Eric Lee / smarc-ti-linux-kernel

Download as
Browse Code ยป

Commit 74e42361fa3bc102647ad1e1ec7c21b747658843

Authored by David Sterba
Committed by Greg Kroah-Hartman
1 parent b82eaa1b48
Exists in smarc-ti-linux-3.14.y and in 1 other branch processor-sdk-linux-01.00.00

btrfs: set proper message level for skinny metadata 

commit 5efa0490cc94aee06cd8d282683e22a8ce0a0026 upstream.

This has been confusing people for too long, the message is really just
informative.

Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Showing 1 changed file with 1 additions and 1 deletions Inline Diff

1 /* 1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved. 2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * 3 *
4 * This program is free software; you can redistribute it and/or 4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public 5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation. 6 * License v2 as published by the Free Software Foundation.
7 * 7 *
8 * This program is distributed in the hope that it will be useful, 8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details. 11 * General Public License for more details.
12 * 12 *
13 * You should have received a copy of the GNU General Public 13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the 14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA. 16 * Boston, MA 021110-1307, USA.
17 */ 17 */
18 18
19 #include <linux/fs.h> 19 #include <linux/fs.h>
20 #include <linux/blkdev.h> 20 #include <linux/blkdev.h>
21 #include <linux/scatterlist.h> 21 #include <linux/scatterlist.h>
22 #include <linux/swap.h> 22 #include <linux/swap.h>
23 #include <linux/radix-tree.h> 23 #include <linux/radix-tree.h>
24 #include <linux/writeback.h> 24 #include <linux/writeback.h>
25 #include <linux/buffer_head.h> 25 #include <linux/buffer_head.h>
26 #include <linux/workqueue.h> 26 #include <linux/workqueue.h>
27 #include <linux/kthread.h> 27 #include <linux/kthread.h>
28 #include <linux/freezer.h> 28 #include <linux/freezer.h>
29 #include <linux/slab.h> 29 #include <linux/slab.h>
30 #include <linux/migrate.h> 30 #include <linux/migrate.h>
31 #include <linux/ratelimit.h> 31 #include <linux/ratelimit.h>
32 #include <linux/uuid.h> 32 #include <linux/uuid.h>
33 #include <linux/semaphore.h> 33 #include <linux/semaphore.h>
34 #include <asm/unaligned.h> 34 #include <asm/unaligned.h>
35 #include "ctree.h" 35 #include "ctree.h"
36 #include "disk-io.h" 36 #include "disk-io.h"
37 #include "hash.h" 37 #include "hash.h"
38 #include "transaction.h" 38 #include "transaction.h"
39 #include "btrfs_inode.h" 39 #include "btrfs_inode.h"
40 #include "volumes.h" 40 #include "volumes.h"
41 #include "print-tree.h" 41 #include "print-tree.h"
42 #include "async-thread.h" 42 #include "async-thread.h"
43 #include "locking.h" 43 #include "locking.h"
44 #include "tree-log.h" 44 #include "tree-log.h"
45 #include "free-space-cache.h" 45 #include "free-space-cache.h"
46 #include "inode-map.h" 46 #include "inode-map.h"
47 #include "check-integrity.h" 47 #include "check-integrity.h"
48 #include "rcu-string.h" 48 #include "rcu-string.h"
49 #include "dev-replace.h" 49 #include "dev-replace.h"
50 #include "raid56.h" 50 #include "raid56.h"
51 #include "sysfs.h" 51 #include "sysfs.h"
52 52
53 #ifdef CONFIG_X86 53 #ifdef CONFIG_X86
54 #include <asm/cpufeature.h> 54 #include <asm/cpufeature.h>
55 #endif 55 #endif
56 56
57 static struct extent_io_ops btree_extent_io_ops; 57 static struct extent_io_ops btree_extent_io_ops;
58 static void end_workqueue_fn(struct btrfs_work *work); 58 static void end_workqueue_fn(struct btrfs_work *work);
59 static void free_fs_root(struct btrfs_root *root); 59 static void free_fs_root(struct btrfs_root *root);
60 static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 60 static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
61 int read_only); 61 int read_only);
62 static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, 62 static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
63 struct btrfs_root *root); 63 struct btrfs_root *root);
64 static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 64 static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
65 static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 65 static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
66 struct btrfs_root *root); 66 struct btrfs_root *root);
67 static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 67 static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
68 static int btrfs_destroy_marked_extents(struct btrfs_root *root, 68 static int btrfs_destroy_marked_extents(struct btrfs_root *root,
69 struct extent_io_tree *dirty_pages, 69 struct extent_io_tree *dirty_pages,
70 int mark); 70 int mark);
71 static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 71 static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
72 struct extent_io_tree *pinned_extents); 72 struct extent_io_tree *pinned_extents);
73 static int btrfs_cleanup_transaction(struct btrfs_root *root); 73 static int btrfs_cleanup_transaction(struct btrfs_root *root);
74 static void btrfs_error_commit_super(struct btrfs_root *root); 74 static void btrfs_error_commit_super(struct btrfs_root *root);
75 75
76 /* 76 /*
77 * end_io_wq structs are used to do processing in task context when an IO is 77 * end_io_wq structs are used to do processing in task context when an IO is
78 * complete. This is used during reads to verify checksums, and it is used 78 * complete. This is used during reads to verify checksums, and it is used
79 * by writes to insert metadata for new file extents after IO is complete. 79 * by writes to insert metadata for new file extents after IO is complete.
80 */ 80 */
81 struct end_io_wq { 81 struct end_io_wq {
82 struct bio *bio; 82 struct bio *bio;
83 bio_end_io_t *end_io; 83 bio_end_io_t *end_io;
84 void *private; 84 void *private;
85 struct btrfs_fs_info *info; 85 struct btrfs_fs_info *info;
86 int error; 86 int error;
87 int metadata; 87 int metadata;
88 struct list_head list; 88 struct list_head list;
89 struct btrfs_work work; 89 struct btrfs_work work;
90 }; 90 };
91 91
92 /* 92 /*
93 * async submit bios are used to offload expensive checksumming 93 * async submit bios are used to offload expensive checksumming
94 * onto the worker threads. They checksum file and metadata bios 94 * onto the worker threads. They checksum file and metadata bios
95 * just before they are sent down the IO stack. 95 * just before they are sent down the IO stack.
96 */ 96 */
97 struct async_submit_bio { 97 struct async_submit_bio {
98 struct inode *inode; 98 struct inode *inode;
99 struct bio *bio; 99 struct bio *bio;
100 struct list_head list; 100 struct list_head list;
101 extent_submit_bio_hook_t *submit_bio_start; 101 extent_submit_bio_hook_t *submit_bio_start;
102 extent_submit_bio_hook_t *submit_bio_done; 102 extent_submit_bio_hook_t *submit_bio_done;
103 int rw; 103 int rw;
104 int mirror_num; 104 int mirror_num;
105 unsigned long bio_flags; 105 unsigned long bio_flags;
106 /* 106 /*
107 * bio_offset is optional, can be used if the pages in the bio 107 * bio_offset is optional, can be used if the pages in the bio
108 * can't tell us where in the file the bio should go 108 * can't tell us where in the file the bio should go
109 */ 109 */
110 u64 bio_offset; 110 u64 bio_offset;
111 struct btrfs_work work; 111 struct btrfs_work work;
112 int error; 112 int error;
113 }; 113 };
114 114
115 /* 115 /*
116 * Lockdep class keys for extent_buffer->lock's in this root. For a given 116 * Lockdep class keys for extent_buffer->lock's in this root. For a given
117 * eb, the lockdep key is determined by the btrfs_root it belongs to and 117 * eb, the lockdep key is determined by the btrfs_root it belongs to and
118 * the level the eb occupies in the tree. 118 * the level the eb occupies in the tree.
119 * 119 *
120 * Different roots are used for different purposes and may nest inside each 120 * Different roots are used for different purposes and may nest inside each
121 * other and they require separate keysets. As lockdep keys should be 121 * other and they require separate keysets. As lockdep keys should be
122 * static, assign keysets according to the purpose of the root as indicated 122 * static, assign keysets according to the purpose of the root as indicated
123 * by btrfs_root->objectid. This ensures that all special purpose roots 123 * by btrfs_root->objectid. This ensures that all special purpose roots
124 * have separate keysets. 124 * have separate keysets.
125 * 125 *
126 * Lock-nesting across peer nodes is always done with the immediate parent 126 * Lock-nesting across peer nodes is always done with the immediate parent
127 * node locked thus preventing deadlock. As lockdep doesn't know this, use 127 * node locked thus preventing deadlock. As lockdep doesn't know this, use
128 * subclass to avoid triggering lockdep warning in such cases. 128 * subclass to avoid triggering lockdep warning in such cases.
129 * 129 *
130 * The key is set by the readpage_end_io_hook after the buffer has passed 130 * The key is set by the readpage_end_io_hook after the buffer has passed
131 * csum validation but before the pages are unlocked. It is also set by 131 * csum validation but before the pages are unlocked. It is also set by
132 * btrfs_init_new_buffer on freshly allocated blocks. 132 * btrfs_init_new_buffer on freshly allocated blocks.
133 * 133 *
134 * We also add a check to make sure the highest level of the tree is the 134 * We also add a check to make sure the highest level of the tree is the
135 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 135 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
136 * needs update as well. 136 * needs update as well.
137 */ 137 */
138 #ifdef CONFIG_DEBUG_LOCK_ALLOC 138 #ifdef CONFIG_DEBUG_LOCK_ALLOC
139 # if BTRFS_MAX_LEVEL != 8 139 # if BTRFS_MAX_LEVEL != 8
140 # error 140 # error
141 # endif 141 # endif
142 142
143 static struct btrfs_lockdep_keyset { 143 static struct btrfs_lockdep_keyset {
144 u64 id; /* root objectid */ 144 u64 id; /* root objectid */
145 const char *name_stem; /* lock name stem */ 145 const char *name_stem; /* lock name stem */
146 char names[BTRFS_MAX_LEVEL + 1][20]; 146 char names[BTRFS_MAX_LEVEL + 1][20];
147 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 147 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
148 } btrfs_lockdep_keysets[] = { 148 } btrfs_lockdep_keysets[] = {
149 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 149 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
150 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 150 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
151 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 151 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
152 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 152 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
153 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 153 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
154 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 154 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
155 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 155 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
156 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 156 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
157 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 157 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
158 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 158 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
159 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 159 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
160 { .id = 0, .name_stem = "tree" }, 160 { .id = 0, .name_stem = "tree" },
161 }; 161 };
162 162
163 void __init btrfs_init_lockdep(void) 163 void __init btrfs_init_lockdep(void)
164 { 164 {
165 int i, j; 165 int i, j;
166 166
167 /* initialize lockdep class names */ 167 /* initialize lockdep class names */
168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
170 170
171 for (j = 0; j < ARRAY_SIZE(ks->names); j++) 171 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
172 snprintf(ks->names[j], sizeof(ks->names[j]), 172 snprintf(ks->names[j], sizeof(ks->names[j]),
173 "btrfs-%s-%02d", ks->name_stem, j); 173 "btrfs-%s-%02d", ks->name_stem, j);
174 } 174 }
175 } 175 }
176 176
177 void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 177 void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
178 int level) 178 int level)
179 { 179 {
180 struct btrfs_lockdep_keyset *ks; 180 struct btrfs_lockdep_keyset *ks;
181 181
182 BUG_ON(level >= ARRAY_SIZE(ks->keys)); 182 BUG_ON(level >= ARRAY_SIZE(ks->keys));
183 183
184 /* find the matching keyset, id 0 is the default entry */ 184 /* find the matching keyset, id 0 is the default entry */
185 for (ks = btrfs_lockdep_keysets; ks->id; ks++) 185 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
186 if (ks->id == objectid) 186 if (ks->id == objectid)
187 break; 187 break;
188 188
189 lockdep_set_class_and_name(&eb->lock, 189 lockdep_set_class_and_name(&eb->lock,
190 &ks->keys[level], ks->names[level]); 190 &ks->keys[level], ks->names[level]);
191 } 191 }
192 192
193 #endif 193 #endif
194 194
195 /* 195 /*
196 * extents on the btree inode are pretty simple, there's one extent 196 * extents on the btree inode are pretty simple, there's one extent
197 * that covers the entire device 197 * that covers the entire device
198 */ 198 */
199 static struct extent_map *btree_get_extent(struct inode *inode, 199 static struct extent_map *btree_get_extent(struct inode *inode,
200 struct page *page, size_t pg_offset, u64 start, u64 len, 200 struct page *page, size_t pg_offset, u64 start, u64 len,
201 int create) 201 int create)
202 { 202 {
203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
204 struct extent_map *em; 204 struct extent_map *em;
205 int ret; 205 int ret;
206 206
207 read_lock(&em_tree->lock); 207 read_lock(&em_tree->lock);
208 em = lookup_extent_mapping(em_tree, start, len); 208 em = lookup_extent_mapping(em_tree, start, len);
209 if (em) { 209 if (em) {
210 em->bdev = 210 em->bdev =
211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
212 read_unlock(&em_tree->lock); 212 read_unlock(&em_tree->lock);
213 goto out; 213 goto out;
214 } 214 }
215 read_unlock(&em_tree->lock); 215 read_unlock(&em_tree->lock);
216 216
217 em = alloc_extent_map(); 217 em = alloc_extent_map();
218 if (!em) { 218 if (!em) {
219 em = ERR_PTR(-ENOMEM); 219 em = ERR_PTR(-ENOMEM);
220 goto out; 220 goto out;
221 } 221 }
222 em->start = 0; 222 em->start = 0;
223 em->len = (u64)-1; 223 em->len = (u64)-1;
224 em->block_len = (u64)-1; 224 em->block_len = (u64)-1;
225 em->block_start = 0; 225 em->block_start = 0;
226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
227 227
228 write_lock(&em_tree->lock); 228 write_lock(&em_tree->lock);
229 ret = add_extent_mapping(em_tree, em, 0); 229 ret = add_extent_mapping(em_tree, em, 0);
230 if (ret == -EEXIST) { 230 if (ret == -EEXIST) {
231 free_extent_map(em); 231 free_extent_map(em);
232 em = lookup_extent_mapping(em_tree, start, len); 232 em = lookup_extent_mapping(em_tree, start, len);
233 if (!em) 233 if (!em)
234 em = ERR_PTR(-EIO); 234 em = ERR_PTR(-EIO);
235 } else if (ret) { 235 } else if (ret) {
236 free_extent_map(em); 236 free_extent_map(em);
237 em = ERR_PTR(ret); 237 em = ERR_PTR(ret);
238 } 238 }
239 write_unlock(&em_tree->lock); 239 write_unlock(&em_tree->lock);
240 240
241 out: 241 out:
242 return em; 242 return em;
243 } 243 }
244 244
245 u32 btrfs_csum_data(char *data, u32 seed, size_t len) 245 u32 btrfs_csum_data(char *data, u32 seed, size_t len)
246 { 246 {
247 return btrfs_crc32c(seed, data, len); 247 return btrfs_crc32c(seed, data, len);
248 } 248 }
249 249
250 void btrfs_csum_final(u32 crc, char *result) 250 void btrfs_csum_final(u32 crc, char *result)
251 { 251 {
252 put_unaligned_le32(~crc, result); 252 put_unaligned_le32(~crc, result);
253 } 253 }
254 254
255 /* 255 /*
256 * compute the csum for a btree block, and either verify it or write it 256 * compute the csum for a btree block, and either verify it or write it
257 * into the csum field of the block. 257 * into the csum field of the block.
258 */ 258 */
259 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, 259 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
260 int verify) 260 int verify)
261 { 261 {
262 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); 262 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
263 char *result = NULL; 263 char *result = NULL;
264 unsigned long len; 264 unsigned long len;
265 unsigned long cur_len; 265 unsigned long cur_len;
266 unsigned long offset = BTRFS_CSUM_SIZE; 266 unsigned long offset = BTRFS_CSUM_SIZE;
267 char *kaddr; 267 char *kaddr;
268 unsigned long map_start; 268 unsigned long map_start;
269 unsigned long map_len; 269 unsigned long map_len;
270 int err; 270 int err;
271 u32 crc = ~(u32)0; 271 u32 crc = ~(u32)0;
272 unsigned long inline_result; 272 unsigned long inline_result;
273 273
274 len = buf->len - offset; 274 len = buf->len - offset;
275 while (len > 0) { 275 while (len > 0) {
276 err = map_private_extent_buffer(buf, offset, 32, 276 err = map_private_extent_buffer(buf, offset, 32,
277 &kaddr, &map_start, &map_len); 277 &kaddr, &map_start, &map_len);
278 if (err) 278 if (err)
279 return 1; 279 return 1;
280 cur_len = min(len, map_len - (offset - map_start)); 280 cur_len = min(len, map_len - (offset - map_start));
281 crc = btrfs_csum_data(kaddr + offset - map_start, 281 crc = btrfs_csum_data(kaddr + offset - map_start,
282 crc, cur_len); 282 crc, cur_len);
283 len -= cur_len; 283 len -= cur_len;
284 offset += cur_len; 284 offset += cur_len;
285 } 285 }
286 if (csum_size > sizeof(inline_result)) { 286 if (csum_size > sizeof(inline_result)) {
287 result = kzalloc(csum_size * sizeof(char), GFP_NOFS); 287 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
288 if (!result) 288 if (!result)
289 return 1; 289 return 1;
290 } else { 290 } else {
291 result = (char *)&inline_result; 291 result = (char *)&inline_result;
292 } 292 }
293 293
294 btrfs_csum_final(crc, result); 294 btrfs_csum_final(crc, result);
295 295
296 if (verify) { 296 if (verify) {
297 if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 297 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
298 u32 val; 298 u32 val;
299 u32 found = 0; 299 u32 found = 0;
300 memcpy(&found, result, csum_size); 300 memcpy(&found, result, csum_size);
301 301
302 read_extent_buffer(buf, &val, 0, csum_size); 302 read_extent_buffer(buf, &val, 0, csum_size);
303 printk_ratelimited(KERN_INFO 303 printk_ratelimited(KERN_INFO
304 "BTRFS: %s checksum verify failed on %llu wanted %X found %X " 304 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
305 "level %d\n", 305 "level %d\n",
306 root->fs_info->sb->s_id, buf->start, 306 root->fs_info->sb->s_id, buf->start,
307 val, found, btrfs_header_level(buf)); 307 val, found, btrfs_header_level(buf));
308 if (result != (char *)&inline_result) 308 if (result != (char *)&inline_result)
309 kfree(result); 309 kfree(result);
310 return 1; 310 return 1;
311 } 311 }
312 } else { 312 } else {
313 write_extent_buffer(buf, result, 0, csum_size); 313 write_extent_buffer(buf, result, 0, csum_size);
314 } 314 }
315 if (result != (char *)&inline_result) 315 if (result != (char *)&inline_result)
316 kfree(result); 316 kfree(result);
317 return 0; 317 return 0;
318 } 318 }
319 319
320 /* 320 /*
321 * we can't consider a given block up to date unless the transid of the 321 * we can't consider a given block up to date unless the transid of the
322 * block matches the transid in the parent node's pointer. This is how we 322 * block matches the transid in the parent node's pointer. This is how we
323 * detect blocks that either didn't get written at all or got written 323 * detect blocks that either didn't get written at all or got written
324 * in the wrong place. 324 * in the wrong place.
325 */ 325 */
326 static int verify_parent_transid(struct extent_io_tree *io_tree, 326 static int verify_parent_transid(struct extent_io_tree *io_tree,
327 struct extent_buffer *eb, u64 parent_transid, 327 struct extent_buffer *eb, u64 parent_transid,
328 int atomic) 328 int atomic)
329 { 329 {
330 struct extent_state *cached_state = NULL; 330 struct extent_state *cached_state = NULL;
331 int ret; 331 int ret;
332 332
333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 return 0; 334 return 0;
335 335
336 if (atomic) 336 if (atomic)
337 return -EAGAIN; 337 return -EAGAIN;
338 338
339 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 339 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
340 0, &cached_state); 340 0, &cached_state);
341 if (extent_buffer_uptodate(eb) && 341 if (extent_buffer_uptodate(eb) &&
342 btrfs_header_generation(eb) == parent_transid) { 342 btrfs_header_generation(eb) == parent_transid) {
343 ret = 0; 343 ret = 0;
344 goto out; 344 goto out;
345 } 345 }
346 printk_ratelimited("parent transid verify failed on %llu wanted %llu " 346 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
347 "found %llu\n", 347 "found %llu\n",
348 eb->start, parent_transid, btrfs_header_generation(eb)); 348 eb->start, parent_transid, btrfs_header_generation(eb));
349 ret = 1; 349 ret = 1;
350 clear_extent_buffer_uptodate(eb); 350 clear_extent_buffer_uptodate(eb);
351 out: 351 out:
352 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 352 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
353 &cached_state, GFP_NOFS); 353 &cached_state, GFP_NOFS);
354 return ret; 354 return ret;
355 } 355 }
356 356
357 /* 357 /*
358 * Return 0 if the superblock checksum type matches the checksum value of that 358 * Return 0 if the superblock checksum type matches the checksum value of that
359 * algorithm. Pass the raw disk superblock data. 359 * algorithm. Pass the raw disk superblock data.
360 */ 360 */
361 static int btrfs_check_super_csum(char *raw_disk_sb) 361 static int btrfs_check_super_csum(char *raw_disk_sb)
362 { 362 {
363 struct btrfs_super_block *disk_sb = 363 struct btrfs_super_block *disk_sb =
364 (struct btrfs_super_block *)raw_disk_sb; 364 (struct btrfs_super_block *)raw_disk_sb;
365 u16 csum_type = btrfs_super_csum_type(disk_sb); 365 u16 csum_type = btrfs_super_csum_type(disk_sb);
366 int ret = 0; 366 int ret = 0;
367 367
368 if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 368 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
369 u32 crc = ~(u32)0; 369 u32 crc = ~(u32)0;
370 const int csum_size = sizeof(crc); 370 const int csum_size = sizeof(crc);
371 char result[csum_size]; 371 char result[csum_size];
372 372
373 /* 373 /*
374 * The super_block structure does not span the whole 374 * The super_block structure does not span the whole
375 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 375 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
376 * is filled with zeros and is included in the checkum. 376 * is filled with zeros and is included in the checkum.
377 */ 377 */
378 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 378 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
379 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 379 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
380 btrfs_csum_final(crc, result); 380 btrfs_csum_final(crc, result);
381 381
382 if (memcmp(raw_disk_sb, result, csum_size)) 382 if (memcmp(raw_disk_sb, result, csum_size))
383 ret = 1; 383 ret = 1;
384 384
385 if (ret && btrfs_super_generation(disk_sb) < 10) { 385 if (ret && btrfs_super_generation(disk_sb) < 10) {
386 printk(KERN_WARNING 386 printk(KERN_WARNING
387 "BTRFS: super block crcs don't match, older mkfs detected\n"); 387 "BTRFS: super block crcs don't match, older mkfs detected\n");
388 ret = 0; 388 ret = 0;
389 } 389 }
390 } 390 }
391 391
392 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 392 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
393 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", 393 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
394 csum_type); 394 csum_type);
395 ret = 1; 395 ret = 1;
396 } 396 }
397 397
398 return ret; 398 return ret;
399 } 399 }
400 400
401 /* 401 /*
402 * helper to read a given tree block, doing retries as required when 402 * helper to read a given tree block, doing retries as required when
403 * the checksums don't match and we have alternate mirrors to try. 403 * the checksums don't match and we have alternate mirrors to try.
404 */ 404 */
405 static int btree_read_extent_buffer_pages(struct btrfs_root *root, 405 static int btree_read_extent_buffer_pages(struct btrfs_root *root,
406 struct extent_buffer *eb, 406 struct extent_buffer *eb,
407 u64 start, u64 parent_transid) 407 u64 start, u64 parent_transid)
408 { 408 {
409 struct extent_io_tree *io_tree; 409 struct extent_io_tree *io_tree;
410 int failed = 0; 410 int failed = 0;
411 int ret; 411 int ret;
412 int num_copies = 0; 412 int num_copies = 0;
413 int mirror_num = 0; 413 int mirror_num = 0;
414 int failed_mirror = 0; 414 int failed_mirror = 0;
415 415
416 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 416 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
417 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 417 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
418 while (1) { 418 while (1) {
419 ret = read_extent_buffer_pages(io_tree, eb, start, 419 ret = read_extent_buffer_pages(io_tree, eb, start,
420 WAIT_COMPLETE, 420 WAIT_COMPLETE,
421 btree_get_extent, mirror_num); 421 btree_get_extent, mirror_num);
422 if (!ret) { 422 if (!ret) {
423 if (!verify_parent_transid(io_tree, eb, 423 if (!verify_parent_transid(io_tree, eb,
424 parent_transid, 0)) 424 parent_transid, 0))
425 break; 425 break;
426 else 426 else
427 ret = -EIO; 427 ret = -EIO;
428 } 428 }
429 429
430 /* 430 /*
431 * This buffer's crc is fine, but its contents are corrupted, so 431 * This buffer's crc is fine, but its contents are corrupted, so
432 * there is no reason to read the other copies, they won't be 432 * there is no reason to read the other copies, they won't be
433 * any less wrong. 433 * any less wrong.
434 */ 434 */
435 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 435 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
436 break; 436 break;
437 437
438 num_copies = btrfs_num_copies(root->fs_info, 438 num_copies = btrfs_num_copies(root->fs_info,
439 eb->start, eb->len); 439 eb->start, eb->len);
440 if (num_copies == 1) 440 if (num_copies == 1)
441 break; 441 break;
442 442
443 if (!failed_mirror) { 443 if (!failed_mirror) {
444 failed = 1; 444 failed = 1;
445 failed_mirror = eb->read_mirror; 445 failed_mirror = eb->read_mirror;
446 } 446 }
447 447
448 mirror_num++; 448 mirror_num++;
449 if (mirror_num == failed_mirror) 449 if (mirror_num == failed_mirror)
450 mirror_num++; 450 mirror_num++;
451 451
452 if (mirror_num > num_copies) 452 if (mirror_num > num_copies)
453 break; 453 break;
454 } 454 }
455 455
456 if (failed && !ret && failed_mirror) 456 if (failed && !ret && failed_mirror)
457 repair_eb_io_failure(root, eb, failed_mirror); 457 repair_eb_io_failure(root, eb, failed_mirror);
458 458
459 return ret; 459 return ret;
460 } 460 }
461 461
462 /* 462 /*
463 * checksum a dirty tree block before IO. This has extra checks to make sure 463 * checksum a dirty tree block before IO. This has extra checks to make sure
464 * we only fill in the checksum field in the first page of a multi-page block 464 * we only fill in the checksum field in the first page of a multi-page block
465 */ 465 */
466 466
467 static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) 467 static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
468 { 468 {
469 u64 start = page_offset(page); 469 u64 start = page_offset(page);
470 u64 found_start; 470 u64 found_start;
471 struct extent_buffer *eb; 471 struct extent_buffer *eb;
472 472
473 eb = (struct extent_buffer *)page->private; 473 eb = (struct extent_buffer *)page->private;
474 if (page != eb->pages[0]) 474 if (page != eb->pages[0])
475 return 0; 475 return 0;
476 found_start = btrfs_header_bytenr(eb); 476 found_start = btrfs_header_bytenr(eb);
477 if (WARN_ON(found_start != start || !PageUptodate(page))) 477 if (WARN_ON(found_start != start || !PageUptodate(page)))
478 return 0; 478 return 0;
479 csum_tree_block(root, eb, 0); 479 csum_tree_block(root, eb, 0);
480 return 0; 480 return 0;
481 } 481 }
482 482
483 static int check_tree_block_fsid(struct btrfs_root *root, 483 static int check_tree_block_fsid(struct btrfs_root *root,
484 struct extent_buffer *eb) 484 struct extent_buffer *eb)
485 { 485 {
486 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 486 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
487 u8 fsid[BTRFS_UUID_SIZE]; 487 u8 fsid[BTRFS_UUID_SIZE];
488 int ret = 1; 488 int ret = 1;
489 489
490 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 490 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
491 while (fs_devices) { 491 while (fs_devices) {
492 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 492 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
493 ret = 0; 493 ret = 0;
494 break; 494 break;
495 } 495 }
496 fs_devices = fs_devices->seed; 496 fs_devices = fs_devices->seed;
497 } 497 }
498 return ret; 498 return ret;
499 } 499 }
500 500
501 #define CORRUPT(reason, eb, root, slot) \ 501 #define CORRUPT(reason, eb, root, slot) \
502 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ 502 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
503 "root=%llu, slot=%d", reason, \ 503 "root=%llu, slot=%d", reason, \
504 btrfs_header_bytenr(eb), root->objectid, slot) 504 btrfs_header_bytenr(eb), root->objectid, slot)
505 505
506 static noinline int check_leaf(struct btrfs_root *root, 506 static noinline int check_leaf(struct btrfs_root *root,
507 struct extent_buffer *leaf) 507 struct extent_buffer *leaf)
508 { 508 {
509 struct btrfs_key key; 509 struct btrfs_key key;
510 struct btrfs_key leaf_key; 510 struct btrfs_key leaf_key;
511 u32 nritems = btrfs_header_nritems(leaf); 511 u32 nritems = btrfs_header_nritems(leaf);
512 int slot; 512 int slot;
513 513
514 if (nritems == 0) 514 if (nritems == 0)
515 return 0; 515 return 0;
516 516
517 /* Check the 0 item */ 517 /* Check the 0 item */
518 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 518 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
519 BTRFS_LEAF_DATA_SIZE(root)) { 519 BTRFS_LEAF_DATA_SIZE(root)) {
520 CORRUPT("invalid item offset size pair", leaf, root, 0); 520 CORRUPT("invalid item offset size pair", leaf, root, 0);
521 return -EIO; 521 return -EIO;
522 } 522 }
523 523
524 /* 524 /*
525 * Check to make sure each items keys are in the correct order and their 525 * Check to make sure each items keys are in the correct order and their
526 * offsets make sense. We only have to loop through nritems-1 because 526 * offsets make sense. We only have to loop through nritems-1 because
527 * we check the current slot against the next slot, which verifies the 527 * we check the current slot against the next slot, which verifies the
528 * next slot's offset+size makes sense and that the current's slot 528 * next slot's offset+size makes sense and that the current's slot
529 * offset is correct. 529 * offset is correct.
530 */ 530 */
531 for (slot = 0; slot < nritems - 1; slot++) { 531 for (slot = 0; slot < nritems - 1; slot++) {
532 btrfs_item_key_to_cpu(leaf, &leaf_key, slot); 532 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
533 btrfs_item_key_to_cpu(leaf, &key, slot + 1); 533 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
534 534
535 /* Make sure the keys are in the right order */ 535 /* Make sure the keys are in the right order */
536 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { 536 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
537 CORRUPT("bad key order", leaf, root, slot); 537 CORRUPT("bad key order", leaf, root, slot);
538 return -EIO; 538 return -EIO;
539 } 539 }
540 540
541 /* 541 /*
542 * Make sure the offset and ends are right, remember that the 542 * Make sure the offset and ends are right, remember that the
543 * item data starts at the end of the leaf and grows towards the 543 * item data starts at the end of the leaf and grows towards the
544 * front. 544 * front.
545 */ 545 */
546 if (btrfs_item_offset_nr(leaf, slot) != 546 if (btrfs_item_offset_nr(leaf, slot) !=
547 btrfs_item_end_nr(leaf, slot + 1)) { 547 btrfs_item_end_nr(leaf, slot + 1)) {
548 CORRUPT("slot offset bad", leaf, root, slot); 548 CORRUPT("slot offset bad", leaf, root, slot);
549 return -EIO; 549 return -EIO;
550 } 550 }
551 551
552 /* 552 /*
553 * Check to make sure that we don't point outside of the leaf, 553 * Check to make sure that we don't point outside of the leaf,
554 * just incase all the items are consistent to eachother, but 554 * just incase all the items are consistent to eachother, but
555 * all point outside of the leaf. 555 * all point outside of the leaf.
556 */ 556 */
557 if (btrfs_item_end_nr(leaf, slot) > 557 if (btrfs_item_end_nr(leaf, slot) >
558 BTRFS_LEAF_DATA_SIZE(root)) { 558 BTRFS_LEAF_DATA_SIZE(root)) {
559 CORRUPT("slot end outside of leaf", leaf, root, slot); 559 CORRUPT("slot end outside of leaf", leaf, root, slot);
560 return -EIO; 560 return -EIO;
561 } 561 }
562 } 562 }
563 563
564 return 0; 564 return 0;
565 } 565 }
566 566
567 static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 567 static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
568 u64 phy_offset, struct page *page, 568 u64 phy_offset, struct page *page,
569 u64 start, u64 end, int mirror) 569 u64 start, u64 end, int mirror)
570 { 570 {
571 u64 found_start; 571 u64 found_start;
572 int found_level; 572 int found_level;
573 struct extent_buffer *eb; 573 struct extent_buffer *eb;
574 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 574 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
575 int ret = 0; 575 int ret = 0;
576 int reads_done; 576 int reads_done;
577 577
578 if (!page->private) 578 if (!page->private)
579 goto out; 579 goto out;
580 580
581 eb = (struct extent_buffer *)page->private; 581 eb = (struct extent_buffer *)page->private;
582 582
583 /* the pending IO might have been the only thing that kept this buffer 583 /* the pending IO might have been the only thing that kept this buffer
584 * in memory. Make sure we have a ref for all this other checks 584 * in memory. Make sure we have a ref for all this other checks
585 */ 585 */
586 extent_buffer_get(eb); 586 extent_buffer_get(eb);
587 587
588 reads_done = atomic_dec_and_test(&eb->io_pages); 588 reads_done = atomic_dec_and_test(&eb->io_pages);
589 if (!reads_done) 589 if (!reads_done)
590 goto err; 590 goto err;
591 591
592 eb->read_mirror = mirror; 592 eb->read_mirror = mirror;
593 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { 593 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
594 ret = -EIO; 594 ret = -EIO;
595 goto err; 595 goto err;
596 } 596 }
597 597
598 found_start = btrfs_header_bytenr(eb); 598 found_start = btrfs_header_bytenr(eb);
599 if (found_start != eb->start) { 599 if (found_start != eb->start) {
600 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start " 600 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
601 "%llu %llu\n", 601 "%llu %llu\n",
602 found_start, eb->start); 602 found_start, eb->start);
603 ret = -EIO; 603 ret = -EIO;
604 goto err; 604 goto err;
605 } 605 }
606 if (check_tree_block_fsid(root, eb)) { 606 if (check_tree_block_fsid(root, eb)) {
607 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n", 607 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
608 eb->start); 608 eb->start);
609 ret = -EIO; 609 ret = -EIO;
610 goto err; 610 goto err;
611 } 611 }
612 found_level = btrfs_header_level(eb); 612 found_level = btrfs_header_level(eb);
613 if (found_level >= BTRFS_MAX_LEVEL) { 613 if (found_level >= BTRFS_MAX_LEVEL) {
614 btrfs_info(root->fs_info, "bad tree block level %d", 614 btrfs_info(root->fs_info, "bad tree block level %d",
615 (int)btrfs_header_level(eb)); 615 (int)btrfs_header_level(eb));
616 ret = -EIO; 616 ret = -EIO;
617 goto err; 617 goto err;
618 } 618 }
619 619
620 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 620 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
621 eb, found_level); 621 eb, found_level);
622 622
623 ret = csum_tree_block(root, eb, 1); 623 ret = csum_tree_block(root, eb, 1);
624 if (ret) { 624 if (ret) {
625 ret = -EIO; 625 ret = -EIO;
626 goto err; 626 goto err;
627 } 627 }
628 628
629 /* 629 /*
630 * If this is a leaf block and it is corrupt, set the corrupt bit so 630 * If this is a leaf block and it is corrupt, set the corrupt bit so
631 * that we don't try and read the other copies of this block, just 631 * that we don't try and read the other copies of this block, just
632 * return -EIO. 632 * return -EIO.
633 */ 633 */
634 if (found_level == 0 && check_leaf(root, eb)) { 634 if (found_level == 0 && check_leaf(root, eb)) {
635 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 635 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
636 ret = -EIO; 636 ret = -EIO;
637 } 637 }
638 638
639 if (!ret) 639 if (!ret)
640 set_extent_buffer_uptodate(eb); 640 set_extent_buffer_uptodate(eb);
641 err: 641 err:
642 if (reads_done && 642 if (reads_done &&
643 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 643 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
644 btree_readahead_hook(root, eb, eb->start, ret); 644 btree_readahead_hook(root, eb, eb->start, ret);
645 645
646 if (ret) { 646 if (ret) {
647 /* 647 /*
648 * our io error hook is going to dec the io pages 648 * our io error hook is going to dec the io pages
649 * again, we have to make sure it has something 649 * again, we have to make sure it has something
650 * to decrement 650 * to decrement
651 */ 651 */
652 atomic_inc(&eb->io_pages); 652 atomic_inc(&eb->io_pages);
653 clear_extent_buffer_uptodate(eb); 653 clear_extent_buffer_uptodate(eb);
654 } 654 }
655 free_extent_buffer(eb); 655 free_extent_buffer(eb);
656 out: 656 out:
657 return ret; 657 return ret;
658 } 658 }
659 659
660 static int btree_io_failed_hook(struct page *page, int failed_mirror) 660 static int btree_io_failed_hook(struct page *page, int failed_mirror)
661 { 661 {
662 struct extent_buffer *eb; 662 struct extent_buffer *eb;
663 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 663 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
664 664
665 eb = (struct extent_buffer *)page->private; 665 eb = (struct extent_buffer *)page->private;
666 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); 666 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
667 eb->read_mirror = failed_mirror; 667 eb->read_mirror = failed_mirror;
668 atomic_dec(&eb->io_pages); 668 atomic_dec(&eb->io_pages);
669 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 669 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
670 btree_readahead_hook(root, eb, eb->start, -EIO); 670 btree_readahead_hook(root, eb, eb->start, -EIO);
671 return -EIO; /* we fixed nothing */ 671 return -EIO; /* we fixed nothing */
672 } 672 }
673 673
674 static void end_workqueue_bio(struct bio *bio, int err) 674 static void end_workqueue_bio(struct bio *bio, int err)
675 { 675 {
676 struct end_io_wq *end_io_wq = bio->bi_private; 676 struct end_io_wq *end_io_wq = bio->bi_private;
677 struct btrfs_fs_info *fs_info; 677 struct btrfs_fs_info *fs_info;
678 678
679 fs_info = end_io_wq->info; 679 fs_info = end_io_wq->info;
680 end_io_wq->error = err; 680 end_io_wq->error = err;
681 end_io_wq->work.func = end_workqueue_fn; 681 end_io_wq->work.func = end_workqueue_fn;
682 end_io_wq->work.flags = 0; 682 end_io_wq->work.flags = 0;
683 683
684 if (bio->bi_rw & REQ_WRITE) { 684 if (bio->bi_rw & REQ_WRITE) {
685 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 685 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
686 btrfs_queue_worker(&fs_info->endio_meta_write_workers, 686 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
687 &end_io_wq->work); 687 &end_io_wq->work);
688 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 688 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
689 btrfs_queue_worker(&fs_info->endio_freespace_worker, 689 btrfs_queue_worker(&fs_info->endio_freespace_worker,
690 &end_io_wq->work); 690 &end_io_wq->work);
691 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 691 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
692 btrfs_queue_worker(&fs_info->endio_raid56_workers, 692 btrfs_queue_worker(&fs_info->endio_raid56_workers,
693 &end_io_wq->work); 693 &end_io_wq->work);
694 else 694 else
695 btrfs_queue_worker(&fs_info->endio_write_workers, 695 btrfs_queue_worker(&fs_info->endio_write_workers,
696 &end_io_wq->work); 696 &end_io_wq->work);
697 } else { 697 } else {
698 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 698 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
699 btrfs_queue_worker(&fs_info->endio_raid56_workers, 699 btrfs_queue_worker(&fs_info->endio_raid56_workers,
700 &end_io_wq->work); 700 &end_io_wq->work);
701 else if (end_io_wq->metadata) 701 else if (end_io_wq->metadata)
702 btrfs_queue_worker(&fs_info->endio_meta_workers, 702 btrfs_queue_worker(&fs_info->endio_meta_workers,
703 &end_io_wq->work); 703 &end_io_wq->work);
704 else 704 else
705 btrfs_queue_worker(&fs_info->endio_workers, 705 btrfs_queue_worker(&fs_info->endio_workers,
706 &end_io_wq->work); 706 &end_io_wq->work);
707 } 707 }
708 } 708 }
709 709
710 /* 710 /*
711 * For the metadata arg you want 711 * For the metadata arg you want
712 * 712 *
713 * 0 - if data 713 * 0 - if data
714 * 1 - if normal metadta 714 * 1 - if normal metadta
715 * 2 - if writing to the free space cache area 715 * 2 - if writing to the free space cache area
716 * 3 - raid parity work 716 * 3 - raid parity work
717 */ 717 */
718 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 718 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
719 int metadata) 719 int metadata)
720 { 720 {
721 struct end_io_wq *end_io_wq; 721 struct end_io_wq *end_io_wq;
722 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); 722 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
723 if (!end_io_wq) 723 if (!end_io_wq)
724 return -ENOMEM; 724 return -ENOMEM;
725 725
726 end_io_wq->private = bio->bi_private; 726 end_io_wq->private = bio->bi_private;
727 end_io_wq->end_io = bio->bi_end_io; 727 end_io_wq->end_io = bio->bi_end_io;
728 end_io_wq->info = info; 728 end_io_wq->info = info;
729 end_io_wq->error = 0; 729 end_io_wq->error = 0;
730 end_io_wq->bio = bio; 730 end_io_wq->bio = bio;
731 end_io_wq->metadata = metadata; 731 end_io_wq->metadata = metadata;
732 732
733 bio->bi_private = end_io_wq; 733 bio->bi_private = end_io_wq;
734 bio->bi_end_io = end_workqueue_bio; 734 bio->bi_end_io = end_workqueue_bio;
735 return 0; 735 return 0;
736 } 736 }
737 737
738 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 738 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
739 { 739 {
740 unsigned long limit = min_t(unsigned long, 740 unsigned long limit = min_t(unsigned long,
741 info->workers.max_workers, 741 info->workers.max_workers,
742 info->fs_devices->open_devices); 742 info->fs_devices->open_devices);
743 return 256 * limit; 743 return 256 * limit;
744 } 744 }
745 745
746 static void run_one_async_start(struct btrfs_work *work) 746 static void run_one_async_start(struct btrfs_work *work)
747 { 747 {
748 struct async_submit_bio *async; 748 struct async_submit_bio *async;
749 int ret; 749 int ret;
750 750
751 async = container_of(work, struct async_submit_bio, work); 751 async = container_of(work, struct async_submit_bio, work);
752 ret = async->submit_bio_start(async->inode, async->rw, async->bio, 752 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
753 async->mirror_num, async->bio_flags, 753 async->mirror_num, async->bio_flags,
754 async->bio_offset); 754 async->bio_offset);
755 if (ret) 755 if (ret)
756 async->error = ret; 756 async->error = ret;
757 } 757 }
758 758
759 static void run_one_async_done(struct btrfs_work *work) 759 static void run_one_async_done(struct btrfs_work *work)
760 { 760 {
761 struct btrfs_fs_info *fs_info; 761 struct btrfs_fs_info *fs_info;
762 struct async_submit_bio *async; 762 struct async_submit_bio *async;
763 int limit; 763 int limit;
764 764
765 async = container_of(work, struct async_submit_bio, work); 765 async = container_of(work, struct async_submit_bio, work);
766 fs_info = BTRFS_I(async->inode)->root->fs_info; 766 fs_info = BTRFS_I(async->inode)->root->fs_info;
767 767
768 limit = btrfs_async_submit_limit(fs_info); 768 limit = btrfs_async_submit_limit(fs_info);
769 limit = limit * 2 / 3; 769 limit = limit * 2 / 3;
770 770
771 if (atomic_dec_return(&fs_info->nr_async_submits) < limit && 771 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
772 waitqueue_active(&fs_info->async_submit_wait)) 772 waitqueue_active(&fs_info->async_submit_wait))
773 wake_up(&fs_info->async_submit_wait); 773 wake_up(&fs_info->async_submit_wait);
774 774
775 /* If an error occured we just want to clean up the bio and move on */ 775 /* If an error occured we just want to clean up the bio and move on */
776 if (async->error) { 776 if (async->error) {
777 bio_endio(async->bio, async->error); 777 bio_endio(async->bio, async->error);
778 return; 778 return;
779 } 779 }
780 780
781 async->submit_bio_done(async->inode, async->rw, async->bio, 781 async->submit_bio_done(async->inode, async->rw, async->bio,
782 async->mirror_num, async->bio_flags, 782 async->mirror_num, async->bio_flags,
783 async->bio_offset); 783 async->bio_offset);
784 } 784 }
785 785
786 static void run_one_async_free(struct btrfs_work *work) 786 static void run_one_async_free(struct btrfs_work *work)
787 { 787 {
788 struct async_submit_bio *async; 788 struct async_submit_bio *async;
789 789
790 async = container_of(work, struct async_submit_bio, work); 790 async = container_of(work, struct async_submit_bio, work);
791 kfree(async); 791 kfree(async);
792 } 792 }
793 793
794 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, 794 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
795 int rw, struct bio *bio, int mirror_num, 795 int rw, struct bio *bio, int mirror_num,
796 unsigned long bio_flags, 796 unsigned long bio_flags,
797 u64 bio_offset, 797 u64 bio_offset,
798 extent_submit_bio_hook_t *submit_bio_start, 798 extent_submit_bio_hook_t *submit_bio_start,
799 extent_submit_bio_hook_t *submit_bio_done) 799 extent_submit_bio_hook_t *submit_bio_done)
800 { 800 {
801 struct async_submit_bio *async; 801 struct async_submit_bio *async;
802 802
803 async = kmalloc(sizeof(*async), GFP_NOFS); 803 async = kmalloc(sizeof(*async), GFP_NOFS);
804 if (!async) 804 if (!async)
805 return -ENOMEM; 805 return -ENOMEM;
806 806
807 async->inode = inode; 807 async->inode = inode;
808 async->rw = rw; 808 async->rw = rw;
809 async->bio = bio; 809 async->bio = bio;
810 async->mirror_num = mirror_num; 810 async->mirror_num = mirror_num;
811 async->submit_bio_start = submit_bio_start; 811 async->submit_bio_start = submit_bio_start;
812 async->submit_bio_done = submit_bio_done; 812 async->submit_bio_done = submit_bio_done;
813 813
814 async->work.func = run_one_async_start; 814 async->work.func = run_one_async_start;
815 async->work.ordered_func = run_one_async_done; 815 async->work.ordered_func = run_one_async_done;
816 async->work.ordered_free = run_one_async_free; 816 async->work.ordered_free = run_one_async_free;
817 817
818 async->work.flags = 0; 818 async->work.flags = 0;
819 async->bio_flags = bio_flags; 819 async->bio_flags = bio_flags;
820 async->bio_offset = bio_offset; 820 async->bio_offset = bio_offset;
821 821
822 async->error = 0; 822 async->error = 0;
823 823
824 atomic_inc(&fs_info->nr_async_submits); 824 atomic_inc(&fs_info->nr_async_submits);
825 825
826 if (rw & REQ_SYNC) 826 if (rw & REQ_SYNC)
827 btrfs_set_work_high_prio(&async->work); 827 btrfs_set_work_high_prio(&async->work);
828 828
829 btrfs_queue_worker(&fs_info->workers, &async->work); 829 btrfs_queue_worker(&fs_info->workers, &async->work);
830 830
831 while (atomic_read(&fs_info->async_submit_draining) && 831 while (atomic_read(&fs_info->async_submit_draining) &&
832 atomic_read(&fs_info->nr_async_submits)) { 832 atomic_read(&fs_info->nr_async_submits)) {
833 wait_event(fs_info->async_submit_wait, 833 wait_event(fs_info->async_submit_wait,
834 (atomic_read(&fs_info->nr_async_submits) == 0)); 834 (atomic_read(&fs_info->nr_async_submits) == 0));
835 } 835 }
836 836
837 return 0; 837 return 0;
838 } 838 }
839 839
840 static int btree_csum_one_bio(struct bio *bio) 840 static int btree_csum_one_bio(struct bio *bio)
841 { 841 {
842 struct bio_vec *bvec; 842 struct bio_vec *bvec;
843 struct btrfs_root *root; 843 struct btrfs_root *root;
844 int i, ret = 0; 844 int i, ret = 0;
845 845
846 bio_for_each_segment_all(bvec, bio, i) { 846 bio_for_each_segment_all(bvec, bio, i) {
847 root = BTRFS_I(bvec->bv_page->mapping->host)->root; 847 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
848 ret = csum_dirty_buffer(root, bvec->bv_page); 848 ret = csum_dirty_buffer(root, bvec->bv_page);
849 if (ret) 849 if (ret)
850 break; 850 break;
851 } 851 }
852 852
853 return ret; 853 return ret;
854 } 854 }
855 855
856 static int __btree_submit_bio_start(struct inode *inode, int rw, 856 static int __btree_submit_bio_start(struct inode *inode, int rw,
857 struct bio *bio, int mirror_num, 857 struct bio *bio, int mirror_num,
858 unsigned long bio_flags, 858 unsigned long bio_flags,
859 u64 bio_offset) 859 u64 bio_offset)
860 { 860 {
861 /* 861 /*
862 * when we're called for a write, we're already in the async 862 * when we're called for a write, we're already in the async
863 * submission context. Just jump into btrfs_map_bio 863 * submission context. Just jump into btrfs_map_bio
864 */ 864 */
865 return btree_csum_one_bio(bio); 865 return btree_csum_one_bio(bio);
866 } 866 }
867 867
868 static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 868 static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
869 int mirror_num, unsigned long bio_flags, 869 int mirror_num, unsigned long bio_flags,
870 u64 bio_offset) 870 u64 bio_offset)
871 { 871 {
872 int ret; 872 int ret;
873 873
874 /* 874 /*
875 * when we're called for a write, we're already in the async 875 * when we're called for a write, we're already in the async
876 * submission context. Just jump into btrfs_map_bio 876 * submission context. Just jump into btrfs_map_bio
877 */ 877 */
878 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 878 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
879 if (ret) 879 if (ret)
880 bio_endio(bio, ret); 880 bio_endio(bio, ret);
881 return ret; 881 return ret;
882 } 882 }
883 883
884 static int check_async_write(struct inode *inode, unsigned long bio_flags) 884 static int check_async_write(struct inode *inode, unsigned long bio_flags)
885 { 885 {
886 if (bio_flags & EXTENT_BIO_TREE_LOG) 886 if (bio_flags & EXTENT_BIO_TREE_LOG)
887 return 0; 887 return 0;
888 #ifdef CONFIG_X86 888 #ifdef CONFIG_X86
889 if (cpu_has_xmm4_2) 889 if (cpu_has_xmm4_2)
890 return 0; 890 return 0;
891 #endif 891 #endif
892 return 1; 892 return 1;
893 } 893 }
894 894
895 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 895 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
896 int mirror_num, unsigned long bio_flags, 896 int mirror_num, unsigned long bio_flags,
897 u64 bio_offset) 897 u64 bio_offset)
898 { 898 {
899 int async = check_async_write(inode, bio_flags); 899 int async = check_async_write(inode, bio_flags);
900 int ret; 900 int ret;
901 901
902 if (!(rw & REQ_WRITE)) { 902 if (!(rw & REQ_WRITE)) {
903 /* 903 /*
904 * called for a read, do the setup so that checksum validation 904 * called for a read, do the setup so that checksum validation
905 * can happen in the async kernel threads 905 * can happen in the async kernel threads
906 */ 906 */
907 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 907 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
908 bio, 1); 908 bio, 1);
909 if (ret) 909 if (ret)
910 goto out_w_error; 910 goto out_w_error;
911 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 911 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
912 mirror_num, 0); 912 mirror_num, 0);
913 } else if (!async) { 913 } else if (!async) {
914 ret = btree_csum_one_bio(bio); 914 ret = btree_csum_one_bio(bio);
915 if (ret) 915 if (ret)
916 goto out_w_error; 916 goto out_w_error;
917 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 917 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
918 mirror_num, 0); 918 mirror_num, 0);
919 } else { 919 } else {
920 /* 920 /*
921 * kthread helpers are used to submit writes so that 921 * kthread helpers are used to submit writes so that
922 * checksumming can happen in parallel across all CPUs 922 * checksumming can happen in parallel across all CPUs
923 */ 923 */
924 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 924 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
925 inode, rw, bio, mirror_num, 0, 925 inode, rw, bio, mirror_num, 0,
926 bio_offset, 926 bio_offset,
927 __btree_submit_bio_start, 927 __btree_submit_bio_start,
928 __btree_submit_bio_done); 928 __btree_submit_bio_done);
929 } 929 }
930 930
931 if (ret) { 931 if (ret) {
932 out_w_error: 932 out_w_error:
933 bio_endio(bio, ret); 933 bio_endio(bio, ret);
934 } 934 }
935 return ret; 935 return ret;
936 } 936 }
937 937
938 #ifdef CONFIG_MIGRATION 938 #ifdef CONFIG_MIGRATION
939 static int btree_migratepage(struct address_space *mapping, 939 static int btree_migratepage(struct address_space *mapping,
940 struct page *newpage, struct page *page, 940 struct page *newpage, struct page *page,
941 enum migrate_mode mode) 941 enum migrate_mode mode)
942 { 942 {
943 /* 943 /*
944 * we can't safely write a btree page from here, 944 * we can't safely write a btree page from here,
945 * we haven't done the locking hook 945 * we haven't done the locking hook
946 */ 946 */
947 if (PageDirty(page)) 947 if (PageDirty(page))
948 return -EAGAIN; 948 return -EAGAIN;
949 /* 949 /*
950 * Buffers may be managed in a filesystem specific way. 950 * Buffers may be managed in a filesystem specific way.
951 * We must have no buffers or drop them. 951 * We must have no buffers or drop them.
952 */ 952 */
953 if (page_has_private(page) && 953 if (page_has_private(page) &&
954 !try_to_release_page(page, GFP_KERNEL)) 954 !try_to_release_page(page, GFP_KERNEL))
955 return -EAGAIN; 955 return -EAGAIN;
956 return migrate_page(mapping, newpage, page, mode); 956 return migrate_page(mapping, newpage, page, mode);
957 } 957 }
958 #endif 958 #endif
959 959
960 960
961 static int btree_writepages(struct address_space *mapping, 961 static int btree_writepages(struct address_space *mapping,
962 struct writeback_control *wbc) 962 struct writeback_control *wbc)
963 { 963 {
964 struct btrfs_fs_info *fs_info; 964 struct btrfs_fs_info *fs_info;
965 int ret; 965 int ret;
966 966
967 if (wbc->sync_mode == WB_SYNC_NONE) { 967 if (wbc->sync_mode == WB_SYNC_NONE) {
968 968
969 if (wbc->for_kupdate) 969 if (wbc->for_kupdate)
970 return 0; 970 return 0;
971 971
972 fs_info = BTRFS_I(mapping->host)->root->fs_info; 972 fs_info = BTRFS_I(mapping->host)->root->fs_info;
973 /* this is a bit racy, but that's ok */ 973 /* this is a bit racy, but that's ok */
974 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 974 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
975 BTRFS_DIRTY_METADATA_THRESH); 975 BTRFS_DIRTY_METADATA_THRESH);
976 if (ret < 0) 976 if (ret < 0)
977 return 0; 977 return 0;
978 } 978 }
979 return btree_write_cache_pages(mapping, wbc); 979 return btree_write_cache_pages(mapping, wbc);
980 } 980 }
981 981
982 static int btree_readpage(struct file *file, struct page *page) 982 static int btree_readpage(struct file *file, struct page *page)
983 { 983 {
984 struct extent_io_tree *tree; 984 struct extent_io_tree *tree;
985 tree = &BTRFS_I(page->mapping->host)->io_tree; 985 tree = &BTRFS_I(page->mapping->host)->io_tree;
986 return extent_read_full_page(tree, page, btree_get_extent, 0); 986 return extent_read_full_page(tree, page, btree_get_extent, 0);
987 } 987 }
988 988
989 static int btree_releasepage(struct page *page, gfp_t gfp_flags) 989 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
990 { 990 {
991 if (PageWriteback(page) || PageDirty(page)) 991 if (PageWriteback(page) || PageDirty(page))
992 return 0; 992 return 0;
993 993
994 return try_release_extent_buffer(page); 994 return try_release_extent_buffer(page);
995 } 995 }
996 996
997 static void btree_invalidatepage(struct page *page, unsigned int offset, 997 static void btree_invalidatepage(struct page *page, unsigned int offset,
998 unsigned int length) 998 unsigned int length)
999 { 999 {
1000 struct extent_io_tree *tree; 1000 struct extent_io_tree *tree;
1001 tree = &BTRFS_I(page->mapping->host)->io_tree; 1001 tree = &BTRFS_I(page->mapping->host)->io_tree;
1002 extent_invalidatepage(tree, page, offset); 1002 extent_invalidatepage(tree, page, offset);
1003 btree_releasepage(page, GFP_NOFS); 1003 btree_releasepage(page, GFP_NOFS);
1004 if (PagePrivate(page)) { 1004 if (PagePrivate(page)) {
1005 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1005 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1006 "page private not zero on page %llu", 1006 "page private not zero on page %llu",
1007 (unsigned long long)page_offset(page)); 1007 (unsigned long long)page_offset(page));
1008 ClearPagePrivate(page); 1008 ClearPagePrivate(page);
1009 set_page_private(page, 0); 1009 set_page_private(page, 0);
1010 page_cache_release(page); 1010 page_cache_release(page);
1011 } 1011 }
1012 } 1012 }
1013 1013
1014 static int btree_set_page_dirty(struct page *page) 1014 static int btree_set_page_dirty(struct page *page)
1015 { 1015 {
1016 #ifdef DEBUG 1016 #ifdef DEBUG
1017 struct extent_buffer *eb; 1017 struct extent_buffer *eb;
1018 1018
1019 BUG_ON(!PagePrivate(page)); 1019 BUG_ON(!PagePrivate(page));
1020 eb = (struct extent_buffer *)page->private; 1020 eb = (struct extent_buffer *)page->private;
1021 BUG_ON(!eb); 1021 BUG_ON(!eb);
1022 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 1022 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1023 BUG_ON(!atomic_read(&eb->refs)); 1023 BUG_ON(!atomic_read(&eb->refs));
1024 btrfs_assert_tree_locked(eb); 1024 btrfs_assert_tree_locked(eb);
1025 #endif 1025 #endif
1026 return __set_page_dirty_nobuffers(page); 1026 return __set_page_dirty_nobuffers(page);
1027 } 1027 }
1028 1028
1029 static const struct address_space_operations btree_aops = { 1029 static const struct address_space_operations btree_aops = {
1030 .readpage = btree_readpage, 1030 .readpage = btree_readpage,
1031 .writepages = btree_writepages, 1031 .writepages = btree_writepages,
1032 .releasepage = btree_releasepage, 1032 .releasepage = btree_releasepage,
1033 .invalidatepage = btree_invalidatepage, 1033 .invalidatepage = btree_invalidatepage,
1034 #ifdef CONFIG_MIGRATION 1034 #ifdef CONFIG_MIGRATION
1035 .migratepage = btree_migratepage, 1035 .migratepage = btree_migratepage,
1036 #endif 1036 #endif
1037 .set_page_dirty = btree_set_page_dirty, 1037 .set_page_dirty = btree_set_page_dirty,
1038 }; 1038 };
1039 1039
1040 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1040 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1041 u64 parent_transid) 1041 u64 parent_transid)
1042 { 1042 {
1043 struct extent_buffer *buf = NULL; 1043 struct extent_buffer *buf = NULL;
1044 struct inode *btree_inode = root->fs_info->btree_inode; 1044 struct inode *btree_inode = root->fs_info->btree_inode;
1045 int ret = 0; 1045 int ret = 0;
1046 1046
1047 buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1047 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1048 if (!buf) 1048 if (!buf)
1049 return 0; 1049 return 0;
1050 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1050 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
1051 buf, 0, WAIT_NONE, btree_get_extent, 0); 1051 buf, 0, WAIT_NONE, btree_get_extent, 0);
1052 free_extent_buffer(buf); 1052 free_extent_buffer(buf);
1053 return ret; 1053 return ret;
1054 } 1054 }
1055 1055
1056 int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1056 int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1057 int mirror_num, struct extent_buffer **eb) 1057 int mirror_num, struct extent_buffer **eb)
1058 { 1058 {
1059 struct extent_buffer *buf = NULL; 1059 struct extent_buffer *buf = NULL;
1060 struct inode *btree_inode = root->fs_info->btree_inode; 1060 struct inode *btree_inode = root->fs_info->btree_inode;
1061 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1061 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1062 int ret; 1062 int ret;
1063 1063
1064 buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1064 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1065 if (!buf) 1065 if (!buf)
1066 return 0; 1066 return 0;
1067 1067
1068 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1068 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1069 1069
1070 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1070 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1071 btree_get_extent, mirror_num); 1071 btree_get_extent, mirror_num);
1072 if (ret) { 1072 if (ret) {
1073 free_extent_buffer(buf); 1073 free_extent_buffer(buf);
1074 return ret; 1074 return ret;
1075 } 1075 }
1076 1076
1077 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1077 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1078 free_extent_buffer(buf); 1078 free_extent_buffer(buf);
1079 return -EIO; 1079 return -EIO;
1080 } else if (extent_buffer_uptodate(buf)) { 1080 } else if (extent_buffer_uptodate(buf)) {
1081 *eb = buf; 1081 *eb = buf;
1082 } else { 1082 } else {
1083 free_extent_buffer(buf); 1083 free_extent_buffer(buf);
1084 } 1084 }
1085 return 0; 1085 return 0;
1086 } 1086 }
1087 1087
1088 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, 1088 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1089 u64 bytenr, u32 blocksize) 1089 u64 bytenr, u32 blocksize)
1090 { 1090 {
1091 return find_extent_buffer(root->fs_info, bytenr); 1091 return find_extent_buffer(root->fs_info, bytenr);
1092 } 1092 }
1093 1093
1094 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 1094 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1095 u64 bytenr, u32 blocksize) 1095 u64 bytenr, u32 blocksize)
1096 { 1096 {
1097 return alloc_extent_buffer(root->fs_info, bytenr, blocksize); 1097 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
1098 } 1098 }
1099 1099
1100 1100
1101 int btrfs_write_tree_block(struct extent_buffer *buf) 1101 int btrfs_write_tree_block(struct extent_buffer *buf)
1102 { 1102 {
1103 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 1103 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
1104 buf->start + buf->len - 1); 1104 buf->start + buf->len - 1);
1105 } 1105 }
1106 1106
1107 int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1107 int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1108 { 1108 {
1109 return filemap_fdatawait_range(buf->pages[0]->mapping, 1109 return filemap_fdatawait_range(buf->pages[0]->mapping,
1110 buf->start, buf->start + buf->len - 1); 1110 buf->start, buf->start + buf->len - 1);
1111 } 1111 }
1112 1112
1113 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1113 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
1114 u32 blocksize, u64 parent_transid) 1114 u32 blocksize, u64 parent_transid)
1115 { 1115 {
1116 struct extent_buffer *buf = NULL; 1116 struct extent_buffer *buf = NULL;
1117 int ret; 1117 int ret;
1118 1118
1119 buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1119 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1120 if (!buf) 1120 if (!buf)
1121 return NULL; 1121 return NULL;
1122 1122
1123 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 1123 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
1124 if (ret) { 1124 if (ret) {
1125 free_extent_buffer(buf); 1125 free_extent_buffer(buf);
1126 return NULL; 1126 return NULL;
1127 } 1127 }
1128 return buf; 1128 return buf;
1129 1129
1130 } 1130 }
1131 1131
1132 void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, 1132 void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1133 struct extent_buffer *buf) 1133 struct extent_buffer *buf)
1134 { 1134 {
1135 struct btrfs_fs_info *fs_info = root->fs_info; 1135 struct btrfs_fs_info *fs_info = root->fs_info;
1136 1136
1137 if (btrfs_header_generation(buf) == 1137 if (btrfs_header_generation(buf) ==
1138 fs_info->running_transaction->transid) { 1138 fs_info->running_transaction->transid) {
1139 btrfs_assert_tree_locked(buf); 1139 btrfs_assert_tree_locked(buf);
1140 1140
1141 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1141 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1142 __percpu_counter_add(&fs_info->dirty_metadata_bytes, 1142 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1143 -buf->len, 1143 -buf->len,
1144 fs_info->dirty_metadata_batch); 1144 fs_info->dirty_metadata_batch);
1145 /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1145 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1146 btrfs_set_lock_blocking(buf); 1146 btrfs_set_lock_blocking(buf);
1147 clear_extent_buffer_dirty(buf); 1147 clear_extent_buffer_dirty(buf);
1148 } 1148 }
1149 } 1149 }
1150 } 1150 }
1151 1151
1152 static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, 1152 static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1153 u32 stripesize, struct btrfs_root *root, 1153 u32 stripesize, struct btrfs_root *root,
1154 struct btrfs_fs_info *fs_info, 1154 struct btrfs_fs_info *fs_info,
1155 u64 objectid) 1155 u64 objectid)
1156 { 1156 {
1157 root->node = NULL; 1157 root->node = NULL;
1158 root->commit_root = NULL; 1158 root->commit_root = NULL;
1159 root->sectorsize = sectorsize; 1159 root->sectorsize = sectorsize;
1160 root->nodesize = nodesize; 1160 root->nodesize = nodesize;
1161 root->leafsize = leafsize; 1161 root->leafsize = leafsize;
1162 root->stripesize = stripesize; 1162 root->stripesize = stripesize;
1163 root->ref_cows = 0; 1163 root->ref_cows = 0;
1164 root->track_dirty = 0; 1164 root->track_dirty = 0;
1165 root->in_radix = 0; 1165 root->in_radix = 0;
1166 root->orphan_item_inserted = 0; 1166 root->orphan_item_inserted = 0;
1167 root->orphan_cleanup_state = 0; 1167 root->orphan_cleanup_state = 0;
1168 1168
1169 root->objectid = objectid; 1169 root->objectid = objectid;
1170 root->last_trans = 0; 1170 root->last_trans = 0;
1171 root->highest_objectid = 0; 1171 root->highest_objectid = 0;
1172 root->nr_delalloc_inodes = 0; 1172 root->nr_delalloc_inodes = 0;
1173 root->nr_ordered_extents = 0; 1173 root->nr_ordered_extents = 0;
1174 root->name = NULL; 1174 root->name = NULL;
1175 root->inode_tree = RB_ROOT; 1175 root->inode_tree = RB_ROOT;
1176 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1176 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
1177 root->block_rsv = NULL; 1177 root->block_rsv = NULL;
1178 root->orphan_block_rsv = NULL; 1178 root->orphan_block_rsv = NULL;
1179 1179
1180 INIT_LIST_HEAD(&root->dirty_list); 1180 INIT_LIST_HEAD(&root->dirty_list);
1181 INIT_LIST_HEAD(&root->root_list); 1181 INIT_LIST_HEAD(&root->root_list);
1182 INIT_LIST_HEAD(&root->delalloc_inodes); 1182 INIT_LIST_HEAD(&root->delalloc_inodes);
1183 INIT_LIST_HEAD(&root->delalloc_root); 1183 INIT_LIST_HEAD(&root->delalloc_root);
1184 INIT_LIST_HEAD(&root->ordered_extents); 1184 INIT_LIST_HEAD(&root->ordered_extents);
1185 INIT_LIST_HEAD(&root->ordered_root); 1185 INIT_LIST_HEAD(&root->ordered_root);
1186 INIT_LIST_HEAD(&root->logged_list[0]); 1186 INIT_LIST_HEAD(&root->logged_list[0]);
1187 INIT_LIST_HEAD(&root->logged_list[1]); 1187 INIT_LIST_HEAD(&root->logged_list[1]);
1188 spin_lock_init(&root->orphan_lock); 1188 spin_lock_init(&root->orphan_lock);
1189 spin_lock_init(&root->inode_lock); 1189 spin_lock_init(&root->inode_lock);
1190 spin_lock_init(&root->delalloc_lock); 1190 spin_lock_init(&root->delalloc_lock);
1191 spin_lock_init(&root->ordered_extent_lock); 1191 spin_lock_init(&root->ordered_extent_lock);
1192 spin_lock_init(&root->accounting_lock); 1192 spin_lock_init(&root->accounting_lock);
1193 spin_lock_init(&root->log_extents_lock[0]); 1193 spin_lock_init(&root->log_extents_lock[0]);
1194 spin_lock_init(&root->log_extents_lock[1]); 1194 spin_lock_init(&root->log_extents_lock[1]);
1195 mutex_init(&root->objectid_mutex); 1195 mutex_init(&root->objectid_mutex);
1196 mutex_init(&root->log_mutex); 1196 mutex_init(&root->log_mutex);
1197 init_waitqueue_head(&root->log_writer_wait); 1197 init_waitqueue_head(&root->log_writer_wait);
1198 init_waitqueue_head(&root->log_commit_wait[0]); 1198 init_waitqueue_head(&root->log_commit_wait[0]);
1199 init_waitqueue_head(&root->log_commit_wait[1]); 1199 init_waitqueue_head(&root->log_commit_wait[1]);
1200 atomic_set(&root->log_commit[0], 0); 1200 atomic_set(&root->log_commit[0], 0);
1201 atomic_set(&root->log_commit[1], 0); 1201 atomic_set(&root->log_commit[1], 0);
1202 atomic_set(&root->log_writers, 0); 1202 atomic_set(&root->log_writers, 0);
1203 atomic_set(&root->log_batch, 0); 1203 atomic_set(&root->log_batch, 0);
1204 atomic_set(&root->orphan_inodes, 0); 1204 atomic_set(&root->orphan_inodes, 0);
1205 atomic_set(&root->refs, 1); 1205 atomic_set(&root->refs, 1);
1206 root->log_transid = 0; 1206 root->log_transid = 0;
1207 root->last_log_commit = 0; 1207 root->last_log_commit = 0;
1208 if (fs_info) 1208 if (fs_info)
1209 extent_io_tree_init(&root->dirty_log_pages, 1209 extent_io_tree_init(&root->dirty_log_pages,
1210 fs_info->btree_inode->i_mapping); 1210 fs_info->btree_inode->i_mapping);
1211 1211
1212 memset(&root->root_key, 0, sizeof(root->root_key)); 1212 memset(&root->root_key, 0, sizeof(root->root_key));
1213 memset(&root->root_item, 0, sizeof(root->root_item)); 1213 memset(&root->root_item, 0, sizeof(root->root_item));
1214 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 1214 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
1215 memset(&root->root_kobj, 0, sizeof(root->root_kobj)); 1215 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
1216 if (fs_info) 1216 if (fs_info)
1217 root->defrag_trans_start = fs_info->generation; 1217 root->defrag_trans_start = fs_info->generation;
1218 else 1218 else
1219 root->defrag_trans_start = 0; 1219 root->defrag_trans_start = 0;
1220 init_completion(&root->kobj_unregister); 1220 init_completion(&root->kobj_unregister);
1221 root->defrag_running = 0; 1221 root->defrag_running = 0;
1222 root->root_key.objectid = objectid; 1222 root->root_key.objectid = objectid;
1223 root->anon_dev = 0; 1223 root->anon_dev = 0;
1224 1224
1225 spin_lock_init(&root->root_item_lock); 1225 spin_lock_init(&root->root_item_lock);
1226 } 1226 }
1227 1227
1228 static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) 1228 static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
1229 { 1229 {
1230 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); 1230 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1231 if (root) 1231 if (root)
1232 root->fs_info = fs_info; 1232 root->fs_info = fs_info;
1233 return root; 1233 return root;
1234 } 1234 }
1235 1235
1236 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 1236 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1237 /* Should only be used by the testing infrastructure */ 1237 /* Should only be used by the testing infrastructure */
1238 struct btrfs_root *btrfs_alloc_dummy_root(void) 1238 struct btrfs_root *btrfs_alloc_dummy_root(void)
1239 { 1239 {
1240 struct btrfs_root *root; 1240 struct btrfs_root *root;
1241 1241
1242 root = btrfs_alloc_root(NULL); 1242 root = btrfs_alloc_root(NULL);
1243 if (!root) 1243 if (!root)
1244 return ERR_PTR(-ENOMEM); 1244 return ERR_PTR(-ENOMEM);
1245 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1); 1245 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1246 root->dummy_root = 1; 1246 root->dummy_root = 1;
1247 1247
1248 return root; 1248 return root;
1249 } 1249 }
1250 #endif 1250 #endif
1251 1251
1252 struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 1252 struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1253 struct btrfs_fs_info *fs_info, 1253 struct btrfs_fs_info *fs_info,
1254 u64 objectid) 1254 u64 objectid)
1255 { 1255 {
1256 struct extent_buffer *leaf; 1256 struct extent_buffer *leaf;
1257 struct btrfs_root *tree_root = fs_info->tree_root; 1257 struct btrfs_root *tree_root = fs_info->tree_root;
1258 struct btrfs_root *root; 1258 struct btrfs_root *root;
1259 struct btrfs_key key; 1259 struct btrfs_key key;
1260 int ret = 0; 1260 int ret = 0;
1261 uuid_le uuid; 1261 uuid_le uuid;
1262 1262
1263 root = btrfs_alloc_root(fs_info); 1263 root = btrfs_alloc_root(fs_info);
1264 if (!root) 1264 if (!root)
1265 return ERR_PTR(-ENOMEM); 1265 return ERR_PTR(-ENOMEM);
1266 1266
1267 __setup_root(tree_root->nodesize, tree_root->leafsize, 1267 __setup_root(tree_root->nodesize, tree_root->leafsize,
1268 tree_root->sectorsize, tree_root->stripesize, 1268 tree_root->sectorsize, tree_root->stripesize,
1269 root, fs_info, objectid); 1269 root, fs_info, objectid);
1270 root->root_key.objectid = objectid; 1270 root->root_key.objectid = objectid;
1271 root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1271 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1272 root->root_key.offset = 0; 1272 root->root_key.offset = 0;
1273 1273
1274 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 1274 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1275 0, objectid, NULL, 0, 0, 0); 1275 0, objectid, NULL, 0, 0, 0);
1276 if (IS_ERR(leaf)) { 1276 if (IS_ERR(leaf)) {
1277 ret = PTR_ERR(leaf); 1277 ret = PTR_ERR(leaf);
1278 leaf = NULL; 1278 leaf = NULL;
1279 goto fail; 1279 goto fail;
1280 } 1280 }
1281 1281
1282 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 1282 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1283 btrfs_set_header_bytenr(leaf, leaf->start); 1283 btrfs_set_header_bytenr(leaf, leaf->start);
1284 btrfs_set_header_generation(leaf, trans->transid); 1284 btrfs_set_header_generation(leaf, trans->transid);
1285 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 1285 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1286 btrfs_set_header_owner(leaf, objectid); 1286 btrfs_set_header_owner(leaf, objectid);
1287 root->node = leaf; 1287 root->node = leaf;
1288 1288
1289 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 1289 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
1290 BTRFS_FSID_SIZE); 1290 BTRFS_FSID_SIZE);
1291 write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 1291 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1292 btrfs_header_chunk_tree_uuid(leaf), 1292 btrfs_header_chunk_tree_uuid(leaf),
1293 BTRFS_UUID_SIZE); 1293 BTRFS_UUID_SIZE);
1294 btrfs_mark_buffer_dirty(leaf); 1294 btrfs_mark_buffer_dirty(leaf);
1295 1295
1296 root->commit_root = btrfs_root_node(root); 1296 root->commit_root = btrfs_root_node(root);
1297 root->track_dirty = 1; 1297 root->track_dirty = 1;
1298 1298
1299 1299
1300 root->root_item.flags = 0; 1300 root->root_item.flags = 0;
1301 root->root_item.byte_limit = 0; 1301 root->root_item.byte_limit = 0;
1302 btrfs_set_root_bytenr(&root->root_item, leaf->start); 1302 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1303 btrfs_set_root_generation(&root->root_item, trans->transid); 1303 btrfs_set_root_generation(&root->root_item, trans->transid);
1304 btrfs_set_root_level(&root->root_item, 0); 1304 btrfs_set_root_level(&root->root_item, 0);
1305 btrfs_set_root_refs(&root->root_item, 1); 1305 btrfs_set_root_refs(&root->root_item, 1);
1306 btrfs_set_root_used(&root->root_item, leaf->len); 1306 btrfs_set_root_used(&root->root_item, leaf->len);
1307 btrfs_set_root_last_snapshot(&root->root_item, 0); 1307 btrfs_set_root_last_snapshot(&root->root_item, 0);
1308 btrfs_set_root_dirid(&root->root_item, 0); 1308 btrfs_set_root_dirid(&root->root_item, 0);
1309 uuid_le_gen(&uuid); 1309 uuid_le_gen(&uuid);
1310 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 1310 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
1311 root->root_item.drop_level = 0; 1311 root->root_item.drop_level = 0;
1312 1312
1313 key.objectid = objectid; 1313 key.objectid = objectid;
1314 key.type = BTRFS_ROOT_ITEM_KEY; 1314 key.type = BTRFS_ROOT_ITEM_KEY;
1315 key.offset = 0; 1315 key.offset = 0;
1316 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 1316 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1317 if (ret) 1317 if (ret)
1318 goto fail; 1318 goto fail;
1319 1319
1320 btrfs_tree_unlock(leaf); 1320 btrfs_tree_unlock(leaf);
1321 1321
1322 return root; 1322 return root;
1323 1323
1324 fail: 1324 fail:
1325 if (leaf) { 1325 if (leaf) {
1326 btrfs_tree_unlock(leaf); 1326 btrfs_tree_unlock(leaf);
1327 free_extent_buffer(leaf); 1327 free_extent_buffer(leaf);
1328 } 1328 }
1329 kfree(root); 1329 kfree(root);
1330 1330
1331 return ERR_PTR(ret); 1331 return ERR_PTR(ret);
1332 } 1332 }
1333 1333
1334 static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1334 static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1335 struct btrfs_fs_info *fs_info) 1335 struct btrfs_fs_info *fs_info)
1336 { 1336 {
1337 struct btrfs_root *root; 1337 struct btrfs_root *root;
1338 struct btrfs_root *tree_root = fs_info->tree_root; 1338 struct btrfs_root *tree_root = fs_info->tree_root;
1339 struct extent_buffer *leaf; 1339 struct extent_buffer *leaf;
1340 1340
1341 root = btrfs_alloc_root(fs_info); 1341 root = btrfs_alloc_root(fs_info);
1342 if (!root) 1342 if (!root)
1343 return ERR_PTR(-ENOMEM); 1343 return ERR_PTR(-ENOMEM);
1344 1344
1345 __setup_root(tree_root->nodesize, tree_root->leafsize, 1345 __setup_root(tree_root->nodesize, tree_root->leafsize,
1346 tree_root->sectorsize, tree_root->stripesize, 1346 tree_root->sectorsize, tree_root->stripesize,
1347 root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1347 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1348 1348
1349 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1349 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1350 root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1350 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1351 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 1351 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
1352 /* 1352 /*
1353 * log trees do not get reference counted because they go away 1353 * log trees do not get reference counted because they go away
1354 * before a real commit is actually done. They do store pointers 1354 * before a real commit is actually done. They do store pointers
1355 * to file data extents, and those reference counts still get 1355 * to file data extents, and those reference counts still get
1356 * updated (along with back refs to the log tree). 1356 * updated (along with back refs to the log tree).
1357 */ 1357 */
1358 root->ref_cows = 0; 1358 root->ref_cows = 0;
1359 1359
1360 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 1360 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1361 BTRFS_TREE_LOG_OBJECTID, NULL, 1361 BTRFS_TREE_LOG_OBJECTID, NULL,
1362 0, 0, 0); 1362 0, 0, 0);
1363 if (IS_ERR(leaf)) { 1363 if (IS_ERR(leaf)) {
1364 kfree(root); 1364 kfree(root);
1365 return ERR_CAST(leaf); 1365 return ERR_CAST(leaf);
1366 } 1366 }
1367 1367
1368 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 1368 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1369 btrfs_set_header_bytenr(leaf, leaf->start); 1369 btrfs_set_header_bytenr(leaf, leaf->start);
1370 btrfs_set_header_generation(leaf, trans->transid); 1370 btrfs_set_header_generation(leaf, trans->transid);
1371 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 1371 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1372 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 1372 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
1373 root->node = leaf; 1373 root->node = leaf;
1374 1374
1375 write_extent_buffer(root->node, root->fs_info->fsid, 1375 write_extent_buffer(root->node, root->fs_info->fsid,
1376 btrfs_header_fsid(), BTRFS_FSID_SIZE); 1376 btrfs_header_fsid(), BTRFS_FSID_SIZE);
1377 btrfs_mark_buffer_dirty(root->node); 1377 btrfs_mark_buffer_dirty(root->node);
1378 btrfs_tree_unlock(root->node); 1378 btrfs_tree_unlock(root->node);
1379 return root; 1379 return root;
1380 } 1380 }
1381 1381
1382 int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 1382 int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1383 struct btrfs_fs_info *fs_info) 1383 struct btrfs_fs_info *fs_info)
1384 { 1384 {
1385 struct btrfs_root *log_root; 1385 struct btrfs_root *log_root;
1386 1386
1387 log_root = alloc_log_tree(trans, fs_info); 1387 log_root = alloc_log_tree(trans, fs_info);
1388 if (IS_ERR(log_root)) 1388 if (IS_ERR(log_root))
1389 return PTR_ERR(log_root); 1389 return PTR_ERR(log_root);
1390 WARN_ON(fs_info->log_root_tree); 1390 WARN_ON(fs_info->log_root_tree);
1391 fs_info->log_root_tree = log_root; 1391 fs_info->log_root_tree = log_root;
1392 return 0; 1392 return 0;
1393 } 1393 }
1394 1394
1395 int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 1395 int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1396 struct btrfs_root *root) 1396 struct btrfs_root *root)
1397 { 1397 {
1398 struct btrfs_root *log_root; 1398 struct btrfs_root *log_root;
1399 struct btrfs_inode_item *inode_item; 1399 struct btrfs_inode_item *inode_item;
1400 1400
1401 log_root = alloc_log_tree(trans, root->fs_info); 1401 log_root = alloc_log_tree(trans, root->fs_info);
1402 if (IS_ERR(log_root)) 1402 if (IS_ERR(log_root))
1403 return PTR_ERR(log_root); 1403 return PTR_ERR(log_root);
1404 1404
1405 log_root->last_trans = trans->transid; 1405 log_root->last_trans = trans->transid;
1406 log_root->root_key.offset = root->root_key.objectid; 1406 log_root->root_key.offset = root->root_key.objectid;
1407 1407
1408 inode_item = &log_root->root_item.inode; 1408 inode_item = &log_root->root_item.inode;
1409 btrfs_set_stack_inode_generation(inode_item, 1); 1409 btrfs_set_stack_inode_generation(inode_item, 1);
1410 btrfs_set_stack_inode_size(inode_item, 3); 1410 btrfs_set_stack_inode_size(inode_item, 3);
1411 btrfs_set_stack_inode_nlink(inode_item, 1); 1411 btrfs_set_stack_inode_nlink(inode_item, 1);
1412 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); 1412 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1413 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 1413 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1414 1414
1415 btrfs_set_root_node(&log_root->root_item, log_root->node); 1415 btrfs_set_root_node(&log_root->root_item, log_root->node);
1416 1416
1417 WARN_ON(root->log_root); 1417 WARN_ON(root->log_root);
1418 root->log_root = log_root; 1418 root->log_root = log_root;
1419 root->log_transid = 0; 1419 root->log_transid = 0;
1420 root->last_log_commit = 0; 1420 root->last_log_commit = 0;
1421 return 0; 1421 return 0;
1422 } 1422 }
1423 1423
1424 static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1424 static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1425 struct btrfs_key *key) 1425 struct btrfs_key *key)
1426 { 1426 {
1427 struct btrfs_root *root; 1427 struct btrfs_root *root;
1428 struct btrfs_fs_info *fs_info = tree_root->fs_info; 1428 struct btrfs_fs_info *fs_info = tree_root->fs_info;
1429 struct btrfs_path *path; 1429 struct btrfs_path *path;
1430 u64 generation; 1430 u64 generation;
1431 u32 blocksize; 1431 u32 blocksize;
1432 int ret; 1432 int ret;
1433 1433
1434 path = btrfs_alloc_path(); 1434 path = btrfs_alloc_path();
1435 if (!path) 1435 if (!path)
1436 return ERR_PTR(-ENOMEM); 1436 return ERR_PTR(-ENOMEM);
1437 1437
1438 root = btrfs_alloc_root(fs_info); 1438 root = btrfs_alloc_root(fs_info);
1439 if (!root) { 1439 if (!root) {
1440 ret = -ENOMEM; 1440 ret = -ENOMEM;
1441 goto alloc_fail; 1441 goto alloc_fail;
1442 } 1442 }
1443 1443
1444 __setup_root(tree_root->nodesize, tree_root->leafsize, 1444 __setup_root(tree_root->nodesize, tree_root->leafsize,
1445 tree_root->sectorsize, tree_root->stripesize, 1445 tree_root->sectorsize, tree_root->stripesize,
1446 root, fs_info, key->objectid); 1446 root, fs_info, key->objectid);
1447 1447
1448 ret = btrfs_find_root(tree_root, key, path, 1448 ret = btrfs_find_root(tree_root, key, path,
1449 &root->root_item, &root->root_key); 1449 &root->root_item, &root->root_key);
1450 if (ret) { 1450 if (ret) {
1451 if (ret > 0) 1451 if (ret > 0)
1452 ret = -ENOENT; 1452 ret = -ENOENT;
1453 goto find_fail; 1453 goto find_fail;
1454 } 1454 }
1455 1455
1456 generation = btrfs_root_generation(&root->root_item); 1456 generation = btrfs_root_generation(&root->root_item);
1457 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); 1457 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1458 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 1458 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
1459 blocksize, generation); 1459 blocksize, generation);
1460 if (!root->node) { 1460 if (!root->node) {
1461 ret = -ENOMEM; 1461 ret = -ENOMEM;
1462 goto find_fail; 1462 goto find_fail;
1463 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1463 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1464 ret = -EIO; 1464 ret = -EIO;
1465 goto read_fail; 1465 goto read_fail;
1466 } 1466 }
1467 root->commit_root = btrfs_root_node(root); 1467 root->commit_root = btrfs_root_node(root);
1468 out: 1468 out:
1469 btrfs_free_path(path); 1469 btrfs_free_path(path);
1470 return root; 1470 return root;
1471 1471
1472 read_fail: 1472 read_fail:
1473 free_extent_buffer(root->node); 1473 free_extent_buffer(root->node);
1474 find_fail: 1474 find_fail:
1475 kfree(root); 1475 kfree(root);
1476 alloc_fail: 1476 alloc_fail:
1477 root = ERR_PTR(ret); 1477 root = ERR_PTR(ret);
1478 goto out; 1478 goto out;
1479 } 1479 }
1480 1480
1481 struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1481 struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1482 struct btrfs_key *location) 1482 struct btrfs_key *location)
1483 { 1483 {
1484 struct btrfs_root *root; 1484 struct btrfs_root *root;
1485 1485
1486 root = btrfs_read_tree_root(tree_root, location); 1486 root = btrfs_read_tree_root(tree_root, location);
1487 if (IS_ERR(root)) 1487 if (IS_ERR(root))
1488 return root; 1488 return root;
1489 1489
1490 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 1490 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
1491 root->ref_cows = 1; 1491 root->ref_cows = 1;
1492 btrfs_check_and_init_root_item(&root->root_item); 1492 btrfs_check_and_init_root_item(&root->root_item);
1493 } 1493 }
1494 1494
1495 return root; 1495 return root;
1496 } 1496 }
1497 1497
1498 int btrfs_init_fs_root(struct btrfs_root *root) 1498 int btrfs_init_fs_root(struct btrfs_root *root)
1499 { 1499 {
1500 int ret; 1500 int ret;
1501 1501
1502 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1502 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1503 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1503 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1504 GFP_NOFS); 1504 GFP_NOFS);
1505 if (!root->free_ino_pinned || !root->free_ino_ctl) { 1505 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1506 ret = -ENOMEM; 1506 ret = -ENOMEM;
1507 goto fail; 1507 goto fail;
1508 } 1508 }
1509 1509
1510 btrfs_init_free_ino_ctl(root); 1510 btrfs_init_free_ino_ctl(root);
1511 mutex_init(&root->fs_commit_mutex); 1511 mutex_init(&root->fs_commit_mutex);
1512 spin_lock_init(&root->cache_lock); 1512 spin_lock_init(&root->cache_lock);
1513 init_waitqueue_head(&root->cache_wait); 1513 init_waitqueue_head(&root->cache_wait);
1514 1514
1515 ret = get_anon_bdev(&root->anon_dev); 1515 ret = get_anon_bdev(&root->anon_dev);
1516 if (ret) 1516 if (ret)
1517 goto fail; 1517 goto fail;
1518 return 0; 1518 return 0;
1519 fail: 1519 fail:
1520 kfree(root->free_ino_ctl); 1520 kfree(root->free_ino_ctl);
1521 kfree(root->free_ino_pinned); 1521 kfree(root->free_ino_pinned);
1522 return ret; 1522 return ret;
1523 } 1523 }
1524 1524
1525 static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1525 static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1526 u64 root_id) 1526 u64 root_id)
1527 { 1527 {
1528 struct btrfs_root *root; 1528 struct btrfs_root *root;
1529 1529
1530 spin_lock(&fs_info->fs_roots_radix_lock); 1530 spin_lock(&fs_info->fs_roots_radix_lock);
1531 root = radix_tree_lookup(&fs_info->fs_roots_radix, 1531 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1532 (unsigned long)root_id); 1532 (unsigned long)root_id);
1533 spin_unlock(&fs_info->fs_roots_radix_lock); 1533 spin_unlock(&fs_info->fs_roots_radix_lock);
1534 return root; 1534 return root;
1535 } 1535 }
1536 1536
1537 int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1537 int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1538 struct btrfs_root *root) 1538 struct btrfs_root *root)
1539 { 1539 {
1540 int ret; 1540 int ret;
1541 1541
1542 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 1542 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1543 if (ret) 1543 if (ret)
1544 return ret; 1544 return ret;
1545 1545
1546 spin_lock(&fs_info->fs_roots_radix_lock); 1546 spin_lock(&fs_info->fs_roots_radix_lock);
1547 ret = radix_tree_insert(&fs_info->fs_roots_radix, 1547 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1548 (unsigned long)root->root_key.objectid, 1548 (unsigned long)root->root_key.objectid,
1549 root); 1549 root);
1550 if (ret == 0) 1550 if (ret == 0)
1551 root->in_radix = 1; 1551 root->in_radix = 1;
1552 spin_unlock(&fs_info->fs_roots_radix_lock); 1552 spin_unlock(&fs_info->fs_roots_radix_lock);
1553 radix_tree_preload_end(); 1553 radix_tree_preload_end();
1554 1554
1555 return ret; 1555 return ret;
1556 } 1556 }
1557 1557
1558 struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1558 struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1559 struct btrfs_key *location, 1559 struct btrfs_key *location,
1560 bool check_ref) 1560 bool check_ref)
1561 { 1561 {
1562 struct btrfs_root *root; 1562 struct btrfs_root *root;
1563 int ret; 1563 int ret;
1564 1564
1565 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1565 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1566 return fs_info->tree_root; 1566 return fs_info->tree_root;
1567 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1567 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1568 return fs_info->extent_root; 1568 return fs_info->extent_root;
1569 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 1569 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1570 return fs_info->chunk_root; 1570 return fs_info->chunk_root;
1571 if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 1571 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1572 return fs_info->dev_root; 1572 return fs_info->dev_root;
1573 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 1573 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1574 return fs_info->csum_root; 1574 return fs_info->csum_root;
1575 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1575 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1576 return fs_info->quota_root ? fs_info->quota_root : 1576 return fs_info->quota_root ? fs_info->quota_root :
1577 ERR_PTR(-ENOENT); 1577 ERR_PTR(-ENOENT);
1578 if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1578 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1579 return fs_info->uuid_root ? fs_info->uuid_root : 1579 return fs_info->uuid_root ? fs_info->uuid_root :
1580 ERR_PTR(-ENOENT); 1580 ERR_PTR(-ENOENT);
1581 again: 1581 again:
1582 root = btrfs_lookup_fs_root(fs_info, location->objectid); 1582 root = btrfs_lookup_fs_root(fs_info, location->objectid);
1583 if (root) { 1583 if (root) {
1584 if (check_ref && btrfs_root_refs(&root->root_item) == 0) 1584 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
1585 return ERR_PTR(-ENOENT); 1585 return ERR_PTR(-ENOENT);
1586 return root; 1586 return root;
1587 } 1587 }
1588 1588
1589 root = btrfs_read_fs_root(fs_info->tree_root, location); 1589 root = btrfs_read_fs_root(fs_info->tree_root, location);
1590 if (IS_ERR(root)) 1590 if (IS_ERR(root))
1591 return root; 1591 return root;
1592 1592
1593 if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1593 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
1594 ret = -ENOENT; 1594 ret = -ENOENT;
1595 goto fail; 1595 goto fail;
1596 } 1596 }
1597 1597
1598 ret = btrfs_init_fs_root(root); 1598 ret = btrfs_init_fs_root(root);
1599 if (ret) 1599 if (ret)
1600 goto fail; 1600 goto fail;
1601 1601
1602 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID, 1602 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1603 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL); 1603 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
1604 if (ret < 0) 1604 if (ret < 0)
1605 goto fail; 1605 goto fail;
1606 if (ret == 0) 1606 if (ret == 0)
1607 root->orphan_item_inserted = 1; 1607 root->orphan_item_inserted = 1;
1608 1608
1609 ret = btrfs_insert_fs_root(fs_info, root); 1609 ret = btrfs_insert_fs_root(fs_info, root);
1610 if (ret) { 1610 if (ret) {
1611 if (ret == -EEXIST) { 1611 if (ret == -EEXIST) {
1612 free_fs_root(root); 1612 free_fs_root(root);
1613 goto again; 1613 goto again;
1614 } 1614 }
1615 goto fail; 1615 goto fail;
1616 } 1616 }
1617 return root; 1617 return root;
1618 fail: 1618 fail:
1619 free_fs_root(root); 1619 free_fs_root(root);
1620 return ERR_PTR(ret); 1620 return ERR_PTR(ret);
1621 } 1621 }
1622 1622
1623 static int btrfs_congested_fn(void *congested_data, int bdi_bits) 1623 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1624 { 1624 {
1625 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 1625 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1626 int ret = 0; 1626 int ret = 0;
1627 struct btrfs_device *device; 1627 struct btrfs_device *device;
1628 struct backing_dev_info *bdi; 1628 struct backing_dev_info *bdi;
1629 1629
1630 rcu_read_lock(); 1630 rcu_read_lock();
1631 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1631 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
1632 if (!device->bdev) 1632 if (!device->bdev)
1633 continue; 1633 continue;
1634 bdi = blk_get_backing_dev_info(device->bdev); 1634 bdi = blk_get_backing_dev_info(device->bdev);
1635 if (bdi && bdi_congested(bdi, bdi_bits)) { 1635 if (bdi && bdi_congested(bdi, bdi_bits)) {
1636 ret = 1; 1636 ret = 1;
1637 break; 1637 break;
1638 } 1638 }
1639 } 1639 }
1640 rcu_read_unlock(); 1640 rcu_read_unlock();
1641 return ret; 1641 return ret;
1642 } 1642 }
1643 1643
1644 /* 1644 /*
1645 * If this fails, caller must call bdi_destroy() to get rid of the 1645 * If this fails, caller must call bdi_destroy() to get rid of the
1646 * bdi again. 1646 * bdi again.
1647 */ 1647 */
1648 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 1648 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1649 { 1649 {
1650 int err; 1650 int err;
1651 1651
1652 bdi->capabilities = BDI_CAP_MAP_COPY; 1652 bdi->capabilities = BDI_CAP_MAP_COPY;
1653 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); 1653 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
1654 if (err) 1654 if (err)
1655 return err; 1655 return err;
1656 1656
1657 bdi->ra_pages = default_backing_dev_info.ra_pages; 1657 bdi->ra_pages = default_backing_dev_info.ra_pages;
1658 bdi->congested_fn = btrfs_congested_fn; 1658 bdi->congested_fn = btrfs_congested_fn;
1659 bdi->congested_data = info; 1659 bdi->congested_data = info;
1660 return 0; 1660 return 0;
1661 } 1661 }
1662 1662
1663 /* 1663 /*
1664 * called by the kthread helper functions to finally call the bio end_io 1664 * called by the kthread helper functions to finally call the bio end_io
1665 * functions. This is where read checksum verification actually happens 1665 * functions. This is where read checksum verification actually happens
1666 */ 1666 */
1667 static void end_workqueue_fn(struct btrfs_work *work) 1667 static void end_workqueue_fn(struct btrfs_work *work)
1668 { 1668 {
1669 struct bio *bio; 1669 struct bio *bio;
1670 struct end_io_wq *end_io_wq; 1670 struct end_io_wq *end_io_wq;
1671 int error; 1671 int error;
1672 1672
1673 end_io_wq = container_of(work, struct end_io_wq, work); 1673 end_io_wq = container_of(work, struct end_io_wq, work);
1674 bio = end_io_wq->bio; 1674 bio = end_io_wq->bio;
1675 1675
1676 error = end_io_wq->error; 1676 error = end_io_wq->error;
1677 bio->bi_private = end_io_wq->private; 1677 bio->bi_private = end_io_wq->private;
1678 bio->bi_end_io = end_io_wq->end_io; 1678 bio->bi_end_io = end_io_wq->end_io;
1679 kfree(end_io_wq); 1679 kfree(end_io_wq);
1680 bio_endio_nodec(bio, error); 1680 bio_endio_nodec(bio, error);
1681 } 1681 }
1682 1682
1683 static int cleaner_kthread(void *arg) 1683 static int cleaner_kthread(void *arg)
1684 { 1684 {
1685 struct btrfs_root *root = arg; 1685 struct btrfs_root *root = arg;
1686 int again; 1686 int again;
1687 1687
1688 do { 1688 do {
1689 again = 0; 1689 again = 0;
1690 1690
1691 /* Make the cleaner go to sleep early. */ 1691 /* Make the cleaner go to sleep early. */
1692 if (btrfs_need_cleaner_sleep(root)) 1692 if (btrfs_need_cleaner_sleep(root))
1693 goto sleep; 1693 goto sleep;
1694 1694
1695 if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1695 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1696 goto sleep; 1696 goto sleep;
1697 1697
1698 /* 1698 /*
1699 * Avoid the problem that we change the status of the fs 1699 * Avoid the problem that we change the status of the fs
1700 * during the above check and trylock. 1700 * during the above check and trylock.
1701 */ 1701 */
1702 if (btrfs_need_cleaner_sleep(root)) { 1702 if (btrfs_need_cleaner_sleep(root)) {
1703 mutex_unlock(&root->fs_info->cleaner_mutex); 1703 mutex_unlock(&root->fs_info->cleaner_mutex);
1704 goto sleep; 1704 goto sleep;
1705 } 1705 }
1706 1706
1707 btrfs_run_delayed_iputs(root); 1707 btrfs_run_delayed_iputs(root);
1708 again = btrfs_clean_one_deleted_snapshot(root); 1708 again = btrfs_clean_one_deleted_snapshot(root);
1709 mutex_unlock(&root->fs_info->cleaner_mutex); 1709 mutex_unlock(&root->fs_info->cleaner_mutex);
1710 1710
1711 /* 1711 /*
1712 * The defragger has dealt with the R/O remount and umount, 1712 * The defragger has dealt with the R/O remount and umount,
1713 * needn't do anything special here. 1713 * needn't do anything special here.
1714 */ 1714 */
1715 btrfs_run_defrag_inodes(root->fs_info); 1715 btrfs_run_defrag_inodes(root->fs_info);
1716 sleep: 1716 sleep:
1717 if (!try_to_freeze() && !again) { 1717 if (!try_to_freeze() && !again) {
1718 set_current_state(TASK_INTERRUPTIBLE); 1718 set_current_state(TASK_INTERRUPTIBLE);
1719 if (!kthread_should_stop()) 1719 if (!kthread_should_stop())
1720 schedule(); 1720 schedule();
1721 __set_current_state(TASK_RUNNING); 1721 __set_current_state(TASK_RUNNING);
1722 } 1722 }
1723 } while (!kthread_should_stop()); 1723 } while (!kthread_should_stop());
1724 return 0; 1724 return 0;
1725 } 1725 }
1726 1726
1727 static int transaction_kthread(void *arg) 1727 static int transaction_kthread(void *arg)
1728 { 1728 {
1729 struct btrfs_root *root = arg; 1729 struct btrfs_root *root = arg;
1730 struct btrfs_trans_handle *trans; 1730 struct btrfs_trans_handle *trans;
1731 struct btrfs_transaction *cur; 1731 struct btrfs_transaction *cur;
1732 u64 transid; 1732 u64 transid;
1733 unsigned long now; 1733 unsigned long now;
1734 unsigned long delay; 1734 unsigned long delay;
1735 bool cannot_commit; 1735 bool cannot_commit;
1736 1736
1737 do { 1737 do {
1738 cannot_commit = false; 1738 cannot_commit = false;
1739 delay = HZ * root->fs_info->commit_interval; 1739 delay = HZ * root->fs_info->commit_interval;
1740 mutex_lock(&root->fs_info->transaction_kthread_mutex); 1740 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1741 1741
1742 spin_lock(&root->fs_info->trans_lock); 1742 spin_lock(&root->fs_info->trans_lock);
1743 cur = root->fs_info->running_transaction; 1743 cur = root->fs_info->running_transaction;
1744 if (!cur) { 1744 if (!cur) {
1745 spin_unlock(&root->fs_info->trans_lock); 1745 spin_unlock(&root->fs_info->trans_lock);
1746 goto sleep; 1746 goto sleep;
1747 } 1747 }
1748 1748
1749 now = get_seconds(); 1749 now = get_seconds();
1750 if (cur->state < TRANS_STATE_BLOCKED && 1750 if (cur->state < TRANS_STATE_BLOCKED &&
1751 (now < cur->start_time || 1751 (now < cur->start_time ||
1752 now - cur->start_time < root->fs_info->commit_interval)) { 1752 now - cur->start_time < root->fs_info->commit_interval)) {
1753 spin_unlock(&root->fs_info->trans_lock); 1753 spin_unlock(&root->fs_info->trans_lock);
1754 delay = HZ * 5; 1754 delay = HZ * 5;
1755 goto sleep; 1755 goto sleep;
1756 } 1756 }
1757 transid = cur->transid; 1757 transid = cur->transid;
1758 spin_unlock(&root->fs_info->trans_lock); 1758 spin_unlock(&root->fs_info->trans_lock);
1759 1759
1760 /* If the file system is aborted, this will always fail. */ 1760 /* If the file system is aborted, this will always fail. */
1761 trans = btrfs_attach_transaction(root); 1761 trans = btrfs_attach_transaction(root);
1762 if (IS_ERR(trans)) { 1762 if (IS_ERR(trans)) {
1763 if (PTR_ERR(trans) != -ENOENT) 1763 if (PTR_ERR(trans) != -ENOENT)
1764 cannot_commit = true; 1764 cannot_commit = true;
1765 goto sleep; 1765 goto sleep;
1766 } 1766 }
1767 if (transid == trans->transid) { 1767 if (transid == trans->transid) {
1768 btrfs_commit_transaction(trans, root); 1768 btrfs_commit_transaction(trans, root);
1769 } else { 1769 } else {
1770 btrfs_end_transaction(trans, root); 1770 btrfs_end_transaction(trans, root);
1771 } 1771 }
1772 sleep: 1772 sleep:
1773 wake_up_process(root->fs_info->cleaner_kthread); 1773 wake_up_process(root->fs_info->cleaner_kthread);
1774 mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1774 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1775 1775
1776 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 1776 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1777 &root->fs_info->fs_state))) 1777 &root->fs_info->fs_state)))
1778 btrfs_cleanup_transaction(root); 1778 btrfs_cleanup_transaction(root);
1779 if (!try_to_freeze()) { 1779 if (!try_to_freeze()) {
1780 set_current_state(TASK_INTERRUPTIBLE); 1780 set_current_state(TASK_INTERRUPTIBLE);
1781 if (!kthread_should_stop() && 1781 if (!kthread_should_stop() &&
1782 (!btrfs_transaction_blocked(root->fs_info) || 1782 (!btrfs_transaction_blocked(root->fs_info) ||
1783 cannot_commit)) 1783 cannot_commit))
1784 schedule_timeout(delay); 1784 schedule_timeout(delay);
1785 __set_current_state(TASK_RUNNING); 1785 __set_current_state(TASK_RUNNING);
1786 } 1786 }
1787 } while (!kthread_should_stop()); 1787 } while (!kthread_should_stop());
1788 return 0; 1788 return 0;
1789 } 1789 }
1790 1790
1791 /* 1791 /*
1792 * this will find the highest generation in the array of 1792 * this will find the highest generation in the array of
1793 * root backups. The index of the highest array is returned, 1793 * root backups. The index of the highest array is returned,
1794 * or -1 if we can't find anything. 1794 * or -1 if we can't find anything.
1795 * 1795 *
1796 * We check to make sure the array is valid by comparing the 1796 * We check to make sure the array is valid by comparing the
1797 * generation of the latest root in the array with the generation 1797 * generation of the latest root in the array with the generation
1798 * in the super block. If they don't match we pitch it. 1798 * in the super block. If they don't match we pitch it.
1799 */ 1799 */
1800 static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1800 static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1801 { 1801 {
1802 u64 cur; 1802 u64 cur;
1803 int newest_index = -1; 1803 int newest_index = -1;
1804 struct btrfs_root_backup *root_backup; 1804 struct btrfs_root_backup *root_backup;
1805 int i; 1805 int i;
1806 1806
1807 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1807 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1808 root_backup = info->super_copy->super_roots + i; 1808 root_backup = info->super_copy->super_roots + i;
1809 cur = btrfs_backup_tree_root_gen(root_backup); 1809 cur = btrfs_backup_tree_root_gen(root_backup);
1810 if (cur == newest_gen) 1810 if (cur == newest_gen)
1811 newest_index = i; 1811 newest_index = i;
1812 } 1812 }
1813 1813
1814 /* check to see if we actually wrapped around */ 1814 /* check to see if we actually wrapped around */
1815 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1815 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1816 root_backup = info->super_copy->super_roots; 1816 root_backup = info->super_copy->super_roots;
1817 cur = btrfs_backup_tree_root_gen(root_backup); 1817 cur = btrfs_backup_tree_root_gen(root_backup);
1818 if (cur == newest_gen) 1818 if (cur == newest_gen)
1819 newest_index = 0; 1819 newest_index = 0;
1820 } 1820 }
1821 return newest_index; 1821 return newest_index;
1822 } 1822 }
1823 1823
1824 1824
1825 /* 1825 /*
1826 * find the oldest backup so we know where to store new entries 1826 * find the oldest backup so we know where to store new entries
1827 * in the backup array. This will set the backup_root_index 1827 * in the backup array. This will set the backup_root_index
1828 * field in the fs_info struct 1828 * field in the fs_info struct
1829 */ 1829 */
1830 static void find_oldest_super_backup(struct btrfs_fs_info *info, 1830 static void find_oldest_super_backup(struct btrfs_fs_info *info,
1831 u64 newest_gen) 1831 u64 newest_gen)
1832 { 1832 {
1833 int newest_index = -1; 1833 int newest_index = -1;
1834 1834
1835 newest_index = find_newest_super_backup(info, newest_gen); 1835 newest_index = find_newest_super_backup(info, newest_gen);
1836 /* if there was garbage in there, just move along */ 1836 /* if there was garbage in there, just move along */
1837 if (newest_index == -1) { 1837 if (newest_index == -1) {
1838 info->backup_root_index = 0; 1838 info->backup_root_index = 0;
1839 } else { 1839 } else {
1840 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1840 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1841 } 1841 }
1842 } 1842 }
1843 1843
1844 /* 1844 /*
1845 * copy all the root pointers into the super backup array. 1845 * copy all the root pointers into the super backup array.
1846 * this will bump the backup pointer by one when it is 1846 * this will bump the backup pointer by one when it is
1847 * done 1847 * done
1848 */ 1848 */
1849 static void backup_super_roots(struct btrfs_fs_info *info) 1849 static void backup_super_roots(struct btrfs_fs_info *info)
1850 { 1850 {
1851 int next_backup; 1851 int next_backup;
1852 struct btrfs_root_backup *root_backup; 1852 struct btrfs_root_backup *root_backup;
1853 int last_backup; 1853 int last_backup;
1854 1854
1855 next_backup = info->backup_root_index; 1855 next_backup = info->backup_root_index;
1856 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1856 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1857 BTRFS_NUM_BACKUP_ROOTS; 1857 BTRFS_NUM_BACKUP_ROOTS;
1858 1858
1859 /* 1859 /*
1860 * just overwrite the last backup if we're at the same generation 1860 * just overwrite the last backup if we're at the same generation
1861 * this happens only at umount 1861 * this happens only at umount
1862 */ 1862 */
1863 root_backup = info->super_for_commit->super_roots + last_backup; 1863 root_backup = info->super_for_commit->super_roots + last_backup;
1864 if (btrfs_backup_tree_root_gen(root_backup) == 1864 if (btrfs_backup_tree_root_gen(root_backup) ==
1865 btrfs_header_generation(info->tree_root->node)) 1865 btrfs_header_generation(info->tree_root->node))
1866 next_backup = last_backup; 1866 next_backup = last_backup;
1867 1867
1868 root_backup = info->super_for_commit->super_roots + next_backup; 1868 root_backup = info->super_for_commit->super_roots + next_backup;
1869 1869
1870 /* 1870 /*
1871 * make sure all of our padding and empty slots get zero filled 1871 * make sure all of our padding and empty slots get zero filled
1872 * regardless of which ones we use today 1872 * regardless of which ones we use today
1873 */ 1873 */
1874 memset(root_backup, 0, sizeof(*root_backup)); 1874 memset(root_backup, 0, sizeof(*root_backup));
1875 1875
1876 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1876 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1877 1877
1878 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1878 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1879 btrfs_set_backup_tree_root_gen(root_backup, 1879 btrfs_set_backup_tree_root_gen(root_backup,
1880 btrfs_header_generation(info->tree_root->node)); 1880 btrfs_header_generation(info->tree_root->node));
1881 1881
1882 btrfs_set_backup_tree_root_level(root_backup, 1882 btrfs_set_backup_tree_root_level(root_backup,
1883 btrfs_header_level(info->tree_root->node)); 1883 btrfs_header_level(info->tree_root->node));
1884 1884
1885 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1885 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1886 btrfs_set_backup_chunk_root_gen(root_backup, 1886 btrfs_set_backup_chunk_root_gen(root_backup,
1887 btrfs_header_generation(info->chunk_root->node)); 1887 btrfs_header_generation(info->chunk_root->node));
1888 btrfs_set_backup_chunk_root_level(root_backup, 1888 btrfs_set_backup_chunk_root_level(root_backup,
1889 btrfs_header_level(info->chunk_root->node)); 1889 btrfs_header_level(info->chunk_root->node));
1890 1890
1891 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1891 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1892 btrfs_set_backup_extent_root_gen(root_backup, 1892 btrfs_set_backup_extent_root_gen(root_backup,
1893 btrfs_header_generation(info->extent_root->node)); 1893 btrfs_header_generation(info->extent_root->node));
1894 btrfs_set_backup_extent_root_level(root_backup, 1894 btrfs_set_backup_extent_root_level(root_backup,
1895 btrfs_header_level(info->extent_root->node)); 1895 btrfs_header_level(info->extent_root->node));
1896 1896
1897 /* 1897 /*
1898 * we might commit during log recovery, which happens before we set 1898 * we might commit during log recovery, which happens before we set
1899 * the fs_root. Make sure it is valid before we fill it in. 1899 * the fs_root. Make sure it is valid before we fill it in.
1900 */ 1900 */
1901 if (info->fs_root && info->fs_root->node) { 1901 if (info->fs_root && info->fs_root->node) {
1902 btrfs_set_backup_fs_root(root_backup, 1902 btrfs_set_backup_fs_root(root_backup,
1903 info->fs_root->node->start); 1903 info->fs_root->node->start);
1904 btrfs_set_backup_fs_root_gen(root_backup, 1904 btrfs_set_backup_fs_root_gen(root_backup,
1905 btrfs_header_generation(info->fs_root->node)); 1905 btrfs_header_generation(info->fs_root->node));
1906 btrfs_set_backup_fs_root_level(root_backup, 1906 btrfs_set_backup_fs_root_level(root_backup,
1907 btrfs_header_level(info->fs_root->node)); 1907 btrfs_header_level(info->fs_root->node));
1908 } 1908 }
1909 1909
1910 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1910 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1911 btrfs_set_backup_dev_root_gen(root_backup, 1911 btrfs_set_backup_dev_root_gen(root_backup,
1912 btrfs_header_generation(info->dev_root->node)); 1912 btrfs_header_generation(info->dev_root->node));
1913 btrfs_set_backup_dev_root_level(root_backup, 1913 btrfs_set_backup_dev_root_level(root_backup,
1914 btrfs_header_level(info->dev_root->node)); 1914 btrfs_header_level(info->dev_root->node));
1915 1915
1916 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1916 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1917 btrfs_set_backup_csum_root_gen(root_backup, 1917 btrfs_set_backup_csum_root_gen(root_backup,
1918 btrfs_header_generation(info->csum_root->node)); 1918 btrfs_header_generation(info->csum_root->node));
1919 btrfs_set_backup_csum_root_level(root_backup, 1919 btrfs_set_backup_csum_root_level(root_backup,
1920 btrfs_header_level(info->csum_root->node)); 1920 btrfs_header_level(info->csum_root->node));
1921 1921
1922 btrfs_set_backup_total_bytes(root_backup, 1922 btrfs_set_backup_total_bytes(root_backup,
1923 btrfs_super_total_bytes(info->super_copy)); 1923 btrfs_super_total_bytes(info->super_copy));
1924 btrfs_set_backup_bytes_used(root_backup, 1924 btrfs_set_backup_bytes_used(root_backup,
1925 btrfs_super_bytes_used(info->super_copy)); 1925 btrfs_super_bytes_used(info->super_copy));
1926 btrfs_set_backup_num_devices(root_backup, 1926 btrfs_set_backup_num_devices(root_backup,
1927 btrfs_super_num_devices(info->super_copy)); 1927 btrfs_super_num_devices(info->super_copy));
1928 1928
1929 /* 1929 /*
1930 * if we don't copy this out to the super_copy, it won't get remembered 1930 * if we don't copy this out to the super_copy, it won't get remembered
1931 * for the next commit 1931 * for the next commit
1932 */ 1932 */
1933 memcpy(&info->super_copy->super_roots, 1933 memcpy(&info->super_copy->super_roots,
1934 &info->super_for_commit->super_roots, 1934 &info->super_for_commit->super_roots,
1935 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1935 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1936 } 1936 }
1937 1937
1938 /* 1938 /*
1939 * this copies info out of the root backup array and back into 1939 * this copies info out of the root backup array and back into
1940 * the in-memory super block. It is meant to help iterate through 1940 * the in-memory super block. It is meant to help iterate through
1941 * the array, so you send it the number of backups you've already 1941 * the array, so you send it the number of backups you've already
1942 * tried and the last backup index you used. 1942 * tried and the last backup index you used.
1943 * 1943 *
1944 * this returns -1 when it has tried all the backups 1944 * this returns -1 when it has tried all the backups
1945 */ 1945 */
1946 static noinline int next_root_backup(struct btrfs_fs_info *info, 1946 static noinline int next_root_backup(struct btrfs_fs_info *info,
1947 struct btrfs_super_block *super, 1947 struct btrfs_super_block *super,
1948 int *num_backups_tried, int *backup_index) 1948 int *num_backups_tried, int *backup_index)
1949 { 1949 {
1950 struct btrfs_root_backup *root_backup; 1950 struct btrfs_root_backup *root_backup;
1951 int newest = *backup_index; 1951 int newest = *backup_index;
1952 1952
1953 if (*num_backups_tried == 0) { 1953 if (*num_backups_tried == 0) {
1954 u64 gen = btrfs_super_generation(super); 1954 u64 gen = btrfs_super_generation(super);
1955 1955
1956 newest = find_newest_super_backup(info, gen); 1956 newest = find_newest_super_backup(info, gen);
1957 if (newest == -1) 1957 if (newest == -1)
1958 return -1; 1958 return -1;
1959 1959
1960 *backup_index = newest; 1960 *backup_index = newest;
1961 *num_backups_tried = 1; 1961 *num_backups_tried = 1;
1962 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 1962 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1963 /* we've tried all the backups, all done */ 1963 /* we've tried all the backups, all done */
1964 return -1; 1964 return -1;
1965 } else { 1965 } else {
1966 /* jump to the next oldest backup */ 1966 /* jump to the next oldest backup */
1967 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 1967 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1968 BTRFS_NUM_BACKUP_ROOTS; 1968 BTRFS_NUM_BACKUP_ROOTS;
1969 *backup_index = newest; 1969 *backup_index = newest;
1970 *num_backups_tried += 1; 1970 *num_backups_tried += 1;
1971 } 1971 }
1972 root_backup = super->super_roots + newest; 1972 root_backup = super->super_roots + newest;
1973 1973
1974 btrfs_set_super_generation(super, 1974 btrfs_set_super_generation(super,
1975 btrfs_backup_tree_root_gen(root_backup)); 1975 btrfs_backup_tree_root_gen(root_backup));
1976 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1976 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1977 btrfs_set_super_root_level(super, 1977 btrfs_set_super_root_level(super,
1978 btrfs_backup_tree_root_level(root_backup)); 1978 btrfs_backup_tree_root_level(root_backup));
1979 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1979 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1980 1980
1981 /* 1981 /*
1982 * fixme: the total bytes and num_devices need to match or we should 1982 * fixme: the total bytes and num_devices need to match or we should
1983 * need a fsck 1983 * need a fsck
1984 */ 1984 */
1985 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1985 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1986 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1986 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1987 return 0; 1987 return 0;
1988 } 1988 }
1989 1989
1990 /* helper to cleanup workers */ 1990 /* helper to cleanup workers */
1991 static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 1991 static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
1992 { 1992 {
1993 btrfs_stop_workers(&fs_info->generic_worker); 1993 btrfs_stop_workers(&fs_info->generic_worker);
1994 btrfs_stop_workers(&fs_info->fixup_workers); 1994 btrfs_stop_workers(&fs_info->fixup_workers);
1995 btrfs_stop_workers(&fs_info->delalloc_workers); 1995 btrfs_stop_workers(&fs_info->delalloc_workers);
1996 btrfs_stop_workers(&fs_info->workers); 1996 btrfs_stop_workers(&fs_info->workers);
1997 btrfs_stop_workers(&fs_info->endio_workers); 1997 btrfs_stop_workers(&fs_info->endio_workers);
1998 btrfs_stop_workers(&fs_info->endio_meta_workers); 1998 btrfs_stop_workers(&fs_info->endio_meta_workers);
1999 btrfs_stop_workers(&fs_info->endio_raid56_workers); 1999 btrfs_stop_workers(&fs_info->endio_raid56_workers);
2000 btrfs_stop_workers(&fs_info->rmw_workers); 2000 btrfs_stop_workers(&fs_info->rmw_workers);
2001 btrfs_stop_workers(&fs_info->endio_meta_write_workers); 2001 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2002 btrfs_stop_workers(&fs_info->endio_write_workers); 2002 btrfs_stop_workers(&fs_info->endio_write_workers);
2003 btrfs_stop_workers(&fs_info->endio_freespace_worker); 2003 btrfs_stop_workers(&fs_info->endio_freespace_worker);
2004 btrfs_stop_workers(&fs_info->submit_workers); 2004 btrfs_stop_workers(&fs_info->submit_workers);
2005 btrfs_stop_workers(&fs_info->delayed_workers); 2005 btrfs_stop_workers(&fs_info->delayed_workers);
2006 btrfs_stop_workers(&fs_info->caching_workers); 2006 btrfs_stop_workers(&fs_info->caching_workers);
2007 btrfs_stop_workers(&fs_info->readahead_workers); 2007 btrfs_stop_workers(&fs_info->readahead_workers);
2008 btrfs_stop_workers(&fs_info->flush_workers); 2008 btrfs_stop_workers(&fs_info->flush_workers);
2009 btrfs_stop_workers(&fs_info->qgroup_rescan_workers); 2009 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
2010 } 2010 }
2011 2011
2012 static void free_root_extent_buffers(struct btrfs_root *root) 2012 static void free_root_extent_buffers(struct btrfs_root *root)
2013 { 2013 {
2014 if (root) { 2014 if (root) {
2015 free_extent_buffer(root->node); 2015 free_extent_buffer(root->node);
2016 free_extent_buffer(root->commit_root); 2016 free_extent_buffer(root->commit_root);
2017 root->node = NULL; 2017 root->node = NULL;
2018 root->commit_root = NULL; 2018 root->commit_root = NULL;
2019 } 2019 }
2020 } 2020 }
2021 2021
2022 /* helper to cleanup tree roots */ 2022 /* helper to cleanup tree roots */
2023 static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2023 static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2024 { 2024 {
2025 free_root_extent_buffers(info->tree_root); 2025 free_root_extent_buffers(info->tree_root);
2026 2026
2027 free_root_extent_buffers(info->dev_root); 2027 free_root_extent_buffers(info->dev_root);
2028 free_root_extent_buffers(info->extent_root); 2028 free_root_extent_buffers(info->extent_root);
2029 free_root_extent_buffers(info->csum_root); 2029 free_root_extent_buffers(info->csum_root);
2030 free_root_extent_buffers(info->quota_root); 2030 free_root_extent_buffers(info->quota_root);
2031 free_root_extent_buffers(info->uuid_root); 2031 free_root_extent_buffers(info->uuid_root);
2032 if (chunk_root) 2032 if (chunk_root)
2033 free_root_extent_buffers(info->chunk_root); 2033 free_root_extent_buffers(info->chunk_root);
2034 } 2034 }
2035 2035
2036 static void del_fs_roots(struct btrfs_fs_info *fs_info) 2036 static void del_fs_roots(struct btrfs_fs_info *fs_info)
2037 { 2037 {
2038 int ret; 2038 int ret;
2039 struct btrfs_root *gang[8]; 2039 struct btrfs_root *gang[8];
2040 int i; 2040 int i;
2041 2041
2042 while (!list_empty(&fs_info->dead_roots)) { 2042 while (!list_empty(&fs_info->dead_roots)) {
2043 gang[0] = list_entry(fs_info->dead_roots.next, 2043 gang[0] = list_entry(fs_info->dead_roots.next,
2044 struct btrfs_root, root_list); 2044 struct btrfs_root, root_list);
2045 list_del(&gang[0]->root_list); 2045 list_del(&gang[0]->root_list);
2046 2046
2047 if (gang[0]->in_radix) { 2047 if (gang[0]->in_radix) {
2048 btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2048 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
2049 } else { 2049 } else {
2050 free_extent_buffer(gang[0]->node); 2050 free_extent_buffer(gang[0]->node);
2051 free_extent_buffer(gang[0]->commit_root); 2051 free_extent_buffer(gang[0]->commit_root);
2052 btrfs_put_fs_root(gang[0]); 2052 btrfs_put_fs_root(gang[0]);
2053 } 2053 }
2054 } 2054 }
2055 2055
2056 while (1) { 2056 while (1) {
2057 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2057 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2058 (void **)gang, 0, 2058 (void **)gang, 0,
2059 ARRAY_SIZE(gang)); 2059 ARRAY_SIZE(gang));
2060 if (!ret) 2060 if (!ret)
2061 break; 2061 break;
2062 for (i = 0; i < ret; i++) 2062 for (i = 0; i < ret; i++)
2063 btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2063 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
2064 } 2064 }
2065 2065
2066 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 2066 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2067 btrfs_free_log_root_tree(NULL, fs_info); 2067 btrfs_free_log_root_tree(NULL, fs_info);
2068 btrfs_destroy_pinned_extent(fs_info->tree_root, 2068 btrfs_destroy_pinned_extent(fs_info->tree_root,
2069 fs_info->pinned_extents); 2069 fs_info->pinned_extents);
2070 } 2070 }
2071 } 2071 }
2072 2072
2073 int open_ctree(struct super_block *sb, 2073 int open_ctree(struct super_block *sb,
2074 struct btrfs_fs_devices *fs_devices, 2074 struct btrfs_fs_devices *fs_devices,
2075 char *options) 2075 char *options)
2076 { 2076 {
2077 u32 sectorsize; 2077 u32 sectorsize;
2078 u32 nodesize; 2078 u32 nodesize;
2079 u32 leafsize; 2079 u32 leafsize;
2080 u32 blocksize; 2080 u32 blocksize;
2081 u32 stripesize; 2081 u32 stripesize;
2082 u64 generation; 2082 u64 generation;
2083 u64 features; 2083 u64 features;
2084 struct btrfs_key location; 2084 struct btrfs_key location;
2085 struct buffer_head *bh; 2085 struct buffer_head *bh;
2086 struct btrfs_super_block *disk_super; 2086 struct btrfs_super_block *disk_super;
2087 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2087 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2088 struct btrfs_root *tree_root; 2088 struct btrfs_root *tree_root;
2089 struct btrfs_root *extent_root; 2089 struct btrfs_root *extent_root;
2090 struct btrfs_root *csum_root; 2090 struct btrfs_root *csum_root;
2091 struct btrfs_root *chunk_root; 2091 struct btrfs_root *chunk_root;
2092 struct btrfs_root *dev_root; 2092 struct btrfs_root *dev_root;
2093 struct btrfs_root *quota_root; 2093 struct btrfs_root *quota_root;
2094 struct btrfs_root *uuid_root; 2094 struct btrfs_root *uuid_root;
2095 struct btrfs_root *log_tree_root; 2095 struct btrfs_root *log_tree_root;
2096 int ret; 2096 int ret;
2097 int err = -EINVAL; 2097 int err = -EINVAL;
2098 int num_backups_tried = 0; 2098 int num_backups_tried = 0;
2099 int backup_index = 0; 2099 int backup_index = 0;
2100 bool create_uuid_tree; 2100 bool create_uuid_tree;
2101 bool check_uuid_tree; 2101 bool check_uuid_tree;
2102 2102
2103 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); 2103 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
2104 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); 2104 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
2105 if (!tree_root || !chunk_root) { 2105 if (!tree_root || !chunk_root) {
2106 err = -ENOMEM; 2106 err = -ENOMEM;
2107 goto fail; 2107 goto fail;
2108 } 2108 }
2109 2109
2110 ret = init_srcu_struct(&fs_info->subvol_srcu); 2110 ret = init_srcu_struct(&fs_info->subvol_srcu);
2111 if (ret) { 2111 if (ret) {
2112 err = ret; 2112 err = ret;
2113 goto fail; 2113 goto fail;
2114 } 2114 }
2115 2115
2116 ret = setup_bdi(fs_info, &fs_info->bdi); 2116 ret = setup_bdi(fs_info, &fs_info->bdi);
2117 if (ret) { 2117 if (ret) {
2118 err = ret; 2118 err = ret;
2119 goto fail_srcu; 2119 goto fail_srcu;
2120 } 2120 }
2121 2121
2122 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); 2122 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2123 if (ret) { 2123 if (ret) {
2124 err = ret; 2124 err = ret;
2125 goto fail_bdi; 2125 goto fail_bdi;
2126 } 2126 }
2127 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * 2127 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2128 (1 + ilog2(nr_cpu_ids)); 2128 (1 + ilog2(nr_cpu_ids));
2129 2129
2130 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); 2130 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2131 if (ret) { 2131 if (ret) {
2132 err = ret; 2132 err = ret;
2133 goto fail_dirty_metadata_bytes; 2133 goto fail_dirty_metadata_bytes;
2134 } 2134 }
2135 2135
2136 fs_info->btree_inode = new_inode(sb); 2136 fs_info->btree_inode = new_inode(sb);
2137 if (!fs_info->btree_inode) { 2137 if (!fs_info->btree_inode) {
2138 err = -ENOMEM; 2138 err = -ENOMEM;
2139 goto fail_delalloc_bytes; 2139 goto fail_delalloc_bytes;
2140 } 2140 }
2141 2141
2142 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 2142 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
2143 2143
2144 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2144 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
2145 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 2145 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
2146 INIT_LIST_HEAD(&fs_info->trans_list); 2146 INIT_LIST_HEAD(&fs_info->trans_list);
2147 INIT_LIST_HEAD(&fs_info->dead_roots); 2147 INIT_LIST_HEAD(&fs_info->dead_roots);
2148 INIT_LIST_HEAD(&fs_info->delayed_iputs); 2148 INIT_LIST_HEAD(&fs_info->delayed_iputs);
2149 INIT_LIST_HEAD(&fs_info->delalloc_roots); 2149 INIT_LIST_HEAD(&fs_info->delalloc_roots);
2150 INIT_LIST_HEAD(&fs_info->caching_block_groups); 2150 INIT_LIST_HEAD(&fs_info->caching_block_groups);
2151 spin_lock_init(&fs_info->delalloc_root_lock); 2151 spin_lock_init(&fs_info->delalloc_root_lock);
2152 spin_lock_init(&fs_info->trans_lock); 2152 spin_lock_init(&fs_info->trans_lock);
2153 spin_lock_init(&fs_info->fs_roots_radix_lock); 2153 spin_lock_init(&fs_info->fs_roots_radix_lock);
2154 spin_lock_init(&fs_info->delayed_iput_lock); 2154 spin_lock_init(&fs_info->delayed_iput_lock);
2155 spin_lock_init(&fs_info->defrag_inodes_lock); 2155 spin_lock_init(&fs_info->defrag_inodes_lock);
2156 spin_lock_init(&fs_info->free_chunk_lock); 2156 spin_lock_init(&fs_info->free_chunk_lock);
2157 spin_lock_init(&fs_info->tree_mod_seq_lock); 2157 spin_lock_init(&fs_info->tree_mod_seq_lock);
2158 spin_lock_init(&fs_info->super_lock); 2158 spin_lock_init(&fs_info->super_lock);
2159 spin_lock_init(&fs_info->buffer_lock); 2159 spin_lock_init(&fs_info->buffer_lock);
2160 rwlock_init(&fs_info->tree_mod_log_lock); 2160 rwlock_init(&fs_info->tree_mod_log_lock);
2161 mutex_init(&fs_info->reloc_mutex); 2161 mutex_init(&fs_info->reloc_mutex);
2162 seqlock_init(&fs_info->profiles_lock); 2162 seqlock_init(&fs_info->profiles_lock);
2163 2163
2164 init_completion(&fs_info->kobj_unregister); 2164 init_completion(&fs_info->kobj_unregister);
2165 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 2165 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
2166 INIT_LIST_HEAD(&fs_info->space_info); 2166 INIT_LIST_HEAD(&fs_info->space_info);
2167 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 2167 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
2168 btrfs_mapping_init(&fs_info->mapping_tree); 2168 btrfs_mapping_init(&fs_info->mapping_tree);
2169 btrfs_init_block_rsv(&fs_info->global_block_rsv, 2169 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2170 BTRFS_BLOCK_RSV_GLOBAL); 2170 BTRFS_BLOCK_RSV_GLOBAL);
2171 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 2171 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2172 BTRFS_BLOCK_RSV_DELALLOC); 2172 BTRFS_BLOCK_RSV_DELALLOC);
2173 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 2173 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2174 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 2174 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2175 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 2175 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2176 btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 2176 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2177 BTRFS_BLOCK_RSV_DELOPS); 2177 BTRFS_BLOCK_RSV_DELOPS);
2178 atomic_set(&fs_info->nr_async_submits, 0); 2178 atomic_set(&fs_info->nr_async_submits, 0);
2179 atomic_set(&fs_info->async_delalloc_pages, 0); 2179 atomic_set(&fs_info->async_delalloc_pages, 0);
2180 atomic_set(&fs_info->async_submit_draining, 0); 2180 atomic_set(&fs_info->async_submit_draining, 0);
2181 atomic_set(&fs_info->nr_async_bios, 0); 2181 atomic_set(&fs_info->nr_async_bios, 0);
2182 atomic_set(&fs_info->defrag_running, 0); 2182 atomic_set(&fs_info->defrag_running, 0);
2183 atomic64_set(&fs_info->tree_mod_seq, 0); 2183 atomic64_set(&fs_info->tree_mod_seq, 0);
2184 fs_info->sb = sb; 2184 fs_info->sb = sb;
2185 fs_info->max_inline = 8192 * 1024; 2185 fs_info->max_inline = 8192 * 1024;
2186 fs_info->metadata_ratio = 0; 2186 fs_info->metadata_ratio = 0;
2187 fs_info->defrag_inodes = RB_ROOT; 2187 fs_info->defrag_inodes = RB_ROOT;
2188 fs_info->free_chunk_space = 0; 2188 fs_info->free_chunk_space = 0;
2189 fs_info->tree_mod_log = RB_ROOT; 2189 fs_info->tree_mod_log = RB_ROOT;
2190 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2190 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2191 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64); 2191 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
2192 /* readahead state */ 2192 /* readahead state */
2193 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); 2193 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2194 spin_lock_init(&fs_info->reada_lock); 2194 spin_lock_init(&fs_info->reada_lock);
2195 2195
2196 fs_info->thread_pool_size = min_t(unsigned long, 2196 fs_info->thread_pool_size = min_t(unsigned long,
2197 num_online_cpus() + 2, 8); 2197 num_online_cpus() + 2, 8);
2198 2198
2199 INIT_LIST_HEAD(&fs_info->ordered_roots); 2199 INIT_LIST_HEAD(&fs_info->ordered_roots);
2200 spin_lock_init(&fs_info->ordered_root_lock); 2200 spin_lock_init(&fs_info->ordered_root_lock);
2201 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2201 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2202 GFP_NOFS); 2202 GFP_NOFS);
2203 if (!fs_info->delayed_root) { 2203 if (!fs_info->delayed_root) {
2204 err = -ENOMEM; 2204 err = -ENOMEM;
2205 goto fail_iput; 2205 goto fail_iput;
2206 } 2206 }
2207 btrfs_init_delayed_root(fs_info->delayed_root); 2207 btrfs_init_delayed_root(fs_info->delayed_root);
2208 2208
2209 mutex_init(&fs_info->scrub_lock); 2209 mutex_init(&fs_info->scrub_lock);
2210 atomic_set(&fs_info->scrubs_running, 0); 2210 atomic_set(&fs_info->scrubs_running, 0);
2211 atomic_set(&fs_info->scrub_pause_req, 0); 2211 atomic_set(&fs_info->scrub_pause_req, 0);
2212 atomic_set(&fs_info->scrubs_paused, 0); 2212 atomic_set(&fs_info->scrubs_paused, 0);
2213 atomic_set(&fs_info->scrub_cancel_req, 0); 2213 atomic_set(&fs_info->scrub_cancel_req, 0);
2214 init_waitqueue_head(&fs_info->scrub_pause_wait); 2214 init_waitqueue_head(&fs_info->scrub_pause_wait);
2215 fs_info->scrub_workers_refcnt = 0; 2215 fs_info->scrub_workers_refcnt = 0;
2216 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 2216 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2217 fs_info->check_integrity_print_mask = 0; 2217 fs_info->check_integrity_print_mask = 0;
2218 #endif 2218 #endif
2219 2219
2220 spin_lock_init(&fs_info->balance_lock); 2220 spin_lock_init(&fs_info->balance_lock);
2221 mutex_init(&fs_info->balance_mutex); 2221 mutex_init(&fs_info->balance_mutex);
2222 atomic_set(&fs_info->balance_running, 0); 2222 atomic_set(&fs_info->balance_running, 0);
2223 atomic_set(&fs_info->balance_pause_req, 0); 2223 atomic_set(&fs_info->balance_pause_req, 0);
2224 atomic_set(&fs_info->balance_cancel_req, 0); 2224 atomic_set(&fs_info->balance_cancel_req, 0);
2225 fs_info->balance_ctl = NULL; 2225 fs_info->balance_ctl = NULL;
2226 init_waitqueue_head(&fs_info->balance_wait_q); 2226 init_waitqueue_head(&fs_info->balance_wait_q);
2227 2227
2228 sb->s_blocksize = 4096; 2228 sb->s_blocksize = 4096;
2229 sb->s_blocksize_bits = blksize_bits(4096); 2229 sb->s_blocksize_bits = blksize_bits(4096);
2230 sb->s_bdi = &fs_info->bdi; 2230 sb->s_bdi = &fs_info->bdi;
2231 2231
2232 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2232 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2233 set_nlink(fs_info->btree_inode, 1); 2233 set_nlink(fs_info->btree_inode, 1);
2234 /* 2234 /*
2235 * we set the i_size on the btree inode to the max possible int. 2235 * we set the i_size on the btree inode to the max possible int.
2236 * the real end of the address space is determined by all of 2236 * the real end of the address space is determined by all of
2237 * the devices in the system 2237 * the devices in the system
2238 */ 2238 */
2239 fs_info->btree_inode->i_size = OFFSET_MAX; 2239 fs_info->btree_inode->i_size = OFFSET_MAX;
2240 fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 2240 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
2241 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; 2241 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2242 2242
2243 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2243 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2244 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2244 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2245 fs_info->btree_inode->i_mapping); 2245 fs_info->btree_inode->i_mapping);
2246 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2246 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2247 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 2247 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2248 2248
2249 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2249 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2250 2250
2251 BTRFS_I(fs_info->btree_inode)->root = tree_root; 2251 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2252 memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 2252 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2253 sizeof(struct btrfs_key)); 2253 sizeof(struct btrfs_key));
2254 set_bit(BTRFS_INODE_DUMMY, 2254 set_bit(BTRFS_INODE_DUMMY,
2255 &BTRFS_I(fs_info->btree_inode)->runtime_flags); 2255 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2256 btrfs_insert_inode_hash(fs_info->btree_inode); 2256 btrfs_insert_inode_hash(fs_info->btree_inode);
2257 2257
2258 spin_lock_init(&fs_info->block_group_cache_lock); 2258 spin_lock_init(&fs_info->block_group_cache_lock);
2259 fs_info->block_group_cache_tree = RB_ROOT; 2259 fs_info->block_group_cache_tree = RB_ROOT;
2260 fs_info->first_logical_byte = (u64)-1; 2260 fs_info->first_logical_byte = (u64)-1;
2261 2261
2262 extent_io_tree_init(&fs_info->freed_extents[0], 2262 extent_io_tree_init(&fs_info->freed_extents[0],
2263 fs_info->btree_inode->i_mapping); 2263 fs_info->btree_inode->i_mapping);
2264 extent_io_tree_init(&fs_info->freed_extents[1], 2264 extent_io_tree_init(&fs_info->freed_extents[1],
2265 fs_info->btree_inode->i_mapping); 2265 fs_info->btree_inode->i_mapping);
2266 fs_info->pinned_extents = &fs_info->freed_extents[0]; 2266 fs_info->pinned_extents = &fs_info->freed_extents[0];
2267 fs_info->do_barriers = 1; 2267 fs_info->do_barriers = 1;
2268 2268
2269 2269
2270 mutex_init(&fs_info->ordered_operations_mutex); 2270 mutex_init(&fs_info->ordered_operations_mutex);
2271 mutex_init(&fs_info->ordered_extent_flush_mutex); 2271 mutex_init(&fs_info->ordered_extent_flush_mutex);
2272 mutex_init(&fs_info->tree_log_mutex); 2272 mutex_init(&fs_info->tree_log_mutex);
2273 mutex_init(&fs_info->chunk_mutex); 2273 mutex_init(&fs_info->chunk_mutex);
2274 mutex_init(&fs_info->transaction_kthread_mutex); 2274 mutex_init(&fs_info->transaction_kthread_mutex);
2275 mutex_init(&fs_info->cleaner_mutex); 2275 mutex_init(&fs_info->cleaner_mutex);
2276 mutex_init(&fs_info->volume_mutex); 2276 mutex_init(&fs_info->volume_mutex);
2277 init_rwsem(&fs_info->extent_commit_sem); 2277 init_rwsem(&fs_info->extent_commit_sem);
2278 init_rwsem(&fs_info->cleanup_work_sem); 2278 init_rwsem(&fs_info->cleanup_work_sem);
2279 init_rwsem(&fs_info->subvol_sem); 2279 init_rwsem(&fs_info->subvol_sem);
2280 sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2280 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
2281 fs_info->dev_replace.lock_owner = 0; 2281 fs_info->dev_replace.lock_owner = 0;
2282 atomic_set(&fs_info->dev_replace.nesting_level, 0); 2282 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2283 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2283 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2284 mutex_init(&fs_info->dev_replace.lock_management_lock); 2284 mutex_init(&fs_info->dev_replace.lock_management_lock);
2285 mutex_init(&fs_info->dev_replace.lock); 2285 mutex_init(&fs_info->dev_replace.lock);
2286 2286
2287 spin_lock_init(&fs_info->qgroup_lock); 2287 spin_lock_init(&fs_info->qgroup_lock);
2288 mutex_init(&fs_info->qgroup_ioctl_lock); 2288 mutex_init(&fs_info->qgroup_ioctl_lock);
2289 fs_info->qgroup_tree = RB_ROOT; 2289 fs_info->qgroup_tree = RB_ROOT;
2290 INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2290 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2291 fs_info->qgroup_seq = 1; 2291 fs_info->qgroup_seq = 1;
2292 fs_info->quota_enabled = 0; 2292 fs_info->quota_enabled = 0;
2293 fs_info->pending_quota_state = 0; 2293 fs_info->pending_quota_state = 0;
2294 fs_info->qgroup_ulist = NULL; 2294 fs_info->qgroup_ulist = NULL;
2295 mutex_init(&fs_info->qgroup_rescan_lock); 2295 mutex_init(&fs_info->qgroup_rescan_lock);
2296 2296
2297 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2297 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2298 btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2298 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2299 2299
2300 init_waitqueue_head(&fs_info->transaction_throttle); 2300 init_waitqueue_head(&fs_info->transaction_throttle);
2301 init_waitqueue_head(&fs_info->transaction_wait); 2301 init_waitqueue_head(&fs_info->transaction_wait);
2302 init_waitqueue_head(&fs_info->transaction_blocked_wait); 2302 init_waitqueue_head(&fs_info->transaction_blocked_wait);
2303 init_waitqueue_head(&fs_info->async_submit_wait); 2303 init_waitqueue_head(&fs_info->async_submit_wait);
2304 2304
2305 ret = btrfs_alloc_stripe_hash_table(fs_info); 2305 ret = btrfs_alloc_stripe_hash_table(fs_info);
2306 if (ret) { 2306 if (ret) {
2307 err = ret; 2307 err = ret;
2308 goto fail_alloc; 2308 goto fail_alloc;
2309 } 2309 }
2310 2310
2311 __setup_root(4096, 4096, 4096, 4096, tree_root, 2311 __setup_root(4096, 4096, 4096, 4096, tree_root,
2312 fs_info, BTRFS_ROOT_TREE_OBJECTID); 2312 fs_info, BTRFS_ROOT_TREE_OBJECTID);
2313 2313
2314 invalidate_bdev(fs_devices->latest_bdev); 2314 invalidate_bdev(fs_devices->latest_bdev);
2315 2315
2316 /* 2316 /*
2317 * Read super block and check the signature bytes only 2317 * Read super block and check the signature bytes only
2318 */ 2318 */
2319 bh = btrfs_read_dev_super(fs_devices->latest_bdev); 2319 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
2320 if (!bh) { 2320 if (!bh) {
2321 err = -EINVAL; 2321 err = -EINVAL;
2322 goto fail_alloc; 2322 goto fail_alloc;
2323 } 2323 }
2324 2324
2325 /* 2325 /*
2326 * We want to check superblock checksum, the type is stored inside. 2326 * We want to check superblock checksum, the type is stored inside.
2327 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 2327 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2328 */ 2328 */
2329 if (btrfs_check_super_csum(bh->b_data)) { 2329 if (btrfs_check_super_csum(bh->b_data)) {
2330 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); 2330 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
2331 err = -EINVAL; 2331 err = -EINVAL;
2332 goto fail_alloc; 2332 goto fail_alloc;
2333 } 2333 }
2334 2334
2335 /* 2335 /*
2336 * super_copy is zeroed at allocation time and we never touch the 2336 * super_copy is zeroed at allocation time and we never touch the
2337 * following bytes up to INFO_SIZE, the checksum is calculated from 2337 * following bytes up to INFO_SIZE, the checksum is calculated from
2338 * the whole block of INFO_SIZE 2338 * the whole block of INFO_SIZE
2339 */ 2339 */
2340 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 2340 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2341 memcpy(fs_info->super_for_commit, fs_info->super_copy, 2341 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2342 sizeof(*fs_info->super_for_commit)); 2342 sizeof(*fs_info->super_for_commit));
2343 brelse(bh); 2343 brelse(bh);
2344 2344
2345 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 2345 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
2346 2346
2347 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 2347 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2348 if (ret) { 2348 if (ret) {
2349 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); 2349 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
2350 err = -EINVAL; 2350 err = -EINVAL;
2351 goto fail_alloc; 2351 goto fail_alloc;
2352 } 2352 }
2353 2353
2354 disk_super = fs_info->super_copy; 2354 disk_super = fs_info->super_copy;
2355 if (!btrfs_super_root(disk_super)) 2355 if (!btrfs_super_root(disk_super))
2356 goto fail_alloc; 2356 goto fail_alloc;
2357 2357
2358 /* check FS state, whether FS is broken. */ 2358 /* check FS state, whether FS is broken. */
2359 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 2359 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2360 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2360 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
2361 2361
2362 /* 2362 /*
2363 * run through our array of backup supers and setup 2363 * run through our array of backup supers and setup
2364 * our ring pointer to the oldest one 2364 * our ring pointer to the oldest one
2365 */ 2365 */
2366 generation = btrfs_super_generation(disk_super); 2366 generation = btrfs_super_generation(disk_super);
2367 find_oldest_super_backup(fs_info, generation); 2367 find_oldest_super_backup(fs_info, generation);
2368 2368
2369 /* 2369 /*
2370 * In the long term, we'll store the compression type in the super 2370 * In the long term, we'll store the compression type in the super
2371 * block, and it'll be used for per file compression control. 2371 * block, and it'll be used for per file compression control.
2372 */ 2372 */
2373 fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 2373 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2374 2374
2375 ret = btrfs_parse_options(tree_root, options); 2375 ret = btrfs_parse_options(tree_root, options);
2376 if (ret) { 2376 if (ret) {
2377 err = ret; 2377 err = ret;
2378 goto fail_alloc; 2378 goto fail_alloc;
2379 } 2379 }
2380 2380
2381 features = btrfs_super_incompat_flags(disk_super) & 2381 features = btrfs_super_incompat_flags(disk_super) &
2382 ~BTRFS_FEATURE_INCOMPAT_SUPP; 2382 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2383 if (features) { 2383 if (features) {
2384 printk(KERN_ERR "BTRFS: couldn't mount because of " 2384 printk(KERN_ERR "BTRFS: couldn't mount because of "
2385 "unsupported optional features (%Lx).\n", 2385 "unsupported optional features (%Lx).\n",
2386 features); 2386 features);
2387 err = -EINVAL; 2387 err = -EINVAL;
2388 goto fail_alloc; 2388 goto fail_alloc;
2389 } 2389 }
2390 2390
2391 if (btrfs_super_leafsize(disk_super) != 2391 if (btrfs_super_leafsize(disk_super) !=
2392 btrfs_super_nodesize(disk_super)) { 2392 btrfs_super_nodesize(disk_super)) {
2393 printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2393 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2394 "blocksizes don't match. node %d leaf %d\n", 2394 "blocksizes don't match. node %d leaf %d\n",
2395 btrfs_super_nodesize(disk_super), 2395 btrfs_super_nodesize(disk_super),
2396 btrfs_super_leafsize(disk_super)); 2396 btrfs_super_leafsize(disk_super));
2397 err = -EINVAL; 2397 err = -EINVAL;
2398 goto fail_alloc; 2398 goto fail_alloc;
2399 } 2399 }
2400 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { 2400 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2401 printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2401 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2402 "blocksize (%d) was too large\n", 2402 "blocksize (%d) was too large\n",
2403 btrfs_super_leafsize(disk_super)); 2403 btrfs_super_leafsize(disk_super));
2404 err = -EINVAL; 2404 err = -EINVAL;
2405 goto fail_alloc; 2405 goto fail_alloc;
2406 } 2406 }
2407 2407
2408 features = btrfs_super_incompat_flags(disk_super); 2408 features = btrfs_super_incompat_flags(disk_super);
2409 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 2409 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
2410 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2410 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
2411 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2411 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
2412 2412
2413 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 2413 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
2414 printk(KERN_ERR "BTRFS: has skinny extents\n"); 2414 printk(KERN_INFO "BTRFS: has skinny extents\n");
2415 2415
2416 /* 2416 /*
2417 * flag our filesystem as having big metadata blocks if 2417 * flag our filesystem as having big metadata blocks if
2418 * they are bigger than the page size 2418 * they are bigger than the page size
2419 */ 2419 */
2420 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { 2420 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2421 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2421 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2422 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); 2422 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
2423 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2423 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2424 } 2424 }
2425 2425
2426 nodesize = btrfs_super_nodesize(disk_super); 2426 nodesize = btrfs_super_nodesize(disk_super);
2427 leafsize = btrfs_super_leafsize(disk_super); 2427 leafsize = btrfs_super_leafsize(disk_super);
2428 sectorsize = btrfs_super_sectorsize(disk_super); 2428 sectorsize = btrfs_super_sectorsize(disk_super);
2429 stripesize = btrfs_super_stripesize(disk_super); 2429 stripesize = btrfs_super_stripesize(disk_super);
2430 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); 2430 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
2431 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2431 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
2432 2432
2433 /* 2433 /*
2434 * mixed block groups end up with duplicate but slightly offset 2434 * mixed block groups end up with duplicate but slightly offset
2435 * extent buffers for the same range. It leads to corruptions 2435 * extent buffers for the same range. It leads to corruptions
2436 */ 2436 */
2437 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2437 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2438 (sectorsize != leafsize)) { 2438 (sectorsize != leafsize)) {
2439 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes " 2439 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
2440 "are not allowed for mixed block groups on %s\n", 2440 "are not allowed for mixed block groups on %s\n",
2441 sb->s_id); 2441 sb->s_id);
2442 goto fail_alloc; 2442 goto fail_alloc;
2443 } 2443 }
2444 2444
2445 /* 2445 /*
2446 * Needn't use the lock because there is no other task which will 2446 * Needn't use the lock because there is no other task which will
2447 * update the flag. 2447 * update the flag.
2448 */ 2448 */
2449 btrfs_set_super_incompat_flags(disk_super, features); 2449 btrfs_set_super_incompat_flags(disk_super, features);
2450 2450
2451 features = btrfs_super_compat_ro_flags(disk_super) & 2451 features = btrfs_super_compat_ro_flags(disk_super) &
2452 ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2452 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2453 if (!(sb->s_flags & MS_RDONLY) && features) { 2453 if (!(sb->s_flags & MS_RDONLY) && features) {
2454 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2454 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2455 "unsupported option features (%Lx).\n", 2455 "unsupported option features (%Lx).\n",
2456 features); 2456 features);
2457 err = -EINVAL; 2457 err = -EINVAL;
2458 goto fail_alloc; 2458 goto fail_alloc;
2459 } 2459 }
2460 2460
2461 btrfs_init_workers(&fs_info->generic_worker, 2461 btrfs_init_workers(&fs_info->generic_worker,
2462 "genwork", 1, NULL); 2462 "genwork", 1, NULL);
2463 2463
2464 btrfs_init_workers(&fs_info->workers, "worker", 2464 btrfs_init_workers(&fs_info->workers, "worker",
2465 fs_info->thread_pool_size, 2465 fs_info->thread_pool_size,
2466 &fs_info->generic_worker); 2466 &fs_info->generic_worker);
2467 2467
2468 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", 2468 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
2469 fs_info->thread_pool_size, NULL); 2469 fs_info->thread_pool_size, NULL);
2470 2470
2471 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc", 2471 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
2472 fs_info->thread_pool_size, NULL); 2472 fs_info->thread_pool_size, NULL);
2473 2473
2474 btrfs_init_workers(&fs_info->submit_workers, "submit", 2474 btrfs_init_workers(&fs_info->submit_workers, "submit",
2475 min_t(u64, fs_devices->num_devices, 2475 min_t(u64, fs_devices->num_devices,
2476 fs_info->thread_pool_size), NULL); 2476 fs_info->thread_pool_size), NULL);
2477 2477
2478 btrfs_init_workers(&fs_info->caching_workers, "cache", 2478 btrfs_init_workers(&fs_info->caching_workers, "cache",
2479 fs_info->thread_pool_size, NULL); 2479 fs_info->thread_pool_size, NULL);
2480 2480
2481 /* a higher idle thresh on the submit workers makes it much more 2481 /* a higher idle thresh on the submit workers makes it much more
2482 * likely that bios will be send down in a sane order to the 2482 * likely that bios will be send down in a sane order to the
2483 * devices 2483 * devices
2484 */ 2484 */
2485 fs_info->submit_workers.idle_thresh = 64; 2485 fs_info->submit_workers.idle_thresh = 64;
2486 2486
2487 fs_info->workers.idle_thresh = 16; 2487 fs_info->workers.idle_thresh = 16;
2488 fs_info->workers.ordered = 1; 2488 fs_info->workers.ordered = 1;
2489 2489
2490 fs_info->delalloc_workers.idle_thresh = 2; 2490 fs_info->delalloc_workers.idle_thresh = 2;
2491 fs_info->delalloc_workers.ordered = 1; 2491 fs_info->delalloc_workers.ordered = 1;
2492 2492
2493 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1, 2493 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2494 &fs_info->generic_worker); 2494 &fs_info->generic_worker);
2495 btrfs_init_workers(&fs_info->endio_workers, "endio", 2495 btrfs_init_workers(&fs_info->endio_workers, "endio",
2496 fs_info->thread_pool_size, 2496 fs_info->thread_pool_size,
2497 &fs_info->generic_worker); 2497 &fs_info->generic_worker);
2498 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", 2498 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
2499 fs_info->thread_pool_size, 2499 fs_info->thread_pool_size,
2500 &fs_info->generic_worker); 2500 &fs_info->generic_worker);
2501 btrfs_init_workers(&fs_info->endio_meta_write_workers, 2501 btrfs_init_workers(&fs_info->endio_meta_write_workers,
2502 "endio-meta-write", fs_info->thread_pool_size, 2502 "endio-meta-write", fs_info->thread_pool_size,
2503 &fs_info->generic_worker); 2503 &fs_info->generic_worker);
2504 btrfs_init_workers(&fs_info->endio_raid56_workers, 2504 btrfs_init_workers(&fs_info->endio_raid56_workers,
2505 "endio-raid56", fs_info->thread_pool_size, 2505 "endio-raid56", fs_info->thread_pool_size,
2506 &fs_info->generic_worker); 2506 &fs_info->generic_worker);
2507 btrfs_init_workers(&fs_info->rmw_workers, 2507 btrfs_init_workers(&fs_info->rmw_workers,
2508 "rmw", fs_info->thread_pool_size, 2508 "rmw", fs_info->thread_pool_size,
2509 &fs_info->generic_worker); 2509 &fs_info->generic_worker);
2510 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", 2510 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
2511 fs_info->thread_pool_size, 2511 fs_info->thread_pool_size,
2512 &fs_info->generic_worker); 2512 &fs_info->generic_worker);
2513 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write", 2513 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2514 1, &fs_info->generic_worker); 2514 1, &fs_info->generic_worker);
2515 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta", 2515 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2516 fs_info->thread_pool_size, 2516 fs_info->thread_pool_size,
2517 &fs_info->generic_worker); 2517 &fs_info->generic_worker);
2518 btrfs_init_workers(&fs_info->readahead_workers, "readahead", 2518 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2519 fs_info->thread_pool_size, 2519 fs_info->thread_pool_size,
2520 &fs_info->generic_worker); 2520 &fs_info->generic_worker);
2521 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1, 2521 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2522 &fs_info->generic_worker); 2522 &fs_info->generic_worker);
2523 2523
2524 /* 2524 /*
2525 * endios are largely parallel and should have a very 2525 * endios are largely parallel and should have a very
2526 * low idle thresh 2526 * low idle thresh
2527 */ 2527 */
2528 fs_info->endio_workers.idle_thresh = 4; 2528 fs_info->endio_workers.idle_thresh = 4;
2529 fs_info->endio_meta_workers.idle_thresh = 4; 2529 fs_info->endio_meta_workers.idle_thresh = 4;
2530 fs_info->endio_raid56_workers.idle_thresh = 4; 2530 fs_info->endio_raid56_workers.idle_thresh = 4;
2531 fs_info->rmw_workers.idle_thresh = 2; 2531 fs_info->rmw_workers.idle_thresh = 2;
2532 2532
2533 fs_info->endio_write_workers.idle_thresh = 2; 2533 fs_info->endio_write_workers.idle_thresh = 2;
2534 fs_info->endio_meta_write_workers.idle_thresh = 2; 2534 fs_info->endio_meta_write_workers.idle_thresh = 2;
2535 fs_info->readahead_workers.idle_thresh = 2; 2535 fs_info->readahead_workers.idle_thresh = 2;
2536 2536
2537 /* 2537 /*
2538 * btrfs_start_workers can really only fail because of ENOMEM so just 2538 * btrfs_start_workers can really only fail because of ENOMEM so just
2539 * return -ENOMEM if any of these fail. 2539 * return -ENOMEM if any of these fail.
2540 */ 2540 */
2541 ret = btrfs_start_workers(&fs_info->workers); 2541 ret = btrfs_start_workers(&fs_info->workers);
2542 ret |= btrfs_start_workers(&fs_info->generic_worker); 2542 ret |= btrfs_start_workers(&fs_info->generic_worker);
2543 ret |= btrfs_start_workers(&fs_info->submit_workers); 2543 ret |= btrfs_start_workers(&fs_info->submit_workers);
2544 ret |= btrfs_start_workers(&fs_info->delalloc_workers); 2544 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2545 ret |= btrfs_start_workers(&fs_info->fixup_workers); 2545 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2546 ret |= btrfs_start_workers(&fs_info->endio_workers); 2546 ret |= btrfs_start_workers(&fs_info->endio_workers);
2547 ret |= btrfs_start_workers(&fs_info->endio_meta_workers); 2547 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
2548 ret |= btrfs_start_workers(&fs_info->rmw_workers); 2548 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2549 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers); 2549 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
2550 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers); 2550 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2551 ret |= btrfs_start_workers(&fs_info->endio_write_workers); 2551 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2552 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker); 2552 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2553 ret |= btrfs_start_workers(&fs_info->delayed_workers); 2553 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2554 ret |= btrfs_start_workers(&fs_info->caching_workers); 2554 ret |= btrfs_start_workers(&fs_info->caching_workers);
2555 ret |= btrfs_start_workers(&fs_info->readahead_workers); 2555 ret |= btrfs_start_workers(&fs_info->readahead_workers);
2556 ret |= btrfs_start_workers(&fs_info->flush_workers); 2556 ret |= btrfs_start_workers(&fs_info->flush_workers);
2557 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers); 2557 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
2558 if (ret) { 2558 if (ret) {
2559 err = -ENOMEM; 2559 err = -ENOMEM;
2560 goto fail_sb_buffer; 2560 goto fail_sb_buffer;
2561 } 2561 }
2562 2562
2563 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2563 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
2564 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2564 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2565 4 * 1024 * 1024 / PAGE_CACHE_SIZE); 2565 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
2566 2566
2567 tree_root->nodesize = nodesize; 2567 tree_root->nodesize = nodesize;
2568 tree_root->leafsize = leafsize; 2568 tree_root->leafsize = leafsize;
2569 tree_root->sectorsize = sectorsize; 2569 tree_root->sectorsize = sectorsize;
2570 tree_root->stripesize = stripesize; 2570 tree_root->stripesize = stripesize;
2571 2571
2572 sb->s_blocksize = sectorsize; 2572 sb->s_blocksize = sectorsize;
2573 sb->s_blocksize_bits = blksize_bits(sectorsize); 2573 sb->s_blocksize_bits = blksize_bits(sectorsize);
2574 2574
2575 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { 2575 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
2576 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id); 2576 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
2577 goto fail_sb_buffer; 2577 goto fail_sb_buffer;
2578 } 2578 }
2579 2579
2580 if (sectorsize != PAGE_SIZE) { 2580 if (sectorsize != PAGE_SIZE) {
2581 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) " 2581 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
2582 "found on %s\n", (unsigned long)sectorsize, sb->s_id); 2582 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
2583 goto fail_sb_buffer; 2583 goto fail_sb_buffer;
2584 } 2584 }
2585 2585
2586 mutex_lock(&fs_info->chunk_mutex); 2586 mutex_lock(&fs_info->chunk_mutex);
2587 ret = btrfs_read_sys_array(tree_root); 2587 ret = btrfs_read_sys_array(tree_root);
2588 mutex_unlock(&fs_info->chunk_mutex); 2588 mutex_unlock(&fs_info->chunk_mutex);
2589 if (ret) { 2589 if (ret) {
2590 printk(KERN_WARNING "BTRFS: failed to read the system " 2590 printk(KERN_WARNING "BTRFS: failed to read the system "
2591 "array on %s\n", sb->s_id); 2591 "array on %s\n", sb->s_id);
2592 goto fail_sb_buffer; 2592 goto fail_sb_buffer;
2593 } 2593 }
2594 2594
2595 blocksize = btrfs_level_size(tree_root, 2595 blocksize = btrfs_level_size(tree_root,
2596 btrfs_super_chunk_root_level(disk_super)); 2596 btrfs_super_chunk_root_level(disk_super));
2597 generation = btrfs_super_chunk_root_generation(disk_super); 2597 generation = btrfs_super_chunk_root_generation(disk_super);
2598 2598
2599 __setup_root(nodesize, leafsize, sectorsize, stripesize, 2599 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2600 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 2600 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2601 2601
2602 chunk_root->node = read_tree_block(chunk_root, 2602 chunk_root->node = read_tree_block(chunk_root,
2603 btrfs_super_chunk_root(disk_super), 2603 btrfs_super_chunk_root(disk_super),
2604 blocksize, generation); 2604 blocksize, generation);
2605 if (!chunk_root->node || 2605 if (!chunk_root->node ||
2606 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { 2606 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2607 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n", 2607 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
2608 sb->s_id); 2608 sb->s_id);
2609 goto fail_tree_roots; 2609 goto fail_tree_roots;
2610 } 2610 }
2611 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 2611 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2612 chunk_root->commit_root = btrfs_root_node(chunk_root); 2612 chunk_root->commit_root = btrfs_root_node(chunk_root);
2613 2613
2614 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2614 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
2615 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2615 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
2616 2616
2617 ret = btrfs_read_chunk_tree(chunk_root); 2617 ret = btrfs_read_chunk_tree(chunk_root);
2618 if (ret) { 2618 if (ret) {
2619 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n", 2619 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
2620 sb->s_id); 2620 sb->s_id);
2621 goto fail_tree_roots; 2621 goto fail_tree_roots;
2622 } 2622 }
2623 2623
2624 /* 2624 /*
2625 * keep the device that is marked to be the target device for the 2625 * keep the device that is marked to be the target device for the
2626 * dev_replace procedure 2626 * dev_replace procedure
2627 */ 2627 */
2628 btrfs_close_extra_devices(fs_info, fs_devices, 0); 2628 btrfs_close_extra_devices(fs_info, fs_devices, 0);
2629 2629
2630 if (!fs_devices->latest_bdev) { 2630 if (!fs_devices->latest_bdev) {
2631 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n", 2631 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
2632 sb->s_id); 2632 sb->s_id);
2633 goto fail_tree_roots; 2633 goto fail_tree_roots;
2634 } 2634 }
2635 2635
2636 retry_root_backup: 2636 retry_root_backup:
2637 blocksize = btrfs_level_size(tree_root, 2637 blocksize = btrfs_level_size(tree_root,
2638 btrfs_super_root_level(disk_super)); 2638 btrfs_super_root_level(disk_super));
2639 generation = btrfs_super_generation(disk_super); 2639 generation = btrfs_super_generation(disk_super);
2640 2640
2641 tree_root->node = read_tree_block(tree_root, 2641 tree_root->node = read_tree_block(tree_root,
2642 btrfs_super_root(disk_super), 2642 btrfs_super_root(disk_super),
2643 blocksize, generation); 2643 blocksize, generation);
2644 if (!tree_root->node || 2644 if (!tree_root->node ||
2645 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { 2645 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
2646 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", 2646 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
2647 sb->s_id); 2647 sb->s_id);
2648 2648
2649 goto recovery_tree_root; 2649 goto recovery_tree_root;
2650 } 2650 }
2651 2651
2652 btrfs_set_root_node(&tree_root->root_item, tree_root->node); 2652 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2653 tree_root->commit_root = btrfs_root_node(tree_root); 2653 tree_root->commit_root = btrfs_root_node(tree_root);
2654 btrfs_set_root_refs(&tree_root->root_item, 1); 2654 btrfs_set_root_refs(&tree_root->root_item, 1);
2655 2655
2656 location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 2656 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2657 location.type = BTRFS_ROOT_ITEM_KEY; 2657 location.type = BTRFS_ROOT_ITEM_KEY;
2658 location.offset = 0; 2658 location.offset = 0;
2659 2659
2660 extent_root = btrfs_read_tree_root(tree_root, &location); 2660 extent_root = btrfs_read_tree_root(tree_root, &location);
2661 if (IS_ERR(extent_root)) { 2661 if (IS_ERR(extent_root)) {
2662 ret = PTR_ERR(extent_root); 2662 ret = PTR_ERR(extent_root);
2663 goto recovery_tree_root; 2663 goto recovery_tree_root;
2664 } 2664 }
2665 extent_root->track_dirty = 1; 2665 extent_root->track_dirty = 1;
2666 fs_info->extent_root = extent_root; 2666 fs_info->extent_root = extent_root;
2667 2667
2668 location.objectid = BTRFS_DEV_TREE_OBJECTID; 2668 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2669 dev_root = btrfs_read_tree_root(tree_root, &location); 2669 dev_root = btrfs_read_tree_root(tree_root, &location);
2670 if (IS_ERR(dev_root)) { 2670 if (IS_ERR(dev_root)) {
2671 ret = PTR_ERR(dev_root); 2671 ret = PTR_ERR(dev_root);
2672 goto recovery_tree_root; 2672 goto recovery_tree_root;
2673 } 2673 }
2674 dev_root->track_dirty = 1; 2674 dev_root->track_dirty = 1;
2675 fs_info->dev_root = dev_root; 2675 fs_info->dev_root = dev_root;
2676 btrfs_init_devices_late(fs_info); 2676 btrfs_init_devices_late(fs_info);
2677 2677
2678 location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2678 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2679 csum_root = btrfs_read_tree_root(tree_root, &location); 2679 csum_root = btrfs_read_tree_root(tree_root, &location);
2680 if (IS_ERR(csum_root)) { 2680 if (IS_ERR(csum_root)) {
2681 ret = PTR_ERR(csum_root); 2681 ret = PTR_ERR(csum_root);
2682 goto recovery_tree_root; 2682 goto recovery_tree_root;
2683 } 2683 }
2684 csum_root->track_dirty = 1; 2684 csum_root->track_dirty = 1;
2685 fs_info->csum_root = csum_root; 2685 fs_info->csum_root = csum_root;
2686 2686
2687 location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2687 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2688 quota_root = btrfs_read_tree_root(tree_root, &location); 2688 quota_root = btrfs_read_tree_root(tree_root, &location);
2689 if (!IS_ERR(quota_root)) { 2689 if (!IS_ERR(quota_root)) {
2690 quota_root->track_dirty = 1; 2690 quota_root->track_dirty = 1;
2691 fs_info->quota_enabled = 1; 2691 fs_info->quota_enabled = 1;
2692 fs_info->pending_quota_state = 1; 2692 fs_info->pending_quota_state = 1;
2693 fs_info->quota_root = quota_root; 2693 fs_info->quota_root = quota_root;
2694 } 2694 }
2695 2695
2696 location.objectid = BTRFS_UUID_TREE_OBJECTID; 2696 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2697 uuid_root = btrfs_read_tree_root(tree_root, &location); 2697 uuid_root = btrfs_read_tree_root(tree_root, &location);
2698 if (IS_ERR(uuid_root)) { 2698 if (IS_ERR(uuid_root)) {
2699 ret = PTR_ERR(uuid_root); 2699 ret = PTR_ERR(uuid_root);
2700 if (ret != -ENOENT) 2700 if (ret != -ENOENT)
2701 goto recovery_tree_root; 2701 goto recovery_tree_root;
2702 create_uuid_tree = true; 2702 create_uuid_tree = true;
2703 check_uuid_tree = false; 2703 check_uuid_tree = false;
2704 } else { 2704 } else {
2705 uuid_root->track_dirty = 1; 2705 uuid_root->track_dirty = 1;
2706 fs_info->uuid_root = uuid_root; 2706 fs_info->uuid_root = uuid_root;
2707 create_uuid_tree = false; 2707 create_uuid_tree = false;
2708 check_uuid_tree = 2708 check_uuid_tree =
2709 generation != btrfs_super_uuid_tree_generation(disk_super); 2709 generation != btrfs_super_uuid_tree_generation(disk_super);
2710 } 2710 }
2711 2711
2712 fs_info->generation = generation; 2712 fs_info->generation = generation;
2713 fs_info->last_trans_committed = generation; 2713 fs_info->last_trans_committed = generation;
2714 2714
2715 ret = btrfs_recover_balance(fs_info); 2715 ret = btrfs_recover_balance(fs_info);
2716 if (ret) { 2716 if (ret) {
2717 printk(KERN_WARNING "BTRFS: failed to recover balance\n"); 2717 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
2718 goto fail_block_groups; 2718 goto fail_block_groups;
2719 } 2719 }
2720 2720
2721 ret = btrfs_init_dev_stats(fs_info); 2721 ret = btrfs_init_dev_stats(fs_info);
2722 if (ret) { 2722 if (ret) {
2723 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", 2723 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
2724 ret); 2724 ret);
2725 goto fail_block_groups; 2725 goto fail_block_groups;
2726 } 2726 }
2727 2727
2728 ret = btrfs_init_dev_replace(fs_info); 2728 ret = btrfs_init_dev_replace(fs_info);
2729 if (ret) { 2729 if (ret) {
2730 pr_err("BTRFS: failed to init dev_replace: %d\n", ret); 2730 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
2731 goto fail_block_groups; 2731 goto fail_block_groups;
2732 } 2732 }
2733 2733
2734 btrfs_close_extra_devices(fs_info, fs_devices, 1); 2734 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2735 2735
2736 ret = btrfs_sysfs_add_one(fs_info); 2736 ret = btrfs_sysfs_add_one(fs_info);
2737 if (ret) { 2737 if (ret) {
2738 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); 2738 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
2739 goto fail_block_groups; 2739 goto fail_block_groups;
2740 } 2740 }
2741 2741
2742 ret = btrfs_init_space_info(fs_info); 2742 ret = btrfs_init_space_info(fs_info);
2743 if (ret) { 2743 if (ret) {
2744 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); 2744 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2745 goto fail_sysfs; 2745 goto fail_sysfs;
2746 } 2746 }
2747 2747
2748 ret = btrfs_read_block_groups(extent_root); 2748 ret = btrfs_read_block_groups(extent_root);
2749 if (ret) { 2749 if (ret) {
2750 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); 2750 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2751 goto fail_sysfs; 2751 goto fail_sysfs;
2752 } 2752 }
2753 fs_info->num_tolerated_disk_barrier_failures = 2753 fs_info->num_tolerated_disk_barrier_failures =
2754 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2754 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
2755 if (fs_info->fs_devices->missing_devices > 2755 if (fs_info->fs_devices->missing_devices >
2756 fs_info->num_tolerated_disk_barrier_failures && 2756 fs_info->num_tolerated_disk_barrier_failures &&
2757 !(sb->s_flags & MS_RDONLY)) { 2757 !(sb->s_flags & MS_RDONLY)) {
2758 printk(KERN_WARNING "BTRFS: " 2758 printk(KERN_WARNING "BTRFS: "
2759 "too many missing devices, writeable mount is not allowed\n"); 2759 "too many missing devices, writeable mount is not allowed\n");
2760 goto fail_sysfs; 2760 goto fail_sysfs;
2761 } 2761 }
2762 2762
2763 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2763 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2764 "btrfs-cleaner"); 2764 "btrfs-cleaner");
2765 if (IS_ERR(fs_info->cleaner_kthread)) 2765 if (IS_ERR(fs_info->cleaner_kthread))
2766 goto fail_sysfs; 2766 goto fail_sysfs;
2767 2767
2768 fs_info->transaction_kthread = kthread_run(transaction_kthread, 2768 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2769 tree_root, 2769 tree_root,
2770 "btrfs-transaction"); 2770 "btrfs-transaction");
2771 if (IS_ERR(fs_info->transaction_kthread)) 2771 if (IS_ERR(fs_info->transaction_kthread))
2772 goto fail_cleaner; 2772 goto fail_cleaner;
2773 2773
2774 if (!btrfs_test_opt(tree_root, SSD) && 2774 if (!btrfs_test_opt(tree_root, SSD) &&
2775 !btrfs_test_opt(tree_root, NOSSD) && 2775 !btrfs_test_opt(tree_root, NOSSD) &&
2776 !fs_info->fs_devices->rotating) { 2776 !fs_info->fs_devices->rotating) {
2777 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " 2777 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
2778 "mode\n"); 2778 "mode\n");
2779 btrfs_set_opt(fs_info->mount_opt, SSD); 2779 btrfs_set_opt(fs_info->mount_opt, SSD);
2780 } 2780 }
2781 2781
2782 /* Set the real inode map cache flag */ 2782 /* Set the real inode map cache flag */
2783 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE)) 2783 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2784 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE); 2784 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2785 2785
2786 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 2786 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2787 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 2787 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2788 ret = btrfsic_mount(tree_root, fs_devices, 2788 ret = btrfsic_mount(tree_root, fs_devices,
2789 btrfs_test_opt(tree_root, 2789 btrfs_test_opt(tree_root,
2790 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 2790 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2791 1 : 0, 2791 1 : 0,
2792 fs_info->check_integrity_print_mask); 2792 fs_info->check_integrity_print_mask);
2793 if (ret) 2793 if (ret)
2794 printk(KERN_WARNING "BTRFS: failed to initialize" 2794 printk(KERN_WARNING "BTRFS: failed to initialize"
2795 " integrity check module %s\n", sb->s_id); 2795 " integrity check module %s\n", sb->s_id);
2796 } 2796 }
2797 #endif 2797 #endif
2798 ret = btrfs_read_qgroup_config(fs_info); 2798 ret = btrfs_read_qgroup_config(fs_info);
2799 if (ret) 2799 if (ret)
2800 goto fail_trans_kthread; 2800 goto fail_trans_kthread;
2801 2801
2802 /* do not make disk changes in broken FS */ 2802 /* do not make disk changes in broken FS */
2803 if (btrfs_super_log_root(disk_super) != 0) { 2803 if (btrfs_super_log_root(disk_super) != 0) {
2804 u64 bytenr = btrfs_super_log_root(disk_super); 2804 u64 bytenr = btrfs_super_log_root(disk_super);
2805 2805
2806 if (fs_devices->rw_devices == 0) { 2806 if (fs_devices->rw_devices == 0) {
2807 printk(KERN_WARNING "BTRFS: log replay required " 2807 printk(KERN_WARNING "BTRFS: log replay required "
2808 "on RO media\n"); 2808 "on RO media\n");
2809 err = -EIO; 2809 err = -EIO;
2810 goto fail_qgroup; 2810 goto fail_qgroup;
2811 } 2811 }
2812 blocksize = 2812 blocksize =
2813 btrfs_level_size(tree_root, 2813 btrfs_level_size(tree_root,
2814 btrfs_super_log_root_level(disk_super)); 2814 btrfs_super_log_root_level(disk_super));
2815 2815
2816 log_tree_root = btrfs_alloc_root(fs_info); 2816 log_tree_root = btrfs_alloc_root(fs_info);
2817 if (!log_tree_root) { 2817 if (!log_tree_root) {
2818 err = -ENOMEM; 2818 err = -ENOMEM;
2819 goto fail_qgroup; 2819 goto fail_qgroup;
2820 } 2820 }
2821 2821
2822 __setup_root(nodesize, leafsize, sectorsize, stripesize, 2822 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2823 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 2823 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2824 2824
2825 log_tree_root->node = read_tree_block(tree_root, bytenr, 2825 log_tree_root->node = read_tree_block(tree_root, bytenr,
2826 blocksize, 2826 blocksize,
2827 generation + 1); 2827 generation + 1);
2828 if (!log_tree_root->node || 2828 if (!log_tree_root->node ||
2829 !extent_buffer_uptodate(log_tree_root->node)) { 2829 !extent_buffer_uptodate(log_tree_root->node)) {
2830 printk(KERN_ERR "BTRFS: failed to read log tree\n"); 2830 printk(KERN_ERR "BTRFS: failed to read log tree\n");
2831 free_extent_buffer(log_tree_root->node); 2831 free_extent_buffer(log_tree_root->node);
2832 kfree(log_tree_root); 2832 kfree(log_tree_root);
2833 goto fail_trans_kthread; 2833 goto fail_trans_kthread;
2834 } 2834 }
2835 /* returns with log_tree_root freed on success */ 2835 /* returns with log_tree_root freed on success */
2836 ret = btrfs_recover_log_trees(log_tree_root); 2836 ret = btrfs_recover_log_trees(log_tree_root);
2837 if (ret) { 2837 if (ret) {
2838 btrfs_error(tree_root->fs_info, ret, 2838 btrfs_error(tree_root->fs_info, ret,
2839 "Failed to recover log tree"); 2839 "Failed to recover log tree");
2840 free_extent_buffer(log_tree_root->node); 2840 free_extent_buffer(log_tree_root->node);
2841 kfree(log_tree_root); 2841 kfree(log_tree_root);
2842 goto fail_trans_kthread; 2842 goto fail_trans_kthread;
2843 } 2843 }
2844 2844
2845 if (sb->s_flags & MS_RDONLY) { 2845 if (sb->s_flags & MS_RDONLY) {
2846 ret = btrfs_commit_super(tree_root); 2846 ret = btrfs_commit_super(tree_root);
2847 if (ret) 2847 if (ret)
2848 goto fail_trans_kthread; 2848 goto fail_trans_kthread;
2849 } 2849 }
2850 } 2850 }
2851 2851
2852 ret = btrfs_find_orphan_roots(tree_root); 2852 ret = btrfs_find_orphan_roots(tree_root);
2853 if (ret) 2853 if (ret)
2854 goto fail_trans_kthread; 2854 goto fail_trans_kthread;
2855 2855
2856 if (!(sb->s_flags & MS_RDONLY)) { 2856 if (!(sb->s_flags & MS_RDONLY)) {
2857 ret = btrfs_cleanup_fs_roots(fs_info); 2857 ret = btrfs_cleanup_fs_roots(fs_info);
2858 if (ret) 2858 if (ret)
2859 goto fail_trans_kthread; 2859 goto fail_trans_kthread;
2860 2860
2861 ret = btrfs_recover_relocation(tree_root); 2861 ret = btrfs_recover_relocation(tree_root);
2862 if (ret < 0) { 2862 if (ret < 0) {
2863 printk(KERN_WARNING 2863 printk(KERN_WARNING
2864 "BTRFS: failed to recover relocation\n"); 2864 "BTRFS: failed to recover relocation\n");
2865 err = -EINVAL; 2865 err = -EINVAL;
2866 goto fail_qgroup; 2866 goto fail_qgroup;
2867 } 2867 }
2868 } 2868 }
2869 2869
2870 location.objectid = BTRFS_FS_TREE_OBJECTID; 2870 location.objectid = BTRFS_FS_TREE_OBJECTID;
2871 location.type = BTRFS_ROOT_ITEM_KEY; 2871 location.type = BTRFS_ROOT_ITEM_KEY;
2872 location.offset = 0; 2872 location.offset = 0;
2873 2873
2874 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 2874 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2875 if (IS_ERR(fs_info->fs_root)) { 2875 if (IS_ERR(fs_info->fs_root)) {
2876 err = PTR_ERR(fs_info->fs_root); 2876 err = PTR_ERR(fs_info->fs_root);
2877 goto fail_qgroup; 2877 goto fail_qgroup;
2878 } 2878 }
2879 2879
2880 if (sb->s_flags & MS_RDONLY) 2880 if (sb->s_flags & MS_RDONLY)
2881 return 0; 2881 return 0;
2882 2882
2883 down_read(&fs_info->cleanup_work_sem); 2883 down_read(&fs_info->cleanup_work_sem);
2884 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 2884 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2885 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 2885 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
2886 up_read(&fs_info->cleanup_work_sem); 2886 up_read(&fs_info->cleanup_work_sem);
2887 close_ctree(tree_root); 2887 close_ctree(tree_root);
2888 return ret; 2888 return ret;
2889 } 2889 }
2890 up_read(&fs_info->cleanup_work_sem); 2890 up_read(&fs_info->cleanup_work_sem);
2891 2891
2892 ret = btrfs_resume_balance_async(fs_info); 2892 ret = btrfs_resume_balance_async(fs_info);
2893 if (ret) { 2893 if (ret) {
2894 printk(KERN_WARNING "BTRFS: failed to resume balance\n"); 2894 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2895 close_ctree(tree_root); 2895 close_ctree(tree_root);
2896 return ret; 2896 return ret;
2897 } 2897 }
2898 2898
2899 ret = btrfs_resume_dev_replace_async(fs_info); 2899 ret = btrfs_resume_dev_replace_async(fs_info);
2900 if (ret) { 2900 if (ret) {
2901 pr_warn("BTRFS: failed to resume dev_replace\n"); 2901 pr_warn("BTRFS: failed to resume dev_replace\n");
2902 close_ctree(tree_root); 2902 close_ctree(tree_root);
2903 return ret; 2903 return ret;
2904 } 2904 }
2905 2905
2906 btrfs_qgroup_rescan_resume(fs_info); 2906 btrfs_qgroup_rescan_resume(fs_info);
2907 2907
2908 if (create_uuid_tree) { 2908 if (create_uuid_tree) {
2909 pr_info("BTRFS: creating UUID tree\n"); 2909 pr_info("BTRFS: creating UUID tree\n");
2910 ret = btrfs_create_uuid_tree(fs_info); 2910 ret = btrfs_create_uuid_tree(fs_info);
2911 if (ret) { 2911 if (ret) {
2912 pr_warn("BTRFS: failed to create the UUID tree %d\n", 2912 pr_warn("BTRFS: failed to create the UUID tree %d\n",
2913 ret); 2913 ret);
2914 close_ctree(tree_root); 2914 close_ctree(tree_root);
2915 return ret; 2915 return ret;
2916 } 2916 }
2917 } else if (check_uuid_tree || 2917 } else if (check_uuid_tree ||
2918 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) { 2918 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
2919 pr_info("BTRFS: checking UUID tree\n"); 2919 pr_info("BTRFS: checking UUID tree\n");
2920 ret = btrfs_check_uuid_tree(fs_info); 2920 ret = btrfs_check_uuid_tree(fs_info);
2921 if (ret) { 2921 if (ret) {
2922 pr_warn("BTRFS: failed to check the UUID tree %d\n", 2922 pr_warn("BTRFS: failed to check the UUID tree %d\n",
2923 ret); 2923 ret);
2924 close_ctree(tree_root); 2924 close_ctree(tree_root);
2925 return ret; 2925 return ret;
2926 } 2926 }
2927 } else { 2927 } else {
2928 fs_info->update_uuid_tree_gen = 1; 2928 fs_info->update_uuid_tree_gen = 1;
2929 } 2929 }
2930 2930
2931 return 0; 2931 return 0;
2932 2932
2933 fail_qgroup: 2933 fail_qgroup:
2934 btrfs_free_qgroup_config(fs_info); 2934 btrfs_free_qgroup_config(fs_info);
2935 fail_trans_kthread: 2935 fail_trans_kthread:
2936 kthread_stop(fs_info->transaction_kthread); 2936 kthread_stop(fs_info->transaction_kthread);
2937 btrfs_cleanup_transaction(fs_info->tree_root); 2937 btrfs_cleanup_transaction(fs_info->tree_root);
2938 del_fs_roots(fs_info); 2938 del_fs_roots(fs_info);
2939 fail_cleaner: 2939 fail_cleaner:
2940 kthread_stop(fs_info->cleaner_kthread); 2940 kthread_stop(fs_info->cleaner_kthread);
2941 2941
2942 /* 2942 /*
2943 * make sure we're done with the btree inode before we stop our 2943 * make sure we're done with the btree inode before we stop our
2944 * kthreads 2944 * kthreads
2945 */ 2945 */
2946 filemap_write_and_wait(fs_info->btree_inode->i_mapping); 2946 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2947 2947
2948 fail_sysfs: 2948 fail_sysfs:
2949 btrfs_sysfs_remove_one(fs_info); 2949 btrfs_sysfs_remove_one(fs_info);
2950 2950
2951 fail_block_groups: 2951 fail_block_groups:
2952 btrfs_put_block_group_cache(fs_info); 2952 btrfs_put_block_group_cache(fs_info);
2953 btrfs_free_block_groups(fs_info); 2953 btrfs_free_block_groups(fs_info);
2954 2954
2955 fail_tree_roots: 2955 fail_tree_roots:
2956 free_root_pointers(fs_info, 1); 2956 free_root_pointers(fs_info, 1);
2957 invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 2957 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2958 2958
2959 fail_sb_buffer: 2959 fail_sb_buffer:
2960 btrfs_stop_all_workers(fs_info); 2960 btrfs_stop_all_workers(fs_info);
2961 fail_alloc: 2961 fail_alloc:
2962 fail_iput: 2962 fail_iput:
2963 btrfs_mapping_tree_free(&fs_info->mapping_tree); 2963 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2964 2964
2965 iput(fs_info->btree_inode); 2965 iput(fs_info->btree_inode);
2966 fail_delalloc_bytes: 2966 fail_delalloc_bytes:
2967 percpu_counter_destroy(&fs_info->delalloc_bytes); 2967 percpu_counter_destroy(&fs_info->delalloc_bytes);
2968 fail_dirty_metadata_bytes: 2968 fail_dirty_metadata_bytes:
2969 percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 2969 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
2970 fail_bdi: 2970 fail_bdi:
2971 bdi_destroy(&fs_info->bdi); 2971 bdi_destroy(&fs_info->bdi);
2972 fail_srcu: 2972 fail_srcu:
2973 cleanup_srcu_struct(&fs_info->subvol_srcu); 2973 cleanup_srcu_struct(&fs_info->subvol_srcu);
2974 fail: 2974 fail:
2975 btrfs_free_stripe_hash_table(fs_info); 2975 btrfs_free_stripe_hash_table(fs_info);
2976 btrfs_close_devices(fs_info->fs_devices); 2976 btrfs_close_devices(fs_info->fs_devices);
2977 return err; 2977 return err;
2978 2978
2979 recovery_tree_root: 2979 recovery_tree_root:
2980 if (!btrfs_test_opt(tree_root, RECOVERY)) 2980 if (!btrfs_test_opt(tree_root, RECOVERY))
2981 goto fail_tree_roots; 2981 goto fail_tree_roots;
2982 2982
2983 free_root_pointers(fs_info, 0); 2983 free_root_pointers(fs_info, 0);
2984 2984
2985 /* don't use the log in recovery mode, it won't be valid */ 2985 /* don't use the log in recovery mode, it won't be valid */
2986 btrfs_set_super_log_root(disk_super, 0); 2986 btrfs_set_super_log_root(disk_super, 0);
2987 2987
2988 /* we can't trust the free space cache either */ 2988 /* we can't trust the free space cache either */
2989 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2989 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2990 2990
2991 ret = next_root_backup(fs_info, fs_info->super_copy, 2991 ret = next_root_backup(fs_info, fs_info->super_copy,
2992 &num_backups_tried, &backup_index); 2992 &num_backups_tried, &backup_index);
2993 if (ret == -1) 2993 if (ret == -1)
2994 goto fail_block_groups; 2994 goto fail_block_groups;
2995 goto retry_root_backup; 2995 goto retry_root_backup;
2996 } 2996 }
2997 2997
2998 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 2998 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2999 { 2999 {
3000 if (uptodate) { 3000 if (uptodate) {
3001 set_buffer_uptodate(bh); 3001 set_buffer_uptodate(bh);
3002 } else { 3002 } else {
3003 struct btrfs_device *device = (struct btrfs_device *) 3003 struct btrfs_device *device = (struct btrfs_device *)
3004 bh->b_private; 3004 bh->b_private;
3005 3005
3006 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to " 3006 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
3007 "I/O error on %s\n", 3007 "I/O error on %s\n",
3008 rcu_str_deref(device->name)); 3008 rcu_str_deref(device->name));
3009 /* note, we dont' set_buffer_write_io_error because we have 3009 /* note, we dont' set_buffer_write_io_error because we have
3010 * our own ways of dealing with the IO errors 3010 * our own ways of dealing with the IO errors
3011 */ 3011 */
3012 clear_buffer_uptodate(bh); 3012 clear_buffer_uptodate(bh);
3013 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3013 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
3014 } 3014 }
3015 unlock_buffer(bh); 3015 unlock_buffer(bh);
3016 put_bh(bh); 3016 put_bh(bh);
3017 } 3017 }
3018 3018
3019 struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3019 struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3020 { 3020 {
3021 struct buffer_head *bh; 3021 struct buffer_head *bh;
3022 struct buffer_head *latest = NULL; 3022 struct buffer_head *latest = NULL;
3023 struct btrfs_super_block *super; 3023 struct btrfs_super_block *super;
3024 int i; 3024 int i;
3025 u64 transid = 0; 3025 u64 transid = 0;
3026 u64 bytenr; 3026 u64 bytenr;
3027 3027
3028 /* we would like to check all the supers, but that would make 3028 /* we would like to check all the supers, but that would make
3029 * a btrfs mount succeed after a mkfs from a different FS. 3029 * a btrfs mount succeed after a mkfs from a different FS.
3030 * So, we need to add a special mount option to scan for 3030 * So, we need to add a special mount option to scan for
3031 * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3031 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3032 */ 3032 */
3033 for (i = 0; i < 1; i++) { 3033 for (i = 0; i < 1; i++) {
3034 bytenr = btrfs_sb_offset(i); 3034 bytenr = btrfs_sb_offset(i);
3035 if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3035 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3036 i_size_read(bdev->bd_inode)) 3036 i_size_read(bdev->bd_inode))
3037 break; 3037 break;
3038 bh = __bread(bdev, bytenr / 4096, 3038 bh = __bread(bdev, bytenr / 4096,
3039 BTRFS_SUPER_INFO_SIZE); 3039 BTRFS_SUPER_INFO_SIZE);
3040 if (!bh) 3040 if (!bh)
3041 continue; 3041 continue;
3042 3042
3043 super = (struct btrfs_super_block *)bh->b_data; 3043 super = (struct btrfs_super_block *)bh->b_data;
3044 if (btrfs_super_bytenr(super) != bytenr || 3044 if (btrfs_super_bytenr(super) != bytenr ||
3045 btrfs_super_magic(super) != BTRFS_MAGIC) { 3045 btrfs_super_magic(super) != BTRFS_MAGIC) {
3046 brelse(bh); 3046 brelse(bh);
3047 continue; 3047 continue;
3048 } 3048 }
3049 3049
3050 if (!latest || btrfs_super_generation(super) > transid) { 3050 if (!latest || btrfs_super_generation(super) > transid) {
3051 brelse(latest); 3051 brelse(latest);
3052 latest = bh; 3052 latest = bh;
3053 transid = btrfs_super_generation(super); 3053 transid = btrfs_super_generation(super);
3054 } else { 3054 } else {
3055 brelse(bh); 3055 brelse(bh);
3056 } 3056 }
3057 } 3057 }
3058 return latest; 3058 return latest;
3059 } 3059 }
3060 3060
3061 /* 3061 /*
3062 * this should be called twice, once with wait == 0 and 3062 * this should be called twice, once with wait == 0 and
3063 * once with wait == 1. When wait == 0 is done, all the buffer heads 3063 * once with wait == 1. When wait == 0 is done, all the buffer heads
3064 * we write are pinned. 3064 * we write are pinned.
3065 * 3065 *
3066 * They are released when wait == 1 is done. 3066 * They are released when wait == 1 is done.
3067 * max_mirrors must be the same for both runs, and it indicates how 3067 * max_mirrors must be the same for both runs, and it indicates how
3068 * many supers on this one device should be written. 3068 * many supers on this one device should be written.
3069 * 3069 *
3070 * max_mirrors == 0 means to write them all. 3070 * max_mirrors == 0 means to write them all.
3071 */ 3071 */
3072 static int write_dev_supers(struct btrfs_device *device, 3072 static int write_dev_supers(struct btrfs_device *device,
3073 struct btrfs_super_block *sb, 3073 struct btrfs_super_block *sb,
3074 int do_barriers, int wait, int max_mirrors) 3074 int do_barriers, int wait, int max_mirrors)
3075 { 3075 {
3076 struct buffer_head *bh; 3076 struct buffer_head *bh;
3077 int i; 3077 int i;
3078 int ret; 3078 int ret;
3079 int errors = 0; 3079 int errors = 0;
3080 u32 crc; 3080 u32 crc;
3081 u64 bytenr; 3081 u64 bytenr;
3082 3082
3083 if (max_mirrors == 0) 3083 if (max_mirrors == 0)
3084 max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3084 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3085 3085
3086 for (i = 0; i < max_mirrors; i++) { 3086 for (i = 0; i < max_mirrors; i++) {
3087 bytenr = btrfs_sb_offset(i); 3087 bytenr = btrfs_sb_offset(i);
3088 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) 3088 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3089 break; 3089 break;
3090 3090
3091 if (wait) { 3091 if (wait) {
3092 bh = __find_get_block(device->bdev, bytenr / 4096, 3092 bh = __find_get_block(device->bdev, bytenr / 4096,
3093 BTRFS_SUPER_INFO_SIZE); 3093 BTRFS_SUPER_INFO_SIZE);
3094 if (!bh) { 3094 if (!bh) {
3095 errors++; 3095 errors++;
3096 continue; 3096 continue;
3097 } 3097 }
3098 wait_on_buffer(bh); 3098 wait_on_buffer(bh);
3099 if (!buffer_uptodate(bh)) 3099 if (!buffer_uptodate(bh))
3100 errors++; 3100 errors++;
3101 3101
3102 /* drop our reference */ 3102 /* drop our reference */
3103 brelse(bh); 3103 brelse(bh);
3104 3104
3105 /* drop the reference from the wait == 0 run */ 3105 /* drop the reference from the wait == 0 run */
3106 brelse(bh); 3106 brelse(bh);
3107 continue; 3107 continue;
3108 } else { 3108 } else {
3109 btrfs_set_super_bytenr(sb, bytenr); 3109 btrfs_set_super_bytenr(sb, bytenr);
3110 3110
3111 crc = ~(u32)0; 3111 crc = ~(u32)0;
3112 crc = btrfs_csum_data((char *)sb + 3112 crc = btrfs_csum_data((char *)sb +
3113 BTRFS_CSUM_SIZE, crc, 3113 BTRFS_CSUM_SIZE, crc,
3114 BTRFS_SUPER_INFO_SIZE - 3114 BTRFS_SUPER_INFO_SIZE -
3115 BTRFS_CSUM_SIZE); 3115 BTRFS_CSUM_SIZE);
3116 btrfs_csum_final(crc, sb->csum); 3116 btrfs_csum_final(crc, sb->csum);
3117 3117
3118 /* 3118 /*
3119 * one reference for us, and we leave it for the 3119 * one reference for us, and we leave it for the
3120 * caller 3120 * caller
3121 */ 3121 */
3122 bh = __getblk(device->bdev, bytenr / 4096, 3122 bh = __getblk(device->bdev, bytenr / 4096,
3123 BTRFS_SUPER_INFO_SIZE); 3123 BTRFS_SUPER_INFO_SIZE);
3124 if (!bh) { 3124 if (!bh) {
3125 printk(KERN_ERR "BTRFS: couldn't get super " 3125 printk(KERN_ERR "BTRFS: couldn't get super "
3126 "buffer head for bytenr %Lu\n", bytenr); 3126 "buffer head for bytenr %Lu\n", bytenr);
3127 errors++; 3127 errors++;
3128 continue; 3128 continue;
3129 } 3129 }
3130 3130
3131 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3131 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3132 3132
3133 /* one reference for submit_bh */ 3133 /* one reference for submit_bh */
3134 get_bh(bh); 3134 get_bh(bh);
3135 3135
3136 set_buffer_uptodate(bh); 3136 set_buffer_uptodate(bh);
3137 lock_buffer(bh); 3137 lock_buffer(bh);
3138 bh->b_end_io = btrfs_end_buffer_write_sync; 3138 bh->b_end_io = btrfs_end_buffer_write_sync;
3139 bh->b_private = device; 3139 bh->b_private = device;
3140 } 3140 }
3141 3141
3142 /* 3142 /*
3143 * we fua the first super. The others we allow 3143 * we fua the first super. The others we allow
3144 * to go down lazy. 3144 * to go down lazy.
3145 */ 3145 */
3146 if (i == 0) 3146 if (i == 0)
3147 ret = btrfsic_submit_bh(WRITE_FUA, bh); 3147 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3148 else 3148 else
3149 ret = btrfsic_submit_bh(WRITE_SYNC, bh); 3149 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
3150 if (ret) 3150 if (ret)
3151 errors++; 3151 errors++;
3152 } 3152 }
3153 return errors < i ? 0 : -1; 3153 return errors < i ? 0 : -1;
3154 } 3154 }
3155 3155
3156 /* 3156 /*
3157 * endio for the write_dev_flush, this will wake anyone waiting 3157 * endio for the write_dev_flush, this will wake anyone waiting
3158 * for the barrier when it is done 3158 * for the barrier when it is done
3159 */ 3159 */
3160 static void btrfs_end_empty_barrier(struct bio *bio, int err) 3160 static void btrfs_end_empty_barrier(struct bio *bio, int err)
3161 { 3161 {
3162 if (err) { 3162 if (err) {
3163 if (err == -EOPNOTSUPP) 3163 if (err == -EOPNOTSUPP)
3164 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); 3164 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3165 clear_bit(BIO_UPTODATE, &bio->bi_flags); 3165 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3166 } 3166 }
3167 if (bio->bi_private) 3167 if (bio->bi_private)
3168 complete(bio->bi_private); 3168 complete(bio->bi_private);
3169 bio_put(bio); 3169 bio_put(bio);
3170 } 3170 }
3171 3171
3172 /* 3172 /*
3173 * trigger flushes for one the devices. If you pass wait == 0, the flushes are 3173 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3174 * sent down. With wait == 1, it waits for the previous flush. 3174 * sent down. With wait == 1, it waits for the previous flush.
3175 * 3175 *
3176 * any device where the flush fails with eopnotsupp are flagged as not-barrier 3176 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3177 * capable 3177 * capable
3178 */ 3178 */
3179 static int write_dev_flush(struct btrfs_device *device, int wait) 3179 static int write_dev_flush(struct btrfs_device *device, int wait)
3180 { 3180 {
3181 struct bio *bio; 3181 struct bio *bio;
3182 int ret = 0; 3182 int ret = 0;
3183 3183
3184 if (device->nobarriers) 3184 if (device->nobarriers)
3185 return 0; 3185 return 0;
3186 3186
3187 if (wait) { 3187 if (wait) {
3188 bio = device->flush_bio; 3188 bio = device->flush_bio;
3189 if (!bio) 3189 if (!bio)
3190 return 0; 3190 return 0;
3191 3191
3192 wait_for_completion(&device->flush_wait); 3192 wait_for_completion(&device->flush_wait);
3193 3193
3194 if (bio_flagged(bio, BIO_EOPNOTSUPP)) { 3194 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
3195 printk_in_rcu("BTRFS: disabling barriers on dev %s\n", 3195 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
3196 rcu_str_deref(device->name)); 3196 rcu_str_deref(device->name));
3197 device->nobarriers = 1; 3197 device->nobarriers = 1;
3198 } else if (!bio_flagged(bio, BIO_UPTODATE)) { 3198 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
3199 ret = -EIO; 3199 ret = -EIO;
3200 btrfs_dev_stat_inc_and_print(device, 3200 btrfs_dev_stat_inc_and_print(device,
3201 BTRFS_DEV_STAT_FLUSH_ERRS); 3201 BTRFS_DEV_STAT_FLUSH_ERRS);
3202 } 3202 }
3203 3203
3204 /* drop the reference from the wait == 0 run */ 3204 /* drop the reference from the wait == 0 run */
3205 bio_put(bio); 3205 bio_put(bio);
3206 device->flush_bio = NULL; 3206 device->flush_bio = NULL;
3207 3207
3208 return ret; 3208 return ret;
3209 } 3209 }
3210 3210
3211 /* 3211 /*
3212 * one reference for us, and we leave it for the 3212 * one reference for us, and we leave it for the
3213 * caller 3213 * caller
3214 */ 3214 */
3215 device->flush_bio = NULL; 3215 device->flush_bio = NULL;
3216 bio = btrfs_io_bio_alloc(GFP_NOFS, 0); 3216 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
3217 if (!bio) 3217 if (!bio)
3218 return -ENOMEM; 3218 return -ENOMEM;
3219 3219
3220 bio->bi_end_io = btrfs_end_empty_barrier; 3220 bio->bi_end_io = btrfs_end_empty_barrier;
3221 bio->bi_bdev = device->bdev; 3221 bio->bi_bdev = device->bdev;
3222 init_completion(&device->flush_wait); 3222 init_completion(&device->flush_wait);
3223 bio->bi_private = &device->flush_wait; 3223 bio->bi_private = &device->flush_wait;
3224 device->flush_bio = bio; 3224 device->flush_bio = bio;
3225 3225
3226 bio_get(bio); 3226 bio_get(bio);
3227 btrfsic_submit_bio(WRITE_FLUSH, bio); 3227 btrfsic_submit_bio(WRITE_FLUSH, bio);
3228 3228
3229 return 0; 3229 return 0;
3230 } 3230 }
3231 3231
3232 /* 3232 /*
3233 * send an empty flush down to each device in parallel, 3233 * send an empty flush down to each device in parallel,
3234 * then wait for them 3234 * then wait for them
3235 */ 3235 */
3236 static int barrier_all_devices(struct btrfs_fs_info *info) 3236 static int barrier_all_devices(struct btrfs_fs_info *info)
3237 { 3237 {
3238 struct list_head *head; 3238 struct list_head *head;
3239 struct btrfs_device *dev; 3239 struct btrfs_device *dev;
3240 int errors_send = 0; 3240 int errors_send = 0;
3241 int errors_wait = 0; 3241 int errors_wait = 0;
3242 int ret; 3242 int ret;
3243 3243
3244 /* send down all the barriers */ 3244 /* send down all the barriers */
3245 head = &info->fs_devices->devices; 3245 head = &info->fs_devices->devices;
3246 list_for_each_entry_rcu(dev, head, dev_list) { 3246 list_for_each_entry_rcu(dev, head, dev_list) {
3247 if (dev->missing) 3247 if (dev->missing)
3248 continue; 3248 continue;
3249 if (!dev->bdev) { 3249 if (!dev->bdev) {
3250 errors_send++; 3250 errors_send++;
3251 continue; 3251 continue;
3252 } 3252 }
3253 if (!dev->in_fs_metadata || !dev->writeable) 3253 if (!dev->in_fs_metadata || !dev->writeable)
3254 continue; 3254 continue;
3255 3255
3256 ret = write_dev_flush(dev, 0); 3256 ret = write_dev_flush(dev, 0);
3257 if (ret) 3257 if (ret)
3258 errors_send++; 3258 errors_send++;
3259 } 3259 }
3260 3260
3261 /* wait for all the barriers */ 3261 /* wait for all the barriers */
3262 list_for_each_entry_rcu(dev, head, dev_list) { 3262 list_for_each_entry_rcu(dev, head, dev_list) {
3263 if (dev->missing) 3263 if (dev->missing)
3264 continue; 3264 continue;
3265 if (!dev->bdev) { 3265 if (!dev->bdev) {
3266 errors_wait++; 3266 errors_wait++;
3267 continue; 3267 continue;
3268 } 3268 }
3269 if (!dev->in_fs_metadata || !dev->writeable) 3269 if (!dev->in_fs_metadata || !dev->writeable)
3270 continue; 3270 continue;
3271 3271
3272 ret = write_dev_flush(dev, 1); 3272 ret = write_dev_flush(dev, 1);
3273 if (ret) 3273 if (ret)
3274 errors_wait++; 3274 errors_wait++;
3275 } 3275 }
3276 if (errors_send > info->num_tolerated_disk_barrier_failures || 3276 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3277 errors_wait > info->num_tolerated_disk_barrier_failures) 3277 errors_wait > info->num_tolerated_disk_barrier_failures)
3278 return -EIO; 3278 return -EIO;
3279 return 0; 3279 return 0;
3280 } 3280 }
3281 3281
3282 int btrfs_calc_num_tolerated_disk_barrier_failures( 3282 int btrfs_calc_num_tolerated_disk_barrier_failures(
3283 struct btrfs_fs_info *fs_info) 3283 struct btrfs_fs_info *fs_info)
3284 { 3284 {
3285 struct btrfs_ioctl_space_info space; 3285 struct btrfs_ioctl_space_info space;
3286 struct btrfs_space_info *sinfo; 3286 struct btrfs_space_info *sinfo;
3287 u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 3287 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3288 BTRFS_BLOCK_GROUP_SYSTEM, 3288 BTRFS_BLOCK_GROUP_SYSTEM,
3289 BTRFS_BLOCK_GROUP_METADATA, 3289 BTRFS_BLOCK_GROUP_METADATA,
3290 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 3290 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3291 int num_types = 4; 3291 int num_types = 4;
3292 int i; 3292 int i;
3293 int c; 3293 int c;
3294 int num_tolerated_disk_barrier_failures = 3294 int num_tolerated_disk_barrier_failures =
3295 (int)fs_info->fs_devices->num_devices; 3295 (int)fs_info->fs_devices->num_devices;
3296 3296
3297 for (i = 0; i < num_types; i++) { 3297 for (i = 0; i < num_types; i++) {
3298 struct btrfs_space_info *tmp; 3298 struct btrfs_space_info *tmp;
3299 3299
3300 sinfo = NULL; 3300 sinfo = NULL;
3301 rcu_read_lock(); 3301 rcu_read_lock();
3302 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 3302 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3303 if (tmp->flags == types[i]) { 3303 if (tmp->flags == types[i]) {
3304 sinfo = tmp; 3304 sinfo = tmp;
3305 break; 3305 break;
3306 } 3306 }
3307 } 3307 }
3308 rcu_read_unlock(); 3308 rcu_read_unlock();
3309 3309
3310 if (!sinfo) 3310 if (!sinfo)
3311 continue; 3311 continue;
3312 3312
3313 down_read(&sinfo->groups_sem); 3313 down_read(&sinfo->groups_sem);
3314 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 3314 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3315 if (!list_empty(&sinfo->block_groups[c])) { 3315 if (!list_empty(&sinfo->block_groups[c])) {
3316 u64 flags; 3316 u64 flags;
3317 3317
3318 btrfs_get_block_group_info( 3318 btrfs_get_block_group_info(
3319 &sinfo->block_groups[c], &space); 3319 &sinfo->block_groups[c], &space);
3320 if (space.total_bytes == 0 || 3320 if (space.total_bytes == 0 ||
3321 space.used_bytes == 0) 3321 space.used_bytes == 0)
3322 continue; 3322 continue;
3323 flags = space.flags; 3323 flags = space.flags;
3324 /* 3324 /*
3325 * return 3325 * return
3326 * 0: if dup, single or RAID0 is configured for 3326 * 0: if dup, single or RAID0 is configured for
3327 * any of metadata, system or data, else 3327 * any of metadata, system or data, else
3328 * 1: if RAID5 is configured, or if RAID1 or 3328 * 1: if RAID5 is configured, or if RAID1 or
3329 * RAID10 is configured and only two mirrors 3329 * RAID10 is configured and only two mirrors
3330 * are used, else 3330 * are used, else
3331 * 2: if RAID6 is configured, else 3331 * 2: if RAID6 is configured, else
3332 * num_mirrors - 1: if RAID1 or RAID10 is 3332 * num_mirrors - 1: if RAID1 or RAID10 is
3333 * configured and more than 3333 * configured and more than
3334 * 2 mirrors are used. 3334 * 2 mirrors are used.
3335 */ 3335 */
3336 if (num_tolerated_disk_barrier_failures > 0 && 3336 if (num_tolerated_disk_barrier_failures > 0 &&
3337 ((flags & (BTRFS_BLOCK_GROUP_DUP | 3337 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3338 BTRFS_BLOCK_GROUP_RAID0)) || 3338 BTRFS_BLOCK_GROUP_RAID0)) ||
3339 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) 3339 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3340 == 0))) 3340 == 0)))
3341 num_tolerated_disk_barrier_failures = 0; 3341 num_tolerated_disk_barrier_failures = 0;
3342 else if (num_tolerated_disk_barrier_failures > 1) { 3342 else if (num_tolerated_disk_barrier_failures > 1) {
3343 if (flags & (BTRFS_BLOCK_GROUP_RAID1 | 3343 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3344 BTRFS_BLOCK_GROUP_RAID5 | 3344 BTRFS_BLOCK_GROUP_RAID5 |
3345 BTRFS_BLOCK_GROUP_RAID10)) { 3345 BTRFS_BLOCK_GROUP_RAID10)) {
3346 num_tolerated_disk_barrier_failures = 1; 3346 num_tolerated_disk_barrier_failures = 1;
3347 } else if (flags & 3347 } else if (flags &
3348 BTRFS_BLOCK_GROUP_RAID6) { 3348 BTRFS_BLOCK_GROUP_RAID6) {
3349 num_tolerated_disk_barrier_failures = 2; 3349 num_tolerated_disk_barrier_failures = 2;
3350 } 3350 }
3351 } 3351 }
3352 } 3352 }
3353 } 3353 }
3354 up_read(&sinfo->groups_sem); 3354 up_read(&sinfo->groups_sem);
3355 } 3355 }
3356 3356
3357 return num_tolerated_disk_barrier_failures; 3357 return num_tolerated_disk_barrier_failures;
3358 } 3358 }
3359 3359
3360 static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3360 static int write_all_supers(struct btrfs_root *root, int max_mirrors)
3361 { 3361 {
3362 struct list_head *head; 3362 struct list_head *head;
3363 struct btrfs_device *dev; 3363 struct btrfs_device *dev;
3364 struct btrfs_super_block *sb; 3364 struct btrfs_super_block *sb;
3365 struct btrfs_dev_item *dev_item; 3365 struct btrfs_dev_item *dev_item;
3366 int ret; 3366 int ret;
3367 int do_barriers; 3367 int do_barriers;
3368 int max_errors; 3368 int max_errors;
3369 int total_errors = 0; 3369 int total_errors = 0;
3370 u64 flags; 3370 u64 flags;
3371 3371
3372 do_barriers = !btrfs_test_opt(root, NOBARRIER); 3372 do_barriers = !btrfs_test_opt(root, NOBARRIER);
3373 backup_super_roots(root->fs_info); 3373 backup_super_roots(root->fs_info);
3374 3374
3375 sb = root->fs_info->super_for_commit; 3375 sb = root->fs_info->super_for_commit;
3376 dev_item = &sb->dev_item; 3376 dev_item = &sb->dev_item;
3377 3377
3378 mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3378 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
3379 head = &root->fs_info->fs_devices->devices; 3379 head = &root->fs_info->fs_devices->devices;
3380 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3380 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
3381 3381
3382 if (do_barriers) { 3382 if (do_barriers) {
3383 ret = barrier_all_devices(root->fs_info); 3383 ret = barrier_all_devices(root->fs_info);
3384 if (ret) { 3384 if (ret) {
3385 mutex_unlock( 3385 mutex_unlock(
3386 &root->fs_info->fs_devices->device_list_mutex); 3386 &root->fs_info->fs_devices->device_list_mutex);
3387 btrfs_error(root->fs_info, ret, 3387 btrfs_error(root->fs_info, ret,
3388 "errors while submitting device barriers."); 3388 "errors while submitting device barriers.");
3389 return ret; 3389 return ret;
3390 } 3390 }
3391 } 3391 }
3392 3392
3393 list_for_each_entry_rcu(dev, head, dev_list) { 3393 list_for_each_entry_rcu(dev, head, dev_list) {
3394 if (!dev->bdev) { 3394 if (!dev->bdev) {
3395 total_errors++; 3395 total_errors++;
3396 continue; 3396 continue;
3397 } 3397 }
3398 if (!dev->in_fs_metadata || !dev->writeable) 3398 if (!dev->in_fs_metadata || !dev->writeable)
3399 continue; 3399 continue;
3400 3400
3401 btrfs_set_stack_device_generation(dev_item, 0); 3401 btrfs_set_stack_device_generation(dev_item, 0);
3402 btrfs_set_stack_device_type(dev_item, dev->type); 3402 btrfs_set_stack_device_type(dev_item, dev->type);
3403 btrfs_set_stack_device_id(dev_item, dev->devid); 3403 btrfs_set_stack_device_id(dev_item, dev->devid);
3404 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); 3404 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3405 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); 3405 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3406 btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3406 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3407 btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3407 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3408 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3408 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3409 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 3409 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
3410 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3410 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
3411 3411
3412 flags = btrfs_super_flags(sb); 3412 flags = btrfs_super_flags(sb);
3413 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3413 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3414 3414
3415 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3415 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
3416 if (ret) 3416 if (ret)
3417 total_errors++; 3417 total_errors++;
3418 } 3418 }
3419 if (total_errors > max_errors) { 3419 if (total_errors > max_errors) {
3420 btrfs_err(root->fs_info, "%d errors while writing supers", 3420 btrfs_err(root->fs_info, "%d errors while writing supers",
3421 total_errors); 3421 total_errors);
3422 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3422 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
3423 3423
3424 /* FUA is masked off if unsupported and can't be the reason */ 3424 /* FUA is masked off if unsupported and can't be the reason */
3425 btrfs_error(root->fs_info, -EIO, 3425 btrfs_error(root->fs_info, -EIO,
3426 "%d errors while writing supers", total_errors); 3426 "%d errors while writing supers", total_errors);
3427 return -EIO; 3427 return -EIO;
3428 } 3428 }
3429 3429
3430 total_errors = 0; 3430 total_errors = 0;
3431 list_for_each_entry_rcu(dev, head, dev_list) { 3431 list_for_each_entry_rcu(dev, head, dev_list) {
3432 if (!dev->bdev) 3432 if (!dev->bdev)
3433 continue; 3433 continue;
3434 if (!dev->in_fs_metadata || !dev->writeable) 3434 if (!dev->in_fs_metadata || !dev->writeable)
3435 continue; 3435 continue;
3436 3436
3437 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3437 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3438 if (ret) 3438 if (ret)
3439 total_errors++; 3439 total_errors++;
3440 } 3440 }
3441 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3441 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
3442 if (total_errors > max_errors) { 3442 if (total_errors > max_errors) {
3443 btrfs_error(root->fs_info, -EIO, 3443 btrfs_error(root->fs_info, -EIO,
3444 "%d errors while writing supers", total_errors); 3444 "%d errors while writing supers", total_errors);
3445 return -EIO; 3445 return -EIO;
3446 } 3446 }
3447 return 0; 3447 return 0;
3448 } 3448 }
3449 3449
3450 int write_ctree_super(struct btrfs_trans_handle *trans, 3450 int write_ctree_super(struct btrfs_trans_handle *trans,
3451 struct btrfs_root *root, int max_mirrors) 3451 struct btrfs_root *root, int max_mirrors)
3452 { 3452 {
3453 return write_all_supers(root, max_mirrors); 3453 return write_all_supers(root, max_mirrors);
3454 } 3454 }
3455 3455
3456 /* Drop a fs root from the radix tree and free it. */ 3456 /* Drop a fs root from the radix tree and free it. */
3457 void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3457 void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3458 struct btrfs_root *root) 3458 struct btrfs_root *root)
3459 { 3459 {
3460 spin_lock(&fs_info->fs_roots_radix_lock); 3460 spin_lock(&fs_info->fs_roots_radix_lock);
3461 radix_tree_delete(&fs_info->fs_roots_radix, 3461 radix_tree_delete(&fs_info->fs_roots_radix,
3462 (unsigned long)root->root_key.objectid); 3462 (unsigned long)root->root_key.objectid);
3463 spin_unlock(&fs_info->fs_roots_radix_lock); 3463 spin_unlock(&fs_info->fs_roots_radix_lock);
3464 3464
3465 if (btrfs_root_refs(&root->root_item) == 0) 3465 if (btrfs_root_refs(&root->root_item) == 0)
3466 synchronize_srcu(&fs_info->subvol_srcu); 3466 synchronize_srcu(&fs_info->subvol_srcu);
3467 3467
3468 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 3468 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3469 btrfs_free_log(NULL, root); 3469 btrfs_free_log(NULL, root);
3470 3470
3471 __btrfs_remove_free_space_cache(root->free_ino_pinned); 3471 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3472 __btrfs_remove_free_space_cache(root->free_ino_ctl); 3472 __btrfs_remove_free_space_cache(root->free_ino_ctl);
3473 free_fs_root(root); 3473 free_fs_root(root);
3474 } 3474 }
3475 3475
3476 static void free_fs_root(struct btrfs_root *root) 3476 static void free_fs_root(struct btrfs_root *root)
3477 { 3477 {
3478 iput(root->cache_inode); 3478 iput(root->cache_inode);
3479 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 3479 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
3480 btrfs_free_block_rsv(root, root->orphan_block_rsv); 3480 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3481 root->orphan_block_rsv = NULL; 3481 root->orphan_block_rsv = NULL;
3482 if (root->anon_dev) 3482 if (root->anon_dev)
3483 free_anon_bdev(root->anon_dev); 3483 free_anon_bdev(root->anon_dev);
3484 free_extent_buffer(root->node); 3484 free_extent_buffer(root->node);
3485 free_extent_buffer(root->commit_root); 3485 free_extent_buffer(root->commit_root);
3486 kfree(root->free_ino_ctl); 3486 kfree(root->free_ino_ctl);
3487 kfree(root->free_ino_pinned); 3487 kfree(root->free_ino_pinned);
3488 kfree(root->name); 3488 kfree(root->name);
3489 btrfs_put_fs_root(root); 3489 btrfs_put_fs_root(root);
3490 } 3490 }
3491 3491
3492 void btrfs_free_fs_root(struct btrfs_root *root) 3492 void btrfs_free_fs_root(struct btrfs_root *root)
3493 { 3493 {
3494 free_fs_root(root); 3494 free_fs_root(root);
3495 } 3495 }
3496 3496
3497 int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3497 int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3498 { 3498 {
3499 u64 root_objectid = 0; 3499 u64 root_objectid = 0;
3500 struct btrfs_root *gang[8]; 3500 struct btrfs_root *gang[8];
3501 int i; 3501 int i;
3502 int ret; 3502 int ret;
3503 3503
3504 while (1) { 3504 while (1) {
3505 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3505 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3506 (void **)gang, root_objectid, 3506 (void **)gang, root_objectid,
3507 ARRAY_SIZE(gang)); 3507 ARRAY_SIZE(gang));
3508 if (!ret) 3508 if (!ret)
3509 break; 3509 break;
3510 3510
3511 root_objectid = gang[ret - 1]->root_key.objectid + 1; 3511 root_objectid = gang[ret - 1]->root_key.objectid + 1;
3512 for (i = 0; i < ret; i++) { 3512 for (i = 0; i < ret; i++) {
3513 int err; 3513 int err;
3514 3514
3515 root_objectid = gang[i]->root_key.objectid; 3515 root_objectid = gang[i]->root_key.objectid;
3516 err = btrfs_orphan_cleanup(gang[i]); 3516 err = btrfs_orphan_cleanup(gang[i]);
3517 if (err) 3517 if (err)
3518 return err; 3518 return err;
3519 } 3519 }
3520 root_objectid++; 3520 root_objectid++;
3521 } 3521 }
3522 return 0; 3522 return 0;
3523 } 3523 }
3524 3524
3525 int btrfs_commit_super(struct btrfs_root *root) 3525 int btrfs_commit_super(struct btrfs_root *root)
3526 { 3526 {
3527 struct btrfs_trans_handle *trans; 3527 struct btrfs_trans_handle *trans;
3528 3528
3529 mutex_lock(&root->fs_info->cleaner_mutex); 3529 mutex_lock(&root->fs_info->cleaner_mutex);
3530 btrfs_run_delayed_iputs(root); 3530 btrfs_run_delayed_iputs(root);
3531 mutex_unlock(&root->fs_info->cleaner_mutex); 3531 mutex_unlock(&root->fs_info->cleaner_mutex);
3532 wake_up_process(root->fs_info->cleaner_kthread); 3532 wake_up_process(root->fs_info->cleaner_kthread);
3533 3533
3534 /* wait until ongoing cleanup work done */ 3534 /* wait until ongoing cleanup work done */
3535 down_write(&root->fs_info->cleanup_work_sem); 3535 down_write(&root->fs_info->cleanup_work_sem);
3536 up_write(&root->fs_info->cleanup_work_sem); 3536 up_write(&root->fs_info->cleanup_work_sem);
3537 3537
3538 trans = btrfs_join_transaction(root); 3538 trans = btrfs_join_transaction(root);
3539 if (IS_ERR(trans)) 3539 if (IS_ERR(trans))
3540 return PTR_ERR(trans); 3540 return PTR_ERR(trans);
3541 return btrfs_commit_transaction(trans, root); 3541 return btrfs_commit_transaction(trans, root);
3542 } 3542 }
3543 3543
3544 int close_ctree(struct btrfs_root *root) 3544 int close_ctree(struct btrfs_root *root)
3545 { 3545 {
3546 struct btrfs_fs_info *fs_info = root->fs_info; 3546 struct btrfs_fs_info *fs_info = root->fs_info;
3547 int ret; 3547 int ret;
3548 3548
3549 fs_info->closing = 1; 3549 fs_info->closing = 1;
3550 smp_mb(); 3550 smp_mb();
3551 3551
3552 /* wait for the uuid_scan task to finish */ 3552 /* wait for the uuid_scan task to finish */
3553 down(&fs_info->uuid_tree_rescan_sem); 3553 down(&fs_info->uuid_tree_rescan_sem);
3554 /* avoid complains from lockdep et al., set sem back to initial state */ 3554 /* avoid complains from lockdep et al., set sem back to initial state */
3555 up(&fs_info->uuid_tree_rescan_sem); 3555 up(&fs_info->uuid_tree_rescan_sem);
3556 3556
3557 /* pause restriper - we want to resume on mount */ 3557 /* pause restriper - we want to resume on mount */
3558 btrfs_pause_balance(fs_info); 3558 btrfs_pause_balance(fs_info);
3559 3559
3560 btrfs_dev_replace_suspend_for_unmount(fs_info); 3560 btrfs_dev_replace_suspend_for_unmount(fs_info);
3561 3561
3562 btrfs_scrub_cancel(fs_info); 3562 btrfs_scrub_cancel(fs_info);
3563 3563
3564 /* wait for any defraggers to finish */ 3564 /* wait for any defraggers to finish */
3565 wait_event(fs_info->transaction_wait, 3565 wait_event(fs_info->transaction_wait,
3566 (atomic_read(&fs_info->defrag_running) == 0)); 3566 (atomic_read(&fs_info->defrag_running) == 0));
3567 3567
3568 /* clear out the rbtree of defraggable inodes */ 3568 /* clear out the rbtree of defraggable inodes */
3569 btrfs_cleanup_defrag_inodes(fs_info); 3569 btrfs_cleanup_defrag_inodes(fs_info);
3570 3570
3571 if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3571 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
3572 ret = btrfs_commit_super(root); 3572 ret = btrfs_commit_super(root);
3573 if (ret) 3573 if (ret)
3574 btrfs_err(root->fs_info, "commit super ret %d", ret); 3574 btrfs_err(root->fs_info, "commit super ret %d", ret);
3575 } 3575 }
3576 3576
3577 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 3577 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3578 btrfs_error_commit_super(root); 3578 btrfs_error_commit_super(root);
3579 3579
3580 kthread_stop(fs_info->transaction_kthread); 3580 kthread_stop(fs_info->transaction_kthread);
3581 kthread_stop(fs_info->cleaner_kthread); 3581 kthread_stop(fs_info->cleaner_kthread);
3582 3582
3583 fs_info->closing = 2; 3583 fs_info->closing = 2;
3584 smp_mb(); 3584 smp_mb();
3585 3585
3586 btrfs_free_qgroup_config(root->fs_info); 3586 btrfs_free_qgroup_config(root->fs_info);
3587 3587
3588 if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 3588 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
3589 btrfs_info(root->fs_info, "at unmount delalloc count %lld", 3589 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
3590 percpu_counter_sum(&fs_info->delalloc_bytes)); 3590 percpu_counter_sum(&fs_info->delalloc_bytes));
3591 } 3591 }
3592 3592
3593 btrfs_sysfs_remove_one(fs_info); 3593 btrfs_sysfs_remove_one(fs_info);
3594 3594
3595 del_fs_roots(fs_info); 3595 del_fs_roots(fs_info);
3596 3596
3597 btrfs_put_block_group_cache(fs_info); 3597 btrfs_put_block_group_cache(fs_info);
3598 3598
3599 btrfs_free_block_groups(fs_info); 3599 btrfs_free_block_groups(fs_info);
3600 3600
3601 /* 3601 /*
3602 * we must make sure there is not any read request to 3602 * we must make sure there is not any read request to
3603 * submit after we stopping all workers. 3603 * submit after we stopping all workers.
3604 */ 3604 */
3605 invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3605 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
3606 btrfs_stop_all_workers(fs_info); 3606 btrfs_stop_all_workers(fs_info);
3607 3607
3608 free_root_pointers(fs_info, 1); 3608 free_root_pointers(fs_info, 1);
3609 3609
3610 iput(fs_info->btree_inode); 3610 iput(fs_info->btree_inode);
3611 3611
3612 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3612 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3613 if (btrfs_test_opt(root, CHECK_INTEGRITY)) 3613 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3614 btrfsic_unmount(root, fs_info->fs_devices); 3614 btrfsic_unmount(root, fs_info->fs_devices);
3615 #endif 3615 #endif
3616 3616
3617 btrfs_close_devices(fs_info->fs_devices); 3617 btrfs_close_devices(fs_info->fs_devices);
3618 btrfs_mapping_tree_free(&fs_info->mapping_tree); 3618 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3619 3619
3620 percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3620 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
3621 percpu_counter_destroy(&fs_info->delalloc_bytes); 3621 percpu_counter_destroy(&fs_info->delalloc_bytes);
3622 bdi_destroy(&fs_info->bdi); 3622 bdi_destroy(&fs_info->bdi);
3623 cleanup_srcu_struct(&fs_info->subvol_srcu); 3623 cleanup_srcu_struct(&fs_info->subvol_srcu);
3624 3624
3625 btrfs_free_stripe_hash_table(fs_info); 3625 btrfs_free_stripe_hash_table(fs_info);
3626 3626
3627 btrfs_free_block_rsv(root, root->orphan_block_rsv); 3627 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3628 root->orphan_block_rsv = NULL; 3628 root->orphan_block_rsv = NULL;
3629 3629
3630 return 0; 3630 return 0;
3631 } 3631 }
3632 3632
3633 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3633 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3634 int atomic) 3634 int atomic)
3635 { 3635 {
3636 int ret; 3636 int ret;
3637 struct inode *btree_inode = buf->pages[0]->mapping->host; 3637 struct inode *btree_inode = buf->pages[0]->mapping->host;
3638 3638
3639 ret = extent_buffer_uptodate(buf); 3639 ret = extent_buffer_uptodate(buf);
3640 if (!ret) 3640 if (!ret)
3641 return ret; 3641 return ret;
3642 3642
3643 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3643 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
3644 parent_transid, atomic); 3644 parent_transid, atomic);
3645 if (ret == -EAGAIN) 3645 if (ret == -EAGAIN)
3646 return ret; 3646 return ret;
3647 return !ret; 3647 return !ret;
3648 } 3648 }
3649 3649
3650 int btrfs_set_buffer_uptodate(struct extent_buffer *buf) 3650 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3651 { 3651 {
3652 return set_extent_buffer_uptodate(buf); 3652 return set_extent_buffer_uptodate(buf);
3653 } 3653 }
3654 3654
3655 void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 3655 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3656 { 3656 {
3657 struct btrfs_root *root; 3657 struct btrfs_root *root;
3658 u64 transid = btrfs_header_generation(buf); 3658 u64 transid = btrfs_header_generation(buf);
3659 int was_dirty; 3659 int was_dirty;
3660 3660
3661 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 3661 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3662 /* 3662 /*
3663 * This is a fast path so only do this check if we have sanity tests 3663 * This is a fast path so only do this check if we have sanity tests
3664 * enabled. Normal people shouldn't be marking dummy buffers as dirty 3664 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3665 * outside of the sanity tests. 3665 * outside of the sanity tests.
3666 */ 3666 */
3667 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 3667 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3668 return; 3668 return;
3669 #endif 3669 #endif
3670 root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3670 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
3671 btrfs_assert_tree_locked(buf); 3671 btrfs_assert_tree_locked(buf);
3672 if (transid != root->fs_info->generation) 3672 if (transid != root->fs_info->generation)
3673 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3673 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
3674 "found %llu running %llu\n", 3674 "found %llu running %llu\n",
3675 buf->start, transid, root->fs_info->generation); 3675 buf->start, transid, root->fs_info->generation);
3676 was_dirty = set_extent_buffer_dirty(buf); 3676 was_dirty = set_extent_buffer_dirty(buf);
3677 if (!was_dirty) 3677 if (!was_dirty)
3678 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 3678 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3679 buf->len, 3679 buf->len,
3680 root->fs_info->dirty_metadata_batch); 3680 root->fs_info->dirty_metadata_batch);
3681 } 3681 }
3682 3682
3683 static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 3683 static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3684 int flush_delayed) 3684 int flush_delayed)
3685 { 3685 {
3686 /* 3686 /*
3687 * looks as though older kernels can get into trouble with 3687 * looks as though older kernels can get into trouble with
3688 * this code, they end up stuck in balance_dirty_pages forever 3688 * this code, they end up stuck in balance_dirty_pages forever
3689 */ 3689 */
3690 int ret; 3690 int ret;
3691 3691
3692 if (current->flags & PF_MEMALLOC) 3692 if (current->flags & PF_MEMALLOC)
3693 return; 3693 return;
3694 3694
3695 if (flush_delayed) 3695 if (flush_delayed)
3696 btrfs_balance_delayed_items(root); 3696 btrfs_balance_delayed_items(root);
3697 3697
3698 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 3698 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3699 BTRFS_DIRTY_METADATA_THRESH); 3699 BTRFS_DIRTY_METADATA_THRESH);
3700 if (ret > 0) { 3700 if (ret > 0) {
3701 balance_dirty_pages_ratelimited( 3701 balance_dirty_pages_ratelimited(
3702 root->fs_info->btree_inode->i_mapping); 3702 root->fs_info->btree_inode->i_mapping);
3703 } 3703 }
3704 return; 3704 return;
3705 } 3705 }
3706 3706
3707 void btrfs_btree_balance_dirty(struct btrfs_root *root) 3707 void btrfs_btree_balance_dirty(struct btrfs_root *root)
3708 { 3708 {
3709 __btrfs_btree_balance_dirty(root, 1); 3709 __btrfs_btree_balance_dirty(root, 1);
3710 } 3710 }
3711 3711
3712 void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 3712 void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3713 { 3713 {
3714 __btrfs_btree_balance_dirty(root, 0); 3714 __btrfs_btree_balance_dirty(root, 0);
3715 } 3715 }
3716 3716
3717 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 3717 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
3718 { 3718 {
3719 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3719 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
3720 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 3720 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
3721 } 3721 }
3722 3722
3723 static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 3723 static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
3724 int read_only) 3724 int read_only)
3725 { 3725 {
3726 /* 3726 /*
3727 * Placeholder for checks 3727 * Placeholder for checks
3728 */ 3728 */
3729 return 0; 3729 return 0;
3730 } 3730 }
3731 3731
3732 static void btrfs_error_commit_super(struct btrfs_root *root) 3732 static void btrfs_error_commit_super(struct btrfs_root *root)
3733 { 3733 {
3734 mutex_lock(&root->fs_info->cleaner_mutex); 3734 mutex_lock(&root->fs_info->cleaner_mutex);
3735 btrfs_run_delayed_iputs(root); 3735 btrfs_run_delayed_iputs(root);
3736 mutex_unlock(&root->fs_info->cleaner_mutex); 3736 mutex_unlock(&root->fs_info->cleaner_mutex);
3737 3737
3738 down_write(&root->fs_info->cleanup_work_sem); 3738 down_write(&root->fs_info->cleanup_work_sem);
3739 up_write(&root->fs_info->cleanup_work_sem); 3739 up_write(&root->fs_info->cleanup_work_sem);
3740 3740
3741 /* cleanup FS via transaction */ 3741 /* cleanup FS via transaction */
3742 btrfs_cleanup_transaction(root); 3742 btrfs_cleanup_transaction(root);
3743 } 3743 }
3744 3744
3745 static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, 3745 static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3746 struct btrfs_root *root) 3746 struct btrfs_root *root)
3747 { 3747 {
3748 struct btrfs_inode *btrfs_inode; 3748 struct btrfs_inode *btrfs_inode;
3749 struct list_head splice; 3749 struct list_head splice;
3750 3750
3751 INIT_LIST_HEAD(&splice); 3751 INIT_LIST_HEAD(&splice);
3752 3752
3753 mutex_lock(&root->fs_info->ordered_operations_mutex); 3753 mutex_lock(&root->fs_info->ordered_operations_mutex);
3754 spin_lock(&root->fs_info->ordered_root_lock); 3754 spin_lock(&root->fs_info->ordered_root_lock);
3755 3755
3756 list_splice_init(&t->ordered_operations, &splice); 3756 list_splice_init(&t->ordered_operations, &splice);
3757 while (!list_empty(&splice)) { 3757 while (!list_empty(&splice)) {
3758 btrfs_inode = list_entry(splice.next, struct btrfs_inode, 3758 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3759 ordered_operations); 3759 ordered_operations);
3760 3760
3761 list_del_init(&btrfs_inode->ordered_operations); 3761 list_del_init(&btrfs_inode->ordered_operations);
3762 spin_unlock(&root->fs_info->ordered_root_lock); 3762 spin_unlock(&root->fs_info->ordered_root_lock);
3763 3763
3764 btrfs_invalidate_inodes(btrfs_inode->root); 3764 btrfs_invalidate_inodes(btrfs_inode->root);
3765 3765
3766 spin_lock(&root->fs_info->ordered_root_lock); 3766 spin_lock(&root->fs_info->ordered_root_lock);
3767 } 3767 }
3768 3768
3769 spin_unlock(&root->fs_info->ordered_root_lock); 3769 spin_unlock(&root->fs_info->ordered_root_lock);
3770 mutex_unlock(&root->fs_info->ordered_operations_mutex); 3770 mutex_unlock(&root->fs_info->ordered_operations_mutex);
3771 } 3771 }
3772 3772
3773 static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 3773 static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
3774 { 3774 {
3775 struct btrfs_ordered_extent *ordered; 3775 struct btrfs_ordered_extent *ordered;
3776 3776
3777 spin_lock(&root->ordered_extent_lock); 3777 spin_lock(&root->ordered_extent_lock);
3778 /* 3778 /*
3779 * This will just short circuit the ordered completion stuff which will 3779 * This will just short circuit the ordered completion stuff which will
3780 * make sure the ordered extent gets properly cleaned up. 3780 * make sure the ordered extent gets properly cleaned up.
3781 */ 3781 */
3782 list_for_each_entry(ordered, &root->ordered_extents, 3782 list_for_each_entry(ordered, &root->ordered_extents,
3783 root_extent_list) 3783 root_extent_list)
3784 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 3784 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
3785 spin_unlock(&root->ordered_extent_lock); 3785 spin_unlock(&root->ordered_extent_lock);
3786 } 3786 }
3787 3787
3788 static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 3788 static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3789 { 3789 {
3790 struct btrfs_root *root; 3790 struct btrfs_root *root;
3791 struct list_head splice; 3791 struct list_head splice;
3792 3792
3793 INIT_LIST_HEAD(&splice); 3793 INIT_LIST_HEAD(&splice);
3794 3794
3795 spin_lock(&fs_info->ordered_root_lock); 3795 spin_lock(&fs_info->ordered_root_lock);
3796 list_splice_init(&fs_info->ordered_roots, &splice); 3796 list_splice_init(&fs_info->ordered_roots, &splice);
3797 while (!list_empty(&splice)) { 3797 while (!list_empty(&splice)) {
3798 root = list_first_entry(&splice, struct btrfs_root, 3798 root = list_first_entry(&splice, struct btrfs_root,
3799 ordered_root); 3799 ordered_root);
3800 list_move_tail(&root->ordered_root, 3800 list_move_tail(&root->ordered_root,
3801 &fs_info->ordered_roots); 3801 &fs_info->ordered_roots);
3802 3802
3803 btrfs_destroy_ordered_extents(root); 3803 btrfs_destroy_ordered_extents(root);
3804 3804
3805 cond_resched_lock(&fs_info->ordered_root_lock); 3805 cond_resched_lock(&fs_info->ordered_root_lock);
3806 } 3806 }
3807 spin_unlock(&fs_info->ordered_root_lock); 3807 spin_unlock(&fs_info->ordered_root_lock);
3808 } 3808 }
3809 3809
3810 static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 3810 static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3811 struct btrfs_root *root) 3811 struct btrfs_root *root)
3812 { 3812 {
3813 struct rb_node *node; 3813 struct rb_node *node;
3814 struct btrfs_delayed_ref_root *delayed_refs; 3814 struct btrfs_delayed_ref_root *delayed_refs;
3815 struct btrfs_delayed_ref_node *ref; 3815 struct btrfs_delayed_ref_node *ref;
3816 int ret = 0; 3816 int ret = 0;
3817 3817
3818 delayed_refs = &trans->delayed_refs; 3818 delayed_refs = &trans->delayed_refs;
3819 3819
3820 spin_lock(&delayed_refs->lock); 3820 spin_lock(&delayed_refs->lock);
3821 if (atomic_read(&delayed_refs->num_entries) == 0) { 3821 if (atomic_read(&delayed_refs->num_entries) == 0) {
3822 spin_unlock(&delayed_refs->lock); 3822 spin_unlock(&delayed_refs->lock);
3823 btrfs_info(root->fs_info, "delayed_refs has NO entry"); 3823 btrfs_info(root->fs_info, "delayed_refs has NO entry");
3824 return ret; 3824 return ret;
3825 } 3825 }
3826 3826
3827 while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 3827 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3828 struct btrfs_delayed_ref_head *head; 3828 struct btrfs_delayed_ref_head *head;
3829 bool pin_bytes = false; 3829 bool pin_bytes = false;
3830 3830
3831 head = rb_entry(node, struct btrfs_delayed_ref_head, 3831 head = rb_entry(node, struct btrfs_delayed_ref_head,
3832 href_node); 3832 href_node);
3833 if (!mutex_trylock(&head->mutex)) { 3833 if (!mutex_trylock(&head->mutex)) {
3834 atomic_inc(&head->node.refs); 3834 atomic_inc(&head->node.refs);
3835 spin_unlock(&delayed_refs->lock); 3835 spin_unlock(&delayed_refs->lock);
3836 3836
3837 mutex_lock(&head->mutex); 3837 mutex_lock(&head->mutex);
3838 mutex_unlock(&head->mutex); 3838 mutex_unlock(&head->mutex);
3839 btrfs_put_delayed_ref(&head->node); 3839 btrfs_put_delayed_ref(&head->node);
3840 spin_lock(&delayed_refs->lock); 3840 spin_lock(&delayed_refs->lock);
3841 continue; 3841 continue;
3842 } 3842 }
3843 spin_lock(&head->lock); 3843 spin_lock(&head->lock);
3844 while ((node = rb_first(&head->ref_root)) != NULL) { 3844 while ((node = rb_first(&head->ref_root)) != NULL) {
3845 ref = rb_entry(node, struct btrfs_delayed_ref_node, 3845 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3846 rb_node); 3846 rb_node);
3847 ref->in_tree = 0; 3847 ref->in_tree = 0;
3848 rb_erase(&ref->rb_node, &head->ref_root); 3848 rb_erase(&ref->rb_node, &head->ref_root);
3849 atomic_dec(&delayed_refs->num_entries); 3849 atomic_dec(&delayed_refs->num_entries);
3850 btrfs_put_delayed_ref(ref); 3850 btrfs_put_delayed_ref(ref);
3851 } 3851 }
3852 if (head->must_insert_reserved) 3852 if (head->must_insert_reserved)
3853 pin_bytes = true; 3853 pin_bytes = true;
3854 btrfs_free_delayed_extent_op(head->extent_op); 3854 btrfs_free_delayed_extent_op(head->extent_op);
3855 delayed_refs->num_heads--; 3855 delayed_refs->num_heads--;
3856 if (head->processing == 0) 3856 if (head->processing == 0)
3857 delayed_refs->num_heads_ready--; 3857 delayed_refs->num_heads_ready--;
3858 atomic_dec(&delayed_refs->num_entries); 3858 atomic_dec(&delayed_refs->num_entries);
3859 head->node.in_tree = 0; 3859 head->node.in_tree = 0;
3860 rb_erase(&head->href_node, &delayed_refs->href_root); 3860 rb_erase(&head->href_node, &delayed_refs->href_root);
3861 spin_unlock(&head->lock); 3861 spin_unlock(&head->lock);
3862 spin_unlock(&delayed_refs->lock); 3862 spin_unlock(&delayed_refs->lock);
3863 mutex_unlock(&head->mutex); 3863 mutex_unlock(&head->mutex);
3864 3864
3865 if (pin_bytes) 3865 if (pin_bytes)
3866 btrfs_pin_extent(root, head->node.bytenr, 3866 btrfs_pin_extent(root, head->node.bytenr,
3867 head->node.num_bytes, 1); 3867 head->node.num_bytes, 1);
3868 btrfs_put_delayed_ref(&head->node); 3868 btrfs_put_delayed_ref(&head->node);
3869 cond_resched(); 3869 cond_resched();
3870 spin_lock(&delayed_refs->lock); 3870 spin_lock(&delayed_refs->lock);
3871 } 3871 }
3872 3872
3873 spin_unlock(&delayed_refs->lock); 3873 spin_unlock(&delayed_refs->lock);
3874 3874
3875 return ret; 3875 return ret;
3876 } 3876 }
3877 3877
3878 static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 3878 static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
3879 { 3879 {
3880 struct btrfs_inode *btrfs_inode; 3880 struct btrfs_inode *btrfs_inode;
3881 struct list_head splice; 3881 struct list_head splice;
3882 3882
3883 INIT_LIST_HEAD(&splice); 3883 INIT_LIST_HEAD(&splice);
3884 3884
3885 spin_lock(&root->delalloc_lock); 3885 spin_lock(&root->delalloc_lock);
3886 list_splice_init(&root->delalloc_inodes, &splice); 3886 list_splice_init(&root->delalloc_inodes, &splice);
3887 3887
3888 while (!list_empty(&splice)) { 3888 while (!list_empty(&splice)) {
3889 btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 3889 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3890 delalloc_inodes); 3890 delalloc_inodes);
3891 3891
3892 list_del_init(&btrfs_inode->delalloc_inodes); 3892 list_del_init(&btrfs_inode->delalloc_inodes);
3893 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 3893 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3894 &btrfs_inode->runtime_flags); 3894 &btrfs_inode->runtime_flags);
3895 spin_unlock(&root->delalloc_lock); 3895 spin_unlock(&root->delalloc_lock);
3896 3896
3897 btrfs_invalidate_inodes(btrfs_inode->root); 3897 btrfs_invalidate_inodes(btrfs_inode->root);
3898 3898
3899 spin_lock(&root->delalloc_lock); 3899 spin_lock(&root->delalloc_lock);
3900 } 3900 }
3901 3901
3902 spin_unlock(&root->delalloc_lock); 3902 spin_unlock(&root->delalloc_lock);
3903 } 3903 }
3904 3904
3905 static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 3905 static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3906 { 3906 {
3907 struct btrfs_root *root; 3907 struct btrfs_root *root;
3908 struct list_head splice; 3908 struct list_head splice;
3909 3909
3910 INIT_LIST_HEAD(&splice); 3910 INIT_LIST_HEAD(&splice);
3911 3911
3912 spin_lock(&fs_info->delalloc_root_lock); 3912 spin_lock(&fs_info->delalloc_root_lock);
3913 list_splice_init(&fs_info->delalloc_roots, &splice); 3913 list_splice_init(&fs_info->delalloc_roots, &splice);
3914 while (!list_empty(&splice)) { 3914 while (!list_empty(&splice)) {
3915 root = list_first_entry(&splice, struct btrfs_root, 3915 root = list_first_entry(&splice, struct btrfs_root,
3916 delalloc_root); 3916 delalloc_root);
3917 list_del_init(&root->delalloc_root); 3917 list_del_init(&root->delalloc_root);
3918 root = btrfs_grab_fs_root(root); 3918 root = btrfs_grab_fs_root(root);
3919 BUG_ON(!root); 3919 BUG_ON(!root);
3920 spin_unlock(&fs_info->delalloc_root_lock); 3920 spin_unlock(&fs_info->delalloc_root_lock);
3921 3921
3922 btrfs_destroy_delalloc_inodes(root); 3922 btrfs_destroy_delalloc_inodes(root);
3923 btrfs_put_fs_root(root); 3923 btrfs_put_fs_root(root);
3924 3924
3925 spin_lock(&fs_info->delalloc_root_lock); 3925 spin_lock(&fs_info->delalloc_root_lock);
3926 } 3926 }
3927 spin_unlock(&fs_info->delalloc_root_lock); 3927 spin_unlock(&fs_info->delalloc_root_lock);
3928 } 3928 }
3929 3929
3930 static int btrfs_destroy_marked_extents(struct btrfs_root *root, 3930 static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3931 struct extent_io_tree *dirty_pages, 3931 struct extent_io_tree *dirty_pages,
3932 int mark) 3932 int mark)
3933 { 3933 {
3934 int ret; 3934 int ret;
3935 struct extent_buffer *eb; 3935 struct extent_buffer *eb;
3936 u64 start = 0; 3936 u64 start = 0;
3937 u64 end; 3937 u64 end;
3938 3938
3939 while (1) { 3939 while (1) {
3940 ret = find_first_extent_bit(dirty_pages, start, &start, &end, 3940 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
3941 mark, NULL); 3941 mark, NULL);
3942 if (ret) 3942 if (ret)
3943 break; 3943 break;
3944 3944
3945 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); 3945 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3946 while (start <= end) { 3946 while (start <= end) {
3947 eb = btrfs_find_tree_block(root, start, 3947 eb = btrfs_find_tree_block(root, start,
3948 root->leafsize); 3948 root->leafsize);
3949 start += root->leafsize; 3949 start += root->leafsize;
3950 if (!eb) 3950 if (!eb)
3951 continue; 3951 continue;
3952 wait_on_extent_buffer_writeback(eb); 3952 wait_on_extent_buffer_writeback(eb);
3953 3953
3954 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 3954 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3955 &eb->bflags)) 3955 &eb->bflags))
3956 clear_extent_buffer_dirty(eb); 3956 clear_extent_buffer_dirty(eb);
3957 free_extent_buffer_stale(eb); 3957 free_extent_buffer_stale(eb);
3958 } 3958 }
3959 } 3959 }
3960 3960
3961 return ret; 3961 return ret;
3962 } 3962 }
3963 3963
3964 static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 3964 static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3965 struct extent_io_tree *pinned_extents) 3965 struct extent_io_tree *pinned_extents)
3966 { 3966 {
3967 struct extent_io_tree *unpin; 3967 struct extent_io_tree *unpin;
3968 u64 start; 3968 u64 start;
3969 u64 end; 3969 u64 end;
3970 int ret; 3970 int ret;
3971 bool loop = true; 3971 bool loop = true;
3972 3972
3973 unpin = pinned_extents; 3973 unpin = pinned_extents;
3974 again: 3974 again:
3975 while (1) { 3975 while (1) {
3976 ret = find_first_extent_bit(unpin, 0, &start, &end, 3976 ret = find_first_extent_bit(unpin, 0, &start, &end,
3977 EXTENT_DIRTY, NULL); 3977 EXTENT_DIRTY, NULL);
3978 if (ret) 3978 if (ret)
3979 break; 3979 break;
3980 3980
3981 clear_extent_dirty(unpin, start, end, GFP_NOFS); 3981 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3982 btrfs_error_unpin_extent_range(root, start, end); 3982 btrfs_error_unpin_extent_range(root, start, end);
3983 cond_resched(); 3983 cond_resched();
3984 } 3984 }
3985 3985
3986 if (loop) { 3986 if (loop) {
3987 if (unpin == &root->fs_info->freed_extents[0]) 3987 if (unpin == &root->fs_info->freed_extents[0])
3988 unpin = &root->fs_info->freed_extents[1]; 3988 unpin = &root->fs_info->freed_extents[1];
3989 else 3989 else
3990 unpin = &root->fs_info->freed_extents[0]; 3990 unpin = &root->fs_info->freed_extents[0];
3991 loop = false; 3991 loop = false;
3992 goto again; 3992 goto again;
3993 } 3993 }
3994 3994
3995 return 0; 3995 return 0;
3996 } 3996 }
3997 3997
3998 void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 3998 void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3999 struct btrfs_root *root) 3999 struct btrfs_root *root)
4000 { 4000 {
4001 btrfs_destroy_ordered_operations(cur_trans, root); 4001 btrfs_destroy_ordered_operations(cur_trans, root);
4002 4002
4003 btrfs_destroy_delayed_refs(cur_trans, root); 4003 btrfs_destroy_delayed_refs(cur_trans, root);
4004 4004
4005 cur_trans->state = TRANS_STATE_COMMIT_START; 4005 cur_trans->state = TRANS_STATE_COMMIT_START;
4006 wake_up(&root->fs_info->transaction_blocked_wait); 4006 wake_up(&root->fs_info->transaction_blocked_wait);
4007 4007
4008 cur_trans->state = TRANS_STATE_UNBLOCKED; 4008 cur_trans->state = TRANS_STATE_UNBLOCKED;
4009 wake_up(&root->fs_info->transaction_wait); 4009 wake_up(&root->fs_info->transaction_wait);
4010 4010
4011 btrfs_destroy_delayed_inodes(root); 4011 btrfs_destroy_delayed_inodes(root);
4012 btrfs_assert_delayed_root_empty(root); 4012 btrfs_assert_delayed_root_empty(root);
4013 4013
4014 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 4014 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4015 EXTENT_DIRTY); 4015 EXTENT_DIRTY);
4016 btrfs_destroy_pinned_extent(root, 4016 btrfs_destroy_pinned_extent(root,
4017 root->fs_info->pinned_extents); 4017 root->fs_info->pinned_extents);
4018 4018
4019 cur_trans->state =TRANS_STATE_COMPLETED; 4019 cur_trans->state =TRANS_STATE_COMPLETED;
4020 wake_up(&cur_trans->commit_wait); 4020 wake_up(&cur_trans->commit_wait);
4021 4021
4022 /* 4022 /*
4023 memset(cur_trans, 0, sizeof(*cur_trans)); 4023 memset(cur_trans, 0, sizeof(*cur_trans));
4024 kmem_cache_free(btrfs_transaction_cachep, cur_trans); 4024 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4025 */ 4025 */
4026 } 4026 }
4027 4027
4028 static int btrfs_cleanup_transaction(struct btrfs_root *root) 4028 static int btrfs_cleanup_transaction(struct btrfs_root *root)
4029 { 4029 {
4030 struct btrfs_transaction *t; 4030 struct btrfs_transaction *t;
4031 4031
4032 mutex_lock(&root->fs_info->transaction_kthread_mutex); 4032 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4033 4033
4034 spin_lock(&root->fs_info->trans_lock); 4034 spin_lock(&root->fs_info->trans_lock);
4035 while (!list_empty(&root->fs_info->trans_list)) { 4035 while (!list_empty(&root->fs_info->trans_list)) {
4036 t = list_first_entry(&root->fs_info->trans_list, 4036 t = list_first_entry(&root->fs_info->trans_list,
4037 struct btrfs_transaction, list); 4037 struct btrfs_transaction, list);
4038 if (t->state >= TRANS_STATE_COMMIT_START) { 4038 if (t->state >= TRANS_STATE_COMMIT_START) {
4039 atomic_inc(&t->use_count); 4039 atomic_inc(&t->use_count);
4040 spin_unlock(&root->fs_info->trans_lock); 4040 spin_unlock(&root->fs_info->trans_lock);
4041 btrfs_wait_for_commit(root, t->transid); 4041 btrfs_wait_for_commit(root, t->transid);
4042 btrfs_put_transaction(t); 4042 btrfs_put_transaction(t);
4043 spin_lock(&root->fs_info->trans_lock); 4043 spin_lock(&root->fs_info->trans_lock);
4044 continue; 4044 continue;
4045 } 4045 }
4046 if (t == root->fs_info->running_transaction) { 4046 if (t == root->fs_info->running_transaction) {
4047 t->state = TRANS_STATE_COMMIT_DOING; 4047 t->state = TRANS_STATE_COMMIT_DOING;
4048 spin_unlock(&root->fs_info->trans_lock); 4048 spin_unlock(&root->fs_info->trans_lock);
4049 /* 4049 /*
4050 * We wait for 0 num_writers since we don't hold a trans 4050 * We wait for 0 num_writers since we don't hold a trans
4051 * handle open currently for this transaction. 4051 * handle open currently for this transaction.
4052 */ 4052 */
4053 wait_event(t->writer_wait, 4053 wait_event(t->writer_wait,
4054 atomic_read(&t->num_writers) == 0); 4054 atomic_read(&t->num_writers) == 0);
4055 } else { 4055 } else {
4056 spin_unlock(&root->fs_info->trans_lock); 4056 spin_unlock(&root->fs_info->trans_lock);
4057 } 4057 }
4058 btrfs_cleanup_one_transaction(t, root); 4058 btrfs_cleanup_one_transaction(t, root);
4059 4059
4060 spin_lock(&root->fs_info->trans_lock); 4060 spin_lock(&root->fs_info->trans_lock);
4061 if (t == root->fs_info->running_transaction) 4061 if (t == root->fs_info->running_transaction)
4062 root->fs_info->running_transaction = NULL; 4062 root->fs_info->running_transaction = NULL;
4063 list_del_init(&t->list); 4063 list_del_init(&t->list);
4064 spin_unlock(&root->fs_info->trans_lock); 4064 spin_unlock(&root->fs_info->trans_lock);
4065 4065
4066 btrfs_put_transaction(t); 4066 btrfs_put_transaction(t);
4067 trace_btrfs_transaction_commit(root); 4067 trace_btrfs_transaction_commit(root);
4068 spin_lock(&root->fs_info->trans_lock); 4068 spin_lock(&root->fs_info->trans_lock);
4069 } 4069 }
4070 spin_unlock(&root->fs_info->trans_lock); 4070 spin_unlock(&root->fs_info->trans_lock);
4071 btrfs_destroy_all_ordered_extents(root->fs_info); 4071 btrfs_destroy_all_ordered_extents(root->fs_info);
4072 btrfs_destroy_delayed_inodes(root); 4072 btrfs_destroy_delayed_inodes(root);
4073 btrfs_assert_delayed_root_empty(root); 4073 btrfs_assert_delayed_root_empty(root);
4074 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4074 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4075 btrfs_destroy_all_delalloc_inodes(root->fs_info); 4075 btrfs_destroy_all_delalloc_inodes(root->fs_info);
4076 mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4076 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
4077 4077
4078 return 0; 4078 return 0;
4079 } 4079 }
4080 4080
4081 static struct extent_io_ops btree_extent_io_ops = { 4081 static struct extent_io_ops btree_extent_io_ops = {
4082 .readpage_end_io_hook = btree_readpage_end_io_hook, 4082 .readpage_end_io_hook = btree_readpage_end_io_hook,
4083 .readpage_io_failed_hook = btree_io_failed_hook, 4083 .readpage_io_failed_hook = btree_io_failed_hook,
4084 .submit_bio_hook = btree_submit_bio_hook, 4084 .submit_bio_hook = btree_submit_bio_hook,
4085 /* note we're sharing with inode.c for the merge bio hook */ 4085 /* note we're sharing with inode.c for the merge bio hook */
4086 .merge_bio_hook = btrfs_merge_bio_hook, 4086 .merge_bio_hook = btrfs_merge_bio_hook,
4087 }; 4087 };
4088 4088