]>
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> |
0b86a832 CM |
21 | #include "ctree.h" |
22 | #include "extent_map.h" | |
23 | #include "disk-io.h" | |
24 | #include "transaction.h" | |
25 | #include "print-tree.h" | |
26 | #include "volumes.h" | |
27 | ||
28 | struct map_lookup { | |
29 | struct btrfs_device *dev; | |
30 | u64 physical; | |
31 | }; | |
8a4b83cc CM |
32 | static DEFINE_MUTEX(uuid_mutex); |
33 | static LIST_HEAD(fs_uuids); | |
34 | ||
35 | int btrfs_cleanup_fs_uuids(void) | |
36 | { | |
37 | struct btrfs_fs_devices *fs_devices; | |
38 | struct list_head *uuid_cur; | |
39 | struct list_head *devices_cur; | |
40 | struct btrfs_device *dev; | |
41 | ||
42 | list_for_each(uuid_cur, &fs_uuids) { | |
43 | fs_devices = list_entry(uuid_cur, struct btrfs_fs_devices, | |
44 | list); | |
45 | while(!list_empty(&fs_devices->devices)) { | |
46 | devices_cur = fs_devices->devices.next; | |
47 | dev = list_entry(devices_cur, struct btrfs_device, | |
48 | dev_list); | |
49 | printk("uuid cleanup finds %s\n", dev->name); | |
50 | if (dev->bdev) { | |
51 | printk("closing\n"); | |
52 | close_bdev_excl(dev->bdev); | |
53 | } | |
54 | list_del(&dev->dev_list); | |
55 | kfree(dev); | |
56 | } | |
57 | } | |
58 | return 0; | |
59 | } | |
60 | ||
61 | static struct btrfs_device *__find_device(struct list_head *head, u64 devid) | |
62 | { | |
63 | struct btrfs_device *dev; | |
64 | struct list_head *cur; | |
65 | ||
66 | list_for_each(cur, head) { | |
67 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
68 | if (dev->devid == devid) | |
69 | return dev; | |
70 | } | |
71 | return NULL; | |
72 | } | |
73 | ||
74 | static struct btrfs_fs_devices *find_fsid(u8 *fsid) | |
75 | { | |
76 | struct list_head *cur; | |
77 | struct btrfs_fs_devices *fs_devices; | |
78 | ||
79 | list_for_each(cur, &fs_uuids) { | |
80 | fs_devices = list_entry(cur, struct btrfs_fs_devices, list); | |
81 | if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0) | |
82 | return fs_devices; | |
83 | } | |
84 | return NULL; | |
85 | } | |
86 | ||
87 | static int device_list_add(const char *path, | |
88 | struct btrfs_super_block *disk_super, | |
89 | u64 devid, struct btrfs_fs_devices **fs_devices_ret) | |
90 | { | |
91 | struct btrfs_device *device; | |
92 | struct btrfs_fs_devices *fs_devices; | |
93 | u64 found_transid = btrfs_super_generation(disk_super); | |
94 | ||
95 | fs_devices = find_fsid(disk_super->fsid); | |
96 | if (!fs_devices) { | |
97 | fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS); | |
98 | if (!fs_devices) | |
99 | return -ENOMEM; | |
100 | INIT_LIST_HEAD(&fs_devices->devices); | |
101 | list_add(&fs_devices->list, &fs_uuids); | |
102 | memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE); | |
103 | fs_devices->latest_devid = devid; | |
104 | fs_devices->latest_trans = found_transid; | |
105 | fs_devices->lowest_devid = (u64)-1; | |
106 | fs_devices->num_devices = 0; | |
107 | device = NULL; | |
108 | } else { | |
109 | device = __find_device(&fs_devices->devices, devid); | |
110 | } | |
111 | if (!device) { | |
112 | device = kzalloc(sizeof(*device), GFP_NOFS); | |
113 | if (!device) { | |
114 | /* we can safely leave the fs_devices entry around */ | |
115 | return -ENOMEM; | |
116 | } | |
117 | device->devid = devid; | |
118 | device->name = kstrdup(path, GFP_NOFS); | |
119 | if (!device->name) { | |
120 | kfree(device); | |
121 | return -ENOMEM; | |
122 | } | |
123 | list_add(&device->dev_list, &fs_devices->devices); | |
124 | fs_devices->num_devices++; | |
125 | } | |
126 | ||
127 | if (found_transid > fs_devices->latest_trans) { | |
128 | fs_devices->latest_devid = devid; | |
129 | fs_devices->latest_trans = found_transid; | |
130 | } | |
131 | if (fs_devices->lowest_devid > devid) { | |
132 | fs_devices->lowest_devid = devid; | |
133 | printk("lowest devid now %Lu\n", devid); | |
134 | } | |
135 | *fs_devices_ret = fs_devices; | |
136 | return 0; | |
137 | } | |
138 | ||
139 | int btrfs_close_devices(struct btrfs_fs_devices *fs_devices) | |
140 | { | |
141 | struct list_head *head = &fs_devices->devices; | |
142 | struct list_head *cur; | |
143 | struct btrfs_device *device; | |
144 | ||
145 | mutex_lock(&uuid_mutex); | |
146 | list_for_each(cur, head) { | |
147 | device = list_entry(cur, struct btrfs_device, dev_list); | |
148 | if (device->bdev) { | |
149 | close_bdev_excl(device->bdev); | |
150 | printk("close devices closes %s\n", device->name); | |
151 | } | |
152 | device->bdev = NULL; | |
153 | } | |
154 | mutex_unlock(&uuid_mutex); | |
155 | return 0; | |
156 | } | |
157 | ||
158 | int btrfs_open_devices(struct btrfs_fs_devices *fs_devices, | |
159 | int flags, void *holder) | |
160 | { | |
161 | struct block_device *bdev; | |
162 | struct list_head *head = &fs_devices->devices; | |
163 | struct list_head *cur; | |
164 | struct btrfs_device *device; | |
165 | int ret; | |
166 | ||
167 | mutex_lock(&uuid_mutex); | |
168 | list_for_each(cur, head) { | |
169 | device = list_entry(cur, struct btrfs_device, dev_list); | |
170 | bdev = open_bdev_excl(device->name, flags, holder); | |
171 | printk("opening %s devid %Lu\n", device->name, device->devid); | |
172 | if (IS_ERR(bdev)) { | |
173 | printk("open %s failed\n", device->name); | |
174 | ret = PTR_ERR(bdev); | |
175 | goto fail; | |
176 | } | |
177 | if (device->devid == fs_devices->latest_devid) | |
178 | fs_devices->latest_bdev = bdev; | |
179 | if (device->devid == fs_devices->lowest_devid) { | |
180 | fs_devices->lowest_bdev = bdev; | |
181 | printk("lowest bdev %s\n", device->name); | |
182 | } | |
183 | device->bdev = bdev; | |
184 | } | |
185 | mutex_unlock(&uuid_mutex); | |
186 | return 0; | |
187 | fail: | |
188 | mutex_unlock(&uuid_mutex); | |
189 | btrfs_close_devices(fs_devices); | |
190 | return ret; | |
191 | } | |
192 | ||
193 | int btrfs_scan_one_device(const char *path, int flags, void *holder, | |
194 | struct btrfs_fs_devices **fs_devices_ret) | |
195 | { | |
196 | struct btrfs_super_block *disk_super; | |
197 | struct block_device *bdev; | |
198 | struct buffer_head *bh; | |
199 | int ret; | |
200 | u64 devid; | |
201 | ||
202 | mutex_lock(&uuid_mutex); | |
203 | ||
204 | printk("scan one opens %s\n", path); | |
205 | bdev = open_bdev_excl(path, flags, holder); | |
206 | ||
207 | if (IS_ERR(bdev)) { | |
208 | printk("open failed\n"); | |
209 | ret = PTR_ERR(bdev); | |
210 | goto error; | |
211 | } | |
212 | ||
213 | ret = set_blocksize(bdev, 4096); | |
214 | if (ret) | |
215 | goto error_close; | |
216 | bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
217 | if (!bh) { | |
218 | ret = -EIO; | |
219 | goto error_close; | |
220 | } | |
221 | disk_super = (struct btrfs_super_block *)bh->b_data; | |
222 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, | |
223 | sizeof(disk_super->magic))) { | |
224 | printk("no btrfs found on %s\n", path); | |
225 | ret = -ENOENT; | |
226 | goto error_brelse; | |
227 | } | |
228 | devid = le64_to_cpu(disk_super->dev_item.devid); | |
229 | printk("found device %Lu on %s\n", devid, path); | |
230 | ret = device_list_add(path, disk_super, devid, fs_devices_ret); | |
231 | ||
232 | error_brelse: | |
233 | brelse(bh); | |
234 | error_close: | |
235 | close_bdev_excl(bdev); | |
236 | printk("scan one closes bdev %s\n", path); | |
237 | error: | |
238 | mutex_unlock(&uuid_mutex); | |
239 | return ret; | |
240 | } | |
0b86a832 CM |
241 | |
242 | /* | |
243 | * this uses a pretty simple search, the expectation is that it is | |
244 | * called very infrequently and that a given device has a small number | |
245 | * of extents | |
246 | */ | |
247 | static int find_free_dev_extent(struct btrfs_trans_handle *trans, | |
248 | struct btrfs_device *device, | |
249 | struct btrfs_path *path, | |
250 | u64 num_bytes, u64 *start) | |
251 | { | |
252 | struct btrfs_key key; | |
253 | struct btrfs_root *root = device->dev_root; | |
254 | struct btrfs_dev_extent *dev_extent = NULL; | |
255 | u64 hole_size = 0; | |
256 | u64 last_byte = 0; | |
257 | u64 search_start = 0; | |
258 | u64 search_end = device->total_bytes; | |
259 | int ret; | |
260 | int slot = 0; | |
261 | int start_found; | |
262 | struct extent_buffer *l; | |
263 | ||
264 | start_found = 0; | |
265 | path->reada = 2; | |
266 | ||
267 | /* FIXME use last free of some kind */ | |
268 | ||
8a4b83cc CM |
269 | /* we don't want to overwrite the superblock on the drive, |
270 | * so we make sure to start at an offset of at least 1MB | |
271 | */ | |
272 | search_start = max((u64)1024 * 1024, search_start); | |
0b86a832 CM |
273 | key.objectid = device->devid; |
274 | key.offset = search_start; | |
275 | key.type = BTRFS_DEV_EXTENT_KEY; | |
276 | ret = btrfs_search_slot(trans, root, &key, path, 0, 0); | |
277 | if (ret < 0) | |
278 | goto error; | |
279 | ret = btrfs_previous_item(root, path, 0, key.type); | |
280 | if (ret < 0) | |
281 | goto error; | |
282 | l = path->nodes[0]; | |
283 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
284 | while (1) { | |
285 | l = path->nodes[0]; | |
286 | slot = path->slots[0]; | |
287 | if (slot >= btrfs_header_nritems(l)) { | |
288 | ret = btrfs_next_leaf(root, path); | |
289 | if (ret == 0) | |
290 | continue; | |
291 | if (ret < 0) | |
292 | goto error; | |
293 | no_more_items: | |
294 | if (!start_found) { | |
295 | if (search_start >= search_end) { | |
296 | ret = -ENOSPC; | |
297 | goto error; | |
298 | } | |
299 | *start = search_start; | |
300 | start_found = 1; | |
301 | goto check_pending; | |
302 | } | |
303 | *start = last_byte > search_start ? | |
304 | last_byte : search_start; | |
305 | if (search_end <= *start) { | |
306 | ret = -ENOSPC; | |
307 | goto error; | |
308 | } | |
309 | goto check_pending; | |
310 | } | |
311 | btrfs_item_key_to_cpu(l, &key, slot); | |
312 | ||
313 | if (key.objectid < device->devid) | |
314 | goto next; | |
315 | ||
316 | if (key.objectid > device->devid) | |
317 | goto no_more_items; | |
318 | ||
319 | if (key.offset >= search_start && key.offset > last_byte && | |
320 | start_found) { | |
321 | if (last_byte < search_start) | |
322 | last_byte = search_start; | |
323 | hole_size = key.offset - last_byte; | |
324 | if (key.offset > last_byte && | |
325 | hole_size >= num_bytes) { | |
326 | *start = last_byte; | |
327 | goto check_pending; | |
328 | } | |
329 | } | |
330 | if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) { | |
331 | goto next; | |
332 | } | |
333 | ||
334 | start_found = 1; | |
335 | dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent); | |
336 | last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent); | |
337 | next: | |
338 | path->slots[0]++; | |
339 | cond_resched(); | |
340 | } | |
341 | check_pending: | |
342 | /* we have to make sure we didn't find an extent that has already | |
343 | * been allocated by the map tree or the original allocation | |
344 | */ | |
345 | btrfs_release_path(root, path); | |
346 | BUG_ON(*start < search_start); | |
347 | ||
6324fbf3 | 348 | if (*start + num_bytes > search_end) { |
0b86a832 CM |
349 | ret = -ENOSPC; |
350 | goto error; | |
351 | } | |
352 | /* check for pending inserts here */ | |
353 | return 0; | |
354 | ||
355 | error: | |
356 | btrfs_release_path(root, path); | |
357 | return ret; | |
358 | } | |
359 | ||
360 | int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans, | |
361 | struct btrfs_device *device, | |
362 | u64 owner, u64 num_bytes, u64 *start) | |
363 | { | |
364 | int ret; | |
365 | struct btrfs_path *path; | |
366 | struct btrfs_root *root = device->dev_root; | |
367 | struct btrfs_dev_extent *extent; | |
368 | struct extent_buffer *leaf; | |
369 | struct btrfs_key key; | |
370 | ||
371 | path = btrfs_alloc_path(); | |
372 | if (!path) | |
373 | return -ENOMEM; | |
374 | ||
375 | ret = find_free_dev_extent(trans, device, path, num_bytes, start); | |
6324fbf3 | 376 | if (ret) { |
0b86a832 | 377 | goto err; |
6324fbf3 | 378 | } |
0b86a832 CM |
379 | |
380 | key.objectid = device->devid; | |
381 | key.offset = *start; | |
382 | key.type = BTRFS_DEV_EXTENT_KEY; | |
383 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
384 | sizeof(*extent)); | |
385 | BUG_ON(ret); | |
386 | ||
387 | leaf = path->nodes[0]; | |
388 | extent = btrfs_item_ptr(leaf, path->slots[0], | |
389 | struct btrfs_dev_extent); | |
390 | btrfs_set_dev_extent_owner(leaf, extent, owner); | |
391 | btrfs_set_dev_extent_length(leaf, extent, num_bytes); | |
392 | btrfs_mark_buffer_dirty(leaf); | |
393 | err: | |
394 | btrfs_free_path(path); | |
395 | return ret; | |
396 | } | |
397 | ||
398 | static int find_next_chunk(struct btrfs_root *root, u64 *objectid) | |
399 | { | |
400 | struct btrfs_path *path; | |
401 | int ret; | |
402 | struct btrfs_key key; | |
403 | struct btrfs_key found_key; | |
404 | ||
405 | path = btrfs_alloc_path(); | |
406 | BUG_ON(!path); | |
407 | ||
408 | key.objectid = (u64)-1; | |
409 | key.offset = (u64)-1; | |
410 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
411 | ||
412 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
413 | if (ret < 0) | |
414 | goto error; | |
415 | ||
416 | BUG_ON(ret == 0); | |
417 | ||
418 | ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY); | |
419 | if (ret) { | |
420 | *objectid = 0; | |
421 | } else { | |
422 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
423 | path->slots[0]); | |
424 | *objectid = found_key.objectid + found_key.offset; | |
425 | } | |
426 | ret = 0; | |
427 | error: | |
428 | btrfs_free_path(path); | |
429 | return ret; | |
430 | } | |
431 | ||
0b86a832 CM |
432 | static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path, |
433 | u64 *objectid) | |
434 | { | |
435 | int ret; | |
436 | struct btrfs_key key; | |
437 | struct btrfs_key found_key; | |
438 | ||
439 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
440 | key.type = BTRFS_DEV_ITEM_KEY; | |
441 | key.offset = (u64)-1; | |
442 | ||
443 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
444 | if (ret < 0) | |
445 | goto error; | |
446 | ||
447 | BUG_ON(ret == 0); | |
448 | ||
449 | ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID, | |
450 | BTRFS_DEV_ITEM_KEY); | |
451 | if (ret) { | |
452 | *objectid = 1; | |
453 | } else { | |
454 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
455 | path->slots[0]); | |
456 | *objectid = found_key.offset + 1; | |
457 | } | |
458 | ret = 0; | |
459 | error: | |
460 | btrfs_release_path(root, path); | |
461 | return ret; | |
462 | } | |
463 | ||
464 | /* | |
465 | * the device information is stored in the chunk root | |
466 | * the btrfs_device struct should be fully filled in | |
467 | */ | |
468 | int btrfs_add_device(struct btrfs_trans_handle *trans, | |
469 | struct btrfs_root *root, | |
470 | struct btrfs_device *device) | |
471 | { | |
472 | int ret; | |
473 | struct btrfs_path *path; | |
474 | struct btrfs_dev_item *dev_item; | |
475 | struct extent_buffer *leaf; | |
476 | struct btrfs_key key; | |
477 | unsigned long ptr; | |
478 | u64 free_devid; | |
479 | ||
480 | root = root->fs_info->chunk_root; | |
481 | ||
482 | path = btrfs_alloc_path(); | |
483 | if (!path) | |
484 | return -ENOMEM; | |
485 | ||
486 | ret = find_next_devid(root, path, &free_devid); | |
487 | if (ret) | |
488 | goto out; | |
489 | ||
490 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
491 | key.type = BTRFS_DEV_ITEM_KEY; | |
492 | key.offset = free_devid; | |
493 | ||
494 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
0d81ba5d | 495 | sizeof(*dev_item)); |
0b86a832 CM |
496 | if (ret) |
497 | goto out; | |
498 | ||
499 | leaf = path->nodes[0]; | |
500 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
501 | ||
8a4b83cc | 502 | device->devid = free_devid; |
0b86a832 CM |
503 | btrfs_set_device_id(leaf, dev_item, device->devid); |
504 | btrfs_set_device_type(leaf, dev_item, device->type); | |
505 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
506 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
507 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
508 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
509 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
510 | ||
0b86a832 CM |
511 | ptr = (unsigned long)btrfs_device_uuid(dev_item); |
512 | write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); | |
513 | btrfs_mark_buffer_dirty(leaf); | |
514 | ret = 0; | |
515 | ||
516 | out: | |
517 | btrfs_free_path(path); | |
518 | return ret; | |
519 | } | |
520 | int btrfs_update_device(struct btrfs_trans_handle *trans, | |
521 | struct btrfs_device *device) | |
522 | { | |
523 | int ret; | |
524 | struct btrfs_path *path; | |
525 | struct btrfs_root *root; | |
526 | struct btrfs_dev_item *dev_item; | |
527 | struct extent_buffer *leaf; | |
528 | struct btrfs_key key; | |
529 | ||
530 | root = device->dev_root->fs_info->chunk_root; | |
531 | ||
532 | path = btrfs_alloc_path(); | |
533 | if (!path) | |
534 | return -ENOMEM; | |
535 | ||
536 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
537 | key.type = BTRFS_DEV_ITEM_KEY; | |
538 | key.offset = device->devid; | |
539 | ||
540 | ret = btrfs_search_slot(trans, root, &key, path, 0, 1); | |
541 | if (ret < 0) | |
542 | goto out; | |
543 | ||
544 | if (ret > 0) { | |
545 | ret = -ENOENT; | |
546 | goto out; | |
547 | } | |
548 | ||
549 | leaf = path->nodes[0]; | |
550 | dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item); | |
551 | ||
552 | btrfs_set_device_id(leaf, dev_item, device->devid); | |
553 | btrfs_set_device_type(leaf, dev_item, device->type); | |
554 | btrfs_set_device_io_align(leaf, dev_item, device->io_align); | |
555 | btrfs_set_device_io_width(leaf, dev_item, device->io_width); | |
556 | btrfs_set_device_sector_size(leaf, dev_item, device->sector_size); | |
0b86a832 CM |
557 | btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes); |
558 | btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used); | |
559 | btrfs_mark_buffer_dirty(leaf); | |
560 | ||
561 | out: | |
562 | btrfs_free_path(path); | |
563 | return ret; | |
564 | } | |
565 | ||
566 | int btrfs_add_system_chunk(struct btrfs_trans_handle *trans, | |
567 | struct btrfs_root *root, | |
568 | struct btrfs_key *key, | |
569 | struct btrfs_chunk *chunk, int item_size) | |
570 | { | |
571 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
572 | struct btrfs_disk_key disk_key; | |
573 | u32 array_size; | |
574 | u8 *ptr; | |
575 | ||
576 | array_size = btrfs_super_sys_array_size(super_copy); | |
577 | if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) | |
578 | return -EFBIG; | |
579 | ||
580 | ptr = super_copy->sys_chunk_array + array_size; | |
581 | btrfs_cpu_key_to_disk(&disk_key, key); | |
582 | memcpy(ptr, &disk_key, sizeof(disk_key)); | |
583 | ptr += sizeof(disk_key); | |
584 | memcpy(ptr, chunk, item_size); | |
585 | item_size += sizeof(disk_key); | |
586 | btrfs_set_super_sys_array_size(super_copy, array_size + item_size); | |
587 | return 0; | |
588 | } | |
589 | ||
590 | int btrfs_alloc_chunk(struct btrfs_trans_handle *trans, | |
591 | struct btrfs_root *extent_root, u64 *start, | |
6324fbf3 | 592 | u64 *num_bytes, u64 type) |
0b86a832 CM |
593 | { |
594 | u64 dev_offset; | |
595 | struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root; | |
596 | struct btrfs_stripe *stripes; | |
597 | struct btrfs_device *device = NULL; | |
598 | struct btrfs_chunk *chunk; | |
6324fbf3 | 599 | struct list_head private_devs; |
8a4b83cc | 600 | struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices; |
6324fbf3 | 601 | struct list_head *cur; |
0b86a832 CM |
602 | struct extent_map_tree *em_tree; |
603 | struct map_lookup *map; | |
604 | struct extent_map *em; | |
605 | u64 physical; | |
606 | u64 calc_size = 1024 * 1024 * 1024; | |
6324fbf3 CM |
607 | u64 avail; |
608 | u64 max_avail = 0; | |
609 | int num_stripes = 1; | |
610 | int looped = 0; | |
0b86a832 | 611 | int ret; |
6324fbf3 | 612 | int index; |
0b86a832 CM |
613 | struct btrfs_key key; |
614 | ||
6324fbf3 CM |
615 | if (list_empty(dev_list)) |
616 | return -ENOSPC; | |
617 | again: | |
618 | INIT_LIST_HEAD(&private_devs); | |
619 | cur = dev_list->next; | |
620 | index = 0; | |
621 | /* build a private list of devices we will allocate from */ | |
622 | while(index < num_stripes) { | |
623 | device = list_entry(cur, struct btrfs_device, dev_list); | |
624 | avail = device->total_bytes - device->bytes_used; | |
625 | cur = cur->next; | |
626 | if (avail > max_avail) | |
627 | max_avail = avail; | |
628 | if (avail >= calc_size) { | |
629 | list_move_tail(&device->dev_list, &private_devs); | |
630 | index++; | |
631 | } | |
632 | if (cur == dev_list) | |
633 | break; | |
634 | } | |
635 | if (index < num_stripes) { | |
636 | list_splice(&private_devs, dev_list); | |
637 | if (!looped && max_avail > 0) { | |
638 | looped = 1; | |
639 | calc_size = max_avail; | |
640 | goto again; | |
641 | } | |
642 | return -ENOSPC; | |
643 | } | |
0b86a832 CM |
644 | |
645 | ret = find_next_chunk(chunk_root, &key.objectid); | |
646 | if (ret) | |
647 | return ret; | |
648 | ||
0b86a832 CM |
649 | chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS); |
650 | if (!chunk) | |
651 | return -ENOMEM; | |
652 | ||
653 | stripes = &chunk->stripe; | |
654 | ||
655 | *num_bytes = calc_size; | |
6324fbf3 | 656 | index = 0; |
0b86a832 | 657 | while(index < num_stripes) { |
6324fbf3 CM |
658 | BUG_ON(list_empty(&private_devs)); |
659 | cur = private_devs.next; | |
660 | device = list_entry(cur, struct btrfs_device, dev_list); | |
661 | list_move_tail(&device->dev_list, dev_list); | |
0b86a832 CM |
662 | |
663 | ret = btrfs_alloc_dev_extent(trans, device, | |
664 | key.objectid, | |
665 | calc_size, &dev_offset); | |
666 | BUG_ON(ret); | |
8a4b83cc | 667 | printk("alloc chunk size %Lu from dev %Lu\n", calc_size, device->devid); |
0b86a832 CM |
668 | device->bytes_used += calc_size; |
669 | ret = btrfs_update_device(trans, device); | |
670 | BUG_ON(ret); | |
671 | ||
672 | btrfs_set_stack_stripe_devid(stripes + index, device->devid); | |
673 | btrfs_set_stack_stripe_offset(stripes + index, dev_offset); | |
674 | physical = dev_offset; | |
675 | index++; | |
676 | } | |
6324fbf3 | 677 | BUG_ON(!list_empty(&private_devs)); |
0b86a832 CM |
678 | |
679 | /* key.objectid was set above */ | |
680 | key.offset = *num_bytes; | |
681 | key.type = BTRFS_CHUNK_ITEM_KEY; | |
682 | btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid); | |
683 | btrfs_set_stack_chunk_stripe_len(chunk, 64 * 1024); | |
684 | btrfs_set_stack_chunk_type(chunk, type); | |
685 | btrfs_set_stack_chunk_num_stripes(chunk, num_stripes); | |
686 | btrfs_set_stack_chunk_io_align(chunk, extent_root->sectorsize); | |
687 | btrfs_set_stack_chunk_io_width(chunk, extent_root->sectorsize); | |
688 | btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize); | |
689 | ||
690 | ret = btrfs_insert_item(trans, chunk_root, &key, chunk, | |
691 | btrfs_chunk_item_size(num_stripes)); | |
692 | BUG_ON(ret); | |
693 | *start = key.objectid; | |
694 | ||
695 | em = alloc_extent_map(GFP_NOFS); | |
696 | if (!em) | |
697 | return -ENOMEM; | |
698 | map = kmalloc(sizeof(*map), GFP_NOFS); | |
699 | if (!map) { | |
700 | free_extent_map(em); | |
701 | return -ENOMEM; | |
702 | } | |
703 | ||
704 | em->bdev = (struct block_device *)map; | |
705 | em->start = key.objectid; | |
706 | em->len = key.offset; | |
707 | em->block_start = 0; | |
708 | ||
709 | map->physical = physical; | |
710 | map->dev = device; | |
711 | ||
712 | if (!map->dev) { | |
713 | kfree(map); | |
714 | free_extent_map(em); | |
715 | return -EIO; | |
716 | } | |
717 | kfree(chunk); | |
718 | ||
719 | em_tree = &extent_root->fs_info->mapping_tree.map_tree; | |
720 | spin_lock(&em_tree->lock); | |
721 | ret = add_extent_mapping(em_tree, em); | |
722 | BUG_ON(ret); | |
723 | spin_unlock(&em_tree->lock); | |
724 | free_extent_map(em); | |
725 | return ret; | |
726 | } | |
727 | ||
728 | void btrfs_mapping_init(struct btrfs_mapping_tree *tree) | |
729 | { | |
730 | extent_map_tree_init(&tree->map_tree, GFP_NOFS); | |
731 | } | |
732 | ||
733 | void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree) | |
734 | { | |
735 | struct extent_map *em; | |
736 | ||
737 | while(1) { | |
738 | spin_lock(&tree->map_tree.lock); | |
739 | em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1); | |
740 | if (em) | |
741 | remove_extent_mapping(&tree->map_tree, em); | |
742 | spin_unlock(&tree->map_tree.lock); | |
743 | if (!em) | |
744 | break; | |
745 | kfree(em->bdev); | |
746 | /* once for us */ | |
747 | free_extent_map(em); | |
748 | /* once for the tree */ | |
749 | free_extent_map(em); | |
750 | } | |
751 | } | |
752 | ||
753 | int btrfs_map_block(struct btrfs_mapping_tree *map_tree, | |
754 | u64 logical, u64 *phys, u64 *length, | |
755 | struct btrfs_device **dev) | |
756 | { | |
757 | struct extent_map *em; | |
758 | struct map_lookup *map; | |
759 | struct extent_map_tree *em_tree = &map_tree->map_tree; | |
760 | u64 offset; | |
761 | ||
762 | ||
763 | spin_lock(&em_tree->lock); | |
764 | em = lookup_extent_mapping(em_tree, logical, *length); | |
765 | BUG_ON(!em); | |
766 | ||
767 | BUG_ON(em->start > logical || em->start + em->len < logical); | |
768 | map = (struct map_lookup *)em->bdev; | |
769 | offset = logical - em->start; | |
770 | *phys = map->physical + offset; | |
771 | *length = em->len - offset; | |
772 | *dev = map->dev; | |
773 | free_extent_map(em); | |
774 | spin_unlock(&em_tree->lock); | |
775 | return 0; | |
776 | } | |
777 | ||
778 | int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio) | |
779 | { | |
780 | struct btrfs_mapping_tree *map_tree; | |
781 | struct btrfs_device *dev; | |
782 | u64 logical = bio->bi_sector << 9; | |
783 | u64 physical; | |
784 | u64 length = 0; | |
785 | u64 map_length; | |
786 | struct bio_vec *bvec; | |
787 | int i; | |
788 | int ret; | |
789 | ||
790 | bio_for_each_segment(bvec, bio, i) { | |
791 | length += bvec->bv_len; | |
792 | } | |
793 | map_tree = &root->fs_info->mapping_tree; | |
794 | map_length = length; | |
795 | ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev); | |
239b14b3 CM |
796 | if (map_length < length) { |
797 | printk("mapping failed logical %Lu bio len %Lu physical %Lu " | |
798 | "len %Lu\n", logical, length, physical, map_length); | |
799 | BUG(); | |
800 | } | |
0b86a832 CM |
801 | BUG_ON(map_length < length); |
802 | bio->bi_sector = physical >> 9; | |
803 | bio->bi_bdev = dev->bdev; | |
804 | submit_bio(rw, bio); | |
805 | return 0; | |
806 | } | |
807 | ||
808 | struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid) | |
809 | { | |
8a4b83cc | 810 | struct list_head *head = &root->fs_info->fs_devices->devices; |
0b86a832 | 811 | |
8a4b83cc | 812 | return __find_device(head, devid); |
0b86a832 CM |
813 | } |
814 | ||
815 | static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, | |
816 | struct extent_buffer *leaf, | |
817 | struct btrfs_chunk *chunk) | |
818 | { | |
819 | struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree; | |
820 | struct map_lookup *map; | |
821 | struct extent_map *em; | |
822 | u64 logical; | |
823 | u64 length; | |
824 | u64 devid; | |
825 | int ret; | |
826 | ||
827 | logical = key->objectid; | |
828 | length = key->offset; | |
829 | spin_lock(&map_tree->map_tree.lock); | |
830 | em = lookup_extent_mapping(&map_tree->map_tree, logical, 1); | |
831 | ||
832 | /* already mapped? */ | |
833 | if (em && em->start <= logical && em->start + em->len > logical) { | |
834 | free_extent_map(em); | |
835 | spin_unlock(&map_tree->map_tree.lock); | |
836 | return 0; | |
837 | } else if (em) { | |
838 | free_extent_map(em); | |
839 | } | |
840 | spin_unlock(&map_tree->map_tree.lock); | |
841 | ||
842 | map = kzalloc(sizeof(*map), GFP_NOFS); | |
843 | if (!map) | |
844 | return -ENOMEM; | |
845 | ||
846 | em = alloc_extent_map(GFP_NOFS); | |
847 | if (!em) | |
848 | return -ENOMEM; | |
849 | map = kmalloc(sizeof(*map), GFP_NOFS); | |
850 | if (!map) { | |
851 | free_extent_map(em); | |
852 | return -ENOMEM; | |
853 | } | |
854 | ||
855 | em->bdev = (struct block_device *)map; | |
856 | em->start = logical; | |
857 | em->len = length; | |
858 | em->block_start = 0; | |
859 | ||
860 | map->physical = btrfs_stripe_offset_nr(leaf, chunk, 0); | |
861 | devid = btrfs_stripe_devid_nr(leaf, chunk, 0); | |
862 | map->dev = btrfs_find_device(root, devid); | |
863 | if (!map->dev) { | |
864 | kfree(map); | |
865 | free_extent_map(em); | |
866 | return -EIO; | |
867 | } | |
868 | ||
869 | spin_lock(&map_tree->map_tree.lock); | |
870 | ret = add_extent_mapping(&map_tree->map_tree, em); | |
871 | BUG_ON(ret); | |
872 | spin_unlock(&map_tree->map_tree.lock); | |
873 | free_extent_map(em); | |
874 | ||
875 | return 0; | |
876 | } | |
877 | ||
878 | static int fill_device_from_item(struct extent_buffer *leaf, | |
879 | struct btrfs_dev_item *dev_item, | |
880 | struct btrfs_device *device) | |
881 | { | |
882 | unsigned long ptr; | |
0b86a832 CM |
883 | |
884 | device->devid = btrfs_device_id(leaf, dev_item); | |
885 | device->total_bytes = btrfs_device_total_bytes(leaf, dev_item); | |
886 | device->bytes_used = btrfs_device_bytes_used(leaf, dev_item); | |
887 | device->type = btrfs_device_type(leaf, dev_item); | |
888 | device->io_align = btrfs_device_io_align(leaf, dev_item); | |
889 | device->io_width = btrfs_device_io_width(leaf, dev_item); | |
890 | device->sector_size = btrfs_device_sector_size(leaf, dev_item); | |
0b86a832 CM |
891 | |
892 | ptr = (unsigned long)btrfs_device_uuid(dev_item); | |
893 | read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE); | |
894 | ||
0b86a832 CM |
895 | return 0; |
896 | } | |
897 | ||
0d81ba5d | 898 | static int read_one_dev(struct btrfs_root *root, |
0b86a832 CM |
899 | struct extent_buffer *leaf, |
900 | struct btrfs_dev_item *dev_item) | |
901 | { | |
902 | struct btrfs_device *device; | |
903 | u64 devid; | |
904 | int ret; | |
905 | ||
906 | devid = btrfs_device_id(leaf, dev_item); | |
6324fbf3 CM |
907 | device = btrfs_find_device(root, devid); |
908 | if (!device) { | |
8a4b83cc | 909 | printk("warning devid %Lu not found already\n", devid); |
6324fbf3 CM |
910 | device = kmalloc(sizeof(*device), GFP_NOFS); |
911 | if (!device) | |
912 | return -ENOMEM; | |
8a4b83cc CM |
913 | list_add(&device->dev_list, |
914 | &root->fs_info->fs_devices->devices); | |
6324fbf3 | 915 | } |
0b86a832 CM |
916 | |
917 | fill_device_from_item(leaf, dev_item, device); | |
918 | device->dev_root = root->fs_info->dev_root; | |
0b86a832 CM |
919 | ret = 0; |
920 | #if 0 | |
921 | ret = btrfs_open_device(device); | |
922 | if (ret) { | |
923 | kfree(device); | |
924 | } | |
925 | #endif | |
926 | return ret; | |
927 | } | |
928 | ||
0d81ba5d CM |
929 | int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf) |
930 | { | |
931 | struct btrfs_dev_item *dev_item; | |
932 | ||
933 | dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block, | |
934 | dev_item); | |
935 | return read_one_dev(root, buf, dev_item); | |
936 | } | |
937 | ||
0b86a832 CM |
938 | int btrfs_read_sys_array(struct btrfs_root *root) |
939 | { | |
940 | struct btrfs_super_block *super_copy = &root->fs_info->super_copy; | |
941 | struct extent_buffer *sb = root->fs_info->sb_buffer; | |
942 | struct btrfs_disk_key *disk_key; | |
0b86a832 CM |
943 | struct btrfs_chunk *chunk; |
944 | struct btrfs_key key; | |
945 | u32 num_stripes; | |
946 | u32 array_size; | |
947 | u32 len = 0; | |
948 | u8 *ptr; | |
949 | unsigned long sb_ptr; | |
950 | u32 cur; | |
951 | int ret; | |
0b86a832 CM |
952 | |
953 | array_size = btrfs_super_sys_array_size(super_copy); | |
954 | ||
955 | /* | |
956 | * we do this loop twice, once for the device items and | |
957 | * once for all of the chunks. This way there are device | |
958 | * structs filled in for every chunk | |
959 | */ | |
0b86a832 CM |
960 | ptr = super_copy->sys_chunk_array; |
961 | sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array); | |
962 | cur = 0; | |
963 | ||
964 | while (cur < array_size) { | |
965 | disk_key = (struct btrfs_disk_key *)ptr; | |
966 | btrfs_disk_key_to_cpu(&key, disk_key); | |
967 | ||
968 | len = sizeof(*disk_key); | |
969 | ptr += len; | |
970 | sb_ptr += len; | |
971 | cur += len; | |
972 | ||
0d81ba5d | 973 | if (key.type == BTRFS_CHUNK_ITEM_KEY) { |
0b86a832 | 974 | chunk = (struct btrfs_chunk *)sb_ptr; |
0d81ba5d CM |
975 | ret = read_one_chunk(root, &key, sb, chunk); |
976 | BUG_ON(ret); | |
0b86a832 CM |
977 | num_stripes = btrfs_chunk_num_stripes(sb, chunk); |
978 | len = btrfs_chunk_item_size(num_stripes); | |
979 | } else { | |
980 | BUG(); | |
981 | } | |
982 | ptr += len; | |
983 | sb_ptr += len; | |
984 | cur += len; | |
985 | } | |
0b86a832 CM |
986 | return 0; |
987 | } | |
988 | ||
989 | int btrfs_read_chunk_tree(struct btrfs_root *root) | |
990 | { | |
991 | struct btrfs_path *path; | |
992 | struct extent_buffer *leaf; | |
993 | struct btrfs_key key; | |
994 | struct btrfs_key found_key; | |
995 | int ret; | |
996 | int slot; | |
997 | ||
998 | root = root->fs_info->chunk_root; | |
999 | ||
1000 | path = btrfs_alloc_path(); | |
1001 | if (!path) | |
1002 | return -ENOMEM; | |
1003 | ||
1004 | /* first we search for all of the device items, and then we | |
1005 | * read in all of the chunk items. This way we can create chunk | |
1006 | * mappings that reference all of the devices that are afound | |
1007 | */ | |
1008 | key.objectid = BTRFS_DEV_ITEMS_OBJECTID; | |
1009 | key.offset = 0; | |
1010 | key.type = 0; | |
1011 | again: | |
1012 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
1013 | while(1) { | |
1014 | leaf = path->nodes[0]; | |
1015 | slot = path->slots[0]; | |
1016 | if (slot >= btrfs_header_nritems(leaf)) { | |
1017 | ret = btrfs_next_leaf(root, path); | |
1018 | if (ret == 0) | |
1019 | continue; | |
1020 | if (ret < 0) | |
1021 | goto error; | |
1022 | break; | |
1023 | } | |
1024 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1025 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1026 | if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID) | |
1027 | break; | |
1028 | if (found_key.type == BTRFS_DEV_ITEM_KEY) { | |
1029 | struct btrfs_dev_item *dev_item; | |
1030 | dev_item = btrfs_item_ptr(leaf, slot, | |
1031 | struct btrfs_dev_item); | |
0d81ba5d | 1032 | ret = read_one_dev(root, leaf, dev_item); |
0b86a832 CM |
1033 | BUG_ON(ret); |
1034 | } | |
1035 | } else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) { | |
1036 | struct btrfs_chunk *chunk; | |
1037 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
1038 | ret = read_one_chunk(root, &found_key, leaf, chunk); | |
1039 | } | |
1040 | path->slots[0]++; | |
1041 | } | |
1042 | if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) { | |
1043 | key.objectid = 0; | |
1044 | btrfs_release_path(root, path); | |
1045 | goto again; | |
1046 | } | |
1047 | ||
1048 | btrfs_free_path(path); | |
1049 | ret = 0; | |
1050 | error: | |
1051 | return ret; | |
1052 | } | |
1053 |