blob: 1e60d00d4ea7c42104614ede9e203a1f56e6408a [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050030#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020031#include <linux/ratelimit.h>
Stefan Behrens6463fe52013-04-19 15:08:05 +000032#include <linux/uuid.h>
Stefan Behrens803b2f52013-08-15 17:11:21 +020033#include <linux/semaphore.h>
David Sterba7e75bf32011-03-18 22:56:43 +000034#include <asm/unaligned.h>
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +000037#include "hash.h"
Chris Masone089f052007-03-16 16:20:31 -040038#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040039#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040040#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040041#include "print-tree.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080045#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010046#include "check-integrity.h"
Josef Bacik606686e2012-06-04 14:03:51 -040047#include "rcu-string.h"
Stefan Behrens8dabb742012-11-06 13:15:27 +010048#include "dev-replace.h"
David Woodhouse53b381b2013-01-29 18:40:14 -050049#include "raid56.h"
Jeff Mahoney5ac1d202013-11-01 13:06:58 -040050#include "sysfs.h"
Josef Bacikfcebe452014-05-13 17:30:47 -070051#include "qgroup.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050052
Josef Bacikde0022b2012-09-25 14:25:58 -040053#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
David Sterbae8c9f182015-01-02 18:23:10 +010057static const struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040058static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040059static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010060static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080061 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010062static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080063static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010065static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080066static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 struct extent_io_tree *dirty_pages,
68 int mark);
69static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000071static int btrfs_cleanup_transaction(struct btrfs_root *root);
72static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040073
Chris Masond352ac62008-09-29 15:18:18 -040074/*
David Sterba97eb6b62014-07-30 00:55:42 +020075 * btrfs_end_io_wq structs are used to do processing in task context when an IO
76 * is complete. This is used during reads to verify checksums, and it is used
Chris Masond352ac62008-09-29 15:18:18 -040077 * by writes to insert metadata for new file extents after IO is complete.
78 */
David Sterba97eb6b62014-07-30 00:55:42 +020079struct btrfs_end_io_wq {
Chris Masonce9adaa2008-04-09 16:28:12 -040080 struct bio *bio;
81 bio_end_io_t *end_io;
82 void *private;
83 struct btrfs_fs_info *info;
84 int error;
David Sterbabfebd8b2014-07-30 00:25:45 +020085 enum btrfs_wq_endio_type metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040086 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040087 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040088};
Chris Mason0da54682007-11-07 21:08:01 -050089
David Sterba97eb6b62014-07-30 00:55:42 +020090static struct kmem_cache *btrfs_end_io_wq_cache;
91
92int __init btrfs_end_io_wq_init(void)
93{
94 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
95 sizeof(struct btrfs_end_io_wq),
96 0,
97 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
98 NULL);
99 if (!btrfs_end_io_wq_cache)
100 return -ENOMEM;
101 return 0;
102}
103
104void btrfs_end_io_wq_exit(void)
105{
106 if (btrfs_end_io_wq_cache)
107 kmem_cache_destroy(btrfs_end_io_wq_cache);
108}
109
Chris Masond352ac62008-09-29 15:18:18 -0400110/*
111 * async submit bios are used to offload expensive checksumming
112 * onto the worker threads. They checksum file and metadata bios
113 * just before they are sent down the IO stack.
114 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400115struct async_submit_bio {
116 struct inode *inode;
117 struct bio *bio;
118 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500119 extent_submit_bio_hook_t *submit_bio_start;
120 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400121 int rw;
122 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400123 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400124 /*
125 * bio_offset is optional, can be used if the pages in the bio
126 * can't tell us where in the file the bio should go
127 */
128 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400129 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100130 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400131};
132
Chris Mason85d4e462011-07-26 16:11:19 -0400133/*
134 * Lockdep class keys for extent_buffer->lock's in this root. For a given
135 * eb, the lockdep key is determined by the btrfs_root it belongs to and
136 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500137 *
Chris Mason85d4e462011-07-26 16:11:19 -0400138 * Different roots are used for different purposes and may nest inside each
139 * other and they require separate keysets. As lockdep keys should be
140 * static, assign keysets according to the purpose of the root as indicated
141 * by btrfs_root->objectid. This ensures that all special purpose roots
142 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500143 *
Chris Mason85d4e462011-07-26 16:11:19 -0400144 * Lock-nesting across peer nodes is always done with the immediate parent
145 * node locked thus preventing deadlock. As lockdep doesn't know this, use
146 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500147 *
Chris Mason85d4e462011-07-26 16:11:19 -0400148 * The key is set by the readpage_end_io_hook after the buffer has passed
149 * csum validation but before the pages are unlocked. It is also set by
150 * btrfs_init_new_buffer on freshly allocated blocks.
151 *
152 * We also add a check to make sure the highest level of the tree is the
153 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
154 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500155 */
156#ifdef CONFIG_DEBUG_LOCK_ALLOC
157# if BTRFS_MAX_LEVEL != 8
158# error
159# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400160
161static struct btrfs_lockdep_keyset {
162 u64 id; /* root objectid */
163 const char *name_stem; /* lock name stem */
164 char names[BTRFS_MAX_LEVEL + 1][20];
165 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
166} btrfs_lockdep_keysets[] = {
167 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
168 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
169 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
170 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
171 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
172 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000173 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400174 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
175 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
176 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200177 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
Chris Mason85d4e462011-07-26 16:11:19 -0400178 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500179};
Chris Mason85d4e462011-07-26 16:11:19 -0400180
181void __init btrfs_init_lockdep(void)
182{
183 int i, j;
184
185 /* initialize lockdep class names */
186 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
187 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
188
189 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
190 snprintf(ks->names[j], sizeof(ks->names[j]),
191 "btrfs-%s-%02d", ks->name_stem, j);
192 }
193}
194
195void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
196 int level)
197{
198 struct btrfs_lockdep_keyset *ks;
199
200 BUG_ON(level >= ARRAY_SIZE(ks->keys));
201
202 /* find the matching keyset, id 0 is the default entry */
203 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
204 if (ks->id == objectid)
205 break;
206
207 lockdep_set_class_and_name(&eb->lock,
208 &ks->keys[level], ks->names[level]);
209}
210
Chris Mason4008c042009-02-12 14:09:45 -0500211#endif
212
Chris Masond352ac62008-09-29 15:18:18 -0400213/*
214 * extents on the btree inode are pretty simple, there's one extent
215 * that covers the entire device
216 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500217static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200218 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500219 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400220{
221 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
222 struct extent_map *em;
223 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400224
Chris Mason890871b2009-09-02 16:24:52 -0400225 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500226 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400227 if (em) {
228 em->bdev =
229 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400230 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400231 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400232 }
Chris Mason890871b2009-09-02 16:24:52 -0400233 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400234
David Sterba172ddd62011-04-21 00:48:27 +0200235 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400236 if (!em) {
237 em = ERR_PTR(-ENOMEM);
238 goto out;
239 }
240 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400241 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400242 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400243 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400244 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500245
Chris Mason890871b2009-09-02 16:24:52 -0400246 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400247 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400248 if (ret == -EEXIST) {
249 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400250 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600251 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600252 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400253 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400254 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600255 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400256 }
Chris Mason890871b2009-09-02 16:24:52 -0400257 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400258
Chris Mason5f39d392007-10-15 16:14:19 -0400259out:
260 return em;
261}
262
Liu Bob0496682013-03-14 14:57:45 +0000263u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400264{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000265 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400266}
267
268void btrfs_csum_final(u32 crc, char *result)
269{
David Sterba7e75bf32011-03-18 22:56:43 +0000270 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400271}
272
Chris Masond352ac62008-09-29 15:18:18 -0400273/*
274 * compute the csum for a btree block, and either verify it or write it
275 * into the csum field of the block.
276 */
Daniel Dressler01d58472014-11-21 17:15:07 +0900277static int csum_tree_block(struct btrfs_fs_info *fs_info,
278 struct extent_buffer *buf,
Chris Mason19c00dd2007-10-15 16:19:22 -0400279 int verify)
280{
Daniel Dressler01d58472014-11-21 17:15:07 +0900281 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500282 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400283 unsigned long len;
284 unsigned long cur_len;
285 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400286 char *kaddr;
287 unsigned long map_start;
288 unsigned long map_len;
289 int err;
290 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500291 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400292
293 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500294 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400295 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400296 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500297 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400298 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400299 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000300 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400301 crc, cur_len);
302 len -= cur_len;
303 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400304 }
Josef Bacik607d4322008-12-02 07:17:45 -0500305 if (csum_size > sizeof(inline_result)) {
David Sterba31e818f2015-02-20 18:00:26 +0100306 result = kzalloc(csum_size, GFP_NOFS);
Josef Bacik607d4322008-12-02 07:17:45 -0500307 if (!result)
308 return 1;
309 } else {
310 result = (char *)&inline_result;
311 }
312
Chris Mason19c00dd2007-10-15 16:19:22 -0400313 btrfs_csum_final(crc, result);
314
315 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500316 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500317 u32 val;
318 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500319 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500320
Josef Bacik607d4322008-12-02 07:17:45 -0500321 read_extent_buffer(buf, &val, 0, csum_size);
David Sterbaf0954c62014-12-19 18:38:44 +0100322 printk_ratelimited(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -0500323 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
324 "level %d\n",
Daniel Dressler01d58472014-11-21 17:15:07 +0900325 fs_info->sb->s_id, buf->start,
Frank Holtonefe120a2013-12-20 11:37:06 -0500326 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500327 if (result != (char *)&inline_result)
328 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400329 return 1;
330 }
331 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500332 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400333 }
Josef Bacik607d4322008-12-02 07:17:45 -0500334 if (result != (char *)&inline_result)
335 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400336 return 0;
337}
338
Chris Masond352ac62008-09-29 15:18:18 -0400339/*
340 * we can't consider a given block up to date unless the transid of the
341 * block matches the transid in the parent node's pointer. This is how we
342 * detect blocks that either didn't get written at all or got written
343 * in the wrong place.
344 */
Chris Mason1259ab72008-05-12 13:39:03 -0400345static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400346 struct extent_buffer *eb, u64 parent_transid,
347 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400348{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000349 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400350 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200351 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400352
353 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
354 return 0;
355
Chris Masonb9fab912012-05-06 07:23:47 -0400356 if (atomic)
357 return -EAGAIN;
358
Josef Bacika26e8c92014-03-28 17:07:27 -0400359 if (need_lock) {
360 btrfs_tree_read_lock(eb);
361 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
362 }
363
Josef Bacik2ac55d42010-02-03 19:33:23 +0000364 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100365 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400366 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400367 btrfs_header_generation(eb) == parent_transid) {
368 ret = 0;
369 goto out;
370 }
David Sterba68b663d2014-12-19 18:38:37 +0100371 printk_ratelimited(KERN_ERR
372 "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n",
Wang Shilong29549ae2014-07-04 17:59:06 +0800373 eb->fs_info->sb->s_id, eb->start,
374 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400375 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400376
377 /*
378 * Things reading via commit roots that don't have normal protection,
379 * like send, can have a really old block in cache that may point at a
380 * block that has been free'd and re-allocated. So don't clear uptodate
381 * if we find an eb that is under IO (dirty/writeback) because we could
382 * end up reading in the stale data and then writing it back out and
383 * making everybody very sad.
384 */
385 if (!extent_buffer_under_io(eb))
386 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400387out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000388 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
389 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700390 if (need_lock)
391 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400392 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400393}
394
Chris Masond352ac62008-09-29 15:18:18 -0400395/*
David Sterba1104a882013-03-06 15:57:46 +0100396 * Return 0 if the superblock checksum type matches the checksum value of that
397 * algorithm. Pass the raw disk superblock data.
398 */
399static int btrfs_check_super_csum(char *raw_disk_sb)
400{
401 struct btrfs_super_block *disk_sb =
402 (struct btrfs_super_block *)raw_disk_sb;
403 u16 csum_type = btrfs_super_csum_type(disk_sb);
404 int ret = 0;
405
406 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
407 u32 crc = ~(u32)0;
408 const int csum_size = sizeof(crc);
409 char result[csum_size];
410
411 /*
412 * The super_block structure does not span the whole
413 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
414 * is filled with zeros and is included in the checkum.
415 */
416 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
417 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
418 btrfs_csum_final(crc, result);
419
420 if (memcmp(raw_disk_sb, result, csum_size))
421 ret = 1;
422 }
423
424 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500425 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100426 csum_type);
427 ret = 1;
428 }
429
430 return ret;
431}
432
433/*
Chris Masond352ac62008-09-29 15:18:18 -0400434 * helper to read a given tree block, doing retries as required when
435 * the checksums don't match and we have alternate mirrors to try.
436 */
Chris Masonf1885912008-04-09 16:28:12 -0400437static int btree_read_extent_buffer_pages(struct btrfs_root *root,
438 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400439 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400440{
441 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400442 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400443 int ret;
444 int num_copies = 0;
445 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400446 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400447
Josef Bacika826d6d2011-03-16 13:42:43 -0400448 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400449 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
450 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200451 ret = read_extent_buffer_pages(io_tree, eb, start,
452 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400453 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600454 if (!ret) {
455 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400456 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600457 break;
458 else
459 ret = -EIO;
460 }
Chris Masond3977122009-01-05 21:25:51 -0500461
Josef Bacika826d6d2011-03-16 13:42:43 -0400462 /*
463 * This buffer's crc is fine, but its contents are corrupted, so
464 * there is no reason to read the other copies, they won't be
465 * any less wrong.
466 */
467 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400468 break;
469
Stefan Behrens5d964052012-11-05 14:59:07 +0100470 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400471 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400472 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400473 break;
Chris Mason42352982008-04-28 16:40:52 -0400474
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400475 if (!failed_mirror) {
476 failed = 1;
477 failed_mirror = eb->read_mirror;
478 }
479
Chris Masonf1885912008-04-09 16:28:12 -0400480 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400481 if (mirror_num == failed_mirror)
482 mirror_num++;
483
Chris Mason42352982008-04-28 16:40:52 -0400484 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400485 break;
Chris Masonf1885912008-04-09 16:28:12 -0400486 }
Josef Bacikea466792012-03-26 21:57:36 -0400487
Stefan Behrensc0901582012-07-10 07:30:17 -0600488 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400489 repair_eb_io_failure(root, eb, failed_mirror);
490
491 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400492}
Chris Mason19c00dd2007-10-15 16:19:22 -0400493
Chris Masond352ac62008-09-29 15:18:18 -0400494/*
Chris Masond3977122009-01-05 21:25:51 -0500495 * checksum a dirty tree block before IO. This has extra checks to make sure
496 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400497 */
Chris Masond3977122009-01-05 21:25:51 -0500498
Daniel Dressler01d58472014-11-21 17:15:07 +0900499static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400500{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000501 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400502 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400503 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400504
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500505 eb = (struct extent_buffer *)page->private;
506 if (page != eb->pages[0])
507 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400508 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530509 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500510 return 0;
Daniel Dressler01d58472014-11-21 17:15:07 +0900511 csum_tree_block(fs_info, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400512 return 0;
513}
514
Daniel Dressler01d58472014-11-21 17:15:07 +0900515static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
Yan Zheng2b820322008-11-17 21:11:30 -0500516 struct extent_buffer *eb)
517{
Daniel Dressler01d58472014-11-21 17:15:07 +0900518 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
Yan Zheng2b820322008-11-17 21:11:30 -0500519 u8 fsid[BTRFS_UUID_SIZE];
520 int ret = 1;
521
Ross Kirk0a4e5582013-09-24 10:12:38 +0100522 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500523 while (fs_devices) {
524 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
525 ret = 0;
526 break;
527 }
528 fs_devices = fs_devices->seed;
529 }
530 return ret;
531}
532
Josef Bacika826d6d2011-03-16 13:42:43 -0400533#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500534 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
535 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200536 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400537
538static noinline int check_leaf(struct btrfs_root *root,
539 struct extent_buffer *leaf)
540{
541 struct btrfs_key key;
542 struct btrfs_key leaf_key;
543 u32 nritems = btrfs_header_nritems(leaf);
544 int slot;
545
546 if (nritems == 0)
547 return 0;
548
549 /* Check the 0 item */
550 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
551 BTRFS_LEAF_DATA_SIZE(root)) {
552 CORRUPT("invalid item offset size pair", leaf, root, 0);
553 return -EIO;
554 }
555
556 /*
557 * Check to make sure each items keys are in the correct order and their
558 * offsets make sense. We only have to loop through nritems-1 because
559 * we check the current slot against the next slot, which verifies the
560 * next slot's offset+size makes sense and that the current's slot
561 * offset is correct.
562 */
563 for (slot = 0; slot < nritems - 1; slot++) {
564 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
565 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
566
567 /* Make sure the keys are in the right order */
568 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
569 CORRUPT("bad key order", leaf, root, slot);
570 return -EIO;
571 }
572
573 /*
574 * Make sure the offset and ends are right, remember that the
575 * item data starts at the end of the leaf and grows towards the
576 * front.
577 */
578 if (btrfs_item_offset_nr(leaf, slot) !=
579 btrfs_item_end_nr(leaf, slot + 1)) {
580 CORRUPT("slot offset bad", leaf, root, slot);
581 return -EIO;
582 }
583
584 /*
585 * Check to make sure that we don't point outside of the leaf,
586 * just incase all the items are consistent to eachother, but
587 * all point outside of the leaf.
588 */
589 if (btrfs_item_end_nr(leaf, slot) >
590 BTRFS_LEAF_DATA_SIZE(root)) {
591 CORRUPT("slot end outside of leaf", leaf, root, slot);
592 return -EIO;
593 }
594 }
595
596 return 0;
597}
598
Miao Xiefacc8a222013-07-25 19:22:34 +0800599static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
600 u64 phy_offset, struct page *page,
601 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400602{
Chris Masonce9adaa2008-04-09 16:28:12 -0400603 u64 found_start;
604 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400605 struct extent_buffer *eb;
606 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400607 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400608 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400609
Chris Masonce9adaa2008-04-09 16:28:12 -0400610 if (!page->private)
611 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500612
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500613 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500614
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400615 /* the pending IO might have been the only thing that kept this buffer
616 * in memory. Make sure we have a ref for all this other checks
617 */
618 extent_buffer_get(eb);
619
620 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400621 if (!reads_done)
622 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400623
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400624 eb->read_mirror = mirror;
Filipe Manana656f30d2014-09-26 12:25:56 +0100625 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
Josef Bacikea466792012-03-26 21:57:36 -0400626 ret = -EIO;
627 goto err;
628 }
629
Chris Masonce9adaa2008-04-09 16:28:12 -0400630 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400631 if (found_start != eb->start) {
David Sterba68b663d2014-12-19 18:38:37 +0100632 printk_ratelimited(KERN_ERR "BTRFS (device %s): bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400633 "%llu %llu\n",
Wang Shilong29549ae2014-07-04 17:59:06 +0800634 eb->fs_info->sb->s_id, found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400635 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400636 goto err;
637 }
Daniel Dressler01d58472014-11-21 17:15:07 +0900638 if (check_tree_block_fsid(root->fs_info, eb)) {
David Sterba68b663d2014-12-19 18:38:37 +0100639 printk_ratelimited(KERN_ERR "BTRFS (device %s): bad fsid on block %llu\n",
Wang Shilong29549ae2014-07-04 17:59:06 +0800640 eb->fs_info->sb->s_id, eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400641 ret = -EIO;
642 goto err;
643 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400644 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400645 if (found_level >= BTRFS_MAX_LEVEL) {
David Sterba68b663d2014-12-19 18:38:37 +0100646 btrfs_err(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400647 (int)btrfs_header_level(eb));
648 ret = -EIO;
649 goto err;
650 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400651
Chris Mason85d4e462011-07-26 16:11:19 -0400652 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
653 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500654
Daniel Dressler01d58472014-11-21 17:15:07 +0900655 ret = csum_tree_block(root->fs_info, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400656 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400657 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400658 goto err;
659 }
660
661 /*
662 * If this is a leaf block and it is corrupt, set the corrupt bit so
663 * that we don't try and read the other copies of this block, just
664 * return -EIO.
665 */
666 if (found_level == 0 && check_leaf(root, eb)) {
667 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
668 ret = -EIO;
669 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400670
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400671 if (!ret)
672 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400673err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400674 if (reads_done &&
675 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200676 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200677
David Woodhouse53b381b2013-01-29 18:40:14 -0500678 if (ret) {
679 /*
680 * our io error hook is going to dec the io pages
681 * again, we have to make sure it has something
682 * to decrement
683 */
684 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400685 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500686 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400687 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400688out:
Chris Masonf1885912008-04-09 16:28:12 -0400689 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400690}
691
Josef Bacikea466792012-03-26 21:57:36 -0400692static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200693{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200694 struct extent_buffer *eb;
695 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
696
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500697 eb = (struct extent_buffer *)page->private;
Filipe Manana656f30d2014-09-26 12:25:56 +0100698 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400699 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500700 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400701 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200702 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200703 return -EIO; /* we fixed nothing */
704}
705
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200706static void end_workqueue_bio(struct bio *bio)
Chris Masonce9adaa2008-04-09 16:28:12 -0400707{
David Sterba97eb6b62014-07-30 00:55:42 +0200708 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400709 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800710 struct btrfs_workqueue *wq;
711 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400712
Chris Masonce9adaa2008-04-09 16:28:12 -0400713 fs_info = end_io_wq->info;
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200714 end_io_wq->error = bio->bi_error;
Chris Masond20f7042008-12-08 16:58:54 -0500715
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200716 if (bio->bi_rw & REQ_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800717 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
718 wq = fs_info->endio_meta_write_workers;
719 func = btrfs_endio_meta_write_helper;
720 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
721 wq = fs_info->endio_freespace_worker;
722 func = btrfs_freespace_write_helper;
723 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
724 wq = fs_info->endio_raid56_workers;
725 func = btrfs_endio_raid56_helper;
726 } else {
727 wq = fs_info->endio_write_workers;
728 func = btrfs_endio_write_helper;
729 }
Chris Masond20f7042008-12-08 16:58:54 -0500730 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800731 if (unlikely(end_io_wq->metadata ==
732 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
733 wq = fs_info->endio_repair_workers;
734 func = btrfs_endio_repair_helper;
735 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800736 wq = fs_info->endio_raid56_workers;
737 func = btrfs_endio_raid56_helper;
738 } else if (end_io_wq->metadata) {
739 wq = fs_info->endio_meta_workers;
740 func = btrfs_endio_meta_helper;
741 } else {
742 wq = fs_info->endio_workers;
743 func = btrfs_endio_helper;
744 }
Chris Masond20f7042008-12-08 16:58:54 -0500745 }
Liu Bo9e0af232014-08-15 23:36:53 +0800746
747 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
748 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400749}
750
Chris Mason22c59942008-04-09 16:28:12 -0400751int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200752 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400753{
David Sterba97eb6b62014-07-30 00:55:42 +0200754 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800755
David Sterba97eb6b62014-07-30 00:55:42 +0200756 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400757 if (!end_io_wq)
758 return -ENOMEM;
759
760 end_io_wq->private = bio->bi_private;
761 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400762 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400763 end_io_wq->error = 0;
764 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400765 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400766
767 bio->bi_private = end_io_wq;
768 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400769 return 0;
770}
771
Chris Masonb64a2852008-08-20 13:39:41 -0400772unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400773{
774 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800775 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400776 info->fs_devices->open_devices);
777 return 256 * limit;
778}
779
Chris Mason4a69a412008-11-06 22:03:00 -0500780static void run_one_async_start(struct btrfs_work *work)
781{
Chris Mason4a69a412008-11-06 22:03:00 -0500782 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100783 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500784
785 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100786 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
787 async->mirror_num, async->bio_flags,
788 async->bio_offset);
789 if (ret)
790 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500791}
792
793static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400794{
795 struct btrfs_fs_info *fs_info;
796 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400797 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400798
799 async = container_of(work, struct async_submit_bio, work);
800 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400801
Chris Masonb64a2852008-08-20 13:39:41 -0400802 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400803 limit = limit * 2 / 3;
804
Josef Bacik66657b32012-08-01 15:36:24 -0400805 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400806 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400807 wake_up(&fs_info->async_submit_wait);
808
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100809 /* If an error occured we just want to clean up the bio and move on */
810 if (async->error) {
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200811 async->bio->bi_error = async->error;
812 bio_endio(async->bio);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100813 return;
814 }
815
Chris Mason4a69a412008-11-06 22:03:00 -0500816 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400817 async->mirror_num, async->bio_flags,
818 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500819}
820
821static void run_one_async_free(struct btrfs_work *work)
822{
823 struct async_submit_bio *async;
824
825 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400826 kfree(async);
827}
828
Chris Mason44b8bd72008-04-16 11:14:51 -0400829int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
830 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400831 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400832 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500833 extent_submit_bio_hook_t *submit_bio_start,
834 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400835{
836 struct async_submit_bio *async;
837
838 async = kmalloc(sizeof(*async), GFP_NOFS);
839 if (!async)
840 return -ENOMEM;
841
842 async->inode = inode;
843 async->rw = rw;
844 async->bio = bio;
845 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500846 async->submit_bio_start = submit_bio_start;
847 async->submit_bio_done = submit_bio_done;
848
Liu Bo9e0af232014-08-15 23:36:53 +0800849 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800850 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500851
Chris Masonc8b97812008-10-29 14:49:59 -0400852 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400853 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400854
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100855 async->error = 0;
856
Chris Masoncb03c742008-05-15 16:15:45 -0400857 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400858
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200859 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800860 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400861
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800862 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400863
Chris Masond3977122009-01-05 21:25:51 -0500864 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500865 atomic_read(&fs_info->nr_async_submits)) {
866 wait_event(fs_info->async_submit_wait,
867 (atomic_read(&fs_info->nr_async_submits) == 0));
868 }
869
Chris Mason44b8bd72008-04-16 11:14:51 -0400870 return 0;
871}
872
Chris Masonce3ed712008-09-23 13:14:12 -0400873static int btree_csum_one_bio(struct bio *bio)
874{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800875 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400876 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800877 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400878
Kent Overstreet2c30c712013-11-07 12:20:26 -0800879 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400880 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Daniel Dressler01d58472014-11-21 17:15:07 +0900881 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100882 if (ret)
883 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400884 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800885
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100886 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400887}
888
Chris Mason4a69a412008-11-06 22:03:00 -0500889static int __btree_submit_bio_start(struct inode *inode, int rw,
890 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400891 unsigned long bio_flags,
892 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400893{
Chris Mason8b712842008-06-11 16:50:36 -0400894 /*
895 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400896 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400897 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100898 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500899}
Chris Mason22c59942008-04-09 16:28:12 -0400900
Chris Mason4a69a412008-11-06 22:03:00 -0500901static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400902 int mirror_num, unsigned long bio_flags,
903 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500904{
Stefan Behrens61891922012-11-05 18:51:52 +0100905 int ret;
906
Chris Mason8b712842008-06-11 16:50:36 -0400907 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500908 * when we're called for a write, we're already in the async
909 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400910 */
Stefan Behrens61891922012-11-05 18:51:52 +0100911 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200912 if (ret) {
913 bio->bi_error = ret;
914 bio_endio(bio);
915 }
Stefan Behrens61891922012-11-05 18:51:52 +0100916 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400917}
918
Josef Bacikde0022b2012-09-25 14:25:58 -0400919static int check_async_write(struct inode *inode, unsigned long bio_flags)
920{
921 if (bio_flags & EXTENT_BIO_TREE_LOG)
922 return 0;
923#ifdef CONFIG_X86
924 if (cpu_has_xmm4_2)
925 return 0;
926#endif
927 return 1;
928}
929
Chris Mason44b8bd72008-04-16 11:14:51 -0400930static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400931 int mirror_num, unsigned long bio_flags,
932 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400933{
Josef Bacikde0022b2012-09-25 14:25:58 -0400934 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500935 int ret;
936
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200937 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500938 /*
939 * called for a read, do the setup so that checksum validation
940 * can happen in the async kernel threads
941 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400942 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +0200943 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -0400944 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100945 goto out_w_error;
946 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
947 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400948 } else if (!async) {
949 ret = btree_csum_one_bio(bio);
950 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100951 goto out_w_error;
952 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
953 mirror_num, 0);
954 } else {
955 /*
956 * kthread helpers are used to submit writes so that
957 * checksumming can happen in parallel across all CPUs
958 */
959 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
960 inode, rw, bio, mirror_num, 0,
961 bio_offset,
962 __btree_submit_bio_start,
963 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400964 }
Chris Masond313d7a2009-04-20 15:50:09 -0400965
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200966 if (ret)
967 goto out_w_error;
968 return 0;
969
Stefan Behrens61891922012-11-05 18:51:52 +0100970out_w_error:
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200971 bio->bi_error = ret;
972 bio_endio(bio);
Stefan Behrens61891922012-11-05 18:51:52 +0100973 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400974}
975
Jan Beulich3dd14622010-12-07 14:54:09 +0000976#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500977static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800978 struct page *newpage, struct page *page,
979 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500980{
981 /*
982 * we can't safely write a btree page from here,
983 * we haven't done the locking hook
984 */
985 if (PageDirty(page))
986 return -EAGAIN;
987 /*
988 * Buffers may be managed in a filesystem specific way.
989 * We must have no buffers or drop them.
990 */
991 if (page_has_private(page) &&
992 !try_to_release_page(page, GFP_KERNEL))
993 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800994 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500995}
Jan Beulich3dd14622010-12-07 14:54:09 +0000996#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500997
Chris Mason0da54682007-11-07 21:08:01 -0500998
999static int btree_writepages(struct address_space *mapping,
1000 struct writeback_control *wbc)
1001{
Miao Xiee2d84522013-01-29 10:09:20 +00001002 struct btrfs_fs_info *fs_info;
1003 int ret;
1004
Chris Masond8d5f3e2007-12-11 12:42:00 -05001005 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001006
1007 if (wbc->for_kupdate)
1008 return 0;
1009
Miao Xiee2d84522013-01-29 10:09:20 +00001010 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001011 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001012 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1013 BTRFS_DIRTY_METADATA_THRESH);
1014 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001015 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001016 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001017 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001018}
1019
Christoph Hellwigb2950862008-12-02 09:54:17 -05001020static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001021{
Chris Masond1310b22008-01-24 16:13:08 -05001022 struct extent_io_tree *tree;
1023 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001024 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001025}
1026
Chris Mason70dec802008-01-29 09:59:12 -05001027static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001028{
Chris Mason98509cf2008-09-11 15:51:43 -04001029 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001030 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001031
David Sterbaf7a52a42013-04-26 14:56:29 +00001032 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001033}
Chris Mason123abc82007-03-14 14:14:43 -04001034
Lukas Czernerd47992f2013-05-21 23:17:23 -04001035static void btree_invalidatepage(struct page *page, unsigned int offset,
1036 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001037{
Chris Masond1310b22008-01-24 16:13:08 -05001038 struct extent_io_tree *tree;
1039 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001040 extent_invalidatepage(tree, page, offset);
1041 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001042 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001043 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1044 "page private not zero on page %llu",
1045 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001046 ClearPagePrivate(page);
1047 set_page_private(page, 0);
1048 page_cache_release(page);
1049 }
Chris Masond98237b2007-03-28 13:57:48 -04001050}
1051
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001052static int btree_set_page_dirty(struct page *page)
1053{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001054#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001055 struct extent_buffer *eb;
1056
1057 BUG_ON(!PagePrivate(page));
1058 eb = (struct extent_buffer *)page->private;
1059 BUG_ON(!eb);
1060 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1061 BUG_ON(!atomic_read(&eb->refs));
1062 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001063#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001064 return __set_page_dirty_nobuffers(page);
1065}
1066
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001067static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001068 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001069 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001070 .releasepage = btree_releasepage,
1071 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001072#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001073 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001074#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001075 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001076};
1077
David Sterbad3e46fe2014-06-15 02:04:19 +02001078void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
Chris Mason090d1872007-05-01 08:53:32 -04001079{
Chris Mason5f39d392007-10-15 16:14:19 -04001080 struct extent_buffer *buf = NULL;
1081 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason090d1872007-05-01 08:53:32 -04001082
David Sterbaa83fffb2014-06-15 02:39:54 +02001083 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001084 if (!buf)
David Sterba6197d862014-06-15 00:49:36 +02001085 return;
Chris Masond1310b22008-01-24 16:13:08 -05001086 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001087 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001088 free_extent_buffer(buf);
Chris Mason090d1872007-05-01 08:53:32 -04001089}
1090
David Sterbac0dcaa42014-06-15 02:24:21 +02001091int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
Arne Jansenab0fff02011-05-23 14:25:41 +02001092 int mirror_num, struct extent_buffer **eb)
1093{
1094 struct extent_buffer *buf = NULL;
1095 struct inode *btree_inode = root->fs_info->btree_inode;
1096 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1097 int ret;
1098
David Sterbaa83fffb2014-06-15 02:39:54 +02001099 buf = btrfs_find_create_tree_block(root, bytenr);
Arne Jansenab0fff02011-05-23 14:25:41 +02001100 if (!buf)
1101 return 0;
1102
1103 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1104
1105 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1106 btree_get_extent, mirror_num);
1107 if (ret) {
1108 free_extent_buffer(buf);
1109 return ret;
1110 }
1111
1112 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1113 free_extent_buffer(buf);
1114 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001115 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001116 *eb = buf;
1117 } else {
1118 free_extent_buffer(buf);
1119 }
1120 return 0;
1121}
1122
Daniel Dressler01d58472014-11-21 17:15:07 +09001123struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
David Sterba0308af42014-06-15 01:43:40 +02001124 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001125{
Daniel Dressler01d58472014-11-21 17:15:07 +09001126 return find_extent_buffer(fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001127}
1128
1129struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
David Sterbaa83fffb2014-06-15 02:39:54 +02001130 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001131{
David Sterbafccb84c2014-09-29 23:53:21 +02001132 if (btrfs_test_is_dummy_root(root))
David Sterbace3e6982014-06-15 03:00:04 +02001133 return alloc_test_extent_buffer(root->fs_info, bytenr);
1134 return alloc_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001135}
1136
1137
Chris Masone02119d2008-09-05 16:13:11 -04001138int btrfs_write_tree_block(struct extent_buffer *buf)
1139{
Chris Mason727011e2010-08-06 13:21:20 -04001140 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001141 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001142}
1143
1144int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1145{
Chris Mason727011e2010-08-06 13:21:20 -04001146 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001147 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001148}
1149
Chris Masondb945352007-10-15 16:15:53 -04001150struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
David Sterbace86cd52014-06-15 01:07:32 +02001151 u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001152{
Chris Mason5f39d392007-10-15 16:14:19 -04001153 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001154 int ret;
1155
David Sterbaa83fffb2014-06-15 02:39:54 +02001156 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001157 if (!buf)
Liu Bo64c043d2015-05-25 17:30:15 +08001158 return ERR_PTR(-ENOMEM);
Chris Masone4204de2008-01-08 15:46:27 -05001159
Chris Masonca7a79a2008-05-12 12:59:19 -04001160 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001161 if (ret) {
1162 free_extent_buffer(buf);
Liu Bo64c043d2015-05-25 17:30:15 +08001163 return ERR_PTR(ret);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001164 }
Chris Mason5f39d392007-10-15 16:14:19 -04001165 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001166
Chris Masoneb60cea2007-02-02 09:18:22 -05001167}
1168
Daniel Dressler01d58472014-11-21 17:15:07 +09001169void clean_tree_block(struct btrfs_trans_handle *trans,
1170 struct btrfs_fs_info *fs_info,
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001171 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001172{
Chris Mason55c69072008-01-09 15:55:33 -05001173 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001174 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001175 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001176
Chris Masonb9473432009-03-13 11:00:37 -04001177 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001178 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1179 -buf->len,
1180 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001181 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1182 btrfs_set_lock_blocking(buf);
1183 clear_extent_buffer_dirty(buf);
1184 }
Chris Mason925baed2008-06-25 16:01:30 -04001185 }
Chris Mason5f39d392007-10-15 16:14:19 -04001186}
1187
Miao Xie8257b2d2014-03-06 13:38:19 +08001188static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1189{
1190 struct btrfs_subvolume_writers *writers;
1191 int ret;
1192
1193 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1194 if (!writers)
1195 return ERR_PTR(-ENOMEM);
1196
Tejun Heo908c7f12014-09-08 09:51:29 +09001197 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
Miao Xie8257b2d2014-03-06 13:38:19 +08001198 if (ret < 0) {
1199 kfree(writers);
1200 return ERR_PTR(ret);
1201 }
1202
1203 init_waitqueue_head(&writers->wait);
1204 return writers;
1205}
1206
1207static void
1208btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1209{
1210 percpu_counter_destroy(&writers->counter);
1211 kfree(writers);
1212}
1213
David Sterba707e8a02014-06-04 19:22:26 +02001214static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1215 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
Jeff Mahoney143bede2012-03-01 14:56:26 +01001216 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001217{
Chris Masoncfaa7292007-02-21 17:04:57 -05001218 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001219 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001220 root->sectorsize = sectorsize;
1221 root->nodesize = nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001222 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001223 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001224 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001225
Chris Mason0f7d52f2007-04-09 10:42:37 -04001226 root->objectid = objectid;
1227 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001228 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001229 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001230 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001231 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001232 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001233 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001234 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001235 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001236
1237 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001238 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001239 INIT_LIST_HEAD(&root->delalloc_inodes);
1240 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001241 INIT_LIST_HEAD(&root->ordered_extents);
1242 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001243 INIT_LIST_HEAD(&root->logged_list[0]);
1244 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001245 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001246 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001247 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001248 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001249 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001250 spin_lock_init(&root->log_extents_lock[0]);
1251 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001252 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001253 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001254 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001255 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001256 init_waitqueue_head(&root->log_writer_wait);
1257 init_waitqueue_head(&root->log_commit_wait[0]);
1258 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001259 INIT_LIST_HEAD(&root->log_ctxs[0]);
1260 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f182009-01-21 12:54:03 -05001261 atomic_set(&root->log_commit[0], 0);
1262 atomic_set(&root->log_commit[1], 0);
1263 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001264 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001265 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001266 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001267 atomic_set(&root->will_be_snapshoted, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001268 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001269 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001270 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001271 if (fs_info)
1272 extent_io_tree_init(&root->dirty_log_pages,
1273 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001274
Chris Mason3768f362007-03-13 16:47:54 -04001275 memset(&root->root_key, 0, sizeof(root->root_key));
1276 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001277 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001278 if (fs_info)
1279 root->defrag_trans_start = fs_info->generation;
1280 else
1281 root->defrag_trans_start = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001282 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001283 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001284
Anand Jain5f3ab902012-12-07 09:28:54 +00001285 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001286}
1287
Al Virof84a8bd2011-11-17 15:57:57 -05001288static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001289{
1290 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1291 if (root)
1292 root->fs_info = fs_info;
1293 return root;
1294}
1295
Josef Bacik06ea65a2013-09-19 16:07:01 -04001296#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1297/* Should only be used by the testing infrastructure */
1298struct btrfs_root *btrfs_alloc_dummy_root(void)
1299{
1300 struct btrfs_root *root;
1301
1302 root = btrfs_alloc_root(NULL);
1303 if (!root)
1304 return ERR_PTR(-ENOMEM);
David Sterba707e8a02014-06-04 19:22:26 +02001305 __setup_root(4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001306 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001307 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001308
1309 return root;
1310}
1311#endif
1312
Arne Jansen20897f52011-09-13 12:44:20 +02001313struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1314 struct btrfs_fs_info *fs_info,
1315 u64 objectid)
1316{
1317 struct extent_buffer *leaf;
1318 struct btrfs_root *tree_root = fs_info->tree_root;
1319 struct btrfs_root *root;
1320 struct btrfs_key key;
1321 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001322 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001323
1324 root = btrfs_alloc_root(fs_info);
1325 if (!root)
1326 return ERR_PTR(-ENOMEM);
1327
David Sterba707e8a02014-06-04 19:22:26 +02001328 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1329 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001330 root->root_key.objectid = objectid;
1331 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1332 root->root_key.offset = 0;
1333
David Sterba4d75f8a2014-06-15 01:54:12 +02001334 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Arne Jansen20897f52011-09-13 12:44:20 +02001335 if (IS_ERR(leaf)) {
1336 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001337 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001338 goto fail;
1339 }
1340
Arne Jansen20897f52011-09-13 12:44:20 +02001341 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1342 btrfs_set_header_bytenr(leaf, leaf->start);
1343 btrfs_set_header_generation(leaf, trans->transid);
1344 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1345 btrfs_set_header_owner(leaf, objectid);
1346 root->node = leaf;
1347
Ross Kirk0a4e5582013-09-24 10:12:38 +01001348 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001349 BTRFS_FSID_SIZE);
1350 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001351 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001352 BTRFS_UUID_SIZE);
1353 btrfs_mark_buffer_dirty(leaf);
1354
1355 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001356 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001357
1358 root->root_item.flags = 0;
1359 root->root_item.byte_limit = 0;
1360 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1361 btrfs_set_root_generation(&root->root_item, trans->transid);
1362 btrfs_set_root_level(&root->root_item, 0);
1363 btrfs_set_root_refs(&root->root_item, 1);
1364 btrfs_set_root_used(&root->root_item, leaf->len);
1365 btrfs_set_root_last_snapshot(&root->root_item, 0);
1366 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001367 uuid_le_gen(&uuid);
1368 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001369 root->root_item.drop_level = 0;
1370
1371 key.objectid = objectid;
1372 key.type = BTRFS_ROOT_ITEM_KEY;
1373 key.offset = 0;
1374 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1375 if (ret)
1376 goto fail;
1377
1378 btrfs_tree_unlock(leaf);
1379
Arne Jansen20897f52011-09-13 12:44:20 +02001380 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001381
1382fail:
1383 if (leaf) {
1384 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001385 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001386 free_extent_buffer(leaf);
1387 }
1388 kfree(root);
1389
1390 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001391}
1392
Yan Zheng7237f182009-01-21 12:54:03 -05001393static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1394 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001395{
1396 struct btrfs_root *root;
1397 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001398 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001399
Al Viro6f07e422011-11-17 00:46:16 -05001400 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001401 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001402 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001403
David Sterba707e8a02014-06-04 19:22:26 +02001404 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1405 tree_root->stripesize, root, fs_info,
1406 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001407
1408 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1409 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1410 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001411
Yan Zheng7237f182009-01-21 12:54:03 -05001412 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001413 * DON'T set REF_COWS for log trees
1414 *
Yan Zheng7237f182009-01-21 12:54:03 -05001415 * log trees do not get reference counted because they go away
1416 * before a real commit is actually done. They do store pointers
1417 * to file data extents, and those reference counts still get
1418 * updated (along with back refs to the log tree).
1419 */
Chris Masone02119d2008-09-05 16:13:11 -04001420
David Sterba4d75f8a2014-06-15 01:54:12 +02001421 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1422 NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001423 if (IS_ERR(leaf)) {
1424 kfree(root);
1425 return ERR_CAST(leaf);
1426 }
Chris Masone02119d2008-09-05 16:13:11 -04001427
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001428 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1429 btrfs_set_header_bytenr(leaf, leaf->start);
1430 btrfs_set_header_generation(leaf, trans->transid);
1431 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1432 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001433 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001434
1435 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001436 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001437 btrfs_mark_buffer_dirty(root->node);
1438 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001439 return root;
1440}
1441
1442int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1443 struct btrfs_fs_info *fs_info)
1444{
1445 struct btrfs_root *log_root;
1446
1447 log_root = alloc_log_tree(trans, fs_info);
1448 if (IS_ERR(log_root))
1449 return PTR_ERR(log_root);
1450 WARN_ON(fs_info->log_root_tree);
1451 fs_info->log_root_tree = log_root;
1452 return 0;
1453}
1454
1455int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1456 struct btrfs_root *root)
1457{
1458 struct btrfs_root *log_root;
1459 struct btrfs_inode_item *inode_item;
1460
1461 log_root = alloc_log_tree(trans, root->fs_info);
1462 if (IS_ERR(log_root))
1463 return PTR_ERR(log_root);
1464
1465 log_root->last_trans = trans->transid;
1466 log_root->root_key.offset = root->root_key.objectid;
1467
1468 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001469 btrfs_set_stack_inode_generation(inode_item, 1);
1470 btrfs_set_stack_inode_size(inode_item, 3);
1471 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001472 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001473 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f182009-01-21 12:54:03 -05001474
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001475 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001476
1477 WARN_ON(root->log_root);
1478 root->log_root = log_root;
1479 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001480 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001481 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001482 return 0;
1483}
1484
Stefan Behrens35a36212013-08-14 18:12:25 +02001485static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1486 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001487{
1488 struct btrfs_root *root;
1489 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001490 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001491 u64 generation;
Miao Xiecb517ea2013-05-15 07:48:19 +00001492 int ret;
1493
1494 path = btrfs_alloc_path();
1495 if (!path)
1496 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001497
Al Viro6f07e422011-11-17 00:46:16 -05001498 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001499 if (!root) {
1500 ret = -ENOMEM;
1501 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001502 }
1503
David Sterba707e8a02014-06-04 19:22:26 +02001504 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1505 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001506
Miao Xiecb517ea2013-05-15 07:48:19 +00001507 ret = btrfs_find_root(tree_root, key, path,
1508 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001509 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001510 if (ret > 0)
1511 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001512 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001513 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001514
Yan Zheng84234f32008-10-29 14:49:05 -04001515 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001516 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
David Sterbace86cd52014-06-15 01:07:32 +02001517 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08001518 if (IS_ERR(root->node)) {
1519 ret = PTR_ERR(root->node);
Miao Xiecb517ea2013-05-15 07:48:19 +00001520 goto find_fail;
1521 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1522 ret = -EIO;
Liu Bo64c043d2015-05-25 17:30:15 +08001523 free_extent_buffer(root->node);
1524 goto find_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001525 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001526 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001527out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001528 btrfs_free_path(path);
1529 return root;
1530
Miao Xiecb517ea2013-05-15 07:48:19 +00001531find_fail:
1532 kfree(root);
1533alloc_fail:
1534 root = ERR_PTR(ret);
1535 goto out;
1536}
1537
1538struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1539 struct btrfs_key *location)
1540{
1541 struct btrfs_root *root;
1542
1543 root = btrfs_read_tree_root(tree_root, location);
1544 if (IS_ERR(root))
1545 return root;
1546
1547 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001548 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001549 btrfs_check_and_init_root_item(&root->root_item);
1550 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001551
Chris Mason5eda7b52007-06-22 14:16:25 -04001552 return root;
1553}
1554
Miao Xiecb517ea2013-05-15 07:48:19 +00001555int btrfs_init_fs_root(struct btrfs_root *root)
1556{
1557 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001558 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001559
1560 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1561 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1562 GFP_NOFS);
1563 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1564 ret = -ENOMEM;
1565 goto fail;
1566 }
1567
Miao Xie8257b2d2014-03-06 13:38:19 +08001568 writers = btrfs_alloc_subvolume_writers();
1569 if (IS_ERR(writers)) {
1570 ret = PTR_ERR(writers);
1571 goto fail;
1572 }
1573 root->subv_writers = writers;
1574
Miao Xiecb517ea2013-05-15 07:48:19 +00001575 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001576 spin_lock_init(&root->ino_cache_lock);
1577 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001578
1579 ret = get_anon_bdev(&root->anon_dev);
1580 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001581 goto free_writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001582 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001583
1584free_writers:
1585 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001586fail:
1587 kfree(root->free_ino_ctl);
1588 kfree(root->free_ino_pinned);
1589 return ret;
1590}
1591
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001592static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1593 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001594{
1595 struct btrfs_root *root;
1596
1597 spin_lock(&fs_info->fs_roots_radix_lock);
1598 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1599 (unsigned long)root_id);
1600 spin_unlock(&fs_info->fs_roots_radix_lock);
1601 return root;
1602}
1603
1604int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1605 struct btrfs_root *root)
1606{
1607 int ret;
1608
1609 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1610 if (ret)
1611 return ret;
1612
1613 spin_lock(&fs_info->fs_roots_radix_lock);
1614 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1615 (unsigned long)root->root_key.objectid,
1616 root);
1617 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001618 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001619 spin_unlock(&fs_info->fs_roots_radix_lock);
1620 radix_tree_preload_end();
1621
1622 return ret;
1623}
1624
Miao Xiec00869f2013-09-25 21:47:44 +08001625struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1626 struct btrfs_key *location,
1627 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001628{
1629 struct btrfs_root *root;
David Sterba381cf652015-01-02 18:45:16 +01001630 struct btrfs_path *path;
David Sterba1d4c08e2015-01-02 19:36:14 +01001631 struct btrfs_key key;
Chris Mason5eda7b52007-06-22 14:16:25 -04001632 int ret;
1633
Chris Masonedbd8d42007-12-21 16:27:24 -05001634 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1635 return fs_info->tree_root;
1636 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1637 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001638 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1639 return fs_info->chunk_root;
1640 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1641 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001642 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1643 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001644 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1645 return fs_info->quota_root ? fs_info->quota_root :
1646 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001647 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1648 return fs_info->uuid_root ? fs_info->uuid_root :
1649 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001650again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001651 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001652 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001653 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001654 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001655 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001656 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001657
Miao Xiecb517ea2013-05-15 07:48:19 +00001658 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001659 if (IS_ERR(root))
1660 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001661
Miao Xiec00869f2013-09-25 21:47:44 +08001662 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001663 ret = -ENOENT;
1664 goto fail;
1665 }
1666
Miao Xiecb517ea2013-05-15 07:48:19 +00001667 ret = btrfs_init_fs_root(root);
1668 if (ret)
1669 goto fail;
1670
David Sterba381cf652015-01-02 18:45:16 +01001671 path = btrfs_alloc_path();
1672 if (!path) {
1673 ret = -ENOMEM;
1674 goto fail;
1675 }
David Sterba1d4c08e2015-01-02 19:36:14 +01001676 key.objectid = BTRFS_ORPHAN_OBJECTID;
1677 key.type = BTRFS_ORPHAN_ITEM_KEY;
1678 key.offset = location->objectid;
1679
1680 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
David Sterba381cf652015-01-02 18:45:16 +01001681 btrfs_free_path(path);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001682 if (ret < 0)
1683 goto fail;
1684 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001685 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001686
Miao Xiecb517ea2013-05-15 07:48:19 +00001687 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001688 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001689 if (ret == -EEXIST) {
1690 free_fs_root(root);
1691 goto again;
1692 }
1693 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001694 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001695 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001696fail:
1697 free_fs_root(root);
1698 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001699}
1700
Chris Mason04160082008-03-26 10:28:07 -04001701static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1702{
1703 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1704 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001705 struct btrfs_device *device;
1706 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001707
Xiao Guangrong1f781602011-04-20 10:09:16 +00001708 rcu_read_lock();
1709 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001710 if (!device->bdev)
1711 continue;
Chris Mason04160082008-03-26 10:28:07 -04001712 bdi = blk_get_backing_dev_info(device->bdev);
Tejun Heoff9ea322014-09-08 08:03:56 +09001713 if (bdi_congested(bdi, bdi_bits)) {
Chris Mason04160082008-03-26 10:28:07 -04001714 ret = 1;
1715 break;
1716 }
1717 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001718 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001719 return ret;
1720}
1721
Chris Mason04160082008-03-26 10:28:07 -04001722static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1723{
Jens Axboead081f12009-06-12 14:43:40 +02001724 int err;
1725
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001726 err = bdi_setup_and_register(bdi, "btrfs");
Jens Axboead081f12009-06-12 14:43:40 +02001727 if (err)
1728 return err;
1729
Christoph Hellwigdf0ce262015-01-14 10:42:41 +01001730 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE;
Chris Mason04160082008-03-26 10:28:07 -04001731 bdi->congested_fn = btrfs_congested_fn;
1732 bdi->congested_data = info;
Chris Masonda2f0f72015-07-02 13:57:22 -07001733 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
Chris Mason04160082008-03-26 10:28:07 -04001734 return 0;
1735}
1736
Chris Mason8b712842008-06-11 16:50:36 -04001737/*
1738 * called by the kthread helper functions to finally call the bio end_io
1739 * functions. This is where read checksum verification actually happens
1740 */
1741static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001742{
Chris Masonce9adaa2008-04-09 16:28:12 -04001743 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001744 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001745
David Sterba97eb6b62014-07-30 00:55:42 +02001746 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001747 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001748
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001749 bio->bi_error = end_io_wq->error;
Chris Mason8b712842008-06-11 16:50:36 -04001750 bio->bi_private = end_io_wq->private;
1751 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001752 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001753 bio_endio(bio);
Chris Mason44b8bd72008-04-16 11:14:51 -04001754}
1755
Chris Masona74a4b92008-06-25 16:01:31 -04001756static int cleaner_kthread(void *arg)
1757{
1758 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001759 int again;
Filipe Mananada288d22015-06-13 06:55:31 +01001760 struct btrfs_trans_handle *trans;
Chris Masona74a4b92008-06-25 16:01:31 -04001761
1762 do {
Miao Xied0278242013-05-14 10:20:40 +00001763 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001764
Miao Xied0278242013-05-14 10:20:40 +00001765 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001766 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001767 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001768
Miao Xied0278242013-05-14 10:20:40 +00001769 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1770 goto sleep;
1771
Miao Xiedc7f3702013-05-14 10:20:42 +00001772 /*
1773 * Avoid the problem that we change the status of the fs
1774 * during the above check and trylock.
1775 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001776 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001777 mutex_unlock(&root->fs_info->cleaner_mutex);
1778 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001779 }
1780
Miao Xied0278242013-05-14 10:20:40 +00001781 btrfs_run_delayed_iputs(root);
1782 again = btrfs_clean_one_deleted_snapshot(root);
1783 mutex_unlock(&root->fs_info->cleaner_mutex);
1784
1785 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001786 * The defragger has dealt with the R/O remount and umount,
1787 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001788 */
1789 btrfs_run_defrag_inodes(root->fs_info);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01001790
1791 /*
1792 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1793 * with relocation (btrfs_relocate_chunk) and relocation
1794 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1795 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1796 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1797 * unused block groups.
1798 */
1799 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001800sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001801 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001802 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001803 if (!kthread_should_stop())
1804 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001805 __set_current_state(TASK_RUNNING);
1806 }
1807 } while (!kthread_should_stop());
Filipe Mananada288d22015-06-13 06:55:31 +01001808
1809 /*
1810 * Transaction kthread is stopped before us and wakes us up.
1811 * However we might have started a new transaction and COWed some
1812 * tree blocks when deleting unused block groups for example. So
1813 * make sure we commit the transaction we started to have a clean
1814 * shutdown when evicting the btree inode - if it has dirty pages
1815 * when we do the final iput() on it, eviction will trigger a
1816 * writeback for it which will fail with null pointer dereferences
1817 * since work queues and other resources were already released and
1818 * destroyed by the time the iput/eviction/writeback is made.
1819 */
1820 trans = btrfs_attach_transaction(root);
1821 if (IS_ERR(trans)) {
1822 if (PTR_ERR(trans) != -ENOENT)
1823 btrfs_err(root->fs_info,
1824 "cleaner transaction attach returned %ld",
1825 PTR_ERR(trans));
1826 } else {
1827 int ret;
1828
1829 ret = btrfs_commit_transaction(trans, root);
1830 if (ret)
1831 btrfs_err(root->fs_info,
1832 "cleaner open transaction commit returned %d",
1833 ret);
1834 }
1835
Chris Masona74a4b92008-06-25 16:01:31 -04001836 return 0;
1837}
1838
1839static int transaction_kthread(void *arg)
1840{
1841 struct btrfs_root *root = arg;
1842 struct btrfs_trans_handle *trans;
1843 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001844 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001845 unsigned long now;
1846 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001847 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001848
1849 do {
Jan Kara914b2002012-03-12 16:05:50 +01001850 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001851 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001852 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1853
Josef Bacika4abeea2011-04-11 17:25:13 -04001854 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001855 cur = root->fs_info->running_transaction;
1856 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001857 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001858 goto sleep;
1859 }
Yan Zheng31153d82008-07-28 15:32:19 -04001860
Chris Masona74a4b92008-06-25 16:01:31 -04001861 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001862 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001863 (now < cur->start_time ||
1864 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001865 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001866 delay = HZ * 5;
1867 goto sleep;
1868 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001869 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001870 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001871
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001872 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001873 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001874 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001875 if (PTR_ERR(trans) != -ENOENT)
1876 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001877 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001878 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001879 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001880 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001881 } else {
1882 btrfs_end_transaction(trans, root);
1883 }
Chris Masona74a4b92008-06-25 16:01:31 -04001884sleep:
1885 wake_up_process(root->fs_info->cleaner_kthread);
1886 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1887
Josef Bacik4e121c02013-09-27 16:32:39 -04001888 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1889 &root->fs_info->fs_state)))
1890 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001891 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001892 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001893 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001894 (!btrfs_transaction_blocked(root->fs_info) ||
1895 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001896 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001897 __set_current_state(TASK_RUNNING);
1898 }
1899 } while (!kthread_should_stop());
1900 return 0;
1901}
1902
Chris Masonaf31f5e2011-11-03 15:17:42 -04001903/*
1904 * this will find the highest generation in the array of
1905 * root backups. The index of the highest array is returned,
1906 * or -1 if we can't find anything.
1907 *
1908 * We check to make sure the array is valid by comparing the
1909 * generation of the latest root in the array with the generation
1910 * in the super block. If they don't match we pitch it.
1911 */
1912static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1913{
1914 u64 cur;
1915 int newest_index = -1;
1916 struct btrfs_root_backup *root_backup;
1917 int i;
1918
1919 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1920 root_backup = info->super_copy->super_roots + i;
1921 cur = btrfs_backup_tree_root_gen(root_backup);
1922 if (cur == newest_gen)
1923 newest_index = i;
1924 }
1925
1926 /* check to see if we actually wrapped around */
1927 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1928 root_backup = info->super_copy->super_roots;
1929 cur = btrfs_backup_tree_root_gen(root_backup);
1930 if (cur == newest_gen)
1931 newest_index = 0;
1932 }
1933 return newest_index;
1934}
1935
1936
1937/*
1938 * find the oldest backup so we know where to store new entries
1939 * in the backup array. This will set the backup_root_index
1940 * field in the fs_info struct
1941 */
1942static void find_oldest_super_backup(struct btrfs_fs_info *info,
1943 u64 newest_gen)
1944{
1945 int newest_index = -1;
1946
1947 newest_index = find_newest_super_backup(info, newest_gen);
1948 /* if there was garbage in there, just move along */
1949 if (newest_index == -1) {
1950 info->backup_root_index = 0;
1951 } else {
1952 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1953 }
1954}
1955
1956/*
1957 * copy all the root pointers into the super backup array.
1958 * this will bump the backup pointer by one when it is
1959 * done
1960 */
1961static void backup_super_roots(struct btrfs_fs_info *info)
1962{
1963 int next_backup;
1964 struct btrfs_root_backup *root_backup;
1965 int last_backup;
1966
1967 next_backup = info->backup_root_index;
1968 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1969 BTRFS_NUM_BACKUP_ROOTS;
1970
1971 /*
1972 * just overwrite the last backup if we're at the same generation
1973 * this happens only at umount
1974 */
1975 root_backup = info->super_for_commit->super_roots + last_backup;
1976 if (btrfs_backup_tree_root_gen(root_backup) ==
1977 btrfs_header_generation(info->tree_root->node))
1978 next_backup = last_backup;
1979
1980 root_backup = info->super_for_commit->super_roots + next_backup;
1981
1982 /*
1983 * make sure all of our padding and empty slots get zero filled
1984 * regardless of which ones we use today
1985 */
1986 memset(root_backup, 0, sizeof(*root_backup));
1987
1988 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1989
1990 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1991 btrfs_set_backup_tree_root_gen(root_backup,
1992 btrfs_header_generation(info->tree_root->node));
1993
1994 btrfs_set_backup_tree_root_level(root_backup,
1995 btrfs_header_level(info->tree_root->node));
1996
1997 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1998 btrfs_set_backup_chunk_root_gen(root_backup,
1999 btrfs_header_generation(info->chunk_root->node));
2000 btrfs_set_backup_chunk_root_level(root_backup,
2001 btrfs_header_level(info->chunk_root->node));
2002
2003 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2004 btrfs_set_backup_extent_root_gen(root_backup,
2005 btrfs_header_generation(info->extent_root->node));
2006 btrfs_set_backup_extent_root_level(root_backup,
2007 btrfs_header_level(info->extent_root->node));
2008
Chris Mason7c7e82a2011-11-06 18:50:56 -05002009 /*
2010 * we might commit during log recovery, which happens before we set
2011 * the fs_root. Make sure it is valid before we fill it in.
2012 */
2013 if (info->fs_root && info->fs_root->node) {
2014 btrfs_set_backup_fs_root(root_backup,
2015 info->fs_root->node->start);
2016 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002017 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002018 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002019 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002020 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002021
2022 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2023 btrfs_set_backup_dev_root_gen(root_backup,
2024 btrfs_header_generation(info->dev_root->node));
2025 btrfs_set_backup_dev_root_level(root_backup,
2026 btrfs_header_level(info->dev_root->node));
2027
2028 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2029 btrfs_set_backup_csum_root_gen(root_backup,
2030 btrfs_header_generation(info->csum_root->node));
2031 btrfs_set_backup_csum_root_level(root_backup,
2032 btrfs_header_level(info->csum_root->node));
2033
2034 btrfs_set_backup_total_bytes(root_backup,
2035 btrfs_super_total_bytes(info->super_copy));
2036 btrfs_set_backup_bytes_used(root_backup,
2037 btrfs_super_bytes_used(info->super_copy));
2038 btrfs_set_backup_num_devices(root_backup,
2039 btrfs_super_num_devices(info->super_copy));
2040
2041 /*
2042 * if we don't copy this out to the super_copy, it won't get remembered
2043 * for the next commit
2044 */
2045 memcpy(&info->super_copy->super_roots,
2046 &info->super_for_commit->super_roots,
2047 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2048}
2049
2050/*
2051 * this copies info out of the root backup array and back into
2052 * the in-memory super block. It is meant to help iterate through
2053 * the array, so you send it the number of backups you've already
2054 * tried and the last backup index you used.
2055 *
2056 * this returns -1 when it has tried all the backups
2057 */
2058static noinline int next_root_backup(struct btrfs_fs_info *info,
2059 struct btrfs_super_block *super,
2060 int *num_backups_tried, int *backup_index)
2061{
2062 struct btrfs_root_backup *root_backup;
2063 int newest = *backup_index;
2064
2065 if (*num_backups_tried == 0) {
2066 u64 gen = btrfs_super_generation(super);
2067
2068 newest = find_newest_super_backup(info, gen);
2069 if (newest == -1)
2070 return -1;
2071
2072 *backup_index = newest;
2073 *num_backups_tried = 1;
2074 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2075 /* we've tried all the backups, all done */
2076 return -1;
2077 } else {
2078 /* jump to the next oldest backup */
2079 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2080 BTRFS_NUM_BACKUP_ROOTS;
2081 *backup_index = newest;
2082 *num_backups_tried += 1;
2083 }
2084 root_backup = super->super_roots + newest;
2085
2086 btrfs_set_super_generation(super,
2087 btrfs_backup_tree_root_gen(root_backup));
2088 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2089 btrfs_set_super_root_level(super,
2090 btrfs_backup_tree_root_level(root_backup));
2091 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2092
2093 /*
2094 * fixme: the total bytes and num_devices need to match or we should
2095 * need a fsck
2096 */
2097 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2098 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2099 return 0;
2100}
2101
Liu Bo7abadb62013-03-17 02:10:31 +00002102/* helper to cleanup workers */
2103static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2104{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002105 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002106 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002107 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002108 btrfs_destroy_workqueue(fs_info->endio_workers);
2109 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2110 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002111 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002112 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002113 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2114 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2115 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002116 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002117 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002118 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002119 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002120 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002121 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002122 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002123}
2124
Rashika2e9f5952013-10-31 02:45:20 +05302125static void free_root_extent_buffers(struct btrfs_root *root)
2126{
2127 if (root) {
2128 free_extent_buffer(root->node);
2129 free_extent_buffer(root->commit_root);
2130 root->node = NULL;
2131 root->commit_root = NULL;
2132 }
2133}
2134
Chris Masonaf31f5e2011-11-03 15:17:42 -04002135/* helper to cleanup tree roots */
2136static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2137{
Rashika2e9f5952013-10-31 02:45:20 +05302138 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002139
Rashika2e9f5952013-10-31 02:45:20 +05302140 free_root_extent_buffers(info->dev_root);
2141 free_root_extent_buffers(info->extent_root);
2142 free_root_extent_buffers(info->csum_root);
2143 free_root_extent_buffers(info->quota_root);
2144 free_root_extent_buffers(info->uuid_root);
2145 if (chunk_root)
2146 free_root_extent_buffers(info->chunk_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002147}
2148
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002149void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002150{
2151 int ret;
2152 struct btrfs_root *gang[8];
2153 int i;
2154
2155 while (!list_empty(&fs_info->dead_roots)) {
2156 gang[0] = list_entry(fs_info->dead_roots.next,
2157 struct btrfs_root, root_list);
2158 list_del(&gang[0]->root_list);
2159
Miao Xie27cdeb72014-04-02 19:51:05 +08002160 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002161 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002162 } else {
2163 free_extent_buffer(gang[0]->node);
2164 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002165 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002166 }
2167 }
2168
2169 while (1) {
2170 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2171 (void **)gang, 0,
2172 ARRAY_SIZE(gang));
2173 if (!ret)
2174 break;
2175 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002176 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002177 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002178
2179 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2180 btrfs_free_log_root_tree(NULL, fs_info);
2181 btrfs_destroy_pinned_extent(fs_info->tree_root,
2182 fs_info->pinned_extents);
2183 }
Josef Bacik171f6532013-04-24 16:35:41 -04002184}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002185
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002186static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2187{
2188 mutex_init(&fs_info->scrub_lock);
2189 atomic_set(&fs_info->scrubs_running, 0);
2190 atomic_set(&fs_info->scrub_pause_req, 0);
2191 atomic_set(&fs_info->scrubs_paused, 0);
2192 atomic_set(&fs_info->scrub_cancel_req, 0);
2193 init_waitqueue_head(&fs_info->scrub_pause_wait);
2194 fs_info->scrub_workers_refcnt = 0;
2195}
2196
Eric Sandeen779a65a2014-08-01 18:12:39 -05002197static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2198{
2199 spin_lock_init(&fs_info->balance_lock);
2200 mutex_init(&fs_info->balance_mutex);
2201 atomic_set(&fs_info->balance_running, 0);
2202 atomic_set(&fs_info->balance_pause_req, 0);
2203 atomic_set(&fs_info->balance_cancel_req, 0);
2204 fs_info->balance_ctl = NULL;
2205 init_waitqueue_head(&fs_info->balance_wait_q);
2206}
2207
Eric Sandeenf37938e2014-08-01 18:12:40 -05002208static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2209 struct btrfs_root *tree_root)
2210{
2211 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2212 set_nlink(fs_info->btree_inode, 1);
2213 /*
2214 * we set the i_size on the btree inode to the max possible int.
2215 * the real end of the address space is determined by all of
2216 * the devices in the system
2217 */
2218 fs_info->btree_inode->i_size = OFFSET_MAX;
2219 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Eric Sandeenf37938e2014-08-01 18:12:40 -05002220
2221 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2222 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2223 fs_info->btree_inode->i_mapping);
2224 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2225 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2226
2227 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2228
2229 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2230 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2231 sizeof(struct btrfs_key));
2232 set_bit(BTRFS_INODE_DUMMY,
2233 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2234 btrfs_insert_inode_hash(fs_info->btree_inode);
2235}
2236
Eric Sandeenad618362014-08-01 18:12:41 -05002237static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2238{
2239 fs_info->dev_replace.lock_owner = 0;
2240 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2241 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2242 mutex_init(&fs_info->dev_replace.lock_management_lock);
2243 mutex_init(&fs_info->dev_replace.lock);
2244 init_waitqueue_head(&fs_info->replace_wait);
2245}
2246
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002247static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2248{
2249 spin_lock_init(&fs_info->qgroup_lock);
2250 mutex_init(&fs_info->qgroup_ioctl_lock);
2251 fs_info->qgroup_tree = RB_ROOT;
2252 fs_info->qgroup_op_tree = RB_ROOT;
2253 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2254 fs_info->qgroup_seq = 1;
2255 fs_info->quota_enabled = 0;
2256 fs_info->pending_quota_state = 0;
2257 fs_info->qgroup_ulist = NULL;
2258 mutex_init(&fs_info->qgroup_rescan_lock);
2259}
2260
Eric Sandeen2a458192015-02-16 16:29:26 +01002261static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2262 struct btrfs_fs_devices *fs_devices)
2263{
2264 int max_active = fs_info->thread_pool_size;
David Sterba6f011052015-02-16 18:34:01 +01002265 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Eric Sandeen2a458192015-02-16 16:29:26 +01002266
2267 fs_info->workers =
2268 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2269 max_active, 16);
2270
2271 fs_info->delalloc_workers =
2272 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
2273
2274 fs_info->flush_workers =
2275 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
2276
2277 fs_info->caching_workers =
2278 btrfs_alloc_workqueue("cache", flags, max_active, 0);
2279
2280 /*
2281 * a higher idle thresh on the submit workers makes it much more
2282 * likely that bios will be send down in a sane order to the
2283 * devices
2284 */
2285 fs_info->submit_workers =
2286 btrfs_alloc_workqueue("submit", flags,
2287 min_t(u64, fs_devices->num_devices,
2288 max_active), 64);
2289
2290 fs_info->fixup_workers =
2291 btrfs_alloc_workqueue("fixup", flags, 1, 0);
2292
2293 /*
2294 * endios are largely parallel and should have a very
2295 * low idle thresh
2296 */
2297 fs_info->endio_workers =
2298 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2299 fs_info->endio_meta_workers =
2300 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2301 fs_info->endio_meta_write_workers =
2302 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2303 fs_info->endio_raid56_workers =
2304 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
2305 fs_info->endio_repair_workers =
2306 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
2307 fs_info->rmw_workers =
2308 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
2309 fs_info->endio_write_workers =
2310 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2311 fs_info->endio_freespace_worker =
2312 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
2313 fs_info->delayed_workers =
2314 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
2315 fs_info->readahead_workers =
2316 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
2317 fs_info->qgroup_rescan_workers =
2318 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
2319 fs_info->extent_workers =
2320 btrfs_alloc_workqueue("extent-refs", flags,
2321 min_t(u64, fs_devices->num_devices,
2322 max_active), 8);
2323
2324 if (!(fs_info->workers && fs_info->delalloc_workers &&
2325 fs_info->submit_workers && fs_info->flush_workers &&
2326 fs_info->endio_workers && fs_info->endio_meta_workers &&
2327 fs_info->endio_meta_write_workers &&
2328 fs_info->endio_repair_workers &&
2329 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2330 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2331 fs_info->caching_workers && fs_info->readahead_workers &&
2332 fs_info->fixup_workers && fs_info->delayed_workers &&
2333 fs_info->extent_workers &&
2334 fs_info->qgroup_rescan_workers)) {
2335 return -ENOMEM;
2336 }
2337
2338 return 0;
2339}
2340
Eric Sandeen63443bf2014-08-01 18:12:46 -05002341static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2342 struct btrfs_fs_devices *fs_devices)
2343{
2344 int ret;
2345 struct btrfs_root *tree_root = fs_info->tree_root;
2346 struct btrfs_root *log_tree_root;
2347 struct btrfs_super_block *disk_super = fs_info->super_copy;
2348 u64 bytenr = btrfs_super_log_root(disk_super);
2349
2350 if (fs_devices->rw_devices == 0) {
2351 printk(KERN_WARNING "BTRFS: log replay required "
2352 "on RO media\n");
2353 return -EIO;
2354 }
2355
2356 log_tree_root = btrfs_alloc_root(fs_info);
2357 if (!log_tree_root)
2358 return -ENOMEM;
2359
2360 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2361 tree_root->stripesize, log_tree_root, fs_info,
2362 BTRFS_TREE_LOG_OBJECTID);
2363
2364 log_tree_root->node = read_tree_block(tree_root, bytenr,
2365 fs_info->generation + 1);
Liu Bo64c043d2015-05-25 17:30:15 +08002366 if (IS_ERR(log_tree_root->node)) {
2367 printk(KERN_ERR "BTRFS: failed to read log tree\n");
Liu Bo0eeff232015-06-11 14:16:44 +08002368 ret = PTR_ERR(log_tree_root->node);
Liu Bo64c043d2015-05-25 17:30:15 +08002369 kfree(log_tree_root);
Liu Bo0eeff232015-06-11 14:16:44 +08002370 return ret;
Liu Bo64c043d2015-05-25 17:30:15 +08002371 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05002372 printk(KERN_ERR "BTRFS: failed to read log tree\n");
2373 free_extent_buffer(log_tree_root->node);
2374 kfree(log_tree_root);
2375 return -EIO;
2376 }
2377 /* returns with log_tree_root freed on success */
2378 ret = btrfs_recover_log_trees(log_tree_root);
2379 if (ret) {
2380 btrfs_error(tree_root->fs_info, ret,
2381 "Failed to recover log tree");
2382 free_extent_buffer(log_tree_root->node);
2383 kfree(log_tree_root);
2384 return ret;
2385 }
2386
2387 if (fs_info->sb->s_flags & MS_RDONLY) {
2388 ret = btrfs_commit_super(tree_root);
2389 if (ret)
2390 return ret;
2391 }
2392
2393 return 0;
2394}
2395
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002396static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2397 struct btrfs_root *tree_root)
2398{
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002399 struct btrfs_root *root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002400 struct btrfs_key location;
2401 int ret;
2402
2403 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2404 location.type = BTRFS_ROOT_ITEM_KEY;
2405 location.offset = 0;
2406
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002407 root = btrfs_read_tree_root(tree_root, &location);
2408 if (IS_ERR(root))
2409 return PTR_ERR(root);
2410 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2411 fs_info->extent_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002412
2413 location.objectid = BTRFS_DEV_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002414 root = btrfs_read_tree_root(tree_root, &location);
2415 if (IS_ERR(root))
2416 return PTR_ERR(root);
2417 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2418 fs_info->dev_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002419 btrfs_init_devices_late(fs_info);
2420
2421 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002422 root = btrfs_read_tree_root(tree_root, &location);
2423 if (IS_ERR(root))
2424 return PTR_ERR(root);
2425 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2426 fs_info->csum_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002427
2428 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002429 root = btrfs_read_tree_root(tree_root, &location);
2430 if (!IS_ERR(root)) {
2431 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002432 fs_info->quota_enabled = 1;
2433 fs_info->pending_quota_state = 1;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002434 fs_info->quota_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002435 }
2436
2437 location.objectid = BTRFS_UUID_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002438 root = btrfs_read_tree_root(tree_root, &location);
2439 if (IS_ERR(root)) {
2440 ret = PTR_ERR(root);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002441 if (ret != -ENOENT)
2442 return ret;
2443 } else {
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002444 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2445 fs_info->uuid_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002446 }
2447
2448 return 0;
2449}
2450
Al Viroad2b2c82011-11-17 01:10:02 -05002451int open_ctree(struct super_block *sb,
2452 struct btrfs_fs_devices *fs_devices,
2453 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002454{
Chris Masondb945352007-10-15 16:15:53 -04002455 u32 sectorsize;
2456 u32 nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002457 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002458 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002459 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002460 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002461 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002462 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002463 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002464 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002465 struct btrfs_root *chunk_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002466 int ret;
Yane58ca022008-04-01 11:21:34 -04002467 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002468 int num_backups_tried = 0;
2469 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002470 int max_active;
Chris Mason4543df72008-06-11 21:47:56 -04002471
Al Virof84a8bd2011-11-17 15:57:57 -05002472 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002473 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002474 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002475 err = -ENOMEM;
2476 goto fail;
2477 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002478
2479 ret = init_srcu_struct(&fs_info->subvol_srcu);
2480 if (ret) {
2481 err = ret;
2482 goto fail;
2483 }
2484
2485 ret = setup_bdi(fs_info, &fs_info->bdi);
2486 if (ret) {
2487 err = ret;
2488 goto fail_srcu;
2489 }
2490
Tejun Heo908c7f12014-09-08 09:51:29 +09002491 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
Miao Xiee2d84522013-01-29 10:09:20 +00002492 if (ret) {
2493 err = ret;
2494 goto fail_bdi;
2495 }
2496 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2497 (1 + ilog2(nr_cpu_ids));
2498
Tejun Heo908c7f12014-09-08 09:51:29 +09002499 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
Miao Xie963d6782013-01-29 10:10:51 +00002500 if (ret) {
2501 err = ret;
2502 goto fail_dirty_metadata_bytes;
2503 }
2504
Tejun Heo908c7f12014-09-08 09:51:29 +09002505 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
Miao Xiec404e0d2014-01-30 16:46:55 +08002506 if (ret) {
2507 err = ret;
2508 goto fail_delalloc_bytes;
2509 }
2510
Yan, Zheng76dda932009-09-21 16:00:26 -04002511 fs_info->btree_inode = new_inode(sb);
2512 if (!fs_info->btree_inode) {
2513 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002514 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002515 }
2516
Chris Masona6591712011-07-19 12:04:14 -04002517 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002518
Yan, Zheng76dda932009-09-21 16:00:26 -04002519 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002520 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002521 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002522 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002523 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002524 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002525 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002526 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002527 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002528 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002529 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002530 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002531 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002532 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002533 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002534 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002535 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002536 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002537 rwlock_init(&fs_info->tree_mod_log_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002538 mutex_init(&fs_info->unused_bg_unpin_mutex);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01002539 mutex_init(&fs_info->delete_unused_bgs_mutex);
Chris Mason75857172011-06-13 20:00:16 -04002540 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002541 mutex_init(&fs_info->delalloc_root_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002542 seqlock_init(&fs_info->profiles_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002543 init_rwsem(&fs_info->delayed_iput_sem);
Chris Mason19c00dd2007-10-15 16:19:22 -04002544
Chris Mason0b86a832008-03-24 15:01:56 -04002545 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002546 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002547 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002548 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002549 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002550 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2551 BTRFS_BLOCK_RSV_GLOBAL);
2552 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2553 BTRFS_BLOCK_RSV_DELALLOC);
2554 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2555 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2556 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2557 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2558 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002559 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002560 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002561 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002562 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002563 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002564 atomic_set(&fs_info->qgroup_op_seq, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002565 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002566 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002567 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002568 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002569 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002570 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002571 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002572 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
David Sterbaf8c269d2015-01-16 17:21:12 +01002573 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
Arne Jansen90519d62011-05-23 14:30:00 +02002574 /* readahead state */
2575 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2576 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002577
Chris Masonb34b0862008-12-19 15:43:22 -05002578 fs_info->thread_pool_size = min_t(unsigned long,
2579 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002580
Miao Xie199c2a92013-05-15 07:48:23 +00002581 INIT_LIST_HEAD(&fs_info->ordered_roots);
2582 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002583 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2584 GFP_NOFS);
2585 if (!fs_info->delayed_root) {
2586 err = -ENOMEM;
2587 goto fail_iput;
2588 }
2589 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002590
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002591 btrfs_init_scrub(fs_info);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002592#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2593 fs_info->check_integrity_print_mask = 0;
2594#endif
Eric Sandeen779a65a2014-08-01 18:12:39 -05002595 btrfs_init_balance(fs_info);
Miao Xie21c7e752014-05-13 17:29:04 -07002596 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Arne Jansena2de7332011-03-08 14:14:00 +01002597
Chris Masona061fc82008-05-07 11:43:44 -04002598 sb->s_blocksize = 4096;
2599 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002600 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04002601
Eric Sandeenf37938e2014-08-01 18:12:40 -05002602 btrfs_init_btree_inode(fs_info, tree_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002603
Chris Masone02119d2008-09-05 16:13:11 -04002604 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002605 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002606 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002607
Yan Zheng11833d62009-09-11 16:11:19 -04002608 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002609 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002610 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002611 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002612 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002613 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002614
Chris Masond98237b2007-03-28 13:57:48 -04002615
Chris Mason5a3f23d2009-03-31 13:27:11 -04002616 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002617 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002618 mutex_init(&fs_info->chunk_mutex);
2619 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002620 mutex_init(&fs_info->cleaner_mutex);
2621 mutex_init(&fs_info->volume_mutex);
Chris Mason1bbc6212015-04-06 12:46:08 -07002622 mutex_init(&fs_info->ro_block_group_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002623 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002624 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002625 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002626 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002627
Eric Sandeenad618362014-08-01 18:12:41 -05002628 btrfs_init_dev_replace_locks(fs_info);
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002629 btrfs_init_qgroup(fs_info);
Arne Jansen416ac512011-09-13 12:56:09 +02002630
Chris Masonfa9c0d792009-04-03 09:47:43 -04002631 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2632 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2633
Chris Mason7d9eb122008-07-08 14:19:17 -04002634 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002635 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002636 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002637 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002638
Filipe Manana04216822014-11-27 21:14:15 +00002639 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2640
David Woodhouse53b381b2013-01-29 18:40:14 -05002641 ret = btrfs_alloc_stripe_hash_table(fs_info);
2642 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002643 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002644 goto fail_alloc;
2645 }
2646
David Sterba707e8a02014-06-04 19:22:26 +02002647 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002648 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002649
Chris Mason3c4bb262012-03-27 18:56:56 -04002650 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002651
2652 /*
2653 * Read super block and check the signature bytes only
2654 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002655 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002656 if (!bh) {
2657 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002658 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002659 }
Chris Mason39279cc2007-06-12 06:35:45 -04002660
David Sterba1104a882013-03-06 15:57:46 +01002661 /*
2662 * We want to check superblock checksum, the type is stored inside.
2663 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2664 */
2665 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002666 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002667 err = -EINVAL;
2668 goto fail_alloc;
2669 }
2670
2671 /*
2672 * super_copy is zeroed at allocation time and we never touch the
2673 * following bytes up to INFO_SIZE, the checksum is calculated from
2674 * the whole block of INFO_SIZE
2675 */
David Sterba6c417612011-04-13 15:41:04 +02002676 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2677 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2678 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002679 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002680
David Sterba6c417612011-04-13 15:41:04 +02002681 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002682
David Sterba1104a882013-03-06 15:57:46 +01002683 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2684 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002685 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002686 err = -EINVAL;
2687 goto fail_alloc;
2688 }
2689
David Sterba6c417612011-04-13 15:41:04 +02002690 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002691 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002692 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002693
liuboacce9522011-01-06 19:30:25 +08002694 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002695 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2696 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002697
Liu Bo75e7cb72011-03-22 10:12:20 +00002698 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002699 * run through our array of backup supers and setup
2700 * our ring pointer to the oldest one
2701 */
2702 generation = btrfs_super_generation(disk_super);
2703 find_oldest_super_backup(fs_info, generation);
2704
2705 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002706 * In the long term, we'll store the compression type in the super
2707 * block, and it'll be used for per file compression control.
2708 */
2709 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2710
Yan Zheng2b820322008-11-17 21:11:30 -05002711 ret = btrfs_parse_options(tree_root, options);
2712 if (ret) {
2713 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002714 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002715 }
Chris Masondfe25022008-05-13 13:46:40 -04002716
Josef Bacikf2b636e2008-12-02 06:36:08 -05002717 features = btrfs_super_incompat_flags(disk_super) &
2718 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2719 if (features) {
2720 printk(KERN_ERR "BTRFS: couldn't mount because of "
2721 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002722 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002723 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002724 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002725 }
2726
David Sterba707e8a02014-06-04 19:22:26 +02002727 /*
2728 * Leafsize and nodesize were always equal, this is only a sanity check.
2729 */
2730 if (le32_to_cpu(disk_super->__unused_leafsize) !=
Chris Mason727011e2010-08-06 13:21:20 -04002731 btrfs_super_nodesize(disk_super)) {
2732 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2733 "blocksizes don't match. node %d leaf %d\n",
2734 btrfs_super_nodesize(disk_super),
David Sterba707e8a02014-06-04 19:22:26 +02002735 le32_to_cpu(disk_super->__unused_leafsize));
Chris Mason727011e2010-08-06 13:21:20 -04002736 err = -EINVAL;
2737 goto fail_alloc;
2738 }
David Sterba707e8a02014-06-04 19:22:26 +02002739 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002740 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2741 "blocksize (%d) was too large\n",
David Sterba707e8a02014-06-04 19:22:26 +02002742 btrfs_super_nodesize(disk_super));
Chris Mason727011e2010-08-06 13:21:20 -04002743 err = -EINVAL;
2744 goto fail_alloc;
2745 }
2746
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002747 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002748 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002749 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002750 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002751
Josef Bacik3173a182013-03-07 14:22:04 -05002752 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
David Sterba5efa0492014-12-19 18:38:47 +01002753 printk(KERN_INFO "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002754
Chris Mason727011e2010-08-06 13:21:20 -04002755 /*
2756 * flag our filesystem as having big metadata blocks if
2757 * they are bigger than the page size
2758 */
David Sterba707e8a02014-06-04 19:22:26 +02002759 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002760 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002761 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002762 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2763 }
2764
Chris Masonbc3f1162012-03-29 17:02:47 -04002765 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002766 sectorsize = btrfs_super_sectorsize(disk_super);
2767 stripesize = btrfs_super_stripesize(disk_super);
David Sterba707e8a02014-06-04 19:22:26 +02002768 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002769 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002770
2771 /*
2772 * mixed block groups end up with duplicate but slightly offset
2773 * extent buffers for the same range. It leads to corruptions
2774 */
2775 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002776 (sectorsize != nodesize)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002777 printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002778 "are not allowed for mixed block groups on %s\n",
2779 sb->s_id);
2780 goto fail_alloc;
2781 }
2782
Miao Xieceda0862013-04-11 10:30:16 +00002783 /*
2784 * Needn't use the lock because there is no other task which will
2785 * update the flag.
2786 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002787 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002788
Josef Bacikf2b636e2008-12-02 06:36:08 -05002789 features = btrfs_super_compat_ro_flags(disk_super) &
2790 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2791 if (!(sb->s_flags & MS_RDONLY) && features) {
2792 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2793 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002794 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002795 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002796 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002797 }
Chris Mason61d92c32009-10-02 19:11:56 -04002798
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002799 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002800
Eric Sandeen2a458192015-02-16 16:29:26 +01002801 ret = btrfs_init_workqueues(fs_info, fs_devices);
2802 if (ret) {
2803 err = ret;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002804 goto fail_sb_buffer;
2805 }
Chris Mason4543df72008-06-11 21:47:56 -04002806
Chris Mason4575c9c2008-04-18 16:13:31 -04002807 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002808 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2809 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002810
Chris Masondb945352007-10-15 16:15:53 -04002811 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002812 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002813 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002814
2815 sb->s_blocksize = sectorsize;
2816 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002817
Qu Wenruo3cae2102013-07-16 11:19:18 +08002818 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002819 printk(KERN_ERR "BTRFS: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002820 goto fail_sb_buffer;
2821 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002822
Liu Bo8d082fb2012-04-03 09:56:53 +08002823 if (sectorsize != PAGE_SIZE) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002824 printk(KERN_ERR "BTRFS: incompatible sector size (%lu) "
Liu Bo8d082fb2012-04-03 09:56:53 +08002825 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002826 goto fail_sb_buffer;
2827 }
2828
Chris Mason925baed2008-06-25 16:01:30 -04002829 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002830 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002831 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002832 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002833 printk(KERN_ERR "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002834 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002835 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002836 }
Chris Mason0b86a832008-03-24 15:01:56 -04002837
Yan Zheng84234f32008-10-29 14:49:05 -04002838 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002839
David Sterba707e8a02014-06-04 19:22:26 +02002840 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2841 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002842
2843 chunk_root->node = read_tree_block(chunk_root,
2844 btrfs_super_chunk_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002845 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002846 if (IS_ERR(chunk_root->node) ||
2847 !extent_buffer_uptodate(chunk_root->node)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002848 printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002849 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302850 if (!IS_ERR(chunk_root->node))
2851 free_extent_buffer(chunk_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002852 chunk_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002853 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002854 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002855 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2856 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002857
Chris Masone17cade2008-04-15 15:41:47 -04002858 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002859 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002860
Chris Mason0b86a832008-03-24 15:01:56 -04002861 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002862 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002863 printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002864 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002865 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002866 }
Chris Mason0b86a832008-03-24 15:01:56 -04002867
Stefan Behrens8dabb742012-11-06 13:15:27 +01002868 /*
2869 * keep the device that is marked to be the target device for the
2870 * dev_replace procedure
2871 */
Eric Sandeen9eaed212014-08-01 18:12:35 -05002872 btrfs_close_extra_devices(fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002873
Chris Masona6b0d5c2012-02-20 20:53:43 -05002874 if (!fs_devices->latest_bdev) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002875 printk(KERN_ERR "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002876 sb->s_id);
2877 goto fail_tree_roots;
2878 }
2879
Chris Masonaf31f5e2011-11-03 15:17:42 -04002880retry_root_backup:
Yan Zheng84234f32008-10-29 14:49:05 -04002881 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002882
Chris Masone20d96d2007-03-22 12:13:20 -04002883 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002884 btrfs_super_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002885 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002886 if (IS_ERR(tree_root->node) ||
2887 !extent_buffer_uptodate(tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002888 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002889 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302890 if (!IS_ERR(tree_root->node))
2891 free_extent_buffer(tree_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002892 tree_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002893 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002894 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002895
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002896 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2897 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002898 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002899
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002900 ret = btrfs_read_roots(fs_info, tree_root);
2901 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002902 goto recovery_tree_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002903
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002904 fs_info->generation = generation;
2905 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002906
Ilya Dryomov68310a52012-06-22 12:24:12 -06002907 ret = btrfs_recover_balance(fs_info);
2908 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002909 printk(KERN_ERR "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002910 goto fail_block_groups;
2911 }
2912
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002913 ret = btrfs_init_dev_stats(fs_info);
2914 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002915 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002916 ret);
2917 goto fail_block_groups;
2918 }
2919
Stefan Behrens8dabb742012-11-06 13:15:27 +01002920 ret = btrfs_init_dev_replace(fs_info);
2921 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002922 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002923 goto fail_block_groups;
2924 }
2925
Eric Sandeen9eaed212014-08-01 18:12:35 -05002926 btrfs_close_extra_devices(fs_devices, 1);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002927
Anand Jainb7c35e82015-03-10 06:38:38 +08002928 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
2929 if (ret) {
2930 pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret);
2931 goto fail_block_groups;
2932 }
2933
2934 ret = btrfs_sysfs_add_device(fs_devices);
2935 if (ret) {
2936 pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret);
2937 goto fail_fsdev_sysfs;
2938 }
2939
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002940 ret = btrfs_sysfs_add_one(fs_info);
2941 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002942 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08002943 goto fail_fsdev_sysfs;
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002944 }
2945
liuboc59021f2011-03-07 02:13:14 +00002946 ret = btrfs_init_space_info(fs_info);
2947 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002948 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002949 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002950 }
2951
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002952 ret = btrfs_read_block_groups(fs_info->extent_root);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002953 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002954 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002955 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002956 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002957 fs_info->num_tolerated_disk_barrier_failures =
2958 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002959 if (fs_info->fs_devices->missing_devices >
2960 fs_info->num_tolerated_disk_barrier_failures &&
2961 !(sb->s_flags & MS_RDONLY)) {
Zhao Lei78fa1772015-07-20 17:54:50 +08002962 pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n",
2963 fs_info->fs_devices->missing_devices,
2964 fs_info->num_tolerated_disk_barrier_failures);
Anand Jain2365dd32014-01-22 11:15:51 +08002965 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002966 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002967
Chris Masona74a4b92008-06-25 16:01:31 -04002968 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2969 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002970 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08002971 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04002972
2973 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2974 tree_root,
2975 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002976 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002977 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002978
Chris Masonc2898112009-06-10 09:51:32 -04002979 if (!btrfs_test_opt(tree_root, SSD) &&
2980 !btrfs_test_opt(tree_root, NOSSD) &&
2981 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002982 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04002983 "mode\n");
2984 btrfs_set_opt(fs_info->mount_opt, SSD);
2985 }
2986
David Sterba572d9ab2014-02-05 15:26:17 +01002987 /*
2988 * Mount does not set all options immediatelly, we can do it now and do
2989 * not have to wait for transaction commit
2990 */
2991 btrfs_apply_pending_changes(fs_info);
Qu Wenruo3818aea2014-01-13 13:36:06 +08002992
Stefan Behrens21adbd52011-11-09 13:44:05 +01002993#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2994 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2995 ret = btrfsic_mount(tree_root, fs_devices,
2996 btrfs_test_opt(tree_root,
2997 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2998 1 : 0,
2999 fs_info->check_integrity_print_mask);
3000 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003001 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01003002 " integrity check module %s\n", sb->s_id);
3003 }
3004#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02003005 ret = btrfs_read_qgroup_config(fs_info);
3006 if (ret)
3007 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01003008
liuboacce9522011-01-06 19:30:25 +08003009 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06003010 if (btrfs_super_log_root(disk_super) != 0) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003011 ret = btrfs_replay_log(fs_info, fs_devices);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003012 if (ret) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003013 err = ret;
Wang Shilong28c16cb2014-04-23 19:33:35 +08003014 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003015 }
Chris Masone02119d2008-09-05 16:13:11 -04003016 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003017
Yan, Zheng76dda932009-09-21 16:00:26 -04003018 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003019 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003020 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04003021
Chris Mason7c2ca462008-11-19 15:13:35 -05003022 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04003023 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06003024 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003025 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04003026
Wang Shilong5f316482014-06-26 11:08:16 +08003027 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003028 ret = btrfs_recover_relocation(tree_root);
Wang Shilong5f316482014-06-26 11:08:16 +08003029 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00003030 if (ret < 0) {
3031 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05003032 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00003033 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02003034 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00003035 }
Chris Mason7c2ca462008-11-19 15:13:35 -05003036 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003037
Chris Mason3de45862008-11-17 21:02:50 -05003038 location.objectid = BTRFS_FS_TREE_OBJECTID;
3039 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00003040 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05003041
Chris Mason3de45862008-11-17 21:02:50 -05003042 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003043 if (IS_ERR(fs_info->fs_root)) {
3044 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003045 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003046 }
Chris Masonc2898112009-06-10 09:51:32 -04003047
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003048 if (sb->s_flags & MS_RDONLY)
3049 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02003050
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003051 down_read(&fs_info->cleanup_work_sem);
3052 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3053 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00003054 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003055 close_ctree(tree_root);
3056 return ret;
3057 }
3058 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02003059
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003060 ret = btrfs_resume_balance_async(fs_info);
3061 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003062 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003063 close_ctree(tree_root);
3064 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00003065 }
3066
Stefan Behrens8dabb742012-11-06 13:15:27 +01003067 ret = btrfs_resume_dev_replace_async(fs_info);
3068 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003069 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01003070 close_ctree(tree_root);
3071 return ret;
3072 }
3073
Jan Schmidtb382a322013-05-28 15:47:24 +00003074 btrfs_qgroup_rescan_resume(fs_info);
3075
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003076 if (!fs_info->uuid_root) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003077 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003078 ret = btrfs_create_uuid_tree(fs_info);
3079 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003080 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003081 ret);
3082 close_ctree(tree_root);
3083 return ret;
3084 }
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003085 } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) ||
3086 fs_info->generation !=
3087 btrfs_super_uuid_tree_generation(disk_super)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003088 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02003089 ret = btrfs_check_uuid_tree(fs_info);
3090 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003091 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02003092 ret);
3093 close_ctree(tree_root);
3094 return ret;
3095 }
3096 } else {
3097 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003098 }
3099
Josef Bacik47ab2a62014-09-18 11:20:02 -04003100 fs_info->open = 1;
3101
Al Viroad2b2c82011-11-17 01:10:02 -05003102 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04003103
Arne Jansenbcef60f2011-09-13 15:23:30 +02003104fail_qgroup:
3105 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05003106fail_trans_kthread:
3107 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04003108 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003109 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04003110fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04003111 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05003112
3113 /*
3114 * make sure we're done with the btree inode before we stop our
3115 * kthreads
3116 */
3117 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05003118
Anand Jain2365dd32014-01-22 11:15:51 +08003119fail_sysfs:
3120 btrfs_sysfs_remove_one(fs_info);
3121
Anand Jainb7c35e82015-03-10 06:38:38 +08003122fail_fsdev_sysfs:
3123 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3124
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003125fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003126 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003127 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003128
3129fail_tree_roots:
3130 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003131 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003132
Chris Mason39279cc2007-06-12 06:35:45 -04003133fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003134 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003135fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003136fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003137 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3138
Chris Mason4543df72008-06-11 21:47:56 -04003139 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003140fail_bio_counter:
3141 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003142fail_delalloc_bytes:
3143 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003144fail_dirty_metadata_bytes:
3145 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003146fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003147 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003148fail_srcu:
3149 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003150fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003151 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003152 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003153 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003154
3155recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003156 if (!btrfs_test_opt(tree_root, RECOVERY))
3157 goto fail_tree_roots;
3158
3159 free_root_pointers(fs_info, 0);
3160
3161 /* don't use the log in recovery mode, it won't be valid */
3162 btrfs_set_super_log_root(disk_super, 0);
3163
3164 /* we can't trust the free space cache either */
3165 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3166
3167 ret = next_root_backup(fs_info, fs_info->super_copy,
3168 &num_backups_tried, &backup_index);
3169 if (ret == -1)
3170 goto fail_block_groups;
3171 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003172}
3173
Chris Masonf2984462008-04-10 16:19:33 -04003174static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3175{
Chris Masonf2984462008-04-10 16:19:33 -04003176 if (uptodate) {
3177 set_buffer_uptodate(bh);
3178 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003179 struct btrfs_device *device = (struct btrfs_device *)
3180 bh->b_private;
3181
Frank Holtonefe120a2013-12-20 11:37:06 -05003182 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
Josef Bacik606686e2012-06-04 14:03:51 -04003183 "I/O error on %s\n",
3184 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003185 /* note, we dont' set_buffer_write_io_error because we have
3186 * our own ways of dealing with the IO errors
3187 */
Chris Masonf2984462008-04-10 16:19:33 -04003188 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003189 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003190 }
3191 unlock_buffer(bh);
3192 put_bh(bh);
3193}
3194
Yan Zhenga512bbf2008-12-08 16:46:26 -05003195struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3196{
3197 struct buffer_head *bh;
3198 struct buffer_head *latest = NULL;
3199 struct btrfs_super_block *super;
3200 int i;
3201 u64 transid = 0;
3202 u64 bytenr;
3203
3204 /* we would like to check all the supers, but that would make
3205 * a btrfs mount succeed after a mkfs from a different FS.
3206 * So, we need to add a special mount option to scan for
3207 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3208 */
3209 for (i = 0; i < 1; i++) {
3210 bytenr = btrfs_sb_offset(i);
Anand Jain8068a472013-07-26 01:29:35 +08003211 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3212 i_size_read(bdev->bd_inode))
Yan Zhenga512bbf2008-12-08 16:46:26 -05003213 break;
Anand Jain8068a472013-07-26 01:29:35 +08003214 bh = __bread(bdev, bytenr / 4096,
3215 BTRFS_SUPER_INFO_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003216 if (!bh)
3217 continue;
3218
3219 super = (struct btrfs_super_block *)bh->b_data;
3220 if (btrfs_super_bytenr(super) != bytenr ||
Qu Wenruo3cae2102013-07-16 11:19:18 +08003221 btrfs_super_magic(super) != BTRFS_MAGIC) {
Yan Zhenga512bbf2008-12-08 16:46:26 -05003222 brelse(bh);
3223 continue;
3224 }
3225
3226 if (!latest || btrfs_super_generation(super) > transid) {
3227 brelse(latest);
3228 latest = bh;
3229 transid = btrfs_super_generation(super);
3230 } else {
3231 brelse(bh);
3232 }
3233 }
3234 return latest;
3235}
3236
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003237/*
3238 * this should be called twice, once with wait == 0 and
3239 * once with wait == 1. When wait == 0 is done, all the buffer heads
3240 * we write are pinned.
3241 *
3242 * They are released when wait == 1 is done.
3243 * max_mirrors must be the same for both runs, and it indicates how
3244 * many supers on this one device should be written.
3245 *
3246 * max_mirrors == 0 means to write them all.
3247 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003248static int write_dev_supers(struct btrfs_device *device,
3249 struct btrfs_super_block *sb,
3250 int do_barriers, int wait, int max_mirrors)
3251{
3252 struct buffer_head *bh;
3253 int i;
3254 int ret;
3255 int errors = 0;
3256 u32 crc;
3257 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003258
3259 if (max_mirrors == 0)
3260 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3261
Yan Zhenga512bbf2008-12-08 16:46:26 -05003262 for (i = 0; i < max_mirrors; i++) {
3263 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003264 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3265 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003266 break;
3267
3268 if (wait) {
3269 bh = __find_get_block(device->bdev, bytenr / 4096,
3270 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003271 if (!bh) {
3272 errors++;
3273 continue;
3274 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003275 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003276 if (!buffer_uptodate(bh))
3277 errors++;
3278
3279 /* drop our reference */
3280 brelse(bh);
3281
3282 /* drop the reference from the wait == 0 run */
3283 brelse(bh);
3284 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003285 } else {
3286 btrfs_set_super_bytenr(sb, bytenr);
3287
3288 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003289 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003290 BTRFS_CSUM_SIZE, crc,
3291 BTRFS_SUPER_INFO_SIZE -
3292 BTRFS_CSUM_SIZE);
3293 btrfs_csum_final(crc, sb->csum);
3294
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003295 /*
3296 * one reference for us, and we leave it for the
3297 * caller
3298 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003299 bh = __getblk(device->bdev, bytenr / 4096,
3300 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003301 if (!bh) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003302 printk(KERN_ERR "BTRFS: couldn't get super "
Josef Bacik634554d2013-04-29 10:05:57 -04003303 "buffer head for bytenr %Lu\n", bytenr);
3304 errors++;
3305 continue;
3306 }
3307
Yan Zhenga512bbf2008-12-08 16:46:26 -05003308 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3309
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003310 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003311 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003312
3313 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003314 lock_buffer(bh);
3315 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003316 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003317 }
3318
Chris Mason387125f2011-11-18 15:07:51 -05003319 /*
3320 * we fua the first super. The others we allow
3321 * to go down lazy.
3322 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003323 if (i == 0)
3324 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3325 else
3326 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003327 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003328 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003329 }
3330 return errors < i ? 0 : -1;
3331}
3332
Chris Mason387125f2011-11-18 15:07:51 -05003333/*
3334 * endio for the write_dev_flush, this will wake anyone waiting
3335 * for the barrier when it is done
3336 */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003337static void btrfs_end_empty_barrier(struct bio *bio)
Chris Mason387125f2011-11-18 15:07:51 -05003338{
Chris Mason387125f2011-11-18 15:07:51 -05003339 if (bio->bi_private)
3340 complete(bio->bi_private);
3341 bio_put(bio);
3342}
3343
3344/*
3345 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3346 * sent down. With wait == 1, it waits for the previous flush.
3347 *
3348 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3349 * capable
3350 */
3351static int write_dev_flush(struct btrfs_device *device, int wait)
3352{
3353 struct bio *bio;
3354 int ret = 0;
3355
3356 if (device->nobarriers)
3357 return 0;
3358
3359 if (wait) {
3360 bio = device->flush_bio;
3361 if (!bio)
3362 return 0;
3363
3364 wait_for_completion(&device->flush_wait);
3365
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003366 if (bio->bi_error) {
3367 ret = bio->bi_error;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003368 btrfs_dev_stat_inc_and_print(device,
3369 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003370 }
3371
3372 /* drop the reference from the wait == 0 run */
3373 bio_put(bio);
3374 device->flush_bio = NULL;
3375
3376 return ret;
3377 }
3378
3379 /*
3380 * one reference for us, and we leave it for the
3381 * caller
3382 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003383 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003384 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003385 if (!bio)
3386 return -ENOMEM;
3387
3388 bio->bi_end_io = btrfs_end_empty_barrier;
3389 bio->bi_bdev = device->bdev;
3390 init_completion(&device->flush_wait);
3391 bio->bi_private = &device->flush_wait;
3392 device->flush_bio = bio;
3393
3394 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003395 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003396
3397 return 0;
3398}
3399
3400/*
3401 * send an empty flush down to each device in parallel,
3402 * then wait for them
3403 */
3404static int barrier_all_devices(struct btrfs_fs_info *info)
3405{
3406 struct list_head *head;
3407 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003408 int errors_send = 0;
3409 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003410 int ret;
3411
3412 /* send down all the barriers */
3413 head = &info->fs_devices->devices;
3414 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003415 if (dev->missing)
3416 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003417 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003418 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003419 continue;
3420 }
3421 if (!dev->in_fs_metadata || !dev->writeable)
3422 continue;
3423
3424 ret = write_dev_flush(dev, 0);
3425 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003426 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003427 }
3428
3429 /* wait for all the barriers */
3430 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003431 if (dev->missing)
3432 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003433 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003434 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003435 continue;
3436 }
3437 if (!dev->in_fs_metadata || !dev->writeable)
3438 continue;
3439
3440 ret = write_dev_flush(dev, 1);
3441 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003442 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003443 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003444 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3445 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003446 return -EIO;
3447 return 0;
3448}
3449
Zhao Lei943c6e92015-08-19 15:54:15 +08003450int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3451{
3452 if ((flags & (BTRFS_BLOCK_GROUP_DUP |
3453 BTRFS_BLOCK_GROUP_RAID0 |
3454 BTRFS_AVAIL_ALLOC_BIT_SINGLE)) ||
3455 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0))
3456 return 0;
3457
3458 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3459 BTRFS_BLOCK_GROUP_RAID5 |
3460 BTRFS_BLOCK_GROUP_RAID10))
3461 return 1;
3462
3463 if (flags & BTRFS_BLOCK_GROUP_RAID6)
3464 return 2;
3465
3466 pr_warn("BTRFS: unknown raid type: %llu\n", flags);
3467 return 0;
3468}
3469
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003470int btrfs_calc_num_tolerated_disk_barrier_failures(
3471 struct btrfs_fs_info *fs_info)
3472{
3473 struct btrfs_ioctl_space_info space;
3474 struct btrfs_space_info *sinfo;
3475 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3476 BTRFS_BLOCK_GROUP_SYSTEM,
3477 BTRFS_BLOCK_GROUP_METADATA,
3478 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003479 int i;
3480 int c;
3481 int num_tolerated_disk_barrier_failures =
3482 (int)fs_info->fs_devices->num_devices;
3483
Zhao Lei2c458042015-07-16 15:00:46 +08003484 for (i = 0; i < ARRAY_SIZE(types); i++) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003485 struct btrfs_space_info *tmp;
3486
3487 sinfo = NULL;
3488 rcu_read_lock();
3489 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3490 if (tmp->flags == types[i]) {
3491 sinfo = tmp;
3492 break;
3493 }
3494 }
3495 rcu_read_unlock();
3496
3497 if (!sinfo)
3498 continue;
3499
3500 down_read(&sinfo->groups_sem);
3501 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
Zhao Lei2c458042015-07-16 15:00:46 +08003502 u64 flags;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003503
Zhao Lei2c458042015-07-16 15:00:46 +08003504 if (list_empty(&sinfo->block_groups[c]))
3505 continue;
3506
3507 btrfs_get_block_group_info(&sinfo->block_groups[c],
3508 &space);
3509 if (space.total_bytes == 0 || space.used_bytes == 0)
3510 continue;
3511 flags = space.flags;
Zhao Lei943c6e92015-08-19 15:54:15 +08003512
3513 num_tolerated_disk_barrier_failures = min(
3514 num_tolerated_disk_barrier_failures,
3515 btrfs_get_num_tolerated_disk_barrier_failures(
3516 flags));
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003517 }
3518 up_read(&sinfo->groups_sem);
3519 }
3520
3521 return num_tolerated_disk_barrier_failures;
3522}
3523
Eric Sandeen48a3b632013-04-25 20:41:01 +00003524static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003525{
Chris Masone5e9a522009-06-10 15:17:02 -04003526 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003527 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003528 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003529 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003530 int ret;
3531 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003532 int max_errors;
3533 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003534 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003535
3536 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003537 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003538
David Sterba6c417612011-04-13 15:41:04 +02003539 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003540 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003541
Chris Mason174ba502011-05-27 10:03:58 -04003542 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003543 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003544 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003545
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003546 if (do_barriers) {
3547 ret = barrier_all_devices(root->fs_info);
3548 if (ret) {
3549 mutex_unlock(
3550 &root->fs_info->fs_devices->device_list_mutex);
3551 btrfs_error(root->fs_info, ret,
3552 "errors while submitting device barriers.");
3553 return ret;
3554 }
3555 }
Chris Mason387125f2011-11-18 15:07:51 -05003556
Xiao Guangrong1f781602011-04-20 10:09:16 +00003557 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003558 if (!dev->bdev) {
3559 total_errors++;
3560 continue;
3561 }
Yan Zheng2b820322008-11-17 21:11:30 -05003562 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003563 continue;
3564
Yan Zheng2b820322008-11-17 21:11:30 -05003565 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003566 btrfs_set_stack_device_type(dev_item, dev->type);
3567 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003568 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003569 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003570 btrfs_set_stack_device_bytes_used(dev_item,
3571 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003572 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3573 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3574 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3575 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003576 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003577
Chris Masona061fc82008-05-07 11:43:44 -04003578 flags = btrfs_super_flags(sb);
3579 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003580
Yan Zhenga512bbf2008-12-08 16:46:26 -05003581 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003582 if (ret)
3583 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003584 }
Chris Masona236aed2008-04-29 09:38:00 -04003585 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003586 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003587 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003588 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003589
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003590 /* FUA is masked off if unsupported and can't be the reason */
3591 btrfs_error(root->fs_info, -EIO,
3592 "%d errors while writing supers", total_errors);
3593 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003594 }
Chris Masonf2984462008-04-10 16:19:33 -04003595
Yan Zhenga512bbf2008-12-08 16:46:26 -05003596 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003597 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003598 if (!dev->bdev)
3599 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003600 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003601 continue;
3602
Yan Zhenga512bbf2008-12-08 16:46:26 -05003603 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3604 if (ret)
3605 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003606 }
Chris Mason174ba502011-05-27 10:03:58 -04003607 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003608 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003609 btrfs_error(root->fs_info, -EIO,
3610 "%d errors while writing supers", total_errors);
3611 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003612 }
Chris Masonf2984462008-04-10 16:19:33 -04003613 return 0;
3614}
3615
Yan Zhenga512bbf2008-12-08 16:46:26 -05003616int write_ctree_super(struct btrfs_trans_handle *trans,
3617 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003618{
Rashikaf570e7572013-10-31 16:50:42 +05303619 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003620}
3621
Miao Xiecb517ea2013-05-15 07:48:19 +00003622/* Drop a fs root from the radix tree and free it. */
3623void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3624 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003625{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003626 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003627 radix_tree_delete(&fs_info->fs_roots_radix,
3628 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003629 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003630
3631 if (btrfs_root_refs(&root->root_item) == 0)
3632 synchronize_srcu(&fs_info->subvol_srcu);
3633
Liu Bo1a4319c2014-01-13 19:53:53 +08003634 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003635 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003636
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003637 if (root->free_ino_pinned)
3638 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3639 if (root->free_ino_ctl)
3640 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003641 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003642}
3643
3644static void free_fs_root(struct btrfs_root *root)
3645{
David Sterba57cdc8d2014-02-05 02:37:48 +01003646 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003647 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003648 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3649 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003650 if (root->anon_dev)
3651 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003652 if (root->subv_writers)
3653 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003654 free_extent_buffer(root->node);
3655 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003656 kfree(root->free_ino_ctl);
3657 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003658 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003659 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003660}
3661
Miao Xiecb517ea2013-05-15 07:48:19 +00003662void btrfs_free_fs_root(struct btrfs_root *root)
3663{
3664 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003665}
3666
Yan Zhengc146afa2008-11-12 14:34:12 -05003667int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3668{
3669 u64 root_objectid = 0;
3670 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003671 int i = 0;
3672 int err = 0;
3673 unsigned int ret = 0;
3674 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003675
3676 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003677 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003678 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3679 (void **)gang, root_objectid,
3680 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003681 if (!ret) {
3682 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003683 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003684 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003685 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003686
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003687 for (i = 0; i < ret; i++) {
3688 /* Avoid to grab roots in dead_roots */
3689 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3690 gang[i] = NULL;
3691 continue;
3692 }
3693 /* grab all the search result for later use */
3694 gang[i] = btrfs_grab_fs_root(gang[i]);
3695 }
3696 srcu_read_unlock(&fs_info->subvol_srcu, index);
3697
3698 for (i = 0; i < ret; i++) {
3699 if (!gang[i])
3700 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003701 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003702 err = btrfs_orphan_cleanup(gang[i]);
3703 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003704 break;
3705 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003706 }
3707 root_objectid++;
3708 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003709
3710 /* release the uncleaned roots due to error */
3711 for (; i < ret; i++) {
3712 if (gang[i])
3713 btrfs_put_fs_root(gang[i]);
3714 }
3715 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003716}
3717
3718int btrfs_commit_super(struct btrfs_root *root)
3719{
3720 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003721
3722 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003723 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003724 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003725 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003726
3727 /* wait until ongoing cleanup work done */
3728 down_write(&root->fs_info->cleanup_work_sem);
3729 up_write(&root->fs_info->cleanup_work_sem);
3730
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003731 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003732 if (IS_ERR(trans))
3733 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003734 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003735}
3736
David Sterba3abdbd72014-06-04 18:10:45 +02003737void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003738{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003739 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003740 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003741
Chris Masonfacda1e2007-06-08 18:11:48 -04003742 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003743 smp_mb();
3744
Stefan Behrens803b2f52013-08-15 17:11:21 +02003745 /* wait for the uuid_scan task to finish */
3746 down(&fs_info->uuid_tree_rescan_sem);
3747 /* avoid complains from lockdep et al., set sem back to initial state */
3748 up(&fs_info->uuid_tree_rescan_sem);
3749
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003750 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003751 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003752
Stefan Behrens8dabb742012-11-06 13:15:27 +01003753 btrfs_dev_replace_suspend_for_unmount(fs_info);
3754
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003755 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003756
3757 /* wait for any defraggers to finish */
3758 wait_event(fs_info->transaction_wait,
3759 (atomic_read(&fs_info->defrag_running) == 0));
3760
3761 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003762 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003763
Miao Xie21c7e752014-05-13 17:29:04 -07003764 cancel_work_sync(&fs_info->async_reclaim_work);
3765
Yan Zhengc146afa2008-11-12 14:34:12 -05003766 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003767 /*
3768 * If the cleaner thread is stopped and there are
3769 * block groups queued for removal, the deletion will be
3770 * skipped when we quit the cleaner thread.
3771 */
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003772 btrfs_delete_unused_bgs(root->fs_info);
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003773
liuboacce9522011-01-06 19:30:25 +08003774 ret = btrfs_commit_super(root);
3775 if (ret)
Eric Sandeen04892342014-08-01 18:12:36 -05003776 btrfs_err(fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003777 }
3778
Miao Xie87533c42013-01-29 10:14:48 +00003779 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003780 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003781
Al Viroe3029d92011-11-17 00:56:18 -05003782 kthread_stop(fs_info->transaction_kthread);
3783 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003784
Yan Zhengf25784b2009-07-28 08:41:57 -04003785 fs_info->closing = 2;
3786 smp_mb();
3787
Eric Sandeen04892342014-08-01 18:12:36 -05003788 btrfs_free_qgroup_config(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003789
Miao Xie963d6782013-01-29 10:10:51 +00003790 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Eric Sandeen04892342014-08-01 18:12:36 -05003791 btrfs_info(fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003792 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003793 }
Yanbcc63ab2008-07-30 16:29:20 -04003794
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003795 btrfs_sysfs_remove_one(fs_info);
Anand Jainb7c35e82015-03-10 06:38:38 +08003796 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003797
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003798 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003799
Liu Bo1a4319c2014-01-13 19:53:53 +08003800 btrfs_put_block_group_cache(fs_info);
3801
Josef Bacik2b1360d2013-10-07 15:14:44 -04003802 btrfs_free_block_groups(fs_info);
3803
Wang Shilongde348ee2014-04-09 19:23:22 +08003804 /*
3805 * we must make sure there is not any read request to
3806 * submit after we stopping all workers.
3807 */
3808 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003809 btrfs_stop_all_workers(fs_info);
3810
Josef Bacik47ab2a62014-09-18 11:20:02 -04003811 fs_info->open = 0;
Josef Bacik13e6c372013-05-30 16:55:44 -04003812 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003813
Josef Bacik13e6c372013-05-30 16:55:44 -04003814 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003815
Stefan Behrens21adbd52011-11-09 13:44:05 +01003816#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3817 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3818 btrfsic_unmount(root, fs_info->fs_devices);
3819#endif
3820
Chris Masondfe25022008-05-13 13:46:40 -04003821 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003822 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003823
Miao Xiee2d84522013-01-29 10:09:20 +00003824 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003825 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003826 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003827 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003828 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003829
David Woodhouse53b381b2013-01-29 18:40:14 -05003830 btrfs_free_stripe_hash_table(fs_info);
3831
Chris Masoncdfb0802015-04-06 18:17:00 -07003832 __btrfs_free_block_rsv(root->orphan_block_rsv);
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003833 root->orphan_block_rsv = NULL;
Filipe Manana04216822014-11-27 21:14:15 +00003834
3835 lock_chunks(root);
3836 while (!list_empty(&fs_info->pinned_chunks)) {
3837 struct extent_map *em;
3838
3839 em = list_first_entry(&fs_info->pinned_chunks,
3840 struct extent_map, list);
3841 list_del_init(&em->list);
3842 free_extent_map(em);
3843 }
3844 unlock_chunks(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003845}
3846
Chris Masonb9fab912012-05-06 07:23:47 -04003847int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3848 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003849{
Chris Mason1259ab72008-05-12 13:39:03 -04003850 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003851 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003852
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003853 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003854 if (!ret)
3855 return ret;
3856
3857 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003858 parent_transid, atomic);
3859 if (ret == -EAGAIN)
3860 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003861 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003862}
Chris Mason6702ed42007-08-07 16:15:09 -04003863
Chris Mason5f39d392007-10-15 16:14:19 -04003864int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3865{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003866 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003867}
3868
3869void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3870{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003871 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003872 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003873 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003874
Josef Bacik06ea65a2013-09-19 16:07:01 -04003875#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3876 /*
3877 * This is a fast path so only do this check if we have sanity tests
3878 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3879 * outside of the sanity tests.
3880 */
3881 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3882 return;
3883#endif
3884 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003885 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003886 if (transid != root->fs_info->generation)
3887 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003888 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003889 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003890 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003891 if (!was_dirty)
3892 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3893 buf->len,
3894 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003895#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3896 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3897 btrfs_print_leaf(root, buf);
3898 ASSERT(0);
3899 }
3900#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05003901}
3902
Liu Bob53d3f52012-11-14 14:34:34 +00003903static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3904 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003905{
Chris Mason188de642008-05-09 11:52:25 -04003906 /*
3907 * looks as though older kernels can get into trouble with
3908 * this code, they end up stuck in balance_dirty_pages forever
3909 */
Miao Xiee2d84522013-01-29 10:09:20 +00003910 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04003911
Jens Axboe6933c022009-03-17 09:36:37 +01003912 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003913 return;
3914
Liu Bob53d3f52012-11-14 14:34:34 +00003915 if (flush_delayed)
3916 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003917
Miao Xiee2d84522013-01-29 10:09:20 +00003918 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3919 BTRFS_DIRTY_METADATA_THRESH);
3920 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08003921 balance_dirty_pages_ratelimited(
3922 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08003923 }
3924 return;
3925}
3926
Liu Bob53d3f52012-11-14 14:34:34 +00003927void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08003928{
Liu Bob53d3f52012-11-14 14:34:34 +00003929 __btrfs_btree_balance_dirty(root, 1);
3930}
Miao Xie16cdcec2011-04-22 18:12:22 +08003931
Liu Bob53d3f52012-11-14 14:34:34 +00003932void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3933{
3934 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04003935}
Chris Mason6b800532007-10-15 16:17:34 -04003936
Chris Masonca7a79a2008-05-12 12:59:19 -04003937int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003938{
Chris Mason727011e2010-08-06 13:21:20 -04003939 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003940 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003941}
Chris Mason0da54682007-11-07 21:08:01 -05003942
David Sterbafcd1f0652012-03-06 00:06:18 +01003943static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003944 int read_only)
3945{
David Sterbac9260932014-09-30 19:16:47 +02003946 struct btrfs_super_block *sb = fs_info->super_copy;
3947 int ret = 0;
3948
David Sterba21e76262014-10-27 13:52:21 +01003949 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
3950 printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n",
3951 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02003952 ret = -EINVAL;
3953 }
David Sterba21e76262014-10-27 13:52:21 +01003954 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
3955 printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n",
3956 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02003957 ret = -EINVAL;
3958 }
David Sterba21e76262014-10-27 13:52:21 +01003959 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
3960 printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n",
3961 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02003962 ret = -EINVAL;
3963 }
3964
David Sterba1104a882013-03-06 15:57:46 +01003965 /*
David Sterbac9260932014-09-30 19:16:47 +02003966 * The common minimum, we don't know if we can trust the nodesize/sectorsize
3967 * items yet, they'll be verified later. Issue just a warning.
David Sterba1104a882013-03-06 15:57:46 +01003968 */
David Sterba21e76262014-10-27 13:52:21 +01003969 if (!IS_ALIGNED(btrfs_super_root(sb), 4096))
David Sterbac9260932014-09-30 19:16:47 +02003970 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
David Sterbacd743fa2014-10-31 19:40:14 +01003971 btrfs_super_root(sb));
David Sterba21e76262014-10-27 13:52:21 +01003972 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), 4096))
David Sterbacd743fa2014-10-31 19:40:14 +01003973 printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n",
3974 btrfs_super_chunk_root(sb));
David Sterba21e76262014-10-27 13:52:21 +01003975 if (!IS_ALIGNED(btrfs_super_log_root(sb), 4096))
David Sterbacd743fa2014-10-31 19:40:14 +01003976 printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01003977 btrfs_super_log_root(sb));
David Sterbac9260932014-09-30 19:16:47 +02003978
David Sterba75d6ad32014-10-31 17:18:08 +01003979 /*
3980 * Check the lower bound, the alignment and other constraints are
3981 * checked later.
3982 */
3983 if (btrfs_super_nodesize(sb) < 4096) {
3984 printk(KERN_ERR "BTRFS: nodesize too small: %u < 4096\n",
3985 btrfs_super_nodesize(sb));
3986 ret = -EINVAL;
3987 }
3988 if (btrfs_super_sectorsize(sb) < 4096) {
3989 printk(KERN_ERR "BTRFS: sectorsize too small: %u < 4096\n",
3990 btrfs_super_sectorsize(sb));
3991 ret = -EINVAL;
3992 }
3993
David Sterbac9260932014-09-30 19:16:47 +02003994 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
3995 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
3996 fs_info->fsid, sb->dev_item.fsid);
3997 ret = -EINVAL;
3998 }
3999
4000 /*
4001 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4002 * done later
4003 */
David Sterba21e76262014-10-27 13:52:21 +01004004 if (btrfs_super_num_devices(sb) > (1UL << 31))
David Sterbac9260932014-09-30 19:16:47 +02004005 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004006 btrfs_super_num_devices(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004007 if (btrfs_super_num_devices(sb) == 0) {
4008 printk(KERN_ERR "BTRFS: number of devices is 0\n");
4009 ret = -EINVAL;
4010 }
David Sterbac9260932014-09-30 19:16:47 +02004011
David Sterba21e76262014-10-27 13:52:21 +01004012 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
David Sterbac9260932014-09-30 19:16:47 +02004013 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
David Sterba21e76262014-10-27 13:52:21 +01004014 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
David Sterbac9260932014-09-30 19:16:47 +02004015 ret = -EINVAL;
4016 }
4017
4018 /*
David Sterbace7fca52014-10-31 18:42:05 +01004019 * Obvious sys_chunk_array corruptions, it must hold at least one key
4020 * and one chunk
4021 */
4022 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
4023 printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
4024 btrfs_super_sys_array_size(sb),
4025 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
4026 ret = -EINVAL;
4027 }
4028 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4029 + sizeof(struct btrfs_chunk)) {
Fabian Frederickd2207122015-02-14 13:10:22 +01004030 printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n",
David Sterbace7fca52014-10-31 18:42:05 +01004031 btrfs_super_sys_array_size(sb),
4032 sizeof(struct btrfs_disk_key)
4033 + sizeof(struct btrfs_chunk));
4034 ret = -EINVAL;
4035 }
4036
4037 /*
David Sterbac9260932014-09-30 19:16:47 +02004038 * The generation is a global counter, we'll trust it more than the others
4039 * but it's still possible that it's the one that's wrong.
4040 */
David Sterba21e76262014-10-27 13:52:21 +01004041 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
David Sterbac9260932014-09-30 19:16:47 +02004042 printk(KERN_WARNING
4043 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004044 btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb));
4045 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4046 && btrfs_super_cache_generation(sb) != (u64)-1)
David Sterbac9260932014-09-30 19:16:47 +02004047 printk(KERN_WARNING
4048 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004049 btrfs_super_generation(sb), btrfs_super_cache_generation(sb));
David Sterbac9260932014-09-30 19:16:47 +02004050
4051 return ret;
liuboacce9522011-01-06 19:30:25 +08004052}
4053
Eric Sandeen48a3b632013-04-25 20:41:01 +00004054static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004055{
liuboacce9522011-01-06 19:30:25 +08004056 mutex_lock(&root->fs_info->cleaner_mutex);
4057 btrfs_run_delayed_iputs(root);
4058 mutex_unlock(&root->fs_info->cleaner_mutex);
4059
4060 down_write(&root->fs_info->cleanup_work_sem);
4061 up_write(&root->fs_info->cleanup_work_sem);
4062
4063 /* cleanup FS via transaction */
4064 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08004065}
4066
Jeff Mahoney143bede2012-03-01 14:56:26 +01004067static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004068{
liuboacce9522011-01-06 19:30:25 +08004069 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08004070
Miao Xie199c2a92013-05-15 07:48:23 +00004071 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05004072 /*
4073 * This will just short circuit the ordered completion stuff which will
4074 * make sure the ordered extent gets properly cleaned up.
4075 */
Miao Xie199c2a92013-05-15 07:48:23 +00004076 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05004077 root_extent_list)
4078 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00004079 spin_unlock(&root->ordered_extent_lock);
4080}
4081
4082static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4083{
4084 struct btrfs_root *root;
4085 struct list_head splice;
4086
4087 INIT_LIST_HEAD(&splice);
4088
4089 spin_lock(&fs_info->ordered_root_lock);
4090 list_splice_init(&fs_info->ordered_roots, &splice);
4091 while (!list_empty(&splice)) {
4092 root = list_first_entry(&splice, struct btrfs_root,
4093 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04004094 list_move_tail(&root->ordered_root,
4095 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00004096
Liu Bo2a85d9c2014-02-10 17:07:16 +08004097 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004098 btrfs_destroy_ordered_extents(root);
4099
Liu Bo2a85d9c2014-02-10 17:07:16 +08004100 cond_resched();
4101 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004102 }
4103 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08004104}
4105
Stefan Behrens35a36212013-08-14 18:12:25 +02004106static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4107 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004108{
4109 struct rb_node *node;
4110 struct btrfs_delayed_ref_root *delayed_refs;
4111 struct btrfs_delayed_ref_node *ref;
4112 int ret = 0;
4113
4114 delayed_refs = &trans->delayed_refs;
4115
4116 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004117 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04004118 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05004119 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08004120 return ret;
4121 }
4122
Josef Bacikd7df2c72014-01-23 09:21:38 -05004123 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4124 struct btrfs_delayed_ref_head *head;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004125 struct btrfs_delayed_ref_node *tmp;
Josef Bacike78417d2013-06-03 16:42:36 -04004126 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08004127
Josef Bacikd7df2c72014-01-23 09:21:38 -05004128 head = rb_entry(node, struct btrfs_delayed_ref_head,
4129 href_node);
4130 if (!mutex_trylock(&head->mutex)) {
4131 atomic_inc(&head->node.refs);
4132 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08004133
Josef Bacikd7df2c72014-01-23 09:21:38 -05004134 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04004135 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004136 btrfs_put_delayed_ref(&head->node);
4137 spin_lock(&delayed_refs->lock);
4138 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04004139 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05004140 spin_lock(&head->lock);
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004141 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4142 list) {
Josef Bacikd7df2c72014-01-23 09:21:38 -05004143 ref->in_tree = 0;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004144 list_del(&ref->list);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004145 atomic_dec(&delayed_refs->num_entries);
4146 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004147 }
4148 if (head->must_insert_reserved)
4149 pin_bytes = true;
4150 btrfs_free_delayed_extent_op(head->extent_op);
4151 delayed_refs->num_heads--;
4152 if (head->processing == 0)
4153 delayed_refs->num_heads_ready--;
4154 atomic_dec(&delayed_refs->num_entries);
4155 head->node.in_tree = 0;
4156 rb_erase(&head->href_node, &delayed_refs->href_root);
4157 spin_unlock(&head->lock);
4158 spin_unlock(&delayed_refs->lock);
4159 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08004160
Josef Bacikd7df2c72014-01-23 09:21:38 -05004161 if (pin_bytes)
4162 btrfs_pin_extent(root, head->node.bytenr,
4163 head->node.num_bytes, 1);
4164 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08004165 cond_resched();
4166 spin_lock(&delayed_refs->lock);
4167 }
4168
4169 spin_unlock(&delayed_refs->lock);
4170
4171 return ret;
4172}
4173
Jeff Mahoney143bede2012-03-01 14:56:26 +01004174static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004175{
4176 struct btrfs_inode *btrfs_inode;
4177 struct list_head splice;
4178
4179 INIT_LIST_HEAD(&splice);
4180
Miao Xieeb73c1b2013-05-15 07:48:22 +00004181 spin_lock(&root->delalloc_lock);
4182 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08004183
4184 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00004185 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4186 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08004187
4188 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00004189 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4190 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00004191 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004192
4193 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00004194
Miao Xieeb73c1b2013-05-15 07:48:22 +00004195 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004196 }
4197
Miao Xieeb73c1b2013-05-15 07:48:22 +00004198 spin_unlock(&root->delalloc_lock);
4199}
4200
4201static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4202{
4203 struct btrfs_root *root;
4204 struct list_head splice;
4205
4206 INIT_LIST_HEAD(&splice);
4207
4208 spin_lock(&fs_info->delalloc_root_lock);
4209 list_splice_init(&fs_info->delalloc_roots, &splice);
4210 while (!list_empty(&splice)) {
4211 root = list_first_entry(&splice, struct btrfs_root,
4212 delalloc_root);
4213 list_del_init(&root->delalloc_root);
4214 root = btrfs_grab_fs_root(root);
4215 BUG_ON(!root);
4216 spin_unlock(&fs_info->delalloc_root_lock);
4217
4218 btrfs_destroy_delalloc_inodes(root);
4219 btrfs_put_fs_root(root);
4220
4221 spin_lock(&fs_info->delalloc_root_lock);
4222 }
4223 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004224}
4225
4226static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4227 struct extent_io_tree *dirty_pages,
4228 int mark)
4229{
4230 int ret;
liuboacce9522011-01-06 19:30:25 +08004231 struct extent_buffer *eb;
4232 u64 start = 0;
4233 u64 end;
liuboacce9522011-01-06 19:30:25 +08004234
4235 while (1) {
4236 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004237 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004238 if (ret)
4239 break;
4240
4241 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4242 while (start <= end) {
Daniel Dressler01d58472014-11-21 17:15:07 +09004243 eb = btrfs_find_tree_block(root->fs_info, start);
David Sterba707e8a02014-06-04 19:22:26 +02004244 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004245 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004246 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004247 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004248
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004249 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4250 &eb->bflags))
4251 clear_extent_buffer_dirty(eb);
4252 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004253 }
4254 }
4255
4256 return ret;
4257}
4258
4259static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4260 struct extent_io_tree *pinned_extents)
4261{
4262 struct extent_io_tree *unpin;
4263 u64 start;
4264 u64 end;
4265 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004266 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004267
4268 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004269again:
liuboacce9522011-01-06 19:30:25 +08004270 while (1) {
4271 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004272 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004273 if (ret)
4274 break;
4275
liuboacce9522011-01-06 19:30:25 +08004276 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4277 btrfs_error_unpin_extent_range(root, start, end);
4278 cond_resched();
4279 }
4280
Liu Boed0eaa12012-06-14 02:23:21 -06004281 if (loop) {
4282 if (unpin == &root->fs_info->freed_extents[0])
4283 unpin = &root->fs_info->freed_extents[1];
4284 else
4285 unpin = &root->fs_info->freed_extents[0];
4286 loop = false;
4287 goto again;
4288 }
4289
liuboacce9522011-01-06 19:30:25 +08004290 return 0;
4291}
4292
Josef Bacik50d9aa92014-11-21 14:52:38 -05004293static void btrfs_free_pending_ordered(struct btrfs_transaction *cur_trans,
4294 struct btrfs_fs_info *fs_info)
4295{
4296 struct btrfs_ordered_extent *ordered;
4297
4298 spin_lock(&fs_info->trans_lock);
4299 while (!list_empty(&cur_trans->pending_ordered)) {
4300 ordered = list_first_entry(&cur_trans->pending_ordered,
4301 struct btrfs_ordered_extent,
4302 trans_list);
4303 list_del_init(&ordered->trans_list);
4304 spin_unlock(&fs_info->trans_lock);
4305
4306 btrfs_put_ordered_extent(ordered);
4307 spin_lock(&fs_info->trans_lock);
4308 }
4309 spin_unlock(&fs_info->trans_lock);
4310}
4311
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004312void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4313 struct btrfs_root *root)
4314{
4315 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004316
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004317 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004318 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004319
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004320 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004321 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004322
Josef Bacik50d9aa92014-11-21 14:52:38 -05004323 btrfs_free_pending_ordered(cur_trans, root->fs_info);
Miao Xie67cde342012-06-14 02:23:22 -06004324 btrfs_destroy_delayed_inodes(root);
4325 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004326
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004327 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4328 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004329 btrfs_destroy_pinned_extent(root,
4330 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004331
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004332 cur_trans->state =TRANS_STATE_COMPLETED;
4333 wake_up(&cur_trans->commit_wait);
4334
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004335 /*
4336 memset(cur_trans, 0, sizeof(*cur_trans));
4337 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4338 */
4339}
4340
Eric Sandeen48a3b632013-04-25 20:41:01 +00004341static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004342{
4343 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004344
liuboacce9522011-01-06 19:30:25 +08004345 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4346
Josef Bacika4abeea2011-04-11 17:25:13 -04004347 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004348 while (!list_empty(&root->fs_info->trans_list)) {
4349 t = list_first_entry(&root->fs_info->trans_list,
4350 struct btrfs_transaction, list);
4351 if (t->state >= TRANS_STATE_COMMIT_START) {
4352 atomic_inc(&t->use_count);
4353 spin_unlock(&root->fs_info->trans_lock);
4354 btrfs_wait_for_commit(root, t->transid);
4355 btrfs_put_transaction(t);
4356 spin_lock(&root->fs_info->trans_lock);
4357 continue;
4358 }
4359 if (t == root->fs_info->running_transaction) {
4360 t->state = TRANS_STATE_COMMIT_DOING;
4361 spin_unlock(&root->fs_info->trans_lock);
4362 /*
4363 * We wait for 0 num_writers since we don't hold a trans
4364 * handle open currently for this transaction.
4365 */
4366 wait_event(t->writer_wait,
4367 atomic_read(&t->num_writers) == 0);
4368 } else {
4369 spin_unlock(&root->fs_info->trans_lock);
4370 }
4371 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004372
Josef Bacik724e2312013-09-30 11:36:38 -04004373 spin_lock(&root->fs_info->trans_lock);
4374 if (t == root->fs_info->running_transaction)
4375 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004376 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004377 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004378
Josef Bacik724e2312013-09-30 11:36:38 -04004379 btrfs_put_transaction(t);
4380 trace_btrfs_transaction_commit(root);
4381 spin_lock(&root->fs_info->trans_lock);
4382 }
4383 spin_unlock(&root->fs_info->trans_lock);
4384 btrfs_destroy_all_ordered_extents(root->fs_info);
4385 btrfs_destroy_delayed_inodes(root);
4386 btrfs_assert_delayed_root_empty(root);
4387 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4388 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004389 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004390
4391 return 0;
4392}
4393
David Sterbae8c9f182015-01-02 18:23:10 +01004394static const struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004395 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004396 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004397 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004398 /* note we're sharing with inode.c for the merge bio hook */
4399 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004400};