blob: 0e4fb4a5c7f0a40a7152e6f6883a8672737b4ff7 [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 Mason8b712842008-06-11 16:50:36 -040042#include "async-thread.h"
Chris Mason925baed2008-06-25 16:01:30 -040043#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040044#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040045#include "free-space-cache.h"
Li Zefan581bb052011-04-20 10:06:11 +080046#include "inode-map.h"
Stefan Behrens21adbd52011-11-09 13:44:05 +010047#include "check-integrity.h"
Josef Bacik606686e2012-06-04 14:03:51 -040048#include "rcu-string.h"
Stefan Behrens8dabb742012-11-06 13:15:27 +010049#include "dev-replace.h"
David Woodhouse53b381b2013-01-29 18:40:14 -050050#include "raid56.h"
Jeff Mahoney5ac1d202013-11-01 13:06:58 -040051#include "sysfs.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;
Qu Wenruod458b052014-02-28 10:46:19 +080058static 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);
Josef Bacik569e0f32013-02-13 11:09:14 -050062static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
63 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010064static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080065static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
66 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010067static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080068static int btrfs_destroy_marked_extents(struct btrfs_root *root,
69 struct extent_io_tree *dirty_pages,
70 int mark);
71static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
72 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000073static int btrfs_cleanup_transaction(struct btrfs_root *root);
74static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040075
Chris Masond352ac62008-09-29 15:18:18 -040076/*
77 * end_io_wq structs are used to do processing in task context when an IO is
78 * complete. This is used during reads to verify checksums, and it is used
79 * by writes to insert metadata for new file extents after IO is complete.
80 */
Chris Masonce9adaa2008-04-09 16:28:12 -040081struct end_io_wq {
82 struct bio *bio;
83 bio_end_io_t *end_io;
84 void *private;
85 struct btrfs_fs_info *info;
86 int error;
Chris Mason22c59942008-04-09 16:28:12 -040087 int metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040088 struct list_head list;
Qu Wenruod458b052014-02-28 10:46:19 +080089 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040090};
Chris Mason0da54682007-11-07 21:08:01 -050091
Chris Masond352ac62008-09-29 15:18:18 -040092/*
93 * async submit bios are used to offload expensive checksumming
94 * onto the worker threads. They checksum file and metadata bios
95 * just before they are sent down the IO stack.
96 */
Chris Mason44b8bd72008-04-16 11:14:51 -040097struct async_submit_bio {
98 struct inode *inode;
99 struct bio *bio;
100 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500101 extent_submit_bio_hook_t *submit_bio_start;
102 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400103 int rw;
104 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400105 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400106 /*
107 * bio_offset is optional, can be used if the pages in the bio
108 * can't tell us where in the file the bio should go
109 */
110 u64 bio_offset;
Qu Wenruod458b052014-02-28 10:46:19 +0800111 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100112 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400113};
114
Chris Mason85d4e462011-07-26 16:11:19 -0400115/*
116 * Lockdep class keys for extent_buffer->lock's in this root. For a given
117 * eb, the lockdep key is determined by the btrfs_root it belongs to and
118 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500119 *
Chris Mason85d4e462011-07-26 16:11:19 -0400120 * Different roots are used for different purposes and may nest inside each
121 * other and they require separate keysets. As lockdep keys should be
122 * static, assign keysets according to the purpose of the root as indicated
123 * by btrfs_root->objectid. This ensures that all special purpose roots
124 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500125 *
Chris Mason85d4e462011-07-26 16:11:19 -0400126 * Lock-nesting across peer nodes is always done with the immediate parent
127 * node locked thus preventing deadlock. As lockdep doesn't know this, use
128 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500129 *
Chris Mason85d4e462011-07-26 16:11:19 -0400130 * The key is set by the readpage_end_io_hook after the buffer has passed
131 * csum validation but before the pages are unlocked. It is also set by
132 * btrfs_init_new_buffer on freshly allocated blocks.
133 *
134 * We also add a check to make sure the highest level of the tree is the
135 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
136 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500137 */
138#ifdef CONFIG_DEBUG_LOCK_ALLOC
139# if BTRFS_MAX_LEVEL != 8
140# error
141# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400142
143static struct btrfs_lockdep_keyset {
144 u64 id; /* root objectid */
145 const char *name_stem; /* lock name stem */
146 char names[BTRFS_MAX_LEVEL + 1][20];
147 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
148} btrfs_lockdep_keysets[] = {
149 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
150 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
151 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
152 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
153 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
154 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000155 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400156 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
157 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
158 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200159 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
Chris Mason85d4e462011-07-26 16:11:19 -0400160 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500161};
Chris Mason85d4e462011-07-26 16:11:19 -0400162
163void __init btrfs_init_lockdep(void)
164{
165 int i, j;
166
167 /* initialize lockdep class names */
168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
170
171 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
172 snprintf(ks->names[j], sizeof(ks->names[j]),
173 "btrfs-%s-%02d", ks->name_stem, j);
174 }
175}
176
177void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
178 int level)
179{
180 struct btrfs_lockdep_keyset *ks;
181
182 BUG_ON(level >= ARRAY_SIZE(ks->keys));
183
184 /* find the matching keyset, id 0 is the default entry */
185 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
186 if (ks->id == objectid)
187 break;
188
189 lockdep_set_class_and_name(&eb->lock,
190 &ks->keys[level], ks->names[level]);
191}
192
Chris Mason4008c042009-02-12 14:09:45 -0500193#endif
194
Chris Masond352ac62008-09-29 15:18:18 -0400195/*
196 * extents on the btree inode are pretty simple, there's one extent
197 * that covers the entire device
198 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500199static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200200 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500201 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400202{
203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
204 struct extent_map *em;
205 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400206
Chris Mason890871b2009-09-02 16:24:52 -0400207 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500208 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400209 if (em) {
210 em->bdev =
211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400212 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400213 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400214 }
Chris Mason890871b2009-09-02 16:24:52 -0400215 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400216
David Sterba172ddd62011-04-21 00:48:27 +0200217 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400218 if (!em) {
219 em = ERR_PTR(-ENOMEM);
220 goto out;
221 }
222 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400223 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400224 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400225 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500227
Chris Mason890871b2009-09-02 16:24:52 -0400228 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400229 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400230 if (ret == -EEXIST) {
231 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400232 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600233 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600234 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400235 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400236 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600237 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400238 }
Chris Mason890871b2009-09-02 16:24:52 -0400239 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400240
Chris Mason5f39d392007-10-15 16:14:19 -0400241out:
242 return em;
243}
244
Liu Bob0496682013-03-14 14:57:45 +0000245u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400246{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000247 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400248}
249
250void btrfs_csum_final(u32 crc, char *result)
251{
David Sterba7e75bf32011-03-18 22:56:43 +0000252 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400253}
254
Chris Masond352ac62008-09-29 15:18:18 -0400255/*
256 * compute the csum for a btree block, and either verify it or write it
257 * into the csum field of the block.
258 */
Chris Mason19c00dd2007-10-15 16:19:22 -0400259static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
260 int verify)
261{
David Sterba6c417612011-04-13 15:41:04 +0200262 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500263 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400264 unsigned long len;
265 unsigned long cur_len;
266 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400267 char *kaddr;
268 unsigned long map_start;
269 unsigned long map_len;
270 int err;
271 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500272 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400273
274 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500275 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400276 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400277 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500278 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400279 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400280 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000281 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400282 crc, cur_len);
283 len -= cur_len;
284 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400285 }
Josef Bacik607d4322008-12-02 07:17:45 -0500286 if (csum_size > sizeof(inline_result)) {
287 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
288 if (!result)
289 return 1;
290 } else {
291 result = (char *)&inline_result;
292 }
293
Chris Mason19c00dd2007-10-15 16:19:22 -0400294 btrfs_csum_final(crc, result);
295
296 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500297 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500298 u32 val;
299 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500300 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500301
Josef Bacik607d4322008-12-02 07:17:45 -0500302 read_extent_buffer(buf, &val, 0, csum_size);
Frank Holtonefe120a2013-12-20 11:37:06 -0500303 printk_ratelimited(KERN_INFO
304 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
305 "level %d\n",
306 root->fs_info->sb->s_id, buf->start,
307 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500308 if (result != (char *)&inline_result)
309 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400310 return 1;
311 }
312 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500313 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400314 }
Josef Bacik607d4322008-12-02 07:17:45 -0500315 if (result != (char *)&inline_result)
316 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400317 return 0;
318}
319
Chris Masond352ac62008-09-29 15:18:18 -0400320/*
321 * we can't consider a given block up to date unless the transid of the
322 * block matches the transid in the parent node's pointer. This is how we
323 * detect blocks that either didn't get written at all or got written
324 * in the wrong place.
325 */
Chris Mason1259ab72008-05-12 13:39:03 -0400326static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400327 struct extent_buffer *eb, u64 parent_transid,
328 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400329{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000330 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400331 int ret;
Josef Bacika26e8c92014-03-28 17:07:27 -0400332 bool need_lock = (current->journal_info ==
333 (void *)BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400334
335 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
336 return 0;
337
Chris Masonb9fab912012-05-06 07:23:47 -0400338 if (atomic)
339 return -EAGAIN;
340
Josef Bacika26e8c92014-03-28 17:07:27 -0400341 if (need_lock) {
342 btrfs_tree_read_lock(eb);
343 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
344 }
345
Josef Bacik2ac55d42010-02-03 19:33:23 +0000346 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
Jeff Mahoneyd0082372012-03-01 14:57:19 +0100347 0, &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400348 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400349 btrfs_header_generation(eb) == parent_transid) {
350 ret = 0;
351 goto out;
352 }
David Sterba7a36dde2011-05-06 15:33:15 +0200353 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
Chris Mason193f2842009-04-27 07:29:05 -0400354 "found %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200355 eb->start, parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400356 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400357
358 /*
359 * Things reading via commit roots that don't have normal protection,
360 * like send, can have a really old block in cache that may point at a
361 * block that has been free'd and re-allocated. So don't clear uptodate
362 * if we find an eb that is under IO (dirty/writeback) because we could
363 * end up reading in the stale data and then writing it back out and
364 * making everybody very sad.
365 */
366 if (!extent_buffer_under_io(eb))
367 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400368out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000369 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
370 &cached_state, GFP_NOFS);
Josef Bacika26e8c92014-03-28 17:07:27 -0400371 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400372 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400373}
374
Chris Masond352ac62008-09-29 15:18:18 -0400375/*
David Sterba1104a882013-03-06 15:57:46 +0100376 * Return 0 if the superblock checksum type matches the checksum value of that
377 * algorithm. Pass the raw disk superblock data.
378 */
379static int btrfs_check_super_csum(char *raw_disk_sb)
380{
381 struct btrfs_super_block *disk_sb =
382 (struct btrfs_super_block *)raw_disk_sb;
383 u16 csum_type = btrfs_super_csum_type(disk_sb);
384 int ret = 0;
385
386 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
387 u32 crc = ~(u32)0;
388 const int csum_size = sizeof(crc);
389 char result[csum_size];
390
391 /*
392 * The super_block structure does not span the whole
393 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
394 * is filled with zeros and is included in the checkum.
395 */
396 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
397 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
398 btrfs_csum_final(crc, result);
399
400 if (memcmp(raw_disk_sb, result, csum_size))
401 ret = 1;
Chris Mason667e7d92013-05-07 11:00:13 -0400402
403 if (ret && btrfs_super_generation(disk_sb) < 10) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500404 printk(KERN_WARNING
405 "BTRFS: super block crcs don't match, older mkfs detected\n");
Chris Mason667e7d92013-05-07 11:00:13 -0400406 ret = 0;
407 }
David Sterba1104a882013-03-06 15:57:46 +0100408 }
409
410 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500411 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100412 csum_type);
413 ret = 1;
414 }
415
416 return ret;
417}
418
419/*
Chris Masond352ac62008-09-29 15:18:18 -0400420 * helper to read a given tree block, doing retries as required when
421 * the checksums don't match and we have alternate mirrors to try.
422 */
Chris Masonf1885912008-04-09 16:28:12 -0400423static int btree_read_extent_buffer_pages(struct btrfs_root *root,
424 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400425 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400426{
427 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400428 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400429 int ret;
430 int num_copies = 0;
431 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400432 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400433
Josef Bacika826d6d2011-03-16 13:42:43 -0400434 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400435 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
436 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200437 ret = read_extent_buffer_pages(io_tree, eb, start,
438 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400439 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600440 if (!ret) {
441 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400442 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600443 break;
444 else
445 ret = -EIO;
446 }
Chris Masond3977122009-01-05 21:25:51 -0500447
Josef Bacika826d6d2011-03-16 13:42:43 -0400448 /*
449 * This buffer's crc is fine, but its contents are corrupted, so
450 * there is no reason to read the other copies, they won't be
451 * any less wrong.
452 */
453 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400454 break;
455
Stefan Behrens5d964052012-11-05 14:59:07 +0100456 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400457 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400458 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400459 break;
Chris Mason42352982008-04-28 16:40:52 -0400460
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400461 if (!failed_mirror) {
462 failed = 1;
463 failed_mirror = eb->read_mirror;
464 }
465
Chris Masonf1885912008-04-09 16:28:12 -0400466 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400467 if (mirror_num == failed_mirror)
468 mirror_num++;
469
Chris Mason42352982008-04-28 16:40:52 -0400470 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400471 break;
Chris Masonf1885912008-04-09 16:28:12 -0400472 }
Josef Bacikea466792012-03-26 21:57:36 -0400473
Stefan Behrensc0901582012-07-10 07:30:17 -0600474 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400475 repair_eb_io_failure(root, eb, failed_mirror);
476
477 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400478}
Chris Mason19c00dd2007-10-15 16:19:22 -0400479
Chris Masond352ac62008-09-29 15:18:18 -0400480/*
Chris Masond3977122009-01-05 21:25:51 -0500481 * checksum a dirty tree block before IO. This has extra checks to make sure
482 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400483 */
Chris Masond3977122009-01-05 21:25:51 -0500484
Christoph Hellwigb2950862008-12-02 09:54:17 -0500485static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400486{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000487 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400488 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400489 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400490
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500491 eb = (struct extent_buffer *)page->private;
492 if (page != eb->pages[0])
493 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400494 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530495 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500496 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400497 csum_tree_block(root, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400498 return 0;
499}
500
Yan Zheng2b820322008-11-17 21:11:30 -0500501static int check_tree_block_fsid(struct btrfs_root *root,
502 struct extent_buffer *eb)
503{
504 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
505 u8 fsid[BTRFS_UUID_SIZE];
506 int ret = 1;
507
Ross Kirk0a4e5582013-09-24 10:12:38 +0100508 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500509 while (fs_devices) {
510 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
511 ret = 0;
512 break;
513 }
514 fs_devices = fs_devices->seed;
515 }
516 return ret;
517}
518
Josef Bacika826d6d2011-03-16 13:42:43 -0400519#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500520 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
521 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200522 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400523
524static noinline int check_leaf(struct btrfs_root *root,
525 struct extent_buffer *leaf)
526{
527 struct btrfs_key key;
528 struct btrfs_key leaf_key;
529 u32 nritems = btrfs_header_nritems(leaf);
530 int slot;
531
532 if (nritems == 0)
533 return 0;
534
535 /* Check the 0 item */
536 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
537 BTRFS_LEAF_DATA_SIZE(root)) {
538 CORRUPT("invalid item offset size pair", leaf, root, 0);
539 return -EIO;
540 }
541
542 /*
543 * Check to make sure each items keys are in the correct order and their
544 * offsets make sense. We only have to loop through nritems-1 because
545 * we check the current slot against the next slot, which verifies the
546 * next slot's offset+size makes sense and that the current's slot
547 * offset is correct.
548 */
549 for (slot = 0; slot < nritems - 1; slot++) {
550 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
551 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
552
553 /* Make sure the keys are in the right order */
554 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
555 CORRUPT("bad key order", leaf, root, slot);
556 return -EIO;
557 }
558
559 /*
560 * Make sure the offset and ends are right, remember that the
561 * item data starts at the end of the leaf and grows towards the
562 * front.
563 */
564 if (btrfs_item_offset_nr(leaf, slot) !=
565 btrfs_item_end_nr(leaf, slot + 1)) {
566 CORRUPT("slot offset bad", leaf, root, slot);
567 return -EIO;
568 }
569
570 /*
571 * Check to make sure that we don't point outside of the leaf,
572 * just incase all the items are consistent to eachother, but
573 * all point outside of the leaf.
574 */
575 if (btrfs_item_end_nr(leaf, slot) >
576 BTRFS_LEAF_DATA_SIZE(root)) {
577 CORRUPT("slot end outside of leaf", leaf, root, slot);
578 return -EIO;
579 }
580 }
581
582 return 0;
583}
584
Miao Xiefacc8a222013-07-25 19:22:34 +0800585static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
586 u64 phy_offset, struct page *page,
587 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400588{
Chris Masonce9adaa2008-04-09 16:28:12 -0400589 u64 found_start;
590 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400591 struct extent_buffer *eb;
592 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400593 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400594 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400595
Chris Masonce9adaa2008-04-09 16:28:12 -0400596 if (!page->private)
597 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500598
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500599 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500600
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400601 /* the pending IO might have been the only thing that kept this buffer
602 * in memory. Make sure we have a ref for all this other checks
603 */
604 extent_buffer_get(eb);
605
606 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400607 if (!reads_done)
608 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400609
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400610 eb->read_mirror = mirror;
Josef Bacikea466792012-03-26 21:57:36 -0400611 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
612 ret = -EIO;
613 goto err;
614 }
615
Chris Masonce9adaa2008-04-09 16:28:12 -0400616 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400617 if (found_start != eb->start) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500618 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
Chris Mason193f2842009-04-27 07:29:05 -0400619 "%llu %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200620 found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400621 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400622 goto err;
623 }
Yan Zheng2b820322008-11-17 21:11:30 -0500624 if (check_tree_block_fsid(root, eb)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500625 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200626 eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400627 ret = -EIO;
628 goto err;
629 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400630 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400631 if (found_level >= BTRFS_MAX_LEVEL) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500632 btrfs_info(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400633 (int)btrfs_header_level(eb));
634 ret = -EIO;
635 goto err;
636 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400637
Chris Mason85d4e462011-07-26 16:11:19 -0400638 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
639 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500640
Chris Masonce9adaa2008-04-09 16:28:12 -0400641 ret = csum_tree_block(root, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400642 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400643 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400644 goto err;
645 }
646
647 /*
648 * If this is a leaf block and it is corrupt, set the corrupt bit so
649 * that we don't try and read the other copies of this block, just
650 * return -EIO.
651 */
652 if (found_level == 0 && check_leaf(root, eb)) {
653 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
654 ret = -EIO;
655 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400656
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400657 if (!ret)
658 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400659err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400660 if (reads_done &&
661 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200662 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200663
David Woodhouse53b381b2013-01-29 18:40:14 -0500664 if (ret) {
665 /*
666 * our io error hook is going to dec the io pages
667 * again, we have to make sure it has something
668 * to decrement
669 */
670 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400671 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500672 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400673 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400674out:
Chris Masonf1885912008-04-09 16:28:12 -0400675 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400676}
677
Josef Bacikea466792012-03-26 21:57:36 -0400678static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200679{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200680 struct extent_buffer *eb;
681 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
682
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500683 eb = (struct extent_buffer *)page->private;
Josef Bacikea466792012-03-26 21:57:36 -0400684 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400685 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500686 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400687 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200688 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200689 return -EIO; /* we fixed nothing */
690}
691
Chris Masonce9adaa2008-04-09 16:28:12 -0400692static void end_workqueue_bio(struct bio *bio, int err)
Chris Masonce9adaa2008-04-09 16:28:12 -0400693{
694 struct end_io_wq *end_io_wq = bio->bi_private;
695 struct btrfs_fs_info *fs_info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400696
Chris Masonce9adaa2008-04-09 16:28:12 -0400697 fs_info = end_io_wq->info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400698 end_io_wq->error = err;
Qu Wenruofccb5d82014-02-28 10:46:10 +0800699 btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL);
Chris Masond20f7042008-12-08 16:58:54 -0500700
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200701 if (bio->bi_rw & REQ_WRITE) {
David Woodhouse53b381b2013-01-29 18:40:14 -0500702 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800703 btrfs_queue_work(fs_info->endio_meta_write_workers,
704 &end_io_wq->work);
David Woodhouse53b381b2013-01-29 18:40:14 -0500705 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800706 btrfs_queue_work(fs_info->endio_freespace_worker,
707 &end_io_wq->work);
David Woodhouse53b381b2013-01-29 18:40:14 -0500708 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800709 btrfs_queue_work(fs_info->endio_raid56_workers,
710 &end_io_wq->work);
Chris Masoncad321a2008-12-17 14:51:42 -0500711 else
Qu Wenruofccb5d82014-02-28 10:46:10 +0800712 btrfs_queue_work(fs_info->endio_write_workers,
713 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500714 } else {
David Woodhouse53b381b2013-01-29 18:40:14 -0500715 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800716 btrfs_queue_work(fs_info->endio_raid56_workers,
717 &end_io_wq->work);
David Woodhouse53b381b2013-01-29 18:40:14 -0500718 else if (end_io_wq->metadata)
Qu Wenruofccb5d82014-02-28 10:46:10 +0800719 btrfs_queue_work(fs_info->endio_meta_workers,
720 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500721 else
Qu Wenruofccb5d82014-02-28 10:46:10 +0800722 btrfs_queue_work(fs_info->endio_workers,
723 &end_io_wq->work);
Chris Masond20f7042008-12-08 16:58:54 -0500724 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400725}
726
Josef Bacik0cb59c92010-07-02 12:14:14 -0400727/*
728 * For the metadata arg you want
729 *
730 * 0 - if data
731 * 1 - if normal metadta
732 * 2 - if writing to the free space cache area
David Woodhouse53b381b2013-01-29 18:40:14 -0500733 * 3 - raid parity work
Josef Bacik0cb59c92010-07-02 12:14:14 -0400734 */
Chris Mason22c59942008-04-09 16:28:12 -0400735int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
736 int metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400737{
Chris Masonce9adaa2008-04-09 16:28:12 -0400738 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -0400739 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
740 if (!end_io_wq)
741 return -ENOMEM;
742
743 end_io_wq->private = bio->bi_private;
744 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400745 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400746 end_io_wq->error = 0;
747 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400748 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400749
750 bio->bi_private = end_io_wq;
751 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400752 return 0;
753}
754
Chris Masonb64a2852008-08-20 13:39:41 -0400755unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400756{
757 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800758 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400759 info->fs_devices->open_devices);
760 return 256 * limit;
761}
762
Qu Wenruod458b052014-02-28 10:46:19 +0800763static void run_one_async_start(struct btrfs_work *work)
Chris Mason4a69a412008-11-06 22:03:00 -0500764{
Chris Mason4a69a412008-11-06 22:03:00 -0500765 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100766 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500767
768 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100769 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
770 async->mirror_num, async->bio_flags,
771 async->bio_offset);
772 if (ret)
773 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500774}
775
Qu Wenruod458b052014-02-28 10:46:19 +0800776static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400777{
778 struct btrfs_fs_info *fs_info;
779 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400780 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400781
782 async = container_of(work, struct async_submit_bio, work);
783 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400784
Chris Masonb64a2852008-08-20 13:39:41 -0400785 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400786 limit = limit * 2 / 3;
787
Josef Bacik66657b32012-08-01 15:36:24 -0400788 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400789 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400790 wake_up(&fs_info->async_submit_wait);
791
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100792 /* If an error occured we just want to clean up the bio and move on */
793 if (async->error) {
794 bio_endio(async->bio, async->error);
795 return;
796 }
797
Chris Mason4a69a412008-11-06 22:03:00 -0500798 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400799 async->mirror_num, async->bio_flags,
800 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500801}
802
Qu Wenruod458b052014-02-28 10:46:19 +0800803static void run_one_async_free(struct btrfs_work *work)
Chris Mason4a69a412008-11-06 22:03:00 -0500804{
805 struct async_submit_bio *async;
806
807 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400808 kfree(async);
809}
810
Chris Mason44b8bd72008-04-16 11:14:51 -0400811int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
812 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400813 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400814 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500815 extent_submit_bio_hook_t *submit_bio_start,
816 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400817{
818 struct async_submit_bio *async;
819
820 async = kmalloc(sizeof(*async), GFP_NOFS);
821 if (!async)
822 return -ENOMEM;
823
824 async->inode = inode;
825 async->rw = rw;
826 async->bio = bio;
827 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500828 async->submit_bio_start = submit_bio_start;
829 async->submit_bio_done = submit_bio_done;
830
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800831 btrfs_init_work(&async->work, run_one_async_start,
832 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500833
Chris Masonc8b97812008-10-29 14:49:59 -0400834 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400835 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400836
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100837 async->error = 0;
838
Chris Masoncb03c742008-05-15 16:15:45 -0400839 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400840
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200841 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800842 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400843
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800844 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400845
Chris Masond3977122009-01-05 21:25:51 -0500846 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500847 atomic_read(&fs_info->nr_async_submits)) {
848 wait_event(fs_info->async_submit_wait,
849 (atomic_read(&fs_info->nr_async_submits) == 0));
850 }
851
Chris Mason44b8bd72008-04-16 11:14:51 -0400852 return 0;
853}
854
Chris Masonce3ed712008-09-23 13:14:12 -0400855static int btree_csum_one_bio(struct bio *bio)
856{
857 struct bio_vec *bvec = bio->bi_io_vec;
858 int bio_index = 0;
859 struct btrfs_root *root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100860 int ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400861
862 WARN_ON(bio->bi_vcnt <= 0);
Chris Masond3977122009-01-05 21:25:51 -0500863 while (bio_index < bio->bi_vcnt) {
Chris Masonce3ed712008-09-23 13:14:12 -0400864 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100865 ret = csum_dirty_buffer(root, bvec->bv_page);
866 if (ret)
867 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400868 bio_index++;
869 bvec++;
870 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100871 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400872}
873
Chris Mason4a69a412008-11-06 22:03:00 -0500874static int __btree_submit_bio_start(struct inode *inode, int rw,
875 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400876 unsigned long bio_flags,
877 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400878{
Chris Mason8b712842008-06-11 16:50:36 -0400879 /*
880 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400881 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400882 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100883 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500884}
Chris Mason22c59942008-04-09 16:28:12 -0400885
Chris Mason4a69a412008-11-06 22:03:00 -0500886static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400887 int mirror_num, unsigned long bio_flags,
888 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500889{
Stefan Behrens61891922012-11-05 18:51:52 +0100890 int ret;
891
Chris Mason8b712842008-06-11 16:50:36 -0400892 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500893 * when we're called for a write, we're already in the async
894 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400895 */
Stefan Behrens61891922012-11-05 18:51:52 +0100896 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
897 if (ret)
898 bio_endio(bio, ret);
899 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400900}
901
Josef Bacikde0022b2012-09-25 14:25:58 -0400902static int check_async_write(struct inode *inode, unsigned long bio_flags)
903{
904 if (bio_flags & EXTENT_BIO_TREE_LOG)
905 return 0;
906#ifdef CONFIG_X86
907 if (cpu_has_xmm4_2)
908 return 0;
909#endif
910 return 1;
911}
912
Chris Mason44b8bd72008-04-16 11:14:51 -0400913static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400914 int mirror_num, unsigned long bio_flags,
915 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400916{
Josef Bacikde0022b2012-09-25 14:25:58 -0400917 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500918 int ret;
919
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200920 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500921 /*
922 * called for a read, do the setup so that checksum validation
923 * can happen in the async kernel threads
924 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400925 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
926 bio, 1);
Chris Mason1d4284b2012-03-28 20:31:37 -0400927 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100928 goto out_w_error;
929 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
930 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400931 } else if (!async) {
932 ret = btree_csum_one_bio(bio);
933 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100934 goto out_w_error;
935 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
936 mirror_num, 0);
937 } else {
938 /*
939 * kthread helpers are used to submit writes so that
940 * checksumming can happen in parallel across all CPUs
941 */
942 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
943 inode, rw, bio, mirror_num, 0,
944 bio_offset,
945 __btree_submit_bio_start,
946 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400947 }
Chris Masond313d7a2009-04-20 15:50:09 -0400948
Stefan Behrens61891922012-11-05 18:51:52 +0100949 if (ret) {
950out_w_error:
951 bio_endio(bio, ret);
952 }
953 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400954}
955
Jan Beulich3dd14622010-12-07 14:54:09 +0000956#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500957static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800958 struct page *newpage, struct page *page,
959 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500960{
961 /*
962 * we can't safely write a btree page from here,
963 * we haven't done the locking hook
964 */
965 if (PageDirty(page))
966 return -EAGAIN;
967 /*
968 * Buffers may be managed in a filesystem specific way.
969 * We must have no buffers or drop them.
970 */
971 if (page_has_private(page) &&
972 !try_to_release_page(page, GFP_KERNEL))
973 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800974 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500975}
Jan Beulich3dd14622010-12-07 14:54:09 +0000976#endif
Chris Mason784b4e22010-11-21 22:20:49 -0500977
Chris Mason0da54682007-11-07 21:08:01 -0500978
979static int btree_writepages(struct address_space *mapping,
980 struct writeback_control *wbc)
981{
Miao Xiee2d84522013-01-29 10:09:20 +0000982 struct btrfs_fs_info *fs_info;
983 int ret;
984
Chris Masond8d5f3e2007-12-11 12:42:00 -0500985 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -0800986
987 if (wbc->for_kupdate)
988 return 0;
989
Miao Xiee2d84522013-01-29 10:09:20 +0000990 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -0400991 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +0000992 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
993 BTRFS_DIRTY_METADATA_THRESH);
994 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -0800995 return 0;
Chris Mason793955b2007-11-26 16:34:41 -0800996 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400997 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -0500998}
999
Christoph Hellwigb2950862008-12-02 09:54:17 -05001000static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001001{
Chris Masond1310b22008-01-24 16:13:08 -05001002 struct extent_io_tree *tree;
1003 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001004 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001005}
1006
Chris Mason70dec802008-01-29 09:59:12 -05001007static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001008{
Chris Mason98509cfc2008-09-11 15:51:43 -04001009 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001010 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001011
David Sterbaf7a52a42013-04-26 14:56:29 +00001012 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001013}
Chris Mason123abc82007-03-14 14:14:43 -04001014
Lukas Czernerd47992f2013-05-21 23:17:23 -04001015static void btree_invalidatepage(struct page *page, unsigned int offset,
1016 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001017{
Chris Masond1310b22008-01-24 16:13:08 -05001018 struct extent_io_tree *tree;
1019 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001020 extent_invalidatepage(tree, page, offset);
1021 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001022 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001023 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1024 "page private not zero on page %llu",
1025 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001026 ClearPagePrivate(page);
1027 set_page_private(page, 0);
1028 page_cache_release(page);
1029 }
Chris Masond98237b2007-03-28 13:57:48 -04001030}
1031
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001032static int btree_set_page_dirty(struct page *page)
1033{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001034#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001035 struct extent_buffer *eb;
1036
1037 BUG_ON(!PagePrivate(page));
1038 eb = (struct extent_buffer *)page->private;
1039 BUG_ON(!eb);
1040 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1041 BUG_ON(!atomic_read(&eb->refs));
1042 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001043#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001044 return __set_page_dirty_nobuffers(page);
1045}
1046
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001047static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001048 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001049 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001050 .releasepage = btree_releasepage,
1051 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001052#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001053 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001054#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001055 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001056};
1057
Chris Masonca7a79a2008-05-12 12:59:19 -04001058int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1059 u64 parent_transid)
Chris Mason090d1872007-05-01 08:53:32 -04001060{
Chris Mason5f39d392007-10-15 16:14:19 -04001061 struct extent_buffer *buf = NULL;
1062 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Masonde428b62007-05-18 13:28:27 -04001063 int ret = 0;
Chris Mason090d1872007-05-01 08:53:32 -04001064
Chris Masondb945352007-10-15 16:15:53 -04001065 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001066 if (!buf)
Chris Mason090d1872007-05-01 08:53:32 -04001067 return 0;
Chris Masond1310b22008-01-24 16:13:08 -05001068 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001069 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001070 free_extent_buffer(buf);
Chris Masonde428b62007-05-18 13:28:27 -04001071 return ret;
Chris Mason090d1872007-05-01 08:53:32 -04001072}
1073
Arne Jansenab0fff02011-05-23 14:25:41 +02001074int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1075 int mirror_num, struct extent_buffer **eb)
1076{
1077 struct extent_buffer *buf = NULL;
1078 struct inode *btree_inode = root->fs_info->btree_inode;
1079 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1080 int ret;
1081
1082 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1083 if (!buf)
1084 return 0;
1085
1086 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1087
1088 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1089 btree_get_extent, mirror_num);
1090 if (ret) {
1091 free_extent_buffer(buf);
1092 return ret;
1093 }
1094
1095 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1096 free_extent_buffer(buf);
1097 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001098 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001099 *eb = buf;
1100 } else {
1101 free_extent_buffer(buf);
1102 }
1103 return 0;
1104}
1105
Chris Mason0999df52008-04-01 13:48:14 -04001106struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1107 u64 bytenr, u32 blocksize)
1108{
Josef Bacikf28491e2013-12-16 13:24:27 -05001109 return find_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001110}
1111
1112struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1113 u64 bytenr, u32 blocksize)
1114{
Josef Bacikf28491e2013-12-16 13:24:27 -05001115 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
Chris Mason0999df52008-04-01 13:48:14 -04001116}
1117
1118
Chris Masone02119d2008-09-05 16:13:11 -04001119int btrfs_write_tree_block(struct extent_buffer *buf)
1120{
Chris Mason727011e2010-08-06 13:21:20 -04001121 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001122 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001123}
1124
1125int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1126{
Chris Mason727011e2010-08-06 13:21:20 -04001127 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001128 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001129}
1130
Chris Masondb945352007-10-15 16:15:53 -04001131struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
Chris Masonca7a79a2008-05-12 12:59:19 -04001132 u32 blocksize, u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001133{
Chris Mason5f39d392007-10-15 16:14:19 -04001134 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001135 int ret;
1136
Chris Masondb945352007-10-15 16:15:53 -04001137 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
Chris Mason5f39d392007-10-15 16:14:19 -04001138 if (!buf)
1139 return NULL;
Chris Masone4204de2008-01-08 15:46:27 -05001140
Chris Masonca7a79a2008-05-12 12:59:19 -04001141 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001142 if (ret) {
1143 free_extent_buffer(buf);
1144 return NULL;
1145 }
Chris Mason5f39d392007-10-15 16:14:19 -04001146 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001147
Chris Masoneb60cea2007-02-02 09:18:22 -05001148}
1149
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001150void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1151 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001152{
Miao Xiee2d84522013-01-29 10:09:20 +00001153 struct btrfs_fs_info *fs_info = root->fs_info;
1154
Chris Mason55c69072008-01-09 15:55:33 -05001155 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001156 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001157 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001158
Chris Masonb9473432009-03-13 11:00:37 -04001159 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001160 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1161 -buf->len,
1162 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001163 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1164 btrfs_set_lock_blocking(buf);
1165 clear_extent_buffer_dirty(buf);
1166 }
Chris Mason925baed2008-06-25 16:01:30 -04001167 }
Chris Mason5f39d392007-10-15 16:14:19 -04001168}
1169
Miao Xie8257b2d2014-03-06 13:38:19 +08001170static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1171{
1172 struct btrfs_subvolume_writers *writers;
1173 int ret;
1174
1175 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1176 if (!writers)
1177 return ERR_PTR(-ENOMEM);
1178
1179 ret = percpu_counter_init(&writers->counter, 0);
1180 if (ret < 0) {
1181 kfree(writers);
1182 return ERR_PTR(ret);
1183 }
1184
1185 init_waitqueue_head(&writers->wait);
1186 return writers;
1187}
1188
1189static void
1190btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1191{
1192 percpu_counter_destroy(&writers->counter);
1193 kfree(writers);
1194}
1195
Jeff Mahoney143bede2012-03-01 14:56:26 +01001196static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1197 u32 stripesize, struct btrfs_root *root,
1198 struct btrfs_fs_info *fs_info,
1199 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001200{
Chris Masoncfaa7292007-02-21 17:04:57 -05001201 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001202 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001203 root->sectorsize = sectorsize;
1204 root->nodesize = nodesize;
1205 root->leafsize = leafsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001206 root->stripesize = stripesize;
Chris Mason123abc82007-03-14 14:14:43 -04001207 root->ref_cows = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001208 root->track_dirty = 0;
Yan, Zhengc71bf092009-11-12 09:34:40 +00001209 root->in_radix = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001210 root->orphan_item_inserted = 0;
1211 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001212
Chris Mason0f7d52f2007-04-09 10:42:37 -04001213 root->objectid = objectid;
1214 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001215 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001216 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001217 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001218 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001219 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001220 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001221 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001222 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001223
1224 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001225 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001226 INIT_LIST_HEAD(&root->delalloc_inodes);
1227 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001228 INIT_LIST_HEAD(&root->ordered_extents);
1229 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001230 INIT_LIST_HEAD(&root->logged_list[0]);
1231 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001232 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001233 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001234 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001235 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001236 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001237 spin_lock_init(&root->log_extents_lock[0]);
1238 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001239 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001240 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001241 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001242 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f1832009-01-21 12:54:03 -05001243 init_waitqueue_head(&root->log_writer_wait);
1244 init_waitqueue_head(&root->log_commit_wait[0]);
1245 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001246 INIT_LIST_HEAD(&root->log_ctxs[0]);
1247 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f1832009-01-21 12:54:03 -05001248 atomic_set(&root->log_commit[0], 0);
1249 atomic_set(&root->log_commit[1], 0);
1250 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001251 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001252 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001253 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001254 atomic_set(&root->will_be_snapshoted, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001255 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001256 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001257 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001258 if (fs_info)
1259 extent_io_tree_init(&root->dirty_log_pages,
1260 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001261
Chris Mason3768f362007-03-13 16:47:54 -04001262 memset(&root->root_key, 0, sizeof(root->root_key));
1263 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001264 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik58176a92007-08-29 15:47:34 -04001265 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001266 if (fs_info)
1267 root->defrag_trans_start = fs_info->generation;
1268 else
1269 root->defrag_trans_start = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001270 init_completion(&root->kobj_unregister);
Chris Mason6702ed42007-08-07 16:15:09 -04001271 root->defrag_running = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001272 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001273 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001274
Anand Jain5f3ab902012-12-07 09:28:54 +00001275 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001276}
1277
Al Virof84a8bd2011-11-17 15:57:57 -05001278static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001279{
1280 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1281 if (root)
1282 root->fs_info = fs_info;
1283 return root;
1284}
1285
Josef Bacik06ea65a2013-09-19 16:07:01 -04001286#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1287/* Should only be used by the testing infrastructure */
1288struct btrfs_root *btrfs_alloc_dummy_root(void)
1289{
1290 struct btrfs_root *root;
1291
1292 root = btrfs_alloc_root(NULL);
1293 if (!root)
1294 return ERR_PTR(-ENOMEM);
1295 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1296 root->dummy_root = 1;
1297
1298 return root;
1299}
1300#endif
1301
Arne Jansen20897f52011-09-13 12:44:20 +02001302struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1303 struct btrfs_fs_info *fs_info,
1304 u64 objectid)
1305{
1306 struct extent_buffer *leaf;
1307 struct btrfs_root *tree_root = fs_info->tree_root;
1308 struct btrfs_root *root;
1309 struct btrfs_key key;
1310 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001311 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001312
1313 root = btrfs_alloc_root(fs_info);
1314 if (!root)
1315 return ERR_PTR(-ENOMEM);
1316
1317 __setup_root(tree_root->nodesize, tree_root->leafsize,
1318 tree_root->sectorsize, tree_root->stripesize,
1319 root, fs_info, objectid);
1320 root->root_key.objectid = objectid;
1321 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1322 root->root_key.offset = 0;
1323
1324 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1325 0, objectid, NULL, 0, 0, 0);
1326 if (IS_ERR(leaf)) {
1327 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001328 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001329 goto fail;
1330 }
1331
Arne Jansen20897f52011-09-13 12:44:20 +02001332 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1333 btrfs_set_header_bytenr(leaf, leaf->start);
1334 btrfs_set_header_generation(leaf, trans->transid);
1335 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1336 btrfs_set_header_owner(leaf, objectid);
1337 root->node = leaf;
1338
Ross Kirk0a4e5582013-09-24 10:12:38 +01001339 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001340 BTRFS_FSID_SIZE);
1341 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001342 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001343 BTRFS_UUID_SIZE);
1344 btrfs_mark_buffer_dirty(leaf);
1345
1346 root->commit_root = btrfs_root_node(root);
1347 root->track_dirty = 1;
1348
1349
1350 root->root_item.flags = 0;
1351 root->root_item.byte_limit = 0;
1352 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1353 btrfs_set_root_generation(&root->root_item, trans->transid);
1354 btrfs_set_root_level(&root->root_item, 0);
1355 btrfs_set_root_refs(&root->root_item, 1);
1356 btrfs_set_root_used(&root->root_item, leaf->len);
1357 btrfs_set_root_last_snapshot(&root->root_item, 0);
1358 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001359 uuid_le_gen(&uuid);
1360 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001361 root->root_item.drop_level = 0;
1362
1363 key.objectid = objectid;
1364 key.type = BTRFS_ROOT_ITEM_KEY;
1365 key.offset = 0;
1366 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1367 if (ret)
1368 goto fail;
1369
1370 btrfs_tree_unlock(leaf);
1371
Arne Jansen20897f52011-09-13 12:44:20 +02001372 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001373
1374fail:
1375 if (leaf) {
1376 btrfs_tree_unlock(leaf);
1377 free_extent_buffer(leaf);
1378 }
1379 kfree(root);
1380
1381 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001382}
1383
Yan Zheng7237f1832009-01-21 12:54:03 -05001384static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1385 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001386{
1387 struct btrfs_root *root;
1388 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f1832009-01-21 12:54:03 -05001389 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001390
Al Viro6f07e422011-11-17 00:46:16 -05001391 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001392 if (!root)
Yan Zheng7237f1832009-01-21 12:54:03 -05001393 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001394
1395 __setup_root(tree_root->nodesize, tree_root->leafsize,
1396 tree_root->sectorsize, tree_root->stripesize,
1397 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1398
1399 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1400 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1401 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Yan Zheng7237f1832009-01-21 12:54:03 -05001402 /*
1403 * log trees do not get reference counted because they go away
1404 * before a real commit is actually done. They do store pointers
1405 * to file data extents, and those reference counts still get
1406 * updated (along with back refs to the log tree).
1407 */
Chris Masone02119d2008-09-05 16:13:11 -04001408 root->ref_cows = 0;
1409
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001410 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
Arne Jansen66d7e7f2011-09-12 15:26:38 +02001411 BTRFS_TREE_LOG_OBJECTID, NULL,
Jan Schmidt5581a512012-05-16 17:04:52 +02001412 0, 0, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001413 if (IS_ERR(leaf)) {
1414 kfree(root);
1415 return ERR_CAST(leaf);
1416 }
Chris Masone02119d2008-09-05 16:13:11 -04001417
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001418 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1419 btrfs_set_header_bytenr(leaf, leaf->start);
1420 btrfs_set_header_generation(leaf, trans->transid);
1421 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1422 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f1832009-01-21 12:54:03 -05001423 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001424
1425 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001426 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001427 btrfs_mark_buffer_dirty(root->node);
1428 btrfs_tree_unlock(root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001429 return root;
1430}
1431
1432int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1433 struct btrfs_fs_info *fs_info)
1434{
1435 struct btrfs_root *log_root;
1436
1437 log_root = alloc_log_tree(trans, fs_info);
1438 if (IS_ERR(log_root))
1439 return PTR_ERR(log_root);
1440 WARN_ON(fs_info->log_root_tree);
1441 fs_info->log_root_tree = log_root;
1442 return 0;
1443}
1444
1445int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1446 struct btrfs_root *root)
1447{
1448 struct btrfs_root *log_root;
1449 struct btrfs_inode_item *inode_item;
1450
1451 log_root = alloc_log_tree(trans, root->fs_info);
1452 if (IS_ERR(log_root))
1453 return PTR_ERR(log_root);
1454
1455 log_root->last_trans = trans->transid;
1456 log_root->root_key.offset = root->root_key.objectid;
1457
1458 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001459 btrfs_set_stack_inode_generation(inode_item, 1);
1460 btrfs_set_stack_inode_size(inode_item, 3);
1461 btrfs_set_stack_inode_nlink(inode_item, 1);
1462 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1463 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f1832009-01-21 12:54:03 -05001464
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001465 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001466
1467 WARN_ON(root->log_root);
1468 root->log_root = log_root;
1469 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001470 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001471 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001472 return 0;
1473}
1474
Stefan Behrens35a36212013-08-14 18:12:25 +02001475static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1476 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001477{
1478 struct btrfs_root *root;
1479 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001480 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001481 u64 generation;
Chris Masondb945352007-10-15 16:15:53 -04001482 u32 blocksize;
Miao Xiecb517ea2013-05-15 07:48:19 +00001483 int ret;
1484
1485 path = btrfs_alloc_path();
1486 if (!path)
1487 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001488
Al Viro6f07e422011-11-17 00:46:16 -05001489 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001490 if (!root) {
1491 ret = -ENOMEM;
1492 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001493 }
1494
Chris Masondb945352007-10-15 16:15:53 -04001495 __setup_root(tree_root->nodesize, tree_root->leafsize,
Chris Mason87ee04e2007-11-30 11:30:34 -05001496 tree_root->sectorsize, tree_root->stripesize,
Miao Xiecb517ea2013-05-15 07:48:19 +00001497 root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001498
Miao Xiecb517ea2013-05-15 07:48:19 +00001499 ret = btrfs_find_root(tree_root, key, path,
1500 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001501 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001502 if (ret > 0)
1503 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001504 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001505 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001506
Yan Zheng84234f32008-10-29 14:49:05 -04001507 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001508 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1509 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
Yan Zheng84234f32008-10-29 14:49:05 -04001510 blocksize, generation);
Miao Xiecb517ea2013-05-15 07:48:19 +00001511 if (!root->node) {
1512 ret = -ENOMEM;
1513 goto find_fail;
1514 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1515 ret = -EIO;
1516 goto read_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001517 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001518 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001519out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001520 btrfs_free_path(path);
1521 return root;
1522
1523read_fail:
1524 free_extent_buffer(root->node);
1525find_fail:
1526 kfree(root);
1527alloc_fail:
1528 root = ERR_PTR(ret);
1529 goto out;
1530}
1531
1532struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1533 struct btrfs_key *location)
1534{
1535 struct btrfs_root *root;
1536
1537 root = btrfs_read_tree_root(tree_root, location);
1538 if (IS_ERR(root))
1539 return root;
1540
1541 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Chris Masone02119d2008-09-05 16:13:11 -04001542 root->ref_cows = 1;
Li Zefan08fe4db2011-03-28 02:01:25 +00001543 btrfs_check_and_init_root_item(&root->root_item);
1544 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001545
Chris Mason5eda7b52007-06-22 14:16:25 -04001546 return root;
1547}
1548
Miao Xiecb517ea2013-05-15 07:48:19 +00001549int btrfs_init_fs_root(struct btrfs_root *root)
1550{
1551 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001552 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001553
1554 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1555 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1556 GFP_NOFS);
1557 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1558 ret = -ENOMEM;
1559 goto fail;
1560 }
1561
Miao Xie8257b2d2014-03-06 13:38:19 +08001562 writers = btrfs_alloc_subvolume_writers();
1563 if (IS_ERR(writers)) {
1564 ret = PTR_ERR(writers);
1565 goto fail;
1566 }
1567 root->subv_writers = writers;
1568
Miao Xiecb517ea2013-05-15 07:48:19 +00001569 btrfs_init_free_ino_ctl(root);
Miao Xiecb517ea2013-05-15 07:48:19 +00001570 spin_lock_init(&root->cache_lock);
1571 init_waitqueue_head(&root->cache_wait);
1572
1573 ret = get_anon_bdev(&root->anon_dev);
1574 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001575 goto free_writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001576 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001577
1578free_writers:
1579 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001580fail:
1581 kfree(root->free_ino_ctl);
1582 kfree(root->free_ino_pinned);
1583 return ret;
1584}
1585
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001586static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1587 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001588{
1589 struct btrfs_root *root;
1590
1591 spin_lock(&fs_info->fs_roots_radix_lock);
1592 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1593 (unsigned long)root_id);
1594 spin_unlock(&fs_info->fs_roots_radix_lock);
1595 return root;
1596}
1597
1598int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1599 struct btrfs_root *root)
1600{
1601 int ret;
1602
1603 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1604 if (ret)
1605 return ret;
1606
1607 spin_lock(&fs_info->fs_roots_radix_lock);
1608 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1609 (unsigned long)root->root_key.objectid,
1610 root);
1611 if (ret == 0)
1612 root->in_radix = 1;
1613 spin_unlock(&fs_info->fs_roots_radix_lock);
1614 radix_tree_preload_end();
1615
1616 return ret;
1617}
1618
Miao Xiec00869f2013-09-25 21:47:44 +08001619struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1620 struct btrfs_key *location,
1621 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001622{
1623 struct btrfs_root *root;
1624 int ret;
1625
Chris Masonedbd8d42007-12-21 16:27:24 -05001626 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1627 return fs_info->tree_root;
1628 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1629 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001630 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1631 return fs_info->chunk_root;
1632 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1633 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001634 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1635 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001636 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1637 return fs_info->quota_root ? fs_info->quota_root :
1638 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001639 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1640 return fs_info->uuid_root ? fs_info->uuid_root :
1641 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001642again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001643 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001644 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001645 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001646 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001647 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001648 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001649
Miao Xiecb517ea2013-05-15 07:48:19 +00001650 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001651 if (IS_ERR(root))
1652 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001653
Miao Xiec00869f2013-09-25 21:47:44 +08001654 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001655 ret = -ENOENT;
1656 goto fail;
1657 }
1658
Miao Xiecb517ea2013-05-15 07:48:19 +00001659 ret = btrfs_init_fs_root(root);
1660 if (ret)
1661 goto fail;
1662
Kelley Nielsen3f870c22013-11-04 19:37:39 -08001663 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1664 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001665 if (ret < 0)
1666 goto fail;
1667 if (ret == 0)
1668 root->orphan_item_inserted = 1;
1669
Miao Xiecb517ea2013-05-15 07:48:19 +00001670 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001671 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001672 if (ret == -EEXIST) {
1673 free_fs_root(root);
1674 goto again;
1675 }
1676 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001677 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001678 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001679fail:
1680 free_fs_root(root);
1681 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001682}
1683
Chris Mason04160082008-03-26 10:28:07 -04001684static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1685{
1686 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1687 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001688 struct btrfs_device *device;
1689 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001690
Xiao Guangrong1f781602011-04-20 10:09:16 +00001691 rcu_read_lock();
1692 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001693 if (!device->bdev)
1694 continue;
Chris Mason04160082008-03-26 10:28:07 -04001695 bdi = blk_get_backing_dev_info(device->bdev);
1696 if (bdi && bdi_congested(bdi, bdi_bits)) {
1697 ret = 1;
1698 break;
1699 }
1700 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001701 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001702 return ret;
1703}
1704
Chris Mason38b66982008-04-22 09:22:11 -04001705/*
Jens Axboead081f12009-06-12 14:43:40 +02001706 * If this fails, caller must call bdi_destroy() to get rid of the
1707 * bdi again.
1708 */
Chris Mason04160082008-03-26 10:28:07 -04001709static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1710{
Jens Axboead081f12009-06-12 14:43:40 +02001711 int err;
1712
1713 bdi->capabilities = BDI_CAP_MAP_COPY;
Jens Axboee6d086d2010-04-26 10:27:54 +02001714 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
Jens Axboead081f12009-06-12 14:43:40 +02001715 if (err)
1716 return err;
1717
Chris Mason4575c9c2008-04-18 16:13:31 -04001718 bdi->ra_pages = default_backing_dev_info.ra_pages;
Chris Mason04160082008-03-26 10:28:07 -04001719 bdi->congested_fn = btrfs_congested_fn;
1720 bdi->congested_data = info;
1721 return 0;
1722}
1723
Chris Mason8b712842008-06-11 16:50:36 -04001724/*
1725 * called by the kthread helper functions to finally call the bio end_io
1726 * functions. This is where read checksum verification actually happens
1727 */
Qu Wenruod458b052014-02-28 10:46:19 +08001728static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001729{
Chris Masonce9adaa2008-04-09 16:28:12 -04001730 struct bio *bio;
Chris Mason8b712842008-06-11 16:50:36 -04001731 struct end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001732 int error;
Chris Masonce9adaa2008-04-09 16:28:12 -04001733
Chris Mason8b712842008-06-11 16:50:36 -04001734 end_io_wq = container_of(work, struct end_io_wq, work);
1735 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001736
Chris Mason8b712842008-06-11 16:50:36 -04001737 error = end_io_wq->error;
1738 bio->bi_private = end_io_wq->private;
1739 bio->bi_end_io = end_io_wq->end_io;
1740 kfree(end_io_wq);
Chris Mason8b712842008-06-11 16:50:36 -04001741 bio_endio(bio, error);
Chris Mason44b8bd72008-04-16 11:14:51 -04001742}
1743
Chris Masona74a4b92008-06-25 16:01:31 -04001744static int cleaner_kthread(void *arg)
1745{
1746 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001747 int again;
Chris Masona74a4b92008-06-25 16:01:31 -04001748
1749 do {
Miao Xied0278242013-05-14 10:20:40 +00001750 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001751
Miao Xied0278242013-05-14 10:20:40 +00001752 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001753 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001754 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001755
Miao Xied0278242013-05-14 10:20:40 +00001756 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1757 goto sleep;
1758
Miao Xiedc7f3702013-05-14 10:20:42 +00001759 /*
1760 * Avoid the problem that we change the status of the fs
1761 * during the above check and trylock.
1762 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001763 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001764 mutex_unlock(&root->fs_info->cleaner_mutex);
1765 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001766 }
1767
Miao Xied0278242013-05-14 10:20:40 +00001768 btrfs_run_delayed_iputs(root);
1769 again = btrfs_clean_one_deleted_snapshot(root);
1770 mutex_unlock(&root->fs_info->cleaner_mutex);
1771
1772 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001773 * The defragger has dealt with the R/O remount and umount,
1774 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001775 */
1776 btrfs_run_defrag_inodes(root->fs_info);
1777sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001778 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001779 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001780 if (!kthread_should_stop())
1781 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001782 __set_current_state(TASK_RUNNING);
1783 }
1784 } while (!kthread_should_stop());
1785 return 0;
1786}
1787
1788static int transaction_kthread(void *arg)
1789{
1790 struct btrfs_root *root = arg;
1791 struct btrfs_trans_handle *trans;
1792 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001793 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001794 unsigned long now;
1795 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001796 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001797
1798 do {
Jan Kara914b2002012-03-12 16:05:50 +01001799 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001800 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001801 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1802
Josef Bacika4abeea2011-04-11 17:25:13 -04001803 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001804 cur = root->fs_info->running_transaction;
1805 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001806 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001807 goto sleep;
1808 }
Yan Zheng31153d82008-07-28 15:32:19 -04001809
Chris Masona74a4b92008-06-25 16:01:31 -04001810 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001811 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001812 (now < cur->start_time ||
1813 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001814 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001815 delay = HZ * 5;
1816 goto sleep;
1817 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001818 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001819 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001820
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001821 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001822 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001823 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001824 if (PTR_ERR(trans) != -ENOENT)
1825 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001826 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001827 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001828 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001829 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001830 } else {
1831 btrfs_end_transaction(trans, root);
1832 }
Chris Masona74a4b92008-06-25 16:01:31 -04001833sleep:
1834 wake_up_process(root->fs_info->cleaner_kthread);
1835 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1836
Josef Bacik4e121c02013-09-27 16:32:39 -04001837 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1838 &root->fs_info->fs_state)))
1839 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001840 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001841 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001842 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001843 (!btrfs_transaction_blocked(root->fs_info) ||
1844 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001845 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001846 __set_current_state(TASK_RUNNING);
1847 }
1848 } while (!kthread_should_stop());
1849 return 0;
1850}
1851
Chris Masonaf31f5e2011-11-03 15:17:42 -04001852/*
1853 * this will find the highest generation in the array of
1854 * root backups. The index of the highest array is returned,
1855 * or -1 if we can't find anything.
1856 *
1857 * We check to make sure the array is valid by comparing the
1858 * generation of the latest root in the array with the generation
1859 * in the super block. If they don't match we pitch it.
1860 */
1861static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1862{
1863 u64 cur;
1864 int newest_index = -1;
1865 struct btrfs_root_backup *root_backup;
1866 int i;
1867
1868 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1869 root_backup = info->super_copy->super_roots + i;
1870 cur = btrfs_backup_tree_root_gen(root_backup);
1871 if (cur == newest_gen)
1872 newest_index = i;
1873 }
1874
1875 /* check to see if we actually wrapped around */
1876 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1877 root_backup = info->super_copy->super_roots;
1878 cur = btrfs_backup_tree_root_gen(root_backup);
1879 if (cur == newest_gen)
1880 newest_index = 0;
1881 }
1882 return newest_index;
1883}
1884
1885
1886/*
1887 * find the oldest backup so we know where to store new entries
1888 * in the backup array. This will set the backup_root_index
1889 * field in the fs_info struct
1890 */
1891static void find_oldest_super_backup(struct btrfs_fs_info *info,
1892 u64 newest_gen)
1893{
1894 int newest_index = -1;
1895
1896 newest_index = find_newest_super_backup(info, newest_gen);
1897 /* if there was garbage in there, just move along */
1898 if (newest_index == -1) {
1899 info->backup_root_index = 0;
1900 } else {
1901 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1902 }
1903}
1904
1905/*
1906 * copy all the root pointers into the super backup array.
1907 * this will bump the backup pointer by one when it is
1908 * done
1909 */
1910static void backup_super_roots(struct btrfs_fs_info *info)
1911{
1912 int next_backup;
1913 struct btrfs_root_backup *root_backup;
1914 int last_backup;
1915
1916 next_backup = info->backup_root_index;
1917 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1918 BTRFS_NUM_BACKUP_ROOTS;
1919
1920 /*
1921 * just overwrite the last backup if we're at the same generation
1922 * this happens only at umount
1923 */
1924 root_backup = info->super_for_commit->super_roots + last_backup;
1925 if (btrfs_backup_tree_root_gen(root_backup) ==
1926 btrfs_header_generation(info->tree_root->node))
1927 next_backup = last_backup;
1928
1929 root_backup = info->super_for_commit->super_roots + next_backup;
1930
1931 /*
1932 * make sure all of our padding and empty slots get zero filled
1933 * regardless of which ones we use today
1934 */
1935 memset(root_backup, 0, sizeof(*root_backup));
1936
1937 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1938
1939 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1940 btrfs_set_backup_tree_root_gen(root_backup,
1941 btrfs_header_generation(info->tree_root->node));
1942
1943 btrfs_set_backup_tree_root_level(root_backup,
1944 btrfs_header_level(info->tree_root->node));
1945
1946 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1947 btrfs_set_backup_chunk_root_gen(root_backup,
1948 btrfs_header_generation(info->chunk_root->node));
1949 btrfs_set_backup_chunk_root_level(root_backup,
1950 btrfs_header_level(info->chunk_root->node));
1951
1952 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1953 btrfs_set_backup_extent_root_gen(root_backup,
1954 btrfs_header_generation(info->extent_root->node));
1955 btrfs_set_backup_extent_root_level(root_backup,
1956 btrfs_header_level(info->extent_root->node));
1957
Chris Mason7c7e82a2011-11-06 18:50:56 -05001958 /*
1959 * we might commit during log recovery, which happens before we set
1960 * the fs_root. Make sure it is valid before we fill it in.
1961 */
1962 if (info->fs_root && info->fs_root->node) {
1963 btrfs_set_backup_fs_root(root_backup,
1964 info->fs_root->node->start);
1965 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001966 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001967 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04001968 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05001969 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04001970
1971 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1972 btrfs_set_backup_dev_root_gen(root_backup,
1973 btrfs_header_generation(info->dev_root->node));
1974 btrfs_set_backup_dev_root_level(root_backup,
1975 btrfs_header_level(info->dev_root->node));
1976
1977 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1978 btrfs_set_backup_csum_root_gen(root_backup,
1979 btrfs_header_generation(info->csum_root->node));
1980 btrfs_set_backup_csum_root_level(root_backup,
1981 btrfs_header_level(info->csum_root->node));
1982
1983 btrfs_set_backup_total_bytes(root_backup,
1984 btrfs_super_total_bytes(info->super_copy));
1985 btrfs_set_backup_bytes_used(root_backup,
1986 btrfs_super_bytes_used(info->super_copy));
1987 btrfs_set_backup_num_devices(root_backup,
1988 btrfs_super_num_devices(info->super_copy));
1989
1990 /*
1991 * if we don't copy this out to the super_copy, it won't get remembered
1992 * for the next commit
1993 */
1994 memcpy(&info->super_copy->super_roots,
1995 &info->super_for_commit->super_roots,
1996 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1997}
1998
1999/*
2000 * this copies info out of the root backup array and back into
2001 * the in-memory super block. It is meant to help iterate through
2002 * the array, so you send it the number of backups you've already
2003 * tried and the last backup index you used.
2004 *
2005 * this returns -1 when it has tried all the backups
2006 */
2007static noinline int next_root_backup(struct btrfs_fs_info *info,
2008 struct btrfs_super_block *super,
2009 int *num_backups_tried, int *backup_index)
2010{
2011 struct btrfs_root_backup *root_backup;
2012 int newest = *backup_index;
2013
2014 if (*num_backups_tried == 0) {
2015 u64 gen = btrfs_super_generation(super);
2016
2017 newest = find_newest_super_backup(info, gen);
2018 if (newest == -1)
2019 return -1;
2020
2021 *backup_index = newest;
2022 *num_backups_tried = 1;
2023 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2024 /* we've tried all the backups, all done */
2025 return -1;
2026 } else {
2027 /* jump to the next oldest backup */
2028 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2029 BTRFS_NUM_BACKUP_ROOTS;
2030 *backup_index = newest;
2031 *num_backups_tried += 1;
2032 }
2033 root_backup = super->super_roots + newest;
2034
2035 btrfs_set_super_generation(super,
2036 btrfs_backup_tree_root_gen(root_backup));
2037 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2038 btrfs_set_super_root_level(super,
2039 btrfs_backup_tree_root_level(root_backup));
2040 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2041
2042 /*
2043 * fixme: the total bytes and num_devices need to match or we should
2044 * need a fsck
2045 */
2046 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2047 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2048 return 0;
2049}
2050
Liu Bo7abadb62013-03-17 02:10:31 +00002051/* helper to cleanup workers */
2052static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2053{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002054 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002055 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002056 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002057 btrfs_destroy_workqueue(fs_info->endio_workers);
2058 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2059 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002060 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002061 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2062 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2063 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002064 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002065 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002066 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002067 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002068 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002069 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002070}
2071
Rashika2e9f5952013-10-31 02:45:20 +05302072static void free_root_extent_buffers(struct btrfs_root *root)
2073{
2074 if (root) {
2075 free_extent_buffer(root->node);
2076 free_extent_buffer(root->commit_root);
2077 root->node = NULL;
2078 root->commit_root = NULL;
2079 }
2080}
2081
Chris Masonaf31f5e2011-11-03 15:17:42 -04002082/* helper to cleanup tree roots */
2083static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2084{
Rashika2e9f5952013-10-31 02:45:20 +05302085 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002086
Rashika2e9f5952013-10-31 02:45:20 +05302087 free_root_extent_buffers(info->dev_root);
2088 free_root_extent_buffers(info->extent_root);
2089 free_root_extent_buffers(info->csum_root);
2090 free_root_extent_buffers(info->quota_root);
2091 free_root_extent_buffers(info->uuid_root);
2092 if (chunk_root)
2093 free_root_extent_buffers(info->chunk_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002094}
2095
Josef Bacik171f6532013-04-24 16:35:41 -04002096static void del_fs_roots(struct btrfs_fs_info *fs_info)
2097{
2098 int ret;
2099 struct btrfs_root *gang[8];
2100 int i;
2101
2102 while (!list_empty(&fs_info->dead_roots)) {
2103 gang[0] = list_entry(fs_info->dead_roots.next,
2104 struct btrfs_root, root_list);
2105 list_del(&gang[0]->root_list);
2106
2107 if (gang[0]->in_radix) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002108 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002109 } else {
2110 free_extent_buffer(gang[0]->node);
2111 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002112 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002113 }
2114 }
2115
2116 while (1) {
2117 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2118 (void **)gang, 0,
2119 ARRAY_SIZE(gang));
2120 if (!ret)
2121 break;
2122 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002123 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002124 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002125
2126 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2127 btrfs_free_log_root_tree(NULL, fs_info);
2128 btrfs_destroy_pinned_extent(fs_info->tree_root,
2129 fs_info->pinned_extents);
2130 }
Josef Bacik171f6532013-04-24 16:35:41 -04002131}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002132
Al Viroad2b2c82011-11-17 01:10:02 -05002133int open_ctree(struct super_block *sb,
2134 struct btrfs_fs_devices *fs_devices,
2135 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002136{
Chris Masondb945352007-10-15 16:15:53 -04002137 u32 sectorsize;
2138 u32 nodesize;
2139 u32 leafsize;
2140 u32 blocksize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002141 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002142 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002143 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002144 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002145 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002146 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002147 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002148 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002149 struct btrfs_root *extent_root;
2150 struct btrfs_root *csum_root;
2151 struct btrfs_root *chunk_root;
2152 struct btrfs_root *dev_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002153 struct btrfs_root *quota_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002154 struct btrfs_root *uuid_root;
Chris Masone02119d2008-09-05 16:13:11 -04002155 struct btrfs_root *log_tree_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002156 int ret;
Yane58ca022008-04-01 11:21:34 -04002157 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002158 int num_backups_tried = 0;
2159 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002160 int max_active;
2161 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Stefan Behrens70f80172013-08-15 17:11:23 +02002162 bool create_uuid_tree;
2163 bool check_uuid_tree;
Chris Mason4543df72008-06-11 21:47:56 -04002164
Al Virof84a8bd2011-11-17 15:57:57 -05002165 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002166 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002167 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002168 err = -ENOMEM;
2169 goto fail;
2170 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002171
2172 ret = init_srcu_struct(&fs_info->subvol_srcu);
2173 if (ret) {
2174 err = ret;
2175 goto fail;
2176 }
2177
2178 ret = setup_bdi(fs_info, &fs_info->bdi);
2179 if (ret) {
2180 err = ret;
2181 goto fail_srcu;
2182 }
2183
Miao Xiee2d84522013-01-29 10:09:20 +00002184 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2185 if (ret) {
2186 err = ret;
2187 goto fail_bdi;
2188 }
2189 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2190 (1 + ilog2(nr_cpu_ids));
2191
Miao Xie963d6782013-01-29 10:10:51 +00002192 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2193 if (ret) {
2194 err = ret;
2195 goto fail_dirty_metadata_bytes;
2196 }
2197
Miao Xiec404e0d2014-01-30 16:46:55 +08002198 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2199 if (ret) {
2200 err = ret;
2201 goto fail_delalloc_bytes;
2202 }
2203
Yan, Zheng76dda932009-09-21 16:00:26 -04002204 fs_info->btree_inode = new_inode(sb);
2205 if (!fs_info->btree_inode) {
2206 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002207 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002208 }
2209
Chris Masona6591712011-07-19 12:04:14 -04002210 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002211
Yan, Zheng76dda932009-09-21 16:00:26 -04002212 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002213 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002214 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002215 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002216 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002217 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002218 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002219 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002220 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002221 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002222 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002223 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002224 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002225 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002226 spin_lock_init(&fs_info->super_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002227 spin_lock_init(&fs_info->buffer_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002228 rwlock_init(&fs_info->tree_mod_log_lock);
Chris Mason75857172011-06-13 20:00:16 -04002229 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002230 mutex_init(&fs_info->delalloc_root_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002231 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002232
Josef Bacik58176a92007-08-29 15:47:34 -04002233 init_completion(&fs_info->kobj_unregister);
Chris Mason0b86a832008-03-24 15:01:56 -04002234 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002235 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002236 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Chris Mason0b86a832008-03-24 15:01:56 -04002237 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002238 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2239 BTRFS_BLOCK_RSV_GLOBAL);
2240 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2241 BTRFS_BLOCK_RSV_DELALLOC);
2242 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2243 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2244 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2245 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2246 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002247 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002248 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002249 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002250 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002251 atomic_set(&fs_info->defrag_running, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002252 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002253 fs_info->sb = sb;
Chris Mason6f568d32008-01-29 16:03:38 -05002254 fs_info->max_inline = 8192 * 1024;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002255 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002256 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002257 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002258 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002259 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
Josef Bacik0a2b2a82014-01-23 10:54:11 -05002260 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
Arne Jansen90519d62011-05-23 14:30:00 +02002261 /* readahead state */
2262 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2263 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002264
Chris Masonb34b0862008-12-19 15:43:22 -05002265 fs_info->thread_pool_size = min_t(unsigned long,
2266 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002267
Miao Xie199c2a92013-05-15 07:48:23 +00002268 INIT_LIST_HEAD(&fs_info->ordered_roots);
2269 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002270 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2271 GFP_NOFS);
2272 if (!fs_info->delayed_root) {
2273 err = -ENOMEM;
2274 goto fail_iput;
2275 }
2276 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002277
Arne Jansena2de7332011-03-08 14:14:00 +01002278 mutex_init(&fs_info->scrub_lock);
2279 atomic_set(&fs_info->scrubs_running, 0);
2280 atomic_set(&fs_info->scrub_pause_req, 0);
2281 atomic_set(&fs_info->scrubs_paused, 0);
2282 atomic_set(&fs_info->scrub_cancel_req, 0);
Miao Xiec404e0d2014-01-30 16:46:55 +08002283 init_waitqueue_head(&fs_info->replace_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002284 init_waitqueue_head(&fs_info->scrub_pause_wait);
Arne Jansena2de7332011-03-08 14:14:00 +01002285 fs_info->scrub_workers_refcnt = 0;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002286#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2287 fs_info->check_integrity_print_mask = 0;
2288#endif
Arne Jansena2de7332011-03-08 14:14:00 +01002289
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002290 spin_lock_init(&fs_info->balance_lock);
2291 mutex_init(&fs_info->balance_mutex);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002292 atomic_set(&fs_info->balance_running, 0);
2293 atomic_set(&fs_info->balance_pause_req, 0);
Ilya Dryomova7e99c62012-01-16 22:04:49 +02002294 atomic_set(&fs_info->balance_cancel_req, 0);
Ilya Dryomovc9e9f972012-01-16 22:04:47 +02002295 fs_info->balance_ctl = NULL;
Ilya Dryomov837d5b62012-01-16 22:04:49 +02002296 init_waitqueue_head(&fs_info->balance_wait_q);
Chris Masona061fc82008-05-07 11:43:44 -04002297
Chris Mason0afbaf82008-04-18 14:17:20 -04002298 sb->s_blocksize = 4096;
2299 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002300 sb->s_bdi = &fs_info->bdi;
Chris Mason0afbaf82008-04-18 14:17:20 -04002301
Yan, Zheng76dda932009-09-21 16:00:26 -04002302 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
Miklos Szeredibfe86842011-10-28 14:13:29 +02002303 set_nlink(fs_info->btree_inode, 1);
Chris Mason0afbaf82008-04-18 14:17:20 -04002304 /*
2305 * we set the i_size on the btree inode to the max possible int.
2306 * the real end of the address space is determined by all of
2307 * the devices in the system
2308 */
2309 fs_info->btree_inode->i_size = OFFSET_MAX;
Chris Masond98237b2007-03-28 13:57:48 -04002310 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Chris Mason04160082008-03-26 10:28:07 -04002311 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2312
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002313 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
Chris Masond1310b22008-01-24 16:13:08 -05002314 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
David Sterbaf993c882011-04-20 23:35:57 +02002315 fs_info->btree_inode->i_mapping);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04002316 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
David Sterbaa8067e02011-04-21 00:34:43 +02002317 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
Chris Mason0da54682007-11-07 21:08:01 -05002318
Chris Masond1310b22008-01-24 16:13:08 -05002319 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2320
Chris Mason0f7d52f2007-04-09 10:42:37 -04002321 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2322 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2323 sizeof(struct btrfs_key));
Josef Bacik72ac3c02012-05-23 14:13:11 -04002324 set_bit(BTRFS_INODE_DUMMY,
2325 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
Filipe David Borba Manana778ba822013-10-06 22:22:33 +01002326 btrfs_insert_inode_hash(fs_info->btree_inode);
Chris Mason39279cc2007-06-12 06:35:45 -04002327
Chris Masone02119d2008-09-05 16:13:11 -04002328 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002329 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002330 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002331
Yan Zheng11833d62009-09-11 16:11:19 -04002332 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002333 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002334 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002335 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002336 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002337 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002338
Chris Masond98237b2007-03-28 13:57:48 -04002339
Chris Mason5a3f23d2009-03-31 13:27:11 -04002340 mutex_init(&fs_info->ordered_operations_mutex);
Josef Bacik9ffba8c2013-08-14 11:33:56 -04002341 mutex_init(&fs_info->ordered_extent_flush_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002342 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002343 mutex_init(&fs_info->chunk_mutex);
2344 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002345 mutex_init(&fs_info->cleaner_mutex);
2346 mutex_init(&fs_info->volume_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002347 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002348 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002349 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002350 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Stefan Behrense922e082012-11-05 17:26:40 +01002351 fs_info->dev_replace.lock_owner = 0;
2352 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2353 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2354 mutex_init(&fs_info->dev_replace.lock_management_lock);
2355 mutex_init(&fs_info->dev_replace.lock);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002356
Arne Jansen416ac512011-09-13 12:56:09 +02002357 spin_lock_init(&fs_info->qgroup_lock);
Wang Shilongf2f6ed32013-04-07 10:50:16 +00002358 mutex_init(&fs_info->qgroup_ioctl_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002359 fs_info->qgroup_tree = RB_ROOT;
2360 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2361 fs_info->qgroup_seq = 1;
2362 fs_info->quota_enabled = 0;
2363 fs_info->pending_quota_state = 0;
Wang Shilong1e8f9152013-05-06 11:03:27 +00002364 fs_info->qgroup_ulist = NULL;
Jan Schmidt2f232032013-04-25 16:04:51 +00002365 mutex_init(&fs_info->qgroup_rescan_lock);
Arne Jansen416ac512011-09-13 12:56:09 +02002366
Chris Masonfa9c0d792009-04-03 09:47:43 -04002367 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2368 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2369
Chris Mason7d9eb122008-07-08 14:19:17 -04002370 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002371 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002372 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002373 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002374
David Woodhouse53b381b2013-01-29 18:40:14 -05002375 ret = btrfs_alloc_stripe_hash_table(fs_info);
2376 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002377 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002378 goto fail_alloc;
2379 }
2380
Chris Mason0b86a832008-03-24 15:01:56 -04002381 __setup_root(4096, 4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002382 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002383
Chris Mason3c4bb262012-03-27 18:56:56 -04002384 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002385
2386 /*
2387 * Read super block and check the signature bytes only
2388 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002389 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Dave Young20b45072011-01-08 10:09:13 +00002390 if (!bh) {
2391 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002392 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002393 }
Chris Mason39279cc2007-06-12 06:35:45 -04002394
David Sterba1104a882013-03-06 15:57:46 +01002395 /*
2396 * We want to check superblock checksum, the type is stored inside.
2397 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2398 */
2399 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002400 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002401 err = -EINVAL;
2402 goto fail_alloc;
2403 }
2404
2405 /*
2406 * super_copy is zeroed at allocation time and we never touch the
2407 * following bytes up to INFO_SIZE, the checksum is calculated from
2408 * the whole block of INFO_SIZE
2409 */
David Sterba6c417612011-04-13 15:41:04 +02002410 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2411 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2412 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002413 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002414
David Sterba6c417612011-04-13 15:41:04 +02002415 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002416
David Sterba1104a882013-03-06 15:57:46 +01002417 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2418 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002419 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002420 err = -EINVAL;
2421 goto fail_alloc;
2422 }
2423
David Sterba6c417612011-04-13 15:41:04 +02002424 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002425 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002426 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002427
liuboacce9522011-01-06 19:30:25 +08002428 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002429 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2430 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002431
Liu Bo75e7cb72011-03-22 10:12:20 +00002432 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002433 * run through our array of backup supers and setup
2434 * our ring pointer to the oldest one
2435 */
2436 generation = btrfs_super_generation(disk_super);
2437 find_oldest_super_backup(fs_info, generation);
2438
2439 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002440 * In the long term, we'll store the compression type in the super
2441 * block, and it'll be used for per file compression control.
2442 */
2443 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2444
Yan Zheng2b820322008-11-17 21:11:30 -05002445 ret = btrfs_parse_options(tree_root, options);
2446 if (ret) {
2447 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002448 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002449 }
Chris Masondfe25022008-05-13 13:46:40 -04002450
Josef Bacikf2b636e2008-12-02 06:36:08 -05002451 features = btrfs_super_incompat_flags(disk_super) &
2452 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2453 if (features) {
2454 printk(KERN_ERR "BTRFS: couldn't mount because of "
2455 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002456 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002457 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002458 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002459 }
2460
Chris Mason727011e2010-08-06 13:21:20 -04002461 if (btrfs_super_leafsize(disk_super) !=
2462 btrfs_super_nodesize(disk_super)) {
2463 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2464 "blocksizes don't match. node %d leaf %d\n",
2465 btrfs_super_nodesize(disk_super),
2466 btrfs_super_leafsize(disk_super));
2467 err = -EINVAL;
2468 goto fail_alloc;
2469 }
2470 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2471 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2472 "blocksize (%d) was too large\n",
2473 btrfs_super_leafsize(disk_super));
2474 err = -EINVAL;
2475 goto fail_alloc;
2476 }
2477
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002478 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002479 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002480 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002481 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002482
Josef Bacik3173a182013-03-07 14:22:04 -05002483 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
Frank Holtonefe120a2013-12-20 11:37:06 -05002484 printk(KERN_ERR "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002485
Chris Mason727011e2010-08-06 13:21:20 -04002486 /*
2487 * flag our filesystem as having big metadata blocks if
2488 * they are bigger than the page size
2489 */
2490 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2491 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002492 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002493 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2494 }
2495
Chris Masonbc3f1162012-03-29 17:02:47 -04002496 nodesize = btrfs_super_nodesize(disk_super);
2497 leafsize = btrfs_super_leafsize(disk_super);
2498 sectorsize = btrfs_super_sectorsize(disk_super);
2499 stripesize = btrfs_super_stripesize(disk_super);
Miao Xiee2d84522013-01-29 10:09:20 +00002500 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002501 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002502
2503 /*
2504 * mixed block groups end up with duplicate but slightly offset
2505 * extent buffers for the same range. It leads to corruptions
2506 */
2507 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2508 (sectorsize != leafsize)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002509 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002510 "are not allowed for mixed block groups on %s\n",
2511 sb->s_id);
2512 goto fail_alloc;
2513 }
2514
Miao Xieceda0862013-04-11 10:30:16 +00002515 /*
2516 * Needn't use the lock because there is no other task which will
2517 * update the flag.
2518 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002519 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002520
Josef Bacikf2b636e2008-12-02 06:36:08 -05002521 features = btrfs_super_compat_ro_flags(disk_super) &
2522 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2523 if (!(sb->s_flags & MS_RDONLY) && features) {
2524 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2525 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002526 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002527 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002528 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002529 }
Chris Mason61d92c32009-10-02 19:11:56 -04002530
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002531 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002532
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002533 fs_info->workers =
2534 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2535 max_active, 16);
Chris Masonc8b97812008-10-29 14:49:59 -04002536
Qu Wenruoafe3d242014-02-28 10:46:07 +08002537 fs_info->delalloc_workers =
2538 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
Chris Mason771ed682008-11-06 22:02:51 -05002539
Qu Wenruoa44903a2014-02-28 10:46:09 +08002540 fs_info->flush_workers =
2541 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002542
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002543 fs_info->caching_workers =
2544 btrfs_alloc_workqueue("cache", flags, max_active, 0);
Josef Bacikbab39bf2011-06-30 14:42:28 -04002545
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002546 /*
2547 * a higher idle thresh on the submit workers makes it much more
Chris Mason61b49442008-07-31 15:42:53 -04002548 * likely that bios will be send down in a sane order to the
2549 * devices
2550 */
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002551 fs_info->submit_workers =
2552 btrfs_alloc_workqueue("submit", flags,
2553 min_t(u64, fs_devices->num_devices,
2554 max_active), 64);
Chris Mason53863232008-08-15 15:34:18 -04002555
Qu Wenruodc6e3202014-02-28 10:46:14 +08002556 fs_info->fixup_workers =
2557 btrfs_alloc_workqueue("fixup", flags, 1, 0);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002558
2559 /*
2560 * endios are largely parallel and should have a very
2561 * low idle thresh
2562 */
2563 fs_info->endio_workers =
2564 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2565 fs_info->endio_meta_workers =
2566 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2567 fs_info->endio_meta_write_workers =
2568 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2569 fs_info->endio_raid56_workers =
2570 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002571 fs_info->rmw_workers =
2572 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002573 fs_info->endio_write_workers =
2574 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2575 fs_info->endio_freespace_worker =
2576 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002577 fs_info->delayed_workers =
2578 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002579 fs_info->readahead_workers =
2580 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002581 fs_info->qgroup_rescan_workers =
2582 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
Chris Mason61b49442008-07-31 15:42:53 -04002583
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002584 if (!(fs_info->workers && fs_info->delalloc_workers &&
Qu Wenruofccb5d82014-02-28 10:46:10 +08002585 fs_info->submit_workers && fs_info->flush_workers &&
2586 fs_info->endio_workers && fs_info->endio_meta_workers &&
2587 fs_info->endio_meta_write_workers &&
2588 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002589 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
Qu Wenruodc6e3202014-02-28 10:46:14 +08002590 fs_info->caching_workers && fs_info->readahead_workers &&
Qu Wenruofc97fab2014-02-28 10:46:16 +08002591 fs_info->fixup_workers && fs_info->delayed_workers &&
2592 fs_info->qgroup_rescan_workers)) {
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002593 err = -ENOMEM;
2594 goto fail_sb_buffer;
2595 }
Chris Mason4543df72008-06-11 21:47:56 -04002596
Chris Mason4575c9c2008-04-18 16:13:31 -04002597 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002598 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2599 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002600
Chris Masondb945352007-10-15 16:15:53 -04002601 tree_root->nodesize = nodesize;
2602 tree_root->leafsize = leafsize;
2603 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002604 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002605
2606 sb->s_blocksize = sectorsize;
2607 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002608
Qu Wenruo3cae2102013-07-16 11:19:18 +08002609 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002610 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002611 goto fail_sb_buffer;
2612 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002613
Liu Bo8d082fb2012-04-03 09:56:53 +08002614 if (sectorsize != PAGE_SIZE) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002615 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
Liu Bo8d082fb2012-04-03 09:56:53 +08002616 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002617 goto fail_sb_buffer;
2618 }
2619
Chris Mason925baed2008-06-25 16:01:30 -04002620 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002621 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002622 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002623 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002624 printk(KERN_WARNING "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002625 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002626 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002627 }
Chris Mason0b86a832008-03-24 15:01:56 -04002628
2629 blocksize = btrfs_level_size(tree_root,
2630 btrfs_super_chunk_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002631 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002632
2633 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2634 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2635
2636 chunk_root->node = read_tree_block(chunk_root,
2637 btrfs_super_chunk_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002638 blocksize, generation);
Josef Bacik416bc652013-04-23 14:17:42 -04002639 if (!chunk_root->node ||
2640 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002641 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002642 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002643 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002644 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002645 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2646 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002647
Chris Masone17cade2008-04-15 15:41:47 -04002648 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002649 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002650
Chris Mason0b86a832008-03-24 15:01:56 -04002651 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002652 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002653 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002654 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002655 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002656 }
Chris Mason0b86a832008-03-24 15:01:56 -04002657
Stefan Behrens8dabb742012-11-06 13:15:27 +01002658 /*
2659 * keep the device that is marked to be the target device for the
2660 * dev_replace procedure
2661 */
2662 btrfs_close_extra_devices(fs_info, fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002663
Chris Masona6b0d5c2012-02-20 20:53:43 -05002664 if (!fs_devices->latest_bdev) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002665 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002666 sb->s_id);
2667 goto fail_tree_roots;
2668 }
2669
Chris Masonaf31f5e2011-11-03 15:17:42 -04002670retry_root_backup:
Chris Masondb945352007-10-15 16:15:53 -04002671 blocksize = btrfs_level_size(tree_root,
2672 btrfs_super_root_level(disk_super));
Yan Zheng84234f32008-10-29 14:49:05 -04002673 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002674
Chris Masone20d96d2007-03-22 12:13:20 -04002675 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002676 btrfs_super_root(disk_super),
Yan Zheng84234f32008-10-29 14:49:05 -04002677 blocksize, generation);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002678 if (!tree_root->node ||
2679 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002680 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002681 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002682
2683 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002684 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002685
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002686 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2687 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002688 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002689
Miao Xiecb517ea2013-05-15 07:48:19 +00002690 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2691 location.type = BTRFS_ROOT_ITEM_KEY;
2692 location.offset = 0;
2693
2694 extent_root = btrfs_read_tree_root(tree_root, &location);
2695 if (IS_ERR(extent_root)) {
2696 ret = PTR_ERR(extent_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002697 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002698 }
Chris Mason0b86a832008-03-24 15:01:56 -04002699 extent_root->track_dirty = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002700 fs_info->extent_root = extent_root;
Chris Mason0b86a832008-03-24 15:01:56 -04002701
Miao Xiecb517ea2013-05-15 07:48:19 +00002702 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2703 dev_root = btrfs_read_tree_root(tree_root, &location);
2704 if (IS_ERR(dev_root)) {
2705 ret = PTR_ERR(dev_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002706 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002707 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002708 dev_root->track_dirty = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002709 fs_info->dev_root = dev_root;
2710 btrfs_init_devices_late(fs_info);
Chris Mason3768f362007-03-13 16:47:54 -04002711
Miao Xiecb517ea2013-05-15 07:48:19 +00002712 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2713 csum_root = btrfs_read_tree_root(tree_root, &location);
2714 if (IS_ERR(csum_root)) {
2715 ret = PTR_ERR(csum_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002716 goto recovery_tree_root;
Miao Xiecb517ea2013-05-15 07:48:19 +00002717 }
Chris Masond20f7042008-12-08 16:58:54 -05002718 csum_root->track_dirty = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002719 fs_info->csum_root = csum_root;
Chris Masond20f7042008-12-08 16:58:54 -05002720
Miao Xiecb517ea2013-05-15 07:48:19 +00002721 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2722 quota_root = btrfs_read_tree_root(tree_root, &location);
2723 if (!IS_ERR(quota_root)) {
Arne Jansenbcef60f2011-09-13 15:23:30 +02002724 quota_root->track_dirty = 1;
2725 fs_info->quota_enabled = 1;
2726 fs_info->pending_quota_state = 1;
Miao Xiecb517ea2013-05-15 07:48:19 +00002727 fs_info->quota_root = quota_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002728 }
2729
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002730 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2731 uuid_root = btrfs_read_tree_root(tree_root, &location);
2732 if (IS_ERR(uuid_root)) {
2733 ret = PTR_ERR(uuid_root);
2734 if (ret != -ENOENT)
2735 goto recovery_tree_root;
2736 create_uuid_tree = true;
Stefan Behrens70f80172013-08-15 17:11:23 +02002737 check_uuid_tree = false;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002738 } else {
2739 uuid_root->track_dirty = 1;
2740 fs_info->uuid_root = uuid_root;
Stefan Behrens70f80172013-08-15 17:11:23 +02002741 create_uuid_tree = false;
2742 check_uuid_tree =
2743 generation != btrfs_super_uuid_tree_generation(disk_super);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002744 }
2745
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002746 fs_info->generation = generation;
2747 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002748
Ilya Dryomov68310a52012-06-22 12:24:12 -06002749 ret = btrfs_recover_balance(fs_info);
2750 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002751 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002752 goto fail_block_groups;
2753 }
2754
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002755 ret = btrfs_init_dev_stats(fs_info);
2756 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002757 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002758 ret);
2759 goto fail_block_groups;
2760 }
2761
Stefan Behrens8dabb742012-11-06 13:15:27 +01002762 ret = btrfs_init_dev_replace(fs_info);
2763 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002764 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002765 goto fail_block_groups;
2766 }
2767
2768 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2769
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002770 ret = btrfs_sysfs_add_one(fs_info);
2771 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002772 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002773 goto fail_block_groups;
2774 }
2775
liuboc59021f2011-03-07 02:13:14 +00002776 ret = btrfs_init_space_info(fs_info);
2777 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002778 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002779 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002780 }
2781
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002782 ret = btrfs_read_block_groups(extent_root);
2783 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002784 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002785 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002786 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002787 fs_info->num_tolerated_disk_barrier_failures =
2788 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002789 if (fs_info->fs_devices->missing_devices >
2790 fs_info->num_tolerated_disk_barrier_failures &&
2791 !(sb->s_flags & MS_RDONLY)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002792 printk(KERN_WARNING "BTRFS: "
2793 "too many missing devices, writeable mount is not allowed\n");
Anand Jain2365dd32014-01-22 11:15:51 +08002794 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002795 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002796
Chris Masona74a4b92008-06-25 16:01:31 -04002797 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2798 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002799 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08002800 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04002801
2802 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2803 tree_root,
2804 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002805 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04002806 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04002807
Chris Masonc2898112009-06-10 09:51:32 -04002808 if (!btrfs_test_opt(tree_root, SSD) &&
2809 !btrfs_test_opt(tree_root, NOSSD) &&
2810 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002811 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04002812 "mode\n");
2813 btrfs_set_opt(fs_info->mount_opt, SSD);
2814 }
2815
Qu Wenruo3818aea2014-01-13 13:36:06 +08002816 /* Set the real inode map cache flag */
2817 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2818 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2819
Stefan Behrens21adbd52011-11-09 13:44:05 +01002820#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2821 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2822 ret = btrfsic_mount(tree_root, fs_devices,
2823 btrfs_test_opt(tree_root,
2824 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2825 1 : 0,
2826 fs_info->check_integrity_print_mask);
2827 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05002828 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01002829 " integrity check module %s\n", sb->s_id);
2830 }
2831#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02002832 ret = btrfs_read_qgroup_config(fs_info);
2833 if (ret)
2834 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01002835
liuboacce9522011-01-06 19:30:25 +08002836 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06002837 if (btrfs_super_log_root(disk_super) != 0) {
Chris Masone02119d2008-09-05 16:13:11 -04002838 u64 bytenr = btrfs_super_log_root(disk_super);
Chris Masond18a2c42008-04-04 15:40:00 -04002839
Chris Mason7c2ca462008-11-19 15:13:35 -05002840 if (fs_devices->rw_devices == 0) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002841 printk(KERN_WARNING "BTRFS: log replay required "
Chris Masond3977122009-01-05 21:25:51 -05002842 "on RO media\n");
Chris Mason7c2ca462008-11-19 15:13:35 -05002843 err = -EIO;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002844 goto fail_qgroup;
Chris Mason7c2ca462008-11-19 15:13:35 -05002845 }
Chris Masone02119d2008-09-05 16:13:11 -04002846 blocksize =
2847 btrfs_level_size(tree_root,
2848 btrfs_super_log_root_level(disk_super));
2849
Al Viro6f07e422011-11-17 00:46:16 -05002850 log_tree_root = btrfs_alloc_root(fs_info);
Dan Carpenter676e4c82010-05-29 09:43:07 +00002851 if (!log_tree_root) {
2852 err = -ENOMEM;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002853 goto fail_qgroup;
Dan Carpenter676e4c82010-05-29 09:43:07 +00002854 }
Chris Masone02119d2008-09-05 16:13:11 -04002855
2856 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2857 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2858
2859 log_tree_root->node = read_tree_block(tree_root, bytenr,
Yan Zheng84234f32008-10-29 14:49:05 -04002860 blocksize,
2861 generation + 1);
Josef Bacik416bc652013-04-23 14:17:42 -04002862 if (!log_tree_root->node ||
2863 !extent_buffer_uptodate(log_tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002864 printk(KERN_ERR "BTRFS: failed to read log tree\n");
Josef Bacik416bc652013-04-23 14:17:42 -04002865 free_extent_buffer(log_tree_root->node);
2866 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002867 goto fail_qgroup;
Josef Bacik416bc652013-04-23 14:17:42 -04002868 }
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002869 /* returns with log_tree_root freed on success */
Chris Masone02119d2008-09-05 16:13:11 -04002870 ret = btrfs_recover_log_trees(log_tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002871 if (ret) {
2872 btrfs_error(tree_root->fs_info, ret,
2873 "Failed to recover log tree");
2874 free_extent_buffer(log_tree_root->node);
2875 kfree(log_tree_root);
Wang Shilong28c16cb2014-04-23 19:33:35 +08002876 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002877 }
Yan Zhenge556ce22008-11-20 10:25:19 -05002878
2879 if (sb->s_flags & MS_RDONLY) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002880 ret = btrfs_commit_super(tree_root);
2881 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002882 goto fail_qgroup;
Yan Zhenge556ce22008-11-20 10:25:19 -05002883 }
Chris Masone02119d2008-09-05 16:13:11 -04002884 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002885
Yan, Zheng76dda932009-09-21 16:00:26 -04002886 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01002887 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002888 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04002889
Chris Mason7c2ca462008-11-19 15:13:35 -05002890 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04002891 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06002892 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08002893 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04002894
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002895 ret = btrfs_recover_relocation(tree_root);
Miao Xied7ce5842010-02-02 08:46:44 +00002896 if (ret < 0) {
2897 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05002898 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00002899 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02002900 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00002901 }
Chris Mason7c2ca462008-11-19 15:13:35 -05002902 }
Zheng Yan1a40e232008-09-26 10:09:34 -04002903
Chris Mason3de45862008-11-17 21:02:50 -05002904 location.objectid = BTRFS_FS_TREE_OBJECTID;
2905 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00002906 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05002907
Chris Mason3de45862008-11-17 21:02:50 -05002908 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002909 if (IS_ERR(fs_info->fs_root)) {
2910 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02002911 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00002912 }
Chris Masonc2898112009-06-10 09:51:32 -04002913
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002914 if (sb->s_flags & MS_RDONLY)
2915 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02002916
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002917 down_read(&fs_info->cleanup_work_sem);
2918 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2919 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00002920 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002921 close_ctree(tree_root);
2922 return ret;
2923 }
2924 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02002925
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002926 ret = btrfs_resume_balance_async(fs_info);
2927 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002928 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06002929 close_ctree(tree_root);
2930 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00002931 }
2932
Stefan Behrens8dabb742012-11-06 13:15:27 +01002933 ret = btrfs_resume_dev_replace_async(fs_info);
2934 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002935 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01002936 close_ctree(tree_root);
2937 return ret;
2938 }
2939
Jan Schmidtb382a322013-05-28 15:47:24 +00002940 btrfs_qgroup_rescan_resume(fs_info);
2941
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002942 if (create_uuid_tree) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002943 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002944 ret = btrfs_create_uuid_tree(fs_info);
2945 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002946 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002947 ret);
2948 close_ctree(tree_root);
2949 return ret;
2950 }
Stefan Behrensf420ee12013-08-15 17:11:24 +02002951 } else if (check_uuid_tree ||
2952 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002953 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02002954 ret = btrfs_check_uuid_tree(fs_info);
2955 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002956 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02002957 ret);
2958 close_ctree(tree_root);
2959 return ret;
2960 }
2961 } else {
2962 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002963 }
2964
Al Viroad2b2c82011-11-17 01:10:02 -05002965 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04002966
Arne Jansenbcef60f2011-09-13 15:23:30 +02002967fail_qgroup:
2968 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05002969fail_trans_kthread:
2970 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04002971 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacik7b5ff902013-06-06 10:29:40 -04002972 del_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04002973fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04002974 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05002975
2976 /*
2977 * make sure we're done with the btree inode before we stop our
2978 * kthreads
2979 */
2980 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05002981
Anand Jain2365dd32014-01-22 11:15:51 +08002982fail_sysfs:
2983 btrfs_sysfs_remove_one(fs_info);
2984
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002985fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04002986 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002987 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002988
2989fail_tree_roots:
2990 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00002991 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002992
Chris Mason39279cc2007-06-12 06:35:45 -04002993fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00002994 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08002995fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04002996fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02002997 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2998
Chris Mason4543df72008-06-11 21:47:56 -04002999 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003000fail_bio_counter:
3001 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003002fail_delalloc_bytes:
3003 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003004fail_dirty_metadata_bytes:
3005 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003006fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003007 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003008fail_srcu:
3009 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003010fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003011 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003012 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003013 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003014
3015recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003016 if (!btrfs_test_opt(tree_root, RECOVERY))
3017 goto fail_tree_roots;
3018
3019 free_root_pointers(fs_info, 0);
3020
3021 /* don't use the log in recovery mode, it won't be valid */
3022 btrfs_set_super_log_root(disk_super, 0);
3023
3024 /* we can't trust the free space cache either */
3025 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3026
3027 ret = next_root_backup(fs_info, fs_info->super_copy,
3028 &num_backups_tried, &backup_index);
3029 if (ret == -1)
3030 goto fail_block_groups;
3031 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003032}
3033
Chris Masonf2984462008-04-10 16:19:33 -04003034static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3035{
Chris Masonf2984462008-04-10 16:19:33 -04003036 if (uptodate) {
3037 set_buffer_uptodate(bh);
3038 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003039 struct btrfs_device *device = (struct btrfs_device *)
3040 bh->b_private;
3041
Frank Holtonefe120a2013-12-20 11:37:06 -05003042 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
Josef Bacik606686e2012-06-04 14:03:51 -04003043 "I/O error on %s\n",
3044 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003045 /* note, we dont' set_buffer_write_io_error because we have
3046 * our own ways of dealing with the IO errors
3047 */
Chris Masonf2984462008-04-10 16:19:33 -04003048 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003049 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003050 }
3051 unlock_buffer(bh);
3052 put_bh(bh);
3053}
3054
Yan Zhenga512bbf2008-12-08 16:46:26 -05003055struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3056{
3057 struct buffer_head *bh;
3058 struct buffer_head *latest = NULL;
3059 struct btrfs_super_block *super;
3060 int i;
3061 u64 transid = 0;
3062 u64 bytenr;
3063
3064 /* we would like to check all the supers, but that would make
3065 * a btrfs mount succeed after a mkfs from a different FS.
3066 * So, we need to add a special mount option to scan for
3067 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3068 */
3069 for (i = 0; i < 1; i++) {
3070 bytenr = btrfs_sb_offset(i);
Anand Jain8068a472013-07-26 01:29:35 +08003071 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3072 i_size_read(bdev->bd_inode))
Yan Zhenga512bbf2008-12-08 16:46:26 -05003073 break;
Anand Jain8068a472013-07-26 01:29:35 +08003074 bh = __bread(bdev, bytenr / 4096,
3075 BTRFS_SUPER_INFO_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003076 if (!bh)
3077 continue;
3078
3079 super = (struct btrfs_super_block *)bh->b_data;
3080 if (btrfs_super_bytenr(super) != bytenr ||
Qu Wenruo3cae2102013-07-16 11:19:18 +08003081 btrfs_super_magic(super) != BTRFS_MAGIC) {
Yan Zhenga512bbf2008-12-08 16:46:26 -05003082 brelse(bh);
3083 continue;
3084 }
3085
3086 if (!latest || btrfs_super_generation(super) > transid) {
3087 brelse(latest);
3088 latest = bh;
3089 transid = btrfs_super_generation(super);
3090 } else {
3091 brelse(bh);
3092 }
3093 }
3094 return latest;
3095}
3096
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003097/*
3098 * this should be called twice, once with wait == 0 and
3099 * once with wait == 1. When wait == 0 is done, all the buffer heads
3100 * we write are pinned.
3101 *
3102 * They are released when wait == 1 is done.
3103 * max_mirrors must be the same for both runs, and it indicates how
3104 * many supers on this one device should be written.
3105 *
3106 * max_mirrors == 0 means to write them all.
3107 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003108static int write_dev_supers(struct btrfs_device *device,
3109 struct btrfs_super_block *sb,
3110 int do_barriers, int wait, int max_mirrors)
3111{
3112 struct buffer_head *bh;
3113 int i;
3114 int ret;
3115 int errors = 0;
3116 u32 crc;
3117 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003118
3119 if (max_mirrors == 0)
3120 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3121
Yan Zhenga512bbf2008-12-08 16:46:26 -05003122 for (i = 0; i < max_mirrors; i++) {
3123 bytenr = btrfs_sb_offset(i);
3124 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3125 break;
3126
3127 if (wait) {
3128 bh = __find_get_block(device->bdev, bytenr / 4096,
3129 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003130 if (!bh) {
3131 errors++;
3132 continue;
3133 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003134 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003135 if (!buffer_uptodate(bh))
3136 errors++;
3137
3138 /* drop our reference */
3139 brelse(bh);
3140
3141 /* drop the reference from the wait == 0 run */
3142 brelse(bh);
3143 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003144 } else {
3145 btrfs_set_super_bytenr(sb, bytenr);
3146
3147 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003148 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003149 BTRFS_CSUM_SIZE, crc,
3150 BTRFS_SUPER_INFO_SIZE -
3151 BTRFS_CSUM_SIZE);
3152 btrfs_csum_final(crc, sb->csum);
3153
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003154 /*
3155 * one reference for us, and we leave it for the
3156 * caller
3157 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003158 bh = __getblk(device->bdev, bytenr / 4096,
3159 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003160 if (!bh) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003161 printk(KERN_ERR "BTRFS: couldn't get super "
Josef Bacik634554d2013-04-29 10:05:57 -04003162 "buffer head for bytenr %Lu\n", bytenr);
3163 errors++;
3164 continue;
3165 }
3166
Yan Zhenga512bbf2008-12-08 16:46:26 -05003167 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3168
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003169 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003170 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003171
3172 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003173 lock_buffer(bh);
3174 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003175 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003176 }
3177
Chris Mason387125f2011-11-18 15:07:51 -05003178 /*
3179 * we fua the first super. The others we allow
3180 * to go down lazy.
3181 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003182 if (i == 0)
3183 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3184 else
3185 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003186 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003187 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003188 }
3189 return errors < i ? 0 : -1;
3190}
3191
Chris Mason387125f2011-11-18 15:07:51 -05003192/*
3193 * endio for the write_dev_flush, this will wake anyone waiting
3194 * for the barrier when it is done
3195 */
3196static void btrfs_end_empty_barrier(struct bio *bio, int err)
3197{
3198 if (err) {
3199 if (err == -EOPNOTSUPP)
3200 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3201 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3202 }
3203 if (bio->bi_private)
3204 complete(bio->bi_private);
3205 bio_put(bio);
3206}
3207
3208/*
3209 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3210 * sent down. With wait == 1, it waits for the previous flush.
3211 *
3212 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3213 * capable
3214 */
3215static int write_dev_flush(struct btrfs_device *device, int wait)
3216{
3217 struct bio *bio;
3218 int ret = 0;
3219
3220 if (device->nobarriers)
3221 return 0;
3222
3223 if (wait) {
3224 bio = device->flush_bio;
3225 if (!bio)
3226 return 0;
3227
3228 wait_for_completion(&device->flush_wait);
3229
3230 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003231 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
Josef Bacik606686e2012-06-04 14:03:51 -04003232 rcu_str_deref(device->name));
Chris Mason387125f2011-11-18 15:07:51 -05003233 device->nobarriers = 1;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003234 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
Chris Mason387125f2011-11-18 15:07:51 -05003235 ret = -EIO;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003236 btrfs_dev_stat_inc_and_print(device,
3237 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003238 }
3239
3240 /* drop the reference from the wait == 0 run */
3241 bio_put(bio);
3242 device->flush_bio = NULL;
3243
3244 return ret;
3245 }
3246
3247 /*
3248 * one reference for us, and we leave it for the
3249 * caller
3250 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003251 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003252 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003253 if (!bio)
3254 return -ENOMEM;
3255
3256 bio->bi_end_io = btrfs_end_empty_barrier;
3257 bio->bi_bdev = device->bdev;
3258 init_completion(&device->flush_wait);
3259 bio->bi_private = &device->flush_wait;
3260 device->flush_bio = bio;
3261
3262 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003263 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003264
3265 return 0;
3266}
3267
3268/*
3269 * send an empty flush down to each device in parallel,
3270 * then wait for them
3271 */
3272static int barrier_all_devices(struct btrfs_fs_info *info)
3273{
3274 struct list_head *head;
3275 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003276 int errors_send = 0;
3277 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003278 int ret;
3279
3280 /* send down all the barriers */
3281 head = &info->fs_devices->devices;
3282 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003283 if (dev->missing)
3284 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003285 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003286 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003287 continue;
3288 }
3289 if (!dev->in_fs_metadata || !dev->writeable)
3290 continue;
3291
3292 ret = write_dev_flush(dev, 0);
3293 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003294 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003295 }
3296
3297 /* wait for all the barriers */
3298 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003299 if (dev->missing)
3300 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003301 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003302 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003303 continue;
3304 }
3305 if (!dev->in_fs_metadata || !dev->writeable)
3306 continue;
3307
3308 ret = write_dev_flush(dev, 1);
3309 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003310 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003311 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003312 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3313 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003314 return -EIO;
3315 return 0;
3316}
3317
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003318int btrfs_calc_num_tolerated_disk_barrier_failures(
3319 struct btrfs_fs_info *fs_info)
3320{
3321 struct btrfs_ioctl_space_info space;
3322 struct btrfs_space_info *sinfo;
3323 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3324 BTRFS_BLOCK_GROUP_SYSTEM,
3325 BTRFS_BLOCK_GROUP_METADATA,
3326 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3327 int num_types = 4;
3328 int i;
3329 int c;
3330 int num_tolerated_disk_barrier_failures =
3331 (int)fs_info->fs_devices->num_devices;
3332
3333 for (i = 0; i < num_types; i++) {
3334 struct btrfs_space_info *tmp;
3335
3336 sinfo = NULL;
3337 rcu_read_lock();
3338 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3339 if (tmp->flags == types[i]) {
3340 sinfo = tmp;
3341 break;
3342 }
3343 }
3344 rcu_read_unlock();
3345
3346 if (!sinfo)
3347 continue;
3348
3349 down_read(&sinfo->groups_sem);
3350 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3351 if (!list_empty(&sinfo->block_groups[c])) {
3352 u64 flags;
3353
3354 btrfs_get_block_group_info(
3355 &sinfo->block_groups[c], &space);
3356 if (space.total_bytes == 0 ||
3357 space.used_bytes == 0)
3358 continue;
3359 flags = space.flags;
3360 /*
3361 * return
3362 * 0: if dup, single or RAID0 is configured for
3363 * any of metadata, system or data, else
3364 * 1: if RAID5 is configured, or if RAID1 or
3365 * RAID10 is configured and only two mirrors
3366 * are used, else
3367 * 2: if RAID6 is configured, else
3368 * num_mirrors - 1: if RAID1 or RAID10 is
3369 * configured and more than
3370 * 2 mirrors are used.
3371 */
3372 if (num_tolerated_disk_barrier_failures > 0 &&
3373 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3374 BTRFS_BLOCK_GROUP_RAID0)) ||
3375 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3376 == 0)))
3377 num_tolerated_disk_barrier_failures = 0;
David Woodhouse53b381b2013-01-29 18:40:14 -05003378 else if (num_tolerated_disk_barrier_failures > 1) {
3379 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3380 BTRFS_BLOCK_GROUP_RAID5 |
3381 BTRFS_BLOCK_GROUP_RAID10)) {
3382 num_tolerated_disk_barrier_failures = 1;
3383 } else if (flags &
Henrik Nordvik15b0a892013-04-29 18:09:23 +00003384 BTRFS_BLOCK_GROUP_RAID6) {
David Woodhouse53b381b2013-01-29 18:40:14 -05003385 num_tolerated_disk_barrier_failures = 2;
3386 }
3387 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003388 }
3389 }
3390 up_read(&sinfo->groups_sem);
3391 }
3392
3393 return num_tolerated_disk_barrier_failures;
3394}
3395
Eric Sandeen48a3b632013-04-25 20:41:01 +00003396static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003397{
Chris Masone5e9a522009-06-10 15:17:02 -04003398 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003399 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003400 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003401 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003402 int ret;
3403 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003404 int max_errors;
3405 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003406 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003407
3408 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003409 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003410
David Sterba6c417612011-04-13 15:41:04 +02003411 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003412 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003413
Chris Mason174ba502011-05-27 10:03:58 -04003414 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003415 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003416 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003417
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003418 if (do_barriers) {
3419 ret = barrier_all_devices(root->fs_info);
3420 if (ret) {
3421 mutex_unlock(
3422 &root->fs_info->fs_devices->device_list_mutex);
3423 btrfs_error(root->fs_info, ret,
3424 "errors while submitting device barriers.");
3425 return ret;
3426 }
3427 }
Chris Mason387125f2011-11-18 15:07:51 -05003428
Xiao Guangrong1f781602011-04-20 10:09:16 +00003429 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003430 if (!dev->bdev) {
3431 total_errors++;
3432 continue;
3433 }
Yan Zheng2b820322008-11-17 21:11:30 -05003434 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003435 continue;
3436
Yan Zheng2b820322008-11-17 21:11:30 -05003437 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003438 btrfs_set_stack_device_type(dev_item, dev->type);
3439 btrfs_set_stack_device_id(dev_item, dev->devid);
3440 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3441 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3442 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3443 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3444 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3445 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003446 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003447
Chris Masona061fc82008-05-07 11:43:44 -04003448 flags = btrfs_super_flags(sb);
3449 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003450
Yan Zhenga512bbf2008-12-08 16:46:26 -05003451 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003452 if (ret)
3453 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003454 }
Chris Masona236aed2008-04-29 09:38:00 -04003455 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003456 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003457 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003458 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003459
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003460 /* FUA is masked off if unsupported and can't be the reason */
3461 btrfs_error(root->fs_info, -EIO,
3462 "%d errors while writing supers", total_errors);
3463 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003464 }
Chris Masonf2984462008-04-10 16:19:33 -04003465
Yan Zhenga512bbf2008-12-08 16:46:26 -05003466 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003467 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003468 if (!dev->bdev)
3469 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003470 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003471 continue;
3472
Yan Zhenga512bbf2008-12-08 16:46:26 -05003473 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3474 if (ret)
3475 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003476 }
Chris Mason174ba502011-05-27 10:03:58 -04003477 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003478 if (total_errors > max_errors) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003479 btrfs_error(root->fs_info, -EIO,
3480 "%d errors while writing supers", total_errors);
3481 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003482 }
Chris Masonf2984462008-04-10 16:19:33 -04003483 return 0;
3484}
3485
Yan Zhenga512bbf2008-12-08 16:46:26 -05003486int write_ctree_super(struct btrfs_trans_handle *trans,
3487 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003488{
Rashikaf570e7572013-10-31 16:50:42 +05303489 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003490}
3491
Miao Xiecb517ea2013-05-15 07:48:19 +00003492/* Drop a fs root from the radix tree and free it. */
3493void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3494 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003495{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003496 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003497 radix_tree_delete(&fs_info->fs_roots_radix,
3498 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003499 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003500
3501 if (btrfs_root_refs(&root->root_item) == 0)
3502 synchronize_srcu(&fs_info->subvol_srcu);
3503
Liu Bo1a4319c2014-01-13 19:53:53 +08003504 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003505 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003506
Li Zefan581bb052011-04-20 10:06:11 +08003507 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3508 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003509 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003510}
3511
3512static void free_fs_root(struct btrfs_root *root)
3513{
Li Zefan82d59022011-04-20 10:33:24 +08003514 iput(root->cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003515 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003516 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3517 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003518 if (root->anon_dev)
3519 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003520 if (root->subv_writers)
3521 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003522 free_extent_buffer(root->node);
3523 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003524 kfree(root->free_ino_ctl);
3525 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003526 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003527 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003528}
3529
Miao Xiecb517ea2013-05-15 07:48:19 +00003530void btrfs_free_fs_root(struct btrfs_root *root)
3531{
3532 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003533}
3534
Yan Zhengc146afa2008-11-12 14:34:12 -05003535int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3536{
3537 u64 root_objectid = 0;
3538 struct btrfs_root *gang[8];
3539 int i;
3540 int ret;
3541
3542 while (1) {
3543 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3544 (void **)gang, root_objectid,
3545 ARRAY_SIZE(gang));
3546 if (!ret)
3547 break;
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003548
3549 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Yan Zhengc146afa2008-11-12 14:34:12 -05003550 for (i = 0; i < ret; i++) {
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003551 int err;
3552
Yan Zhengc146afa2008-11-12 14:34:12 -05003553 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003554 err = btrfs_orphan_cleanup(gang[i]);
3555 if (err)
3556 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003557 }
3558 root_objectid++;
3559 }
3560 return 0;
3561}
3562
3563int btrfs_commit_super(struct btrfs_root *root)
3564{
3565 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003566
3567 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00003568 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003569 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003570 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003571
3572 /* wait until ongoing cleanup work done */
3573 down_write(&root->fs_info->cleanup_work_sem);
3574 up_write(&root->fs_info->cleanup_work_sem);
3575
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003576 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003577 if (IS_ERR(trans))
3578 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003579 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003580}
3581
Chris Masone20d96d2007-03-22 12:13:20 -04003582int close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003583{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003584 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003585 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003586
Chris Masonfacda1e2007-06-08 18:11:48 -04003587 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003588 smp_mb();
3589
Stefan Behrens803b2f52013-08-15 17:11:21 +02003590 /* wait for the uuid_scan task to finish */
3591 down(&fs_info->uuid_tree_rescan_sem);
3592 /* avoid complains from lockdep et al., set sem back to initial state */
3593 up(&fs_info->uuid_tree_rescan_sem);
3594
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003595 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003596 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003597
Stefan Behrens8dabb742012-11-06 13:15:27 +01003598 btrfs_dev_replace_suspend_for_unmount(fs_info);
3599
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003600 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003601
3602 /* wait for any defraggers to finish */
3603 wait_event(fs_info->transaction_wait,
3604 (atomic_read(&fs_info->defrag_running) == 0));
3605
3606 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003607 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003608
Yan Zhengc146afa2008-11-12 14:34:12 -05003609 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
liuboacce9522011-01-06 19:30:25 +08003610 ret = btrfs_commit_super(root);
3611 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003612 btrfs_err(root->fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003613 }
3614
Miao Xie87533c42013-01-29 10:14:48 +00003615 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003616 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003617
Al Viroe3029d92011-11-17 00:56:18 -05003618 kthread_stop(fs_info->transaction_kthread);
3619 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003620
Yan Zhengf25784b2009-07-28 08:41:57 -04003621 fs_info->closing = 2;
3622 smp_mb();
3623
Arne Jansenbcef60f2011-09-13 15:23:30 +02003624 btrfs_free_qgroup_config(root->fs_info);
3625
Miao Xie963d6782013-01-29 10:10:51 +00003626 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003627 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003628 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003629 }
Yanbcc63ab2008-07-30 16:29:20 -04003630
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003631 btrfs_sysfs_remove_one(fs_info);
3632
Chris Mason0f7d52f2007-04-09 10:42:37 -04003633 del_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003634
Liu Bo1a4319c2014-01-13 19:53:53 +08003635 btrfs_put_block_group_cache(fs_info);
3636
Josef Bacik2b1360d2013-10-07 15:14:44 -04003637 btrfs_free_block_groups(fs_info);
3638
Josef Bacik96192492013-10-16 13:53:28 -04003639 btrfs_stop_all_workers(fs_info);
3640
Josef Bacik13e6c372013-05-30 16:55:44 -04003641 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003642
Josef Bacik13e6c372013-05-30 16:55:44 -04003643 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003644
Stefan Behrens21adbd52011-11-09 13:44:05 +01003645#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3646 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3647 btrfsic_unmount(root, fs_info->fs_devices);
3648#endif
3649
Chris Masondfe25022008-05-13 13:46:40 -04003650 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003651 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003652
Miao Xiee2d84522013-01-29 10:09:20 +00003653 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003654 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003655 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003656 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003657 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003658
David Woodhouse53b381b2013-01-29 18:40:14 -05003659 btrfs_free_stripe_hash_table(fs_info);
3660
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003661 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3662 root->orphan_block_rsv = NULL;
3663
Chris Masoneb60cea2007-02-02 09:18:22 -05003664 return 0;
3665}
3666
Chris Masonb9fab912012-05-06 07:23:47 -04003667int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3668 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003669{
Chris Mason1259ab72008-05-12 13:39:03 -04003670 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003671 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003672
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003673 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003674 if (!ret)
3675 return ret;
3676
3677 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003678 parent_transid, atomic);
3679 if (ret == -EAGAIN)
3680 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003681 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003682}
Chris Mason6702ed42007-08-07 16:15:09 -04003683
Chris Mason5f39d392007-10-15 16:14:19 -04003684int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3685{
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003686 return set_extent_buffer_uptodate(buf);
Chris Mason5f39d392007-10-15 16:14:19 -04003687}
3688
3689void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3690{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003691 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003692 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003693 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003694
Josef Bacik06ea65a2013-09-19 16:07:01 -04003695#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3696 /*
3697 * This is a fast path so only do this check if we have sanity tests
3698 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3699 * outside of the sanity tests.
3700 */
3701 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3702 return;
3703#endif
3704 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003705 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003706 if (transid != root->fs_info->generation)
3707 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003708 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003709 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003710 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003711 if (!was_dirty)
3712 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3713 buf->len,
3714 root->fs_info->dirty_metadata_batch);
Chris Masoneb60cea2007-02-02 09:18:22 -05003715}
3716
Liu Bob53d3f52012-11-14 14:34:34 +00003717static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3718 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003719{
Chris Mason188de642008-05-09 11:52:25 -04003720 /*
3721 * looks as though older kernels can get into trouble with
3722 * this code, they end up stuck in balance_dirty_pages forever
3723 */
Miao Xiee2d84522013-01-29 10:09:20 +00003724 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04003725
Jens Axboe6933c022009-03-17 09:36:37 +01003726 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003727 return;
3728
Liu Bob53d3f52012-11-14 14:34:34 +00003729 if (flush_delayed)
3730 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003731
Miao Xiee2d84522013-01-29 10:09:20 +00003732 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3733 BTRFS_DIRTY_METADATA_THRESH);
3734 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08003735 balance_dirty_pages_ratelimited(
3736 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08003737 }
3738 return;
3739}
3740
Liu Bob53d3f52012-11-14 14:34:34 +00003741void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08003742{
Liu Bob53d3f52012-11-14 14:34:34 +00003743 __btrfs_btree_balance_dirty(root, 1);
3744}
Miao Xie16cdcec2011-04-22 18:12:22 +08003745
Liu Bob53d3f52012-11-14 14:34:34 +00003746void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3747{
3748 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04003749}
Chris Mason6b800532007-10-15 16:17:34 -04003750
Chris Masonca7a79a2008-05-12 12:59:19 -04003751int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04003752{
Chris Mason727011e2010-08-06 13:21:20 -04003753 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003754 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04003755}
Chris Mason0da54682007-11-07 21:08:01 -05003756
David Sterbafcd1f0652012-03-06 00:06:18 +01003757static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08003758 int read_only)
3759{
David Sterba1104a882013-03-06 15:57:46 +01003760 /*
3761 * Placeholder for checks
3762 */
David Sterbafcd1f0652012-03-06 00:06:18 +01003763 return 0;
liuboacce9522011-01-06 19:30:25 +08003764}
3765
Eric Sandeen48a3b632013-04-25 20:41:01 +00003766static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003767{
liuboacce9522011-01-06 19:30:25 +08003768 mutex_lock(&root->fs_info->cleaner_mutex);
3769 btrfs_run_delayed_iputs(root);
3770 mutex_unlock(&root->fs_info->cleaner_mutex);
3771
3772 down_write(&root->fs_info->cleanup_work_sem);
3773 up_write(&root->fs_info->cleanup_work_sem);
3774
3775 /* cleanup FS via transaction */
3776 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08003777}
3778
Josef Bacik569e0f32013-02-13 11:09:14 -05003779static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3780 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003781{
3782 struct btrfs_inode *btrfs_inode;
3783 struct list_head splice;
3784
3785 INIT_LIST_HEAD(&splice);
3786
3787 mutex_lock(&root->fs_info->ordered_operations_mutex);
Miao Xie199c2a92013-05-15 07:48:23 +00003788 spin_lock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003789
Josef Bacik569e0f32013-02-13 11:09:14 -05003790 list_splice_init(&t->ordered_operations, &splice);
liuboacce9522011-01-06 19:30:25 +08003791 while (!list_empty(&splice)) {
3792 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3793 ordered_operations);
3794
3795 list_del_init(&btrfs_inode->ordered_operations);
Miao Xie199c2a92013-05-15 07:48:23 +00003796 spin_unlock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003797
3798 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00003799
Miao Xie199c2a92013-05-15 07:48:23 +00003800 spin_lock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003801 }
3802
Miao Xie199c2a92013-05-15 07:48:23 +00003803 spin_unlock(&root->fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003804 mutex_unlock(&root->fs_info->ordered_operations_mutex);
liuboacce9522011-01-06 19:30:25 +08003805}
3806
Jeff Mahoney143bede2012-03-01 14:56:26 +01003807static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003808{
liuboacce9522011-01-06 19:30:25 +08003809 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08003810
Miao Xie199c2a92013-05-15 07:48:23 +00003811 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05003812 /*
3813 * This will just short circuit the ordered completion stuff which will
3814 * make sure the ordered extent gets properly cleaned up.
3815 */
Miao Xie199c2a92013-05-15 07:48:23 +00003816 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05003817 root_extent_list)
3818 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00003819 spin_unlock(&root->ordered_extent_lock);
3820}
3821
3822static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3823{
3824 struct btrfs_root *root;
3825 struct list_head splice;
3826
3827 INIT_LIST_HEAD(&splice);
3828
3829 spin_lock(&fs_info->ordered_root_lock);
3830 list_splice_init(&fs_info->ordered_roots, &splice);
3831 while (!list_empty(&splice)) {
3832 root = list_first_entry(&splice, struct btrfs_root,
3833 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04003834 list_move_tail(&root->ordered_root,
3835 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00003836
Liu Bo2a85d9c2014-02-10 17:07:16 +08003837 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003838 btrfs_destroy_ordered_extents(root);
3839
Liu Bo2a85d9c2014-02-10 17:07:16 +08003840 cond_resched();
3841 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00003842 }
3843 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08003844}
3845
Stefan Behrens35a36212013-08-14 18:12:25 +02003846static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3847 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003848{
3849 struct rb_node *node;
3850 struct btrfs_delayed_ref_root *delayed_refs;
3851 struct btrfs_delayed_ref_node *ref;
3852 int ret = 0;
3853
3854 delayed_refs = &trans->delayed_refs;
3855
3856 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003857 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04003858 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05003859 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08003860 return ret;
3861 }
3862
Josef Bacikd7df2c72014-01-23 09:21:38 -05003863 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3864 struct btrfs_delayed_ref_head *head;
Josef Bacike78417d2013-06-03 16:42:36 -04003865 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08003866
Josef Bacikd7df2c72014-01-23 09:21:38 -05003867 head = rb_entry(node, struct btrfs_delayed_ref_head,
3868 href_node);
3869 if (!mutex_trylock(&head->mutex)) {
3870 atomic_inc(&head->node.refs);
3871 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08003872
Josef Bacikd7df2c72014-01-23 09:21:38 -05003873 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04003874 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003875 btrfs_put_delayed_ref(&head->node);
3876 spin_lock(&delayed_refs->lock);
3877 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04003878 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05003879 spin_lock(&head->lock);
3880 while ((node = rb_first(&head->ref_root)) != NULL) {
3881 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3882 rb_node);
3883 ref->in_tree = 0;
3884 rb_erase(&ref->rb_node, &head->ref_root);
3885 atomic_dec(&delayed_refs->num_entries);
3886 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05003887 }
3888 if (head->must_insert_reserved)
3889 pin_bytes = true;
3890 btrfs_free_delayed_extent_op(head->extent_op);
3891 delayed_refs->num_heads--;
3892 if (head->processing == 0)
3893 delayed_refs->num_heads_ready--;
3894 atomic_dec(&delayed_refs->num_entries);
3895 head->node.in_tree = 0;
3896 rb_erase(&head->href_node, &delayed_refs->href_root);
3897 spin_unlock(&head->lock);
3898 spin_unlock(&delayed_refs->lock);
3899 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08003900
Josef Bacikd7df2c72014-01-23 09:21:38 -05003901 if (pin_bytes)
3902 btrfs_pin_extent(root, head->node.bytenr,
3903 head->node.num_bytes, 1);
3904 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08003905 cond_resched();
3906 spin_lock(&delayed_refs->lock);
3907 }
3908
3909 spin_unlock(&delayed_refs->lock);
3910
3911 return ret;
3912}
3913
Jeff Mahoney143bede2012-03-01 14:56:26 +01003914static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08003915{
3916 struct btrfs_inode *btrfs_inode;
3917 struct list_head splice;
3918
3919 INIT_LIST_HEAD(&splice);
3920
Miao Xieeb73c1b2013-05-15 07:48:22 +00003921 spin_lock(&root->delalloc_lock);
3922 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08003923
3924 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00003925 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3926 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08003927
3928 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00003929 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3930 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00003931 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003932
3933 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00003934
Miao Xieeb73c1b2013-05-15 07:48:22 +00003935 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08003936 }
3937
Miao Xieeb73c1b2013-05-15 07:48:22 +00003938 spin_unlock(&root->delalloc_lock);
3939}
3940
3941static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3942{
3943 struct btrfs_root *root;
3944 struct list_head splice;
3945
3946 INIT_LIST_HEAD(&splice);
3947
3948 spin_lock(&fs_info->delalloc_root_lock);
3949 list_splice_init(&fs_info->delalloc_roots, &splice);
3950 while (!list_empty(&splice)) {
3951 root = list_first_entry(&splice, struct btrfs_root,
3952 delalloc_root);
3953 list_del_init(&root->delalloc_root);
3954 root = btrfs_grab_fs_root(root);
3955 BUG_ON(!root);
3956 spin_unlock(&fs_info->delalloc_root_lock);
3957
3958 btrfs_destroy_delalloc_inodes(root);
3959 btrfs_put_fs_root(root);
3960
3961 spin_lock(&fs_info->delalloc_root_lock);
3962 }
3963 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08003964}
3965
3966static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3967 struct extent_io_tree *dirty_pages,
3968 int mark)
3969{
3970 int ret;
liuboacce9522011-01-06 19:30:25 +08003971 struct extent_buffer *eb;
3972 u64 start = 0;
3973 u64 end;
liuboacce9522011-01-06 19:30:25 +08003974
3975 while (1) {
3976 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04003977 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08003978 if (ret)
3979 break;
3980
3981 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3982 while (start <= end) {
Josef Bacikfd8b2b62013-04-24 16:41:19 -04003983 eb = btrfs_find_tree_block(root, start,
3984 root->leafsize);
Josef Bacik69a85bd2013-05-08 13:30:11 -04003985 start += root->leafsize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04003986 if (!eb)
liuboacce9522011-01-06 19:30:25 +08003987 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04003988 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08003989
Josef Bacikfd8b2b62013-04-24 16:41:19 -04003990 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3991 &eb->bflags))
3992 clear_extent_buffer_dirty(eb);
3993 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08003994 }
3995 }
3996
3997 return ret;
3998}
3999
4000static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4001 struct extent_io_tree *pinned_extents)
4002{
4003 struct extent_io_tree *unpin;
4004 u64 start;
4005 u64 end;
4006 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004007 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004008
4009 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004010again:
liuboacce9522011-01-06 19:30:25 +08004011 while (1) {
4012 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004013 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004014 if (ret)
4015 break;
4016
4017 /* opt_discard */
Li Dongyang5378e602011-03-24 10:24:27 +00004018 if (btrfs_test_opt(root, DISCARD))
4019 ret = btrfs_error_discard_extent(root, start,
4020 end + 1 - start,
4021 NULL);
liuboacce9522011-01-06 19:30:25 +08004022
4023 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4024 btrfs_error_unpin_extent_range(root, start, end);
4025 cond_resched();
4026 }
4027
Liu Boed0eaa12012-06-14 02:23:21 -06004028 if (loop) {
4029 if (unpin == &root->fs_info->freed_extents[0])
4030 unpin = &root->fs_info->freed_extents[1];
4031 else
4032 unpin = &root->fs_info->freed_extents[0];
4033 loop = false;
4034 goto again;
4035 }
4036
liuboacce9522011-01-06 19:30:25 +08004037 return 0;
4038}
4039
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004040void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4041 struct btrfs_root *root)
4042{
Josef Bacik724e2312013-09-30 11:36:38 -04004043 btrfs_destroy_ordered_operations(cur_trans, root);
4044
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004045 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004046
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004047 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004048 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004049
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004050 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004051 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004052
Miao Xie67cde342012-06-14 02:23:22 -06004053 btrfs_destroy_delayed_inodes(root);
4054 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004055
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004056 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4057 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004058 btrfs_destroy_pinned_extent(root,
4059 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004060
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004061 cur_trans->state =TRANS_STATE_COMPLETED;
4062 wake_up(&cur_trans->commit_wait);
4063
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004064 /*
4065 memset(cur_trans, 0, sizeof(*cur_trans));
4066 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4067 */
4068}
4069
Eric Sandeen48a3b632013-04-25 20:41:01 +00004070static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004071{
4072 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004073
liuboacce9522011-01-06 19:30:25 +08004074 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4075
Josef Bacika4abeea2011-04-11 17:25:13 -04004076 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004077 while (!list_empty(&root->fs_info->trans_list)) {
4078 t = list_first_entry(&root->fs_info->trans_list,
4079 struct btrfs_transaction, list);
4080 if (t->state >= TRANS_STATE_COMMIT_START) {
4081 atomic_inc(&t->use_count);
4082 spin_unlock(&root->fs_info->trans_lock);
4083 btrfs_wait_for_commit(root, t->transid);
4084 btrfs_put_transaction(t);
4085 spin_lock(&root->fs_info->trans_lock);
4086 continue;
4087 }
4088 if (t == root->fs_info->running_transaction) {
4089 t->state = TRANS_STATE_COMMIT_DOING;
4090 spin_unlock(&root->fs_info->trans_lock);
4091 /*
4092 * We wait for 0 num_writers since we don't hold a trans
4093 * handle open currently for this transaction.
4094 */
4095 wait_event(t->writer_wait,
4096 atomic_read(&t->num_writers) == 0);
4097 } else {
4098 spin_unlock(&root->fs_info->trans_lock);
4099 }
4100 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004101
Josef Bacik724e2312013-09-30 11:36:38 -04004102 spin_lock(&root->fs_info->trans_lock);
4103 if (t == root->fs_info->running_transaction)
4104 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004105 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004106 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004107
Josef Bacik724e2312013-09-30 11:36:38 -04004108 btrfs_put_transaction(t);
4109 trace_btrfs_transaction_commit(root);
4110 spin_lock(&root->fs_info->trans_lock);
4111 }
4112 spin_unlock(&root->fs_info->trans_lock);
4113 btrfs_destroy_all_ordered_extents(root->fs_info);
4114 btrfs_destroy_delayed_inodes(root);
4115 btrfs_assert_delayed_root_empty(root);
4116 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4117 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004118 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004119
4120 return 0;
4121}
4122
Chris Masond1310b22008-01-24 16:13:08 -05004123static struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004124 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004125 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004126 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004127 /* note we're sharing with inode.c for the merge bio hook */
4128 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004129};