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; | |
cea9e445 | 36 | struct btrfs_bio_stripe stripes[]; |
593060d7 CM |
37 | }; |
38 | ||
39 | #define map_lookup_size(n) (sizeof(struct map_lookup) + \ | |
cea9e445 | 40 | (sizeof(struct btrfs_bio_stripe) * (n))) |
593060d7 | 41 | |
8a4b83cc CM |
42 | static DEFINE_MUTEX(uuid_mutex); |
43 | static LIST_HEAD(fs_uuids); | |
44 | ||
45 | int btrfs_cleanup_fs_uuids(void) | |
46 | { | |
47 | struct btrfs_fs_devices *fs_devices; | |
48 | struct list_head *uuid_cur; | |
49 | struct list_head *devices_cur; | |
50 | struct btrfs_device *dev; | |
51 | ||
52 | list_for_each(uuid_cur, &fs_uuids) { | |
53 | fs_devices = list_entry(uuid_cur, struct btrfs_fs_devices, | |
54 | list); | |
55 | while(!list_empty(&fs_devices->devices)) { | |
56 | devices_cur = fs_devices->devices.next; | |
57 | dev = list_entry(devices_cur, struct btrfs_device, | |
58 | dev_list); | |
59 | printk("uuid cleanup finds %s\n", dev->name); | |
60 | if (dev->bdev) { | |
61 | printk("closing\n"); | |
62 | close_bdev_excl(dev->bdev); | |
63 | } | |
64 | list_del(&dev->dev_list); | |
65 | kfree(dev); | |
66 | } | |
67 | } | |
68 | return 0; | |
69 | } | |
70 | ||
71 | static struct btrfs_device *__find_device(struct list_head *head, u64 devid) | |
72 | { | |
73 | struct btrfs_device *dev; | |
74 | struct list_head *cur; | |
75 | ||
76 | list_for_each(cur, head) { | |
77 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
78 | if (dev->devid == devid) | |
79 | return dev; | |
80 | } | |
81 | return NULL; | |
82 | } | |
83 | ||
84 | static struct btrfs_fs_devices *find_fsid(u8 *fsid) | |
85 | { | |
86 | struct list_head *cur; | |
87 | struct btrfs_fs_devices *fs_devices; | |
88 | ||
89 | list_for_each(cur, &fs_uuids) { | |
90 | fs_devices = list_entry(cur, struct btrfs_fs_devices, list); | |
91 | if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0) | |
92 | return fs_devices; | |
93 | } | |
94 | return NULL; | |
95 | } | |
96 | ||
97 | static int device_list_add(const char *path, | |
98 | struct btrfs_super_block *disk_super, | |
99 | u64 devid, struct btrfs_fs_devices **fs_devices_ret) | |
100 | { | |
101 | struct btrfs_device *device; | |
102 | struct btrfs_fs_devices *fs_devices; | |
103 | u64 found_transid = btrfs_super_generation(disk_super); | |
104 | ||
105 | fs_devices = find_fsid(disk_super->fsid); | |
106 | if (!fs_devices) { | |
107 | fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS); | |
108 | if (!fs_devices) | |
109 | return -ENOMEM; | |
110 | INIT_LIST_HEAD(&fs_devices->devices); | |
111 | list_add(&fs_devices->list, &fs_uuids); | |
112 | memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE); | |
113 | fs_devices->latest_devid = devid; | |
114 | fs_devices->latest_trans = found_transid; | |
115 | fs_devices->lowest_devid = (u64)-1; | |
116 | fs_devices->num_devices = 0; | |
117 | device = NULL; | |
118 | } else { | |
119 | device = __find_device(&fs_devices->devices, devid); | |
120 | } | |
121 | if (!device) { | |
122 | device = kzalloc(sizeof(*device), GFP_NOFS); | |
123 | if (!device) { | |
124 | /* we can safely leave the fs_devices entry around */ | |
125 | return -ENOMEM; | |
126 | } | |
127 | device->devid = devid; | |
f2984462 | 128 | device->barriers = 1; |
b248a415 | 129 | spin_lock_init(&device->io_lock); |
8a4b83cc CM |
130 | device->name = kstrdup(path, GFP_NOFS); |
131 | if (!device->name) { | |
132 | kfree(device); | |
133 | return -ENOMEM; | |
134 | } | |
135 | list_add(&device->dev_list, &fs_devices->devices); | |
136 | fs_devices->num_devices++; | |
137 | } | |
138 | ||
139 | if (found_transid > fs_devices->latest_trans) { | |
140 | fs_devices->latest_devid = devid; | |
141 | fs_devices->latest_trans = found_transid; | |
142 | } | |
143 | if (fs_devices->lowest_devid > devid) { | |
144 | fs_devices->lowest_devid = devid; | |
145 | printk("lowest devid now %Lu\n", devid); | |
146 | } | |
147 | *fs_devices_ret = fs_devices; | |
148 | return 0; | |
149 | } | |
150 | ||
151 | int btrfs_close_devices(struct btrfs_fs_devices *fs_devices) | |
152 | { | |
153 | struct list_head *head = &fs_devices->devices; | |
154 | struct list_head *cur; | |
155 | struct btrfs_device *device; | |
156 | ||
157 | mutex_lock(&uuid_mutex); | |
158 | list_for_each(cur, head) { | |
159 | device = list_entry(cur, struct btrfs_device, dev_list); | |
160 | if (device->bdev) { | |
161 | close_bdev_excl(device->bdev); | |
162 | printk("close devices closes %s\n", device->name); | |
163 | } | |
164 | device->bdev = NULL; | |
165 | } | |
166 | mutex_unlock(&uuid_mutex); | |
167 | return 0; | |
168 | } | |
169 | ||
170 | int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, | |
171 | int flags, void *holder) | |
172 | { | |
173 | struct block_device *bdev; | |
174 | struct list_head *head = &fs_devices->devices; | |
175 | struct list_head *cur; | |
176 | struct btrfs_device *device; | |
177 | int ret; | |
178 | ||
179 | mutex_lock(&uuid_mutex); | |
180 | list_for_each(cur, head) { | |
181 | device = list_entry(cur, struct btrfs_device, dev_list); | |
182 | bdev = open_bdev_excl(device->name, flags, holder); | |
e17cade2 | 183 | |
8a4b83cc CM |
184 | if (IS_ERR(bdev)) { |
185 | printk("open %s failed\n", device->name); | |
186 | ret = PTR_ERR(bdev); | |
187 | goto fail; | |
188 | } | |
189 | if (device->devid == fs_devices->latest_devid) | |
190 | fs_devices->latest_bdev = bdev; | |
191 | if (device->devid == fs_devices->lowest_devid) { | |
192 | fs_devices->lowest_bdev = bdev; | |
8a4b83cc CM |
193 | } |
194 | device->bdev = bdev; | |
195 | } | |
196 | mutex_unlock(&uuid_mutex); | |
197 | return 0; | |
198 | fail: | |
199 | mutex_unlock(&uuid_mutex); | |
200 | btrfs_close_devices(fs_devices); | |
201 | return ret; | |
202 | } | |
203 | ||
204 | int btrfs_scan_one_device(const char *path, int flags, void *holder, | |
205 | struct btrfs_fs_devices **fs_devices_ret) | |
206 | { | |
207 | struct btrfs_super_block *disk_super; | |
208 | struct block_device *bdev; | |
209 | struct buffer_head *bh; | |
210 | int ret; | |
211 | u64 devid; | |
f2984462 | 212 | u64 transid; |
8a4b83cc CM |
213 | |
214 | mutex_lock(&uuid_mutex); | |
215 | ||
216 | printk("scan one opens %s\n", path); | |
217 | bdev = open_bdev_excl(path, flags, holder); | |
218 | ||
219 | if (IS_ERR(bdev)) { | |
220 | printk("open failed\n"); | |
221 | ret = PTR_ERR(bdev); | |
222 | goto error; | |
223 | } | |
224 | ||
225 | ret = set_blocksize(bdev, 4096); | |
226 | if (ret) | |
227 | goto error_close; | |
228 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
229 | if (!bh) { | |
230 | ret = -EIO; | |
231 | goto error_close; | |
232 | } | |
233 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
234 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
235 | sizeof(disk_super->magic))) { | |
236 | printk("no btrfs found on %s\n", path); | |
e58ca020 | 237 | ret = -EINVAL; |
8a4b83cc CM |
238 | goto error_brelse; |
239 | } | |
240 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
f2984462 CM |
241 | transid = btrfs_super_generation(disk_super); |
242 | printk("found device %Lu transid %Lu on %s\n", devid, transid, path); | |
8a4b83cc CM |
243 | ret = device_list_add(path, disk_super, devid, fs_devices_ret); |
244 | ||
245 | error_brelse: | |
246 | brelse(bh); | |
247 | error_close: | |
248 | close_bdev_excl(bdev); | |
8a4b83cc CM |
249 | error: |
250 | mutex_unlock(&uuid_mutex); | |
251 | return ret; | |
252 | } | |
0b86a832 CM |
253 | |
254 | /* | |
255 | * this uses a pretty simple search, the expectation is that it is | |
256 | * called very infrequently and that a given device has a small number | |
257 | * of extents | |
258 | */ | |
259 | static int find_free_dev_extent(struct btrfs_trans_handle *trans, | |
260 | struct btrfs_device *device, | |
261 | struct btrfs_path *path, | |
262 | u64 num_bytes, u64 *start) | |
263 | { | |
264 | struct btrfs_key key; | |
265 | struct btrfs_root *root = device->dev_root; | |
266 | struct btrfs_dev_extent *dev_extent = NULL; | |
267 | u64 hole_size = 0; | |
268 | u64 last_byte = 0; | |
269 | u64 search_start = 0; | |
270 | u64 search_end = device->total_bytes; | |
271 | int ret; | |
272 | int slot = 0; | |
273 | int start_found; | |
274 | struct extent_buffer *l; | |
275 | ||
276 | start_found = 0; | |
277 | path->reada = 2; | |
278 | ||
279 | /* FIXME use last free of some kind */ | |
280 | ||
8a4b83cc CM |
281 | /* we don't want to overwrite the superblock on the drive, |
282 | * so we make sure to start at an offset of at least 1MB | |
283 | */ | |
284 | search_start = max((u64)1024 * 1024, search_start); | |
0b86a832 CM |
285 | key.objectid = device->devid; |
286 | key.offset = search_start; | |
287 | key.type = BTRFS_DEV_EXTENT_KEY; | |
288 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
289 | if (ret < 0) | |
290 | goto error; | |
291 | ret = btrfs_previous_item(root, path, 0, key.type); | |
292 | if (ret < 0) | |
293 | goto error; | |
294 | l = path->nodes[0]; | |
295 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
296 | while (1) { | |
297 | l = path->nodes[0]; | |
298 | slot = path->slots[0]; | |
299 | if (slot >= btrfs_header_nritems(l)) { | |
300 | ret = btrfs_next_leaf(root, path); | |
301 | if (ret == 0) | |
302 | continue; | |
303 | if (ret < 0) | |
304 | goto error; | |
305 | no_more_items: | |
306 | if (!start_found) { | |
307 | if (search_start >= search_end) { | |
308 | ret = -ENOSPC; | |
309 | goto error; | |
310 | } | |
311 | *start = search_start; | |
312 | start_found = 1; | |
313 | goto check_pending; | |
314 | } | |
315 | *start = last_byte > search_start ? | |
316 | last_byte : search_start; | |
317 | if (search_end <= *start) { | |
318 | ret = -ENOSPC; | |
319 | goto error; | |
320 | } | |
321 | goto check_pending; | |
322 | } | |
323 | btrfs_item_key_to_cpu(l, &key, slot); | |
324 | ||
325 | if (key.objectid < device->devid) | |
326 | goto next; | |
327 | ||
328 | if (key.objectid > device->devid) | |
329 | goto no_more_items; | |
330 | ||
331 | if (key.offset >= search_start && key.offset > last_byte && | |
332 | start_found) { | |
333 | if (last_byte < search_start) | |
334 | last_byte = search_start; | |
335 | hole_size = key.offset - last_byte; | |
336 | if (key.offset > last_byte && | |
337 | hole_size >= num_bytes) { | |
338 | *start = last_byte; | |
339 | goto check_pending; | |
340 | } | |
341 | } | |
342 | if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { | |
343 | goto next; | |
344 | } | |
345 | ||
346 | start_found = 1; | |
347 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
348 | last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); | |
349 | next: | |
350 | path->slots[0]++; | |
351 | cond_resched(); | |
352 | } | |
353 | check_pending: | |
354 | /* we have to make sure we didn't find an extent that has already | |
355 | * been allocated by the map tree or the original allocation | |
356 | */ | |
357 | btrfs_release_path(root, path); | |
358 | BUG_ON(*start < search_start); | |
359 | ||
6324fbf3 | 360 | if (*start + num_bytes > search_end) { |
0b86a832 CM |
361 | ret = -ENOSPC; |
362 | goto error; | |
363 | } | |
364 | /* check for pending inserts here */ | |
365 | return 0; | |
366 | ||
367 | error: | |
368 | btrfs_release_path(root, path); | |
369 | return ret; | |
370 | } | |
371 | ||
372 | int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, | |
373 | struct btrfs_device *device, | |
e17cade2 CM |
374 | u64 chunk_tree, u64 chunk_objectid, |
375 | u64 chunk_offset, | |
376 | u64 num_bytes, u64 *start) | |
0b86a832 CM |
377 | { |
378 | int ret; | |
379 | struct btrfs_path *path; | |
380 | struct btrfs_root *root = device->dev_root; | |
381 | struct btrfs_dev_extent *extent; | |
382 | struct extent_buffer *leaf; | |
383 | struct btrfs_key key; | |
384 | ||
385 | path = btrfs_alloc_path(); | |
386 | if (!path) | |
387 | return -ENOMEM; | |
388 | ||
389 | ret = find_free_dev_extent(trans, device, path, num_bytes, start); | |
6324fbf3 | 390 | if (ret) { |
0b86a832 | 391 | goto err; |
6324fbf3 | 392 | } |
0b86a832 CM |
393 | |
394 | key.objectid = device->devid; | |
395 | key.offset = *start; | |
396 | key.type = BTRFS_DEV_EXTENT_KEY; | |
397 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
398 | sizeof(*extent)); | |
399 | BUG_ON(ret); | |
400 | ||
401 | leaf = path->nodes[0]; | |
402 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
403 | struct btrfs_dev_extent); | |
e17cade2 CM |
404 | btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree); |
405 | btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid); | |
406 | btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset); | |
407 | ||
408 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, | |
409 | (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent), | |
410 | BTRFS_UUID_SIZE); | |
411 | ||
0b86a832 CM |
412 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); |
413 | btrfs_mark_buffer_dirty(leaf); | |
414 | err: | |
415 | btrfs_free_path(path); | |
416 | return ret; | |
417 | } | |
418 | ||
e17cade2 | 419 | static int find_next_chunk(struct btrfs_root *root, u64 objectid, u64 *offset) |
0b86a832 CM |
420 | { |
421 | struct btrfs_path *path; | |
422 | int ret; | |
423 | struct btrfs_key key; | |
e17cade2 | 424 | struct btrfs_chunk *chunk; |
0b86a832 CM |
425 | struct btrfs_key found_key; |
426 | ||
427 | path = btrfs_alloc_path(); | |
428 | BUG_ON(!path); | |
429 | ||
e17cade2 | 430 | key.objectid = objectid; |
0b86a832 CM |
431 | key.offset = (u64)-1; |
432 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
433 | ||
434 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
435 | if (ret < 0) | |
436 | goto error; | |
437 | ||
438 | BUG_ON(ret == 0); | |
439 | ||
440 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
441 | if (ret) { | |
e17cade2 | 442 | *offset = 0; |
0b86a832 CM |
443 | } else { |
444 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
445 | path->slots[0]); | |
e17cade2 CM |
446 | if (found_key.objectid != objectid) |
447 | *offset = 0; | |
448 | else { | |
449 | chunk = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
450 | struct btrfs_chunk); | |
451 | *offset = found_key.offset + | |
452 | btrfs_chunk_length(path->nodes[0], chunk); | |
453 | } | |
0b86a832 CM |
454 | } |
455 | ret = 0; | |
456 | error: | |
457 | btrfs_free_path(path); | |
458 | return ret; | |
459 | } | |
460 | ||
0b86a832 CM |
461 | static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, |
462 | u64 *objectid) | |
463 | { | |
464 | int ret; | |
465 | struct btrfs_key key; | |
466 | struct btrfs_key found_key; | |
467 | ||
468 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
469 | key.type = BTRFS_DEV_ITEM_KEY; | |
470 | key.offset = (u64)-1; | |
471 | ||
472 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
473 | if (ret < 0) | |
474 | goto error; | |
475 | ||
476 | BUG_ON(ret == 0); | |
477 | ||
478 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
479 | BTRFS_DEV_ITEM_KEY); | |
480 | if (ret) { | |
481 | *objectid = 1; | |
482 | } else { | |
483 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
484 | path->slots[0]); | |
485 | *objectid = found_key.offset + 1; | |
486 | } | |
487 | ret = 0; | |
488 | error: | |
489 | btrfs_release_path(root, path); | |
490 | return ret; | |
491 | } | |
492 | ||
493 | /* | |
494 | * the device information is stored in the chunk root | |
495 | * the btrfs_device struct should be fully filled in | |
496 | */ | |
497 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
498 | struct btrfs_root *root, | |
499 | struct btrfs_device *device) | |
500 | { | |
501 | int ret; | |
502 | struct btrfs_path *path; | |
503 | struct btrfs_dev_item *dev_item; | |
504 | struct extent_buffer *leaf; | |
505 | struct btrfs_key key; | |
506 | unsigned long ptr; | |
507 | u64 free_devid; | |
508 | ||
509 | root = root->fs_info->chunk_root; | |
510 | ||
511 | path = btrfs_alloc_path(); | |
512 | if (!path) | |
513 | return -ENOMEM; | |
514 | ||
515 | ret = find_next_devid(root, path, &free_devid); | |
516 | if (ret) | |
517 | goto out; | |
518 | ||
519 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
520 | key.type = BTRFS_DEV_ITEM_KEY; | |
521 | key.offset = free_devid; | |
522 | ||
523 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
0d81ba5d | 524 | sizeof(*dev_item)); |
0b86a832 CM |
525 | if (ret) |
526 | goto out; | |
527 | ||
528 | leaf = path->nodes[0]; | |
529 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
530 | ||
8a4b83cc | 531 | device->devid = free_devid; |
0b86a832 CM |
532 | btrfs_set_device_id(leaf, dev_item, device->devid); |
533 | btrfs_set_device_type(leaf, dev_item, device->type); | |
534 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
535 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
536 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
537 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
538 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
e17cade2 CM |
539 | btrfs_set_device_group(leaf, dev_item, 0); |
540 | btrfs_set_device_seek_speed(leaf, dev_item, 0); | |
541 | btrfs_set_device_bandwidth(leaf, dev_item, 0); | |
0b86a832 | 542 | |
0b86a832 | 543 | ptr = (unsigned long)btrfs_device_uuid(dev_item); |
e17cade2 | 544 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 CM |
545 | btrfs_mark_buffer_dirty(leaf); |
546 | ret = 0; | |
547 | ||
548 | out: | |
549 | btrfs_free_path(path); | |
550 | return ret; | |
551 | } | |
552 | int btrfs_update_device(struct btrfs_trans_handle *trans, | |
553 | struct btrfs_device *device) | |
554 | { | |
555 | int ret; | |
556 | struct btrfs_path *path; | |
557 | struct btrfs_root *root; | |
558 | struct btrfs_dev_item *dev_item; | |
559 | struct extent_buffer *leaf; | |
560 | struct btrfs_key key; | |
561 | ||
562 | root = device->dev_root->fs_info->chunk_root; | |
563 | ||
564 | path = btrfs_alloc_path(); | |
565 | if (!path) | |
566 | return -ENOMEM; | |
567 | ||
568 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
569 | key.type = BTRFS_DEV_ITEM_KEY; | |
570 | key.offset = device->devid; | |
571 | ||
572 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
573 | if (ret < 0) | |
574 | goto out; | |
575 | ||
576 | if (ret > 0) { | |
577 | ret = -ENOENT; | |
578 | goto out; | |
579 | } | |
580 | ||
581 | leaf = path->nodes[0]; | |
582 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
583 | ||
584 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
585 | btrfs_set_device_type(leaf, dev_item, device->type); | |
586 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
587 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
588 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
589 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
590 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
591 | btrfs_mark_buffer_dirty(leaf); | |
592 | ||
593 | out: | |
594 | btrfs_free_path(path); | |
595 | return ret; | |
596 | } | |
597 | ||
598 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, | |
599 | struct btrfs_root *root, | |
600 | struct btrfs_key *key, | |
601 | struct btrfs_chunk *chunk, int item_size) | |
602 | { | |
603 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
604 | struct btrfs_disk_key disk_key; | |
605 | u32 array_size; | |
606 | u8 *ptr; | |
607 | ||
608 | array_size = btrfs_super_sys_array_size(super_copy); | |
609 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
610 | return -EFBIG; | |
611 | ||
612 | ptr = super_copy->sys_chunk_array + array_size; | |
613 | btrfs_cpu_key_to_disk(&disk_key, key); | |
614 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
615 | ptr += sizeof(disk_key); | |
616 | memcpy(ptr, chunk, item_size); | |
617 | item_size += sizeof(disk_key); | |
618 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
619 | return 0; | |
620 | } | |
621 | ||
622 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, | |
623 | struct btrfs_root *extent_root, u64 *start, | |
6324fbf3 | 624 | u64 *num_bytes, u64 type) |
0b86a832 CM |
625 | { |
626 | u64 dev_offset; | |
593060d7 | 627 | struct btrfs_fs_info *info = extent_root->fs_info; |
0b86a832 CM |
628 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; |
629 | struct btrfs_stripe *stripes; | |
630 | struct btrfs_device *device = NULL; | |
631 | struct btrfs_chunk *chunk; | |
6324fbf3 | 632 | struct list_head private_devs; |
8a4b83cc | 633 | struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices; |
6324fbf3 | 634 | struct list_head *cur; |
0b86a832 CM |
635 | struct extent_map_tree *em_tree; |
636 | struct map_lookup *map; | |
637 | struct extent_map *em; | |
638 | u64 physical; | |
639 | u64 calc_size = 1024 * 1024 * 1024; | |
611f0e00 | 640 | u64 min_free = calc_size; |
6324fbf3 CM |
641 | u64 avail; |
642 | u64 max_avail = 0; | |
643 | int num_stripes = 1; | |
644 | int looped = 0; | |
0b86a832 | 645 | int ret; |
6324fbf3 | 646 | int index; |
593060d7 | 647 | int stripe_len = 64 * 1024; |
0b86a832 CM |
648 | struct btrfs_key key; |
649 | ||
6324fbf3 CM |
650 | if (list_empty(dev_list)) |
651 | return -ENOSPC; | |
593060d7 | 652 | |
8790d502 | 653 | if (type & (BTRFS_BLOCK_GROUP_RAID0)) |
593060d7 | 654 | num_stripes = btrfs_super_num_devices(&info->super_copy); |
611f0e00 CM |
655 | if (type & (BTRFS_BLOCK_GROUP_DUP)) |
656 | num_stripes = 2; | |
8790d502 CM |
657 | if (type & (BTRFS_BLOCK_GROUP_RAID1)) { |
658 | num_stripes = min_t(u64, 2, | |
659 | btrfs_super_num_devices(&info->super_copy)); | |
660 | } | |
6324fbf3 CM |
661 | again: |
662 | INIT_LIST_HEAD(&private_devs); | |
663 | cur = dev_list->next; | |
664 | index = 0; | |
611f0e00 CM |
665 | |
666 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
667 | min_free = calc_size * 2; | |
668 | ||
6324fbf3 CM |
669 | /* build a private list of devices we will allocate from */ |
670 | while(index < num_stripes) { | |
671 | device = list_entry(cur, struct btrfs_device, dev_list); | |
611f0e00 | 672 | |
6324fbf3 CM |
673 | avail = device->total_bytes - device->bytes_used; |
674 | cur = cur->next; | |
675 | if (avail > max_avail) | |
676 | max_avail = avail; | |
611f0e00 | 677 | if (avail >= min_free) { |
6324fbf3 CM |
678 | list_move_tail(&device->dev_list, &private_devs); |
679 | index++; | |
611f0e00 CM |
680 | if (type & BTRFS_BLOCK_GROUP_DUP) |
681 | index++; | |
6324fbf3 CM |
682 | } |
683 | if (cur == dev_list) | |
684 | break; | |
685 | } | |
686 | if (index < num_stripes) { | |
687 | list_splice(&private_devs, dev_list); | |
688 | if (!looped && max_avail > 0) { | |
689 | looped = 1; | |
690 | calc_size = max_avail; | |
691 | goto again; | |
692 | } | |
693 | return -ENOSPC; | |
694 | } | |
0b86a832 | 695 | |
e17cade2 CM |
696 | key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; |
697 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
698 | ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID, | |
699 | &key.offset); | |
0b86a832 CM |
700 | if (ret) |
701 | return ret; | |
702 | ||
0b86a832 CM |
703 | chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); |
704 | if (!chunk) | |
705 | return -ENOMEM; | |
706 | ||
593060d7 CM |
707 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); |
708 | if (!map) { | |
709 | kfree(chunk); | |
710 | return -ENOMEM; | |
711 | } | |
712 | ||
0b86a832 CM |
713 | stripes = &chunk->stripe; |
714 | ||
611f0e00 | 715 | if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP)) |
8790d502 CM |
716 | *num_bytes = calc_size; |
717 | else | |
718 | *num_bytes = calc_size * num_stripes; | |
719 | ||
6324fbf3 | 720 | index = 0; |
e17cade2 | 721 | printk("new chunk type %Lu start %Lu size %Lu\n", type, key.offset, *num_bytes); |
0b86a832 | 722 | while(index < num_stripes) { |
e17cade2 | 723 | struct btrfs_stripe *stripe; |
6324fbf3 CM |
724 | BUG_ON(list_empty(&private_devs)); |
725 | cur = private_devs.next; | |
726 | device = list_entry(cur, struct btrfs_device, dev_list); | |
611f0e00 CM |
727 | |
728 | /* loop over this device again if we're doing a dup group */ | |
729 | if (!(type & BTRFS_BLOCK_GROUP_DUP) || | |
730 | (index == num_stripes - 1)) | |
731 | list_move_tail(&device->dev_list, dev_list); | |
0b86a832 CM |
732 | |
733 | ret = btrfs_alloc_dev_extent(trans, device, | |
e17cade2 CM |
734 | info->chunk_root->root_key.objectid, |
735 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, key.offset, | |
736 | calc_size, &dev_offset); | |
0b86a832 | 737 | BUG_ON(ret); |
e17cade2 | 738 | printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key.offset, calc_size, device->devid, type); |
0b86a832 CM |
739 | device->bytes_used += calc_size; |
740 | ret = btrfs_update_device(trans, device); | |
741 | BUG_ON(ret); | |
742 | ||
593060d7 CM |
743 | map->stripes[index].dev = device; |
744 | map->stripes[index].physical = dev_offset; | |
e17cade2 CM |
745 | stripe = stripes + index; |
746 | btrfs_set_stack_stripe_devid(stripe, device->devid); | |
747 | btrfs_set_stack_stripe_offset(stripe, dev_offset); | |
748 | memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE); | |
0b86a832 CM |
749 | physical = dev_offset; |
750 | index++; | |
751 | } | |
6324fbf3 | 752 | BUG_ON(!list_empty(&private_devs)); |
0b86a832 | 753 | |
e17cade2 CM |
754 | /* key was set above */ |
755 | btrfs_set_stack_chunk_length(chunk, *num_bytes); | |
0b86a832 | 756 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); |
593060d7 | 757 | btrfs_set_stack_chunk_stripe_len(chunk, stripe_len); |
0b86a832 CM |
758 | btrfs_set_stack_chunk_type(chunk, type); |
759 | btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); | |
593060d7 CM |
760 | btrfs_set_stack_chunk_io_align(chunk, stripe_len); |
761 | btrfs_set_stack_chunk_io_width(chunk, stripe_len); | |
0b86a832 | 762 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); |
593060d7 CM |
763 | map->sector_size = extent_root->sectorsize; |
764 | map->stripe_len = stripe_len; | |
765 | map->io_align = stripe_len; | |
766 | map->io_width = stripe_len; | |
767 | map->type = type; | |
768 | map->num_stripes = num_stripes; | |
0b86a832 CM |
769 | |
770 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, | |
771 | btrfs_chunk_item_size(num_stripes)); | |
772 | BUG_ON(ret); | |
e17cade2 | 773 | *start = key.offset;; |
0b86a832 CM |
774 | |
775 | em = alloc_extent_map(GFP_NOFS); | |
776 | if (!em) | |
777 | return -ENOMEM; | |
0b86a832 | 778 | em->bdev = (struct block_device *)map; |
e17cade2 CM |
779 | em->start = key.offset; |
780 | em->len = *num_bytes; | |
0b86a832 CM |
781 | em->block_start = 0; |
782 | ||
0b86a832 CM |
783 | kfree(chunk); |
784 | ||
785 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; | |
786 | spin_lock(&em_tree->lock); | |
787 | ret = add_extent_mapping(em_tree, em); | |
0b86a832 | 788 | spin_unlock(&em_tree->lock); |
b248a415 | 789 | BUG_ON(ret); |
0b86a832 CM |
790 | free_extent_map(em); |
791 | return ret; | |
792 | } | |
793 | ||
794 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
795 | { | |
796 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
797 | } | |
798 | ||
799 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
800 | { | |
801 | struct extent_map *em; | |
802 | ||
803 | while(1) { | |
804 | spin_lock(&tree->map_tree.lock); | |
805 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
806 | if (em) | |
807 | remove_extent_mapping(&tree->map_tree, em); | |
808 | spin_unlock(&tree->map_tree.lock); | |
809 | if (!em) | |
810 | break; | |
811 | kfree(em->bdev); | |
812 | /* once for us */ | |
813 | free_extent_map(em); | |
814 | /* once for the tree */ | |
815 | free_extent_map(em); | |
816 | } | |
817 | } | |
818 | ||
f188591e CM |
819 | int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) |
820 | { | |
821 | struct extent_map *em; | |
822 | struct map_lookup *map; | |
823 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
824 | int ret; | |
825 | ||
826 | spin_lock(&em_tree->lock); | |
827 | em = lookup_extent_mapping(em_tree, logical, len); | |
b248a415 | 828 | spin_unlock(&em_tree->lock); |
f188591e CM |
829 | BUG_ON(!em); |
830 | ||
831 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
832 | map = (struct map_lookup *)em->bdev; | |
833 | if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1)) | |
834 | ret = map->num_stripes; | |
835 | else | |
836 | ret = 1; | |
837 | free_extent_map(em); | |
f188591e CM |
838 | return ret; |
839 | } | |
840 | ||
8790d502 | 841 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
cea9e445 | 842 | u64 logical, u64 *length, |
f188591e | 843 | struct btrfs_multi_bio **multi_ret, int mirror_num) |
0b86a832 CM |
844 | { |
845 | struct extent_map *em; | |
846 | struct map_lookup *map; | |
847 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
848 | u64 offset; | |
593060d7 CM |
849 | u64 stripe_offset; |
850 | u64 stripe_nr; | |
cea9e445 | 851 | int stripes_allocated = 8; |
593060d7 | 852 | int stripe_index; |
cea9e445 CM |
853 | int i; |
854 | struct btrfs_multi_bio *multi = NULL; | |
0b86a832 | 855 | |
cea9e445 CM |
856 | if (multi_ret && !(rw & (1 << BIO_RW))) { |
857 | stripes_allocated = 1; | |
858 | } | |
859 | again: | |
860 | if (multi_ret) { | |
861 | multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), | |
862 | GFP_NOFS); | |
863 | if (!multi) | |
864 | return -ENOMEM; | |
865 | } | |
0b86a832 CM |
866 | |
867 | spin_lock(&em_tree->lock); | |
868 | em = lookup_extent_mapping(em_tree, logical, *length); | |
b248a415 | 869 | spin_unlock(&em_tree->lock); |
0b86a832 CM |
870 | BUG_ON(!em); |
871 | ||
872 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
873 | map = (struct map_lookup *)em->bdev; | |
874 | offset = logical - em->start; | |
593060d7 | 875 | |
f188591e CM |
876 | if (mirror_num > map->num_stripes) |
877 | mirror_num = 0; | |
878 | ||
cea9e445 CM |
879 | /* if our multi bio struct is too small, back off and try again */ |
880 | if (multi_ret && (rw & (1 << BIO_RW)) && | |
881 | stripes_allocated < map->num_stripes && | |
882 | ((map->type & BTRFS_BLOCK_GROUP_RAID1) || | |
883 | (map->type & BTRFS_BLOCK_GROUP_DUP))) { | |
884 | stripes_allocated = map->num_stripes; | |
cea9e445 CM |
885 | free_extent_map(em); |
886 | kfree(multi); | |
887 | goto again; | |
888 | } | |
593060d7 CM |
889 | stripe_nr = offset; |
890 | /* | |
891 | * stripe_nr counts the total number of stripes we have to stride | |
892 | * to get to this block | |
893 | */ | |
894 | do_div(stripe_nr, map->stripe_len); | |
895 | ||
896 | stripe_offset = stripe_nr * map->stripe_len; | |
897 | BUG_ON(offset < stripe_offset); | |
898 | ||
899 | /* stripe_offset is the offset of this block in its stripe*/ | |
900 | stripe_offset = offset - stripe_offset; | |
901 | ||
cea9e445 CM |
902 | if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | |
903 | BTRFS_BLOCK_GROUP_DUP)) { | |
904 | /* we limit the length of each bio to what fits in a stripe */ | |
905 | *length = min_t(u64, em->len - offset, | |
906 | map->stripe_len - stripe_offset); | |
907 | } else { | |
908 | *length = em->len - offset; | |
909 | } | |
910 | if (!multi_ret) | |
911 | goto out; | |
912 | ||
913 | multi->num_stripes = 1; | |
914 | stripe_index = 0; | |
8790d502 | 915 | if (map->type & BTRFS_BLOCK_GROUP_RAID1) { |
8790d502 | 916 | if (rw & (1 << BIO_RW)) |
cea9e445 | 917 | multi->num_stripes = map->num_stripes; |
f188591e CM |
918 | else if (mirror_num) { |
919 | stripe_index = mirror_num - 1; | |
920 | } else { | |
8790d502 CM |
921 | int i; |
922 | u64 least = (u64)-1; | |
923 | struct btrfs_device *cur; | |
924 | ||
925 | for (i = 0; i < map->num_stripes; i++) { | |
926 | cur = map->stripes[i].dev; | |
927 | spin_lock(&cur->io_lock); | |
928 | if (cur->total_ios < least) { | |
929 | least = cur->total_ios; | |
930 | stripe_index = i; | |
931 | } | |
932 | spin_unlock(&cur->io_lock); | |
933 | } | |
8790d502 | 934 | } |
611f0e00 | 935 | } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { |
cea9e445 CM |
936 | if (rw & (1 << BIO_RW)) |
937 | multi->num_stripes = map->num_stripes; | |
f188591e CM |
938 | else if (mirror_num) |
939 | stripe_index = mirror_num - 1; | |
8790d502 CM |
940 | } else { |
941 | /* | |
942 | * after this do_div call, stripe_nr is the number of stripes | |
943 | * on this device we have to walk to find the data, and | |
944 | * stripe_index is the number of our device in the stripe array | |
945 | */ | |
946 | stripe_index = do_div(stripe_nr, map->num_stripes); | |
947 | } | |
593060d7 | 948 | BUG_ON(stripe_index >= map->num_stripes); |
cea9e445 CM |
949 | BUG_ON(stripe_index != 0 && multi->num_stripes > 1); |
950 | ||
951 | for (i = 0; i < multi->num_stripes; i++) { | |
952 | multi->stripes[i].physical = | |
953 | map->stripes[stripe_index].physical + stripe_offset + | |
954 | stripe_nr * map->stripe_len; | |
955 | multi->stripes[i].dev = map->stripes[stripe_index].dev; | |
956 | stripe_index++; | |
593060d7 | 957 | } |
cea9e445 CM |
958 | *multi_ret = multi; |
959 | out: | |
0b86a832 | 960 | free_extent_map(em); |
0b86a832 CM |
961 | return 0; |
962 | } | |
963 | ||
8790d502 CM |
964 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
965 | static void end_bio_multi_stripe(struct bio *bio, int err) | |
966 | #else | |
967 | static int end_bio_multi_stripe(struct bio *bio, | |
968 | unsigned int bytes_done, int err) | |
969 | #endif | |
970 | { | |
cea9e445 | 971 | struct btrfs_multi_bio *multi = bio->bi_private; |
8790d502 CM |
972 | |
973 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
974 | if (bio->bi_size) | |
975 | return 1; | |
976 | #endif | |
977 | if (err) | |
978 | multi->error = err; | |
979 | ||
cea9e445 | 980 | if (atomic_dec_and_test(&multi->stripes_pending)) { |
8790d502 CM |
981 | bio->bi_private = multi->private; |
982 | bio->bi_end_io = multi->end_io; | |
983 | ||
984 | if (!err && multi->error) | |
985 | err = multi->error; | |
986 | kfree(multi); | |
987 | ||
73f61b2a M |
988 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) |
989 | bio_endio(bio, bio->bi_size, err); | |
990 | #else | |
8790d502 | 991 | bio_endio(bio, err); |
73f61b2a | 992 | #endif |
8790d502 CM |
993 | } else { |
994 | bio_put(bio); | |
995 | } | |
996 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
997 | return 0; | |
998 | #endif | |
999 | } | |
1000 | ||
f188591e CM |
1001 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, |
1002 | int mirror_num) | |
0b86a832 CM |
1003 | { |
1004 | struct btrfs_mapping_tree *map_tree; | |
1005 | struct btrfs_device *dev; | |
8790d502 | 1006 | struct bio *first_bio = bio; |
0b86a832 | 1007 | u64 logical = bio->bi_sector << 9; |
0b86a832 CM |
1008 | u64 length = 0; |
1009 | u64 map_length; | |
1010 | struct bio_vec *bvec; | |
cea9e445 | 1011 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 CM |
1012 | int i; |
1013 | int ret; | |
8790d502 CM |
1014 | int dev_nr = 0; |
1015 | int total_devs = 1; | |
0b86a832 CM |
1016 | |
1017 | bio_for_each_segment(bvec, bio, i) { | |
1018 | length += bvec->bv_len; | |
1019 | } | |
8790d502 | 1020 | |
0b86a832 CM |
1021 | map_tree = &root->fs_info->mapping_tree; |
1022 | map_length = length; | |
cea9e445 | 1023 | |
f188591e CM |
1024 | ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi, |
1025 | mirror_num); | |
cea9e445 CM |
1026 | BUG_ON(ret); |
1027 | ||
1028 | total_devs = multi->num_stripes; | |
1029 | if (map_length < length) { | |
1030 | printk("mapping failed logical %Lu bio len %Lu " | |
1031 | "len %Lu\n", logical, length, map_length); | |
1032 | BUG(); | |
1033 | } | |
1034 | multi->end_io = first_bio->bi_end_io; | |
1035 | multi->private = first_bio->bi_private; | |
1036 | atomic_set(&multi->stripes_pending, multi->num_stripes); | |
1037 | ||
8790d502 | 1038 | while(dev_nr < total_devs) { |
8790d502 | 1039 | if (total_devs > 1) { |
8790d502 CM |
1040 | if (dev_nr < total_devs - 1) { |
1041 | bio = bio_clone(first_bio, GFP_NOFS); | |
1042 | BUG_ON(!bio); | |
1043 | } else { | |
1044 | bio = first_bio; | |
1045 | } | |
1046 | bio->bi_private = multi; | |
1047 | bio->bi_end_io = end_bio_multi_stripe; | |
1048 | } | |
cea9e445 CM |
1049 | bio->bi_sector = multi->stripes[dev_nr].physical >> 9; |
1050 | dev = multi->stripes[dev_nr].dev; | |
8790d502 CM |
1051 | bio->bi_bdev = dev->bdev; |
1052 | spin_lock(&dev->io_lock); | |
1053 | dev->total_ios++; | |
1054 | spin_unlock(&dev->io_lock); | |
1055 | submit_bio(rw, bio); | |
1056 | dev_nr++; | |
1057 | } | |
cea9e445 CM |
1058 | if (total_devs == 1) |
1059 | kfree(multi); | |
0b86a832 CM |
1060 | return 0; |
1061 | } | |
1062 | ||
1063 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid) | |
1064 | { | |
8a4b83cc | 1065 | struct list_head *head = &root->fs_info->fs_devices->devices; |
0b86a832 | 1066 | |
8a4b83cc | 1067 | return __find_device(head, devid); |
0b86a832 CM |
1068 | } |
1069 | ||
1070 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, | |
1071 | struct extent_buffer *leaf, | |
1072 | struct btrfs_chunk *chunk) | |
1073 | { | |
1074 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
1075 | struct map_lookup *map; | |
1076 | struct extent_map *em; | |
1077 | u64 logical; | |
1078 | u64 length; | |
1079 | u64 devid; | |
593060d7 | 1080 | int num_stripes; |
0b86a832 | 1081 | int ret; |
593060d7 | 1082 | int i; |
0b86a832 | 1083 | |
e17cade2 CM |
1084 | logical = key->offset; |
1085 | length = btrfs_chunk_length(leaf, chunk); | |
0b86a832 CM |
1086 | spin_lock(&map_tree->map_tree.lock); |
1087 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
b248a415 | 1088 | spin_unlock(&map_tree->map_tree.lock); |
0b86a832 CM |
1089 | |
1090 | /* already mapped? */ | |
1091 | if (em && em->start <= logical && em->start + em->len > logical) { | |
1092 | free_extent_map(em); | |
0b86a832 CM |
1093 | return 0; |
1094 | } else if (em) { | |
1095 | free_extent_map(em); | |
1096 | } | |
0b86a832 CM |
1097 | |
1098 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
1099 | if (!map) | |
1100 | return -ENOMEM; | |
1101 | ||
1102 | em = alloc_extent_map(GFP_NOFS); | |
1103 | if (!em) | |
1104 | return -ENOMEM; | |
593060d7 CM |
1105 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); |
1106 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); | |
0b86a832 CM |
1107 | if (!map) { |
1108 | free_extent_map(em); | |
1109 | return -ENOMEM; | |
1110 | } | |
1111 | ||
1112 | em->bdev = (struct block_device *)map; | |
1113 | em->start = logical; | |
1114 | em->len = length; | |
1115 | em->block_start = 0; | |
1116 | ||
593060d7 CM |
1117 | map->num_stripes = num_stripes; |
1118 | map->io_width = btrfs_chunk_io_width(leaf, chunk); | |
1119 | map->io_align = btrfs_chunk_io_align(leaf, chunk); | |
1120 | map->sector_size = btrfs_chunk_sector_size(leaf, chunk); | |
1121 | map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
1122 | map->type = btrfs_chunk_type(leaf, chunk); | |
1123 | for (i = 0; i < num_stripes; i++) { | |
1124 | map->stripes[i].physical = | |
1125 | btrfs_stripe_offset_nr(leaf, chunk, i); | |
1126 | devid = btrfs_stripe_devid_nr(leaf, chunk, i); | |
1127 | map->stripes[i].dev = btrfs_find_device(root, devid); | |
1128 | if (!map->stripes[i].dev) { | |
1129 | kfree(map); | |
1130 | free_extent_map(em); | |
1131 | return -EIO; | |
1132 | } | |
0b86a832 CM |
1133 | } |
1134 | ||
1135 | spin_lock(&map_tree->map_tree.lock); | |
1136 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
0b86a832 | 1137 | spin_unlock(&map_tree->map_tree.lock); |
b248a415 | 1138 | BUG_ON(ret); |
0b86a832 CM |
1139 | free_extent_map(em); |
1140 | ||
1141 | return 0; | |
1142 | } | |
1143 | ||
1144 | static int fill_device_from_item(struct extent_buffer *leaf, | |
1145 | struct btrfs_dev_item *dev_item, | |
1146 | struct btrfs_device *device) | |
1147 | { | |
1148 | unsigned long ptr; | |
0b86a832 CM |
1149 | |
1150 | device->devid = btrfs_device_id(leaf, dev_item); | |
1151 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
1152 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
1153 | device->type = btrfs_device_type(leaf, dev_item); | |
1154 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
1155 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
1156 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
0b86a832 CM |
1157 | |
1158 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
e17cade2 | 1159 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 | 1160 | |
0b86a832 CM |
1161 | return 0; |
1162 | } | |
1163 | ||
0d81ba5d | 1164 | static int read_one_dev(struct btrfs_root *root, |
0b86a832 CM |
1165 | struct extent_buffer *leaf, |
1166 | struct btrfs_dev_item *dev_item) | |
1167 | { | |
1168 | struct btrfs_device *device; | |
1169 | u64 devid; | |
1170 | int ret; | |
0b86a832 | 1171 | devid = btrfs_device_id(leaf, dev_item); |
6324fbf3 CM |
1172 | device = btrfs_find_device(root, devid); |
1173 | if (!device) { | |
8a4b83cc | 1174 | printk("warning devid %Lu not found already\n", devid); |
f2984462 | 1175 | device = kzalloc(sizeof(*device), GFP_NOFS); |
6324fbf3 CM |
1176 | if (!device) |
1177 | return -ENOMEM; | |
8a4b83cc CM |
1178 | list_add(&device->dev_list, |
1179 | &root->fs_info->fs_devices->devices); | |
b248a415 | 1180 | device->barriers = 1; |
8790d502 | 1181 | spin_lock_init(&device->io_lock); |
6324fbf3 | 1182 | } |
0b86a832 CM |
1183 | |
1184 | fill_device_from_item(leaf, dev_item, device); | |
1185 | device->dev_root = root->fs_info->dev_root; | |
0b86a832 CM |
1186 | ret = 0; |
1187 | #if 0 | |
1188 | ret = btrfs_open_device(device); | |
1189 | if (ret) { | |
1190 | kfree(device); | |
1191 | } | |
1192 | #endif | |
1193 | return ret; | |
1194 | } | |
1195 | ||
0d81ba5d CM |
1196 | int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) |
1197 | { | |
1198 | struct btrfs_dev_item *dev_item; | |
1199 | ||
1200 | dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, | |
1201 | dev_item); | |
1202 | return read_one_dev(root, buf, dev_item); | |
1203 | } | |
1204 | ||
0b86a832 CM |
1205 | int btrfs_read_sys_array(struct btrfs_root *root) |
1206 | { | |
1207 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1208 | struct extent_buffer *sb = root->fs_info->sb_buffer; | |
1209 | struct btrfs_disk_key *disk_key; | |
0b86a832 CM |
1210 | struct btrfs_chunk *chunk; |
1211 | struct btrfs_key key; | |
1212 | u32 num_stripes; | |
1213 | u32 array_size; | |
1214 | u32 len = 0; | |
1215 | u8 *ptr; | |
1216 | unsigned long sb_ptr; | |
1217 | u32 cur; | |
1218 | int ret; | |
0b86a832 CM |
1219 | |
1220 | array_size = btrfs_super_sys_array_size(super_copy); | |
1221 | ||
1222 | /* | |
1223 | * we do this loop twice, once for the device items and | |
1224 | * once for all of the chunks. This way there are device | |
1225 | * structs filled in for every chunk | |
1226 | */ | |
0b86a832 CM |
1227 | ptr = super_copy->sys_chunk_array; |
1228 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
1229 | cur = 0; | |
1230 | ||
1231 | while (cur < array_size) { | |
1232 | disk_key = (struct btrfs_disk_key *)ptr; | |
1233 | btrfs_disk_key_to_cpu(&key, disk_key); | |
1234 | ||
1235 | len = sizeof(*disk_key); | |
1236 | ptr += len; | |
1237 | sb_ptr += len; | |
1238 | cur += len; | |
1239 | ||
0d81ba5d | 1240 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { |
0b86a832 | 1241 | chunk = (struct btrfs_chunk *)sb_ptr; |
0d81ba5d CM |
1242 | ret = read_one_chunk(root, &key, sb, chunk); |
1243 | BUG_ON(ret); | |
0b86a832 CM |
1244 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); |
1245 | len = btrfs_chunk_item_size(num_stripes); | |
1246 | } else { | |
1247 | BUG(); | |
1248 | } | |
1249 | ptr += len; | |
1250 | sb_ptr += len; | |
1251 | cur += len; | |
1252 | } | |
0b86a832 CM |
1253 | return 0; |
1254 | } | |
1255 | ||
1256 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
1257 | { | |
1258 | struct btrfs_path *path; | |
1259 | struct extent_buffer *leaf; | |
1260 | struct btrfs_key key; | |
1261 | struct btrfs_key found_key; | |
1262 | int ret; | |
1263 | int slot; | |
1264 | ||
1265 | root = root->fs_info->chunk_root; | |
1266 | ||
1267 | path = btrfs_alloc_path(); | |
1268 | if (!path) | |
1269 | return -ENOMEM; | |
1270 | ||
1271 | /* first we search for all of the device items, and then we | |
1272 | * read in all of the chunk items. This way we can create chunk | |
1273 | * mappings that reference all of the devices that are afound | |
1274 | */ | |
1275 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1276 | key.offset = 0; | |
1277 | key.type = 0; | |
1278 | again: | |
1279 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1280 | while(1) { | |
1281 | leaf = path->nodes[0]; | |
1282 | slot = path->slots[0]; | |
1283 | if (slot >= btrfs_header_nritems(leaf)) { | |
1284 | ret = btrfs_next_leaf(root, path); | |
1285 | if (ret == 0) | |
1286 | continue; | |
1287 | if (ret < 0) | |
1288 | goto error; | |
1289 | break; | |
1290 | } | |
1291 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1292 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1293 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
1294 | break; | |
1295 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
1296 | struct btrfs_dev_item *dev_item; | |
1297 | dev_item = btrfs_item_ptr(leaf, slot, | |
1298 | struct btrfs_dev_item); | |
0d81ba5d | 1299 | ret = read_one_dev(root, leaf, dev_item); |
0b86a832 CM |
1300 | BUG_ON(ret); |
1301 | } | |
1302 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
1303 | struct btrfs_chunk *chunk; | |
1304 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
1305 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
1306 | } | |
1307 | path->slots[0]++; | |
1308 | } | |
1309 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1310 | key.objectid = 0; | |
1311 | btrfs_release_path(root, path); | |
1312 | goto again; | |
1313 | } | |
1314 | ||
1315 | btrfs_free_path(path); | |
1316 | ret = 0; | |
1317 | error: | |
1318 | return ret; | |
1319 | } | |
1320 |