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