Btrfs: Better block record keeping, real mkfs
[linux-2.6-block.git] / fs / btrfs / disk-io.c
CommitLineData
eb60ceac
CM
1#define _XOPEN_SOURCE 500
2#include <stdio.h>
3#include <stdlib.h>
4#include <sys/types.h>
5#include <sys/stat.h>
6#include <fcntl.h>
7#include <unistd.h>
8#include "kerncompat.h"
9#include "radix-tree.h"
10#include "ctree.h"
11#include "disk-io.h"
e089f05c 12#include "transaction.h"
eb60ceac
CM
13
14static int allocated_blocks = 0;
ed2ff2cb 15int cache_max = 10000;
eb60ceac 16
234b63a0 17static int check_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
eb60ceac 18{
7518a238 19 if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
9a8dd150 20 BUG();
7518a238
CM
21 if (root->node && btrfs_header_parentid(&buf->node.header) !=
22 btrfs_header_parentid(&root->node->node.header))
9a8dd150
CM
23 BUG();
24 return 0;
eb60ceac
CM
25}
26
234b63a0 27static int free_some_buffers(struct btrfs_root *root)
ed2ff2cb
CM
28{
29 struct list_head *node, *next;
234b63a0 30 struct btrfs_buffer *b;
9f5fae2f 31 if (root->fs_info->cache_size < cache_max)
ed2ff2cb 32 return 0;
9f5fae2f 33 list_for_each_safe(node, next, &root->fs_info->cache) {
234b63a0 34 b = list_entry(node, struct btrfs_buffer, cache);
ed2ff2cb
CM
35 if (b->count == 1) {
36 BUG_ON(!list_empty(&b->dirty));
37 list_del_init(&b->cache);
234b63a0 38 btrfs_block_release(root, b);
9f5fae2f 39 if (root->fs_info->cache_size < cache_max)
77ce6846 40 break;
ed2ff2cb
CM
41 }
42 }
43 return 0;
44}
45
234b63a0 46struct btrfs_buffer *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
eb60ceac 47{
234b63a0 48 struct btrfs_buffer *buf;
eb60ceac 49 int ret;
123abc88
CM
50
51 buf = malloc(sizeof(struct btrfs_buffer) + root->blocksize);
eb60ceac
CM
52 if (!buf)
53 return buf;
54 allocated_blocks++;
55 buf->blocknr = blocknr;
ed2ff2cb
CM
56 buf->count = 2;
57 INIT_LIST_HEAD(&buf->dirty);
58 free_some_buffers(root);
eb60ceac 59 radix_tree_preload(GFP_KERNEL);
9f5fae2f 60 ret = radix_tree_insert(&root->fs_info->cache_radix, blocknr, buf);
eb60ceac 61 radix_tree_preload_end();
9f5fae2f
CM
62 list_add_tail(&buf->cache, &root->fs_info->cache);
63 root->fs_info->cache_size++;
eb60ceac
CM
64 if (ret) {
65 free(buf);
66 return NULL;
67 }
68 return buf;
69}
70
234b63a0 71struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr)
eb60ceac 72{
234b63a0 73 struct btrfs_buffer *buf;
9f5fae2f 74 buf = radix_tree_lookup(&root->fs_info->cache_radix, blocknr);
9a8dd150
CM
75 if (buf) {
76 buf->count++;
77 } else {
78 buf = alloc_tree_block(root, blocknr);
79 if (!buf) {
80 BUG();
81 return NULL;
82 }
eb60ceac 83 }
eb60ceac
CM
84 return buf;
85}
86
234b63a0 87struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
eb60ceac 88{
123abc88 89 loff_t offset = blocknr * root->blocksize;
234b63a0 90 struct btrfs_buffer *buf;
eb60ceac
CM
91 int ret;
92
9f5fae2f 93 buf = radix_tree_lookup(&root->fs_info->cache_radix, blocknr);
eb60ceac
CM
94 if (buf) {
95 buf->count++;
9a8dd150
CM
96 } else {
97 buf = alloc_tree_block(root, blocknr);
98 if (!buf)
99 return NULL;
9f5fae2f
CM
100 ret = pread(root->fs_info->fp, &buf->node, root->blocksize,
101 offset);
123abc88 102 if (ret != root->blocksize) {
9a8dd150
CM
103 free(buf);
104 return NULL;
105 }
eb60ceac 106 }
9a8dd150 107 if (check_tree_block(root, buf))
cfaa7295 108 BUG();
eb60ceac
CM
109 return buf;
110}
111
e089f05c
CM
112int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
113 struct btrfs_buffer *buf)
ed2ff2cb
CM
114{
115 if (!list_empty(&buf->dirty))
116 return 0;
9f5fae2f 117 list_add_tail(&buf->dirty, &root->fs_info->trans);
ed2ff2cb
CM
118 buf->count++;
119 return 0;
120}
121
e089f05c
CM
122int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
123 struct btrfs_buffer *buf)
ed2ff2cb
CM
124{
125 if (!list_empty(&buf->dirty)) {
126 list_del_init(&buf->dirty);
234b63a0 127 btrfs_block_release(root, buf);
ed2ff2cb
CM
128 }
129 return 0;
130}
131
e089f05c
CM
132int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
133 struct btrfs_buffer *buf)
eb60ceac
CM
134{
135 u64 blocknr = buf->blocknr;
123abc88 136 loff_t offset = blocknr * root->blocksize;
eb60ceac
CM
137 int ret;
138
7518a238 139 if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
eb60ceac 140 BUG();
9f5fae2f 141 ret = pwrite(root->fs_info->fp, &buf->node, root->blocksize, offset);
123abc88 142 if (ret != root->blocksize)
eb60ceac 143 return ret;
eb60ceac
CM
144 return 0;
145}
146
e089f05c
CM
147static int __commit_transaction(struct btrfs_trans_handle *trans, struct
148 btrfs_root *root)
ed2ff2cb 149{
234b63a0 150 struct btrfs_buffer *b;
ed2ff2cb
CM
151 int ret = 0;
152 int wret;
9f5fae2f
CM
153 while(!list_empty(&root->fs_info->trans)) {
154 b = list_entry(root->fs_info->trans.next, struct btrfs_buffer,
155 dirty);
ed2ff2cb 156 list_del_init(&b->dirty);
e089f05c 157 wret = write_tree_block(trans, root, b);
ed2ff2cb
CM
158 if (wret)
159 ret = wret;
234b63a0 160 btrfs_block_release(root, b);
ed2ff2cb
CM
161 }
162 return ret;
163}
164
9f5fae2f
CM
165static int commit_tree_roots(struct btrfs_trans_handle *trans,
166 struct btrfs_fs_info *fs_info)
3768f368
CM
167{
168 int ret;
169 u64 old_extent_block;
9f5fae2f
CM
170 struct btrfs_root *tree_root = fs_info->tree_root;
171 struct btrfs_root *extent_root = fs_info->extent_root;
172 struct btrfs_root *inode_root = fs_info->inode_root;
173
174 btrfs_set_root_blocknr(&inode_root->root_item,
175 inode_root->node->blocknr);
176 ret = btrfs_update_root(trans, tree_root,
177 &inode_root->root_key,
178 &inode_root->root_item);
179 BUG_ON(ret);
3768f368
CM
180 while(1) {
181 old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
182 if (old_extent_block == extent_root->node->blocknr)
183 break;
184 btrfs_set_root_blocknr(&extent_root->root_item,
185 extent_root->node->blocknr);
e089f05c 186 ret = btrfs_update_root(trans, tree_root,
3768f368
CM
187 &extent_root->root_key,
188 &extent_root->root_item);
189 BUG_ON(ret);
190 }
3768f368
CM
191 return 0;
192}
193
e089f05c
CM
194int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
195 btrfs_root *root, struct btrfs_super_block *s)
ed2ff2cb 196{
a28ec197 197 int ret = 0;
3768f368
CM
198 struct btrfs_buffer *snap = root->commit_root;
199 struct btrfs_key snap_key;
a28ec197 200
3768f368
CM
201 if (root->commit_root == root->node)
202 return 0;
203
204 memcpy(&snap_key, &root->root_key, sizeof(snap_key));
205 root->root_key.offset++;
206
207 btrfs_set_root_blocknr(&root->root_item, root->node->blocknr);
9f5fae2f
CM
208 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
209 &root->root_key, &root->root_item);
210 BUG_ON(ret);
211
212 ret = commit_tree_roots(trans, root->fs_info);
3768f368
CM
213 BUG_ON(ret);
214
9f5fae2f 215 ret = __commit_transaction(trans, root);
3768f368
CM
216 BUG_ON(ret);
217
e089f05c 218 write_ctree_super(trans, root, s);
9f5fae2f
CM
219 btrfs_finish_extent_commit(trans, root->fs_info->extent_root);
220 btrfs_finish_extent_commit(trans, root->fs_info->tree_root);
3768f368
CM
221
222 root->commit_root = root->node;
223 root->node->count++;
e089f05c 224 ret = btrfs_drop_snapshot(trans, root, snap);
3768f368
CM
225 BUG_ON(ret);
226
9f5fae2f 227 ret = btrfs_del_root(trans, root->fs_info->tree_root, &snap_key);
3768f368 228 BUG_ON(ret);
293ffd5f 229 root->fs_info->generation = root->root_key.offset + 1;
3768f368 230
ed2ff2cb
CM
231 return ret;
232}
233
123abc88 234static int __setup_root(struct btrfs_super_block *super,
9f5fae2f
CM
235 struct btrfs_root *root,
236 struct btrfs_fs_info *fs_info,
237 u64 objectid, int fp)
d97e63b6 238{
cfaa7295 239 root->node = NULL;
a28ec197 240 root->commit_root = NULL;
123abc88
CM
241 root->blocksize = btrfs_super_blocksize(super);
242 root->ref_cows = 0;
9f5fae2f 243 root->fs_info = fs_info;
3768f368
CM
244 memset(&root->root_key, 0, sizeof(root->root_key));
245 memset(&root->root_item, 0, sizeof(root->root_item));
246 return 0;
247}
248
123abc88 249static int find_and_setup_root(struct btrfs_super_block *super,
9f5fae2f
CM
250 struct btrfs_root *tree_root,
251 struct btrfs_fs_info *fs_info,
252 u64 objectid,
123abc88 253 struct btrfs_root *root, int fp)
3768f368
CM
254{
255 int ret;
256
9f5fae2f 257 __setup_root(super, root, fs_info, objectid, fp);
3768f368
CM
258 ret = btrfs_find_last_root(tree_root, objectid,
259 &root->root_item, &root->root_key);
260 BUG_ON(ret);
261
262 root->node = read_tree_block(root,
263 btrfs_root_blocknr(&root->root_item));
3768f368 264 BUG_ON(!root->node);
d97e63b6
CM
265 return 0;
266}
267
234b63a0 268struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
eb60ceac 269{
234b63a0
CM
270 struct btrfs_root *root = malloc(sizeof(struct btrfs_root));
271 struct btrfs_root *extent_root = malloc(sizeof(struct btrfs_root));
3768f368 272 struct btrfs_root *tree_root = malloc(sizeof(struct btrfs_root));
9f5fae2f
CM
273 struct btrfs_root *inode_root = malloc(sizeof(struct btrfs_root));
274 struct btrfs_fs_info *fs_info = malloc(sizeof(*fs_info));
eb60ceac 275 int fp;
eb60ceac
CM
276 int ret;
277
c673024a 278 fp = open(filename, O_CREAT | O_RDWR, 0600);
eb60ceac
CM
279 if (fp < 0) {
280 free(root);
281 return NULL;
282 }
9f5fae2f
CM
283 INIT_RADIX_TREE(&fs_info->cache_radix, GFP_KERNEL);
284 INIT_RADIX_TREE(&fs_info->pinned_radix, GFP_KERNEL);
285 INIT_LIST_HEAD(&fs_info->trans);
286 INIT_LIST_HEAD(&fs_info->cache);
287 fs_info->cache_size = 0;
288 fs_info->fp = fp;
289 fs_info->running_transaction = NULL;
290 fs_info->fs_root = root;
291 fs_info->tree_root = tree_root;
292 fs_info->extent_root = extent_root;
293 fs_info->inode_root = inode_root;
294 fs_info->last_inode_alloc = 0;
295 fs_info->last_inode_alloc_dirid = 0;
1261ec42 296 fs_info->disk_super = super;
9f5fae2f
CM
297 memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
298 memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
3768f368 299
234b63a0 300 ret = pread(fp, super, sizeof(struct btrfs_super_block),
123abc88 301 BTRFS_SUPER_INFO_OFFSET);
3768f368 302 if (ret == 0 || btrfs_super_root(super) == 0) {
1261ec42
CM
303 BUG();
304 return NULL;
d97e63b6
CM
305 }
306 BUG_ON(ret < 0);
3768f368 307
9f5fae2f 308 __setup_root(super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID, fp);
3768f368
CM
309 tree_root->node = read_tree_block(tree_root, btrfs_super_root(super));
310 BUG_ON(!tree_root->node);
311
9f5fae2f
CM
312 ret = find_and_setup_root(super, tree_root, fs_info,
313 BTRFS_EXTENT_TREE_OBJECTID, extent_root, fp);
3768f368
CM
314 BUG_ON(ret);
315
9f5fae2f
CM
316 ret = find_and_setup_root(super, tree_root, fs_info,
317 BTRFS_INODE_MAP_OBJECTID, inode_root, fp);
318 BUG_ON(ret);
319
320 ret = find_and_setup_root(super, tree_root, fs_info,
321 BTRFS_FS_TREE_OBJECTID, root, fp);
3768f368
CM
322 BUG_ON(ret);
323
a28ec197
CM
324 root->commit_root = root->node;
325 root->node->count++;
3768f368 326 root->ref_cows = 1;
293ffd5f 327 root->fs_info->generation = root->root_key.offset + 1;
eb60ceac
CM
328 return root;
329}
330
e089f05c
CM
331int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
332 *root, struct btrfs_super_block *s)
eb60ceac
CM
333{
334 int ret;
9f5fae2f
CM
335 btrfs_set_super_root(s, root->fs_info->tree_root->node->blocknr);
336 ret = pwrite(root->fs_info->fp, s, sizeof(*s),
123abc88 337 BTRFS_SUPER_INFO_OFFSET);
cfaa7295
CM
338 if (ret != sizeof(*s)) {
339 fprintf(stderr, "failed to write new super block err %d\n", ret);
eb60ceac 340 return ret;
cfaa7295
CM
341 }
342 return 0;
343}
344
234b63a0 345static int drop_cache(struct btrfs_root *root)
ed2ff2cb 346{
9f5fae2f
CM
347 while(!list_empty(&root->fs_info->cache)) {
348 struct btrfs_buffer *b = list_entry(root->fs_info->cache.next,
349 struct btrfs_buffer,
350 cache);
ed2ff2cb 351 list_del_init(&b->cache);
234b63a0 352 btrfs_block_release(root, b);
ed2ff2cb
CM
353 }
354 return 0;
355}
234b63a0 356int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s)
cfaa7295 357{
3768f368 358 int ret;
e089f05c
CM
359 struct btrfs_trans_handle *trans;
360
9f5fae2f 361 trans = root->fs_info->running_transaction;
e089f05c 362 btrfs_commit_transaction(trans, root, s);
9f5fae2f
CM
363 ret = commit_tree_roots(trans, root->fs_info);
364 BUG_ON(ret);
365 ret = __commit_transaction(trans, root);
3768f368 366 BUG_ON(ret);
e089f05c 367 write_ctree_super(trans, root, s);
ed2ff2cb 368 drop_cache(root);
9f5fae2f 369 BUG_ON(!list_empty(&root->fs_info->trans));
ed2ff2cb 370
9f5fae2f 371 close(root->fs_info->fp);
cfaa7295 372 if (root->node)
234b63a0 373 btrfs_block_release(root, root->node);
9f5fae2f
CM
374 if (root->fs_info->extent_root->node)
375 btrfs_block_release(root->fs_info->extent_root,
376 root->fs_info->extent_root->node);
377 if (root->fs_info->inode_root->node)
378 btrfs_block_release(root->fs_info->inode_root,
379 root->fs_info->inode_root->node);
380 if (root->fs_info->tree_root->node)
381 btrfs_block_release(root->fs_info->tree_root,
382 root->fs_info->tree_root->node);
234b63a0 383 btrfs_block_release(root, root->commit_root);
cfaa7295
CM
384 free(root);
385 printf("on close %d blocks are allocated\n", allocated_blocks);
eb60ceac
CM
386 return 0;
387}
388
234b63a0 389void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf)
eb60ceac
CM
390{
391 buf->count--;
cfaa7295
CM
392 if (buf->count < 0)
393 BUG();
eb60ceac 394 if (buf->count == 0) {
02217ed2
CM
395 BUG_ON(!list_empty(&buf->cache));
396 BUG_ON(!list_empty(&buf->dirty));
9f5fae2f
CM
397 if (!radix_tree_lookup(&root->fs_info->cache_radix,
398 buf->blocknr))
eb60ceac 399 BUG();
9f5fae2f 400 radix_tree_delete(&root->fs_info->cache_radix, buf->blocknr);
eb60ceac
CM
401 memset(buf, 0, sizeof(*buf));
402 free(buf);
403 BUG_ON(allocated_blocks == 0);
404 allocated_blocks--;
9f5fae2f
CM
405 BUG_ON(root->fs_info->cache_size == 0);
406 root->fs_info->cache_size--;
eb60ceac
CM
407 }
408}
409