Commit | Line | Data |
---|---|---|
0b86a832 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | #include <linux/sched.h> | |
19 | #include <linux/bio.h> | |
20 | #include "ctree.h" | |
21 | #include "extent_map.h" | |
22 | #include "disk-io.h" | |
23 | #include "transaction.h" | |
24 | #include "print-tree.h" | |
25 | #include "volumes.h" | |
26 | ||
27 | struct map_lookup { | |
28 | struct btrfs_device *dev; | |
29 | u64 physical; | |
30 | }; | |
31 | ||
32 | /* | |
33 | * this uses a pretty simple search, the expectation is that it is | |
34 | * called very infrequently and that a given device has a small number | |
35 | * of extents | |
36 | */ | |
37 | static int find_free_dev_extent(struct btrfs_trans_handle *trans, | |
38 | struct btrfs_device *device, | |
39 | struct btrfs_path *path, | |
40 | u64 num_bytes, u64 *start) | |
41 | { | |
42 | struct btrfs_key key; | |
43 | struct btrfs_root *root = device->dev_root; | |
44 | struct btrfs_dev_extent *dev_extent = NULL; | |
45 | u64 hole_size = 0; | |
46 | u64 last_byte = 0; | |
47 | u64 search_start = 0; | |
48 | u64 search_end = device->total_bytes; | |
49 | int ret; | |
50 | int slot = 0; | |
51 | int start_found; | |
52 | struct extent_buffer *l; | |
53 | ||
54 | start_found = 0; | |
55 | path->reada = 2; | |
56 | ||
57 | /* FIXME use last free of some kind */ | |
58 | ||
59 | key.objectid = device->devid; | |
60 | key.offset = search_start; | |
61 | key.type = BTRFS_DEV_EXTENT_KEY; | |
62 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
63 | if (ret < 0) | |
64 | goto error; | |
65 | ret = btrfs_previous_item(root, path, 0, key.type); | |
66 | if (ret < 0) | |
67 | goto error; | |
68 | l = path->nodes[0]; | |
69 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
70 | while (1) { | |
71 | l = path->nodes[0]; | |
72 | slot = path->slots[0]; | |
73 | if (slot >= btrfs_header_nritems(l)) { | |
74 | ret = btrfs_next_leaf(root, path); | |
75 | if (ret == 0) | |
76 | continue; | |
77 | if (ret < 0) | |
78 | goto error; | |
79 | no_more_items: | |
80 | if (!start_found) { | |
81 | if (search_start >= search_end) { | |
82 | ret = -ENOSPC; | |
83 | goto error; | |
84 | } | |
85 | *start = search_start; | |
86 | start_found = 1; | |
87 | goto check_pending; | |
88 | } | |
89 | *start = last_byte > search_start ? | |
90 | last_byte : search_start; | |
91 | if (search_end <= *start) { | |
92 | ret = -ENOSPC; | |
93 | goto error; | |
94 | } | |
95 | goto check_pending; | |
96 | } | |
97 | btrfs_item_key_to_cpu(l, &key, slot); | |
98 | ||
99 | if (key.objectid < device->devid) | |
100 | goto next; | |
101 | ||
102 | if (key.objectid > device->devid) | |
103 | goto no_more_items; | |
104 | ||
105 | if (key.offset >= search_start && key.offset > last_byte && | |
106 | start_found) { | |
107 | if (last_byte < search_start) | |
108 | last_byte = search_start; | |
109 | hole_size = key.offset - last_byte; | |
110 | if (key.offset > last_byte && | |
111 | hole_size >= num_bytes) { | |
112 | *start = last_byte; | |
113 | goto check_pending; | |
114 | } | |
115 | } | |
116 | if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { | |
117 | goto next; | |
118 | } | |
119 | ||
120 | start_found = 1; | |
121 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
122 | last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); | |
123 | next: | |
124 | path->slots[0]++; | |
125 | cond_resched(); | |
126 | } | |
127 | check_pending: | |
128 | /* we have to make sure we didn't find an extent that has already | |
129 | * been allocated by the map tree or the original allocation | |
130 | */ | |
131 | btrfs_release_path(root, path); | |
132 | BUG_ON(*start < search_start); | |
133 | ||
6324fbf3 | 134 | if (*start + num_bytes > search_end) { |
0b86a832 CM |
135 | ret = -ENOSPC; |
136 | goto error; | |
137 | } | |
138 | /* check for pending inserts here */ | |
139 | return 0; | |
140 | ||
141 | error: | |
142 | btrfs_release_path(root, path); | |
143 | return ret; | |
144 | } | |
145 | ||
146 | int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, | |
147 | struct btrfs_device *device, | |
148 | u64 owner, u64 num_bytes, u64 *start) | |
149 | { | |
150 | int ret; | |
151 | struct btrfs_path *path; | |
152 | struct btrfs_root *root = device->dev_root; | |
153 | struct btrfs_dev_extent *extent; | |
154 | struct extent_buffer *leaf; | |
155 | struct btrfs_key key; | |
156 | ||
157 | path = btrfs_alloc_path(); | |
158 | if (!path) | |
159 | return -ENOMEM; | |
160 | ||
161 | ret = find_free_dev_extent(trans, device, path, num_bytes, start); | |
6324fbf3 | 162 | if (ret) { |
0b86a832 | 163 | goto err; |
6324fbf3 | 164 | } |
0b86a832 CM |
165 | |
166 | key.objectid = device->devid; | |
167 | key.offset = *start; | |
168 | key.type = BTRFS_DEV_EXTENT_KEY; | |
169 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
170 | sizeof(*extent)); | |
171 | BUG_ON(ret); | |
172 | ||
173 | leaf = path->nodes[0]; | |
174 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
175 | struct btrfs_dev_extent); | |
176 | btrfs_set_dev_extent_owner(leaf, extent, owner); | |
177 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); | |
178 | btrfs_mark_buffer_dirty(leaf); | |
179 | err: | |
180 | btrfs_free_path(path); | |
181 | return ret; | |
182 | } | |
183 | ||
184 | static int find_next_chunk(struct btrfs_root *root, u64 *objectid) | |
185 | { | |
186 | struct btrfs_path *path; | |
187 | int ret; | |
188 | struct btrfs_key key; | |
189 | struct btrfs_key found_key; | |
190 | ||
191 | path = btrfs_alloc_path(); | |
192 | BUG_ON(!path); | |
193 | ||
194 | key.objectid = (u64)-1; | |
195 | key.offset = (u64)-1; | |
196 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
197 | ||
198 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
199 | if (ret < 0) | |
200 | goto error; | |
201 | ||
202 | BUG_ON(ret == 0); | |
203 | ||
204 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
205 | if (ret) { | |
206 | *objectid = 0; | |
207 | } else { | |
208 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
209 | path->slots[0]); | |
210 | *objectid = found_key.objectid + found_key.offset; | |
211 | } | |
212 | ret = 0; | |
213 | error: | |
214 | btrfs_free_path(path); | |
215 | return ret; | |
216 | } | |
217 | ||
0b86a832 CM |
218 | static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, |
219 | u64 *objectid) | |
220 | { | |
221 | int ret; | |
222 | struct btrfs_key key; | |
223 | struct btrfs_key found_key; | |
224 | ||
225 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
226 | key.type = BTRFS_DEV_ITEM_KEY; | |
227 | key.offset = (u64)-1; | |
228 | ||
229 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
230 | if (ret < 0) | |
231 | goto error; | |
232 | ||
233 | BUG_ON(ret == 0); | |
234 | ||
235 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
236 | BTRFS_DEV_ITEM_KEY); | |
237 | if (ret) { | |
238 | *objectid = 1; | |
239 | } else { | |
240 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
241 | path->slots[0]); | |
242 | *objectid = found_key.offset + 1; | |
243 | } | |
244 | ret = 0; | |
245 | error: | |
246 | btrfs_release_path(root, path); | |
247 | return ret; | |
248 | } | |
249 | ||
250 | /* | |
251 | * the device information is stored in the chunk root | |
252 | * the btrfs_device struct should be fully filled in | |
253 | */ | |
254 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
255 | struct btrfs_root *root, | |
256 | struct btrfs_device *device) | |
257 | { | |
258 | int ret; | |
259 | struct btrfs_path *path; | |
260 | struct btrfs_dev_item *dev_item; | |
261 | struct extent_buffer *leaf; | |
262 | struct btrfs_key key; | |
263 | unsigned long ptr; | |
264 | u64 free_devid; | |
265 | ||
266 | root = root->fs_info->chunk_root; | |
267 | ||
268 | path = btrfs_alloc_path(); | |
269 | if (!path) | |
270 | return -ENOMEM; | |
271 | ||
272 | ret = find_next_devid(root, path, &free_devid); | |
273 | if (ret) | |
274 | goto out; | |
275 | ||
276 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
277 | key.type = BTRFS_DEV_ITEM_KEY; | |
278 | key.offset = free_devid; | |
279 | ||
280 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
281 | sizeof(*dev_item) + device->name_len); | |
282 | if (ret) | |
283 | goto out; | |
284 | ||
285 | leaf = path->nodes[0]; | |
286 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
287 | ||
288 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
289 | btrfs_set_device_type(leaf, dev_item, device->type); | |
290 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
291 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
292 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
293 | btrfs_set_device_rdev(leaf, dev_item, device->rdev); | |
294 | btrfs_set_device_partition(leaf, dev_item, device->partition); | |
295 | btrfs_set_device_name_len(leaf, dev_item, device->name_len); | |
296 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); | |
297 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
298 | ||
299 | ptr = (unsigned long)btrfs_device_name(dev_item); | |
300 | write_extent_buffer(leaf, device->name, ptr, device->name_len); | |
301 | ||
302 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
303 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); | |
304 | btrfs_mark_buffer_dirty(leaf); | |
305 | ret = 0; | |
306 | ||
307 | out: | |
308 | btrfs_free_path(path); | |
309 | return ret; | |
310 | } | |
311 | int btrfs_update_device(struct btrfs_trans_handle *trans, | |
312 | struct btrfs_device *device) | |
313 | { | |
314 | int ret; | |
315 | struct btrfs_path *path; | |
316 | struct btrfs_root *root; | |
317 | struct btrfs_dev_item *dev_item; | |
318 | struct extent_buffer *leaf; | |
319 | struct btrfs_key key; | |
320 | ||
321 | root = device->dev_root->fs_info->chunk_root; | |
322 | ||
323 | path = btrfs_alloc_path(); | |
324 | if (!path) | |
325 | return -ENOMEM; | |
326 | ||
327 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
328 | key.type = BTRFS_DEV_ITEM_KEY; | |
329 | key.offset = device->devid; | |
330 | ||
331 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
332 | if (ret < 0) | |
333 | goto out; | |
334 | ||
335 | if (ret > 0) { | |
336 | ret = -ENOENT; | |
337 | goto out; | |
338 | } | |
339 | ||
340 | leaf = path->nodes[0]; | |
341 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
342 | ||
343 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
344 | btrfs_set_device_type(leaf, dev_item, device->type); | |
345 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
346 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
347 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
348 | btrfs_set_device_rdev(leaf, dev_item, device->rdev); | |
349 | btrfs_set_device_partition(leaf, dev_item, device->partition); | |
350 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); | |
351 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
352 | btrfs_mark_buffer_dirty(leaf); | |
353 | ||
354 | out: | |
355 | btrfs_free_path(path); | |
356 | return ret; | |
357 | } | |
358 | ||
359 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, | |
360 | struct btrfs_root *root, | |
361 | struct btrfs_key *key, | |
362 | struct btrfs_chunk *chunk, int item_size) | |
363 | { | |
364 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
365 | struct btrfs_disk_key disk_key; | |
366 | u32 array_size; | |
367 | u8 *ptr; | |
368 | ||
369 | array_size = btrfs_super_sys_array_size(super_copy); | |
370 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
371 | return -EFBIG; | |
372 | ||
373 | ptr = super_copy->sys_chunk_array + array_size; | |
374 | btrfs_cpu_key_to_disk(&disk_key, key); | |
375 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
376 | ptr += sizeof(disk_key); | |
377 | memcpy(ptr, chunk, item_size); | |
378 | item_size += sizeof(disk_key); | |
379 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
380 | return 0; | |
381 | } | |
382 | ||
383 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, | |
384 | struct btrfs_root *extent_root, u64 *start, | |
6324fbf3 | 385 | u64 *num_bytes, u64 type) |
0b86a832 CM |
386 | { |
387 | u64 dev_offset; | |
388 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; | |
389 | struct btrfs_stripe *stripes; | |
390 | struct btrfs_device *device = NULL; | |
391 | struct btrfs_chunk *chunk; | |
6324fbf3 | 392 | struct list_head private_devs; |
0b86a832 | 393 | struct list_head *dev_list = &extent_root->fs_info->devices; |
6324fbf3 | 394 | struct list_head *cur; |
0b86a832 CM |
395 | struct extent_map_tree *em_tree; |
396 | struct map_lookup *map; | |
397 | struct extent_map *em; | |
398 | u64 physical; | |
399 | u64 calc_size = 1024 * 1024 * 1024; | |
6324fbf3 CM |
400 | u64 avail; |
401 | u64 max_avail = 0; | |
402 | int num_stripes = 1; | |
403 | int looped = 0; | |
0b86a832 | 404 | int ret; |
6324fbf3 | 405 | int index; |
0b86a832 CM |
406 | struct btrfs_key key; |
407 | ||
6324fbf3 CM |
408 | if (list_empty(dev_list)) |
409 | return -ENOSPC; | |
410 | again: | |
411 | INIT_LIST_HEAD(&private_devs); | |
412 | cur = dev_list->next; | |
413 | index = 0; | |
414 | /* build a private list of devices we will allocate from */ | |
415 | while(index < num_stripes) { | |
416 | device = list_entry(cur, struct btrfs_device, dev_list); | |
417 | avail = device->total_bytes - device->bytes_used; | |
418 | cur = cur->next; | |
419 | if (avail > max_avail) | |
420 | max_avail = avail; | |
421 | if (avail >= calc_size) { | |
422 | list_move_tail(&device->dev_list, &private_devs); | |
423 | index++; | |
424 | } | |
425 | if (cur == dev_list) | |
426 | break; | |
427 | } | |
428 | if (index < num_stripes) { | |
429 | list_splice(&private_devs, dev_list); | |
430 | if (!looped && max_avail > 0) { | |
431 | looped = 1; | |
432 | calc_size = max_avail; | |
433 | goto again; | |
434 | } | |
435 | return -ENOSPC; | |
436 | } | |
0b86a832 CM |
437 | |
438 | ret = find_next_chunk(chunk_root, &key.objectid); | |
439 | if (ret) | |
440 | return ret; | |
441 | ||
0b86a832 CM |
442 | chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); |
443 | if (!chunk) | |
444 | return -ENOMEM; | |
445 | ||
446 | stripes = &chunk->stripe; | |
447 | ||
448 | *num_bytes = calc_size; | |
6324fbf3 | 449 | index = 0; |
0b86a832 | 450 | while(index < num_stripes) { |
6324fbf3 CM |
451 | BUG_ON(list_empty(&private_devs)); |
452 | cur = private_devs.next; | |
453 | device = list_entry(cur, struct btrfs_device, dev_list); | |
454 | list_move_tail(&device->dev_list, dev_list); | |
0b86a832 CM |
455 | |
456 | ret = btrfs_alloc_dev_extent(trans, device, | |
457 | key.objectid, | |
458 | calc_size, &dev_offset); | |
459 | BUG_ON(ret); | |
460 | ||
461 | device->bytes_used += calc_size; | |
462 | ret = btrfs_update_device(trans, device); | |
463 | BUG_ON(ret); | |
464 | ||
465 | btrfs_set_stack_stripe_devid(stripes + index, device->devid); | |
466 | btrfs_set_stack_stripe_offset(stripes + index, dev_offset); | |
467 | physical = dev_offset; | |
468 | index++; | |
469 | } | |
6324fbf3 | 470 | BUG_ON(!list_empty(&private_devs)); |
0b86a832 CM |
471 | |
472 | /* key.objectid was set above */ | |
473 | key.offset = *num_bytes; | |
474 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
475 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); | |
476 | btrfs_set_stack_chunk_stripe_len(chunk, 64 * 1024); | |
477 | btrfs_set_stack_chunk_type(chunk, type); | |
478 | btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); | |
479 | btrfs_set_stack_chunk_io_align(chunk, extent_root->sectorsize); | |
480 | btrfs_set_stack_chunk_io_width(chunk, extent_root->sectorsize); | |
481 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); | |
482 | ||
483 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, | |
484 | btrfs_chunk_item_size(num_stripes)); | |
485 | BUG_ON(ret); | |
486 | *start = key.objectid; | |
487 | ||
488 | em = alloc_extent_map(GFP_NOFS); | |
489 | if (!em) | |
490 | return -ENOMEM; | |
491 | map = kmalloc(sizeof(*map), GFP_NOFS); | |
492 | if (!map) { | |
493 | free_extent_map(em); | |
494 | return -ENOMEM; | |
495 | } | |
496 | ||
497 | em->bdev = (struct block_device *)map; | |
498 | em->start = key.objectid; | |
499 | em->len = key.offset; | |
500 | em->block_start = 0; | |
501 | ||
502 | map->physical = physical; | |
503 | map->dev = device; | |
504 | ||
505 | if (!map->dev) { | |
506 | kfree(map); | |
507 | free_extent_map(em); | |
508 | return -EIO; | |
509 | } | |
510 | kfree(chunk); | |
511 | ||
512 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; | |
513 | spin_lock(&em_tree->lock); | |
514 | ret = add_extent_mapping(em_tree, em); | |
515 | BUG_ON(ret); | |
516 | spin_unlock(&em_tree->lock); | |
517 | free_extent_map(em); | |
518 | return ret; | |
519 | } | |
520 | ||
521 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
522 | { | |
523 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
524 | } | |
525 | ||
526 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
527 | { | |
528 | struct extent_map *em; | |
529 | ||
530 | while(1) { | |
531 | spin_lock(&tree->map_tree.lock); | |
532 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
533 | if (em) | |
534 | remove_extent_mapping(&tree->map_tree, em); | |
535 | spin_unlock(&tree->map_tree.lock); | |
536 | if (!em) | |
537 | break; | |
538 | kfree(em->bdev); | |
539 | /* once for us */ | |
540 | free_extent_map(em); | |
541 | /* once for the tree */ | |
542 | free_extent_map(em); | |
543 | } | |
544 | } | |
545 | ||
546 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, | |
547 | u64 logical, u64 *phys, u64 *length, | |
548 | struct btrfs_device **dev) | |
549 | { | |
550 | struct extent_map *em; | |
551 | struct map_lookup *map; | |
552 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
553 | u64 offset; | |
554 | ||
555 | ||
556 | spin_lock(&em_tree->lock); | |
557 | em = lookup_extent_mapping(em_tree, logical, *length); | |
558 | BUG_ON(!em); | |
559 | ||
560 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
561 | map = (struct map_lookup *)em->bdev; | |
562 | offset = logical - em->start; | |
563 | *phys = map->physical + offset; | |
564 | *length = em->len - offset; | |
565 | *dev = map->dev; | |
566 | free_extent_map(em); | |
567 | spin_unlock(&em_tree->lock); | |
568 | return 0; | |
569 | } | |
570 | ||
571 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) | |
572 | { | |
573 | struct btrfs_mapping_tree *map_tree; | |
574 | struct btrfs_device *dev; | |
575 | u64 logical = bio->bi_sector << 9; | |
576 | u64 physical; | |
577 | u64 length = 0; | |
578 | u64 map_length; | |
579 | struct bio_vec *bvec; | |
580 | int i; | |
581 | int ret; | |
582 | ||
583 | bio_for_each_segment(bvec, bio, i) { | |
584 | length += bvec->bv_len; | |
585 | } | |
586 | map_tree = &root->fs_info->mapping_tree; | |
587 | map_length = length; | |
588 | ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); | |
589 | BUG_ON(map_length < length); | |
590 | bio->bi_sector = physical >> 9; | |
591 | bio->bi_bdev = dev->bdev; | |
592 | submit_bio(rw, bio); | |
593 | return 0; | |
594 | } | |
595 | ||
596 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid) | |
597 | { | |
598 | struct btrfs_device *dev; | |
599 | struct list_head *cur = root->fs_info->devices.next; | |
600 | struct list_head *head = &root->fs_info->devices; | |
601 | ||
602 | while(cur != head) { | |
603 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
604 | if (dev->devid == devid) | |
605 | return dev; | |
606 | cur = cur->next; | |
607 | } | |
608 | return NULL; | |
609 | } | |
610 | ||
611 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, | |
612 | struct extent_buffer *leaf, | |
613 | struct btrfs_chunk *chunk) | |
614 | { | |
615 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
616 | struct map_lookup *map; | |
617 | struct extent_map *em; | |
618 | u64 logical; | |
619 | u64 length; | |
620 | u64 devid; | |
621 | int ret; | |
622 | ||
623 | logical = key->objectid; | |
624 | length = key->offset; | |
625 | spin_lock(&map_tree->map_tree.lock); | |
626 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
627 | ||
628 | /* already mapped? */ | |
629 | if (em && em->start <= logical && em->start + em->len > logical) { | |
630 | free_extent_map(em); | |
631 | spin_unlock(&map_tree->map_tree.lock); | |
632 | return 0; | |
633 | } else if (em) { | |
634 | free_extent_map(em); | |
635 | } | |
636 | spin_unlock(&map_tree->map_tree.lock); | |
637 | ||
638 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
639 | if (!map) | |
640 | return -ENOMEM; | |
641 | ||
642 | em = alloc_extent_map(GFP_NOFS); | |
643 | if (!em) | |
644 | return -ENOMEM; | |
645 | map = kmalloc(sizeof(*map), GFP_NOFS); | |
646 | if (!map) { | |
647 | free_extent_map(em); | |
648 | return -ENOMEM; | |
649 | } | |
650 | ||
651 | em->bdev = (struct block_device *)map; | |
652 | em->start = logical; | |
653 | em->len = length; | |
654 | em->block_start = 0; | |
655 | ||
656 | map->physical = btrfs_stripe_offset_nr(leaf, chunk, 0); | |
657 | devid = btrfs_stripe_devid_nr(leaf, chunk, 0); | |
658 | map->dev = btrfs_find_device(root, devid); | |
659 | if (!map->dev) { | |
660 | kfree(map); | |
661 | free_extent_map(em); | |
662 | return -EIO; | |
663 | } | |
664 | ||
665 | spin_lock(&map_tree->map_tree.lock); | |
666 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
667 | BUG_ON(ret); | |
668 | spin_unlock(&map_tree->map_tree.lock); | |
669 | free_extent_map(em); | |
670 | ||
671 | return 0; | |
672 | } | |
673 | ||
674 | static int fill_device_from_item(struct extent_buffer *leaf, | |
675 | struct btrfs_dev_item *dev_item, | |
676 | struct btrfs_device *device) | |
677 | { | |
678 | unsigned long ptr; | |
679 | char *name; | |
680 | ||
681 | device->devid = btrfs_device_id(leaf, dev_item); | |
682 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
683 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
684 | device->type = btrfs_device_type(leaf, dev_item); | |
685 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
686 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
687 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
688 | device->rdev = btrfs_device_rdev(leaf, dev_item); | |
689 | device->partition = btrfs_device_partition(leaf, dev_item); | |
690 | device->name_len = btrfs_device_name_len(leaf, dev_item); | |
691 | ||
692 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
693 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); | |
694 | ||
695 | name = kmalloc(device->name_len + 1, GFP_NOFS); | |
696 | if (!name) | |
697 | return -ENOMEM; | |
698 | device->name = name; | |
699 | ptr = (unsigned long)btrfs_device_name(dev_item); | |
700 | read_extent_buffer(leaf, name, ptr, device->name_len); | |
701 | name[device->name_len] = '\0'; | |
702 | return 0; | |
703 | } | |
704 | ||
705 | static int read_one_dev(struct btrfs_root *root, struct btrfs_key *key, | |
706 | struct extent_buffer *leaf, | |
707 | struct btrfs_dev_item *dev_item) | |
708 | { | |
709 | struct btrfs_device *device; | |
710 | u64 devid; | |
711 | int ret; | |
712 | ||
713 | devid = btrfs_device_id(leaf, dev_item); | |
6324fbf3 CM |
714 | device = btrfs_find_device(root, devid); |
715 | if (!device) { | |
716 | device = kmalloc(sizeof(*device), GFP_NOFS); | |
717 | if (!device) | |
718 | return -ENOMEM; | |
719 | list_add(&device->dev_list, &root->fs_info->devices); | |
720 | } | |
0b86a832 CM |
721 | |
722 | fill_device_from_item(leaf, dev_item, device); | |
723 | device->dev_root = root->fs_info->dev_root; | |
724 | device->bdev = root->fs_info->sb->s_bdev; | |
0b86a832 CM |
725 | memcpy(&device->dev_key, key, sizeof(*key)); |
726 | ret = 0; | |
727 | #if 0 | |
728 | ret = btrfs_open_device(device); | |
729 | if (ret) { | |
730 | kfree(device); | |
731 | } | |
732 | #endif | |
733 | return ret; | |
734 | } | |
735 | ||
736 | int btrfs_read_sys_array(struct btrfs_root *root) | |
737 | { | |
738 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
739 | struct extent_buffer *sb = root->fs_info->sb_buffer; | |
740 | struct btrfs_disk_key *disk_key; | |
741 | struct btrfs_dev_item *dev_item; | |
742 | struct btrfs_chunk *chunk; | |
743 | struct btrfs_key key; | |
744 | u32 num_stripes; | |
745 | u32 array_size; | |
746 | u32 len = 0; | |
747 | u8 *ptr; | |
748 | unsigned long sb_ptr; | |
749 | u32 cur; | |
750 | int ret; | |
751 | int dev_only = 1; | |
752 | ||
753 | array_size = btrfs_super_sys_array_size(super_copy); | |
754 | ||
755 | /* | |
756 | * we do this loop twice, once for the device items and | |
757 | * once for all of the chunks. This way there are device | |
758 | * structs filled in for every chunk | |
759 | */ | |
760 | again: | |
761 | ptr = super_copy->sys_chunk_array; | |
762 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
763 | cur = 0; | |
764 | ||
765 | while (cur < array_size) { | |
766 | disk_key = (struct btrfs_disk_key *)ptr; | |
767 | btrfs_disk_key_to_cpu(&key, disk_key); | |
768 | ||
769 | len = sizeof(*disk_key); | |
770 | ptr += len; | |
771 | sb_ptr += len; | |
772 | cur += len; | |
773 | ||
774 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID && | |
775 | key.type == BTRFS_DEV_ITEM_KEY) { | |
776 | dev_item = (struct btrfs_dev_item *)sb_ptr; | |
777 | if (dev_only) { | |
778 | ret = read_one_dev(root, &key, sb, dev_item); | |
779 | BUG_ON(ret); | |
780 | } | |
781 | len = sizeof(*dev_item); | |
782 | len += btrfs_device_name_len(sb, dev_item); | |
783 | } else if (key.type == BTRFS_CHUNK_ITEM_KEY) { | |
784 | ||
785 | chunk = (struct btrfs_chunk *)sb_ptr; | |
786 | if (!dev_only) { | |
787 | ret = read_one_chunk(root, &key, sb, chunk); | |
788 | BUG_ON(ret); | |
789 | } | |
790 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); | |
791 | len = btrfs_chunk_item_size(num_stripes); | |
792 | } else { | |
793 | BUG(); | |
794 | } | |
795 | ptr += len; | |
796 | sb_ptr += len; | |
797 | cur += len; | |
798 | } | |
799 | if (dev_only == 1) { | |
800 | dev_only = 0; | |
801 | goto again; | |
802 | } | |
803 | return 0; | |
804 | } | |
805 | ||
806 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
807 | { | |
808 | struct btrfs_path *path; | |
809 | struct extent_buffer *leaf; | |
810 | struct btrfs_key key; | |
811 | struct btrfs_key found_key; | |
812 | int ret; | |
813 | int slot; | |
814 | ||
815 | root = root->fs_info->chunk_root; | |
816 | ||
817 | path = btrfs_alloc_path(); | |
818 | if (!path) | |
819 | return -ENOMEM; | |
820 | ||
821 | /* first we search for all of the device items, and then we | |
822 | * read in all of the chunk items. This way we can create chunk | |
823 | * mappings that reference all of the devices that are afound | |
824 | */ | |
825 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
826 | key.offset = 0; | |
827 | key.type = 0; | |
828 | again: | |
829 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
830 | while(1) { | |
831 | leaf = path->nodes[0]; | |
832 | slot = path->slots[0]; | |
833 | if (slot >= btrfs_header_nritems(leaf)) { | |
834 | ret = btrfs_next_leaf(root, path); | |
835 | if (ret == 0) | |
836 | continue; | |
837 | if (ret < 0) | |
838 | goto error; | |
839 | break; | |
840 | } | |
841 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
842 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
843 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
844 | break; | |
845 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
846 | struct btrfs_dev_item *dev_item; | |
847 | dev_item = btrfs_item_ptr(leaf, slot, | |
848 | struct btrfs_dev_item); | |
849 | ret = read_one_dev(root, &found_key, leaf, | |
850 | dev_item); | |
851 | BUG_ON(ret); | |
852 | } | |
853 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
854 | struct btrfs_chunk *chunk; | |
855 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
856 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
857 | } | |
858 | path->slots[0]++; | |
859 | } | |
860 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
861 | key.objectid = 0; | |
862 | btrfs_release_path(root, path); | |
863 | goto again; | |
864 | } | |
865 | ||
866 | btrfs_free_path(path); | |
867 | ret = 0; | |
868 | error: | |
869 | return ret; | |
870 | } | |
871 |