]>
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> | |
8a4b83cc | 20 | #include <linux/buffer_head.h> |
f2d8d74d | 21 | #include <linux/blkdev.h> |
593060d7 | 22 | #include <asm/div64.h> |
0b86a832 CM |
23 | #include "ctree.h" |
24 | #include "extent_map.h" | |
25 | #include "disk-io.h" | |
26 | #include "transaction.h" | |
27 | #include "print-tree.h" | |
28 | #include "volumes.h" | |
29 | ||
593060d7 CM |
30 | struct map_lookup { |
31 | u64 type; | |
32 | int io_align; | |
33 | int io_width; | |
34 | int stripe_len; | |
35 | int sector_size; | |
36 | int num_stripes; | |
321aecc6 | 37 | int sub_stripes; |
cea9e445 | 38 | struct btrfs_bio_stripe stripes[]; |
593060d7 CM |
39 | }; |
40 | ||
41 | #define map_lookup_size(n) (sizeof(struct map_lookup) + \ | |
cea9e445 | 42 | (sizeof(struct btrfs_bio_stripe) * (n))) |
593060d7 | 43 | |
8a4b83cc CM |
44 | static DEFINE_MUTEX(uuid_mutex); |
45 | static LIST_HEAD(fs_uuids); | |
46 | ||
47 | int btrfs_cleanup_fs_uuids(void) | |
48 | { | |
49 | struct btrfs_fs_devices *fs_devices; | |
50 | struct list_head *uuid_cur; | |
51 | struct list_head *devices_cur; | |
52 | struct btrfs_device *dev; | |
53 | ||
54 | list_for_each(uuid_cur, &fs_uuids) { | |
55 | fs_devices = list_entry(uuid_cur, struct btrfs_fs_devices, | |
56 | list); | |
57 | while(!list_empty(&fs_devices->devices)) { | |
58 | devices_cur = fs_devices->devices.next; | |
59 | dev = list_entry(devices_cur, struct btrfs_device, | |
60 | dev_list); | |
8a4b83cc | 61 | if (dev->bdev) { |
8a4b83cc CM |
62 | close_bdev_excl(dev->bdev); |
63 | } | |
64 | list_del(&dev->dev_list); | |
65 | kfree(dev); | |
66 | } | |
67 | } | |
68 | return 0; | |
69 | } | |
70 | ||
a443755f CM |
71 | static struct btrfs_device *__find_device(struct list_head *head, u64 devid, |
72 | u8 *uuid) | |
8a4b83cc CM |
73 | { |
74 | struct btrfs_device *dev; | |
75 | struct list_head *cur; | |
76 | ||
77 | list_for_each(cur, head) { | |
78 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
a443755f | 79 | if (dev->devid == devid && |
8f18cf13 | 80 | (!uuid || !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE))) { |
8a4b83cc | 81 | return dev; |
a443755f | 82 | } |
8a4b83cc CM |
83 | } |
84 | return NULL; | |
85 | } | |
86 | ||
87 | static struct btrfs_fs_devices *find_fsid(u8 *fsid) | |
88 | { | |
89 | struct list_head *cur; | |
90 | struct btrfs_fs_devices *fs_devices; | |
91 | ||
92 | list_for_each(cur, &fs_uuids) { | |
93 | fs_devices = list_entry(cur, struct btrfs_fs_devices, list); | |
94 | if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0) | |
95 | return fs_devices; | |
96 | } | |
97 | return NULL; | |
98 | } | |
99 | ||
100 | static int device_list_add(const char *path, | |
101 | struct btrfs_super_block *disk_super, | |
102 | u64 devid, struct btrfs_fs_devices **fs_devices_ret) | |
103 | { | |
104 | struct btrfs_device *device; | |
105 | struct btrfs_fs_devices *fs_devices; | |
106 | u64 found_transid = btrfs_super_generation(disk_super); | |
107 | ||
108 | fs_devices = find_fsid(disk_super->fsid); | |
109 | if (!fs_devices) { | |
110 | fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS); | |
111 | if (!fs_devices) | |
112 | return -ENOMEM; | |
113 | INIT_LIST_HEAD(&fs_devices->devices); | |
b3075717 | 114 | INIT_LIST_HEAD(&fs_devices->alloc_list); |
8a4b83cc CM |
115 | list_add(&fs_devices->list, &fs_uuids); |
116 | memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE); | |
117 | fs_devices->latest_devid = devid; | |
118 | fs_devices->latest_trans = found_transid; | |
119 | fs_devices->lowest_devid = (u64)-1; | |
120 | fs_devices->num_devices = 0; | |
121 | device = NULL; | |
122 | } else { | |
a443755f CM |
123 | device = __find_device(&fs_devices->devices, devid, |
124 | disk_super->dev_item.uuid); | |
8a4b83cc CM |
125 | } |
126 | if (!device) { | |
127 | device = kzalloc(sizeof(*device), GFP_NOFS); | |
128 | if (!device) { | |
129 | /* we can safely leave the fs_devices entry around */ | |
130 | return -ENOMEM; | |
131 | } | |
132 | device->devid = devid; | |
a443755f CM |
133 | memcpy(device->uuid, disk_super->dev_item.uuid, |
134 | BTRFS_UUID_SIZE); | |
f2984462 | 135 | device->barriers = 1; |
b248a415 | 136 | spin_lock_init(&device->io_lock); |
8a4b83cc CM |
137 | device->name = kstrdup(path, GFP_NOFS); |
138 | if (!device->name) { | |
139 | kfree(device); | |
140 | return -ENOMEM; | |
141 | } | |
142 | list_add(&device->dev_list, &fs_devices->devices); | |
b3075717 | 143 | list_add(&device->dev_alloc_list, &fs_devices->alloc_list); |
8a4b83cc CM |
144 | fs_devices->num_devices++; |
145 | } | |
146 | ||
147 | if (found_transid > fs_devices->latest_trans) { | |
148 | fs_devices->latest_devid = devid; | |
149 | fs_devices->latest_trans = found_transid; | |
150 | } | |
151 | if (fs_devices->lowest_devid > devid) { | |
152 | fs_devices->lowest_devid = devid; | |
8a4b83cc CM |
153 | } |
154 | *fs_devices_ret = fs_devices; | |
155 | return 0; | |
156 | } | |
157 | ||
158 | int btrfs_close_devices(struct btrfs_fs_devices *fs_devices) | |
159 | { | |
160 | struct list_head *head = &fs_devices->devices; | |
161 | struct list_head *cur; | |
162 | struct btrfs_device *device; | |
163 | ||
164 | mutex_lock(&uuid_mutex); | |
165 | list_for_each(cur, head) { | |
166 | device = list_entry(cur, struct btrfs_device, dev_list); | |
167 | if (device->bdev) { | |
168 | close_bdev_excl(device->bdev); | |
8a4b83cc CM |
169 | } |
170 | device->bdev = NULL; | |
171 | } | |
172 | mutex_unlock(&uuid_mutex); | |
173 | return 0; | |
174 | } | |
175 | ||
176 | int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, | |
177 | int flags, void *holder) | |
178 | { | |
179 | struct block_device *bdev; | |
180 | struct list_head *head = &fs_devices->devices; | |
181 | struct list_head *cur; | |
182 | struct btrfs_device *device; | |
183 | int ret; | |
184 | ||
185 | mutex_lock(&uuid_mutex); | |
186 | list_for_each(cur, head) { | |
187 | device = list_entry(cur, struct btrfs_device, dev_list); | |
188 | bdev = open_bdev_excl(device->name, flags, holder); | |
e17cade2 | 189 | |
8a4b83cc CM |
190 | if (IS_ERR(bdev)) { |
191 | printk("open %s failed\n", device->name); | |
192 | ret = PTR_ERR(bdev); | |
193 | goto fail; | |
194 | } | |
195 | if (device->devid == fs_devices->latest_devid) | |
196 | fs_devices->latest_bdev = bdev; | |
197 | if (device->devid == fs_devices->lowest_devid) { | |
198 | fs_devices->lowest_bdev = bdev; | |
8a4b83cc CM |
199 | } |
200 | device->bdev = bdev; | |
201 | } | |
202 | mutex_unlock(&uuid_mutex); | |
203 | return 0; | |
204 | fail: | |
205 | mutex_unlock(&uuid_mutex); | |
206 | btrfs_close_devices(fs_devices); | |
207 | return ret; | |
208 | } | |
209 | ||
210 | int btrfs_scan_one_device(const char *path, int flags, void *holder, | |
211 | struct btrfs_fs_devices **fs_devices_ret) | |
212 | { | |
213 | struct btrfs_super_block *disk_super; | |
214 | struct block_device *bdev; | |
215 | struct buffer_head *bh; | |
216 | int ret; | |
217 | u64 devid; | |
f2984462 | 218 | u64 transid; |
8a4b83cc CM |
219 | |
220 | mutex_lock(&uuid_mutex); | |
221 | ||
8a4b83cc CM |
222 | bdev = open_bdev_excl(path, flags, holder); |
223 | ||
224 | if (IS_ERR(bdev)) { | |
8a4b83cc CM |
225 | ret = PTR_ERR(bdev); |
226 | goto error; | |
227 | } | |
228 | ||
229 | ret = set_blocksize(bdev, 4096); | |
230 | if (ret) | |
231 | goto error_close; | |
232 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
233 | if (!bh) { | |
234 | ret = -EIO; | |
235 | goto error_close; | |
236 | } | |
237 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
238 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
239 | sizeof(disk_super->magic))) { | |
e58ca020 | 240 | ret = -EINVAL; |
8a4b83cc CM |
241 | goto error_brelse; |
242 | } | |
243 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
f2984462 | 244 | transid = btrfs_super_generation(disk_super); |
7ae9c09d CM |
245 | if (disk_super->label[0]) |
246 | printk("device label %s ", disk_super->label); | |
247 | else { | |
248 | /* FIXME, make a readl uuid parser */ | |
249 | printk("device fsid %llx-%llx ", | |
250 | *(unsigned long long *)disk_super->fsid, | |
251 | *(unsigned long long *)(disk_super->fsid + 8)); | |
252 | } | |
253 | printk("devid %Lu transid %Lu %s\n", devid, transid, path); | |
8a4b83cc CM |
254 | ret = device_list_add(path, disk_super, devid, fs_devices_ret); |
255 | ||
256 | error_brelse: | |
257 | brelse(bh); | |
258 | error_close: | |
259 | close_bdev_excl(bdev); | |
8a4b83cc CM |
260 | error: |
261 | mutex_unlock(&uuid_mutex); | |
262 | return ret; | |
263 | } | |
0b86a832 CM |
264 | |
265 | /* | |
266 | * this uses a pretty simple search, the expectation is that it is | |
267 | * called very infrequently and that a given device has a small number | |
268 | * of extents | |
269 | */ | |
270 | static int find_free_dev_extent(struct btrfs_trans_handle *trans, | |
271 | struct btrfs_device *device, | |
272 | struct btrfs_path *path, | |
273 | u64 num_bytes, u64 *start) | |
274 | { | |
275 | struct btrfs_key key; | |
276 | struct btrfs_root *root = device->dev_root; | |
277 | struct btrfs_dev_extent *dev_extent = NULL; | |
278 | u64 hole_size = 0; | |
279 | u64 last_byte = 0; | |
280 | u64 search_start = 0; | |
281 | u64 search_end = device->total_bytes; | |
282 | int ret; | |
283 | int slot = 0; | |
284 | int start_found; | |
285 | struct extent_buffer *l; | |
286 | ||
287 | start_found = 0; | |
288 | path->reada = 2; | |
289 | ||
290 | /* FIXME use last free of some kind */ | |
291 | ||
8a4b83cc CM |
292 | /* we don't want to overwrite the superblock on the drive, |
293 | * so we make sure to start at an offset of at least 1MB | |
294 | */ | |
295 | search_start = max((u64)1024 * 1024, search_start); | |
8f18cf13 CM |
296 | |
297 | if (root->fs_info->alloc_start + num_bytes <= device->total_bytes) | |
298 | search_start = max(root->fs_info->alloc_start, search_start); | |
299 | ||
0b86a832 CM |
300 | key.objectid = device->devid; |
301 | key.offset = search_start; | |
302 | key.type = BTRFS_DEV_EXTENT_KEY; | |
303 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
304 | if (ret < 0) | |
305 | goto error; | |
306 | ret = btrfs_previous_item(root, path, 0, key.type); | |
307 | if (ret < 0) | |
308 | goto error; | |
309 | l = path->nodes[0]; | |
310 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
311 | while (1) { | |
312 | l = path->nodes[0]; | |
313 | slot = path->slots[0]; | |
314 | if (slot >= btrfs_header_nritems(l)) { | |
315 | ret = btrfs_next_leaf(root, path); | |
316 | if (ret == 0) | |
317 | continue; | |
318 | if (ret < 0) | |
319 | goto error; | |
320 | no_more_items: | |
321 | if (!start_found) { | |
322 | if (search_start >= search_end) { | |
323 | ret = -ENOSPC; | |
324 | goto error; | |
325 | } | |
326 | *start = search_start; | |
327 | start_found = 1; | |
328 | goto check_pending; | |
329 | } | |
330 | *start = last_byte > search_start ? | |
331 | last_byte : search_start; | |
332 | if (search_end <= *start) { | |
333 | ret = -ENOSPC; | |
334 | goto error; | |
335 | } | |
336 | goto check_pending; | |
337 | } | |
338 | btrfs_item_key_to_cpu(l, &key, slot); | |
339 | ||
340 | if (key.objectid < device->devid) | |
341 | goto next; | |
342 | ||
343 | if (key.objectid > device->devid) | |
344 | goto no_more_items; | |
345 | ||
346 | if (key.offset >= search_start && key.offset > last_byte && | |
347 | start_found) { | |
348 | if (last_byte < search_start) | |
349 | last_byte = search_start; | |
350 | hole_size = key.offset - last_byte; | |
351 | if (key.offset > last_byte && | |
352 | hole_size >= num_bytes) { | |
353 | *start = last_byte; | |
354 | goto check_pending; | |
355 | } | |
356 | } | |
357 | if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { | |
358 | goto next; | |
359 | } | |
360 | ||
361 | start_found = 1; | |
362 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
363 | last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); | |
364 | next: | |
365 | path->slots[0]++; | |
366 | cond_resched(); | |
367 | } | |
368 | check_pending: | |
369 | /* we have to make sure we didn't find an extent that has already | |
370 | * been allocated by the map tree or the original allocation | |
371 | */ | |
372 | btrfs_release_path(root, path); | |
373 | BUG_ON(*start < search_start); | |
374 | ||
6324fbf3 | 375 | if (*start + num_bytes > search_end) { |
0b86a832 CM |
376 | ret = -ENOSPC; |
377 | goto error; | |
378 | } | |
379 | /* check for pending inserts here */ | |
380 | return 0; | |
381 | ||
382 | error: | |
383 | btrfs_release_path(root, path); | |
384 | return ret; | |
385 | } | |
386 | ||
8f18cf13 CM |
387 | int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, |
388 | struct btrfs_device *device, | |
389 | u64 start) | |
390 | { | |
391 | int ret; | |
392 | struct btrfs_path *path; | |
393 | struct btrfs_root *root = device->dev_root; | |
394 | struct btrfs_key key; | |
395 | ||
396 | path = btrfs_alloc_path(); | |
397 | if (!path) | |
398 | return -ENOMEM; | |
399 | ||
400 | key.objectid = device->devid; | |
401 | key.offset = start; | |
402 | key.type = BTRFS_DEV_EXTENT_KEY; | |
403 | ||
404 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); | |
405 | BUG_ON(ret); | |
406 | ||
407 | ret = btrfs_del_item(trans, root, path); | |
408 | BUG_ON(ret); | |
409 | ||
410 | btrfs_free_path(path); | |
411 | return ret; | |
412 | } | |
413 | ||
0b86a832 CM |
414 | int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, |
415 | struct btrfs_device *device, | |
e17cade2 CM |
416 | u64 chunk_tree, u64 chunk_objectid, |
417 | u64 chunk_offset, | |
418 | u64 num_bytes, u64 *start) | |
0b86a832 CM |
419 | { |
420 | int ret; | |
421 | struct btrfs_path *path; | |
422 | struct btrfs_root *root = device->dev_root; | |
423 | struct btrfs_dev_extent *extent; | |
424 | struct extent_buffer *leaf; | |
425 | struct btrfs_key key; | |
426 | ||
427 | path = btrfs_alloc_path(); | |
428 | if (!path) | |
429 | return -ENOMEM; | |
430 | ||
431 | ret = find_free_dev_extent(trans, device, path, num_bytes, start); | |
6324fbf3 | 432 | if (ret) { |
0b86a832 | 433 | goto err; |
6324fbf3 | 434 | } |
0b86a832 CM |
435 | |
436 | key.objectid = device->devid; | |
437 | key.offset = *start; | |
438 | key.type = BTRFS_DEV_EXTENT_KEY; | |
439 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
440 | sizeof(*extent)); | |
441 | BUG_ON(ret); | |
442 | ||
443 | leaf = path->nodes[0]; | |
444 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
445 | struct btrfs_dev_extent); | |
e17cade2 CM |
446 | btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree); |
447 | btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid); | |
448 | btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset); | |
449 | ||
450 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, | |
451 | (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent), | |
452 | BTRFS_UUID_SIZE); | |
453 | ||
0b86a832 CM |
454 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); |
455 | btrfs_mark_buffer_dirty(leaf); | |
456 | err: | |
457 | btrfs_free_path(path); | |
458 | return ret; | |
459 | } | |
460 | ||
e17cade2 | 461 | static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset) |
0b86a832 CM |
462 | { |
463 | struct btrfs_path *path; | |
464 | int ret; | |
465 | struct btrfs_key key; | |
e17cade2 | 466 | struct btrfs_chunk *chunk; |
0b86a832 CM |
467 | struct btrfs_key found_key; |
468 | ||
469 | path = btrfs_alloc_path(); | |
470 | BUG_ON(!path); | |
471 | ||
e17cade2 | 472 | key.objectid = objectid; |
0b86a832 CM |
473 | key.offset = (u64)-1; |
474 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
475 | ||
476 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
477 | if (ret < 0) | |
478 | goto error; | |
479 | ||
480 | BUG_ON(ret == 0); | |
481 | ||
482 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
483 | if (ret) { | |
e17cade2 | 484 | *offset = 0; |
0b86a832 CM |
485 | } else { |
486 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
487 | path->slots[0]); | |
e17cade2 CM |
488 | if (found_key.objectid != objectid) |
489 | *offset = 0; | |
490 | else { | |
491 | chunk = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
492 | struct btrfs_chunk); | |
493 | *offset = found_key.offset + | |
494 | btrfs_chunk_length(path->nodes[0], chunk); | |
495 | } | |
0b86a832 CM |
496 | } |
497 | ret = 0; | |
498 | error: | |
499 | btrfs_free_path(path); | |
500 | return ret; | |
501 | } | |
502 | ||
0b86a832 CM |
503 | static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, |
504 | u64 *objectid) | |
505 | { | |
506 | int ret; | |
507 | struct btrfs_key key; | |
508 | struct btrfs_key found_key; | |
509 | ||
510 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
511 | key.type = BTRFS_DEV_ITEM_KEY; | |
512 | key.offset = (u64)-1; | |
513 | ||
514 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
515 | if (ret < 0) | |
516 | goto error; | |
517 | ||
518 | BUG_ON(ret == 0); | |
519 | ||
520 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
521 | BTRFS_DEV_ITEM_KEY); | |
522 | if (ret) { | |
523 | *objectid = 1; | |
524 | } else { | |
525 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
526 | path->slots[0]); | |
527 | *objectid = found_key.offset + 1; | |
528 | } | |
529 | ret = 0; | |
530 | error: | |
531 | btrfs_release_path(root, path); | |
532 | return ret; | |
533 | } | |
534 | ||
535 | /* | |
536 | * the device information is stored in the chunk root | |
537 | * the btrfs_device struct should be fully filled in | |
538 | */ | |
539 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
540 | struct btrfs_root *root, | |
541 | struct btrfs_device *device) | |
542 | { | |
543 | int ret; | |
544 | struct btrfs_path *path; | |
545 | struct btrfs_dev_item *dev_item; | |
546 | struct extent_buffer *leaf; | |
547 | struct btrfs_key key; | |
548 | unsigned long ptr; | |
549 | u64 free_devid; | |
550 | ||
551 | root = root->fs_info->chunk_root; | |
552 | ||
553 | path = btrfs_alloc_path(); | |
554 | if (!path) | |
555 | return -ENOMEM; | |
556 | ||
557 | ret = find_next_devid(root, path, &free_devid); | |
558 | if (ret) | |
559 | goto out; | |
560 | ||
561 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
562 | key.type = BTRFS_DEV_ITEM_KEY; | |
563 | key.offset = free_devid; | |
564 | ||
565 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
0d81ba5d | 566 | sizeof(*dev_item)); |
0b86a832 CM |
567 | if (ret) |
568 | goto out; | |
569 | ||
570 | leaf = path->nodes[0]; | |
571 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
572 | ||
8a4b83cc | 573 | device->devid = free_devid; |
0b86a832 CM |
574 | btrfs_set_device_id(leaf, dev_item, device->devid); |
575 | btrfs_set_device_type(leaf, dev_item, device->type); | |
576 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
577 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
578 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
579 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
580 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
e17cade2 CM |
581 | btrfs_set_device_group(leaf, dev_item, 0); |
582 | btrfs_set_device_seek_speed(leaf, dev_item, 0); | |
583 | btrfs_set_device_bandwidth(leaf, dev_item, 0); | |
0b86a832 | 584 | |
0b86a832 | 585 | ptr = (unsigned long)btrfs_device_uuid(dev_item); |
e17cade2 | 586 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 CM |
587 | btrfs_mark_buffer_dirty(leaf); |
588 | ret = 0; | |
589 | ||
590 | out: | |
591 | btrfs_free_path(path); | |
592 | return ret; | |
593 | } | |
8f18cf13 | 594 | |
0b86a832 CM |
595 | int btrfs_update_device(struct btrfs_trans_handle *trans, |
596 | struct btrfs_device *device) | |
597 | { | |
598 | int ret; | |
599 | struct btrfs_path *path; | |
600 | struct btrfs_root *root; | |
601 | struct btrfs_dev_item *dev_item; | |
602 | struct extent_buffer *leaf; | |
603 | struct btrfs_key key; | |
604 | ||
605 | root = device->dev_root->fs_info->chunk_root; | |
606 | ||
607 | path = btrfs_alloc_path(); | |
608 | if (!path) | |
609 | return -ENOMEM; | |
610 | ||
611 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
612 | key.type = BTRFS_DEV_ITEM_KEY; | |
613 | key.offset = device->devid; | |
614 | ||
615 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
616 | if (ret < 0) | |
617 | goto out; | |
618 | ||
619 | if (ret > 0) { | |
620 | ret = -ENOENT; | |
621 | goto out; | |
622 | } | |
623 | ||
624 | leaf = path->nodes[0]; | |
625 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
626 | ||
627 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
628 | btrfs_set_device_type(leaf, dev_item, device->type); | |
629 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
630 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
631 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
632 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
633 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
634 | btrfs_mark_buffer_dirty(leaf); | |
635 | ||
636 | out: | |
637 | btrfs_free_path(path); | |
638 | return ret; | |
639 | } | |
640 | ||
8f18cf13 CM |
641 | int btrfs_grow_device(struct btrfs_trans_handle *trans, |
642 | struct btrfs_device *device, u64 new_size) | |
643 | { | |
644 | struct btrfs_super_block *super_copy = | |
645 | &device->dev_root->fs_info->super_copy; | |
646 | u64 old_total = btrfs_super_total_bytes(super_copy); | |
647 | u64 diff = new_size - device->total_bytes; | |
648 | ||
649 | btrfs_set_super_total_bytes(super_copy, old_total + diff); | |
650 | return btrfs_update_device(trans, device); | |
651 | } | |
652 | ||
653 | static int btrfs_free_chunk(struct btrfs_trans_handle *trans, | |
654 | struct btrfs_root *root, | |
655 | u64 chunk_tree, u64 chunk_objectid, | |
656 | u64 chunk_offset) | |
657 | { | |
658 | int ret; | |
659 | struct btrfs_path *path; | |
660 | struct btrfs_key key; | |
661 | ||
662 | root = root->fs_info->chunk_root; | |
663 | path = btrfs_alloc_path(); | |
664 | if (!path) | |
665 | return -ENOMEM; | |
666 | ||
667 | key.objectid = chunk_objectid; | |
668 | key.offset = chunk_offset; | |
669 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
670 | ||
671 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); | |
672 | BUG_ON(ret); | |
673 | ||
674 | ret = btrfs_del_item(trans, root, path); | |
675 | BUG_ON(ret); | |
676 | ||
677 | btrfs_free_path(path); | |
678 | return 0; | |
679 | } | |
680 | ||
681 | int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64 | |
682 | chunk_offset) | |
683 | { | |
684 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
685 | struct btrfs_disk_key *disk_key; | |
686 | struct btrfs_chunk *chunk; | |
687 | u8 *ptr; | |
688 | int ret = 0; | |
689 | u32 num_stripes; | |
690 | u32 array_size; | |
691 | u32 len = 0; | |
692 | u32 cur; | |
693 | struct btrfs_key key; | |
694 | ||
695 | array_size = btrfs_super_sys_array_size(super_copy); | |
696 | ||
697 | ptr = super_copy->sys_chunk_array; | |
698 | cur = 0; | |
699 | ||
700 | while (cur < array_size) { | |
701 | disk_key = (struct btrfs_disk_key *)ptr; | |
702 | btrfs_disk_key_to_cpu(&key, disk_key); | |
703 | ||
704 | len = sizeof(*disk_key); | |
705 | ||
706 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { | |
707 | chunk = (struct btrfs_chunk *)(ptr + len); | |
708 | num_stripes = btrfs_stack_chunk_num_stripes(chunk); | |
709 | len += btrfs_chunk_item_size(num_stripes); | |
710 | } else { | |
711 | ret = -EIO; | |
712 | break; | |
713 | } | |
714 | if (key.objectid == chunk_objectid && | |
715 | key.offset == chunk_offset) { | |
716 | memmove(ptr, ptr + len, array_size - (cur + len)); | |
717 | array_size -= len; | |
718 | btrfs_set_super_sys_array_size(super_copy, array_size); | |
719 | } else { | |
720 | ptr += len; | |
721 | cur += len; | |
722 | } | |
723 | } | |
724 | return ret; | |
725 | } | |
726 | ||
727 | ||
728 | int btrfs_relocate_chunk(struct btrfs_root *root, | |
729 | u64 chunk_tree, u64 chunk_objectid, | |
730 | u64 chunk_offset) | |
731 | { | |
732 | struct extent_map_tree *em_tree; | |
733 | struct btrfs_root *extent_root; | |
734 | struct btrfs_trans_handle *trans; | |
735 | struct extent_map *em; | |
736 | struct map_lookup *map; | |
737 | int ret; | |
738 | int i; | |
739 | ||
740 | root = root->fs_info->chunk_root; | |
741 | extent_root = root->fs_info->extent_root; | |
742 | em_tree = &root->fs_info->mapping_tree.map_tree; | |
743 | ||
744 | /* step one, relocate all the extents inside this chunk */ | |
745 | ret = btrfs_shrink_extent_tree(extent_root, chunk_offset); | |
746 | BUG_ON(ret); | |
747 | ||
748 | trans = btrfs_start_transaction(root, 1); | |
749 | BUG_ON(!trans); | |
750 | ||
751 | /* | |
752 | * step two, delete the device extents and the | |
753 | * chunk tree entries | |
754 | */ | |
755 | spin_lock(&em_tree->lock); | |
756 | em = lookup_extent_mapping(em_tree, chunk_offset, 1); | |
757 | spin_unlock(&em_tree->lock); | |
758 | ||
759 | BUG_ON(em->start > chunk_offset || em->start + em->len < chunk_offset); | |
760 | map = (struct map_lookup *)em->bdev; | |
761 | ||
762 | for (i = 0; i < map->num_stripes; i++) { | |
763 | ret = btrfs_free_dev_extent(trans, map->stripes[i].dev, | |
764 | map->stripes[i].physical); | |
765 | BUG_ON(ret); | |
766 | } | |
767 | ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid, | |
768 | chunk_offset); | |
769 | ||
770 | BUG_ON(ret); | |
771 | ||
772 | if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) { | |
773 | ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset); | |
774 | BUG_ON(ret); | |
775 | goto out; | |
776 | } | |
777 | ||
778 | ||
779 | ||
780 | spin_lock(&em_tree->lock); | |
781 | remove_extent_mapping(em_tree, em); | |
782 | kfree(map); | |
783 | em->bdev = NULL; | |
784 | ||
785 | /* once for the tree */ | |
786 | free_extent_map(em); | |
787 | spin_unlock(&em_tree->lock); | |
788 | ||
789 | out: | |
790 | /* once for us */ | |
791 | free_extent_map(em); | |
792 | ||
793 | btrfs_end_transaction(trans, root); | |
794 | return 0; | |
795 | } | |
796 | ||
797 | /* | |
798 | * shrinking a device means finding all of the device extents past | |
799 | * the new size, and then following the back refs to the chunks. | |
800 | * The chunk relocation code actually frees the device extent | |
801 | */ | |
802 | int btrfs_shrink_device(struct btrfs_device *device, u64 new_size) | |
803 | { | |
804 | struct btrfs_trans_handle *trans; | |
805 | struct btrfs_root *root = device->dev_root; | |
806 | struct btrfs_dev_extent *dev_extent = NULL; | |
807 | struct btrfs_path *path; | |
808 | u64 length; | |
809 | u64 chunk_tree; | |
810 | u64 chunk_objectid; | |
811 | u64 chunk_offset; | |
812 | int ret; | |
813 | int slot; | |
814 | struct extent_buffer *l; | |
815 | struct btrfs_key key; | |
816 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
817 | u64 old_total = btrfs_super_total_bytes(super_copy); | |
818 | u64 diff = device->total_bytes - new_size; | |
819 | ||
820 | ||
821 | path = btrfs_alloc_path(); | |
822 | if (!path) | |
823 | return -ENOMEM; | |
824 | ||
825 | trans = btrfs_start_transaction(root, 1); | |
826 | if (!trans) { | |
827 | ret = -ENOMEM; | |
828 | goto done; | |
829 | } | |
830 | ||
831 | path->reada = 2; | |
832 | ||
833 | device->total_bytes = new_size; | |
834 | ret = btrfs_update_device(trans, device); | |
835 | if (ret) { | |
836 | btrfs_end_transaction(trans, root); | |
837 | goto done; | |
838 | } | |
839 | WARN_ON(diff > old_total); | |
840 | btrfs_set_super_total_bytes(super_copy, old_total - diff); | |
841 | btrfs_end_transaction(trans, root); | |
842 | ||
843 | key.objectid = device->devid; | |
844 | key.offset = (u64)-1; | |
845 | key.type = BTRFS_DEV_EXTENT_KEY; | |
846 | ||
847 | while (1) { | |
848 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
849 | if (ret < 0) | |
850 | goto done; | |
851 | ||
852 | ret = btrfs_previous_item(root, path, 0, key.type); | |
853 | if (ret < 0) | |
854 | goto done; | |
855 | if (ret) { | |
856 | ret = 0; | |
857 | goto done; | |
858 | } | |
859 | ||
860 | l = path->nodes[0]; | |
861 | slot = path->slots[0]; | |
862 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
863 | ||
864 | if (key.objectid != device->devid) | |
865 | goto done; | |
866 | ||
867 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
868 | length = btrfs_dev_extent_length(l, dev_extent); | |
869 | ||
870 | if (key.offset + length <= new_size) | |
871 | goto done; | |
872 | ||
873 | chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent); | |
874 | chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent); | |
875 | chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent); | |
876 | btrfs_release_path(root, path); | |
877 | ||
878 | ret = btrfs_relocate_chunk(root, chunk_tree, chunk_objectid, | |
879 | chunk_offset); | |
880 | if (ret) | |
881 | goto done; | |
882 | } | |
883 | ||
884 | done: | |
885 | btrfs_free_path(path); | |
886 | return ret; | |
887 | } | |
888 | ||
0b86a832 CM |
889 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, |
890 | struct btrfs_root *root, | |
891 | struct btrfs_key *key, | |
892 | struct btrfs_chunk *chunk, int item_size) | |
893 | { | |
894 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
895 | struct btrfs_disk_key disk_key; | |
896 | u32 array_size; | |
897 | u8 *ptr; | |
898 | ||
899 | array_size = btrfs_super_sys_array_size(super_copy); | |
900 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
901 | return -EFBIG; | |
902 | ||
903 | ptr = super_copy->sys_chunk_array + array_size; | |
904 | btrfs_cpu_key_to_disk(&disk_key, key); | |
905 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
906 | ptr += sizeof(disk_key); | |
907 | memcpy(ptr, chunk, item_size); | |
908 | item_size += sizeof(disk_key); | |
909 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
910 | return 0; | |
911 | } | |
912 | ||
9b3f68b9 CM |
913 | static u64 div_factor(u64 num, int factor) |
914 | { | |
915 | if (factor == 10) | |
916 | return num; | |
917 | num *= factor; | |
918 | do_div(num, 10); | |
919 | return num; | |
920 | } | |
921 | ||
922 | static u64 chunk_bytes_by_type(u64 type, u64 calc_size, int num_stripes, | |
923 | int sub_stripes) | |
924 | { | |
925 | if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP)) | |
926 | return calc_size; | |
927 | else if (type & BTRFS_BLOCK_GROUP_RAID10) | |
928 | return calc_size * (num_stripes / sub_stripes); | |
929 | else | |
930 | return calc_size * num_stripes; | |
931 | } | |
932 | ||
933 | ||
0b86a832 CM |
934 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, |
935 | struct btrfs_root *extent_root, u64 *start, | |
6324fbf3 | 936 | u64 *num_bytes, u64 type) |
0b86a832 CM |
937 | { |
938 | u64 dev_offset; | |
593060d7 | 939 | struct btrfs_fs_info *info = extent_root->fs_info; |
0b86a832 | 940 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; |
8f18cf13 | 941 | struct btrfs_path *path; |
0b86a832 CM |
942 | struct btrfs_stripe *stripes; |
943 | struct btrfs_device *device = NULL; | |
944 | struct btrfs_chunk *chunk; | |
6324fbf3 | 945 | struct list_head private_devs; |
b3075717 | 946 | struct list_head *dev_list; |
6324fbf3 | 947 | struct list_head *cur; |
0b86a832 CM |
948 | struct extent_map_tree *em_tree; |
949 | struct map_lookup *map; | |
950 | struct extent_map *em; | |
a40a90a0 | 951 | int min_stripe_size = 1 * 1024 * 1024; |
0b86a832 CM |
952 | u64 physical; |
953 | u64 calc_size = 1024 * 1024 * 1024; | |
9b3f68b9 CM |
954 | u64 max_chunk_size = calc_size; |
955 | u64 min_free; | |
6324fbf3 CM |
956 | u64 avail; |
957 | u64 max_avail = 0; | |
9b3f68b9 | 958 | u64 percent_max; |
6324fbf3 | 959 | int num_stripes = 1; |
a40a90a0 | 960 | int min_stripes = 1; |
321aecc6 | 961 | int sub_stripes = 0; |
6324fbf3 | 962 | int looped = 0; |
0b86a832 | 963 | int ret; |
6324fbf3 | 964 | int index; |
593060d7 | 965 | int stripe_len = 64 * 1024; |
0b86a832 CM |
966 | struct btrfs_key key; |
967 | ||
b3075717 | 968 | dev_list = &extent_root->fs_info->fs_devices->alloc_list; |
6324fbf3 CM |
969 | if (list_empty(dev_list)) |
970 | return -ENOSPC; | |
593060d7 | 971 | |
a40a90a0 | 972 | if (type & (BTRFS_BLOCK_GROUP_RAID0)) { |
593060d7 | 973 | num_stripes = btrfs_super_num_devices(&info->super_copy); |
a40a90a0 CM |
974 | min_stripes = 2; |
975 | } | |
976 | if (type & (BTRFS_BLOCK_GROUP_DUP)) { | |
611f0e00 | 977 | num_stripes = 2; |
a40a90a0 CM |
978 | min_stripes = 2; |
979 | } | |
8790d502 CM |
980 | if (type & (BTRFS_BLOCK_GROUP_RAID1)) { |
981 | num_stripes = min_t(u64, 2, | |
982 | btrfs_super_num_devices(&info->super_copy)); | |
9b3f68b9 CM |
983 | if (num_stripes < 2) |
984 | return -ENOSPC; | |
a40a90a0 | 985 | min_stripes = 2; |
8790d502 | 986 | } |
321aecc6 CM |
987 | if (type & (BTRFS_BLOCK_GROUP_RAID10)) { |
988 | num_stripes = btrfs_super_num_devices(&info->super_copy); | |
989 | if (num_stripes < 4) | |
990 | return -ENOSPC; | |
991 | num_stripes &= ~(u32)1; | |
992 | sub_stripes = 2; | |
a40a90a0 | 993 | min_stripes = 4; |
321aecc6 | 994 | } |
9b3f68b9 CM |
995 | |
996 | if (type & BTRFS_BLOCK_GROUP_DATA) { | |
997 | max_chunk_size = 10 * calc_size; | |
a40a90a0 | 998 | min_stripe_size = 64 * 1024 * 1024; |
9b3f68b9 CM |
999 | } else if (type & BTRFS_BLOCK_GROUP_METADATA) { |
1000 | max_chunk_size = 4 * calc_size; | |
a40a90a0 CM |
1001 | min_stripe_size = 32 * 1024 * 1024; |
1002 | } else if (type & BTRFS_BLOCK_GROUP_SYSTEM) { | |
1003 | calc_size = 8 * 1024 * 1024; | |
1004 | max_chunk_size = calc_size * 2; | |
1005 | min_stripe_size = 1 * 1024 * 1024; | |
9b3f68b9 CM |
1006 | } |
1007 | ||
8f18cf13 CM |
1008 | path = btrfs_alloc_path(); |
1009 | if (!path) | |
1010 | return -ENOMEM; | |
1011 | ||
9b3f68b9 CM |
1012 | /* we don't want a chunk larger than 10% of the FS */ |
1013 | percent_max = div_factor(btrfs_super_total_bytes(&info->super_copy), 1); | |
1014 | max_chunk_size = min(percent_max, max_chunk_size); | |
1015 | ||
a40a90a0 | 1016 | again: |
9b3f68b9 CM |
1017 | if (calc_size * num_stripes > max_chunk_size) { |
1018 | calc_size = max_chunk_size; | |
1019 | do_div(calc_size, num_stripes); | |
1020 | do_div(calc_size, stripe_len); | |
1021 | calc_size *= stripe_len; | |
1022 | } | |
1023 | /* we don't want tiny stripes */ | |
a40a90a0 | 1024 | calc_size = max_t(u64, min_stripe_size, calc_size); |
9b3f68b9 | 1025 | |
9b3f68b9 CM |
1026 | do_div(calc_size, stripe_len); |
1027 | calc_size *= stripe_len; | |
1028 | ||
6324fbf3 CM |
1029 | INIT_LIST_HEAD(&private_devs); |
1030 | cur = dev_list->next; | |
1031 | index = 0; | |
611f0e00 CM |
1032 | |
1033 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
1034 | min_free = calc_size * 2; | |
9b3f68b9 CM |
1035 | else |
1036 | min_free = calc_size; | |
611f0e00 | 1037 | |
ad5bd91e CM |
1038 | /* we add 1MB because we never use the first 1MB of the device */ |
1039 | min_free += 1024 * 1024; | |
1040 | ||
6324fbf3 CM |
1041 | /* build a private list of devices we will allocate from */ |
1042 | while(index < num_stripes) { | |
b3075717 | 1043 | device = list_entry(cur, struct btrfs_device, dev_alloc_list); |
611f0e00 | 1044 | |
6324fbf3 CM |
1045 | avail = device->total_bytes - device->bytes_used; |
1046 | cur = cur->next; | |
8f18cf13 | 1047 | |
611f0e00 | 1048 | if (avail >= min_free) { |
8f18cf13 CM |
1049 | u64 ignored_start = 0; |
1050 | ret = find_free_dev_extent(trans, device, path, | |
1051 | min_free, | |
1052 | &ignored_start); | |
1053 | if (ret == 0) { | |
1054 | list_move_tail(&device->dev_alloc_list, | |
1055 | &private_devs); | |
611f0e00 | 1056 | index++; |
8f18cf13 CM |
1057 | if (type & BTRFS_BLOCK_GROUP_DUP) |
1058 | index++; | |
1059 | } | |
a40a90a0 CM |
1060 | } else if (avail > max_avail) |
1061 | max_avail = avail; | |
6324fbf3 CM |
1062 | if (cur == dev_list) |
1063 | break; | |
1064 | } | |
1065 | if (index < num_stripes) { | |
1066 | list_splice(&private_devs, dev_list); | |
a40a90a0 CM |
1067 | if (index >= min_stripes) { |
1068 | num_stripes = index; | |
1069 | if (type & (BTRFS_BLOCK_GROUP_RAID10)) { | |
1070 | num_stripes /= sub_stripes; | |
1071 | num_stripes *= sub_stripes; | |
1072 | } | |
1073 | looped = 1; | |
1074 | goto again; | |
1075 | } | |
6324fbf3 CM |
1076 | if (!looped && max_avail > 0) { |
1077 | looped = 1; | |
1078 | calc_size = max_avail; | |
1079 | goto again; | |
1080 | } | |
8f18cf13 | 1081 | btrfs_free_path(path); |
6324fbf3 CM |
1082 | return -ENOSPC; |
1083 | } | |
e17cade2 CM |
1084 | key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; |
1085 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
1086 | ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID, | |
1087 | &key.offset); | |
8f18cf13 CM |
1088 | if (ret) { |
1089 | btrfs_free_path(path); | |
0b86a832 | 1090 | return ret; |
8f18cf13 | 1091 | } |
0b86a832 | 1092 | |
0b86a832 | 1093 | chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); |
8f18cf13 CM |
1094 | if (!chunk) { |
1095 | btrfs_free_path(path); | |
0b86a832 | 1096 | return -ENOMEM; |
8f18cf13 | 1097 | } |
0b86a832 | 1098 | |
593060d7 CM |
1099 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); |
1100 | if (!map) { | |
1101 | kfree(chunk); | |
8f18cf13 | 1102 | btrfs_free_path(path); |
593060d7 CM |
1103 | return -ENOMEM; |
1104 | } | |
8f18cf13 CM |
1105 | btrfs_free_path(path); |
1106 | path = NULL; | |
593060d7 | 1107 | |
0b86a832 | 1108 | stripes = &chunk->stripe; |
9b3f68b9 CM |
1109 | *num_bytes = chunk_bytes_by_type(type, calc_size, |
1110 | num_stripes, sub_stripes); | |
0b86a832 | 1111 | |
6324fbf3 | 1112 | index = 0; |
e17cade2 | 1113 | printk("new chunk type %Lu start %Lu size %Lu\n", type, key.offset, *num_bytes); |
0b86a832 | 1114 | while(index < num_stripes) { |
e17cade2 | 1115 | struct btrfs_stripe *stripe; |
6324fbf3 CM |
1116 | BUG_ON(list_empty(&private_devs)); |
1117 | cur = private_devs.next; | |
b3075717 | 1118 | device = list_entry(cur, struct btrfs_device, dev_alloc_list); |
611f0e00 CM |
1119 | |
1120 | /* loop over this device again if we're doing a dup group */ | |
1121 | if (!(type & BTRFS_BLOCK_GROUP_DUP) || | |
1122 | (index == num_stripes - 1)) | |
b3075717 | 1123 | list_move_tail(&device->dev_alloc_list, dev_list); |
0b86a832 CM |
1124 | |
1125 | ret = btrfs_alloc_dev_extent(trans, device, | |
e17cade2 CM |
1126 | info->chunk_root->root_key.objectid, |
1127 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset, | |
1128 | calc_size, &dev_offset); | |
0b86a832 | 1129 | BUG_ON(ret); |
e17cade2 | 1130 | printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key.offset, calc_size, device->devid, type); |
0b86a832 CM |
1131 | device->bytes_used += calc_size; |
1132 | ret = btrfs_update_device(trans, device); | |
1133 | BUG_ON(ret); | |
1134 | ||
593060d7 CM |
1135 | map->stripes[index].dev = device; |
1136 | map->stripes[index].physical = dev_offset; | |
e17cade2 CM |
1137 | stripe = stripes + index; |
1138 | btrfs_set_stack_stripe_devid(stripe, device->devid); | |
1139 | btrfs_set_stack_stripe_offset(stripe, dev_offset); | |
1140 | memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE); | |
0b86a832 CM |
1141 | physical = dev_offset; |
1142 | index++; | |
1143 | } | |
6324fbf3 | 1144 | BUG_ON(!list_empty(&private_devs)); |
0b86a832 | 1145 | |
e17cade2 CM |
1146 | /* key was set above */ |
1147 | btrfs_set_stack_chunk_length(chunk, *num_bytes); | |
0b86a832 | 1148 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); |
593060d7 | 1149 | btrfs_set_stack_chunk_stripe_len(chunk, stripe_len); |
0b86a832 CM |
1150 | btrfs_set_stack_chunk_type(chunk, type); |
1151 | btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); | |
593060d7 CM |
1152 | btrfs_set_stack_chunk_io_align(chunk, stripe_len); |
1153 | btrfs_set_stack_chunk_io_width(chunk, stripe_len); | |
0b86a832 | 1154 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); |
321aecc6 | 1155 | btrfs_set_stack_chunk_sub_stripes(chunk, sub_stripes); |
593060d7 CM |
1156 | map->sector_size = extent_root->sectorsize; |
1157 | map->stripe_len = stripe_len; | |
1158 | map->io_align = stripe_len; | |
1159 | map->io_width = stripe_len; | |
1160 | map->type = type; | |
1161 | map->num_stripes = num_stripes; | |
321aecc6 | 1162 | map->sub_stripes = sub_stripes; |
0b86a832 CM |
1163 | |
1164 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, | |
1165 | btrfs_chunk_item_size(num_stripes)); | |
1166 | BUG_ON(ret); | |
e17cade2 | 1167 | *start = key.offset;; |
0b86a832 CM |
1168 | |
1169 | em = alloc_extent_map(GFP_NOFS); | |
1170 | if (!em) | |
1171 | return -ENOMEM; | |
0b86a832 | 1172 | em->bdev = (struct block_device *)map; |
e17cade2 CM |
1173 | em->start = key.offset; |
1174 | em->len = *num_bytes; | |
0b86a832 CM |
1175 | em->block_start = 0; |
1176 | ||
8f18cf13 CM |
1177 | if (type & BTRFS_BLOCK_GROUP_SYSTEM) { |
1178 | ret = btrfs_add_system_chunk(trans, chunk_root, &key, | |
1179 | chunk, btrfs_chunk_item_size(num_stripes)); | |
1180 | BUG_ON(ret); | |
1181 | } | |
0b86a832 CM |
1182 | kfree(chunk); |
1183 | ||
1184 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; | |
1185 | spin_lock(&em_tree->lock); | |
1186 | ret = add_extent_mapping(em_tree, em); | |
0b86a832 | 1187 | spin_unlock(&em_tree->lock); |
b248a415 | 1188 | BUG_ON(ret); |
0b86a832 CM |
1189 | free_extent_map(em); |
1190 | return ret; | |
1191 | } | |
1192 | ||
1193 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
1194 | { | |
1195 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
1196 | } | |
1197 | ||
1198 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
1199 | { | |
1200 | struct extent_map *em; | |
1201 | ||
1202 | while(1) { | |
1203 | spin_lock(&tree->map_tree.lock); | |
1204 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
1205 | if (em) | |
1206 | remove_extent_mapping(&tree->map_tree, em); | |
1207 | spin_unlock(&tree->map_tree.lock); | |
1208 | if (!em) | |
1209 | break; | |
1210 | kfree(em->bdev); | |
1211 | /* once for us */ | |
1212 | free_extent_map(em); | |
1213 | /* once for the tree */ | |
1214 | free_extent_map(em); | |
1215 | } | |
1216 | } | |
1217 | ||
f188591e CM |
1218 | int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) |
1219 | { | |
1220 | struct extent_map *em; | |
1221 | struct map_lookup *map; | |
1222 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
1223 | int ret; | |
1224 | ||
1225 | spin_lock(&em_tree->lock); | |
1226 | em = lookup_extent_mapping(em_tree, logical, len); | |
b248a415 | 1227 | spin_unlock(&em_tree->lock); |
f188591e CM |
1228 | BUG_ON(!em); |
1229 | ||
1230 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
1231 | map = (struct map_lookup *)em->bdev; | |
1232 | if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1)) | |
1233 | ret = map->num_stripes; | |
321aecc6 CM |
1234 | else if (map->type & BTRFS_BLOCK_GROUP_RAID10) |
1235 | ret = map->sub_stripes; | |
f188591e CM |
1236 | else |
1237 | ret = 1; | |
1238 | free_extent_map(em); | |
f188591e CM |
1239 | return ret; |
1240 | } | |
1241 | ||
f2d8d74d CM |
1242 | static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
1243 | u64 logical, u64 *length, | |
1244 | struct btrfs_multi_bio **multi_ret, | |
1245 | int mirror_num, struct page *unplug_page) | |
0b86a832 CM |
1246 | { |
1247 | struct extent_map *em; | |
1248 | struct map_lookup *map; | |
1249 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
1250 | u64 offset; | |
593060d7 CM |
1251 | u64 stripe_offset; |
1252 | u64 stripe_nr; | |
cea9e445 | 1253 | int stripes_allocated = 8; |
321aecc6 | 1254 | int stripes_required = 1; |
593060d7 | 1255 | int stripe_index; |
cea9e445 | 1256 | int i; |
f2d8d74d | 1257 | int num_stripes; |
cea9e445 | 1258 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 | 1259 | |
cea9e445 CM |
1260 | if (multi_ret && !(rw & (1 << BIO_RW))) { |
1261 | stripes_allocated = 1; | |
1262 | } | |
1263 | again: | |
1264 | if (multi_ret) { | |
1265 | multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), | |
1266 | GFP_NOFS); | |
1267 | if (!multi) | |
1268 | return -ENOMEM; | |
1269 | } | |
0b86a832 CM |
1270 | |
1271 | spin_lock(&em_tree->lock); | |
1272 | em = lookup_extent_mapping(em_tree, logical, *length); | |
b248a415 | 1273 | spin_unlock(&em_tree->lock); |
f2d8d74d CM |
1274 | |
1275 | if (!em && unplug_page) | |
1276 | return 0; | |
1277 | ||
3b951516 CM |
1278 | if (!em) { |
1279 | printk("unable to find logical %Lu\n", logical); | |
f2d8d74d | 1280 | BUG(); |
3b951516 | 1281 | } |
0b86a832 CM |
1282 | |
1283 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
1284 | map = (struct map_lookup *)em->bdev; | |
1285 | offset = logical - em->start; | |
593060d7 | 1286 | |
f188591e CM |
1287 | if (mirror_num > map->num_stripes) |
1288 | mirror_num = 0; | |
1289 | ||
cea9e445 | 1290 | /* if our multi bio struct is too small, back off and try again */ |
321aecc6 CM |
1291 | if (rw & (1 << BIO_RW)) { |
1292 | if (map->type & (BTRFS_BLOCK_GROUP_RAID1 | | |
1293 | BTRFS_BLOCK_GROUP_DUP)) { | |
1294 | stripes_required = map->num_stripes; | |
1295 | } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { | |
1296 | stripes_required = map->sub_stripes; | |
1297 | } | |
1298 | } | |
1299 | if (multi_ret && rw == WRITE && | |
1300 | stripes_allocated < stripes_required) { | |
cea9e445 | 1301 | stripes_allocated = map->num_stripes; |
cea9e445 CM |
1302 | free_extent_map(em); |
1303 | kfree(multi); | |
1304 | goto again; | |
1305 | } | |
593060d7 CM |
1306 | stripe_nr = offset; |
1307 | /* | |
1308 | * stripe_nr counts the total number of stripes we have to stride | |
1309 | * to get to this block | |
1310 | */ | |
1311 | do_div(stripe_nr, map->stripe_len); | |
1312 | ||
1313 | stripe_offset = stripe_nr * map->stripe_len; | |
1314 | BUG_ON(offset < stripe_offset); | |
1315 | ||
1316 | /* stripe_offset is the offset of this block in its stripe*/ | |
1317 | stripe_offset = offset - stripe_offset; | |
1318 | ||
cea9e445 | 1319 | if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | |
321aecc6 | 1320 | BTRFS_BLOCK_GROUP_RAID10 | |
cea9e445 CM |
1321 | BTRFS_BLOCK_GROUP_DUP)) { |
1322 | /* we limit the length of each bio to what fits in a stripe */ | |
1323 | *length = min_t(u64, em->len - offset, | |
1324 | map->stripe_len - stripe_offset); | |
1325 | } else { | |
1326 | *length = em->len - offset; | |
1327 | } | |
f2d8d74d CM |
1328 | |
1329 | if (!multi_ret && !unplug_page) | |
cea9e445 CM |
1330 | goto out; |
1331 | ||
f2d8d74d | 1332 | num_stripes = 1; |
cea9e445 | 1333 | stripe_index = 0; |
8790d502 | 1334 | if (map->type & BTRFS_BLOCK_GROUP_RAID1) { |
f2d8d74d CM |
1335 | if (unplug_page || (rw & (1 << BIO_RW))) |
1336 | num_stripes = map->num_stripes; | |
f188591e CM |
1337 | else if (mirror_num) { |
1338 | stripe_index = mirror_num - 1; | |
1339 | } else { | |
3c12ac72 CM |
1340 | u64 orig_stripe_nr = stripe_nr; |
1341 | stripe_index = do_div(orig_stripe_nr, num_stripes); | |
8790d502 | 1342 | } |
611f0e00 | 1343 | } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { |
cea9e445 | 1344 | if (rw & (1 << BIO_RW)) |
f2d8d74d | 1345 | num_stripes = map->num_stripes; |
f188591e CM |
1346 | else if (mirror_num) |
1347 | stripe_index = mirror_num - 1; | |
321aecc6 CM |
1348 | } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { |
1349 | int factor = map->num_stripes / map->sub_stripes; | |
321aecc6 CM |
1350 | |
1351 | stripe_index = do_div(stripe_nr, factor); | |
1352 | stripe_index *= map->sub_stripes; | |
1353 | ||
f2d8d74d CM |
1354 | if (unplug_page || (rw & (1 << BIO_RW))) |
1355 | num_stripes = map->sub_stripes; | |
321aecc6 CM |
1356 | else if (mirror_num) |
1357 | stripe_index += mirror_num - 1; | |
3c12ac72 CM |
1358 | else { |
1359 | u64 orig_stripe_nr = stripe_nr; | |
1360 | stripe_index += do_div(orig_stripe_nr, | |
1361 | map->sub_stripes); | |
1362 | } | |
8790d502 CM |
1363 | } else { |
1364 | /* | |
1365 | * after this do_div call, stripe_nr is the number of stripes | |
1366 | * on this device we have to walk to find the data, and | |
1367 | * stripe_index is the number of our device in the stripe array | |
1368 | */ | |
1369 | stripe_index = do_div(stripe_nr, map->num_stripes); | |
1370 | } | |
593060d7 | 1371 | BUG_ON(stripe_index >= map->num_stripes); |
cea9e445 | 1372 | |
f2d8d74d CM |
1373 | for (i = 0; i < num_stripes; i++) { |
1374 | if (unplug_page) { | |
1375 | struct btrfs_device *device; | |
1376 | struct backing_dev_info *bdi; | |
1377 | ||
1378 | device = map->stripes[stripe_index].dev; | |
1379 | bdi = blk_get_backing_dev_info(device->bdev); | |
1380 | if (bdi->unplug_io_fn) { | |
1381 | bdi->unplug_io_fn(bdi, unplug_page); | |
1382 | } | |
1383 | } else { | |
1384 | multi->stripes[i].physical = | |
1385 | map->stripes[stripe_index].physical + | |
1386 | stripe_offset + stripe_nr * map->stripe_len; | |
1387 | multi->stripes[i].dev = map->stripes[stripe_index].dev; | |
1388 | } | |
cea9e445 | 1389 | stripe_index++; |
593060d7 | 1390 | } |
f2d8d74d CM |
1391 | if (multi_ret) { |
1392 | *multi_ret = multi; | |
1393 | multi->num_stripes = num_stripes; | |
1394 | } | |
cea9e445 | 1395 | out: |
0b86a832 | 1396 | free_extent_map(em); |
0b86a832 CM |
1397 | return 0; |
1398 | } | |
1399 | ||
f2d8d74d CM |
1400 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
1401 | u64 logical, u64 *length, | |
1402 | struct btrfs_multi_bio **multi_ret, int mirror_num) | |
1403 | { | |
1404 | return __btrfs_map_block(map_tree, rw, logical, length, multi_ret, | |
1405 | mirror_num, NULL); | |
1406 | } | |
1407 | ||
1408 | int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, | |
1409 | u64 logical, struct page *page) | |
1410 | { | |
1411 | u64 length = PAGE_CACHE_SIZE; | |
1412 | return __btrfs_map_block(map_tree, READ, logical, &length, | |
1413 | NULL, 0, page); | |
1414 | } | |
1415 | ||
1416 | ||
8790d502 CM |
1417 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
1418 | static void end_bio_multi_stripe(struct bio *bio, int err) | |
1419 | #else | |
1420 | static int end_bio_multi_stripe(struct bio *bio, | |
1421 | unsigned int bytes_done, int err) | |
1422 | #endif | |
1423 | { | |
cea9e445 | 1424 | struct btrfs_multi_bio *multi = bio->bi_private; |
8790d502 CM |
1425 | |
1426 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1427 | if (bio->bi_size) | |
1428 | return 1; | |
1429 | #endif | |
1430 | if (err) | |
1431 | multi->error = err; | |
1432 | ||
cea9e445 | 1433 | if (atomic_dec_and_test(&multi->stripes_pending)) { |
8790d502 CM |
1434 | bio->bi_private = multi->private; |
1435 | bio->bi_end_io = multi->end_io; | |
1436 | ||
1437 | if (!err && multi->error) | |
1438 | err = multi->error; | |
1439 | kfree(multi); | |
1440 | ||
73f61b2a M |
1441 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) |
1442 | bio_endio(bio, bio->bi_size, err); | |
1443 | #else | |
8790d502 | 1444 | bio_endio(bio, err); |
73f61b2a | 1445 | #endif |
8790d502 CM |
1446 | } else { |
1447 | bio_put(bio); | |
1448 | } | |
1449 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1450 | return 0; | |
1451 | #endif | |
1452 | } | |
1453 | ||
f188591e CM |
1454 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, |
1455 | int mirror_num) | |
0b86a832 CM |
1456 | { |
1457 | struct btrfs_mapping_tree *map_tree; | |
1458 | struct btrfs_device *dev; | |
8790d502 | 1459 | struct bio *first_bio = bio; |
0b86a832 | 1460 | u64 logical = bio->bi_sector << 9; |
0b86a832 CM |
1461 | u64 length = 0; |
1462 | u64 map_length; | |
cea9e445 | 1463 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 | 1464 | int ret; |
8790d502 CM |
1465 | int dev_nr = 0; |
1466 | int total_devs = 1; | |
0b86a832 | 1467 | |
f2d8d74d | 1468 | length = bio->bi_size; |
0b86a832 CM |
1469 | map_tree = &root->fs_info->mapping_tree; |
1470 | map_length = length; | |
cea9e445 | 1471 | |
f188591e CM |
1472 | ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi, |
1473 | mirror_num); | |
cea9e445 CM |
1474 | BUG_ON(ret); |
1475 | ||
1476 | total_devs = multi->num_stripes; | |
1477 | if (map_length < length) { | |
1478 | printk("mapping failed logical %Lu bio len %Lu " | |
1479 | "len %Lu\n", logical, length, map_length); | |
1480 | BUG(); | |
1481 | } | |
1482 | multi->end_io = first_bio->bi_end_io; | |
1483 | multi->private = first_bio->bi_private; | |
1484 | atomic_set(&multi->stripes_pending, multi->num_stripes); | |
1485 | ||
8790d502 | 1486 | while(dev_nr < total_devs) { |
8790d502 | 1487 | if (total_devs > 1) { |
8790d502 CM |
1488 | if (dev_nr < total_devs - 1) { |
1489 | bio = bio_clone(first_bio, GFP_NOFS); | |
1490 | BUG_ON(!bio); | |
1491 | } else { | |
1492 | bio = first_bio; | |
1493 | } | |
1494 | bio->bi_private = multi; | |
1495 | bio->bi_end_io = end_bio_multi_stripe; | |
1496 | } | |
cea9e445 CM |
1497 | bio->bi_sector = multi->stripes[dev_nr].physical >> 9; |
1498 | dev = multi->stripes[dev_nr].dev; | |
e1c4b745 | 1499 | |
8790d502 CM |
1500 | bio->bi_bdev = dev->bdev; |
1501 | spin_lock(&dev->io_lock); | |
1502 | dev->total_ios++; | |
1503 | spin_unlock(&dev->io_lock); | |
1504 | submit_bio(rw, bio); | |
1505 | dev_nr++; | |
1506 | } | |
cea9e445 CM |
1507 | if (total_devs == 1) |
1508 | kfree(multi); | |
0b86a832 CM |
1509 | return 0; |
1510 | } | |
1511 | ||
a443755f CM |
1512 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, |
1513 | u8 *uuid) | |
0b86a832 | 1514 | { |
8a4b83cc | 1515 | struct list_head *head = &root->fs_info->fs_devices->devices; |
0b86a832 | 1516 | |
a443755f | 1517 | return __find_device(head, devid, uuid); |
0b86a832 CM |
1518 | } |
1519 | ||
1520 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, | |
1521 | struct extent_buffer *leaf, | |
1522 | struct btrfs_chunk *chunk) | |
1523 | { | |
1524 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
1525 | struct map_lookup *map; | |
1526 | struct extent_map *em; | |
1527 | u64 logical; | |
1528 | u64 length; | |
1529 | u64 devid; | |
a443755f | 1530 | u8 uuid[BTRFS_UUID_SIZE]; |
593060d7 | 1531 | int num_stripes; |
0b86a832 | 1532 | int ret; |
593060d7 | 1533 | int i; |
0b86a832 | 1534 | |
e17cade2 CM |
1535 | logical = key->offset; |
1536 | length = btrfs_chunk_length(leaf, chunk); | |
0b86a832 CM |
1537 | spin_lock(&map_tree->map_tree.lock); |
1538 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
b248a415 | 1539 | spin_unlock(&map_tree->map_tree.lock); |
0b86a832 CM |
1540 | |
1541 | /* already mapped? */ | |
1542 | if (em && em->start <= logical && em->start + em->len > logical) { | |
1543 | free_extent_map(em); | |
0b86a832 CM |
1544 | return 0; |
1545 | } else if (em) { | |
1546 | free_extent_map(em); | |
1547 | } | |
0b86a832 CM |
1548 | |
1549 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
1550 | if (!map) | |
1551 | return -ENOMEM; | |
1552 | ||
1553 | em = alloc_extent_map(GFP_NOFS); | |
1554 | if (!em) | |
1555 | return -ENOMEM; | |
593060d7 CM |
1556 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); |
1557 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); | |
0b86a832 CM |
1558 | if (!map) { |
1559 | free_extent_map(em); | |
1560 | return -ENOMEM; | |
1561 | } | |
1562 | ||
1563 | em->bdev = (struct block_device *)map; | |
1564 | em->start = logical; | |
1565 | em->len = length; | |
1566 | em->block_start = 0; | |
1567 | ||
593060d7 CM |
1568 | map->num_stripes = num_stripes; |
1569 | map->io_width = btrfs_chunk_io_width(leaf, chunk); | |
1570 | map->io_align = btrfs_chunk_io_align(leaf, chunk); | |
1571 | map->sector_size = btrfs_chunk_sector_size(leaf, chunk); | |
1572 | map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
1573 | map->type = btrfs_chunk_type(leaf, chunk); | |
321aecc6 | 1574 | map->sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk); |
593060d7 CM |
1575 | for (i = 0; i < num_stripes; i++) { |
1576 | map->stripes[i].physical = | |
1577 | btrfs_stripe_offset_nr(leaf, chunk, i); | |
1578 | devid = btrfs_stripe_devid_nr(leaf, chunk, i); | |
a443755f CM |
1579 | read_extent_buffer(leaf, uuid, (unsigned long) |
1580 | btrfs_stripe_dev_uuid_nr(chunk, i), | |
1581 | BTRFS_UUID_SIZE); | |
1582 | map->stripes[i].dev = btrfs_find_device(root, devid, uuid); | |
593060d7 CM |
1583 | if (!map->stripes[i].dev) { |
1584 | kfree(map); | |
1585 | free_extent_map(em); | |
1586 | return -EIO; | |
1587 | } | |
0b86a832 CM |
1588 | } |
1589 | ||
1590 | spin_lock(&map_tree->map_tree.lock); | |
1591 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
0b86a832 | 1592 | spin_unlock(&map_tree->map_tree.lock); |
b248a415 | 1593 | BUG_ON(ret); |
0b86a832 CM |
1594 | free_extent_map(em); |
1595 | ||
1596 | return 0; | |
1597 | } | |
1598 | ||
1599 | static int fill_device_from_item(struct extent_buffer *leaf, | |
1600 | struct btrfs_dev_item *dev_item, | |
1601 | struct btrfs_device *device) | |
1602 | { | |
1603 | unsigned long ptr; | |
0b86a832 CM |
1604 | |
1605 | device->devid = btrfs_device_id(leaf, dev_item); | |
1606 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
1607 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
1608 | device->type = btrfs_device_type(leaf, dev_item); | |
1609 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
1610 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
1611 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
0b86a832 CM |
1612 | |
1613 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
e17cade2 | 1614 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 | 1615 | |
0b86a832 CM |
1616 | return 0; |
1617 | } | |
1618 | ||
0d81ba5d | 1619 | static int read_one_dev(struct btrfs_root *root, |
0b86a832 CM |
1620 | struct extent_buffer *leaf, |
1621 | struct btrfs_dev_item *dev_item) | |
1622 | { | |
1623 | struct btrfs_device *device; | |
1624 | u64 devid; | |
1625 | int ret; | |
a443755f CM |
1626 | u8 dev_uuid[BTRFS_UUID_SIZE]; |
1627 | ||
0b86a832 | 1628 | devid = btrfs_device_id(leaf, dev_item); |
a443755f CM |
1629 | read_extent_buffer(leaf, dev_uuid, |
1630 | (unsigned long)btrfs_device_uuid(dev_item), | |
1631 | BTRFS_UUID_SIZE); | |
1632 | device = btrfs_find_device(root, devid, dev_uuid); | |
6324fbf3 | 1633 | if (!device) { |
8a4b83cc | 1634 | printk("warning devid %Lu not found already\n", devid); |
f2984462 | 1635 | device = kzalloc(sizeof(*device), GFP_NOFS); |
6324fbf3 CM |
1636 | if (!device) |
1637 | return -ENOMEM; | |
8a4b83cc CM |
1638 | list_add(&device->dev_list, |
1639 | &root->fs_info->fs_devices->devices); | |
b3075717 CM |
1640 | list_add(&device->dev_alloc_list, |
1641 | &root->fs_info->fs_devices->alloc_list); | |
b248a415 | 1642 | device->barriers = 1; |
8790d502 | 1643 | spin_lock_init(&device->io_lock); |
6324fbf3 | 1644 | } |
0b86a832 CM |
1645 | |
1646 | fill_device_from_item(leaf, dev_item, device); | |
1647 | device->dev_root = root->fs_info->dev_root; | |
0b86a832 CM |
1648 | ret = 0; |
1649 | #if 0 | |
1650 | ret = btrfs_open_device(device); | |
1651 | if (ret) { | |
1652 | kfree(device); | |
1653 | } | |
1654 | #endif | |
1655 | return ret; | |
1656 | } | |
1657 | ||
0d81ba5d CM |
1658 | int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) |
1659 | { | |
1660 | struct btrfs_dev_item *dev_item; | |
1661 | ||
1662 | dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, | |
1663 | dev_item); | |
1664 | return read_one_dev(root, buf, dev_item); | |
1665 | } | |
1666 | ||
0b86a832 CM |
1667 | int btrfs_read_sys_array(struct btrfs_root *root) |
1668 | { | |
1669 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1670 | struct extent_buffer *sb = root->fs_info->sb_buffer; | |
1671 | struct btrfs_disk_key *disk_key; | |
0b86a832 | 1672 | struct btrfs_chunk *chunk; |
84eed90f CM |
1673 | u8 *ptr; |
1674 | unsigned long sb_ptr; | |
1675 | int ret = 0; | |
0b86a832 CM |
1676 | u32 num_stripes; |
1677 | u32 array_size; | |
1678 | u32 len = 0; | |
0b86a832 | 1679 | u32 cur; |
84eed90f | 1680 | struct btrfs_key key; |
0b86a832 CM |
1681 | |
1682 | array_size = btrfs_super_sys_array_size(super_copy); | |
1683 | ||
0b86a832 CM |
1684 | ptr = super_copy->sys_chunk_array; |
1685 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
1686 | cur = 0; | |
1687 | ||
1688 | while (cur < array_size) { | |
1689 | disk_key = (struct btrfs_disk_key *)ptr; | |
1690 | btrfs_disk_key_to_cpu(&key, disk_key); | |
1691 | ||
1692 | len = sizeof(*disk_key); | |
1693 | ptr += len; | |
1694 | sb_ptr += len; | |
1695 | cur += len; | |
1696 | ||
0d81ba5d | 1697 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { |
0b86a832 | 1698 | chunk = (struct btrfs_chunk *)sb_ptr; |
0d81ba5d | 1699 | ret = read_one_chunk(root, &key, sb, chunk); |
84eed90f CM |
1700 | if (ret) |
1701 | break; | |
0b86a832 CM |
1702 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); |
1703 | len = btrfs_chunk_item_size(num_stripes); | |
1704 | } else { | |
84eed90f CM |
1705 | ret = -EIO; |
1706 | break; | |
0b86a832 CM |
1707 | } |
1708 | ptr += len; | |
1709 | sb_ptr += len; | |
1710 | cur += len; | |
1711 | } | |
84eed90f | 1712 | return ret; |
0b86a832 CM |
1713 | } |
1714 | ||
1715 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
1716 | { | |
1717 | struct btrfs_path *path; | |
1718 | struct extent_buffer *leaf; | |
1719 | struct btrfs_key key; | |
1720 | struct btrfs_key found_key; | |
1721 | int ret; | |
1722 | int slot; | |
1723 | ||
1724 | root = root->fs_info->chunk_root; | |
1725 | ||
1726 | path = btrfs_alloc_path(); | |
1727 | if (!path) | |
1728 | return -ENOMEM; | |
1729 | ||
1730 | /* first we search for all of the device items, and then we | |
1731 | * read in all of the chunk items. This way we can create chunk | |
1732 | * mappings that reference all of the devices that are afound | |
1733 | */ | |
1734 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1735 | key.offset = 0; | |
1736 | key.type = 0; | |
1737 | again: | |
1738 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1739 | while(1) { | |
1740 | leaf = path->nodes[0]; | |
1741 | slot = path->slots[0]; | |
1742 | if (slot >= btrfs_header_nritems(leaf)) { | |
1743 | ret = btrfs_next_leaf(root, path); | |
1744 | if (ret == 0) | |
1745 | continue; | |
1746 | if (ret < 0) | |
1747 | goto error; | |
1748 | break; | |
1749 | } | |
1750 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1751 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1752 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
1753 | break; | |
1754 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
1755 | struct btrfs_dev_item *dev_item; | |
1756 | dev_item = btrfs_item_ptr(leaf, slot, | |
1757 | struct btrfs_dev_item); | |
0d81ba5d | 1758 | ret = read_one_dev(root, leaf, dev_item); |
0b86a832 CM |
1759 | BUG_ON(ret); |
1760 | } | |
1761 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
1762 | struct btrfs_chunk *chunk; | |
1763 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
1764 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
1765 | } | |
1766 | path->slots[0]++; | |
1767 | } | |
1768 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1769 | key.objectid = 0; | |
1770 | btrfs_release_path(root, path); | |
1771 | goto again; | |
1772 | } | |
1773 | ||
1774 | btrfs_free_path(path); | |
1775 | ret = 0; | |
1776 | error: | |
1777 | return ret; | |
1778 | } | |
1779 |