]>
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> |
593060d7 | 21 | #include <asm/div64.h> |
0b86a832 CM |
22 | #include "ctree.h" |
23 | #include "extent_map.h" | |
24 | #include "disk-io.h" | |
25 | #include "transaction.h" | |
26 | #include "print-tree.h" | |
27 | #include "volumes.h" | |
28 | ||
593060d7 CM |
29 | struct map_lookup { |
30 | u64 type; | |
31 | int io_align; | |
32 | int io_width; | |
33 | int stripe_len; | |
34 | int sector_size; | |
35 | int num_stripes; | |
321aecc6 | 36 | int sub_stripes; |
cea9e445 | 37 | struct btrfs_bio_stripe stripes[]; |
593060d7 CM |
38 | }; |
39 | ||
40 | #define map_lookup_size(n) (sizeof(struct map_lookup) + \ | |
cea9e445 | 41 | (sizeof(struct btrfs_bio_stripe) * (n))) |
593060d7 | 42 | |
8a4b83cc CM |
43 | static DEFINE_MUTEX(uuid_mutex); |
44 | static LIST_HEAD(fs_uuids); | |
45 | ||
46 | int btrfs_cleanup_fs_uuids(void) | |
47 | { | |
48 | struct btrfs_fs_devices *fs_devices; | |
49 | struct list_head *uuid_cur; | |
50 | struct list_head *devices_cur; | |
51 | struct btrfs_device *dev; | |
52 | ||
53 | list_for_each(uuid_cur, &fs_uuids) { | |
54 | fs_devices = list_entry(uuid_cur, struct btrfs_fs_devices, | |
55 | list); | |
56 | while(!list_empty(&fs_devices->devices)) { | |
57 | devices_cur = fs_devices->devices.next; | |
58 | dev = list_entry(devices_cur, struct btrfs_device, | |
59 | dev_list); | |
60 | printk("uuid cleanup finds %s\n", dev->name); | |
61 | if (dev->bdev) { | |
62 | printk("closing\n"); | |
63 | close_bdev_excl(dev->bdev); | |
64 | } | |
65 | list_del(&dev->dev_list); | |
66 | kfree(dev); | |
67 | } | |
68 | } | |
69 | return 0; | |
70 | } | |
71 | ||
a443755f CM |
72 | static struct btrfs_device *__find_device(struct list_head *head, u64 devid, |
73 | u8 *uuid) | |
8a4b83cc CM |
74 | { |
75 | struct btrfs_device *dev; | |
76 | struct list_head *cur; | |
77 | ||
78 | list_for_each(cur, head) { | |
79 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
a443755f CM |
80 | if (dev->devid == devid && |
81 | !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE)) { | |
8a4b83cc | 82 | return dev; |
a443755f | 83 | } |
8a4b83cc CM |
84 | } |
85 | return NULL; | |
86 | } | |
87 | ||
88 | static struct btrfs_fs_devices *find_fsid(u8 *fsid) | |
89 | { | |
90 | struct list_head *cur; | |
91 | struct btrfs_fs_devices *fs_devices; | |
92 | ||
93 | list_for_each(cur, &fs_uuids) { | |
94 | fs_devices = list_entry(cur, struct btrfs_fs_devices, list); | |
95 | if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0) | |
96 | return fs_devices; | |
97 | } | |
98 | return NULL; | |
99 | } | |
100 | ||
101 | static int device_list_add(const char *path, | |
102 | struct btrfs_super_block *disk_super, | |
103 | u64 devid, struct btrfs_fs_devices **fs_devices_ret) | |
104 | { | |
105 | struct btrfs_device *device; | |
106 | struct btrfs_fs_devices *fs_devices; | |
107 | u64 found_transid = btrfs_super_generation(disk_super); | |
108 | ||
109 | fs_devices = find_fsid(disk_super->fsid); | |
110 | if (!fs_devices) { | |
111 | fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS); | |
112 | if (!fs_devices) | |
113 | return -ENOMEM; | |
114 | INIT_LIST_HEAD(&fs_devices->devices); | |
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); | |
143 | fs_devices->num_devices++; | |
144 | } | |
145 | ||
146 | if (found_transid > fs_devices->latest_trans) { | |
147 | fs_devices->latest_devid = devid; | |
148 | fs_devices->latest_trans = found_transid; | |
149 | } | |
150 | if (fs_devices->lowest_devid > devid) { | |
151 | fs_devices->lowest_devid = devid; | |
152 | printk("lowest devid now %Lu\n", devid); | |
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); | |
169 | printk("close devices closes %s\n", device->name); | |
170 | } | |
171 | device->bdev = NULL; | |
172 | } | |
173 | mutex_unlock(&uuid_mutex); | |
174 | return 0; | |
175 | } | |
176 | ||
177 | int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, | |
178 | int flags, void *holder) | |
179 | { | |
180 | struct block_device *bdev; | |
181 | struct list_head *head = &fs_devices->devices; | |
182 | struct list_head *cur; | |
183 | struct btrfs_device *device; | |
184 | int ret; | |
185 | ||
186 | mutex_lock(&uuid_mutex); | |
187 | list_for_each(cur, head) { | |
188 | device = list_entry(cur, struct btrfs_device, dev_list); | |
189 | bdev = open_bdev_excl(device->name, flags, holder); | |
e17cade2 | 190 | |
8a4b83cc CM |
191 | if (IS_ERR(bdev)) { |
192 | printk("open %s failed\n", device->name); | |
193 | ret = PTR_ERR(bdev); | |
194 | goto fail; | |
195 | } | |
196 | if (device->devid == fs_devices->latest_devid) | |
197 | fs_devices->latest_bdev = bdev; | |
198 | if (device->devid == fs_devices->lowest_devid) { | |
199 | fs_devices->lowest_bdev = bdev; | |
8a4b83cc CM |
200 | } |
201 | device->bdev = bdev; | |
202 | } | |
203 | mutex_unlock(&uuid_mutex); | |
204 | return 0; | |
205 | fail: | |
206 | mutex_unlock(&uuid_mutex); | |
207 | btrfs_close_devices(fs_devices); | |
208 | return ret; | |
209 | } | |
210 | ||
211 | int btrfs_scan_one_device(const char *path, int flags, void *holder, | |
212 | struct btrfs_fs_devices **fs_devices_ret) | |
213 | { | |
214 | struct btrfs_super_block *disk_super; | |
215 | struct block_device *bdev; | |
216 | struct buffer_head *bh; | |
217 | int ret; | |
218 | u64 devid; | |
f2984462 | 219 | u64 transid; |
8a4b83cc CM |
220 | |
221 | mutex_lock(&uuid_mutex); | |
222 | ||
223 | printk("scan one opens %s\n", path); | |
224 | bdev = open_bdev_excl(path, flags, holder); | |
225 | ||
226 | if (IS_ERR(bdev)) { | |
227 | printk("open failed\n"); | |
228 | ret = PTR_ERR(bdev); | |
229 | goto error; | |
230 | } | |
231 | ||
232 | ret = set_blocksize(bdev, 4096); | |
233 | if (ret) | |
234 | goto error_close; | |
235 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
236 | if (!bh) { | |
237 | ret = -EIO; | |
238 | goto error_close; | |
239 | } | |
240 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
241 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
242 | sizeof(disk_super->magic))) { | |
243 | printk("no btrfs found on %s\n", path); | |
e58ca020 | 244 | ret = -EINVAL; |
8a4b83cc CM |
245 | goto error_brelse; |
246 | } | |
247 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
f2984462 CM |
248 | transid = btrfs_super_generation(disk_super); |
249 | printk("found device %Lu transid %Lu on %s\n", devid, transid, path); | |
8a4b83cc CM |
250 | ret = device_list_add(path, disk_super, devid, fs_devices_ret); |
251 | ||
252 | error_brelse: | |
253 | brelse(bh); | |
254 | error_close: | |
255 | close_bdev_excl(bdev); | |
8a4b83cc CM |
256 | error: |
257 | mutex_unlock(&uuid_mutex); | |
258 | return ret; | |
259 | } | |
0b86a832 CM |
260 | |
261 | /* | |
262 | * this uses a pretty simple search, the expectation is that it is | |
263 | * called very infrequently and that a given device has a small number | |
264 | * of extents | |
265 | */ | |
266 | static int find_free_dev_extent(struct btrfs_trans_handle *trans, | |
267 | struct btrfs_device *device, | |
268 | struct btrfs_path *path, | |
269 | u64 num_bytes, u64 *start) | |
270 | { | |
271 | struct btrfs_key key; | |
272 | struct btrfs_root *root = device->dev_root; | |
273 | struct btrfs_dev_extent *dev_extent = NULL; | |
274 | u64 hole_size = 0; | |
275 | u64 last_byte = 0; | |
276 | u64 search_start = 0; | |
277 | u64 search_end = device->total_bytes; | |
278 | int ret; | |
279 | int slot = 0; | |
280 | int start_found; | |
281 | struct extent_buffer *l; | |
282 | ||
283 | start_found = 0; | |
284 | path->reada = 2; | |
285 | ||
286 | /* FIXME use last free of some kind */ | |
287 | ||
8a4b83cc CM |
288 | /* we don't want to overwrite the superblock on the drive, |
289 | * so we make sure to start at an offset of at least 1MB | |
290 | */ | |
291 | search_start = max((u64)1024 * 1024, search_start); | |
0b86a832 CM |
292 | key.objectid = device->devid; |
293 | key.offset = search_start; | |
294 | key.type = BTRFS_DEV_EXTENT_KEY; | |
295 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
296 | if (ret < 0) | |
297 | goto error; | |
298 | ret = btrfs_previous_item(root, path, 0, key.type); | |
299 | if (ret < 0) | |
300 | goto error; | |
301 | l = path->nodes[0]; | |
302 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
303 | while (1) { | |
304 | l = path->nodes[0]; | |
305 | slot = path->slots[0]; | |
306 | if (slot >= btrfs_header_nritems(l)) { | |
307 | ret = btrfs_next_leaf(root, path); | |
308 | if (ret == 0) | |
309 | continue; | |
310 | if (ret < 0) | |
311 | goto error; | |
312 | no_more_items: | |
313 | if (!start_found) { | |
314 | if (search_start >= search_end) { | |
315 | ret = -ENOSPC; | |
316 | goto error; | |
317 | } | |
318 | *start = search_start; | |
319 | start_found = 1; | |
320 | goto check_pending; | |
321 | } | |
322 | *start = last_byte > search_start ? | |
323 | last_byte : search_start; | |
324 | if (search_end <= *start) { | |
325 | ret = -ENOSPC; | |
326 | goto error; | |
327 | } | |
328 | goto check_pending; | |
329 | } | |
330 | btrfs_item_key_to_cpu(l, &key, slot); | |
331 | ||
332 | if (key.objectid < device->devid) | |
333 | goto next; | |
334 | ||
335 | if (key.objectid > device->devid) | |
336 | goto no_more_items; | |
337 | ||
338 | if (key.offset >= search_start && key.offset > last_byte && | |
339 | start_found) { | |
340 | if (last_byte < search_start) | |
341 | last_byte = search_start; | |
342 | hole_size = key.offset - last_byte; | |
343 | if (key.offset > last_byte && | |
344 | hole_size >= num_bytes) { | |
345 | *start = last_byte; | |
346 | goto check_pending; | |
347 | } | |
348 | } | |
349 | if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { | |
350 | goto next; | |
351 | } | |
352 | ||
353 | start_found = 1; | |
354 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
355 | last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); | |
356 | next: | |
357 | path->slots[0]++; | |
358 | cond_resched(); | |
359 | } | |
360 | check_pending: | |
361 | /* we have to make sure we didn't find an extent that has already | |
362 | * been allocated by the map tree or the original allocation | |
363 | */ | |
364 | btrfs_release_path(root, path); | |
365 | BUG_ON(*start < search_start); | |
366 | ||
6324fbf3 | 367 | if (*start + num_bytes > search_end) { |
0b86a832 CM |
368 | ret = -ENOSPC; |
369 | goto error; | |
370 | } | |
371 | /* check for pending inserts here */ | |
372 | return 0; | |
373 | ||
374 | error: | |
375 | btrfs_release_path(root, path); | |
376 | return ret; | |
377 | } | |
378 | ||
379 | int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, | |
380 | struct btrfs_device *device, | |
e17cade2 CM |
381 | u64 chunk_tree, u64 chunk_objectid, |
382 | u64 chunk_offset, | |
383 | u64 num_bytes, u64 *start) | |
0b86a832 CM |
384 | { |
385 | int ret; | |
386 | struct btrfs_path *path; | |
387 | struct btrfs_root *root = device->dev_root; | |
388 | struct btrfs_dev_extent *extent; | |
389 | struct extent_buffer *leaf; | |
390 | struct btrfs_key key; | |
391 | ||
392 | path = btrfs_alloc_path(); | |
393 | if (!path) | |
394 | return -ENOMEM; | |
395 | ||
396 | ret = find_free_dev_extent(trans, device, path, num_bytes, start); | |
6324fbf3 | 397 | if (ret) { |
0b86a832 | 398 | goto err; |
6324fbf3 | 399 | } |
0b86a832 CM |
400 | |
401 | key.objectid = device->devid; | |
402 | key.offset = *start; | |
403 | key.type = BTRFS_DEV_EXTENT_KEY; | |
404 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
405 | sizeof(*extent)); | |
406 | BUG_ON(ret); | |
407 | ||
408 | leaf = path->nodes[0]; | |
409 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
410 | struct btrfs_dev_extent); | |
e17cade2 CM |
411 | btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree); |
412 | btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid); | |
413 | btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset); | |
414 | ||
415 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, | |
416 | (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent), | |
417 | BTRFS_UUID_SIZE); | |
418 | ||
0b86a832 CM |
419 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); |
420 | btrfs_mark_buffer_dirty(leaf); | |
421 | err: | |
422 | btrfs_free_path(path); | |
423 | return ret; | |
424 | } | |
425 | ||
e17cade2 | 426 | static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset) |
0b86a832 CM |
427 | { |
428 | struct btrfs_path *path; | |
429 | int ret; | |
430 | struct btrfs_key key; | |
e17cade2 | 431 | struct btrfs_chunk *chunk; |
0b86a832 CM |
432 | struct btrfs_key found_key; |
433 | ||
434 | path = btrfs_alloc_path(); | |
435 | BUG_ON(!path); | |
436 | ||
e17cade2 | 437 | key.objectid = objectid; |
0b86a832 CM |
438 | key.offset = (u64)-1; |
439 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
440 | ||
441 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
442 | if (ret < 0) | |
443 | goto error; | |
444 | ||
445 | BUG_ON(ret == 0); | |
446 | ||
447 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
448 | if (ret) { | |
e17cade2 | 449 | *offset = 0; |
0b86a832 CM |
450 | } else { |
451 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
452 | path->slots[0]); | |
e17cade2 CM |
453 | if (found_key.objectid != objectid) |
454 | *offset = 0; | |
455 | else { | |
456 | chunk = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
457 | struct btrfs_chunk); | |
458 | *offset = found_key.offset + | |
459 | btrfs_chunk_length(path->nodes[0], chunk); | |
460 | } | |
0b86a832 CM |
461 | } |
462 | ret = 0; | |
463 | error: | |
464 | btrfs_free_path(path); | |
465 | return ret; | |
466 | } | |
467 | ||
0b86a832 CM |
468 | static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, |
469 | u64 *objectid) | |
470 | { | |
471 | int ret; | |
472 | struct btrfs_key key; | |
473 | struct btrfs_key found_key; | |
474 | ||
475 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
476 | key.type = BTRFS_DEV_ITEM_KEY; | |
477 | key.offset = (u64)-1; | |
478 | ||
479 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
480 | if (ret < 0) | |
481 | goto error; | |
482 | ||
483 | BUG_ON(ret == 0); | |
484 | ||
485 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
486 | BTRFS_DEV_ITEM_KEY); | |
487 | if (ret) { | |
488 | *objectid = 1; | |
489 | } else { | |
490 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
491 | path->slots[0]); | |
492 | *objectid = found_key.offset + 1; | |
493 | } | |
494 | ret = 0; | |
495 | error: | |
496 | btrfs_release_path(root, path); | |
497 | return ret; | |
498 | } | |
499 | ||
500 | /* | |
501 | * the device information is stored in the chunk root | |
502 | * the btrfs_device struct should be fully filled in | |
503 | */ | |
504 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
505 | struct btrfs_root *root, | |
506 | struct btrfs_device *device) | |
507 | { | |
508 | int ret; | |
509 | struct btrfs_path *path; | |
510 | struct btrfs_dev_item *dev_item; | |
511 | struct extent_buffer *leaf; | |
512 | struct btrfs_key key; | |
513 | unsigned long ptr; | |
514 | u64 free_devid; | |
515 | ||
516 | root = root->fs_info->chunk_root; | |
517 | ||
518 | path = btrfs_alloc_path(); | |
519 | if (!path) | |
520 | return -ENOMEM; | |
521 | ||
522 | ret = find_next_devid(root, path, &free_devid); | |
523 | if (ret) | |
524 | goto out; | |
525 | ||
526 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
527 | key.type = BTRFS_DEV_ITEM_KEY; | |
528 | key.offset = free_devid; | |
529 | ||
530 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
0d81ba5d | 531 | sizeof(*dev_item)); |
0b86a832 CM |
532 | if (ret) |
533 | goto out; | |
534 | ||
535 | leaf = path->nodes[0]; | |
536 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
537 | ||
8a4b83cc | 538 | device->devid = free_devid; |
0b86a832 CM |
539 | btrfs_set_device_id(leaf, dev_item, device->devid); |
540 | btrfs_set_device_type(leaf, dev_item, device->type); | |
541 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
542 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
543 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
544 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
545 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
e17cade2 CM |
546 | btrfs_set_device_group(leaf, dev_item, 0); |
547 | btrfs_set_device_seek_speed(leaf, dev_item, 0); | |
548 | btrfs_set_device_bandwidth(leaf, dev_item, 0); | |
0b86a832 | 549 | |
0b86a832 | 550 | ptr = (unsigned long)btrfs_device_uuid(dev_item); |
e17cade2 | 551 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 CM |
552 | btrfs_mark_buffer_dirty(leaf); |
553 | ret = 0; | |
554 | ||
555 | out: | |
556 | btrfs_free_path(path); | |
557 | return ret; | |
558 | } | |
559 | int btrfs_update_device(struct btrfs_trans_handle *trans, | |
560 | struct btrfs_device *device) | |
561 | { | |
562 | int ret; | |
563 | struct btrfs_path *path; | |
564 | struct btrfs_root *root; | |
565 | struct btrfs_dev_item *dev_item; | |
566 | struct extent_buffer *leaf; | |
567 | struct btrfs_key key; | |
568 | ||
569 | root = device->dev_root->fs_info->chunk_root; | |
570 | ||
571 | path = btrfs_alloc_path(); | |
572 | if (!path) | |
573 | return -ENOMEM; | |
574 | ||
575 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
576 | key.type = BTRFS_DEV_ITEM_KEY; | |
577 | key.offset = device->devid; | |
578 | ||
579 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
580 | if (ret < 0) | |
581 | goto out; | |
582 | ||
583 | if (ret > 0) { | |
584 | ret = -ENOENT; | |
585 | goto out; | |
586 | } | |
587 | ||
588 | leaf = path->nodes[0]; | |
589 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
590 | ||
591 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
592 | btrfs_set_device_type(leaf, dev_item, device->type); | |
593 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
594 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
595 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
596 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
597 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
598 | btrfs_mark_buffer_dirty(leaf); | |
599 | ||
600 | out: | |
601 | btrfs_free_path(path); | |
602 | return ret; | |
603 | } | |
604 | ||
605 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, | |
606 | struct btrfs_root *root, | |
607 | struct btrfs_key *key, | |
608 | struct btrfs_chunk *chunk, int item_size) | |
609 | { | |
610 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
611 | struct btrfs_disk_key disk_key; | |
612 | u32 array_size; | |
613 | u8 *ptr; | |
614 | ||
615 | array_size = btrfs_super_sys_array_size(super_copy); | |
616 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
617 | return -EFBIG; | |
618 | ||
619 | ptr = super_copy->sys_chunk_array + array_size; | |
620 | btrfs_cpu_key_to_disk(&disk_key, key); | |
621 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
622 | ptr += sizeof(disk_key); | |
623 | memcpy(ptr, chunk, item_size); | |
624 | item_size += sizeof(disk_key); | |
625 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
626 | return 0; | |
627 | } | |
628 | ||
629 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, | |
630 | struct btrfs_root *extent_root, u64 *start, | |
6324fbf3 | 631 | u64 *num_bytes, u64 type) |
0b86a832 CM |
632 | { |
633 | u64 dev_offset; | |
593060d7 | 634 | struct btrfs_fs_info *info = extent_root->fs_info; |
0b86a832 CM |
635 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; |
636 | struct btrfs_stripe *stripes; | |
637 | struct btrfs_device *device = NULL; | |
638 | struct btrfs_chunk *chunk; | |
6324fbf3 | 639 | struct list_head private_devs; |
8a4b83cc | 640 | struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices; |
6324fbf3 | 641 | struct list_head *cur; |
0b86a832 CM |
642 | struct extent_map_tree *em_tree; |
643 | struct map_lookup *map; | |
644 | struct extent_map *em; | |
645 | u64 physical; | |
646 | u64 calc_size = 1024 * 1024 * 1024; | |
611f0e00 | 647 | u64 min_free = calc_size; |
6324fbf3 CM |
648 | u64 avail; |
649 | u64 max_avail = 0; | |
650 | int num_stripes = 1; | |
321aecc6 | 651 | int sub_stripes = 0; |
6324fbf3 | 652 | int looped = 0; |
0b86a832 | 653 | int ret; |
6324fbf3 | 654 | int index; |
593060d7 | 655 | int stripe_len = 64 * 1024; |
0b86a832 CM |
656 | struct btrfs_key key; |
657 | ||
6324fbf3 CM |
658 | if (list_empty(dev_list)) |
659 | return -ENOSPC; | |
593060d7 | 660 | |
8790d502 | 661 | if (type & (BTRFS_BLOCK_GROUP_RAID0)) |
593060d7 | 662 | num_stripes = btrfs_super_num_devices(&info->super_copy); |
611f0e00 CM |
663 | if (type & (BTRFS_BLOCK_GROUP_DUP)) |
664 | num_stripes = 2; | |
8790d502 CM |
665 | if (type & (BTRFS_BLOCK_GROUP_RAID1)) { |
666 | num_stripes = min_t(u64, 2, | |
667 | btrfs_super_num_devices(&info->super_copy)); | |
668 | } | |
321aecc6 CM |
669 | if (type & (BTRFS_BLOCK_GROUP_RAID10)) { |
670 | num_stripes = btrfs_super_num_devices(&info->super_copy); | |
671 | if (num_stripes < 4) | |
672 | return -ENOSPC; | |
673 | num_stripes &= ~(u32)1; | |
674 | sub_stripes = 2; | |
675 | } | |
6324fbf3 CM |
676 | again: |
677 | INIT_LIST_HEAD(&private_devs); | |
678 | cur = dev_list->next; | |
679 | index = 0; | |
611f0e00 CM |
680 | |
681 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
682 | min_free = calc_size * 2; | |
683 | ||
6324fbf3 CM |
684 | /* build a private list of devices we will allocate from */ |
685 | while(index < num_stripes) { | |
686 | device = list_entry(cur, struct btrfs_device, dev_list); | |
611f0e00 | 687 | |
6324fbf3 CM |
688 | avail = device->total_bytes - device->bytes_used; |
689 | cur = cur->next; | |
690 | if (avail > max_avail) | |
691 | max_avail = avail; | |
611f0e00 | 692 | if (avail >= min_free) { |
6324fbf3 CM |
693 | list_move_tail(&device->dev_list, &private_devs); |
694 | index++; | |
611f0e00 CM |
695 | if (type & BTRFS_BLOCK_GROUP_DUP) |
696 | index++; | |
6324fbf3 CM |
697 | } |
698 | if (cur == dev_list) | |
699 | break; | |
700 | } | |
701 | if (index < num_stripes) { | |
702 | list_splice(&private_devs, dev_list); | |
703 | if (!looped && max_avail > 0) { | |
704 | looped = 1; | |
705 | calc_size = max_avail; | |
706 | goto again; | |
707 | } | |
708 | return -ENOSPC; | |
709 | } | |
0b86a832 | 710 | |
e17cade2 CM |
711 | key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; |
712 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
713 | ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID, | |
714 | &key.offset); | |
0b86a832 CM |
715 | if (ret) |
716 | return ret; | |
717 | ||
0b86a832 CM |
718 | chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); |
719 | if (!chunk) | |
720 | return -ENOMEM; | |
721 | ||
593060d7 CM |
722 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); |
723 | if (!map) { | |
724 | kfree(chunk); | |
725 | return -ENOMEM; | |
726 | } | |
727 | ||
0b86a832 CM |
728 | stripes = &chunk->stripe; |
729 | ||
611f0e00 | 730 | if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP)) |
8790d502 | 731 | *num_bytes = calc_size; |
321aecc6 | 732 | else if (type & BTRFS_BLOCK_GROUP_RAID10) |
7bf3b490 | 733 | *num_bytes = calc_size * (num_stripes / sub_stripes); |
8790d502 CM |
734 | else |
735 | *num_bytes = calc_size * num_stripes; | |
736 | ||
6324fbf3 | 737 | index = 0; |
e17cade2 | 738 | printk("new chunk type %Lu start %Lu size %Lu\n", type, key.offset, *num_bytes); |
0b86a832 | 739 | while(index < num_stripes) { |
e17cade2 | 740 | struct btrfs_stripe *stripe; |
6324fbf3 CM |
741 | BUG_ON(list_empty(&private_devs)); |
742 | cur = private_devs.next; | |
743 | device = list_entry(cur, struct btrfs_device, dev_list); | |
611f0e00 CM |
744 | |
745 | /* loop over this device again if we're doing a dup group */ | |
746 | if (!(type & BTRFS_BLOCK_GROUP_DUP) || | |
747 | (index == num_stripes - 1)) | |
748 | list_move_tail(&device->dev_list, dev_list); | |
0b86a832 CM |
749 | |
750 | ret = btrfs_alloc_dev_extent(trans, device, | |
e17cade2 CM |
751 | info->chunk_root->root_key.objectid, |
752 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset, | |
753 | calc_size, &dev_offset); | |
0b86a832 | 754 | BUG_ON(ret); |
e17cade2 | 755 | printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key.offset, calc_size, device->devid, type); |
0b86a832 CM |
756 | device->bytes_used += calc_size; |
757 | ret = btrfs_update_device(trans, device); | |
758 | BUG_ON(ret); | |
759 | ||
593060d7 CM |
760 | map->stripes[index].dev = device; |
761 | map->stripes[index].physical = dev_offset; | |
e17cade2 CM |
762 | stripe = stripes + index; |
763 | btrfs_set_stack_stripe_devid(stripe, device->devid); | |
764 | btrfs_set_stack_stripe_offset(stripe, dev_offset); | |
765 | memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE); | |
0b86a832 CM |
766 | physical = dev_offset; |
767 | index++; | |
768 | } | |
6324fbf3 | 769 | BUG_ON(!list_empty(&private_devs)); |
0b86a832 | 770 | |
e17cade2 CM |
771 | /* key was set above */ |
772 | btrfs_set_stack_chunk_length(chunk, *num_bytes); | |
0b86a832 | 773 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); |
593060d7 | 774 | btrfs_set_stack_chunk_stripe_len(chunk, stripe_len); |
0b86a832 CM |
775 | btrfs_set_stack_chunk_type(chunk, type); |
776 | btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); | |
593060d7 CM |
777 | btrfs_set_stack_chunk_io_align(chunk, stripe_len); |
778 | btrfs_set_stack_chunk_io_width(chunk, stripe_len); | |
0b86a832 | 779 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); |
321aecc6 | 780 | btrfs_set_stack_chunk_sub_stripes(chunk, sub_stripes); |
593060d7 CM |
781 | map->sector_size = extent_root->sectorsize; |
782 | map->stripe_len = stripe_len; | |
783 | map->io_align = stripe_len; | |
784 | map->io_width = stripe_len; | |
785 | map->type = type; | |
786 | map->num_stripes = num_stripes; | |
321aecc6 | 787 | map->sub_stripes = sub_stripes; |
0b86a832 CM |
788 | |
789 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, | |
790 | btrfs_chunk_item_size(num_stripes)); | |
791 | BUG_ON(ret); | |
e17cade2 | 792 | *start = key.offset;; |
0b86a832 CM |
793 | |
794 | em = alloc_extent_map(GFP_NOFS); | |
795 | if (!em) | |
796 | return -ENOMEM; | |
0b86a832 | 797 | em->bdev = (struct block_device *)map; |
e17cade2 CM |
798 | em->start = key.offset; |
799 | em->len = *num_bytes; | |
0b86a832 CM |
800 | em->block_start = 0; |
801 | ||
0b86a832 CM |
802 | kfree(chunk); |
803 | ||
804 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; | |
805 | spin_lock(&em_tree->lock); | |
806 | ret = add_extent_mapping(em_tree, em); | |
0b86a832 | 807 | spin_unlock(&em_tree->lock); |
b248a415 | 808 | BUG_ON(ret); |
0b86a832 CM |
809 | free_extent_map(em); |
810 | return ret; | |
811 | } | |
812 | ||
813 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
814 | { | |
815 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
816 | } | |
817 | ||
818 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
819 | { | |
820 | struct extent_map *em; | |
821 | ||
822 | while(1) { | |
823 | spin_lock(&tree->map_tree.lock); | |
824 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
825 | if (em) | |
826 | remove_extent_mapping(&tree->map_tree, em); | |
827 | spin_unlock(&tree->map_tree.lock); | |
828 | if (!em) | |
829 | break; | |
830 | kfree(em->bdev); | |
831 | /* once for us */ | |
832 | free_extent_map(em); | |
833 | /* once for the tree */ | |
834 | free_extent_map(em); | |
835 | } | |
836 | } | |
837 | ||
f188591e CM |
838 | int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) |
839 | { | |
840 | struct extent_map *em; | |
841 | struct map_lookup *map; | |
842 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
843 | int ret; | |
844 | ||
845 | spin_lock(&em_tree->lock); | |
846 | em = lookup_extent_mapping(em_tree, logical, len); | |
b248a415 | 847 | spin_unlock(&em_tree->lock); |
f188591e CM |
848 | BUG_ON(!em); |
849 | ||
850 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
851 | map = (struct map_lookup *)em->bdev; | |
852 | if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1)) | |
853 | ret = map->num_stripes; | |
321aecc6 CM |
854 | else if (map->type & BTRFS_BLOCK_GROUP_RAID10) |
855 | ret = map->sub_stripes; | |
f188591e CM |
856 | else |
857 | ret = 1; | |
858 | free_extent_map(em); | |
f188591e CM |
859 | return ret; |
860 | } | |
861 | ||
8790d502 | 862 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
cea9e445 | 863 | u64 logical, u64 *length, |
f188591e | 864 | struct btrfs_multi_bio **multi_ret, int mirror_num) |
0b86a832 CM |
865 | { |
866 | struct extent_map *em; | |
867 | struct map_lookup *map; | |
868 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
869 | u64 offset; | |
593060d7 CM |
870 | u64 stripe_offset; |
871 | u64 stripe_nr; | |
cea9e445 | 872 | int stripes_allocated = 8; |
321aecc6 | 873 | int stripes_required = 1; |
593060d7 | 874 | int stripe_index; |
cea9e445 CM |
875 | int i; |
876 | struct btrfs_multi_bio *multi = NULL; | |
0b86a832 | 877 | |
cea9e445 CM |
878 | if (multi_ret && !(rw & (1 << BIO_RW))) { |
879 | stripes_allocated = 1; | |
880 | } | |
881 | again: | |
882 | if (multi_ret) { | |
883 | multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), | |
884 | GFP_NOFS); | |
885 | if (!multi) | |
886 | return -ENOMEM; | |
887 | } | |
0b86a832 CM |
888 | |
889 | spin_lock(&em_tree->lock); | |
890 | em = lookup_extent_mapping(em_tree, logical, *length); | |
b248a415 | 891 | spin_unlock(&em_tree->lock); |
3b951516 CM |
892 | if (!em) { |
893 | printk("unable to find logical %Lu\n", logical); | |
894 | } | |
0b86a832 CM |
895 | BUG_ON(!em); |
896 | ||
897 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
898 | map = (struct map_lookup *)em->bdev; | |
899 | offset = logical - em->start; | |
593060d7 | 900 | |
f188591e CM |
901 | if (mirror_num > map->num_stripes) |
902 | mirror_num = 0; | |
903 | ||
cea9e445 | 904 | /* if our multi bio struct is too small, back off and try again */ |
321aecc6 CM |
905 | if (rw & (1 << BIO_RW)) { |
906 | if (map->type & (BTRFS_BLOCK_GROUP_RAID1 | | |
907 | BTRFS_BLOCK_GROUP_DUP)) { | |
908 | stripes_required = map->num_stripes; | |
909 | } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { | |
910 | stripes_required = map->sub_stripes; | |
911 | } | |
912 | } | |
913 | if (multi_ret && rw == WRITE && | |
914 | stripes_allocated < stripes_required) { | |
cea9e445 | 915 | stripes_allocated = map->num_stripes; |
cea9e445 CM |
916 | free_extent_map(em); |
917 | kfree(multi); | |
918 | goto again; | |
919 | } | |
593060d7 CM |
920 | stripe_nr = offset; |
921 | /* | |
922 | * stripe_nr counts the total number of stripes we have to stride | |
923 | * to get to this block | |
924 | */ | |
925 | do_div(stripe_nr, map->stripe_len); | |
926 | ||
927 | stripe_offset = stripe_nr * map->stripe_len; | |
928 | BUG_ON(offset < stripe_offset); | |
929 | ||
930 | /* stripe_offset is the offset of this block in its stripe*/ | |
931 | stripe_offset = offset - stripe_offset; | |
932 | ||
cea9e445 | 933 | if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | |
321aecc6 | 934 | BTRFS_BLOCK_GROUP_RAID10 | |
cea9e445 CM |
935 | BTRFS_BLOCK_GROUP_DUP)) { |
936 | /* we limit the length of each bio to what fits in a stripe */ | |
937 | *length = min_t(u64, em->len - offset, | |
938 | map->stripe_len - stripe_offset); | |
939 | } else { | |
940 | *length = em->len - offset; | |
941 | } | |
942 | if (!multi_ret) | |
943 | goto out; | |
944 | ||
945 | multi->num_stripes = 1; | |
946 | stripe_index = 0; | |
8790d502 | 947 | if (map->type & BTRFS_BLOCK_GROUP_RAID1) { |
8790d502 | 948 | if (rw & (1 << BIO_RW)) |
cea9e445 | 949 | multi->num_stripes = map->num_stripes; |
f188591e CM |
950 | else if (mirror_num) { |
951 | stripe_index = mirror_num - 1; | |
952 | } else { | |
8790d502 CM |
953 | int i; |
954 | u64 least = (u64)-1; | |
955 | struct btrfs_device *cur; | |
956 | ||
957 | for (i = 0; i < map->num_stripes; i++) { | |
958 | cur = map->stripes[i].dev; | |
959 | spin_lock(&cur->io_lock); | |
960 | if (cur->total_ios < least) { | |
961 | least = cur->total_ios; | |
962 | stripe_index = i; | |
963 | } | |
964 | spin_unlock(&cur->io_lock); | |
965 | } | |
8790d502 | 966 | } |
611f0e00 | 967 | } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { |
cea9e445 CM |
968 | if (rw & (1 << BIO_RW)) |
969 | multi->num_stripes = map->num_stripes; | |
f188591e CM |
970 | else if (mirror_num) |
971 | stripe_index = mirror_num - 1; | |
321aecc6 CM |
972 | } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { |
973 | int factor = map->num_stripes / map->sub_stripes; | |
974 | int orig_stripe_nr = stripe_nr; | |
975 | ||
976 | stripe_index = do_div(stripe_nr, factor); | |
977 | stripe_index *= map->sub_stripes; | |
978 | ||
979 | if (rw & (1 << BIO_RW)) | |
980 | multi->num_stripes = map->sub_stripes; | |
981 | else if (mirror_num) | |
982 | stripe_index += mirror_num - 1; | |
983 | else | |
984 | stripe_index += orig_stripe_nr % map->sub_stripes; | |
8790d502 CM |
985 | } else { |
986 | /* | |
987 | * after this do_div call, stripe_nr is the number of stripes | |
988 | * on this device we have to walk to find the data, and | |
989 | * stripe_index is the number of our device in the stripe array | |
990 | */ | |
991 | stripe_index = do_div(stripe_nr, map->num_stripes); | |
992 | } | |
593060d7 | 993 | BUG_ON(stripe_index >= map->num_stripes); |
cea9e445 CM |
994 | |
995 | for (i = 0; i < multi->num_stripes; i++) { | |
996 | multi->stripes[i].physical = | |
997 | map->stripes[stripe_index].physical + stripe_offset + | |
998 | stripe_nr * map->stripe_len; | |
999 | multi->stripes[i].dev = map->stripes[stripe_index].dev; | |
1000 | stripe_index++; | |
593060d7 | 1001 | } |
cea9e445 CM |
1002 | *multi_ret = multi; |
1003 | out: | |
0b86a832 | 1004 | free_extent_map(em); |
0b86a832 CM |
1005 | return 0; |
1006 | } | |
1007 | ||
8790d502 CM |
1008 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
1009 | static void end_bio_multi_stripe(struct bio *bio, int err) | |
1010 | #else | |
1011 | static int end_bio_multi_stripe(struct bio *bio, | |
1012 | unsigned int bytes_done, int err) | |
1013 | #endif | |
1014 | { | |
cea9e445 | 1015 | struct btrfs_multi_bio *multi = bio->bi_private; |
8790d502 CM |
1016 | |
1017 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1018 | if (bio->bi_size) | |
1019 | return 1; | |
1020 | #endif | |
1021 | if (err) | |
1022 | multi->error = err; | |
1023 | ||
cea9e445 | 1024 | if (atomic_dec_and_test(&multi->stripes_pending)) { |
8790d502 CM |
1025 | bio->bi_private = multi->private; |
1026 | bio->bi_end_io = multi->end_io; | |
1027 | ||
1028 | if (!err && multi->error) | |
1029 | err = multi->error; | |
1030 | kfree(multi); | |
1031 | ||
73f61b2a M |
1032 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) |
1033 | bio_endio(bio, bio->bi_size, err); | |
1034 | #else | |
8790d502 | 1035 | bio_endio(bio, err); |
73f61b2a | 1036 | #endif |
8790d502 CM |
1037 | } else { |
1038 | bio_put(bio); | |
1039 | } | |
1040 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1041 | return 0; | |
1042 | #endif | |
1043 | } | |
1044 | ||
f188591e CM |
1045 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, |
1046 | int mirror_num) | |
0b86a832 CM |
1047 | { |
1048 | struct btrfs_mapping_tree *map_tree; | |
1049 | struct btrfs_device *dev; | |
8790d502 | 1050 | struct bio *first_bio = bio; |
0b86a832 | 1051 | u64 logical = bio->bi_sector << 9; |
0b86a832 CM |
1052 | u64 length = 0; |
1053 | u64 map_length; | |
1054 | struct bio_vec *bvec; | |
cea9e445 | 1055 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 CM |
1056 | int i; |
1057 | int ret; | |
8790d502 CM |
1058 | int dev_nr = 0; |
1059 | int total_devs = 1; | |
0b86a832 CM |
1060 | |
1061 | bio_for_each_segment(bvec, bio, i) { | |
1062 | length += bvec->bv_len; | |
1063 | } | |
8790d502 | 1064 | |
0b86a832 CM |
1065 | map_tree = &root->fs_info->mapping_tree; |
1066 | map_length = length; | |
cea9e445 | 1067 | |
f188591e CM |
1068 | ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi, |
1069 | mirror_num); | |
cea9e445 CM |
1070 | BUG_ON(ret); |
1071 | ||
1072 | total_devs = multi->num_stripes; | |
1073 | if (map_length < length) { | |
1074 | printk("mapping failed logical %Lu bio len %Lu " | |
1075 | "len %Lu\n", logical, length, map_length); | |
1076 | BUG(); | |
1077 | } | |
1078 | multi->end_io = first_bio->bi_end_io; | |
1079 | multi->private = first_bio->bi_private; | |
1080 | atomic_set(&multi->stripes_pending, multi->num_stripes); | |
1081 | ||
8790d502 | 1082 | while(dev_nr < total_devs) { |
8790d502 | 1083 | if (total_devs > 1) { |
8790d502 CM |
1084 | if (dev_nr < total_devs - 1) { |
1085 | bio = bio_clone(first_bio, GFP_NOFS); | |
1086 | BUG_ON(!bio); | |
1087 | } else { | |
1088 | bio = first_bio; | |
1089 | } | |
1090 | bio->bi_private = multi; | |
1091 | bio->bi_end_io = end_bio_multi_stripe; | |
1092 | } | |
cea9e445 CM |
1093 | bio->bi_sector = multi->stripes[dev_nr].physical >> 9; |
1094 | dev = multi->stripes[dev_nr].dev; | |
8790d502 CM |
1095 | bio->bi_bdev = dev->bdev; |
1096 | spin_lock(&dev->io_lock); | |
1097 | dev->total_ios++; | |
1098 | spin_unlock(&dev->io_lock); | |
1099 | submit_bio(rw, bio); | |
1100 | dev_nr++; | |
1101 | } | |
cea9e445 CM |
1102 | if (total_devs == 1) |
1103 | kfree(multi); | |
0b86a832 CM |
1104 | return 0; |
1105 | } | |
1106 | ||
a443755f CM |
1107 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, |
1108 | u8 *uuid) | |
0b86a832 | 1109 | { |
8a4b83cc | 1110 | struct list_head *head = &root->fs_info->fs_devices->devices; |
0b86a832 | 1111 | |
a443755f | 1112 | return __find_device(head, devid, uuid); |
0b86a832 CM |
1113 | } |
1114 | ||
1115 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, | |
1116 | struct extent_buffer *leaf, | |
1117 | struct btrfs_chunk *chunk) | |
1118 | { | |
1119 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
1120 | struct map_lookup *map; | |
1121 | struct extent_map *em; | |
1122 | u64 logical; | |
1123 | u64 length; | |
1124 | u64 devid; | |
a443755f | 1125 | u8 uuid[BTRFS_UUID_SIZE]; |
593060d7 | 1126 | int num_stripes; |
0b86a832 | 1127 | int ret; |
593060d7 | 1128 | int i; |
0b86a832 | 1129 | |
e17cade2 CM |
1130 | logical = key->offset; |
1131 | length = btrfs_chunk_length(leaf, chunk); | |
0b86a832 CM |
1132 | spin_lock(&map_tree->map_tree.lock); |
1133 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
b248a415 | 1134 | spin_unlock(&map_tree->map_tree.lock); |
0b86a832 CM |
1135 | |
1136 | /* already mapped? */ | |
1137 | if (em && em->start <= logical && em->start + em->len > logical) { | |
1138 | free_extent_map(em); | |
0b86a832 CM |
1139 | return 0; |
1140 | } else if (em) { | |
1141 | free_extent_map(em); | |
1142 | } | |
0b86a832 CM |
1143 | |
1144 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
1145 | if (!map) | |
1146 | return -ENOMEM; | |
1147 | ||
1148 | em = alloc_extent_map(GFP_NOFS); | |
1149 | if (!em) | |
1150 | return -ENOMEM; | |
593060d7 CM |
1151 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); |
1152 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); | |
0b86a832 CM |
1153 | if (!map) { |
1154 | free_extent_map(em); | |
1155 | return -ENOMEM; | |
1156 | } | |
1157 | ||
1158 | em->bdev = (struct block_device *)map; | |
1159 | em->start = logical; | |
1160 | em->len = length; | |
1161 | em->block_start = 0; | |
1162 | ||
593060d7 CM |
1163 | map->num_stripes = num_stripes; |
1164 | map->io_width = btrfs_chunk_io_width(leaf, chunk); | |
1165 | map->io_align = btrfs_chunk_io_align(leaf, chunk); | |
1166 | map->sector_size = btrfs_chunk_sector_size(leaf, chunk); | |
1167 | map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
1168 | map->type = btrfs_chunk_type(leaf, chunk); | |
321aecc6 | 1169 | map->sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk); |
593060d7 CM |
1170 | for (i = 0; i < num_stripes; i++) { |
1171 | map->stripes[i].physical = | |
1172 | btrfs_stripe_offset_nr(leaf, chunk, i); | |
1173 | devid = btrfs_stripe_devid_nr(leaf, chunk, i); | |
a443755f CM |
1174 | read_extent_buffer(leaf, uuid, (unsigned long) |
1175 | btrfs_stripe_dev_uuid_nr(chunk, i), | |
1176 | BTRFS_UUID_SIZE); | |
1177 | map->stripes[i].dev = btrfs_find_device(root, devid, uuid); | |
593060d7 CM |
1178 | if (!map->stripes[i].dev) { |
1179 | kfree(map); | |
1180 | free_extent_map(em); | |
1181 | return -EIO; | |
1182 | } | |
0b86a832 CM |
1183 | } |
1184 | ||
1185 | spin_lock(&map_tree->map_tree.lock); | |
1186 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
0b86a832 | 1187 | spin_unlock(&map_tree->map_tree.lock); |
b248a415 | 1188 | BUG_ON(ret); |
0b86a832 CM |
1189 | free_extent_map(em); |
1190 | ||
1191 | return 0; | |
1192 | } | |
1193 | ||
1194 | static int fill_device_from_item(struct extent_buffer *leaf, | |
1195 | struct btrfs_dev_item *dev_item, | |
1196 | struct btrfs_device *device) | |
1197 | { | |
1198 | unsigned long ptr; | |
0b86a832 CM |
1199 | |
1200 | device->devid = btrfs_device_id(leaf, dev_item); | |
1201 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
1202 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
1203 | device->type = btrfs_device_type(leaf, dev_item); | |
1204 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
1205 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
1206 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
0b86a832 CM |
1207 | |
1208 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
e17cade2 | 1209 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 | 1210 | |
0b86a832 CM |
1211 | return 0; |
1212 | } | |
1213 | ||
0d81ba5d | 1214 | static int read_one_dev(struct btrfs_root *root, |
0b86a832 CM |
1215 | struct extent_buffer *leaf, |
1216 | struct btrfs_dev_item *dev_item) | |
1217 | { | |
1218 | struct btrfs_device *device; | |
1219 | u64 devid; | |
1220 | int ret; | |
a443755f CM |
1221 | u8 dev_uuid[BTRFS_UUID_SIZE]; |
1222 | ||
0b86a832 | 1223 | devid = btrfs_device_id(leaf, dev_item); |
a443755f CM |
1224 | read_extent_buffer(leaf, dev_uuid, |
1225 | (unsigned long)btrfs_device_uuid(dev_item), | |
1226 | BTRFS_UUID_SIZE); | |
1227 | device = btrfs_find_device(root, devid, dev_uuid); | |
6324fbf3 | 1228 | if (!device) { |
8a4b83cc | 1229 | printk("warning devid %Lu not found already\n", devid); |
f2984462 | 1230 | device = kzalloc(sizeof(*device), GFP_NOFS); |
6324fbf3 CM |
1231 | if (!device) |
1232 | return -ENOMEM; | |
8a4b83cc CM |
1233 | list_add(&device->dev_list, |
1234 | &root->fs_info->fs_devices->devices); | |
b248a415 | 1235 | device->barriers = 1; |
8790d502 | 1236 | spin_lock_init(&device->io_lock); |
6324fbf3 | 1237 | } |
0b86a832 CM |
1238 | |
1239 | fill_device_from_item(leaf, dev_item, device); | |
1240 | device->dev_root = root->fs_info->dev_root; | |
0b86a832 CM |
1241 | ret = 0; |
1242 | #if 0 | |
1243 | ret = btrfs_open_device(device); | |
1244 | if (ret) { | |
1245 | kfree(device); | |
1246 | } | |
1247 | #endif | |
1248 | return ret; | |
1249 | } | |
1250 | ||
0d81ba5d CM |
1251 | int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) |
1252 | { | |
1253 | struct btrfs_dev_item *dev_item; | |
1254 | ||
1255 | dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, | |
1256 | dev_item); | |
1257 | return read_one_dev(root, buf, dev_item); | |
1258 | } | |
1259 | ||
0b86a832 CM |
1260 | int btrfs_read_sys_array(struct btrfs_root *root) |
1261 | { | |
1262 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1263 | struct extent_buffer *sb = root->fs_info->sb_buffer; | |
1264 | struct btrfs_disk_key *disk_key; | |
0b86a832 CM |
1265 | struct btrfs_chunk *chunk; |
1266 | struct btrfs_key key; | |
1267 | u32 num_stripes; | |
1268 | u32 array_size; | |
1269 | u32 len = 0; | |
1270 | u8 *ptr; | |
1271 | unsigned long sb_ptr; | |
1272 | u32 cur; | |
1273 | int ret; | |
0b86a832 CM |
1274 | |
1275 | array_size = btrfs_super_sys_array_size(super_copy); | |
1276 | ||
1277 | /* | |
1278 | * we do this loop twice, once for the device items and | |
1279 | * once for all of the chunks. This way there are device | |
1280 | * structs filled in for every chunk | |
1281 | */ | |
0b86a832 CM |
1282 | ptr = super_copy->sys_chunk_array; |
1283 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
1284 | cur = 0; | |
1285 | ||
1286 | while (cur < array_size) { | |
1287 | disk_key = (struct btrfs_disk_key *)ptr; | |
1288 | btrfs_disk_key_to_cpu(&key, disk_key); | |
1289 | ||
1290 | len = sizeof(*disk_key); | |
1291 | ptr += len; | |
1292 | sb_ptr += len; | |
1293 | cur += len; | |
1294 | ||
0d81ba5d | 1295 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { |
0b86a832 | 1296 | chunk = (struct btrfs_chunk *)sb_ptr; |
0d81ba5d CM |
1297 | ret = read_one_chunk(root, &key, sb, chunk); |
1298 | BUG_ON(ret); | |
0b86a832 CM |
1299 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); |
1300 | len = btrfs_chunk_item_size(num_stripes); | |
1301 | } else { | |
1302 | BUG(); | |
1303 | } | |
1304 | ptr += len; | |
1305 | sb_ptr += len; | |
1306 | cur += len; | |
1307 | } | |
0b86a832 CM |
1308 | return 0; |
1309 | } | |
1310 | ||
1311 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
1312 | { | |
1313 | struct btrfs_path *path; | |
1314 | struct extent_buffer *leaf; | |
1315 | struct btrfs_key key; | |
1316 | struct btrfs_key found_key; | |
1317 | int ret; | |
1318 | int slot; | |
1319 | ||
1320 | root = root->fs_info->chunk_root; | |
1321 | ||
1322 | path = btrfs_alloc_path(); | |
1323 | if (!path) | |
1324 | return -ENOMEM; | |
1325 | ||
1326 | /* first we search for all of the device items, and then we | |
1327 | * read in all of the chunk items. This way we can create chunk | |
1328 | * mappings that reference all of the devices that are afound | |
1329 | */ | |
1330 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1331 | key.offset = 0; | |
1332 | key.type = 0; | |
1333 | again: | |
1334 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1335 | while(1) { | |
1336 | leaf = path->nodes[0]; | |
1337 | slot = path->slots[0]; | |
1338 | if (slot >= btrfs_header_nritems(leaf)) { | |
1339 | ret = btrfs_next_leaf(root, path); | |
1340 | if (ret == 0) | |
1341 | continue; | |
1342 | if (ret < 0) | |
1343 | goto error; | |
1344 | break; | |
1345 | } | |
1346 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1347 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1348 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
1349 | break; | |
1350 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
1351 | struct btrfs_dev_item *dev_item; | |
1352 | dev_item = btrfs_item_ptr(leaf, slot, | |
1353 | struct btrfs_dev_item); | |
0d81ba5d | 1354 | ret = read_one_dev(root, leaf, dev_item); |
0b86a832 CM |
1355 | BUG_ON(ret); |
1356 | } | |
1357 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
1358 | struct btrfs_chunk *chunk; | |
1359 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
1360 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
1361 | } | |
1362 | path->slots[0]++; | |
1363 | } | |
1364 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1365 | key.objectid = 0; | |
1366 | btrfs_release_path(root, path); | |
1367 | goto again; | |
1368 | } | |
1369 | ||
1370 | btrfs_free_path(path); | |
1371 | ret = 0; | |
1372 | error: | |
1373 | return ret; | |
1374 | } | |
1375 |