]>
Commit | Line | Data |
---|---|---|
0b86a832 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | #include <linux/sched.h> | |
19 | #include <linux/bio.h> | |
8a4b83cc | 20 | #include <linux/buffer_head.h> |
f2d8d74d | 21 | #include <linux/blkdev.h> |
788f20eb | 22 | #include <linux/random.h> |
593060d7 | 23 | #include <asm/div64.h> |
0b86a832 CM |
24 | #include "ctree.h" |
25 | #include "extent_map.h" | |
26 | #include "disk-io.h" | |
27 | #include "transaction.h" | |
28 | #include "print-tree.h" | |
29 | #include "volumes.h" | |
8b712842 | 30 | #include "async-thread.h" |
0b86a832 | 31 | |
593060d7 CM |
32 | struct map_lookup { |
33 | u64 type; | |
34 | int io_align; | |
35 | int io_width; | |
36 | int stripe_len; | |
37 | int sector_size; | |
38 | int num_stripes; | |
321aecc6 | 39 | int sub_stripes; |
cea9e445 | 40 | struct btrfs_bio_stripe stripes[]; |
593060d7 CM |
41 | }; |
42 | ||
2b82032c YZ |
43 | static int init_first_rw_device(struct btrfs_trans_handle *trans, |
44 | struct btrfs_root *root, | |
45 | struct btrfs_device *device); | |
46 | static int btrfs_relocate_sys_chunks(struct btrfs_root *root); | |
47 | ||
48 | ||
593060d7 | 49 | #define map_lookup_size(n) (sizeof(struct map_lookup) + \ |
cea9e445 | 50 | (sizeof(struct btrfs_bio_stripe) * (n))) |
593060d7 | 51 | |
8a4b83cc CM |
52 | static DEFINE_MUTEX(uuid_mutex); |
53 | static LIST_HEAD(fs_uuids); | |
54 | ||
a061fc8d CM |
55 | void btrfs_lock_volumes(void) |
56 | { | |
57 | mutex_lock(&uuid_mutex); | |
58 | } | |
59 | ||
60 | void btrfs_unlock_volumes(void) | |
61 | { | |
62 | mutex_unlock(&uuid_mutex); | |
63 | } | |
64 | ||
7d9eb12c CM |
65 | static void lock_chunks(struct btrfs_root *root) |
66 | { | |
7d9eb12c CM |
67 | mutex_lock(&root->fs_info->chunk_mutex); |
68 | } | |
69 | ||
70 | static void unlock_chunks(struct btrfs_root *root) | |
71 | { | |
7d9eb12c CM |
72 | mutex_unlock(&root->fs_info->chunk_mutex); |
73 | } | |
74 | ||
8a4b83cc CM |
75 | int btrfs_cleanup_fs_uuids(void) |
76 | { | |
77 | struct btrfs_fs_devices *fs_devices; | |
8a4b83cc CM |
78 | struct btrfs_device *dev; |
79 | ||
2b82032c YZ |
80 | while (!list_empty(&fs_uuids)) { |
81 | fs_devices = list_entry(fs_uuids.next, | |
82 | struct btrfs_fs_devices, list); | |
83 | list_del(&fs_devices->list); | |
8a4b83cc | 84 | while(!list_empty(&fs_devices->devices)) { |
2b82032c YZ |
85 | dev = list_entry(fs_devices->devices.next, |
86 | struct btrfs_device, dev_list); | |
8a4b83cc | 87 | if (dev->bdev) { |
15916de8 | 88 | close_bdev_exclusive(dev->bdev, dev->mode); |
a0af469b | 89 | fs_devices->open_devices--; |
8a4b83cc | 90 | } |
2b82032c YZ |
91 | fs_devices->num_devices--; |
92 | if (dev->writeable) | |
93 | fs_devices->rw_devices--; | |
8a4b83cc | 94 | list_del(&dev->dev_list); |
2b82032c | 95 | list_del(&dev->dev_alloc_list); |
dfe25020 | 96 | kfree(dev->name); |
8a4b83cc CM |
97 | kfree(dev); |
98 | } | |
2b82032c YZ |
99 | WARN_ON(fs_devices->num_devices); |
100 | WARN_ON(fs_devices->open_devices); | |
101 | WARN_ON(fs_devices->rw_devices); | |
102 | kfree(fs_devices); | |
8a4b83cc CM |
103 | } |
104 | return 0; | |
105 | } | |
106 | ||
a1b32a59 CM |
107 | static noinline struct btrfs_device *__find_device(struct list_head *head, |
108 | u64 devid, u8 *uuid) | |
8a4b83cc CM |
109 | { |
110 | struct btrfs_device *dev; | |
111 | struct list_head *cur; | |
112 | ||
113 | list_for_each(cur, head) { | |
114 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
a443755f | 115 | if (dev->devid == devid && |
8f18cf13 | 116 | (!uuid || !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE))) { |
8a4b83cc | 117 | return dev; |
a443755f | 118 | } |
8a4b83cc CM |
119 | } |
120 | return NULL; | |
121 | } | |
122 | ||
a1b32a59 | 123 | static noinline struct btrfs_fs_devices *find_fsid(u8 *fsid) |
8a4b83cc CM |
124 | { |
125 | struct list_head *cur; | |
126 | struct btrfs_fs_devices *fs_devices; | |
127 | ||
128 | list_for_each(cur, &fs_uuids) { | |
129 | fs_devices = list_entry(cur, struct btrfs_fs_devices, list); | |
130 | if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0) | |
131 | return fs_devices; | |
132 | } | |
133 | return NULL; | |
134 | } | |
135 | ||
8b712842 CM |
136 | /* |
137 | * we try to collect pending bios for a device so we don't get a large | |
138 | * number of procs sending bios down to the same device. This greatly | |
139 | * improves the schedulers ability to collect and merge the bios. | |
140 | * | |
141 | * But, it also turns into a long list of bios to process and that is sure | |
142 | * to eventually make the worker thread block. The solution here is to | |
143 | * make some progress and then put this work struct back at the end of | |
144 | * the list if the block device is congested. This way, multiple devices | |
145 | * can make progress from a single worker thread. | |
146 | */ | |
a1b32a59 | 147 | static int noinline run_scheduled_bios(struct btrfs_device *device) |
8b712842 CM |
148 | { |
149 | struct bio *pending; | |
150 | struct backing_dev_info *bdi; | |
b64a2851 | 151 | struct btrfs_fs_info *fs_info; |
8b712842 CM |
152 | struct bio *tail; |
153 | struct bio *cur; | |
154 | int again = 0; | |
155 | unsigned long num_run = 0; | |
b64a2851 | 156 | unsigned long limit; |
8b712842 CM |
157 | |
158 | bdi = device->bdev->bd_inode->i_mapping->backing_dev_info; | |
b64a2851 CM |
159 | fs_info = device->dev_root->fs_info; |
160 | limit = btrfs_async_submit_limit(fs_info); | |
161 | limit = limit * 2 / 3; | |
162 | ||
8b712842 CM |
163 | loop: |
164 | spin_lock(&device->io_lock); | |
165 | ||
166 | /* take all the bios off the list at once and process them | |
167 | * later on (without the lock held). But, remember the | |
168 | * tail and other pointers so the bios can be properly reinserted | |
169 | * into the list if we hit congestion | |
170 | */ | |
171 | pending = device->pending_bios; | |
172 | tail = device->pending_bio_tail; | |
173 | WARN_ON(pending && !tail); | |
174 | device->pending_bios = NULL; | |
175 | device->pending_bio_tail = NULL; | |
176 | ||
177 | /* | |
178 | * if pending was null this time around, no bios need processing | |
179 | * at all and we can stop. Otherwise it'll loop back up again | |
180 | * and do an additional check so no bios are missed. | |
181 | * | |
182 | * device->running_pending is used to synchronize with the | |
183 | * schedule_bio code. | |
184 | */ | |
185 | if (pending) { | |
186 | again = 1; | |
187 | device->running_pending = 1; | |
188 | } else { | |
189 | again = 0; | |
190 | device->running_pending = 0; | |
191 | } | |
192 | spin_unlock(&device->io_lock); | |
193 | ||
194 | while(pending) { | |
195 | cur = pending; | |
196 | pending = pending->bi_next; | |
197 | cur->bi_next = NULL; | |
b64a2851 CM |
198 | atomic_dec(&fs_info->nr_async_bios); |
199 | ||
200 | if (atomic_read(&fs_info->nr_async_bios) < limit && | |
201 | waitqueue_active(&fs_info->async_submit_wait)) | |
202 | wake_up(&fs_info->async_submit_wait); | |
492bb6de CM |
203 | |
204 | BUG_ON(atomic_read(&cur->bi_cnt) == 0); | |
205 | bio_get(cur); | |
8b712842 | 206 | submit_bio(cur->bi_rw, cur); |
492bb6de | 207 | bio_put(cur); |
8b712842 CM |
208 | num_run++; |
209 | ||
210 | /* | |
211 | * we made progress, there is more work to do and the bdi | |
212 | * is now congested. Back off and let other work structs | |
213 | * run instead | |
214 | */ | |
5f2cc086 CM |
215 | if (pending && bdi_write_congested(bdi) && |
216 | fs_info->fs_devices->open_devices > 1) { | |
8b712842 CM |
217 | struct bio *old_head; |
218 | ||
219 | spin_lock(&device->io_lock); | |
492bb6de | 220 | |
8b712842 CM |
221 | old_head = device->pending_bios; |
222 | device->pending_bios = pending; | |
223 | if (device->pending_bio_tail) | |
224 | tail->bi_next = old_head; | |
225 | else | |
226 | device->pending_bio_tail = tail; | |
227 | ||
228 | spin_unlock(&device->io_lock); | |
229 | btrfs_requeue_work(&device->work); | |
230 | goto done; | |
231 | } | |
232 | } | |
233 | if (again) | |
234 | goto loop; | |
235 | done: | |
236 | return 0; | |
237 | } | |
238 | ||
239 | void pending_bios_fn(struct btrfs_work *work) | |
240 | { | |
241 | struct btrfs_device *device; | |
242 | ||
243 | device = container_of(work, struct btrfs_device, work); | |
244 | run_scheduled_bios(device); | |
245 | } | |
246 | ||
a1b32a59 | 247 | static noinline int device_list_add(const char *path, |
8a4b83cc CM |
248 | struct btrfs_super_block *disk_super, |
249 | u64 devid, struct btrfs_fs_devices **fs_devices_ret) | |
250 | { | |
251 | struct btrfs_device *device; | |
252 | struct btrfs_fs_devices *fs_devices; | |
253 | u64 found_transid = btrfs_super_generation(disk_super); | |
254 | ||
255 | fs_devices = find_fsid(disk_super->fsid); | |
256 | if (!fs_devices) { | |
515dc322 | 257 | fs_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS); |
8a4b83cc CM |
258 | if (!fs_devices) |
259 | return -ENOMEM; | |
260 | INIT_LIST_HEAD(&fs_devices->devices); | |
b3075717 | 261 | INIT_LIST_HEAD(&fs_devices->alloc_list); |
8a4b83cc CM |
262 | list_add(&fs_devices->list, &fs_uuids); |
263 | memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE); | |
264 | fs_devices->latest_devid = devid; | |
265 | fs_devices->latest_trans = found_transid; | |
8a4b83cc CM |
266 | device = NULL; |
267 | } else { | |
a443755f CM |
268 | device = __find_device(&fs_devices->devices, devid, |
269 | disk_super->dev_item.uuid); | |
8a4b83cc CM |
270 | } |
271 | if (!device) { | |
2b82032c YZ |
272 | if (fs_devices->opened) |
273 | return -EBUSY; | |
274 | ||
8a4b83cc CM |
275 | device = kzalloc(sizeof(*device), GFP_NOFS); |
276 | if (!device) { | |
277 | /* we can safely leave the fs_devices entry around */ | |
278 | return -ENOMEM; | |
279 | } | |
280 | device->devid = devid; | |
8b712842 | 281 | device->work.func = pending_bios_fn; |
a443755f CM |
282 | memcpy(device->uuid, disk_super->dev_item.uuid, |
283 | BTRFS_UUID_SIZE); | |
f2984462 | 284 | device->barriers = 1; |
b248a415 | 285 | spin_lock_init(&device->io_lock); |
8a4b83cc CM |
286 | device->name = kstrdup(path, GFP_NOFS); |
287 | if (!device->name) { | |
288 | kfree(device); | |
289 | return -ENOMEM; | |
290 | } | |
2b82032c | 291 | INIT_LIST_HEAD(&device->dev_alloc_list); |
8a4b83cc | 292 | list_add(&device->dev_list, &fs_devices->devices); |
2b82032c | 293 | device->fs_devices = fs_devices; |
8a4b83cc CM |
294 | fs_devices->num_devices++; |
295 | } | |
296 | ||
297 | if (found_transid > fs_devices->latest_trans) { | |
298 | fs_devices->latest_devid = devid; | |
299 | fs_devices->latest_trans = found_transid; | |
300 | } | |
8a4b83cc CM |
301 | *fs_devices_ret = fs_devices; |
302 | return 0; | |
303 | } | |
304 | ||
dfe25020 CM |
305 | int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices) |
306 | { | |
2b82032c | 307 | struct list_head *tmp; |
dfe25020 CM |
308 | struct list_head *cur; |
309 | struct btrfs_device *device; | |
2b82032c | 310 | int seed_devices = 0; |
dfe25020 CM |
311 | |
312 | mutex_lock(&uuid_mutex); | |
313 | again: | |
2b82032c | 314 | list_for_each_safe(cur, tmp, &fs_devices->devices) { |
dfe25020 | 315 | device = list_entry(cur, struct btrfs_device, dev_list); |
2b82032c YZ |
316 | if (device->in_fs_metadata) |
317 | continue; | |
318 | ||
319 | if (device->bdev) { | |
15916de8 | 320 | close_bdev_exclusive(device->bdev, device->mode); |
2b82032c YZ |
321 | device->bdev = NULL; |
322 | fs_devices->open_devices--; | |
323 | } | |
324 | if (device->writeable) { | |
325 | list_del_init(&device->dev_alloc_list); | |
326 | device->writeable = 0; | |
327 | fs_devices->rw_devices--; | |
328 | } | |
329 | if (!seed_devices) { | |
330 | list_del_init(&device->dev_list); | |
dfe25020 CM |
331 | fs_devices->num_devices--; |
332 | kfree(device->name); | |
333 | kfree(device); | |
dfe25020 CM |
334 | } |
335 | } | |
2b82032c YZ |
336 | |
337 | if (fs_devices->seed) { | |
338 | fs_devices = fs_devices->seed; | |
339 | seed_devices = 1; | |
340 | goto again; | |
341 | } | |
342 | ||
dfe25020 CM |
343 | mutex_unlock(&uuid_mutex); |
344 | return 0; | |
345 | } | |
a0af469b | 346 | |
2b82032c | 347 | static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices) |
8a4b83cc | 348 | { |
2b82032c | 349 | struct btrfs_fs_devices *seed_devices; |
8a4b83cc CM |
350 | struct list_head *cur; |
351 | struct btrfs_device *device; | |
2b82032c YZ |
352 | again: |
353 | if (--fs_devices->opened > 0) | |
354 | return 0; | |
8a4b83cc | 355 | |
2b82032c | 356 | list_for_each(cur, &fs_devices->devices) { |
8a4b83cc CM |
357 | device = list_entry(cur, struct btrfs_device, dev_list); |
358 | if (device->bdev) { | |
15916de8 | 359 | close_bdev_exclusive(device->bdev, device->mode); |
a0af469b | 360 | fs_devices->open_devices--; |
8a4b83cc | 361 | } |
2b82032c YZ |
362 | if (device->writeable) { |
363 | list_del_init(&device->dev_alloc_list); | |
364 | fs_devices->rw_devices--; | |
365 | } | |
366 | ||
8a4b83cc | 367 | device->bdev = NULL; |
2b82032c | 368 | device->writeable = 0; |
dfe25020 | 369 | device->in_fs_metadata = 0; |
8a4b83cc | 370 | } |
2b82032c YZ |
371 | fs_devices->opened = 0; |
372 | fs_devices->seeding = 0; | |
373 | fs_devices->sprouted = 0; | |
374 | ||
375 | seed_devices = fs_devices->seed; | |
376 | fs_devices->seed = NULL; | |
377 | if (seed_devices) { | |
378 | fs_devices = seed_devices; | |
379 | goto again; | |
380 | } | |
8a4b83cc CM |
381 | return 0; |
382 | } | |
383 | ||
2b82032c YZ |
384 | int btrfs_close_devices(struct btrfs_fs_devices *fs_devices) |
385 | { | |
386 | int ret; | |
387 | ||
388 | mutex_lock(&uuid_mutex); | |
389 | ret = __btrfs_close_devices(fs_devices); | |
390 | mutex_unlock(&uuid_mutex); | |
391 | return ret; | |
392 | } | |
393 | ||
15916de8 CM |
394 | int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices, |
395 | int flags, void *holder) | |
8a4b83cc CM |
396 | { |
397 | struct block_device *bdev; | |
398 | struct list_head *head = &fs_devices->devices; | |
399 | struct list_head *cur; | |
400 | struct btrfs_device *device; | |
a0af469b CM |
401 | struct block_device *latest_bdev = NULL; |
402 | struct buffer_head *bh; | |
403 | struct btrfs_super_block *disk_super; | |
404 | u64 latest_devid = 0; | |
405 | u64 latest_transid = 0; | |
a0af469b | 406 | u64 devid; |
2b82032c | 407 | int seeding = 1; |
a0af469b | 408 | int ret = 0; |
8a4b83cc | 409 | |
8a4b83cc CM |
410 | list_for_each(cur, head) { |
411 | device = list_entry(cur, struct btrfs_device, dev_list); | |
c1c4d91c CM |
412 | if (device->bdev) |
413 | continue; | |
dfe25020 CM |
414 | if (!device->name) |
415 | continue; | |
416 | ||
15916de8 | 417 | bdev = open_bdev_exclusive(device->name, flags, holder); |
8a4b83cc CM |
418 | if (IS_ERR(bdev)) { |
419 | printk("open %s failed\n", device->name); | |
a0af469b | 420 | goto error; |
8a4b83cc | 421 | } |
a061fc8d | 422 | set_blocksize(bdev, 4096); |
a0af469b CM |
423 | |
424 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
425 | if (!bh) | |
426 | goto error_close; | |
427 | ||
428 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
429 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
430 | sizeof(disk_super->magic))) | |
431 | goto error_brelse; | |
432 | ||
433 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
434 | if (devid != device->devid) | |
435 | goto error_brelse; | |
436 | ||
2b82032c YZ |
437 | if (memcmp(device->uuid, disk_super->dev_item.uuid, |
438 | BTRFS_UUID_SIZE)) | |
439 | goto error_brelse; | |
440 | ||
441 | device->generation = btrfs_super_generation(disk_super); | |
442 | if (!latest_transid || device->generation > latest_transid) { | |
a0af469b | 443 | latest_devid = devid; |
2b82032c | 444 | latest_transid = device->generation; |
a0af469b CM |
445 | latest_bdev = bdev; |
446 | } | |
447 | ||
2b82032c YZ |
448 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_SEEDING) { |
449 | device->writeable = 0; | |
450 | } else { | |
451 | device->writeable = !bdev_read_only(bdev); | |
452 | seeding = 0; | |
453 | } | |
454 | ||
8a4b83cc | 455 | device->bdev = bdev; |
dfe25020 | 456 | device->in_fs_metadata = 0; |
15916de8 CM |
457 | device->mode = flags; |
458 | ||
a0af469b | 459 | fs_devices->open_devices++; |
2b82032c YZ |
460 | if (device->writeable) { |
461 | fs_devices->rw_devices++; | |
462 | list_add(&device->dev_alloc_list, | |
463 | &fs_devices->alloc_list); | |
464 | } | |
a0af469b | 465 | continue; |
a061fc8d | 466 | |
a0af469b CM |
467 | error_brelse: |
468 | brelse(bh); | |
469 | error_close: | |
15916de8 | 470 | close_bdev_exclusive(bdev, MS_RDONLY); |
a0af469b CM |
471 | error: |
472 | continue; | |
8a4b83cc | 473 | } |
a0af469b CM |
474 | if (fs_devices->open_devices == 0) { |
475 | ret = -EIO; | |
476 | goto out; | |
477 | } | |
2b82032c YZ |
478 | fs_devices->seeding = seeding; |
479 | fs_devices->opened = 1; | |
a0af469b CM |
480 | fs_devices->latest_bdev = latest_bdev; |
481 | fs_devices->latest_devid = latest_devid; | |
482 | fs_devices->latest_trans = latest_transid; | |
2b82032c | 483 | fs_devices->total_rw_bytes = 0; |
a0af469b | 484 | out: |
2b82032c YZ |
485 | return ret; |
486 | } | |
487 | ||
488 | int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, | |
489 | int flags, void *holder) | |
490 | { | |
491 | int ret; | |
492 | ||
493 | mutex_lock(&uuid_mutex); | |
494 | if (fs_devices->opened) { | |
495 | if (fs_devices->sprouted) { | |
496 | ret = -EBUSY; | |
497 | } else { | |
498 | fs_devices->opened++; | |
499 | ret = 0; | |
500 | } | |
501 | } else { | |
15916de8 | 502 | ret = __btrfs_open_devices(fs_devices, flags, holder); |
2b82032c | 503 | } |
8a4b83cc | 504 | mutex_unlock(&uuid_mutex); |
8a4b83cc CM |
505 | return ret; |
506 | } | |
507 | ||
508 | int btrfs_scan_one_device(const char *path, int flags, void *holder, | |
509 | struct btrfs_fs_devices **fs_devices_ret) | |
510 | { | |
511 | struct btrfs_super_block *disk_super; | |
512 | struct block_device *bdev; | |
513 | struct buffer_head *bh; | |
514 | int ret; | |
515 | u64 devid; | |
f2984462 | 516 | u64 transid; |
8a4b83cc CM |
517 | |
518 | mutex_lock(&uuid_mutex); | |
519 | ||
15916de8 | 520 | bdev = open_bdev_exclusive(path, flags, holder); |
8a4b83cc CM |
521 | |
522 | if (IS_ERR(bdev)) { | |
8a4b83cc CM |
523 | ret = PTR_ERR(bdev); |
524 | goto error; | |
525 | } | |
526 | ||
527 | ret = set_blocksize(bdev, 4096); | |
528 | if (ret) | |
529 | goto error_close; | |
530 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
531 | if (!bh) { | |
532 | ret = -EIO; | |
533 | goto error_close; | |
534 | } | |
535 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
536 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
537 | sizeof(disk_super->magic))) { | |
e58ca020 | 538 | ret = -EINVAL; |
8a4b83cc CM |
539 | goto error_brelse; |
540 | } | |
541 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
f2984462 | 542 | transid = btrfs_super_generation(disk_super); |
7ae9c09d CM |
543 | if (disk_super->label[0]) |
544 | printk("device label %s ", disk_super->label); | |
545 | else { | |
546 | /* FIXME, make a readl uuid parser */ | |
547 | printk("device fsid %llx-%llx ", | |
548 | *(unsigned long long *)disk_super->fsid, | |
549 | *(unsigned long long *)(disk_super->fsid + 8)); | |
550 | } | |
551 | printk("devid %Lu transid %Lu %s\n", devid, transid, path); | |
8a4b83cc CM |
552 | ret = device_list_add(path, disk_super, devid, fs_devices_ret); |
553 | ||
554 | error_brelse: | |
555 | brelse(bh); | |
556 | error_close: | |
15916de8 | 557 | close_bdev_exclusive(bdev, flags); |
8a4b83cc CM |
558 | error: |
559 | mutex_unlock(&uuid_mutex); | |
560 | return ret; | |
561 | } | |
0b86a832 CM |
562 | |
563 | /* | |
564 | * this uses a pretty simple search, the expectation is that it is | |
565 | * called very infrequently and that a given device has a small number | |
566 | * of extents | |
567 | */ | |
a1b32a59 CM |
568 | static noinline int find_free_dev_extent(struct btrfs_trans_handle *trans, |
569 | struct btrfs_device *device, | |
a1b32a59 | 570 | u64 num_bytes, u64 *start) |
0b86a832 CM |
571 | { |
572 | struct btrfs_key key; | |
573 | struct btrfs_root *root = device->dev_root; | |
574 | struct btrfs_dev_extent *dev_extent = NULL; | |
2b82032c | 575 | struct btrfs_path *path; |
0b86a832 CM |
576 | u64 hole_size = 0; |
577 | u64 last_byte = 0; | |
578 | u64 search_start = 0; | |
579 | u64 search_end = device->total_bytes; | |
580 | int ret; | |
581 | int slot = 0; | |
582 | int start_found; | |
583 | struct extent_buffer *l; | |
584 | ||
2b82032c YZ |
585 | path = btrfs_alloc_path(); |
586 | if (!path) | |
587 | return -ENOMEM; | |
0b86a832 | 588 | path->reada = 2; |
2b82032c | 589 | start_found = 0; |
0b86a832 CM |
590 | |
591 | /* FIXME use last free of some kind */ | |
592 | ||
8a4b83cc CM |
593 | /* we don't want to overwrite the superblock on the drive, |
594 | * so we make sure to start at an offset of at least 1MB | |
595 | */ | |
596 | search_start = max((u64)1024 * 1024, search_start); | |
8f18cf13 CM |
597 | |
598 | if (root->fs_info->alloc_start + num_bytes <= device->total_bytes) | |
599 | search_start = max(root->fs_info->alloc_start, search_start); | |
600 | ||
0b86a832 CM |
601 | key.objectid = device->devid; |
602 | key.offset = search_start; | |
603 | key.type = BTRFS_DEV_EXTENT_KEY; | |
604 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
605 | if (ret < 0) | |
606 | goto error; | |
607 | ret = btrfs_previous_item(root, path, 0, key.type); | |
608 | if (ret < 0) | |
609 | goto error; | |
610 | l = path->nodes[0]; | |
611 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
612 | while (1) { | |
613 | l = path->nodes[0]; | |
614 | slot = path->slots[0]; | |
615 | if (slot >= btrfs_header_nritems(l)) { | |
616 | ret = btrfs_next_leaf(root, path); | |
617 | if (ret == 0) | |
618 | continue; | |
619 | if (ret < 0) | |
620 | goto error; | |
621 | no_more_items: | |
622 | if (!start_found) { | |
623 | if (search_start >= search_end) { | |
624 | ret = -ENOSPC; | |
625 | goto error; | |
626 | } | |
627 | *start = search_start; | |
628 | start_found = 1; | |
629 | goto check_pending; | |
630 | } | |
631 | *start = last_byte > search_start ? | |
632 | last_byte : search_start; | |
633 | if (search_end <= *start) { | |
634 | ret = -ENOSPC; | |
635 | goto error; | |
636 | } | |
637 | goto check_pending; | |
638 | } | |
639 | btrfs_item_key_to_cpu(l, &key, slot); | |
640 | ||
641 | if (key.objectid < device->devid) | |
642 | goto next; | |
643 | ||
644 | if (key.objectid > device->devid) | |
645 | goto no_more_items; | |
646 | ||
647 | if (key.offset >= search_start && key.offset > last_byte && | |
648 | start_found) { | |
649 | if (last_byte < search_start) | |
650 | last_byte = search_start; | |
651 | hole_size = key.offset - last_byte; | |
652 | if (key.offset > last_byte && | |
653 | hole_size >= num_bytes) { | |
654 | *start = last_byte; | |
655 | goto check_pending; | |
656 | } | |
657 | } | |
658 | if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { | |
659 | goto next; | |
660 | } | |
661 | ||
662 | start_found = 1; | |
663 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
664 | last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); | |
665 | next: | |
666 | path->slots[0]++; | |
667 | cond_resched(); | |
668 | } | |
669 | check_pending: | |
670 | /* we have to make sure we didn't find an extent that has already | |
671 | * been allocated by the map tree or the original allocation | |
672 | */ | |
0b86a832 CM |
673 | BUG_ON(*start < search_start); |
674 | ||
6324fbf3 | 675 | if (*start + num_bytes > search_end) { |
0b86a832 CM |
676 | ret = -ENOSPC; |
677 | goto error; | |
678 | } | |
679 | /* check for pending inserts here */ | |
2b82032c | 680 | ret = 0; |
0b86a832 CM |
681 | |
682 | error: | |
2b82032c | 683 | btrfs_free_path(path); |
0b86a832 CM |
684 | return ret; |
685 | } | |
686 | ||
8f18cf13 CM |
687 | int btrfs_free_dev_extent(struct btrfs_trans_handle *trans, |
688 | struct btrfs_device *device, | |
689 | u64 start) | |
690 | { | |
691 | int ret; | |
692 | struct btrfs_path *path; | |
693 | struct btrfs_root *root = device->dev_root; | |
694 | struct btrfs_key key; | |
a061fc8d CM |
695 | struct btrfs_key found_key; |
696 | struct extent_buffer *leaf = NULL; | |
697 | struct btrfs_dev_extent *extent = NULL; | |
8f18cf13 CM |
698 | |
699 | path = btrfs_alloc_path(); | |
700 | if (!path) | |
701 | return -ENOMEM; | |
702 | ||
703 | key.objectid = device->devid; | |
704 | key.offset = start; | |
705 | key.type = BTRFS_DEV_EXTENT_KEY; | |
706 | ||
707 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); | |
a061fc8d CM |
708 | if (ret > 0) { |
709 | ret = btrfs_previous_item(root, path, key.objectid, | |
710 | BTRFS_DEV_EXTENT_KEY); | |
711 | BUG_ON(ret); | |
712 | leaf = path->nodes[0]; | |
713 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
714 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
715 | struct btrfs_dev_extent); | |
716 | BUG_ON(found_key.offset > start || found_key.offset + | |
717 | btrfs_dev_extent_length(leaf, extent) < start); | |
718 | ret = 0; | |
719 | } else if (ret == 0) { | |
720 | leaf = path->nodes[0]; | |
721 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
722 | struct btrfs_dev_extent); | |
723 | } | |
8f18cf13 CM |
724 | BUG_ON(ret); |
725 | ||
dfe25020 CM |
726 | if (device->bytes_used > 0) |
727 | device->bytes_used -= btrfs_dev_extent_length(leaf, extent); | |
8f18cf13 CM |
728 | ret = btrfs_del_item(trans, root, path); |
729 | BUG_ON(ret); | |
730 | ||
731 | btrfs_free_path(path); | |
732 | return ret; | |
733 | } | |
734 | ||
2b82032c | 735 | int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, |
0b86a832 | 736 | struct btrfs_device *device, |
e17cade2 | 737 | u64 chunk_tree, u64 chunk_objectid, |
2b82032c | 738 | u64 chunk_offset, u64 start, u64 num_bytes) |
0b86a832 CM |
739 | { |
740 | int ret; | |
741 | struct btrfs_path *path; | |
742 | struct btrfs_root *root = device->dev_root; | |
743 | struct btrfs_dev_extent *extent; | |
744 | struct extent_buffer *leaf; | |
745 | struct btrfs_key key; | |
746 | ||
dfe25020 | 747 | WARN_ON(!device->in_fs_metadata); |
0b86a832 CM |
748 | path = btrfs_alloc_path(); |
749 | if (!path) | |
750 | return -ENOMEM; | |
751 | ||
0b86a832 | 752 | key.objectid = device->devid; |
2b82032c | 753 | key.offset = start; |
0b86a832 CM |
754 | key.type = BTRFS_DEV_EXTENT_KEY; |
755 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
756 | sizeof(*extent)); | |
757 | BUG_ON(ret); | |
758 | ||
759 | leaf = path->nodes[0]; | |
760 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
761 | struct btrfs_dev_extent); | |
e17cade2 CM |
762 | btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree); |
763 | btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid); | |
764 | btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset); | |
765 | ||
766 | write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid, | |
767 | (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent), | |
768 | BTRFS_UUID_SIZE); | |
769 | ||
0b86a832 CM |
770 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); |
771 | btrfs_mark_buffer_dirty(leaf); | |
0b86a832 CM |
772 | btrfs_free_path(path); |
773 | return ret; | |
774 | } | |
775 | ||
a1b32a59 CM |
776 | static noinline int find_next_chunk(struct btrfs_root *root, |
777 | u64 objectid, u64 *offset) | |
0b86a832 CM |
778 | { |
779 | struct btrfs_path *path; | |
780 | int ret; | |
781 | struct btrfs_key key; | |
e17cade2 | 782 | struct btrfs_chunk *chunk; |
0b86a832 CM |
783 | struct btrfs_key found_key; |
784 | ||
785 | path = btrfs_alloc_path(); | |
786 | BUG_ON(!path); | |
787 | ||
e17cade2 | 788 | key.objectid = objectid; |
0b86a832 CM |
789 | key.offset = (u64)-1; |
790 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
791 | ||
792 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
793 | if (ret < 0) | |
794 | goto error; | |
795 | ||
796 | BUG_ON(ret == 0); | |
797 | ||
798 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
799 | if (ret) { | |
e17cade2 | 800 | *offset = 0; |
0b86a832 CM |
801 | } else { |
802 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
803 | path->slots[0]); | |
e17cade2 CM |
804 | if (found_key.objectid != objectid) |
805 | *offset = 0; | |
806 | else { | |
807 | chunk = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
808 | struct btrfs_chunk); | |
809 | *offset = found_key.offset + | |
810 | btrfs_chunk_length(path->nodes[0], chunk); | |
811 | } | |
0b86a832 CM |
812 | } |
813 | ret = 0; | |
814 | error: | |
815 | btrfs_free_path(path); | |
816 | return ret; | |
817 | } | |
818 | ||
2b82032c | 819 | static noinline int find_next_devid(struct btrfs_root *root, u64 *objectid) |
0b86a832 CM |
820 | { |
821 | int ret; | |
822 | struct btrfs_key key; | |
823 | struct btrfs_key found_key; | |
2b82032c YZ |
824 | struct btrfs_path *path; |
825 | ||
826 | root = root->fs_info->chunk_root; | |
827 | ||
828 | path = btrfs_alloc_path(); | |
829 | if (!path) | |
830 | return -ENOMEM; | |
0b86a832 CM |
831 | |
832 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
833 | key.type = BTRFS_DEV_ITEM_KEY; | |
834 | key.offset = (u64)-1; | |
835 | ||
836 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
837 | if (ret < 0) | |
838 | goto error; | |
839 | ||
840 | BUG_ON(ret == 0); | |
841 | ||
842 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
843 | BTRFS_DEV_ITEM_KEY); | |
844 | if (ret) { | |
845 | *objectid = 1; | |
846 | } else { | |
847 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
848 | path->slots[0]); | |
849 | *objectid = found_key.offset + 1; | |
850 | } | |
851 | ret = 0; | |
852 | error: | |
2b82032c | 853 | btrfs_free_path(path); |
0b86a832 CM |
854 | return ret; |
855 | } | |
856 | ||
857 | /* | |
858 | * the device information is stored in the chunk root | |
859 | * the btrfs_device struct should be fully filled in | |
860 | */ | |
861 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
862 | struct btrfs_root *root, | |
863 | struct btrfs_device *device) | |
864 | { | |
865 | int ret; | |
866 | struct btrfs_path *path; | |
867 | struct btrfs_dev_item *dev_item; | |
868 | struct extent_buffer *leaf; | |
869 | struct btrfs_key key; | |
870 | unsigned long ptr; | |
0b86a832 CM |
871 | |
872 | root = root->fs_info->chunk_root; | |
873 | ||
874 | path = btrfs_alloc_path(); | |
875 | if (!path) | |
876 | return -ENOMEM; | |
877 | ||
0b86a832 CM |
878 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; |
879 | key.type = BTRFS_DEV_ITEM_KEY; | |
2b82032c | 880 | key.offset = device->devid; |
0b86a832 CM |
881 | |
882 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
0d81ba5d | 883 | sizeof(*dev_item)); |
0b86a832 CM |
884 | if (ret) |
885 | goto out; | |
886 | ||
887 | leaf = path->nodes[0]; | |
888 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
889 | ||
890 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
2b82032c | 891 | btrfs_set_device_generation(leaf, dev_item, 0); |
0b86a832 CM |
892 | btrfs_set_device_type(leaf, dev_item, device->type); |
893 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
894 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
895 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
896 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
897 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
e17cade2 CM |
898 | btrfs_set_device_group(leaf, dev_item, 0); |
899 | btrfs_set_device_seek_speed(leaf, dev_item, 0); | |
900 | btrfs_set_device_bandwidth(leaf, dev_item, 0); | |
0b86a832 | 901 | |
0b86a832 | 902 | ptr = (unsigned long)btrfs_device_uuid(dev_item); |
e17cade2 | 903 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
2b82032c YZ |
904 | ptr = (unsigned long)btrfs_device_fsid(dev_item); |
905 | write_extent_buffer(leaf, root->fs_info->fsid, ptr, BTRFS_UUID_SIZE); | |
0b86a832 | 906 | btrfs_mark_buffer_dirty(leaf); |
0b86a832 | 907 | |
2b82032c | 908 | ret = 0; |
0b86a832 CM |
909 | out: |
910 | btrfs_free_path(path); | |
911 | return ret; | |
912 | } | |
8f18cf13 | 913 | |
a061fc8d CM |
914 | static int btrfs_rm_dev_item(struct btrfs_root *root, |
915 | struct btrfs_device *device) | |
916 | { | |
917 | int ret; | |
918 | struct btrfs_path *path; | |
a061fc8d | 919 | struct btrfs_key key; |
a061fc8d CM |
920 | struct btrfs_trans_handle *trans; |
921 | ||
922 | root = root->fs_info->chunk_root; | |
923 | ||
924 | path = btrfs_alloc_path(); | |
925 | if (!path) | |
926 | return -ENOMEM; | |
927 | ||
928 | trans = btrfs_start_transaction(root, 1); | |
929 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
930 | key.type = BTRFS_DEV_ITEM_KEY; | |
931 | key.offset = device->devid; | |
7d9eb12c | 932 | lock_chunks(root); |
a061fc8d CM |
933 | |
934 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); | |
935 | if (ret < 0) | |
936 | goto out; | |
937 | ||
938 | if (ret > 0) { | |
939 | ret = -ENOENT; | |
940 | goto out; | |
941 | } | |
942 | ||
943 | ret = btrfs_del_item(trans, root, path); | |
944 | if (ret) | |
945 | goto out; | |
a061fc8d CM |
946 | out: |
947 | btrfs_free_path(path); | |
7d9eb12c | 948 | unlock_chunks(root); |
a061fc8d CM |
949 | btrfs_commit_transaction(trans, root); |
950 | return ret; | |
951 | } | |
952 | ||
953 | int btrfs_rm_device(struct btrfs_root *root, char *device_path) | |
954 | { | |
955 | struct btrfs_device *device; | |
2b82032c | 956 | struct btrfs_device *next_device; |
a061fc8d | 957 | struct block_device *bdev; |
dfe25020 | 958 | struct buffer_head *bh = NULL; |
a061fc8d CM |
959 | struct btrfs_super_block *disk_super; |
960 | u64 all_avail; | |
961 | u64 devid; | |
2b82032c YZ |
962 | u64 num_devices; |
963 | u8 *dev_uuid; | |
a061fc8d CM |
964 | int ret = 0; |
965 | ||
a061fc8d | 966 | mutex_lock(&uuid_mutex); |
7d9eb12c | 967 | mutex_lock(&root->fs_info->volume_mutex); |
a061fc8d CM |
968 | |
969 | all_avail = root->fs_info->avail_data_alloc_bits | | |
970 | root->fs_info->avail_system_alloc_bits | | |
971 | root->fs_info->avail_metadata_alloc_bits; | |
972 | ||
973 | if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) && | |
2b82032c | 974 | root->fs_info->fs_devices->rw_devices <= 4) { |
a061fc8d CM |
975 | printk("btrfs: unable to go below four devices on raid10\n"); |
976 | ret = -EINVAL; | |
977 | goto out; | |
978 | } | |
979 | ||
980 | if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) && | |
2b82032c | 981 | root->fs_info->fs_devices->rw_devices <= 2) { |
a061fc8d CM |
982 | printk("btrfs: unable to go below two devices on raid1\n"); |
983 | ret = -EINVAL; | |
984 | goto out; | |
985 | } | |
986 | ||
dfe25020 CM |
987 | if (strcmp(device_path, "missing") == 0) { |
988 | struct list_head *cur; | |
989 | struct list_head *devices; | |
990 | struct btrfs_device *tmp; | |
a061fc8d | 991 | |
dfe25020 CM |
992 | device = NULL; |
993 | devices = &root->fs_info->fs_devices->devices; | |
994 | list_for_each(cur, devices) { | |
995 | tmp = list_entry(cur, struct btrfs_device, dev_list); | |
996 | if (tmp->in_fs_metadata && !tmp->bdev) { | |
997 | device = tmp; | |
998 | break; | |
999 | } | |
1000 | } | |
1001 | bdev = NULL; | |
1002 | bh = NULL; | |
1003 | disk_super = NULL; | |
1004 | if (!device) { | |
1005 | printk("btrfs: no missing devices found to remove\n"); | |
1006 | goto out; | |
1007 | } | |
dfe25020 | 1008 | } else { |
15916de8 | 1009 | bdev = open_bdev_exclusive(device_path, MS_RDONLY, |
dfe25020 CM |
1010 | root->fs_info->bdev_holder); |
1011 | if (IS_ERR(bdev)) { | |
1012 | ret = PTR_ERR(bdev); | |
1013 | goto out; | |
1014 | } | |
a061fc8d | 1015 | |
2b82032c | 1016 | set_blocksize(bdev, 4096); |
dfe25020 CM |
1017 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); |
1018 | if (!bh) { | |
1019 | ret = -EIO; | |
1020 | goto error_close; | |
1021 | } | |
1022 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
1023 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
2b82032c | 1024 | sizeof(disk_super->magic))) { |
dfe25020 CM |
1025 | ret = -ENOENT; |
1026 | goto error_brelse; | |
1027 | } | |
1028 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
2b82032c YZ |
1029 | dev_uuid = disk_super->dev_item.uuid; |
1030 | device = btrfs_find_device(root, devid, dev_uuid, | |
1031 | disk_super->fsid); | |
dfe25020 CM |
1032 | if (!device) { |
1033 | ret = -ENOENT; | |
1034 | goto error_brelse; | |
1035 | } | |
2b82032c | 1036 | } |
dfe25020 | 1037 | |
2b82032c YZ |
1038 | if (device->writeable && root->fs_info->fs_devices->rw_devices == 1) { |
1039 | printk("btrfs: unable to remove the only writeable device\n"); | |
1040 | ret = -EINVAL; | |
1041 | goto error_brelse; | |
1042 | } | |
1043 | ||
1044 | if (device->writeable) { | |
1045 | list_del_init(&device->dev_alloc_list); | |
1046 | root->fs_info->fs_devices->rw_devices--; | |
dfe25020 | 1047 | } |
a061fc8d CM |
1048 | |
1049 | ret = btrfs_shrink_device(device, 0); | |
1050 | if (ret) | |
1051 | goto error_brelse; | |
1052 | ||
a061fc8d CM |
1053 | ret = btrfs_rm_dev_item(root->fs_info->chunk_root, device); |
1054 | if (ret) | |
1055 | goto error_brelse; | |
1056 | ||
2b82032c YZ |
1057 | device->in_fs_metadata = 0; |
1058 | if (device->fs_devices == root->fs_info->fs_devices) { | |
1059 | list_del_init(&device->dev_list); | |
1060 | root->fs_info->fs_devices->num_devices--; | |
1061 | if (device->bdev) | |
1062 | device->fs_devices->open_devices--; | |
1063 | } | |
1064 | ||
1065 | next_device = list_entry(root->fs_info->fs_devices->devices.next, | |
1066 | struct btrfs_device, dev_list); | |
1067 | if (device->bdev == root->fs_info->sb->s_bdev) | |
1068 | root->fs_info->sb->s_bdev = next_device->bdev; | |
1069 | if (device->bdev == root->fs_info->fs_devices->latest_bdev) | |
1070 | root->fs_info->fs_devices->latest_bdev = next_device->bdev; | |
1071 | ||
1072 | num_devices = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; | |
1073 | btrfs_set_super_num_devices(&root->fs_info->super_copy, num_devices); | |
1074 | ||
1075 | if (device->fs_devices != root->fs_info->fs_devices) { | |
1076 | BUG_ON(device->writeable); | |
1077 | brelse(bh); | |
1078 | if (bdev) | |
15916de8 | 1079 | close_bdev_exclusive(bdev, MS_RDONLY); |
2b82032c YZ |
1080 | |
1081 | if (device->bdev) { | |
15916de8 | 1082 | close_bdev_exclusive(device->bdev, device->mode); |
2b82032c YZ |
1083 | device->bdev = NULL; |
1084 | device->fs_devices->open_devices--; | |
1085 | } | |
1086 | if (device->fs_devices->open_devices == 0) { | |
1087 | struct btrfs_fs_devices *fs_devices; | |
1088 | fs_devices = root->fs_info->fs_devices; | |
1089 | while (fs_devices) { | |
1090 | if (fs_devices->seed == device->fs_devices) | |
1091 | break; | |
1092 | fs_devices = fs_devices->seed; | |
1093 | } | |
1094 | fs_devices->seed = device->fs_devices->seed; | |
1095 | device->fs_devices->seed = NULL; | |
1096 | __btrfs_close_devices(device->fs_devices); | |
1097 | } | |
1098 | ret = 0; | |
1099 | goto out; | |
1100 | } | |
1101 | ||
1102 | /* | |
1103 | * at this point, the device is zero sized. We want to | |
1104 | * remove it from the devices list and zero out the old super | |
1105 | */ | |
1106 | if (device->writeable) { | |
dfe25020 CM |
1107 | /* make sure this device isn't detected as part of |
1108 | * the FS anymore | |
1109 | */ | |
1110 | memset(&disk_super->magic, 0, sizeof(disk_super->magic)); | |
1111 | set_buffer_dirty(bh); | |
1112 | sync_dirty_buffer(bh); | |
dfe25020 | 1113 | } |
2b82032c | 1114 | brelse(bh); |
a061fc8d | 1115 | |
dfe25020 CM |
1116 | if (device->bdev) { |
1117 | /* one close for the device struct or super_block */ | |
15916de8 | 1118 | close_bdev_exclusive(device->bdev, device->mode); |
dfe25020 CM |
1119 | } |
1120 | if (bdev) { | |
1121 | /* one close for us */ | |
15916de8 | 1122 | close_bdev_exclusive(bdev, MS_RDONLY); |
dfe25020 | 1123 | } |
a061fc8d CM |
1124 | kfree(device->name); |
1125 | kfree(device); | |
1126 | ret = 0; | |
1127 | goto out; | |
1128 | ||
1129 | error_brelse: | |
1130 | brelse(bh); | |
1131 | error_close: | |
dfe25020 | 1132 | if (bdev) |
15916de8 | 1133 | close_bdev_exclusive(bdev, MS_RDONLY); |
a061fc8d | 1134 | out: |
7d9eb12c | 1135 | mutex_unlock(&root->fs_info->volume_mutex); |
a061fc8d | 1136 | mutex_unlock(&uuid_mutex); |
a061fc8d CM |
1137 | return ret; |
1138 | } | |
1139 | ||
2b82032c YZ |
1140 | /* |
1141 | * does all the dirty work required for changing file system's UUID. | |
1142 | */ | |
1143 | static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans, | |
1144 | struct btrfs_root *root) | |
1145 | { | |
1146 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
1147 | struct btrfs_fs_devices *old_devices; | |
1148 | struct btrfs_super_block *disk_super = &root->fs_info->super_copy; | |
1149 | struct btrfs_device *device; | |
1150 | u64 super_flags; | |
1151 | ||
1152 | BUG_ON(!mutex_is_locked(&uuid_mutex)); | |
1153 | if (!fs_devices->seeding || fs_devices->opened != 1) | |
1154 | return -EINVAL; | |
1155 | ||
1156 | old_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS); | |
1157 | if (!old_devices) | |
1158 | return -ENOMEM; | |
1159 | ||
1160 | memcpy(old_devices, fs_devices, sizeof(*old_devices)); | |
1161 | old_devices->opened = 1; | |
1162 | old_devices->sprouted = 1; | |
1163 | INIT_LIST_HEAD(&old_devices->devices); | |
1164 | INIT_LIST_HEAD(&old_devices->alloc_list); | |
1165 | list_splice_init(&fs_devices->devices, &old_devices->devices); | |
1166 | list_splice_init(&fs_devices->alloc_list, &old_devices->alloc_list); | |
1167 | list_for_each_entry(device, &old_devices->devices, dev_list) { | |
1168 | device->fs_devices = old_devices; | |
1169 | } | |
1170 | list_add(&old_devices->list, &fs_uuids); | |
1171 | ||
1172 | fs_devices->seeding = 0; | |
1173 | fs_devices->num_devices = 0; | |
1174 | fs_devices->open_devices = 0; | |
1175 | fs_devices->seed = old_devices; | |
1176 | ||
1177 | generate_random_uuid(fs_devices->fsid); | |
1178 | memcpy(root->fs_info->fsid, fs_devices->fsid, BTRFS_FSID_SIZE); | |
1179 | memcpy(disk_super->fsid, fs_devices->fsid, BTRFS_FSID_SIZE); | |
1180 | super_flags = btrfs_super_flags(disk_super) & | |
1181 | ~BTRFS_SUPER_FLAG_SEEDING; | |
1182 | btrfs_set_super_flags(disk_super, super_flags); | |
1183 | ||
1184 | return 0; | |
1185 | } | |
1186 | ||
1187 | /* | |
1188 | * strore the expected generation for seed devices in device items. | |
1189 | */ | |
1190 | static int btrfs_finish_sprout(struct btrfs_trans_handle *trans, | |
1191 | struct btrfs_root *root) | |
1192 | { | |
1193 | struct btrfs_path *path; | |
1194 | struct extent_buffer *leaf; | |
1195 | struct btrfs_dev_item *dev_item; | |
1196 | struct btrfs_device *device; | |
1197 | struct btrfs_key key; | |
1198 | u8 fs_uuid[BTRFS_UUID_SIZE]; | |
1199 | u8 dev_uuid[BTRFS_UUID_SIZE]; | |
1200 | u64 devid; | |
1201 | int ret; | |
1202 | ||
1203 | path = btrfs_alloc_path(); | |
1204 | if (!path) | |
1205 | return -ENOMEM; | |
1206 | ||
1207 | root = root->fs_info->chunk_root; | |
1208 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1209 | key.offset = 0; | |
1210 | key.type = BTRFS_DEV_ITEM_KEY; | |
1211 | ||
1212 | while (1) { | |
1213 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
1214 | if (ret < 0) | |
1215 | goto error; | |
1216 | ||
1217 | leaf = path->nodes[0]; | |
1218 | next_slot: | |
1219 | if (path->slots[0] >= btrfs_header_nritems(leaf)) { | |
1220 | ret = btrfs_next_leaf(root, path); | |
1221 | if (ret > 0) | |
1222 | break; | |
1223 | if (ret < 0) | |
1224 | goto error; | |
1225 | leaf = path->nodes[0]; | |
1226 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
1227 | btrfs_release_path(root, path); | |
1228 | continue; | |
1229 | } | |
1230 | ||
1231 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
1232 | if (key.objectid != BTRFS_DEV_ITEMS_OBJECTID || | |
1233 | key.type != BTRFS_DEV_ITEM_KEY) | |
1234 | break; | |
1235 | ||
1236 | dev_item = btrfs_item_ptr(leaf, path->slots[0], | |
1237 | struct btrfs_dev_item); | |
1238 | devid = btrfs_device_id(leaf, dev_item); | |
1239 | read_extent_buffer(leaf, dev_uuid, | |
1240 | (unsigned long)btrfs_device_uuid(dev_item), | |
1241 | BTRFS_UUID_SIZE); | |
1242 | read_extent_buffer(leaf, fs_uuid, | |
1243 | (unsigned long)btrfs_device_fsid(dev_item), | |
1244 | BTRFS_UUID_SIZE); | |
1245 | device = btrfs_find_device(root, devid, dev_uuid, fs_uuid); | |
1246 | BUG_ON(!device); | |
1247 | ||
1248 | if (device->fs_devices->seeding) { | |
1249 | btrfs_set_device_generation(leaf, dev_item, | |
1250 | device->generation); | |
1251 | btrfs_mark_buffer_dirty(leaf); | |
1252 | } | |
1253 | ||
1254 | path->slots[0]++; | |
1255 | goto next_slot; | |
1256 | } | |
1257 | ret = 0; | |
1258 | error: | |
1259 | btrfs_free_path(path); | |
1260 | return ret; | |
1261 | } | |
1262 | ||
788f20eb CM |
1263 | int btrfs_init_new_device(struct btrfs_root *root, char *device_path) |
1264 | { | |
1265 | struct btrfs_trans_handle *trans; | |
1266 | struct btrfs_device *device; | |
1267 | struct block_device *bdev; | |
1268 | struct list_head *cur; | |
1269 | struct list_head *devices; | |
2b82032c | 1270 | struct super_block *sb = root->fs_info->sb; |
788f20eb | 1271 | u64 total_bytes; |
2b82032c | 1272 | int seeding_dev = 0; |
788f20eb CM |
1273 | int ret = 0; |
1274 | ||
2b82032c YZ |
1275 | if ((sb->s_flags & MS_RDONLY) && !root->fs_info->fs_devices->seeding) |
1276 | return -EINVAL; | |
788f20eb | 1277 | |
15916de8 | 1278 | bdev = open_bdev_exclusive(device_path, 0, root->fs_info->bdev_holder); |
788f20eb CM |
1279 | if (!bdev) { |
1280 | return -EIO; | |
1281 | } | |
a2135011 | 1282 | |
2b82032c YZ |
1283 | if (root->fs_info->fs_devices->seeding) { |
1284 | seeding_dev = 1; | |
1285 | down_write(&sb->s_umount); | |
1286 | mutex_lock(&uuid_mutex); | |
1287 | } | |
1288 | ||
8c8bee1d | 1289 | filemap_write_and_wait(bdev->bd_inode->i_mapping); |
7d9eb12c | 1290 | mutex_lock(&root->fs_info->volume_mutex); |
a2135011 | 1291 | |
788f20eb CM |
1292 | devices = &root->fs_info->fs_devices->devices; |
1293 | list_for_each(cur, devices) { | |
1294 | device = list_entry(cur, struct btrfs_device, dev_list); | |
1295 | if (device->bdev == bdev) { | |
1296 | ret = -EEXIST; | |
2b82032c | 1297 | goto error; |
788f20eb CM |
1298 | } |
1299 | } | |
1300 | ||
1301 | device = kzalloc(sizeof(*device), GFP_NOFS); | |
1302 | if (!device) { | |
1303 | /* we can safely leave the fs_devices entry around */ | |
1304 | ret = -ENOMEM; | |
2b82032c | 1305 | goto error; |
788f20eb CM |
1306 | } |
1307 | ||
788f20eb CM |
1308 | device->name = kstrdup(device_path, GFP_NOFS); |
1309 | if (!device->name) { | |
1310 | kfree(device); | |
2b82032c YZ |
1311 | ret = -ENOMEM; |
1312 | goto error; | |
788f20eb | 1313 | } |
2b82032c YZ |
1314 | |
1315 | ret = find_next_devid(root, &device->devid); | |
1316 | if (ret) { | |
1317 | kfree(device); | |
1318 | goto error; | |
1319 | } | |
1320 | ||
1321 | trans = btrfs_start_transaction(root, 1); | |
1322 | lock_chunks(root); | |
1323 | ||
1324 | device->barriers = 1; | |
1325 | device->writeable = 1; | |
1326 | device->work.func = pending_bios_fn; | |
1327 | generate_random_uuid(device->uuid); | |
1328 | spin_lock_init(&device->io_lock); | |
1329 | device->generation = trans->transid; | |
788f20eb CM |
1330 | device->io_width = root->sectorsize; |
1331 | device->io_align = root->sectorsize; | |
1332 | device->sector_size = root->sectorsize; | |
1333 | device->total_bytes = i_size_read(bdev->bd_inode); | |
1334 | device->dev_root = root->fs_info->dev_root; | |
1335 | device->bdev = bdev; | |
dfe25020 | 1336 | device->in_fs_metadata = 1; |
15916de8 | 1337 | device->mode = 0; |
2b82032c | 1338 | set_blocksize(device->bdev, 4096); |
788f20eb | 1339 | |
2b82032c YZ |
1340 | if (seeding_dev) { |
1341 | sb->s_flags &= ~MS_RDONLY; | |
1342 | ret = btrfs_prepare_sprout(trans, root); | |
1343 | BUG_ON(ret); | |
1344 | } | |
788f20eb | 1345 | |
2b82032c YZ |
1346 | device->fs_devices = root->fs_info->fs_devices; |
1347 | list_add(&device->dev_list, &root->fs_info->fs_devices->devices); | |
1348 | list_add(&device->dev_alloc_list, | |
1349 | &root->fs_info->fs_devices->alloc_list); | |
1350 | root->fs_info->fs_devices->num_devices++; | |
1351 | root->fs_info->fs_devices->open_devices++; | |
1352 | root->fs_info->fs_devices->rw_devices++; | |
1353 | root->fs_info->fs_devices->total_rw_bytes += device->total_bytes; | |
325cd4ba | 1354 | |
788f20eb CM |
1355 | total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
1356 | btrfs_set_super_total_bytes(&root->fs_info->super_copy, | |
1357 | total_bytes + device->total_bytes); | |
1358 | ||
1359 | total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy); | |
1360 | btrfs_set_super_num_devices(&root->fs_info->super_copy, | |
1361 | total_bytes + 1); | |
1362 | ||
2b82032c YZ |
1363 | if (seeding_dev) { |
1364 | ret = init_first_rw_device(trans, root, device); | |
1365 | BUG_ON(ret); | |
1366 | ret = btrfs_finish_sprout(trans, root); | |
1367 | BUG_ON(ret); | |
1368 | } else { | |
1369 | ret = btrfs_add_device(trans, root, device); | |
1370 | } | |
1371 | ||
7d9eb12c | 1372 | unlock_chunks(root); |
2b82032c | 1373 | btrfs_commit_transaction(trans, root); |
a2135011 | 1374 | |
2b82032c YZ |
1375 | if (seeding_dev) { |
1376 | mutex_unlock(&uuid_mutex); | |
1377 | up_write(&sb->s_umount); | |
788f20eb | 1378 | |
2b82032c YZ |
1379 | ret = btrfs_relocate_sys_chunks(root); |
1380 | BUG_ON(ret); | |
1381 | } | |
1382 | out: | |
1383 | mutex_unlock(&root->fs_info->volume_mutex); | |
1384 | return ret; | |
1385 | error: | |
15916de8 | 1386 | close_bdev_exclusive(bdev, 0); |
2b82032c YZ |
1387 | if (seeding_dev) { |
1388 | mutex_unlock(&uuid_mutex); | |
1389 | up_write(&sb->s_umount); | |
1390 | } | |
788f20eb CM |
1391 | goto out; |
1392 | } | |
1393 | ||
a1b32a59 CM |
1394 | int noinline btrfs_update_device(struct btrfs_trans_handle *trans, |
1395 | struct btrfs_device *device) | |
0b86a832 CM |
1396 | { |
1397 | int ret; | |
1398 | struct btrfs_path *path; | |
1399 | struct btrfs_root *root; | |
1400 | struct btrfs_dev_item *dev_item; | |
1401 | struct extent_buffer *leaf; | |
1402 | struct btrfs_key key; | |
1403 | ||
1404 | root = device->dev_root->fs_info->chunk_root; | |
1405 | ||
1406 | path = btrfs_alloc_path(); | |
1407 | if (!path) | |
1408 | return -ENOMEM; | |
1409 | ||
1410 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1411 | key.type = BTRFS_DEV_ITEM_KEY; | |
1412 | key.offset = device->devid; | |
1413 | ||
1414 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
1415 | if (ret < 0) | |
1416 | goto out; | |
1417 | ||
1418 | if (ret > 0) { | |
1419 | ret = -ENOENT; | |
1420 | goto out; | |
1421 | } | |
1422 | ||
1423 | leaf = path->nodes[0]; | |
1424 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
1425 | ||
1426 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
1427 | btrfs_set_device_type(leaf, dev_item, device->type); | |
1428 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
1429 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
1430 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
1431 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
1432 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
1433 | btrfs_mark_buffer_dirty(leaf); | |
1434 | ||
1435 | out: | |
1436 | btrfs_free_path(path); | |
1437 | return ret; | |
1438 | } | |
1439 | ||
7d9eb12c | 1440 | static int __btrfs_grow_device(struct btrfs_trans_handle *trans, |
8f18cf13 CM |
1441 | struct btrfs_device *device, u64 new_size) |
1442 | { | |
1443 | struct btrfs_super_block *super_copy = | |
1444 | &device->dev_root->fs_info->super_copy; | |
1445 | u64 old_total = btrfs_super_total_bytes(super_copy); | |
1446 | u64 diff = new_size - device->total_bytes; | |
1447 | ||
2b82032c YZ |
1448 | if (!device->writeable) |
1449 | return -EACCES; | |
1450 | if (new_size <= device->total_bytes) | |
1451 | return -EINVAL; | |
1452 | ||
8f18cf13 | 1453 | btrfs_set_super_total_bytes(super_copy, old_total + diff); |
2b82032c YZ |
1454 | device->fs_devices->total_rw_bytes += diff; |
1455 | ||
1456 | device->total_bytes = new_size; | |
8f18cf13 CM |
1457 | return btrfs_update_device(trans, device); |
1458 | } | |
1459 | ||
7d9eb12c CM |
1460 | int btrfs_grow_device(struct btrfs_trans_handle *trans, |
1461 | struct btrfs_device *device, u64 new_size) | |
1462 | { | |
1463 | int ret; | |
1464 | lock_chunks(device->dev_root); | |
1465 | ret = __btrfs_grow_device(trans, device, new_size); | |
1466 | unlock_chunks(device->dev_root); | |
1467 | return ret; | |
1468 | } | |
1469 | ||
8f18cf13 CM |
1470 | static int btrfs_free_chunk(struct btrfs_trans_handle *trans, |
1471 | struct btrfs_root *root, | |
1472 | u64 chunk_tree, u64 chunk_objectid, | |
1473 | u64 chunk_offset) | |
1474 | { | |
1475 | int ret; | |
1476 | struct btrfs_path *path; | |
1477 | struct btrfs_key key; | |
1478 | ||
1479 | root = root->fs_info->chunk_root; | |
1480 | path = btrfs_alloc_path(); | |
1481 | if (!path) | |
1482 | return -ENOMEM; | |
1483 | ||
1484 | key.objectid = chunk_objectid; | |
1485 | key.offset = chunk_offset; | |
1486 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
1487 | ||
1488 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); | |
1489 | BUG_ON(ret); | |
1490 | ||
1491 | ret = btrfs_del_item(trans, root, path); | |
1492 | BUG_ON(ret); | |
1493 | ||
1494 | btrfs_free_path(path); | |
1495 | return 0; | |
1496 | } | |
1497 | ||
1498 | int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64 | |
1499 | chunk_offset) | |
1500 | { | |
1501 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1502 | struct btrfs_disk_key *disk_key; | |
1503 | struct btrfs_chunk *chunk; | |
1504 | u8 *ptr; | |
1505 | int ret = 0; | |
1506 | u32 num_stripes; | |
1507 | u32 array_size; | |
1508 | u32 len = 0; | |
1509 | u32 cur; | |
1510 | struct btrfs_key key; | |
1511 | ||
1512 | array_size = btrfs_super_sys_array_size(super_copy); | |
1513 | ||
1514 | ptr = super_copy->sys_chunk_array; | |
1515 | cur = 0; | |
1516 | ||
1517 | while (cur < array_size) { | |
1518 | disk_key = (struct btrfs_disk_key *)ptr; | |
1519 | btrfs_disk_key_to_cpu(&key, disk_key); | |
1520 | ||
1521 | len = sizeof(*disk_key); | |
1522 | ||
1523 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { | |
1524 | chunk = (struct btrfs_chunk *)(ptr + len); | |
1525 | num_stripes = btrfs_stack_chunk_num_stripes(chunk); | |
1526 | len += btrfs_chunk_item_size(num_stripes); | |
1527 | } else { | |
1528 | ret = -EIO; | |
1529 | break; | |
1530 | } | |
1531 | if (key.objectid == chunk_objectid && | |
1532 | key.offset == chunk_offset) { | |
1533 | memmove(ptr, ptr + len, array_size - (cur + len)); | |
1534 | array_size -= len; | |
1535 | btrfs_set_super_sys_array_size(super_copy, array_size); | |
1536 | } else { | |
1537 | ptr += len; | |
1538 | cur += len; | |
1539 | } | |
1540 | } | |
1541 | return ret; | |
1542 | } | |
1543 | ||
8f18cf13 CM |
1544 | int btrfs_relocate_chunk(struct btrfs_root *root, |
1545 | u64 chunk_tree, u64 chunk_objectid, | |
1546 | u64 chunk_offset) | |
1547 | { | |
1548 | struct extent_map_tree *em_tree; | |
1549 | struct btrfs_root *extent_root; | |
1550 | struct btrfs_trans_handle *trans; | |
1551 | struct extent_map *em; | |
1552 | struct map_lookup *map; | |
1553 | int ret; | |
1554 | int i; | |
1555 | ||
323da79c CM |
1556 | printk("btrfs relocating chunk %llu\n", |
1557 | (unsigned long long)chunk_offset); | |
8f18cf13 CM |
1558 | root = root->fs_info->chunk_root; |
1559 | extent_root = root->fs_info->extent_root; | |
1560 | em_tree = &root->fs_info->mapping_tree.map_tree; | |
1561 | ||
1562 | /* step one, relocate all the extents inside this chunk */ | |
1a40e23b | 1563 | ret = btrfs_relocate_block_group(extent_root, chunk_offset); |
8f18cf13 CM |
1564 | BUG_ON(ret); |
1565 | ||
1566 | trans = btrfs_start_transaction(root, 1); | |
1567 | BUG_ON(!trans); | |
1568 | ||
7d9eb12c CM |
1569 | lock_chunks(root); |
1570 | ||
8f18cf13 CM |
1571 | /* |
1572 | * step two, delete the device extents and the | |
1573 | * chunk tree entries | |
1574 | */ | |
1575 | spin_lock(&em_tree->lock); | |
1576 | em = lookup_extent_mapping(em_tree, chunk_offset, 1); | |
1577 | spin_unlock(&em_tree->lock); | |
1578 | ||
a061fc8d CM |
1579 | BUG_ON(em->start > chunk_offset || |
1580 | em->start + em->len < chunk_offset); | |
8f18cf13 CM |
1581 | map = (struct map_lookup *)em->bdev; |
1582 | ||
1583 | for (i = 0; i < map->num_stripes; i++) { | |
1584 | ret = btrfs_free_dev_extent(trans, map->stripes[i].dev, | |
1585 | map->stripes[i].physical); | |
1586 | BUG_ON(ret); | |
a061fc8d | 1587 | |
dfe25020 CM |
1588 | if (map->stripes[i].dev) { |
1589 | ret = btrfs_update_device(trans, map->stripes[i].dev); | |
1590 | BUG_ON(ret); | |
1591 | } | |
8f18cf13 CM |
1592 | } |
1593 | ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid, | |
1594 | chunk_offset); | |
1595 | ||
1596 | BUG_ON(ret); | |
1597 | ||
1598 | if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) { | |
1599 | ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset); | |
1600 | BUG_ON(ret); | |
8f18cf13 CM |
1601 | } |
1602 | ||
2b82032c YZ |
1603 | ret = btrfs_remove_block_group(trans, extent_root, chunk_offset); |
1604 | BUG_ON(ret); | |
1605 | ||
1606 | spin_lock(&em_tree->lock); | |
1607 | remove_extent_mapping(em_tree, em); | |
1608 | spin_unlock(&em_tree->lock); | |
1609 | ||
1610 | kfree(map); | |
1611 | em->bdev = NULL; | |
1612 | ||
1613 | /* once for the tree */ | |
1614 | free_extent_map(em); | |
1615 | /* once for us */ | |
1616 | free_extent_map(em); | |
1617 | ||
1618 | unlock_chunks(root); | |
1619 | btrfs_end_transaction(trans, root); | |
1620 | return 0; | |
1621 | } | |
1622 | ||
1623 | static int btrfs_relocate_sys_chunks(struct btrfs_root *root) | |
1624 | { | |
1625 | struct btrfs_root *chunk_root = root->fs_info->chunk_root; | |
1626 | struct btrfs_path *path; | |
1627 | struct extent_buffer *leaf; | |
1628 | struct btrfs_chunk *chunk; | |
1629 | struct btrfs_key key; | |
1630 | struct btrfs_key found_key; | |
1631 | u64 chunk_tree = chunk_root->root_key.objectid; | |
1632 | u64 chunk_type; | |
1633 | int ret; | |
1634 | ||
1635 | path = btrfs_alloc_path(); | |
1636 | if (!path) | |
1637 | return -ENOMEM; | |
1638 | ||
1639 | key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; | |
1640 | key.offset = (u64)-1; | |
1641 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
1642 | ||
1643 | while (1) { | |
1644 | ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0); | |
1645 | if (ret < 0) | |
1646 | goto error; | |
1647 | BUG_ON(ret == 0); | |
1648 | ||
1649 | ret = btrfs_previous_item(chunk_root, path, key.objectid, | |
1650 | key.type); | |
1651 | if (ret < 0) | |
1652 | goto error; | |
1653 | if (ret > 0) | |
1654 | break; | |
1a40e23b | 1655 | |
2b82032c YZ |
1656 | leaf = path->nodes[0]; |
1657 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
1a40e23b | 1658 | |
2b82032c YZ |
1659 | chunk = btrfs_item_ptr(leaf, path->slots[0], |
1660 | struct btrfs_chunk); | |
1661 | chunk_type = btrfs_chunk_type(leaf, chunk); | |
1662 | btrfs_release_path(chunk_root, path); | |
8f18cf13 | 1663 | |
2b82032c YZ |
1664 | if (chunk_type & BTRFS_BLOCK_GROUP_SYSTEM) { |
1665 | ret = btrfs_relocate_chunk(chunk_root, chunk_tree, | |
1666 | found_key.objectid, | |
1667 | found_key.offset); | |
1668 | BUG_ON(ret); | |
1669 | } | |
8f18cf13 | 1670 | |
2b82032c YZ |
1671 | if (found_key.offset == 0) |
1672 | break; | |
1673 | key.offset = found_key.offset - 1; | |
1674 | } | |
1675 | ret = 0; | |
1676 | error: | |
1677 | btrfs_free_path(path); | |
1678 | return ret; | |
8f18cf13 CM |
1679 | } |
1680 | ||
ec44a35c CM |
1681 | static u64 div_factor(u64 num, int factor) |
1682 | { | |
1683 | if (factor == 10) | |
1684 | return num; | |
1685 | num *= factor; | |
1686 | do_div(num, 10); | |
1687 | return num; | |
1688 | } | |
1689 | ||
ec44a35c CM |
1690 | int btrfs_balance(struct btrfs_root *dev_root) |
1691 | { | |
1692 | int ret; | |
1693 | struct list_head *cur; | |
1694 | struct list_head *devices = &dev_root->fs_info->fs_devices->devices; | |
1695 | struct btrfs_device *device; | |
1696 | u64 old_size; | |
1697 | u64 size_to_free; | |
1698 | struct btrfs_path *path; | |
1699 | struct btrfs_key key; | |
1700 | struct btrfs_chunk *chunk; | |
1701 | struct btrfs_root *chunk_root = dev_root->fs_info->chunk_root; | |
1702 | struct btrfs_trans_handle *trans; | |
1703 | struct btrfs_key found_key; | |
1704 | ||
2b82032c YZ |
1705 | if (dev_root->fs_info->sb->s_flags & MS_RDONLY) |
1706 | return -EROFS; | |
ec44a35c | 1707 | |
7d9eb12c | 1708 | mutex_lock(&dev_root->fs_info->volume_mutex); |
ec44a35c CM |
1709 | dev_root = dev_root->fs_info->dev_root; |
1710 | ||
ec44a35c CM |
1711 | /* step one make some room on all the devices */ |
1712 | list_for_each(cur, devices) { | |
1713 | device = list_entry(cur, struct btrfs_device, dev_list); | |
1714 | old_size = device->total_bytes; | |
1715 | size_to_free = div_factor(old_size, 1); | |
1716 | size_to_free = min(size_to_free, (u64)1 * 1024 * 1024); | |
2b82032c YZ |
1717 | if (!device->writeable || |
1718 | device->total_bytes - device->bytes_used > size_to_free) | |
ec44a35c CM |
1719 | continue; |
1720 | ||
1721 | ret = btrfs_shrink_device(device, old_size - size_to_free); | |
1722 | BUG_ON(ret); | |
1723 | ||
1724 | trans = btrfs_start_transaction(dev_root, 1); | |
1725 | BUG_ON(!trans); | |
1726 | ||
1727 | ret = btrfs_grow_device(trans, device, old_size); | |
1728 | BUG_ON(ret); | |
1729 | ||
1730 | btrfs_end_transaction(trans, dev_root); | |
1731 | } | |
1732 | ||
1733 | /* step two, relocate all the chunks */ | |
1734 | path = btrfs_alloc_path(); | |
1735 | BUG_ON(!path); | |
1736 | ||
1737 | key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; | |
1738 | key.offset = (u64)-1; | |
1739 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
1740 | ||
1741 | while(1) { | |
1742 | ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0); | |
1743 | if (ret < 0) | |
1744 | goto error; | |
1745 | ||
1746 | /* | |
1747 | * this shouldn't happen, it means the last relocate | |
1748 | * failed | |
1749 | */ | |
1750 | if (ret == 0) | |
1751 | break; | |
1752 | ||
1753 | ret = btrfs_previous_item(chunk_root, path, 0, | |
1754 | BTRFS_CHUNK_ITEM_KEY); | |
7d9eb12c | 1755 | if (ret) |
ec44a35c | 1756 | break; |
7d9eb12c | 1757 | |
ec44a35c CM |
1758 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, |
1759 | path->slots[0]); | |
1760 | if (found_key.objectid != key.objectid) | |
1761 | break; | |
7d9eb12c | 1762 | |
ec44a35c CM |
1763 | chunk = btrfs_item_ptr(path->nodes[0], |
1764 | path->slots[0], | |
1765 | struct btrfs_chunk); | |
1766 | key.offset = found_key.offset; | |
1767 | /* chunk zero is special */ | |
1768 | if (key.offset == 0) | |
1769 | break; | |
1770 | ||
7d9eb12c | 1771 | btrfs_release_path(chunk_root, path); |
ec44a35c CM |
1772 | ret = btrfs_relocate_chunk(chunk_root, |
1773 | chunk_root->root_key.objectid, | |
1774 | found_key.objectid, | |
1775 | found_key.offset); | |
1776 | BUG_ON(ret); | |
ec44a35c CM |
1777 | } |
1778 | ret = 0; | |
1779 | error: | |
1780 | btrfs_free_path(path); | |
7d9eb12c | 1781 | mutex_unlock(&dev_root->fs_info->volume_mutex); |
ec44a35c CM |
1782 | return ret; |
1783 | } | |
1784 | ||
8f18cf13 CM |
1785 | /* |
1786 | * shrinking a device means finding all of the device extents past | |
1787 | * the new size, and then following the back refs to the chunks. | |
1788 | * The chunk relocation code actually frees the device extent | |
1789 | */ | |
1790 | int btrfs_shrink_device(struct btrfs_device *device, u64 new_size) | |
1791 | { | |
1792 | struct btrfs_trans_handle *trans; | |
1793 | struct btrfs_root *root = device->dev_root; | |
1794 | struct btrfs_dev_extent *dev_extent = NULL; | |
1795 | struct btrfs_path *path; | |
1796 | u64 length; | |
1797 | u64 chunk_tree; | |
1798 | u64 chunk_objectid; | |
1799 | u64 chunk_offset; | |
1800 | int ret; | |
1801 | int slot; | |
1802 | struct extent_buffer *l; | |
1803 | struct btrfs_key key; | |
1804 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1805 | u64 old_total = btrfs_super_total_bytes(super_copy); | |
1806 | u64 diff = device->total_bytes - new_size; | |
1807 | ||
2b82032c YZ |
1808 | if (new_size >= device->total_bytes) |
1809 | return -EINVAL; | |
8f18cf13 CM |
1810 | |
1811 | path = btrfs_alloc_path(); | |
1812 | if (!path) | |
1813 | return -ENOMEM; | |
1814 | ||
1815 | trans = btrfs_start_transaction(root, 1); | |
1816 | if (!trans) { | |
1817 | ret = -ENOMEM; | |
1818 | goto done; | |
1819 | } | |
1820 | ||
1821 | path->reada = 2; | |
1822 | ||
7d9eb12c CM |
1823 | lock_chunks(root); |
1824 | ||
8f18cf13 | 1825 | device->total_bytes = new_size; |
2b82032c YZ |
1826 | if (device->writeable) |
1827 | device->fs_devices->total_rw_bytes -= diff; | |
8f18cf13 CM |
1828 | ret = btrfs_update_device(trans, device); |
1829 | if (ret) { | |
7d9eb12c | 1830 | unlock_chunks(root); |
8f18cf13 CM |
1831 | btrfs_end_transaction(trans, root); |
1832 | goto done; | |
1833 | } | |
1834 | WARN_ON(diff > old_total); | |
1835 | btrfs_set_super_total_bytes(super_copy, old_total - diff); | |
7d9eb12c | 1836 | unlock_chunks(root); |
8f18cf13 CM |
1837 | btrfs_end_transaction(trans, root); |
1838 | ||
1839 | key.objectid = device->devid; | |
1840 | key.offset = (u64)-1; | |
1841 | key.type = BTRFS_DEV_EXTENT_KEY; | |
1842 | ||
1843 | while (1) { | |
1844 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1845 | if (ret < 0) | |
1846 | goto done; | |
1847 | ||
1848 | ret = btrfs_previous_item(root, path, 0, key.type); | |
1849 | if (ret < 0) | |
1850 | goto done; | |
1851 | if (ret) { | |
1852 | ret = 0; | |
1853 | goto done; | |
1854 | } | |
1855 | ||
1856 | l = path->nodes[0]; | |
1857 | slot = path->slots[0]; | |
1858 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
1859 | ||
1860 | if (key.objectid != device->devid) | |
1861 | goto done; | |
1862 | ||
1863 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
1864 | length = btrfs_dev_extent_length(l, dev_extent); | |
1865 | ||
1866 | if (key.offset + length <= new_size) | |
1867 | goto done; | |
1868 | ||
1869 | chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent); | |
1870 | chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent); | |
1871 | chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent); | |
1872 | btrfs_release_path(root, path); | |
1873 | ||
1874 | ret = btrfs_relocate_chunk(root, chunk_tree, chunk_objectid, | |
1875 | chunk_offset); | |
1876 | if (ret) | |
1877 | goto done; | |
1878 | } | |
1879 | ||
1880 | done: | |
1881 | btrfs_free_path(path); | |
1882 | return ret; | |
1883 | } | |
1884 | ||
0b86a832 CM |
1885 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, |
1886 | struct btrfs_root *root, | |
1887 | struct btrfs_key *key, | |
1888 | struct btrfs_chunk *chunk, int item_size) | |
1889 | { | |
1890 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
1891 | struct btrfs_disk_key disk_key; | |
1892 | u32 array_size; | |
1893 | u8 *ptr; | |
1894 | ||
1895 | array_size = btrfs_super_sys_array_size(super_copy); | |
1896 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
1897 | return -EFBIG; | |
1898 | ||
1899 | ptr = super_copy->sys_chunk_array + array_size; | |
1900 | btrfs_cpu_key_to_disk(&disk_key, key); | |
1901 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
1902 | ptr += sizeof(disk_key); | |
1903 | memcpy(ptr, chunk, item_size); | |
1904 | item_size += sizeof(disk_key); | |
1905 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
1906 | return 0; | |
1907 | } | |
1908 | ||
a1b32a59 CM |
1909 | static u64 noinline chunk_bytes_by_type(u64 type, u64 calc_size, |
1910 | int num_stripes, int sub_stripes) | |
9b3f68b9 CM |
1911 | { |
1912 | if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP)) | |
1913 | return calc_size; | |
1914 | else if (type & BTRFS_BLOCK_GROUP_RAID10) | |
1915 | return calc_size * (num_stripes / sub_stripes); | |
1916 | else | |
1917 | return calc_size * num_stripes; | |
1918 | } | |
1919 | ||
2b82032c YZ |
1920 | static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, |
1921 | struct btrfs_root *extent_root, | |
1922 | struct map_lookup **map_ret, | |
1923 | u64 *num_bytes, u64 *stripe_size, | |
1924 | u64 start, u64 type) | |
0b86a832 | 1925 | { |
593060d7 | 1926 | struct btrfs_fs_info *info = extent_root->fs_info; |
0b86a832 | 1927 | struct btrfs_device *device = NULL; |
2b82032c | 1928 | struct btrfs_fs_devices *fs_devices = info->fs_devices; |
6324fbf3 | 1929 | struct list_head *cur; |
2b82032c | 1930 | struct map_lookup *map = NULL; |
0b86a832 | 1931 | struct extent_map_tree *em_tree; |
0b86a832 | 1932 | struct extent_map *em; |
2b82032c | 1933 | struct list_head private_devs; |
a40a90a0 | 1934 | int min_stripe_size = 1 * 1024 * 1024; |
0b86a832 | 1935 | u64 calc_size = 1024 * 1024 * 1024; |
9b3f68b9 CM |
1936 | u64 max_chunk_size = calc_size; |
1937 | u64 min_free; | |
6324fbf3 CM |
1938 | u64 avail; |
1939 | u64 max_avail = 0; | |
2b82032c | 1940 | u64 dev_offset; |
6324fbf3 | 1941 | int num_stripes = 1; |
a40a90a0 | 1942 | int min_stripes = 1; |
321aecc6 | 1943 | int sub_stripes = 0; |
6324fbf3 | 1944 | int looped = 0; |
0b86a832 | 1945 | int ret; |
6324fbf3 | 1946 | int index; |
593060d7 | 1947 | int stripe_len = 64 * 1024; |
0b86a832 | 1948 | |
ec44a35c CM |
1949 | if ((type & BTRFS_BLOCK_GROUP_RAID1) && |
1950 | (type & BTRFS_BLOCK_GROUP_DUP)) { | |
1951 | WARN_ON(1); | |
1952 | type &= ~BTRFS_BLOCK_GROUP_DUP; | |
1953 | } | |
2b82032c | 1954 | if (list_empty(&fs_devices->alloc_list)) |
6324fbf3 | 1955 | return -ENOSPC; |
593060d7 | 1956 | |
a40a90a0 | 1957 | if (type & (BTRFS_BLOCK_GROUP_RAID0)) { |
2b82032c | 1958 | num_stripes = fs_devices->rw_devices; |
a40a90a0 CM |
1959 | min_stripes = 2; |
1960 | } | |
1961 | if (type & (BTRFS_BLOCK_GROUP_DUP)) { | |
611f0e00 | 1962 | num_stripes = 2; |
a40a90a0 CM |
1963 | min_stripes = 2; |
1964 | } | |
8790d502 | 1965 | if (type & (BTRFS_BLOCK_GROUP_RAID1)) { |
2b82032c | 1966 | num_stripes = min_t(u64, 2, fs_devices->rw_devices); |
9b3f68b9 CM |
1967 | if (num_stripes < 2) |
1968 | return -ENOSPC; | |
a40a90a0 | 1969 | min_stripes = 2; |
8790d502 | 1970 | } |
321aecc6 | 1971 | if (type & (BTRFS_BLOCK_GROUP_RAID10)) { |
2b82032c | 1972 | num_stripes = fs_devices->rw_devices; |
321aecc6 CM |
1973 | if (num_stripes < 4) |
1974 | return -ENOSPC; | |
1975 | num_stripes &= ~(u32)1; | |
1976 | sub_stripes = 2; | |
a40a90a0 | 1977 | min_stripes = 4; |
321aecc6 | 1978 | } |
9b3f68b9 CM |
1979 | |
1980 | if (type & BTRFS_BLOCK_GROUP_DATA) { | |
1981 | max_chunk_size = 10 * calc_size; | |
a40a90a0 | 1982 | min_stripe_size = 64 * 1024 * 1024; |
9b3f68b9 CM |
1983 | } else if (type & BTRFS_BLOCK_GROUP_METADATA) { |
1984 | max_chunk_size = 4 * calc_size; | |
a40a90a0 CM |
1985 | min_stripe_size = 32 * 1024 * 1024; |
1986 | } else if (type & BTRFS_BLOCK_GROUP_SYSTEM) { | |
1987 | calc_size = 8 * 1024 * 1024; | |
1988 | max_chunk_size = calc_size * 2; | |
1989 | min_stripe_size = 1 * 1024 * 1024; | |
9b3f68b9 CM |
1990 | } |
1991 | ||
2b82032c YZ |
1992 | /* we don't want a chunk larger than 10% of writeable space */ |
1993 | max_chunk_size = min(div_factor(fs_devices->total_rw_bytes, 1), | |
1994 | max_chunk_size); | |
9b3f68b9 | 1995 | |
a40a90a0 | 1996 | again: |
2b82032c YZ |
1997 | if (!map || map->num_stripes != num_stripes) { |
1998 | kfree(map); | |
1999 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); | |
2000 | if (!map) | |
2001 | return -ENOMEM; | |
2002 | map->num_stripes = num_stripes; | |
2003 | } | |
2004 | ||
9b3f68b9 CM |
2005 | if (calc_size * num_stripes > max_chunk_size) { |
2006 | calc_size = max_chunk_size; | |
2007 | do_div(calc_size, num_stripes); | |
2008 | do_div(calc_size, stripe_len); | |
2009 | calc_size *= stripe_len; | |
2010 | } | |
2011 | /* we don't want tiny stripes */ | |
a40a90a0 | 2012 | calc_size = max_t(u64, min_stripe_size, calc_size); |
9b3f68b9 | 2013 | |
9b3f68b9 CM |
2014 | do_div(calc_size, stripe_len); |
2015 | calc_size *= stripe_len; | |
2016 | ||
2b82032c | 2017 | cur = fs_devices->alloc_list.next; |
6324fbf3 | 2018 | index = 0; |
611f0e00 CM |
2019 | |
2020 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
2021 | min_free = calc_size * 2; | |
9b3f68b9 CM |
2022 | else |
2023 | min_free = calc_size; | |
611f0e00 | 2024 | |
0f9dd46c JB |
2025 | /* |
2026 | * we add 1MB because we never use the first 1MB of the device, unless | |
2027 | * we've looped, then we are likely allocating the maximum amount of | |
2028 | * space left already | |
2029 | */ | |
2030 | if (!looped) | |
2031 | min_free += 1024 * 1024; | |
ad5bd91e | 2032 | |
2b82032c | 2033 | INIT_LIST_HEAD(&private_devs); |
6324fbf3 | 2034 | while(index < num_stripes) { |
b3075717 | 2035 | device = list_entry(cur, struct btrfs_device, dev_alloc_list); |
2b82032c | 2036 | BUG_ON(!device->writeable); |
dfe25020 CM |
2037 | if (device->total_bytes > device->bytes_used) |
2038 | avail = device->total_bytes - device->bytes_used; | |
2039 | else | |
2040 | avail = 0; | |
6324fbf3 | 2041 | cur = cur->next; |
8f18cf13 | 2042 | |
dfe25020 | 2043 | if (device->in_fs_metadata && avail >= min_free) { |
2b82032c YZ |
2044 | ret = find_free_dev_extent(trans, device, |
2045 | min_free, &dev_offset); | |
8f18cf13 CM |
2046 | if (ret == 0) { |
2047 | list_move_tail(&device->dev_alloc_list, | |
2048 | &private_devs); | |
2b82032c YZ |
2049 | map->stripes[index].dev = device; |
2050 | map->stripes[index].physical = dev_offset; | |
611f0e00 | 2051 | index++; |
2b82032c YZ |
2052 | if (type & BTRFS_BLOCK_GROUP_DUP) { |
2053 | map->stripes[index].dev = device; | |
2054 | map->stripes[index].physical = | |
2055 | dev_offset + calc_size; | |
8f18cf13 | 2056 | index++; |
2b82032c | 2057 | } |
8f18cf13 | 2058 | } |
dfe25020 | 2059 | } else if (device->in_fs_metadata && avail > max_avail) |
a40a90a0 | 2060 | max_avail = avail; |
2b82032c | 2061 | if (cur == &fs_devices->alloc_list) |
6324fbf3 CM |
2062 | break; |
2063 | } | |
2b82032c | 2064 | list_splice(&private_devs, &fs_devices->alloc_list); |
6324fbf3 | 2065 | if (index < num_stripes) { |
a40a90a0 CM |
2066 | if (index >= min_stripes) { |
2067 | num_stripes = index; | |
2068 | if (type & (BTRFS_BLOCK_GROUP_RAID10)) { | |
2069 | num_stripes /= sub_stripes; | |
2070 | num_stripes *= sub_stripes; | |
2071 | } | |
2072 | looped = 1; | |
2073 | goto again; | |
2074 | } | |
6324fbf3 CM |
2075 | if (!looped && max_avail > 0) { |
2076 | looped = 1; | |
2077 | calc_size = max_avail; | |
2078 | goto again; | |
2079 | } | |
2b82032c | 2080 | kfree(map); |
6324fbf3 CM |
2081 | return -ENOSPC; |
2082 | } | |
2b82032c YZ |
2083 | map->sector_size = extent_root->sectorsize; |
2084 | map->stripe_len = stripe_len; | |
2085 | map->io_align = stripe_len; | |
2086 | map->io_width = stripe_len; | |
2087 | map->type = type; | |
2088 | map->num_stripes = num_stripes; | |
2089 | map->sub_stripes = sub_stripes; | |
0b86a832 | 2090 | |
2b82032c YZ |
2091 | *map_ret = map; |
2092 | *stripe_size = calc_size; | |
2093 | *num_bytes = chunk_bytes_by_type(type, calc_size, | |
2094 | num_stripes, sub_stripes); | |
0b86a832 | 2095 | |
2b82032c YZ |
2096 | em = alloc_extent_map(GFP_NOFS); |
2097 | if (!em) { | |
2098 | kfree(map); | |
593060d7 CM |
2099 | return -ENOMEM; |
2100 | } | |
2b82032c YZ |
2101 | em->bdev = (struct block_device *)map; |
2102 | em->start = start; | |
2103 | em->len = *num_bytes; | |
2104 | em->block_start = 0; | |
2105 | em->block_len = em->len; | |
593060d7 | 2106 | |
2b82032c YZ |
2107 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; |
2108 | spin_lock(&em_tree->lock); | |
2109 | ret = add_extent_mapping(em_tree, em); | |
2110 | spin_unlock(&em_tree->lock); | |
2111 | BUG_ON(ret); | |
2112 | free_extent_map(em); | |
0b86a832 | 2113 | |
2b82032c YZ |
2114 | ret = btrfs_make_block_group(trans, extent_root, 0, type, |
2115 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, | |
2116 | start, *num_bytes); | |
2117 | BUG_ON(ret); | |
611f0e00 | 2118 | |
2b82032c YZ |
2119 | index = 0; |
2120 | while (index < map->num_stripes) { | |
2121 | device = map->stripes[index].dev; | |
2122 | dev_offset = map->stripes[index].physical; | |
0b86a832 CM |
2123 | |
2124 | ret = btrfs_alloc_dev_extent(trans, device, | |
2b82032c YZ |
2125 | info->chunk_root->root_key.objectid, |
2126 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, | |
2127 | start, dev_offset, calc_size); | |
0b86a832 | 2128 | BUG_ON(ret); |
2b82032c YZ |
2129 | index++; |
2130 | } | |
2131 | ||
2132 | return 0; | |
2133 | } | |
2134 | ||
2135 | static int __finish_chunk_alloc(struct btrfs_trans_handle *trans, | |
2136 | struct btrfs_root *extent_root, | |
2137 | struct map_lookup *map, u64 chunk_offset, | |
2138 | u64 chunk_size, u64 stripe_size) | |
2139 | { | |
2140 | u64 dev_offset; | |
2141 | struct btrfs_key key; | |
2142 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; | |
2143 | struct btrfs_device *device; | |
2144 | struct btrfs_chunk *chunk; | |
2145 | struct btrfs_stripe *stripe; | |
2146 | size_t item_size = btrfs_chunk_item_size(map->num_stripes); | |
2147 | int index = 0; | |
2148 | int ret; | |
2149 | ||
2150 | chunk = kzalloc(item_size, GFP_NOFS); | |
2151 | if (!chunk) | |
2152 | return -ENOMEM; | |
2153 | ||
2154 | index = 0; | |
2155 | while (index < map->num_stripes) { | |
2156 | device = map->stripes[index].dev; | |
2157 | device->bytes_used += stripe_size; | |
0b86a832 CM |
2158 | ret = btrfs_update_device(trans, device); |
2159 | BUG_ON(ret); | |
2b82032c YZ |
2160 | index++; |
2161 | } | |
2162 | ||
2163 | index = 0; | |
2164 | stripe = &chunk->stripe; | |
2165 | while (index < map->num_stripes) { | |
2166 | device = map->stripes[index].dev; | |
2167 | dev_offset = map->stripes[index].physical; | |
0b86a832 | 2168 | |
e17cade2 CM |
2169 | btrfs_set_stack_stripe_devid(stripe, device->devid); |
2170 | btrfs_set_stack_stripe_offset(stripe, dev_offset); | |
2171 | memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 2172 | stripe++; |
0b86a832 CM |
2173 | index++; |
2174 | } | |
2175 | ||
2b82032c | 2176 | btrfs_set_stack_chunk_length(chunk, chunk_size); |
0b86a832 | 2177 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); |
2b82032c YZ |
2178 | btrfs_set_stack_chunk_stripe_len(chunk, map->stripe_len); |
2179 | btrfs_set_stack_chunk_type(chunk, map->type); | |
2180 | btrfs_set_stack_chunk_num_stripes(chunk, map->num_stripes); | |
2181 | btrfs_set_stack_chunk_io_align(chunk, map->stripe_len); | |
2182 | btrfs_set_stack_chunk_io_width(chunk, map->stripe_len); | |
0b86a832 | 2183 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); |
2b82032c | 2184 | btrfs_set_stack_chunk_sub_stripes(chunk, map->sub_stripes); |
0b86a832 | 2185 | |
2b82032c YZ |
2186 | key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID; |
2187 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
2188 | key.offset = chunk_offset; | |
0b86a832 | 2189 | |
2b82032c YZ |
2190 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, item_size); |
2191 | BUG_ON(ret); | |
0b86a832 | 2192 | |
2b82032c YZ |
2193 | if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) { |
2194 | ret = btrfs_add_system_chunk(trans, chunk_root, &key, chunk, | |
2195 | item_size); | |
8f18cf13 CM |
2196 | BUG_ON(ret); |
2197 | } | |
0b86a832 | 2198 | kfree(chunk); |
2b82032c YZ |
2199 | return 0; |
2200 | } | |
0b86a832 | 2201 | |
2b82032c YZ |
2202 | /* |
2203 | * Chunk allocation falls into two parts. The first part does works | |
2204 | * that make the new allocated chunk useable, but not do any operation | |
2205 | * that modifies the chunk tree. The second part does the works that | |
2206 | * require modifying the chunk tree. This division is important for the | |
2207 | * bootstrap process of adding storage to a seed btrfs. | |
2208 | */ | |
2209 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, | |
2210 | struct btrfs_root *extent_root, u64 type) | |
2211 | { | |
2212 | u64 chunk_offset; | |
2213 | u64 chunk_size; | |
2214 | u64 stripe_size; | |
2215 | struct map_lookup *map; | |
2216 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; | |
2217 | int ret; | |
2218 | ||
2219 | ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID, | |
2220 | &chunk_offset); | |
2221 | if (ret) | |
2222 | return ret; | |
2223 | ||
2224 | ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size, | |
2225 | &stripe_size, chunk_offset, type); | |
2226 | if (ret) | |
2227 | return ret; | |
2228 | ||
2229 | ret = __finish_chunk_alloc(trans, extent_root, map, chunk_offset, | |
2230 | chunk_size, stripe_size); | |
2231 | BUG_ON(ret); | |
2232 | return 0; | |
2233 | } | |
2234 | ||
2235 | static int noinline init_first_rw_device(struct btrfs_trans_handle *trans, | |
2236 | struct btrfs_root *root, | |
2237 | struct btrfs_device *device) | |
2238 | { | |
2239 | u64 chunk_offset; | |
2240 | u64 sys_chunk_offset; | |
2241 | u64 chunk_size; | |
2242 | u64 sys_chunk_size; | |
2243 | u64 stripe_size; | |
2244 | u64 sys_stripe_size; | |
2245 | u64 alloc_profile; | |
2246 | struct map_lookup *map; | |
2247 | struct map_lookup *sys_map; | |
2248 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2249 | struct btrfs_root *extent_root = fs_info->extent_root; | |
2250 | int ret; | |
2251 | ||
2252 | ret = find_next_chunk(fs_info->chunk_root, | |
2253 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, &chunk_offset); | |
2254 | BUG_ON(ret); | |
2255 | ||
2256 | alloc_profile = BTRFS_BLOCK_GROUP_METADATA | | |
2257 | (fs_info->metadata_alloc_profile & | |
2258 | fs_info->avail_metadata_alloc_bits); | |
2259 | alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile); | |
2260 | ||
2261 | ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size, | |
2262 | &stripe_size, chunk_offset, alloc_profile); | |
2263 | BUG_ON(ret); | |
2264 | ||
2265 | sys_chunk_offset = chunk_offset + chunk_size; | |
2266 | ||
2267 | alloc_profile = BTRFS_BLOCK_GROUP_SYSTEM | | |
2268 | (fs_info->system_alloc_profile & | |
2269 | fs_info->avail_system_alloc_bits); | |
2270 | alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile); | |
2271 | ||
2272 | ret = __btrfs_alloc_chunk(trans, extent_root, &sys_map, | |
2273 | &sys_chunk_size, &sys_stripe_size, | |
2274 | sys_chunk_offset, alloc_profile); | |
2275 | BUG_ON(ret); | |
2276 | ||
2277 | ret = btrfs_add_device(trans, fs_info->chunk_root, device); | |
2278 | BUG_ON(ret); | |
2279 | ||
2280 | /* | |
2281 | * Modifying chunk tree needs allocating new blocks from both | |
2282 | * system block group and metadata block group. So we only can | |
2283 | * do operations require modifying the chunk tree after both | |
2284 | * block groups were created. | |
2285 | */ | |
2286 | ret = __finish_chunk_alloc(trans, extent_root, map, chunk_offset, | |
2287 | chunk_size, stripe_size); | |
2288 | BUG_ON(ret); | |
2289 | ||
2290 | ret = __finish_chunk_alloc(trans, extent_root, sys_map, | |
2291 | sys_chunk_offset, sys_chunk_size, | |
2292 | sys_stripe_size); | |
b248a415 | 2293 | BUG_ON(ret); |
2b82032c YZ |
2294 | return 0; |
2295 | } | |
2296 | ||
2297 | int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset) | |
2298 | { | |
2299 | struct extent_map *em; | |
2300 | struct map_lookup *map; | |
2301 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
2302 | int readonly = 0; | |
2303 | int i; | |
2304 | ||
2305 | spin_lock(&map_tree->map_tree.lock); | |
2306 | em = lookup_extent_mapping(&map_tree->map_tree, chunk_offset, 1); | |
2307 | spin_unlock(&map_tree->map_tree.lock); | |
2308 | if (!em) | |
2309 | return 1; | |
2310 | ||
2311 | map = (struct map_lookup *)em->bdev; | |
2312 | for (i = 0; i < map->num_stripes; i++) { | |
2313 | if (!map->stripes[i].dev->writeable) { | |
2314 | readonly = 1; | |
2315 | break; | |
2316 | } | |
2317 | } | |
0b86a832 | 2318 | free_extent_map(em); |
2b82032c | 2319 | return readonly; |
0b86a832 CM |
2320 | } |
2321 | ||
2322 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
2323 | { | |
2324 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
2325 | } | |
2326 | ||
2327 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
2328 | { | |
2329 | struct extent_map *em; | |
2330 | ||
2331 | while(1) { | |
2332 | spin_lock(&tree->map_tree.lock); | |
2333 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
2334 | if (em) | |
2335 | remove_extent_mapping(&tree->map_tree, em); | |
2336 | spin_unlock(&tree->map_tree.lock); | |
2337 | if (!em) | |
2338 | break; | |
2339 | kfree(em->bdev); | |
2340 | /* once for us */ | |
2341 | free_extent_map(em); | |
2342 | /* once for the tree */ | |
2343 | free_extent_map(em); | |
2344 | } | |
2345 | } | |
2346 | ||
f188591e CM |
2347 | int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len) |
2348 | { | |
2349 | struct extent_map *em; | |
2350 | struct map_lookup *map; | |
2351 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
2352 | int ret; | |
2353 | ||
2354 | spin_lock(&em_tree->lock); | |
2355 | em = lookup_extent_mapping(em_tree, logical, len); | |
b248a415 | 2356 | spin_unlock(&em_tree->lock); |
f188591e CM |
2357 | BUG_ON(!em); |
2358 | ||
2359 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
2360 | map = (struct map_lookup *)em->bdev; | |
2361 | if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1)) | |
2362 | ret = map->num_stripes; | |
321aecc6 CM |
2363 | else if (map->type & BTRFS_BLOCK_GROUP_RAID10) |
2364 | ret = map->sub_stripes; | |
f188591e CM |
2365 | else |
2366 | ret = 1; | |
2367 | free_extent_map(em); | |
f188591e CM |
2368 | return ret; |
2369 | } | |
2370 | ||
dfe25020 CM |
2371 | static int find_live_mirror(struct map_lookup *map, int first, int num, |
2372 | int optimal) | |
2373 | { | |
2374 | int i; | |
2375 | if (map->stripes[optimal].dev->bdev) | |
2376 | return optimal; | |
2377 | for (i = first; i < first + num; i++) { | |
2378 | if (map->stripes[i].dev->bdev) | |
2379 | return i; | |
2380 | } | |
2381 | /* we couldn't find one that doesn't fail. Just return something | |
2382 | * and the io error handling code will clean up eventually | |
2383 | */ | |
2384 | return optimal; | |
2385 | } | |
2386 | ||
f2d8d74d CM |
2387 | static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
2388 | u64 logical, u64 *length, | |
2389 | struct btrfs_multi_bio **multi_ret, | |
2390 | int mirror_num, struct page *unplug_page) | |
0b86a832 CM |
2391 | { |
2392 | struct extent_map *em; | |
2393 | struct map_lookup *map; | |
2394 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
2395 | u64 offset; | |
593060d7 CM |
2396 | u64 stripe_offset; |
2397 | u64 stripe_nr; | |
cea9e445 | 2398 | int stripes_allocated = 8; |
321aecc6 | 2399 | int stripes_required = 1; |
593060d7 | 2400 | int stripe_index; |
cea9e445 | 2401 | int i; |
f2d8d74d | 2402 | int num_stripes; |
a236aed1 | 2403 | int max_errors = 0; |
cea9e445 | 2404 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 | 2405 | |
cea9e445 CM |
2406 | if (multi_ret && !(rw & (1 << BIO_RW))) { |
2407 | stripes_allocated = 1; | |
2408 | } | |
2409 | again: | |
2410 | if (multi_ret) { | |
2411 | multi = kzalloc(btrfs_multi_bio_size(stripes_allocated), | |
2412 | GFP_NOFS); | |
2413 | if (!multi) | |
2414 | return -ENOMEM; | |
a236aed1 CM |
2415 | |
2416 | atomic_set(&multi->error, 0); | |
cea9e445 | 2417 | } |
0b86a832 CM |
2418 | |
2419 | spin_lock(&em_tree->lock); | |
2420 | em = lookup_extent_mapping(em_tree, logical, *length); | |
b248a415 | 2421 | spin_unlock(&em_tree->lock); |
f2d8d74d CM |
2422 | |
2423 | if (!em && unplug_page) | |
2424 | return 0; | |
2425 | ||
3b951516 | 2426 | if (!em) { |
a061fc8d | 2427 | printk("unable to find logical %Lu len %Lu\n", logical, *length); |
f2d8d74d | 2428 | BUG(); |
3b951516 | 2429 | } |
0b86a832 CM |
2430 | |
2431 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
2432 | map = (struct map_lookup *)em->bdev; | |
2433 | offset = logical - em->start; | |
593060d7 | 2434 | |
f188591e CM |
2435 | if (mirror_num > map->num_stripes) |
2436 | mirror_num = 0; | |
2437 | ||
cea9e445 | 2438 | /* if our multi bio struct is too small, back off and try again */ |
321aecc6 CM |
2439 | if (rw & (1 << BIO_RW)) { |
2440 | if (map->type & (BTRFS_BLOCK_GROUP_RAID1 | | |
2441 | BTRFS_BLOCK_GROUP_DUP)) { | |
2442 | stripes_required = map->num_stripes; | |
a236aed1 | 2443 | max_errors = 1; |
321aecc6 CM |
2444 | } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { |
2445 | stripes_required = map->sub_stripes; | |
a236aed1 | 2446 | max_errors = 1; |
321aecc6 CM |
2447 | } |
2448 | } | |
2449 | if (multi_ret && rw == WRITE && | |
2450 | stripes_allocated < stripes_required) { | |
cea9e445 | 2451 | stripes_allocated = map->num_stripes; |
cea9e445 CM |
2452 | free_extent_map(em); |
2453 | kfree(multi); | |
2454 | goto again; | |
2455 | } | |
593060d7 CM |
2456 | stripe_nr = offset; |
2457 | /* | |
2458 | * stripe_nr counts the total number of stripes we have to stride | |
2459 | * to get to this block | |
2460 | */ | |
2461 | do_div(stripe_nr, map->stripe_len); | |
2462 | ||
2463 | stripe_offset = stripe_nr * map->stripe_len; | |
2464 | BUG_ON(offset < stripe_offset); | |
2465 | ||
2466 | /* stripe_offset is the offset of this block in its stripe*/ | |
2467 | stripe_offset = offset - stripe_offset; | |
2468 | ||
cea9e445 | 2469 | if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 | |
321aecc6 | 2470 | BTRFS_BLOCK_GROUP_RAID10 | |
cea9e445 CM |
2471 | BTRFS_BLOCK_GROUP_DUP)) { |
2472 | /* we limit the length of each bio to what fits in a stripe */ | |
2473 | *length = min_t(u64, em->len - offset, | |
2474 | map->stripe_len - stripe_offset); | |
2475 | } else { | |
2476 | *length = em->len - offset; | |
2477 | } | |
f2d8d74d CM |
2478 | |
2479 | if (!multi_ret && !unplug_page) | |
cea9e445 CM |
2480 | goto out; |
2481 | ||
f2d8d74d | 2482 | num_stripes = 1; |
cea9e445 | 2483 | stripe_index = 0; |
8790d502 | 2484 | if (map->type & BTRFS_BLOCK_GROUP_RAID1) { |
f2d8d74d CM |
2485 | if (unplug_page || (rw & (1 << BIO_RW))) |
2486 | num_stripes = map->num_stripes; | |
2fff734f | 2487 | else if (mirror_num) |
f188591e | 2488 | stripe_index = mirror_num - 1; |
dfe25020 CM |
2489 | else { |
2490 | stripe_index = find_live_mirror(map, 0, | |
2491 | map->num_stripes, | |
2492 | current->pid % map->num_stripes); | |
2493 | } | |
2fff734f | 2494 | |
611f0e00 | 2495 | } else if (map->type & BTRFS_BLOCK_GROUP_DUP) { |
cea9e445 | 2496 | if (rw & (1 << BIO_RW)) |
f2d8d74d | 2497 | num_stripes = map->num_stripes; |
f188591e CM |
2498 | else if (mirror_num) |
2499 | stripe_index = mirror_num - 1; | |
2fff734f | 2500 | |
321aecc6 CM |
2501 | } else if (map->type & BTRFS_BLOCK_GROUP_RAID10) { |
2502 | int factor = map->num_stripes / map->sub_stripes; | |
321aecc6 CM |
2503 | |
2504 | stripe_index = do_div(stripe_nr, factor); | |
2505 | stripe_index *= map->sub_stripes; | |
2506 | ||
f2d8d74d CM |
2507 | if (unplug_page || (rw & (1 << BIO_RW))) |
2508 | num_stripes = map->sub_stripes; | |
321aecc6 CM |
2509 | else if (mirror_num) |
2510 | stripe_index += mirror_num - 1; | |
dfe25020 CM |
2511 | else { |
2512 | stripe_index = find_live_mirror(map, stripe_index, | |
2513 | map->sub_stripes, stripe_index + | |
2514 | current->pid % map->sub_stripes); | |
2515 | } | |
8790d502 CM |
2516 | } else { |
2517 | /* | |
2518 | * after this do_div call, stripe_nr is the number of stripes | |
2519 | * on this device we have to walk to find the data, and | |
2520 | * stripe_index is the number of our device in the stripe array | |
2521 | */ | |
2522 | stripe_index = do_div(stripe_nr, map->num_stripes); | |
2523 | } | |
593060d7 | 2524 | BUG_ON(stripe_index >= map->num_stripes); |
cea9e445 | 2525 | |
f2d8d74d CM |
2526 | for (i = 0; i < num_stripes; i++) { |
2527 | if (unplug_page) { | |
2528 | struct btrfs_device *device; | |
2529 | struct backing_dev_info *bdi; | |
2530 | ||
2531 | device = map->stripes[stripe_index].dev; | |
dfe25020 CM |
2532 | if (device->bdev) { |
2533 | bdi = blk_get_backing_dev_info(device->bdev); | |
2534 | if (bdi->unplug_io_fn) { | |
2535 | bdi->unplug_io_fn(bdi, unplug_page); | |
2536 | } | |
f2d8d74d CM |
2537 | } |
2538 | } else { | |
2539 | multi->stripes[i].physical = | |
2540 | map->stripes[stripe_index].physical + | |
2541 | stripe_offset + stripe_nr * map->stripe_len; | |
2542 | multi->stripes[i].dev = map->stripes[stripe_index].dev; | |
2543 | } | |
cea9e445 | 2544 | stripe_index++; |
593060d7 | 2545 | } |
f2d8d74d CM |
2546 | if (multi_ret) { |
2547 | *multi_ret = multi; | |
2548 | multi->num_stripes = num_stripes; | |
a236aed1 | 2549 | multi->max_errors = max_errors; |
f2d8d74d | 2550 | } |
cea9e445 | 2551 | out: |
0b86a832 | 2552 | free_extent_map(em); |
0b86a832 CM |
2553 | return 0; |
2554 | } | |
2555 | ||
f2d8d74d CM |
2556 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw, |
2557 | u64 logical, u64 *length, | |
2558 | struct btrfs_multi_bio **multi_ret, int mirror_num) | |
2559 | { | |
2560 | return __btrfs_map_block(map_tree, rw, logical, length, multi_ret, | |
2561 | mirror_num, NULL); | |
2562 | } | |
2563 | ||
2564 | int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree, | |
2565 | u64 logical, struct page *page) | |
2566 | { | |
2567 | u64 length = PAGE_CACHE_SIZE; | |
2568 | return __btrfs_map_block(map_tree, READ, logical, &length, | |
2569 | NULL, 0, page); | |
2570 | } | |
2571 | ||
2572 | ||
8790d502 | 2573 | static void end_bio_multi_stripe(struct bio *bio, int err) |
8790d502 | 2574 | { |
cea9e445 | 2575 | struct btrfs_multi_bio *multi = bio->bi_private; |
7d2b4daa | 2576 | int is_orig_bio = 0; |
8790d502 | 2577 | |
8790d502 | 2578 | if (err) |
a236aed1 | 2579 | atomic_inc(&multi->error); |
8790d502 | 2580 | |
7d2b4daa CM |
2581 | if (bio == multi->orig_bio) |
2582 | is_orig_bio = 1; | |
2583 | ||
cea9e445 | 2584 | if (atomic_dec_and_test(&multi->stripes_pending)) { |
7d2b4daa CM |
2585 | if (!is_orig_bio) { |
2586 | bio_put(bio); | |
2587 | bio = multi->orig_bio; | |
2588 | } | |
8790d502 CM |
2589 | bio->bi_private = multi->private; |
2590 | bio->bi_end_io = multi->end_io; | |
a236aed1 CM |
2591 | /* only send an error to the higher layers if it is |
2592 | * beyond the tolerance of the multi-bio | |
2593 | */ | |
1259ab75 | 2594 | if (atomic_read(&multi->error) > multi->max_errors) { |
a236aed1 | 2595 | err = -EIO; |
1259ab75 CM |
2596 | } else if (err) { |
2597 | /* | |
2598 | * this bio is actually up to date, we didn't | |
2599 | * go over the max number of errors | |
2600 | */ | |
2601 | set_bit(BIO_UPTODATE, &bio->bi_flags); | |
a236aed1 | 2602 | err = 0; |
1259ab75 | 2603 | } |
8790d502 CM |
2604 | kfree(multi); |
2605 | ||
2606 | bio_endio(bio, err); | |
7d2b4daa | 2607 | } else if (!is_orig_bio) { |
8790d502 CM |
2608 | bio_put(bio); |
2609 | } | |
8790d502 CM |
2610 | } |
2611 | ||
8b712842 CM |
2612 | struct async_sched { |
2613 | struct bio *bio; | |
2614 | int rw; | |
2615 | struct btrfs_fs_info *info; | |
2616 | struct btrfs_work work; | |
2617 | }; | |
2618 | ||
2619 | /* | |
2620 | * see run_scheduled_bios for a description of why bios are collected for | |
2621 | * async submit. | |
2622 | * | |
2623 | * This will add one bio to the pending list for a device and make sure | |
2624 | * the work struct is scheduled. | |
2625 | */ | |
a1b32a59 CM |
2626 | static int noinline schedule_bio(struct btrfs_root *root, |
2627 | struct btrfs_device *device, | |
2628 | int rw, struct bio *bio) | |
8b712842 CM |
2629 | { |
2630 | int should_queue = 1; | |
2631 | ||
2632 | /* don't bother with additional async steps for reads, right now */ | |
2633 | if (!(rw & (1 << BIO_RW))) { | |
492bb6de | 2634 | bio_get(bio); |
8b712842 | 2635 | submit_bio(rw, bio); |
492bb6de | 2636 | bio_put(bio); |
8b712842 CM |
2637 | return 0; |
2638 | } | |
2639 | ||
2640 | /* | |
0986fe9e | 2641 | * nr_async_bios allows us to reliably return congestion to the |
8b712842 CM |
2642 | * higher layers. Otherwise, the async bio makes it appear we have |
2643 | * made progress against dirty pages when we've really just put it | |
2644 | * on a queue for later | |
2645 | */ | |
0986fe9e | 2646 | atomic_inc(&root->fs_info->nr_async_bios); |
492bb6de | 2647 | WARN_ON(bio->bi_next); |
8b712842 CM |
2648 | bio->bi_next = NULL; |
2649 | bio->bi_rw |= rw; | |
2650 | ||
2651 | spin_lock(&device->io_lock); | |
2652 | ||
2653 | if (device->pending_bio_tail) | |
2654 | device->pending_bio_tail->bi_next = bio; | |
2655 | ||
2656 | device->pending_bio_tail = bio; | |
2657 | if (!device->pending_bios) | |
2658 | device->pending_bios = bio; | |
2659 | if (device->running_pending) | |
2660 | should_queue = 0; | |
2661 | ||
2662 | spin_unlock(&device->io_lock); | |
2663 | ||
2664 | if (should_queue) | |
1cc127b5 CM |
2665 | btrfs_queue_worker(&root->fs_info->submit_workers, |
2666 | &device->work); | |
8b712842 CM |
2667 | return 0; |
2668 | } | |
2669 | ||
f188591e | 2670 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio, |
8b712842 | 2671 | int mirror_num, int async_submit) |
0b86a832 CM |
2672 | { |
2673 | struct btrfs_mapping_tree *map_tree; | |
2674 | struct btrfs_device *dev; | |
8790d502 | 2675 | struct bio *first_bio = bio; |
a62b9401 | 2676 | u64 logical = (u64)bio->bi_sector << 9; |
0b86a832 CM |
2677 | u64 length = 0; |
2678 | u64 map_length; | |
cea9e445 | 2679 | struct btrfs_multi_bio *multi = NULL; |
0b86a832 | 2680 | int ret; |
8790d502 CM |
2681 | int dev_nr = 0; |
2682 | int total_devs = 1; | |
0b86a832 | 2683 | |
f2d8d74d | 2684 | length = bio->bi_size; |
0b86a832 CM |
2685 | map_tree = &root->fs_info->mapping_tree; |
2686 | map_length = length; | |
cea9e445 | 2687 | |
f188591e CM |
2688 | ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi, |
2689 | mirror_num); | |
cea9e445 CM |
2690 | BUG_ON(ret); |
2691 | ||
2692 | total_devs = multi->num_stripes; | |
2693 | if (map_length < length) { | |
2694 | printk("mapping failed logical %Lu bio len %Lu " | |
2695 | "len %Lu\n", logical, length, map_length); | |
2696 | BUG(); | |
2697 | } | |
2698 | multi->end_io = first_bio->bi_end_io; | |
2699 | multi->private = first_bio->bi_private; | |
7d2b4daa | 2700 | multi->orig_bio = first_bio; |
cea9e445 CM |
2701 | atomic_set(&multi->stripes_pending, multi->num_stripes); |
2702 | ||
8790d502 | 2703 | while(dev_nr < total_devs) { |
8790d502 | 2704 | if (total_devs > 1) { |
8790d502 CM |
2705 | if (dev_nr < total_devs - 1) { |
2706 | bio = bio_clone(first_bio, GFP_NOFS); | |
2707 | BUG_ON(!bio); | |
2708 | } else { | |
2709 | bio = first_bio; | |
2710 | } | |
2711 | bio->bi_private = multi; | |
2712 | bio->bi_end_io = end_bio_multi_stripe; | |
2713 | } | |
cea9e445 CM |
2714 | bio->bi_sector = multi->stripes[dev_nr].physical >> 9; |
2715 | dev = multi->stripes[dev_nr].dev; | |
2b82032c | 2716 | BUG_ON(rw == WRITE && !dev->writeable); |
dfe25020 CM |
2717 | if (dev && dev->bdev) { |
2718 | bio->bi_bdev = dev->bdev; | |
8b712842 CM |
2719 | if (async_submit) |
2720 | schedule_bio(root, dev, rw, bio); | |
2721 | else | |
2722 | submit_bio(rw, bio); | |
dfe25020 CM |
2723 | } else { |
2724 | bio->bi_bdev = root->fs_info->fs_devices->latest_bdev; | |
2725 | bio->bi_sector = logical >> 9; | |
dfe25020 | 2726 | bio_endio(bio, -EIO); |
dfe25020 | 2727 | } |
8790d502 CM |
2728 | dev_nr++; |
2729 | } | |
cea9e445 CM |
2730 | if (total_devs == 1) |
2731 | kfree(multi); | |
0b86a832 CM |
2732 | return 0; |
2733 | } | |
2734 | ||
a443755f | 2735 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid, |
2b82032c | 2736 | u8 *uuid, u8 *fsid) |
0b86a832 | 2737 | { |
2b82032c YZ |
2738 | struct btrfs_device *device; |
2739 | struct btrfs_fs_devices *cur_devices; | |
2740 | ||
2741 | cur_devices = root->fs_info->fs_devices; | |
2742 | while (cur_devices) { | |
2743 | if (!fsid || | |
2744 | !memcmp(cur_devices->fsid, fsid, BTRFS_UUID_SIZE)) { | |
2745 | device = __find_device(&cur_devices->devices, | |
2746 | devid, uuid); | |
2747 | if (device) | |
2748 | return device; | |
2749 | } | |
2750 | cur_devices = cur_devices->seed; | |
2751 | } | |
2752 | return NULL; | |
0b86a832 CM |
2753 | } |
2754 | ||
dfe25020 CM |
2755 | static struct btrfs_device *add_missing_dev(struct btrfs_root *root, |
2756 | u64 devid, u8 *dev_uuid) | |
2757 | { | |
2758 | struct btrfs_device *device; | |
2759 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
2760 | ||
2761 | device = kzalloc(sizeof(*device), GFP_NOFS); | |
7cbd8a83 | 2762 | if (!device) |
2763 | return NULL; | |
dfe25020 CM |
2764 | list_add(&device->dev_list, |
2765 | &fs_devices->devices); | |
dfe25020 CM |
2766 | device->barriers = 1; |
2767 | device->dev_root = root->fs_info->dev_root; | |
2768 | device->devid = devid; | |
8b712842 | 2769 | device->work.func = pending_bios_fn; |
dfe25020 CM |
2770 | fs_devices->num_devices++; |
2771 | spin_lock_init(&device->io_lock); | |
2772 | memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE); | |
2773 | return device; | |
2774 | } | |
2775 | ||
0b86a832 CM |
2776 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, |
2777 | struct extent_buffer *leaf, | |
2778 | struct btrfs_chunk *chunk) | |
2779 | { | |
2780 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
2781 | struct map_lookup *map; | |
2782 | struct extent_map *em; | |
2783 | u64 logical; | |
2784 | u64 length; | |
2785 | u64 devid; | |
a443755f | 2786 | u8 uuid[BTRFS_UUID_SIZE]; |
593060d7 | 2787 | int num_stripes; |
0b86a832 | 2788 | int ret; |
593060d7 | 2789 | int i; |
0b86a832 | 2790 | |
e17cade2 CM |
2791 | logical = key->offset; |
2792 | length = btrfs_chunk_length(leaf, chunk); | |
a061fc8d | 2793 | |
0b86a832 CM |
2794 | spin_lock(&map_tree->map_tree.lock); |
2795 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
b248a415 | 2796 | spin_unlock(&map_tree->map_tree.lock); |
0b86a832 CM |
2797 | |
2798 | /* already mapped? */ | |
2799 | if (em && em->start <= logical && em->start + em->len > logical) { | |
2800 | free_extent_map(em); | |
0b86a832 CM |
2801 | return 0; |
2802 | } else if (em) { | |
2803 | free_extent_map(em); | |
2804 | } | |
0b86a832 CM |
2805 | |
2806 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
2807 | if (!map) | |
2808 | return -ENOMEM; | |
2809 | ||
2810 | em = alloc_extent_map(GFP_NOFS); | |
2811 | if (!em) | |
2812 | return -ENOMEM; | |
593060d7 CM |
2813 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); |
2814 | map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS); | |
0b86a832 CM |
2815 | if (!map) { |
2816 | free_extent_map(em); | |
2817 | return -ENOMEM; | |
2818 | } | |
2819 | ||
2820 | em->bdev = (struct block_device *)map; | |
2821 | em->start = logical; | |
2822 | em->len = length; | |
2823 | em->block_start = 0; | |
c8b97818 | 2824 | em->block_len = em->len; |
0b86a832 | 2825 | |
593060d7 CM |
2826 | map->num_stripes = num_stripes; |
2827 | map->io_width = btrfs_chunk_io_width(leaf, chunk); | |
2828 | map->io_align = btrfs_chunk_io_align(leaf, chunk); | |
2829 | map->sector_size = btrfs_chunk_sector_size(leaf, chunk); | |
2830 | map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
2831 | map->type = btrfs_chunk_type(leaf, chunk); | |
321aecc6 | 2832 | map->sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk); |
593060d7 CM |
2833 | for (i = 0; i < num_stripes; i++) { |
2834 | map->stripes[i].physical = | |
2835 | btrfs_stripe_offset_nr(leaf, chunk, i); | |
2836 | devid = btrfs_stripe_devid_nr(leaf, chunk, i); | |
a443755f CM |
2837 | read_extent_buffer(leaf, uuid, (unsigned long) |
2838 | btrfs_stripe_dev_uuid_nr(chunk, i), | |
2839 | BTRFS_UUID_SIZE); | |
2b82032c YZ |
2840 | map->stripes[i].dev = btrfs_find_device(root, devid, uuid, |
2841 | NULL); | |
dfe25020 | 2842 | if (!map->stripes[i].dev && !btrfs_test_opt(root, DEGRADED)) { |
593060d7 CM |
2843 | kfree(map); |
2844 | free_extent_map(em); | |
2845 | return -EIO; | |
2846 | } | |
dfe25020 CM |
2847 | if (!map->stripes[i].dev) { |
2848 | map->stripes[i].dev = | |
2849 | add_missing_dev(root, devid, uuid); | |
2850 | if (!map->stripes[i].dev) { | |
2851 | kfree(map); | |
2852 | free_extent_map(em); | |
2853 | return -EIO; | |
2854 | } | |
2855 | } | |
2856 | map->stripes[i].dev->in_fs_metadata = 1; | |
0b86a832 CM |
2857 | } |
2858 | ||
2859 | spin_lock(&map_tree->map_tree.lock); | |
2860 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
0b86a832 | 2861 | spin_unlock(&map_tree->map_tree.lock); |
b248a415 | 2862 | BUG_ON(ret); |
0b86a832 CM |
2863 | free_extent_map(em); |
2864 | ||
2865 | return 0; | |
2866 | } | |
2867 | ||
2868 | static int fill_device_from_item(struct extent_buffer *leaf, | |
2869 | struct btrfs_dev_item *dev_item, | |
2870 | struct btrfs_device *device) | |
2871 | { | |
2872 | unsigned long ptr; | |
0b86a832 CM |
2873 | |
2874 | device->devid = btrfs_device_id(leaf, dev_item); | |
2875 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
2876 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
2877 | device->type = btrfs_device_type(leaf, dev_item); | |
2878 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
2879 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
2880 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
0b86a832 CM |
2881 | |
2882 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
e17cade2 | 2883 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE); |
0b86a832 | 2884 | |
0b86a832 CM |
2885 | return 0; |
2886 | } | |
2887 | ||
2b82032c YZ |
2888 | static int open_seed_devices(struct btrfs_root *root, u8 *fsid) |
2889 | { | |
2890 | struct btrfs_fs_devices *fs_devices; | |
2891 | int ret; | |
2892 | ||
2893 | mutex_lock(&uuid_mutex); | |
2894 | ||
2895 | fs_devices = root->fs_info->fs_devices->seed; | |
2896 | while (fs_devices) { | |
2897 | if (!memcmp(fs_devices->fsid, fsid, BTRFS_UUID_SIZE)) { | |
2898 | ret = 0; | |
2899 | goto out; | |
2900 | } | |
2901 | fs_devices = fs_devices->seed; | |
2902 | } | |
2903 | ||
2904 | fs_devices = find_fsid(fsid); | |
2905 | if (!fs_devices) { | |
2906 | ret = -ENOENT; | |
2907 | goto out; | |
2908 | } | |
2909 | if (fs_devices->opened) { | |
2910 | ret = -EBUSY; | |
2911 | goto out; | |
2912 | } | |
2913 | ||
15916de8 CM |
2914 | ret = __btrfs_open_devices(fs_devices, MS_RDONLY, |
2915 | root->fs_info->bdev_holder); | |
2b82032c YZ |
2916 | if (ret) |
2917 | goto out; | |
2918 | ||
2919 | if (!fs_devices->seeding) { | |
2920 | __btrfs_close_devices(fs_devices); | |
2921 | ret = -EINVAL; | |
2922 | goto out; | |
2923 | } | |
2924 | ||
2925 | fs_devices->seed = root->fs_info->fs_devices->seed; | |
2926 | root->fs_info->fs_devices->seed = fs_devices; | |
2927 | fs_devices->sprouted = 1; | |
2928 | out: | |
2929 | mutex_unlock(&uuid_mutex); | |
2930 | return ret; | |
2931 | } | |
2932 | ||
0d81ba5d | 2933 | static int read_one_dev(struct btrfs_root *root, |
0b86a832 CM |
2934 | struct extent_buffer *leaf, |
2935 | struct btrfs_dev_item *dev_item) | |
2936 | { | |
2937 | struct btrfs_device *device; | |
2938 | u64 devid; | |
2939 | int ret; | |
2b82032c YZ |
2940 | int seed_devices = 0; |
2941 | u8 fs_uuid[BTRFS_UUID_SIZE]; | |
a443755f CM |
2942 | u8 dev_uuid[BTRFS_UUID_SIZE]; |
2943 | ||
0b86a832 | 2944 | devid = btrfs_device_id(leaf, dev_item); |
a443755f CM |
2945 | read_extent_buffer(leaf, dev_uuid, |
2946 | (unsigned long)btrfs_device_uuid(dev_item), | |
2947 | BTRFS_UUID_SIZE); | |
2b82032c YZ |
2948 | read_extent_buffer(leaf, fs_uuid, |
2949 | (unsigned long)btrfs_device_fsid(dev_item), | |
2950 | BTRFS_UUID_SIZE); | |
2951 | ||
2952 | if (memcmp(fs_uuid, root->fs_info->fsid, BTRFS_UUID_SIZE)) { | |
2953 | ret = open_seed_devices(root, fs_uuid); | |
2954 | if (ret) | |
2955 | return ret; | |
2956 | seed_devices = 1; | |
2957 | } | |
2958 | ||
2959 | device = btrfs_find_device(root, devid, dev_uuid, fs_uuid); | |
2960 | if (!device || !device->bdev) { | |
2961 | if (!btrfs_test_opt(root, DEGRADED) || seed_devices) | |
2962 | return -EIO; | |
2963 | ||
2964 | if (!device) { | |
2965 | printk("warning devid %Lu missing\n", devid); | |
2966 | device = add_missing_dev(root, devid, dev_uuid); | |
2967 | if (!device) | |
2968 | return -ENOMEM; | |
2969 | } | |
2970 | } | |
2971 | ||
2972 | if (device->fs_devices != root->fs_info->fs_devices) { | |
2973 | BUG_ON(device->writeable); | |
2974 | if (device->generation != | |
2975 | btrfs_device_generation(leaf, dev_item)) | |
2976 | return -EINVAL; | |
6324fbf3 | 2977 | } |
0b86a832 CM |
2978 | |
2979 | fill_device_from_item(leaf, dev_item, device); | |
2980 | device->dev_root = root->fs_info->dev_root; | |
dfe25020 | 2981 | device->in_fs_metadata = 1; |
2b82032c YZ |
2982 | if (device->writeable) |
2983 | device->fs_devices->total_rw_bytes += device->total_bytes; | |
0b86a832 CM |
2984 | ret = 0; |
2985 | #if 0 | |
2986 | ret = btrfs_open_device(device); | |
2987 | if (ret) { | |
2988 | kfree(device); | |
2989 | } | |
2990 | #endif | |
2991 | return ret; | |
2992 | } | |
2993 | ||
0d81ba5d CM |
2994 | int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) |
2995 | { | |
2996 | struct btrfs_dev_item *dev_item; | |
2997 | ||
2998 | dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, | |
2999 | dev_item); | |
3000 | return read_one_dev(root, buf, dev_item); | |
3001 | } | |
3002 | ||
0b86a832 CM |
3003 | int btrfs_read_sys_array(struct btrfs_root *root) |
3004 | { | |
3005 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
a061fc8d | 3006 | struct extent_buffer *sb; |
0b86a832 | 3007 | struct btrfs_disk_key *disk_key; |
0b86a832 | 3008 | struct btrfs_chunk *chunk; |
84eed90f CM |
3009 | u8 *ptr; |
3010 | unsigned long sb_ptr; | |
3011 | int ret = 0; | |
0b86a832 CM |
3012 | u32 num_stripes; |
3013 | u32 array_size; | |
3014 | u32 len = 0; | |
0b86a832 | 3015 | u32 cur; |
84eed90f | 3016 | struct btrfs_key key; |
0b86a832 | 3017 | |
a061fc8d CM |
3018 | sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET, |
3019 | BTRFS_SUPER_INFO_SIZE); | |
3020 | if (!sb) | |
3021 | return -ENOMEM; | |
3022 | btrfs_set_buffer_uptodate(sb); | |
3023 | write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE); | |
0b86a832 CM |
3024 | array_size = btrfs_super_sys_array_size(super_copy); |
3025 | ||
0b86a832 CM |
3026 | ptr = super_copy->sys_chunk_array; |
3027 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
3028 | cur = 0; | |
3029 | ||
3030 | while (cur < array_size) { | |
3031 | disk_key = (struct btrfs_disk_key *)ptr; | |
3032 | btrfs_disk_key_to_cpu(&key, disk_key); | |
3033 | ||
a061fc8d | 3034 | len = sizeof(*disk_key); ptr += len; |
0b86a832 CM |
3035 | sb_ptr += len; |
3036 | cur += len; | |
3037 | ||
0d81ba5d | 3038 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { |
0b86a832 | 3039 | chunk = (struct btrfs_chunk *)sb_ptr; |
0d81ba5d | 3040 | ret = read_one_chunk(root, &key, sb, chunk); |
84eed90f CM |
3041 | if (ret) |
3042 | break; | |
0b86a832 CM |
3043 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); |
3044 | len = btrfs_chunk_item_size(num_stripes); | |
3045 | } else { | |
84eed90f CM |
3046 | ret = -EIO; |
3047 | break; | |
0b86a832 CM |
3048 | } |
3049 | ptr += len; | |
3050 | sb_ptr += len; | |
3051 | cur += len; | |
3052 | } | |
a061fc8d | 3053 | free_extent_buffer(sb); |
84eed90f | 3054 | return ret; |
0b86a832 CM |
3055 | } |
3056 | ||
3057 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
3058 | { | |
3059 | struct btrfs_path *path; | |
3060 | struct extent_buffer *leaf; | |
3061 | struct btrfs_key key; | |
3062 | struct btrfs_key found_key; | |
3063 | int ret; | |
3064 | int slot; | |
3065 | ||
3066 | root = root->fs_info->chunk_root; | |
3067 | ||
3068 | path = btrfs_alloc_path(); | |
3069 | if (!path) | |
3070 | return -ENOMEM; | |
3071 | ||
3072 | /* first we search for all of the device items, and then we | |
3073 | * read in all of the chunk items. This way we can create chunk | |
3074 | * mappings that reference all of the devices that are afound | |
3075 | */ | |
3076 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
3077 | key.offset = 0; | |
3078 | key.type = 0; | |
3079 | again: | |
3080 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
3081 | while(1) { | |
3082 | leaf = path->nodes[0]; | |
3083 | slot = path->slots[0]; | |
3084 | if (slot >= btrfs_header_nritems(leaf)) { | |
3085 | ret = btrfs_next_leaf(root, path); | |
3086 | if (ret == 0) | |
3087 | continue; | |
3088 | if (ret < 0) | |
3089 | goto error; | |
3090 | break; | |
3091 | } | |
3092 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
3093 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
3094 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
3095 | break; | |
3096 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
3097 | struct btrfs_dev_item *dev_item; | |
3098 | dev_item = btrfs_item_ptr(leaf, slot, | |
3099 | struct btrfs_dev_item); | |
0d81ba5d | 3100 | ret = read_one_dev(root, leaf, dev_item); |
2b82032c YZ |
3101 | if (ret) |
3102 | goto error; | |
0b86a832 CM |
3103 | } |
3104 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
3105 | struct btrfs_chunk *chunk; | |
3106 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
3107 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
2b82032c YZ |
3108 | if (ret) |
3109 | goto error; | |
0b86a832 CM |
3110 | } |
3111 | path->slots[0]++; | |
3112 | } | |
3113 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
3114 | key.objectid = 0; | |
3115 | btrfs_release_path(root, path); | |
3116 | goto again; | |
3117 | } | |
0b86a832 CM |
3118 | ret = 0; |
3119 | error: | |
2b82032c | 3120 | btrfs_free_path(path); |
0b86a832 CM |
3121 | return ret; |
3122 | } |