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; |
8a4b83cc CM |
129 | device->name = kstrdup(path, GFP_NOFS); |
130 | if (!device->name) { | |
131 | kfree(device); | |
132 | return -ENOMEM; | |
133 | } | |
134 | list_add(&device->dev_list, &fs_devices->devices); | |
135 | fs_devices->num_devices++; | |
136 | } | |
137 | ||
138 | if (found_transid > fs_devices->latest_trans) { | |
139 | fs_devices->latest_devid = devid; | |
140 | fs_devices->latest_trans = found_transid; | |
141 | } | |
142 | if (fs_devices->lowest_devid > devid) { | |
143 | fs_devices->lowest_devid = devid; | |
144 | printk("lowest devid now %Lu\n", devid); | |
145 | } | |
146 | *fs_devices_ret = fs_devices; | |
147 | return 0; | |
148 | } | |
149 | ||
150 | int btrfs_close_devices(struct btrfs_fs_devices *fs_devices) | |
151 | { | |
152 | struct list_head *head = &fs_devices->devices; | |
153 | struct list_head *cur; | |
154 | struct btrfs_device *device; | |
155 | ||
156 | mutex_lock(&uuid_mutex); | |
157 | list_for_each(cur, head) { | |
158 | device = list_entry(cur, struct btrfs_device, dev_list); | |
159 | if (device->bdev) { | |
160 | close_bdev_excl(device->bdev); | |
161 | printk("close devices closes %s\n", device->name); | |
162 | } | |
163 | device->bdev = NULL; | |
164 | } | |
165 | mutex_unlock(&uuid_mutex); | |
166 | return 0; | |
167 | } | |
168 | ||
169 | int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, | |
170 | int flags, void *holder) | |
171 | { | |
172 | struct block_device *bdev; | |
173 | struct list_head *head = &fs_devices->devices; | |
174 | struct list_head *cur; | |
175 | struct btrfs_device *device; | |
176 | int ret; | |
177 | ||
178 | mutex_lock(&uuid_mutex); | |
179 | list_for_each(cur, head) { | |
180 | device = list_entry(cur, struct btrfs_device, dev_list); | |
181 | bdev = open_bdev_excl(device->name, flags, holder); | |
182 | printk("opening %s devid %Lu\n", device->name, device->devid); | |
183 | if (IS_ERR(bdev)) { | |
184 | printk("open %s failed\n", device->name); | |
185 | ret = PTR_ERR(bdev); | |
186 | goto fail; | |
187 | } | |
188 | if (device->devid == fs_devices->latest_devid) | |
189 | fs_devices->latest_bdev = bdev; | |
190 | if (device->devid == fs_devices->lowest_devid) { | |
191 | fs_devices->lowest_bdev = bdev; | |
192 | printk("lowest bdev %s\n", device->name); | |
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, | |
374 | u64 owner, u64 num_bytes, u64 *start) | |
375 | { | |
376 | int ret; | |
377 | struct btrfs_path *path; | |
378 | struct btrfs_root *root = device->dev_root; | |
379 | struct btrfs_dev_extent *extent; | |
380 | struct extent_buffer *leaf; | |
381 | struct btrfs_key key; | |
382 | ||
383 | path = btrfs_alloc_path(); | |
384 | if (!path) | |
385 | return -ENOMEM; | |
386 | ||
387 | ret = find_free_dev_extent(trans, device, path, num_bytes, start); | |
6324fbf3 | 388 | if (ret) { |
0b86a832 | 389 | goto err; |
6324fbf3 | 390 | } |
0b86a832 CM |
391 | |
392 | key.objectid = device->devid; | |
393 | key.offset = *start; | |
394 | key.type = BTRFS_DEV_EXTENT_KEY; | |
395 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
396 | sizeof(*extent)); | |
397 | BUG_ON(ret); | |
398 | ||
399 | leaf = path->nodes[0]; | |
400 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
401 | struct btrfs_dev_extent); | |
402 | btrfs_set_dev_extent_owner(leaf, extent, owner); | |
403 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); | |
404 | btrfs_mark_buffer_dirty(leaf); | |
405 | err: | |
406 | btrfs_free_path(path); | |
407 | return ret; | |
408 | } | |
409 | ||
410 | static int find_next_chunk(struct btrfs_root *root, u64 *objectid) | |
411 | { | |
412 | struct btrfs_path *path; | |
413 | int ret; | |
414 | struct btrfs_key key; | |
415 | struct btrfs_key found_key; | |
416 | ||
417 | path = btrfs_alloc_path(); | |
418 | BUG_ON(!path); | |
419 | ||
420 | key.objectid = (u64)-1; | |
421 | key.offset = (u64)-1; | |
422 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
423 | ||
424 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
425 | if (ret < 0) | |
426 | goto error; | |
427 | ||
428 | BUG_ON(ret == 0); | |
429 | ||
430 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
431 | if (ret) { | |
432 | *objectid = 0; | |
433 | } else { | |
434 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
435 | path->slots[0]); | |
436 | *objectid = found_key.objectid + found_key.offset; | |
437 | } | |
438 | ret = 0; | |
439 | error: | |
440 | btrfs_free_path(path); | |
441 | return ret; | |
442 | } | |
443 | ||
0b86a832 CM |
444 | static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, |
445 | u64 *objectid) | |
446 | { | |
447 | int ret; | |
448 | struct btrfs_key key; | |
449 | struct btrfs_key found_key; | |
450 | ||
451 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
452 | key.type = BTRFS_DEV_ITEM_KEY; | |
453 | key.offset = (u64)-1; | |
454 | ||
455 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
456 | if (ret < 0) | |
457 | goto error; | |
458 | ||
459 | BUG_ON(ret == 0); | |
460 | ||
461 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
462 | BTRFS_DEV_ITEM_KEY); | |
463 | if (ret) { | |
464 | *objectid = 1; | |
465 | } else { | |
466 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
467 | path->slots[0]); | |
468 | *objectid = found_key.offset + 1; | |
469 | } | |
470 | ret = 0; | |
471 | error: | |
472 | btrfs_release_path(root, path); | |
473 | return ret; | |
474 | } | |
475 | ||
476 | /* | |
477 | * the device information is stored in the chunk root | |
478 | * the btrfs_device struct should be fully filled in | |
479 | */ | |
480 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
481 | struct btrfs_root *root, | |
482 | struct btrfs_device *device) | |
483 | { | |
484 | int ret; | |
485 | struct btrfs_path *path; | |
486 | struct btrfs_dev_item *dev_item; | |
487 | struct extent_buffer *leaf; | |
488 | struct btrfs_key key; | |
489 | unsigned long ptr; | |
490 | u64 free_devid; | |
491 | ||
492 | root = root->fs_info->chunk_root; | |
493 | ||
494 | path = btrfs_alloc_path(); | |
495 | if (!path) | |
496 | return -ENOMEM; | |
497 | ||
498 | ret = find_next_devid(root, path, &free_devid); | |
499 | if (ret) | |
500 | goto out; | |
501 | ||
502 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
503 | key.type = BTRFS_DEV_ITEM_KEY; | |
504 | key.offset = free_devid; | |
505 | ||
506 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
0d81ba5d | 507 | sizeof(*dev_item)); |
0b86a832 CM |
508 | if (ret) |
509 | goto out; | |
510 | ||
511 | leaf = path->nodes[0]; | |
512 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
513 | ||
8a4b83cc | 514 | device->devid = free_devid; |
0b86a832 CM |
515 | btrfs_set_device_id(leaf, dev_item, device->devid); |
516 | btrfs_set_device_type(leaf, dev_item, device->type); | |
517 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
518 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
519 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
520 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
521 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
522 | ||
0b86a832 CM |
523 | ptr = (unsigned long)btrfs_device_uuid(dev_item); |
524 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); | |
525 | btrfs_mark_buffer_dirty(leaf); | |
526 | ret = 0; | |
527 | ||
528 | out: | |
529 | btrfs_free_path(path); | |
530 | return ret; | |
531 | } | |
532 | int btrfs_update_device(struct btrfs_trans_handle *trans, | |
533 | struct btrfs_device *device) | |
534 | { | |
535 | int ret; | |
536 | struct btrfs_path *path; | |
537 | struct btrfs_root *root; | |
538 | struct btrfs_dev_item *dev_item; | |
539 | struct extent_buffer *leaf; | |
540 | struct btrfs_key key; | |
541 | ||
542 | root = device->dev_root->fs_info->chunk_root; | |
543 | ||
544 | path = btrfs_alloc_path(); | |
545 | if (!path) | |
546 | return -ENOMEM; | |
547 | ||
548 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
549 | key.type = BTRFS_DEV_ITEM_KEY; | |
550 | key.offset = device->devid; | |
551 | ||
552 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
553 | if (ret < 0) | |
554 | goto out; | |
555 | ||
556 | if (ret > 0) { | |
557 | ret = -ENOENT; | |
558 | goto out; | |
559 | } | |
560 | ||
561 | leaf = path->nodes[0]; | |
562 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
563 | ||
564 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
565 | btrfs_set_device_type(leaf, dev_item, device->type); | |
566 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
567 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
568 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
569 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
570 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
571 | btrfs_mark_buffer_dirty(leaf); | |
572 | ||
573 | out: | |
574 | btrfs_free_path(path); | |
575 | return ret; | |
576 | } | |
577 | ||
578 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, | |
579 | struct btrfs_root *root, | |
580 | struct btrfs_key *key, | |
581 | struct btrfs_chunk *chunk, int item_size) | |
582 | { | |
583 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
584 | struct btrfs_disk_key disk_key; | |
585 | u32 array_size; | |
586 | u8 *ptr; | |
587 | ||
588 | array_size = btrfs_super_sys_array_size(super_copy); | |
589 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
590 | return -EFBIG; | |
591 | ||
592 | ptr = super_copy->sys_chunk_array + array_size; | |
593 | btrfs_cpu_key_to_disk(&disk_key, key); | |
594 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
595 | ptr += sizeof(disk_key); | |
596 | memcpy(ptr, chunk, item_size); | |
597 | item_size += sizeof(disk_key); | |
598 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
599 | return 0; | |
600 | } | |
601 | ||
602 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, | |
603 | struct btrfs_root *extent_root, u64 *start, | |
6324fbf3 | 604 | u64 *num_bytes, u64 type) |
0b86a832 CM |
605 | { |
606 | u64 dev_offset; | |
593060d7 | 607 | struct btrfs_fs_info *info = extent_root->fs_info; |
0b86a832 CM |
608 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; |
609 | struct btrfs_stripe *stripes; | |
610 | struct btrfs_device *device = NULL; | |
611 | struct btrfs_chunk *chunk; | |
6324fbf3 | 612 | struct list_head private_devs; |
8a4b83cc | 613 | struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices; |
6324fbf3 | 614 | struct list_head *cur; |
0b86a832 CM |
615 | struct extent_map_tree *em_tree; |
616 | struct map_lookup *map; | |
617 | struct extent_map *em; | |
618 | u64 physical; | |
619 | u64 calc_size = 1024 * 1024 * 1024; | |
611f0e00 | 620 | u64 min_free = calc_size; |
6324fbf3 CM |
621 | u64 avail; |
622 | u64 max_avail = 0; | |
623 | int num_stripes = 1; | |
624 | int looped = 0; | |
0b86a832 | 625 | int ret; |
6324fbf3 | 626 | int index; |
593060d7 | 627 | int stripe_len = 64 * 1024; |
0b86a832 CM |
628 | struct btrfs_key key; |
629 | ||
6324fbf3 CM |
630 | if (list_empty(dev_list)) |
631 | return -ENOSPC; | |
593060d7 | 632 | |
8790d502 | 633 | if (type & (BTRFS_BLOCK_GROUP_RAID0)) |
593060d7 | 634 | num_stripes = btrfs_super_num_devices(&info->super_copy); |
611f0e00 CM |
635 | if (type & (BTRFS_BLOCK_GROUP_DUP)) |
636 | num_stripes = 2; | |
8790d502 CM |
637 | if (type & (BTRFS_BLOCK_GROUP_RAID1)) { |
638 | num_stripes = min_t(u64, 2, | |
639 | btrfs_super_num_devices(&info->super_copy)); | |
640 | } | |
6324fbf3 CM |
641 | again: |
642 | INIT_LIST_HEAD(&private_devs); | |
643 | cur = dev_list->next; | |
644 | index = 0; | |
611f0e00 CM |
645 | |
646 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
647 | min_free = calc_size * 2; | |
648 | ||
6324fbf3 CM |
649 | /* build a private list of devices we will allocate from */ |
650 | while(index < num_stripes) { | |
651 | device = list_entry(cur, struct btrfs_device, dev_list); | |
611f0e00 | 652 | |
6324fbf3 CM |
653 | avail = device->total_bytes - device->bytes_used; |
654 | cur = cur->next; | |
655 | if (avail > max_avail) | |
656 | max_avail = avail; | |
611f0e00 | 657 | if (avail >= min_free) { |
6324fbf3 CM |
658 | list_move_tail(&device->dev_list, &private_devs); |
659 | index++; | |
611f0e00 CM |
660 | if (type & BTRFS_BLOCK_GROUP_DUP) |
661 | index++; | |
6324fbf3 CM |
662 | } |
663 | if (cur == dev_list) | |
664 | break; | |
665 | } | |
666 | if (index < num_stripes) { | |
667 | list_splice(&private_devs, dev_list); | |
668 | if (!looped && max_avail > 0) { | |
669 | looped = 1; | |
670 | calc_size = max_avail; | |
671 | goto again; | |
672 | } | |
673 | return -ENOSPC; | |
674 | } | |
0b86a832 CM |
675 | |
676 | ret = find_next_chunk(chunk_root, &key.objectid); | |
677 | if (ret) | |
678 | return ret; | |
679 | ||
0b86a832 CM |
680 | chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); |
681 | if (!chunk) | |
682 | return -ENOMEM; | |
683 | ||
593060d7 CM |
684 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); |
685 | if (!map) { | |
686 | kfree(chunk); | |
687 | return -ENOMEM; | |
688 | } | |
689 | ||
0b86a832 CM |
690 | stripes = &chunk->stripe; |
691 | ||
611f0e00 | 692 | if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP)) |
8790d502 CM |
693 | *num_bytes = calc_size; |
694 | else | |
695 | *num_bytes = calc_size * num_stripes; | |
696 | ||
6324fbf3 | 697 | index = 0; |
611f0e00 | 698 | printk("new chunk type %Lu start %Lu size %Lu\n", type, key.objectid, *num_bytes); |
0b86a832 | 699 | while(index < num_stripes) { |
6324fbf3 CM |
700 | BUG_ON(list_empty(&private_devs)); |
701 | cur = private_devs.next; | |
702 | device = list_entry(cur, struct btrfs_device, dev_list); | |
611f0e00 CM |
703 | |
704 | /* loop over this device again if we're doing a dup group */ | |
705 | if (!(type & BTRFS_BLOCK_GROUP_DUP) || | |
706 | (index == num_stripes - 1)) | |
707 | list_move_tail(&device->dev_list, dev_list); | |
0b86a832 CM |
708 | |
709 | ret = btrfs_alloc_dev_extent(trans, device, | |
710 | key.objectid, | |
711 | calc_size, &dev_offset); | |
712 | BUG_ON(ret); | |
8790d502 | 713 | printk("alloc chunk start %Lu size %Lu from dev %Lu type %Lu\n", key.objectid, calc_size, device->devid, type); |
0b86a832 CM |
714 | device->bytes_used += calc_size; |
715 | ret = btrfs_update_device(trans, device); | |
716 | BUG_ON(ret); | |
717 | ||
593060d7 CM |
718 | map->stripes[index].dev = device; |
719 | map->stripes[index].physical = dev_offset; | |
0b86a832 CM |
720 | btrfs_set_stack_stripe_devid(stripes + index, device->devid); |
721 | btrfs_set_stack_stripe_offset(stripes + index, dev_offset); | |
722 | physical = dev_offset; | |
723 | index++; | |
724 | } | |
6324fbf3 | 725 | BUG_ON(!list_empty(&private_devs)); |
0b86a832 CM |
726 | |
727 | /* key.objectid was set above */ | |
728 | key.offset = *num_bytes; | |
729 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
730 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); | |
593060d7 | 731 | btrfs_set_stack_chunk_stripe_len(chunk, stripe_len); |
0b86a832 CM |
732 | btrfs_set_stack_chunk_type(chunk, type); |
733 | btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); | |
593060d7 CM |
734 | btrfs_set_stack_chunk_io_align(chunk, stripe_len); |
735 | btrfs_set_stack_chunk_io_width(chunk, stripe_len); | |
0b86a832 | 736 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); |
593060d7 CM |
737 | map->sector_size = extent_root->sectorsize; |
738 | map->stripe_len = stripe_len; | |
739 | map->io_align = stripe_len; | |
740 | map->io_width = stripe_len; | |
741 | map->type = type; | |
742 | map->num_stripes = num_stripes; | |
0b86a832 CM |
743 | |
744 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, | |
745 | btrfs_chunk_item_size(num_stripes)); | |
746 | BUG_ON(ret); | |
747 | *start = key.objectid; | |
748 | ||
749 | em = alloc_extent_map(GFP_NOFS); | |
750 | if (!em) | |
751 | return -ENOMEM; | |
0b86a832 CM |
752 | em->bdev = (struct block_device *)map; |
753 | em->start = key.objectid; | |
754 | em->len = key.offset; | |
755 | em->block_start = 0; | |
756 | ||
0b86a832 CM |
757 | kfree(chunk); |
758 | ||
759 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; | |
760 | spin_lock(&em_tree->lock); | |
761 | ret = add_extent_mapping(em_tree, em); | |
762 | BUG_ON(ret); | |
763 | spin_unlock(&em_tree->lock); | |
764 | free_extent_map(em); | |
765 | return ret; | |
766 | } | |
767 | ||
768 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
769 | { | |
770 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
771 | } | |
772 | ||
773 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
774 | { | |
775 | struct extent_map *em; | |
776 | ||
777 | while(1) { | |
778 | spin_lock(&tree->map_tree.lock); | |
779 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
780 | if (em) | |
781 | remove_extent_mapping(&tree->map_tree, em); | |
782 | spin_unlock(&tree->map_tree.lock); | |
783 | if (!em) | |
784 | break; | |
785 | kfree(em->bdev); | |
786 | /* once for us */ | |
787 | free_extent_map(em); | |
788 | /* once for the tree */ | |
789 | free_extent_map(em); | |
790 | } | |
791 | } | |
792 | ||
f188591e CM |
793 | int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) |
794 | { | |
795 | struct extent_map *em; | |
796 | struct map_lookup *map; | |
797 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
798 | int ret; | |
799 | ||
800 | spin_lock(&em_tree->lock); | |
801 | em = lookup_extent_mapping(em_tree, logical, len); | |
802 | BUG_ON(!em); | |
803 | ||
804 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
805 | map = (struct map_lookup *)em->bdev; | |
806 | if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1)) | |
807 | ret = map->num_stripes; | |
808 | else | |
809 | ret = 1; | |
810 | free_extent_map(em); | |
811 | spin_unlock(&em_tree->lock); | |
812 | return ret; | |
813 | } | |
814 | ||
8790d502 | 815 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
cea9e445 | 816 | u64 logical, u64 *length, |
f188591e | 817 | struct btrfs_multi_bio **multi_ret, int mirror_num) |
0b86a832 CM |
818 | { |
819 | struct extent_map *em; | |
820 | struct map_lookup *map; | |
821 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
822 | u64 offset; | |
593060d7 CM |
823 | u64 stripe_offset; |
824 | u64 stripe_nr; | |
cea9e445 | 825 | int stripes_allocated = 8; |
593060d7 | 826 | int stripe_index; |
cea9e445 CM |
827 | int i; |
828 | struct btrfs_multi_bio *multi = NULL; | |
0b86a832 | 829 | |
cea9e445 CM |
830 | if (multi_ret && !(rw & (1 << BIO_RW))) { |
831 | stripes_allocated = 1; | |
832 | } | |
833 | again: | |
834 | if (multi_ret) { | |
835 | multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), | |
836 | GFP_NOFS); | |
837 | if (!multi) | |
838 | return -ENOMEM; | |
839 | } | |
0b86a832 CM |
840 | |
841 | spin_lock(&em_tree->lock); | |
842 | em = lookup_extent_mapping(em_tree, logical, *length); | |
843 | BUG_ON(!em); | |
844 | ||
845 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
846 | map = (struct map_lookup *)em->bdev; | |
847 | offset = logical - em->start; | |
593060d7 | 848 | |
f188591e CM |
849 | if (mirror_num > map->num_stripes) |
850 | mirror_num = 0; | |
851 | ||
cea9e445 CM |
852 | /* if our multi bio struct is too small, back off and try again */ |
853 | if (multi_ret && (rw & (1 << BIO_RW)) && | |
854 | stripes_allocated < map->num_stripes && | |
855 | ((map->type & BTRFS_BLOCK_GROUP_RAID1) || | |
856 | (map->type & BTRFS_BLOCK_GROUP_DUP))) { | |
857 | stripes_allocated = map->num_stripes; | |
858 | spin_unlock(&em_tree->lock); | |
859 | free_extent_map(em); | |
860 | kfree(multi); | |
861 | goto again; | |
862 | } | |
593060d7 CM |
863 | stripe_nr = offset; |
864 | /* | |
865 | * stripe_nr counts the total number of stripes we have to stride | |
866 | * to get to this block | |
867 | */ | |
868 | do_div(stripe_nr, map->stripe_len); | |
869 | ||
870 | stripe_offset = stripe_nr * map->stripe_len; | |
871 | BUG_ON(offset < stripe_offset); | |
872 | ||
873 | /* stripe_offset is the offset of this block in its stripe*/ | |
874 | stripe_offset = offset - stripe_offset; | |
875 | ||
cea9e445 CM |
876 | if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | |
877 | BTRFS_BLOCK_GROUP_DUP)) { | |
878 | /* we limit the length of each bio to what fits in a stripe */ | |
879 | *length = min_t(u64, em->len - offset, | |
880 | map->stripe_len - stripe_offset); | |
881 | } else { | |
882 | *length = em->len - offset; | |
883 | } | |
884 | if (!multi_ret) | |
885 | goto out; | |
886 | ||
887 | multi->num_stripes = 1; | |
888 | stripe_index = 0; | |
8790d502 | 889 | if (map->type & BTRFS_BLOCK_GROUP_RAID1) { |
8790d502 | 890 | if (rw & (1 << BIO_RW)) |
cea9e445 | 891 | multi->num_stripes = map->num_stripes; |
f188591e CM |
892 | else if (mirror_num) { |
893 | stripe_index = mirror_num - 1; | |
894 | } else { | |
8790d502 CM |
895 | int i; |
896 | u64 least = (u64)-1; | |
897 | struct btrfs_device *cur; | |
898 | ||
899 | for (i = 0; i < map->num_stripes; i++) { | |
900 | cur = map->stripes[i].dev; | |
901 | spin_lock(&cur->io_lock); | |
902 | if (cur->total_ios < least) { | |
903 | least = cur->total_ios; | |
904 | stripe_index = i; | |
905 | } | |
906 | spin_unlock(&cur->io_lock); | |
907 | } | |
8790d502 | 908 | } |
611f0e00 | 909 | } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { |
cea9e445 CM |
910 | if (rw & (1 << BIO_RW)) |
911 | multi->num_stripes = map->num_stripes; | |
f188591e CM |
912 | else if (mirror_num) |
913 | stripe_index = mirror_num - 1; | |
8790d502 CM |
914 | } else { |
915 | /* | |
916 | * after this do_div call, stripe_nr is the number of stripes | |
917 | * on this device we have to walk to find the data, and | |
918 | * stripe_index is the number of our device in the stripe array | |
919 | */ | |
920 | stripe_index = do_div(stripe_nr, map->num_stripes); | |
921 | } | |
593060d7 | 922 | BUG_ON(stripe_index >= map->num_stripes); |
cea9e445 CM |
923 | BUG_ON(stripe_index != 0 && multi->num_stripes > 1); |
924 | ||
925 | for (i = 0; i < multi->num_stripes; i++) { | |
926 | multi->stripes[i].physical = | |
927 | map->stripes[stripe_index].physical + stripe_offset + | |
928 | stripe_nr * map->stripe_len; | |
929 | multi->stripes[i].dev = map->stripes[stripe_index].dev; | |
930 | stripe_index++; | |
593060d7 | 931 | } |
cea9e445 CM |
932 | *multi_ret = multi; |
933 | out: | |
0b86a832 CM |
934 | free_extent_map(em); |
935 | spin_unlock(&em_tree->lock); | |
936 | return 0; | |
937 | } | |
938 | ||
8790d502 CM |
939 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
940 | static void end_bio_multi_stripe(struct bio *bio, int err) | |
941 | #else | |
942 | static int end_bio_multi_stripe(struct bio *bio, | |
943 | unsigned int bytes_done, int err) | |
944 | #endif | |
945 | { | |
cea9e445 | 946 | struct btrfs_multi_bio *multi = bio->bi_private; |
8790d502 CM |
947 | |
948 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
949 | if (bio->bi_size) | |
950 | return 1; | |
951 | #endif | |
952 | if (err) | |
953 | multi->error = err; | |
954 | ||
cea9e445 | 955 | if (atomic_dec_and_test(&multi->stripes_pending)) { |
8790d502 CM |
956 | bio->bi_private = multi->private; |
957 | bio->bi_end_io = multi->end_io; | |
958 | ||
959 | if (!err && multi->error) | |
960 | err = multi->error; | |
961 | kfree(multi); | |
962 | ||
963 | bio_endio(bio, err); | |
964 | } else { | |
965 | bio_put(bio); | |
966 | } | |
967 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
968 | return 0; | |
969 | #endif | |
970 | } | |
971 | ||
f188591e CM |
972 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, |
973 | int mirror_num) | |
0b86a832 CM |
974 | { |
975 | struct btrfs_mapping_tree *map_tree; | |
976 | struct btrfs_device *dev; | |
8790d502 | 977 | struct bio *first_bio = bio; |
0b86a832 | 978 | u64 logical = bio->bi_sector << 9; |
0b86a832 CM |
979 | u64 length = 0; |
980 | u64 map_length; | |
981 | struct bio_vec *bvec; | |
cea9e445 | 982 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 CM |
983 | int i; |
984 | int ret; | |
8790d502 CM |
985 | int dev_nr = 0; |
986 | int total_devs = 1; | |
0b86a832 CM |
987 | |
988 | bio_for_each_segment(bvec, bio, i) { | |
989 | length += bvec->bv_len; | |
990 | } | |
8790d502 | 991 | |
0b86a832 CM |
992 | map_tree = &root->fs_info->mapping_tree; |
993 | map_length = length; | |
cea9e445 | 994 | |
f188591e CM |
995 | ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi, |
996 | mirror_num); | |
cea9e445 CM |
997 | BUG_ON(ret); |
998 | ||
999 | total_devs = multi->num_stripes; | |
1000 | if (map_length < length) { | |
1001 | printk("mapping failed logical %Lu bio len %Lu " | |
1002 | "len %Lu\n", logical, length, map_length); | |
1003 | BUG(); | |
1004 | } | |
1005 | multi->end_io = first_bio->bi_end_io; | |
1006 | multi->private = first_bio->bi_private; | |
1007 | atomic_set(&multi->stripes_pending, multi->num_stripes); | |
1008 | ||
8790d502 | 1009 | while(dev_nr < total_devs) { |
8790d502 | 1010 | if (total_devs > 1) { |
8790d502 CM |
1011 | if (dev_nr < total_devs - 1) { |
1012 | bio = bio_clone(first_bio, GFP_NOFS); | |
1013 | BUG_ON(!bio); | |
1014 | } else { | |
1015 | bio = first_bio; | |
1016 | } | |
1017 | bio->bi_private = multi; | |
1018 | bio->bi_end_io = end_bio_multi_stripe; | |
1019 | } | |
cea9e445 CM |
1020 | bio->bi_sector = multi->stripes[dev_nr].physical >> 9; |
1021 | dev = multi->stripes[dev_nr].dev; | |
8790d502 CM |
1022 | bio->bi_bdev = dev->bdev; |
1023 | spin_lock(&dev->io_lock); | |
1024 | dev->total_ios++; | |
1025 | spin_unlock(&dev->io_lock); | |
1026 | submit_bio(rw, bio); | |
1027 | dev_nr++; | |
1028 | } | |
cea9e445 CM |
1029 | if (total_devs == 1) |
1030 | kfree(multi); | |
0b86a832 CM |
1031 | return 0; |
1032 | } | |
1033 | ||
1034 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid) | |
1035 | { | |
8a4b83cc | 1036 | struct list_head *head = &root->fs_info->fs_devices->devices; |
0b86a832 | 1037 | |
8a4b83cc | 1038 | return __find_device(head, devid); |
0b86a832 CM |
1039 | } |
1040 | ||
1041 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, | |
1042 | struct extent_buffer *leaf, | |
1043 | struct btrfs_chunk *chunk) | |
1044 | { | |
1045 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
1046 | struct map_lookup *map; | |
1047 | struct extent_map *em; | |
1048 | u64 logical; | |
1049 | u64 length; | |
1050 | u64 devid; | |
593060d7 | 1051 | int num_stripes; |
0b86a832 | 1052 | int ret; |
593060d7 | 1053 | int i; |
0b86a832 CM |
1054 | |
1055 | logical = key->objectid; | |
1056 | length = key->offset; | |
1057 | spin_lock(&map_tree->map_tree.lock); | |
1058 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
1059 | ||
1060 | /* already mapped? */ | |
1061 | if (em && em->start <= logical && em->start + em->len > logical) { | |
1062 | free_extent_map(em); | |
1063 | spin_unlock(&map_tree->map_tree.lock); | |
1064 | return 0; | |
1065 | } else if (em) { | |
1066 | free_extent_map(em); | |
1067 | } | |
1068 | spin_unlock(&map_tree->map_tree.lock); | |
1069 | ||
1070 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
1071 | if (!map) | |
1072 | return -ENOMEM; | |
1073 | ||
1074 | em = alloc_extent_map(GFP_NOFS); | |
1075 | if (!em) | |
1076 | return -ENOMEM; | |
593060d7 CM |
1077 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); |
1078 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); | |
0b86a832 CM |
1079 | if (!map) { |
1080 | free_extent_map(em); | |
1081 | return -ENOMEM; | |
1082 | } | |
1083 | ||
1084 | em->bdev = (struct block_device *)map; | |
1085 | em->start = logical; | |
1086 | em->len = length; | |
1087 | em->block_start = 0; | |
1088 | ||
593060d7 CM |
1089 | map->num_stripes = num_stripes; |
1090 | map->io_width = btrfs_chunk_io_width(leaf, chunk); | |
1091 | map->io_align = btrfs_chunk_io_align(leaf, chunk); | |
1092 | map->sector_size = btrfs_chunk_sector_size(leaf, chunk); | |
1093 | map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
1094 | map->type = btrfs_chunk_type(leaf, chunk); | |
1095 | for (i = 0; i < num_stripes; i++) { | |
1096 | map->stripes[i].physical = | |
1097 | btrfs_stripe_offset_nr(leaf, chunk, i); | |
1098 | devid = btrfs_stripe_devid_nr(leaf, chunk, i); | |
1099 | map->stripes[i].dev = btrfs_find_device(root, devid); | |
1100 | if (!map->stripes[i].dev) { | |
1101 | kfree(map); | |
1102 | free_extent_map(em); | |
1103 | return -EIO; | |
1104 | } | |
0b86a832 CM |
1105 | } |
1106 | ||
1107 | spin_lock(&map_tree->map_tree.lock); | |
1108 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
1109 | BUG_ON(ret); | |
1110 | spin_unlock(&map_tree->map_tree.lock); | |
1111 | free_extent_map(em); | |
1112 | ||
1113 | return 0; | |
1114 | } | |
1115 | ||
1116 | static int fill_device_from_item(struct extent_buffer *leaf, | |
1117 | struct btrfs_dev_item *dev_item, | |
1118 | struct btrfs_device *device) | |
1119 | { | |
1120 | unsigned long ptr; | |
0b86a832 CM |
1121 | |
1122 | device->devid = btrfs_device_id(leaf, dev_item); | |
1123 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
1124 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
1125 | device->type = btrfs_device_type(leaf, dev_item); | |
1126 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
1127 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
1128 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
0b86a832 CM |
1129 | |
1130 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
1131 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); | |
1132 | ||
0b86a832 CM |
1133 | return 0; |
1134 | } | |
1135 | ||
0d81ba5d | 1136 | static int read_one_dev(struct btrfs_root *root, |
0b86a832 CM |
1137 | struct extent_buffer *leaf, |
1138 | struct btrfs_dev_item *dev_item) | |
1139 | { | |
1140 | struct btrfs_device *device; | |
1141 | u64 devid; | |
1142 | int ret; | |
1143 | ||
1144 | devid = btrfs_device_id(leaf, dev_item); | |
6324fbf3 CM |
1145 | device = btrfs_find_device(root, devid); |
1146 | if (!device) { | |
8a4b83cc | 1147 | printk("warning devid %Lu not found already\n", devid); |
f2984462 | 1148 | device = kzalloc(sizeof(*device), GFP_NOFS); |
6324fbf3 CM |
1149 | if (!device) |
1150 | return -ENOMEM; | |
8a4b83cc CM |
1151 | list_add(&device->dev_list, |
1152 | &root->fs_info->fs_devices->devices); | |
8790d502 CM |
1153 | device->total_ios = 0; |
1154 | spin_lock_init(&device->io_lock); | |
6324fbf3 | 1155 | } |
0b86a832 CM |
1156 | |
1157 | fill_device_from_item(leaf, dev_item, device); | |
1158 | device->dev_root = root->fs_info->dev_root; | |
0b86a832 CM |
1159 | ret = 0; |
1160 | #if 0 | |
1161 | ret = btrfs_open_device(device); | |
1162 | if (ret) { | |
1163 | kfree(device); | |
1164 | } | |
1165 | #endif | |
1166 | return ret; | |
1167 | } | |
1168 | ||
0d81ba5d CM |
1169 | int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) |
1170 | { | |
1171 | struct btrfs_dev_item *dev_item; | |
1172 | ||
1173 | dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, | |
1174 | dev_item); | |
1175 | return read_one_dev(root, buf, dev_item); | |
1176 | } | |
1177 | ||
0b86a832 CM |
1178 | int btrfs_read_sys_array(struct btrfs_root *root) |
1179 | { | |
1180 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1181 | struct extent_buffer *sb = root->fs_info->sb_buffer; | |
1182 | struct btrfs_disk_key *disk_key; | |
0b86a832 CM |
1183 | struct btrfs_chunk *chunk; |
1184 | struct btrfs_key key; | |
1185 | u32 num_stripes; | |
1186 | u32 array_size; | |
1187 | u32 len = 0; | |
1188 | u8 *ptr; | |
1189 | unsigned long sb_ptr; | |
1190 | u32 cur; | |
1191 | int ret; | |
0b86a832 CM |
1192 | |
1193 | array_size = btrfs_super_sys_array_size(super_copy); | |
1194 | ||
1195 | /* | |
1196 | * we do this loop twice, once for the device items and | |
1197 | * once for all of the chunks. This way there are device | |
1198 | * structs filled in for every chunk | |
1199 | */ | |
0b86a832 CM |
1200 | ptr = super_copy->sys_chunk_array; |
1201 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
1202 | cur = 0; | |
1203 | ||
1204 | while (cur < array_size) { | |
1205 | disk_key = (struct btrfs_disk_key *)ptr; | |
1206 | btrfs_disk_key_to_cpu(&key, disk_key); | |
1207 | ||
1208 | len = sizeof(*disk_key); | |
1209 | ptr += len; | |
1210 | sb_ptr += len; | |
1211 | cur += len; | |
1212 | ||
0d81ba5d | 1213 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { |
0b86a832 | 1214 | chunk = (struct btrfs_chunk *)sb_ptr; |
0d81ba5d CM |
1215 | ret = read_one_chunk(root, &key, sb, chunk); |
1216 | BUG_ON(ret); | |
0b86a832 CM |
1217 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); |
1218 | len = btrfs_chunk_item_size(num_stripes); | |
1219 | } else { | |
1220 | BUG(); | |
1221 | } | |
1222 | ptr += len; | |
1223 | sb_ptr += len; | |
1224 | cur += len; | |
1225 | } | |
0b86a832 CM |
1226 | return 0; |
1227 | } | |
1228 | ||
1229 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
1230 | { | |
1231 | struct btrfs_path *path; | |
1232 | struct extent_buffer *leaf; | |
1233 | struct btrfs_key key; | |
1234 | struct btrfs_key found_key; | |
1235 | int ret; | |
1236 | int slot; | |
1237 | ||
1238 | root = root->fs_info->chunk_root; | |
1239 | ||
1240 | path = btrfs_alloc_path(); | |
1241 | if (!path) | |
1242 | return -ENOMEM; | |
1243 | ||
1244 | /* first we search for all of the device items, and then we | |
1245 | * read in all of the chunk items. This way we can create chunk | |
1246 | * mappings that reference all of the devices that are afound | |
1247 | */ | |
1248 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1249 | key.offset = 0; | |
1250 | key.type = 0; | |
1251 | again: | |
1252 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1253 | while(1) { | |
1254 | leaf = path->nodes[0]; | |
1255 | slot = path->slots[0]; | |
1256 | if (slot >= btrfs_header_nritems(leaf)) { | |
1257 | ret = btrfs_next_leaf(root, path); | |
1258 | if (ret == 0) | |
1259 | continue; | |
1260 | if (ret < 0) | |
1261 | goto error; | |
1262 | break; | |
1263 | } | |
1264 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1265 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1266 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
1267 | break; | |
1268 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
1269 | struct btrfs_dev_item *dev_item; | |
1270 | dev_item = btrfs_item_ptr(leaf, slot, | |
1271 | struct btrfs_dev_item); | |
0d81ba5d | 1272 | ret = read_one_dev(root, leaf, dev_item); |
0b86a832 CM |
1273 | BUG_ON(ret); |
1274 | } | |
1275 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
1276 | struct btrfs_chunk *chunk; | |
1277 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
1278 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
1279 | } | |
1280 | path->slots[0]++; | |
1281 | } | |
1282 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1283 | key.objectid = 0; | |
1284 | btrfs_release_path(root, path); | |
1285 | goto again; | |
1286 | } | |
1287 | ||
1288 | btrfs_free_path(path); | |
1289 | ret = 0; | |
1290 | error: | |
1291 | return ret; | |
1292 | } | |
1293 |