blob: 4545e2e2ad45a556113eecf30b1a7d49ee7ca260 [file] [log] [blame]
Chris Mason6cbd5572007-06-12 09:07:21 -04001/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
Chris Masone20d96d2007-03-22 12:13:20 -040019#include <linux/fs.h>
Chris Masond98237b2007-03-28 13:57:48 -040020#include <linux/blkdev.h>
Chris Mason87cbda52007-03-28 19:44:27 -040021#include <linux/scatterlist.h>
Chris Mason22b0ebd2007-03-30 08:47:31 -040022#include <linux/swap.h>
Chris Mason0f7d52f2007-04-09 10:42:37 -040023#include <linux/radix-tree.h>
Chris Mason35b7e472007-05-02 15:53:43 -040024#include <linux/writeback.h>
Chris Masond3977122009-01-05 21:25:51 -050025#include <linux/buffer_head.h>
Chris Masonce9adaa2008-04-09 16:28:12 -040026#include <linux/workqueue.h>
Chris Masona74a4b92008-06-25 16:01:31 -040027#include <linux/kthread.h>
Chris Mason4b4e25f2008-11-20 10:22:27 -050028#include <linux/freezer.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/slab.h>
Chris Mason784b4e22010-11-21 22:20:49 -050030#include <linux/migrate.h>
David Sterba7a36dde2011-05-06 15:33:15 +020031#include <linux/ratelimit.h>
Stefan Behrens6463fe52013-04-19 15:08:05 +000032#include <linux/uuid.h>
Stefan Behrens803b2f52013-08-15 17:11:21 +020033#include <linux/semaphore.h>
David Sterba7e75bf32011-03-18 22:56:43 +000034#include <asm/unaligned.h>
Chris Masoneb60cea2007-02-02 09:18:22 -050035#include "ctree.h"
36#include "disk-io.h"
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +000037#include "hash.h"
Chris Masone089f052007-03-16 16:20:31 -040038#include "transaction.h"
Chris Mason0f7d52f2007-04-09 10:42:37 -040039#include "btrfs_inode.h"
Chris Mason0b86a832008-03-24 15:01:56 -040040#include "volumes.h"
Chris Masondb945352007-10-15 16:15:53 -040041#include "print-tree.h"
Chris Mason925baed2008-06-25 16:01:30 -040042#include "locking.h"
Chris Masone02119d2008-09-05 16:13:11 -040043#include "tree-log.h"
Chris Masonfa9c0d792009-04-03 09:47:43 -040044#include "free-space-cache.h"
Omar Sandoval70f6d822015-09-29 20:50:38 -070045#include "free-space-tree.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"
Josef Bacikfcebe452014-05-13 17:30:47 -070052#include "qgroup.h"
Chris Masoneb60cea2007-02-02 09:18:22 -050053
Josef Bacikde0022b2012-09-25 14:25:58 -040054#ifdef CONFIG_X86
55#include <asm/cpufeature.h>
56#endif
57
Qu Wenruo319e4d02015-12-15 09:14:36 +080058#define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\
59 BTRFS_HEADER_FLAG_RELOC |\
60 BTRFS_SUPER_FLAG_ERROR |\
61 BTRFS_SUPER_FLAG_SEEDING |\
62 BTRFS_SUPER_FLAG_METADUMP)
63
David Sterbae8c9f182015-01-02 18:23:10 +010064static const struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040065static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040066static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010067static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080068 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010069static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080070static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
71 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010072static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080073static int btrfs_destroy_marked_extents(struct btrfs_root *root,
74 struct extent_io_tree *dirty_pages,
75 int mark);
76static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
77 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000078static int btrfs_cleanup_transaction(struct btrfs_root *root);
79static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040080
Chris Masond352ac62008-09-29 15:18:18 -040081/*
David Sterba97eb6b62014-07-30 00:55:42 +020082 * btrfs_end_io_wq structs are used to do processing in task context when an IO
83 * is complete. This is used during reads to verify checksums, and it is used
Chris Masond352ac62008-09-29 15:18:18 -040084 * by writes to insert metadata for new file extents after IO is complete.
85 */
David Sterba97eb6b62014-07-30 00:55:42 +020086struct btrfs_end_io_wq {
Chris Masonce9adaa2008-04-09 16:28:12 -040087 struct bio *bio;
88 bio_end_io_t *end_io;
89 void *private;
90 struct btrfs_fs_info *info;
91 int error;
David Sterbabfebd8b2014-07-30 00:25:45 +020092 enum btrfs_wq_endio_type metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040093 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040094 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040095};
Chris Mason0da54682007-11-07 21:08:01 -050096
David Sterba97eb6b62014-07-30 00:55:42 +020097static struct kmem_cache *btrfs_end_io_wq_cache;
98
99int __init btrfs_end_io_wq_init(void)
100{
101 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
102 sizeof(struct btrfs_end_io_wq),
103 0,
104 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
105 NULL);
106 if (!btrfs_end_io_wq_cache)
107 return -ENOMEM;
108 return 0;
109}
110
111void btrfs_end_io_wq_exit(void)
112{
113 if (btrfs_end_io_wq_cache)
114 kmem_cache_destroy(btrfs_end_io_wq_cache);
115}
116
Chris Masond352ac62008-09-29 15:18:18 -0400117/*
118 * async submit bios are used to offload expensive checksumming
119 * onto the worker threads. They checksum file and metadata bios
120 * just before they are sent down the IO stack.
121 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400122struct async_submit_bio {
123 struct inode *inode;
124 struct bio *bio;
125 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500126 extent_submit_bio_hook_t *submit_bio_start;
127 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400128 int rw;
129 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400130 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400131 /*
132 * bio_offset is optional, can be used if the pages in the bio
133 * can't tell us where in the file the bio should go
134 */
135 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400136 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100137 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400138};
139
Chris Mason85d4e462011-07-26 16:11:19 -0400140/*
141 * Lockdep class keys for extent_buffer->lock's in this root. For a given
142 * eb, the lockdep key is determined by the btrfs_root it belongs to and
143 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500144 *
Chris Mason85d4e462011-07-26 16:11:19 -0400145 * Different roots are used for different purposes and may nest inside each
146 * other and they require separate keysets. As lockdep keys should be
147 * static, assign keysets according to the purpose of the root as indicated
148 * by btrfs_root->objectid. This ensures that all special purpose roots
149 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500150 *
Chris Mason85d4e462011-07-26 16:11:19 -0400151 * Lock-nesting across peer nodes is always done with the immediate parent
152 * node locked thus preventing deadlock. As lockdep doesn't know this, use
153 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500154 *
Chris Mason85d4e462011-07-26 16:11:19 -0400155 * The key is set by the readpage_end_io_hook after the buffer has passed
156 * csum validation but before the pages are unlocked. It is also set by
157 * btrfs_init_new_buffer on freshly allocated blocks.
158 *
159 * We also add a check to make sure the highest level of the tree is the
160 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
161 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500162 */
163#ifdef CONFIG_DEBUG_LOCK_ALLOC
164# if BTRFS_MAX_LEVEL != 8
165# error
166# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400167
168static struct btrfs_lockdep_keyset {
169 u64 id; /* root objectid */
170 const char *name_stem; /* lock name stem */
171 char names[BTRFS_MAX_LEVEL + 1][20];
172 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
173} btrfs_lockdep_keysets[] = {
174 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
175 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
176 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
177 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
178 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
179 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000180 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400181 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
182 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
183 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200184 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
David Sterba6b20e0a2016-01-25 16:30:22 +0100185 { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" },
Chris Mason85d4e462011-07-26 16:11:19 -0400186 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500187};
Chris Mason85d4e462011-07-26 16:11:19 -0400188
189void __init btrfs_init_lockdep(void)
190{
191 int i, j;
192
193 /* initialize lockdep class names */
194 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
195 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
196
197 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
198 snprintf(ks->names[j], sizeof(ks->names[j]),
199 "btrfs-%s-%02d", ks->name_stem, j);
200 }
201}
202
203void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
204 int level)
205{
206 struct btrfs_lockdep_keyset *ks;
207
208 BUG_ON(level >= ARRAY_SIZE(ks->keys));
209
210 /* find the matching keyset, id 0 is the default entry */
211 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
212 if (ks->id == objectid)
213 break;
214
215 lockdep_set_class_and_name(&eb->lock,
216 &ks->keys[level], ks->names[level]);
217}
218
Chris Mason4008c042009-02-12 14:09:45 -0500219#endif
220
Chris Masond352ac62008-09-29 15:18:18 -0400221/*
222 * extents on the btree inode are pretty simple, there's one extent
223 * that covers the entire device
224 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500225static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200226 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500227 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400228{
229 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
230 struct extent_map *em;
231 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400232
Chris Mason890871b2009-09-02 16:24:52 -0400233 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500234 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400235 if (em) {
236 em->bdev =
237 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400238 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400239 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400240 }
Chris Mason890871b2009-09-02 16:24:52 -0400241 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400242
David Sterba172ddd62011-04-21 00:48:27 +0200243 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400244 if (!em) {
245 em = ERR_PTR(-ENOMEM);
246 goto out;
247 }
248 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400249 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400250 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400251 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400252 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500253
Chris Mason890871b2009-09-02 16:24:52 -0400254 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400255 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400256 if (ret == -EEXIST) {
257 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400258 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600259 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600260 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400261 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400262 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600263 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400264 }
Chris Mason890871b2009-09-02 16:24:52 -0400265 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400266
Chris Mason5f39d392007-10-15 16:14:19 -0400267out:
268 return em;
269}
270
Liu Bob0496682013-03-14 14:57:45 +0000271u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400272{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000273 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400274}
275
276void btrfs_csum_final(u32 crc, char *result)
277{
David Sterba7e75bf32011-03-18 22:56:43 +0000278 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400279}
280
Chris Masond352ac62008-09-29 15:18:18 -0400281/*
282 * compute the csum for a btree block, and either verify it or write it
283 * into the csum field of the block.
284 */
Daniel Dressler01d58472014-11-21 17:15:07 +0900285static int csum_tree_block(struct btrfs_fs_info *fs_info,
286 struct extent_buffer *buf,
Chris Mason19c00dd2007-10-15 16:19:22 -0400287 int verify)
288{
Daniel Dressler01d58472014-11-21 17:15:07 +0900289 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500290 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400291 unsigned long len;
292 unsigned long cur_len;
293 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400294 char *kaddr;
295 unsigned long map_start;
296 unsigned long map_len;
297 int err;
298 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500299 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400300
301 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500302 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400303 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400304 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500305 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400306 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400307 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000308 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400309 crc, cur_len);
310 len -= cur_len;
311 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400312 }
Josef Bacik607d4322008-12-02 07:17:45 -0500313 if (csum_size > sizeof(inline_result)) {
David Sterba31e818f2015-02-20 18:00:26 +0100314 result = kzalloc(csum_size, GFP_NOFS);
Josef Bacik607d4322008-12-02 07:17:45 -0500315 if (!result)
316 return 1;
317 } else {
318 result = (char *)&inline_result;
319 }
320
Chris Mason19c00dd2007-10-15 16:19:22 -0400321 btrfs_csum_final(crc, result);
322
323 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500324 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500325 u32 val;
326 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500327 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500328
Josef Bacik607d4322008-12-02 07:17:45 -0500329 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba94647322015-10-08 11:01:36 +0200330 btrfs_warn_rl(fs_info,
331 "%s checksum verify failed on %llu wanted %X found %X "
332 "level %d",
Daniel Dressler01d58472014-11-21 17:15:07 +0900333 fs_info->sb->s_id, buf->start,
Frank Holtonefe120a2013-12-20 11:37:06 -0500334 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500335 if (result != (char *)&inline_result)
336 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400337 return 1;
338 }
339 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500340 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400341 }
Josef Bacik607d4322008-12-02 07:17:45 -0500342 if (result != (char *)&inline_result)
343 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400344 return 0;
345}
346
Chris Masond352ac62008-09-29 15:18:18 -0400347/*
348 * we can't consider a given block up to date unless the transid of the
349 * block matches the transid in the parent node's pointer. This is how we
350 * detect blocks that either didn't get written at all or got written
351 * in the wrong place.
352 */
Chris Mason1259ab72008-05-12 13:39:03 -0400353static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400354 struct extent_buffer *eb, u64 parent_transid,
355 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400356{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000357 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400358 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200359 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400360
361 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
362 return 0;
363
Chris Masonb9fab912012-05-06 07:23:47 -0400364 if (atomic)
365 return -EAGAIN;
366
Josef Bacika26e8c92014-03-28 17:07:27 -0400367 if (need_lock) {
368 btrfs_tree_read_lock(eb);
369 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
370 }
371
Josef Bacik2ac55d42010-02-03 19:33:23 +0000372 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
David Sterbaff13db42015-12-03 14:30:40 +0100373 &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400374 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400375 btrfs_header_generation(eb) == parent_transid) {
376 ret = 0;
377 goto out;
378 }
David Sterba94647322015-10-08 11:01:36 +0200379 btrfs_err_rl(eb->fs_info,
380 "parent transid verify failed on %llu wanted %llu found %llu",
381 eb->start,
Wang Shilong29549ae2014-07-04 17:59:06 +0800382 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400383 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400384
385 /*
386 * Things reading via commit roots that don't have normal protection,
387 * like send, can have a really old block in cache that may point at a
388 * block that has been free'd and re-allocated. So don't clear uptodate
389 * if we find an eb that is under IO (dirty/writeback) because we could
390 * end up reading in the stale data and then writing it back out and
391 * making everybody very sad.
392 */
393 if (!extent_buffer_under_io(eb))
394 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400395out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000396 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
397 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700398 if (need_lock)
399 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400400 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400401}
402
Chris Masond352ac62008-09-29 15:18:18 -0400403/*
David Sterba1104a882013-03-06 15:57:46 +0100404 * Return 0 if the superblock checksum type matches the checksum value of that
405 * algorithm. Pass the raw disk superblock data.
406 */
407static int btrfs_check_super_csum(char *raw_disk_sb)
408{
409 struct btrfs_super_block *disk_sb =
410 (struct btrfs_super_block *)raw_disk_sb;
411 u16 csum_type = btrfs_super_csum_type(disk_sb);
412 int ret = 0;
413
414 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
415 u32 crc = ~(u32)0;
416 const int csum_size = sizeof(crc);
417 char result[csum_size];
418
419 /*
420 * The super_block structure does not span the whole
421 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
422 * is filled with zeros and is included in the checkum.
423 */
424 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
425 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
426 btrfs_csum_final(crc, result);
427
428 if (memcmp(raw_disk_sb, result, csum_size))
429 ret = 1;
430 }
431
432 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500433 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100434 csum_type);
435 ret = 1;
436 }
437
438 return ret;
439}
440
441/*
Chris Masond352ac62008-09-29 15:18:18 -0400442 * helper to read a given tree block, doing retries as required when
443 * the checksums don't match and we have alternate mirrors to try.
444 */
Chris Masonf1885912008-04-09 16:28:12 -0400445static int btree_read_extent_buffer_pages(struct btrfs_root *root,
446 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400447 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400448{
449 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400450 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400451 int ret;
452 int num_copies = 0;
453 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400454 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400455
Josef Bacika826d6d2011-03-16 13:42:43 -0400456 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400457 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
458 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200459 ret = read_extent_buffer_pages(io_tree, eb, start,
460 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400461 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600462 if (!ret) {
463 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400464 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600465 break;
466 else
467 ret = -EIO;
468 }
Chris Masond3977122009-01-05 21:25:51 -0500469
Josef Bacika826d6d2011-03-16 13:42:43 -0400470 /*
471 * This buffer's crc is fine, but its contents are corrupted, so
472 * there is no reason to read the other copies, they won't be
473 * any less wrong.
474 */
475 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400476 break;
477
Stefan Behrens5d964052012-11-05 14:59:07 +0100478 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400479 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400480 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400481 break;
Chris Mason42352982008-04-28 16:40:52 -0400482
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400483 if (!failed_mirror) {
484 failed = 1;
485 failed_mirror = eb->read_mirror;
486 }
487
Chris Masonf1885912008-04-09 16:28:12 -0400488 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400489 if (mirror_num == failed_mirror)
490 mirror_num++;
491
Chris Mason42352982008-04-28 16:40:52 -0400492 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400493 break;
Chris Masonf1885912008-04-09 16:28:12 -0400494 }
Josef Bacikea466792012-03-26 21:57:36 -0400495
Stefan Behrensc0901582012-07-10 07:30:17 -0600496 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400497 repair_eb_io_failure(root, eb, failed_mirror);
498
499 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400500}
Chris Mason19c00dd2007-10-15 16:19:22 -0400501
Chris Masond352ac62008-09-29 15:18:18 -0400502/*
Chris Masond3977122009-01-05 21:25:51 -0500503 * checksum a dirty tree block before IO. This has extra checks to make sure
504 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400505 */
Chris Masond3977122009-01-05 21:25:51 -0500506
Daniel Dressler01d58472014-11-21 17:15:07 +0900507static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400508{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000509 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400510 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400511 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400512
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500513 eb = (struct extent_buffer *)page->private;
514 if (page != eb->pages[0])
515 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400516 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530517 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500518 return 0;
Daniel Dressler01d58472014-11-21 17:15:07 +0900519 csum_tree_block(fs_info, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400520 return 0;
521}
522
Daniel Dressler01d58472014-11-21 17:15:07 +0900523static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
Yan Zheng2b820322008-11-17 21:11:30 -0500524 struct extent_buffer *eb)
525{
Daniel Dressler01d58472014-11-21 17:15:07 +0900526 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
Yan Zheng2b820322008-11-17 21:11:30 -0500527 u8 fsid[BTRFS_UUID_SIZE];
528 int ret = 1;
529
Ross Kirk0a4e5582013-09-24 10:12:38 +0100530 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500531 while (fs_devices) {
532 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
533 ret = 0;
534 break;
535 }
536 fs_devices = fs_devices->seed;
537 }
538 return ret;
539}
540
Josef Bacika826d6d2011-03-16 13:42:43 -0400541#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500542 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
543 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200544 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400545
546static noinline int check_leaf(struct btrfs_root *root,
547 struct extent_buffer *leaf)
548{
549 struct btrfs_key key;
550 struct btrfs_key leaf_key;
551 u32 nritems = btrfs_header_nritems(leaf);
552 int slot;
553
554 if (nritems == 0)
555 return 0;
556
557 /* Check the 0 item */
558 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
559 BTRFS_LEAF_DATA_SIZE(root)) {
560 CORRUPT("invalid item offset size pair", leaf, root, 0);
561 return -EIO;
562 }
563
564 /*
565 * Check to make sure each items keys are in the correct order and their
566 * offsets make sense. We only have to loop through nritems-1 because
567 * we check the current slot against the next slot, which verifies the
568 * next slot's offset+size makes sense and that the current's slot
569 * offset is correct.
570 */
571 for (slot = 0; slot < nritems - 1; slot++) {
572 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
573 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
574
575 /* Make sure the keys are in the right order */
576 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
577 CORRUPT("bad key order", leaf, root, slot);
578 return -EIO;
579 }
580
581 /*
582 * Make sure the offset and ends are right, remember that the
583 * item data starts at the end of the leaf and grows towards the
584 * front.
585 */
586 if (btrfs_item_offset_nr(leaf, slot) !=
587 btrfs_item_end_nr(leaf, slot + 1)) {
588 CORRUPT("slot offset bad", leaf, root, slot);
589 return -EIO;
590 }
591
592 /*
593 * Check to make sure that we don't point outside of the leaf,
594 * just incase all the items are consistent to eachother, but
595 * all point outside of the leaf.
596 */
597 if (btrfs_item_end_nr(leaf, slot) >
598 BTRFS_LEAF_DATA_SIZE(root)) {
599 CORRUPT("slot end outside of leaf", leaf, root, slot);
600 return -EIO;
601 }
602 }
603
604 return 0;
605}
606
Miao Xiefacc8a222013-07-25 19:22:34 +0800607static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
608 u64 phy_offset, struct page *page,
609 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400610{
Chris Masonce9adaa2008-04-09 16:28:12 -0400611 u64 found_start;
612 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400613 struct extent_buffer *eb;
614 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400615 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400616 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400617
Chris Masonce9adaa2008-04-09 16:28:12 -0400618 if (!page->private)
619 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500620
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500621 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500622
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400623 /* the pending IO might have been the only thing that kept this buffer
624 * in memory. Make sure we have a ref for all this other checks
625 */
626 extent_buffer_get(eb);
627
628 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400629 if (!reads_done)
630 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400631
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400632 eb->read_mirror = mirror;
Filipe Manana656f30d2014-09-26 12:25:56 +0100633 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
Josef Bacikea466792012-03-26 21:57:36 -0400634 ret = -EIO;
635 goto err;
636 }
637
Chris Masonce9adaa2008-04-09 16:28:12 -0400638 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400639 if (found_start != eb->start) {
David Sterba94647322015-10-08 11:01:36 +0200640 btrfs_err_rl(eb->fs_info, "bad tree block start %llu %llu",
641 found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400642 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400643 goto err;
644 }
Daniel Dressler01d58472014-11-21 17:15:07 +0900645 if (check_tree_block_fsid(root->fs_info, eb)) {
David Sterba94647322015-10-08 11:01:36 +0200646 btrfs_err_rl(eb->fs_info, "bad fsid on block %llu",
647 eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400648 ret = -EIO;
649 goto err;
650 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400651 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400652 if (found_level >= BTRFS_MAX_LEVEL) {
David Sterba68b663d2014-12-19 18:38:37 +0100653 btrfs_err(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400654 (int)btrfs_header_level(eb));
655 ret = -EIO;
656 goto err;
657 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400658
Chris Mason85d4e462011-07-26 16:11:19 -0400659 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
660 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500661
Daniel Dressler01d58472014-11-21 17:15:07 +0900662 ret = csum_tree_block(root->fs_info, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400663 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400664 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400665 goto err;
666 }
667
668 /*
669 * If this is a leaf block and it is corrupt, set the corrupt bit so
670 * that we don't try and read the other copies of this block, just
671 * return -EIO.
672 */
673 if (found_level == 0 && check_leaf(root, eb)) {
674 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
675 ret = -EIO;
676 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400677
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400678 if (!ret)
679 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400680err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400681 if (reads_done &&
682 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200683 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200684
David Woodhouse53b381b2013-01-29 18:40:14 -0500685 if (ret) {
686 /*
687 * our io error hook is going to dec the io pages
688 * again, we have to make sure it has something
689 * to decrement
690 */
691 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400692 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500693 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400694 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400695out:
Chris Masonf1885912008-04-09 16:28:12 -0400696 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400697}
698
Josef Bacikea466792012-03-26 21:57:36 -0400699static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200700{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200701 struct extent_buffer *eb;
702 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
703
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500704 eb = (struct extent_buffer *)page->private;
Filipe Manana656f30d2014-09-26 12:25:56 +0100705 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400706 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500707 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400708 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200709 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200710 return -EIO; /* we fixed nothing */
711}
712
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200713static void end_workqueue_bio(struct bio *bio)
Chris Masonce9adaa2008-04-09 16:28:12 -0400714{
David Sterba97eb6b62014-07-30 00:55:42 +0200715 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400716 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800717 struct btrfs_workqueue *wq;
718 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400719
Chris Masonce9adaa2008-04-09 16:28:12 -0400720 fs_info = end_io_wq->info;
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200721 end_io_wq->error = bio->bi_error;
Chris Masond20f7042008-12-08 16:58:54 -0500722
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200723 if (bio->bi_rw & REQ_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800724 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
725 wq = fs_info->endio_meta_write_workers;
726 func = btrfs_endio_meta_write_helper;
727 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
728 wq = fs_info->endio_freespace_worker;
729 func = btrfs_freespace_write_helper;
730 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
731 wq = fs_info->endio_raid56_workers;
732 func = btrfs_endio_raid56_helper;
733 } else {
734 wq = fs_info->endio_write_workers;
735 func = btrfs_endio_write_helper;
736 }
Chris Masond20f7042008-12-08 16:58:54 -0500737 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800738 if (unlikely(end_io_wq->metadata ==
739 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
740 wq = fs_info->endio_repair_workers;
741 func = btrfs_endio_repair_helper;
742 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800743 wq = fs_info->endio_raid56_workers;
744 func = btrfs_endio_raid56_helper;
745 } else if (end_io_wq->metadata) {
746 wq = fs_info->endio_meta_workers;
747 func = btrfs_endio_meta_helper;
748 } else {
749 wq = fs_info->endio_workers;
750 func = btrfs_endio_helper;
751 }
Chris Masond20f7042008-12-08 16:58:54 -0500752 }
Liu Bo9e0af232014-08-15 23:36:53 +0800753
754 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
755 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400756}
757
Chris Mason22c59942008-04-09 16:28:12 -0400758int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200759 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400760{
David Sterba97eb6b62014-07-30 00:55:42 +0200761 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800762
David Sterba97eb6b62014-07-30 00:55:42 +0200763 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400764 if (!end_io_wq)
765 return -ENOMEM;
766
767 end_io_wq->private = bio->bi_private;
768 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400769 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400770 end_io_wq->error = 0;
771 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400772 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400773
774 bio->bi_private = end_io_wq;
775 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400776 return 0;
777}
778
Chris Masonb64a2852008-08-20 13:39:41 -0400779unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400780{
781 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800782 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400783 info->fs_devices->open_devices);
784 return 256 * limit;
785}
786
Chris Mason4a69a412008-11-06 22:03:00 -0500787static void run_one_async_start(struct btrfs_work *work)
788{
Chris Mason4a69a412008-11-06 22:03:00 -0500789 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100790 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500791
792 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100793 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
794 async->mirror_num, async->bio_flags,
795 async->bio_offset);
796 if (ret)
797 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500798}
799
800static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400801{
802 struct btrfs_fs_info *fs_info;
803 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400804 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400805
806 async = container_of(work, struct async_submit_bio, work);
807 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400808
Chris Masonb64a2852008-08-20 13:39:41 -0400809 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400810 limit = limit * 2 / 3;
811
David Sterbaee863952015-02-16 19:41:40 +0100812 /*
813 * atomic_dec_return implies a barrier for waitqueue_active
814 */
Josef Bacik66657b32012-08-01 15:36:24 -0400815 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400816 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400817 wake_up(&fs_info->async_submit_wait);
818
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100819 /* If an error occured we just want to clean up the bio and move on */
820 if (async->error) {
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200821 async->bio->bi_error = async->error;
822 bio_endio(async->bio);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100823 return;
824 }
825
Chris Mason4a69a412008-11-06 22:03:00 -0500826 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400827 async->mirror_num, async->bio_flags,
828 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500829}
830
831static void run_one_async_free(struct btrfs_work *work)
832{
833 struct async_submit_bio *async;
834
835 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400836 kfree(async);
837}
838
Chris Mason44b8bd72008-04-16 11:14:51 -0400839int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
840 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400841 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400842 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500843 extent_submit_bio_hook_t *submit_bio_start,
844 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400845{
846 struct async_submit_bio *async;
847
848 async = kmalloc(sizeof(*async), GFP_NOFS);
849 if (!async)
850 return -ENOMEM;
851
852 async->inode = inode;
853 async->rw = rw;
854 async->bio = bio;
855 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500856 async->submit_bio_start = submit_bio_start;
857 async->submit_bio_done = submit_bio_done;
858
Liu Bo9e0af232014-08-15 23:36:53 +0800859 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800860 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500861
Chris Masonc8b97812008-10-29 14:49:59 -0400862 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400863 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400864
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100865 async->error = 0;
866
Chris Masoncb03c742008-05-15 16:15:45 -0400867 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400868
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200869 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800870 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400871
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800872 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400873
Chris Masond3977122009-01-05 21:25:51 -0500874 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500875 atomic_read(&fs_info->nr_async_submits)) {
876 wait_event(fs_info->async_submit_wait,
877 (atomic_read(&fs_info->nr_async_submits) == 0));
878 }
879
Chris Mason44b8bd72008-04-16 11:14:51 -0400880 return 0;
881}
882
Chris Masonce3ed712008-09-23 13:14:12 -0400883static int btree_csum_one_bio(struct bio *bio)
884{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800885 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400886 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800887 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400888
Kent Overstreet2c30c712013-11-07 12:20:26 -0800889 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400890 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Daniel Dressler01d58472014-11-21 17:15:07 +0900891 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100892 if (ret)
893 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400894 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800895
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100896 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400897}
898
Chris Mason4a69a412008-11-06 22:03:00 -0500899static int __btree_submit_bio_start(struct inode *inode, int rw,
900 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400901 unsigned long bio_flags,
902 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400903{
Chris Mason8b712842008-06-11 16:50:36 -0400904 /*
905 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400906 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400907 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100908 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500909}
Chris Mason22c59942008-04-09 16:28:12 -0400910
Chris Mason4a69a412008-11-06 22:03:00 -0500911static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400912 int mirror_num, unsigned long bio_flags,
913 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500914{
Stefan Behrens61891922012-11-05 18:51:52 +0100915 int ret;
916
Chris Mason8b712842008-06-11 16:50:36 -0400917 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500918 * when we're called for a write, we're already in the async
919 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400920 */
Stefan Behrens61891922012-11-05 18:51:52 +0100921 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200922 if (ret) {
923 bio->bi_error = ret;
924 bio_endio(bio);
925 }
Stefan Behrens61891922012-11-05 18:51:52 +0100926 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400927}
928
Josef Bacikde0022b2012-09-25 14:25:58 -0400929static int check_async_write(struct inode *inode, unsigned long bio_flags)
930{
931 if (bio_flags & EXTENT_BIO_TREE_LOG)
932 return 0;
933#ifdef CONFIG_X86
Borislav Petkov362f9242015-12-07 10:39:41 +0100934 if (static_cpu_has_safe(X86_FEATURE_XMM4_2))
Josef Bacikde0022b2012-09-25 14:25:58 -0400935 return 0;
936#endif
937 return 1;
938}
939
Chris Mason44b8bd72008-04-16 11:14:51 -0400940static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400941 int mirror_num, unsigned long bio_flags,
942 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400943{
Josef Bacikde0022b2012-09-25 14:25:58 -0400944 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500945 int ret;
946
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200947 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500948 /*
949 * called for a read, do the setup so that checksum validation
950 * can happen in the async kernel threads
951 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400952 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +0200953 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -0400954 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100955 goto out_w_error;
956 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
957 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400958 } else if (!async) {
959 ret = btree_csum_one_bio(bio);
960 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100961 goto out_w_error;
962 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
963 mirror_num, 0);
964 } else {
965 /*
966 * kthread helpers are used to submit writes so that
967 * checksumming can happen in parallel across all CPUs
968 */
969 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
970 inode, rw, bio, mirror_num, 0,
971 bio_offset,
972 __btree_submit_bio_start,
973 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400974 }
Chris Masond313d7a2009-04-20 15:50:09 -0400975
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200976 if (ret)
977 goto out_w_error;
978 return 0;
979
Stefan Behrens61891922012-11-05 18:51:52 +0100980out_w_error:
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200981 bio->bi_error = ret;
982 bio_endio(bio);
Stefan Behrens61891922012-11-05 18:51:52 +0100983 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400984}
985
Jan Beulich3dd14622010-12-07 14:54:09 +0000986#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500987static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800988 struct page *newpage, struct page *page,
989 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500990{
991 /*
992 * we can't safely write a btree page from here,
993 * we haven't done the locking hook
994 */
995 if (PageDirty(page))
996 return -EAGAIN;
997 /*
998 * Buffers may be managed in a filesystem specific way.
999 * We must have no buffers or drop them.
1000 */
1001 if (page_has_private(page) &&
1002 !try_to_release_page(page, GFP_KERNEL))
1003 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -08001004 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -05001005}
Jan Beulich3dd14622010-12-07 14:54:09 +00001006#endif
Chris Mason784b4e22010-11-21 22:20:49 -05001007
Chris Mason0da54682007-11-07 21:08:01 -05001008
1009static int btree_writepages(struct address_space *mapping,
1010 struct writeback_control *wbc)
1011{
Miao Xiee2d84522013-01-29 10:09:20 +00001012 struct btrfs_fs_info *fs_info;
1013 int ret;
1014
Chris Masond8d5f3e2007-12-11 12:42:00 -05001015 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001016
1017 if (wbc->for_kupdate)
1018 return 0;
1019
Miao Xiee2d84522013-01-29 10:09:20 +00001020 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001021 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001022 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1023 BTRFS_DIRTY_METADATA_THRESH);
1024 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001025 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001026 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001027 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001028}
1029
Christoph Hellwigb2950862008-12-02 09:54:17 -05001030static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001031{
Chris Masond1310b22008-01-24 16:13:08 -05001032 struct extent_io_tree *tree;
1033 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001034 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001035}
1036
Chris Mason70dec802008-01-29 09:59:12 -05001037static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001038{
Chris Mason98509cf2008-09-11 15:51:43 -04001039 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001040 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001041
David Sterbaf7a52a42013-04-26 14:56:29 +00001042 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001043}
Chris Mason123abc82007-03-14 14:14:43 -04001044
Lukas Czernerd47992f2013-05-21 23:17:23 -04001045static void btree_invalidatepage(struct page *page, unsigned int offset,
1046 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001047{
Chris Masond1310b22008-01-24 16:13:08 -05001048 struct extent_io_tree *tree;
1049 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001050 extent_invalidatepage(tree, page, offset);
1051 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001052 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001053 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1054 "page private not zero on page %llu",
1055 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001056 ClearPagePrivate(page);
1057 set_page_private(page, 0);
1058 page_cache_release(page);
1059 }
Chris Masond98237b2007-03-28 13:57:48 -04001060}
1061
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001062static int btree_set_page_dirty(struct page *page)
1063{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001064#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001065 struct extent_buffer *eb;
1066
1067 BUG_ON(!PagePrivate(page));
1068 eb = (struct extent_buffer *)page->private;
1069 BUG_ON(!eb);
1070 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1071 BUG_ON(!atomic_read(&eb->refs));
1072 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001073#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001074 return __set_page_dirty_nobuffers(page);
1075}
1076
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001077static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001078 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001079 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001080 .releasepage = btree_releasepage,
1081 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001082#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001083 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001084#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001085 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001086};
1087
David Sterbad3e46fe2014-06-15 02:04:19 +02001088void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
Chris Mason090d1872007-05-01 08:53:32 -04001089{
Chris Mason5f39d392007-10-15 16:14:19 -04001090 struct extent_buffer *buf = NULL;
1091 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason090d1872007-05-01 08:53:32 -04001092
David Sterbaa83fffb2014-06-15 02:39:54 +02001093 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001094 if (!buf)
David Sterba6197d862014-06-15 00:49:36 +02001095 return;
Chris Masond1310b22008-01-24 16:13:08 -05001096 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001097 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001098 free_extent_buffer(buf);
Chris Mason090d1872007-05-01 08:53:32 -04001099}
1100
David Sterbac0dcaa42014-06-15 02:24:21 +02001101int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
Arne Jansenab0fff02011-05-23 14:25:41 +02001102 int mirror_num, struct extent_buffer **eb)
1103{
1104 struct extent_buffer *buf = NULL;
1105 struct inode *btree_inode = root->fs_info->btree_inode;
1106 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1107 int ret;
1108
David Sterbaa83fffb2014-06-15 02:39:54 +02001109 buf = btrfs_find_create_tree_block(root, bytenr);
Arne Jansenab0fff02011-05-23 14:25:41 +02001110 if (!buf)
1111 return 0;
1112
1113 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1114
1115 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1116 btree_get_extent, mirror_num);
1117 if (ret) {
1118 free_extent_buffer(buf);
1119 return ret;
1120 }
1121
1122 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1123 free_extent_buffer(buf);
1124 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001125 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001126 *eb = buf;
1127 } else {
1128 free_extent_buffer(buf);
1129 }
1130 return 0;
1131}
1132
Daniel Dressler01d58472014-11-21 17:15:07 +09001133struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
David Sterba0308af42014-06-15 01:43:40 +02001134 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001135{
Daniel Dressler01d58472014-11-21 17:15:07 +09001136 return find_extent_buffer(fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001137}
1138
1139struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
David Sterbaa83fffb2014-06-15 02:39:54 +02001140 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001141{
David Sterbafccb84c2014-09-29 23:53:21 +02001142 if (btrfs_test_is_dummy_root(root))
David Sterbace3e6982014-06-15 03:00:04 +02001143 return alloc_test_extent_buffer(root->fs_info, bytenr);
1144 return alloc_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001145}
1146
1147
Chris Masone02119d2008-09-05 16:13:11 -04001148int btrfs_write_tree_block(struct extent_buffer *buf)
1149{
Chris Mason727011e2010-08-06 13:21:20 -04001150 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001151 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001152}
1153
1154int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1155{
Chris Mason727011e2010-08-06 13:21:20 -04001156 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001157 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001158}
1159
Chris Masondb945352007-10-15 16:15:53 -04001160struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
David Sterbace86cd52014-06-15 01:07:32 +02001161 u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001162{
Chris Mason5f39d392007-10-15 16:14:19 -04001163 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001164 int ret;
1165
David Sterbaa83fffb2014-06-15 02:39:54 +02001166 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001167 if (!buf)
Liu Bo64c043d2015-05-25 17:30:15 +08001168 return ERR_PTR(-ENOMEM);
Chris Masone4204de2008-01-08 15:46:27 -05001169
Chris Masonca7a79a2008-05-12 12:59:19 -04001170 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001171 if (ret) {
1172 free_extent_buffer(buf);
Liu Bo64c043d2015-05-25 17:30:15 +08001173 return ERR_PTR(ret);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001174 }
Chris Mason5f39d392007-10-15 16:14:19 -04001175 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001176
Chris Masoneb60cea2007-02-02 09:18:22 -05001177}
1178
Daniel Dressler01d58472014-11-21 17:15:07 +09001179void clean_tree_block(struct btrfs_trans_handle *trans,
1180 struct btrfs_fs_info *fs_info,
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001181 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001182{
Chris Mason55c69072008-01-09 15:55:33 -05001183 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001184 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001185 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001186
Chris Masonb9473432009-03-13 11:00:37 -04001187 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001188 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1189 -buf->len,
1190 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001191 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1192 btrfs_set_lock_blocking(buf);
1193 clear_extent_buffer_dirty(buf);
1194 }
Chris Mason925baed2008-06-25 16:01:30 -04001195 }
Chris Mason5f39d392007-10-15 16:14:19 -04001196}
1197
Miao Xie8257b2d2014-03-06 13:38:19 +08001198static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1199{
1200 struct btrfs_subvolume_writers *writers;
1201 int ret;
1202
1203 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1204 if (!writers)
1205 return ERR_PTR(-ENOMEM);
1206
Tejun Heo908c7f12014-09-08 09:51:29 +09001207 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
Miao Xie8257b2d2014-03-06 13:38:19 +08001208 if (ret < 0) {
1209 kfree(writers);
1210 return ERR_PTR(ret);
1211 }
1212
1213 init_waitqueue_head(&writers->wait);
1214 return writers;
1215}
1216
1217static void
1218btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1219{
1220 percpu_counter_destroy(&writers->counter);
1221 kfree(writers);
1222}
1223
David Sterba707e8a02014-06-04 19:22:26 +02001224static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1225 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
Jeff Mahoney143bede2012-03-01 14:56:26 +01001226 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001227{
Chris Masoncfaa7292007-02-21 17:04:57 -05001228 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001229 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001230 root->sectorsize = sectorsize;
1231 root->nodesize = nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001232 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001233 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001234 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001235
Chris Mason0f7d52f2007-04-09 10:42:37 -04001236 root->objectid = objectid;
1237 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001238 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001239 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001240 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001241 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001242 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001243 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001244 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001245 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001246
1247 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001248 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001249 INIT_LIST_HEAD(&root->delalloc_inodes);
1250 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001251 INIT_LIST_HEAD(&root->ordered_extents);
1252 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001253 INIT_LIST_HEAD(&root->logged_list[0]);
1254 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001255 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001256 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001257 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001258 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001259 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001260 spin_lock_init(&root->log_extents_lock[0]);
1261 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001262 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001263 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001264 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001265 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f182009-01-21 12:54:03 -05001266 init_waitqueue_head(&root->log_writer_wait);
1267 init_waitqueue_head(&root->log_commit_wait[0]);
1268 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001269 INIT_LIST_HEAD(&root->log_ctxs[0]);
1270 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f182009-01-21 12:54:03 -05001271 atomic_set(&root->log_commit[0], 0);
1272 atomic_set(&root->log_commit[1], 0);
1273 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001274 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001275 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001276 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001277 atomic_set(&root->will_be_snapshoted, 0);
Qu Wenruo55eeaf02015-09-08 17:08:38 +08001278 atomic_set(&root->qgroup_meta_rsv, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001279 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001280 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001281 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001282 if (fs_info)
1283 extent_io_tree_init(&root->dirty_log_pages,
1284 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001285
Chris Mason3768f362007-03-13 16:47:54 -04001286 memset(&root->root_key, 0, sizeof(root->root_key));
1287 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001288 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001289 if (fs_info)
1290 root->defrag_trans_start = fs_info->generation;
1291 else
1292 root->defrag_trans_start = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001293 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001294 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001295
Anand Jain5f3ab902012-12-07 09:28:54 +00001296 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001297}
1298
Al Virof84a8bd2011-11-17 15:57:57 -05001299static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001300{
1301 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1302 if (root)
1303 root->fs_info = fs_info;
1304 return root;
1305}
1306
Josef Bacik06ea65a2013-09-19 16:07:01 -04001307#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1308/* Should only be used by the testing infrastructure */
1309struct btrfs_root *btrfs_alloc_dummy_root(void)
1310{
1311 struct btrfs_root *root;
1312
1313 root = btrfs_alloc_root(NULL);
1314 if (!root)
1315 return ERR_PTR(-ENOMEM);
David Sterba707e8a02014-06-04 19:22:26 +02001316 __setup_root(4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001317 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001318 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001319
1320 return root;
1321}
1322#endif
1323
Arne Jansen20897f52011-09-13 12:44:20 +02001324struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1325 struct btrfs_fs_info *fs_info,
1326 u64 objectid)
1327{
1328 struct extent_buffer *leaf;
1329 struct btrfs_root *tree_root = fs_info->tree_root;
1330 struct btrfs_root *root;
1331 struct btrfs_key key;
1332 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001333 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001334
1335 root = btrfs_alloc_root(fs_info);
1336 if (!root)
1337 return ERR_PTR(-ENOMEM);
1338
David Sterba707e8a02014-06-04 19:22:26 +02001339 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1340 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001341 root->root_key.objectid = objectid;
1342 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1343 root->root_key.offset = 0;
1344
David Sterba4d75f8a2014-06-15 01:54:12 +02001345 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Arne Jansen20897f52011-09-13 12:44:20 +02001346 if (IS_ERR(leaf)) {
1347 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001348 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001349 goto fail;
1350 }
1351
Arne Jansen20897f52011-09-13 12:44:20 +02001352 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1353 btrfs_set_header_bytenr(leaf, leaf->start);
1354 btrfs_set_header_generation(leaf, trans->transid);
1355 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1356 btrfs_set_header_owner(leaf, objectid);
1357 root->node = leaf;
1358
Ross Kirk0a4e5582013-09-24 10:12:38 +01001359 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001360 BTRFS_FSID_SIZE);
1361 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001362 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001363 BTRFS_UUID_SIZE);
1364 btrfs_mark_buffer_dirty(leaf);
1365
1366 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001367 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001368
1369 root->root_item.flags = 0;
1370 root->root_item.byte_limit = 0;
1371 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1372 btrfs_set_root_generation(&root->root_item, trans->transid);
1373 btrfs_set_root_level(&root->root_item, 0);
1374 btrfs_set_root_refs(&root->root_item, 1);
1375 btrfs_set_root_used(&root->root_item, leaf->len);
1376 btrfs_set_root_last_snapshot(&root->root_item, 0);
1377 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001378 uuid_le_gen(&uuid);
1379 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001380 root->root_item.drop_level = 0;
1381
1382 key.objectid = objectid;
1383 key.type = BTRFS_ROOT_ITEM_KEY;
1384 key.offset = 0;
1385 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1386 if (ret)
1387 goto fail;
1388
1389 btrfs_tree_unlock(leaf);
1390
Arne Jansen20897f52011-09-13 12:44:20 +02001391 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001392
1393fail:
1394 if (leaf) {
1395 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001396 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001397 free_extent_buffer(leaf);
1398 }
1399 kfree(root);
1400
1401 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001402}
1403
Yan Zheng7237f182009-01-21 12:54:03 -05001404static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1405 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001406{
1407 struct btrfs_root *root;
1408 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f182009-01-21 12:54:03 -05001409 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001410
Al Viro6f07e422011-11-17 00:46:16 -05001411 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001412 if (!root)
Yan Zheng7237f182009-01-21 12:54:03 -05001413 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001414
David Sterba707e8a02014-06-04 19:22:26 +02001415 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1416 tree_root->stripesize, root, fs_info,
1417 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001418
1419 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1420 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1421 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001422
Yan Zheng7237f182009-01-21 12:54:03 -05001423 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001424 * DON'T set REF_COWS for log trees
1425 *
Yan Zheng7237f182009-01-21 12:54:03 -05001426 * log trees do not get reference counted because they go away
1427 * before a real commit is actually done. They do store pointers
1428 * to file data extents, and those reference counts still get
1429 * updated (along with back refs to the log tree).
1430 */
Chris Masone02119d2008-09-05 16:13:11 -04001431
David Sterba4d75f8a2014-06-15 01:54:12 +02001432 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1433 NULL, 0, 0, 0);
Yan Zheng7237f182009-01-21 12:54:03 -05001434 if (IS_ERR(leaf)) {
1435 kfree(root);
1436 return ERR_CAST(leaf);
1437 }
Chris Masone02119d2008-09-05 16:13:11 -04001438
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001439 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1440 btrfs_set_header_bytenr(leaf, leaf->start);
1441 btrfs_set_header_generation(leaf, trans->transid);
1442 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1443 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f182009-01-21 12:54:03 -05001444 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001445
1446 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001447 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001448 btrfs_mark_buffer_dirty(root->node);
1449 btrfs_tree_unlock(root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001450 return root;
1451}
1452
1453int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1454 struct btrfs_fs_info *fs_info)
1455{
1456 struct btrfs_root *log_root;
1457
1458 log_root = alloc_log_tree(trans, fs_info);
1459 if (IS_ERR(log_root))
1460 return PTR_ERR(log_root);
1461 WARN_ON(fs_info->log_root_tree);
1462 fs_info->log_root_tree = log_root;
1463 return 0;
1464}
1465
1466int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1467 struct btrfs_root *root)
1468{
1469 struct btrfs_root *log_root;
1470 struct btrfs_inode_item *inode_item;
1471
1472 log_root = alloc_log_tree(trans, root->fs_info);
1473 if (IS_ERR(log_root))
1474 return PTR_ERR(log_root);
1475
1476 log_root->last_trans = trans->transid;
1477 log_root->root_key.offset = root->root_key.objectid;
1478
1479 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001480 btrfs_set_stack_inode_generation(inode_item, 1);
1481 btrfs_set_stack_inode_size(inode_item, 3);
1482 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001483 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001484 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f182009-01-21 12:54:03 -05001485
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001486 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f182009-01-21 12:54:03 -05001487
1488 WARN_ON(root->log_root);
1489 root->log_root = log_root;
1490 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001491 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001492 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001493 return 0;
1494}
1495
Stefan Behrens35a36212013-08-14 18:12:25 +02001496static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1497 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001498{
1499 struct btrfs_root *root;
1500 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001501 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001502 u64 generation;
Miao Xiecb517ea2013-05-15 07:48:19 +00001503 int ret;
1504
1505 path = btrfs_alloc_path();
1506 if (!path)
1507 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001508
Al Viro6f07e422011-11-17 00:46:16 -05001509 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001510 if (!root) {
1511 ret = -ENOMEM;
1512 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001513 }
1514
David Sterba707e8a02014-06-04 19:22:26 +02001515 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1516 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001517
Miao Xiecb517ea2013-05-15 07:48:19 +00001518 ret = btrfs_find_root(tree_root, key, path,
1519 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001520 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001521 if (ret > 0)
1522 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001523 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001524 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001525
Yan Zheng84234f32008-10-29 14:49:05 -04001526 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001527 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
David Sterbace86cd52014-06-15 01:07:32 +02001528 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08001529 if (IS_ERR(root->node)) {
1530 ret = PTR_ERR(root->node);
Miao Xiecb517ea2013-05-15 07:48:19 +00001531 goto find_fail;
1532 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1533 ret = -EIO;
Liu Bo64c043d2015-05-25 17:30:15 +08001534 free_extent_buffer(root->node);
1535 goto find_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001536 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001537 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001538out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001539 btrfs_free_path(path);
1540 return root;
1541
Miao Xiecb517ea2013-05-15 07:48:19 +00001542find_fail:
1543 kfree(root);
1544alloc_fail:
1545 root = ERR_PTR(ret);
1546 goto out;
1547}
1548
1549struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1550 struct btrfs_key *location)
1551{
1552 struct btrfs_root *root;
1553
1554 root = btrfs_read_tree_root(tree_root, location);
1555 if (IS_ERR(root))
1556 return root;
1557
1558 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001559 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001560 btrfs_check_and_init_root_item(&root->root_item);
1561 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001562
Chris Mason5eda7b52007-06-22 14:16:25 -04001563 return root;
1564}
1565
Miao Xiecb517ea2013-05-15 07:48:19 +00001566int btrfs_init_fs_root(struct btrfs_root *root)
1567{
1568 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001569 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001570
1571 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1572 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1573 GFP_NOFS);
1574 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1575 ret = -ENOMEM;
1576 goto fail;
1577 }
1578
Miao Xie8257b2d2014-03-06 13:38:19 +08001579 writers = btrfs_alloc_subvolume_writers();
1580 if (IS_ERR(writers)) {
1581 ret = PTR_ERR(writers);
1582 goto fail;
1583 }
1584 root->subv_writers = writers;
1585
Miao Xiecb517ea2013-05-15 07:48:19 +00001586 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001587 spin_lock_init(&root->ino_cache_lock);
1588 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001589
1590 ret = get_anon_bdev(&root->anon_dev);
1591 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001592 goto free_writers;
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301593
1594 mutex_lock(&root->objectid_mutex);
1595 ret = btrfs_find_highest_objectid(root,
1596 &root->highest_objectid);
1597 if (ret) {
1598 mutex_unlock(&root->objectid_mutex);
1599 goto free_root_dev;
1600 }
1601
1602 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1603
1604 mutex_unlock(&root->objectid_mutex);
1605
Miao Xiecb517ea2013-05-15 07:48:19 +00001606 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001607
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301608free_root_dev:
1609 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08001610free_writers:
1611 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001612fail:
1613 kfree(root->free_ino_ctl);
1614 kfree(root->free_ino_pinned);
1615 return ret;
1616}
1617
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001618static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1619 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001620{
1621 struct btrfs_root *root;
1622
1623 spin_lock(&fs_info->fs_roots_radix_lock);
1624 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1625 (unsigned long)root_id);
1626 spin_unlock(&fs_info->fs_roots_radix_lock);
1627 return root;
1628}
1629
1630int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1631 struct btrfs_root *root)
1632{
1633 int ret;
1634
1635 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1636 if (ret)
1637 return ret;
1638
1639 spin_lock(&fs_info->fs_roots_radix_lock);
1640 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1641 (unsigned long)root->root_key.objectid,
1642 root);
1643 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001644 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001645 spin_unlock(&fs_info->fs_roots_radix_lock);
1646 radix_tree_preload_end();
1647
1648 return ret;
1649}
1650
Miao Xiec00869f2013-09-25 21:47:44 +08001651struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1652 struct btrfs_key *location,
1653 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001654{
1655 struct btrfs_root *root;
David Sterba381cf652015-01-02 18:45:16 +01001656 struct btrfs_path *path;
David Sterba1d4c08e2015-01-02 19:36:14 +01001657 struct btrfs_key key;
Chris Mason5eda7b52007-06-22 14:16:25 -04001658 int ret;
1659
Chris Masonedbd8d42007-12-21 16:27:24 -05001660 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1661 return fs_info->tree_root;
1662 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1663 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001664 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1665 return fs_info->chunk_root;
1666 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1667 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001668 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1669 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001670 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1671 return fs_info->quota_root ? fs_info->quota_root :
1672 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001673 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1674 return fs_info->uuid_root ? fs_info->uuid_root :
1675 ERR_PTR(-ENOENT);
Omar Sandoval70f6d822015-09-29 20:50:38 -07001676 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1677 return fs_info->free_space_root ? fs_info->free_space_root :
1678 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001679again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001680 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001681 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001682 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001683 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001684 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001685 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001686
Miao Xiecb517ea2013-05-15 07:48:19 +00001687 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001688 if (IS_ERR(root))
1689 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001690
Miao Xiec00869f2013-09-25 21:47:44 +08001691 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001692 ret = -ENOENT;
1693 goto fail;
1694 }
1695
Miao Xiecb517ea2013-05-15 07:48:19 +00001696 ret = btrfs_init_fs_root(root);
1697 if (ret)
1698 goto fail;
1699
David Sterba381cf652015-01-02 18:45:16 +01001700 path = btrfs_alloc_path();
1701 if (!path) {
1702 ret = -ENOMEM;
1703 goto fail;
1704 }
David Sterba1d4c08e2015-01-02 19:36:14 +01001705 key.objectid = BTRFS_ORPHAN_OBJECTID;
1706 key.type = BTRFS_ORPHAN_ITEM_KEY;
1707 key.offset = location->objectid;
1708
1709 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
David Sterba381cf652015-01-02 18:45:16 +01001710 btrfs_free_path(path);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001711 if (ret < 0)
1712 goto fail;
1713 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001714 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001715
Miao Xiecb517ea2013-05-15 07:48:19 +00001716 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001717 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001718 if (ret == -EEXIST) {
1719 free_fs_root(root);
1720 goto again;
1721 }
1722 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001723 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001724 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001725fail:
1726 free_fs_root(root);
1727 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001728}
1729
Chris Mason04160082008-03-26 10:28:07 -04001730static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1731{
1732 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1733 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001734 struct btrfs_device *device;
1735 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001736
Xiao Guangrong1f781602011-04-20 10:09:16 +00001737 rcu_read_lock();
1738 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001739 if (!device->bdev)
1740 continue;
Chris Mason04160082008-03-26 10:28:07 -04001741 bdi = blk_get_backing_dev_info(device->bdev);
Tejun Heoff9ea322014-09-08 08:03:56 +09001742 if (bdi_congested(bdi, bdi_bits)) {
Chris Mason04160082008-03-26 10:28:07 -04001743 ret = 1;
1744 break;
1745 }
1746 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001747 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001748 return ret;
1749}
1750
Chris Mason04160082008-03-26 10:28:07 -04001751static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1752{
Jens Axboead081f12009-06-12 14:43:40 +02001753 int err;
1754
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001755 err = bdi_setup_and_register(bdi, "btrfs");
Jens Axboead081f12009-06-12 14:43:40 +02001756 if (err)
1757 return err;
1758
Christoph Hellwigdf0ce262015-01-14 10:42:41 +01001759 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE;
Chris Mason04160082008-03-26 10:28:07 -04001760 bdi->congested_fn = btrfs_congested_fn;
1761 bdi->congested_data = info;
Chris Masonda2f0f72015-07-02 13:57:22 -07001762 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
Chris Mason04160082008-03-26 10:28:07 -04001763 return 0;
1764}
1765
Chris Mason8b712842008-06-11 16:50:36 -04001766/*
1767 * called by the kthread helper functions to finally call the bio end_io
1768 * functions. This is where read checksum verification actually happens
1769 */
1770static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001771{
Chris Masonce9adaa2008-04-09 16:28:12 -04001772 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001773 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001774
David Sterba97eb6b62014-07-30 00:55:42 +02001775 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001776 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001777
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001778 bio->bi_error = end_io_wq->error;
Chris Mason8b712842008-06-11 16:50:36 -04001779 bio->bi_private = end_io_wq->private;
1780 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001781 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001782 bio_endio(bio);
Chris Mason44b8bd72008-04-16 11:14:51 -04001783}
1784
Chris Masona74a4b92008-06-25 16:01:31 -04001785static int cleaner_kthread(void *arg)
1786{
1787 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001788 int again;
Filipe Mananada288d22015-06-13 06:55:31 +01001789 struct btrfs_trans_handle *trans;
Chris Masona74a4b92008-06-25 16:01:31 -04001790
1791 do {
Miao Xied0278242013-05-14 10:20:40 +00001792 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001793
Miao Xied0278242013-05-14 10:20:40 +00001794 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001795 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001796 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001797
Miao Xied0278242013-05-14 10:20:40 +00001798 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1799 goto sleep;
1800
Miao Xiedc7f3702013-05-14 10:20:42 +00001801 /*
1802 * Avoid the problem that we change the status of the fs
1803 * during the above check and trylock.
1804 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001805 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001806 mutex_unlock(&root->fs_info->cleaner_mutex);
1807 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001808 }
1809
Filipe Mananac2d6cb12016-01-15 11:05:12 +00001810 mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex);
Miao Xied0278242013-05-14 10:20:40 +00001811 btrfs_run_delayed_iputs(root);
Filipe Mananac2d6cb12016-01-15 11:05:12 +00001812 mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex);
1813
Miao Xied0278242013-05-14 10:20:40 +00001814 again = btrfs_clean_one_deleted_snapshot(root);
1815 mutex_unlock(&root->fs_info->cleaner_mutex);
1816
1817 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001818 * The defragger has dealt with the R/O remount and umount,
1819 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001820 */
1821 btrfs_run_defrag_inodes(root->fs_info);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01001822
1823 /*
1824 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1825 * with relocation (btrfs_relocate_chunk) and relocation
1826 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1827 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1828 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1829 * unused block groups.
1830 */
1831 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001832sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001833 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001834 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001835 if (!kthread_should_stop())
1836 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001837 __set_current_state(TASK_RUNNING);
1838 }
1839 } while (!kthread_should_stop());
Filipe Mananada288d22015-06-13 06:55:31 +01001840
1841 /*
1842 * Transaction kthread is stopped before us and wakes us up.
1843 * However we might have started a new transaction and COWed some
1844 * tree blocks when deleting unused block groups for example. So
1845 * make sure we commit the transaction we started to have a clean
1846 * shutdown when evicting the btree inode - if it has dirty pages
1847 * when we do the final iput() on it, eviction will trigger a
1848 * writeback for it which will fail with null pointer dereferences
1849 * since work queues and other resources were already released and
1850 * destroyed by the time the iput/eviction/writeback is made.
1851 */
1852 trans = btrfs_attach_transaction(root);
1853 if (IS_ERR(trans)) {
1854 if (PTR_ERR(trans) != -ENOENT)
1855 btrfs_err(root->fs_info,
1856 "cleaner transaction attach returned %ld",
1857 PTR_ERR(trans));
1858 } else {
1859 int ret;
1860
1861 ret = btrfs_commit_transaction(trans, root);
1862 if (ret)
1863 btrfs_err(root->fs_info,
1864 "cleaner open transaction commit returned %d",
1865 ret);
1866 }
1867
Chris Masona74a4b92008-06-25 16:01:31 -04001868 return 0;
1869}
1870
1871static int transaction_kthread(void *arg)
1872{
1873 struct btrfs_root *root = arg;
1874 struct btrfs_trans_handle *trans;
1875 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001876 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001877 unsigned long now;
1878 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001879 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001880
1881 do {
Jan Kara914b2002012-03-12 16:05:50 +01001882 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001883 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001884 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1885
Josef Bacika4abeea2011-04-11 17:25:13 -04001886 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001887 cur = root->fs_info->running_transaction;
1888 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001889 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001890 goto sleep;
1891 }
Yan Zheng31153d82008-07-28 15:32:19 -04001892
Chris Masona74a4b92008-06-25 16:01:31 -04001893 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001894 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001895 (now < cur->start_time ||
1896 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001897 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001898 delay = HZ * 5;
1899 goto sleep;
1900 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001901 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001902 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001903
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001904 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001905 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001906 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001907 if (PTR_ERR(trans) != -ENOENT)
1908 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001909 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001910 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001911 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001912 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001913 } else {
1914 btrfs_end_transaction(trans, root);
1915 }
Chris Masona74a4b92008-06-25 16:01:31 -04001916sleep:
1917 wake_up_process(root->fs_info->cleaner_kthread);
1918 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1919
Josef Bacik4e121c02013-09-27 16:32:39 -04001920 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1921 &root->fs_info->fs_state)))
1922 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001923 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001924 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001925 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001926 (!btrfs_transaction_blocked(root->fs_info) ||
1927 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001928 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001929 __set_current_state(TASK_RUNNING);
1930 }
1931 } while (!kthread_should_stop());
1932 return 0;
1933}
1934
Chris Masonaf31f5e2011-11-03 15:17:42 -04001935/*
1936 * this will find the highest generation in the array of
1937 * root backups. The index of the highest array is returned,
1938 * or -1 if we can't find anything.
1939 *
1940 * We check to make sure the array is valid by comparing the
1941 * generation of the latest root in the array with the generation
1942 * in the super block. If they don't match we pitch it.
1943 */
1944static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1945{
1946 u64 cur;
1947 int newest_index = -1;
1948 struct btrfs_root_backup *root_backup;
1949 int i;
1950
1951 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1952 root_backup = info->super_copy->super_roots + i;
1953 cur = btrfs_backup_tree_root_gen(root_backup);
1954 if (cur == newest_gen)
1955 newest_index = i;
1956 }
1957
1958 /* check to see if we actually wrapped around */
1959 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1960 root_backup = info->super_copy->super_roots;
1961 cur = btrfs_backup_tree_root_gen(root_backup);
1962 if (cur == newest_gen)
1963 newest_index = 0;
1964 }
1965 return newest_index;
1966}
1967
1968
1969/*
1970 * find the oldest backup so we know where to store new entries
1971 * in the backup array. This will set the backup_root_index
1972 * field in the fs_info struct
1973 */
1974static void find_oldest_super_backup(struct btrfs_fs_info *info,
1975 u64 newest_gen)
1976{
1977 int newest_index = -1;
1978
1979 newest_index = find_newest_super_backup(info, newest_gen);
1980 /* if there was garbage in there, just move along */
1981 if (newest_index == -1) {
1982 info->backup_root_index = 0;
1983 } else {
1984 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1985 }
1986}
1987
1988/*
1989 * copy all the root pointers into the super backup array.
1990 * this will bump the backup pointer by one when it is
1991 * done
1992 */
1993static void backup_super_roots(struct btrfs_fs_info *info)
1994{
1995 int next_backup;
1996 struct btrfs_root_backup *root_backup;
1997 int last_backup;
1998
1999 next_backup = info->backup_root_index;
2000 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
2001 BTRFS_NUM_BACKUP_ROOTS;
2002
2003 /*
2004 * just overwrite the last backup if we're at the same generation
2005 * this happens only at umount
2006 */
2007 root_backup = info->super_for_commit->super_roots + last_backup;
2008 if (btrfs_backup_tree_root_gen(root_backup) ==
2009 btrfs_header_generation(info->tree_root->node))
2010 next_backup = last_backup;
2011
2012 root_backup = info->super_for_commit->super_roots + next_backup;
2013
2014 /*
2015 * make sure all of our padding and empty slots get zero filled
2016 * regardless of which ones we use today
2017 */
2018 memset(root_backup, 0, sizeof(*root_backup));
2019
2020 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
2021
2022 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
2023 btrfs_set_backup_tree_root_gen(root_backup,
2024 btrfs_header_generation(info->tree_root->node));
2025
2026 btrfs_set_backup_tree_root_level(root_backup,
2027 btrfs_header_level(info->tree_root->node));
2028
2029 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
2030 btrfs_set_backup_chunk_root_gen(root_backup,
2031 btrfs_header_generation(info->chunk_root->node));
2032 btrfs_set_backup_chunk_root_level(root_backup,
2033 btrfs_header_level(info->chunk_root->node));
2034
2035 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2036 btrfs_set_backup_extent_root_gen(root_backup,
2037 btrfs_header_generation(info->extent_root->node));
2038 btrfs_set_backup_extent_root_level(root_backup,
2039 btrfs_header_level(info->extent_root->node));
2040
Chris Mason7c7e82a2011-11-06 18:50:56 -05002041 /*
2042 * we might commit during log recovery, which happens before we set
2043 * the fs_root. Make sure it is valid before we fill it in.
2044 */
2045 if (info->fs_root && info->fs_root->node) {
2046 btrfs_set_backup_fs_root(root_backup,
2047 info->fs_root->node->start);
2048 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002049 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002050 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002051 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002052 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002053
2054 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2055 btrfs_set_backup_dev_root_gen(root_backup,
2056 btrfs_header_generation(info->dev_root->node));
2057 btrfs_set_backup_dev_root_level(root_backup,
2058 btrfs_header_level(info->dev_root->node));
2059
2060 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2061 btrfs_set_backup_csum_root_gen(root_backup,
2062 btrfs_header_generation(info->csum_root->node));
2063 btrfs_set_backup_csum_root_level(root_backup,
2064 btrfs_header_level(info->csum_root->node));
2065
2066 btrfs_set_backup_total_bytes(root_backup,
2067 btrfs_super_total_bytes(info->super_copy));
2068 btrfs_set_backup_bytes_used(root_backup,
2069 btrfs_super_bytes_used(info->super_copy));
2070 btrfs_set_backup_num_devices(root_backup,
2071 btrfs_super_num_devices(info->super_copy));
2072
2073 /*
2074 * if we don't copy this out to the super_copy, it won't get remembered
2075 * for the next commit
2076 */
2077 memcpy(&info->super_copy->super_roots,
2078 &info->super_for_commit->super_roots,
2079 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2080}
2081
2082/*
2083 * this copies info out of the root backup array and back into
2084 * the in-memory super block. It is meant to help iterate through
2085 * the array, so you send it the number of backups you've already
2086 * tried and the last backup index you used.
2087 *
2088 * this returns -1 when it has tried all the backups
2089 */
2090static noinline int next_root_backup(struct btrfs_fs_info *info,
2091 struct btrfs_super_block *super,
2092 int *num_backups_tried, int *backup_index)
2093{
2094 struct btrfs_root_backup *root_backup;
2095 int newest = *backup_index;
2096
2097 if (*num_backups_tried == 0) {
2098 u64 gen = btrfs_super_generation(super);
2099
2100 newest = find_newest_super_backup(info, gen);
2101 if (newest == -1)
2102 return -1;
2103
2104 *backup_index = newest;
2105 *num_backups_tried = 1;
2106 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2107 /* we've tried all the backups, all done */
2108 return -1;
2109 } else {
2110 /* jump to the next oldest backup */
2111 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2112 BTRFS_NUM_BACKUP_ROOTS;
2113 *backup_index = newest;
2114 *num_backups_tried += 1;
2115 }
2116 root_backup = super->super_roots + newest;
2117
2118 btrfs_set_super_generation(super,
2119 btrfs_backup_tree_root_gen(root_backup));
2120 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2121 btrfs_set_super_root_level(super,
2122 btrfs_backup_tree_root_level(root_backup));
2123 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2124
2125 /*
2126 * fixme: the total bytes and num_devices need to match or we should
2127 * need a fsck
2128 */
2129 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2130 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2131 return 0;
2132}
2133
Liu Bo7abadb62013-03-17 02:10:31 +00002134/* helper to cleanup workers */
2135static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2136{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002137 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002138 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002139 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002140 btrfs_destroy_workqueue(fs_info->endio_workers);
2141 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2142 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002143 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002144 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002145 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2146 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2147 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002148 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002149 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002150 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002151 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002152 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002153 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002154 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002155}
2156
Rashika2e9f5952013-10-31 02:45:20 +05302157static void free_root_extent_buffers(struct btrfs_root *root)
2158{
2159 if (root) {
2160 free_extent_buffer(root->node);
2161 free_extent_buffer(root->commit_root);
2162 root->node = NULL;
2163 root->commit_root = NULL;
2164 }
2165}
2166
Chris Masonaf31f5e2011-11-03 15:17:42 -04002167/* helper to cleanup tree roots */
2168static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2169{
Rashika2e9f5952013-10-31 02:45:20 +05302170 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002171
Rashika2e9f5952013-10-31 02:45:20 +05302172 free_root_extent_buffers(info->dev_root);
2173 free_root_extent_buffers(info->extent_root);
2174 free_root_extent_buffers(info->csum_root);
2175 free_root_extent_buffers(info->quota_root);
2176 free_root_extent_buffers(info->uuid_root);
2177 if (chunk_root)
2178 free_root_extent_buffers(info->chunk_root);
Omar Sandoval70f6d822015-09-29 20:50:38 -07002179 free_root_extent_buffers(info->free_space_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002180}
2181
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002182void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002183{
2184 int ret;
2185 struct btrfs_root *gang[8];
2186 int i;
2187
2188 while (!list_empty(&fs_info->dead_roots)) {
2189 gang[0] = list_entry(fs_info->dead_roots.next,
2190 struct btrfs_root, root_list);
2191 list_del(&gang[0]->root_list);
2192
Miao Xie27cdeb72014-04-02 19:51:05 +08002193 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002194 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002195 } else {
2196 free_extent_buffer(gang[0]->node);
2197 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002198 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002199 }
2200 }
2201
2202 while (1) {
2203 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2204 (void **)gang, 0,
2205 ARRAY_SIZE(gang));
2206 if (!ret)
2207 break;
2208 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002209 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002210 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002211
2212 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2213 btrfs_free_log_root_tree(NULL, fs_info);
2214 btrfs_destroy_pinned_extent(fs_info->tree_root,
2215 fs_info->pinned_extents);
2216 }
Josef Bacik171f6532013-04-24 16:35:41 -04002217}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002218
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002219static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2220{
2221 mutex_init(&fs_info->scrub_lock);
2222 atomic_set(&fs_info->scrubs_running, 0);
2223 atomic_set(&fs_info->scrub_pause_req, 0);
2224 atomic_set(&fs_info->scrubs_paused, 0);
2225 atomic_set(&fs_info->scrub_cancel_req, 0);
2226 init_waitqueue_head(&fs_info->scrub_pause_wait);
2227 fs_info->scrub_workers_refcnt = 0;
2228}
2229
Eric Sandeen779a65a2014-08-01 18:12:39 -05002230static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2231{
2232 spin_lock_init(&fs_info->balance_lock);
2233 mutex_init(&fs_info->balance_mutex);
2234 atomic_set(&fs_info->balance_running, 0);
2235 atomic_set(&fs_info->balance_pause_req, 0);
2236 atomic_set(&fs_info->balance_cancel_req, 0);
2237 fs_info->balance_ctl = NULL;
2238 init_waitqueue_head(&fs_info->balance_wait_q);
2239}
2240
Eric Sandeenf37938e2014-08-01 18:12:40 -05002241static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2242 struct btrfs_root *tree_root)
2243{
2244 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2245 set_nlink(fs_info->btree_inode, 1);
2246 /*
2247 * we set the i_size on the btree inode to the max possible int.
2248 * the real end of the address space is determined by all of
2249 * the devices in the system
2250 */
2251 fs_info->btree_inode->i_size = OFFSET_MAX;
2252 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Eric Sandeenf37938e2014-08-01 18:12:40 -05002253
2254 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2255 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2256 fs_info->btree_inode->i_mapping);
2257 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2258 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2259
2260 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2261
2262 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2263 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2264 sizeof(struct btrfs_key));
2265 set_bit(BTRFS_INODE_DUMMY,
2266 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2267 btrfs_insert_inode_hash(fs_info->btree_inode);
2268}
2269
Eric Sandeenad618362014-08-01 18:12:41 -05002270static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2271{
2272 fs_info->dev_replace.lock_owner = 0;
2273 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2274 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2275 mutex_init(&fs_info->dev_replace.lock_management_lock);
2276 mutex_init(&fs_info->dev_replace.lock);
2277 init_waitqueue_head(&fs_info->replace_wait);
2278}
2279
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002280static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2281{
2282 spin_lock_init(&fs_info->qgroup_lock);
2283 mutex_init(&fs_info->qgroup_ioctl_lock);
2284 fs_info->qgroup_tree = RB_ROOT;
2285 fs_info->qgroup_op_tree = RB_ROOT;
2286 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2287 fs_info->qgroup_seq = 1;
2288 fs_info->quota_enabled = 0;
2289 fs_info->pending_quota_state = 0;
2290 fs_info->qgroup_ulist = NULL;
2291 mutex_init(&fs_info->qgroup_rescan_lock);
2292}
2293
Eric Sandeen2a458192015-02-16 16:29:26 +01002294static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2295 struct btrfs_fs_devices *fs_devices)
2296{
2297 int max_active = fs_info->thread_pool_size;
David Sterba6f011052015-02-16 18:34:01 +01002298 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Eric Sandeen2a458192015-02-16 16:29:26 +01002299
2300 fs_info->workers =
2301 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2302 max_active, 16);
2303
2304 fs_info->delalloc_workers =
2305 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
2306
2307 fs_info->flush_workers =
2308 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
2309
2310 fs_info->caching_workers =
2311 btrfs_alloc_workqueue("cache", flags, max_active, 0);
2312
2313 /*
2314 * a higher idle thresh on the submit workers makes it much more
2315 * likely that bios will be send down in a sane order to the
2316 * devices
2317 */
2318 fs_info->submit_workers =
2319 btrfs_alloc_workqueue("submit", flags,
2320 min_t(u64, fs_devices->num_devices,
2321 max_active), 64);
2322
2323 fs_info->fixup_workers =
2324 btrfs_alloc_workqueue("fixup", flags, 1, 0);
2325
2326 /*
2327 * endios are largely parallel and should have a very
2328 * low idle thresh
2329 */
2330 fs_info->endio_workers =
2331 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2332 fs_info->endio_meta_workers =
2333 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2334 fs_info->endio_meta_write_workers =
2335 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2336 fs_info->endio_raid56_workers =
2337 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
2338 fs_info->endio_repair_workers =
2339 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
2340 fs_info->rmw_workers =
2341 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
2342 fs_info->endio_write_workers =
2343 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2344 fs_info->endio_freespace_worker =
2345 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
2346 fs_info->delayed_workers =
2347 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
2348 fs_info->readahead_workers =
2349 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
2350 fs_info->qgroup_rescan_workers =
2351 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
2352 fs_info->extent_workers =
2353 btrfs_alloc_workqueue("extent-refs", flags,
2354 min_t(u64, fs_devices->num_devices,
2355 max_active), 8);
2356
2357 if (!(fs_info->workers && fs_info->delalloc_workers &&
2358 fs_info->submit_workers && fs_info->flush_workers &&
2359 fs_info->endio_workers && fs_info->endio_meta_workers &&
2360 fs_info->endio_meta_write_workers &&
2361 fs_info->endio_repair_workers &&
2362 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2363 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2364 fs_info->caching_workers && fs_info->readahead_workers &&
2365 fs_info->fixup_workers && fs_info->delayed_workers &&
2366 fs_info->extent_workers &&
2367 fs_info->qgroup_rescan_workers)) {
2368 return -ENOMEM;
2369 }
2370
2371 return 0;
2372}
2373
Eric Sandeen63443bf2014-08-01 18:12:46 -05002374static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2375 struct btrfs_fs_devices *fs_devices)
2376{
2377 int ret;
2378 struct btrfs_root *tree_root = fs_info->tree_root;
2379 struct btrfs_root *log_tree_root;
2380 struct btrfs_super_block *disk_super = fs_info->super_copy;
2381 u64 bytenr = btrfs_super_log_root(disk_super);
2382
2383 if (fs_devices->rw_devices == 0) {
David Sterbaf14d1042015-10-08 11:37:06 +02002384 btrfs_warn(fs_info, "log replay required on RO media");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002385 return -EIO;
2386 }
2387
2388 log_tree_root = btrfs_alloc_root(fs_info);
2389 if (!log_tree_root)
2390 return -ENOMEM;
2391
2392 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2393 tree_root->stripesize, log_tree_root, fs_info,
2394 BTRFS_TREE_LOG_OBJECTID);
2395
2396 log_tree_root->node = read_tree_block(tree_root, bytenr,
2397 fs_info->generation + 1);
Liu Bo64c043d2015-05-25 17:30:15 +08002398 if (IS_ERR(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002399 btrfs_warn(fs_info, "failed to read log tree");
Liu Bo0eeff232015-06-11 14:16:44 +08002400 ret = PTR_ERR(log_tree_root->node);
Liu Bo64c043d2015-05-25 17:30:15 +08002401 kfree(log_tree_root);
Liu Bo0eeff232015-06-11 14:16:44 +08002402 return ret;
Liu Bo64c043d2015-05-25 17:30:15 +08002403 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002404 btrfs_err(fs_info, "failed to read log tree");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002405 free_extent_buffer(log_tree_root->node);
2406 kfree(log_tree_root);
2407 return -EIO;
2408 }
2409 /* returns with log_tree_root freed on success */
2410 ret = btrfs_recover_log_trees(log_tree_root);
2411 if (ret) {
Anand Jaina4553fe2015-09-25 14:43:01 +08002412 btrfs_std_error(tree_root->fs_info, ret,
Eric Sandeen63443bf2014-08-01 18:12:46 -05002413 "Failed to recover log tree");
2414 free_extent_buffer(log_tree_root->node);
2415 kfree(log_tree_root);
2416 return ret;
2417 }
2418
2419 if (fs_info->sb->s_flags & MS_RDONLY) {
2420 ret = btrfs_commit_super(tree_root);
2421 if (ret)
2422 return ret;
2423 }
2424
2425 return 0;
2426}
2427
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002428static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2429 struct btrfs_root *tree_root)
2430{
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002431 struct btrfs_root *root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002432 struct btrfs_key location;
2433 int ret;
2434
2435 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2436 location.type = BTRFS_ROOT_ITEM_KEY;
2437 location.offset = 0;
2438
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002439 root = btrfs_read_tree_root(tree_root, &location);
2440 if (IS_ERR(root))
2441 return PTR_ERR(root);
2442 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2443 fs_info->extent_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002444
2445 location.objectid = BTRFS_DEV_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002446 root = btrfs_read_tree_root(tree_root, &location);
2447 if (IS_ERR(root))
2448 return PTR_ERR(root);
2449 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2450 fs_info->dev_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002451 btrfs_init_devices_late(fs_info);
2452
2453 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002454 root = btrfs_read_tree_root(tree_root, &location);
2455 if (IS_ERR(root))
2456 return PTR_ERR(root);
2457 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2458 fs_info->csum_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002459
2460 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002461 root = btrfs_read_tree_root(tree_root, &location);
2462 if (!IS_ERR(root)) {
2463 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002464 fs_info->quota_enabled = 1;
2465 fs_info->pending_quota_state = 1;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002466 fs_info->quota_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002467 }
2468
2469 location.objectid = BTRFS_UUID_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002470 root = btrfs_read_tree_root(tree_root, &location);
2471 if (IS_ERR(root)) {
2472 ret = PTR_ERR(root);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002473 if (ret != -ENOENT)
2474 return ret;
2475 } else {
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002476 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2477 fs_info->uuid_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002478 }
2479
Omar Sandoval70f6d822015-09-29 20:50:38 -07002480 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2481 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2482 root = btrfs_read_tree_root(tree_root, &location);
2483 if (IS_ERR(root))
2484 return PTR_ERR(root);
2485 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2486 fs_info->free_space_root = root;
2487 }
2488
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002489 return 0;
2490}
2491
Al Viroad2b2c82011-11-17 01:10:02 -05002492int open_ctree(struct super_block *sb,
2493 struct btrfs_fs_devices *fs_devices,
2494 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002495{
Chris Masondb945352007-10-15 16:15:53 -04002496 u32 sectorsize;
2497 u32 nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002498 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002499 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002500 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002501 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002502 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002503 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002504 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002505 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002506 struct btrfs_root *chunk_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002507 int ret;
Yane58ca022008-04-01 11:21:34 -04002508 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002509 int num_backups_tried = 0;
2510 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002511 int max_active;
Chris Mason4543df72008-06-11 21:47:56 -04002512
Al Virof84a8bd2011-11-17 15:57:57 -05002513 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002514 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002515 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002516 err = -ENOMEM;
2517 goto fail;
2518 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002519
2520 ret = init_srcu_struct(&fs_info->subvol_srcu);
2521 if (ret) {
2522 err = ret;
2523 goto fail;
2524 }
2525
2526 ret = setup_bdi(fs_info, &fs_info->bdi);
2527 if (ret) {
2528 err = ret;
2529 goto fail_srcu;
2530 }
2531
Tejun Heo908c7f12014-09-08 09:51:29 +09002532 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
Miao Xiee2d84522013-01-29 10:09:20 +00002533 if (ret) {
2534 err = ret;
2535 goto fail_bdi;
2536 }
2537 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2538 (1 + ilog2(nr_cpu_ids));
2539
Tejun Heo908c7f12014-09-08 09:51:29 +09002540 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
Miao Xie963d6782013-01-29 10:10:51 +00002541 if (ret) {
2542 err = ret;
2543 goto fail_dirty_metadata_bytes;
2544 }
2545
Tejun Heo908c7f12014-09-08 09:51:29 +09002546 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
Miao Xiec404e0d2014-01-30 16:46:55 +08002547 if (ret) {
2548 err = ret;
2549 goto fail_delalloc_bytes;
2550 }
2551
Yan, Zheng76dda932009-09-21 16:00:26 -04002552 fs_info->btree_inode = new_inode(sb);
2553 if (!fs_info->btree_inode) {
2554 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002555 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002556 }
2557
Chris Masona6591712011-07-19 12:04:14 -04002558 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002559
Yan, Zheng76dda932009-09-21 16:00:26 -04002560 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002561 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002562 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002563 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002564 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002565 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002566 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002567 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002568 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002569 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00002570 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002571 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002572 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002573 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002574 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002575 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002576 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002577 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002578 rwlock_init(&fs_info->tree_mod_log_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002579 mutex_init(&fs_info->unused_bg_unpin_mutex);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01002580 mutex_init(&fs_info->delete_unused_bgs_mutex);
Chris Mason75857172011-06-13 20:00:16 -04002581 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002582 mutex_init(&fs_info->delalloc_root_mutex);
Filipe Mananac2d6cb12016-01-15 11:05:12 +00002583 mutex_init(&fs_info->cleaner_delayed_iput_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002584 seqlock_init(&fs_info->profiles_lock);
Chris Mason19c00dd2007-10-15 16:19:22 -04002585
Chris Mason0b86a832008-03-24 15:01:56 -04002586 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002587 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002588 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002589 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002590 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002591 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2592 BTRFS_BLOCK_RSV_GLOBAL);
2593 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2594 BTRFS_BLOCK_RSV_DELALLOC);
2595 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2596 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2597 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2598 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2599 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002600 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002601 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002602 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002603 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002604 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002605 atomic_set(&fs_info->qgroup_op_seq, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002606 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002607 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002608 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002609 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002610 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002611 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002612 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002613 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
David Sterbaf8c269d2015-01-16 17:21:12 +01002614 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
Arne Jansen90519d62011-05-23 14:30:00 +02002615 /* readahead state */
Mel Gormand0164ad2015-11-06 16:28:21 -08002616 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
Arne Jansen90519d62011-05-23 14:30:00 +02002617 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002618
Chris Masonb34b0862008-12-19 15:43:22 -05002619 fs_info->thread_pool_size = min_t(unsigned long,
2620 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002621
Miao Xie199c2a92013-05-15 07:48:23 +00002622 INIT_LIST_HEAD(&fs_info->ordered_roots);
2623 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002624 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2625 GFP_NOFS);
2626 if (!fs_info->delayed_root) {
2627 err = -ENOMEM;
2628 goto fail_iput;
2629 }
2630 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002631
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002632 btrfs_init_scrub(fs_info);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002633#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2634 fs_info->check_integrity_print_mask = 0;
2635#endif
Eric Sandeen779a65a2014-08-01 18:12:39 -05002636 btrfs_init_balance(fs_info);
Miao Xie21c7e752014-05-13 17:29:04 -07002637 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Arne Jansena2de7332011-03-08 14:14:00 +01002638
Chris Masona061fc82008-05-07 11:43:44 -04002639 sb->s_blocksize = 4096;
2640 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002641 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04002642
Eric Sandeenf37938e2014-08-01 18:12:40 -05002643 btrfs_init_btree_inode(fs_info, tree_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002644
Chris Masone02119d2008-09-05 16:13:11 -04002645 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002646 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002647 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002648
Yan Zheng11833d62009-09-11 16:11:19 -04002649 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002650 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002651 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002652 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002653 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002654 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002655
Chris Masond98237b2007-03-28 13:57:48 -04002656
Chris Mason5a3f23d2009-03-31 13:27:11 -04002657 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002658 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002659 mutex_init(&fs_info->chunk_mutex);
2660 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002661 mutex_init(&fs_info->cleaner_mutex);
2662 mutex_init(&fs_info->volume_mutex);
Chris Mason1bbc6212015-04-06 12:46:08 -07002663 mutex_init(&fs_info->ro_block_group_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002664 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002665 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002666 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002667 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002668
Eric Sandeenad618362014-08-01 18:12:41 -05002669 btrfs_init_dev_replace_locks(fs_info);
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002670 btrfs_init_qgroup(fs_info);
Arne Jansen416ac512011-09-13 12:56:09 +02002671
Chris Masonfa9c0d792009-04-03 09:47:43 -04002672 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2673 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2674
Chris Mason7d9eb122008-07-08 14:19:17 -04002675 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002676 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002677 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002678 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002679
Filipe Manana04216822014-11-27 21:14:15 +00002680 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2681
David Woodhouse53b381b2013-01-29 18:40:14 -05002682 ret = btrfs_alloc_stripe_hash_table(fs_info);
2683 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002684 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002685 goto fail_alloc;
2686 }
2687
David Sterba707e8a02014-06-04 19:22:26 +02002688 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002689 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002690
Chris Mason3c4bb262012-03-27 18:56:56 -04002691 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002692
2693 /*
2694 * Read super block and check the signature bytes only
2695 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002696 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Anand Jain92fc03f2015-08-14 18:32:51 +08002697 if (IS_ERR(bh)) {
2698 err = PTR_ERR(bh);
Miao Xie16cdcec2011-04-22 18:12:22 +08002699 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002700 }
Chris Mason39279cc2007-06-12 06:35:45 -04002701
David Sterba1104a882013-03-06 15:57:46 +01002702 /*
2703 * We want to check superblock checksum, the type is stored inside.
2704 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2705 */
2706 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002707 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002708 err = -EINVAL;
Anand Jainb2acddd2015-10-07 17:23:23 +08002709 brelse(bh);
David Sterba1104a882013-03-06 15:57:46 +01002710 goto fail_alloc;
2711 }
2712
2713 /*
2714 * super_copy is zeroed at allocation time and we never touch the
2715 * following bytes up to INFO_SIZE, the checksum is calculated from
2716 * the whole block of INFO_SIZE
2717 */
David Sterba6c417612011-04-13 15:41:04 +02002718 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2719 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2720 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002721 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002722
David Sterba6c417612011-04-13 15:41:04 +02002723 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002724
David Sterba1104a882013-03-06 15:57:46 +01002725 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2726 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002727 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002728 err = -EINVAL;
2729 goto fail_alloc;
2730 }
2731
David Sterba6c417612011-04-13 15:41:04 +02002732 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002733 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002734 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002735
liuboacce9522011-01-06 19:30:25 +08002736 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002737 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2738 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002739
Liu Bo75e7cb72011-03-22 10:12:20 +00002740 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002741 * run through our array of backup supers and setup
2742 * our ring pointer to the oldest one
2743 */
2744 generation = btrfs_super_generation(disk_super);
2745 find_oldest_super_backup(fs_info, generation);
2746
2747 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002748 * In the long term, we'll store the compression type in the super
2749 * block, and it'll be used for per file compression control.
2750 */
2751 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2752
Yan Zheng2b820322008-11-17 21:11:30 -05002753 ret = btrfs_parse_options(tree_root, options);
2754 if (ret) {
2755 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002756 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002757 }
Chris Masondfe25022008-05-13 13:46:40 -04002758
Josef Bacikf2b636e2008-12-02 06:36:08 -05002759 features = btrfs_super_incompat_flags(disk_super) &
2760 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2761 if (features) {
2762 printk(KERN_ERR "BTRFS: couldn't mount because of "
2763 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002764 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002765 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002766 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002767 }
2768
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002769 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002770 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002771 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002772 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002773
Josef Bacik3173a182013-03-07 14:22:04 -05002774 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
David Sterba5efa0492014-12-19 18:38:47 +01002775 printk(KERN_INFO "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002776
Chris Mason727011e2010-08-06 13:21:20 -04002777 /*
2778 * flag our filesystem as having big metadata blocks if
2779 * they are bigger than the page size
2780 */
David Sterba707e8a02014-06-04 19:22:26 +02002781 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002782 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002783 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002784 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2785 }
2786
Chris Masonbc3f1162012-03-29 17:02:47 -04002787 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002788 sectorsize = btrfs_super_sectorsize(disk_super);
2789 stripesize = btrfs_super_stripesize(disk_super);
David Sterba707e8a02014-06-04 19:22:26 +02002790 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002791 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002792
2793 /*
2794 * mixed block groups end up with duplicate but slightly offset
2795 * extent buffers for the same range. It leads to corruptions
2796 */
2797 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002798 (sectorsize != nodesize)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002799 printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002800 "are not allowed for mixed block groups on %s\n",
2801 sb->s_id);
2802 goto fail_alloc;
2803 }
2804
Miao Xieceda0862013-04-11 10:30:16 +00002805 /*
2806 * Needn't use the lock because there is no other task which will
2807 * update the flag.
2808 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002809 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002810
Josef Bacikf2b636e2008-12-02 06:36:08 -05002811 features = btrfs_super_compat_ro_flags(disk_super) &
2812 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2813 if (!(sb->s_flags & MS_RDONLY) && features) {
2814 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2815 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002816 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002817 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002818 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002819 }
Chris Mason61d92c32009-10-02 19:11:56 -04002820
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002821 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002822
Eric Sandeen2a458192015-02-16 16:29:26 +01002823 ret = btrfs_init_workqueues(fs_info, fs_devices);
2824 if (ret) {
2825 err = ret;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002826 goto fail_sb_buffer;
2827 }
Chris Mason4543df72008-06-11 21:47:56 -04002828
Chris Mason4575c9c2008-04-18 16:13:31 -04002829 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002830 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
Byongho Leeee221842015-12-15 01:42:10 +09002831 SZ_4M / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002832
Chris Masondb945352007-10-15 16:15:53 -04002833 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002834 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002835 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002836
2837 sb->s_blocksize = sectorsize;
2838 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002839
Chris Mason925baed2008-06-25 16:01:30 -04002840 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002841 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002842 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002843 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002844 printk(KERN_ERR "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002845 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002846 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002847 }
Chris Mason0b86a832008-03-24 15:01:56 -04002848
Yan Zheng84234f32008-10-29 14:49:05 -04002849 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002850
David Sterba707e8a02014-06-04 19:22:26 +02002851 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2852 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002853
2854 chunk_root->node = read_tree_block(chunk_root,
2855 btrfs_super_chunk_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002856 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002857 if (IS_ERR(chunk_root->node) ||
2858 !extent_buffer_uptodate(chunk_root->node)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002859 printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002860 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302861 if (!IS_ERR(chunk_root->node))
2862 free_extent_buffer(chunk_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002863 chunk_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002864 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002865 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002866 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2867 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002868
Chris Masone17cade2008-04-15 15:41:47 -04002869 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002870 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002871
Chris Mason0b86a832008-03-24 15:01:56 -04002872 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002873 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002874 printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002875 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002876 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002877 }
Chris Mason0b86a832008-03-24 15:01:56 -04002878
Stefan Behrens8dabb742012-11-06 13:15:27 +01002879 /*
2880 * keep the device that is marked to be the target device for the
2881 * dev_replace procedure
2882 */
Eric Sandeen9eaed212014-08-01 18:12:35 -05002883 btrfs_close_extra_devices(fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002884
Chris Masona6b0d5c2012-02-20 20:53:43 -05002885 if (!fs_devices->latest_bdev) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002886 printk(KERN_ERR "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002887 sb->s_id);
2888 goto fail_tree_roots;
2889 }
2890
Chris Masonaf31f5e2011-11-03 15:17:42 -04002891retry_root_backup:
Yan Zheng84234f32008-10-29 14:49:05 -04002892 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002893
Chris Masone20d96d2007-03-22 12:13:20 -04002894 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002895 btrfs_super_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002896 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002897 if (IS_ERR(tree_root->node) ||
2898 !extent_buffer_uptodate(tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002899 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002900 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302901 if (!IS_ERR(tree_root->node))
2902 free_extent_buffer(tree_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002903 tree_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002904 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002905 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002906
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002907 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2908 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002909 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002910
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05302911 mutex_lock(&tree_root->objectid_mutex);
2912 ret = btrfs_find_highest_objectid(tree_root,
2913 &tree_root->highest_objectid);
2914 if (ret) {
2915 mutex_unlock(&tree_root->objectid_mutex);
2916 goto recovery_tree_root;
2917 }
2918
2919 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
2920
2921 mutex_unlock(&tree_root->objectid_mutex);
2922
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002923 ret = btrfs_read_roots(fs_info, tree_root);
2924 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002925 goto recovery_tree_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002926
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002927 fs_info->generation = generation;
2928 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002929
Ilya Dryomov68310a52012-06-22 12:24:12 -06002930 ret = btrfs_recover_balance(fs_info);
2931 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002932 printk(KERN_ERR "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002933 goto fail_block_groups;
2934 }
2935
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002936 ret = btrfs_init_dev_stats(fs_info);
2937 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002938 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002939 ret);
2940 goto fail_block_groups;
2941 }
2942
Stefan Behrens8dabb742012-11-06 13:15:27 +01002943 ret = btrfs_init_dev_replace(fs_info);
2944 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002945 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002946 goto fail_block_groups;
2947 }
2948
Eric Sandeen9eaed212014-08-01 18:12:35 -05002949 btrfs_close_extra_devices(fs_devices, 1);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002950
Anand Jainb7c35e82015-03-10 06:38:38 +08002951 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
2952 if (ret) {
2953 pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret);
2954 goto fail_block_groups;
2955 }
2956
2957 ret = btrfs_sysfs_add_device(fs_devices);
2958 if (ret) {
2959 pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret);
2960 goto fail_fsdev_sysfs;
2961 }
2962
Anand Jain96f31362015-08-14 18:32:46 +08002963 ret = btrfs_sysfs_add_mounted(fs_info);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002964 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002965 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08002966 goto fail_fsdev_sysfs;
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002967 }
2968
liuboc59021f2011-03-07 02:13:14 +00002969 ret = btrfs_init_space_info(fs_info);
2970 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002971 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002972 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002973 }
2974
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002975 ret = btrfs_read_block_groups(fs_info->extent_root);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002976 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002977 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002978 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002979 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02002980 fs_info->num_tolerated_disk_barrier_failures =
2981 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002982 if (fs_info->fs_devices->missing_devices >
2983 fs_info->num_tolerated_disk_barrier_failures &&
2984 !(sb->s_flags & MS_RDONLY)) {
Zhao Lei78fa1772015-07-20 17:54:50 +08002985 pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n",
2986 fs_info->fs_devices->missing_devices,
2987 fs_info->num_tolerated_disk_barrier_failures);
Anand Jain2365dd32014-01-22 11:15:51 +08002988 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00002989 }
Chris Mason9078a3e2007-04-26 16:46:15 -04002990
Chris Masona74a4b92008-06-25 16:01:31 -04002991 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2992 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002993 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08002994 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04002995
2996 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2997 tree_root,
2998 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05002999 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04003000 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04003001
Chris Masonc2898112009-06-10 09:51:32 -04003002 if (!btrfs_test_opt(tree_root, SSD) &&
3003 !btrfs_test_opt(tree_root, NOSSD) &&
3004 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003005 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04003006 "mode\n");
3007 btrfs_set_opt(fs_info->mount_opt, SSD);
3008 }
3009
David Sterba572d9ab2014-02-05 15:26:17 +01003010 /*
3011 * Mount does not set all options immediatelly, we can do it now and do
3012 * not have to wait for transaction commit
3013 */
3014 btrfs_apply_pending_changes(fs_info);
Qu Wenruo3818aea2014-01-13 13:36:06 +08003015
Stefan Behrens21adbd52011-11-09 13:44:05 +01003016#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3017 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
3018 ret = btrfsic_mount(tree_root, fs_devices,
3019 btrfs_test_opt(tree_root,
3020 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3021 1 : 0,
3022 fs_info->check_integrity_print_mask);
3023 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003024 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01003025 " integrity check module %s\n", sb->s_id);
3026 }
3027#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02003028 ret = btrfs_read_qgroup_config(fs_info);
3029 if (ret)
3030 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01003031
liuboacce9522011-01-06 19:30:25 +08003032 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06003033 if (btrfs_super_log_root(disk_super) != 0) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003034 ret = btrfs_replay_log(fs_info, fs_devices);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003035 if (ret) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003036 err = ret;
Wang Shilong28c16cb2014-04-23 19:33:35 +08003037 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003038 }
Chris Masone02119d2008-09-05 16:13:11 -04003039 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003040
Yan, Zheng76dda932009-09-21 16:00:26 -04003041 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003042 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003043 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04003044
Chris Mason7c2ca462008-11-19 15:13:35 -05003045 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04003046 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06003047 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003048 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04003049
Wang Shilong5f316482014-06-26 11:08:16 +08003050 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003051 ret = btrfs_recover_relocation(tree_root);
Wang Shilong5f316482014-06-26 11:08:16 +08003052 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00003053 if (ret < 0) {
3054 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05003055 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00003056 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02003057 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00003058 }
Chris Mason7c2ca462008-11-19 15:13:35 -05003059 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003060
Chris Mason3de45862008-11-17 21:02:50 -05003061 location.objectid = BTRFS_FS_TREE_OBJECTID;
3062 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00003063 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05003064
Chris Mason3de45862008-11-17 21:02:50 -05003065 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003066 if (IS_ERR(fs_info->fs_root)) {
3067 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003068 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003069 }
Chris Masonc2898112009-06-10 09:51:32 -04003070
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003071 if (sb->s_flags & MS_RDONLY)
3072 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02003073
Chris Mason511711a2015-12-30 07:52:35 -08003074 if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) &&
3075 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3076 pr_info("BTRFS: creating free space tree\n");
3077 ret = btrfs_create_free_space_tree(fs_info);
3078 if (ret) {
3079 pr_warn("BTRFS: failed to create free space tree %d\n",
3080 ret);
3081 close_ctree(tree_root);
3082 return ret;
3083 }
3084 }
3085
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003086 down_read(&fs_info->cleanup_work_sem);
3087 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3088 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00003089 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003090 close_ctree(tree_root);
3091 return ret;
3092 }
3093 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02003094
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003095 ret = btrfs_resume_balance_async(fs_info);
3096 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003097 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003098 close_ctree(tree_root);
3099 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00003100 }
3101
Stefan Behrens8dabb742012-11-06 13:15:27 +01003102 ret = btrfs_resume_dev_replace_async(fs_info);
3103 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003104 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01003105 close_ctree(tree_root);
3106 return ret;
3107 }
3108
Jan Schmidtb382a322013-05-28 15:47:24 +00003109 btrfs_qgroup_rescan_resume(fs_info);
3110
Omar Sandoval70f6d822015-09-29 20:50:38 -07003111 if (btrfs_test_opt(tree_root, CLEAR_CACHE) &&
3112 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3113 pr_info("BTRFS: clearing free space tree\n");
3114 ret = btrfs_clear_free_space_tree(fs_info);
3115 if (ret) {
3116 pr_warn("BTRFS: failed to clear free space tree %d\n",
3117 ret);
3118 close_ctree(tree_root);
3119 return ret;
3120 }
3121 }
3122
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003123 if (!fs_info->uuid_root) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003124 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003125 ret = btrfs_create_uuid_tree(fs_info);
3126 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003127 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003128 ret);
3129 close_ctree(tree_root);
3130 return ret;
3131 }
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003132 } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) ||
3133 fs_info->generation !=
3134 btrfs_super_uuid_tree_generation(disk_super)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003135 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02003136 ret = btrfs_check_uuid_tree(fs_info);
3137 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003138 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02003139 ret);
3140 close_ctree(tree_root);
3141 return ret;
3142 }
3143 } else {
3144 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003145 }
3146
Josef Bacik47ab2a62014-09-18 11:20:02 -04003147 fs_info->open = 1;
3148
Al Viroad2b2c82011-11-17 01:10:02 -05003149 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04003150
Arne Jansenbcef60f2011-09-13 15:23:30 +02003151fail_qgroup:
3152 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05003153fail_trans_kthread:
3154 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04003155 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003156 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04003157fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04003158 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05003159
3160 /*
3161 * make sure we're done with the btree inode before we stop our
3162 * kthreads
3163 */
3164 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05003165
Anand Jain2365dd32014-01-22 11:15:51 +08003166fail_sysfs:
Anand Jain6618a592015-08-14 18:32:47 +08003167 btrfs_sysfs_remove_mounted(fs_info);
Anand Jain2365dd32014-01-22 11:15:51 +08003168
Anand Jainb7c35e82015-03-10 06:38:38 +08003169fail_fsdev_sysfs:
3170 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3171
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003172fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003173 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003174 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003175
3176fail_tree_roots:
3177 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003178 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003179
Chris Mason39279cc2007-06-12 06:35:45 -04003180fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003181 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003182fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003183fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003184 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3185
Chris Mason4543df72008-06-11 21:47:56 -04003186 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003187fail_bio_counter:
3188 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003189fail_delalloc_bytes:
3190 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003191fail_dirty_metadata_bytes:
3192 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003193fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003194 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003195fail_srcu:
3196 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003197fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003198 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003199 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003200 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003201
3202recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003203 if (!btrfs_test_opt(tree_root, RECOVERY))
3204 goto fail_tree_roots;
3205
3206 free_root_pointers(fs_info, 0);
3207
3208 /* don't use the log in recovery mode, it won't be valid */
3209 btrfs_set_super_log_root(disk_super, 0);
3210
3211 /* we can't trust the free space cache either */
3212 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3213
3214 ret = next_root_backup(fs_info, fs_info->super_copy,
3215 &num_backups_tried, &backup_index);
3216 if (ret == -1)
3217 goto fail_block_groups;
3218 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003219}
3220
Chris Masonf2984462008-04-10 16:19:33 -04003221static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3222{
Chris Masonf2984462008-04-10 16:19:33 -04003223 if (uptodate) {
3224 set_buffer_uptodate(bh);
3225 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003226 struct btrfs_device *device = (struct btrfs_device *)
3227 bh->b_private;
3228
David Sterbab14af3b2015-10-08 10:43:10 +02003229 btrfs_warn_rl_in_rcu(device->dev_root->fs_info,
3230 "lost page write due to IO error on %s",
Josef Bacik606686e2012-06-04 14:03:51 -04003231 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003232 /* note, we dont' set_buffer_write_io_error because we have
3233 * our own ways of dealing with the IO errors
3234 */
Chris Masonf2984462008-04-10 16:19:33 -04003235 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003236 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003237 }
3238 unlock_buffer(bh);
3239 put_bh(bh);
3240}
3241
Anand Jain29c36d72015-08-14 18:32:58 +08003242int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3243 struct buffer_head **bh_ret)
3244{
3245 struct buffer_head *bh;
3246 struct btrfs_super_block *super;
3247 u64 bytenr;
3248
3249 bytenr = btrfs_sb_offset(copy_num);
3250 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3251 return -EINVAL;
3252
3253 bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE);
3254 /*
3255 * If we fail to read from the underlying devices, as of now
3256 * the best option we have is to mark it EIO.
3257 */
3258 if (!bh)
3259 return -EIO;
3260
3261 super = (struct btrfs_super_block *)bh->b_data;
3262 if (btrfs_super_bytenr(super) != bytenr ||
3263 btrfs_super_magic(super) != BTRFS_MAGIC) {
3264 brelse(bh);
3265 return -EINVAL;
3266 }
3267
3268 *bh_ret = bh;
3269 return 0;
3270}
3271
3272
Yan Zhenga512bbf2008-12-08 16:46:26 -05003273struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3274{
3275 struct buffer_head *bh;
3276 struct buffer_head *latest = NULL;
3277 struct btrfs_super_block *super;
3278 int i;
3279 u64 transid = 0;
Anand Jain92fc03f2015-08-14 18:32:51 +08003280 int ret = -EINVAL;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003281
3282 /* we would like to check all the supers, but that would make
3283 * a btrfs mount succeed after a mkfs from a different FS.
3284 * So, we need to add a special mount option to scan for
3285 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3286 */
3287 for (i = 0; i < 1; i++) {
Anand Jain29c36d72015-08-14 18:32:58 +08003288 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3289 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003290 continue;
3291
3292 super = (struct btrfs_super_block *)bh->b_data;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003293
3294 if (!latest || btrfs_super_generation(super) > transid) {
3295 brelse(latest);
3296 latest = bh;
3297 transid = btrfs_super_generation(super);
3298 } else {
3299 brelse(bh);
3300 }
3301 }
Anand Jain92fc03f2015-08-14 18:32:51 +08003302
3303 if (!latest)
3304 return ERR_PTR(ret);
3305
Yan Zhenga512bbf2008-12-08 16:46:26 -05003306 return latest;
3307}
3308
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003309/*
3310 * this should be called twice, once with wait == 0 and
3311 * once with wait == 1. When wait == 0 is done, all the buffer heads
3312 * we write are pinned.
3313 *
3314 * They are released when wait == 1 is done.
3315 * max_mirrors must be the same for both runs, and it indicates how
3316 * many supers on this one device should be written.
3317 *
3318 * max_mirrors == 0 means to write them all.
3319 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003320static int write_dev_supers(struct btrfs_device *device,
3321 struct btrfs_super_block *sb,
3322 int do_barriers, int wait, int max_mirrors)
3323{
3324 struct buffer_head *bh;
3325 int i;
3326 int ret;
3327 int errors = 0;
3328 u32 crc;
3329 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003330
3331 if (max_mirrors == 0)
3332 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3333
Yan Zhenga512bbf2008-12-08 16:46:26 -05003334 for (i = 0; i < max_mirrors; i++) {
3335 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003336 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3337 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003338 break;
3339
3340 if (wait) {
3341 bh = __find_get_block(device->bdev, bytenr / 4096,
3342 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003343 if (!bh) {
3344 errors++;
3345 continue;
3346 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003347 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003348 if (!buffer_uptodate(bh))
3349 errors++;
3350
3351 /* drop our reference */
3352 brelse(bh);
3353
3354 /* drop the reference from the wait == 0 run */
3355 brelse(bh);
3356 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003357 } else {
3358 btrfs_set_super_bytenr(sb, bytenr);
3359
3360 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003361 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003362 BTRFS_CSUM_SIZE, crc,
3363 BTRFS_SUPER_INFO_SIZE -
3364 BTRFS_CSUM_SIZE);
3365 btrfs_csum_final(crc, sb->csum);
3366
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003367 /*
3368 * one reference for us, and we leave it for the
3369 * caller
3370 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003371 bh = __getblk(device->bdev, bytenr / 4096,
3372 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003373 if (!bh) {
David Sterbaf14d1042015-10-08 11:37:06 +02003374 btrfs_err(device->dev_root->fs_info,
3375 "couldn't get super buffer head for bytenr %llu",
3376 bytenr);
Josef Bacik634554d2013-04-29 10:05:57 -04003377 errors++;
3378 continue;
3379 }
3380
Yan Zhenga512bbf2008-12-08 16:46:26 -05003381 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3382
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003383 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003384 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003385
3386 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003387 lock_buffer(bh);
3388 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003389 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003390 }
3391
Chris Mason387125f2011-11-18 15:07:51 -05003392 /*
3393 * we fua the first super. The others we allow
3394 * to go down lazy.
3395 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003396 if (i == 0)
3397 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3398 else
3399 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003400 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003401 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003402 }
3403 return errors < i ? 0 : -1;
3404}
3405
Chris Mason387125f2011-11-18 15:07:51 -05003406/*
3407 * endio for the write_dev_flush, this will wake anyone waiting
3408 * for the barrier when it is done
3409 */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003410static void btrfs_end_empty_barrier(struct bio *bio)
Chris Mason387125f2011-11-18 15:07:51 -05003411{
Chris Mason387125f2011-11-18 15:07:51 -05003412 if (bio->bi_private)
3413 complete(bio->bi_private);
3414 bio_put(bio);
3415}
3416
3417/*
3418 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3419 * sent down. With wait == 1, it waits for the previous flush.
3420 *
3421 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3422 * capable
3423 */
3424static int write_dev_flush(struct btrfs_device *device, int wait)
3425{
3426 struct bio *bio;
3427 int ret = 0;
3428
3429 if (device->nobarriers)
3430 return 0;
3431
3432 if (wait) {
3433 bio = device->flush_bio;
3434 if (!bio)
3435 return 0;
3436
3437 wait_for_completion(&device->flush_wait);
3438
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003439 if (bio->bi_error) {
3440 ret = bio->bi_error;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003441 btrfs_dev_stat_inc_and_print(device,
3442 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003443 }
3444
3445 /* drop the reference from the wait == 0 run */
3446 bio_put(bio);
3447 device->flush_bio = NULL;
3448
3449 return ret;
3450 }
3451
3452 /*
3453 * one reference for us, and we leave it for the
3454 * caller
3455 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003456 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003457 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003458 if (!bio)
3459 return -ENOMEM;
3460
3461 bio->bi_end_io = btrfs_end_empty_barrier;
3462 bio->bi_bdev = device->bdev;
3463 init_completion(&device->flush_wait);
3464 bio->bi_private = &device->flush_wait;
3465 device->flush_bio = bio;
3466
3467 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003468 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003469
3470 return 0;
3471}
3472
3473/*
3474 * send an empty flush down to each device in parallel,
3475 * then wait for them
3476 */
3477static int barrier_all_devices(struct btrfs_fs_info *info)
3478{
3479 struct list_head *head;
3480 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003481 int errors_send = 0;
3482 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003483 int ret;
3484
3485 /* send down all the barriers */
3486 head = &info->fs_devices->devices;
3487 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003488 if (dev->missing)
3489 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003490 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003491 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003492 continue;
3493 }
3494 if (!dev->in_fs_metadata || !dev->writeable)
3495 continue;
3496
3497 ret = write_dev_flush(dev, 0);
3498 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003499 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003500 }
3501
3502 /* wait for all the barriers */
3503 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003504 if (dev->missing)
3505 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003506 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003507 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003508 continue;
3509 }
3510 if (!dev->in_fs_metadata || !dev->writeable)
3511 continue;
3512
3513 ret = write_dev_flush(dev, 1);
3514 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003515 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003516 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003517 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3518 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003519 return -EIO;
3520 return 0;
3521}
3522
Zhao Lei943c6e92015-08-19 15:54:15 +08003523int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3524{
Zhao Lei8789f4f2015-09-15 21:08:07 +08003525 int raid_type;
3526 int min_tolerated = INT_MAX;
Zhao Lei943c6e92015-08-19 15:54:15 +08003527
Zhao Lei8789f4f2015-09-15 21:08:07 +08003528 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3529 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3530 min_tolerated = min(min_tolerated,
3531 btrfs_raid_array[BTRFS_RAID_SINGLE].
3532 tolerated_failures);
Zhao Lei943c6e92015-08-19 15:54:15 +08003533
Zhao Lei8789f4f2015-09-15 21:08:07 +08003534 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3535 if (raid_type == BTRFS_RAID_SINGLE)
3536 continue;
3537 if (!(flags & btrfs_raid_group[raid_type]))
3538 continue;
3539 min_tolerated = min(min_tolerated,
3540 btrfs_raid_array[raid_type].
3541 tolerated_failures);
3542 }
Zhao Lei943c6e92015-08-19 15:54:15 +08003543
Zhao Lei8789f4f2015-09-15 21:08:07 +08003544 if (min_tolerated == INT_MAX) {
3545 pr_warn("BTRFS: unknown raid flag: %llu\n", flags);
3546 min_tolerated = 0;
3547 }
3548
3549 return min_tolerated;
Zhao Lei943c6e92015-08-19 15:54:15 +08003550}
3551
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003552int btrfs_calc_num_tolerated_disk_barrier_failures(
3553 struct btrfs_fs_info *fs_info)
3554{
3555 struct btrfs_ioctl_space_info space;
3556 struct btrfs_space_info *sinfo;
3557 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3558 BTRFS_BLOCK_GROUP_SYSTEM,
3559 BTRFS_BLOCK_GROUP_METADATA,
3560 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003561 int i;
3562 int c;
3563 int num_tolerated_disk_barrier_failures =
3564 (int)fs_info->fs_devices->num_devices;
3565
Zhao Lei2c458042015-07-16 15:00:46 +08003566 for (i = 0; i < ARRAY_SIZE(types); i++) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003567 struct btrfs_space_info *tmp;
3568
3569 sinfo = NULL;
3570 rcu_read_lock();
3571 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3572 if (tmp->flags == types[i]) {
3573 sinfo = tmp;
3574 break;
3575 }
3576 }
3577 rcu_read_unlock();
3578
3579 if (!sinfo)
3580 continue;
3581
3582 down_read(&sinfo->groups_sem);
3583 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
Zhao Lei2c458042015-07-16 15:00:46 +08003584 u64 flags;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003585
Zhao Lei2c458042015-07-16 15:00:46 +08003586 if (list_empty(&sinfo->block_groups[c]))
3587 continue;
3588
3589 btrfs_get_block_group_info(&sinfo->block_groups[c],
3590 &space);
3591 if (space.total_bytes == 0 || space.used_bytes == 0)
3592 continue;
3593 flags = space.flags;
Zhao Lei943c6e92015-08-19 15:54:15 +08003594
3595 num_tolerated_disk_barrier_failures = min(
3596 num_tolerated_disk_barrier_failures,
3597 btrfs_get_num_tolerated_disk_barrier_failures(
3598 flags));
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003599 }
3600 up_read(&sinfo->groups_sem);
3601 }
3602
3603 return num_tolerated_disk_barrier_failures;
3604}
3605
Eric Sandeen48a3b632013-04-25 20:41:01 +00003606static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003607{
Chris Masone5e9a522009-06-10 15:17:02 -04003608 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003609 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003610 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003611 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003612 int ret;
3613 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003614 int max_errors;
3615 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003616 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003617
3618 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003619 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003620
David Sterba6c417612011-04-13 15:41:04 +02003621 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003622 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003623
Chris Mason174ba502011-05-27 10:03:58 -04003624 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003625 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003626 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003627
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003628 if (do_barriers) {
3629 ret = barrier_all_devices(root->fs_info);
3630 if (ret) {
3631 mutex_unlock(
3632 &root->fs_info->fs_devices->device_list_mutex);
Anand Jaina4553fe2015-09-25 14:43:01 +08003633 btrfs_std_error(root->fs_info, ret,
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003634 "errors while submitting device barriers.");
3635 return ret;
3636 }
3637 }
Chris Mason387125f2011-11-18 15:07:51 -05003638
Xiao Guangrong1f781602011-04-20 10:09:16 +00003639 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003640 if (!dev->bdev) {
3641 total_errors++;
3642 continue;
3643 }
Yan Zheng2b820322008-11-17 21:11:30 -05003644 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003645 continue;
3646
Yan Zheng2b820322008-11-17 21:11:30 -05003647 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003648 btrfs_set_stack_device_type(dev_item, dev->type);
3649 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003650 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003651 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003652 btrfs_set_stack_device_bytes_used(dev_item,
3653 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003654 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3655 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3656 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3657 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003658 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003659
Chris Masona061fc82008-05-07 11:43:44 -04003660 flags = btrfs_super_flags(sb);
3661 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003662
Yan Zhenga512bbf2008-12-08 16:46:26 -05003663 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003664 if (ret)
3665 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003666 }
Chris Masona236aed2008-04-29 09:38:00 -04003667 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003668 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003669 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003670 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003671
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003672 /* FUA is masked off if unsupported and can't be the reason */
Anand Jaina4553fe2015-09-25 14:43:01 +08003673 btrfs_std_error(root->fs_info, -EIO,
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003674 "%d errors while writing supers", total_errors);
3675 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003676 }
Chris Masonf2984462008-04-10 16:19:33 -04003677
Yan Zhenga512bbf2008-12-08 16:46:26 -05003678 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003679 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003680 if (!dev->bdev)
3681 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003682 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003683 continue;
3684
Yan Zhenga512bbf2008-12-08 16:46:26 -05003685 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3686 if (ret)
3687 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003688 }
Chris Mason174ba502011-05-27 10:03:58 -04003689 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003690 if (total_errors > max_errors) {
Anand Jaina4553fe2015-09-25 14:43:01 +08003691 btrfs_std_error(root->fs_info, -EIO,
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003692 "%d errors while writing supers", total_errors);
3693 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003694 }
Chris Masonf2984462008-04-10 16:19:33 -04003695 return 0;
3696}
3697
Yan Zhenga512bbf2008-12-08 16:46:26 -05003698int write_ctree_super(struct btrfs_trans_handle *trans,
3699 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003700{
Rashikaf570e7572013-10-31 16:50:42 +05303701 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003702}
3703
Miao Xiecb517ea2013-05-15 07:48:19 +00003704/* Drop a fs root from the radix tree and free it. */
3705void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3706 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003707{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003708 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003709 radix_tree_delete(&fs_info->fs_roots_radix,
3710 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003711 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003712
3713 if (btrfs_root_refs(&root->root_item) == 0)
3714 synchronize_srcu(&fs_info->subvol_srcu);
3715
Liu Bo1a4319c2014-01-13 19:53:53 +08003716 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003717 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003718
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003719 if (root->free_ino_pinned)
3720 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3721 if (root->free_ino_ctl)
3722 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003723 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003724}
3725
3726static void free_fs_root(struct btrfs_root *root)
3727{
David Sterba57cdc8d2014-02-05 02:37:48 +01003728 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003729 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003730 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3731 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003732 if (root->anon_dev)
3733 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003734 if (root->subv_writers)
3735 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003736 free_extent_buffer(root->node);
3737 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003738 kfree(root->free_ino_ctl);
3739 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003740 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003741 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003742}
3743
Miao Xiecb517ea2013-05-15 07:48:19 +00003744void btrfs_free_fs_root(struct btrfs_root *root)
3745{
3746 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003747}
3748
Yan Zhengc146afa2008-11-12 14:34:12 -05003749int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3750{
3751 u64 root_objectid = 0;
3752 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003753 int i = 0;
3754 int err = 0;
3755 unsigned int ret = 0;
3756 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003757
3758 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003759 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003760 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3761 (void **)gang, root_objectid,
3762 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003763 if (!ret) {
3764 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003765 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003766 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003767 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003768
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003769 for (i = 0; i < ret; i++) {
3770 /* Avoid to grab roots in dead_roots */
3771 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3772 gang[i] = NULL;
3773 continue;
3774 }
3775 /* grab all the search result for later use */
3776 gang[i] = btrfs_grab_fs_root(gang[i]);
3777 }
3778 srcu_read_unlock(&fs_info->subvol_srcu, index);
3779
3780 for (i = 0; i < ret; i++) {
3781 if (!gang[i])
3782 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003783 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003784 err = btrfs_orphan_cleanup(gang[i]);
3785 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003786 break;
3787 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003788 }
3789 root_objectid++;
3790 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003791
3792 /* release the uncleaned roots due to error */
3793 for (; i < ret; i++) {
3794 if (gang[i])
3795 btrfs_put_fs_root(gang[i]);
3796 }
3797 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003798}
3799
3800int btrfs_commit_super(struct btrfs_root *root)
3801{
3802 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003803
3804 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf02009-11-12 09:36:34 +00003805 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003806 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003807 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003808
3809 /* wait until ongoing cleanup work done */
3810 down_write(&root->fs_info->cleanup_work_sem);
3811 up_write(&root->fs_info->cleanup_work_sem);
3812
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003813 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003814 if (IS_ERR(trans))
3815 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003816 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003817}
3818
David Sterba3abdbd72014-06-04 18:10:45 +02003819void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003820{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003821 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003822 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003823
Chris Masonfacda1e2007-06-08 18:11:48 -04003824 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003825 smp_mb();
3826
Justin Maggard7343dd62015-11-04 15:56:16 -08003827 /* wait for the qgroup rescan worker to stop */
3828 btrfs_qgroup_wait_for_completion(fs_info);
3829
Stefan Behrens803b2f52013-08-15 17:11:21 +02003830 /* wait for the uuid_scan task to finish */
3831 down(&fs_info->uuid_tree_rescan_sem);
3832 /* avoid complains from lockdep et al., set sem back to initial state */
3833 up(&fs_info->uuid_tree_rescan_sem);
3834
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003835 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003836 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003837
Stefan Behrens8dabb742012-11-06 13:15:27 +01003838 btrfs_dev_replace_suspend_for_unmount(fs_info);
3839
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003840 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003841
3842 /* wait for any defraggers to finish */
3843 wait_event(fs_info->transaction_wait,
3844 (atomic_read(&fs_info->defrag_running) == 0));
3845
3846 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003847 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003848
Miao Xie21c7e752014-05-13 17:29:04 -07003849 cancel_work_sync(&fs_info->async_reclaim_work);
3850
Yan Zhengc146afa2008-11-12 14:34:12 -05003851 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003852 /*
3853 * If the cleaner thread is stopped and there are
3854 * block groups queued for removal, the deletion will be
3855 * skipped when we quit the cleaner thread.
3856 */
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003857 btrfs_delete_unused_bgs(root->fs_info);
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003858
liuboacce9522011-01-06 19:30:25 +08003859 ret = btrfs_commit_super(root);
3860 if (ret)
Eric Sandeen04892342014-08-01 18:12:36 -05003861 btrfs_err(fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003862 }
3863
Miao Xie87533c42013-01-29 10:14:48 +00003864 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003865 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003866
Al Viroe3029d92011-11-17 00:56:18 -05003867 kthread_stop(fs_info->transaction_kthread);
3868 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003869
Yan Zhengf25784b2009-07-28 08:41:57 -04003870 fs_info->closing = 2;
3871 smp_mb();
3872
Eric Sandeen04892342014-08-01 18:12:36 -05003873 btrfs_free_qgroup_config(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003874
Miao Xie963d6782013-01-29 10:10:51 +00003875 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Eric Sandeen04892342014-08-01 18:12:36 -05003876 btrfs_info(fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003877 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003878 }
Yanbcc63ab2008-07-30 16:29:20 -04003879
Anand Jain6618a592015-08-14 18:32:47 +08003880 btrfs_sysfs_remove_mounted(fs_info);
Anand Jainb7c35e82015-03-10 06:38:38 +08003881 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003882
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003883 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003884
Liu Bo1a4319c2014-01-13 19:53:53 +08003885 btrfs_put_block_group_cache(fs_info);
3886
Josef Bacik2b1360d2013-10-07 15:14:44 -04003887 btrfs_free_block_groups(fs_info);
3888
Wang Shilongde348ee2014-04-09 19:23:22 +08003889 /*
3890 * we must make sure there is not any read request to
3891 * submit after we stopping all workers.
3892 */
3893 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003894 btrfs_stop_all_workers(fs_info);
3895
Josef Bacik47ab2a62014-09-18 11:20:02 -04003896 fs_info->open = 0;
Josef Bacik13e6c372013-05-30 16:55:44 -04003897 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003898
Josef Bacik13e6c372013-05-30 16:55:44 -04003899 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003900
Stefan Behrens21adbd52011-11-09 13:44:05 +01003901#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3902 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3903 btrfsic_unmount(root, fs_info->fs_devices);
3904#endif
3905
Chris Masondfe25022008-05-13 13:46:40 -04003906 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003907 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003908
Miao Xiee2d84522013-01-29 10:09:20 +00003909 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003910 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003911 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003912 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003913 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003914
David Woodhouse53b381b2013-01-29 18:40:14 -05003915 btrfs_free_stripe_hash_table(fs_info);
3916
Chris Masoncdfb0802015-04-06 18:17:00 -07003917 __btrfs_free_block_rsv(root->orphan_block_rsv);
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003918 root->orphan_block_rsv = NULL;
Filipe Manana04216822014-11-27 21:14:15 +00003919
3920 lock_chunks(root);
3921 while (!list_empty(&fs_info->pinned_chunks)) {
3922 struct extent_map *em;
3923
3924 em = list_first_entry(&fs_info->pinned_chunks,
3925 struct extent_map, list);
3926 list_del_init(&em->list);
3927 free_extent_map(em);
3928 }
3929 unlock_chunks(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003930}
3931
Chris Masonb9fab912012-05-06 07:23:47 -04003932int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3933 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003934{
Chris Mason1259ab72008-05-12 13:39:03 -04003935 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003936 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003937
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003938 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003939 if (!ret)
3940 return ret;
3941
3942 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003943 parent_transid, atomic);
3944 if (ret == -EAGAIN)
3945 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003946 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003947}
Chris Mason6702ed42007-08-07 16:15:09 -04003948
Chris Mason5f39d392007-10-15 16:14:19 -04003949void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3950{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003951 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003952 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003953 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003954
Josef Bacik06ea65a2013-09-19 16:07:01 -04003955#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3956 /*
3957 * This is a fast path so only do this check if we have sanity tests
3958 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3959 * outside of the sanity tests.
3960 */
3961 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3962 return;
3963#endif
3964 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003965 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003966 if (transid != root->fs_info->generation)
3967 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003968 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003969 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003970 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003971 if (!was_dirty)
3972 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3973 buf->len,
3974 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003975#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3976 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3977 btrfs_print_leaf(root, buf);
3978 ASSERT(0);
3979 }
3980#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05003981}
3982
Liu Bob53d3f52012-11-14 14:34:34 +00003983static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3984 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04003985{
Chris Mason188de642008-05-09 11:52:25 -04003986 /*
3987 * looks as though older kernels can get into trouble with
3988 * this code, they end up stuck in balance_dirty_pages forever
3989 */
Miao Xiee2d84522013-01-29 10:09:20 +00003990 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04003991
Jens Axboe6933c022009-03-17 09:36:37 +01003992 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04003993 return;
3994
Liu Bob53d3f52012-11-14 14:34:34 +00003995 if (flush_delayed)
3996 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08003997
Miao Xiee2d84522013-01-29 10:09:20 +00003998 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3999 BTRFS_DIRTY_METADATA_THRESH);
4000 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08004001 balance_dirty_pages_ratelimited(
4002 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08004003 }
Miao Xie16cdcec2011-04-22 18:12:22 +08004004}
4005
Liu Bob53d3f52012-11-14 14:34:34 +00004006void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08004007{
Liu Bob53d3f52012-11-14 14:34:34 +00004008 __btrfs_btree_balance_dirty(root, 1);
4009}
Miao Xie16cdcec2011-04-22 18:12:22 +08004010
Liu Bob53d3f52012-11-14 14:34:34 +00004011void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
4012{
4013 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04004014}
Chris Mason6b800532007-10-15 16:17:34 -04004015
Chris Masonca7a79a2008-05-12 12:59:19 -04004016int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04004017{
Chris Mason727011e2010-08-06 13:21:20 -04004018 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04004019 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04004020}
Chris Mason0da54682007-11-07 21:08:01 -05004021
David Sterbafcd1f0652012-03-06 00:06:18 +01004022static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08004023 int read_only)
4024{
David Sterbac9260932014-09-30 19:16:47 +02004025 struct btrfs_super_block *sb = fs_info->super_copy;
Qu Wenruo319e4d02015-12-15 09:14:36 +08004026 u64 nodesize = btrfs_super_nodesize(sb);
4027 u64 sectorsize = btrfs_super_sectorsize(sb);
David Sterbac9260932014-09-30 19:16:47 +02004028 int ret = 0;
4029
Qu Wenruo319e4d02015-12-15 09:14:36 +08004030 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
4031 printk(KERN_ERR "BTRFS: no valid FS found\n");
4032 ret = -EINVAL;
4033 }
4034 if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP)
4035 printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n",
4036 btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP);
David Sterba21e76262014-10-27 13:52:21 +01004037 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
4038 printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n",
4039 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004040 ret = -EINVAL;
4041 }
David Sterba21e76262014-10-27 13:52:21 +01004042 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
4043 printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n",
4044 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004045 ret = -EINVAL;
4046 }
David Sterba21e76262014-10-27 13:52:21 +01004047 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
4048 printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n",
4049 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004050 ret = -EINVAL;
4051 }
4052
David Sterba1104a882013-03-06 15:57:46 +01004053 /*
Qu Wenruo319e4d02015-12-15 09:14:36 +08004054 * Check sectorsize and nodesize first, other check will need it.
4055 * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here.
David Sterba1104a882013-03-06 15:57:46 +01004056 */
Qu Wenruo319e4d02015-12-15 09:14:36 +08004057 if (!is_power_of_2(sectorsize) || sectorsize < 4096 ||
4058 sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) {
4059 printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize);
David Sterba75d6ad32014-10-31 17:18:08 +01004060 ret = -EINVAL;
4061 }
Qu Wenruo319e4d02015-12-15 09:14:36 +08004062 /* Only PAGE SIZE is supported yet */
4063 if (sectorsize != PAGE_CACHE_SIZE) {
4064 printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n",
4065 sectorsize, PAGE_CACHE_SIZE);
4066 ret = -EINVAL;
4067 }
4068 if (!is_power_of_2(nodesize) || nodesize < sectorsize ||
4069 nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
4070 printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize);
4071 ret = -EINVAL;
4072 }
4073 if (nodesize != le32_to_cpu(sb->__unused_leafsize)) {
4074 printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n",
4075 le32_to_cpu(sb->__unused_leafsize),
4076 nodesize);
4077 ret = -EINVAL;
4078 }
4079
4080 /* Root alignment check */
4081 if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) {
4082 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
4083 btrfs_super_root(sb));
4084 ret = -EINVAL;
4085 }
4086 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) {
4087 printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n",
4088 btrfs_super_chunk_root(sb));
4089 ret = -EINVAL;
4090 }
4091 if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) {
4092 printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
4093 btrfs_super_log_root(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004094 ret = -EINVAL;
4095 }
4096
David Sterbac9260932014-09-30 19:16:47 +02004097 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
4098 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
4099 fs_info->fsid, sb->dev_item.fsid);
4100 ret = -EINVAL;
4101 }
4102
4103 /*
4104 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4105 * done later
4106 */
David Sterba21e76262014-10-27 13:52:21 +01004107 if (btrfs_super_num_devices(sb) > (1UL << 31))
David Sterbac9260932014-09-30 19:16:47 +02004108 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004109 btrfs_super_num_devices(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004110 if (btrfs_super_num_devices(sb) == 0) {
4111 printk(KERN_ERR "BTRFS: number of devices is 0\n");
4112 ret = -EINVAL;
4113 }
David Sterbac9260932014-09-30 19:16:47 +02004114
David Sterba21e76262014-10-27 13:52:21 +01004115 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
David Sterbac9260932014-09-30 19:16:47 +02004116 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
David Sterba21e76262014-10-27 13:52:21 +01004117 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
David Sterbac9260932014-09-30 19:16:47 +02004118 ret = -EINVAL;
4119 }
4120
4121 /*
David Sterbace7fca52014-10-31 18:42:05 +01004122 * Obvious sys_chunk_array corruptions, it must hold at least one key
4123 * and one chunk
4124 */
4125 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
4126 printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
4127 btrfs_super_sys_array_size(sb),
4128 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
4129 ret = -EINVAL;
4130 }
4131 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4132 + sizeof(struct btrfs_chunk)) {
Fabian Frederickd2207122015-02-14 13:10:22 +01004133 printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n",
David Sterbace7fca52014-10-31 18:42:05 +01004134 btrfs_super_sys_array_size(sb),
4135 sizeof(struct btrfs_disk_key)
4136 + sizeof(struct btrfs_chunk));
4137 ret = -EINVAL;
4138 }
4139
4140 /*
David Sterbac9260932014-09-30 19:16:47 +02004141 * The generation is a global counter, we'll trust it more than the others
4142 * but it's still possible that it's the one that's wrong.
4143 */
David Sterba21e76262014-10-27 13:52:21 +01004144 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
David Sterbac9260932014-09-30 19:16:47 +02004145 printk(KERN_WARNING
4146 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004147 btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb));
4148 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4149 && btrfs_super_cache_generation(sb) != (u64)-1)
David Sterbac9260932014-09-30 19:16:47 +02004150 printk(KERN_WARNING
4151 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004152 btrfs_super_generation(sb), btrfs_super_cache_generation(sb));
David Sterbac9260932014-09-30 19:16:47 +02004153
4154 return ret;
liuboacce9522011-01-06 19:30:25 +08004155}
4156
Eric Sandeen48a3b632013-04-25 20:41:01 +00004157static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004158{
liuboacce9522011-01-06 19:30:25 +08004159 mutex_lock(&root->fs_info->cleaner_mutex);
4160 btrfs_run_delayed_iputs(root);
4161 mutex_unlock(&root->fs_info->cleaner_mutex);
4162
4163 down_write(&root->fs_info->cleanup_work_sem);
4164 up_write(&root->fs_info->cleanup_work_sem);
4165
4166 /* cleanup FS via transaction */
4167 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08004168}
4169
Jeff Mahoney143bede2012-03-01 14:56:26 +01004170static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004171{
liuboacce9522011-01-06 19:30:25 +08004172 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08004173
Miao Xie199c2a92013-05-15 07:48:23 +00004174 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05004175 /*
4176 * This will just short circuit the ordered completion stuff which will
4177 * make sure the ordered extent gets properly cleaned up.
4178 */
Miao Xie199c2a92013-05-15 07:48:23 +00004179 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05004180 root_extent_list)
4181 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00004182 spin_unlock(&root->ordered_extent_lock);
4183}
4184
4185static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4186{
4187 struct btrfs_root *root;
4188 struct list_head splice;
4189
4190 INIT_LIST_HEAD(&splice);
4191
4192 spin_lock(&fs_info->ordered_root_lock);
4193 list_splice_init(&fs_info->ordered_roots, &splice);
4194 while (!list_empty(&splice)) {
4195 root = list_first_entry(&splice, struct btrfs_root,
4196 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04004197 list_move_tail(&root->ordered_root,
4198 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00004199
Liu Bo2a85d9c2014-02-10 17:07:16 +08004200 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004201 btrfs_destroy_ordered_extents(root);
4202
Liu Bo2a85d9c2014-02-10 17:07:16 +08004203 cond_resched();
4204 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004205 }
4206 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08004207}
4208
Stefan Behrens35a36212013-08-14 18:12:25 +02004209static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4210 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004211{
4212 struct rb_node *node;
4213 struct btrfs_delayed_ref_root *delayed_refs;
4214 struct btrfs_delayed_ref_node *ref;
4215 int ret = 0;
4216
4217 delayed_refs = &trans->delayed_refs;
4218
4219 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004220 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04004221 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05004222 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08004223 return ret;
4224 }
4225
Josef Bacikd7df2c72014-01-23 09:21:38 -05004226 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4227 struct btrfs_delayed_ref_head *head;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004228 struct btrfs_delayed_ref_node *tmp;
Josef Bacike78417d2013-06-03 16:42:36 -04004229 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08004230
Josef Bacikd7df2c72014-01-23 09:21:38 -05004231 head = rb_entry(node, struct btrfs_delayed_ref_head,
4232 href_node);
4233 if (!mutex_trylock(&head->mutex)) {
4234 atomic_inc(&head->node.refs);
4235 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08004236
Josef Bacikd7df2c72014-01-23 09:21:38 -05004237 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04004238 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004239 btrfs_put_delayed_ref(&head->node);
4240 spin_lock(&delayed_refs->lock);
4241 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04004242 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05004243 spin_lock(&head->lock);
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004244 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4245 list) {
Josef Bacikd7df2c72014-01-23 09:21:38 -05004246 ref->in_tree = 0;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004247 list_del(&ref->list);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004248 atomic_dec(&delayed_refs->num_entries);
4249 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004250 }
4251 if (head->must_insert_reserved)
4252 pin_bytes = true;
4253 btrfs_free_delayed_extent_op(head->extent_op);
4254 delayed_refs->num_heads--;
4255 if (head->processing == 0)
4256 delayed_refs->num_heads_ready--;
4257 atomic_dec(&delayed_refs->num_entries);
4258 head->node.in_tree = 0;
4259 rb_erase(&head->href_node, &delayed_refs->href_root);
4260 spin_unlock(&head->lock);
4261 spin_unlock(&delayed_refs->lock);
4262 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08004263
Josef Bacikd7df2c72014-01-23 09:21:38 -05004264 if (pin_bytes)
4265 btrfs_pin_extent(root, head->node.bytenr,
4266 head->node.num_bytes, 1);
4267 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08004268 cond_resched();
4269 spin_lock(&delayed_refs->lock);
4270 }
4271
4272 spin_unlock(&delayed_refs->lock);
4273
4274 return ret;
4275}
4276
Jeff Mahoney143bede2012-03-01 14:56:26 +01004277static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004278{
4279 struct btrfs_inode *btrfs_inode;
4280 struct list_head splice;
4281
4282 INIT_LIST_HEAD(&splice);
4283
Miao Xieeb73c1b2013-05-15 07:48:22 +00004284 spin_lock(&root->delalloc_lock);
4285 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08004286
4287 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00004288 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4289 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08004290
4291 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00004292 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4293 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00004294 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004295
4296 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00004297
Miao Xieeb73c1b2013-05-15 07:48:22 +00004298 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004299 }
4300
Miao Xieeb73c1b2013-05-15 07:48:22 +00004301 spin_unlock(&root->delalloc_lock);
4302}
4303
4304static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4305{
4306 struct btrfs_root *root;
4307 struct list_head splice;
4308
4309 INIT_LIST_HEAD(&splice);
4310
4311 spin_lock(&fs_info->delalloc_root_lock);
4312 list_splice_init(&fs_info->delalloc_roots, &splice);
4313 while (!list_empty(&splice)) {
4314 root = list_first_entry(&splice, struct btrfs_root,
4315 delalloc_root);
4316 list_del_init(&root->delalloc_root);
4317 root = btrfs_grab_fs_root(root);
4318 BUG_ON(!root);
4319 spin_unlock(&fs_info->delalloc_root_lock);
4320
4321 btrfs_destroy_delalloc_inodes(root);
4322 btrfs_put_fs_root(root);
4323
4324 spin_lock(&fs_info->delalloc_root_lock);
4325 }
4326 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004327}
4328
4329static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4330 struct extent_io_tree *dirty_pages,
4331 int mark)
4332{
4333 int ret;
liuboacce9522011-01-06 19:30:25 +08004334 struct extent_buffer *eb;
4335 u64 start = 0;
4336 u64 end;
liuboacce9522011-01-06 19:30:25 +08004337
4338 while (1) {
4339 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004340 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004341 if (ret)
4342 break;
4343
4344 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4345 while (start <= end) {
Daniel Dressler01d58472014-11-21 17:15:07 +09004346 eb = btrfs_find_tree_block(root->fs_info, start);
David Sterba707e8a02014-06-04 19:22:26 +02004347 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004348 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004349 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004350 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004351
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004352 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4353 &eb->bflags))
4354 clear_extent_buffer_dirty(eb);
4355 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004356 }
4357 }
4358
4359 return ret;
4360}
4361
4362static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4363 struct extent_io_tree *pinned_extents)
4364{
4365 struct extent_io_tree *unpin;
4366 u64 start;
4367 u64 end;
4368 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004369 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004370
4371 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004372again:
liuboacce9522011-01-06 19:30:25 +08004373 while (1) {
4374 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004375 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004376 if (ret)
4377 break;
4378
liuboacce9522011-01-06 19:30:25 +08004379 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4380 btrfs_error_unpin_extent_range(root, start, end);
4381 cond_resched();
4382 }
4383
Liu Boed0eaa12012-06-14 02:23:21 -06004384 if (loop) {
4385 if (unpin == &root->fs_info->freed_extents[0])
4386 unpin = &root->fs_info->freed_extents[1];
4387 else
4388 unpin = &root->fs_info->freed_extents[0];
4389 loop = false;
4390 goto again;
4391 }
4392
liuboacce9522011-01-06 19:30:25 +08004393 return 0;
4394}
4395
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004396void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4397 struct btrfs_root *root)
4398{
4399 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004400
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004401 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004402 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004403
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004404 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004405 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004406
Miao Xie67cde342012-06-14 02:23:22 -06004407 btrfs_destroy_delayed_inodes(root);
4408 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004409
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004410 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4411 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004412 btrfs_destroy_pinned_extent(root,
4413 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004414
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004415 cur_trans->state =TRANS_STATE_COMPLETED;
4416 wake_up(&cur_trans->commit_wait);
4417
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004418 /*
4419 memset(cur_trans, 0, sizeof(*cur_trans));
4420 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4421 */
4422}
4423
Eric Sandeen48a3b632013-04-25 20:41:01 +00004424static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004425{
4426 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004427
liuboacce9522011-01-06 19:30:25 +08004428 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4429
Josef Bacika4abeea2011-04-11 17:25:13 -04004430 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004431 while (!list_empty(&root->fs_info->trans_list)) {
4432 t = list_first_entry(&root->fs_info->trans_list,
4433 struct btrfs_transaction, list);
4434 if (t->state >= TRANS_STATE_COMMIT_START) {
4435 atomic_inc(&t->use_count);
4436 spin_unlock(&root->fs_info->trans_lock);
4437 btrfs_wait_for_commit(root, t->transid);
4438 btrfs_put_transaction(t);
4439 spin_lock(&root->fs_info->trans_lock);
4440 continue;
4441 }
4442 if (t == root->fs_info->running_transaction) {
4443 t->state = TRANS_STATE_COMMIT_DOING;
4444 spin_unlock(&root->fs_info->trans_lock);
4445 /*
4446 * We wait for 0 num_writers since we don't hold a trans
4447 * handle open currently for this transaction.
4448 */
4449 wait_event(t->writer_wait,
4450 atomic_read(&t->num_writers) == 0);
4451 } else {
4452 spin_unlock(&root->fs_info->trans_lock);
4453 }
4454 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004455
Josef Bacik724e2312013-09-30 11:36:38 -04004456 spin_lock(&root->fs_info->trans_lock);
4457 if (t == root->fs_info->running_transaction)
4458 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004459 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004460 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004461
Josef Bacik724e2312013-09-30 11:36:38 -04004462 btrfs_put_transaction(t);
4463 trace_btrfs_transaction_commit(root);
4464 spin_lock(&root->fs_info->trans_lock);
4465 }
4466 spin_unlock(&root->fs_info->trans_lock);
4467 btrfs_destroy_all_ordered_extents(root->fs_info);
4468 btrfs_destroy_delayed_inodes(root);
4469 btrfs_assert_delayed_root_empty(root);
4470 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4471 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004472 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004473
4474 return 0;
4475}
4476
David Sterbae8c9f182015-01-02 18:23:10 +01004477static const struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004478 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004479 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004480 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004481 /* note we're sharing with inode.c for the merge bio hook */
4482 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004483};