:18:25 +08001399 /* We don't use the stripesize in selftest, set it as sectorsize */
Jeff Mahoney7c0260e2016-06-20 14:14:09 -04001400 __setup_root(nodesize, sectorsize, sectorsize, root, fs_info,
Feifei Xuef9f2db2016-06-01 19:18:28 +08001401 BTRFS_ROOT_TREE_OBJECTID);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001402 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001403
1404 return root;
1405}
1406#endif
1407
Arne Jansen20897f52011-09-13 12:44:20 +02001408struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1409 struct btrfs_fs_info *fs_info,
1410 u64 objectid)
1411{
1412 struct extent_buffer *leaf;
1413 struct btrfs_root *tree_root = fs_info->tree_root;
1414 struct btrfs_root *root;
1415 struct btrfs_key key;
1416 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001417 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001418
David Sterba74e4d822016-02-11 11:01:55 +01001419 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Arne Jansen20897f52011-09-13 12:44:20 +02001420 if (!root)
1421 return ERR_PTR(-ENOMEM);
1422
David Sterba707e8a02014-06-04 19:22:26 +02001423 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1424 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001425 root->root_key.objectid = objectid;
1426 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1427 root->root_key.offset = 0;
1428
David Sterba4d75f8a2014-06-15 01:54:12 +02001429 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Arne Jansen20897f52011-09-13 12:44:20 +02001430 if (IS_ERR(leaf)) {
1431 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001432 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001433 goto fail;
1434 }
1435
Arne Jansen20897f52011-09-13 12:44:20 +02001436 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1437 btrfs_set_header_bytenr(leaf, leaf->start);
1438 btrfs_set_header_generation(leaf, trans->transid);
1439 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1440 btrfs_set_header_owner(leaf, objectid);
1441 root->node = leaf;
1442
Ross Kirk0a4e5582013-09-24 10:12:38 +01001443 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001444 BTRFS_FSID_SIZE);
1445 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001446 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001447 BTRFS_UUID_SIZE);
1448 btrfs_mark_buffer_dirty(leaf);
1449
1450 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001451 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001452
1453 root->root_item.flags = 0;
1454 root->root_item.byte_limit = 0;
1455 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1456 btrfs_set_root_generation(&root->root_item, trans->transid);
1457 btrfs_set_root_level(&root->root_item, 0);
1458 btrfs_set_root_refs(&root->root_item, 1);
1459 btrfs_set_root_used(&root->root_item, leaf->len);
1460 btrfs_set_root_last_snapshot(&root->root_item, 0);
1461 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001462 uuid_le_gen(&uuid);
1463 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001464 root->root_item.drop_level = 0;
1465
1466 key.objectid = objectid;
1467 key.type = BTRFS_ROOT_ITEM_KEY;
1468 key.offset = 0;
1469 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1470 if (ret)
1471 goto fail;
1472
1473 btrfs_tree_unlock(leaf);
1474
Arne Jansen20897f52011-09-13 12:44:20 +02001475 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001476
1477fail:
1478 if (leaf) {
1479 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001480 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001481 free_extent_buffer(leaf);
1482 }
1483 kfree(root);
1484
1485 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001486}
1487
Yan Zheng7237f182009-01-21 12:54:03 -05001488static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1489 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001490{
1491 struct btrfs_root *root;
1492 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001493 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001494
David Sterba74e4d822016-02-11 11:01:55 +01001495 root = btrfs_alloc_root(fs_info, GFP_NOFS);
Chris Masone02119d2008-09-05 16:13:11 -04001496 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001497 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001498
David Sterba707e8a02014-06-04 19:22:26 +02001499 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1500 tree_root->stripesize, root, fs_info,
1501 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001502
1503 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1504 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1505 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001506
Yan Zheng7237f182009-01-21 12:54:03 -05001507 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001508 * DON'T set REF_COWS for log trees
1509 *
Yan Zheng7237f182009-01-21 12:54:03 -05001510 * log trees do not get reference counted because they go away
1511 * before a real commit is actually done. They do store pointers
1512 * to file data extents, and those reference counts still get
1513 * updated (along with back refs to the log tree).
1514 */
Chris Masone02119d2008-09-05 16:13:11 -04001515
David Sterba4d75f8a2014-06-15 01:54:12 +02001516 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1517 NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001518 if (IS_ERR(leaf)) {
1519 kfree(root);
1520 return ERR_CAST(leaf);
1521 }
Chris Masone02119d2008-09-05 16:13:11 -04001522
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001523 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1524 btrfs_set_header_bytenr(leaf, leaf->start);
1525 btrfs_set_header_generation(leaf, trans->transid);
1526 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1527 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001528 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001529
1530 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001531 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001532 btrfs_mark_buffer_dirty(root->node);
1533 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001534 return root;
1535}
1536
1537int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1538 struct btrfs_fs_info *fs_info)
1539{
1540 struct btrfs_root *log_root;
1541
1542 log_root = alloc_log_tree(trans, fs_info);
1543 if (IS_ERR(log_root))
1544 return PTR_ERR(log_root);
1545 WARN_ON(fs_info->log_root_tree);
1546 fs_info->log_root_tree = log_root;
1547 return 0;
1548}
1549
1550int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1551 struct btrfs_root *root)
1552{
1553 struct btrfs_root *log_root;
1554 struct btrfs_inode_item *inode_item;
1555
1556 log_root = alloc_log_tree(trans, root->fs_info);
1557 if (IS_ERR(log_root))
1558 return PTR_ERR(log_root);
1559
1560 log_root->last_trans = trans->transid;
1561 log_root->root_key.offset = root->root_key.objectid;
1562
1563 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001564 btrfs_set_stack_inode_generation(inode_item, 1);
1565 btrfs_set_stack_inode_size(inode_item, 3);
1566 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001567 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001568 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f182009-01-21 12:54:03 -05001569
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001570 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001571
1572 WARN_ON(root->log_root);
1573 root->log_root = log_root;
1574 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001575 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001576 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001577 return 0;
1578}
1579
Stefan Behrens35a36212013-08-14 18:12:25 +02001580static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1581 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001582{
1583 struct btrfs_root *root;
1584 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001585 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001586 u64 generation;
Miao Xiecb517ea2013-05-15 07:48:19 +00001587 int ret;
1588
1589 path = btrfs_alloc_path();
1590 if (!path)
1591 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001592
David Sterba74e4d822016-02-11 11:01:55 +01001593 root = btrfs_alloc_root(fs_info, GFP_NOFS);
Miao Xiecb517ea2013-05-15 07:48:19 +00001594 if (!root) {
1595 ret = -ENOMEM;
1596 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001597 }
1598
David Sterba707e8a02014-06-04 19:22:26 +02001599 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1600 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001601
Miao Xiecb517ea2013-05-15 07:48:19 +00001602 ret = btrfs_find_root(tree_root, key, path,
1603 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001604 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001605 if (ret > 0)
1606 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001607 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001608 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001609
Yan Zheng84234f32008-10-29 14:49:05 -04001610 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001611 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
David Sterbace86cd52014-06-15 01:07:32 +02001612 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08001613 if (IS_ERR(root->node)) {
1614 ret = PTR_ERR(root->node);
Miao Xiecb517ea2013-05-15 07:48:19 +00001615 goto find_fail;
1616 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1617 ret = -EIO;
Liu Bo64c043d2015-05-25 17:30:15 +08001618 free_extent_buffer(root->node);
1619 goto find_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001620 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001621 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001622out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001623 btrfs_free_path(path);
1624 return root;
1625
Miao Xiecb517ea2013-05-15 07:48:19 +00001626find_fail:
1627 kfree(root);
1628alloc_fail:
1629 root = ERR_PTR(ret);
1630 goto out;
1631}
1632
1633struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1634 struct btrfs_key *location)
1635{
1636 struct btrfs_root *root;
1637
1638 root = btrfs_read_tree_root(tree_root, location);
1639 if (IS_ERR(root))
1640 return root;
1641
1642 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001643 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001644 btrfs_check_and_init_root_item(&root->root_item);
1645 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001646
Chris Mason5eda7b52007-06-22 14:16:25 -04001647 return root;
1648}
1649
Miao Xiecb517ea2013-05-15 07:48:19 +00001650int btrfs_init_fs_root(struct btrfs_root *root)
1651{
1652 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001653 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001654
1655 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1656 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1657 GFP_NOFS);
1658 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1659 ret = -ENOMEM;
1660 goto fail;
1661 }
1662
Miao Xie8257b2d2014-03-06 13:38:19 +08001663 writers = btrfs_alloc_subvolume_writers();
1664 if (IS_ERR(writers)) {
1665 ret = PTR_ERR(writers);
1666 goto fail;
1667 }
1668 root->subv_writers = writers;
1669
Miao Xiecb517ea2013-05-15 07:48:19 +00001670 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001671 spin_lock_init(&root->ino_cache_lock);
1672 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001673
1674 ret = get_anon_bdev(&root->anon_dev);
1675 if (ret)
Liu Bo876d2cf2016-06-28 13:44:38 -07001676 goto fail;
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301677
1678 mutex_lock(&root->objectid_mutex);
1679 ret = btrfs_find_highest_objectid(root,
1680 &root->highest_objectid);
1681 if (ret) {
1682 mutex_unlock(&root->objectid_mutex);
Liu Bo876d2cf2016-06-28 13:44:38 -07001683 goto fail;
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301684 }
1685
1686 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1687
1688 mutex_unlock(&root->objectid_mutex);
1689
Miao Xiecb517ea2013-05-15 07:48:19 +00001690 return 0;
1691fail:
Liu Bo876d2cf2016-06-28 13:44:38 -07001692 /* the caller is responsible to call free_fs_root */
Miao Xiecb517ea2013-05-15 07:48:19 +00001693 return ret;
1694}
1695
Jeff Mahoney35bbb972016-08-17 21:58:33 -04001696struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1697 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001698{
1699 struct btrfs_root *root;
1700
1701 spin_lock(&fs_info->fs_roots_radix_lock);
1702 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1703 (unsigned long)root_id);
1704 spin_unlock(&fs_info->fs_roots_radix_lock);
1705 return root;
1706}
1707
1708int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1709 struct btrfs_root *root)
1710{
1711 int ret;
1712
David Sterbae1860a72016-05-09 14:11:38 +02001713 ret = radix_tree_preload(GFP_NOFS);
Miao Xiecb517ea2013-05-15 07:48:19 +00001714 if (ret)
1715 return ret;
1716
1717 spin_lock(&fs_info->fs_roots_radix_lock);
1718 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1719 (unsigned long)root->root_key.objectid,
1720 root);
1721 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001722 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001723 spin_unlock(&fs_info->fs_roots_radix_lock);
1724 radix_tree_preload_end();
1725
1726 return ret;
1727}
1728
Miao Xiec00869f2013-09-25 21:47:44 +08001729struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1730 struct btrfs_key *location,
1731 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001732{
1733 struct btrfs_root *root;
David Sterba381cf652015-01-02 18:45:16 +01001734 struct btrfs_path *path;
David Sterba1d4c08e2015-01-02 19:36:14 +01001735 struct btrfs_key key;
Chris Mason5eda7b52007-06-22 14:16:25 -04001736 int ret;
1737
Chris Masonedbd8d42007-12-21 16:27:24 -05001738 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1739 return fs_info->tree_root;
1740 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1741 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001742 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1743 return fs_info->chunk_root;
1744 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1745 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001746 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1747 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001748 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1749 return fs_info->quota_root ? fs_info->quota_root :
1750 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001751 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1752 return fs_info->uuid_root ? fs_info->uuid_root :
1753 ERR_PTR(-ENOENT);
Omar Sandoval70f6d822015-09-29 20:50:38 -07001754 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1755 return fs_info->free_space_root ? fs_info->free_space_root :
1756 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001757again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001758 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001759 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001760 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001761 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001762 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001763 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001764
Miao Xiecb517ea2013-05-15 07:48:19 +00001765 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001766 if (IS_ERR(root))
1767 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001768
Miao Xiec00869f2013-09-25 21:47:44 +08001769 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001770 ret = -ENOENT;
1771 goto fail;
1772 }
1773
Miao Xiecb517ea2013-05-15 07:48:19 +00001774 ret = btrfs_init_fs_root(root);
1775 if (ret)
1776 goto fail;
1777
David Sterba381cf652015-01-02 18:45:16 +01001778 path = btrfs_alloc_path();
1779 if (!path) {
1780 ret = -ENOMEM;
1781 goto fail;
1782 }
David Sterba1d4c08e2015-01-02 19:36:14 +01001783 key.objectid = BTRFS_ORPHAN_OBJECTID;
1784 key.type = BTRFS_ORPHAN_ITEM_KEY;
1785 key.offset = location->objectid;
1786
1787 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
David Sterba381cf652015-01-02 18:45:16 +01001788 btrfs_free_path(path);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001789 if (ret < 0)
1790 goto fail;
1791 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001792 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001793
Miao Xiecb517ea2013-05-15 07:48:19 +00001794 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001795 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001796 if (ret == -EEXIST) {
1797 free_fs_root(root);
1798 goto again;
1799 }
1800 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001801 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001802 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001803fail:
1804 free_fs_root(root);
1805 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001806}
1807
Chris Mason04160082008-03-26 10:28:07 -04001808static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1809{
1810 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1811 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001812 struct btrfs_device *device;
1813 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001814
Xiao Guangrong1f781602011-04-20 10:09:16 +00001815 rcu_read_lock();
1816 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001817 if (!device->bdev)
1818 continue;
Chris Mason04160082008-03-26 10:28:07 -04001819 bdi = blk_get_backing_dev_info(device->bdev);
Tejun Heoff9ea322014-09-08 08:03:56 +09001820 if (bdi_congested(bdi, bdi_bits)) {
Chris Mason04160082008-03-26 10:28:07 -04001821 ret = 1;
1822 break;
1823 }
1824 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001825 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001826 return ret;
1827}
1828
Chris Mason04160082008-03-26 10:28:07 -04001829static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1830{
Jens Axboead081f12009-06-12 14:43:40 +02001831 int err;
1832
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001833 err = bdi_setup_and_register(bdi, "btrfs");
Jens Axboead081f12009-06-12 14:43:40 +02001834 if (err)
1835 return err;
1836
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03001837 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
Chris Mason04160082008-03-26 10:28:07 -04001838 bdi->congested_fn = btrfs_congested_fn;
1839 bdi->congested_data = info;
Chris Masonda2f0f72015-07-02 13:57:22 -07001840 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
Chris Mason04160082008-03-26 10:28:07 -04001841 return 0;
1842}
1843
Chris Mason8b712842008-06-11 16:50:36 -04001844/*
1845 * called by the kthread helper functions to finally call the bio end_io
1846 * functions. This is where read checksum verification actually happens
1847 */
1848static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001849{
Chris Masonce9adaa2008-04-09 16:28:12 -04001850 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001851 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001852
David Sterba97eb6b62014-07-30 00:55:42 +02001853 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001854 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001855
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001856 bio->bi_error = end_io_wq->error;
Chris Mason8b712842008-06-11 16:50:36 -04001857 bio->bi_private = end_io_wq->private;
1858 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001859 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001860 bio_endio(bio);
Chris Mason44b8bd72008-04-16 11:14:51 -04001861}
1862
Chris Masona74a4b92008-06-25 16:01:31 -04001863static int cleaner_kthread(void *arg)
1864{
1865 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001866 int again;
Filipe Mananada288d22015-06-13 06:55:31 +01001867 struct btrfs_trans_handle *trans;
Chris Masona74a4b92008-06-25 16:01:31 -04001868
1869 do {
Miao Xied0278242013-05-14 10:20:40 +00001870 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001871
Miao Xied0278242013-05-14 10:20:40 +00001872 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001873 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001874 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001875
Zygo Blaxell90c711a2016-06-12 23:39:58 -04001876 /*
1877 * Do not do anything if we might cause open_ctree() to block
1878 * before we have finished mounting the filesystem.
1879 */
Josef Bacikafcdd122016-09-02 15:40:02 -04001880 if (!test_bit(BTRFS_FS_OPEN, &root->fs_info->flags))
Zygo Blaxell90c711a2016-06-12 23:39:58 -04001881 goto sleep;
1882
Miao Xied0278242013-05-14 10:20:40 +00001883 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1884 goto sleep;
1885
Miao Xiedc7f3702013-05-14 10:20:42 +00001886 /*
1887 * Avoid the problem that we change the status of the fs
1888 * during the above check and trylock.
1889 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001890 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001891 mutex_unlock(&root->fs_info->cleaner_mutex);
1892 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001893 }
1894
Filipe Mananac2d6cb12016-01-15 11:05:12 +00001895 mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex);
Miao Xied0278242013-05-14 10:20:40 +00001896 btrfs_run_delayed_iputs(root);
Filipe Mananac2d6cb12016-01-15 11:05:12 +00001897 mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex);
1898
Miao Xied0278242013-05-14 10:20:40 +00001899 again = btrfs_clean_one_deleted_snapshot(root);
1900 mutex_unlock(&root->fs_info->cleaner_mutex);
1901
1902 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001903 * The defragger has dealt with the R/O remount and umount,
1904 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001905 */
1906 btrfs_run_defrag_inodes(root->fs_info);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01001907
1908 /*
1909 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1910 * with relocation (btrfs_relocate_chunk) and relocation
1911 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1912 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1913 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1914 * unused block groups.
1915 */
1916 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001917sleep:
Jiri Kosina838fe182016-03-15 11:28:54 +01001918 if (!again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001919 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001920 if (!kthread_should_stop())
1921 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001922 __set_current_state(TASK_RUNNING);
1923 }
1924 } while (!kthread_should_stop());
Filipe Mananada288d22015-06-13 06:55:31 +01001925
1926 /*
1927 * Transaction kthread is stopped before us and wakes us up.
1928 * However we might have started a new transaction and COWed some
1929 * tree blocks when deleting unused block groups for example. So
1930 * make sure we commit the transaction we started to have a clean
1931 * shutdown when evicting the btree inode - if it has dirty pages
1932 * when we do the final iput() on it, eviction will trigger a
1933 * writeback for it which will fail with null pointer dereferences
1934 * since work queues and other resources were already released and
1935 * destroyed by the time the iput/eviction/writeback is made.
1936 */
1937 trans = btrfs_attach_transaction(root);
1938 if (IS_ERR(trans)) {
1939 if (PTR_ERR(trans) != -ENOENT)
1940 btrfs_err(root->fs_info,
1941 "cleaner transaction attach returned %ld",
1942 PTR_ERR(trans));
1943 } else {
1944 int ret;
1945
1946 ret = btrfs_commit_transaction(trans, root);
1947 if (ret)
1948 btrfs_err(root->fs_info,
1949 "cleaner open transaction commit returned %d",
1950 ret);
1951 }
1952
Chris Masona74a4b92008-06-25 16:01:31 -04001953 return 0;
1954}
1955
1956static int transaction_kthread(void *arg)
1957{
1958 struct btrfs_root *root = arg;
1959 struct btrfs_trans_handle *trans;
1960 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001961 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001962 unsigned long now;
1963 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001964 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001965
1966 do {
Jan Kara914b2002012-03-12 16:05:50 +01001967 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001968 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001969 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1970
Josef Bacika4abeea2011-04-11 17:25:13 -04001971 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001972 cur = root->fs_info->running_transaction;
1973 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001974 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001975 goto sleep;
1976 }
Yan Zheng31153d82008-07-28 15:32:19 -04001977
Chris Masona74a4b92008-06-25 16:01:31 -04001978 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001979 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001980 (now < cur->start_time ||
1981 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001982 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001983 delay = HZ * 5;
1984 goto sleep;
1985 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001986 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001987 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001988
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001989 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001990 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001991 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001992 if (PTR_ERR(trans) != -ENOENT)
1993 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001994 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001995 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001996 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001997 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001998 } else {
1999 btrfs_end_transaction(trans, root);
2000 }
Chris Masona74a4b92008-06-25 16:01:31 -04002001sleep:
2002 wake_up_process(root->fs_info->cleaner_kthread);
2003 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
2004
Josef Bacik4e121c02013-09-27 16:32:39 -04002005 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
2006 &root->fs_info->fs_state)))
2007 btrfs_cleanup_transaction(root);
Jiri Kosinace63f892016-03-15 11:28:59 +01002008 set_current_state(TASK_INTERRUPTIBLE);
2009 if (!kthread_should_stop() &&
2010 (!btrfs_transaction_blocked(root->fs_info) ||
2011 cannot_commit))
2012 schedule_timeout(delay);
2013 __set_current_state(TASK_RUNNING);
Chris Masona74a4b92008-06-25 16:01:31 -04002014 } while (!kthread_should_stop());
2015 return 0;
2016}
2017
Chris Masonaf31f5e2011-11-03 15:17:42 -04002018/*
2019 * this will find the highest generation in the array of
2020 * root backups. The index of the highest array is returned,
2021 * or -1 if we can't find anything.
2022 *
2023 * We check to make sure the array is valid by comparing the
2024 * generation of the latest root in the array with the generation
2025 * in the super block. If they don't match we pitch it.
2026 */
2027static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
2028{
2029 u64 cur;
2030 int newest_index = -1;
2031 struct btrfs_root_backup *root_backup;
2032 int i;
2033
2034 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
2035 root_backup = info->super_copy->super_roots + i;
2036 cur = btrfs_backup_tree_root_gen(root_backup);
2037 if (cur == newest_gen)
2038 newest_index = i;
2039 }
2040
2041 /* check to see if we actually wrapped around */
2042 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
2043 root_backup = info->super_copy->super_roots;
2044 cur = btrfs_backup_tree_root_gen(root_backup);
2045 if (cur == newest_gen)
2046 newest_index = 0;
2047 }
2048 return newest_index;
2049}
2050
2051
2052/*
2053 * find the oldest backup so we know where to store new entries
2054 * in the backup array. This will set the backup_root_index
2055 * field in the fs_info struct
2056 */
2057static void find_oldest_super_backup(struct btrfs_fs_info *info,
2058 u64 newest_gen)
2059{
2060 int newest_index = -1;
2061
2062 newest_index = find_newest_super_backup(info, newest_gen);
2063 /* if there was garbage in there, just move along */
2064 if (newest_index == -1) {
2065 info->backup_root_index = 0;
2066 } else {
2067 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
2068 }
2069}
2070
2071/*
2072 * copy all the root pointers into the super backup array.
2073 * this will bump the backup pointer by one when it is
2074 * done
2075 */
2076static void backup_super_roots(struct btrfs_fs_info *info)
2077{
2078 int next_backup;
2079 struct btrfs_root_backup *root_backup;
2080 int last_backup;
2081
2082 next_backup = info->backup_root_index;
2083 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
2084 BTRFS_NUM_BACKUP_ROOTS;
2085
2086 /*
2087 * just overwrite the last backup if we're at the same generation
2088 * this happens only at umount
2089 */
2090 root_backup = info->super_for_commit->super_roots + last_backup;
2091 if (btrfs_backup_tree_root_gen(root_backup) ==
2092 btrfs_header_generation(info->tree_root->node))
2093 next_backup = last_backup;
2094
2095 root_backup = info->super_for_commit->super_roots + next_backup;
2096
2097 /*
2098 * make sure all of our padding and empty slots get zero filled
2099 * regardless of which ones we use today
2100 */
2101 memset(root_backup, 0, sizeof(*root_backup));
2102
2103 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
2104
2105 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
2106 btrfs_set_backup_tree_root_gen(root_backup,
2107 btrfs_header_generation(info->tree_root->node));
2108
2109 btrfs_set_backup_tree_root_level(root_backup,
2110 btrfs_header_level(info->tree_root->node));
2111
2112 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
2113 btrfs_set_backup_chunk_root_gen(root_backup,
2114 btrfs_header_generation(info->chunk_root->node));
2115 btrfs_set_backup_chunk_root_level(root_backup,
2116 btrfs_header_level(info->chunk_root->node));
2117
2118 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2119 btrfs_set_backup_extent_root_gen(root_backup,
2120 btrfs_header_generation(info->extent_root->node));
2121 btrfs_set_backup_extent_root_level(root_backup,
2122 btrfs_header_level(info->extent_root->node));
2123
Chris Mason7c7e82a2011-11-06 18:50:56 -05002124 /*
2125 * we might commit during log recovery, which happens before we set
2126 * the fs_root. Make sure it is valid before we fill it in.
2127 */
2128 if (info->fs_root && info->fs_root->node) {
2129 btrfs_set_backup_fs_root(root_backup,
2130 info->fs_root->node->start);
2131 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002132 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002133 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002134 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002135 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002136
2137 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2138 btrfs_set_backup_dev_root_gen(root_backup,
2139 btrfs_header_generation(info->dev_root->node));
2140 btrfs_set_backup_dev_root_level(root_backup,
2141 btrfs_header_level(info->dev_root->node));
2142
2143 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2144 btrfs_set_backup_csum_root_gen(root_backup,
2145 btrfs_header_generation(info->csum_root->node));
2146 btrfs_set_backup_csum_root_level(root_backup,
2147 btrfs_header_level(info->csum_root->node));
2148
2149 btrfs_set_backup_total_bytes(root_backup,
2150 btrfs_super_total_bytes(info->super_copy));
2151 btrfs_set_backup_bytes_used(root_backup,
2152 btrfs_super_bytes_used(info->super_copy));
2153 btrfs_set_backup_num_devices(root_backup,
2154 btrfs_super_num_devices(info->super_copy));
2155
2156 /*
2157 * if we don't copy this out to the super_copy, it won't get remembered
2158 * for the next commit
2159 */
2160 memcpy(&info->super_copy->super_roots,
2161 &info->super_for_commit->super_roots,
2162 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2163}
2164
2165/*
2166 * this copies info out of the root backup array and back into
2167 * the in-memory super block. It is meant to help iterate through
2168 * the array, so you send it the number of backups you've already
2169 * tried and the last backup index you used.
2170 *
2171 * this returns -1 when it has tried all the backups
2172 */
2173static noinline int next_root_backup(struct btrfs_fs_info *info,
2174 struct btrfs_super_block *super,
2175 int *num_backups_tried, int *backup_index)
2176{
2177 struct btrfs_root_backup *root_backup;
2178 int newest = *backup_index;
2179
2180 if (*num_backups_tried == 0) {
2181 u64 gen = btrfs_super_generation(super);
2182
2183 newest = find_newest_super_backup(info, gen);
2184 if (newest == -1)
2185 return -1;
2186
2187 *backup_index = newest;
2188 *num_backups_tried = 1;
2189 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2190 /* we've tried all the backups, all done */
2191 return -1;
2192 } else {
2193 /* jump to the next oldest backup */
2194 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2195 BTRFS_NUM_BACKUP_ROOTS;
2196 *backup_index = newest;
2197 *num_backups_tried += 1;
2198 }
2199 root_backup = super->super_roots + newest;
2200
2201 btrfs_set_super_generation(super,
2202 btrfs_backup_tree_root_gen(root_backup));
2203 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2204 btrfs_set_super_root_level(super,
2205 btrfs_backup_tree_root_level(root_backup));
2206 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2207
2208 /*
2209 * fixme: the total bytes and num_devices need to match or we should
2210 * need a fsck
2211 */
2212 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2213 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2214 return 0;
2215}
2216
Liu Bo7abadb62013-03-17 02:10:31 +00002217/* helper to cleanup workers */
2218static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2219{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002220 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002221 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002222 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002223 btrfs_destroy_workqueue(fs_info->endio_workers);
2224 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2225 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002226 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002227 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002228 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2229 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2230 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002231 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002232 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002233 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002234 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002235 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002236 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002237 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002238}
2239
Rashika2e9f5952013-10-31 02:45:20 +05302240static void free_root_extent_buffers(struct btrfs_root *root)
2241{
2242 if (root) {
2243 free_extent_buffer(root->node);
2244 free_extent_buffer(root->commit_root);
2245 root->node = NULL;
2246 root->commit_root = NULL;
2247 }
2248}
2249
Chris Masonaf31f5e2011-11-03 15:17:42 -04002250/* helper to cleanup tree roots */
2251static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2252{
Rashika2e9f5952013-10-31 02:45:20 +05302253 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002254
Rashika2e9f5952013-10-31 02:45:20 +05302255 free_root_extent_buffers(info->dev_root);
2256 free_root_extent_buffers(info->extent_root);
2257 free_root_extent_buffers(info->csum_root);
2258 free_root_extent_buffers(info->quota_root);
2259 free_root_extent_buffers(info->uuid_root);
2260 if (chunk_root)
2261 free_root_extent_buffers(info->chunk_root);
Omar Sandoval70f6d822015-09-29 20:50:38 -07002262 free_root_extent_buffers(info->free_space_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002263}
2264
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002265void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002266{
2267 int ret;
2268 struct btrfs_root *gang[8];
2269 int i;
2270
2271 while (!list_empty(&fs_info->dead_roots)) {
2272 gang[0] = list_entry(fs_info->dead_roots.next,
2273 struct btrfs_root, root_list);
2274 list_del(&gang[0]->root_list);
2275
Miao Xie27cdeb72014-04-02 19:51:05 +08002276 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002277 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002278 } else {
2279 free_extent_buffer(gang[0]->node);
2280 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002281 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002282 }
2283 }
2284
2285 while (1) {
2286 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2287 (void **)gang, 0,
2288 ARRAY_SIZE(gang));
2289 if (!ret)
2290 break;
2291 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002292 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002293 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002294
2295 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2296 btrfs_free_log_root_tree(NULL, fs_info);
2297 btrfs_destroy_pinned_extent(fs_info->tree_root,
2298 fs_info->pinned_extents);
2299 }
Josef Bacik171f6532013-04-24 16:35:41 -04002300}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002301
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002302static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2303{
2304 mutex_init(&fs_info->scrub_lock);
2305 atomic_set(&fs_info->scrubs_running, 0);
2306 atomic_set(&fs_info->scrub_pause_req, 0);
2307 atomic_set(&fs_info->scrubs_paused, 0);
2308 atomic_set(&fs_info->scrub_cancel_req, 0);
2309 init_waitqueue_head(&fs_info->scrub_pause_wait);
2310 fs_info->scrub_workers_refcnt = 0;
2311}
2312
Eric Sandeen779a65a2014-08-01 18:12:39 -05002313static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2314{
2315 spin_lock_init(&fs_info->balance_lock);
2316 mutex_init(&fs_info->balance_mutex);
2317 atomic_set(&fs_info->balance_running, 0);
2318 atomic_set(&fs_info->balance_pause_req, 0);
2319 atomic_set(&fs_info->balance_cancel_req, 0);
2320 fs_info->balance_ctl = NULL;
2321 init_waitqueue_head(&fs_info->balance_wait_q);
2322}
2323
Eric Sandeenf37938e2014-08-01 18:12:40 -05002324static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2325 struct btrfs_root *tree_root)
2326{
2327 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2328 set_nlink(fs_info->btree_inode, 1);
2329 /*
2330 * we set the i_size on the btree inode to the max possible int.
2331 * the real end of the address space is determined by all of
2332 * the devices in the system
2333 */
2334 fs_info->btree_inode->i_size = OFFSET_MAX;
2335 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Eric Sandeenf37938e2014-08-01 18:12:40 -05002336
2337 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2338 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2339 fs_info->btree_inode->i_mapping);
2340 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2341 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2342
2343 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2344
2345 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2346 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2347 sizeof(struct btrfs_key));
2348 set_bit(BTRFS_INODE_DUMMY,
2349 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2350 btrfs_insert_inode_hash(fs_info->btree_inode);
2351}
2352
Eric Sandeenad618362014-08-01 18:12:41 -05002353static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2354{
2355 fs_info->dev_replace.lock_owner = 0;
2356 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2357 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
Liu Bo73beece2015-07-17 16:49:19 +08002358 rwlock_init(&fs_info->dev_replace.lock);
2359 atomic_set(&fs_info->dev_replace.read_locks, 0);
2360 atomic_set(&fs_info->dev_replace.blocking_readers, 0);
Eric Sandeenad618362014-08-01 18:12:41 -05002361 init_waitqueue_head(&fs_info->replace_wait);
Liu Bo73beece2015-07-17 16:49:19 +08002362 init_waitqueue_head(&fs_info->dev_replace.read_lock_wq);
Eric Sandeenad618362014-08-01 18:12:41 -05002363}
2364
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002365static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2366{
2367 spin_lock_init(&fs_info->qgroup_lock);
2368 mutex_init(&fs_info->qgroup_ioctl_lock);
2369 fs_info->qgroup_tree = RB_ROOT;
2370 fs_info->qgroup_op_tree = RB_ROOT;
2371 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2372 fs_info->qgroup_seq = 1;
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002373 fs_info->qgroup_ulist = NULL;
Jeff Mahoneyd2c609b2016-08-15 12:10:33 -04002374 fs_info->qgroup_rescan_running = false;
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002375 mutex_init(&fs_info->qgroup_rescan_lock);
2376}
2377
Eric Sandeen2a458192015-02-16 16:29:26 +01002378static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2379 struct btrfs_fs_devices *fs_devices)
2380{
2381 int max_active = fs_info->thread_pool_size;
David Sterba6f011052015-02-16 18:34:01 +01002382 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Eric Sandeen2a458192015-02-16 16:29:26 +01002383
2384 fs_info->workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002385 btrfs_alloc_workqueue(fs_info, "worker",
2386 flags | WQ_HIGHPRI, max_active, 16);
Eric Sandeen2a458192015-02-16 16:29:26 +01002387
2388 fs_info->delalloc_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002389 btrfs_alloc_workqueue(fs_info, "delalloc",
2390 flags, max_active, 2);
Eric Sandeen2a458192015-02-16 16:29:26 +01002391
2392 fs_info->flush_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002393 btrfs_alloc_workqueue(fs_info, "flush_delalloc",
2394 flags, max_active, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002395
2396 fs_info->caching_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002397 btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002398
2399 /*
2400 * a higher idle thresh on the submit workers makes it much more
2401 * likely that bios will be send down in a sane order to the
2402 * devices
2403 */
2404 fs_info->submit_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002405 btrfs_alloc_workqueue(fs_info, "submit", flags,
Eric Sandeen2a458192015-02-16 16:29:26 +01002406 min_t(u64, fs_devices->num_devices,
2407 max_active), 64);
2408
2409 fs_info->fixup_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002410 btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002411
2412 /*
2413 * endios are largely parallel and should have a very
2414 * low idle thresh
2415 */
2416 fs_info->endio_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002417 btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4);
Eric Sandeen2a458192015-02-16 16:29:26 +01002418 fs_info->endio_meta_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002419 btrfs_alloc_workqueue(fs_info, "endio-meta", flags,
2420 max_active, 4);
Eric Sandeen2a458192015-02-16 16:29:26 +01002421 fs_info->endio_meta_write_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002422 btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags,
2423 max_active, 2);
Eric Sandeen2a458192015-02-16 16:29:26 +01002424 fs_info->endio_raid56_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002425 btrfs_alloc_workqueue(fs_info, "endio-raid56", flags,
2426 max_active, 4);
Eric Sandeen2a458192015-02-16 16:29:26 +01002427 fs_info->endio_repair_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002428 btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002429 fs_info->rmw_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002430 btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2);
Eric Sandeen2a458192015-02-16 16:29:26 +01002431 fs_info->endio_write_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002432 btrfs_alloc_workqueue(fs_info, "endio-write", flags,
2433 max_active, 2);
Eric Sandeen2a458192015-02-16 16:29:26 +01002434 fs_info->endio_freespace_worker =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002435 btrfs_alloc_workqueue(fs_info, "freespace-write", flags,
2436 max_active, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002437 fs_info->delayed_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002438 btrfs_alloc_workqueue(fs_info, "delayed-meta", flags,
2439 max_active, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002440 fs_info->readahead_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002441 btrfs_alloc_workqueue(fs_info, "readahead", flags,
2442 max_active, 2);
Eric Sandeen2a458192015-02-16 16:29:26 +01002443 fs_info->qgroup_rescan_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002444 btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0);
Eric Sandeen2a458192015-02-16 16:29:26 +01002445 fs_info->extent_workers =
Jeff Mahoneycb001092016-06-09 16:22:11 -04002446 btrfs_alloc_workqueue(fs_info, "extent-refs", flags,
Eric Sandeen2a458192015-02-16 16:29:26 +01002447 min_t(u64, fs_devices->num_devices,
2448 max_active), 8);
2449
2450 if (!(fs_info->workers && fs_info->delalloc_workers &&
2451 fs_info->submit_workers && fs_info->flush_workers &&
2452 fs_info->endio_workers && fs_info->endio_meta_workers &&
2453 fs_info->endio_meta_write_workers &&
2454 fs_info->endio_repair_workers &&
2455 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2456 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2457 fs_info->caching_workers && fs_info->readahead_workers &&
2458 fs_info->fixup_workers && fs_info->delayed_workers &&
2459 fs_info->extent_workers &&
2460 fs_info->qgroup_rescan_workers)) {
2461 return -ENOMEM;
2462 }
2463
2464 return 0;
2465}
2466
Eric Sandeen63443bf2014-08-01 18:12:46 -05002467static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2468 struct btrfs_fs_devices *fs_devices)
2469{
2470 int ret;
2471 struct btrfs_root *tree_root = fs_info->tree_root;
2472 struct btrfs_root *log_tree_root;
2473 struct btrfs_super_block *disk_super = fs_info->super_copy;
2474 u64 bytenr = btrfs_super_log_root(disk_super);
2475
2476 if (fs_devices->rw_devices == 0) {
David Sterbaf14d1042015-10-08 11:37:06 +02002477 btrfs_warn(fs_info, "log replay required on RO media");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002478 return -EIO;
2479 }
2480
David Sterba74e4d822016-02-11 11:01:55 +01002481 log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Eric Sandeen63443bf2014-08-01 18:12:46 -05002482 if (!log_tree_root)
2483 return -ENOMEM;
2484
2485 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2486 tree_root->stripesize, log_tree_root, fs_info,
2487 BTRFS_TREE_LOG_OBJECTID);
2488
2489 log_tree_root->node = read_tree_block(tree_root, bytenr,
2490 fs_info->generation + 1);
Liu Bo64c043d2015-05-25 17:30:15 +08002491 if (IS_ERR(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002492 btrfs_warn(fs_info, "failed to read log tree");
Liu Bo0eeff232015-06-11 14:16:44 +08002493 ret = PTR_ERR(log_tree_root->node);
Liu Bo64c043d2015-05-25 17:30:15 +08002494 kfree(log_tree_root);
Liu Bo0eeff232015-06-11 14:16:44 +08002495 return ret;
Liu Bo64c043d2015-05-25 17:30:15 +08002496 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002497 btrfs_err(fs_info, "failed to read log tree");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002498 free_extent_buffer(log_tree_root->node);
2499 kfree(log_tree_root);
2500 return -EIO;
2501 }
2502 /* returns with log_tree_root freed on success */
2503 ret = btrfs_recover_log_trees(log_tree_root);
2504 if (ret) {
Anand Jain34d97002016-03-16 16:43:06 +08002505 btrfs_handle_fs_error(tree_root->fs_info, ret,
Eric Sandeen63443bf2014-08-01 18:12:46 -05002506 "Failed to recover log tree");
2507 free_extent_buffer(log_tree_root->node);
2508 kfree(log_tree_root);
2509 return ret;
2510 }
2511
2512 if (fs_info->sb->s_flags & MS_RDONLY) {
2513 ret = btrfs_commit_super(tree_root);
2514 if (ret)
2515 return ret;
2516 }
2517
2518 return 0;
2519}
2520
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002521static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2522 struct btrfs_root *tree_root)
2523{
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002524 struct btrfs_root *root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002525 struct btrfs_key location;
2526 int ret;
2527
2528 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2529 location.type = BTRFS_ROOT_ITEM_KEY;
2530 location.offset = 0;
2531
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002532 root = btrfs_read_tree_root(tree_root, &location);
2533 if (IS_ERR(root))
2534 return PTR_ERR(root);
2535 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2536 fs_info->extent_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002537
2538 location.objectid = BTRFS_DEV_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002539 root = btrfs_read_tree_root(tree_root, &location);
2540 if (IS_ERR(root))
2541 return PTR_ERR(root);
2542 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2543 fs_info->dev_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002544 btrfs_init_devices_late(fs_info);
2545
2546 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002547 root = btrfs_read_tree_root(tree_root, &location);
2548 if (IS_ERR(root))
2549 return PTR_ERR(root);
2550 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2551 fs_info->csum_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002552
2553 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002554 root = btrfs_read_tree_root(tree_root, &location);
2555 if (!IS_ERR(root)) {
2556 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Josef Bacikafcdd122016-09-02 15:40:02 -04002557 set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002558 fs_info->quota_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002559 }
2560
2561 location.objectid = BTRFS_UUID_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002562 root = btrfs_read_tree_root(tree_root, &location);
2563 if (IS_ERR(root)) {
2564 ret = PTR_ERR(root);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002565 if (ret != -ENOENT)
2566 return ret;
2567 } else {
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002568 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2569 fs_info->uuid_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002570 }
2571
Omar Sandoval70f6d822015-09-29 20:50:38 -07002572 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2573 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2574 root = btrfs_read_tree_root(tree_root, &location);
2575 if (IS_ERR(root))
2576 return PTR_ERR(root);
2577 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2578 fs_info->free_space_root = root;
2579 }
2580
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002581 return 0;
2582}
2583
Al Viroad2b2c82011-11-17 01:10:02 -05002584int open_ctree(struct super_block *sb,
2585 struct btrfs_fs_devices *fs_devices,
2586 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002587{
Chris Masondb945352007-10-15 16:15:53 -04002588 u32 sectorsize;
2589 u32 nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002590 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002591 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002592 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002593 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002594 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002595 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002596 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002597 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002598 struct btrfs_root *chunk_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002599 int ret;
Yane58ca022008-04-01 11:21:34 -04002600 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002601 int num_backups_tried = 0;
2602 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002603 int max_active;
Omar Sandoval6675df32016-09-22 17:24:22 -07002604 int clear_free_space_tree = 0;
Chris Mason4543df72008-06-11 21:47:56 -04002605
David Sterba74e4d822016-02-11 11:01:55 +01002606 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
2607 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
Miao Xiecb517ea2013-05-15 07:48:19 +00002608 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002609 err = -ENOMEM;
2610 goto fail;
2611 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002612
2613 ret = init_srcu_struct(&fs_info->subvol_srcu);
2614 if (ret) {
2615 err = ret;
2616 goto fail;
2617 }
2618
2619 ret = setup_bdi(fs_info, &fs_info->bdi);
2620 if (ret) {
2621 err = ret;
2622 goto fail_srcu;
2623 }
2624
Tejun Heo908c7f12014-09-08 09:51:29 +09002625 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
Miao Xiee2d84522013-01-29 10:09:20 +00002626 if (ret) {
2627 err = ret;
2628 goto fail_bdi;
2629 }
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002630 fs_info->dirty_metadata_batch = PAGE_SIZE *
Miao Xiee2d84522013-01-29 10:09:20 +00002631 (1 + ilog2(nr_cpu_ids));
2632
Tejun Heo908c7f12014-09-08 09:51:29 +09002633 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
Miao Xie963d6782013-01-29 10:10:51 +00002634 if (ret) {
2635 err = ret;
2636 goto fail_dirty_metadata_bytes;
2637 }
2638
Tejun Heo908c7f12014-09-08 09:51:29 +09002639 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
Miao Xiec404e0d2014-01-30 16:46:55 +08002640 if (ret) {
2641 err = ret;
2642 goto fail_delalloc_bytes;
2643 }
2644
Yan, Zheng76dda932009-09-21 16:00:26 -04002645 fs_info->btree_inode = new_inode(sb);
2646 if (!fs_info->btree_inode) {
2647 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002648 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002649 }
2650
Chris Masona6591712011-07-19 12:04:14 -04002651 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002652
Yan, Zheng76dda932009-09-21 16:00:26 -04002653 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002654 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002655 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002656 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002657 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002658 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002659 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002660 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002661 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002662 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002663 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002664 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002665 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002666 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002667 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002668 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002669 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002670 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002671 rwlock_init(&fs_info->tree_mod_log_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002672 mutex_init(&fs_info->unused_bg_unpin_mutex);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01002673 mutex_init(&fs_info->delete_unused_bgs_mutex);
Chris Mason75857172011-06-13 20:00:16 -04002674 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002675 mutex_init(&fs_info->delalloc_root_mutex);
Filipe Mananac2d6cb12016-01-15 11:05:12 +00002676 mutex_init(&fs_info->cleaner_delayed_iput_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002677 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002678
Chris Mason0b86a832008-03-24 15:01:56 -04002679 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002680 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002681 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002682 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002683 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002684 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2685 BTRFS_BLOCK_RSV_GLOBAL);
2686 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2687 BTRFS_BLOCK_RSV_DELALLOC);
2688 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2689 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2690 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2691 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2692 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002693 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002694 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002695 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002696 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002697 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002698 atomic_set(&fs_info->qgroup_op_seq, 0);
Zhao Lei2fefd552016-01-07 18:38:48 +08002699 atomic_set(&fs_info->reada_works_cnt, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002700 atomic64_set(&fs_info->tree_mod_seq, 0);
Wang Xiaoguang9e7cc912016-08-01 13:28:08 +08002701 fs_info->fs_frozen = 0;
Chris Masone20d96d2007-03-22 12:13:20 -04002702 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002703 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002704 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002705 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002706 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002707 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002708 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
David Sterbaf8c269d2015-01-16 17:21:12 +01002709 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
Arne Jansen90519d62011-05-23 14:30:00 +02002710 /* readahead state */
Mel Gormand0164ad2015-11-06 16:28:21 -08002711 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
Arne Jansen90519d62011-05-23 14:30:00 +02002712 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002713
Chris Masonb34b0862008-12-19 15:43:22 -05002714 fs_info->thread_pool_size = min_t(unsigned long,
2715 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002716
Miao Xie199c2a92013-05-15 07:48:23 +00002717 INIT_LIST_HEAD(&fs_info->ordered_roots);
2718 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002719 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
David Sterba74e4d822016-02-11 11:01:55 +01002720 GFP_KERNEL);
Miao Xie16cdcec2011-04-22 18:12:22 +08002721 if (!fs_info->delayed_root) {
2722 err = -ENOMEM;
2723 goto fail_iput;
2724 }
2725 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002726
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002727 btrfs_init_scrub(fs_info);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002728#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2729 fs_info->check_integrity_print_mask = 0;
2730#endif
Eric Sandeen779a65a2014-08-01 18:12:39 -05002731 btrfs_init_balance(fs_info);
Miao Xie21c7e752014-05-13 17:29:04 -07002732 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Arne Jansena2de7332011-03-08 14:14:00 +01002733
Chris Masona061fc82008-05-07 11:43:44 -04002734 sb->s_blocksize = 4096;
2735 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002736 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04002737
Eric Sandeenf37938e2014-08-01 18:12:40 -05002738 btrfs_init_btree_inode(fs_info, tree_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002739
Chris Masone02119d2008-09-05 16:13:11 -04002740 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002741 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002742 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002743
Yan Zheng11833d62009-09-11 16:11:19 -04002744 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002745 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002746 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002747 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002748 fs_info->pinned_extents = &fs_info->freed_extents[0];
Josef Bacikafcdd122016-09-02 15:40:02 -04002749 set_bit(BTRFS_FS_BARRIER, &fs_info->flags);
Chris Mason509659c2007-05-10 12:36:17 -04002750
Chris Mason5a3f23d2009-03-31 13:27:11 -04002751 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason79154b12007-03-22 15:59:16 -04002752 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002753 mutex_init(&fs_info->chunk_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002754 mutex_init(&fs_info->transaction_kthread_mutex);
2755 mutex_init(&fs_info->cleaner_mutex);
Chris Mason7d9eb122008-07-08 14:19:17 -04002756 mutex_init(&fs_info->volume_mutex);
Chris Mason1bbc6212015-04-06 12:46:08 -07002757 mutex_init(&fs_info->ro_block_group_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002758 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002759 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002760 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002761 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002762
Eric Sandeenad618362014-08-01 18:12:41 -05002763 btrfs_init_dev_replace_locks(fs_info);
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002764 btrfs_init_qgroup(fs_info);
Arne Jansen416ac512011-09-13 12:56:09 +02002765
Chris Masonfa9c0d792009-04-03 09:47:43 -04002766 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2767 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2768
Chris Masone6dcd2d2008-07-17 12:53:50 -04002769 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Masonf9295742008-07-17 12:54:14 -04002770 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002771 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002772 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason3768f362007-03-13 16:47:54 -04002773
Filipe Manana04216822014-11-27 21:14:15 +00002774 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2775
David Woodhouse53b381b2013-01-29 18:40:14 -05002776 ret = btrfs_alloc_stripe_hash_table(fs_info);
2777 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002778 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002779 goto fail_alloc;
2780 }
2781
David Sterba707e8a02014-06-04 19:22:26 +02002782 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002783 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002784
Chris Mason3c4bb262012-03-27 18:56:56 -04002785 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002786
2787 /*
2788 * Read super block and check the signature bytes only
2789 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002790 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Anand Jain92fc03f2015-08-14 18:32:51 +08002791 if (IS_ERR(bh)) {
2792 err = PTR_ERR(bh);
Miao Xie16cdcec2011-04-22 18:12:22 +08002793 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002794 }
Chris Mason39279cc2007-06-12 06:35:45 -04002795
David Sterba1104a882013-03-06 15:57:46 +01002796 /*
2797 * We want to check superblock checksum, the type is stored inside.
2798 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2799 */
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04002800 if (btrfs_check_super_csum(fs_info, bh->b_data)) {
David Sterba05135f52016-05-09 11:32:39 +02002801 btrfs_err(fs_info, "superblock checksum mismatch");
David Sterba1104a882013-03-06 15:57:46 +01002802 err = -EINVAL;
Anand Jainb2acddd2015-10-07 17:23:23 +08002803 brelse(bh);
David Sterba1104a882013-03-06 15:57:46 +01002804 goto fail_alloc;
2805 }
2806
2807 /*
2808 * super_copy is zeroed at allocation time and we never touch the
2809 * following bytes up to INFO_SIZE, the checksum is calculated from
2810 * the whole block of INFO_SIZE
2811 */
David Sterba6c417612011-04-13 15:41:04 +02002812 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2813 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2814 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002815 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002816
David Sterba6c417612011-04-13 15:41:04 +02002817 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002818
David Sterba1104a882013-03-06 15:57:46 +01002819 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2820 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02002821 btrfs_err(fs_info, "superblock contains fatal errors");
David Sterba1104a882013-03-06 15:57:46 +01002822 err = -EINVAL;
2823 goto fail_alloc;
2824 }
2825
David Sterba6c417612011-04-13 15:41:04 +02002826 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002827 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002828 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002829
liuboacce9522011-01-06 19:30:25 +08002830 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002831 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2832 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002833
Liu Bo75e7cb72011-03-22 10:12:20 +00002834 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002835 * run through our array of backup supers and setup
2836 * our ring pointer to the oldest one
2837 */
2838 generation = btrfs_super_generation(disk_super);
2839 find_oldest_super_backup(fs_info, generation);
2840
2841 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002842 * In the long term, we'll store the compression type in the super
2843 * block, and it'll be used for per file compression control.
2844 */
2845 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2846
Qu Wenruo96da0912016-01-19 10:23:03 +08002847 ret = btrfs_parse_options(tree_root, options, sb->s_flags);
Yan Zheng2b820322008-11-17 21:11:30 -05002848 if (ret) {
2849 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002850 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002851 }
Chris Masondfe25022008-05-13 13:46:40 -04002852
Josef Bacikf2b636e2008-12-02 06:36:08 -05002853 features = btrfs_super_incompat_flags(disk_super) &
2854 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2855 if (features) {
David Sterba05135f52016-05-09 11:32:39 +02002856 btrfs_err(fs_info,
2857 "cannot mount because of unsupported optional features (%llx)",
2858 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002859 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002860 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002861 }
2862
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002863 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002864 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002865 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002866 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002867
Josef Bacik3173a182013-03-07 14:22:04 -05002868 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
David Sterba05135f52016-05-09 11:32:39 +02002869 btrfs_info(fs_info, "has skinny extents");
Josef Bacik3173a182013-03-07 14:22:04 -05002870
Chris Mason727011e2010-08-06 13:21:20 -04002871 /*
2872 * flag our filesystem as having big metadata blocks if
2873 * they are bigger than the page size
2874 */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002875 if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002876 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
David Sterba05135f52016-05-09 11:32:39 +02002877 btrfs_info(fs_info,
2878 "flagging fs with big metadata feature");
Chris Mason727011e2010-08-06 13:21:20 -04002879 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2880 }
2881
Chris Masonbc3f1162012-03-29 17:02:47 -04002882 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002883 sectorsize = btrfs_super_sectorsize(disk_super);
Chandan Rajendrab7f67052016-06-23 15:16:44 +05302884 stripesize = sectorsize;
David Sterba707e8a02014-06-04 19:22:26 +02002885 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002886 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002887
2888 /*
2889 * mixed block groups end up with duplicate but slightly offset
2890 * extent buffers for the same range. It leads to corruptions
2891 */
2892 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002893 (sectorsize != nodesize)) {
David Sterba05135f52016-05-09 11:32:39 +02002894 btrfs_err(fs_info,
2895"unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups",
2896 nodesize, sectorsize);
Chris Masonbc3f1162012-03-29 17:02:47 -04002897 goto fail_alloc;
2898 }
2899
Miao Xieceda0862013-04-11 10:30:16 +00002900 /*
2901 * Needn't use the lock because there is no other task which will
2902 * update the flag.
2903 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002904 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002905
Josef Bacikf2b636e2008-12-02 06:36:08 -05002906 features = btrfs_super_compat_ro_flags(disk_super) &
2907 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2908 if (!(sb->s_flags & MS_RDONLY) && features) {
David Sterba05135f52016-05-09 11:32:39 +02002909 btrfs_err(fs_info,
2910 "cannot mount read-write because of unsupported optional features (%llx)",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002911 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002912 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002913 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002914 }
Chris Mason61d92c32009-10-02 19:11:56 -04002915
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002916 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002917
Eric Sandeen2a458192015-02-16 16:29:26 +01002918 ret = btrfs_init_workqueues(fs_info, fs_devices);
2919 if (ret) {
2920 err = ret;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002921 goto fail_sb_buffer;
2922 }
Chris Mason4543df72008-06-11 21:47:56 -04002923
Chris Mason4575c9c2008-04-18 16:13:31 -04002924 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002925 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03002926 SZ_4M / PAGE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002927
Chris Masondb945352007-10-15 16:15:53 -04002928 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002929 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002930 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002931
2932 sb->s_blocksize = sectorsize;
2933 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002934
Chris Mason925baed2008-06-25 16:01:30 -04002935 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002936 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002937 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002938 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02002939 btrfs_err(fs_info, "failed to read the system array: %d", ret);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002940 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002941 }
Chris Mason0b86a832008-03-24 15:01:56 -04002942
Yan Zheng84234f32008-10-29 14:49:05 -04002943 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002944
David Sterba707e8a02014-06-04 19:22:26 +02002945 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2946 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002947
2948 chunk_root->node = read_tree_block(chunk_root,
2949 btrfs_super_chunk_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002950 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002951 if (IS_ERR(chunk_root->node) ||
2952 !extent_buffer_uptodate(chunk_root->node)) {
David Sterba05135f52016-05-09 11:32:39 +02002953 btrfs_err(fs_info, "failed to read chunk root");
chandane5fffba2015-10-05 22:14:25 +05302954 if (!IS_ERR(chunk_root->node))
2955 free_extent_buffer(chunk_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002956 chunk_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002957 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002958 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002959 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2960 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002961
Chris Masone17cade2008-04-15 15:41:47 -04002962 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002963 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002964
Chris Mason0b86a832008-03-24 15:01:56 -04002965 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002966 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02002967 btrfs_err(fs_info, "failed to read chunk tree: %d", ret);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002968 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002969 }
Chris Mason0b86a832008-03-24 15:01:56 -04002970
Stefan Behrens8dabb742012-11-06 13:15:27 +01002971 /*
2972 * keep the device that is marked to be the target device for the
2973 * dev_replace procedure
2974 */
Eric Sandeen9eaed212014-08-01 18:12:35 -05002975 btrfs_close_extra_devices(fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002976
Chris Masona6b0d5c2012-02-20 20:53:43 -05002977 if (!fs_devices->latest_bdev) {
David Sterba05135f52016-05-09 11:32:39 +02002978 btrfs_err(fs_info, "failed to read devices");
Chris Masona6b0d5c2012-02-20 20:53:43 -05002979 goto fail_tree_roots;
2980 }
2981
Chris Masonaf31f5e2011-11-03 15:17:42 -04002982retry_root_backup:
Yan Zheng84234f32008-10-29 14:49:05 -04002983 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002984
Chris Masone20d96d2007-03-22 12:13:20 -04002985 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002986 btrfs_super_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002987 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002988 if (IS_ERR(tree_root->node) ||
2989 !extent_buffer_uptodate(tree_root->node)) {
David Sterba05135f52016-05-09 11:32:39 +02002990 btrfs_warn(fs_info, "failed to read tree root");
chandane5fffba2015-10-05 22:14:25 +05302991 if (!IS_ERR(tree_root->node))
2992 free_extent_buffer(tree_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002993 tree_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002994 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002995 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002996
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002997 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2998 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002999 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04003000
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05303001 mutex_lock(&tree_root->objectid_mutex);
3002 ret = btrfs_find_highest_objectid(tree_root,
3003 &tree_root->highest_objectid);
3004 if (ret) {
3005 mutex_unlock(&tree_root->objectid_mutex);
3006 goto recovery_tree_root;
3007 }
3008
3009 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
3010
3011 mutex_unlock(&tree_root->objectid_mutex);
3012
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003013 ret = btrfs_read_roots(fs_info, tree_root);
3014 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04003015 goto recovery_tree_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003016
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003017 fs_info->generation = generation;
3018 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003019
Ilya Dryomov68310a52012-06-22 12:24:12 -06003020 ret = btrfs_recover_balance(fs_info);
3021 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003022 btrfs_err(fs_info, "failed to recover balance: %d", ret);
Ilya Dryomov68310a52012-06-22 12:24:12 -06003023 goto fail_block_groups;
3024 }
3025
Stefan Behrens733f4fb2012-05-25 16:06:10 +02003026 ret = btrfs_init_dev_stats(fs_info);
3027 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003028 btrfs_err(fs_info, "failed to init dev_stats: %d", ret);
Stefan Behrens733f4fb2012-05-25 16:06:10 +02003029 goto fail_block_groups;
3030 }
3031
Stefan Behrens8dabb742012-11-06 13:15:27 +01003032 ret = btrfs_init_dev_replace(fs_info);
3033 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003034 btrfs_err(fs_info, "failed to init dev_replace: %d", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01003035 goto fail_block_groups;
3036 }
3037
Eric Sandeen9eaed212014-08-01 18:12:35 -05003038 btrfs_close_extra_devices(fs_devices, 1);
Stefan Behrens8dabb742012-11-06 13:15:27 +01003039
Anand Jainb7c35e82015-03-10 06:38:38 +08003040 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
3041 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003042 btrfs_err(fs_info, "failed to init sysfs fsid interface: %d",
3043 ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08003044 goto fail_block_groups;
3045 }
3046
3047 ret = btrfs_sysfs_add_device(fs_devices);
3048 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003049 btrfs_err(fs_info, "failed to init sysfs device interface: %d",
3050 ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08003051 goto fail_fsdev_sysfs;
3052 }
3053
Anand Jain96f31362015-08-14 18:32:46 +08003054 ret = btrfs_sysfs_add_mounted(fs_info);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003055 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003056 btrfs_err(fs_info, "failed to init sysfs interface: %d", ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08003057 goto fail_fsdev_sysfs;
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003058 }
3059
liuboc59021f2011-03-07 02:13:14 +00003060 ret = btrfs_init_space_info(fs_info);
3061 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003062 btrfs_err(fs_info, "failed to initialize space info: %d", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08003063 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00003064 }
3065
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003066 ret = btrfs_read_block_groups(fs_info->extent_root);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003067 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003068 btrfs_err(fs_info, "failed to read block groups: %d", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08003069 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003070 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003071 fs_info->num_tolerated_disk_barrier_failures =
3072 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00003073 if (fs_info->fs_devices->missing_devices >
3074 fs_info->num_tolerated_disk_barrier_failures &&
3075 !(sb->s_flags & MS_RDONLY)) {
David Sterba05135f52016-05-09 11:32:39 +02003076 btrfs_warn(fs_info,
3077"missing devices (%llu) exceeds the limit (%d), writeable mount is not allowed",
Zhao Lei78fa1772015-07-20 17:54:50 +08003078 fs_info->fs_devices->missing_devices,
3079 fs_info->num_tolerated_disk_barrier_failures);
Anand Jain2365dd32014-01-22 11:15:51 +08003080 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00003081 }
Chris Mason9078a3e2007-04-26 16:46:15 -04003082
Chris Masona74a4b92008-06-25 16:01:31 -04003083 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
3084 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05003085 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08003086 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04003087
3088 fs_info->transaction_kthread = kthread_run(transaction_kthread,
3089 tree_root,
3090 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05003091 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04003092 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04003093
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003094 if (!btrfs_test_opt(tree_root->fs_info, SSD) &&
3095 !btrfs_test_opt(tree_root->fs_info, NOSSD) &&
Chris Masonc2898112009-06-10 09:51:32 -04003096 !fs_info->fs_devices->rotating) {
David Sterba05135f52016-05-09 11:32:39 +02003097 btrfs_info(fs_info, "detected SSD devices, enabling SSD mode");
Chris Masonc2898112009-06-10 09:51:32 -04003098 btrfs_set_opt(fs_info->mount_opt, SSD);
3099 }
3100
David Sterba572d9ab2014-02-05 15:26:17 +01003101 /*
Nicholas D Steeves01327612016-05-19 21:18:45 -04003102 * Mount does not set all options immediately, we can do it now and do
David Sterba572d9ab2014-02-05 15:26:17 +01003103 * not have to wait for transaction commit
3104 */
3105 btrfs_apply_pending_changes(fs_info);
Qu Wenruo3818aea2014-01-13 13:36:06 +08003106
Stefan Behrens21adbd52011-11-09 13:44:05 +01003107#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003108 if (btrfs_test_opt(tree_root->fs_info, CHECK_INTEGRITY)) {
Stefan Behrens21adbd52011-11-09 13:44:05 +01003109 ret = btrfsic_mount(tree_root, fs_devices,
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003110 btrfs_test_opt(tree_root->fs_info,
Stefan Behrens21adbd52011-11-09 13:44:05 +01003111 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3112 1 : 0,
3113 fs_info->check_integrity_print_mask);
3114 if (ret)
David Sterba05135f52016-05-09 11:32:39 +02003115 btrfs_warn(fs_info,
3116 "failed to initialize integrity check module: %d",
3117 ret);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003118 }
3119#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02003120 ret = btrfs_read_qgroup_config(fs_info);
3121 if (ret)
3122 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01003123
Qu Wenruo96da0912016-01-19 10:23:03 +08003124 /* do not make disk changes in broken FS or nologreplay is given */
3125 if (btrfs_super_log_root(disk_super) != 0 &&
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003126 !btrfs_test_opt(tree_root->fs_info, NOLOGREPLAY)) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003127 ret = btrfs_replay_log(fs_info, fs_devices);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003128 if (ret) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003129 err = ret;
Wang Shilong28c16cb2014-04-23 19:33:35 +08003130 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003131 }
Chris Masone02119d2008-09-05 16:13:11 -04003132 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003133
Yan, Zheng76dda932009-09-21 16:00:26 -04003134 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003135 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003136 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04003137
Chris Mason7c2ca462008-11-19 15:13:35 -05003138 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04003139 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06003140 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003141 goto fail_qgroup;
Zygo Blaxell90c711a2016-06-12 23:39:58 -04003142
3143 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003144 ret = btrfs_recover_relocation(tree_root);
Zygo Blaxell90c711a2016-06-12 23:39:58 -04003145 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00003146 if (ret < 0) {
David Sterba05135f52016-05-09 11:32:39 +02003147 btrfs_warn(fs_info, "failed to recover relocation: %d",
3148 ret);
Miao Xied7ce5842010-02-02 08:46:44 +00003149 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02003150 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00003151 }
Chris Mason7c2ca462008-11-19 15:13:35 -05003152 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003153
Chris Mason3de45862008-11-17 21:02:50 -05003154 location.objectid = BTRFS_FS_TREE_OBJECTID;
3155 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00003156 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05003157
Chris Mason3de45862008-11-17 21:02:50 -05003158 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003159 if (IS_ERR(fs_info->fs_root)) {
3160 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003161 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003162 }
Chris Masonc2898112009-06-10 09:51:32 -04003163
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003164 if (sb->s_flags & MS_RDONLY)
3165 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02003166
Omar Sandovalf8d468a2016-09-22 17:24:21 -07003167 if (btrfs_test_opt(fs_info, CLEAR_CACHE) &&
3168 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
Omar Sandoval6675df32016-09-22 17:24:22 -07003169 clear_free_space_tree = 1;
3170 } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
3171 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) {
3172 btrfs_warn(fs_info, "free space tree is invalid");
3173 clear_free_space_tree = 1;
3174 }
3175
3176 if (clear_free_space_tree) {
Omar Sandovalf8d468a2016-09-22 17:24:21 -07003177 btrfs_info(fs_info, "clearing free space tree");
3178 ret = btrfs_clear_free_space_tree(fs_info);
3179 if (ret) {
3180 btrfs_warn(fs_info,
3181 "failed to clear free space tree: %d", ret);
3182 close_ctree(tree_root);
3183 return ret;
3184 }
3185 }
3186
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003187 if (btrfs_test_opt(tree_root->fs_info, FREE_SPACE_TREE) &&
Chris Mason511711a2015-12-30 07:52:35 -08003188 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
David Sterba05135f52016-05-09 11:32:39 +02003189 btrfs_info(fs_info, "creating free space tree");
Chris Mason511711a2015-12-30 07:52:35 -08003190 ret = btrfs_create_free_space_tree(fs_info);
3191 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003192 btrfs_warn(fs_info,
3193 "failed to create free space tree: %d", ret);
Chris Mason511711a2015-12-30 07:52:35 -08003194 close_ctree(tree_root);
3195 return ret;
3196 }
3197 }
3198
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003199 down_read(&fs_info->cleanup_work_sem);
3200 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3201 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00003202 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003203 close_ctree(tree_root);
3204 return ret;
3205 }
3206 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02003207
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003208 ret = btrfs_resume_balance_async(fs_info);
3209 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003210 btrfs_warn(fs_info, "failed to resume balance: %d", ret);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003211 close_ctree(tree_root);
3212 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00003213 }
3214
Stefan Behrens8dabb742012-11-06 13:15:27 +01003215 ret = btrfs_resume_dev_replace_async(fs_info);
3216 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003217 btrfs_warn(fs_info, "failed to resume device replace: %d", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01003218 close_ctree(tree_root);
3219 return ret;
3220 }
3221
Jan Schmidtb382a322013-05-28 15:47:24 +00003222 btrfs_qgroup_rescan_resume(fs_info);
3223
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003224 if (!fs_info->uuid_root) {
David Sterba05135f52016-05-09 11:32:39 +02003225 btrfs_info(fs_info, "creating UUID tree");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003226 ret = btrfs_create_uuid_tree(fs_info);
3227 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003228 btrfs_warn(fs_info,
3229 "failed to create the UUID tree: %d", ret);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003230 close_ctree(tree_root);
3231 return ret;
3232 }
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003233 } else if (btrfs_test_opt(tree_root->fs_info, RESCAN_UUID_TREE) ||
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003234 fs_info->generation !=
3235 btrfs_super_uuid_tree_generation(disk_super)) {
David Sterba05135f52016-05-09 11:32:39 +02003236 btrfs_info(fs_info, "checking UUID tree");
Stefan Behrens70f80172013-08-15 17:11:23 +02003237 ret = btrfs_check_uuid_tree(fs_info);
3238 if (ret) {
David Sterba05135f52016-05-09 11:32:39 +02003239 btrfs_warn(fs_info,
3240 "failed to check the UUID tree: %d", ret);
Stefan Behrens70f80172013-08-15 17:11:23 +02003241 close_ctree(tree_root);
3242 return ret;
3243 }
3244 } else {
Josef Bacikafcdd122016-09-02 15:40:02 -04003245 set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003246 }
Josef Bacikafcdd122016-09-02 15:40:02 -04003247 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
Josef Bacik47ab2a62014-09-18 11:20:02 -04003248
Qu Wenruo8dcddfa2016-01-19 10:23:02 +08003249 /*
3250 * backuproot only affect mount behavior, and if open_ctree succeeded,
3251 * no need to keep the flag
3252 */
3253 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
3254
Al Viroad2b2c82011-11-17 01:10:02 -05003255 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04003256
Arne Jansenbcef60f2011-09-13 15:23:30 +02003257fail_qgroup:
3258 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05003259fail_trans_kthread:
3260 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04003261 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003262 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04003263fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04003264 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05003265
3266 /*
3267 * make sure we're done with the btree inode before we stop our
3268 * kthreads
3269 */
3270 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05003271
Anand Jain2365dd32014-01-22 11:15:51 +08003272fail_sysfs:
Anand Jain6618a592015-08-14 18:32:47 +08003273 btrfs_sysfs_remove_mounted(fs_info);
Anand Jain2365dd32014-01-22 11:15:51 +08003274
Anand Jainb7c35e82015-03-10 06:38:38 +08003275fail_fsdev_sysfs:
3276 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3277
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003278fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003279 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003280 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003281
3282fail_tree_roots:
3283 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003284 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003285
Chris Mason39279cc2007-06-12 06:35:45 -04003286fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003287 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003288fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003289fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003290 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3291
Chris Mason4543df72008-06-11 21:47:56 -04003292 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003293fail_bio_counter:
3294 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003295fail_delalloc_bytes:
3296 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003297fail_dirty_metadata_bytes:
3298 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003299fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003300 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003301fail_srcu:
3302 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003303fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003304 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003305 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003306 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003307
3308recovery_tree_root:
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003309 if (!btrfs_test_opt(tree_root->fs_info, USEBACKUPROOT))
Chris Masonaf31f5e2011-11-03 15:17:42 -04003310 goto fail_tree_roots;
3311
3312 free_root_pointers(fs_info, 0);
3313
3314 /* don't use the log in recovery mode, it won't be valid */
3315 btrfs_set_super_log_root(disk_super, 0);
3316
3317 /* we can't trust the free space cache either */
3318 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3319
3320 ret = next_root_backup(fs_info, fs_info->super_copy,
3321 &num_backups_tried, &backup_index);
3322 if (ret == -1)
3323 goto fail_block_groups;
3324 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003325}
3326
Chris Masonf2984462008-04-10 16:19:33 -04003327static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3328{
Chris Masonf2984462008-04-10 16:19:33 -04003329 if (uptodate) {
3330 set_buffer_uptodate(bh);
3331 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003332 struct btrfs_device *device = (struct btrfs_device *)
3333 bh->b_private;
3334
David Sterbab14af3b2015-10-08 10:43:10 +02003335 btrfs_warn_rl_in_rcu(device->dev_root->fs_info,
3336 "lost page write due to IO error on %s",
Josef Bacik606686e2012-06-04 14:03:51 -04003337 rcu_str_deref(device->name));
Nicholas D Steeves01327612016-05-19 21:18:45 -04003338 /* note, we don't set_buffer_write_io_error because we have
Chris Mason1259ab72008-05-12 13:39:03 -04003339 * our own ways of dealing with the IO errors
3340 */
Chris Masonf2984462008-04-10 16:19:33 -04003341 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003342 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003343 }
3344 unlock_buffer(bh);
3345 put_bh(bh);
3346}
3347
Anand Jain29c36d72015-08-14 18:32:58 +08003348int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3349 struct buffer_head **bh_ret)
3350{
3351 struct buffer_head *bh;
3352 struct btrfs_super_block *super;
3353 u64 bytenr;
3354
3355 bytenr = btrfs_sb_offset(copy_num);
3356 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3357 return -EINVAL;
3358
3359 bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE);
3360 /*
3361 * If we fail to read from the underlying devices, as of now
3362 * the best option we have is to mark it EIO.
3363 */
3364 if (!bh)
3365 return -EIO;
3366
3367 super = (struct btrfs_super_block *)bh->b_data;
3368 if (btrfs_super_bytenr(super) != bytenr ||
3369 btrfs_super_magic(super) != BTRFS_MAGIC) {
3370 brelse(bh);
3371 return -EINVAL;
3372 }
3373
3374 *bh_ret = bh;
3375 return 0;
3376}
3377
3378
Yan Zhenga512bbf2008-12-08 16:46:26 -05003379struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3380{
3381 struct buffer_head *bh;
3382 struct buffer_head *latest = NULL;
3383 struct btrfs_super_block *super;
3384 int i;
3385 u64 transid = 0;
Anand Jain92fc03f2015-08-14 18:32:51 +08003386 int ret = -EINVAL;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003387
3388 /* we would like to check all the supers, but that would make
3389 * a btrfs mount succeed after a mkfs from a different FS.
3390 * So, we need to add a special mount option to scan for
3391 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3392 */
3393 for (i = 0; i < 1; i++) {
Anand Jain29c36d72015-08-14 18:32:58 +08003394 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3395 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003396 continue;
3397
3398 super = (struct btrfs_super_block *)bh->b_data;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003399
3400 if (!latest || btrfs_super_generation(super) > transid) {
3401 brelse(latest);
3402 latest = bh;
3403 transid = btrfs_super_generation(super);
3404 } else {
3405 brelse(bh);
3406 }
3407 }
Anand Jain92fc03f2015-08-14 18:32:51 +08003408
3409 if (!latest)
3410 return ERR_PTR(ret);
3411
Yan Zhenga512bbf2008-12-08 16:46:26 -05003412 return latest;
3413}
3414
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003415/*
3416 * this should be called twice, once with wait == 0 and
3417 * once with wait == 1. When wait == 0 is done, all the buffer heads
3418 * we write are pinned.
3419 *
3420 * They are released when wait == 1 is done.
3421 * max_mirrors must be the same for both runs, and it indicates how
3422 * many supers on this one device should be written.
3423 *
3424 * max_mirrors == 0 means to write them all.
3425 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003426static int write_dev_supers(struct btrfs_device *device,
3427 struct btrfs_super_block *sb,
3428 int do_barriers, int wait, int max_mirrors)
3429{
3430 struct buffer_head *bh;
3431 int i;
3432 int ret;
3433 int errors = 0;
3434 u32 crc;
3435 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003436
3437 if (max_mirrors == 0)
3438 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3439
Yan Zhenga512bbf2008-12-08 16:46:26 -05003440 for (i = 0; i < max_mirrors; i++) {
3441 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003442 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3443 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003444 break;
3445
3446 if (wait) {
3447 bh = __find_get_block(device->bdev, bytenr / 4096,
3448 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003449 if (!bh) {
3450 errors++;
3451 continue;
3452 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003453 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003454 if (!buffer_uptodate(bh))
3455 errors++;
3456
3457 /* drop our reference */
3458 brelse(bh);
3459
3460 /* drop the reference from the wait == 0 run */
3461 brelse(bh);
3462 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003463 } else {
3464 btrfs_set_super_bytenr(sb, bytenr);
3465
3466 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003467 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003468 BTRFS_CSUM_SIZE, crc,
3469 BTRFS_SUPER_INFO_SIZE -
3470 BTRFS_CSUM_SIZE);
3471 btrfs_csum_final(crc, sb->csum);
3472
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003473 /*
3474 * one reference for us, and we leave it for the
3475 * caller
3476 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003477 bh = __getblk(device->bdev, bytenr / 4096,
3478 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003479 if (!bh) {
David Sterbaf14d1042015-10-08 11:37:06 +02003480 btrfs_err(device->dev_root->fs_info,
3481 "couldn't get super buffer head for bytenr %llu",
3482 bytenr);
Josef Bacik634554d2013-04-29 10:05:57 -04003483 errors++;
3484 continue;
3485 }
3486
Yan Zhenga512bbf2008-12-08 16:46:26 -05003487 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3488
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003489 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003490 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003491
3492 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003493 lock_buffer(bh);
3494 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003495 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003496 }
3497
Chris Mason387125f2011-11-18 15:07:51 -05003498 /*
3499 * we fua the first super. The others we allow
3500 * to go down lazy.
3501 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003502 if (i == 0)
Mike Christie2a222ca2016-06-05 14:31:43 -05003503 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_FUA, bh);
Wang Shilonge8117c22014-01-03 18:22:57 +08003504 else
Mike Christie2a222ca2016-06-05 14:31:43 -05003505 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003506 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003507 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003508 }
3509 return errors < i ? 0 : -1;
3510}
3511
Chris Mason387125f2011-11-18 15:07:51 -05003512/*
3513 * endio for the write_dev_flush, this will wake anyone waiting
3514 * for the barrier when it is done
3515 */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003516static void btrfs_end_empty_barrier(struct bio *bio)
Chris Mason387125f2011-11-18 15:07:51 -05003517{
Chris Mason387125f2011-11-18 15:07:51 -05003518 if (bio->bi_private)
3519 complete(bio->bi_private);
3520 bio_put(bio);
3521}
3522
3523/*
3524 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3525 * sent down. With wait == 1, it waits for the previous flush.
3526 *
3527 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3528 * capable
3529 */
3530static int write_dev_flush(struct btrfs_device *device, int wait)
3531{
3532 struct bio *bio;
3533 int ret = 0;
3534
3535 if (device->nobarriers)
3536 return 0;
3537
3538 if (wait) {
3539 bio = device->flush_bio;
3540 if (!bio)
3541 return 0;
3542
3543 wait_for_completion(&device->flush_wait);
3544
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003545 if (bio->bi_error) {
3546 ret = bio->bi_error;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003547 btrfs_dev_stat_inc_and_print(device,
3548 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003549 }
3550
3551 /* drop the reference from the wait == 0 run */
3552 bio_put(bio);
3553 device->flush_bio = NULL;
3554
3555 return ret;
3556 }
3557
3558 /*
3559 * one reference for us, and we leave it for the
3560 * caller
3561 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003562 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003563 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003564 if (!bio)
3565 return -ENOMEM;
3566
3567 bio->bi_end_io = btrfs_end_empty_barrier;
3568 bio->bi_bdev = device->bdev;
Mike Christie37226b22016-06-05 14:31:52 -05003569 bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH);
Chris Mason387125f2011-11-18 15:07:51 -05003570 init_completion(&device->flush_wait);
3571 bio->bi_private = &device->flush_wait;
3572 device->flush_bio = bio;
3573
3574 bio_get(bio);
Mike Christie4e49ea42016-06-05 14:31:41 -05003575 btrfsic_submit_bio(bio);
Chris Mason387125f2011-11-18 15:07:51 -05003576
3577 return 0;
3578}
3579
3580/*
3581 * send an empty flush down to each device in parallel,
3582 * then wait for them
3583 */
3584static int barrier_all_devices(struct btrfs_fs_info *info)
3585{
3586 struct list_head *head;
3587 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003588 int errors_send = 0;
3589 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003590 int ret;
3591
3592 /* send down all the barriers */
3593 head = &info->fs_devices->devices;
3594 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003595 if (dev->missing)
3596 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003597 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003598 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003599 continue;
3600 }
3601 if (!dev->in_fs_metadata || !dev->writeable)
3602 continue;
3603
3604 ret = write_dev_flush(dev, 0);
3605 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003606 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003607 }
3608
3609 /* wait for all the barriers */
3610 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003611 if (dev->missing)
3612 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003613 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003614 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003615 continue;
3616 }
3617 if (!dev->in_fs_metadata || !dev->writeable)
3618 continue;
3619
3620 ret = write_dev_flush(dev, 1);
3621 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003622 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003623 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003624 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3625 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003626 return -EIO;
3627 return 0;
3628}
3629
Zhao Lei943c6e92015-08-19 15:54:15 +08003630int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3631{
Zhao Lei8789f4f2015-09-15 21:08:07 +08003632 int raid_type;
3633 int min_tolerated = INT_MAX;
Zhao Lei943c6e92015-08-19 15:54:15 +08003634
Zhao Lei8789f4f2015-09-15 21:08:07 +08003635 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3636 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3637 min_tolerated = min(min_tolerated,
3638 btrfs_raid_array[BTRFS_RAID_SINGLE].
3639 tolerated_failures);
Zhao Lei943c6e92015-08-19 15:54:15 +08003640
Zhao Lei8789f4f2015-09-15 21:08:07 +08003641 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3642 if (raid_type == BTRFS_RAID_SINGLE)
3643 continue;
3644 if (!(flags & btrfs_raid_group[raid_type]))
3645 continue;
3646 min_tolerated = min(min_tolerated,
3647 btrfs_raid_array[raid_type].
3648 tolerated_failures);
3649 }
Zhao Lei943c6e92015-08-19 15:54:15 +08003650
Zhao Lei8789f4f2015-09-15 21:08:07 +08003651 if (min_tolerated == INT_MAX) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04003652 pr_warn("BTRFS: unknown raid flag: %llu", flags);
Zhao Lei8789f4f2015-09-15 21:08:07 +08003653 min_tolerated = 0;
3654 }
3655
3656 return min_tolerated;
Zhao Lei943c6e92015-08-19 15:54:15 +08003657}
3658
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003659int btrfs_calc_num_tolerated_disk_barrier_failures(
3660 struct btrfs_fs_info *fs_info)
3661{
3662 struct btrfs_ioctl_space_info space;
3663 struct btrfs_space_info *sinfo;
3664 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3665 BTRFS_BLOCK_GROUP_SYSTEM,
3666 BTRFS_BLOCK_GROUP_METADATA,
3667 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003668 int i;
3669 int c;
3670 int num_tolerated_disk_barrier_failures =
3671 (int)fs_info->fs_devices->num_devices;
3672
Zhao Lei2c458042015-07-16 15:00:46 +08003673 for (i = 0; i < ARRAY_SIZE(types); i++) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003674 struct btrfs_space_info *tmp;
3675
3676 sinfo = NULL;
3677 rcu_read_lock();
3678 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3679 if (tmp->flags == types[i]) {
3680 sinfo = tmp;
3681 break;
3682 }
3683 }
3684 rcu_read_unlock();
3685
3686 if (!sinfo)
3687 continue;
3688
3689 down_read(&sinfo->groups_sem);
3690 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
Zhao Lei2c458042015-07-16 15:00:46 +08003691 u64 flags;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003692
Zhao Lei2c458042015-07-16 15:00:46 +08003693 if (list_empty(&sinfo->block_groups[c]))
3694 continue;
3695
3696 btrfs_get_block_group_info(&sinfo->block_groups[c],
3697 &space);
3698 if (space.total_bytes == 0 || space.used_bytes == 0)
3699 continue;
3700 flags = space.flags;
Zhao Lei943c6e92015-08-19 15:54:15 +08003701
3702 num_tolerated_disk_barrier_failures = min(
3703 num_tolerated_disk_barrier_failures,
3704 btrfs_get_num_tolerated_disk_barrier_failures(
3705 flags));
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003706 }
3707 up_read(&sinfo->groups_sem);
3708 }
3709
3710 return num_tolerated_disk_barrier_failures;
3711}
3712
Eric Sandeen48a3b632013-04-25 20:41:01 +00003713static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003714{
Chris Masone5e9a522009-06-10 15:17:02 -04003715 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003716 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003717 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003718 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003719 int ret;
3720 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003721 int max_errors;
3722 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003723 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003724
Jeff Mahoney3cdde222016-06-09 21:38:35 -04003725 do_barriers = !btrfs_test_opt(root->fs_info, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003726 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003727
David Sterba6c417612011-04-13 15:41:04 +02003728 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003729 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003730
Chris Mason174ba502011-05-27 10:03:58 -04003731 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003732 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003733 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003734
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003735 if (do_barriers) {
3736 ret = barrier_all_devices(root->fs_info);
3737 if (ret) {
3738 mutex_unlock(
3739 &root->fs_info->fs_devices->device_list_mutex);
Anand Jain34d97002016-03-16 16:43:06 +08003740 btrfs_handle_fs_error(root->fs_info, ret,
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003741 "errors while submitting device barriers.");
3742 return ret;
3743 }
3744 }
Chris Mason387125f2011-11-18 15:07:51 -05003745
Xiao Guangrong1f781602011-04-20 10:09:16 +00003746 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003747 if (!dev->bdev) {
3748 total_errors++;
3749 continue;
3750 }
Yan Zheng2b820322008-11-17 21:11:30 -05003751 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003752 continue;
3753
Yan Zheng2b820322008-11-17 21:11:30 -05003754 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003755 btrfs_set_stack_device_type(dev_item, dev->type);
3756 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003757 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003758 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003759 btrfs_set_stack_device_bytes_used(dev_item,
3760 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003761 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3762 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3763 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3764 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003765 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003766
Chris Masona061fc82008-05-07 11:43:44 -04003767 flags = btrfs_super_flags(sb);
3768 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003769
Yan Zhenga512bbf2008-12-08 16:46:26 -05003770 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003771 if (ret)
3772 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003773 }
Chris Masona236aed2008-04-29 09:38:00 -04003774 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003775 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003776 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003777 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003778
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003779 /* FUA is masked off if unsupported and can't be the reason */
Anand Jain34d97002016-03-16 16:43:06 +08003780 btrfs_handle_fs_error(root->fs_info, -EIO,
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003781 "%d errors while writing supers", total_errors);
3782 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003783 }
Chris Masonf2984462008-04-10 16:19:33 -04003784
Yan Zhenga512bbf2008-12-08 16:46:26 -05003785 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003786 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003787 if (!dev->bdev)
3788 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003789 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003790 continue;
3791
Yan Zhenga512bbf2008-12-08 16:46:26 -05003792 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3793 if (ret)
3794 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003795 }
Chris Mason174ba502011-05-27 10:03:58 -04003796 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003797 if (total_errors > max_errors) {
Anand Jain34d97002016-03-16 16:43:06 +08003798 btrfs_handle_fs_error(root->fs_info, -EIO,
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003799 "%d errors while writing supers", total_errors);
3800 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003801 }
Chris Masonf2984462008-04-10 16:19:33 -04003802 return 0;
3803}
3804
Yan Zhenga512bbf2008-12-08 16:46:26 -05003805int write_ctree_super(struct btrfs_trans_handle *trans,
3806 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003807{
Rashikaf570e7572013-10-31 16:50:42 +05303808 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003809}
3810
Miao Xiecb517ea2013-05-15 07:48:19 +00003811/* Drop a fs root from the radix tree and free it. */
3812void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3813 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003814{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003815 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003816 radix_tree_delete(&fs_info->fs_roots_radix,
3817 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003818 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003819
3820 if (btrfs_root_refs(&root->root_item) == 0)
3821 synchronize_srcu(&fs_info->subvol_srcu);
3822
Liu Bo1c1ea4f2016-07-19 15:36:05 -07003823 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
Liu Bo33217192013-02-27 13:28:24 +00003824 btrfs_free_log(NULL, root);
Liu Bo1c1ea4f2016-07-19 15:36:05 -07003825 if (root->reloc_root) {
3826 free_extent_buffer(root->reloc_root->node);
3827 free_extent_buffer(root->reloc_root->commit_root);
3828 btrfs_put_fs_root(root->reloc_root);
3829 root->reloc_root = NULL;
3830 }
3831 }
Liu Bo33217192013-02-27 13:28:24 +00003832
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003833 if (root->free_ino_pinned)
3834 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3835 if (root->free_ino_ctl)
3836 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003837 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003838}
3839
3840static void free_fs_root(struct btrfs_root *root)
3841{
David Sterba57cdc8d2014-02-05 02:37:48 +01003842 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003843 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003844 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3845 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003846 if (root->anon_dev)
3847 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003848 if (root->subv_writers)
3849 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003850 free_extent_buffer(root->node);
3851 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003852 kfree(root->free_ino_ctl);
3853 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003854 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003855 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003856}
3857
Miao Xiecb517ea2013-05-15 07:48:19 +00003858void btrfs_free_fs_root(struct btrfs_root *root)
3859{
3860 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003861}
3862
Yan Zhengc146afa2008-11-12 14:34:12 -05003863int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3864{
3865 u64 root_objectid = 0;
3866 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003867 int i = 0;
3868 int err = 0;
3869 unsigned int ret = 0;
3870 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003871
3872 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003873 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003874 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3875 (void **)gang, root_objectid,
3876 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003877 if (!ret) {
3878 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003879 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003880 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003881 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003882
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003883 for (i = 0; i < ret; i++) {
3884 /* Avoid to grab roots in dead_roots */
3885 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3886 gang[i] = NULL;
3887 continue;
3888 }
3889 /* grab all the search result for later use */
3890 gang[i] = btrfs_grab_fs_root(gang[i]);
3891 }
3892 srcu_read_unlock(&fs_info->subvol_srcu, index);
3893
3894 for (i = 0; i < ret; i++) {
3895 if (!gang[i])
3896 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003897 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003898 err = btrfs_orphan_cleanup(gang[i]);
3899 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003900 break;
3901 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003902 }
3903 root_objectid++;
3904 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003905
3906 /* release the uncleaned roots due to error */
3907 for (; i < ret; i++) {
3908 if (gang[i])
3909 btrfs_put_fs_root(gang[i]);
3910 }
3911 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003912}
3913
3914int btrfs_commit_super(struct btrfs_root *root)
3915{
3916 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003917
3918 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003919 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003920 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003921 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003922
3923 /* wait until ongoing cleanup work done */
3924 down_write(&root->fs_info->cleanup_work_sem);
3925 up_write(&root->fs_info->cleanup_work_sem);
3926
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003927 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003928 if (IS_ERR(trans))
3929 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003930 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003931}
3932
David Sterba3abdbd72014-06-04 18:10:45 +02003933void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003934{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003935 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003936 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003937
Josef Bacikafcdd122016-09-02 15:40:02 -04003938 set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags);
Chris Masona2135012008-06-25 16:01:30 -04003939
Justin Maggard7343dd62015-11-04 15:56:16 -08003940 /* wait for the qgroup rescan worker to stop */
Jeff Mahoneyd06f23d2016-08-08 22:08:06 -04003941 btrfs_qgroup_wait_for_completion(fs_info, false);
Justin Maggard7343dd62015-11-04 15:56:16 -08003942
Stefan Behrens803b2f52013-08-15 17:11:21 +02003943 /* wait for the uuid_scan task to finish */
3944 down(&fs_info->uuid_tree_rescan_sem);
3945 /* avoid complains from lockdep et al., set sem back to initial state */
3946 up(&fs_info->uuid_tree_rescan_sem);
3947
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003948 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003949 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003950
Stefan Behrens8dabb742012-11-06 13:15:27 +01003951 btrfs_dev_replace_suspend_for_unmount(fs_info);
3952
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003953 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003954
3955 /* wait for any defraggers to finish */
3956 wait_event(fs_info->transaction_wait,
3957 (atomic_read(&fs_info->defrag_running) == 0));
3958
3959 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003960 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003961
Miao Xie21c7e752014-05-13 17:29:04 -07003962 cancel_work_sync(&fs_info->async_reclaim_work);
3963
Yan Zhengc146afa2008-11-12 14:34:12 -05003964 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003965 /*
3966 * If the cleaner thread is stopped and there are
3967 * block groups queued for removal, the deletion will be
3968 * skipped when we quit the cleaner thread.
3969 */
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003970 btrfs_delete_unused_bgs(root->fs_info);
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003971
liuboacce9522011-01-06 19:30:25 +08003972 ret = btrfs_commit_super(root);
3973 if (ret)
Eric Sandeen04892342014-08-01 18:12:36 -05003974 btrfs_err(fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003975 }
3976
Miao Xie87533c42013-01-29 10:14:48 +00003977 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003978 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003979
Al Viroe3029d92011-11-17 00:56:18 -05003980 kthread_stop(fs_info->transaction_kthread);
3981 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003982
Josef Bacikafcdd122016-09-02 15:40:02 -04003983 set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags);
Yan Zhengf25784b2009-07-28 08:41:57 -04003984
Eric Sandeen04892342014-08-01 18:12:36 -05003985 btrfs_free_qgroup_config(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003986
Miao Xie963d6782013-01-29 10:10:51 +00003987 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Eric Sandeen04892342014-08-01 18:12:36 -05003988 btrfs_info(fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003989 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003990 }
Yanbcc63ab2008-07-30 16:29:20 -04003991
Anand Jain6618a592015-08-14 18:32:47 +08003992 btrfs_sysfs_remove_mounted(fs_info);
Anand Jainb7c35e82015-03-10 06:38:38 +08003993 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003994
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003995 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003996
Liu Bo1a4319c2014-01-13 19:53:53 +08003997 btrfs_put_block_group_cache(fs_info);
3998
Josef Bacik2b1360d2013-10-07 15:14:44 -04003999 btrfs_free_block_groups(fs_info);
4000
Wang Shilongde348ee2014-04-09 19:23:22 +08004001 /*
4002 * we must make sure there is not any read request to
4003 * submit after we stopping all workers.
4004 */
4005 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04004006 btrfs_stop_all_workers(fs_info);
4007
Josef Bacikafcdd122016-09-02 15:40:02 -04004008 clear_bit(BTRFS_FS_OPEN, &fs_info->flags);
Josef Bacik13e6c372013-05-30 16:55:44 -04004009 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04004010
Josef Bacik13e6c372013-05-30 16:55:44 -04004011 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04004012
Stefan Behrens21adbd52011-11-09 13:44:05 +01004013#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
Jeff Mahoney3cdde222016-06-09 21:38:35 -04004014 if (btrfs_test_opt(root->fs_info, CHECK_INTEGRITY))
Stefan Behrens21adbd52011-11-09 13:44:05 +01004015 btrfsic_unmount(root, fs_info->fs_devices);
4016#endif
4017
Chris Masondfe25022008-05-13 13:46:40 -04004018 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04004019 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04004020
Miao Xiee2d84522013-01-29 10:09:20 +00004021 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00004022 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08004023 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04004024 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04004025 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04004026
David Woodhouse53b381b2013-01-29 18:40:14 -05004027 btrfs_free_stripe_hash_table(fs_info);
4028
Chris Masoncdfb0802015-04-06 18:17:00 -07004029 __btrfs_free_block_rsv(root->orphan_block_rsv);
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01004030 root->orphan_block_rsv = NULL;
Filipe Manana04216822014-11-27 21:14:15 +00004031
4032 lock_chunks(root);
4033 while (!list_empty(&fs_info->pinned_chunks)) {
4034 struct extent_map *em;
4035
4036 em = list_first_entry(&fs_info->pinned_chunks,
4037 struct extent_map, list);
4038 list_del_init(&em->list);
4039 free_extent_map(em);
4040 }
4041 unlock_chunks(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05004042}
4043
Chris Masonb9fab912012-05-06 07:23:47 -04004044int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
4045 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04004046{
Chris Mason1259ab72008-05-12 13:39:03 -04004047 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04004048 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04004049
Josef Bacik0b32f4b2012-03-13 09:38:00 -04004050 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04004051 if (!ret)
4052 return ret;
4053
4054 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04004055 parent_transid, atomic);
4056 if (ret == -EAGAIN)
4057 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04004058 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04004059}
Chris Mason6702ed42007-08-07 16:15:09 -04004060
Chris Mason5f39d392007-10-15 16:14:19 -04004061void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
4062{
Josef Bacik06ea65a2013-09-19 16:07:01 -04004063 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04004064 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04004065 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05004066
Josef Bacik06ea65a2013-09-19 16:07:01 -04004067#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4068 /*
4069 * This is a fast path so only do this check if we have sanity tests
4070 * enabled. Normal people shouldn't be marking dummy buffers as dirty
4071 * outside of the sanity tests.
4072 */
4073 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
4074 return;
4075#endif
4076 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04004077 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00004078 if (transid != root->fs_info->generation)
Jeff Mahoney5d163e02016-09-20 10:05:00 -04004079 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02004080 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04004081 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00004082 if (!was_dirty)
4083 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
4084 buf->len,
4085 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01004086#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
4087 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
4088 btrfs_print_leaf(root, buf);
4089 ASSERT(0);
4090 }
4091#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05004092}
4093
Liu Bob53d3f52012-11-14 14:34:34 +00004094static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
4095 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04004096{
Chris Mason188de642008-05-09 11:52:25 -04004097 /*
4098 * looks as though older kernels can get into trouble with
4099 * this code, they end up stuck in balance_dirty_pages forever
4100 */
Miao Xiee2d84522013-01-29 10:09:20 +00004101 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04004102
Jens Axboe6933c022009-03-17 09:36:37 +01004103 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04004104 return;
4105
Liu Bob53d3f52012-11-14 14:34:34 +00004106 if (flush_delayed)
4107 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08004108
Miao Xiee2d84522013-01-29 10:09:20 +00004109 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
4110 BTRFS_DIRTY_METADATA_THRESH);
4111 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08004112 balance_dirty_pages_ratelimited(
4113 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08004114 }
Miao Xie16cdcec2011-04-22 18:12:22 +08004115}
4116
Liu Bob53d3f52012-11-14 14:34:34 +00004117void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08004118{
Liu Bob53d3f52012-11-14 14:34:34 +00004119 __btrfs_btree_balance_dirty(root, 1);
4120}
Miao Xie16cdcec2011-04-22 18:12:22 +08004121
Liu Bob53d3f52012-11-14 14:34:34 +00004122void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
4123{
4124 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04004125}
Chris Mason6b800532007-10-15 16:17:34 -04004126
Chris Masonca7a79a2008-05-12 12:59:19 -04004127int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04004128{
Chris Mason727011e2010-08-06 13:21:20 -04004129 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik8436ea912016-09-02 15:40:03 -04004130 return btree_read_extent_buffer_pages(root, buf, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04004131}
Chris Mason0da54682007-11-07 21:08:01 -05004132
David Sterbafcd1f0652012-03-06 00:06:18 +01004133static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08004134 int read_only)
4135{
David Sterbac9260932014-09-30 19:16:47 +02004136 struct btrfs_super_block *sb = fs_info->super_copy;
Qu Wenruo319e4d02015-12-15 09:14:36 +08004137 u64 nodesize = btrfs_super_nodesize(sb);
4138 u64 sectorsize = btrfs_super_sectorsize(sb);
David Sterbac9260932014-09-30 19:16:47 +02004139 int ret = 0;
4140
Qu Wenruo319e4d02015-12-15 09:14:36 +08004141 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004142 btrfs_err(fs_info, "no valid FS found");
Qu Wenruo319e4d02015-12-15 09:14:36 +08004143 ret = -EINVAL;
4144 }
4145 if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP)
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004146 btrfs_warn(fs_info, "unrecognized super flag: %llu",
Qu Wenruo319e4d02015-12-15 09:14:36 +08004147 btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP);
David Sterba21e76262014-10-27 13:52:21 +01004148 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004149 btrfs_err(fs_info, "tree_root level too big: %d >= %d",
David Sterba21e76262014-10-27 13:52:21 +01004150 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004151 ret = -EINVAL;
4152 }
David Sterba21e76262014-10-27 13:52:21 +01004153 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004154 btrfs_err(fs_info, "chunk_root level too big: %d >= %d",
David Sterba21e76262014-10-27 13:52:21 +01004155 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004156 ret = -EINVAL;
4157 }
David Sterba21e76262014-10-27 13:52:21 +01004158 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004159 btrfs_err(fs_info, "log_root level too big: %d >= %d",
David Sterba21e76262014-10-27 13:52:21 +01004160 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004161 ret = -EINVAL;
4162 }
4163
David Sterba1104a882013-03-06 15:57:46 +01004164 /*
Qu Wenruo319e4d02015-12-15 09:14:36 +08004165 * Check sectorsize and nodesize first, other check will need it.
4166 * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here.
David Sterba1104a882013-03-06 15:57:46 +01004167 */
Qu Wenruo319e4d02015-12-15 09:14:36 +08004168 if (!is_power_of_2(sectorsize) || sectorsize < 4096 ||
4169 sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004170 btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize);
David Sterba75d6ad32014-10-31 17:18:08 +01004171 ret = -EINVAL;
4172 }
Qu Wenruo319e4d02015-12-15 09:14:36 +08004173 /* Only PAGE SIZE is supported yet */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +03004174 if (sectorsize != PAGE_SIZE) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004175 btrfs_err(fs_info,
4176 "sectorsize %llu not supported yet, only support %lu",
4177 sectorsize, PAGE_SIZE);
Qu Wenruo319e4d02015-12-15 09:14:36 +08004178 ret = -EINVAL;
4179 }
4180 if (!is_power_of_2(nodesize) || nodesize < sectorsize ||
4181 nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004182 btrfs_err(fs_info, "invalid nodesize %llu", nodesize);
Qu Wenruo319e4d02015-12-15 09:14:36 +08004183 ret = -EINVAL;
4184 }
4185 if (nodesize != le32_to_cpu(sb->__unused_leafsize)) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004186 btrfs_err(fs_info, "invalid leafsize %u, should be %llu",
4187 le32_to_cpu(sb->__unused_leafsize), nodesize);
Qu Wenruo319e4d02015-12-15 09:14:36 +08004188 ret = -EINVAL;
4189 }
4190
4191 /* Root alignment check */
4192 if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004193 btrfs_warn(fs_info, "tree_root block unaligned: %llu",
4194 btrfs_super_root(sb));
Qu Wenruo319e4d02015-12-15 09:14:36 +08004195 ret = -EINVAL;
4196 }
4197 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004198 btrfs_warn(fs_info, "chunk_root block unaligned: %llu",
4199 btrfs_super_chunk_root(sb));
Qu Wenruo319e4d02015-12-15 09:14:36 +08004200 ret = -EINVAL;
4201 }
4202 if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004203 btrfs_warn(fs_info, "log_root block unaligned: %llu",
4204 btrfs_super_log_root(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004205 ret = -EINVAL;
4206 }
4207
David Sterbac9260932014-09-30 19:16:47 +02004208 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004209 btrfs_err(fs_info,
4210 "dev_item UUID does not match fsid: %pU != %pU",
4211 fs_info->fsid, sb->dev_item.fsid);
David Sterbac9260932014-09-30 19:16:47 +02004212 ret = -EINVAL;
4213 }
4214
4215 /*
4216 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4217 * done later
4218 */
Liu Bo99e3ecf2016-06-03 12:05:14 -07004219 if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) {
4220 btrfs_err(fs_info, "bytes_used is too small %llu",
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004221 btrfs_super_bytes_used(sb));
Liu Bo99e3ecf2016-06-03 12:05:14 -07004222 ret = -EINVAL;
4223 }
Chandan Rajendrab7f67052016-06-23 15:16:44 +05304224 if (!is_power_of_2(btrfs_super_stripesize(sb))) {
Liu Bo99e3ecf2016-06-03 12:05:14 -07004225 btrfs_err(fs_info, "invalid stripesize %u",
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004226 btrfs_super_stripesize(sb));
Liu Bo99e3ecf2016-06-03 12:05:14 -07004227 ret = -EINVAL;
4228 }
David Sterba21e76262014-10-27 13:52:21 +01004229 if (btrfs_super_num_devices(sb) > (1UL << 31))
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004230 btrfs_warn(fs_info, "suspicious number of devices: %llu",
4231 btrfs_super_num_devices(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004232 if (btrfs_super_num_devices(sb) == 0) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004233 btrfs_err(fs_info, "number of devices is 0");
David Sterba75d6ad32014-10-31 17:18:08 +01004234 ret = -EINVAL;
4235 }
David Sterbac9260932014-09-30 19:16:47 +02004236
David Sterba21e76262014-10-27 13:52:21 +01004237 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004238 btrfs_err(fs_info, "super offset mismatch %llu != %u",
4239 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
David Sterbac9260932014-09-30 19:16:47 +02004240 ret = -EINVAL;
4241 }
4242
4243 /*
David Sterbace7fca52014-10-31 18:42:05 +01004244 * Obvious sys_chunk_array corruptions, it must hold at least one key
4245 * and one chunk
4246 */
4247 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004248 btrfs_err(fs_info, "system chunk array too big %u > %u",
4249 btrfs_super_sys_array_size(sb),
4250 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
David Sterbace7fca52014-10-31 18:42:05 +01004251 ret = -EINVAL;
4252 }
4253 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4254 + sizeof(struct btrfs_chunk)) {
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004255 btrfs_err(fs_info, "system chunk array too small %u < %zu",
4256 btrfs_super_sys_array_size(sb),
4257 sizeof(struct btrfs_disk_key)
4258 + sizeof(struct btrfs_chunk));
David Sterbace7fca52014-10-31 18:42:05 +01004259 ret = -EINVAL;
4260 }
4261
4262 /*
David Sterbac9260932014-09-30 19:16:47 +02004263 * The generation is a global counter, we'll trust it more than the others
4264 * but it's still possible that it's the one that's wrong.
4265 */
David Sterba21e76262014-10-27 13:52:21 +01004266 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004267 btrfs_warn(fs_info,
4268 "suspicious: generation < chunk_root_generation: %llu < %llu",
4269 btrfs_super_generation(sb),
4270 btrfs_super_chunk_root_generation(sb));
David Sterba21e76262014-10-27 13:52:21 +01004271 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4272 && btrfs_super_cache_generation(sb) != (u64)-1)
Jeff Mahoneyab8d0fc2016-09-20 10:05:02 -04004273 btrfs_warn(fs_info,
4274 "suspicious: generation < cache_generation: %llu < %llu",
4275 btrfs_super_generation(sb),
4276 btrfs_super_cache_generation(sb));
David Sterbac9260932014-09-30 19:16:47 +02004277
4278 return ret;
liuboacce9522011-01-06 19:30:25 +08004279}
4280
Eric Sandeen48a3b632013-04-25 20:41:01 +00004281static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004282{
liuboacce9522011-01-06 19:30:25 +08004283 mutex_lock(&root->fs_info->cleaner_mutex);
4284 btrfs_run_delayed_iputs(root);
4285 mutex_unlock(&root->fs_info->cleaner_mutex);
4286
4287 down_write(&root->fs_info->cleanup_work_sem);
4288 up_write(&root->fs_info->cleanup_work_sem);
4289
4290 /* cleanup FS via transaction */
4291 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08004292}
4293
Jeff Mahoney143bede2012-03-01 14:56:26 +01004294static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004295{
liuboacce9522011-01-06 19:30:25 +08004296 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08004297
Miao Xie199c2a92013-05-15 07:48:23 +00004298 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05004299 /*
4300 * This will just short circuit the ordered completion stuff which will
4301 * make sure the ordered extent gets properly cleaned up.
4302 */
Miao Xie199c2a92013-05-15 07:48:23 +00004303 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05004304 root_extent_list)
4305 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00004306 spin_unlock(&root->ordered_extent_lock);
4307}
4308
4309static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4310{
4311 struct btrfs_root *root;
4312 struct list_head splice;
4313
4314 INIT_LIST_HEAD(&splice);
4315
4316 spin_lock(&fs_info->ordered_root_lock);
4317 list_splice_init(&fs_info->ordered_roots, &splice);
4318 while (!list_empty(&splice)) {
4319 root = list_first_entry(&splice, struct btrfs_root,
4320 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04004321 list_move_tail(&root->ordered_root,
4322 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00004323
Liu Bo2a85d9c2014-02-10 17:07:16 +08004324 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004325 btrfs_destroy_ordered_extents(root);
4326
Liu Bo2a85d9c2014-02-10 17:07:16 +08004327 cond_resched();
4328 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004329 }
4330 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08004331}
4332
Stefan Behrens35a36212013-08-14 18:12:25 +02004333static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4334 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004335{
4336 struct rb_node *node;
4337 struct btrfs_delayed_ref_root *delayed_refs;
4338 struct btrfs_delayed_ref_node *ref;
4339 int ret = 0;
4340
4341 delayed_refs = &trans->delayed_refs;
4342
4343 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004344 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04004345 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05004346 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08004347 return ret;
4348 }
4349
Josef Bacikd7df2c72014-01-23 09:21:38 -05004350 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4351 struct btrfs_delayed_ref_head *head;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004352 struct btrfs_delayed_ref_node *tmp;
Josef Bacike78417d2013-06-03 16:42:36 -04004353 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08004354
Josef Bacikd7df2c72014-01-23 09:21:38 -05004355 head = rb_entry(node, struct btrfs_delayed_ref_head,
4356 href_node);
4357 if (!mutex_trylock(&head->mutex)) {
4358 atomic_inc(&head->node.refs);
4359 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08004360
Josef Bacikd7df2c72014-01-23 09:21:38 -05004361 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04004362 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004363 btrfs_put_delayed_ref(&head->node);
4364 spin_lock(&delayed_refs->lock);
4365 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04004366 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05004367 spin_lock(&head->lock);
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004368 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4369 list) {
Josef Bacikd7df2c72014-01-23 09:21:38 -05004370 ref->in_tree = 0;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004371 list_del(&ref->list);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004372 atomic_dec(&delayed_refs->num_entries);
4373 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004374 }
4375 if (head->must_insert_reserved)
4376 pin_bytes = true;
4377 btrfs_free_delayed_extent_op(head->extent_op);
4378 delayed_refs->num_heads--;
4379 if (head->processing == 0)
4380 delayed_refs->num_heads_ready--;
4381 atomic_dec(&delayed_refs->num_entries);
4382 head->node.in_tree = 0;
4383 rb_erase(&head->href_node, &delayed_refs->href_root);
4384 spin_unlock(&head->lock);
4385 spin_unlock(&delayed_refs->lock);
4386 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08004387
Josef Bacikd7df2c72014-01-23 09:21:38 -05004388 if (pin_bytes)
4389 btrfs_pin_extent(root, head->node.bytenr,
4390 head->node.num_bytes, 1);
4391 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08004392 cond_resched();
4393 spin_lock(&delayed_refs->lock);
4394 }
4395
4396 spin_unlock(&delayed_refs->lock);
4397
4398 return ret;
4399}
4400
Jeff Mahoney143bede2012-03-01 14:56:26 +01004401static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004402{
4403 struct btrfs_inode *btrfs_inode;
4404 struct list_head splice;
4405
4406 INIT_LIST_HEAD(&splice);
4407
Miao Xieeb73c1b2013-05-15 07:48:22 +00004408 spin_lock(&root->delalloc_lock);
4409 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08004410
4411 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00004412 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4413 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08004414
4415 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00004416 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4417 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00004418 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004419
4420 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00004421
Miao Xieeb73c1b2013-05-15 07:48:22 +00004422 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004423 }
4424
Miao Xieeb73c1b2013-05-15 07:48:22 +00004425 spin_unlock(&root->delalloc_lock);
4426}
4427
4428static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4429{
4430 struct btrfs_root *root;
4431 struct list_head splice;
4432
4433 INIT_LIST_HEAD(&splice);
4434
4435 spin_lock(&fs_info->delalloc_root_lock);
4436 list_splice_init(&fs_info->delalloc_roots, &splice);
4437 while (!list_empty(&splice)) {
4438 root = list_first_entry(&splice, struct btrfs_root,
4439 delalloc_root);
4440 list_del_init(&root->delalloc_root);
4441 root = btrfs_grab_fs_root(root);
4442 BUG_ON(!root);
4443 spin_unlock(&fs_info->delalloc_root_lock);
4444
4445 btrfs_destroy_delalloc_inodes(root);
4446 btrfs_put_fs_root(root);
4447
4448 spin_lock(&fs_info->delalloc_root_lock);
4449 }
4450 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004451}
4452
4453static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4454 struct extent_io_tree *dirty_pages,
4455 int mark)
4456{
4457 int ret;
liuboacce9522011-01-06 19:30:25 +08004458 struct extent_buffer *eb;
4459 u64 start = 0;
4460 u64 end;
liuboacce9522011-01-06 19:30:25 +08004461
4462 while (1) {
4463 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004464 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004465 if (ret)
4466 break;
4467
David Sterba91166212016-04-26 23:54:39 +02004468 clear_extent_bits(dirty_pages, start, end, mark);
liuboacce9522011-01-06 19:30:25 +08004469 while (start <= end) {
Daniel Dressler01d58472014-11-21 17:15:07 +09004470 eb = btrfs_find_tree_block(root->fs_info, start);
David Sterba707e8a02014-06-04 19:22:26 +02004471 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004472 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004473 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004474 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004475
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004476 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4477 &eb->bflags))
4478 clear_extent_buffer_dirty(eb);
4479 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004480 }
4481 }
4482
4483 return ret;
4484}
4485
4486static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4487 struct extent_io_tree *pinned_extents)
4488{
4489 struct extent_io_tree *unpin;
4490 u64 start;
4491 u64 end;
4492 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004493 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004494
4495 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004496again:
liuboacce9522011-01-06 19:30:25 +08004497 while (1) {
4498 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004499 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004500 if (ret)
4501 break;
4502
David Sterbaaf6f8f62016-04-26 23:54:39 +02004503 clear_extent_dirty(unpin, start, end);
liuboacce9522011-01-06 19:30:25 +08004504 btrfs_error_unpin_extent_range(root, start, end);
4505 cond_resched();
4506 }
4507
Liu Boed0eaa12012-06-14 02:23:21 -06004508 if (loop) {
4509 if (unpin == &root->fs_info->freed_extents[0])
4510 unpin = &root->fs_info->freed_extents[1];
4511 else
4512 unpin = &root->fs_info->freed_extents[0];
4513 loop = false;
4514 goto again;
4515 }
4516
liuboacce9522011-01-06 19:30:25 +08004517 return 0;
4518}
4519
Liu Boc79a1752016-07-20 17:44:12 -07004520static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache)
4521{
4522 struct inode *inode;
4523
4524 inode = cache->io_ctl.inode;
4525 if (inode) {
4526 invalidate_inode_pages2(inode->i_mapping);
4527 BTRFS_I(inode)->generation = 0;
4528 cache->io_ctl.inode = NULL;
4529 iput(inode);
4530 }
4531 btrfs_put_block_group(cache);
4532}
4533
4534void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4535 struct btrfs_root *root)
4536{
4537 struct btrfs_block_group_cache *cache;
4538
4539 spin_lock(&cur_trans->dirty_bgs_lock);
4540 while (!list_empty(&cur_trans->dirty_bgs)) {
4541 cache = list_first_entry(&cur_trans->dirty_bgs,
4542 struct btrfs_block_group_cache,
4543 dirty_list);
4544 if (!cache) {
4545 btrfs_err(root->fs_info,
4546 "orphan block group dirty_bgs list");
4547 spin_unlock(&cur_trans->dirty_bgs_lock);
4548 return;
4549 }
4550
4551 if (!list_empty(&cache->io_list)) {
4552 spin_unlock(&cur_trans->dirty_bgs_lock);
4553 list_del_init(&cache->io_list);
4554 btrfs_cleanup_bg_io(cache);
4555 spin_lock(&cur_trans->dirty_bgs_lock);
4556 }
4557
4558 list_del_init(&cache->dirty_list);
4559 spin_lock(&cache->lock);
4560 cache->disk_cache_state = BTRFS_DC_ERROR;
4561 spin_unlock(&cache->lock);
4562
4563 spin_unlock(&cur_trans->dirty_bgs_lock);
4564 btrfs_put_block_group(cache);
4565 spin_lock(&cur_trans->dirty_bgs_lock);
4566 }
4567 spin_unlock(&cur_trans->dirty_bgs_lock);
4568
4569 while (!list_empty(&cur_trans->io_bgs)) {
4570 cache = list_first_entry(&cur_trans->io_bgs,
4571 struct btrfs_block_group_cache,
4572 io_list);
4573 if (!cache) {
4574 btrfs_err(root->fs_info,
4575 "orphan block group on io_bgs list");
4576 return;
4577 }
4578
4579 list_del_init(&cache->io_list);
4580 spin_lock(&cache->lock);
4581 cache->disk_cache_state = BTRFS_DC_ERROR;
4582 spin_unlock(&cache->lock);
4583 btrfs_cleanup_bg_io(cache);
4584 }
4585}
4586
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004587void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4588 struct btrfs_root *root)
4589{
Liu Boc79a1752016-07-20 17:44:12 -07004590 btrfs_cleanup_dirty_bgs(cur_trans, root);
4591 ASSERT(list_empty(&cur_trans->dirty_bgs));
4592 ASSERT(list_empty(&cur_trans->io_bgs));
4593
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004594 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004595
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004596 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004597 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004598
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004599 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004600 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004601
Miao Xie67cde342012-06-14 02:23:22 -06004602 btrfs_destroy_delayed_inodes(root);
4603 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004604
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004605 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4606 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004607 btrfs_destroy_pinned_extent(root,
4608 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004609
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004610 cur_trans->state =TRANS_STATE_COMPLETED;
4611 wake_up(&cur_trans->commit_wait);
4612
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004613 /*
4614 memset(cur_trans, 0, sizeof(*cur_trans));
4615 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4616 */
4617}
4618
Eric Sandeen48a3b632013-04-25 20:41:01 +00004619static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004620{
4621 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004622
liuboacce9522011-01-06 19:30:25 +08004623 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4624
Josef Bacika4abeea2011-04-11 17:25:13 -04004625 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004626 while (!list_empty(&root->fs_info->trans_list)) {
4627 t = list_first_entry(&root->fs_info->trans_list,
4628 struct btrfs_transaction, list);
4629 if (t->state >= TRANS_STATE_COMMIT_START) {
4630 atomic_inc(&t->use_count);
4631 spin_unlock(&root->fs_info->trans_lock);
4632 btrfs_wait_for_commit(root, t->transid);
4633 btrfs_put_transaction(t);
4634 spin_lock(&root->fs_info->trans_lock);
4635 continue;
4636 }
4637 if (t == root->fs_info->running_transaction) {
4638 t->state = TRANS_STATE_COMMIT_DOING;
4639 spin_unlock(&root->fs_info->trans_lock);
4640 /*
4641 * We wait for 0 num_writers since we don't hold a trans
4642 * handle open currently for this transaction.
4643 */
4644 wait_event(t->writer_wait,
4645 atomic_read(&t->num_writers) == 0);
4646 } else {
4647 spin_unlock(&root->fs_info->trans_lock);
4648 }
4649 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004650
Josef Bacik724e2312013-09-30 11:36:38 -04004651 spin_lock(&root->fs_info->trans_lock);
4652 if (t == root->fs_info->running_transaction)
4653 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004654 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004655 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004656
Josef Bacik724e2312013-09-30 11:36:38 -04004657 btrfs_put_transaction(t);
4658 trace_btrfs_transaction_commit(root);
4659 spin_lock(&root->fs_info->trans_lock);
4660 }
4661 spin_unlock(&root->fs_info->trans_lock);
4662 btrfs_destroy_all_ordered_extents(root->fs_info);
4663 btrfs_destroy_delayed_inodes(root);
4664 btrfs_assert_delayed_root_empty(root);
4665 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4666 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004667 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004668
4669 return 0;
4670}
4671
David Sterbae8c9f182015-01-02 18:23:10 +01004672static const struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004673 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004674 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004675 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004676 /* note we're sharing with inode.c for the merge bio hook */
4677 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004678};