blob: 64f02c3d0dd0f2d1bdba9a891612469fcdf95b22 [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
David Sterbae8c9f182015-01-02 18:23:10 +010058static const struct extent_io_ops btree_extent_io_ops;
Chris Mason8b712842008-06-11 16:50:36 -040059static void end_workqueue_fn(struct btrfs_work *work);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -040060static void free_fs_root(struct btrfs_root *root);
David Sterbafcd1f0652012-03-06 00:06:18 +010061static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +080062 int read_only);
Jeff Mahoney143bede2012-03-01 14:56:26 +010063static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080064static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
65 struct btrfs_root *root);
Jeff Mahoney143bede2012-03-01 14:56:26 +010066static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
liuboacce9522011-01-06 19:30:25 +080067static int btrfs_destroy_marked_extents(struct btrfs_root *root,
68 struct extent_io_tree *dirty_pages,
69 int mark);
70static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
71 struct extent_io_tree *pinned_extents);
Eric Sandeen48a3b632013-04-25 20:41:01 +000072static int btrfs_cleanup_transaction(struct btrfs_root *root);
73static void btrfs_error_commit_super(struct btrfs_root *root);
Chris Masonce9adaa2008-04-09 16:28:12 -040074
Chris Masond352ac62008-09-29 15:18:18 -040075/*
David Sterba97eb6b62014-07-30 00:55:42 +020076 * btrfs_end_io_wq structs are used to do processing in task context when an IO
77 * is complete. This is used during reads to verify checksums, and it is used
Chris Masond352ac62008-09-29 15:18:18 -040078 * by writes to insert metadata for new file extents after IO is complete.
79 */
David Sterba97eb6b62014-07-30 00:55:42 +020080struct btrfs_end_io_wq {
Chris Masonce9adaa2008-04-09 16:28:12 -040081 struct bio *bio;
82 bio_end_io_t *end_io;
83 void *private;
84 struct btrfs_fs_info *info;
85 int error;
David Sterbabfebd8b2014-07-30 00:25:45 +020086 enum btrfs_wq_endio_type metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -040087 struct list_head list;
Chris Mason8b712842008-06-11 16:50:36 -040088 struct btrfs_work work;
Chris Masonce9adaa2008-04-09 16:28:12 -040089};
Chris Mason0da54682007-11-07 21:08:01 -050090
David Sterba97eb6b62014-07-30 00:55:42 +020091static struct kmem_cache *btrfs_end_io_wq_cache;
92
93int __init btrfs_end_io_wq_init(void)
94{
95 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
96 sizeof(struct btrfs_end_io_wq),
97 0,
98 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
99 NULL);
100 if (!btrfs_end_io_wq_cache)
101 return -ENOMEM;
102 return 0;
103}
104
105void btrfs_end_io_wq_exit(void)
106{
107 if (btrfs_end_io_wq_cache)
108 kmem_cache_destroy(btrfs_end_io_wq_cache);
109}
110
Chris Masond352ac62008-09-29 15:18:18 -0400111/*
112 * async submit bios are used to offload expensive checksumming
113 * onto the worker threads. They checksum file and metadata bios
114 * just before they are sent down the IO stack.
115 */
Chris Mason44b8bd72008-04-16 11:14:51 -0400116struct async_submit_bio {
117 struct inode *inode;
118 struct bio *bio;
119 struct list_head list;
Chris Mason4a69a412008-11-06 22:03:00 -0500120 extent_submit_bio_hook_t *submit_bio_start;
121 extent_submit_bio_hook_t *submit_bio_done;
Chris Mason44b8bd72008-04-16 11:14:51 -0400122 int rw;
123 int mirror_num;
Chris Masonc8b97812008-10-29 14:49:59 -0400124 unsigned long bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400125 /*
126 * bio_offset is optional, can be used if the pages in the bio
127 * can't tell us where in the file the bio should go
128 */
129 u64 bio_offset;
Chris Mason8b712842008-06-11 16:50:36 -0400130 struct btrfs_work work;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100131 int error;
Chris Mason44b8bd72008-04-16 11:14:51 -0400132};
133
Chris Mason85d4e462011-07-26 16:11:19 -0400134/*
135 * Lockdep class keys for extent_buffer->lock's in this root. For a given
136 * eb, the lockdep key is determined by the btrfs_root it belongs to and
137 * the level the eb occupies in the tree.
Chris Mason4008c042009-02-12 14:09:45 -0500138 *
Chris Mason85d4e462011-07-26 16:11:19 -0400139 * Different roots are used for different purposes and may nest inside each
140 * other and they require separate keysets. As lockdep keys should be
141 * static, assign keysets according to the purpose of the root as indicated
142 * by btrfs_root->objectid. This ensures that all special purpose roots
143 * have separate keysets.
Chris Mason4008c042009-02-12 14:09:45 -0500144 *
Chris Mason85d4e462011-07-26 16:11:19 -0400145 * Lock-nesting across peer nodes is always done with the immediate parent
146 * node locked thus preventing deadlock. As lockdep doesn't know this, use
147 * subclass to avoid triggering lockdep warning in such cases.
Chris Mason4008c042009-02-12 14:09:45 -0500148 *
Chris Mason85d4e462011-07-26 16:11:19 -0400149 * The key is set by the readpage_end_io_hook after the buffer has passed
150 * csum validation but before the pages are unlocked. It is also set by
151 * btrfs_init_new_buffer on freshly allocated blocks.
152 *
153 * We also add a check to make sure the highest level of the tree is the
154 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
155 * needs update as well.
Chris Mason4008c042009-02-12 14:09:45 -0500156 */
157#ifdef CONFIG_DEBUG_LOCK_ALLOC
158# if BTRFS_MAX_LEVEL != 8
159# error
160# endif
Chris Mason85d4e462011-07-26 16:11:19 -0400161
162static struct btrfs_lockdep_keyset {
163 u64 id; /* root objectid */
164 const char *name_stem; /* lock name stem */
165 char names[BTRFS_MAX_LEVEL + 1][20];
166 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
167} btrfs_lockdep_keysets[] = {
168 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
169 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
170 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
171 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
172 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
173 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
David Sterba60b62972013-04-30 17:29:29 +0000174 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
Chris Mason85d4e462011-07-26 16:11:19 -0400175 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
176 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
177 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
David Sterba13fd8da2013-09-03 18:28:57 +0200178 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
Chris Mason85d4e462011-07-26 16:11:19 -0400179 { .id = 0, .name_stem = "tree" },
Chris Mason4008c042009-02-12 14:09:45 -0500180};
Chris Mason85d4e462011-07-26 16:11:19 -0400181
182void __init btrfs_init_lockdep(void)
183{
184 int i, j;
185
186 /* initialize lockdep class names */
187 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
188 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
189
190 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
191 snprintf(ks->names[j], sizeof(ks->names[j]),
192 "btrfs-%s-%02d", ks->name_stem, j);
193 }
194}
195
196void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
197 int level)
198{
199 struct btrfs_lockdep_keyset *ks;
200
201 BUG_ON(level >= ARRAY_SIZE(ks->keys));
202
203 /* find the matching keyset, id 0 is the default entry */
204 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
205 if (ks->id == objectid)
206 break;
207
208 lockdep_set_class_and_name(&eb->lock,
209 &ks->keys[level], ks->names[level]);
210}
211
Chris Mason4008c042009-02-12 14:09:45 -0500212#endif
213
Chris Masond352ac62008-09-29 15:18:18 -0400214/*
215 * extents on the btree inode are pretty simple, there's one extent
216 * that covers the entire device
217 */
Christoph Hellwigb2950862008-12-02 09:54:17 -0500218static struct extent_map *btree_get_extent(struct inode *inode,
David Sterba306e16c2011-04-19 14:29:38 +0200219 struct page *page, size_t pg_offset, u64 start, u64 len,
Christoph Hellwigb2950862008-12-02 09:54:17 -0500220 int create)
Chris Mason5f39d392007-10-15 16:14:19 -0400221{
222 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
223 struct extent_map *em;
224 int ret;
Chris Masond98237b2007-03-28 13:57:48 -0400225
Chris Mason890871b2009-09-02 16:24:52 -0400226 read_lock(&em_tree->lock);
Chris Masond1310b22008-01-24 16:13:08 -0500227 em = lookup_extent_mapping(em_tree, start, len);
Chris Masona061fc82008-05-07 11:43:44 -0400228 if (em) {
229 em->bdev =
230 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Mason890871b2009-09-02 16:24:52 -0400231 read_unlock(&em_tree->lock);
Chris Mason5f39d392007-10-15 16:14:19 -0400232 goto out;
Chris Masona061fc82008-05-07 11:43:44 -0400233 }
Chris Mason890871b2009-09-02 16:24:52 -0400234 read_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400235
David Sterba172ddd62011-04-21 00:48:27 +0200236 em = alloc_extent_map();
Chris Mason5f39d392007-10-15 16:14:19 -0400237 if (!em) {
238 em = ERR_PTR(-ENOMEM);
239 goto out;
240 }
241 em->start = 0;
Chris Mason0afbaf82008-04-18 14:17:20 -0400242 em->len = (u64)-1;
Chris Masonc8b97812008-10-29 14:49:59 -0400243 em->block_len = (u64)-1;
Chris Mason5f39d392007-10-15 16:14:19 -0400244 em->block_start = 0;
Chris Masona061fc82008-05-07 11:43:44 -0400245 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
Chris Masond1310b22008-01-24 16:13:08 -0500246
Chris Mason890871b2009-09-02 16:24:52 -0400247 write_lock(&em_tree->lock);
Josef Bacik09a2a8f92013-04-05 16:51:15 -0400248 ret = add_extent_mapping(em_tree, em, 0);
Chris Mason5f39d392007-10-15 16:14:19 -0400249 if (ret == -EEXIST) {
250 free_extent_map(em);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400251 em = lookup_extent_mapping(em_tree, start, len);
Tsutomu Itohb4f359a2012-09-13 03:32:54 -0600252 if (!em)
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600253 em = ERR_PTR(-EIO);
Chris Mason5f39d392007-10-15 16:14:19 -0400254 } else if (ret) {
Chris Mason7b13b7b2008-04-18 10:29:50 -0400255 free_extent_map(em);
Tsutomu Itoh0433f202012-09-13 03:32:32 -0600256 em = ERR_PTR(ret);
Chris Mason5f39d392007-10-15 16:14:19 -0400257 }
Chris Mason890871b2009-09-02 16:24:52 -0400258 write_unlock(&em_tree->lock);
Chris Mason7b13b7b2008-04-18 10:29:50 -0400259
Chris Mason5f39d392007-10-15 16:14:19 -0400260out:
261 return em;
262}
263
Liu Bob0496682013-03-14 14:57:45 +0000264u32 btrfs_csum_data(char *data, u32 seed, size_t len)
Chris Mason19c00dd2007-10-15 16:19:22 -0400265{
Filipe David Borba Manana0b947af2014-01-29 21:06:04 +0000266 return btrfs_crc32c(seed, data, len);
Chris Mason19c00dd2007-10-15 16:19:22 -0400267}
268
269void btrfs_csum_final(u32 crc, char *result)
270{
David Sterba7e75bf32011-03-18 22:56:43 +0000271 put_unaligned_le32(~crc, result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400272}
273
Chris Masond352ac62008-09-29 15:18:18 -0400274/*
275 * compute the csum for a btree block, and either verify it or write it
276 * into the csum field of the block.
277 */
Daniel Dressler01d58472014-11-21 17:15:07 +0900278static int csum_tree_block(struct btrfs_fs_info *fs_info,
279 struct extent_buffer *buf,
Chris Mason19c00dd2007-10-15 16:19:22 -0400280 int verify)
281{
Daniel Dressler01d58472014-11-21 17:15:07 +0900282 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
Josef Bacik607d4322008-12-02 07:17:45 -0500283 char *result = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -0400284 unsigned long len;
285 unsigned long cur_len;
286 unsigned long offset = BTRFS_CSUM_SIZE;
Chris Mason19c00dd2007-10-15 16:19:22 -0400287 char *kaddr;
288 unsigned long map_start;
289 unsigned long map_len;
290 int err;
291 u32 crc = ~(u32)0;
Josef Bacik607d4322008-12-02 07:17:45 -0500292 unsigned long inline_result;
Chris Mason19c00dd2007-10-15 16:19:22 -0400293
294 len = buf->len - offset;
Chris Masond3977122009-01-05 21:25:51 -0500295 while (len > 0) {
Chris Mason19c00dd2007-10-15 16:19:22 -0400296 err = map_private_extent_buffer(buf, offset, 32,
Chris Masona6591712011-07-19 12:04:14 -0400297 &kaddr, &map_start, &map_len);
Chris Masond3977122009-01-05 21:25:51 -0500298 if (err)
Chris Mason19c00dd2007-10-15 16:19:22 -0400299 return 1;
Chris Mason19c00dd2007-10-15 16:19:22 -0400300 cur_len = min(len, map_len - (offset - map_start));
Liu Bob0496682013-03-14 14:57:45 +0000301 crc = btrfs_csum_data(kaddr + offset - map_start,
Chris Mason19c00dd2007-10-15 16:19:22 -0400302 crc, cur_len);
303 len -= cur_len;
304 offset += cur_len;
Chris Mason19c00dd2007-10-15 16:19:22 -0400305 }
Josef Bacik607d4322008-12-02 07:17:45 -0500306 if (csum_size > sizeof(inline_result)) {
David Sterba31e818f2015-02-20 18:00:26 +0100307 result = kzalloc(csum_size, GFP_NOFS);
Josef Bacik607d4322008-12-02 07:17:45 -0500308 if (!result)
309 return 1;
310 } else {
311 result = (char *)&inline_result;
312 }
313
Chris Mason19c00dd2007-10-15 16:19:22 -0400314 btrfs_csum_final(crc, result);
315
316 if (verify) {
Josef Bacik607d4322008-12-02 07:17:45 -0500317 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
Chris Masone4204de2008-01-08 15:46:27 -0500318 u32 val;
319 u32 found = 0;
Josef Bacik607d4322008-12-02 07:17:45 -0500320 memcpy(&found, result, csum_size);
Chris Masone4204de2008-01-08 15:46:27 -0500321
Josef Bacik607d4322008-12-02 07:17:45 -0500322 read_extent_buffer(buf, &val, 0, csum_size);
David Sterba94647322015-10-08 11:01:36 +0200323 btrfs_warn_rl(fs_info,
324 "%s checksum verify failed on %llu wanted %X found %X "
325 "level %d",
Daniel Dressler01d58472014-11-21 17:15:07 +0900326 fs_info->sb->s_id, buf->start,
Frank Holtonefe120a2013-12-20 11:37:06 -0500327 val, found, btrfs_header_level(buf));
Josef Bacik607d4322008-12-02 07:17:45 -0500328 if (result != (char *)&inline_result)
329 kfree(result);
Chris Mason19c00dd2007-10-15 16:19:22 -0400330 return 1;
331 }
332 } else {
Josef Bacik607d4322008-12-02 07:17:45 -0500333 write_extent_buffer(buf, result, 0, csum_size);
Chris Mason19c00dd2007-10-15 16:19:22 -0400334 }
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 0;
338}
339
Chris Masond352ac62008-09-29 15:18:18 -0400340/*
341 * we can't consider a given block up to date unless the transid of the
342 * block matches the transid in the parent node's pointer. This is how we
343 * detect blocks that either didn't get written at all or got written
344 * in the wrong place.
345 */
Chris Mason1259ab72008-05-12 13:39:03 -0400346static int verify_parent_transid(struct extent_io_tree *io_tree,
Chris Masonb9fab912012-05-06 07:23:47 -0400347 struct extent_buffer *eb, u64 parent_transid,
348 int atomic)
Chris Mason1259ab72008-05-12 13:39:03 -0400349{
Josef Bacik2ac55d42010-02-03 19:33:23 +0000350 struct extent_state *cached_state = NULL;
Chris Mason1259ab72008-05-12 13:39:03 -0400351 int ret;
David Sterba2755a0d2014-07-31 00:43:18 +0200352 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
Chris Mason1259ab72008-05-12 13:39:03 -0400353
354 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
355 return 0;
356
Chris Masonb9fab912012-05-06 07:23:47 -0400357 if (atomic)
358 return -EAGAIN;
359
Josef Bacika26e8c92014-03-28 17:07:27 -0400360 if (need_lock) {
361 btrfs_tree_read_lock(eb);
362 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
363 }
364
Josef Bacik2ac55d42010-02-03 19:33:23 +0000365 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
David Sterbaff13db42015-12-03 14:30:40 +0100366 &cached_state);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400367 if (extent_buffer_uptodate(eb) &&
Chris Mason1259ab72008-05-12 13:39:03 -0400368 btrfs_header_generation(eb) == parent_transid) {
369 ret = 0;
370 goto out;
371 }
David Sterba94647322015-10-08 11:01:36 +0200372 btrfs_err_rl(eb->fs_info,
373 "parent transid verify failed on %llu wanted %llu found %llu",
374 eb->start,
Wang Shilong29549ae2014-07-04 17:59:06 +0800375 parent_transid, btrfs_header_generation(eb));
Chris Mason1259ab72008-05-12 13:39:03 -0400376 ret = 1;
Josef Bacika26e8c92014-03-28 17:07:27 -0400377
378 /*
379 * Things reading via commit roots that don't have normal protection,
380 * like send, can have a really old block in cache that may point at a
381 * block that has been free'd and re-allocated. So don't clear uptodate
382 * if we find an eb that is under IO (dirty/writeback) because we could
383 * end up reading in the stale data and then writing it back out and
384 * making everybody very sad.
385 */
386 if (!extent_buffer_under_io(eb))
387 clear_extent_buffer_uptodate(eb);
Chris Mason33958dc2008-07-30 10:29:12 -0400388out:
Josef Bacik2ac55d42010-02-03 19:33:23 +0000389 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
390 &cached_state, GFP_NOFS);
Josef Bacik472b9092014-06-25 13:45:41 -0700391 if (need_lock)
392 btrfs_tree_read_unlock_blocking(eb);
Chris Mason1259ab72008-05-12 13:39:03 -0400393 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -0400394}
395
Chris Masond352ac62008-09-29 15:18:18 -0400396/*
David Sterba1104a882013-03-06 15:57:46 +0100397 * Return 0 if the superblock checksum type matches the checksum value of that
398 * algorithm. Pass the raw disk superblock data.
399 */
400static int btrfs_check_super_csum(char *raw_disk_sb)
401{
402 struct btrfs_super_block *disk_sb =
403 (struct btrfs_super_block *)raw_disk_sb;
404 u16 csum_type = btrfs_super_csum_type(disk_sb);
405 int ret = 0;
406
407 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
408 u32 crc = ~(u32)0;
409 const int csum_size = sizeof(crc);
410 char result[csum_size];
411
412 /*
413 * The super_block structure does not span the whole
414 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
415 * is filled with zeros and is included in the checkum.
416 */
417 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
418 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
419 btrfs_csum_final(crc, result);
420
421 if (memcmp(raw_disk_sb, result, csum_size))
422 ret = 1;
423 }
424
425 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
Frank Holtonefe120a2013-12-20 11:37:06 -0500426 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
David Sterba1104a882013-03-06 15:57:46 +0100427 csum_type);
428 ret = 1;
429 }
430
431 return ret;
432}
433
434/*
Chris Masond352ac62008-09-29 15:18:18 -0400435 * helper to read a given tree block, doing retries as required when
436 * the checksums don't match and we have alternate mirrors to try.
437 */
Chris Masonf1885912008-04-09 16:28:12 -0400438static int btree_read_extent_buffer_pages(struct btrfs_root *root,
439 struct extent_buffer *eb,
Chris Masonca7a79a2008-05-12 12:59:19 -0400440 u64 start, u64 parent_transid)
Chris Masonf1885912008-04-09 16:28:12 -0400441{
442 struct extent_io_tree *io_tree;
Josef Bacikea466792012-03-26 21:57:36 -0400443 int failed = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400444 int ret;
445 int num_copies = 0;
446 int mirror_num = 0;
Josef Bacikea466792012-03-26 21:57:36 -0400447 int failed_mirror = 0;
Chris Masonf1885912008-04-09 16:28:12 -0400448
Josef Bacika826d6d2011-03-16 13:42:43 -0400449 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
Chris Masonf1885912008-04-09 16:28:12 -0400450 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
451 while (1) {
Arne Jansenbb82ab82011-06-10 14:06:53 +0200452 ret = read_extent_buffer_pages(io_tree, eb, start,
453 WAIT_COMPLETE,
Chris Masonf1885912008-04-09 16:28:12 -0400454 btree_get_extent, mirror_num);
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600455 if (!ret) {
456 if (!verify_parent_transid(io_tree, eb,
Chris Masonb9fab912012-05-06 07:23:47 -0400457 parent_transid, 0))
Stefan Behrens256dd1b2012-08-10 08:58:21 -0600458 break;
459 else
460 ret = -EIO;
461 }
Chris Masond3977122009-01-05 21:25:51 -0500462
Josef Bacika826d6d2011-03-16 13:42:43 -0400463 /*
464 * This buffer's crc is fine, but its contents are corrupted, so
465 * there is no reason to read the other copies, they won't be
466 * any less wrong.
467 */
468 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
Josef Bacikea466792012-03-26 21:57:36 -0400469 break;
470
Stefan Behrens5d964052012-11-05 14:59:07 +0100471 num_copies = btrfs_num_copies(root->fs_info,
Chris Masonf1885912008-04-09 16:28:12 -0400472 eb->start, eb->len);
Chris Mason42352982008-04-28 16:40:52 -0400473 if (num_copies == 1)
Josef Bacikea466792012-03-26 21:57:36 -0400474 break;
Chris Mason42352982008-04-28 16:40:52 -0400475
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400476 if (!failed_mirror) {
477 failed = 1;
478 failed_mirror = eb->read_mirror;
479 }
480
Chris Masonf1885912008-04-09 16:28:12 -0400481 mirror_num++;
Josef Bacikea466792012-03-26 21:57:36 -0400482 if (mirror_num == failed_mirror)
483 mirror_num++;
484
Chris Mason42352982008-04-28 16:40:52 -0400485 if (mirror_num > num_copies)
Josef Bacikea466792012-03-26 21:57:36 -0400486 break;
Chris Masonf1885912008-04-09 16:28:12 -0400487 }
Josef Bacikea466792012-03-26 21:57:36 -0400488
Stefan Behrensc0901582012-07-10 07:30:17 -0600489 if (failed && !ret && failed_mirror)
Josef Bacikea466792012-03-26 21:57:36 -0400490 repair_eb_io_failure(root, eb, failed_mirror);
491
492 return ret;
Chris Masonf1885912008-04-09 16:28:12 -0400493}
Chris Mason19c00dd2007-10-15 16:19:22 -0400494
Chris Masond352ac62008-09-29 15:18:18 -0400495/*
Chris Masond3977122009-01-05 21:25:51 -0500496 * checksum a dirty tree block before IO. This has extra checks to make sure
497 * we only fill in the checksum field in the first page of a multi-page block
Chris Masond352ac62008-09-29 15:18:18 -0400498 */
Chris Masond3977122009-01-05 21:25:51 -0500499
Daniel Dressler01d58472014-11-21 17:15:07 +0900500static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
Chris Mason19c00dd2007-10-15 16:19:22 -0400501{
Miao Xie4eee4fa2012-12-21 09:17:45 +0000502 u64 start = page_offset(page);
Chris Mason19c00dd2007-10-15 16:19:22 -0400503 u64 found_start;
Chris Mason19c00dd2007-10-15 16:19:22 -0400504 struct extent_buffer *eb;
Chris Masonf1885912008-04-09 16:28:12 -0400505
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500506 eb = (struct extent_buffer *)page->private;
507 if (page != eb->pages[0])
508 return 0;
Chris Mason19c00dd2007-10-15 16:19:22 -0400509 found_start = btrfs_header_bytenr(eb);
Dulshani Gunawardhanafae7f212013-10-31 10:30:08 +0530510 if (WARN_ON(found_start != start || !PageUptodate(page)))
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500511 return 0;
Daniel Dressler01d58472014-11-21 17:15:07 +0900512 csum_tree_block(fs_info, eb, 0);
Chris Mason19c00dd2007-10-15 16:19:22 -0400513 return 0;
514}
515
Daniel Dressler01d58472014-11-21 17:15:07 +0900516static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
Yan Zheng2b820322008-11-17 21:11:30 -0500517 struct extent_buffer *eb)
518{
Daniel Dressler01d58472014-11-21 17:15:07 +0900519 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
Yan Zheng2b820322008-11-17 21:11:30 -0500520 u8 fsid[BTRFS_UUID_SIZE];
521 int ret = 1;
522
Ross Kirk0a4e5582013-09-24 10:12:38 +0100523 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -0500524 while (fs_devices) {
525 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
526 ret = 0;
527 break;
528 }
529 fs_devices = fs_devices->seed;
530 }
531 return ret;
532}
533
Josef Bacika826d6d2011-03-16 13:42:43 -0400534#define CORRUPT(reason, eb, root, slot) \
Frank Holtonefe120a2013-12-20 11:37:06 -0500535 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
536 "root=%llu, slot=%d", reason, \
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +0200537 btrfs_header_bytenr(eb), root->objectid, slot)
Josef Bacika826d6d2011-03-16 13:42:43 -0400538
539static noinline int check_leaf(struct btrfs_root *root,
540 struct extent_buffer *leaf)
541{
542 struct btrfs_key key;
543 struct btrfs_key leaf_key;
544 u32 nritems = btrfs_header_nritems(leaf);
545 int slot;
546
547 if (nritems == 0)
548 return 0;
549
550 /* Check the 0 item */
551 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
552 BTRFS_LEAF_DATA_SIZE(root)) {
553 CORRUPT("invalid item offset size pair", leaf, root, 0);
554 return -EIO;
555 }
556
557 /*
558 * Check to make sure each items keys are in the correct order and their
559 * offsets make sense. We only have to loop through nritems-1 because
560 * we check the current slot against the next slot, which verifies the
561 * next slot's offset+size makes sense and that the current's slot
562 * offset is correct.
563 */
564 for (slot = 0; slot < nritems - 1; slot++) {
565 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
566 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
567
568 /* Make sure the keys are in the right order */
569 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
570 CORRUPT("bad key order", leaf, root, slot);
571 return -EIO;
572 }
573
574 /*
575 * Make sure the offset and ends are right, remember that the
576 * item data starts at the end of the leaf and grows towards the
577 * front.
578 */
579 if (btrfs_item_offset_nr(leaf, slot) !=
580 btrfs_item_end_nr(leaf, slot + 1)) {
581 CORRUPT("slot offset bad", leaf, root, slot);
582 return -EIO;
583 }
584
585 /*
586 * Check to make sure that we don't point outside of the leaf,
587 * just incase all the items are consistent to eachother, but
588 * all point outside of the leaf.
589 */
590 if (btrfs_item_end_nr(leaf, slot) >
591 BTRFS_LEAF_DATA_SIZE(root)) {
592 CORRUPT("slot end outside of leaf", leaf, root, slot);
593 return -EIO;
594 }
595 }
596
597 return 0;
598}
599
Miao Xiefacc8a222013-07-25 19:22:34 +0800600static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
601 u64 phy_offset, struct page *page,
602 u64 start, u64 end, int mirror)
Chris Masonce9adaa2008-04-09 16:28:12 -0400603{
Chris Masonce9adaa2008-04-09 16:28:12 -0400604 u64 found_start;
605 int found_level;
Chris Masonce9adaa2008-04-09 16:28:12 -0400606 struct extent_buffer *eb;
607 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
Chris Masonf1885912008-04-09 16:28:12 -0400608 int ret = 0;
Chris Mason727011e2010-08-06 13:21:20 -0400609 int reads_done;
Chris Masonce9adaa2008-04-09 16:28:12 -0400610
Chris Masonce9adaa2008-04-09 16:28:12 -0400611 if (!page->private)
612 goto out;
Chris Masond3977122009-01-05 21:25:51 -0500613
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500614 eb = (struct extent_buffer *)page->private;
Chris Masond3977122009-01-05 21:25:51 -0500615
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400616 /* the pending IO might have been the only thing that kept this buffer
617 * in memory. Make sure we have a ref for all this other checks
618 */
619 extent_buffer_get(eb);
620
621 reads_done = atomic_dec_and_test(&eb->io_pages);
Chris Mason727011e2010-08-06 13:21:20 -0400622 if (!reads_done)
623 goto err;
Chris Masonf1885912008-04-09 16:28:12 -0400624
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400625 eb->read_mirror = mirror;
Filipe Manana656f30d2014-09-26 12:25:56 +0100626 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
Josef Bacikea466792012-03-26 21:57:36 -0400627 ret = -EIO;
628 goto err;
629 }
630
Chris Masonce9adaa2008-04-09 16:28:12 -0400631 found_start = btrfs_header_bytenr(eb);
Chris Mason727011e2010-08-06 13:21:20 -0400632 if (found_start != eb->start) {
David Sterba94647322015-10-08 11:01:36 +0200633 btrfs_err_rl(eb->fs_info, "bad tree block start %llu %llu",
634 found_start, eb->start);
Chris Masonf1885912008-04-09 16:28:12 -0400635 ret = -EIO;
Chris Masonce9adaa2008-04-09 16:28:12 -0400636 goto err;
637 }
Daniel Dressler01d58472014-11-21 17:15:07 +0900638 if (check_tree_block_fsid(root->fs_info, eb)) {
David Sterba94647322015-10-08 11:01:36 +0200639 btrfs_err_rl(eb->fs_info, "bad fsid on block %llu",
640 eb->start);
Chris Mason1259ab72008-05-12 13:39:03 -0400641 ret = -EIO;
642 goto err;
643 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400644 found_level = btrfs_header_level(eb);
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400645 if (found_level >= BTRFS_MAX_LEVEL) {
David Sterba68b663d2014-12-19 18:38:37 +0100646 btrfs_err(root->fs_info, "bad tree block level %d",
Josef Bacik1c24c3c2013-04-23 11:30:14 -0400647 (int)btrfs_header_level(eb));
648 ret = -EIO;
649 goto err;
650 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400651
Chris Mason85d4e462011-07-26 16:11:19 -0400652 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
653 eb, found_level);
Chris Mason4008c042009-02-12 14:09:45 -0500654
Daniel Dressler01d58472014-11-21 17:15:07 +0900655 ret = csum_tree_block(root->fs_info, eb, 1);
Josef Bacika826d6d2011-03-16 13:42:43 -0400656 if (ret) {
Chris Masonf1885912008-04-09 16:28:12 -0400657 ret = -EIO;
Josef Bacika826d6d2011-03-16 13:42:43 -0400658 goto err;
659 }
660
661 /*
662 * If this is a leaf block and it is corrupt, set the corrupt bit so
663 * that we don't try and read the other copies of this block, just
664 * return -EIO.
665 */
666 if (found_level == 0 && check_leaf(root, eb)) {
667 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
668 ret = -EIO;
669 }
Chris Masonce9adaa2008-04-09 16:28:12 -0400670
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400671 if (!ret)
672 set_extent_buffer_uptodate(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400673err:
Josef Bacik79fb65a2013-04-20 10:18:27 -0400674 if (reads_done &&
675 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200676 btree_readahead_hook(root, eb, eb->start, ret);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200677
David Woodhouse53b381b2013-01-29 18:40:14 -0500678 if (ret) {
679 /*
680 * our io error hook is going to dec the io pages
681 * again, we have to make sure it has something
682 * to decrement
683 */
684 atomic_inc(&eb->io_pages);
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400685 clear_extent_buffer_uptodate(eb);
David Woodhouse53b381b2013-01-29 18:40:14 -0500686 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -0400687 free_extent_buffer(eb);
Chris Masonce9adaa2008-04-09 16:28:12 -0400688out:
Chris Masonf1885912008-04-09 16:28:12 -0400689 return ret;
Chris Masonce9adaa2008-04-09 16:28:12 -0400690}
691
Josef Bacikea466792012-03-26 21:57:36 -0400692static int btree_io_failed_hook(struct page *page, int failed_mirror)
Arne Jansen4bb31e92011-06-10 13:55:54 +0200693{
Arne Jansen4bb31e92011-06-10 13:55:54 +0200694 struct extent_buffer *eb;
695 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
696
Josef Bacik4f2de97a2012-03-07 16:20:05 -0500697 eb = (struct extent_buffer *)page->private;
Filipe Manana656f30d2014-09-26 12:25:56 +0100698 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
Josef Bacik5cf1ab52012-04-16 09:42:26 -0400699 eb->read_mirror = failed_mirror;
David Woodhouse53b381b2013-01-29 18:40:14 -0500700 atomic_dec(&eb->io_pages);
Josef Bacikea466792012-03-26 21:57:36 -0400701 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
Arne Jansen4bb31e92011-06-10 13:55:54 +0200702 btree_readahead_hook(root, eb, eb->start, -EIO);
Arne Jansen4bb31e92011-06-10 13:55:54 +0200703 return -EIO; /* we fixed nothing */
704}
705
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200706static void end_workqueue_bio(struct bio *bio)
Chris Masonce9adaa2008-04-09 16:28:12 -0400707{
David Sterba97eb6b62014-07-30 00:55:42 +0200708 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
Chris Masonce9adaa2008-04-09 16:28:12 -0400709 struct btrfs_fs_info *fs_info;
Liu Bo9e0af232014-08-15 23:36:53 +0800710 struct btrfs_workqueue *wq;
711 btrfs_work_func_t func;
Chris Masonce9adaa2008-04-09 16:28:12 -0400712
Chris Masonce9adaa2008-04-09 16:28:12 -0400713 fs_info = end_io_wq->info;
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200714 end_io_wq->error = bio->bi_error;
Chris Masond20f7042008-12-08 16:58:54 -0500715
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200716 if (bio->bi_rw & REQ_WRITE) {
Liu Bo9e0af232014-08-15 23:36:53 +0800717 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
718 wq = fs_info->endio_meta_write_workers;
719 func = btrfs_endio_meta_write_helper;
720 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
721 wq = fs_info->endio_freespace_worker;
722 func = btrfs_freespace_write_helper;
723 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
724 wq = fs_info->endio_raid56_workers;
725 func = btrfs_endio_raid56_helper;
726 } else {
727 wq = fs_info->endio_write_workers;
728 func = btrfs_endio_write_helper;
729 }
Chris Masond20f7042008-12-08 16:58:54 -0500730 } else {
Miao Xie8b110e32014-09-12 18:44:03 +0800731 if (unlikely(end_io_wq->metadata ==
732 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
733 wq = fs_info->endio_repair_workers;
734 func = btrfs_endio_repair_helper;
735 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
Liu Bo9e0af232014-08-15 23:36:53 +0800736 wq = fs_info->endio_raid56_workers;
737 func = btrfs_endio_raid56_helper;
738 } else if (end_io_wq->metadata) {
739 wq = fs_info->endio_meta_workers;
740 func = btrfs_endio_meta_helper;
741 } else {
742 wq = fs_info->endio_workers;
743 func = btrfs_endio_helper;
744 }
Chris Masond20f7042008-12-08 16:58:54 -0500745 }
Liu Bo9e0af232014-08-15 23:36:53 +0800746
747 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
748 btrfs_queue_work(wq, &end_io_wq->work);
Chris Masonce9adaa2008-04-09 16:28:12 -0400749}
750
Chris Mason22c59942008-04-09 16:28:12 -0400751int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
David Sterbabfebd8b2014-07-30 00:25:45 +0200752 enum btrfs_wq_endio_type metadata)
Chris Mason0b86a832008-03-24 15:01:56 -0400753{
David Sterba97eb6b62014-07-30 00:55:42 +0200754 struct btrfs_end_io_wq *end_io_wq;
Miao Xie8b110e32014-09-12 18:44:03 +0800755
David Sterba97eb6b62014-07-30 00:55:42 +0200756 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
Chris Masonce9adaa2008-04-09 16:28:12 -0400757 if (!end_io_wq)
758 return -ENOMEM;
759
760 end_io_wq->private = bio->bi_private;
761 end_io_wq->end_io = bio->bi_end_io;
Chris Mason22c59942008-04-09 16:28:12 -0400762 end_io_wq->info = info;
Chris Masonce9adaa2008-04-09 16:28:12 -0400763 end_io_wq->error = 0;
764 end_io_wq->bio = bio;
Chris Mason22c59942008-04-09 16:28:12 -0400765 end_io_wq->metadata = metadata;
Chris Masonce9adaa2008-04-09 16:28:12 -0400766
767 bio->bi_private = end_io_wq;
768 bio->bi_end_io = end_workqueue_bio;
Chris Mason22c59942008-04-09 16:28:12 -0400769 return 0;
770}
771
Chris Masonb64a2852008-08-20 13:39:41 -0400772unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
Chris Mason4854ddd2008-08-15 15:34:17 -0400773{
774 unsigned long limit = min_t(unsigned long,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800775 info->thread_pool_size,
Chris Mason4854ddd2008-08-15 15:34:17 -0400776 info->fs_devices->open_devices);
777 return 256 * limit;
778}
779
Chris Mason4a69a412008-11-06 22:03:00 -0500780static void run_one_async_start(struct btrfs_work *work)
781{
Chris Mason4a69a412008-11-06 22:03:00 -0500782 struct async_submit_bio *async;
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100783 int ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500784
785 async = container_of(work, struct async_submit_bio, work);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100786 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
787 async->mirror_num, async->bio_flags,
788 async->bio_offset);
789 if (ret)
790 async->error = ret;
Chris Mason4a69a412008-11-06 22:03:00 -0500791}
792
793static void run_one_async_done(struct btrfs_work *work)
Chris Mason8b712842008-06-11 16:50:36 -0400794{
795 struct btrfs_fs_info *fs_info;
796 struct async_submit_bio *async;
Chris Mason4854ddd2008-08-15 15:34:17 -0400797 int limit;
Chris Mason8b712842008-06-11 16:50:36 -0400798
799 async = container_of(work, struct async_submit_bio, work);
800 fs_info = BTRFS_I(async->inode)->root->fs_info;
Chris Mason4854ddd2008-08-15 15:34:17 -0400801
Chris Masonb64a2852008-08-20 13:39:41 -0400802 limit = btrfs_async_submit_limit(fs_info);
Chris Mason4854ddd2008-08-15 15:34:17 -0400803 limit = limit * 2 / 3;
804
David Sterbaee863952015-02-16 19:41:40 +0100805 /*
806 * atomic_dec_return implies a barrier for waitqueue_active
807 */
Josef Bacik66657b32012-08-01 15:36:24 -0400808 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
Chris Masonb64a2852008-08-20 13:39:41 -0400809 waitqueue_active(&fs_info->async_submit_wait))
Chris Mason4854ddd2008-08-15 15:34:17 -0400810 wake_up(&fs_info->async_submit_wait);
811
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100812 /* If an error occured we just want to clean up the bio and move on */
813 if (async->error) {
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200814 async->bio->bi_error = async->error;
815 bio_endio(async->bio);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100816 return;
817 }
818
Chris Mason4a69a412008-11-06 22:03:00 -0500819 async->submit_bio_done(async->inode, async->rw, async->bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400820 async->mirror_num, async->bio_flags,
821 async->bio_offset);
Chris Mason4a69a412008-11-06 22:03:00 -0500822}
823
824static void run_one_async_free(struct btrfs_work *work)
825{
826 struct async_submit_bio *async;
827
828 async = container_of(work, struct async_submit_bio, work);
Chris Mason8b712842008-06-11 16:50:36 -0400829 kfree(async);
830}
831
Chris Mason44b8bd72008-04-16 11:14:51 -0400832int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
833 int rw, struct bio *bio, int mirror_num,
Chris Masonc8b97812008-10-29 14:49:59 -0400834 unsigned long bio_flags,
Chris Masoneaf25d92010-05-25 09:48:28 -0400835 u64 bio_offset,
Chris Mason4a69a412008-11-06 22:03:00 -0500836 extent_submit_bio_hook_t *submit_bio_start,
837 extent_submit_bio_hook_t *submit_bio_done)
Chris Mason44b8bd72008-04-16 11:14:51 -0400838{
839 struct async_submit_bio *async;
840
841 async = kmalloc(sizeof(*async), GFP_NOFS);
842 if (!async)
843 return -ENOMEM;
844
845 async->inode = inode;
846 async->rw = rw;
847 async->bio = bio;
848 async->mirror_num = mirror_num;
Chris Mason4a69a412008-11-06 22:03:00 -0500849 async->submit_bio_start = submit_bio_start;
850 async->submit_bio_done = submit_bio_done;
851
Liu Bo9e0af232014-08-15 23:36:53 +0800852 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800853 run_one_async_done, run_one_async_free);
Chris Mason4a69a412008-11-06 22:03:00 -0500854
Chris Masonc8b97812008-10-29 14:49:59 -0400855 async->bio_flags = bio_flags;
Chris Masoneaf25d92010-05-25 09:48:28 -0400856 async->bio_offset = bio_offset;
Chris Mason8c8bee12008-09-29 11:19:10 -0400857
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100858 async->error = 0;
859
Chris Masoncb03c742008-05-15 16:15:45 -0400860 atomic_inc(&fs_info->nr_async_submits);
Chris Masond313d7a2009-04-20 15:50:09 -0400861
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200862 if (rw & REQ_SYNC)
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800863 btrfs_set_work_high_priority(&async->work);
Chris Masond313d7a2009-04-20 15:50:09 -0400864
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +0800865 btrfs_queue_work(fs_info->workers, &async->work);
Chris Mason9473f162008-08-28 06:15:24 -0400866
Chris Masond3977122009-01-05 21:25:51 -0500867 while (atomic_read(&fs_info->async_submit_draining) &&
Chris Mason771ed682008-11-06 22:02:51 -0500868 atomic_read(&fs_info->nr_async_submits)) {
869 wait_event(fs_info->async_submit_wait,
870 (atomic_read(&fs_info->nr_async_submits) == 0));
871 }
872
Chris Mason44b8bd72008-04-16 11:14:51 -0400873 return 0;
874}
875
Chris Masonce3ed712008-09-23 13:14:12 -0400876static int btree_csum_one_bio(struct bio *bio)
877{
Kent Overstreet2c30c712013-11-07 12:20:26 -0800878 struct bio_vec *bvec;
Chris Masonce3ed712008-09-23 13:14:12 -0400879 struct btrfs_root *root;
Kent Overstreet2c30c712013-11-07 12:20:26 -0800880 int i, ret = 0;
Chris Masonce3ed712008-09-23 13:14:12 -0400881
Kent Overstreet2c30c712013-11-07 12:20:26 -0800882 bio_for_each_segment_all(bvec, bio, i) {
Chris Masonce3ed712008-09-23 13:14:12 -0400883 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
Daniel Dressler01d58472014-11-21 17:15:07 +0900884 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100885 if (ret)
886 break;
Chris Masonce3ed712008-09-23 13:14:12 -0400887 }
Kent Overstreet2c30c712013-11-07 12:20:26 -0800888
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100889 return ret;
Chris Masonce3ed712008-09-23 13:14:12 -0400890}
891
Chris Mason4a69a412008-11-06 22:03:00 -0500892static int __btree_submit_bio_start(struct inode *inode, int rw,
893 struct bio *bio, int mirror_num,
Chris Masoneaf25d92010-05-25 09:48:28 -0400894 unsigned long bio_flags,
895 u64 bio_offset)
Chris Mason22c59942008-04-09 16:28:12 -0400896{
Chris Mason8b712842008-06-11 16:50:36 -0400897 /*
898 * when we're called for a write, we're already in the async
Chris Mason5443be42008-08-15 15:34:16 -0400899 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400900 */
Jeff Mahoney79787ea2012-03-12 16:03:00 +0100901 return btree_csum_one_bio(bio);
Chris Mason4a69a412008-11-06 22:03:00 -0500902}
Chris Mason22c59942008-04-09 16:28:12 -0400903
Chris Mason4a69a412008-11-06 22:03:00 -0500904static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400905 int mirror_num, unsigned long bio_flags,
906 u64 bio_offset)
Chris Mason4a69a412008-11-06 22:03:00 -0500907{
Stefan Behrens61891922012-11-05 18:51:52 +0100908 int ret;
909
Chris Mason8b712842008-06-11 16:50:36 -0400910 /*
Chris Mason4a69a412008-11-06 22:03:00 -0500911 * when we're called for a write, we're already in the async
912 * submission context. Just jump into btrfs_map_bio
Chris Mason8b712842008-06-11 16:50:36 -0400913 */
Stefan Behrens61891922012-11-05 18:51:52 +0100914 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200915 if (ret) {
916 bio->bi_error = ret;
917 bio_endio(bio);
918 }
Stefan Behrens61891922012-11-05 18:51:52 +0100919 return ret;
Chris Mason0b86a832008-03-24 15:01:56 -0400920}
921
Josef Bacikde0022b2012-09-25 14:25:58 -0400922static int check_async_write(struct inode *inode, unsigned long bio_flags)
923{
924 if (bio_flags & EXTENT_BIO_TREE_LOG)
925 return 0;
926#ifdef CONFIG_X86
927 if (cpu_has_xmm4_2)
928 return 0;
929#endif
930 return 1;
931}
932
Chris Mason44b8bd72008-04-16 11:14:51 -0400933static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
Chris Masoneaf25d92010-05-25 09:48:28 -0400934 int mirror_num, unsigned long bio_flags,
935 u64 bio_offset)
Chris Mason44b8bd72008-04-16 11:14:51 -0400936{
Josef Bacikde0022b2012-09-25 14:25:58 -0400937 int async = check_async_write(inode, bio_flags);
Chris Masoncad321a2008-12-17 14:51:42 -0500938 int ret;
939
Christoph Hellwig7b6d91d2010-08-07 18:20:39 +0200940 if (!(rw & REQ_WRITE)) {
Chris Mason4a69a412008-11-06 22:03:00 -0500941 /*
942 * called for a read, do the setup so that checksum validation
943 * can happen in the async kernel threads
944 */
Chris Masonf3f266a2012-03-23 10:22:46 -0400945 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
David Sterbabfebd8b2014-07-30 00:25:45 +0200946 bio, BTRFS_WQ_ENDIO_METADATA);
Chris Mason1d4284b2012-03-28 20:31:37 -0400947 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100948 goto out_w_error;
949 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
950 mirror_num, 0);
Josef Bacikde0022b2012-09-25 14:25:58 -0400951 } else if (!async) {
952 ret = btree_csum_one_bio(bio);
953 if (ret)
Stefan Behrens61891922012-11-05 18:51:52 +0100954 goto out_w_error;
955 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
956 mirror_num, 0);
957 } else {
958 /*
959 * kthread helpers are used to submit writes so that
960 * checksumming can happen in parallel across all CPUs
961 */
962 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
963 inode, rw, bio, mirror_num, 0,
964 bio_offset,
965 __btree_submit_bio_start,
966 __btree_submit_bio_done);
Chris Mason44b8bd72008-04-16 11:14:51 -0400967 }
Chris Masond313d7a2009-04-20 15:50:09 -0400968
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200969 if (ret)
970 goto out_w_error;
971 return 0;
972
Stefan Behrens61891922012-11-05 18:51:52 +0100973out_w_error:
Christoph Hellwig4246a0b2015-07-20 15:29:37 +0200974 bio->bi_error = ret;
975 bio_endio(bio);
Stefan Behrens61891922012-11-05 18:51:52 +0100976 return ret;
Chris Mason44b8bd72008-04-16 11:14:51 -0400977}
978
Jan Beulich3dd14622010-12-07 14:54:09 +0000979#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -0500980static int btree_migratepage(struct address_space *mapping,
Mel Gormana6bc32b2012-01-12 17:19:43 -0800981 struct page *newpage, struct page *page,
982 enum migrate_mode mode)
Chris Mason784b4e22010-11-21 22:20:49 -0500983{
984 /*
985 * we can't safely write a btree page from here,
986 * we haven't done the locking hook
987 */
988 if (PageDirty(page))
989 return -EAGAIN;
990 /*
991 * Buffers may be managed in a filesystem specific way.
992 * We must have no buffers or drop them.
993 */
994 if (page_has_private(page) &&
995 !try_to_release_page(page, GFP_KERNEL))
996 return -EAGAIN;
Mel Gormana6bc32b2012-01-12 17:19:43 -0800997 return migrate_page(mapping, newpage, page, mode);
Chris Mason784b4e22010-11-21 22:20:49 -0500998}
Jan Beulich3dd14622010-12-07 14:54:09 +0000999#endif
Chris Mason784b4e22010-11-21 22:20:49 -05001000
Chris Mason0da54682007-11-07 21:08:01 -05001001
1002static int btree_writepages(struct address_space *mapping,
1003 struct writeback_control *wbc)
1004{
Miao Xiee2d84522013-01-29 10:09:20 +00001005 struct btrfs_fs_info *fs_info;
1006 int ret;
1007
Chris Masond8d5f3e2007-12-11 12:42:00 -05001008 if (wbc->sync_mode == WB_SYNC_NONE) {
Chris Mason448d6402007-11-27 07:52:01 -08001009
1010 if (wbc->for_kupdate)
1011 return 0;
1012
Miao Xiee2d84522013-01-29 10:09:20 +00001013 fs_info = BTRFS_I(mapping->host)->root->fs_info;
Chris Masonb9473432009-03-13 11:00:37 -04001014 /* this is a bit racy, but that's ok */
Miao Xiee2d84522013-01-29 10:09:20 +00001015 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1016 BTRFS_DIRTY_METADATA_THRESH);
1017 if (ret < 0)
Chris Mason793955b2007-11-26 16:34:41 -08001018 return 0;
Chris Mason793955b2007-11-26 16:34:41 -08001019 }
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001020 return btree_write_cache_pages(mapping, wbc);
Chris Mason0da54682007-11-07 21:08:01 -05001021}
1022
Christoph Hellwigb2950862008-12-02 09:54:17 -05001023static int btree_readpage(struct file *file, struct page *page)
Chris Mason5f39d392007-10-15 16:14:19 -04001024{
Chris Masond1310b22008-01-24 16:13:08 -05001025 struct extent_io_tree *tree;
1026 tree = &BTRFS_I(page->mapping->host)->io_tree;
Jan Schmidt8ddc7d92011-06-13 20:02:58 +02001027 return extent_read_full_page(tree, page, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001028}
1029
Chris Mason70dec802008-01-29 09:59:12 -05001030static int btree_releasepage(struct page *page, gfp_t gfp_flags)
Chris Mason5f39d392007-10-15 16:14:19 -04001031{
Chris Mason98509cfc2008-09-11 15:51:43 -04001032 if (PageWriteback(page) || PageDirty(page))
Chris Masond3977122009-01-05 21:25:51 -05001033 return 0;
David Sterba0c4e5382012-01-26 15:01:12 -05001034
David Sterbaf7a52a42013-04-26 14:56:29 +00001035 return try_release_extent_buffer(page);
Chris Masone20d96d2007-03-22 12:13:20 -04001036}
Chris Mason123abc82007-03-14 14:14:43 -04001037
Lukas Czernerd47992f2013-05-21 23:17:23 -04001038static void btree_invalidatepage(struct page *page, unsigned int offset,
1039 unsigned int length)
Chris Masond98237b2007-03-28 13:57:48 -04001040{
Chris Masond1310b22008-01-24 16:13:08 -05001041 struct extent_io_tree *tree;
1042 tree = &BTRFS_I(page->mapping->host)->io_tree;
Chris Mason5f39d392007-10-15 16:14:19 -04001043 extent_invalidatepage(tree, page, offset);
1044 btree_releasepage(page, GFP_NOFS);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001045 if (PagePrivate(page)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05001046 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1047 "page private not zero on page %llu",
1048 (unsigned long long)page_offset(page));
Chris Mason9ad6b7b2008-04-18 16:11:30 -04001049 ClearPagePrivate(page);
1050 set_page_private(page, 0);
1051 page_cache_release(page);
1052 }
Chris Masond98237b2007-03-28 13:57:48 -04001053}
1054
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001055static int btree_set_page_dirty(struct page *page)
1056{
Josef Bacikbb146eb2012-10-15 13:30:43 -04001057#ifdef DEBUG
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001058 struct extent_buffer *eb;
1059
1060 BUG_ON(!PagePrivate(page));
1061 eb = (struct extent_buffer *)page->private;
1062 BUG_ON(!eb);
1063 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1064 BUG_ON(!atomic_read(&eb->refs));
1065 btrfs_assert_tree_locked(eb);
Josef Bacikbb146eb2012-10-15 13:30:43 -04001066#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001067 return __set_page_dirty_nobuffers(page);
1068}
1069
Alexey Dobriyan7f094102009-09-21 17:01:10 -07001070static const struct address_space_operations btree_aops = {
Chris Masond98237b2007-03-28 13:57:48 -04001071 .readpage = btree_readpage,
Chris Mason0da54682007-11-07 21:08:01 -05001072 .writepages = btree_writepages,
Chris Mason5f39d392007-10-15 16:14:19 -04001073 .releasepage = btree_releasepage,
1074 .invalidatepage = btree_invalidatepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001075#ifdef CONFIG_MIGRATION
Chris Mason784b4e22010-11-21 22:20:49 -05001076 .migratepage = btree_migratepage,
Chris Mason5a92bc82010-11-29 09:49:11 -05001077#endif
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001078 .set_page_dirty = btree_set_page_dirty,
Chris Masond98237b2007-03-28 13:57:48 -04001079};
1080
David Sterbad3e46fe2014-06-15 02:04:19 +02001081void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
Chris Mason090d1872007-05-01 08:53:32 -04001082{
Chris Mason5f39d392007-10-15 16:14:19 -04001083 struct extent_buffer *buf = NULL;
1084 struct inode *btree_inode = root->fs_info->btree_inode;
Chris Mason090d1872007-05-01 08:53:32 -04001085
David Sterbaa83fffb2014-06-15 02:39:54 +02001086 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001087 if (!buf)
David Sterba6197d862014-06-15 00:49:36 +02001088 return;
Chris Masond1310b22008-01-24 16:13:08 -05001089 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
Arne Jansenbb82ab82011-06-10 14:06:53 +02001090 buf, 0, WAIT_NONE, btree_get_extent, 0);
Chris Mason5f39d392007-10-15 16:14:19 -04001091 free_extent_buffer(buf);
Chris Mason090d1872007-05-01 08:53:32 -04001092}
1093
David Sterbac0dcaa42014-06-15 02:24:21 +02001094int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
Arne Jansenab0fff02011-05-23 14:25:41 +02001095 int mirror_num, struct extent_buffer **eb)
1096{
1097 struct extent_buffer *buf = NULL;
1098 struct inode *btree_inode = root->fs_info->btree_inode;
1099 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1100 int ret;
1101
David Sterbaa83fffb2014-06-15 02:39:54 +02001102 buf = btrfs_find_create_tree_block(root, bytenr);
Arne Jansenab0fff02011-05-23 14:25:41 +02001103 if (!buf)
1104 return 0;
1105
1106 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1107
1108 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1109 btree_get_extent, mirror_num);
1110 if (ret) {
1111 free_extent_buffer(buf);
1112 return ret;
1113 }
1114
1115 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1116 free_extent_buffer(buf);
1117 return -EIO;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04001118 } else if (extent_buffer_uptodate(buf)) {
Arne Jansenab0fff02011-05-23 14:25:41 +02001119 *eb = buf;
1120 } else {
1121 free_extent_buffer(buf);
1122 }
1123 return 0;
1124}
1125
Daniel Dressler01d58472014-11-21 17:15:07 +09001126struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
David Sterba0308af42014-06-15 01:43:40 +02001127 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001128{
Daniel Dressler01d58472014-11-21 17:15:07 +09001129 return find_extent_buffer(fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001130}
1131
1132struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
David Sterbaa83fffb2014-06-15 02:39:54 +02001133 u64 bytenr)
Chris Mason0999df52008-04-01 13:48:14 -04001134{
David Sterbafccb84c2014-09-29 23:53:21 +02001135 if (btrfs_test_is_dummy_root(root))
David Sterbace3e6982014-06-15 03:00:04 +02001136 return alloc_test_extent_buffer(root->fs_info, bytenr);
1137 return alloc_extent_buffer(root->fs_info, bytenr);
Chris Mason0999df52008-04-01 13:48:14 -04001138}
1139
1140
Chris Masone02119d2008-09-05 16:13:11 -04001141int btrfs_write_tree_block(struct extent_buffer *buf)
1142{
Chris Mason727011e2010-08-06 13:21:20 -04001143 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001144 buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001145}
1146
1147int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1148{
Chris Mason727011e2010-08-06 13:21:20 -04001149 return filemap_fdatawait_range(buf->pages[0]->mapping,
Christoph Hellwig8aa38c32009-10-01 12:58:30 -04001150 buf->start, buf->start + buf->len - 1);
Chris Masone02119d2008-09-05 16:13:11 -04001151}
1152
Chris Masondb945352007-10-15 16:15:53 -04001153struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
David Sterbace86cd52014-06-15 01:07:32 +02001154 u64 parent_transid)
Chris Masone20d96d2007-03-22 12:13:20 -04001155{
Chris Mason5f39d392007-10-15 16:14:19 -04001156 struct extent_buffer *buf = NULL;
Chris Mason19c00dd2007-10-15 16:19:22 -04001157 int ret;
1158
David Sterbaa83fffb2014-06-15 02:39:54 +02001159 buf = btrfs_find_create_tree_block(root, bytenr);
Chris Mason5f39d392007-10-15 16:14:19 -04001160 if (!buf)
Liu Bo64c043d2015-05-25 17:30:15 +08001161 return ERR_PTR(-ENOMEM);
Chris Masone4204de2008-01-08 15:46:27 -05001162
Chris Masonca7a79a2008-05-12 12:59:19 -04001163 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001164 if (ret) {
1165 free_extent_buffer(buf);
Liu Bo64c043d2015-05-25 17:30:15 +08001166 return ERR_PTR(ret);
Filipe David Borba Manana0f0fe8f2013-07-31 00:39:56 +01001167 }
Chris Mason5f39d392007-10-15 16:14:19 -04001168 return buf;
Chris Masonce9adaa2008-04-09 16:28:12 -04001169
Chris Masoneb60cea2007-02-02 09:18:22 -05001170}
1171
Daniel Dressler01d58472014-11-21 17:15:07 +09001172void clean_tree_block(struct btrfs_trans_handle *trans,
1173 struct btrfs_fs_info *fs_info,
Jeff Mahoneyd5c13f92012-03-01 14:56:27 +01001174 struct extent_buffer *buf)
Chris Masoned2ff2c2007-03-01 18:59:40 -05001175{
Chris Mason55c69072008-01-09 15:55:33 -05001176 if (btrfs_header_generation(buf) ==
Miao Xiee2d84522013-01-29 10:09:20 +00001177 fs_info->running_transaction->transid) {
Chris Masonb9447ef2009-03-09 11:45:38 -04001178 btrfs_assert_tree_locked(buf);
Chris Masonb4ce94d2009-02-04 09:25:08 -05001179
Chris Masonb9473432009-03-13 11:00:37 -04001180 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
Miao Xiee2d84522013-01-29 10:09:20 +00001181 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1182 -buf->len,
1183 fs_info->dirty_metadata_batch);
Josef Baciked7b63e2012-10-15 13:33:54 -04001184 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1185 btrfs_set_lock_blocking(buf);
1186 clear_extent_buffer_dirty(buf);
1187 }
Chris Mason925baed2008-06-25 16:01:30 -04001188 }
Chris Mason5f39d392007-10-15 16:14:19 -04001189}
1190
Miao Xie8257b2d2014-03-06 13:38:19 +08001191static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1192{
1193 struct btrfs_subvolume_writers *writers;
1194 int ret;
1195
1196 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1197 if (!writers)
1198 return ERR_PTR(-ENOMEM);
1199
Tejun Heo908c7f12014-09-08 09:51:29 +09001200 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
Miao Xie8257b2d2014-03-06 13:38:19 +08001201 if (ret < 0) {
1202 kfree(writers);
1203 return ERR_PTR(ret);
1204 }
1205
1206 init_waitqueue_head(&writers->wait);
1207 return writers;
1208}
1209
1210static void
1211btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1212{
1213 percpu_counter_destroy(&writers->counter);
1214 kfree(writers);
1215}
1216
David Sterba707e8a02014-06-04 19:22:26 +02001217static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1218 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
Jeff Mahoney143bede2012-03-01 14:56:26 +01001219 u64 objectid)
Chris Masond97e63b2007-02-20 16:40:44 -05001220{
Chris Masoncfaa7292007-02-21 17:04:57 -05001221 root->node = NULL;
Chris Masona28ec192007-03-06 20:08:01 -05001222 root->commit_root = NULL;
Chris Masondb945352007-10-15 16:15:53 -04001223 root->sectorsize = sectorsize;
1224 root->nodesize = nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05001225 root->stripesize = stripesize;
Miao Xie27cdeb72014-04-02 19:51:05 +08001226 root->state = 0;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001227 root->orphan_cleanup_state = 0;
Chris Mason0b86a832008-03-24 15:01:56 -04001228
Chris Mason0f7d52f2007-04-09 10:42:37 -04001229 root->objectid = objectid;
1230 root->last_trans = 0;
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001231 root->highest_objectid = 0;
Miao Xieeb73c1b2013-05-15 07:48:22 +00001232 root->nr_delalloc_inodes = 0;
Miao Xie199c2a92013-05-15 07:48:23 +00001233 root->nr_ordered_extents = 0;
Josef Bacik58176a92007-08-29 15:47:34 -04001234 root->name = NULL;
Eric Paris6bef4d32010-02-23 19:43:04 +00001235 root->inode_tree = RB_ROOT;
Miao Xie16cdcec2011-04-22 18:12:22 +08001236 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001237 root->block_rsv = NULL;
Yan, Zhengd68fc572010-05-16 10:49:58 -04001238 root->orphan_block_rsv = NULL;
Chris Mason0b86a832008-03-24 15:01:56 -04001239
1240 INIT_LIST_HEAD(&root->dirty_list);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001241 INIT_LIST_HEAD(&root->root_list);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001242 INIT_LIST_HEAD(&root->delalloc_inodes);
1243 INIT_LIST_HEAD(&root->delalloc_root);
Miao Xie199c2a92013-05-15 07:48:23 +00001244 INIT_LIST_HEAD(&root->ordered_extents);
1245 INIT_LIST_HEAD(&root->ordered_root);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001246 INIT_LIST_HEAD(&root->logged_list[0]);
1247 INIT_LIST_HEAD(&root->logged_list[1]);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001248 spin_lock_init(&root->orphan_lock);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001249 spin_lock_init(&root->inode_lock);
Miao Xieeb73c1b2013-05-15 07:48:22 +00001250 spin_lock_init(&root->delalloc_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00001251 spin_lock_init(&root->ordered_extent_lock);
Yan, Zhengf0486c62010-05-16 10:46:25 -04001252 spin_lock_init(&root->accounting_lock);
Josef Bacik2ab28f32012-10-12 15:27:49 -04001253 spin_lock_init(&root->log_extents_lock[0]);
1254 spin_lock_init(&root->log_extents_lock[1]);
Chris Masona2135012008-06-25 16:01:30 -04001255 mutex_init(&root->objectid_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001256 mutex_init(&root->log_mutex);
Miao Xie31f3d252014-03-06 13:55:02 +08001257 mutex_init(&root->ordered_extent_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08001258 mutex_init(&root->delalloc_mutex);
Yan Zheng7237f1832009-01-21 12:54:03 -05001259 init_waitqueue_head(&root->log_writer_wait);
1260 init_waitqueue_head(&root->log_commit_wait[0]);
1261 init_waitqueue_head(&root->log_commit_wait[1]);
Miao Xie8b050d32014-02-20 18:08:58 +08001262 INIT_LIST_HEAD(&root->log_ctxs[0]);
1263 INIT_LIST_HEAD(&root->log_ctxs[1]);
Yan Zheng7237f1832009-01-21 12:54:03 -05001264 atomic_set(&root->log_commit[0], 0);
1265 atomic_set(&root->log_commit[1], 0);
1266 atomic_set(&root->log_writers, 0);
Miao Xie2ecb7922012-09-06 04:04:27 -06001267 atomic_set(&root->log_batch, 0);
Josef Bacik8a35d952012-05-23 14:26:42 -04001268 atomic_set(&root->orphan_inodes, 0);
Miao Xieb0feb9d2013-05-15 07:48:20 +00001269 atomic_set(&root->refs, 1);
Miao Xie8257b2d2014-03-06 13:38:19 +08001270 atomic_set(&root->will_be_snapshoted, 0);
Qu Wenruo55eeaf02015-09-08 17:08:38 +08001271 atomic_set(&root->qgroup_meta_rsv, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001272 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001273 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001274 root->last_log_commit = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001275 if (fs_info)
1276 extent_io_tree_init(&root->dirty_log_pages,
1277 fs_info->btree_inode->i_mapping);
Chris Mason017e5362008-07-28 15:32:51 -04001278
Chris Mason3768f362007-03-13 16:47:54 -04001279 memset(&root->root_key, 0, sizeof(root->root_key));
1280 memset(&root->root_item, 0, sizeof(root->root_item));
Chris Mason6702ed42007-08-07 16:15:09 -04001281 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
Josef Bacik06ea65a2013-09-19 16:07:01 -04001282 if (fs_info)
1283 root->defrag_trans_start = fs_info->generation;
1284 else
1285 root->defrag_trans_start = 0;
Chris Mason4d775672007-04-20 20:23:12 -04001286 root->root_key.objectid = objectid;
Al Viro0ee5dc62011-07-07 15:44:25 -04001287 root->anon_dev = 0;
Alexander Block8ea05e32012-07-25 17:35:53 +02001288
Anand Jain5f3ab902012-12-07 09:28:54 +00001289 spin_lock_init(&root->root_item_lock);
Chris Mason3768f362007-03-13 16:47:54 -04001290}
1291
Al Virof84a8bd2011-11-17 15:57:57 -05001292static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
Al Viro6f07e422011-11-17 00:46:16 -05001293{
1294 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1295 if (root)
1296 root->fs_info = fs_info;
1297 return root;
1298}
1299
Josef Bacik06ea65a2013-09-19 16:07:01 -04001300#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1301/* Should only be used by the testing infrastructure */
1302struct btrfs_root *btrfs_alloc_dummy_root(void)
1303{
1304 struct btrfs_root *root;
1305
1306 root = btrfs_alloc_root(NULL);
1307 if (!root)
1308 return ERR_PTR(-ENOMEM);
David Sterba707e8a02014-06-04 19:22:26 +02001309 __setup_root(4096, 4096, 4096, root, NULL, 1);
Miao Xie27cdeb72014-04-02 19:51:05 +08001310 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04001311 root->alloc_bytenr = 0;
Josef Bacik06ea65a2013-09-19 16:07:01 -04001312
1313 return root;
1314}
1315#endif
1316
Arne Jansen20897f52011-09-13 12:44:20 +02001317struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1318 struct btrfs_fs_info *fs_info,
1319 u64 objectid)
1320{
1321 struct extent_buffer *leaf;
1322 struct btrfs_root *tree_root = fs_info->tree_root;
1323 struct btrfs_root *root;
1324 struct btrfs_key key;
1325 int ret = 0;
Stefan Behrens6463fe52013-04-19 15:08:05 +00001326 uuid_le uuid;
Arne Jansen20897f52011-09-13 12:44:20 +02001327
1328 root = btrfs_alloc_root(fs_info);
1329 if (!root)
1330 return ERR_PTR(-ENOMEM);
1331
David Sterba707e8a02014-06-04 19:22:26 +02001332 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1333 tree_root->stripesize, root, fs_info, objectid);
Arne Jansen20897f52011-09-13 12:44:20 +02001334 root->root_key.objectid = objectid;
1335 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1336 root->root_key.offset = 0;
1337
David Sterba4d75f8a2014-06-15 01:54:12 +02001338 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
Arne Jansen20897f52011-09-13 12:44:20 +02001339 if (IS_ERR(leaf)) {
1340 ret = PTR_ERR(leaf);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001341 leaf = NULL;
Arne Jansen20897f52011-09-13 12:44:20 +02001342 goto fail;
1343 }
1344
Arne Jansen20897f52011-09-13 12:44:20 +02001345 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1346 btrfs_set_header_bytenr(leaf, leaf->start);
1347 btrfs_set_header_generation(leaf, trans->transid);
1348 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1349 btrfs_set_header_owner(leaf, objectid);
1350 root->node = leaf;
1351
Ross Kirk0a4e5582013-09-24 10:12:38 +01001352 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
Arne Jansen20897f52011-09-13 12:44:20 +02001353 BTRFS_FSID_SIZE);
1354 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02001355 btrfs_header_chunk_tree_uuid(leaf),
Arne Jansen20897f52011-09-13 12:44:20 +02001356 BTRFS_UUID_SIZE);
1357 btrfs_mark_buffer_dirty(leaf);
1358
1359 root->commit_root = btrfs_root_node(root);
Miao Xie27cdeb72014-04-02 19:51:05 +08001360 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Arne Jansen20897f52011-09-13 12:44:20 +02001361
1362 root->root_item.flags = 0;
1363 root->root_item.byte_limit = 0;
1364 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1365 btrfs_set_root_generation(&root->root_item, trans->transid);
1366 btrfs_set_root_level(&root->root_item, 0);
1367 btrfs_set_root_refs(&root->root_item, 1);
1368 btrfs_set_root_used(&root->root_item, leaf->len);
1369 btrfs_set_root_last_snapshot(&root->root_item, 0);
1370 btrfs_set_root_dirid(&root->root_item, 0);
Stefan Behrens6463fe52013-04-19 15:08:05 +00001371 uuid_le_gen(&uuid);
1372 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
Arne Jansen20897f52011-09-13 12:44:20 +02001373 root->root_item.drop_level = 0;
1374
1375 key.objectid = objectid;
1376 key.type = BTRFS_ROOT_ITEM_KEY;
1377 key.offset = 0;
1378 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1379 if (ret)
1380 goto fail;
1381
1382 btrfs_tree_unlock(leaf);
1383
Arne Jansen20897f52011-09-13 12:44:20 +02001384 return root;
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001385
1386fail:
1387 if (leaf) {
1388 btrfs_tree_unlock(leaf);
Tsutomu Itoh59885b32014-04-09 09:18:04 +09001389 free_extent_buffer(root->commit_root);
Tsutomu Itoh1dd05682013-03-21 04:32:32 +00001390 free_extent_buffer(leaf);
1391 }
1392 kfree(root);
1393
1394 return ERR_PTR(ret);
Arne Jansen20897f52011-09-13 12:44:20 +02001395}
1396
Yan Zheng7237f1832009-01-21 12:54:03 -05001397static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1398 struct btrfs_fs_info *fs_info)
Chris Mason0f7d52f2007-04-09 10:42:37 -04001399{
1400 struct btrfs_root *root;
1401 struct btrfs_root *tree_root = fs_info->tree_root;
Yan Zheng7237f1832009-01-21 12:54:03 -05001402 struct extent_buffer *leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001403
Al Viro6f07e422011-11-17 00:46:16 -05001404 root = btrfs_alloc_root(fs_info);
Chris Masone02119d2008-09-05 16:13:11 -04001405 if (!root)
Yan Zheng7237f1832009-01-21 12:54:03 -05001406 return ERR_PTR(-ENOMEM);
Chris Masone02119d2008-09-05 16:13:11 -04001407
David Sterba707e8a02014-06-04 19:22:26 +02001408 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1409 tree_root->stripesize, root, fs_info,
1410 BTRFS_TREE_LOG_OBJECTID);
Chris Masone02119d2008-09-05 16:13:11 -04001411
1412 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1413 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1414 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Miao Xie27cdeb72014-04-02 19:51:05 +08001415
Yan Zheng7237f1832009-01-21 12:54:03 -05001416 /*
Miao Xie27cdeb72014-04-02 19:51:05 +08001417 * DON'T set REF_COWS for log trees
1418 *
Yan Zheng7237f1832009-01-21 12:54:03 -05001419 * log trees do not get reference counted because they go away
1420 * before a real commit is actually done. They do store pointers
1421 * to file data extents, and those reference counts still get
1422 * updated (along with back refs to the log tree).
1423 */
Chris Masone02119d2008-09-05 16:13:11 -04001424
David Sterba4d75f8a2014-06-15 01:54:12 +02001425 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1426 NULL, 0, 0, 0);
Yan Zheng7237f1832009-01-21 12:54:03 -05001427 if (IS_ERR(leaf)) {
1428 kfree(root);
1429 return ERR_CAST(leaf);
1430 }
Chris Masone02119d2008-09-05 16:13:11 -04001431
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001432 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1433 btrfs_set_header_bytenr(leaf, leaf->start);
1434 btrfs_set_header_generation(leaf, trans->transid);
1435 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1436 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Yan Zheng7237f1832009-01-21 12:54:03 -05001437 root->node = leaf;
Chris Masone02119d2008-09-05 16:13:11 -04001438
1439 write_extent_buffer(root->node, root->fs_info->fsid,
Ross Kirk0a4e5582013-09-24 10:12:38 +01001440 btrfs_header_fsid(), BTRFS_FSID_SIZE);
Chris Masone02119d2008-09-05 16:13:11 -04001441 btrfs_mark_buffer_dirty(root->node);
1442 btrfs_tree_unlock(root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001443 return root;
1444}
1445
1446int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1447 struct btrfs_fs_info *fs_info)
1448{
1449 struct btrfs_root *log_root;
1450
1451 log_root = alloc_log_tree(trans, fs_info);
1452 if (IS_ERR(log_root))
1453 return PTR_ERR(log_root);
1454 WARN_ON(fs_info->log_root_tree);
1455 fs_info->log_root_tree = log_root;
1456 return 0;
1457}
1458
1459int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1460 struct btrfs_root *root)
1461{
1462 struct btrfs_root *log_root;
1463 struct btrfs_inode_item *inode_item;
1464
1465 log_root = alloc_log_tree(trans, root->fs_info);
1466 if (IS_ERR(log_root))
1467 return PTR_ERR(log_root);
1468
1469 log_root->last_trans = trans->transid;
1470 log_root->root_key.offset = root->root_key.objectid;
1471
1472 inode_item = &log_root->root_item.inode;
Qu Wenruo3cae2102013-07-16 11:19:18 +08001473 btrfs_set_stack_inode_generation(inode_item, 1);
1474 btrfs_set_stack_inode_size(inode_item, 3);
1475 btrfs_set_stack_inode_nlink(inode_item, 1);
David Sterba707e8a02014-06-04 19:22:26 +02001476 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
Qu Wenruo3cae2102013-07-16 11:19:18 +08001477 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
Yan Zheng7237f1832009-01-21 12:54:03 -05001478
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001479 btrfs_set_root_node(&log_root->root_item, log_root->node);
Yan Zheng7237f1832009-01-21 12:54:03 -05001480
1481 WARN_ON(root->log_root);
1482 root->log_root = log_root;
1483 root->log_transid = 0;
Miao Xied1433de2014-02-20 18:08:59 +08001484 root->log_transid_committed = -1;
Chris Mason257c62e2009-10-13 13:21:08 -04001485 root->last_log_commit = 0;
Chris Masone02119d2008-09-05 16:13:11 -04001486 return 0;
1487}
1488
Stefan Behrens35a36212013-08-14 18:12:25 +02001489static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1490 struct btrfs_key *key)
Chris Masone02119d2008-09-05 16:13:11 -04001491{
1492 struct btrfs_root *root;
1493 struct btrfs_fs_info *fs_info = tree_root->fs_info;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001494 struct btrfs_path *path;
Yan Zheng84234f32008-10-29 14:49:05 -04001495 u64 generation;
Miao Xiecb517ea2013-05-15 07:48:19 +00001496 int ret;
1497
1498 path = btrfs_alloc_path();
1499 if (!path)
1500 return ERR_PTR(-ENOMEM);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001501
Al Viro6f07e422011-11-17 00:46:16 -05001502 root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00001503 if (!root) {
1504 ret = -ENOMEM;
1505 goto alloc_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001506 }
1507
David Sterba707e8a02014-06-04 19:22:26 +02001508 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1509 tree_root->stripesize, root, fs_info, key->objectid);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001510
Miao Xiecb517ea2013-05-15 07:48:19 +00001511 ret = btrfs_find_root(tree_root, key, path,
1512 &root->root_item, &root->root_key);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001513 if (ret) {
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001514 if (ret > 0)
1515 ret = -ENOENT;
Miao Xiecb517ea2013-05-15 07:48:19 +00001516 goto find_fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001517 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001518
Yan Zheng84234f32008-10-29 14:49:05 -04001519 generation = btrfs_root_generation(&root->root_item);
Chris Masondb945352007-10-15 16:15:53 -04001520 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
David Sterbace86cd52014-06-15 01:07:32 +02001521 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08001522 if (IS_ERR(root->node)) {
1523 ret = PTR_ERR(root->node);
Miao Xiecb517ea2013-05-15 07:48:19 +00001524 goto find_fail;
1525 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1526 ret = -EIO;
Liu Bo64c043d2015-05-25 17:30:15 +08001527 free_extent_buffer(root->node);
1528 goto find_fail;
Josef Bacik416bc652013-04-23 14:17:42 -04001529 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04001530 root->commit_root = btrfs_root_node(root);
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001531out:
Miao Xiecb517ea2013-05-15 07:48:19 +00001532 btrfs_free_path(path);
1533 return root;
1534
Miao Xiecb517ea2013-05-15 07:48:19 +00001535find_fail:
1536 kfree(root);
1537alloc_fail:
1538 root = ERR_PTR(ret);
1539 goto out;
1540}
1541
1542struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1543 struct btrfs_key *location)
1544{
1545 struct btrfs_root *root;
1546
1547 root = btrfs_read_tree_root(tree_root, location);
1548 if (IS_ERR(root))
1549 return root;
1550
1551 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
Miao Xie27cdeb72014-04-02 19:51:05 +08001552 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
Li Zefan08fe4db2011-03-28 02:01:25 +00001553 btrfs_check_and_init_root_item(&root->root_item);
1554 }
Yan, Zheng13a8a7c2009-09-21 15:56:00 -04001555
Chris Mason5eda7b52007-06-22 14:16:25 -04001556 return root;
1557}
1558
Miao Xiecb517ea2013-05-15 07:48:19 +00001559int btrfs_init_fs_root(struct btrfs_root *root)
1560{
1561 int ret;
Miao Xie8257b2d2014-03-06 13:38:19 +08001562 struct btrfs_subvolume_writers *writers;
Miao Xiecb517ea2013-05-15 07:48:19 +00001563
1564 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1565 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1566 GFP_NOFS);
1567 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1568 ret = -ENOMEM;
1569 goto fail;
1570 }
1571
Miao Xie8257b2d2014-03-06 13:38:19 +08001572 writers = btrfs_alloc_subvolume_writers();
1573 if (IS_ERR(writers)) {
1574 ret = PTR_ERR(writers);
1575 goto fail;
1576 }
1577 root->subv_writers = writers;
1578
Miao Xiecb517ea2013-05-15 07:48:19 +00001579 btrfs_init_free_ino_ctl(root);
David Sterba57cdc8d2014-02-05 02:37:48 +01001580 spin_lock_init(&root->ino_cache_lock);
1581 init_waitqueue_head(&root->ino_cache_wait);
Miao Xiecb517ea2013-05-15 07:48:19 +00001582
1583 ret = get_anon_bdev(&root->anon_dev);
1584 if (ret)
Miao Xie8257b2d2014-03-06 13:38:19 +08001585 goto free_writers;
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301586
1587 mutex_lock(&root->objectid_mutex);
1588 ret = btrfs_find_highest_objectid(root,
1589 &root->highest_objectid);
1590 if (ret) {
1591 mutex_unlock(&root->objectid_mutex);
1592 goto free_root_dev;
1593 }
1594
1595 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1596
1597 mutex_unlock(&root->objectid_mutex);
1598
Miao Xiecb517ea2013-05-15 07:48:19 +00001599 return 0;
Miao Xie8257b2d2014-03-06 13:38:19 +08001600
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05301601free_root_dev:
1602 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08001603free_writers:
1604 btrfs_free_subvolume_writers(root->subv_writers);
Miao Xiecb517ea2013-05-15 07:48:19 +00001605fail:
1606 kfree(root->free_ino_ctl);
1607 kfree(root->free_ino_pinned);
1608 return ret;
1609}
1610
Sergei Trofimovich171170c2013-08-14 23:27:46 +03001611static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1612 u64 root_id)
Miao Xiecb517ea2013-05-15 07:48:19 +00001613{
1614 struct btrfs_root *root;
1615
1616 spin_lock(&fs_info->fs_roots_radix_lock);
1617 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1618 (unsigned long)root_id);
1619 spin_unlock(&fs_info->fs_roots_radix_lock);
1620 return root;
1621}
1622
1623int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1624 struct btrfs_root *root)
1625{
1626 int ret;
1627
1628 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1629 if (ret)
1630 return ret;
1631
1632 spin_lock(&fs_info->fs_roots_radix_lock);
1633 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1634 (unsigned long)root->root_key.objectid,
1635 root);
1636 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001637 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
Miao Xiecb517ea2013-05-15 07:48:19 +00001638 spin_unlock(&fs_info->fs_roots_radix_lock);
1639 radix_tree_preload_end();
1640
1641 return ret;
1642}
1643
Miao Xiec00869f2013-09-25 21:47:44 +08001644struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1645 struct btrfs_key *location,
1646 bool check_ref)
Chris Mason5eda7b52007-06-22 14:16:25 -04001647{
1648 struct btrfs_root *root;
David Sterba381cf652015-01-02 18:45:16 +01001649 struct btrfs_path *path;
David Sterba1d4c08e2015-01-02 19:36:14 +01001650 struct btrfs_key key;
Chris Mason5eda7b52007-06-22 14:16:25 -04001651 int ret;
1652
Chris Masonedbd8d42007-12-21 16:27:24 -05001653 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1654 return fs_info->tree_root;
1655 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1656 return fs_info->extent_root;
Chris Mason8f18cf12008-04-25 16:53:30 -04001657 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1658 return fs_info->chunk_root;
1659 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1660 return fs_info->dev_root;
Yan Zheng0403e472008-12-10 20:32:51 -05001661 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1662 return fs_info->csum_root;
Arne Jansenbcef60f2011-09-13 15:23:30 +02001663 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1664 return fs_info->quota_root ? fs_info->quota_root :
1665 ERR_PTR(-ENOENT);
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02001666 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1667 return fs_info->uuid_root ? fs_info->uuid_root :
1668 ERR_PTR(-ENOENT);
Omar Sandoval70f6d822015-09-29 20:50:38 -07001669 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1670 return fs_info->free_space_root ? fs_info->free_space_root :
1671 ERR_PTR(-ENOENT);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001672again:
Miao Xiecb517ea2013-05-15 07:48:19 +00001673 root = btrfs_lookup_fs_root(fs_info, location->objectid);
Stefan Behrens48475472013-08-23 10:34:42 +02001674 if (root) {
Miao Xiec00869f2013-09-25 21:47:44 +08001675 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
Stefan Behrens48475472013-08-23 10:34:42 +02001676 return ERR_PTR(-ENOENT);
Chris Mason5eda7b52007-06-22 14:16:25 -04001677 return root;
Stefan Behrens48475472013-08-23 10:34:42 +02001678 }
Chris Mason5eda7b52007-06-22 14:16:25 -04001679
Miao Xiecb517ea2013-05-15 07:48:19 +00001680 root = btrfs_read_fs_root(fs_info->tree_root, location);
Chris Mason5eda7b52007-06-22 14:16:25 -04001681 if (IS_ERR(root))
1682 return root;
Chris Mason3394e162008-11-17 20:42:26 -05001683
Miao Xiec00869f2013-09-25 21:47:44 +08001684 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04001685 ret = -ENOENT;
1686 goto fail;
1687 }
1688
Miao Xiecb517ea2013-05-15 07:48:19 +00001689 ret = btrfs_init_fs_root(root);
1690 if (ret)
1691 goto fail;
1692
David Sterba381cf652015-01-02 18:45:16 +01001693 path = btrfs_alloc_path();
1694 if (!path) {
1695 ret = -ENOMEM;
1696 goto fail;
1697 }
David Sterba1d4c08e2015-01-02 19:36:14 +01001698 key.objectid = BTRFS_ORPHAN_OBJECTID;
1699 key.type = BTRFS_ORPHAN_ITEM_KEY;
1700 key.offset = location->objectid;
1701
1702 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
David Sterba381cf652015-01-02 18:45:16 +01001703 btrfs_free_path(path);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001704 if (ret < 0)
1705 goto fail;
1706 if (ret == 0)
Miao Xie27cdeb72014-04-02 19:51:05 +08001707 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
Yan, Zhengd68fc572010-05-16 10:49:58 -04001708
Miao Xiecb517ea2013-05-15 07:48:19 +00001709 ret = btrfs_insert_fs_root(fs_info, root);
Chris Mason0f7d52f2007-04-09 10:42:37 -04001710 if (ret) {
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001711 if (ret == -EEXIST) {
1712 free_fs_root(root);
1713 goto again;
1714 }
1715 goto fail;
Chris Mason0f7d52f2007-04-09 10:42:37 -04001716 }
Chris Masonedbd8d42007-12-21 16:27:24 -05001717 return root;
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04001718fail:
1719 free_fs_root(root);
1720 return ERR_PTR(ret);
Chris Masonedbd8d42007-12-21 16:27:24 -05001721}
1722
Chris Mason04160082008-03-26 10:28:07 -04001723static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1724{
1725 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1726 int ret = 0;
Chris Mason04160082008-03-26 10:28:07 -04001727 struct btrfs_device *device;
1728 struct backing_dev_info *bdi;
Chris Masonb7967db2009-04-27 07:29:04 -04001729
Xiao Guangrong1f781602011-04-20 10:09:16 +00001730 rcu_read_lock();
1731 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04001732 if (!device->bdev)
1733 continue;
Chris Mason04160082008-03-26 10:28:07 -04001734 bdi = blk_get_backing_dev_info(device->bdev);
Tejun Heoff9ea322014-09-08 08:03:56 +09001735 if (bdi_congested(bdi, bdi_bits)) {
Chris Mason04160082008-03-26 10:28:07 -04001736 ret = 1;
1737 break;
1738 }
1739 }
Xiao Guangrong1f781602011-04-20 10:09:16 +00001740 rcu_read_unlock();
Chris Mason04160082008-03-26 10:28:07 -04001741 return ret;
1742}
1743
Chris Mason04160082008-03-26 10:28:07 -04001744static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1745{
Jens Axboead081f12009-06-12 14:43:40 +02001746 int err;
1747
Christoph Hellwigb4caecd2015-01-14 10:42:32 +01001748 err = bdi_setup_and_register(bdi, "btrfs");
Jens Axboead081f12009-06-12 14:43:40 +02001749 if (err)
1750 return err;
1751
Christoph Hellwigdf0ce262015-01-14 10:42:41 +01001752 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE;
Chris Mason04160082008-03-26 10:28:07 -04001753 bdi->congested_fn = btrfs_congested_fn;
1754 bdi->congested_data = info;
Chris Masonda2f0f72015-07-02 13:57:22 -07001755 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
Chris Mason04160082008-03-26 10:28:07 -04001756 return 0;
1757}
1758
Chris Mason8b712842008-06-11 16:50:36 -04001759/*
1760 * called by the kthread helper functions to finally call the bio end_io
1761 * functions. This is where read checksum verification actually happens
1762 */
1763static void end_workqueue_fn(struct btrfs_work *work)
Chris Masonce9adaa2008-04-09 16:28:12 -04001764{
Chris Masonce9adaa2008-04-09 16:28:12 -04001765 struct bio *bio;
David Sterba97eb6b62014-07-30 00:55:42 +02001766 struct btrfs_end_io_wq *end_io_wq;
Chris Masonce9adaa2008-04-09 16:28:12 -04001767
David Sterba97eb6b62014-07-30 00:55:42 +02001768 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
Chris Mason8b712842008-06-11 16:50:36 -04001769 bio = end_io_wq->bio;
Chris Masonce9adaa2008-04-09 16:28:12 -04001770
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001771 bio->bi_error = end_io_wq->error;
Chris Mason8b712842008-06-11 16:50:36 -04001772 bio->bi_private = end_io_wq->private;
1773 bio->bi_end_io = end_io_wq->end_io;
David Sterba97eb6b62014-07-30 00:55:42 +02001774 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001775 bio_endio(bio);
Chris Mason44b8bd72008-04-16 11:14:51 -04001776}
1777
Chris Masona74a4b92008-06-25 16:01:31 -04001778static int cleaner_kthread(void *arg)
1779{
1780 struct btrfs_root *root = arg;
Miao Xied0278242013-05-14 10:20:40 +00001781 int again;
Filipe Mananada288d22015-06-13 06:55:31 +01001782 struct btrfs_trans_handle *trans;
Chris Masona74a4b92008-06-25 16:01:31 -04001783
Jiri Kosina69624912015-10-26 15:06:19 +09001784 set_freezable();
Chris Masona74a4b92008-06-25 16:01:31 -04001785 do {
Miao Xied0278242013-05-14 10:20:40 +00001786 again = 0;
David Sterba9d1a2a32013-03-12 15:13:28 +00001787
Miao Xied0278242013-05-14 10:20:40 +00001788 /* Make the cleaner go to sleep early. */
Miao Xiebabbf1702013-05-14 10:20:43 +00001789 if (btrfs_need_cleaner_sleep(root))
Miao Xied0278242013-05-14 10:20:40 +00001790 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001791
Miao Xied0278242013-05-14 10:20:40 +00001792 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1793 goto sleep;
1794
Miao Xiedc7f3702013-05-14 10:20:42 +00001795 /*
1796 * Avoid the problem that we change the status of the fs
1797 * during the above check and trylock.
1798 */
Miao Xiebabbf1702013-05-14 10:20:43 +00001799 if (btrfs_need_cleaner_sleep(root)) {
Miao Xiedc7f3702013-05-14 10:20:42 +00001800 mutex_unlock(&root->fs_info->cleaner_mutex);
1801 goto sleep;
Chris Masona74a4b92008-06-25 16:01:31 -04001802 }
1803
Miao Xied0278242013-05-14 10:20:40 +00001804 btrfs_run_delayed_iputs(root);
1805 again = btrfs_clean_one_deleted_snapshot(root);
1806 mutex_unlock(&root->fs_info->cleaner_mutex);
1807
1808 /*
Miao Xie05323cd2013-05-14 10:20:41 +00001809 * The defragger has dealt with the R/O remount and umount,
1810 * needn't do anything special here.
Miao Xied0278242013-05-14 10:20:40 +00001811 */
1812 btrfs_run_defrag_inodes(root->fs_info);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01001813
1814 /*
1815 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1816 * with relocation (btrfs_relocate_chunk) and relocation
1817 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1818 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1819 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1820 * unused block groups.
1821 */
1822 btrfs_delete_unused_bgs(root->fs_info);
Miao Xied0278242013-05-14 10:20:40 +00001823sleep:
David Sterba9d1a2a32013-03-12 15:13:28 +00001824 if (!try_to_freeze() && !again) {
Chris Masona74a4b92008-06-25 16:01:31 -04001825 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001826 if (!kthread_should_stop())
1827 schedule();
Chris Masona74a4b92008-06-25 16:01:31 -04001828 __set_current_state(TASK_RUNNING);
1829 }
1830 } while (!kthread_should_stop());
Filipe Mananada288d22015-06-13 06:55:31 +01001831
1832 /*
1833 * Transaction kthread is stopped before us and wakes us up.
1834 * However we might have started a new transaction and COWed some
1835 * tree blocks when deleting unused block groups for example. So
1836 * make sure we commit the transaction we started to have a clean
1837 * shutdown when evicting the btree inode - if it has dirty pages
1838 * when we do the final iput() on it, eviction will trigger a
1839 * writeback for it which will fail with null pointer dereferences
1840 * since work queues and other resources were already released and
1841 * destroyed by the time the iput/eviction/writeback is made.
1842 */
1843 trans = btrfs_attach_transaction(root);
1844 if (IS_ERR(trans)) {
1845 if (PTR_ERR(trans) != -ENOENT)
1846 btrfs_err(root->fs_info,
1847 "cleaner transaction attach returned %ld",
1848 PTR_ERR(trans));
1849 } else {
1850 int ret;
1851
1852 ret = btrfs_commit_transaction(trans, root);
1853 if (ret)
1854 btrfs_err(root->fs_info,
1855 "cleaner open transaction commit returned %d",
1856 ret);
1857 }
1858
Chris Masona74a4b92008-06-25 16:01:31 -04001859 return 0;
1860}
1861
1862static int transaction_kthread(void *arg)
1863{
1864 struct btrfs_root *root = arg;
1865 struct btrfs_trans_handle *trans;
1866 struct btrfs_transaction *cur;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001867 u64 transid;
Chris Masona74a4b92008-06-25 16:01:31 -04001868 unsigned long now;
1869 unsigned long delay;
Jan Kara914b2002012-03-12 16:05:50 +01001870 bool cannot_commit;
Chris Masona74a4b92008-06-25 16:01:31 -04001871
1872 do {
Jan Kara914b2002012-03-12 16:05:50 +01001873 cannot_commit = false;
David Sterba8b87dc12013-08-01 18:14:52 +02001874 delay = HZ * root->fs_info->commit_interval;
Chris Masona74a4b92008-06-25 16:01:31 -04001875 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1876
Josef Bacika4abeea2011-04-11 17:25:13 -04001877 spin_lock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001878 cur = root->fs_info->running_transaction;
1879 if (!cur) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001880 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001881 goto sleep;
1882 }
Yan Zheng31153d82008-07-28 15:32:19 -04001883
Chris Masona74a4b92008-06-25 16:01:31 -04001884 now = get_seconds();
Miao Xie4a9d8bd2013-05-17 03:53:43 +00001885 if (cur->state < TRANS_STATE_BLOCKED &&
David Sterba8b87dc12013-08-01 18:14:52 +02001886 (now < cur->start_time ||
1887 now - cur->start_time < root->fs_info->commit_interval)) {
Josef Bacika4abeea2011-04-11 17:25:13 -04001888 spin_unlock(&root->fs_info->trans_lock);
Chris Masona74a4b92008-06-25 16:01:31 -04001889 delay = HZ * 5;
1890 goto sleep;
1891 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001892 transid = cur->transid;
Josef Bacika4abeea2011-04-11 17:25:13 -04001893 spin_unlock(&root->fs_info->trans_lock);
Chris Mason56bec292009-03-13 10:10:06 -04001894
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001895 /* If the file system is aborted, this will always fail. */
Miao Xie354aa0f2012-09-20 01:54:00 -06001896 trans = btrfs_attach_transaction(root);
Jan Kara914b2002012-03-12 16:05:50 +01001897 if (IS_ERR(trans)) {
Miao Xie354aa0f2012-09-20 01:54:00 -06001898 if (PTR_ERR(trans) != -ENOENT)
1899 cannot_commit = true;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001900 goto sleep;
Jan Kara914b2002012-03-12 16:05:50 +01001901 }
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001902 if (transid == trans->transid) {
Jeff Mahoney79787ea2012-03-12 16:03:00 +01001903 btrfs_commit_transaction(trans, root);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001904 } else {
1905 btrfs_end_transaction(trans, root);
1906 }
Chris Masona74a4b92008-06-25 16:01:31 -04001907sleep:
1908 wake_up_process(root->fs_info->cleaner_kthread);
1909 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1910
Josef Bacik4e121c02013-09-27 16:32:39 -04001911 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1912 &root->fs_info->fs_state)))
1913 btrfs_cleanup_transaction(root);
Tejun Heoa0acae02011-11-21 12:32:22 -08001914 if (!try_to_freeze()) {
Chris Masona74a4b92008-06-25 16:01:31 -04001915 set_current_state(TASK_INTERRUPTIBLE);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001916 if (!kthread_should_stop() &&
Jan Kara914b2002012-03-12 16:05:50 +01001917 (!btrfs_transaction_blocked(root->fs_info) ||
1918 cannot_commit))
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04001919 schedule_timeout(delay);
Chris Masona74a4b92008-06-25 16:01:31 -04001920 __set_current_state(TASK_RUNNING);
1921 }
1922 } while (!kthread_should_stop());
1923 return 0;
1924}
1925
Chris Masonaf31f5e2011-11-03 15:17:42 -04001926/*
1927 * this will find the highest generation in the array of
1928 * root backups. The index of the highest array is returned,
1929 * or -1 if we can't find anything.
1930 *
1931 * We check to make sure the array is valid by comparing the
1932 * generation of the latest root in the array with the generation
1933 * in the super block. If they don't match we pitch it.
1934 */
1935static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1936{
1937 u64 cur;
1938 int newest_index = -1;
1939 struct btrfs_root_backup *root_backup;
1940 int i;
1941
1942 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1943 root_backup = info->super_copy->super_roots + i;
1944 cur = btrfs_backup_tree_root_gen(root_backup);
1945 if (cur == newest_gen)
1946 newest_index = i;
1947 }
1948
1949 /* check to see if we actually wrapped around */
1950 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1951 root_backup = info->super_copy->super_roots;
1952 cur = btrfs_backup_tree_root_gen(root_backup);
1953 if (cur == newest_gen)
1954 newest_index = 0;
1955 }
1956 return newest_index;
1957}
1958
1959
1960/*
1961 * find the oldest backup so we know where to store new entries
1962 * in the backup array. This will set the backup_root_index
1963 * field in the fs_info struct
1964 */
1965static void find_oldest_super_backup(struct btrfs_fs_info *info,
1966 u64 newest_gen)
1967{
1968 int newest_index = -1;
1969
1970 newest_index = find_newest_super_backup(info, newest_gen);
1971 /* if there was garbage in there, just move along */
1972 if (newest_index == -1) {
1973 info->backup_root_index = 0;
1974 } else {
1975 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1976 }
1977}
1978
1979/*
1980 * copy all the root pointers into the super backup array.
1981 * this will bump the backup pointer by one when it is
1982 * done
1983 */
1984static void backup_super_roots(struct btrfs_fs_info *info)
1985{
1986 int next_backup;
1987 struct btrfs_root_backup *root_backup;
1988 int last_backup;
1989
1990 next_backup = info->backup_root_index;
1991 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1992 BTRFS_NUM_BACKUP_ROOTS;
1993
1994 /*
1995 * just overwrite the last backup if we're at the same generation
1996 * this happens only at umount
1997 */
1998 root_backup = info->super_for_commit->super_roots + last_backup;
1999 if (btrfs_backup_tree_root_gen(root_backup) ==
2000 btrfs_header_generation(info->tree_root->node))
2001 next_backup = last_backup;
2002
2003 root_backup = info->super_for_commit->super_roots + next_backup;
2004
2005 /*
2006 * make sure all of our padding and empty slots get zero filled
2007 * regardless of which ones we use today
2008 */
2009 memset(root_backup, 0, sizeof(*root_backup));
2010
2011 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
2012
2013 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
2014 btrfs_set_backup_tree_root_gen(root_backup,
2015 btrfs_header_generation(info->tree_root->node));
2016
2017 btrfs_set_backup_tree_root_level(root_backup,
2018 btrfs_header_level(info->tree_root->node));
2019
2020 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
2021 btrfs_set_backup_chunk_root_gen(root_backup,
2022 btrfs_header_generation(info->chunk_root->node));
2023 btrfs_set_backup_chunk_root_level(root_backup,
2024 btrfs_header_level(info->chunk_root->node));
2025
2026 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2027 btrfs_set_backup_extent_root_gen(root_backup,
2028 btrfs_header_generation(info->extent_root->node));
2029 btrfs_set_backup_extent_root_level(root_backup,
2030 btrfs_header_level(info->extent_root->node));
2031
Chris Mason7c7e82a2011-11-06 18:50:56 -05002032 /*
2033 * we might commit during log recovery, which happens before we set
2034 * the fs_root. Make sure it is valid before we fill it in.
2035 */
2036 if (info->fs_root && info->fs_root->node) {
2037 btrfs_set_backup_fs_root(root_backup,
2038 info->fs_root->node->start);
2039 btrfs_set_backup_fs_root_gen(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002040 btrfs_header_generation(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002041 btrfs_set_backup_fs_root_level(root_backup,
Chris Masonaf31f5e2011-11-03 15:17:42 -04002042 btrfs_header_level(info->fs_root->node));
Chris Mason7c7e82a2011-11-06 18:50:56 -05002043 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002044
2045 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2046 btrfs_set_backup_dev_root_gen(root_backup,
2047 btrfs_header_generation(info->dev_root->node));
2048 btrfs_set_backup_dev_root_level(root_backup,
2049 btrfs_header_level(info->dev_root->node));
2050
2051 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2052 btrfs_set_backup_csum_root_gen(root_backup,
2053 btrfs_header_generation(info->csum_root->node));
2054 btrfs_set_backup_csum_root_level(root_backup,
2055 btrfs_header_level(info->csum_root->node));
2056
2057 btrfs_set_backup_total_bytes(root_backup,
2058 btrfs_super_total_bytes(info->super_copy));
2059 btrfs_set_backup_bytes_used(root_backup,
2060 btrfs_super_bytes_used(info->super_copy));
2061 btrfs_set_backup_num_devices(root_backup,
2062 btrfs_super_num_devices(info->super_copy));
2063
2064 /*
2065 * if we don't copy this out to the super_copy, it won't get remembered
2066 * for the next commit
2067 */
2068 memcpy(&info->super_copy->super_roots,
2069 &info->super_for_commit->super_roots,
2070 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2071}
2072
2073/*
2074 * this copies info out of the root backup array and back into
2075 * the in-memory super block. It is meant to help iterate through
2076 * the array, so you send it the number of backups you've already
2077 * tried and the last backup index you used.
2078 *
2079 * this returns -1 when it has tried all the backups
2080 */
2081static noinline int next_root_backup(struct btrfs_fs_info *info,
2082 struct btrfs_super_block *super,
2083 int *num_backups_tried, int *backup_index)
2084{
2085 struct btrfs_root_backup *root_backup;
2086 int newest = *backup_index;
2087
2088 if (*num_backups_tried == 0) {
2089 u64 gen = btrfs_super_generation(super);
2090
2091 newest = find_newest_super_backup(info, gen);
2092 if (newest == -1)
2093 return -1;
2094
2095 *backup_index = newest;
2096 *num_backups_tried = 1;
2097 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2098 /* we've tried all the backups, all done */
2099 return -1;
2100 } else {
2101 /* jump to the next oldest backup */
2102 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2103 BTRFS_NUM_BACKUP_ROOTS;
2104 *backup_index = newest;
2105 *num_backups_tried += 1;
2106 }
2107 root_backup = super->super_roots + newest;
2108
2109 btrfs_set_super_generation(super,
2110 btrfs_backup_tree_root_gen(root_backup));
2111 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2112 btrfs_set_super_root_level(super,
2113 btrfs_backup_tree_root_level(root_backup));
2114 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2115
2116 /*
2117 * fixme: the total bytes and num_devices need to match or we should
2118 * need a fsck
2119 */
2120 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2121 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2122 return 0;
2123}
2124
Liu Bo7abadb62013-03-17 02:10:31 +00002125/* helper to cleanup workers */
2126static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2127{
Qu Wenruodc6e3202014-02-28 10:46:14 +08002128 btrfs_destroy_workqueue(fs_info->fixup_workers);
Qu Wenruoafe3d242014-02-28 10:46:07 +08002129 btrfs_destroy_workqueue(fs_info->delalloc_workers);
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002130 btrfs_destroy_workqueue(fs_info->workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002131 btrfs_destroy_workqueue(fs_info->endio_workers);
2132 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2133 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
Miao Xie8b110e32014-09-12 18:44:03 +08002134 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
Qu Wenruod05a33a2014-02-28 10:46:11 +08002135 btrfs_destroy_workqueue(fs_info->rmw_workers);
Qu Wenruofccb5d82014-02-28 10:46:10 +08002136 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2137 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2138 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
Qu Wenruoa8c93d42014-02-28 10:46:08 +08002139 btrfs_destroy_workqueue(fs_info->submit_workers);
Qu Wenruo5b3bc442014-02-28 10:46:15 +08002140 btrfs_destroy_workqueue(fs_info->delayed_workers);
Qu Wenruoe66f0bb2014-02-28 10:46:12 +08002141 btrfs_destroy_workqueue(fs_info->caching_workers);
Qu Wenruo736cfa12014-02-28 10:46:13 +08002142 btrfs_destroy_workqueue(fs_info->readahead_workers);
Qu Wenruoa44903a2014-02-28 10:46:09 +08002143 btrfs_destroy_workqueue(fs_info->flush_workers);
Qu Wenruofc97fab2014-02-28 10:46:16 +08002144 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
Chris Masona79b7d42014-05-22 16:18:52 -07002145 btrfs_destroy_workqueue(fs_info->extent_workers);
Liu Bo7abadb62013-03-17 02:10:31 +00002146}
2147
Rashika2e9f5952013-10-31 02:45:20 +05302148static void free_root_extent_buffers(struct btrfs_root *root)
2149{
2150 if (root) {
2151 free_extent_buffer(root->node);
2152 free_extent_buffer(root->commit_root);
2153 root->node = NULL;
2154 root->commit_root = NULL;
2155 }
2156}
2157
Chris Masonaf31f5e2011-11-03 15:17:42 -04002158/* helper to cleanup tree roots */
2159static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2160{
Rashika2e9f5952013-10-31 02:45:20 +05302161 free_root_extent_buffers(info->tree_root);
Josef Bacik655b09f2013-05-17 14:06:51 -04002162
Rashika2e9f5952013-10-31 02:45:20 +05302163 free_root_extent_buffers(info->dev_root);
2164 free_root_extent_buffers(info->extent_root);
2165 free_root_extent_buffers(info->csum_root);
2166 free_root_extent_buffers(info->quota_root);
2167 free_root_extent_buffers(info->uuid_root);
2168 if (chunk_root)
2169 free_root_extent_buffers(info->chunk_root);
Omar Sandoval70f6d822015-09-29 20:50:38 -07002170 free_root_extent_buffers(info->free_space_root);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002171}
2172
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04002173void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
Josef Bacik171f6532013-04-24 16:35:41 -04002174{
2175 int ret;
2176 struct btrfs_root *gang[8];
2177 int i;
2178
2179 while (!list_empty(&fs_info->dead_roots)) {
2180 gang[0] = list_entry(fs_info->dead_roots.next,
2181 struct btrfs_root, root_list);
2182 list_del(&gang[0]->root_list);
2183
Miao Xie27cdeb72014-04-02 19:51:05 +08002184 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
Miao Xiecb517ea2013-05-15 07:48:19 +00002185 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002186 } else {
2187 free_extent_buffer(gang[0]->node);
2188 free_extent_buffer(gang[0]->commit_root);
Miao Xieb0feb9d2013-05-15 07:48:20 +00002189 btrfs_put_fs_root(gang[0]);
Josef Bacik171f6532013-04-24 16:35:41 -04002190 }
2191 }
2192
2193 while (1) {
2194 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2195 (void **)gang, 0,
2196 ARRAY_SIZE(gang));
2197 if (!ret)
2198 break;
2199 for (i = 0; i < ret; i++)
Miao Xiecb517ea2013-05-15 07:48:19 +00002200 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
Josef Bacik171f6532013-04-24 16:35:41 -04002201 }
Liu Bo1a4319c2014-01-13 19:53:53 +08002202
2203 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2204 btrfs_free_log_root_tree(NULL, fs_info);
2205 btrfs_destroy_pinned_extent(fs_info->tree_root,
2206 fs_info->pinned_extents);
2207 }
Josef Bacik171f6532013-04-24 16:35:41 -04002208}
Chris Masonaf31f5e2011-11-03 15:17:42 -04002209
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002210static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2211{
2212 mutex_init(&fs_info->scrub_lock);
2213 atomic_set(&fs_info->scrubs_running, 0);
2214 atomic_set(&fs_info->scrub_pause_req, 0);
2215 atomic_set(&fs_info->scrubs_paused, 0);
2216 atomic_set(&fs_info->scrub_cancel_req, 0);
2217 init_waitqueue_head(&fs_info->scrub_pause_wait);
2218 fs_info->scrub_workers_refcnt = 0;
2219}
2220
Eric Sandeen779a65a2014-08-01 18:12:39 -05002221static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2222{
2223 spin_lock_init(&fs_info->balance_lock);
2224 mutex_init(&fs_info->balance_mutex);
2225 atomic_set(&fs_info->balance_running, 0);
2226 atomic_set(&fs_info->balance_pause_req, 0);
2227 atomic_set(&fs_info->balance_cancel_req, 0);
2228 fs_info->balance_ctl = NULL;
2229 init_waitqueue_head(&fs_info->balance_wait_q);
2230}
2231
Eric Sandeenf37938e2014-08-01 18:12:40 -05002232static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2233 struct btrfs_root *tree_root)
2234{
2235 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2236 set_nlink(fs_info->btree_inode, 1);
2237 /*
2238 * we set the i_size on the btree inode to the max possible int.
2239 * the real end of the address space is determined by all of
2240 * the devices in the system
2241 */
2242 fs_info->btree_inode->i_size = OFFSET_MAX;
2243 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
Eric Sandeenf37938e2014-08-01 18:12:40 -05002244
2245 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2246 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2247 fs_info->btree_inode->i_mapping);
2248 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2249 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2250
2251 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2252
2253 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2254 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2255 sizeof(struct btrfs_key));
2256 set_bit(BTRFS_INODE_DUMMY,
2257 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2258 btrfs_insert_inode_hash(fs_info->btree_inode);
2259}
2260
Eric Sandeenad618362014-08-01 18:12:41 -05002261static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2262{
2263 fs_info->dev_replace.lock_owner = 0;
2264 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2265 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2266 mutex_init(&fs_info->dev_replace.lock_management_lock);
2267 mutex_init(&fs_info->dev_replace.lock);
2268 init_waitqueue_head(&fs_info->replace_wait);
2269}
2270
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002271static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2272{
2273 spin_lock_init(&fs_info->qgroup_lock);
2274 mutex_init(&fs_info->qgroup_ioctl_lock);
2275 fs_info->qgroup_tree = RB_ROOT;
2276 fs_info->qgroup_op_tree = RB_ROOT;
2277 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2278 fs_info->qgroup_seq = 1;
2279 fs_info->quota_enabled = 0;
2280 fs_info->pending_quota_state = 0;
2281 fs_info->qgroup_ulist = NULL;
2282 mutex_init(&fs_info->qgroup_rescan_lock);
2283}
2284
Eric Sandeen2a458192015-02-16 16:29:26 +01002285static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2286 struct btrfs_fs_devices *fs_devices)
2287{
2288 int max_active = fs_info->thread_pool_size;
David Sterba6f011052015-02-16 18:34:01 +01002289 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
Eric Sandeen2a458192015-02-16 16:29:26 +01002290
2291 fs_info->workers =
2292 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2293 max_active, 16);
2294
2295 fs_info->delalloc_workers =
2296 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
2297
2298 fs_info->flush_workers =
2299 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
2300
2301 fs_info->caching_workers =
2302 btrfs_alloc_workqueue("cache", flags, max_active, 0);
2303
2304 /*
2305 * a higher idle thresh on the submit workers makes it much more
2306 * likely that bios will be send down in a sane order to the
2307 * devices
2308 */
2309 fs_info->submit_workers =
2310 btrfs_alloc_workqueue("submit", flags,
2311 min_t(u64, fs_devices->num_devices,
2312 max_active), 64);
2313
2314 fs_info->fixup_workers =
2315 btrfs_alloc_workqueue("fixup", flags, 1, 0);
2316
2317 /*
2318 * endios are largely parallel and should have a very
2319 * low idle thresh
2320 */
2321 fs_info->endio_workers =
2322 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2323 fs_info->endio_meta_workers =
2324 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2325 fs_info->endio_meta_write_workers =
2326 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2327 fs_info->endio_raid56_workers =
2328 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
2329 fs_info->endio_repair_workers =
2330 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
2331 fs_info->rmw_workers =
2332 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
2333 fs_info->endio_write_workers =
2334 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2335 fs_info->endio_freespace_worker =
2336 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
2337 fs_info->delayed_workers =
2338 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
2339 fs_info->readahead_workers =
2340 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
2341 fs_info->qgroup_rescan_workers =
2342 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
2343 fs_info->extent_workers =
2344 btrfs_alloc_workqueue("extent-refs", flags,
2345 min_t(u64, fs_devices->num_devices,
2346 max_active), 8);
2347
2348 if (!(fs_info->workers && fs_info->delalloc_workers &&
2349 fs_info->submit_workers && fs_info->flush_workers &&
2350 fs_info->endio_workers && fs_info->endio_meta_workers &&
2351 fs_info->endio_meta_write_workers &&
2352 fs_info->endio_repair_workers &&
2353 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2354 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2355 fs_info->caching_workers && fs_info->readahead_workers &&
2356 fs_info->fixup_workers && fs_info->delayed_workers &&
2357 fs_info->extent_workers &&
2358 fs_info->qgroup_rescan_workers)) {
2359 return -ENOMEM;
2360 }
2361
2362 return 0;
2363}
2364
Eric Sandeen63443bf2014-08-01 18:12:46 -05002365static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2366 struct btrfs_fs_devices *fs_devices)
2367{
2368 int ret;
2369 struct btrfs_root *tree_root = fs_info->tree_root;
2370 struct btrfs_root *log_tree_root;
2371 struct btrfs_super_block *disk_super = fs_info->super_copy;
2372 u64 bytenr = btrfs_super_log_root(disk_super);
2373
2374 if (fs_devices->rw_devices == 0) {
David Sterbaf14d1042015-10-08 11:37:06 +02002375 btrfs_warn(fs_info, "log replay required on RO media");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002376 return -EIO;
2377 }
2378
2379 log_tree_root = btrfs_alloc_root(fs_info);
2380 if (!log_tree_root)
2381 return -ENOMEM;
2382
2383 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2384 tree_root->stripesize, log_tree_root, fs_info,
2385 BTRFS_TREE_LOG_OBJECTID);
2386
2387 log_tree_root->node = read_tree_block(tree_root, bytenr,
2388 fs_info->generation + 1);
Liu Bo64c043d2015-05-25 17:30:15 +08002389 if (IS_ERR(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002390 btrfs_warn(fs_info, "failed to read log tree");
Liu Bo0eeff232015-06-11 14:16:44 +08002391 ret = PTR_ERR(log_tree_root->node);
Liu Bo64c043d2015-05-25 17:30:15 +08002392 kfree(log_tree_root);
Liu Bo0eeff232015-06-11 14:16:44 +08002393 return ret;
Liu Bo64c043d2015-05-25 17:30:15 +08002394 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
David Sterbaf14d1042015-10-08 11:37:06 +02002395 btrfs_err(fs_info, "failed to read log tree");
Eric Sandeen63443bf2014-08-01 18:12:46 -05002396 free_extent_buffer(log_tree_root->node);
2397 kfree(log_tree_root);
2398 return -EIO;
2399 }
2400 /* returns with log_tree_root freed on success */
2401 ret = btrfs_recover_log_trees(log_tree_root);
2402 if (ret) {
Anand Jaina4553fe2015-09-25 14:43:01 +08002403 btrfs_std_error(tree_root->fs_info, ret,
Eric Sandeen63443bf2014-08-01 18:12:46 -05002404 "Failed to recover log tree");
2405 free_extent_buffer(log_tree_root->node);
2406 kfree(log_tree_root);
2407 return ret;
2408 }
2409
2410 if (fs_info->sb->s_flags & MS_RDONLY) {
2411 ret = btrfs_commit_super(tree_root);
2412 if (ret)
2413 return ret;
2414 }
2415
2416 return 0;
2417}
2418
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002419static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2420 struct btrfs_root *tree_root)
2421{
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002422 struct btrfs_root *root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002423 struct btrfs_key location;
2424 int ret;
2425
2426 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2427 location.type = BTRFS_ROOT_ITEM_KEY;
2428 location.offset = 0;
2429
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002430 root = btrfs_read_tree_root(tree_root, &location);
2431 if (IS_ERR(root))
2432 return PTR_ERR(root);
2433 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2434 fs_info->extent_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002435
2436 location.objectid = BTRFS_DEV_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002437 root = btrfs_read_tree_root(tree_root, &location);
2438 if (IS_ERR(root))
2439 return PTR_ERR(root);
2440 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2441 fs_info->dev_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002442 btrfs_init_devices_late(fs_info);
2443
2444 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002445 root = btrfs_read_tree_root(tree_root, &location);
2446 if (IS_ERR(root))
2447 return PTR_ERR(root);
2448 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2449 fs_info->csum_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002450
2451 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002452 root = btrfs_read_tree_root(tree_root, &location);
2453 if (!IS_ERR(root)) {
2454 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002455 fs_info->quota_enabled = 1;
2456 fs_info->pending_quota_state = 1;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002457 fs_info->quota_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002458 }
2459
2460 location.objectid = BTRFS_UUID_TREE_OBJECTID;
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002461 root = btrfs_read_tree_root(tree_root, &location);
2462 if (IS_ERR(root)) {
2463 ret = PTR_ERR(root);
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002464 if (ret != -ENOENT)
2465 return ret;
2466 } else {
David Sterbaa4f3d2c2015-02-16 18:44:34 +01002467 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2468 fs_info->uuid_root = root;
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002469 }
2470
Omar Sandoval70f6d822015-09-29 20:50:38 -07002471 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2472 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2473 root = btrfs_read_tree_root(tree_root, &location);
2474 if (IS_ERR(root))
2475 return PTR_ERR(root);
2476 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2477 fs_info->free_space_root = root;
2478 }
2479
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002480 return 0;
2481}
2482
Al Viroad2b2c82011-11-17 01:10:02 -05002483int open_ctree(struct super_block *sb,
2484 struct btrfs_fs_devices *fs_devices,
2485 char *options)
Chris Masoneb60cea2007-02-02 09:18:22 -05002486{
Chris Masondb945352007-10-15 16:15:53 -04002487 u32 sectorsize;
2488 u32 nodesize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002489 u32 stripesize;
Yan Zheng84234f32008-10-29 14:49:05 -04002490 u64 generation;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002491 u64 features;
Chris Mason3de45862008-11-17 21:02:50 -05002492 struct btrfs_key location;
Chris Masona061fc82008-05-07 11:43:44 -04002493 struct buffer_head *bh;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002494 struct btrfs_super_block *disk_super;
Al Viro815745c2011-11-17 15:40:49 -05002495 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
Al Virof84a8bd2011-11-17 15:57:57 -05002496 struct btrfs_root *tree_root;
Ilya Dryomov4d34b272011-11-09 00:08:15 +02002497 struct btrfs_root *chunk_root;
Chris Masoneb60cea2007-02-02 09:18:22 -05002498 int ret;
Yane58ca022008-04-01 11:21:34 -04002499 int err = -EINVAL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002500 int num_backups_tried = 0;
2501 int backup_index = 0;
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002502 int max_active;
Chris Mason4543df72008-06-11 21:47:56 -04002503
Al Virof84a8bd2011-11-17 15:57:57 -05002504 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
Al Viro6f07e422011-11-17 00:46:16 -05002505 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
Miao Xiecb517ea2013-05-15 07:48:19 +00002506 if (!tree_root || !chunk_root) {
Chris Mason39279cc2007-06-12 06:35:45 -04002507 err = -ENOMEM;
2508 goto fail;
2509 }
Yan, Zheng76dda932009-09-21 16:00:26 -04002510
2511 ret = init_srcu_struct(&fs_info->subvol_srcu);
2512 if (ret) {
2513 err = ret;
2514 goto fail;
2515 }
2516
2517 ret = setup_bdi(fs_info, &fs_info->bdi);
2518 if (ret) {
2519 err = ret;
2520 goto fail_srcu;
2521 }
2522
Tejun Heo908c7f12014-09-08 09:51:29 +09002523 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
Miao Xiee2d84522013-01-29 10:09:20 +00002524 if (ret) {
2525 err = ret;
2526 goto fail_bdi;
2527 }
2528 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2529 (1 + ilog2(nr_cpu_ids));
2530
Tejun Heo908c7f12014-09-08 09:51:29 +09002531 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
Miao Xie963d6782013-01-29 10:10:51 +00002532 if (ret) {
2533 err = ret;
2534 goto fail_dirty_metadata_bytes;
2535 }
2536
Tejun Heo908c7f12014-09-08 09:51:29 +09002537 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
Miao Xiec404e0d2014-01-30 16:46:55 +08002538 if (ret) {
2539 err = ret;
2540 goto fail_delalloc_bytes;
2541 }
2542
Yan, Zheng76dda932009-09-21 16:00:26 -04002543 fs_info->btree_inode = new_inode(sb);
2544 if (!fs_info->btree_inode) {
2545 err = -ENOMEM;
Miao Xiec404e0d2014-01-30 16:46:55 +08002546 goto fail_bio_counter;
Yan, Zheng76dda932009-09-21 16:00:26 -04002547 }
2548
Chris Masona6591712011-07-19 12:04:14 -04002549 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
Miao Xie1561ded2011-03-27 16:07:36 +08002550
Yan, Zheng76dda932009-09-21 16:00:26 -04002551 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
Josef Bacikf28491e2013-12-16 13:24:27 -05002552 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
Chris Mason8fd17792007-04-19 21:01:03 -04002553 INIT_LIST_HEAD(&fs_info->trans_list);
Chris Masonfacda1e2007-06-08 18:11:48 -04002554 INIT_LIST_HEAD(&fs_info->dead_roots);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002555 INIT_LIST_HEAD(&fs_info->delayed_iputs);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002556 INIT_LIST_HEAD(&fs_info->delalloc_roots);
Yan Zheng11833d62009-09-11 16:11:19 -04002557 INIT_LIST_HEAD(&fs_info->caching_block_groups);
Miao Xieeb73c1b2013-05-15 07:48:22 +00002558 spin_lock_init(&fs_info->delalloc_root_lock);
Josef Bacika4abeea2011-04-11 17:25:13 -04002559 spin_lock_init(&fs_info->trans_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04002560 spin_lock_init(&fs_info->fs_roots_radix_lock);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00002561 spin_lock_init(&fs_info->delayed_iput_lock);
Chris Mason4cb53002011-05-24 15:35:30 -04002562 spin_lock_init(&fs_info->defrag_inodes_lock);
Josef Bacik2bf64752011-09-26 17:12:22 -04002563 spin_lock_init(&fs_info->free_chunk_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002564 spin_lock_init(&fs_info->tree_mod_seq_lock);
Miao Xieceda0862013-04-11 10:30:16 +00002565 spin_lock_init(&fs_info->super_lock);
Josef Bacikfcebe452014-05-13 17:30:47 -07002566 spin_lock_init(&fs_info->qgroup_op_lock);
Josef Bacikf28491e2013-12-16 13:24:27 -05002567 spin_lock_init(&fs_info->buffer_lock);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002568 spin_lock_init(&fs_info->unused_bgs_lock);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002569 rwlock_init(&fs_info->tree_mod_log_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002570 mutex_init(&fs_info->unused_bg_unpin_mutex);
Filipe Manana67c5e7d2015-06-11 00:58:53 +01002571 mutex_init(&fs_info->delete_unused_bgs_mutex);
Chris Mason75857172011-06-13 20:00:16 -04002572 mutex_init(&fs_info->reloc_mutex);
Miao Xie573bfb72014-03-06 13:55:03 +08002573 mutex_init(&fs_info->delalloc_root_mutex);
Miao Xiede98ced2013-01-29 10:13:12 +00002574 seqlock_init(&fs_info->profiles_lock);
Zhao Leid7c15172015-02-26 10:49:20 +08002575 init_rwsem(&fs_info->delayed_iput_sem);
Chris Mason19c00dd2007-10-15 16:19:22 -04002576
Chris Mason0b86a832008-03-24 15:01:56 -04002577 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
Chris Mason6324fbf2008-03-24 15:01:59 -04002578 INIT_LIST_HEAD(&fs_info->space_info);
Jan Schmidtf29021b2012-05-16 17:55:38 +02002579 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
Josef Bacik47ab2a62014-09-18 11:20:02 -04002580 INIT_LIST_HEAD(&fs_info->unused_bgs);
Chris Mason0b86a832008-03-24 15:01:56 -04002581 btrfs_mapping_init(&fs_info->mapping_tree);
Miao Xie66d8f3d2012-09-06 04:02:28 -06002582 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2583 BTRFS_BLOCK_RSV_GLOBAL);
2584 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2585 BTRFS_BLOCK_RSV_DELALLOC);
2586 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2587 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2588 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2589 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2590 BTRFS_BLOCK_RSV_DELOPS);
Chris Masoncb03c742008-05-15 16:15:45 -04002591 atomic_set(&fs_info->nr_async_submits, 0);
Chris Mason771ed682008-11-06 22:02:51 -05002592 atomic_set(&fs_info->async_delalloc_pages, 0);
Chris Mason8c8bee12008-09-29 11:19:10 -04002593 atomic_set(&fs_info->async_submit_draining, 0);
Chris Mason0986fe9e2008-08-15 15:34:15 -04002594 atomic_set(&fs_info->nr_async_bios, 0);
Chris Mason4cb53002011-05-24 15:35:30 -04002595 atomic_set(&fs_info->defrag_running, 0);
Josef Bacikfcebe452014-05-13 17:30:47 -07002596 atomic_set(&fs_info->qgroup_op_seq, 0);
Jan Schmidtfc36ed7e2013-04-24 16:57:33 +00002597 atomic64_set(&fs_info->tree_mod_seq, 0);
Chris Masone20d96d2007-03-22 12:13:20 -04002598 fs_info->sb = sb;
Filipe David Borba Manana95ac5672013-08-08 22:45:48 +01002599 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
Josef Bacik9ed74f22009-09-11 16:12:44 -04002600 fs_info->metadata_ratio = 0;
Chris Mason4cb53002011-05-24 15:35:30 -04002601 fs_info->defrag_inodes = RB_ROOT;
Josef Bacik2bf64752011-09-26 17:12:22 -04002602 fs_info->free_chunk_space = 0;
Jan Schmidtf29021b2012-05-16 17:55:38 +02002603 fs_info->tree_mod_log = RB_ROOT;
David Sterba8b87dc12013-08-01 18:14:52 +02002604 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
David Sterbaf8c269d2015-01-16 17:21:12 +01002605 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
Arne Jansen90519d62011-05-23 14:30:00 +02002606 /* readahead state */
Mel Gormand0164ad2015-11-06 16:28:21 -08002607 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
Arne Jansen90519d62011-05-23 14:30:00 +02002608 spin_lock_init(&fs_info->reada_lock);
Chris Masonc8b97812008-10-29 14:49:59 -04002609
Chris Masonb34b0862008-12-19 15:43:22 -05002610 fs_info->thread_pool_size = min_t(unsigned long,
2611 num_online_cpus() + 2, 8);
Chris Mason0afbaf82008-04-18 14:17:20 -04002612
Miao Xie199c2a92013-05-15 07:48:23 +00002613 INIT_LIST_HEAD(&fs_info->ordered_roots);
2614 spin_lock_init(&fs_info->ordered_root_lock);
Miao Xie16cdcec2011-04-22 18:12:22 +08002615 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2616 GFP_NOFS);
2617 if (!fs_info->delayed_root) {
2618 err = -ENOMEM;
2619 goto fail_iput;
2620 }
2621 btrfs_init_delayed_root(fs_info->delayed_root);
Chris Mason3eaa2882008-07-24 11:57:52 -04002622
Eric Sandeen638aa7e2014-08-01 18:12:38 -05002623 btrfs_init_scrub(fs_info);
Stefan Behrens21adbd52011-11-09 13:44:05 +01002624#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2625 fs_info->check_integrity_print_mask = 0;
2626#endif
Eric Sandeen779a65a2014-08-01 18:12:39 -05002627 btrfs_init_balance(fs_info);
Miao Xie21c7e752014-05-13 17:29:04 -07002628 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
Arne Jansena2de7332011-03-08 14:14:00 +01002629
Chris Masona061fc82008-05-07 11:43:44 -04002630 sb->s_blocksize = 4096;
2631 sb->s_blocksize_bits = blksize_bits(4096);
Jens Axboe32a88aa2009-09-16 15:02:33 +02002632 sb->s_bdi = &fs_info->bdi;
Chris Masona061fc82008-05-07 11:43:44 -04002633
Eric Sandeenf37938e2014-08-01 18:12:40 -05002634 btrfs_init_btree_inode(fs_info, tree_root);
Chris Mason39279cc2007-06-12 06:35:45 -04002635
Chris Masone02119d2008-09-05 16:13:11 -04002636 spin_lock_init(&fs_info->block_group_cache_lock);
Eric Paris6bef4d32010-02-23 19:43:04 +00002637 fs_info->block_group_cache_tree = RB_ROOT;
Liu Boa1897fd2012-12-27 09:01:23 +00002638 fs_info->first_logical_byte = (u64)-1;
Chris Mason925baed2008-06-25 16:01:30 -04002639
Yan Zheng11833d62009-09-11 16:11:19 -04002640 extent_io_tree_init(&fs_info->freed_extents[0],
David Sterbaf993c882011-04-20 23:35:57 +02002641 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002642 extent_io_tree_init(&fs_info->freed_extents[1],
David Sterbaf993c882011-04-20 23:35:57 +02002643 fs_info->btree_inode->i_mapping);
Yan Zheng11833d62009-09-11 16:11:19 -04002644 fs_info->pinned_extents = &fs_info->freed_extents[0];
Chris Masone6dcd2d2008-07-17 12:53:50 -04002645 fs_info->do_barriers = 1;
Chris Masonf9295742008-07-17 12:54:14 -04002646
Chris Masond98237b2007-03-28 13:57:48 -04002647
Chris Mason5a3f23d2009-03-31 13:27:11 -04002648 mutex_init(&fs_info->ordered_operations_mutex);
Chris Mason30ae8462007-03-29 09:59:15 -04002649 mutex_init(&fs_info->tree_log_mutex);
Chris Mason925baed2008-06-25 16:01:30 -04002650 mutex_init(&fs_info->chunk_mutex);
2651 mutex_init(&fs_info->transaction_kthread_mutex);
Chris Masona74a4b92008-06-25 16:01:31 -04002652 mutex_init(&fs_info->cleaner_mutex);
2653 mutex_init(&fs_info->volume_mutex);
Chris Mason1bbc6212015-04-06 12:46:08 -07002654 mutex_init(&fs_info->ro_block_group_mutex);
Josef Bacik9e351cc2014-03-13 15:42:13 -04002655 init_rwsem(&fs_info->commit_root_sem);
Yan, Zhengc71bf092009-11-12 09:34:40 +00002656 init_rwsem(&fs_info->cleanup_work_sem);
Yan, Zheng76dda932009-09-21 16:00:26 -04002657 init_rwsem(&fs_info->subvol_sem);
Stefan Behrens803b2f52013-08-15 17:11:21 +02002658 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
Chris Masonfa9c0d792009-04-03 09:47:43 -04002659
Eric Sandeenad618362014-08-01 18:12:41 -05002660 btrfs_init_dev_replace_locks(fs_info);
Eric Sandeenf9e92e42014-08-01 18:12:42 -05002661 btrfs_init_qgroup(fs_info);
Arne Jansen416ac512011-09-13 12:56:09 +02002662
Chris Masonfa9c0d792009-04-03 09:47:43 -04002663 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2664 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2665
Chris Mason7d9eb122008-07-08 14:19:17 -04002666 init_waitqueue_head(&fs_info->transaction_throttle);
Chris Mason3768f362007-03-13 16:47:54 -04002667 init_waitqueue_head(&fs_info->transaction_wait);
Sage Weilbb9c12c2010-10-29 15:37:34 -04002668 init_waitqueue_head(&fs_info->transaction_blocked_wait);
Chris Mason4854ddd2008-08-15 15:34:17 -04002669 init_waitqueue_head(&fs_info->async_submit_wait);
Chris Mason9a8dd152007-02-23 08:38:36 -05002670
Filipe Manana04216822014-11-27 21:14:15 +00002671 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2672
David Woodhouse53b381b2013-01-29 18:40:14 -05002673 ret = btrfs_alloc_stripe_hash_table(fs_info);
2674 if (ret) {
David Sterba83c82662013-03-01 15:03:00 +00002675 err = ret;
David Woodhouse53b381b2013-01-29 18:40:14 -05002676 goto fail_alloc;
2677 }
2678
David Sterba707e8a02014-06-04 19:22:26 +02002679 __setup_root(4096, 4096, 4096, tree_root,
Chris Mason2c90e5d2007-04-02 10:50:19 -04002680 fs_info, BTRFS_ROOT_TREE_OBJECTID);
Chris Mason7eccb902007-04-11 15:53:25 -04002681
Chris Mason3c4bb262012-03-27 18:56:56 -04002682 invalidate_bdev(fs_devices->latest_bdev);
David Sterba1104a882013-03-06 15:57:46 +01002683
2684 /*
2685 * Read super block and check the signature bytes only
2686 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05002687 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
Anand Jain92fc03f2015-08-14 18:32:51 +08002688 if (IS_ERR(bh)) {
2689 err = PTR_ERR(bh);
Miao Xie16cdcec2011-04-22 18:12:22 +08002690 goto fail_alloc;
Dave Young20b45072011-01-08 10:09:13 +00002691 }
Chris Mason39279cc2007-06-12 06:35:45 -04002692
David Sterba1104a882013-03-06 15:57:46 +01002693 /*
2694 * We want to check superblock checksum, the type is stored inside.
2695 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2696 */
2697 if (btrfs_check_super_csum(bh->b_data)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002698 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
David Sterba1104a882013-03-06 15:57:46 +01002699 err = -EINVAL;
Anand Jainb2acddd2015-10-07 17:23:23 +08002700 brelse(bh);
David Sterba1104a882013-03-06 15:57:46 +01002701 goto fail_alloc;
2702 }
2703
2704 /*
2705 * super_copy is zeroed at allocation time and we never touch the
2706 * following bytes up to INFO_SIZE, the checksum is calculated from
2707 * the whole block of INFO_SIZE
2708 */
David Sterba6c417612011-04-13 15:41:04 +02002709 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2710 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2711 sizeof(*fs_info->super_for_commit));
Chris Masona061fc82008-05-07 11:43:44 -04002712 brelse(bh);
Chris Mason5f39d392007-10-15 16:14:19 -04002713
David Sterba6c417612011-04-13 15:41:04 +02002714 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
Chris Mason0b86a832008-03-24 15:01:56 -04002715
David Sterba1104a882013-03-06 15:57:46 +01002716 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2717 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002718 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
David Sterba1104a882013-03-06 15:57:46 +01002719 err = -EINVAL;
2720 goto fail_alloc;
2721 }
2722
David Sterba6c417612011-04-13 15:41:04 +02002723 disk_super = fs_info->super_copy;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002724 if (!btrfs_super_root(disk_super))
Miao Xie16cdcec2011-04-22 18:12:22 +08002725 goto fail_alloc;
Chris Mason0f7d52f2007-04-09 10:42:37 -04002726
liuboacce9522011-01-06 19:30:25 +08002727 /* check FS state, whether FS is broken. */
Miao Xie87533c42013-01-29 10:14:48 +00002728 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2729 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
liuboacce9522011-01-06 19:30:25 +08002730
Liu Bo75e7cb72011-03-22 10:12:20 +00002731 /*
Chris Masonaf31f5e2011-11-03 15:17:42 -04002732 * run through our array of backup supers and setup
2733 * our ring pointer to the oldest one
2734 */
2735 generation = btrfs_super_generation(disk_super);
2736 find_oldest_super_backup(fs_info, generation);
2737
2738 /*
Liu Bo75e7cb72011-03-22 10:12:20 +00002739 * In the long term, we'll store the compression type in the super
2740 * block, and it'll be used for per file compression control.
2741 */
2742 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2743
Yan Zheng2b820322008-11-17 21:11:30 -05002744 ret = btrfs_parse_options(tree_root, options);
2745 if (ret) {
2746 err = ret;
Miao Xie16cdcec2011-04-22 18:12:22 +08002747 goto fail_alloc;
Yan Zheng2b820322008-11-17 21:11:30 -05002748 }
Chris Masondfe25022008-05-13 13:46:40 -04002749
Josef Bacikf2b636e2008-12-02 06:36:08 -05002750 features = btrfs_super_incompat_flags(disk_super) &
2751 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2752 if (features) {
2753 printk(KERN_ERR "BTRFS: couldn't mount because of "
2754 "unsupported optional features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002755 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002756 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002757 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002758 }
2759
David Sterba707e8a02014-06-04 19:22:26 +02002760 /*
2761 * Leafsize and nodesize were always equal, this is only a sanity check.
2762 */
2763 if (le32_to_cpu(disk_super->__unused_leafsize) !=
Chris Mason727011e2010-08-06 13:21:20 -04002764 btrfs_super_nodesize(disk_super)) {
2765 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2766 "blocksizes don't match. node %d leaf %d\n",
2767 btrfs_super_nodesize(disk_super),
David Sterba707e8a02014-06-04 19:22:26 +02002768 le32_to_cpu(disk_super->__unused_leafsize));
Chris Mason727011e2010-08-06 13:21:20 -04002769 err = -EINVAL;
2770 goto fail_alloc;
2771 }
David Sterba707e8a02014-06-04 19:22:26 +02002772 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002773 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2774 "blocksize (%d) was too large\n",
David Sterba707e8a02014-06-04 19:22:26 +02002775 btrfs_super_nodesize(disk_super));
Chris Mason727011e2010-08-06 13:21:20 -04002776 err = -EINVAL;
2777 goto fail_alloc;
2778 }
2779
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002780 features = btrfs_super_incompat_flags(disk_super);
Li Zefana6fa6fa2010-10-25 15:12:26 +08002781 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
Li Zefan69e380d2012-06-11 17:10:51 +08002782 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
Li Zefana6fa6fa2010-10-25 15:12:26 +08002783 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
Chris Mason727011e2010-08-06 13:21:20 -04002784
Josef Bacik3173a182013-03-07 14:22:04 -05002785 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
David Sterba5efa0492014-12-19 18:38:47 +01002786 printk(KERN_INFO "BTRFS: has skinny extents\n");
Josef Bacik3173a182013-03-07 14:22:04 -05002787
Chris Mason727011e2010-08-06 13:21:20 -04002788 /*
2789 * flag our filesystem as having big metadata blocks if
2790 * they are bigger than the page size
2791 */
David Sterba707e8a02014-06-04 19:22:26 +02002792 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
Chris Mason727011e2010-08-06 13:21:20 -04002793 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
Frank Holtonefe120a2013-12-20 11:37:06 -05002794 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
Chris Mason727011e2010-08-06 13:21:20 -04002795 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2796 }
2797
Chris Masonbc3f1162012-03-29 17:02:47 -04002798 nodesize = btrfs_super_nodesize(disk_super);
Chris Masonbc3f1162012-03-29 17:02:47 -04002799 sectorsize = btrfs_super_sectorsize(disk_super);
2800 stripesize = btrfs_super_stripesize(disk_super);
David Sterba707e8a02014-06-04 19:22:26 +02002801 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
Miao Xie963d6782013-01-29 10:10:51 +00002802 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
Chris Masonbc3f1162012-03-29 17:02:47 -04002803
2804 /*
2805 * mixed block groups end up with duplicate but slightly offset
2806 * extent buffers for the same range. It leads to corruptions
2807 */
2808 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
David Sterba707e8a02014-06-04 19:22:26 +02002809 (sectorsize != nodesize)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002810 printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes "
Chris Masonbc3f1162012-03-29 17:02:47 -04002811 "are not allowed for mixed block groups on %s\n",
2812 sb->s_id);
2813 goto fail_alloc;
2814 }
2815
Miao Xieceda0862013-04-11 10:30:16 +00002816 /*
2817 * Needn't use the lock because there is no other task which will
2818 * update the flag.
2819 */
Li Zefana6fa6fa2010-10-25 15:12:26 +08002820 btrfs_set_super_incompat_flags(disk_super, features);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002821
Josef Bacikf2b636e2008-12-02 06:36:08 -05002822 features = btrfs_super_compat_ro_flags(disk_super) &
2823 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2824 if (!(sb->s_flags & MS_RDONLY) && features) {
2825 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2826 "unsupported option features (%Lx).\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02002827 features);
Josef Bacikf2b636e2008-12-02 06:36:08 -05002828 err = -EINVAL;
Miao Xie16cdcec2011-04-22 18:12:22 +08002829 goto fail_alloc;
Josef Bacikf2b636e2008-12-02 06:36:08 -05002830 }
Chris Mason61d92c32009-10-02 19:11:56 -04002831
Qu Wenruo5cdc7ad32014-02-28 10:46:06 +08002832 max_active = fs_info->thread_pool_size;
Chris Mason61d92c32009-10-02 19:11:56 -04002833
Eric Sandeen2a458192015-02-16 16:29:26 +01002834 ret = btrfs_init_workqueues(fs_info, fs_devices);
2835 if (ret) {
2836 err = ret;
Josef Bacik0dc3b842011-11-18 14:37:27 -05002837 goto fail_sb_buffer;
2838 }
Chris Mason4543df72008-06-11 21:47:56 -04002839
Chris Mason4575c9c2008-04-18 16:13:31 -04002840 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
Chris Masonc8b97812008-10-29 14:49:59 -04002841 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
Byongho Leeee221842015-12-15 01:42:10 +09002842 SZ_4M / PAGE_CACHE_SIZE);
Chris Mason4575c9c2008-04-18 16:13:31 -04002843
Chris Masondb945352007-10-15 16:15:53 -04002844 tree_root->nodesize = nodesize;
Chris Masondb945352007-10-15 16:15:53 -04002845 tree_root->sectorsize = sectorsize;
Chris Mason87ee04e2007-11-30 11:30:34 -05002846 tree_root->stripesize = stripesize;
Chris Masona061fc82008-05-07 11:43:44 -04002847
2848 sb->s_blocksize = sectorsize;
2849 sb->s_blocksize_bits = blksize_bits(sectorsize);
Chris Masondb945352007-10-15 16:15:53 -04002850
Qu Wenruo3cae2102013-07-16 11:19:18 +08002851 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002852 printk(KERN_ERR "BTRFS: valid FS not found on %s\n", sb->s_id);
Chris Mason39279cc2007-06-12 06:35:45 -04002853 goto fail_sb_buffer;
2854 }
Chris Mason19c00dd2007-10-15 16:19:22 -04002855
Liu Bo8d082fb2012-04-03 09:56:53 +08002856 if (sectorsize != PAGE_SIZE) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002857 printk(KERN_ERR "BTRFS: incompatible sector size (%lu) "
Liu Bo8d082fb2012-04-03 09:56:53 +08002858 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
Keith Mannthey941b2dd2011-11-29 17:44:12 -08002859 goto fail_sb_buffer;
2860 }
2861
Chris Mason925baed2008-06-25 16:01:30 -04002862 mutex_lock(&fs_info->chunk_mutex);
Yan Zhenge4404d62008-12-12 10:03:26 -05002863 ret = btrfs_read_sys_array(tree_root);
Chris Mason925baed2008-06-25 16:01:30 -04002864 mutex_unlock(&fs_info->chunk_mutex);
Chris Mason84eed902008-04-25 09:04:37 -04002865 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002866 printk(KERN_ERR "BTRFS: failed to read the system "
Chris Masond3977122009-01-05 21:25:51 -05002867 "array on %s\n", sb->s_id);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002868 goto fail_sb_buffer;
Chris Mason84eed902008-04-25 09:04:37 -04002869 }
Chris Mason0b86a832008-03-24 15:01:56 -04002870
Yan Zheng84234f32008-10-29 14:49:05 -04002871 generation = btrfs_super_chunk_root_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002872
David Sterba707e8a02014-06-04 19:22:26 +02002873 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2874 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
Chris Mason0b86a832008-03-24 15:01:56 -04002875
2876 chunk_root->node = read_tree_block(chunk_root,
2877 btrfs_super_chunk_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002878 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002879 if (IS_ERR(chunk_root->node) ||
2880 !extent_buffer_uptodate(chunk_root->node)) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002881 printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002882 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302883 if (!IS_ERR(chunk_root->node))
2884 free_extent_buffer(chunk_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002885 chunk_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002886 goto fail_tree_roots;
David Woodhouse83121942009-07-22 16:52:13 -04002887 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002888 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2889 chunk_root->commit_root = btrfs_root_node(chunk_root);
Chris Mason0b86a832008-03-24 15:01:56 -04002890
Chris Masone17cade2008-04-15 15:41:47 -04002891 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
Geert Uytterhoevenb308bc22013-08-20 13:20:15 +02002892 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
Chris Masone17cade2008-04-15 15:41:47 -04002893
Chris Mason0b86a832008-03-24 15:01:56 -04002894 ret = btrfs_read_chunk_tree(chunk_root);
Yan Zheng2b820322008-11-17 21:11:30 -05002895 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002896 printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n",
Chris Masond3977122009-01-05 21:25:51 -05002897 sb->s_id);
Chris Masonaf31f5e2011-11-03 15:17:42 -04002898 goto fail_tree_roots;
Yan Zheng2b820322008-11-17 21:11:30 -05002899 }
Chris Mason0b86a832008-03-24 15:01:56 -04002900
Stefan Behrens8dabb742012-11-06 13:15:27 +01002901 /*
2902 * keep the device that is marked to be the target device for the
2903 * dev_replace procedure
2904 */
Eric Sandeen9eaed212014-08-01 18:12:35 -05002905 btrfs_close_extra_devices(fs_devices, 0);
Chris Masondfe25022008-05-13 13:46:40 -04002906
Chris Masona6b0d5c2012-02-20 20:53:43 -05002907 if (!fs_devices->latest_bdev) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002908 printk(KERN_ERR "BTRFS: failed to read devices on %s\n",
Chris Masona6b0d5c2012-02-20 20:53:43 -05002909 sb->s_id);
2910 goto fail_tree_roots;
2911 }
2912
Chris Masonaf31f5e2011-11-03 15:17:42 -04002913retry_root_backup:
Yan Zheng84234f32008-10-29 14:49:05 -04002914 generation = btrfs_super_generation(disk_super);
Chris Mason0b86a832008-03-24 15:01:56 -04002915
Chris Masone20d96d2007-03-22 12:13:20 -04002916 tree_root->node = read_tree_block(tree_root,
Chris Masondb945352007-10-15 16:15:53 -04002917 btrfs_super_root(disk_super),
David Sterbace86cd52014-06-15 01:07:32 +02002918 generation);
Liu Bo64c043d2015-05-25 17:30:15 +08002919 if (IS_ERR(tree_root->node) ||
2920 !extent_buffer_uptodate(tree_root->node)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002921 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
David Woodhouse83121942009-07-22 16:52:13 -04002922 sb->s_id);
chandane5fffba2015-10-05 22:14:25 +05302923 if (!IS_ERR(tree_root->node))
2924 free_extent_buffer(tree_root->node);
Zhao Lei95ab1f62015-07-15 21:02:09 +08002925 tree_root->node = NULL;
Chris Masonaf31f5e2011-11-03 15:17:42 -04002926 goto recovery_tree_root;
David Woodhouse83121942009-07-22 16:52:13 -04002927 }
Chris Masonaf31f5e2011-11-03 15:17:42 -04002928
Yan Zheng5d4f98a2009-06-10 10:45:14 -04002929 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2930 tree_root->commit_root = btrfs_root_node(tree_root);
Stefan Behrens69e9c6c2013-09-05 16:58:43 +02002931 btrfs_set_root_refs(&tree_root->root_item, 1);
Chris Masondb945352007-10-15 16:15:53 -04002932
Chandan Rajendraf32e48e2016-01-07 18:56:59 +05302933 mutex_lock(&tree_root->objectid_mutex);
2934 ret = btrfs_find_highest_objectid(tree_root,
2935 &tree_root->highest_objectid);
2936 if (ret) {
2937 mutex_unlock(&tree_root->objectid_mutex);
2938 goto recovery_tree_root;
2939 }
2940
2941 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
2942
2943 mutex_unlock(&tree_root->objectid_mutex);
2944
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002945 ret = btrfs_read_roots(fs_info, tree_root);
2946 if (ret)
Chris Masonaf31f5e2011-11-03 15:17:42 -04002947 goto recovery_tree_root;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02002948
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002949 fs_info->generation = generation;
2950 fs_info->last_trans_committed = generation;
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04002951
Ilya Dryomov68310a52012-06-22 12:24:12 -06002952 ret = btrfs_recover_balance(fs_info);
2953 if (ret) {
David Sterbaaa8ee312014-12-19 18:38:41 +01002954 printk(KERN_ERR "BTRFS: failed to recover balance\n");
Ilya Dryomov68310a52012-06-22 12:24:12 -06002955 goto fail_block_groups;
2956 }
2957
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002958 ret = btrfs_init_dev_stats(fs_info);
2959 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002960 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
Stefan Behrens733f4fb2012-05-25 16:06:10 +02002961 ret);
2962 goto fail_block_groups;
2963 }
2964
Stefan Behrens8dabb742012-11-06 13:15:27 +01002965 ret = btrfs_init_dev_replace(fs_info);
2966 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002967 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002968 goto fail_block_groups;
2969 }
2970
Eric Sandeen9eaed212014-08-01 18:12:35 -05002971 btrfs_close_extra_devices(fs_devices, 1);
Stefan Behrens8dabb742012-11-06 13:15:27 +01002972
Anand Jainb7c35e82015-03-10 06:38:38 +08002973 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
2974 if (ret) {
2975 pr_err("BTRFS: failed to init sysfs fsid interface: %d\n", ret);
2976 goto fail_block_groups;
2977 }
2978
2979 ret = btrfs_sysfs_add_device(fs_devices);
2980 if (ret) {
2981 pr_err("BTRFS: failed to init sysfs device interface: %d\n", ret);
2982 goto fail_fsdev_sysfs;
2983 }
2984
Anand Jain96f31362015-08-14 18:32:46 +08002985 ret = btrfs_sysfs_add_mounted(fs_info);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002986 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002987 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
Anand Jainb7c35e82015-03-10 06:38:38 +08002988 goto fail_fsdev_sysfs;
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04002989 }
2990
liuboc59021f2011-03-07 02:13:14 +00002991 ret = btrfs_init_space_info(fs_info);
2992 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002993 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08002994 goto fail_sysfs;
liuboc59021f2011-03-07 02:13:14 +00002995 }
2996
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05002997 ret = btrfs_read_block_groups(fs_info->extent_root);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00002998 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05002999 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
Anand Jain2365dd32014-01-22 11:15:51 +08003000 goto fail_sysfs;
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003001 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003002 fs_info->num_tolerated_disk_barrier_failures =
3003 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
Stefan Behrens292fd7f2012-10-30 17:16:16 +00003004 if (fs_info->fs_devices->missing_devices >
3005 fs_info->num_tolerated_disk_barrier_failures &&
3006 !(sb->s_flags & MS_RDONLY)) {
Zhao Lei78fa1772015-07-20 17:54:50 +08003007 pr_warn("BTRFS: missing devices(%llu) exceeds the limit(%d), writeable mount is not allowed\n",
3008 fs_info->fs_devices->missing_devices,
3009 fs_info->num_tolerated_disk_barrier_failures);
Anand Jain2365dd32014-01-22 11:15:51 +08003010 goto fail_sysfs;
Stefan Behrens292fd7f2012-10-30 17:16:16 +00003011 }
Chris Mason9078a3e2007-04-26 16:46:15 -04003012
Chris Masona74a4b92008-06-25 16:01:31 -04003013 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
3014 "btrfs-cleaner");
Qinghuang Feng57506d52009-01-21 10:49:16 -05003015 if (IS_ERR(fs_info->cleaner_kthread))
Anand Jain2365dd32014-01-22 11:15:51 +08003016 goto fail_sysfs;
Chris Masona74a4b92008-06-25 16:01:31 -04003017
3018 fs_info->transaction_kthread = kthread_run(transaction_kthread,
3019 tree_root,
3020 "btrfs-transaction");
Qinghuang Feng57506d52009-01-21 10:49:16 -05003021 if (IS_ERR(fs_info->transaction_kthread))
Chris Mason3f157a22008-06-25 16:01:31 -04003022 goto fail_cleaner;
Chris Masona74a4b92008-06-25 16:01:31 -04003023
Chris Masonc2898112009-06-10 09:51:32 -04003024 if (!btrfs_test_opt(tree_root, SSD) &&
3025 !btrfs_test_opt(tree_root, NOSSD) &&
3026 !fs_info->fs_devices->rotating) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003027 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
Chris Masonc2898112009-06-10 09:51:32 -04003028 "mode\n");
3029 btrfs_set_opt(fs_info->mount_opt, SSD);
3030 }
3031
David Sterba572d9ab2014-02-05 15:26:17 +01003032 /*
3033 * Mount does not set all options immediatelly, we can do it now and do
3034 * not have to wait for transaction commit
3035 */
3036 btrfs_apply_pending_changes(fs_info);
Qu Wenruo3818aea2014-01-13 13:36:06 +08003037
Stefan Behrens21adbd52011-11-09 13:44:05 +01003038#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3039 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
3040 ret = btrfsic_mount(tree_root, fs_devices,
3041 btrfs_test_opt(tree_root,
3042 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3043 1 : 0,
3044 fs_info->check_integrity_print_mask);
3045 if (ret)
Frank Holtonefe120a2013-12-20 11:37:06 -05003046 printk(KERN_WARNING "BTRFS: failed to initialize"
Stefan Behrens21adbd52011-11-09 13:44:05 +01003047 " integrity check module %s\n", sb->s_id);
3048 }
3049#endif
Arne Jansenbcef60f2011-09-13 15:23:30 +02003050 ret = btrfs_read_qgroup_config(fs_info);
3051 if (ret)
3052 goto fail_trans_kthread;
Stefan Behrens21adbd52011-11-09 13:44:05 +01003053
liuboacce9522011-01-06 19:30:25 +08003054 /* do not make disk changes in broken FS */
Stefan Behrens68ce9682012-08-01 05:45:52 -06003055 if (btrfs_super_log_root(disk_super) != 0) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003056 ret = btrfs_replay_log(fs_info, fs_devices);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003057 if (ret) {
Eric Sandeen63443bf2014-08-01 18:12:46 -05003058 err = ret;
Wang Shilong28c16cb2014-04-23 19:33:35 +08003059 goto fail_qgroup;
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003060 }
Chris Masone02119d2008-09-05 16:13:11 -04003061 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003062
Yan, Zheng76dda932009-09-21 16:00:26 -04003063 ret = btrfs_find_orphan_roots(tree_root);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003064 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003065 goto fail_qgroup;
Yan, Zheng76dda932009-09-21 16:00:26 -04003066
Chris Mason7c2ca462008-11-19 15:13:35 -05003067 if (!(sb->s_flags & MS_RDONLY)) {
Yan, Zhengd68fc572010-05-16 10:49:58 -04003068 ret = btrfs_cleanup_fs_roots(fs_info);
Ilya Dryomov44c44af2012-06-22 12:14:13 -06003069 if (ret)
Wang Shilong28c16cb2014-04-23 19:33:35 +08003070 goto fail_qgroup;
Yan, Zhengd68fc572010-05-16 10:49:58 -04003071
Wang Shilong5f316482014-06-26 11:08:16 +08003072 mutex_lock(&fs_info->cleaner_mutex);
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003073 ret = btrfs_recover_relocation(tree_root);
Wang Shilong5f316482014-06-26 11:08:16 +08003074 mutex_unlock(&fs_info->cleaner_mutex);
Miao Xied7ce5842010-02-02 08:46:44 +00003075 if (ret < 0) {
3076 printk(KERN_WARNING
Frank Holtonefe120a2013-12-20 11:37:06 -05003077 "BTRFS: failed to recover relocation\n");
Miao Xied7ce5842010-02-02 08:46:44 +00003078 err = -EINVAL;
Arne Jansenbcef60f2011-09-13 15:23:30 +02003079 goto fail_qgroup;
Miao Xied7ce5842010-02-02 08:46:44 +00003080 }
Chris Mason7c2ca462008-11-19 15:13:35 -05003081 }
Zheng Yan1a40e232008-09-26 10:09:34 -04003082
Chris Mason3de45862008-11-17 21:02:50 -05003083 location.objectid = BTRFS_FS_TREE_OBJECTID;
3084 location.type = BTRFS_ROOT_ITEM_KEY;
Miao Xiecb517ea2013-05-15 07:48:19 +00003085 location.offset = 0;
Chris Mason3de45862008-11-17 21:02:50 -05003086
Chris Mason3de45862008-11-17 21:02:50 -05003087 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003088 if (IS_ERR(fs_info->fs_root)) {
3089 err = PTR_ERR(fs_info->fs_root);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003090 goto fail_qgroup;
Dan Carpenter3140c9a2010-05-29 09:44:10 +00003091 }
Chris Masonc2898112009-06-10 09:51:32 -04003092
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003093 if (sb->s_flags & MS_RDONLY)
3094 return 0;
Ilya Dryomov59641012012-01-16 22:04:48 +02003095
Chris Mason511711a2015-12-30 07:52:35 -08003096 if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) &&
3097 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3098 pr_info("BTRFS: creating free space tree\n");
3099 ret = btrfs_create_free_space_tree(fs_info);
3100 if (ret) {
3101 pr_warn("BTRFS: failed to create free space tree %d\n",
3102 ret);
3103 close_ctree(tree_root);
3104 return ret;
3105 }
3106 }
3107
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003108 down_read(&fs_info->cleanup_work_sem);
3109 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3110 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
Josef Bacike3acc2a2010-01-26 14:30:53 +00003111 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003112 close_ctree(tree_root);
3113 return ret;
3114 }
3115 up_read(&fs_info->cleanup_work_sem);
Ilya Dryomov59641012012-01-16 22:04:48 +02003116
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003117 ret = btrfs_resume_balance_async(fs_info);
3118 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003119 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
Ilya Dryomov2b6ba622012-06-22 12:24:13 -06003120 close_ctree(tree_root);
3121 return ret;
Josef Bacike3acc2a2010-01-26 14:30:53 +00003122 }
3123
Stefan Behrens8dabb742012-11-06 13:15:27 +01003124 ret = btrfs_resume_dev_replace_async(fs_info);
3125 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003126 pr_warn("BTRFS: failed to resume dev_replace\n");
Stefan Behrens8dabb742012-11-06 13:15:27 +01003127 close_ctree(tree_root);
3128 return ret;
3129 }
3130
Jan Schmidtb382a322013-05-28 15:47:24 +00003131 btrfs_qgroup_rescan_resume(fs_info);
3132
Omar Sandoval70f6d822015-09-29 20:50:38 -07003133 if (btrfs_test_opt(tree_root, CLEAR_CACHE) &&
3134 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3135 pr_info("BTRFS: clearing free space tree\n");
3136 ret = btrfs_clear_free_space_tree(fs_info);
3137 if (ret) {
3138 pr_warn("BTRFS: failed to clear free space tree %d\n",
3139 ret);
3140 close_ctree(tree_root);
3141 return ret;
3142 }
3143 }
3144
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003145 if (!fs_info->uuid_root) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003146 pr_info("BTRFS: creating UUID tree\n");
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003147 ret = btrfs_create_uuid_tree(fs_info);
3148 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003149 pr_warn("BTRFS: failed to create the UUID tree %d\n",
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003150 ret);
3151 close_ctree(tree_root);
3152 return ret;
3153 }
Eric Sandeen4bbcaa62014-08-01 18:12:45 -05003154 } else if (btrfs_test_opt(tree_root, RESCAN_UUID_TREE) ||
3155 fs_info->generation !=
3156 btrfs_super_uuid_tree_generation(disk_super)) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003157 pr_info("BTRFS: checking UUID tree\n");
Stefan Behrens70f80172013-08-15 17:11:23 +02003158 ret = btrfs_check_uuid_tree(fs_info);
3159 if (ret) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003160 pr_warn("BTRFS: failed to check the UUID tree %d\n",
Stefan Behrens70f80172013-08-15 17:11:23 +02003161 ret);
3162 close_ctree(tree_root);
3163 return ret;
3164 }
3165 } else {
3166 fs_info->update_uuid_tree_gen = 1;
Stefan Behrensf7a81ea2013-08-15 17:11:19 +02003167 }
3168
Josef Bacik47ab2a62014-09-18 11:20:02 -04003169 fs_info->open = 1;
3170
Al Viroad2b2c82011-11-17 01:10:02 -05003171 return 0;
Chris Mason39279cc2007-06-12 06:35:45 -04003172
Arne Jansenbcef60f2011-09-13 15:23:30 +02003173fail_qgroup:
3174 btrfs_free_qgroup_config(fs_info);
Chris Mason7c2ca462008-11-19 15:13:35 -05003175fail_trans_kthread:
3176 kthread_stop(fs_info->transaction_kthread);
Josef Bacik54067ae2013-04-25 13:44:38 -04003177 btrfs_cleanup_transaction(fs_info->tree_root);
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003178 btrfs_free_fs_roots(fs_info);
Chris Mason3f157a22008-06-25 16:01:31 -04003179fail_cleaner:
Chris Masona74a4b92008-06-25 16:01:31 -04003180 kthread_stop(fs_info->cleaner_kthread);
Chris Mason7c2ca462008-11-19 15:13:35 -05003181
3182 /*
3183 * make sure we're done with the btree inode before we stop our
3184 * kthreads
3185 */
3186 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
Chris Mason7c2ca462008-11-19 15:13:35 -05003187
Anand Jain2365dd32014-01-22 11:15:51 +08003188fail_sysfs:
Anand Jain6618a592015-08-14 18:32:47 +08003189 btrfs_sysfs_remove_mounted(fs_info);
Anand Jain2365dd32014-01-22 11:15:51 +08003190
Anand Jainb7c35e82015-03-10 06:38:38 +08003191fail_fsdev_sysfs:
3192 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3193
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003194fail_block_groups:
Josef Bacik54067ae2013-04-25 13:44:38 -04003195 btrfs_put_block_group_cache(fs_info);
Josef Bacik1b1d1f62010-03-19 20:49:55 +00003196 btrfs_free_block_groups(fs_info);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003197
3198fail_tree_roots:
3199 free_root_pointers(fs_info, 1);
Miao Xie2b8195b2013-02-07 06:01:35 +00003200 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003201
Chris Mason39279cc2007-06-12 06:35:45 -04003202fail_sb_buffer:
Liu Bo7abadb62013-03-17 02:10:31 +00003203 btrfs_stop_all_workers(fs_info);
Miao Xie16cdcec2011-04-22 18:12:22 +08003204fail_alloc:
Chris Mason4543df72008-06-11 21:47:56 -04003205fail_iput:
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003206 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3207
Chris Mason4543df72008-06-11 21:47:56 -04003208 iput(fs_info->btree_inode);
Miao Xiec404e0d2014-01-30 16:46:55 +08003209fail_bio_counter:
3210 percpu_counter_destroy(&fs_info->bio_counter);
Miao Xie963d6782013-01-29 10:10:51 +00003211fail_delalloc_bytes:
3212 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiee2d84522013-01-29 10:09:20 +00003213fail_dirty_metadata_bytes:
3214 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Jens Axboead081f12009-06-12 14:43:40 +02003215fail_bdi:
Qinghuang Feng7e662852009-01-21 10:49:16 -05003216 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003217fail_srcu:
3218 cleanup_srcu_struct(&fs_info->subvol_srcu);
Qinghuang Feng7e662852009-01-21 10:49:16 -05003219fail:
David Woodhouse53b381b2013-01-29 18:40:14 -05003220 btrfs_free_stripe_hash_table(fs_info);
Ilya Dryomov586e46e2011-11-09 13:26:37 +02003221 btrfs_close_devices(fs_info->fs_devices);
Al Viroad2b2c82011-11-17 01:10:02 -05003222 return err;
Chris Masonaf31f5e2011-11-03 15:17:42 -04003223
3224recovery_tree_root:
Chris Masonaf31f5e2011-11-03 15:17:42 -04003225 if (!btrfs_test_opt(tree_root, RECOVERY))
3226 goto fail_tree_roots;
3227
3228 free_root_pointers(fs_info, 0);
3229
3230 /* don't use the log in recovery mode, it won't be valid */
3231 btrfs_set_super_log_root(disk_super, 0);
3232
3233 /* we can't trust the free space cache either */
3234 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3235
3236 ret = next_root_backup(fs_info, fs_info->super_copy,
3237 &num_backups_tried, &backup_index);
3238 if (ret == -1)
3239 goto fail_block_groups;
3240 goto retry_root_backup;
Chris Masoneb60cea2007-02-02 09:18:22 -05003241}
3242
Chris Masonf2984462008-04-10 16:19:33 -04003243static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3244{
Chris Masonf2984462008-04-10 16:19:33 -04003245 if (uptodate) {
3246 set_buffer_uptodate(bh);
3247 } else {
Stefan Behrens442a4f62012-05-25 16:06:08 +02003248 struct btrfs_device *device = (struct btrfs_device *)
3249 bh->b_private;
3250
David Sterbab14af3b2015-10-08 10:43:10 +02003251 btrfs_warn_rl_in_rcu(device->dev_root->fs_info,
3252 "lost page write due to IO error on %s",
Josef Bacik606686e2012-06-04 14:03:51 -04003253 rcu_str_deref(device->name));
Chris Mason1259ab72008-05-12 13:39:03 -04003254 /* note, we dont' set_buffer_write_io_error because we have
3255 * our own ways of dealing with the IO errors
3256 */
Chris Masonf2984462008-04-10 16:19:33 -04003257 clear_buffer_uptodate(bh);
Stefan Behrens442a4f62012-05-25 16:06:08 +02003258 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
Chris Masonf2984462008-04-10 16:19:33 -04003259 }
3260 unlock_buffer(bh);
3261 put_bh(bh);
3262}
3263
Anand Jain29c36d72015-08-14 18:32:58 +08003264int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3265 struct buffer_head **bh_ret)
3266{
3267 struct buffer_head *bh;
3268 struct btrfs_super_block *super;
3269 u64 bytenr;
3270
3271 bytenr = btrfs_sb_offset(copy_num);
3272 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3273 return -EINVAL;
3274
3275 bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE);
3276 /*
3277 * If we fail to read from the underlying devices, as of now
3278 * the best option we have is to mark it EIO.
3279 */
3280 if (!bh)
3281 return -EIO;
3282
3283 super = (struct btrfs_super_block *)bh->b_data;
3284 if (btrfs_super_bytenr(super) != bytenr ||
3285 btrfs_super_magic(super) != BTRFS_MAGIC) {
3286 brelse(bh);
3287 return -EINVAL;
3288 }
3289
3290 *bh_ret = bh;
3291 return 0;
3292}
3293
3294
Yan Zhenga512bbf2008-12-08 16:46:26 -05003295struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3296{
3297 struct buffer_head *bh;
3298 struct buffer_head *latest = NULL;
3299 struct btrfs_super_block *super;
3300 int i;
3301 u64 transid = 0;
Anand Jain92fc03f2015-08-14 18:32:51 +08003302 int ret = -EINVAL;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003303
3304 /* we would like to check all the supers, but that would make
3305 * a btrfs mount succeed after a mkfs from a different FS.
3306 * So, we need to add a special mount option to scan for
3307 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3308 */
3309 for (i = 0; i < 1; i++) {
Anand Jain29c36d72015-08-14 18:32:58 +08003310 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3311 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003312 continue;
3313
3314 super = (struct btrfs_super_block *)bh->b_data;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003315
3316 if (!latest || btrfs_super_generation(super) > transid) {
3317 brelse(latest);
3318 latest = bh;
3319 transid = btrfs_super_generation(super);
3320 } else {
3321 brelse(bh);
3322 }
3323 }
Anand Jain92fc03f2015-08-14 18:32:51 +08003324
3325 if (!latest)
3326 return ERR_PTR(ret);
3327
Yan Zhenga512bbf2008-12-08 16:46:26 -05003328 return latest;
3329}
3330
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003331/*
3332 * this should be called twice, once with wait == 0 and
3333 * once with wait == 1. When wait == 0 is done, all the buffer heads
3334 * we write are pinned.
3335 *
3336 * They are released when wait == 1 is done.
3337 * max_mirrors must be the same for both runs, and it indicates how
3338 * many supers on this one device should be written.
3339 *
3340 * max_mirrors == 0 means to write them all.
3341 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003342static int write_dev_supers(struct btrfs_device *device,
3343 struct btrfs_super_block *sb,
3344 int do_barriers, int wait, int max_mirrors)
3345{
3346 struct buffer_head *bh;
3347 int i;
3348 int ret;
3349 int errors = 0;
3350 u32 crc;
3351 u64 bytenr;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003352
3353 if (max_mirrors == 0)
3354 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3355
Yan Zhenga512bbf2008-12-08 16:46:26 -05003356 for (i = 0; i < max_mirrors; i++) {
3357 bytenr = btrfs_sb_offset(i);
Miao Xie935e5cc2014-09-03 21:35:33 +08003358 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3359 device->commit_total_bytes)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003360 break;
3361
3362 if (wait) {
3363 bh = __find_get_block(device->bdev, bytenr / 4096,
3364 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003365 if (!bh) {
3366 errors++;
3367 continue;
3368 }
Yan Zhenga512bbf2008-12-08 16:46:26 -05003369 wait_on_buffer(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003370 if (!buffer_uptodate(bh))
3371 errors++;
3372
3373 /* drop our reference */
3374 brelse(bh);
3375
3376 /* drop the reference from the wait == 0 run */
3377 brelse(bh);
3378 continue;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003379 } else {
3380 btrfs_set_super_bytenr(sb, bytenr);
3381
3382 crc = ~(u32)0;
Liu Bob0496682013-03-14 14:57:45 +00003383 crc = btrfs_csum_data((char *)sb +
Yan Zhenga512bbf2008-12-08 16:46:26 -05003384 BTRFS_CSUM_SIZE, crc,
3385 BTRFS_SUPER_INFO_SIZE -
3386 BTRFS_CSUM_SIZE);
3387 btrfs_csum_final(crc, sb->csum);
3388
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003389 /*
3390 * one reference for us, and we leave it for the
3391 * caller
3392 */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003393 bh = __getblk(device->bdev, bytenr / 4096,
3394 BTRFS_SUPER_INFO_SIZE);
Josef Bacik634554d2013-04-29 10:05:57 -04003395 if (!bh) {
David Sterbaf14d1042015-10-08 11:37:06 +02003396 btrfs_err(device->dev_root->fs_info,
3397 "couldn't get super buffer head for bytenr %llu",
3398 bytenr);
Josef Bacik634554d2013-04-29 10:05:57 -04003399 errors++;
3400 continue;
3401 }
3402
Yan Zhenga512bbf2008-12-08 16:46:26 -05003403 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3404
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003405 /* one reference for submit_bh */
Yan Zhenga512bbf2008-12-08 16:46:26 -05003406 get_bh(bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003407
3408 set_buffer_uptodate(bh);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003409 lock_buffer(bh);
3410 bh->b_end_io = btrfs_end_buffer_write_sync;
Stefan Behrens442a4f62012-05-25 16:06:08 +02003411 bh->b_private = device;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003412 }
3413
Chris Mason387125f2011-11-18 15:07:51 -05003414 /*
3415 * we fua the first super. The others we allow
3416 * to go down lazy.
3417 */
Wang Shilonge8117c22014-01-03 18:22:57 +08003418 if (i == 0)
3419 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3420 else
3421 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
Hisashi Hifumi4eedeb72009-06-10 15:28:55 -04003422 if (ret)
Yan Zhenga512bbf2008-12-08 16:46:26 -05003423 errors++;
Yan Zhenga512bbf2008-12-08 16:46:26 -05003424 }
3425 return errors < i ? 0 : -1;
3426}
3427
Chris Mason387125f2011-11-18 15:07:51 -05003428/*
3429 * endio for the write_dev_flush, this will wake anyone waiting
3430 * for the barrier when it is done
3431 */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003432static void btrfs_end_empty_barrier(struct bio *bio)
Chris Mason387125f2011-11-18 15:07:51 -05003433{
Chris Mason387125f2011-11-18 15:07:51 -05003434 if (bio->bi_private)
3435 complete(bio->bi_private);
3436 bio_put(bio);
3437}
3438
3439/*
3440 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3441 * sent down. With wait == 1, it waits for the previous flush.
3442 *
3443 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3444 * capable
3445 */
3446static int write_dev_flush(struct btrfs_device *device, int wait)
3447{
3448 struct bio *bio;
3449 int ret = 0;
3450
3451 if (device->nobarriers)
3452 return 0;
3453
3454 if (wait) {
3455 bio = device->flush_bio;
3456 if (!bio)
3457 return 0;
3458
3459 wait_for_completion(&device->flush_wait);
3460
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02003461 if (bio->bi_error) {
3462 ret = bio->bi_error;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003463 btrfs_dev_stat_inc_and_print(device,
3464 BTRFS_DEV_STAT_FLUSH_ERRS);
Chris Mason387125f2011-11-18 15:07:51 -05003465 }
3466
3467 /* drop the reference from the wait == 0 run */
3468 bio_put(bio);
3469 device->flush_bio = NULL;
3470
3471 return ret;
3472 }
3473
3474 /*
3475 * one reference for us, and we leave it for the
3476 * caller
3477 */
Jesper Juhl9c017ab2012-02-27 00:04:56 +01003478 device->flush_bio = NULL;
Chris Mason9be33952013-05-17 18:30:14 -04003479 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
Chris Mason387125f2011-11-18 15:07:51 -05003480 if (!bio)
3481 return -ENOMEM;
3482
3483 bio->bi_end_io = btrfs_end_empty_barrier;
3484 bio->bi_bdev = device->bdev;
3485 init_completion(&device->flush_wait);
3486 bio->bi_private = &device->flush_wait;
3487 device->flush_bio = bio;
3488
3489 bio_get(bio);
Stefan Behrens21adbd52011-11-09 13:44:05 +01003490 btrfsic_submit_bio(WRITE_FLUSH, bio);
Chris Mason387125f2011-11-18 15:07:51 -05003491
3492 return 0;
3493}
3494
3495/*
3496 * send an empty flush down to each device in parallel,
3497 * then wait for them
3498 */
3499static int barrier_all_devices(struct btrfs_fs_info *info)
3500{
3501 struct list_head *head;
3502 struct btrfs_device *dev;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003503 int errors_send = 0;
3504 int errors_wait = 0;
Chris Mason387125f2011-11-18 15:07:51 -05003505 int ret;
3506
3507 /* send down all the barriers */
3508 head = &info->fs_devices->devices;
3509 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003510 if (dev->missing)
3511 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003512 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003513 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003514 continue;
3515 }
3516 if (!dev->in_fs_metadata || !dev->writeable)
3517 continue;
3518
3519 ret = write_dev_flush(dev, 0);
3520 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003521 errors_send++;
Chris Mason387125f2011-11-18 15:07:51 -05003522 }
3523
3524 /* wait for all the barriers */
3525 list_for_each_entry_rcu(dev, head, dev_list) {
Hidetoshi Setof88ba6a2014-02-05 16:34:38 +09003526 if (dev->missing)
3527 continue;
Chris Mason387125f2011-11-18 15:07:51 -05003528 if (!dev->bdev) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003529 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003530 continue;
3531 }
3532 if (!dev->in_fs_metadata || !dev->writeable)
3533 continue;
3534
3535 ret = write_dev_flush(dev, 1);
3536 if (ret)
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003537 errors_wait++;
Chris Mason387125f2011-11-18 15:07:51 -05003538 }
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003539 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3540 errors_wait > info->num_tolerated_disk_barrier_failures)
Chris Mason387125f2011-11-18 15:07:51 -05003541 return -EIO;
3542 return 0;
3543}
3544
Zhao Lei943c6e92015-08-19 15:54:15 +08003545int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3546{
Zhao Lei8789f4f2015-09-15 21:08:07 +08003547 int raid_type;
3548 int min_tolerated = INT_MAX;
Zhao Lei943c6e92015-08-19 15:54:15 +08003549
Zhao Lei8789f4f2015-09-15 21:08:07 +08003550 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3551 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3552 min_tolerated = min(min_tolerated,
3553 btrfs_raid_array[BTRFS_RAID_SINGLE].
3554 tolerated_failures);
Zhao Lei943c6e92015-08-19 15:54:15 +08003555
Zhao Lei8789f4f2015-09-15 21:08:07 +08003556 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3557 if (raid_type == BTRFS_RAID_SINGLE)
3558 continue;
3559 if (!(flags & btrfs_raid_group[raid_type]))
3560 continue;
3561 min_tolerated = min(min_tolerated,
3562 btrfs_raid_array[raid_type].
3563 tolerated_failures);
3564 }
Zhao Lei943c6e92015-08-19 15:54:15 +08003565
Zhao Lei8789f4f2015-09-15 21:08:07 +08003566 if (min_tolerated == INT_MAX) {
3567 pr_warn("BTRFS: unknown raid flag: %llu\n", flags);
3568 min_tolerated = 0;
3569 }
3570
3571 return min_tolerated;
Zhao Lei943c6e92015-08-19 15:54:15 +08003572}
3573
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003574int btrfs_calc_num_tolerated_disk_barrier_failures(
3575 struct btrfs_fs_info *fs_info)
3576{
3577 struct btrfs_ioctl_space_info space;
3578 struct btrfs_space_info *sinfo;
3579 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3580 BTRFS_BLOCK_GROUP_SYSTEM,
3581 BTRFS_BLOCK_GROUP_METADATA,
3582 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003583 int i;
3584 int c;
3585 int num_tolerated_disk_barrier_failures =
3586 (int)fs_info->fs_devices->num_devices;
3587
Zhao Lei2c458042015-07-16 15:00:46 +08003588 for (i = 0; i < ARRAY_SIZE(types); i++) {
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003589 struct btrfs_space_info *tmp;
3590
3591 sinfo = NULL;
3592 rcu_read_lock();
3593 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3594 if (tmp->flags == types[i]) {
3595 sinfo = tmp;
3596 break;
3597 }
3598 }
3599 rcu_read_unlock();
3600
3601 if (!sinfo)
3602 continue;
3603
3604 down_read(&sinfo->groups_sem);
3605 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
Zhao Lei2c458042015-07-16 15:00:46 +08003606 u64 flags;
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003607
Zhao Lei2c458042015-07-16 15:00:46 +08003608 if (list_empty(&sinfo->block_groups[c]))
3609 continue;
3610
3611 btrfs_get_block_group_info(&sinfo->block_groups[c],
3612 &space);
3613 if (space.total_bytes == 0 || space.used_bytes == 0)
3614 continue;
3615 flags = space.flags;
Zhao Lei943c6e92015-08-19 15:54:15 +08003616
3617 num_tolerated_disk_barrier_failures = min(
3618 num_tolerated_disk_barrier_failures,
3619 btrfs_get_num_tolerated_disk_barrier_failures(
3620 flags));
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003621 }
3622 up_read(&sinfo->groups_sem);
3623 }
3624
3625 return num_tolerated_disk_barrier_failures;
3626}
3627
Eric Sandeen48a3b632013-04-25 20:41:01 +00003628static int write_all_supers(struct btrfs_root *root, int max_mirrors)
Chris Masonf2984462008-04-10 16:19:33 -04003629{
Chris Masone5e9a522009-06-10 15:17:02 -04003630 struct list_head *head;
Chris Masonf2984462008-04-10 16:19:33 -04003631 struct btrfs_device *dev;
Chris Masona061fc82008-05-07 11:43:44 -04003632 struct btrfs_super_block *sb;
Chris Masonf2984462008-04-10 16:19:33 -04003633 struct btrfs_dev_item *dev_item;
Chris Masonf2984462008-04-10 16:19:33 -04003634 int ret;
3635 int do_barriers;
Chris Masona236aed2008-04-29 09:38:00 -04003636 int max_errors;
3637 int total_errors = 0;
Chris Masona061fc82008-05-07 11:43:44 -04003638 u64 flags;
Chris Masonf2984462008-04-10 16:19:33 -04003639
3640 do_barriers = !btrfs_test_opt(root, NOBARRIER);
Chris Masonaf31f5e2011-11-03 15:17:42 -04003641 backup_super_roots(root->fs_info);
Chris Masonf2984462008-04-10 16:19:33 -04003642
David Sterba6c417612011-04-13 15:41:04 +02003643 sb = root->fs_info->super_for_commit;
Chris Masona061fc82008-05-07 11:43:44 -04003644 dev_item = &sb->dev_item;
Chris Masone5e9a522009-06-10 15:17:02 -04003645
Chris Mason174ba502011-05-27 10:03:58 -04003646 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masone5e9a522009-06-10 15:17:02 -04003647 head = &root->fs_info->fs_devices->devices;
Filipe David Borba Mananad7306802013-08-09 15:41:36 +01003648 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
Chris Mason387125f2011-11-18 15:07:51 -05003649
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003650 if (do_barriers) {
3651 ret = barrier_all_devices(root->fs_info);
3652 if (ret) {
3653 mutex_unlock(
3654 &root->fs_info->fs_devices->device_list_mutex);
Anand Jaina4553fe2015-09-25 14:43:01 +08003655 btrfs_std_error(root->fs_info, ret,
Stefan Behrens5af3e8c2012-08-01 18:56:49 +02003656 "errors while submitting device barriers.");
3657 return ret;
3658 }
3659 }
Chris Mason387125f2011-11-18 15:07:51 -05003660
Xiao Guangrong1f781602011-04-20 10:09:16 +00003661 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003662 if (!dev->bdev) {
3663 total_errors++;
3664 continue;
3665 }
Yan Zheng2b820322008-11-17 21:11:30 -05003666 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003667 continue;
3668
Yan Zheng2b820322008-11-17 21:11:30 -05003669 btrfs_set_stack_device_generation(dev_item, 0);
Chris Masona061fc82008-05-07 11:43:44 -04003670 btrfs_set_stack_device_type(dev_item, dev->type);
3671 btrfs_set_stack_device_id(dev_item, dev->devid);
Miao Xie7df69d32014-07-24 11:37:13 +08003672 btrfs_set_stack_device_total_bytes(dev_item,
Miao Xie935e5cc2014-09-03 21:35:33 +08003673 dev->commit_total_bytes);
Miao Xiece7213c2014-09-03 21:35:34 +08003674 btrfs_set_stack_device_bytes_used(dev_item,
3675 dev->commit_bytes_used);
Chris Masona061fc82008-05-07 11:43:44 -04003676 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3677 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3678 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3679 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Yan Zheng2b820322008-11-17 21:11:30 -05003680 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Yan Zhenga512bbf2008-12-08 16:46:26 -05003681
Chris Masona061fc82008-05-07 11:43:44 -04003682 flags = btrfs_super_flags(sb);
3683 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
Chris Masonf2984462008-04-10 16:19:33 -04003684
Yan Zhenga512bbf2008-12-08 16:46:26 -05003685 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
Chris Masona236aed2008-04-29 09:38:00 -04003686 if (ret)
3687 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003688 }
Chris Masona236aed2008-04-29 09:38:00 -04003689 if (total_errors > max_errors) {
Frank Holtonefe120a2013-12-20 11:37:06 -05003690 btrfs_err(root->fs_info, "%d errors while writing supers",
Chris Masond3977122009-01-05 21:25:51 -05003691 total_errors);
Stefan Behrensa724b432013-09-11 09:59:22 +02003692 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003693
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003694 /* FUA is masked off if unsupported and can't be the reason */
Anand Jaina4553fe2015-09-25 14:43:01 +08003695 btrfs_std_error(root->fs_info, -EIO,
Stefan Behrens9d565ba2013-08-09 17:08:40 +02003696 "%d errors while writing supers", total_errors);
3697 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003698 }
Chris Masonf2984462008-04-10 16:19:33 -04003699
Yan Zhenga512bbf2008-12-08 16:46:26 -05003700 total_errors = 0;
Xiao Guangrong1f781602011-04-20 10:09:16 +00003701 list_for_each_entry_rcu(dev, head, dev_list) {
Chris Masondfe25022008-05-13 13:46:40 -04003702 if (!dev->bdev)
3703 continue;
Yan Zheng2b820322008-11-17 21:11:30 -05003704 if (!dev->in_fs_metadata || !dev->writeable)
Chris Masondfe25022008-05-13 13:46:40 -04003705 continue;
3706
Yan Zhenga512bbf2008-12-08 16:46:26 -05003707 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3708 if (ret)
3709 total_errors++;
Chris Masonf2984462008-04-10 16:19:33 -04003710 }
Chris Mason174ba502011-05-27 10:03:58 -04003711 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
Chris Masona236aed2008-04-29 09:38:00 -04003712 if (total_errors > max_errors) {
Anand Jaina4553fe2015-09-25 14:43:01 +08003713 btrfs_std_error(root->fs_info, -EIO,
Jeff Mahoney79787ea2012-03-12 16:03:00 +01003714 "%d errors while writing supers", total_errors);
3715 return -EIO;
Chris Masona236aed2008-04-29 09:38:00 -04003716 }
Chris Masonf2984462008-04-10 16:19:33 -04003717 return 0;
3718}
3719
Yan Zhenga512bbf2008-12-08 16:46:26 -05003720int write_ctree_super(struct btrfs_trans_handle *trans,
3721 struct btrfs_root *root, int max_mirrors)
Chris Masoncfaa7292007-02-21 17:04:57 -05003722{
Rashikaf570e7572013-10-31 16:50:42 +05303723 return write_all_supers(root, max_mirrors);
Chris Masoncfaa7292007-02-21 17:04:57 -05003724}
3725
Miao Xiecb517ea2013-05-15 07:48:19 +00003726/* Drop a fs root from the radix tree and free it. */
3727void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3728 struct btrfs_root *root)
Chris Mason2619ba12007-04-10 16:58:11 -04003729{
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003730 spin_lock(&fs_info->fs_roots_radix_lock);
Chris Mason2619ba12007-04-10 16:58:11 -04003731 radix_tree_delete(&fs_info->fs_roots_radix,
3732 (unsigned long)root->root_key.objectid);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003733 spin_unlock(&fs_info->fs_roots_radix_lock);
Yan, Zheng76dda932009-09-21 16:00:26 -04003734
3735 if (btrfs_root_refs(&root->root_item) == 0)
3736 synchronize_srcu(&fs_info->subvol_srcu);
3737
Liu Bo1a4319c2014-01-13 19:53:53 +08003738 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Liu Bo33217192013-02-27 13:28:24 +00003739 btrfs_free_log(NULL, root);
Liu Bo33217192013-02-27 13:28:24 +00003740
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003741 if (root->free_ino_pinned)
3742 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3743 if (root->free_ino_ctl)
3744 __btrfs_remove_free_space_cache(root->free_ino_ctl);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003745 free_fs_root(root);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003746}
3747
3748static void free_fs_root(struct btrfs_root *root)
3749{
David Sterba57cdc8d2014-02-05 02:37:48 +01003750 iput(root->ino_cache_inode);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003751 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003752 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3753 root->orphan_block_rsv = NULL;
Al Viro0ee5dc62011-07-07 15:44:25 -04003754 if (root->anon_dev)
3755 free_anon_bdev(root->anon_dev);
Miao Xie8257b2d2014-03-06 13:38:19 +08003756 if (root->subv_writers)
3757 btrfs_free_subvolume_writers(root->subv_writers);
Yan, Zheng4df27c4d2009-09-21 15:56:00 -04003758 free_extent_buffer(root->node);
3759 free_extent_buffer(root->commit_root);
Li Zefan581bb052011-04-20 10:06:11 +08003760 kfree(root->free_ino_ctl);
3761 kfree(root->free_ino_pinned);
Chris Masond3977122009-01-05 21:25:51 -05003762 kfree(root->name);
Miao Xieb0feb9d2013-05-15 07:48:20 +00003763 btrfs_put_fs_root(root);
Chris Mason2619ba12007-04-10 16:58:11 -04003764}
3765
Miao Xiecb517ea2013-05-15 07:48:19 +00003766void btrfs_free_fs_root(struct btrfs_root *root)
3767{
3768 free_fs_root(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003769}
3770
Yan Zhengc146afa2008-11-12 14:34:12 -05003771int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3772{
3773 u64 root_objectid = 0;
3774 struct btrfs_root *gang[8];
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003775 int i = 0;
3776 int err = 0;
3777 unsigned int ret = 0;
3778 int index;
Yan Zhengc146afa2008-11-12 14:34:12 -05003779
3780 while (1) {
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003781 index = srcu_read_lock(&fs_info->subvol_srcu);
Yan Zhengc146afa2008-11-12 14:34:12 -05003782 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3783 (void **)gang, root_objectid,
3784 ARRAY_SIZE(gang));
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003785 if (!ret) {
3786 srcu_read_unlock(&fs_info->subvol_srcu, index);
Yan Zhengc146afa2008-11-12 14:34:12 -05003787 break;
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003788 }
Yan Zheng5d4f98a2009-06-10 10:45:14 -04003789 root_objectid = gang[ret - 1]->root_key.objectid + 1;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003790
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003791 for (i = 0; i < ret; i++) {
3792 /* Avoid to grab roots in dead_roots */
3793 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3794 gang[i] = NULL;
3795 continue;
3796 }
3797 /* grab all the search result for later use */
3798 gang[i] = btrfs_grab_fs_root(gang[i]);
3799 }
3800 srcu_read_unlock(&fs_info->subvol_srcu, index);
3801
3802 for (i = 0; i < ret; i++) {
3803 if (!gang[i])
3804 continue;
Yan Zhengc146afa2008-11-12 14:34:12 -05003805 root_objectid = gang[i]->root_key.objectid;
Josef Bacik66b4ffd2011-01-31 16:22:42 -05003806 err = btrfs_orphan_cleanup(gang[i]);
3807 if (err)
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003808 break;
3809 btrfs_put_fs_root(gang[i]);
Yan Zhengc146afa2008-11-12 14:34:12 -05003810 }
3811 root_objectid++;
3812 }
Qu Wenruo65d33fd2014-04-22 17:13:51 +08003813
3814 /* release the uncleaned roots due to error */
3815 for (; i < ret; i++) {
3816 if (gang[i])
3817 btrfs_put_fs_root(gang[i]);
3818 }
3819 return err;
Yan Zhengc146afa2008-11-12 14:34:12 -05003820}
3821
3822int btrfs_commit_super(struct btrfs_root *root)
3823{
3824 struct btrfs_trans_handle *trans;
Yan Zhengc146afa2008-11-12 14:34:12 -05003825
3826 mutex_lock(&root->fs_info->cleaner_mutex);
Yan, Zheng24bbcf042009-11-12 09:36:34 +00003827 btrfs_run_delayed_iputs(root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003828 mutex_unlock(&root->fs_info->cleaner_mutex);
David Sterba9d1a2a32013-03-12 15:13:28 +00003829 wake_up_process(root->fs_info->cleaner_kthread);
Yan, Zhengc71bf092009-11-12 09:34:40 +00003830
3831 /* wait until ongoing cleanup work done */
3832 down_write(&root->fs_info->cleanup_work_sem);
3833 up_write(&root->fs_info->cleanup_work_sem);
3834
Josef Bacik7a7eaa42011-04-13 12:54:33 -04003835 trans = btrfs_join_transaction(root);
Tsutomu Itoh3612b492011-01-25 02:51:38 +00003836 if (IS_ERR(trans))
3837 return PTR_ERR(trans);
Liu Bod52c1bc2013-11-05 11:45:53 +08003838 return btrfs_commit_transaction(trans, root);
Yan Zhengc146afa2008-11-12 14:34:12 -05003839}
3840
David Sterba3abdbd72014-06-04 18:10:45 +02003841void close_ctree(struct btrfs_root *root)
Chris Masoneb60cea2007-02-02 09:18:22 -05003842{
Chris Mason0f7d52f2007-04-09 10:42:37 -04003843 struct btrfs_fs_info *fs_info = root->fs_info;
Yan Zhengc146afa2008-11-12 14:34:12 -05003844 int ret;
Chris Masone089f052007-03-16 16:20:31 -04003845
Chris Masonfacda1e2007-06-08 18:11:48 -04003846 fs_info->closing = 1;
Chris Masona2135012008-06-25 16:01:30 -04003847 smp_mb();
3848
Justin Maggard7343dd62015-11-04 15:56:16 -08003849 /* wait for the qgroup rescan worker to stop */
3850 btrfs_qgroup_wait_for_completion(fs_info);
3851
Stefan Behrens803b2f52013-08-15 17:11:21 +02003852 /* wait for the uuid_scan task to finish */
3853 down(&fs_info->uuid_tree_rescan_sem);
3854 /* avoid complains from lockdep et al., set sem back to initial state */
3855 up(&fs_info->uuid_tree_rescan_sem);
3856
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003857 /* pause restriper - we want to resume on mount */
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003858 btrfs_pause_balance(fs_info);
Ilya Dryomov837d5b62012-01-16 22:04:49 +02003859
Stefan Behrens8dabb742012-11-06 13:15:27 +01003860 btrfs_dev_replace_suspend_for_unmount(fs_info);
3861
Stefan Behrensaa1b8cd2012-11-05 17:03:39 +01003862 btrfs_scrub_cancel(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003863
3864 /* wait for any defraggers to finish */
3865 wait_event(fs_info->transaction_wait,
3866 (atomic_read(&fs_info->defrag_running) == 0));
3867
3868 /* clear out the rbtree of defraggable inodes */
Miao Xie26176e72012-11-26 09:26:20 +00003869 btrfs_cleanup_defrag_inodes(fs_info);
Chris Mason4cb53002011-05-24 15:35:30 -04003870
Miao Xie21c7e752014-05-13 17:29:04 -07003871 cancel_work_sync(&fs_info->async_reclaim_work);
3872
Yan Zhengc146afa2008-11-12 14:34:12 -05003873 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003874 /*
3875 * If the cleaner thread is stopped and there are
3876 * block groups queued for removal, the deletion will be
3877 * skipped when we quit the cleaner thread.
3878 */
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003879 btrfs_delete_unused_bgs(root->fs_info);
Jeff Mahoneye44163e2015-06-15 09:41:18 -04003880
liuboacce9522011-01-06 19:30:25 +08003881 ret = btrfs_commit_super(root);
3882 if (ret)
Eric Sandeen04892342014-08-01 18:12:36 -05003883 btrfs_err(fs_info, "commit super ret %d", ret);
liuboacce9522011-01-06 19:30:25 +08003884 }
3885
Miao Xie87533c42013-01-29 10:14:48 +00003886 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
Stefan Behrens68ce9682012-08-01 05:45:52 -06003887 btrfs_error_commit_super(root);
Chris Masoned2ff2c2007-03-01 18:59:40 -05003888
Al Viroe3029d92011-11-17 00:56:18 -05003889 kthread_stop(fs_info->transaction_kthread);
3890 kthread_stop(fs_info->cleaner_kthread);
Yan, Zheng8929ecfa2010-05-16 10:49:58 -04003891
Yan Zhengf25784b2009-07-28 08:41:57 -04003892 fs_info->closing = 2;
3893 smp_mb();
3894
Eric Sandeen04892342014-08-01 18:12:36 -05003895 btrfs_free_qgroup_config(fs_info);
Arne Jansenbcef60f2011-09-13 15:23:30 +02003896
Miao Xie963d6782013-01-29 10:10:51 +00003897 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
Eric Sandeen04892342014-08-01 18:12:36 -05003898 btrfs_info(fs_info, "at unmount delalloc count %lld",
Miao Xie963d6782013-01-29 10:10:51 +00003899 percpu_counter_sum(&fs_info->delalloc_bytes));
Chris Masonb0c68f82008-01-31 11:05:37 -05003900 }
Yanbcc63ab2008-07-30 16:29:20 -04003901
Anand Jain6618a592015-08-14 18:32:47 +08003902 btrfs_sysfs_remove_mounted(fs_info);
Anand Jainb7c35e82015-03-10 06:38:38 +08003903 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
Jeff Mahoney5ac1d202013-11-01 13:06:58 -04003904
Josef Bacikfaa2dbf2014-05-07 17:06:09 -04003905 btrfs_free_fs_roots(fs_info);
Chris Masond10c5f32007-12-17 20:14:04 -05003906
Liu Bo1a4319c2014-01-13 19:53:53 +08003907 btrfs_put_block_group_cache(fs_info);
3908
Josef Bacik2b1360d2013-10-07 15:14:44 -04003909 btrfs_free_block_groups(fs_info);
3910
Wang Shilongde348ee2014-04-09 19:23:22 +08003911 /*
3912 * we must make sure there is not any read request to
3913 * submit after we stopping all workers.
3914 */
3915 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
Josef Bacik96192492013-10-16 13:53:28 -04003916 btrfs_stop_all_workers(fs_info);
3917
Josef Bacik47ab2a62014-09-18 11:20:02 -04003918 fs_info->open = 0;
Josef Bacik13e6c372013-05-30 16:55:44 -04003919 free_root_pointers(fs_info, 1);
Chris Mason9ad6b7b2008-04-18 16:11:30 -04003920
Josef Bacik13e6c372013-05-30 16:55:44 -04003921 iput(fs_info->btree_inode);
Chris Masond6bfde82008-04-30 13:59:35 -04003922
Stefan Behrens21adbd52011-11-09 13:44:05 +01003923#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3924 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3925 btrfsic_unmount(root, fs_info->fs_devices);
3926#endif
3927
Chris Masondfe25022008-05-13 13:46:40 -04003928 btrfs_close_devices(fs_info->fs_devices);
Chris Mason0b86a832008-03-24 15:01:56 -04003929 btrfs_mapping_tree_free(&fs_info->mapping_tree);
Chris Masonb248a412008-04-14 09:48:18 -04003930
Miao Xiee2d84522013-01-29 10:09:20 +00003931 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
Miao Xie963d6782013-01-29 10:10:51 +00003932 percpu_counter_destroy(&fs_info->delalloc_bytes);
Miao Xiec404e0d2014-01-30 16:46:55 +08003933 percpu_counter_destroy(&fs_info->bio_counter);
Chris Mason04160082008-03-26 10:28:07 -04003934 bdi_destroy(&fs_info->bdi);
Yan, Zheng76dda932009-09-21 16:00:26 -04003935 cleanup_srcu_struct(&fs_info->subvol_srcu);
Chris Mason0b86a832008-03-24 15:01:56 -04003936
David Woodhouse53b381b2013-01-29 18:40:14 -05003937 btrfs_free_stripe_hash_table(fs_info);
3938
Chris Masoncdfb0802015-04-06 18:17:00 -07003939 __btrfs_free_block_rsv(root->orphan_block_rsv);
Filipe David Borba Manana1cb048f2013-08-20 00:52:51 +01003940 root->orphan_block_rsv = NULL;
Filipe Manana04216822014-11-27 21:14:15 +00003941
3942 lock_chunks(root);
3943 while (!list_empty(&fs_info->pinned_chunks)) {
3944 struct extent_map *em;
3945
3946 em = list_first_entry(&fs_info->pinned_chunks,
3947 struct extent_map, list);
3948 list_del_init(&em->list);
3949 free_extent_map(em);
3950 }
3951 unlock_chunks(root);
Chris Masoneb60cea2007-02-02 09:18:22 -05003952}
3953
Chris Masonb9fab912012-05-06 07:23:47 -04003954int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3955 int atomic)
Chris Masonccd467d2007-06-28 15:57:36 -04003956{
Chris Mason1259ab72008-05-12 13:39:03 -04003957 int ret;
Chris Mason727011e2010-08-06 13:21:20 -04003958 struct inode *btree_inode = buf->pages[0]->mapping->host;
Chris Mason1259ab72008-05-12 13:39:03 -04003959
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003960 ret = extent_buffer_uptodate(buf);
Chris Mason1259ab72008-05-12 13:39:03 -04003961 if (!ret)
3962 return ret;
3963
3964 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
Chris Masonb9fab912012-05-06 07:23:47 -04003965 parent_transid, atomic);
3966 if (ret == -EAGAIN)
3967 return ret;
Chris Mason1259ab72008-05-12 13:39:03 -04003968 return !ret;
Chris Mason5f39d392007-10-15 16:14:19 -04003969}
Chris Mason6702ed42007-08-07 16:15:09 -04003970
Chris Mason5f39d392007-10-15 16:14:19 -04003971void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3972{
Josef Bacik06ea65a2013-09-19 16:07:01 -04003973 struct btrfs_root *root;
Chris Mason5f39d392007-10-15 16:14:19 -04003974 u64 transid = btrfs_header_generation(buf);
Chris Masonb9473432009-03-13 11:00:37 -04003975 int was_dirty;
Chris Masonb4ce94d2009-02-04 09:25:08 -05003976
Josef Bacik06ea65a2013-09-19 16:07:01 -04003977#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3978 /*
3979 * This is a fast path so only do this check if we have sanity tests
3980 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3981 * outside of the sanity tests.
3982 */
3983 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3984 return;
3985#endif
3986 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Chris Masonb9447ef2009-03-09 11:45:38 -04003987 btrfs_assert_tree_locked(buf);
Julia Lawall31b1a2b2012-11-03 10:58:34 +00003988 if (transid != root->fs_info->generation)
3989 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
Chris Masond3977122009-01-05 21:25:51 -05003990 "found %llu running %llu\n",
Geert Uytterhoevenc1c9ff72013-08-20 13:20:07 +02003991 buf->start, transid, root->fs_info->generation);
Josef Bacik0b32f4b2012-03-13 09:38:00 -04003992 was_dirty = set_extent_buffer_dirty(buf);
Miao Xiee2d84522013-01-29 10:09:20 +00003993 if (!was_dirty)
3994 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3995 buf->len,
3996 root->fs_info->dirty_metadata_batch);
Filipe Manana1f21ef02014-04-09 15:37:06 +01003997#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3998 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3999 btrfs_print_leaf(root, buf);
4000 ASSERT(0);
4001 }
4002#endif
Chris Masoneb60cea2007-02-02 09:18:22 -05004003}
4004
Liu Bob53d3f52012-11-14 14:34:34 +00004005static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
4006 int flush_delayed)
Chris Mason35b7e472007-05-02 15:53:43 -04004007{
Chris Mason188de642008-05-09 11:52:25 -04004008 /*
4009 * looks as though older kernels can get into trouble with
4010 * this code, they end up stuck in balance_dirty_pages forever
4011 */
Miao Xiee2d84522013-01-29 10:09:20 +00004012 int ret;
Chris Masond6bfde82008-04-30 13:59:35 -04004013
Jens Axboe6933c022009-03-17 09:36:37 +01004014 if (current->flags & PF_MEMALLOC)
Chris Masond6bfde82008-04-30 13:59:35 -04004015 return;
4016
Liu Bob53d3f52012-11-14 14:34:34 +00004017 if (flush_delayed)
4018 btrfs_balance_delayed_items(root);
Miao Xie16cdcec2011-04-22 18:12:22 +08004019
Miao Xiee2d84522013-01-29 10:09:20 +00004020 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
4021 BTRFS_DIRTY_METADATA_THRESH);
4022 if (ret > 0) {
Namjae Jeond0e1d662012-12-11 16:00:21 -08004023 balance_dirty_pages_ratelimited(
4024 root->fs_info->btree_inode->i_mapping);
Miao Xie16cdcec2011-04-22 18:12:22 +08004025 }
Miao Xie16cdcec2011-04-22 18:12:22 +08004026}
4027
Liu Bob53d3f52012-11-14 14:34:34 +00004028void btrfs_btree_balance_dirty(struct btrfs_root *root)
Miao Xie16cdcec2011-04-22 18:12:22 +08004029{
Liu Bob53d3f52012-11-14 14:34:34 +00004030 __btrfs_btree_balance_dirty(root, 1);
4031}
Miao Xie16cdcec2011-04-22 18:12:22 +08004032
Liu Bob53d3f52012-11-14 14:34:34 +00004033void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
4034{
4035 __btrfs_btree_balance_dirty(root, 0);
Chris Mason35b7e472007-05-02 15:53:43 -04004036}
Chris Mason6b800532007-10-15 16:17:34 -04004037
Chris Masonca7a79a2008-05-12 12:59:19 -04004038int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
Chris Mason6b800532007-10-15 16:17:34 -04004039{
Chris Mason727011e2010-08-06 13:21:20 -04004040 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
Josef Bacik0b32f4b2012-03-13 09:38:00 -04004041 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
Chris Mason6b800532007-10-15 16:17:34 -04004042}
Chris Mason0da54682007-11-07 21:08:01 -05004043
David Sterbafcd1f0652012-03-06 00:06:18 +01004044static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
liuboacce9522011-01-06 19:30:25 +08004045 int read_only)
4046{
David Sterbac9260932014-09-30 19:16:47 +02004047 struct btrfs_super_block *sb = fs_info->super_copy;
4048 int ret = 0;
4049
David Sterba21e76262014-10-27 13:52:21 +01004050 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
4051 printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n",
4052 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004053 ret = -EINVAL;
4054 }
David Sterba21e76262014-10-27 13:52:21 +01004055 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
4056 printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n",
4057 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004058 ret = -EINVAL;
4059 }
David Sterba21e76262014-10-27 13:52:21 +01004060 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
4061 printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n",
4062 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
David Sterbac9260932014-09-30 19:16:47 +02004063 ret = -EINVAL;
4064 }
4065
David Sterba1104a882013-03-06 15:57:46 +01004066 /*
David Sterbac9260932014-09-30 19:16:47 +02004067 * The common minimum, we don't know if we can trust the nodesize/sectorsize
4068 * items yet, they'll be verified later. Issue just a warning.
David Sterba1104a882013-03-06 15:57:46 +01004069 */
David Sterba21e76262014-10-27 13:52:21 +01004070 if (!IS_ALIGNED(btrfs_super_root(sb), 4096))
David Sterbac9260932014-09-30 19:16:47 +02004071 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
David Sterbacd743fa2014-10-31 19:40:14 +01004072 btrfs_super_root(sb));
David Sterba21e76262014-10-27 13:52:21 +01004073 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), 4096))
David Sterbacd743fa2014-10-31 19:40:14 +01004074 printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n",
4075 btrfs_super_chunk_root(sb));
David Sterba21e76262014-10-27 13:52:21 +01004076 if (!IS_ALIGNED(btrfs_super_log_root(sb), 4096))
David Sterbacd743fa2014-10-31 19:40:14 +01004077 printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004078 btrfs_super_log_root(sb));
David Sterbac9260932014-09-30 19:16:47 +02004079
David Sterba75d6ad32014-10-31 17:18:08 +01004080 /*
4081 * Check the lower bound, the alignment and other constraints are
4082 * checked later.
4083 */
4084 if (btrfs_super_nodesize(sb) < 4096) {
4085 printk(KERN_ERR "BTRFS: nodesize too small: %u < 4096\n",
4086 btrfs_super_nodesize(sb));
4087 ret = -EINVAL;
4088 }
4089 if (btrfs_super_sectorsize(sb) < 4096) {
4090 printk(KERN_ERR "BTRFS: sectorsize too small: %u < 4096\n",
4091 btrfs_super_sectorsize(sb));
4092 ret = -EINVAL;
4093 }
4094
David Sterbac9260932014-09-30 19:16:47 +02004095 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
4096 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
4097 fs_info->fsid, sb->dev_item.fsid);
4098 ret = -EINVAL;
4099 }
4100
4101 /*
4102 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4103 * done later
4104 */
David Sterba21e76262014-10-27 13:52:21 +01004105 if (btrfs_super_num_devices(sb) > (1UL << 31))
David Sterbac9260932014-09-30 19:16:47 +02004106 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004107 btrfs_super_num_devices(sb));
David Sterba75d6ad32014-10-31 17:18:08 +01004108 if (btrfs_super_num_devices(sb) == 0) {
4109 printk(KERN_ERR "BTRFS: number of devices is 0\n");
4110 ret = -EINVAL;
4111 }
David Sterbac9260932014-09-30 19:16:47 +02004112
David Sterba21e76262014-10-27 13:52:21 +01004113 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
David Sterbac9260932014-09-30 19:16:47 +02004114 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
David Sterba21e76262014-10-27 13:52:21 +01004115 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
David Sterbac9260932014-09-30 19:16:47 +02004116 ret = -EINVAL;
4117 }
4118
4119 /*
David Sterbace7fca52014-10-31 18:42:05 +01004120 * Obvious sys_chunk_array corruptions, it must hold at least one key
4121 * and one chunk
4122 */
4123 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
4124 printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
4125 btrfs_super_sys_array_size(sb),
4126 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
4127 ret = -EINVAL;
4128 }
4129 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4130 + sizeof(struct btrfs_chunk)) {
Fabian Frederickd2207122015-02-14 13:10:22 +01004131 printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n",
David Sterbace7fca52014-10-31 18:42:05 +01004132 btrfs_super_sys_array_size(sb),
4133 sizeof(struct btrfs_disk_key)
4134 + sizeof(struct btrfs_chunk));
4135 ret = -EINVAL;
4136 }
4137
4138 /*
David Sterbac9260932014-09-30 19:16:47 +02004139 * The generation is a global counter, we'll trust it more than the others
4140 * but it's still possible that it's the one that's wrong.
4141 */
David Sterba21e76262014-10-27 13:52:21 +01004142 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
David Sterbac9260932014-09-30 19:16:47 +02004143 printk(KERN_WARNING
4144 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004145 btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb));
4146 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4147 && btrfs_super_cache_generation(sb) != (u64)-1)
David Sterbac9260932014-09-30 19:16:47 +02004148 printk(KERN_WARNING
4149 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
David Sterba21e76262014-10-27 13:52:21 +01004150 btrfs_super_generation(sb), btrfs_super_cache_generation(sb));
David Sterbac9260932014-09-30 19:16:47 +02004151
4152 return ret;
liuboacce9522011-01-06 19:30:25 +08004153}
4154
Eric Sandeen48a3b632013-04-25 20:41:01 +00004155static void btrfs_error_commit_super(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004156{
liuboacce9522011-01-06 19:30:25 +08004157 mutex_lock(&root->fs_info->cleaner_mutex);
4158 btrfs_run_delayed_iputs(root);
4159 mutex_unlock(&root->fs_info->cleaner_mutex);
4160
4161 down_write(&root->fs_info->cleanup_work_sem);
4162 up_write(&root->fs_info->cleanup_work_sem);
4163
4164 /* cleanup FS via transaction */
4165 btrfs_cleanup_transaction(root);
liuboacce9522011-01-06 19:30:25 +08004166}
4167
Jeff Mahoney143bede2012-03-01 14:56:26 +01004168static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004169{
liuboacce9522011-01-06 19:30:25 +08004170 struct btrfs_ordered_extent *ordered;
liuboacce9522011-01-06 19:30:25 +08004171
Miao Xie199c2a92013-05-15 07:48:23 +00004172 spin_lock(&root->ordered_extent_lock);
Josef Bacik779880e2013-01-31 14:30:08 -05004173 /*
4174 * This will just short circuit the ordered completion stuff which will
4175 * make sure the ordered extent gets properly cleaned up.
4176 */
Miao Xie199c2a92013-05-15 07:48:23 +00004177 list_for_each_entry(ordered, &root->ordered_extents,
Josef Bacik779880e2013-01-31 14:30:08 -05004178 root_extent_list)
4179 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
Miao Xie199c2a92013-05-15 07:48:23 +00004180 spin_unlock(&root->ordered_extent_lock);
4181}
4182
4183static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4184{
4185 struct btrfs_root *root;
4186 struct list_head splice;
4187
4188 INIT_LIST_HEAD(&splice);
4189
4190 spin_lock(&fs_info->ordered_root_lock);
4191 list_splice_init(&fs_info->ordered_roots, &splice);
4192 while (!list_empty(&splice)) {
4193 root = list_first_entry(&splice, struct btrfs_root,
4194 ordered_root);
Josef Bacik1de2cfd2013-09-27 16:36:02 -04004195 list_move_tail(&root->ordered_root,
4196 &fs_info->ordered_roots);
Miao Xie199c2a92013-05-15 07:48:23 +00004197
Liu Bo2a85d9c2014-02-10 17:07:16 +08004198 spin_unlock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004199 btrfs_destroy_ordered_extents(root);
4200
Liu Bo2a85d9c2014-02-10 17:07:16 +08004201 cond_resched();
4202 spin_lock(&fs_info->ordered_root_lock);
Miao Xie199c2a92013-05-15 07:48:23 +00004203 }
4204 spin_unlock(&fs_info->ordered_root_lock);
liuboacce9522011-01-06 19:30:25 +08004205}
4206
Stefan Behrens35a36212013-08-14 18:12:25 +02004207static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4208 struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004209{
4210 struct rb_node *node;
4211 struct btrfs_delayed_ref_root *delayed_refs;
4212 struct btrfs_delayed_ref_node *ref;
4213 int ret = 0;
4214
4215 delayed_refs = &trans->delayed_refs;
4216
4217 spin_lock(&delayed_refs->lock);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004218 if (atomic_read(&delayed_refs->num_entries) == 0) {
David Sterbacfece4d2011-04-25 19:43:52 -04004219 spin_unlock(&delayed_refs->lock);
Frank Holtonefe120a2013-12-20 11:37:06 -05004220 btrfs_info(root->fs_info, "delayed_refs has NO entry");
liuboacce9522011-01-06 19:30:25 +08004221 return ret;
4222 }
4223
Josef Bacikd7df2c72014-01-23 09:21:38 -05004224 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4225 struct btrfs_delayed_ref_head *head;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004226 struct btrfs_delayed_ref_node *tmp;
Josef Bacike78417d2013-06-03 16:42:36 -04004227 bool pin_bytes = false;
liuboacce9522011-01-06 19:30:25 +08004228
Josef Bacikd7df2c72014-01-23 09:21:38 -05004229 head = rb_entry(node, struct btrfs_delayed_ref_head,
4230 href_node);
4231 if (!mutex_trylock(&head->mutex)) {
4232 atomic_inc(&head->node.refs);
4233 spin_unlock(&delayed_refs->lock);
liuboacce9522011-01-06 19:30:25 +08004234
Josef Bacikd7df2c72014-01-23 09:21:38 -05004235 mutex_lock(&head->mutex);
Josef Bacike78417d2013-06-03 16:42:36 -04004236 mutex_unlock(&head->mutex);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004237 btrfs_put_delayed_ref(&head->node);
4238 spin_lock(&delayed_refs->lock);
4239 continue;
Josef Bacike78417d2013-06-03 16:42:36 -04004240 }
Josef Bacikd7df2c72014-01-23 09:21:38 -05004241 spin_lock(&head->lock);
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004242 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4243 list) {
Josef Bacikd7df2c72014-01-23 09:21:38 -05004244 ref->in_tree = 0;
Qu Wenruoc6fc2452015-03-30 17:03:00 +08004245 list_del(&ref->list);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004246 atomic_dec(&delayed_refs->num_entries);
4247 btrfs_put_delayed_ref(ref);
Josef Bacikd7df2c72014-01-23 09:21:38 -05004248 }
4249 if (head->must_insert_reserved)
4250 pin_bytes = true;
4251 btrfs_free_delayed_extent_op(head->extent_op);
4252 delayed_refs->num_heads--;
4253 if (head->processing == 0)
4254 delayed_refs->num_heads_ready--;
4255 atomic_dec(&delayed_refs->num_entries);
4256 head->node.in_tree = 0;
4257 rb_erase(&head->href_node, &delayed_refs->href_root);
4258 spin_unlock(&head->lock);
4259 spin_unlock(&delayed_refs->lock);
4260 mutex_unlock(&head->mutex);
liuboacce9522011-01-06 19:30:25 +08004261
Josef Bacikd7df2c72014-01-23 09:21:38 -05004262 if (pin_bytes)
4263 btrfs_pin_extent(root, head->node.bytenr,
4264 head->node.num_bytes, 1);
4265 btrfs_put_delayed_ref(&head->node);
liuboacce9522011-01-06 19:30:25 +08004266 cond_resched();
4267 spin_lock(&delayed_refs->lock);
4268 }
4269
4270 spin_unlock(&delayed_refs->lock);
4271
4272 return ret;
4273}
4274
Jeff Mahoney143bede2012-03-01 14:56:26 +01004275static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004276{
4277 struct btrfs_inode *btrfs_inode;
4278 struct list_head splice;
4279
4280 INIT_LIST_HEAD(&splice);
4281
Miao Xieeb73c1b2013-05-15 07:48:22 +00004282 spin_lock(&root->delalloc_lock);
4283 list_splice_init(&root->delalloc_inodes, &splice);
liuboacce9522011-01-06 19:30:25 +08004284
4285 while (!list_empty(&splice)) {
Miao Xieeb73c1b2013-05-15 07:48:22 +00004286 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4287 delalloc_inodes);
liuboacce9522011-01-06 19:30:25 +08004288
4289 list_del_init(&btrfs_inode->delalloc_inodes);
Miao Xiedf0af1a2013-01-29 10:11:59 +00004290 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4291 &btrfs_inode->runtime_flags);
Miao Xieeb73c1b2013-05-15 07:48:22 +00004292 spin_unlock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004293
4294 btrfs_invalidate_inodes(btrfs_inode->root);
Miao Xieb216cbf2013-05-15 07:48:21 +00004295
Miao Xieeb73c1b2013-05-15 07:48:22 +00004296 spin_lock(&root->delalloc_lock);
liuboacce9522011-01-06 19:30:25 +08004297 }
4298
Miao Xieeb73c1b2013-05-15 07:48:22 +00004299 spin_unlock(&root->delalloc_lock);
4300}
4301
4302static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4303{
4304 struct btrfs_root *root;
4305 struct list_head splice;
4306
4307 INIT_LIST_HEAD(&splice);
4308
4309 spin_lock(&fs_info->delalloc_root_lock);
4310 list_splice_init(&fs_info->delalloc_roots, &splice);
4311 while (!list_empty(&splice)) {
4312 root = list_first_entry(&splice, struct btrfs_root,
4313 delalloc_root);
4314 list_del_init(&root->delalloc_root);
4315 root = btrfs_grab_fs_root(root);
4316 BUG_ON(!root);
4317 spin_unlock(&fs_info->delalloc_root_lock);
4318
4319 btrfs_destroy_delalloc_inodes(root);
4320 btrfs_put_fs_root(root);
4321
4322 spin_lock(&fs_info->delalloc_root_lock);
4323 }
4324 spin_unlock(&fs_info->delalloc_root_lock);
liuboacce9522011-01-06 19:30:25 +08004325}
4326
4327static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4328 struct extent_io_tree *dirty_pages,
4329 int mark)
4330{
4331 int ret;
liuboacce9522011-01-06 19:30:25 +08004332 struct extent_buffer *eb;
4333 u64 start = 0;
4334 u64 end;
liuboacce9522011-01-06 19:30:25 +08004335
4336 while (1) {
4337 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004338 mark, NULL);
liuboacce9522011-01-06 19:30:25 +08004339 if (ret)
4340 break;
4341
4342 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4343 while (start <= end) {
Daniel Dressler01d58472014-11-21 17:15:07 +09004344 eb = btrfs_find_tree_block(root->fs_info, start);
David Sterba707e8a02014-06-04 19:22:26 +02004345 start += root->nodesize;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004346 if (!eb)
liuboacce9522011-01-06 19:30:25 +08004347 continue;
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004348 wait_on_extent_buffer_writeback(eb);
liuboacce9522011-01-06 19:30:25 +08004349
Josef Bacikfd8b2b62013-04-24 16:41:19 -04004350 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4351 &eb->bflags))
4352 clear_extent_buffer_dirty(eb);
4353 free_extent_buffer_stale(eb);
liuboacce9522011-01-06 19:30:25 +08004354 }
4355 }
4356
4357 return ret;
4358}
4359
4360static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4361 struct extent_io_tree *pinned_extents)
4362{
4363 struct extent_io_tree *unpin;
4364 u64 start;
4365 u64 end;
4366 int ret;
Liu Boed0eaa12012-06-14 02:23:21 -06004367 bool loop = true;
liuboacce9522011-01-06 19:30:25 +08004368
4369 unpin = pinned_extents;
Liu Boed0eaa12012-06-14 02:23:21 -06004370again:
liuboacce9522011-01-06 19:30:25 +08004371 while (1) {
4372 ret = find_first_extent_bit(unpin, 0, &start, &end,
Josef Bacike6138872012-09-27 17:07:30 -04004373 EXTENT_DIRTY, NULL);
liuboacce9522011-01-06 19:30:25 +08004374 if (ret)
4375 break;
4376
liuboacce9522011-01-06 19:30:25 +08004377 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4378 btrfs_error_unpin_extent_range(root, start, end);
4379 cond_resched();
4380 }
4381
Liu Boed0eaa12012-06-14 02:23:21 -06004382 if (loop) {
4383 if (unpin == &root->fs_info->freed_extents[0])
4384 unpin = &root->fs_info->freed_extents[1];
4385 else
4386 unpin = &root->fs_info->freed_extents[0];
4387 loop = false;
4388 goto again;
4389 }
4390
liuboacce9522011-01-06 19:30:25 +08004391 return 0;
4392}
4393
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004394void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4395 struct btrfs_root *root)
4396{
4397 btrfs_destroy_delayed_refs(cur_trans, root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004398
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004399 cur_trans->state = TRANS_STATE_COMMIT_START;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004400 wake_up(&root->fs_info->transaction_blocked_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004401
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004402 cur_trans->state = TRANS_STATE_UNBLOCKED;
Josef Bacikd7096fc2012-05-31 15:49:57 -04004403 wake_up(&root->fs_info->transaction_wait);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004404
Miao Xie67cde342012-06-14 02:23:22 -06004405 btrfs_destroy_delayed_inodes(root);
4406 btrfs_assert_delayed_root_empty(root);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004407
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004408 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4409 EXTENT_DIRTY);
Liu Bo6e841e32012-06-14 02:23:20 -06004410 btrfs_destroy_pinned_extent(root,
4411 root->fs_info->pinned_extents);
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004412
Miao Xie4a9d8bd2013-05-17 03:53:43 +00004413 cur_trans->state =TRANS_STATE_COMPLETED;
4414 wake_up(&cur_trans->commit_wait);
4415
Jeff Mahoney49b25e02012-03-01 17:24:58 +01004416 /*
4417 memset(cur_trans, 0, sizeof(*cur_trans));
4418 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4419 */
4420}
4421
Eric Sandeen48a3b632013-04-25 20:41:01 +00004422static int btrfs_cleanup_transaction(struct btrfs_root *root)
liuboacce9522011-01-06 19:30:25 +08004423{
4424 struct btrfs_transaction *t;
liuboacce9522011-01-06 19:30:25 +08004425
liuboacce9522011-01-06 19:30:25 +08004426 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4427
Josef Bacika4abeea2011-04-11 17:25:13 -04004428 spin_lock(&root->fs_info->trans_lock);
Josef Bacik724e2312013-09-30 11:36:38 -04004429 while (!list_empty(&root->fs_info->trans_list)) {
4430 t = list_first_entry(&root->fs_info->trans_list,
4431 struct btrfs_transaction, list);
4432 if (t->state >= TRANS_STATE_COMMIT_START) {
4433 atomic_inc(&t->use_count);
4434 spin_unlock(&root->fs_info->trans_lock);
4435 btrfs_wait_for_commit(root, t->transid);
4436 btrfs_put_transaction(t);
4437 spin_lock(&root->fs_info->trans_lock);
4438 continue;
4439 }
4440 if (t == root->fs_info->running_transaction) {
4441 t->state = TRANS_STATE_COMMIT_DOING;
4442 spin_unlock(&root->fs_info->trans_lock);
4443 /*
4444 * We wait for 0 num_writers since we don't hold a trans
4445 * handle open currently for this transaction.
4446 */
4447 wait_event(t->writer_wait,
4448 atomic_read(&t->num_writers) == 0);
4449 } else {
4450 spin_unlock(&root->fs_info->trans_lock);
4451 }
4452 btrfs_cleanup_one_transaction(t, root);
Josef Bacika4abeea2011-04-11 17:25:13 -04004453
Josef Bacik724e2312013-09-30 11:36:38 -04004454 spin_lock(&root->fs_info->trans_lock);
4455 if (t == root->fs_info->running_transaction)
4456 root->fs_info->running_transaction = NULL;
liuboacce9522011-01-06 19:30:25 +08004457 list_del_init(&t->list);
Josef Bacik724e2312013-09-30 11:36:38 -04004458 spin_unlock(&root->fs_info->trans_lock);
liuboacce9522011-01-06 19:30:25 +08004459
Josef Bacik724e2312013-09-30 11:36:38 -04004460 btrfs_put_transaction(t);
4461 trace_btrfs_transaction_commit(root);
4462 spin_lock(&root->fs_info->trans_lock);
4463 }
4464 spin_unlock(&root->fs_info->trans_lock);
4465 btrfs_destroy_all_ordered_extents(root->fs_info);
4466 btrfs_destroy_delayed_inodes(root);
4467 btrfs_assert_delayed_root_empty(root);
4468 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4469 btrfs_destroy_all_delalloc_inodes(root->fs_info);
liuboacce9522011-01-06 19:30:25 +08004470 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
liuboacce9522011-01-06 19:30:25 +08004471
4472 return 0;
4473}
4474
David Sterbae8c9f182015-01-02 18:23:10 +01004475static const struct extent_io_ops btree_extent_io_ops = {
Chris Masonce9adaa2008-04-09 16:28:12 -04004476 .readpage_end_io_hook = btree_readpage_end_io_hook,
Arne Jansen4bb31e92011-06-10 13:55:54 +02004477 .readpage_io_failed_hook = btree_io_failed_hook,
Chris Mason0b86a832008-03-24 15:01:56 -04004478 .submit_bio_hook = btree_submit_bio_hook,
Chris Mason239b14b2008-03-24 15:02:07 -04004479 /* note we're sharing with inode.c for the merge bio hook */
4480 .merge_bio_hook = btrfs_merge_bio_hook,
Chris Mason0da54682007-11-07 21:08:01 -05004481};