blob: 1ad0f47ac850bce13c5bc369416eff171604768c [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050030#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020031#include <linux/ratelimit.h>
Stefan Behrens6463fe52013-04-19 15:08:05 +000032#include <linux/uuid.h>
Stefan Behrens803b2f52013-08-15 17:11:21 +020033#include <linux/semaphore.h>
David Sterba7e75bf32011-03-18 22:56:43 +000034#include <asm/unaligned.h>
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +000037#include "hash.h"
Chris Masone089f052007-03-16 16:20:31 -040038#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040039#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040040#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040041#include "print-tree.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080045#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010046#include "check-integrity.h"
Josef Bacik606686e2012-06-04 14:03:51 -040047#include "rcu-string.h"
Stefan Behrens8dabb742012-11-06 13:15:27 +010048#include "dev-replace.h"
David Woodhouse53b381b2013-01-29 18:40:14 -050049#include "raid56.h"
Jeff Mahoney5ac1d202013-11-01 13:06:58 -040050#include "sysfs.h"
Josef Bacikfcebe452014-05-13 17:30:47 -070051#include "qgroup.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050052
Josef Bacikde0022b2012-09-25 14:25:58 -040053#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
Chris Masond1310b22008-01-24 16:13:08 -050057static struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040058static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040059static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010060static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080061 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010062static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080063static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010065static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080066static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 struct extent_io_tree *dirty_pages,
68 int mark);
69static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000071static int btrfs_cleanup_transaction(struct btrfs_root *root);
72static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040073
Chris Masond352ac62008-09-29 15:18:18 -040074/*
David Sterba97eb6b62014-07-30 00:55:42 +020075 * btrfs_end_io_wq structs are used to do processing in task context when an IO
76 * is complete. This is used during reads to verify checksums, and it is used
Chris Masond352ac62008-09-29 15:18:18 -040077 * by writes to insert metadata for new file extents after IO is complete.
78 */
David Sterba97eb6b62014-07-30 00:55:42 +020079struct btrfs_end_io_wq {
Chris Masonce9adaa2008-04-09 16:28:12 -040080 struct bio *bio;
81 bio_end_io_t *end_io;
82 void *private;
83 struct btrfs_fs_info *info;
84 int error;
David Sterbabfebd8b2014-07-30 00:25:45 +020085 enum btrfs_wq_endio_type metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040086 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040087 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040088};
Chris Mason0da54682007-11-07 21:08:01 -050089
David Sterba97eb6b62014-07-30 00:55:42 +020090static struct kmem_cache *btrfs_end_io_wq_cache;
91
92int __init btrfs_end_io_wq_init(void)
93{
94 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
95 sizeof(struct btrfs_end_io_wq),
96 0,
97 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
98 NULL);
99 if (!btrfs_end_io_wq_cache)
100 return -ENOMEM;
101 return 0;
102}
103
104void btrfs_end_io_wq_exit(void)
105{
106 if (btrfs_end_io_wq_cache)
107 kmem_cache_destroy(btrfs_end_io_wq_cache);
108}
109
Chris Masond352ac62008-09-29 15:18:18 -0400110/*
111 * async submit bios are used to offload expensive checksumming
112 * onto the worker threads. They checksum file and metadata bios
113 * just before they are sent down the IO stack.
114 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400115struct async_submit_bio {
116 struct inode *inode;
117 struct bio *bio;
118 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500119 extent_submit_bio_hook_t *submit_bio_start;
120 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400121 int rw;
122 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400123 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400124 /*
125 * bio_offset is optional, can be used if the pages in the bio
126 * can't tell us where in the file the bio should go
127 */
128 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400129 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100130 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400131};
132
Chris Mason85d4e462011-07-26 16:11:19 -0400133/*
134 * Lockdep class keys for extent_buffer->lock's in this root. For a given
135 * eb, the lockdep key is determined by the btrfs_root it belongs to and
136 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500137 *
Chris Mason85d4e462011-07-26 16:11:19 -0400138 * Different roots are used for different purposes and may nest inside each
139 * other and they require separate keysets. As lockdep keys should be
140 * static, assign keysets according to the purpose of the root as indicated
141 * by btrfs_root->objectid. This ensures that all special purpose roots
142 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500143 *
Chris Mason85d4e462011-07-26 16:11:19 -0400144 * Lock-nesting across peer nodes is always done with the immediate parent
145 * node locked thus preventing deadlock. As lockdep doesn't know this, use
146 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500147 *
Chris Mason85d4e462011-07-26 16:11:19 -0400148 * The key is set by the readpage_end_io_hook after the buffer has passed
149 * csum validation but before the pages are unlocked. It is also set by
150 * btrfs_init_new_buffer on freshly allocated blocks.
151 *
152 * We also add a check to make sure the highest level of the tree is the
153 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
154 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500155 */
156#ifdef CONFIG_DEBUG_LOCK_ALLOC
157# if BTRFS_MAX_LEVEL != 8
158# error
159# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400160
161static struct btrfs_lockdep_keyset {
162 u64 id; /* root objectid */
163 const char *name_stem; /* lock name stem */
164 char names[BTRFS_MAX_LEVEL + 1][20];
165 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
166} btrfs_lockdep_keysets[] = {
167 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
168 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
169 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
170 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
171 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
172 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000173 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400174 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
175 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
176 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200177 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
Chris Mason85d4e462011-07-26 16:11:19 -0400178 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500179};
Chris Mason85d4e462011-07-26 16:11:19 -0400180
181void __init btrfs_init_lockdep(void)
182{
183 int i, j;
184
185 /* initialize lockdep class names */
186 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
187 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
188
189 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
190 snprintf(ks->names[j], sizeof(ks->names[j]),
191 "btrfs-%s-%02d", ks->name_stem, j);
192 }
193}
194
195void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
196 int level)
197{
198 struct btrfs_lockdep_keyset *ks;
199
200 BUG_ON(level >= ARRAY_SIZE(ks->keys));
201
202 /* find the matching keyset, id 0 is the default entry */
203 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
204 if (ks->id == objectid)
205 break;
206
207 lockdep_set_class_and_name(&eb->lock,
208 &ks->keys[level], ks->names[level]);
209}
210
Chris Mason4008c042009-02-12 14:09:45 -0500211#endif
212
Chris Masond352ac62008-09-29 15:18:18 -0400213/*
214 * extents on the btree inode are pretty simple, there's one extent
215 * that covers the entire device
216 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500217static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200218 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500219 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400220{
221 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
222 struct extent_map *em;
223 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400224
Chris Mason890871b2009-09-02 16:24:52 -0400225 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500226 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400227 if (em) {
228 em->bdev =
229 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400230 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400231 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400232 }
Chris Mason890871b2009-09-02 16:24:52 -0400233 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400234
David Sterba172ddd62011-04-21 00:48:27 +0200235 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400236 if (!em) {
237 em = ERR_PTR(-ENOMEM);
238 goto out;
239 }
240 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400241 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400242 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400243 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400244 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500245
Chris Mason890871b2009-09-02 16:24:52 -0400246 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400247 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400248 if (ret == -EEXIST) {
249 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400250 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600251 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600252 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400253 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400254 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600255 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400256 }
Chris Mason890871b2009-09-02 16:24:52 -0400257 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400258
Chris Mason5f39d392007-10-15 16:14:19 -0400259out:
260 return em;
261}
262
Liu Bob0496682013-03-14 14:57:45 +0000263u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400264{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000265 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400266}
267
268void btrfs_csum_final(u32 crc, char *result)
269{
David Sterba7e75bf32011-03-18 22:56:43 +0000270 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400271}
272
Chris Masond352ac62008-09-29 15:18:18 -0400273/*
274 * compute the csum for a btree block, and either verify it or write it
275 * into the csum field of the block.
276 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400277static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
278 int verify)
279{
David Sterba6c417612011-04-13 15:41:04 +0200280 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500281 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400282 unsigned long len;
283 unsigned long cur_len;
284 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400285 char *kaddr;
286 unsigned long map_start;
287 unsigned long map_len;
288 int err;
289 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500290 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400291
292 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500293 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400294 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400295 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500296 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400297 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400298 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000299 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400300 crc, cur_len);
301 len -= cur_len;
302 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400303 }
Josef Bacik607d4322008-12-02 07:17:45 -0500304 if (csum_size > sizeof(inline_result)) {
305 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
306 if (!result)
307 return 1;
308 } else {
309 result = (char *)&inline_result;
310 }
311
Chris Mason19c00dd2007-10-15 16:19:22 -0400312 btrfs_csum_final(crc, result);
313
314 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500315 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500316 u32 val;
317 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500318 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500319
Josef Bacik607d4322008-12-02 07:17:45 -0500320 read_extent_buffer(buf, &val, 0, csum_size);
Frank Holtonefe120a2013-12-20 11:37:06 -0500321 printk_ratelimited(KERN_INFO
322 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
323 "level %d\n",
324 root->fs_info->sb->s_id, buf->start,
325 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500326 if (result != (char *)&inline_result)
327 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400328 return 1;
329 }
330 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500331 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400332 }
Josef Bacik607d4322008-12-02 07:17:45 -0500333 if (result != (char *)&inline_result)
334 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400335 return 0;
336}
337
Chris Masond352ac62008-09-29 15:18:18 -0400338/*
339 * we can't consider a given block up to date unless the transid of the
340 * block matches the transid in the parent node's pointer. This is how we
341 * detect blocks that either didn't get written at all or got written
342 * in the wrong place.
343 */
Chris Mason1259ab72008-05-12 13:39:03 -0400344static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400345 struct extent_buffer *eb, u64 parent_transid,
346 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400347{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000348 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400349 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200350 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400351
352 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
353 return 0;
354
Chris Masonb9fab912012-05-06 07:23:47 -0400355 if (atomic)
356 return -EAGAIN;
357
Josef Bacika26e8c92014-03-28 17:07:27 -0400358 if (need_lock) {
359 btrfs_tree_read_lock(eb);
360 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
361 }
362
Josef Bacik2ac55d42010-02-03 19:33:23 +0000363 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100364 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400365 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400366 btrfs_header_generation(eb) == parent_transid) {
367 ret = 0;
368 goto out;
369 }
Wang Shilong29549ae2014-07-04 17:59:06 +0800370 printk_ratelimited(KERN_INFO "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n",
371 eb->fs_info->sb->s_id, eb->start,
372 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400373 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400374
375 /*
376 * Things reading via commit roots that don't have normal protection,
377 * like send, can have a really old block in cache that may point at a
378 * block that has been free'd and re-allocated. So don't clear uptodate
379 * if we find an eb that is under IO (dirty/writeback) because we could
380 * end up reading in the stale data and then writing it back out and
381 * making everybody very sad.
382 */
383 if (!extent_buffer_under_io(eb))
384 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400385out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000386 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
387 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700388 if (need_lock)
389 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400390 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400391}
392
Chris Masond352ac62008-09-29 15:18:18 -0400393/*
David Sterba1104a882013-03-06 15:57:46 +0100394 * Return 0 if the superblock checksum type matches the checksum value of that
395 * algorithm. Pass the raw disk superblock data.
396 */
397static int btrfs_check_super_csum(char *raw_disk_sb)
398{
399 struct btrfs_super_block *disk_sb =
400 (struct btrfs_super_block *)raw_disk_sb;
401 u16 csum_type = btrfs_super_csum_type(disk_sb);
402 int ret = 0;
403
404 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
405 u32 crc = ~(u32)0;
406 const int csum_size = sizeof(crc);
407 char result[csum_size];
408
409 /*
410 * The super_block structure does not span the whole
411 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
412 * is filled with zeros and is included in the checkum.
413 */
414 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
415 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
416 btrfs_csum_final(crc, result);
417
418 if (memcmp(raw_disk_sb, result, csum_size))
419 ret = 1;
Chris Mason667e7d92013-05-07 11:00:13 -0400420
421 if (ret && btrfs_super_generation(disk_sb) < 10) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500422 printk(KERN_WARNING
423 "BTRFS: super block crcs don't match, older mkfs detected\n");
Chris Mason667e7d92013-05-07 11:00:13 -0400424 ret = 0;
425 }
David Sterba1104a882013-03-06 15:57:46 +0100426 }
427
428 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500429 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100430 csum_type);
431 ret = 1;
432 }
433
434 return ret;
435}
436
437/*
Chris Masond352ac62008-09-29 15:18:18 -0400438 * helper to read a given tree block, doing retries as required when
439 * the checksums don't match and we have alternate mirrors to try.
440 */
Chris Masonf1885912008-04-09 16:28:12 -0400441static int btree_read_extent_buffer_pages(struct btrfs_root *root,
442 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400443 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400444{
445 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400446 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400447 int ret;
448 int num_copies = 0;
449 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400450 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400451
Josef Bacika826d6d2011-03-16 13:42:43 -0400452 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400453 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
454 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200455 ret = read_extent_buffer_pages(io_tree, eb, start,
456 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400457 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600458 if (!ret) {
459 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400460 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600461 break;
462 else
463 ret = -EIO;
464 }
Chris Masond3977122009-01-05 21:25:51 -0500465
Josef Bacika826d6d2011-03-16 13:42:43 -0400466 /*
467 * This buffer's crc is fine, but its contents are corrupted, so
468 * there is no reason to read the other copies, they won't be
469 * any less wrong.
470 */
471 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400472 break;
473
Stefan Behrens5d964052012-11-05 14:59:07 +0100474 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400475 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400476 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400477 break;
Chris Mason42352982008-04-28 16:40:52 -0400478
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400479 if (!failed_mirror) {
480 failed = 1;
481 failed_mirror = eb->read_mirror;
482 }
483
Chris Masonf1885912008-04-09 16:28:12 -0400484 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400485 if (mirror_num == failed_mirror)
486 mirror_num++;
487
Chris Mason42352982008-04-28 16:40:52 -0400488 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400489 break;
Chris Masonf1885912008-04-09 16:28:12 -0400490 }
Josef Bacikea466792012-03-26 21:57:36 -0400491
Stefan Behrensc0901582012-07-10 07:30:17 -0600492 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400493 repair_eb_io_failure(root, eb, failed_mirror);
494
495 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400496}
Chris Mason19c00dd2007-10-15 16:19:22 -0400497
Chris Masond352ac62008-09-29 15:18:18 -0400498/*
Chris Masond3977122009-01-05 21:25:51 -0500499 * checksum a dirty tree block before IO. This has extra checks to make sure
500 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400501 */
Chris Masond3977122009-01-05 21:25:51 -0500502
Christoph Hellwigb2950862008-12-02 09:54:17 -0500503static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400504{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000505 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400506 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400507 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400508
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500509 eb = (struct extent_buffer *)page->private;
510 if (page != eb->pages[0])
511 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400512 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530513 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500514 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400515 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400516 return 0;
517}
518
Yan Zheng2b820322008-11-17 21:11:30 -0500519static int check_tree_block_fsid(struct btrfs_root *root,
520 struct extent_buffer *eb)
521{
522 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
523 u8 fsid[BTRFS_UUID_SIZE];
524 int ret = 1;
525
Ross Kirk0a4e5582013-09-24 10:12:38 +0100526 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500527 while (fs_devices) {
528 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
529 ret = 0;
530 break;
531 }
532 fs_devices = fs_devices->seed;
533 }
534 return ret;
535}
536
Josef Bacika826d6d2011-03-16 13:42:43 -0400537#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500538 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
539 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200540 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400541
542static noinline int check_leaf(struct btrfs_root *root,
543 struct extent_buffer *leaf)
544{
545 struct btrfs_key key;
546 struct btrfs_key leaf_key;
547 u32 nritems = btrfs_header_nritems(leaf);
548 int slot;
549
550 if (nritems == 0)
551 return 0;
552
553 /* Check the 0 item */
554 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
555 BTRFS_LEAF_DATA_SIZE(root)) {
556 CORRUPT("invalid item offset size pair", leaf, root, 0);
557 return -EIO;
558 }
559
560 /*
561 * Check to make sure each items keys are in the correct order and their
562 * offsets make sense. We only have to loop through nritems-1 because
563 * we check the current slot against the next slot, which verifies the
564 * next slot's offset+size makes sense and that the current's slot
565 * offset is correct.
566 */
567 for (slot = 0; slot < nritems - 1; slot++) {
568 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
569 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
570
571 /* Make sure the keys are in the right order */
572 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
573 CORRUPT("bad key order", leaf, root, slot);
574 return -EIO;
575 }
576
577 /*
578 * Make sure the offset and ends are right, remember that the
579 * item data starts at the end of the leaf and grows towards the
580 * front.
581 */
582 if (btrfs_item_offset_nr(leaf, slot) !=
583 btrfs_item_end_nr(leaf, slot + 1)) {
584 CORRUPT("slot offset bad", leaf, root, slot);
585 return -EIO;
586 }
587
588 /*
589 * Check to make sure that we don't point outside of the leaf,
590 * just incase all the items are consistent to eachother, but
591 * all point outside of the leaf.
592 */
593 if (btrfs_item_end_nr(leaf, slot) >
594 BTRFS_LEAF_DATA_SIZE(root)) {
595 CORRUPT("slot end outside of leaf", leaf, root, slot);
596 return -EIO;
597 }
598 }
599
600 return 0;
601}
602
Miao Xiefacc8a222013-07-25 19:22:34 +0800603static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
604 u64 phy_offset, struct page *page,
605 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400606{
Chris Masonce9adaa2008-04-09 16:28:12 -0400607 u64 found_start;
608 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400609 struct extent_buffer *eb;
610 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400611 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400612 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400613
Chris Masonce9adaa2008-04-09 16:28:12 -0400614 if (!page->private)
615 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500616
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500617 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500618
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400619 /* the pending IO might have been the only thing that kept this buffer
620 * in memory. Make sure we have a ref for all this other checks
621 */
622 extent_buffer_get(eb);
623
624 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400625 if (!reads_done)
626 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400627
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400628 eb->read_mirror = mirror;
Filipe Manana656f30d2014-09-26 12:25:56 +0100629 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
Josef Bacikea466792012-03-26 21:57:36 -0400630 ret = -EIO;
631 goto err;
632 }
633
Chris Masonce9adaa2008-04-09 16:28:12 -0400634 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400635 if (found_start != eb->start) {
Wang Shilong29549ae2014-07-04 17:59:06 +0800636 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400637 "%llu %llu\n",
Wang Shilong29549ae2014-07-04 17:59:06 +0800638 eb->fs_info->sb->s_id, found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400639 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400640 goto err;
641 }
Yan Zheng2b820322008-11-17 21:11:30 -0500642 if (check_tree_block_fsid(root, eb)) {
Wang Shilong29549ae2014-07-04 17:59:06 +0800643 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad fsid on block %llu\n",
644 eb->fs_info->sb->s_id, eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400645 ret = -EIO;
646 goto err;
647 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400648 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400649 if (found_level >= BTRFS_MAX_LEVEL) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500650 btrfs_info(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400651 (int)btrfs_header_level(eb));
652 ret = -EIO;
653 goto err;
654 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400655
Chris Mason85d4e462011-07-26 16:11:19 -0400656 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
657 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500658
Chris Masonce9adaa2008-04-09 16:28:12 -0400659 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400660 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400661 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400662 goto err;
663 }
664
665 /*
666 * If this is a leaf block and it is corrupt, set the corrupt bit so
667 * that we don't try and read the other copies of this block, just
668 * return -EIO.
669 */
670 if (found_level == 0 && check_leaf(root, eb)) {
671 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
672 ret = -EIO;
673 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400674
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400675 if (!ret)
676 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400677err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400678 if (reads_done &&
679 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200680 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200681
David Woodhouse53b381b2013-01-29 18:40:14 -0500682 if (ret) {
683 /*
684 * our io error hook is going to dec the io pages
685 * again, we have to make sure it has something
686 * to decrement
687 */
688 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400689 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500690 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400691 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400692out:
Chris Masonf1885912008-04-09 16:28:12 -0400693 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400694}
695
Josef Bacikea466792012-03-26 21:57:36 -0400696static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200697{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200698 struct extent_buffer *eb;
699 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
700
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500701 eb = (struct extent_buffer *)page->private;
Filipe Manana656f30d2014-09-26 12:25:56 +0100702 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400703 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500704 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400705 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200706 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200707 return -EIO; /* we fixed nothing */
708}
709
Chris Masonce9adaa2008-04-09 16:28:12 -0400710static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400711{
David Sterba97eb6b62014-07-30 00:55:42 +0200712 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400713 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800714 struct btrfs_workqueue *wq;
715 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400716
Chris Masonce9adaa2008-04-09 16:28:12 -0400717 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400718 end_io_wq->error = err;
Chris Masond20f7042008-12-08 16:58:54 -0500719
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200720 if (bio->bi_rw & REQ_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800721 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
722 wq = fs_info->endio_meta_write_workers;
723 func = btrfs_endio_meta_write_helper;
724 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
725 wq = fs_info->endio_freespace_worker;
726 func = btrfs_freespace_write_helper;
727 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
728 wq = fs_info->endio_raid56_workers;
729 func = btrfs_endio_raid56_helper;
730 } else {
731 wq = fs_info->endio_write_workers;
732 func = btrfs_endio_write_helper;
733 }
Chris Masond20f7042008-12-08 16:58:54 -0500734 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800735 if (unlikely(end_io_wq->metadata ==
736 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
737 wq = fs_info->endio_repair_workers;
738 func = btrfs_endio_repair_helper;
739 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800740 wq = fs_info->endio_raid56_workers;
741 func = btrfs_endio_raid56_helper;
742 } else if (end_io_wq->metadata) {
743 wq = fs_info->endio_meta_workers;
744 func = btrfs_endio_meta_helper;
745 } else {
746 wq = fs_info->endio_workers;
747 func = btrfs_endio_helper;
748 }
Chris Masond20f7042008-12-08 16:58:54 -0500749 }
Liu Bo9e0af232014-08-15 23:36:53 +0800750
751 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
752 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400753}
754
Chris Mason22c59942008-04-09 16:28:12 -0400755int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200756 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400757{
David Sterba97eb6b62014-07-30 00:55:42 +0200758 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800759
David Sterba97eb6b62014-07-30 00:55:42 +0200760 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400761 if (!end_io_wq)
762 return -ENOMEM;
763
764 end_io_wq->private = bio->bi_private;
765 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400766 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400767 end_io_wq->error = 0;
768 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400769 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400770
771 bio->bi_private = end_io_wq;
772 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400773 return 0;
774}
775
Chris Masonb64a2852008-08-20 13:39:41 -0400776unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400777{
778 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800779 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400780 info->fs_devices->open_devices);
781 return 256 * limit;
782}
783
Chris Mason4a69a412008-11-06 22:03:00 -0500784static void run_one_async_start(struct btrfs_work *work)
785{
Chris Mason4a69a412008-11-06 22:03:00 -0500786 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100787 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500788
789 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100790 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
791 async->mirror_num, async->bio_flags,
792 async->bio_offset);
793 if (ret)
794 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500795}
796
797static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400798{
799 struct btrfs_fs_info *fs_info;
800 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400801 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400802
803 async = container_of(work, struct async_submit_bio, work);
804 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400805
Chris Masonb64a2852008-08-20 13:39:41 -0400806 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400807 limit = limit * 2 / 3;
808
Josef Bacik66657b32012-08-01 15:36:24 -0400809 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400810 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400811 wake_up(&fs_info->async_submit_wait);
812
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100813 /* If an error occured we just want to clean up the bio and move on */
814 if (async->error) {
815 bio_endio(async->bio, async->error);
816 return;
817 }
818
Chris Mason4a69a412008-11-06 22:03:00 -0500819 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400820 async->mirror_num, async->bio_flags,
821 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500822}
823
824static void run_one_async_free(struct btrfs_work *work)
825{
826 struct async_submit_bio *async;
827
828 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400829 kfree(async);
830}
831
Chris Mason44b8bd72008-04-16 11:14:51 -0400832int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
833 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400834 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400835 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500836 extent_submit_bio_hook_t *submit_bio_start,
837 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400838{
839 struct async_submit_bio *async;
840
841 async = kmalloc(sizeof(*async), GFP_NOFS);
842 if (!async)
843 return -ENOMEM;
844
845 async->inode = inode;
846 async->rw = rw;
847 async->bio = bio;
848 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500849 async->submit_bio_start = submit_bio_start;
850 async->submit_bio_done = submit_bio_done;
851
Liu Bo9e0af232014-08-15 23:36:53 +0800852 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800853 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500854
Chris Masonc8b97812008-10-29 14:49:59 -0400855 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400856 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400857
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100858 async->error = 0;
859
Chris Masoncb03c742008-05-15 16:15:45 -0400860 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400861
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200862 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800863 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400864
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800865 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400866
Chris Masond3977122009-01-05 21:25:51 -0500867 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500868 atomic_read(&fs_info->nr_async_submits)) {
869 wait_event(fs_info->async_submit_wait,
870 (atomic_read(&fs_info->nr_async_submits) == 0));
871 }
872
Chris Mason44b8bd72008-04-16 11:14:51 -0400873 return 0;
874}
875
Chris Masonce3ed712008-09-23 13:14:12 -0400876static int btree_csum_one_bio(struct bio *bio)
877{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800878 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400879 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800880 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400881
Kent Overstreet2c30c712013-11-07 12:20:26 -0800882 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400883 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100884 ret = csum_dirty_buffer(root, bvec->bv_page);
885 if (ret)
886 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400887 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800888
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100889 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400890}
891
Chris Mason4a69a412008-11-06 22:03:00 -0500892static int __btree_submit_bio_start(struct inode *inode, int rw,
893 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400894 unsigned long bio_flags,
895 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400896{
Chris Mason8b712842008-06-11 16:50:36 -0400897 /*
898 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400899 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400900 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100901 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500902}
Chris Mason22c59942008-04-09 16:28:12 -0400903
Chris Mason4a69a412008-11-06 22:03:00 -0500904static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400905 int mirror_num, unsigned long bio_flags,
906 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500907{
Stefan Behrens61891922012-11-05 18:51:52 +0100908 int ret;
909
Chris Mason8b712842008-06-11 16:50:36 -0400910 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500911 * when we're called for a write, we're already in the async
912 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400913 */
Stefan Behrens61891922012-11-05 18:51:52 +0100914 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
915 if (ret)
916 bio_endio(bio, ret);
917 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400918}
919
Josef Bacikde0022b2012-09-25 14:25:58 -0400920static int check_async_write(struct inode *inode, unsigned long bio_flags)
921{
922 if (bio_flags & EXTENT_BIO_TREE_LOG)
923 return 0;
924#ifdef CONFIG_X86
925 if (cpu_has_xmm4_2)
926 return 0;
927#endif
928 return 1;
929}
930
Chris Mason44b8bd72008-04-16 11:14:51 -0400931static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400932 int mirror_num, unsigned long bio_flags,
933 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400934{
Josef Bacikde0022b2012-09-25 14:25:58 -0400935 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500936 int ret;
937
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200938 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500939 /*
940 * called for a read, do the setup so that checksum validation
941 * can happen in the async kernel threads
942 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400943 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +0200944 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -0400945 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100946 goto out_w_error;
947 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
948 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400949 } else if (!async) {
950 ret = btree_csum_one_bio(bio);
951 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100952 goto out_w_error;
953 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
954 mirror_num, 0);
955 } else {
956 /*
957 * kthread helpers are used to submit writes so that
958 * checksumming can happen in parallel across all CPUs
959 */
960 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
961 inode, rw, bio, mirror_num, 0,
962 bio_offset,
963 __btree_submit_bio_start,
964 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400965 }
Chris Masond313d7a2009-04-20 15:50:09 -0400966
Stefan Behrens61891922012-11-05 18:51:52 +0100967 if (ret) {
968out_w_error:
969 bio_endio(bio, ret);
970 }
971 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400972}
973
Jan Beulich3dd14622010-12-07 14:54:09 +0000974#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500975static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800976 struct page *newpage, struct page *page,
977 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500978{
979 /*
980 * we can't safely write a btree page from here,
981 * we haven't done the locking hook
982 */
983 if (PageDirty(page))
984 return -EAGAIN;
985 /*
986 * Buffers may be managed in a filesystem specific way.
987 * We must have no buffers or drop them.
988 */
989 if (page_has_private(page) &&
990 !try_to_release_page(page, GFP_KERNEL))
991 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800992 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500993}
Jan Beulich3dd14622010-12-07 14:54:09 +0000994#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500995
Chris Mason0da54682007-11-07 21:08:01 -0500996
997static int btree_writepages(struct address_space *mapping,
998 struct writeback_control *wbc)
999{
Miao Xiee2d84522013-01-29 10:09:20 +00001000 struct btrfs_fs_info *fs_info;
1001 int ret;
1002
Chris Masond8d5f3e2007-12-11 12:42:00 -05001003 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001004
1005 if (wbc->for_kupdate)
1006 return 0;
1007
Miao Xiee2d84522013-01-29 10:09:20 +00001008 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001009 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001010 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1011 BTRFS_DIRTY_METADATA_THRESH);
1012 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001013 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001014 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001015 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001016}
1017
Christoph Hellwigb2950862008-12-02 09:54:17 -05001018static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001019{
Chris Masond1310b22008-01-24 16:13:08 -05001020 struct extent_io_tree *tree;
1021 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001022 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001023}
1024
Chris Mason70dec802008-01-29 09:59:12 -05001025static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001026{
Chris Mason98509cf2008-09-11 15:51:43 -04001027 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001028 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001029
David Sterbaf7a52a42013-04-26 14:56:29 +00001030 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001031}
Chris Mason123abc82007-03-14 14:14:43 -04001032
Lukas Czernerd47992f2013-05-21 23:17:23 -04001033static void btree_invalidatepage(struct page *page, unsigned int offset,
1034 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001035{
Chris Masond1310b22008-01-24 16:13:08 -05001036 struct extent_io_tree *tree;
1037 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001038 extent_invalidatepage(tree, page, offset);
1039 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001040 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001041 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1042 "page private not zero on page %llu",
1043 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001044 ClearPagePrivate(page);
1045 set_page_private(page, 0);
1046 page_cache_release(page);
1047 }
Chris Masond98237b2007-03-28 13:57:48 -04001048}
1049
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001050static int btree_set_page_dirty(struct page *page)
1051{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001052#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001053 struct extent_buffer *eb;
1054
1055 BUG_ON(!PagePrivate(page));
1056 eb = (struct extent_buffer *)page->private;
1057 BUG_ON(!eb);
1058 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1059 BUG_ON(!atomic_read(&eb->refs));
1060 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001061#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001062 return __set_page_dirty_nobuffers(page);
1063}
1064
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001065static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001066 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001067 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001068 .releasepage = btree_releasepage,
1069 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001070#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001071 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001072#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001073 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001074};
1075
David Sterba6197d862014-06-15 00:49:36 +02001076void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
Chris Mason090d1872007-05-01 08:53:32 -04001077{
Chris Mason5f39d392007-10-15 16:14:19 -04001078 struct extent_buffer *buf = NULL;
1079 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason090d1872007-05-01 08:53:32 -04001080
Chris Masondb945352007-10-15 16:15:53 -04001081 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001082 if (!buf)
David Sterba6197d862014-06-15 00:49:36 +02001083 return;
Chris Masond1310b22008-01-24 16:13:08 -05001084 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001085 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001086 free_extent_buffer(buf);
Chris Mason090d1872007-05-01 08:53:32 -04001087}
1088
Arne Jansenab0fff02011-05-23 14:25:41 +02001089int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1090 int mirror_num, struct extent_buffer **eb)
1091{
1092 struct extent_buffer *buf = NULL;
1093 struct inode *btree_inode = root->fs_info->btree_inode;
1094 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1095 int ret;
1096
1097 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1098 if (!buf)
1099 return 0;
1100
1101 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1102
1103 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1104 btree_get_extent, mirror_num);
1105 if (ret) {
1106 free_extent_buffer(buf);
1107 return ret;
1108 }
1109
1110 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1111 free_extent_buffer(buf);
1112 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001113 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001114 *eb = buf;
1115 } else {
1116 free_extent_buffer(buf);
1117 }
1118 return 0;
1119}
1120
Chris Mason0999df52008-04-01 13:48:14 -04001121struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
David Sterba0308af42014-06-15 01:43:40 +02001122 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001123{
Josef Bacikf28491e2013-12-16 13:24:27 -05001124 return find_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001125}
1126
1127struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1128 u64 bytenr, u32 blocksize)
1129{
David Sterbafccb84c2014-09-29 23:53:21 +02001130 if (btrfs_test_is_dummy_root(root))
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001131 return alloc_test_extent_buffer(root->fs_info, bytenr,
1132 blocksize);
Josef Bacikf28491e2013-12-16 13:24:27 -05001133 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001134}
1135
1136
Chris Masone02119d2008-09-05 16:13:11 -04001137int btrfs_write_tree_block(struct extent_buffer *buf)
1138{
Chris Mason727011e2010-08-06 13:21:20 -04001139 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001140 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001141}
1142
1143int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1144{
Chris Mason727011e2010-08-06 13:21:20 -04001145 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001146 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001147}
1148
Chris Masondb945352007-10-15 16:15:53 -04001149struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
David Sterbace86cd52014-06-15 01:07:32 +02001150 u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001151{
Chris Mason5f39d392007-10-15 16:14:19 -04001152 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001153 int ret;
1154
David Sterbace86cd52014-06-15 01:07:32 +02001155 buf = btrfs_find_create_tree_block(root, bytenr, root->nodesize);
Chris Mason5f39d392007-10-15 16:14:19 -04001156 if (!buf)
1157 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001158
Chris Masonca7a79a2008-05-12 12:59:19 -04001159 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001160 if (ret) {
1161 free_extent_buffer(buf);
1162 return NULL;
1163 }
Chris Mason5f39d392007-10-15 16:14:19 -04001164 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001165
Chris Masoneb60cea2007-02-02 09:18:22 -05001166}
1167
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001168void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1169 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001170{
Miao Xiee2d84522013-01-29 10:09:20 +00001171 struct btrfs_fs_info *fs_info = root->fs_info;
1172
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 Bacik58176a92007-08-29 15:47:34 -04001278 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001279 if (fs_info)
1280 root->defrag_trans_start = fs_info->generation;
1281 else
1282 root->defrag_trans_start = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001283 init_completion(&root->kobj_unregister);
Chris Mason4d775672007-04-20 20:23:12 -04001284 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001285 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001286
Anand Jain5f3ab902012-12-07 09:28:54 +00001287 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001288}
1289
Al Virof84a8bd2011-11-17 15:57:57 -05001290static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001291{
1292 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1293 if (root)
1294 root->fs_info = fs_info;
1295 return root;
1296}
1297
Josef Bacik06ea65a2013-09-19 16:07:01 -04001298#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1299/* Should only be used by the testing infrastructure */
1300struct btrfs_root *btrfs_alloc_dummy_root(void)
1301{
1302 struct btrfs_root *root;
1303
1304 root = btrfs_alloc_root(NULL);
1305 if (!root)
1306 return ERR_PTR(-ENOMEM);
David Sterba707e8a02014-06-04 19:22:26 +02001307 __setup_root(4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001308 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001309 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001310
1311 return root;
1312}
1313#endif
1314
Arne Jansen20897f52011-09-13 12:44:20 +02001315struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1316 struct btrfs_fs_info *fs_info,
1317 u64 objectid)
1318{
1319 struct extent_buffer *leaf;
1320 struct btrfs_root *tree_root = fs_info->tree_root;
1321 struct btrfs_root *root;
1322 struct btrfs_key key;
1323 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001324 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001325
1326 root = btrfs_alloc_root(fs_info);
1327 if (!root)
1328 return ERR_PTR(-ENOMEM);
1329
David Sterba707e8a02014-06-04 19:22:26 +02001330 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1331 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001332 root->root_key.objectid = objectid;
1333 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1334 root->root_key.offset = 0;
1335
David Sterba4d75f8a2014-06-15 01:54:12 +02001336 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Arne Jansen20897f52011-09-13 12:44:20 +02001337 if (IS_ERR(leaf)) {
1338 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001339 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001340 goto fail;
1341 }
1342
Arne Jansen20897f52011-09-13 12:44:20 +02001343 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1344 btrfs_set_header_bytenr(leaf, leaf->start);
1345 btrfs_set_header_generation(leaf, trans->transid);
1346 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1347 btrfs_set_header_owner(leaf, objectid);
1348 root->node = leaf;
1349
Ross Kirk0a4e5582013-09-24 10:12:38 +01001350 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001351 BTRFS_FSID_SIZE);
1352 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001353 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001354 BTRFS_UUID_SIZE);
1355 btrfs_mark_buffer_dirty(leaf);
1356
1357 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001358 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001359
1360 root->root_item.flags = 0;
1361 root->root_item.byte_limit = 0;
1362 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1363 btrfs_set_root_generation(&root->root_item, trans->transid);
1364 btrfs_set_root_level(&root->root_item, 0);
1365 btrfs_set_root_refs(&root->root_item, 1);
1366 btrfs_set_root_used(&root->root_item, leaf->len);
1367 btrfs_set_root_last_snapshot(&root->root_item, 0);
1368 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001369 uuid_le_gen(&uuid);
1370 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001371 root->root_item.drop_level = 0;
1372
1373 key.objectid = objectid;
1374 key.type = BTRFS_ROOT_ITEM_KEY;
1375 key.offset = 0;
1376 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1377 if (ret)
1378 goto fail;
1379
1380 btrfs_tree_unlock(leaf);
1381
Arne Jansen20897f52011-09-13 12:44:20 +02001382 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001383
1384fail:
1385 if (leaf) {
1386 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001387 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001388 free_extent_buffer(leaf);
1389 }
1390 kfree(root);
1391
1392 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001393}
1394
Yan Zheng7237f182009-01-21 12:54:03 -05001395static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1396 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001397{
1398 struct btrfs_root *root;
1399 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001400 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001401
Al Viro6f07e422011-11-17 00:46:16 -05001402 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001403 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001404 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001405
David Sterba707e8a02014-06-04 19:22:26 +02001406 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1407 tree_root->stripesize, root, fs_info,
1408 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001409
1410 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1411 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1412 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001413
Yan Zheng7237f182009-01-21 12:54:03 -05001414 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001415 * DON'T set REF_COWS for log trees
1416 *
Yan Zheng7237f182009-01-21 12:54:03 -05001417 * log trees do not get reference counted because they go away
1418 * before a real commit is actually done. They do store pointers
1419 * to file data extents, and those reference counts still get
1420 * updated (along with back refs to the log tree).
1421 */
Chris Masone02119d2008-09-05 16:13:11 -04001422
David Sterba4d75f8a2014-06-15 01:54:12 +02001423 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1424 NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001425 if (IS_ERR(leaf)) {
1426 kfree(root);
1427 return ERR_CAST(leaf);
1428 }
Chris Masone02119d2008-09-05 16:13:11 -04001429
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001430 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1431 btrfs_set_header_bytenr(leaf, leaf->start);
1432 btrfs_set_header_generation(leaf, trans->transid);
1433 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1434 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001435 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001436
1437 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001438 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001439 btrfs_mark_buffer_dirty(root->node);
1440 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001441 return root;
1442}
1443
1444int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1445 struct btrfs_fs_info *fs_info)
1446{
1447 struct btrfs_root *log_root;
1448
1449 log_root = alloc_log_tree(trans, fs_info);
1450 if (IS_ERR(log_root))
1451 return PTR_ERR(log_root);
1452 WARN_ON(fs_info->log_root_tree);
1453 fs_info->log_root_tree = log_root;
1454 return 0;
1455}
1456
1457int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1458 struct btrfs_root *root)
1459{
1460 struct btrfs_root *log_root;
1461 struct btrfs_inode_item *inode_item;
1462
1463 log_root = alloc_log_tree(trans, root->fs_info);
1464 if (IS_ERR(log_root))
1465 return PTR_ERR(log_root);
1466
1467 log_root->last_trans = trans->transid;
1468 log_root->root_key.offset = root->root_key.objectid;
1469
1470 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001471 btrfs_set_stack_inode_generation(inode_item, 1);
1472 btrfs_set_stack_inode_size(inode_item, 3);
1473 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001474 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001475 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f182009-01-21 12:54:03 -05001476
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001477 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001478
1479 WARN_ON(root->log_root);
1480 root->log_root = log_root;
1481 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001482 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001483 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001484 return 0;
1485}
1486
Stefan Behrens35a36212013-08-14 18:12:25 +02001487static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1488 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001489{
1490 struct btrfs_root *root;
1491 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001492 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001493 u64 generation;
Miao Xiecb517ea2013-05-15 07:48:19 +00001494 int ret;
1495
1496 path = btrfs_alloc_path();
1497 if (!path)
1498 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001499
Al Viro6f07e422011-11-17 00:46:16 -05001500 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001501 if (!root) {
1502 ret = -ENOMEM;
1503 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001504 }
1505
David Sterba707e8a02014-06-04 19:22:26 +02001506 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1507 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001508
Miao Xiecb517ea2013-05-15 07:48:19 +00001509 ret = btrfs_find_root(tree_root, key, path,
1510 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001511 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001512 if (ret > 0)
1513 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001514 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001515 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001516
Yan Zheng84234f32008-10-29 14:49:05 -04001517 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001518 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
David Sterbace86cd52014-06-15 01:07:32 +02001519 generation);
Miao Xiecb517ea2013-05-15 07:48:19 +00001520 if (!root->node) {
1521 ret = -ENOMEM;
1522 goto find_fail;
1523 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1524 ret = -EIO;
1525 goto read_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001526 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001527 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001528out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001529 btrfs_free_path(path);
1530 return root;
1531
1532read_fail:
1533 free_extent_buffer(root->node);
1534find_fail:
1535 kfree(root);
1536alloc_fail:
1537 root = ERR_PTR(ret);
1538 goto out;
1539}
1540
1541struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1542 struct btrfs_key *location)
1543{
1544 struct btrfs_root *root;
1545
1546 root = btrfs_read_tree_root(tree_root, location);
1547 if (IS_ERR(root))
1548 return root;
1549
1550 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001551 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001552 btrfs_check_and_init_root_item(&root->root_item);
1553 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001554
Chris Mason5eda7b52007-06-22 14:16:25 -04001555 return root;
1556}
1557
Miao Xiecb517ea2013-05-15 07:48:19 +00001558int btrfs_init_fs_root(struct btrfs_root *root)
1559{
1560 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001561 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001562
1563 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1564 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1565 GFP_NOFS);
1566 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1567 ret = -ENOMEM;
1568 goto fail;
1569 }
1570
Miao Xie8257b2d2014-03-06 13:38:19 +08001571 writers = btrfs_alloc_subvolume_writers();
1572 if (IS_ERR(writers)) {
1573 ret = PTR_ERR(writers);
1574 goto fail;
1575 }
1576 root->subv_writers = writers;
1577
Miao Xiecb517ea2013-05-15 07:48:19 +00001578 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001579 spin_lock_init(&root->ino_cache_lock);
1580 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001581
1582 ret = get_anon_bdev(&root->anon_dev);
1583 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001584 goto free_writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001585 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001586
1587free_writers:
1588 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001589fail:
1590 kfree(root->free_ino_ctl);
1591 kfree(root->free_ino_pinned);
1592 return ret;
1593}
1594
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001595static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1596 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001597{
1598 struct btrfs_root *root;
1599
1600 spin_lock(&fs_info->fs_roots_radix_lock);
1601 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1602 (unsigned long)root_id);
1603 spin_unlock(&fs_info->fs_roots_radix_lock);
1604 return root;
1605}
1606
1607int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1608 struct btrfs_root *root)
1609{
1610 int ret;
1611
1612 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1613 if (ret)
1614 return ret;
1615
1616 spin_lock(&fs_info->fs_roots_radix_lock);
1617 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1618 (unsigned long)root->root_key.objectid,
1619 root);
1620 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001621 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001622 spin_unlock(&fs_info->fs_roots_radix_lock);
1623 radix_tree_preload_end();
1624
1625 return ret;
1626}
1627
Miao Xiec00869f2013-09-25 21:47:44 +08001628struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1629 struct btrfs_key *location,
1630 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001631{
1632 struct btrfs_root *root;
1633 int ret;
1634
Chris Masonedbd8d42007-12-21 16:27:24 -05001635 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1636 return fs_info->tree_root;
1637 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1638 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001639 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1640 return fs_info->chunk_root;
1641 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1642 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001643 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1644 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001645 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1646 return fs_info->quota_root ? fs_info->quota_root :
1647 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001648 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1649 return fs_info->uuid_root ? fs_info->uuid_root :
1650 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001651again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001652 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001653 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001654 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001655 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001656 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001657 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001658
Miao Xiecb517ea2013-05-15 07:48:19 +00001659 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001660 if (IS_ERR(root))
1661 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001662
Miao Xiec00869f2013-09-25 21:47:44 +08001663 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001664 ret = -ENOENT;
1665 goto fail;
1666 }
1667
Miao Xiecb517ea2013-05-15 07:48:19 +00001668 ret = btrfs_init_fs_root(root);
1669 if (ret)
1670 goto fail;
1671
Kelley Nielsen3f870c22013-11-04 19:37:39 -08001672 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1673 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001674 if (ret < 0)
1675 goto fail;
1676 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001677 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001678
Miao Xiecb517ea2013-05-15 07:48:19 +00001679 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001680 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001681 if (ret == -EEXIST) {
1682 free_fs_root(root);
1683 goto again;
1684 }
1685 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001686 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001687 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001688fail:
1689 free_fs_root(root);
1690 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001691}
1692
Chris Mason04160082008-03-26 10:28:07 -04001693static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1694{
1695 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1696 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001697 struct btrfs_device *device;
1698 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001699
Xiao Guangrong1f781602011-04-20 10:09:16 +00001700 rcu_read_lock();
1701 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001702 if (!device->bdev)
1703 continue;
Chris Mason04160082008-03-26 10:28:07 -04001704 bdi = blk_get_backing_dev_info(device->bdev);
Tejun Heoff9ea322014-09-08 08:03:56 +09001705 if (bdi_congested(bdi, bdi_bits)) {
Chris Mason04160082008-03-26 10:28:07 -04001706 ret = 1;
1707 break;
1708 }
1709 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001710 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001711 return ret;
1712}
1713
Chris Mason04160082008-03-26 10:28:07 -04001714static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1715{
Jens Axboead081f12009-06-12 14:43:40 +02001716 int err;
1717
1718 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001719 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001720 if (err)
1721 return err;
1722
Chris Mason4575c9c2008-04-18 16:13:31 -04001723 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001724 bdi->congested_fn = btrfs_congested_fn;
1725 bdi->congested_data = info;
1726 return 0;
1727}
1728
Chris Mason8b712842008-06-11 16:50:36 -04001729/*
1730 * called by the kthread helper functions to finally call the bio end_io
1731 * functions. This is where read checksum verification actually happens
1732 */
1733static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001734{
Chris Masonce9adaa2008-04-09 16:28:12 -04001735 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001736 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001737 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001738
David Sterba97eb6b62014-07-30 00:55:42 +02001739 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001740 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001741
Chris Mason8b712842008-06-11 16:50:36 -04001742 error = end_io_wq->error;
1743 bio->bi_private = end_io_wq->private;
1744 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001745 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Kent Overstreetbc1e79a2013-12-03 13:24:08 -08001746 bio_endio_nodec(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001747}
1748
Chris Masona74a4b92008-06-25 16:01:31 -04001749static int cleaner_kthread(void *arg)
1750{
1751 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001752 int again;
Chris Masona74a4b92008-06-25 16:01:31 -04001753
1754 do {
Miao Xied0278242013-05-14 10:20:40 +00001755 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001756
Miao Xied0278242013-05-14 10:20:40 +00001757 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001758 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001759 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001760
Miao Xied0278242013-05-14 10:20:40 +00001761 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1762 goto sleep;
1763
Miao Xiedc7f3702013-05-14 10:20:42 +00001764 /*
1765 * Avoid the problem that we change the status of the fs
1766 * during the above check and trylock.
1767 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001768 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001769 mutex_unlock(&root->fs_info->cleaner_mutex);
1770 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001771 }
1772
Miao Xied0278242013-05-14 10:20:40 +00001773 btrfs_run_delayed_iputs(root);
Josef Bacik47ab2a62014-09-18 11:20:02 -04001774 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001775 again = btrfs_clean_one_deleted_snapshot(root);
1776 mutex_unlock(&root->fs_info->cleaner_mutex);
1777
1778 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001779 * The defragger has dealt with the R/O remount and umount,
1780 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001781 */
1782 btrfs_run_defrag_inodes(root->fs_info);
1783sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001784 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001785 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001786 if (!kthread_should_stop())
1787 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001788 __set_current_state(TASK_RUNNING);
1789 }
1790 } while (!kthread_should_stop());
1791 return 0;
1792}
1793
1794static int transaction_kthread(void *arg)
1795{
1796 struct btrfs_root *root = arg;
1797 struct btrfs_trans_handle *trans;
1798 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001799 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001800 unsigned long now;
1801 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001802 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001803
1804 do {
Jan Kara914b2002012-03-12 16:05:50 +01001805 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001806 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001807 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1808
Josef Bacika4abeea2011-04-11 17:25:13 -04001809 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001810 cur = root->fs_info->running_transaction;
1811 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001812 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001813 goto sleep;
1814 }
Yan Zheng31153d82008-07-28 15:32:19 -04001815
Chris Masona74a4b92008-06-25 16:01:31 -04001816 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001817 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001818 (now < cur->start_time ||
1819 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001820 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001821 delay = HZ * 5;
1822 goto sleep;
1823 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001824 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001825 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001826
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001827 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001828 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001829 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001830 if (PTR_ERR(trans) != -ENOENT)
1831 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001832 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001833 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001834 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001835 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001836 } else {
1837 btrfs_end_transaction(trans, root);
1838 }
Chris Masona74a4b92008-06-25 16:01:31 -04001839sleep:
1840 wake_up_process(root->fs_info->cleaner_kthread);
1841 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1842
Josef Bacik4e121c02013-09-27 16:32:39 -04001843 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1844 &root->fs_info->fs_state)))
1845 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001846 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001847 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001848 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001849 (!btrfs_transaction_blocked(root->fs_info) ||
1850 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001851 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001852 __set_current_state(TASK_RUNNING);
1853 }
1854 } while (!kthread_should_stop());
1855 return 0;
1856}
1857
Chris Masonaf31f5e2011-11-03 15:17:42 -04001858/*
1859 * this will find the highest generation in the array of
1860 * root backups. The index of the highest array is returned,
1861 * or -1 if we can't find anything.
1862 *
1863 * We check to make sure the array is valid by comparing the
1864 * generation of the latest root in the array with the generation
1865 * in the super block. If they don't match we pitch it.
1866 */
1867static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1868{
1869 u64 cur;
1870 int newest_index = -1;
1871 struct btrfs_root_backup *root_backup;
1872 int i;
1873
1874 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1875 root_backup = info->super_copy->super_roots + i;
1876 cur = btrfs_backup_tree_root_gen(root_backup);
1877 if (cur == newest_gen)
1878 newest_index = i;
1879 }
1880
1881 /* check to see if we actually wrapped around */
1882 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1883 root_backup = info->super_copy->super_roots;
1884 cur = btrfs_backup_tree_root_gen(root_backup);
1885 if (cur == newest_gen)
1886 newest_index = 0;
1887 }
1888 return newest_index;
1889}
1890
1891
1892/*
1893 * find the oldest backup so we know where to store new entries
1894 * in the backup array. This will set the backup_root_index
1895 * field in the fs_info struct
1896 */
1897static void find_oldest_super_backup(struct btrfs_fs_info *info,
1898 u64 newest_gen)
1899{
1900 int newest_index = -1;
1901
1902 newest_index = find_newest_super_backup(info, newest_gen);
1903 /* if there was garbage in there, just move along */
1904 if (newest_index == -1) {
1905 info->backup_root_index = 0;
1906 } else {
1907 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1908 }
1909}
1910
1911/*
1912 * copy all the root pointers into the super backup array.
1913 * this will bump the backup pointer by one when it is
1914 * done
1915 */
1916static void backup_super_roots(struct btrfs_fs_info *info)
1917{
1918 int next_backup;
1919 struct btrfs_root_backup *root_backup;
1920 int last_backup;
1921
1922 next_backup = info->backup_root_index;
1923 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1924 BTRFS_NUM_BACKUP_ROOTS;
1925
1926 /*
1927 * just overwrite the last backup if we're at the same generation
1928 * this happens only at umount
1929 */
1930 root_backup = info->super_for_commit->super_roots + last_backup;
1931 if (btrfs_backup_tree_root_gen(root_backup) ==
1932 btrfs_header_generation(info->tree_root->node))
1933 next_backup = last_backup;
1934
1935 root_backup = info->super_for_commit->super_roots + next_backup;
1936
1937 /*
1938 * make sure all of our padding and empty slots get zero filled
1939 * regardless of which ones we use today
1940 */
1941 memset(root_backup, 0, sizeof(*root_backup));
1942
1943 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1944
1945 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1946 btrfs_set_backup_tree_root_gen(root_backup,
1947 btrfs_header_generation(info->tree_root->node));
1948
1949 btrfs_set_backup_tree_root_level(root_backup,
1950 btrfs_header_level(info->tree_root->node));
1951
1952 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1953 btrfs_set_backup_chunk_root_gen(root_backup,
1954 btrfs_header_generation(info->chunk_root->node));
1955 btrfs_set_backup_chunk_root_level(root_backup,
1956 btrfs_header_level(info->chunk_root->node));
1957
1958 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1959 btrfs_set_backup_extent_root_gen(root_backup,
1960 btrfs_header_generation(info->extent_root->node));
1961 btrfs_set_backup_extent_root_level(root_backup,
1962 btrfs_header_level(info->extent_root->node));
1963
Chris Mason7c7e82a2011-11-06 18:50:56 -05001964 /*
1965 * we might commit during log recovery, which happens before we set
1966 * the fs_root. Make sure it is valid before we fill it in.
1967 */
1968 if (info->fs_root && info->fs_root->node) {
1969 btrfs_set_backup_fs_root(root_backup,
1970 info->fs_root->node->start);
1971 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001972 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001973 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001974 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001975 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001976
1977 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1978 btrfs_set_backup_dev_root_gen(root_backup,
1979 btrfs_header_generation(info->dev_root->node));
1980 btrfs_set_backup_dev_root_level(root_backup,
1981 btrfs_header_level(info->dev_root->node));
1982
1983 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1984 btrfs_set_backup_csum_root_gen(root_backup,
1985 btrfs_header_generation(info->csum_root->node));
1986 btrfs_set_backup_csum_root_level(root_backup,
1987 btrfs_header_level(info->csum_root->node));
1988
1989 btrfs_set_backup_total_bytes(root_backup,
1990 btrfs_super_total_bytes(info->super_copy));
1991 btrfs_set_backup_bytes_used(root_backup,
1992 btrfs_super_bytes_used(info->super_copy));
1993 btrfs_set_backup_num_devices(root_backup,
1994 btrfs_super_num_devices(info->super_copy));
1995
1996 /*
1997 * if we don't copy this out to the super_copy, it won't get remembered
1998 * for the next commit
1999 */
2000 memcpy(&info->super_copy->super_roots,
2001 &info->super_for_commit->super_roots,
2002 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2003}
2004
2005/*
2006 * this copies info out of the root backup array and back into
2007 * the in-memory super block. It is meant to help iterate through
2008 * the array, so you send it the number of backups you've already
2009 * tried and the last backup index you used.
2010 *
2011 * this returns -1 when it has tried all the backups
2012 */
2013static noinline int next_root_backup(struct btrfs_fs_info *info,
2014 struct btrfs_super_block *super,
2015 int *num_backups_tried, int *backup_index)
2016{
2017 struct btrfs_root_backup *root_backup;
2018 int newest = *backup_index;
2019
2020 if (*num_backups_tried == 0) {
2021 u64 gen = btrfs_super_generation(super);
2022
2023 newest = find_newest_super_backup(info, gen);
2024 if (newest == -1)
2025 return -1;
2026
2027 *backup_index = newest;
2028 *num_backups_tried = 1;
2029 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2030 /* we've tried all the backups, all done */
2031 return -1;
2032 } else {
2033 /* jump to the next oldest backup */
2034 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2035 BTRFS_NUM_BACKUP_ROOTS;
2036 *backup_index = newest;
2037 *num_backups_tried += 1;
2038 }
2039 root_backup = super->super_roots + newest;
2040
2041 btrfs_set_super_generation(super,
2042 btrfs_backup_tree_root_gen(root_backup));
2043 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2044 btrfs_set_super_root_level(super,
2045 btrfs_backup_tree_root_level(root_backup));
2046 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2047
2048 /*
2049 * fixme: the total bytes and num_devices need to match or we should
2050 * need a fsck
2051 */
2052 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2053 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2054 return 0;
2055}
2056
Liu Bo7abadb62013-03-17 02:10:31 +00002057/* helper to cleanup workers */
2058static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2059{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002060 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002061 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002062 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002063 btrfs_destroy_workqueue(fs_info->endio_workers);
2064 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2065 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002066 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002067 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002068 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2069 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2070 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002071 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002072 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002073 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002074 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002075 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002076 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002077 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002078}
2079
Rashika2e9f5952013-10-31 02:45:20 +05302080static void free_root_extent_buffers(struct btrfs_root *root)
2081{
2082 if (root) {
2083 free_extent_buffer(root->node);
2084 free_extent_buffer(root->commit_root);
2085 root->node = NULL;
2086 root->commit_root = NULL;
2087 }
2088}
2089
Chris Masonaf31f5e2011-11-03 15:17:42 -04002090/* helper to cleanup tree roots */
2091static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2092{
Rashika2e9f5952013-10-31 02:45:20 +05302093 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002094
Rashika2e9f5952013-10-31 02:45:20 +05302095 free_root_extent_buffers(info->dev_root);
2096 free_root_extent_buffers(info->extent_root);
2097 free_root_extent_buffers(info->csum_root);
2098 free_root_extent_buffers(info->quota_root);
2099 free_root_extent_buffers(info->uuid_root);
2100 if (chunk_root)
2101 free_root_extent_buffers(info->chunk_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002102}
2103
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002104void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002105{
2106 int ret;
2107 struct btrfs_root *gang[8];
2108 int i;
2109
2110 while (!list_empty(&fs_info->dead_roots)) {
2111 gang[0] = list_entry(fs_info->dead_roots.next,
2112 struct btrfs_root, root_list);
2113 list_del(&gang[0]->root_list);
2114
Miao Xie27cdeb72014-04-02 19:51:05 +08002115 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002116 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002117 } else {
2118 free_extent_buffer(gang[0]->node);
2119 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002120 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002121 }
2122 }
2123
2124 while (1) {
2125 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2126 (void **)gang, 0,
2127 ARRAY_SIZE(gang));
2128 if (!ret)
2129 break;
2130 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002131 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002132 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002133
2134 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2135 btrfs_free_log_root_tree(NULL, fs_info);
2136 btrfs_destroy_pinned_extent(fs_info->tree_root,
2137 fs_info->pinned_extents);
2138 }
Josef Bacik171f6532013-04-24 16:35:41 -04002139}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002140
Al Viroad2b2c82011-11-17 01:10:02 -05002141int open_ctree(struct super_block *sb,
2142 struct btrfs_fs_devices *fs_devices,
2143 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002144{
Chris Masondb945352007-10-15 16:15:53 -04002145 u32 sectorsize;
2146 u32 nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002147 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002148 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002149 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002150 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002151 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002152 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002153 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002154 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002155 struct btrfs_root *extent_root;
2156 struct btrfs_root *csum_root;
2157 struct btrfs_root *chunk_root;
2158 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002159 struct btrfs_root *quota_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002160 struct btrfs_root *uuid_root;
Chris Masone02119d2008-09-05 16:13:11 -04002161 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002162 int ret;
Yane58ca022008-04-01 11:21:34 -04002163 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002164 int num_backups_tried = 0;
2165 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002166 int max_active;
2167 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Stefan Behrens70f80172013-08-15 17:11:23 +02002168 bool create_uuid_tree;
2169 bool check_uuid_tree;
Chris Mason4543df72008-06-11 21:47:56 -04002170
Al Virof84a8bd2011-11-17 15:57:57 -05002171 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002172 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002173 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002174 err = -ENOMEM;
2175 goto fail;
2176 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002177
2178 ret = init_srcu_struct(&fs_info->subvol_srcu);
2179 if (ret) {
2180 err = ret;
2181 goto fail;
2182 }
2183
2184 ret = setup_bdi(fs_info, &fs_info->bdi);
2185 if (ret) {
2186 err = ret;
2187 goto fail_srcu;
2188 }
2189
Tejun Heo908c7f12014-09-08 09:51:29 +09002190 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
Miao Xiee2d84522013-01-29 10:09:20 +00002191 if (ret) {
2192 err = ret;
2193 goto fail_bdi;
2194 }
2195 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2196 (1 + ilog2(nr_cpu_ids));
2197
Tejun Heo908c7f12014-09-08 09:51:29 +09002198 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
Miao Xie963d6782013-01-29 10:10:51 +00002199 if (ret) {
2200 err = ret;
2201 goto fail_dirty_metadata_bytes;
2202 }
2203
Tejun Heo908c7f12014-09-08 09:51:29 +09002204 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
Miao Xiec404e0d2014-01-30 16:46:55 +08002205 if (ret) {
2206 err = ret;
2207 goto fail_delalloc_bytes;
2208 }
2209
Yan, Zheng76dda932009-09-21 16:00:26 -04002210 fs_info->btree_inode = new_inode(sb);
2211 if (!fs_info->btree_inode) {
2212 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002213 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002214 }
2215
Chris Masona6591712011-07-19 12:04:14 -04002216 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002217
Yan, Zheng76dda932009-09-21 16:00:26 -04002218 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002219 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002220 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002221 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002222 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002223 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002224 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002225 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002226 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002227 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002228 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002229 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002230 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002231 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002232 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002233 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002234 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002235 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002236 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002237 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002238 mutex_init(&fs_info->delalloc_root_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002239 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002240
Josef Bacik58176a92007-08-29 15:47:34 -04002241 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002242 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002243 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002244 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002245 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002246 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002247 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2248 BTRFS_BLOCK_RSV_GLOBAL);
2249 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2250 BTRFS_BLOCK_RSV_DELALLOC);
2251 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2252 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2253 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2254 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2255 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002256 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002257 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002258 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002259 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002260 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002261 atomic_set(&fs_info->qgroup_op_seq, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002262 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002263 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002264 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002265 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002266 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002267 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002268 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002269 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
Josef Bacik0a2b2a82014-01-23 10:54:11 -05002270 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
Arne Jansen90519d62011-05-23 14:30:00 +02002271 /* readahead state */
2272 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2273 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002274
Chris Masonb34b0862008-12-19 15:43:22 -05002275 fs_info->thread_pool_size = min_t(unsigned long,
2276 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002277
Miao Xie199c2a92013-05-15 07:48:23 +00002278 INIT_LIST_HEAD(&fs_info->ordered_roots);
2279 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002280 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2281 GFP_NOFS);
2282 if (!fs_info->delayed_root) {
2283 err = -ENOMEM;
2284 goto fail_iput;
2285 }
2286 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002287
Arne Jansena2de7332011-03-08 14:14:00 +01002288 mutex_init(&fs_info->scrub_lock);
2289 atomic_set(&fs_info->scrubs_running, 0);
2290 atomic_set(&fs_info->scrub_pause_req, 0);
2291 atomic_set(&fs_info->scrubs_paused, 0);
2292 atomic_set(&fs_info->scrub_cancel_req, 0);
Miao Xiec404e0d2014-01-30 16:46:55 +08002293 init_waitqueue_head(&fs_info->replace_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002294 init_waitqueue_head(&fs_info->scrub_pause_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002295 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002296#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2297 fs_info->check_integrity_print_mask = 0;
2298#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002299
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002300 spin_lock_init(&fs_info->balance_lock);
2301 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002302 atomic_set(&fs_info->balance_running, 0);
2303 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002304 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002305 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002306 init_waitqueue_head(&fs_info->balance_wait_q);
Miao Xie21c7e752014-05-13 17:29:04 -07002307 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Chris Masona061fc82008-05-07 11:43:44 -04002308
Chris Mason0afbaf82008-04-18 14:17:20 -04002309 sb->s_blocksize = 4096;
2310 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002311 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002312
Yan, Zheng76dda932009-09-21 16:00:26 -04002313 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002314 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002315 /*
2316 * we set the i_size on the btree inode to the max possible int.
2317 * the real end of the address space is determined by all of
2318 * the devices in the system
2319 */
2320 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002321 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002322 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2323
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002324 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002325 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002326 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002327 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002328 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002329
Chris Masond1310b22008-01-24 16:13:08 -05002330 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2331
Chris Mason0f7d52f2007-04-09 10:42:37 -04002332 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2333 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2334 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002335 set_bit(BTRFS_INODE_DUMMY,
2336 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Filipe David Borba Manana778ba822013-10-06 22:22:33 +01002337 btrfs_insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002338
Chris Masone02119d2008-09-05 16:13:11 -04002339 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002340 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002341 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002342
Yan Zheng11833d62009-09-11 16:11:19 -04002343 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002344 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002345 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002346 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002347 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002348 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002349
Chris Masond98237b2007-03-28 13:57:48 -04002350
Chris Mason5a3f23d2009-03-31 13:27:11 -04002351 mutex_init(&fs_info->ordered_operations_mutex);
Josef Bacik9ffba8c2013-08-14 11:33:56 -04002352 mutex_init(&fs_info->ordered_extent_flush_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002353 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002354 mutex_init(&fs_info->chunk_mutex);
2355 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002356 mutex_init(&fs_info->cleaner_mutex);
2357 mutex_init(&fs_info->volume_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002358 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002359 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002360 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002361 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Stefan Behrense922e082012-11-05 17:26:40 +01002362 fs_info->dev_replace.lock_owner = 0;
2363 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2364 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2365 mutex_init(&fs_info->dev_replace.lock_management_lock);
2366 mutex_init(&fs_info->dev_replace.lock);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002367
Arne Jansen416ac512011-09-13 12:56:09 +02002368 spin_lock_init(&fs_info->qgroup_lock);
Wang Shilongf2f6ed32013-04-07 10:50:16 +00002369 mutex_init(&fs_info->qgroup_ioctl_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002370 fs_info->qgroup_tree = RB_ROOT;
Josef Bacikfcebe452014-05-13 17:30:47 -07002371 fs_info->qgroup_op_tree = RB_ROOT;
Arne Jansen416ac512011-09-13 12:56:09 +02002372 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2373 fs_info->qgroup_seq = 1;
2374 fs_info->quota_enabled = 0;
2375 fs_info->pending_quota_state = 0;
Wang Shilong1e8f9152013-05-06 11:03:27 +00002376 fs_info->qgroup_ulist = NULL;
Jan Schmidt2f232032013-04-25 16:04:51 +00002377 mutex_init(&fs_info->qgroup_rescan_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002378
Chris Masonfa9c0d792009-04-03 09:47:43 -04002379 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2380 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2381
Chris Mason7d9eb122008-07-08 14:19:17 -04002382 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002383 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002384 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002385 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002386
David Woodhouse53b381b2013-01-29 18:40:14 -05002387 ret = btrfs_alloc_stripe_hash_table(fs_info);
2388 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002389 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002390 goto fail_alloc;
2391 }
2392
David Sterba707e8a02014-06-04 19:22:26 +02002393 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002394 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002395
Chris Mason3c4bb262012-03-27 18:56:56 -04002396 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002397
2398 /*
2399 * Read super block and check the signature bytes only
2400 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002401 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002402 if (!bh) {
2403 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002404 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002405 }
Chris Mason39279cc2007-06-12 06:35:45 -04002406
David Sterba1104a882013-03-06 15:57:46 +01002407 /*
2408 * We want to check superblock checksum, the type is stored inside.
2409 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2410 */
2411 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002412 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002413 err = -EINVAL;
2414 goto fail_alloc;
2415 }
2416
2417 /*
2418 * super_copy is zeroed at allocation time and we never touch the
2419 * following bytes up to INFO_SIZE, the checksum is calculated from
2420 * the whole block of INFO_SIZE
2421 */
David Sterba6c417612011-04-13 15:41:04 +02002422 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2423 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2424 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002425 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002426
David Sterba6c417612011-04-13 15:41:04 +02002427 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002428
David Sterba1104a882013-03-06 15:57:46 +01002429 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2430 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002431 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002432 err = -EINVAL;
2433 goto fail_alloc;
2434 }
2435
David Sterba6c417612011-04-13 15:41:04 +02002436 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002437 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002438 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002439
liuboacce9522011-01-06 19:30:25 +08002440 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002441 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2442 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002443
Liu Bo75e7cb72011-03-22 10:12:20 +00002444 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002445 * run through our array of backup supers and setup
2446 * our ring pointer to the oldest one
2447 */
2448 generation = btrfs_super_generation(disk_super);
2449 find_oldest_super_backup(fs_info, generation);
2450
2451 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002452 * In the long term, we'll store the compression type in the super
2453 * block, and it'll be used for per file compression control.
2454 */
2455 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2456
Yan Zheng2b820322008-11-17 21:11:30 -05002457 ret = btrfs_parse_options(tree_root, options);
2458 if (ret) {
2459 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002460 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002461 }
Chris Masondfe25022008-05-13 13:46:40 -04002462
Josef Bacikf2b636e2008-12-02 06:36:08 -05002463 features = btrfs_super_incompat_flags(disk_super) &
2464 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2465 if (features) {
2466 printk(KERN_ERR "BTRFS: couldn't mount because of "
2467 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002468 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002469 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002470 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002471 }
2472
David Sterba707e8a02014-06-04 19:22:26 +02002473 /*
2474 * Leafsize and nodesize were always equal, this is only a sanity check.
2475 */
2476 if (le32_to_cpu(disk_super->__unused_leafsize) !=
Chris Mason727011e2010-08-06 13:21:20 -04002477 btrfs_super_nodesize(disk_super)) {
2478 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2479 "blocksizes don't match. node %d leaf %d\n",
2480 btrfs_super_nodesize(disk_super),
David Sterba707e8a02014-06-04 19:22:26 +02002481 le32_to_cpu(disk_super->__unused_leafsize));
Chris Mason727011e2010-08-06 13:21:20 -04002482 err = -EINVAL;
2483 goto fail_alloc;
2484 }
David Sterba707e8a02014-06-04 19:22:26 +02002485 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002486 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2487 "blocksize (%d) was too large\n",
David Sterba707e8a02014-06-04 19:22:26 +02002488 btrfs_super_nodesize(disk_super));
Chris Mason727011e2010-08-06 13:21:20 -04002489 err = -EINVAL;
2490 goto fail_alloc;
2491 }
2492
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002493 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002494 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002495 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002496 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002497
Josef Bacik3173a182013-03-07 14:22:04 -05002498 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
Frank Holtonefe120a2013-12-20 11:37:06 -05002499 printk(KERN_ERR "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002500
Chris Mason727011e2010-08-06 13:21:20 -04002501 /*
2502 * flag our filesystem as having big metadata blocks if
2503 * they are bigger than the page size
2504 */
David Sterba707e8a02014-06-04 19:22:26 +02002505 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002506 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002507 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002508 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2509 }
2510
Chris Masonbc3f1162012-03-29 17:02:47 -04002511 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002512 sectorsize = btrfs_super_sectorsize(disk_super);
2513 stripesize = btrfs_super_stripesize(disk_super);
David Sterba707e8a02014-06-04 19:22:26 +02002514 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002515 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002516
2517 /*
2518 * mixed block groups end up with duplicate but slightly offset
2519 * extent buffers for the same range. It leads to corruptions
2520 */
2521 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002522 (sectorsize != nodesize)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002523 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002524 "are not allowed for mixed block groups on %s\n",
2525 sb->s_id);
2526 goto fail_alloc;
2527 }
2528
Miao Xieceda0862013-04-11 10:30:16 +00002529 /*
2530 * Needn't use the lock because there is no other task which will
2531 * update the flag.
2532 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002533 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002534
Josef Bacikf2b636e2008-12-02 06:36:08 -05002535 features = btrfs_super_compat_ro_flags(disk_super) &
2536 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2537 if (!(sb->s_flags & MS_RDONLY) && features) {
2538 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2539 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002540 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002541 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002542 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002543 }
Chris Mason61d92c32009-10-02 19:11:56 -04002544
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002545 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002546
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002547 fs_info->workers =
2548 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2549 max_active, 16);
Chris Masonc8b97812008-10-29 14:49:59 -04002550
Qu Wenruoafe3d242014-02-28 10:46:07 +08002551 fs_info->delalloc_workers =
2552 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
Chris Mason771ed682008-11-06 22:02:51 -05002553
Qu Wenruoa44903a2014-02-28 10:46:09 +08002554 fs_info->flush_workers =
2555 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
Miao Xie8ccf6f192012-10-25 09:28:04 +00002556
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002557 fs_info->caching_workers =
2558 btrfs_alloc_workqueue("cache", flags, max_active, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002559
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002560 /*
2561 * a higher idle thresh on the submit workers makes it much more
Chris Mason61b49442008-07-31 15:42:53 -04002562 * likely that bios will be send down in a sane order to the
2563 * devices
2564 */
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002565 fs_info->submit_workers =
2566 btrfs_alloc_workqueue("submit", flags,
2567 min_t(u64, fs_devices->num_devices,
2568 max_active), 64);
Chris Mason53863232008-08-15 15:34:18 -04002569
Qu Wenruodc6e3202014-02-28 10:46:14 +08002570 fs_info->fixup_workers =
2571 btrfs_alloc_workqueue("fixup", flags, 1, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002572
2573 /*
2574 * endios are largely parallel and should have a very
2575 * low idle thresh
2576 */
Qu Wenruofccb5d82014-02-28 10:46:10 +08002577 fs_info->endio_workers =
2578 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2579 fs_info->endio_meta_workers =
2580 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2581 fs_info->endio_meta_write_workers =
2582 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2583 fs_info->endio_raid56_workers =
2584 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
Miao Xie8b110e32014-09-12 18:44:03 +08002585 fs_info->endio_repair_workers =
2586 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002587 fs_info->rmw_workers =
2588 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002589 fs_info->endio_write_workers =
2590 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2591 fs_info->endio_freespace_worker =
2592 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002593 fs_info->delayed_workers =
2594 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002595 fs_info->readahead_workers =
2596 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002597 fs_info->qgroup_rescan_workers =
2598 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
Chris Masona79b7d42014-05-22 16:18:52 -07002599 fs_info->extent_workers =
2600 btrfs_alloc_workqueue("extent-refs", flags,
2601 min_t(u64, fs_devices->num_devices,
2602 max_active), 8);
Chris Masonb51912c2009-02-04 09:23:24 -05002603
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002604 if (!(fs_info->workers && fs_info->delalloc_workers &&
Qu Wenruofccb5d82014-02-28 10:46:10 +08002605 fs_info->submit_workers && fs_info->flush_workers &&
2606 fs_info->endio_workers && fs_info->endio_meta_workers &&
2607 fs_info->endio_meta_write_workers &&
Miao Xie8b110e32014-09-12 18:44:03 +08002608 fs_info->endio_repair_workers &&
Qu Wenruofccb5d82014-02-28 10:46:10 +08002609 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002610 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
Qu Wenruodc6e3202014-02-28 10:46:14 +08002611 fs_info->caching_workers && fs_info->readahead_workers &&
Qu Wenruofc97fab2014-02-28 10:46:16 +08002612 fs_info->fixup_workers && fs_info->delayed_workers &&
Andrey Utkin56094ee2014-08-09 14:51:15 +03002613 fs_info->extent_workers &&
Qu Wenruofc97fab2014-02-28 10:46:16 +08002614 fs_info->qgroup_rescan_workers)) {
Ilya Dryomovfed425c2012-06-22 12:13:01 -06002615 err = -ENOMEM;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002616 goto fail_sb_buffer;
2617 }
Chris Mason4543df72008-06-11 21:47:56 -04002618
Chris Mason4575c9c2008-04-18 16:13:31 -04002619 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002620 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2621 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002622
Chris Masondb945352007-10-15 16:15:53 -04002623 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002624 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002625 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002626
2627 sb->s_blocksize = sectorsize;
2628 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002629
Qu Wenruo3cae2102013-07-16 11:19:18 +08002630 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002631 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002632 goto fail_sb_buffer;
2633 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002634
Liu Bo8d082fb2012-04-03 09:56:53 +08002635 if (sectorsize != PAGE_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002636 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
Liu Bo8d082fb2012-04-03 09:56:53 +08002637 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002638 goto fail_sb_buffer;
2639 }
2640
Chris Mason925baed2008-06-25 16:01:30 -04002641 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002642 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002643 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002644 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002645 printk(KERN_WARNING "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002646 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002647 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002648 }
Chris Mason0b86a832008-03-24 15:01:56 -04002649
Yan Zheng84234f32008-10-29 14:49:05 -04002650 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002651
David Sterba707e8a02014-06-04 19:22:26 +02002652 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2653 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002654
2655 chunk_root->node = read_tree_block(chunk_root,
2656 btrfs_super_chunk_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002657 generation);
Josef Bacik416bc652013-04-23 14:17:42 -04002658 if (!chunk_root->node ||
2659 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002660 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002661 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002662 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002663 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002664 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2665 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002666
Chris Masone17cade2008-04-15 15:41:47 -04002667 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002668 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002669
Chris Mason0b86a832008-03-24 15:01:56 -04002670 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002671 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002672 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002673 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002674 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002675 }
Chris Mason0b86a832008-03-24 15:01:56 -04002676
Stefan Behrens8dabb742012-11-06 13:15:27 +01002677 /*
2678 * keep the device that is marked to be the target device for the
2679 * dev_replace procedure
2680 */
2681 btrfs_close_extra_devices(fs_info, fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002682
Chris Masona6b0d5c2012-02-20 20:53:43 -05002683 if (!fs_devices->latest_bdev) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002684 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002685 sb->s_id);
2686 goto fail_tree_roots;
2687 }
2688
Chris Masonaf31f5e2011-11-03 15:17:42 -04002689retry_root_backup:
Yan Zheng84234f32008-10-29 14:49:05 -04002690 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002691
Chris Masone20d96d2007-03-22 12:13:20 -04002692 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002693 btrfs_super_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002694 generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002695 if (!tree_root->node ||
2696 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002697 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002698 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002699
2700 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002701 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002702
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002703 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2704 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002705 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002706
Miao Xiecb517ea2013-05-15 07:48:19 +00002707 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2708 location.type = BTRFS_ROOT_ITEM_KEY;
2709 location.offset = 0;
2710
2711 extent_root = btrfs_read_tree_root(tree_root, &location);
2712 if (IS_ERR(extent_root)) {
2713 ret = PTR_ERR(extent_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002714 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002715 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002716 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002717 fs_info->extent_root = extent_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002718
Miao Xiecb517ea2013-05-15 07:48:19 +00002719 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2720 dev_root = btrfs_read_tree_root(tree_root, &location);
2721 if (IS_ERR(dev_root)) {
2722 ret = PTR_ERR(dev_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002723 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002724 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002725 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002726 fs_info->dev_root = dev_root;
2727 btrfs_init_devices_late(fs_info);
Chris Mason3768f362007-03-13 16:47:54 -04002728
Miao Xiecb517ea2013-05-15 07:48:19 +00002729 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2730 csum_root = btrfs_read_tree_root(tree_root, &location);
2731 if (IS_ERR(csum_root)) {
2732 ret = PTR_ERR(csum_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002733 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002734 }
Miao Xie27cdeb72014-04-02 19:51:05 +08002735 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00002736 fs_info->csum_root = csum_root;
Chris Masond20f7042008-12-08 16:58:54 -05002737
Miao Xiecb517ea2013-05-15 07:48:19 +00002738 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2739 quota_root = btrfs_read_tree_root(tree_root, &location);
2740 if (!IS_ERR(quota_root)) {
Miao Xie27cdeb72014-04-02 19:51:05 +08002741 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002742 fs_info->quota_enabled = 1;
2743 fs_info->pending_quota_state = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002744 fs_info->quota_root = quota_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002745 }
2746
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002747 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2748 uuid_root = btrfs_read_tree_root(tree_root, &location);
2749 if (IS_ERR(uuid_root)) {
2750 ret = PTR_ERR(uuid_root);
2751 if (ret != -ENOENT)
2752 goto recovery_tree_root;
2753 create_uuid_tree = true;
Stefan Behrens70f80172013-08-15 17:11:23 +02002754 check_uuid_tree = false;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002755 } else {
Miao Xie27cdeb72014-04-02 19:51:05 +08002756 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002757 fs_info->uuid_root = uuid_root;
Stefan Behrens70f80172013-08-15 17:11:23 +02002758 create_uuid_tree = false;
2759 check_uuid_tree =
2760 generation != btrfs_super_uuid_tree_generation(disk_super);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002761 }
2762
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002763 fs_info->generation = generation;
2764 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002765
Ilya Dryomov68310a52012-06-22 12:24:12 -06002766 ret = btrfs_recover_balance(fs_info);
2767 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002768 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002769 goto fail_block_groups;
2770 }
2771
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002772 ret = btrfs_init_dev_stats(fs_info);
2773 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002774 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002775 ret);
2776 goto fail_block_groups;
2777 }
2778
Stefan Behrens8dabb742012-11-06 13:15:27 +01002779 ret = btrfs_init_dev_replace(fs_info);
2780 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002781 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002782 goto fail_block_groups;
2783 }
2784
2785 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2786
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002787 ret = btrfs_sysfs_add_one(fs_info);
2788 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002789 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002790 goto fail_block_groups;
2791 }
2792
liuboc59021f2011-03-07 02:13:14 +00002793 ret = btrfs_init_space_info(fs_info);
2794 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002795 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002796 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002797 }
2798
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002799 ret = btrfs_read_block_groups(extent_root);
2800 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002801 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002802 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002803 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002804 fs_info->num_tolerated_disk_barrier_failures =
2805 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002806 if (fs_info->fs_devices->missing_devices >
2807 fs_info->num_tolerated_disk_barrier_failures &&
2808 !(sb->s_flags & MS_RDONLY)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002809 printk(KERN_WARNING "BTRFS: "
2810 "too many missing devices, writeable mount is not allowed\n");
Anand Jain2365dd32014-01-22 11:15:51 +08002811 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002812 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002813
Chris Masona74a4b92008-06-25 16:01:31 -04002814 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2815 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002816 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08002817 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04002818
2819 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2820 tree_root,
2821 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002822 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002823 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002824
Chris Masonc2898112009-06-10 09:51:32 -04002825 if (!btrfs_test_opt(tree_root, SSD) &&
2826 !btrfs_test_opt(tree_root, NOSSD) &&
2827 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002828 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04002829 "mode\n");
2830 btrfs_set_opt(fs_info->mount_opt, SSD);
2831 }
2832
Qu Wenruo3818aea2014-01-13 13:36:06 +08002833 /* Set the real inode map cache flag */
2834 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2835 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2836
Stefan Behrens21adbd52011-11-09 13:44:05 +01002837#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2838 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2839 ret = btrfsic_mount(tree_root, fs_devices,
2840 btrfs_test_opt(tree_root,
2841 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2842 1 : 0,
2843 fs_info->check_integrity_print_mask);
2844 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05002845 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01002846 " integrity check module %s\n", sb->s_id);
2847 }
2848#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002849 ret = btrfs_read_qgroup_config(fs_info);
2850 if (ret)
2851 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002852
liuboacce9522011-01-06 19:30:25 +08002853 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002854 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002855 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002856
Chris Mason7c2ca462008-11-19 15:13:35 -05002857 if (fs_devices->rw_devices == 0) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002858 printk(KERN_WARNING "BTRFS: log replay required "
Chris Masond3977122009-01-05 21:25:51 -05002859 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002860 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002861 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002862 }
Chris Masone02119d2008-09-05 16:13:11 -04002863
Al Viro6f07e422011-11-17 00:46:16 -05002864 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002865 if (!log_tree_root) {
2866 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002867 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002868 }
Chris Masone02119d2008-09-05 16:13:11 -04002869
David Sterba707e8a02014-06-04 19:22:26 +02002870 __setup_root(nodesize, sectorsize, stripesize,
Chris Masone02119d2008-09-05 16:13:11 -04002871 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2872
2873 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002874 generation + 1);
Josef Bacik416bc652013-04-23 14:17:42 -04002875 if (!log_tree_root->node ||
2876 !extent_buffer_uptodate(log_tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002877 printk(KERN_ERR "BTRFS: failed to read log tree\n");
Josef Bacik416bc652013-04-23 14:17:42 -04002878 free_extent_buffer(log_tree_root->node);
2879 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002880 goto fail_qgroup;
Josef Bacik416bc652013-04-23 14:17:42 -04002881 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002882 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002883 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002884 if (ret) {
2885 btrfs_error(tree_root->fs_info, ret,
2886 "Failed to recover log tree");
2887 free_extent_buffer(log_tree_root->node);
2888 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002889 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002890 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002891
2892 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002893 ret = btrfs_commit_super(tree_root);
2894 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002895 goto fail_qgroup;
Yan Zhenge556ce22008-11-20 10:25:19 -05002896 }
Chris Masone02119d2008-09-05 16:13:11 -04002897 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002898
Yan, Zheng76dda932009-09-21 16:00:26 -04002899 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002900 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002901 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04002902
Chris Mason7c2ca462008-11-19 15:13:35 -05002903 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002904 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002905 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002906 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002907
Wang Shilong5f316482014-06-26 11:08:16 +08002908 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002909 ret = btrfs_recover_relocation(tree_root);
Wang Shilong5f316482014-06-26 11:08:16 +08002910 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00002911 if (ret < 0) {
2912 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05002913 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00002914 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002915 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002916 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002917 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002918
Chris Mason3de45862008-11-17 21:02:50 -05002919 location.objectid = BTRFS_FS_TREE_OBJECTID;
2920 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00002921 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05002922
Chris Mason3de45862008-11-17 21:02:50 -05002923 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002924 if (IS_ERR(fs_info->fs_root)) {
2925 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002926 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002927 }
Chris Masonc2898112009-06-10 09:51:32 -04002928
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002929 if (sb->s_flags & MS_RDONLY)
2930 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002931
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002932 down_read(&fs_info->cleanup_work_sem);
2933 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2934 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002935 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002936 close_ctree(tree_root);
2937 return ret;
2938 }
2939 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002940
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002941 ret = btrfs_resume_balance_async(fs_info);
2942 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002943 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002944 close_ctree(tree_root);
2945 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002946 }
2947
Stefan Behrens8dabb742012-11-06 13:15:27 +01002948 ret = btrfs_resume_dev_replace_async(fs_info);
2949 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002950 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01002951 close_ctree(tree_root);
2952 return ret;
2953 }
2954
Jan Schmidtb382a322013-05-28 15:47:24 +00002955 btrfs_qgroup_rescan_resume(fs_info);
2956
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002957 if (create_uuid_tree) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002958 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002959 ret = btrfs_create_uuid_tree(fs_info);
2960 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002961 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002962 ret);
2963 close_ctree(tree_root);
2964 return ret;
2965 }
Stefan Behrensf420ee12013-08-15 17:11:24 +02002966 } else if (check_uuid_tree ||
2967 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002968 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02002969 ret = btrfs_check_uuid_tree(fs_info);
2970 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002971 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02002972 ret);
2973 close_ctree(tree_root);
2974 return ret;
2975 }
2976 } else {
2977 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002978 }
2979
Josef Bacik47ab2a62014-09-18 11:20:02 -04002980 fs_info->open = 1;
2981
Al Viroad2b2c82011-11-17 01:10:02 -05002982 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002983
Arne Jansenbcef60f2011-09-13 15:23:30 +02002984fail_qgroup:
2985 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05002986fail_trans_kthread:
2987 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04002988 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002989 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04002990fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002991 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002992
2993 /*
2994 * make sure we're done with the btree inode before we stop our
2995 * kthreads
2996 */
2997 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05002998
Anand Jain2365dd32014-01-22 11:15:51 +08002999fail_sysfs:
3000 btrfs_sysfs_remove_one(fs_info);
3001
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003002fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003003 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003004 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003005
3006fail_tree_roots:
3007 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003008 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003009
Chris Mason39279cc2007-06-12 06:35:45 -04003010fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003011 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003012fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003013fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003014 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3015
Chris Mason4543df72008-06-11 21:47:56 -04003016 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003017fail_bio_counter:
3018 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003019fail_delalloc_bytes:
3020 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003021fail_dirty_metadata_bytes:
3022 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003023fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003024 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003025fail_srcu:
3026 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003027fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003028 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003029 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003030 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003031
3032recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003033 if (!btrfs_test_opt(tree_root, RECOVERY))
3034 goto fail_tree_roots;
3035
3036 free_root_pointers(fs_info, 0);
3037
3038 /* don't use the log in recovery mode, it won't be valid */
3039 btrfs_set_super_log_root(disk_super, 0);
3040
3041 /* we can't trust the free space cache either */
3042 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3043
3044 ret = next_root_backup(fs_info, fs_info->super_copy,
3045 &num_backups_tried, &backup_index);
3046 if (ret == -1)
3047 goto fail_block_groups;
3048 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003049}
3050
Chris Masonf2984462008-04-10 16:19:33 -04003051static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3052{
Chris Masonf2984462008-04-10 16:19:33 -04003053 if (uptodate) {
3054 set_buffer_uptodate(bh);
3055 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003056 struct btrfs_device *device = (struct btrfs_device *)
3057 bh->b_private;
3058
Frank Holtonefe120a2013-12-20 11:37:06 -05003059 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
Josef Bacik606686e2012-06-04 14:03:51 -04003060 "I/O error on %s\n",
3061 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003062 /* note, we dont' set_buffer_write_io_error because we have
3063 * our own ways of dealing with the IO errors
3064 */
Chris Masonf2984462008-04-10 16:19:33 -04003065 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003066 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003067 }
3068 unlock_buffer(bh);
3069 put_bh(bh);
3070}
3071
Yan Zhenga512bbf2008-12-08 16:46:26 -05003072struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3073{
3074 struct buffer_head *bh;
3075 struct buffer_head *latest = NULL;
3076 struct btrfs_super_block *super;
3077 int i;
3078 u64 transid = 0;
3079 u64 bytenr;
3080
3081 /* we would like to check all the supers, but that would make
3082 * a btrfs mount succeed after a mkfs from a different FS.
3083 * So, we need to add a special mount option to scan for
3084 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3085 */
3086 for (i = 0; i < 1; i++) {
3087 bytenr = btrfs_sb_offset(i);
Anand Jain8068a472013-07-26 01:29:35 +08003088 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3089 i_size_read(bdev->bd_inode))
Yan Zhenga512bbf2008-12-08 16:46:26 -05003090 break;
Anand Jain8068a472013-07-26 01:29:35 +08003091 bh = __bread(bdev, bytenr / 4096,
3092 BTRFS_SUPER_INFO_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003093 if (!bh)
3094 continue;
3095
3096 super = (struct btrfs_super_block *)bh->b_data;
3097 if (btrfs_super_bytenr(super) != bytenr ||
Qu Wenruo3cae2102013-07-16 11:19:18 +08003098 btrfs_super_magic(super) != BTRFS_MAGIC) {
Yan Zhenga512bbf2008-12-08 16:46:26 -05003099 brelse(bh);
3100 continue;
3101 }
3102
3103 if (!latest || btrfs_super_generation(super) > transid) {
3104 brelse(latest);
3105 latest = bh;
3106 transid = btrfs_super_generation(super);
3107 } else {
3108 brelse(bh);
3109 }
3110 }
3111 return latest;
3112}
3113
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003114/*
3115 * this should be called twice, once with wait == 0 and
3116 * once with wait == 1. When wait == 0 is done, all the buffer heads
3117 * we write are pinned.
3118 *
3119 * They are released when wait == 1 is done.
3120 * max_mirrors must be the same for both runs, and it indicates how
3121 * many supers on this one device should be written.
3122 *
3123 * max_mirrors == 0 means to write them all.
3124 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003125static int write_dev_supers(struct btrfs_device *device,
3126 struct btrfs_super_block *sb,
3127 int do_barriers, int wait, int max_mirrors)
3128{
3129 struct buffer_head *bh;
3130 int i;
3131 int ret;
3132 int errors = 0;
3133 u32 crc;
3134 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003135
3136 if (max_mirrors == 0)
3137 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3138
Yan Zhenga512bbf2008-12-08 16:46:26 -05003139 for (i = 0; i < max_mirrors; i++) {
3140 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003141 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3142 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003143 break;
3144
3145 if (wait) {
3146 bh = __find_get_block(device->bdev, bytenr / 4096,
3147 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003148 if (!bh) {
3149 errors++;
3150 continue;
3151 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003152 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003153 if (!buffer_uptodate(bh))
3154 errors++;
3155
3156 /* drop our reference */
3157 brelse(bh);
3158
3159 /* drop the reference from the wait == 0 run */
3160 brelse(bh);
3161 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003162 } else {
3163 btrfs_set_super_bytenr(sb, bytenr);
3164
3165 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003166 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003167 BTRFS_CSUM_SIZE, crc,
3168 BTRFS_SUPER_INFO_SIZE -
3169 BTRFS_CSUM_SIZE);
3170 btrfs_csum_final(crc, sb->csum);
3171
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003172 /*
3173 * one reference for us, and we leave it for the
3174 * caller
3175 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003176 bh = __getblk(device->bdev, bytenr / 4096,
3177 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003178 if (!bh) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003179 printk(KERN_ERR "BTRFS: couldn't get super "
Josef Bacik634554d2013-04-29 10:05:57 -04003180 "buffer head for bytenr %Lu\n", bytenr);
3181 errors++;
3182 continue;
3183 }
3184
Yan Zhenga512bbf2008-12-08 16:46:26 -05003185 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3186
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003187 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003188 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003189
3190 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003191 lock_buffer(bh);
3192 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003193 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003194 }
3195
Chris Mason387125f2011-11-18 15:07:51 -05003196 /*
3197 * we fua the first super. The others we allow
3198 * to go down lazy.
3199 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003200 if (i == 0)
3201 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3202 else
3203 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003204 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003205 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003206 }
3207 return errors < i ? 0 : -1;
3208}
3209
Chris Mason387125f2011-11-18 15:07:51 -05003210/*
3211 * endio for the write_dev_flush, this will wake anyone waiting
3212 * for the barrier when it is done
3213 */
3214static void btrfs_end_empty_barrier(struct bio *bio, int err)
3215{
3216 if (err) {
3217 if (err == -EOPNOTSUPP)
3218 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3219 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3220 }
3221 if (bio->bi_private)
3222 complete(bio->bi_private);
3223 bio_put(bio);
3224}
3225
3226/*
3227 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3228 * sent down. With wait == 1, it waits for the previous flush.
3229 *
3230 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3231 * capable
3232 */
3233static int write_dev_flush(struct btrfs_device *device, int wait)
3234{
3235 struct bio *bio;
3236 int ret = 0;
3237
3238 if (device->nobarriers)
3239 return 0;
3240
3241 if (wait) {
3242 bio = device->flush_bio;
3243 if (!bio)
3244 return 0;
3245
3246 wait_for_completion(&device->flush_wait);
3247
3248 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003249 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
Josef Bacik606686e2012-06-04 14:03:51 -04003250 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05003251 device->nobarriers = 1;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003252 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
Chris Mason387125f2011-11-18 15:07:51 -05003253 ret = -EIO;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003254 btrfs_dev_stat_inc_and_print(device,
3255 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003256 }
3257
3258 /* drop the reference from the wait == 0 run */
3259 bio_put(bio);
3260 device->flush_bio = NULL;
3261
3262 return ret;
3263 }
3264
3265 /*
3266 * one reference for us, and we leave it for the
3267 * caller
3268 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003269 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003270 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003271 if (!bio)
3272 return -ENOMEM;
3273
3274 bio->bi_end_io = btrfs_end_empty_barrier;
3275 bio->bi_bdev = device->bdev;
3276 init_completion(&device->flush_wait);
3277 bio->bi_private = &device->flush_wait;
3278 device->flush_bio = bio;
3279
3280 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003281 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003282
3283 return 0;
3284}
3285
3286/*
3287 * send an empty flush down to each device in parallel,
3288 * then wait for them
3289 */
3290static int barrier_all_devices(struct btrfs_fs_info *info)
3291{
3292 struct list_head *head;
3293 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003294 int errors_send = 0;
3295 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003296 int ret;
3297
3298 /* send down all the barriers */
3299 head = &info->fs_devices->devices;
3300 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003301 if (dev->missing)
3302 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003303 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003304 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003305 continue;
3306 }
3307 if (!dev->in_fs_metadata || !dev->writeable)
3308 continue;
3309
3310 ret = write_dev_flush(dev, 0);
3311 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003312 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003313 }
3314
3315 /* wait for all the barriers */
3316 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003317 if (dev->missing)
3318 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003319 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003320 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003321 continue;
3322 }
3323 if (!dev->in_fs_metadata || !dev->writeable)
3324 continue;
3325
3326 ret = write_dev_flush(dev, 1);
3327 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003328 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003329 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003330 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3331 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003332 return -EIO;
3333 return 0;
3334}
3335
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003336int btrfs_calc_num_tolerated_disk_barrier_failures(
3337 struct btrfs_fs_info *fs_info)
3338{
3339 struct btrfs_ioctl_space_info space;
3340 struct btrfs_space_info *sinfo;
3341 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3342 BTRFS_BLOCK_GROUP_SYSTEM,
3343 BTRFS_BLOCK_GROUP_METADATA,
3344 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3345 int num_types = 4;
3346 int i;
3347 int c;
3348 int num_tolerated_disk_barrier_failures =
3349 (int)fs_info->fs_devices->num_devices;
3350
3351 for (i = 0; i < num_types; i++) {
3352 struct btrfs_space_info *tmp;
3353
3354 sinfo = NULL;
3355 rcu_read_lock();
3356 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3357 if (tmp->flags == types[i]) {
3358 sinfo = tmp;
3359 break;
3360 }
3361 }
3362 rcu_read_unlock();
3363
3364 if (!sinfo)
3365 continue;
3366
3367 down_read(&sinfo->groups_sem);
3368 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3369 if (!list_empty(&sinfo->block_groups[c])) {
3370 u64 flags;
3371
3372 btrfs_get_block_group_info(
3373 &sinfo->block_groups[c], &space);
3374 if (space.total_bytes == 0 ||
3375 space.used_bytes == 0)
3376 continue;
3377 flags = space.flags;
3378 /*
3379 * return
3380 * 0: if dup, single or RAID0 is configured for
3381 * any of metadata, system or data, else
3382 * 1: if RAID5 is configured, or if RAID1 or
3383 * RAID10 is configured and only two mirrors
3384 * are used, else
3385 * 2: if RAID6 is configured, else
3386 * num_mirrors - 1: if RAID1 or RAID10 is
3387 * configured and more than
3388 * 2 mirrors are used.
3389 */
3390 if (num_tolerated_disk_barrier_failures > 0 &&
3391 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3392 BTRFS_BLOCK_GROUP_RAID0)) ||
3393 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3394 == 0)))
3395 num_tolerated_disk_barrier_failures = 0;
David Woodhouse53b381b2013-01-29 18:40:14 -05003396 else if (num_tolerated_disk_barrier_failures > 1) {
3397 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3398 BTRFS_BLOCK_GROUP_RAID5 |
3399 BTRFS_BLOCK_GROUP_RAID10)) {
3400 num_tolerated_disk_barrier_failures = 1;
3401 } else if (flags &
Henrik Nordvik15b0a892013-04-29 18:09:23 +00003402 BTRFS_BLOCK_GROUP_RAID6) {
David Woodhouse53b381b2013-01-29 18:40:14 -05003403 num_tolerated_disk_barrier_failures = 2;
3404 }
3405 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003406 }
3407 }
3408 up_read(&sinfo->groups_sem);
3409 }
3410
3411 return num_tolerated_disk_barrier_failures;
3412}
3413
Eric Sandeen48a3b632013-04-25 20:41:01 +00003414static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003415{
Chris Masone5e9a522009-06-10 15:17:02 -04003416 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003417 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003418 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003419 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003420 int ret;
3421 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003422 int max_errors;
3423 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003424 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003425
3426 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003427 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003428
David Sterba6c417612011-04-13 15:41:04 +02003429 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003430 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003431
Chris Mason174ba502011-05-27 10:03:58 -04003432 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003433 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003434 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003435
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003436 if (do_barriers) {
3437 ret = barrier_all_devices(root->fs_info);
3438 if (ret) {
3439 mutex_unlock(
3440 &root->fs_info->fs_devices->device_list_mutex);
3441 btrfs_error(root->fs_info, ret,
3442 "errors while submitting device barriers.");
3443 return ret;
3444 }
3445 }
Chris Mason387125f2011-11-18 15:07:51 -05003446
Xiao Guangrong1f781602011-04-20 10:09:16 +00003447 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003448 if (!dev->bdev) {
3449 total_errors++;
3450 continue;
3451 }
Yan Zheng2b820322008-11-17 21:11:30 -05003452 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003453 continue;
3454
Yan Zheng2b820322008-11-17 21:11:30 -05003455 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003456 btrfs_set_stack_device_type(dev_item, dev->type);
3457 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003458 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003459 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003460 btrfs_set_stack_device_bytes_used(dev_item,
3461 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003462 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3463 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3464 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3465 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003466 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003467
Chris Masona061fc82008-05-07 11:43:44 -04003468 flags = btrfs_super_flags(sb);
3469 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003470
Yan Zhenga512bbf2008-12-08 16:46:26 -05003471 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003472 if (ret)
3473 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003474 }
Chris Masona236aed2008-04-29 09:38:00 -04003475 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003476 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003477 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003478 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003479
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003480 /* FUA is masked off if unsupported and can't be the reason */
3481 btrfs_error(root->fs_info, -EIO,
3482 "%d errors while writing supers", total_errors);
3483 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003484 }
Chris Masonf2984462008-04-10 16:19:33 -04003485
Yan Zhenga512bbf2008-12-08 16:46:26 -05003486 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003487 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003488 if (!dev->bdev)
3489 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003490 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003491 continue;
3492
Yan Zhenga512bbf2008-12-08 16:46:26 -05003493 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3494 if (ret)
3495 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003496 }
Chris Mason174ba502011-05-27 10:03:58 -04003497 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003498 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003499 btrfs_error(root->fs_info, -EIO,
3500 "%d errors while writing supers", total_errors);
3501 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003502 }
Chris Masonf2984462008-04-10 16:19:33 -04003503 return 0;
3504}
3505
Yan Zhenga512bbf2008-12-08 16:46:26 -05003506int write_ctree_super(struct btrfs_trans_handle *trans,
3507 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003508{
Rashikaf570e7572013-10-31 16:50:42 +05303509 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003510}
3511
Miao Xiecb517ea2013-05-15 07:48:19 +00003512/* Drop a fs root from the radix tree and free it. */
3513void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3514 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003515{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003516 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003517 radix_tree_delete(&fs_info->fs_roots_radix,
3518 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003519 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003520
3521 if (btrfs_root_refs(&root->root_item) == 0)
3522 synchronize_srcu(&fs_info->subvol_srcu);
3523
Liu Bo1a4319c2014-01-13 19:53:53 +08003524 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003525 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003526
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003527 if (root->free_ino_pinned)
3528 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3529 if (root->free_ino_ctl)
3530 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003531 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003532}
3533
3534static void free_fs_root(struct btrfs_root *root)
3535{
David Sterba57cdc8d2014-02-05 02:37:48 +01003536 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003537 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003538 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3539 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003540 if (root->anon_dev)
3541 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003542 if (root->subv_writers)
3543 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003544 free_extent_buffer(root->node);
3545 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003546 kfree(root->free_ino_ctl);
3547 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003548 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003549 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003550}
3551
Miao Xiecb517ea2013-05-15 07:48:19 +00003552void btrfs_free_fs_root(struct btrfs_root *root)
3553{
3554 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003555}
3556
Yan Zhengc146afa2008-11-12 14:34:12 -05003557int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3558{
3559 u64 root_objectid = 0;
3560 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003561 int i = 0;
3562 int err = 0;
3563 unsigned int ret = 0;
3564 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003565
3566 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003567 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003568 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3569 (void **)gang, root_objectid,
3570 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003571 if (!ret) {
3572 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003573 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003574 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003575 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003576
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003577 for (i = 0; i < ret; i++) {
3578 /* Avoid to grab roots in dead_roots */
3579 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3580 gang[i] = NULL;
3581 continue;
3582 }
3583 /* grab all the search result for later use */
3584 gang[i] = btrfs_grab_fs_root(gang[i]);
3585 }
3586 srcu_read_unlock(&fs_info->subvol_srcu, index);
3587
3588 for (i = 0; i < ret; i++) {
3589 if (!gang[i])
3590 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003591 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003592 err = btrfs_orphan_cleanup(gang[i]);
3593 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003594 break;
3595 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003596 }
3597 root_objectid++;
3598 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003599
3600 /* release the uncleaned roots due to error */
3601 for (; i < ret; i++) {
3602 if (gang[i])
3603 btrfs_put_fs_root(gang[i]);
3604 }
3605 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003606}
3607
3608int btrfs_commit_super(struct btrfs_root *root)
3609{
3610 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003611
3612 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003613 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003614 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003615 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003616
3617 /* wait until ongoing cleanup work done */
3618 down_write(&root->fs_info->cleanup_work_sem);
3619 up_write(&root->fs_info->cleanup_work_sem);
3620
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003621 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003622 if (IS_ERR(trans))
3623 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003624 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003625}
3626
David Sterba3abdbd72014-06-04 18:10:45 +02003627void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003628{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003629 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003630 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003631
Chris Masonfacda1e2007-06-08 18:11:48 -04003632 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003633 smp_mb();
3634
Stefan Behrens803b2f52013-08-15 17:11:21 +02003635 /* wait for the uuid_scan task to finish */
3636 down(&fs_info->uuid_tree_rescan_sem);
3637 /* avoid complains from lockdep et al., set sem back to initial state */
3638 up(&fs_info->uuid_tree_rescan_sem);
3639
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003640 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003641 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003642
Stefan Behrens8dabb742012-11-06 13:15:27 +01003643 btrfs_dev_replace_suspend_for_unmount(fs_info);
3644
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003645 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003646
3647 /* wait for any defraggers to finish */
3648 wait_event(fs_info->transaction_wait,
3649 (atomic_read(&fs_info->defrag_running) == 0));
3650
3651 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003652 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003653
Miao Xie21c7e752014-05-13 17:29:04 -07003654 cancel_work_sync(&fs_info->async_reclaim_work);
3655
Yan Zhengc146afa2008-11-12 14:34:12 -05003656 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003657 ret = btrfs_commit_super(root);
3658 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003659 btrfs_err(root->fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003660 }
3661
Miao Xie87533c42013-01-29 10:14:48 +00003662 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003663 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003664
Al Viroe3029d92011-11-17 00:56:18 -05003665 kthread_stop(fs_info->transaction_kthread);
3666 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003667
Yan Zhengf25784b2009-07-28 08:41:57 -04003668 fs_info->closing = 2;
3669 smp_mb();
3670
Arne Jansenbcef60f2011-09-13 15:23:30 +02003671 btrfs_free_qgroup_config(root->fs_info);
3672
Miao Xie963d6782013-01-29 10:10:51 +00003673 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003674 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003675 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003676 }
Yanbcc63ab2008-07-30 16:29:20 -04003677
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003678 btrfs_sysfs_remove_one(fs_info);
3679
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003680 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003681
Liu Bo1a4319c2014-01-13 19:53:53 +08003682 btrfs_put_block_group_cache(fs_info);
3683
Josef Bacik2b1360d2013-10-07 15:14:44 -04003684 btrfs_free_block_groups(fs_info);
3685
Wang Shilongde348ee2014-04-09 19:23:22 +08003686 /*
3687 * we must make sure there is not any read request to
3688 * submit after we stopping all workers.
3689 */
3690 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003691 btrfs_stop_all_workers(fs_info);
3692
Josef Bacik47ab2a62014-09-18 11:20:02 -04003693 fs_info->open = 0;
Josef Bacik13e6c372013-05-30 16:55:44 -04003694 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003695
Josef Bacik13e6c372013-05-30 16:55:44 -04003696 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003697
Stefan Behrens21adbd52011-11-09 13:44:05 +01003698#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3699 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3700 btrfsic_unmount(root, fs_info->fs_devices);
3701#endif
3702
Chris Masondfe25022008-05-13 13:46:40 -04003703 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003704 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003705
Miao Xiee2d84522013-01-29 10:09:20 +00003706 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003707 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003708 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003709 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003710 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003711
David Woodhouse53b381b2013-01-29 18:40:14 -05003712 btrfs_free_stripe_hash_table(fs_info);
3713
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003714 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3715 root->orphan_block_rsv = NULL;
Chris Masoneb60cea2007-02-02 09:18:22 -05003716}
3717
Chris Masonb9fab912012-05-06 07:23:47 -04003718int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3719 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003720{
Chris Mason1259ab72008-05-12 13:39:03 -04003721 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003722 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003723
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003724 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003725 if (!ret)
3726 return ret;
3727
3728 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003729 parent_transid, atomic);
3730 if (ret == -EAGAIN)
3731 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003732 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003733}
Chris Mason6702ed42007-08-07 16:15:09 -04003734
Chris Mason5f39d392007-10-15 16:14:19 -04003735int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3736{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003737 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003738}
3739
3740void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3741{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003742 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003743 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003744 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003745
Josef Bacik06ea65a2013-09-19 16:07:01 -04003746#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3747 /*
3748 * This is a fast path so only do this check if we have sanity tests
3749 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3750 * outside of the sanity tests.
3751 */
3752 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3753 return;
3754#endif
3755 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003756 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003757 if (transid != root->fs_info->generation)
3758 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003759 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003760 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003761 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003762 if (!was_dirty)
3763 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3764 buf->len,
3765 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003766#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3767 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3768 btrfs_print_leaf(root, buf);
3769 ASSERT(0);
3770 }
3771#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05003772}
3773
Liu Bob53d3f52012-11-14 14:34:34 +00003774static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3775 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003776{
Chris Mason188de642008-05-09 11:52:25 -04003777 /*
3778 * looks as though older kernels can get into trouble with
3779 * this code, they end up stuck in balance_dirty_pages forever
3780 */
Miao Xiee2d84522013-01-29 10:09:20 +00003781 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04003782
Jens Axboe6933c022009-03-17 09:36:37 +01003783 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003784 return;
3785
Liu Bob53d3f52012-11-14 14:34:34 +00003786 if (flush_delayed)
3787 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003788
Miao Xiee2d84522013-01-29 10:09:20 +00003789 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3790 BTRFS_DIRTY_METADATA_THRESH);
3791 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08003792 balance_dirty_pages_ratelimited(
3793 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08003794 }
3795 return;
3796}
3797
Liu Bob53d3f52012-11-14 14:34:34 +00003798void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08003799{
Liu Bob53d3f52012-11-14 14:34:34 +00003800 __btrfs_btree_balance_dirty(root, 1);
3801}
Miao Xie16cdcec2011-04-22 18:12:22 +08003802
Liu Bob53d3f52012-11-14 14:34:34 +00003803void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3804{
3805 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04003806}
Chris Mason6b800532007-10-15 16:17:34 -04003807
Chris Masonca7a79a2008-05-12 12:59:19 -04003808int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003809{
Chris Mason727011e2010-08-06 13:21:20 -04003810 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003811 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003812}
Chris Mason0da54682007-11-07 21:08:01 -05003813
David Sterbafcd1f0652012-03-06 00:06:18 +01003814static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003815 int read_only)
3816{
David Sterbac9260932014-09-30 19:16:47 +02003817 struct btrfs_super_block *sb = fs_info->super_copy;
3818 int ret = 0;
3819
3820 if (sb->root_level > BTRFS_MAX_LEVEL) {
3821 printk(KERN_ERR "BTRFS: tree_root level too big: %d > %d\n",
3822 sb->root_level, BTRFS_MAX_LEVEL);
3823 ret = -EINVAL;
3824 }
3825 if (sb->chunk_root_level > BTRFS_MAX_LEVEL) {
3826 printk(KERN_ERR "BTRFS: chunk_root level too big: %d > %d\n",
3827 sb->chunk_root_level, BTRFS_MAX_LEVEL);
3828 ret = -EINVAL;
3829 }
3830 if (sb->log_root_level > BTRFS_MAX_LEVEL) {
3831 printk(KERN_ERR "BTRFS: log_root level too big: %d > %d\n",
3832 sb->log_root_level, BTRFS_MAX_LEVEL);
3833 ret = -EINVAL;
3834 }
3835
David Sterba1104a882013-03-06 15:57:46 +01003836 /*
David Sterbac9260932014-09-30 19:16:47 +02003837 * The common minimum, we don't know if we can trust the nodesize/sectorsize
3838 * items yet, they'll be verified later. Issue just a warning.
David Sterba1104a882013-03-06 15:57:46 +01003839 */
David Sterbac9260932014-09-30 19:16:47 +02003840 if (!IS_ALIGNED(sb->root, 4096))
3841 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
3842 sb->root);
3843 if (!IS_ALIGNED(sb->chunk_root, 4096))
3844 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
3845 sb->chunk_root);
3846 if (!IS_ALIGNED(sb->log_root, 4096))
3847 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
3848 sb->log_root);
3849
3850 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
3851 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
3852 fs_info->fsid, sb->dev_item.fsid);
3853 ret = -EINVAL;
3854 }
3855
3856 /*
3857 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
3858 * done later
3859 */
3860 if (sb->num_devices > (1UL << 31))
3861 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
3862 sb->num_devices);
3863
3864 if (sb->bytenr != BTRFS_SUPER_INFO_OFFSET) {
3865 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
3866 sb->bytenr, BTRFS_SUPER_INFO_OFFSET);
3867 ret = -EINVAL;
3868 }
3869
3870 /*
3871 * The generation is a global counter, we'll trust it more than the others
3872 * but it's still possible that it's the one that's wrong.
3873 */
3874 if (sb->generation < sb->chunk_root_generation)
3875 printk(KERN_WARNING
3876 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
3877 sb->generation, sb->chunk_root_generation);
3878 if (sb->generation < sb->cache_generation && sb->cache_generation != (u64)-1)
3879 printk(KERN_WARNING
3880 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
3881 sb->generation, sb->cache_generation);
3882
3883 return ret;
liuboacce9522011-01-06 19:30:25 +08003884}
3885
Eric Sandeen48a3b632013-04-25 20:41:01 +00003886static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003887{
liuboacce9522011-01-06 19:30:25 +08003888 mutex_lock(&root->fs_info->cleaner_mutex);
3889 btrfs_run_delayed_iputs(root);
3890 mutex_unlock(&root->fs_info->cleaner_mutex);
3891
3892 down_write(&root->fs_info->cleanup_work_sem);
3893 up_write(&root->fs_info->cleanup_work_sem);
3894
3895 /* cleanup FS via transaction */
3896 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003897}
3898
Jeff Mahoney143bede2012-03-01 14:56:26 +01003899static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003900{
liuboacce9522011-01-06 19:30:25 +08003901 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08003902
Miao Xie199c2a92013-05-15 07:48:23 +00003903 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05003904 /*
3905 * This will just short circuit the ordered completion stuff which will
3906 * make sure the ordered extent gets properly cleaned up.
3907 */
Miao Xie199c2a92013-05-15 07:48:23 +00003908 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05003909 root_extent_list)
3910 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00003911 spin_unlock(&root->ordered_extent_lock);
3912}
3913
3914static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3915{
3916 struct btrfs_root *root;
3917 struct list_head splice;
3918
3919 INIT_LIST_HEAD(&splice);
3920
3921 spin_lock(&fs_info->ordered_root_lock);
3922 list_splice_init(&fs_info->ordered_roots, &splice);
3923 while (!list_empty(&splice)) {
3924 root = list_first_entry(&splice, struct btrfs_root,
3925 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04003926 list_move_tail(&root->ordered_root,
3927 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00003928
Liu Bo2a85d9c2014-02-10 17:07:16 +08003929 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003930 btrfs_destroy_ordered_extents(root);
3931
Liu Bo2a85d9c2014-02-10 17:07:16 +08003932 cond_resched();
3933 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003934 }
3935 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003936}
3937
Stefan Behrens35a36212013-08-14 18:12:25 +02003938static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3939 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003940{
3941 struct rb_node *node;
3942 struct btrfs_delayed_ref_root *delayed_refs;
3943 struct btrfs_delayed_ref_node *ref;
3944 int ret = 0;
3945
3946 delayed_refs = &trans->delayed_refs;
3947
3948 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003949 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003950 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05003951 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08003952 return ret;
3953 }
3954
Josef Bacikd7df2c72014-01-23 09:21:38 -05003955 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3956 struct btrfs_delayed_ref_head *head;
Josef Bacike78417d2013-06-03 16:42:36 -04003957 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08003958
Josef Bacikd7df2c72014-01-23 09:21:38 -05003959 head = rb_entry(node, struct btrfs_delayed_ref_head,
3960 href_node);
3961 if (!mutex_trylock(&head->mutex)) {
3962 atomic_inc(&head->node.refs);
3963 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003964
Josef Bacikd7df2c72014-01-23 09:21:38 -05003965 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04003966 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003967 btrfs_put_delayed_ref(&head->node);
3968 spin_lock(&delayed_refs->lock);
3969 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04003970 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05003971 spin_lock(&head->lock);
3972 while ((node = rb_first(&head->ref_root)) != NULL) {
3973 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3974 rb_node);
3975 ref->in_tree = 0;
3976 rb_erase(&ref->rb_node, &head->ref_root);
3977 atomic_dec(&delayed_refs->num_entries);
3978 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003979 }
3980 if (head->must_insert_reserved)
3981 pin_bytes = true;
3982 btrfs_free_delayed_extent_op(head->extent_op);
3983 delayed_refs->num_heads--;
3984 if (head->processing == 0)
3985 delayed_refs->num_heads_ready--;
3986 atomic_dec(&delayed_refs->num_entries);
3987 head->node.in_tree = 0;
3988 rb_erase(&head->href_node, &delayed_refs->href_root);
3989 spin_unlock(&head->lock);
3990 spin_unlock(&delayed_refs->lock);
3991 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08003992
Josef Bacikd7df2c72014-01-23 09:21:38 -05003993 if (pin_bytes)
3994 btrfs_pin_extent(root, head->node.bytenr,
3995 head->node.num_bytes, 1);
3996 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08003997 cond_resched();
3998 spin_lock(&delayed_refs->lock);
3999 }
4000
4001 spin_unlock(&delayed_refs->lock);
4002
4003 return ret;
4004}
4005
Jeff Mahoney143bede2012-03-01 14:56:26 +01004006static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004007{
4008 struct btrfs_inode *btrfs_inode;
4009 struct list_head splice;
4010
4011 INIT_LIST_HEAD(&splice);
4012
Miao Xieeb73c1b2013-05-15 07:48:22 +00004013 spin_lock(&root->delalloc_lock);
4014 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08004015
4016 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00004017 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4018 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08004019
4020 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00004021 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4022 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00004023 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004024
4025 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00004026
Miao Xieeb73c1b2013-05-15 07:48:22 +00004027 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004028 }
4029
Miao Xieeb73c1b2013-05-15 07:48:22 +00004030 spin_unlock(&root->delalloc_lock);
4031}
4032
4033static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4034{
4035 struct btrfs_root *root;
4036 struct list_head splice;
4037
4038 INIT_LIST_HEAD(&splice);
4039
4040 spin_lock(&fs_info->delalloc_root_lock);
4041 list_splice_init(&fs_info->delalloc_roots, &splice);
4042 while (!list_empty(&splice)) {
4043 root = list_first_entry(&splice, struct btrfs_root,
4044 delalloc_root);
4045 list_del_init(&root->delalloc_root);
4046 root = btrfs_grab_fs_root(root);
4047 BUG_ON(!root);
4048 spin_unlock(&fs_info->delalloc_root_lock);
4049
4050 btrfs_destroy_delalloc_inodes(root);
4051 btrfs_put_fs_root(root);
4052
4053 spin_lock(&fs_info->delalloc_root_lock);
4054 }
4055 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004056}
4057
4058static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4059 struct extent_io_tree *dirty_pages,
4060 int mark)
4061{
4062 int ret;
liuboacce9522011-01-06 19:30:25 +08004063 struct extent_buffer *eb;
4064 u64 start = 0;
4065 u64 end;
liuboacce9522011-01-06 19:30:25 +08004066
4067 while (1) {
4068 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004069 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004070 if (ret)
4071 break;
4072
4073 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4074 while (start <= end) {
David Sterba0308af42014-06-15 01:43:40 +02004075 eb = btrfs_find_tree_block(root, start);
David Sterba707e8a02014-06-04 19:22:26 +02004076 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004077 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004078 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004079 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004080
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004081 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4082 &eb->bflags))
4083 clear_extent_buffer_dirty(eb);
4084 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004085 }
4086 }
4087
4088 return ret;
4089}
4090
4091static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4092 struct extent_io_tree *pinned_extents)
4093{
4094 struct extent_io_tree *unpin;
4095 u64 start;
4096 u64 end;
4097 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004098 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004099
4100 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004101again:
liuboacce9522011-01-06 19:30:25 +08004102 while (1) {
4103 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004104 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004105 if (ret)
4106 break;
4107
4108 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00004109 if (btrfs_test_opt(root, DISCARD))
4110 ret = btrfs_error_discard_extent(root, start,
4111 end + 1 - start,
4112 NULL);
liuboacce9522011-01-06 19:30:25 +08004113
4114 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4115 btrfs_error_unpin_extent_range(root, start, end);
4116 cond_resched();
4117 }
4118
Liu Boed0eaa12012-06-14 02:23:21 -06004119 if (loop) {
4120 if (unpin == &root->fs_info->freed_extents[0])
4121 unpin = &root->fs_info->freed_extents[1];
4122 else
4123 unpin = &root->fs_info->freed_extents[0];
4124 loop = false;
4125 goto again;
4126 }
4127
liuboacce9522011-01-06 19:30:25 +08004128 return 0;
4129}
4130
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004131void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4132 struct btrfs_root *root)
4133{
4134 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004135
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004136 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004137 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004138
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004139 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004140 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004141
Miao Xie67cde342012-06-14 02:23:22 -06004142 btrfs_destroy_delayed_inodes(root);
4143 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004144
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004145 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4146 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004147 btrfs_destroy_pinned_extent(root,
4148 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004149
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004150 cur_trans->state =TRANS_STATE_COMPLETED;
4151 wake_up(&cur_trans->commit_wait);
4152
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004153 /*
4154 memset(cur_trans, 0, sizeof(*cur_trans));
4155 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4156 */
4157}
4158
Eric Sandeen48a3b632013-04-25 20:41:01 +00004159static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004160{
4161 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004162
liuboacce9522011-01-06 19:30:25 +08004163 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4164
Josef Bacika4abeea2011-04-11 17:25:13 -04004165 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004166 while (!list_empty(&root->fs_info->trans_list)) {
4167 t = list_first_entry(&root->fs_info->trans_list,
4168 struct btrfs_transaction, list);
4169 if (t->state >= TRANS_STATE_COMMIT_START) {
4170 atomic_inc(&t->use_count);
4171 spin_unlock(&root->fs_info->trans_lock);
4172 btrfs_wait_for_commit(root, t->transid);
4173 btrfs_put_transaction(t);
4174 spin_lock(&root->fs_info->trans_lock);
4175 continue;
4176 }
4177 if (t == root->fs_info->running_transaction) {
4178 t->state = TRANS_STATE_COMMIT_DOING;
4179 spin_unlock(&root->fs_info->trans_lock);
4180 /*
4181 * We wait for 0 num_writers since we don't hold a trans
4182 * handle open currently for this transaction.
4183 */
4184 wait_event(t->writer_wait,
4185 atomic_read(&t->num_writers) == 0);
4186 } else {
4187 spin_unlock(&root->fs_info->trans_lock);
4188 }
4189 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004190
Josef Bacik724e2312013-09-30 11:36:38 -04004191 spin_lock(&root->fs_info->trans_lock);
4192 if (t == root->fs_info->running_transaction)
4193 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004194 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004195 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004196
Josef Bacik724e2312013-09-30 11:36:38 -04004197 btrfs_put_transaction(t);
4198 trace_btrfs_transaction_commit(root);
4199 spin_lock(&root->fs_info->trans_lock);
4200 }
4201 spin_unlock(&root->fs_info->trans_lock);
4202 btrfs_destroy_all_ordered_extents(root->fs_info);
4203 btrfs_destroy_delayed_inodes(root);
4204 btrfs_assert_delayed_root_empty(root);
4205 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4206 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004207 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004208
4209 return 0;
4210}
4211
Chris Masond1310b22008-01-24 16:13:08 -05004212static struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004213 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004214 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004215 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004216 /* note we're sharing with inode.c for the merge bio hook */
4217 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004218};