Commit | Line | Data |
---|---|---|
d1310b2e CM |
1 | #include <linux/bitops.h> |
2 | #include <linux/slab.h> | |
3 | #include <linux/bio.h> | |
4 | #include <linux/mm.h> | |
d1310b2e CM |
5 | #include <linux/pagemap.h> |
6 | #include <linux/page-flags.h> | |
7 | #include <linux/module.h> | |
8 | #include <linux/spinlock.h> | |
9 | #include <linux/blkdev.h> | |
10 | #include <linux/swap.h> | |
d1310b2e CM |
11 | #include <linux/writeback.h> |
12 | #include <linux/pagevec.h> | |
13 | #include "extent_io.h" | |
14 | #include "extent_map.h" | |
2db04966 | 15 | #include "compat.h" |
902b22f3 DW |
16 | #include "ctree.h" |
17 | #include "btrfs_inode.h" | |
d1310b2e | 18 | |
d1310b2e CM |
19 | static struct kmem_cache *extent_state_cache; |
20 | static struct kmem_cache *extent_buffer_cache; | |
21 | ||
22 | static LIST_HEAD(buffers); | |
23 | static LIST_HEAD(states); | |
4bef0848 | 24 | |
b47eda86 | 25 | #define LEAK_DEBUG 0 |
3935127c | 26 | #if LEAK_DEBUG |
d397712b | 27 | static DEFINE_SPINLOCK(leak_lock); |
4bef0848 | 28 | #endif |
d1310b2e | 29 | |
d1310b2e CM |
30 | #define BUFFER_LRU_MAX 64 |
31 | ||
32 | struct tree_entry { | |
33 | u64 start; | |
34 | u64 end; | |
d1310b2e CM |
35 | struct rb_node rb_node; |
36 | }; | |
37 | ||
38 | struct extent_page_data { | |
39 | struct bio *bio; | |
40 | struct extent_io_tree *tree; | |
41 | get_extent_t *get_extent; | |
771ed689 CM |
42 | |
43 | /* tells writepage not to lock the state bits for this range | |
44 | * it still does the unlocking | |
45 | */ | |
ffbd517d CM |
46 | unsigned int extent_locked:1; |
47 | ||
48 | /* tells the submit_bio code to use a WRITE_SYNC */ | |
49 | unsigned int sync_io:1; | |
d1310b2e CM |
50 | }; |
51 | ||
52 | int __init extent_io_init(void) | |
53 | { | |
9601e3f6 CH |
54 | extent_state_cache = kmem_cache_create("extent_state", |
55 | sizeof(struct extent_state), 0, | |
56 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); | |
d1310b2e CM |
57 | if (!extent_state_cache) |
58 | return -ENOMEM; | |
59 | ||
9601e3f6 CH |
60 | extent_buffer_cache = kmem_cache_create("extent_buffers", |
61 | sizeof(struct extent_buffer), 0, | |
62 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL); | |
d1310b2e CM |
63 | if (!extent_buffer_cache) |
64 | goto free_state_cache; | |
65 | return 0; | |
66 | ||
67 | free_state_cache: | |
68 | kmem_cache_destroy(extent_state_cache); | |
69 | return -ENOMEM; | |
70 | } | |
71 | ||
72 | void extent_io_exit(void) | |
73 | { | |
74 | struct extent_state *state; | |
2d2ae547 | 75 | struct extent_buffer *eb; |
d1310b2e CM |
76 | |
77 | while (!list_empty(&states)) { | |
2d2ae547 | 78 | state = list_entry(states.next, struct extent_state, leak_list); |
d397712b CM |
79 | printk(KERN_ERR "btrfs state leak: start %llu end %llu " |
80 | "state %lu in tree %p refs %d\n", | |
81 | (unsigned long long)state->start, | |
82 | (unsigned long long)state->end, | |
83 | state->state, state->tree, atomic_read(&state->refs)); | |
2d2ae547 | 84 | list_del(&state->leak_list); |
d1310b2e CM |
85 | kmem_cache_free(extent_state_cache, state); |
86 | ||
87 | } | |
88 | ||
2d2ae547 CM |
89 | while (!list_empty(&buffers)) { |
90 | eb = list_entry(buffers.next, struct extent_buffer, leak_list); | |
d397712b CM |
91 | printk(KERN_ERR "btrfs buffer leak start %llu len %lu " |
92 | "refs %d\n", (unsigned long long)eb->start, | |
93 | eb->len, atomic_read(&eb->refs)); | |
2d2ae547 CM |
94 | list_del(&eb->leak_list); |
95 | kmem_cache_free(extent_buffer_cache, eb); | |
96 | } | |
d1310b2e CM |
97 | if (extent_state_cache) |
98 | kmem_cache_destroy(extent_state_cache); | |
99 | if (extent_buffer_cache) | |
100 | kmem_cache_destroy(extent_buffer_cache); | |
101 | } | |
102 | ||
103 | void extent_io_tree_init(struct extent_io_tree *tree, | |
104 | struct address_space *mapping, gfp_t mask) | |
105 | { | |
6bef4d31 | 106 | tree->state = RB_ROOT; |
19fe0a8b | 107 | INIT_RADIX_TREE(&tree->buffer, GFP_ATOMIC); |
d1310b2e CM |
108 | tree->ops = NULL; |
109 | tree->dirty_bytes = 0; | |
70dec807 | 110 | spin_lock_init(&tree->lock); |
6af118ce | 111 | spin_lock_init(&tree->buffer_lock); |
d1310b2e | 112 | tree->mapping = mapping; |
d1310b2e | 113 | } |
d1310b2e | 114 | |
b2950863 | 115 | static struct extent_state *alloc_extent_state(gfp_t mask) |
d1310b2e CM |
116 | { |
117 | struct extent_state *state; | |
3935127c | 118 | #if LEAK_DEBUG |
2d2ae547 | 119 | unsigned long flags; |
4bef0848 | 120 | #endif |
d1310b2e CM |
121 | |
122 | state = kmem_cache_alloc(extent_state_cache, mask); | |
2b114d1d | 123 | if (!state) |
d1310b2e CM |
124 | return state; |
125 | state->state = 0; | |
d1310b2e | 126 | state->private = 0; |
70dec807 | 127 | state->tree = NULL; |
3935127c | 128 | #if LEAK_DEBUG |
2d2ae547 CM |
129 | spin_lock_irqsave(&leak_lock, flags); |
130 | list_add(&state->leak_list, &states); | |
131 | spin_unlock_irqrestore(&leak_lock, flags); | |
4bef0848 | 132 | #endif |
d1310b2e CM |
133 | atomic_set(&state->refs, 1); |
134 | init_waitqueue_head(&state->wq); | |
135 | return state; | |
136 | } | |
d1310b2e | 137 | |
4845e44f | 138 | void free_extent_state(struct extent_state *state) |
d1310b2e | 139 | { |
d1310b2e CM |
140 | if (!state) |
141 | return; | |
142 | if (atomic_dec_and_test(&state->refs)) { | |
3935127c | 143 | #if LEAK_DEBUG |
2d2ae547 | 144 | unsigned long flags; |
4bef0848 | 145 | #endif |
70dec807 | 146 | WARN_ON(state->tree); |
3935127c | 147 | #if LEAK_DEBUG |
2d2ae547 CM |
148 | spin_lock_irqsave(&leak_lock, flags); |
149 | list_del(&state->leak_list); | |
150 | spin_unlock_irqrestore(&leak_lock, flags); | |
4bef0848 | 151 | #endif |
d1310b2e CM |
152 | kmem_cache_free(extent_state_cache, state); |
153 | } | |
154 | } | |
d1310b2e CM |
155 | |
156 | static struct rb_node *tree_insert(struct rb_root *root, u64 offset, | |
157 | struct rb_node *node) | |
158 | { | |
d397712b CM |
159 | struct rb_node **p = &root->rb_node; |
160 | struct rb_node *parent = NULL; | |
d1310b2e CM |
161 | struct tree_entry *entry; |
162 | ||
d397712b | 163 | while (*p) { |
d1310b2e CM |
164 | parent = *p; |
165 | entry = rb_entry(parent, struct tree_entry, rb_node); | |
166 | ||
167 | if (offset < entry->start) | |
168 | p = &(*p)->rb_left; | |
169 | else if (offset > entry->end) | |
170 | p = &(*p)->rb_right; | |
171 | else | |
172 | return parent; | |
173 | } | |
174 | ||
175 | entry = rb_entry(node, struct tree_entry, rb_node); | |
d1310b2e CM |
176 | rb_link_node(node, parent, p); |
177 | rb_insert_color(node, root); | |
178 | return NULL; | |
179 | } | |
180 | ||
80ea96b1 | 181 | static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset, |
d1310b2e CM |
182 | struct rb_node **prev_ret, |
183 | struct rb_node **next_ret) | |
184 | { | |
80ea96b1 | 185 | struct rb_root *root = &tree->state; |
d397712b | 186 | struct rb_node *n = root->rb_node; |
d1310b2e CM |
187 | struct rb_node *prev = NULL; |
188 | struct rb_node *orig_prev = NULL; | |
189 | struct tree_entry *entry; | |
190 | struct tree_entry *prev_entry = NULL; | |
191 | ||
d397712b | 192 | while (n) { |
d1310b2e CM |
193 | entry = rb_entry(n, struct tree_entry, rb_node); |
194 | prev = n; | |
195 | prev_entry = entry; | |
196 | ||
197 | if (offset < entry->start) | |
198 | n = n->rb_left; | |
199 | else if (offset > entry->end) | |
200 | n = n->rb_right; | |
d397712b | 201 | else |
d1310b2e CM |
202 | return n; |
203 | } | |
204 | ||
205 | if (prev_ret) { | |
206 | orig_prev = prev; | |
d397712b | 207 | while (prev && offset > prev_entry->end) { |
d1310b2e CM |
208 | prev = rb_next(prev); |
209 | prev_entry = rb_entry(prev, struct tree_entry, rb_node); | |
210 | } | |
211 | *prev_ret = prev; | |
212 | prev = orig_prev; | |
213 | } | |
214 | ||
215 | if (next_ret) { | |
216 | prev_entry = rb_entry(prev, struct tree_entry, rb_node); | |
d397712b | 217 | while (prev && offset < prev_entry->start) { |
d1310b2e CM |
218 | prev = rb_prev(prev); |
219 | prev_entry = rb_entry(prev, struct tree_entry, rb_node); | |
220 | } | |
221 | *next_ret = prev; | |
222 | } | |
223 | return NULL; | |
224 | } | |
225 | ||
80ea96b1 CM |
226 | static inline struct rb_node *tree_search(struct extent_io_tree *tree, |
227 | u64 offset) | |
d1310b2e | 228 | { |
70dec807 | 229 | struct rb_node *prev = NULL; |
d1310b2e | 230 | struct rb_node *ret; |
70dec807 | 231 | |
80ea96b1 | 232 | ret = __etree_search(tree, offset, &prev, NULL); |
d397712b | 233 | if (!ret) |
d1310b2e CM |
234 | return prev; |
235 | return ret; | |
236 | } | |
237 | ||
9ed74f2d JB |
238 | static void merge_cb(struct extent_io_tree *tree, struct extent_state *new, |
239 | struct extent_state *other) | |
240 | { | |
241 | if (tree->ops && tree->ops->merge_extent_hook) | |
242 | tree->ops->merge_extent_hook(tree->mapping->host, new, | |
243 | other); | |
244 | } | |
245 | ||
d1310b2e CM |
246 | /* |
247 | * utility function to look for merge candidates inside a given range. | |
248 | * Any extents with matching state are merged together into a single | |
249 | * extent in the tree. Extents with EXTENT_IO in their state field | |
250 | * are not merged because the end_io handlers need to be able to do | |
251 | * operations on them without sleeping (or doing allocations/splits). | |
252 | * | |
253 | * This should be called with the tree lock held. | |
254 | */ | |
255 | static int merge_state(struct extent_io_tree *tree, | |
256 | struct extent_state *state) | |
257 | { | |
258 | struct extent_state *other; | |
259 | struct rb_node *other_node; | |
260 | ||
5b21f2ed | 261 | if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) |
d1310b2e CM |
262 | return 0; |
263 | ||
264 | other_node = rb_prev(&state->rb_node); | |
265 | if (other_node) { | |
266 | other = rb_entry(other_node, struct extent_state, rb_node); | |
267 | if (other->end == state->start - 1 && | |
268 | other->state == state->state) { | |
9ed74f2d | 269 | merge_cb(tree, state, other); |
d1310b2e | 270 | state->start = other->start; |
70dec807 | 271 | other->tree = NULL; |
d1310b2e CM |
272 | rb_erase(&other->rb_node, &tree->state); |
273 | free_extent_state(other); | |
274 | } | |
275 | } | |
276 | other_node = rb_next(&state->rb_node); | |
277 | if (other_node) { | |
278 | other = rb_entry(other_node, struct extent_state, rb_node); | |
279 | if (other->start == state->end + 1 && | |
280 | other->state == state->state) { | |
9ed74f2d | 281 | merge_cb(tree, state, other); |
d1310b2e | 282 | other->start = state->start; |
70dec807 | 283 | state->tree = NULL; |
d1310b2e CM |
284 | rb_erase(&state->rb_node, &tree->state); |
285 | free_extent_state(state); | |
9ed74f2d | 286 | state = NULL; |
d1310b2e CM |
287 | } |
288 | } | |
9ed74f2d | 289 | |
d1310b2e CM |
290 | return 0; |
291 | } | |
292 | ||
9ed74f2d | 293 | static int set_state_cb(struct extent_io_tree *tree, |
0ca1f7ce | 294 | struct extent_state *state, int *bits) |
291d673e CM |
295 | { |
296 | if (tree->ops && tree->ops->set_bit_hook) { | |
9ed74f2d | 297 | return tree->ops->set_bit_hook(tree->mapping->host, |
0ca1f7ce | 298 | state, bits); |
291d673e | 299 | } |
9ed74f2d JB |
300 | |
301 | return 0; | |
291d673e CM |
302 | } |
303 | ||
304 | static void clear_state_cb(struct extent_io_tree *tree, | |
0ca1f7ce | 305 | struct extent_state *state, int *bits) |
291d673e | 306 | { |
9ed74f2d JB |
307 | if (tree->ops && tree->ops->clear_bit_hook) |
308 | tree->ops->clear_bit_hook(tree->mapping->host, state, bits); | |
291d673e CM |
309 | } |
310 | ||
d1310b2e CM |
311 | /* |
312 | * insert an extent_state struct into the tree. 'bits' are set on the | |
313 | * struct before it is inserted. | |
314 | * | |
315 | * This may return -EEXIST if the extent is already there, in which case the | |
316 | * state struct is freed. | |
317 | * | |
318 | * The tree lock is not taken internally. This is a utility function and | |
319 | * probably isn't what you want to call (see set/clear_extent_bit). | |
320 | */ | |
321 | static int insert_state(struct extent_io_tree *tree, | |
322 | struct extent_state *state, u64 start, u64 end, | |
0ca1f7ce | 323 | int *bits) |
d1310b2e CM |
324 | { |
325 | struct rb_node *node; | |
0ca1f7ce | 326 | int bits_to_set = *bits & ~EXTENT_CTLBITS; |
9ed74f2d | 327 | int ret; |
d1310b2e CM |
328 | |
329 | if (end < start) { | |
d397712b CM |
330 | printk(KERN_ERR "btrfs end < start %llu %llu\n", |
331 | (unsigned long long)end, | |
332 | (unsigned long long)start); | |
d1310b2e CM |
333 | WARN_ON(1); |
334 | } | |
d1310b2e CM |
335 | state->start = start; |
336 | state->end = end; | |
9ed74f2d JB |
337 | ret = set_state_cb(tree, state, bits); |
338 | if (ret) | |
339 | return ret; | |
340 | ||
0ca1f7ce | 341 | if (bits_to_set & EXTENT_DIRTY) |
9ed74f2d | 342 | tree->dirty_bytes += end - start + 1; |
0ca1f7ce | 343 | state->state |= bits_to_set; |
d1310b2e CM |
344 | node = tree_insert(&tree->state, end, &state->rb_node); |
345 | if (node) { | |
346 | struct extent_state *found; | |
347 | found = rb_entry(node, struct extent_state, rb_node); | |
d397712b CM |
348 | printk(KERN_ERR "btrfs found node %llu %llu on insert of " |
349 | "%llu %llu\n", (unsigned long long)found->start, | |
350 | (unsigned long long)found->end, | |
351 | (unsigned long long)start, (unsigned long long)end); | |
d1310b2e CM |
352 | free_extent_state(state); |
353 | return -EEXIST; | |
354 | } | |
70dec807 | 355 | state->tree = tree; |
d1310b2e CM |
356 | merge_state(tree, state); |
357 | return 0; | |
358 | } | |
359 | ||
9ed74f2d JB |
360 | static int split_cb(struct extent_io_tree *tree, struct extent_state *orig, |
361 | u64 split) | |
362 | { | |
363 | if (tree->ops && tree->ops->split_extent_hook) | |
364 | return tree->ops->split_extent_hook(tree->mapping->host, | |
365 | orig, split); | |
366 | return 0; | |
367 | } | |
368 | ||
d1310b2e CM |
369 | /* |
370 | * split a given extent state struct in two, inserting the preallocated | |
371 | * struct 'prealloc' as the newly created second half. 'split' indicates an | |
372 | * offset inside 'orig' where it should be split. | |
373 | * | |
374 | * Before calling, | |
375 | * the tree has 'orig' at [orig->start, orig->end]. After calling, there | |
376 | * are two extent state structs in the tree: | |
377 | * prealloc: [orig->start, split - 1] | |
378 | * orig: [ split, orig->end ] | |
379 | * | |
380 | * The tree locks are not taken by this function. They need to be held | |
381 | * by the caller. | |
382 | */ | |
383 | static int split_state(struct extent_io_tree *tree, struct extent_state *orig, | |
384 | struct extent_state *prealloc, u64 split) | |
385 | { | |
386 | struct rb_node *node; | |
9ed74f2d JB |
387 | |
388 | split_cb(tree, orig, split); | |
389 | ||
d1310b2e CM |
390 | prealloc->start = orig->start; |
391 | prealloc->end = split - 1; | |
392 | prealloc->state = orig->state; | |
393 | orig->start = split; | |
394 | ||
395 | node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node); | |
396 | if (node) { | |
d1310b2e CM |
397 | free_extent_state(prealloc); |
398 | return -EEXIST; | |
399 | } | |
70dec807 | 400 | prealloc->tree = tree; |
d1310b2e CM |
401 | return 0; |
402 | } | |
403 | ||
404 | /* | |
405 | * utility function to clear some bits in an extent state struct. | |
406 | * it will optionally wake up any one waiting on this state (wake == 1), or | |
407 | * forcibly remove the state from the tree (delete == 1). | |
408 | * | |
409 | * If no bits are set on the state struct after clearing things, the | |
410 | * struct is freed and removed from the tree | |
411 | */ | |
412 | static int clear_state_bit(struct extent_io_tree *tree, | |
0ca1f7ce YZ |
413 | struct extent_state *state, |
414 | int *bits, int wake) | |
d1310b2e | 415 | { |
0ca1f7ce | 416 | int bits_to_clear = *bits & ~EXTENT_CTLBITS; |
32c00aff | 417 | int ret = state->state & bits_to_clear; |
d1310b2e | 418 | |
0ca1f7ce | 419 | if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) { |
d1310b2e CM |
420 | u64 range = state->end - state->start + 1; |
421 | WARN_ON(range > tree->dirty_bytes); | |
422 | tree->dirty_bytes -= range; | |
423 | } | |
291d673e | 424 | clear_state_cb(tree, state, bits); |
32c00aff | 425 | state->state &= ~bits_to_clear; |
d1310b2e CM |
426 | if (wake) |
427 | wake_up(&state->wq); | |
0ca1f7ce | 428 | if (state->state == 0) { |
70dec807 | 429 | if (state->tree) { |
d1310b2e | 430 | rb_erase(&state->rb_node, &tree->state); |
70dec807 | 431 | state->tree = NULL; |
d1310b2e CM |
432 | free_extent_state(state); |
433 | } else { | |
434 | WARN_ON(1); | |
435 | } | |
436 | } else { | |
437 | merge_state(tree, state); | |
438 | } | |
439 | return ret; | |
440 | } | |
441 | ||
442 | /* | |
443 | * clear some bits on a range in the tree. This may require splitting | |
444 | * or inserting elements in the tree, so the gfp mask is used to | |
445 | * indicate which allocations or sleeping are allowed. | |
446 | * | |
447 | * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove | |
448 | * the given range from the tree regardless of state (ie for truncate). | |
449 | * | |
450 | * the range [start, end] is inclusive. | |
451 | * | |
452 | * This takes the tree lock, and returns < 0 on error, > 0 if any of the | |
453 | * bits were already set, or zero if none of the bits were already set. | |
454 | */ | |
455 | int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, | |
2c64c53d CM |
456 | int bits, int wake, int delete, |
457 | struct extent_state **cached_state, | |
458 | gfp_t mask) | |
d1310b2e CM |
459 | { |
460 | struct extent_state *state; | |
2c64c53d | 461 | struct extent_state *cached; |
d1310b2e | 462 | struct extent_state *prealloc = NULL; |
2c64c53d | 463 | struct rb_node *next_node; |
d1310b2e | 464 | struct rb_node *node; |
5c939df5 | 465 | u64 last_end; |
d1310b2e CM |
466 | int err; |
467 | int set = 0; | |
2ac55d41 | 468 | int clear = 0; |
d1310b2e | 469 | |
0ca1f7ce YZ |
470 | if (delete) |
471 | bits |= ~EXTENT_CTLBITS; | |
472 | bits |= EXTENT_FIRST_DELALLOC; | |
473 | ||
2ac55d41 JB |
474 | if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY)) |
475 | clear = 1; | |
d1310b2e CM |
476 | again: |
477 | if (!prealloc && (mask & __GFP_WAIT)) { | |
478 | prealloc = alloc_extent_state(mask); | |
479 | if (!prealloc) | |
480 | return -ENOMEM; | |
481 | } | |
482 | ||
cad321ad | 483 | spin_lock(&tree->lock); |
2c64c53d CM |
484 | if (cached_state) { |
485 | cached = *cached_state; | |
2ac55d41 JB |
486 | |
487 | if (clear) { | |
488 | *cached_state = NULL; | |
489 | cached_state = NULL; | |
490 | } | |
491 | ||
42daec29 | 492 | if (cached && cached->tree && cached->start == start) { |
2ac55d41 JB |
493 | if (clear) |
494 | atomic_dec(&cached->refs); | |
2c64c53d | 495 | state = cached; |
42daec29 | 496 | goto hit_next; |
2c64c53d | 497 | } |
2ac55d41 JB |
498 | if (clear) |
499 | free_extent_state(cached); | |
2c64c53d | 500 | } |
d1310b2e CM |
501 | /* |
502 | * this search will find the extents that end after | |
503 | * our range starts | |
504 | */ | |
80ea96b1 | 505 | node = tree_search(tree, start); |
d1310b2e CM |
506 | if (!node) |
507 | goto out; | |
508 | state = rb_entry(node, struct extent_state, rb_node); | |
2c64c53d | 509 | hit_next: |
d1310b2e CM |
510 | if (state->start > end) |
511 | goto out; | |
512 | WARN_ON(state->end < start); | |
5c939df5 | 513 | last_end = state->end; |
d1310b2e CM |
514 | |
515 | /* | |
516 | * | ---- desired range ---- | | |
517 | * | state | or | |
518 | * | ------------- state -------------- | | |
519 | * | |
520 | * We need to split the extent we found, and may flip | |
521 | * bits on second half. | |
522 | * | |
523 | * If the extent we found extends past our range, we | |
524 | * just split and search again. It'll get split again | |
525 | * the next time though. | |
526 | * | |
527 | * If the extent we found is inside our range, we clear | |
528 | * the desired bit on it. | |
529 | */ | |
530 | ||
531 | if (state->start < start) { | |
70dec807 CM |
532 | if (!prealloc) |
533 | prealloc = alloc_extent_state(GFP_ATOMIC); | |
d1310b2e CM |
534 | err = split_state(tree, state, prealloc, start); |
535 | BUG_ON(err == -EEXIST); | |
536 | prealloc = NULL; | |
537 | if (err) | |
538 | goto out; | |
539 | if (state->end <= end) { | |
0ca1f7ce | 540 | set |= clear_state_bit(tree, state, &bits, wake); |
5c939df5 YZ |
541 | if (last_end == (u64)-1) |
542 | goto out; | |
543 | start = last_end + 1; | |
d1310b2e CM |
544 | } |
545 | goto search_again; | |
546 | } | |
547 | /* | |
548 | * | ---- desired range ---- | | |
549 | * | state | | |
550 | * We need to split the extent, and clear the bit | |
551 | * on the first half | |
552 | */ | |
553 | if (state->start <= end && state->end > end) { | |
70dec807 CM |
554 | if (!prealloc) |
555 | prealloc = alloc_extent_state(GFP_ATOMIC); | |
d1310b2e CM |
556 | err = split_state(tree, state, prealloc, end + 1); |
557 | BUG_ON(err == -EEXIST); | |
d1310b2e CM |
558 | if (wake) |
559 | wake_up(&state->wq); | |
42daec29 | 560 | |
0ca1f7ce | 561 | set |= clear_state_bit(tree, prealloc, &bits, wake); |
9ed74f2d | 562 | |
d1310b2e CM |
563 | prealloc = NULL; |
564 | goto out; | |
565 | } | |
42daec29 | 566 | |
2c64c53d CM |
567 | if (state->end < end && prealloc && !need_resched()) |
568 | next_node = rb_next(&state->rb_node); | |
569 | else | |
570 | next_node = NULL; | |
42daec29 | 571 | |
0ca1f7ce | 572 | set |= clear_state_bit(tree, state, &bits, wake); |
5c939df5 YZ |
573 | if (last_end == (u64)-1) |
574 | goto out; | |
575 | start = last_end + 1; | |
2c64c53d CM |
576 | if (start <= end && next_node) { |
577 | state = rb_entry(next_node, struct extent_state, | |
578 | rb_node); | |
579 | if (state->start == start) | |
580 | goto hit_next; | |
581 | } | |
d1310b2e CM |
582 | goto search_again; |
583 | ||
584 | out: | |
cad321ad | 585 | spin_unlock(&tree->lock); |
d1310b2e CM |
586 | if (prealloc) |
587 | free_extent_state(prealloc); | |
588 | ||
589 | return set; | |
590 | ||
591 | search_again: | |
592 | if (start > end) | |
593 | goto out; | |
cad321ad | 594 | spin_unlock(&tree->lock); |
d1310b2e CM |
595 | if (mask & __GFP_WAIT) |
596 | cond_resched(); | |
597 | goto again; | |
598 | } | |
d1310b2e CM |
599 | |
600 | static int wait_on_state(struct extent_io_tree *tree, | |
601 | struct extent_state *state) | |
641f5219 CH |
602 | __releases(tree->lock) |
603 | __acquires(tree->lock) | |
d1310b2e CM |
604 | { |
605 | DEFINE_WAIT(wait); | |
606 | prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE); | |
cad321ad | 607 | spin_unlock(&tree->lock); |
d1310b2e | 608 | schedule(); |
cad321ad | 609 | spin_lock(&tree->lock); |
d1310b2e CM |
610 | finish_wait(&state->wq, &wait); |
611 | return 0; | |
612 | } | |
613 | ||
614 | /* | |
615 | * waits for one or more bits to clear on a range in the state tree. | |
616 | * The range [start, end] is inclusive. | |
617 | * The tree lock is taken by this function | |
618 | */ | |
619 | int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits) | |
620 | { | |
621 | struct extent_state *state; | |
622 | struct rb_node *node; | |
623 | ||
cad321ad | 624 | spin_lock(&tree->lock); |
d1310b2e CM |
625 | again: |
626 | while (1) { | |
627 | /* | |
628 | * this search will find all the extents that end after | |
629 | * our range starts | |
630 | */ | |
80ea96b1 | 631 | node = tree_search(tree, start); |
d1310b2e CM |
632 | if (!node) |
633 | break; | |
634 | ||
635 | state = rb_entry(node, struct extent_state, rb_node); | |
636 | ||
637 | if (state->start > end) | |
638 | goto out; | |
639 | ||
640 | if (state->state & bits) { | |
641 | start = state->start; | |
642 | atomic_inc(&state->refs); | |
643 | wait_on_state(tree, state); | |
644 | free_extent_state(state); | |
645 | goto again; | |
646 | } | |
647 | start = state->end + 1; | |
648 | ||
649 | if (start > end) | |
650 | break; | |
651 | ||
652 | if (need_resched()) { | |
cad321ad | 653 | spin_unlock(&tree->lock); |
d1310b2e | 654 | cond_resched(); |
cad321ad | 655 | spin_lock(&tree->lock); |
d1310b2e CM |
656 | } |
657 | } | |
658 | out: | |
cad321ad | 659 | spin_unlock(&tree->lock); |
d1310b2e CM |
660 | return 0; |
661 | } | |
d1310b2e | 662 | |
9ed74f2d | 663 | static int set_state_bits(struct extent_io_tree *tree, |
d1310b2e | 664 | struct extent_state *state, |
0ca1f7ce | 665 | int *bits) |
d1310b2e | 666 | { |
9ed74f2d | 667 | int ret; |
0ca1f7ce | 668 | int bits_to_set = *bits & ~EXTENT_CTLBITS; |
9ed74f2d JB |
669 | |
670 | ret = set_state_cb(tree, state, bits); | |
671 | if (ret) | |
672 | return ret; | |
0ca1f7ce | 673 | if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) { |
d1310b2e CM |
674 | u64 range = state->end - state->start + 1; |
675 | tree->dirty_bytes += range; | |
676 | } | |
0ca1f7ce | 677 | state->state |= bits_to_set; |
9ed74f2d JB |
678 | |
679 | return 0; | |
d1310b2e CM |
680 | } |
681 | ||
2c64c53d CM |
682 | static void cache_state(struct extent_state *state, |
683 | struct extent_state **cached_ptr) | |
684 | { | |
685 | if (cached_ptr && !(*cached_ptr)) { | |
686 | if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) { | |
687 | *cached_ptr = state; | |
688 | atomic_inc(&state->refs); | |
689 | } | |
690 | } | |
691 | } | |
692 | ||
d1310b2e | 693 | /* |
1edbb734 CM |
694 | * set some bits on a range in the tree. This may require allocations or |
695 | * sleeping, so the gfp mask is used to indicate what is allowed. | |
d1310b2e | 696 | * |
1edbb734 CM |
697 | * If any of the exclusive bits are set, this will fail with -EEXIST if some |
698 | * part of the range already has the desired bits set. The start of the | |
699 | * existing range is returned in failed_start in this case. | |
d1310b2e | 700 | * |
1edbb734 | 701 | * [start, end] is inclusive This takes the tree lock. |
d1310b2e | 702 | */ |
1edbb734 | 703 | |
4845e44f CM |
704 | int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, |
705 | int bits, int exclusive_bits, u64 *failed_start, | |
706 | struct extent_state **cached_state, gfp_t mask) | |
d1310b2e CM |
707 | { |
708 | struct extent_state *state; | |
709 | struct extent_state *prealloc = NULL; | |
710 | struct rb_node *node; | |
d1310b2e | 711 | int err = 0; |
d1310b2e CM |
712 | u64 last_start; |
713 | u64 last_end; | |
42daec29 | 714 | |
0ca1f7ce | 715 | bits |= EXTENT_FIRST_DELALLOC; |
d1310b2e CM |
716 | again: |
717 | if (!prealloc && (mask & __GFP_WAIT)) { | |
718 | prealloc = alloc_extent_state(mask); | |
719 | if (!prealloc) | |
720 | return -ENOMEM; | |
721 | } | |
722 | ||
cad321ad | 723 | spin_lock(&tree->lock); |
9655d298 CM |
724 | if (cached_state && *cached_state) { |
725 | state = *cached_state; | |
726 | if (state->start == start && state->tree) { | |
727 | node = &state->rb_node; | |
728 | goto hit_next; | |
729 | } | |
730 | } | |
d1310b2e CM |
731 | /* |
732 | * this search will find all the extents that end after | |
733 | * our range starts. | |
734 | */ | |
80ea96b1 | 735 | node = tree_search(tree, start); |
d1310b2e | 736 | if (!node) { |
0ca1f7ce | 737 | err = insert_state(tree, prealloc, start, end, &bits); |
d1310b2e CM |
738 | prealloc = NULL; |
739 | BUG_ON(err == -EEXIST); | |
740 | goto out; | |
741 | } | |
d1310b2e | 742 | state = rb_entry(node, struct extent_state, rb_node); |
40431d6c | 743 | hit_next: |
d1310b2e CM |
744 | last_start = state->start; |
745 | last_end = state->end; | |
746 | ||
747 | /* | |
748 | * | ---- desired range ---- | | |
749 | * | state | | |
750 | * | |
751 | * Just lock what we found and keep going | |
752 | */ | |
753 | if (state->start == start && state->end <= end) { | |
40431d6c | 754 | struct rb_node *next_node; |
1edbb734 | 755 | if (state->state & exclusive_bits) { |
d1310b2e CM |
756 | *failed_start = state->start; |
757 | err = -EEXIST; | |
758 | goto out; | |
759 | } | |
42daec29 | 760 | |
0ca1f7ce | 761 | err = set_state_bits(tree, state, &bits); |
9ed74f2d JB |
762 | if (err) |
763 | goto out; | |
764 | ||
2c64c53d | 765 | cache_state(state, cached_state); |
d1310b2e | 766 | merge_state(tree, state); |
5c939df5 YZ |
767 | if (last_end == (u64)-1) |
768 | goto out; | |
40431d6c | 769 | |
5c939df5 | 770 | start = last_end + 1; |
40431d6c CM |
771 | if (start < end && prealloc && !need_resched()) { |
772 | next_node = rb_next(node); | |
773 | if (next_node) { | |
774 | state = rb_entry(next_node, struct extent_state, | |
775 | rb_node); | |
776 | if (state->start == start) | |
777 | goto hit_next; | |
778 | } | |
779 | } | |
d1310b2e CM |
780 | goto search_again; |
781 | } | |
782 | ||
783 | /* | |
784 | * | ---- desired range ---- | | |
785 | * | state | | |
786 | * or | |
787 | * | ------------- state -------------- | | |
788 | * | |
789 | * We need to split the extent we found, and may flip bits on | |
790 | * second half. | |
791 | * | |
792 | * If the extent we found extends past our | |
793 | * range, we just split and search again. It'll get split | |
794 | * again the next time though. | |
795 | * | |
796 | * If the extent we found is inside our range, we set the | |
797 | * desired bit on it. | |
798 | */ | |
799 | if (state->start < start) { | |
1edbb734 | 800 | if (state->state & exclusive_bits) { |
d1310b2e CM |
801 | *failed_start = start; |
802 | err = -EEXIST; | |
803 | goto out; | |
804 | } | |
805 | err = split_state(tree, state, prealloc, start); | |
806 | BUG_ON(err == -EEXIST); | |
807 | prealloc = NULL; | |
808 | if (err) | |
809 | goto out; | |
810 | if (state->end <= end) { | |
0ca1f7ce | 811 | err = set_state_bits(tree, state, &bits); |
9ed74f2d JB |
812 | if (err) |
813 | goto out; | |
2c64c53d | 814 | cache_state(state, cached_state); |
d1310b2e | 815 | merge_state(tree, state); |
5c939df5 YZ |
816 | if (last_end == (u64)-1) |
817 | goto out; | |
818 | start = last_end + 1; | |
d1310b2e CM |
819 | } |
820 | goto search_again; | |
821 | } | |
822 | /* | |
823 | * | ---- desired range ---- | | |
824 | * | state | or | state | | |
825 | * | |
826 | * There's a hole, we need to insert something in it and | |
827 | * ignore the extent we found. | |
828 | */ | |
829 | if (state->start > start) { | |
830 | u64 this_end; | |
831 | if (end < last_start) | |
832 | this_end = end; | |
833 | else | |
d397712b | 834 | this_end = last_start - 1; |
d1310b2e | 835 | err = insert_state(tree, prealloc, start, this_end, |
0ca1f7ce | 836 | &bits); |
d1310b2e | 837 | BUG_ON(err == -EEXIST); |
9ed74f2d JB |
838 | if (err) { |
839 | prealloc = NULL; | |
d1310b2e | 840 | goto out; |
9ed74f2d JB |
841 | } |
842 | cache_state(prealloc, cached_state); | |
843 | prealloc = NULL; | |
d1310b2e CM |
844 | start = this_end + 1; |
845 | goto search_again; | |
846 | } | |
847 | /* | |
848 | * | ---- desired range ---- | | |
849 | * | state | | |
850 | * We need to split the extent, and set the bit | |
851 | * on the first half | |
852 | */ | |
853 | if (state->start <= end && state->end > end) { | |
1edbb734 | 854 | if (state->state & exclusive_bits) { |
d1310b2e CM |
855 | *failed_start = start; |
856 | err = -EEXIST; | |
857 | goto out; | |
858 | } | |
859 | err = split_state(tree, state, prealloc, end + 1); | |
860 | BUG_ON(err == -EEXIST); | |
861 | ||
0ca1f7ce | 862 | err = set_state_bits(tree, prealloc, &bits); |
9ed74f2d JB |
863 | if (err) { |
864 | prealloc = NULL; | |
865 | goto out; | |
866 | } | |
2c64c53d | 867 | cache_state(prealloc, cached_state); |
d1310b2e CM |
868 | merge_state(tree, prealloc); |
869 | prealloc = NULL; | |
870 | goto out; | |
871 | } | |
872 | ||
873 | goto search_again; | |
874 | ||
875 | out: | |
cad321ad | 876 | spin_unlock(&tree->lock); |
d1310b2e CM |
877 | if (prealloc) |
878 | free_extent_state(prealloc); | |
879 | ||
880 | return err; | |
881 | ||
882 | search_again: | |
883 | if (start > end) | |
884 | goto out; | |
cad321ad | 885 | spin_unlock(&tree->lock); |
d1310b2e CM |
886 | if (mask & __GFP_WAIT) |
887 | cond_resched(); | |
888 | goto again; | |
889 | } | |
d1310b2e CM |
890 | |
891 | /* wrappers around set/clear extent bit */ | |
892 | int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, | |
893 | gfp_t mask) | |
894 | { | |
895 | return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL, | |
2c64c53d | 896 | NULL, mask); |
d1310b2e | 897 | } |
d1310b2e CM |
898 | |
899 | int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, | |
900 | int bits, gfp_t mask) | |
901 | { | |
902 | return set_extent_bit(tree, start, end, bits, 0, NULL, | |
2c64c53d | 903 | NULL, mask); |
d1310b2e | 904 | } |
d1310b2e CM |
905 | |
906 | int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, | |
907 | int bits, gfp_t mask) | |
908 | { | |
2c64c53d | 909 | return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask); |
d1310b2e | 910 | } |
d1310b2e CM |
911 | |
912 | int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end, | |
2ac55d41 | 913 | struct extent_state **cached_state, gfp_t mask) |
d1310b2e CM |
914 | { |
915 | return set_extent_bit(tree, start, end, | |
40431d6c | 916 | EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE, |
2ac55d41 | 917 | 0, NULL, cached_state, mask); |
d1310b2e | 918 | } |
d1310b2e CM |
919 | |
920 | int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end, | |
921 | gfp_t mask) | |
922 | { | |
923 | return clear_extent_bit(tree, start, end, | |
32c00aff | 924 | EXTENT_DIRTY | EXTENT_DELALLOC | |
0ca1f7ce | 925 | EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask); |
d1310b2e | 926 | } |
d1310b2e CM |
927 | |
928 | int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end, | |
929 | gfp_t mask) | |
930 | { | |
931 | return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, | |
2c64c53d | 932 | NULL, mask); |
d1310b2e | 933 | } |
d1310b2e | 934 | |
b2950863 | 935 | static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end, |
d1310b2e CM |
936 | gfp_t mask) |
937 | { | |
2c64c53d CM |
938 | return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, |
939 | NULL, mask); | |
d1310b2e | 940 | } |
d1310b2e CM |
941 | |
942 | int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end, | |
943 | gfp_t mask) | |
944 | { | |
945 | return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL, | |
2c64c53d | 946 | NULL, mask); |
d1310b2e | 947 | } |
d1310b2e | 948 | |
d397712b | 949 | static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, |
2ac55d41 JB |
950 | u64 end, struct extent_state **cached_state, |
951 | gfp_t mask) | |
d1310b2e | 952 | { |
2c64c53d | 953 | return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, |
2ac55d41 | 954 | cached_state, mask); |
d1310b2e | 955 | } |
d1310b2e | 956 | |
d1310b2e CM |
957 | int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end) |
958 | { | |
959 | return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK); | |
960 | } | |
d1310b2e | 961 | |
d352ac68 CM |
962 | /* |
963 | * either insert or lock state struct between start and end use mask to tell | |
964 | * us if waiting is desired. | |
965 | */ | |
1edbb734 | 966 | int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end, |
2c64c53d | 967 | int bits, struct extent_state **cached_state, gfp_t mask) |
d1310b2e CM |
968 | { |
969 | int err; | |
970 | u64 failed_start; | |
971 | while (1) { | |
1edbb734 | 972 | err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits, |
2c64c53d CM |
973 | EXTENT_LOCKED, &failed_start, |
974 | cached_state, mask); | |
d1310b2e CM |
975 | if (err == -EEXIST && (mask & __GFP_WAIT)) { |
976 | wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED); | |
977 | start = failed_start; | |
978 | } else { | |
979 | break; | |
980 | } | |
981 | WARN_ON(start > end); | |
982 | } | |
983 | return err; | |
984 | } | |
d1310b2e | 985 | |
1edbb734 CM |
986 | int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask) |
987 | { | |
2c64c53d | 988 | return lock_extent_bits(tree, start, end, 0, NULL, mask); |
1edbb734 CM |
989 | } |
990 | ||
25179201 JB |
991 | int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, |
992 | gfp_t mask) | |
993 | { | |
994 | int err; | |
995 | u64 failed_start; | |
996 | ||
2c64c53d CM |
997 | err = set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED, |
998 | &failed_start, NULL, mask); | |
6643558d YZ |
999 | if (err == -EEXIST) { |
1000 | if (failed_start > start) | |
1001 | clear_extent_bit(tree, start, failed_start - 1, | |
2c64c53d | 1002 | EXTENT_LOCKED, 1, 0, NULL, mask); |
25179201 | 1003 | return 0; |
6643558d | 1004 | } |
25179201 JB |
1005 | return 1; |
1006 | } | |
25179201 | 1007 | |
2c64c53d CM |
1008 | int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end, |
1009 | struct extent_state **cached, gfp_t mask) | |
1010 | { | |
1011 | return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached, | |
1012 | mask); | |
1013 | } | |
1014 | ||
d1310b2e CM |
1015 | int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, |
1016 | gfp_t mask) | |
1017 | { | |
2c64c53d CM |
1018 | return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL, |
1019 | mask); | |
d1310b2e | 1020 | } |
d1310b2e CM |
1021 | |
1022 | /* | |
1023 | * helper function to set pages and extents in the tree dirty | |
1024 | */ | |
1025 | int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end) | |
1026 | { | |
1027 | unsigned long index = start >> PAGE_CACHE_SHIFT; | |
1028 | unsigned long end_index = end >> PAGE_CACHE_SHIFT; | |
1029 | struct page *page; | |
1030 | ||
1031 | while (index <= end_index) { | |
1032 | page = find_get_page(tree->mapping, index); | |
1033 | BUG_ON(!page); | |
1034 | __set_page_dirty_nobuffers(page); | |
1035 | page_cache_release(page); | |
1036 | index++; | |
1037 | } | |
d1310b2e CM |
1038 | return 0; |
1039 | } | |
d1310b2e CM |
1040 | |
1041 | /* | |
1042 | * helper function to set both pages and extents in the tree writeback | |
1043 | */ | |
b2950863 | 1044 | static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end) |
d1310b2e CM |
1045 | { |
1046 | unsigned long index = start >> PAGE_CACHE_SHIFT; | |
1047 | unsigned long end_index = end >> PAGE_CACHE_SHIFT; | |
1048 | struct page *page; | |
1049 | ||
1050 | while (index <= end_index) { | |
1051 | page = find_get_page(tree->mapping, index); | |
1052 | BUG_ON(!page); | |
1053 | set_page_writeback(page); | |
1054 | page_cache_release(page); | |
1055 | index++; | |
1056 | } | |
d1310b2e CM |
1057 | return 0; |
1058 | } | |
d1310b2e | 1059 | |
d352ac68 CM |
1060 | /* |
1061 | * find the first offset in the io tree with 'bits' set. zero is | |
1062 | * returned if we find something, and *start_ret and *end_ret are | |
1063 | * set to reflect the state struct that was found. | |
1064 | * | |
1065 | * If nothing was found, 1 is returned, < 0 on error | |
1066 | */ | |
d1310b2e CM |
1067 | int find_first_extent_bit(struct extent_io_tree *tree, u64 start, |
1068 | u64 *start_ret, u64 *end_ret, int bits) | |
1069 | { | |
1070 | struct rb_node *node; | |
1071 | struct extent_state *state; | |
1072 | int ret = 1; | |
1073 | ||
cad321ad | 1074 | spin_lock(&tree->lock); |
d1310b2e CM |
1075 | /* |
1076 | * this search will find all the extents that end after | |
1077 | * our range starts. | |
1078 | */ | |
80ea96b1 | 1079 | node = tree_search(tree, start); |
d397712b | 1080 | if (!node) |
d1310b2e | 1081 | goto out; |
d1310b2e | 1082 | |
d397712b | 1083 | while (1) { |
d1310b2e CM |
1084 | state = rb_entry(node, struct extent_state, rb_node); |
1085 | if (state->end >= start && (state->state & bits)) { | |
1086 | *start_ret = state->start; | |
1087 | *end_ret = state->end; | |
1088 | ret = 0; | |
1089 | break; | |
1090 | } | |
1091 | node = rb_next(node); | |
1092 | if (!node) | |
1093 | break; | |
1094 | } | |
1095 | out: | |
cad321ad | 1096 | spin_unlock(&tree->lock); |
d1310b2e CM |
1097 | return ret; |
1098 | } | |
d1310b2e | 1099 | |
d352ac68 CM |
1100 | /* find the first state struct with 'bits' set after 'start', and |
1101 | * return it. tree->lock must be held. NULL will returned if | |
1102 | * nothing was found after 'start' | |
1103 | */ | |
d7fc640e CM |
1104 | struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree, |
1105 | u64 start, int bits) | |
1106 | { | |
1107 | struct rb_node *node; | |
1108 | struct extent_state *state; | |
1109 | ||
1110 | /* | |
1111 | * this search will find all the extents that end after | |
1112 | * our range starts. | |
1113 | */ | |
1114 | node = tree_search(tree, start); | |
d397712b | 1115 | if (!node) |
d7fc640e | 1116 | goto out; |
d7fc640e | 1117 | |
d397712b | 1118 | while (1) { |
d7fc640e | 1119 | state = rb_entry(node, struct extent_state, rb_node); |
d397712b | 1120 | if (state->end >= start && (state->state & bits)) |
d7fc640e | 1121 | return state; |
d397712b | 1122 | |
d7fc640e CM |
1123 | node = rb_next(node); |
1124 | if (!node) | |
1125 | break; | |
1126 | } | |
1127 | out: | |
1128 | return NULL; | |
1129 | } | |
d7fc640e | 1130 | |
d352ac68 CM |
1131 | /* |
1132 | * find a contiguous range of bytes in the file marked as delalloc, not | |
1133 | * more than 'max_bytes'. start and end are used to return the range, | |
1134 | * | |
1135 | * 1 is returned if we find something, 0 if nothing was in the tree | |
1136 | */ | |
c8b97818 | 1137 | static noinline u64 find_delalloc_range(struct extent_io_tree *tree, |
c2a128d2 JB |
1138 | u64 *start, u64 *end, u64 max_bytes, |
1139 | struct extent_state **cached_state) | |
d1310b2e CM |
1140 | { |
1141 | struct rb_node *node; | |
1142 | struct extent_state *state; | |
1143 | u64 cur_start = *start; | |
1144 | u64 found = 0; | |
1145 | u64 total_bytes = 0; | |
1146 | ||
cad321ad | 1147 | spin_lock(&tree->lock); |
c8b97818 | 1148 | |
d1310b2e CM |
1149 | /* |
1150 | * this search will find all the extents that end after | |
1151 | * our range starts. | |
1152 | */ | |
80ea96b1 | 1153 | node = tree_search(tree, cur_start); |
2b114d1d | 1154 | if (!node) { |
3b951516 CM |
1155 | if (!found) |
1156 | *end = (u64)-1; | |
d1310b2e CM |
1157 | goto out; |
1158 | } | |
1159 | ||
d397712b | 1160 | while (1) { |
d1310b2e | 1161 | state = rb_entry(node, struct extent_state, rb_node); |
5b21f2ed ZY |
1162 | if (found && (state->start != cur_start || |
1163 | (state->state & EXTENT_BOUNDARY))) { | |
d1310b2e CM |
1164 | goto out; |
1165 | } | |
1166 | if (!(state->state & EXTENT_DELALLOC)) { | |
1167 | if (!found) | |
1168 | *end = state->end; | |
1169 | goto out; | |
1170 | } | |
c2a128d2 | 1171 | if (!found) { |
d1310b2e | 1172 | *start = state->start; |
c2a128d2 JB |
1173 | *cached_state = state; |
1174 | atomic_inc(&state->refs); | |
1175 | } | |
d1310b2e CM |
1176 | found++; |
1177 | *end = state->end; | |
1178 | cur_start = state->end + 1; | |
1179 | node = rb_next(node); | |
1180 | if (!node) | |
1181 | break; | |
1182 | total_bytes += state->end - state->start + 1; | |
1183 | if (total_bytes >= max_bytes) | |
1184 | break; | |
1185 | } | |
1186 | out: | |
cad321ad | 1187 | spin_unlock(&tree->lock); |
d1310b2e CM |
1188 | return found; |
1189 | } | |
1190 | ||
c8b97818 CM |
1191 | static noinline int __unlock_for_delalloc(struct inode *inode, |
1192 | struct page *locked_page, | |
1193 | u64 start, u64 end) | |
1194 | { | |
1195 | int ret; | |
1196 | struct page *pages[16]; | |
1197 | unsigned long index = start >> PAGE_CACHE_SHIFT; | |
1198 | unsigned long end_index = end >> PAGE_CACHE_SHIFT; | |
1199 | unsigned long nr_pages = end_index - index + 1; | |
1200 | int i; | |
1201 | ||
1202 | if (index == locked_page->index && end_index == index) | |
1203 | return 0; | |
1204 | ||
d397712b | 1205 | while (nr_pages > 0) { |
c8b97818 | 1206 | ret = find_get_pages_contig(inode->i_mapping, index, |
5b050f04 CM |
1207 | min_t(unsigned long, nr_pages, |
1208 | ARRAY_SIZE(pages)), pages); | |
c8b97818 CM |
1209 | for (i = 0; i < ret; i++) { |
1210 | if (pages[i] != locked_page) | |
1211 | unlock_page(pages[i]); | |
1212 | page_cache_release(pages[i]); | |
1213 | } | |
1214 | nr_pages -= ret; | |
1215 | index += ret; | |
1216 | cond_resched(); | |
1217 | } | |
1218 | return 0; | |
1219 | } | |
1220 | ||
1221 | static noinline int lock_delalloc_pages(struct inode *inode, | |
1222 | struct page *locked_page, | |
1223 | u64 delalloc_start, | |
1224 | u64 delalloc_end) | |
1225 | { | |
1226 | unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT; | |
1227 | unsigned long start_index = index; | |
1228 | unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT; | |
1229 | unsigned long pages_locked = 0; | |
1230 | struct page *pages[16]; | |
1231 | unsigned long nrpages; | |
1232 | int ret; | |
1233 | int i; | |
1234 | ||
1235 | /* the caller is responsible for locking the start index */ | |
1236 | if (index == locked_page->index && index == end_index) | |
1237 | return 0; | |
1238 | ||
1239 | /* skip the page at the start index */ | |
1240 | nrpages = end_index - index + 1; | |
d397712b | 1241 | while (nrpages > 0) { |
c8b97818 | 1242 | ret = find_get_pages_contig(inode->i_mapping, index, |
5b050f04 CM |
1243 | min_t(unsigned long, |
1244 | nrpages, ARRAY_SIZE(pages)), pages); | |
c8b97818 CM |
1245 | if (ret == 0) { |
1246 | ret = -EAGAIN; | |
1247 | goto done; | |
1248 | } | |
1249 | /* now we have an array of pages, lock them all */ | |
1250 | for (i = 0; i < ret; i++) { | |
1251 | /* | |
1252 | * the caller is taking responsibility for | |
1253 | * locked_page | |
1254 | */ | |
771ed689 | 1255 | if (pages[i] != locked_page) { |
c8b97818 | 1256 | lock_page(pages[i]); |
f2b1c41c CM |
1257 | if (!PageDirty(pages[i]) || |
1258 | pages[i]->mapping != inode->i_mapping) { | |
771ed689 CM |
1259 | ret = -EAGAIN; |
1260 | unlock_page(pages[i]); | |
1261 | page_cache_release(pages[i]); | |
1262 | goto done; | |
1263 | } | |
1264 | } | |
c8b97818 | 1265 | page_cache_release(pages[i]); |
771ed689 | 1266 | pages_locked++; |
c8b97818 | 1267 | } |
c8b97818 CM |
1268 | nrpages -= ret; |
1269 | index += ret; | |
1270 | cond_resched(); | |
1271 | } | |
1272 | ret = 0; | |
1273 | done: | |
1274 | if (ret && pages_locked) { | |
1275 | __unlock_for_delalloc(inode, locked_page, | |
1276 | delalloc_start, | |
1277 | ((u64)(start_index + pages_locked - 1)) << | |
1278 | PAGE_CACHE_SHIFT); | |
1279 | } | |
1280 | return ret; | |
1281 | } | |
1282 | ||
1283 | /* | |
1284 | * find a contiguous range of bytes in the file marked as delalloc, not | |
1285 | * more than 'max_bytes'. start and end are used to return the range, | |
1286 | * | |
1287 | * 1 is returned if we find something, 0 if nothing was in the tree | |
1288 | */ | |
1289 | static noinline u64 find_lock_delalloc_range(struct inode *inode, | |
1290 | struct extent_io_tree *tree, | |
1291 | struct page *locked_page, | |
1292 | u64 *start, u64 *end, | |
1293 | u64 max_bytes) | |
1294 | { | |
1295 | u64 delalloc_start; | |
1296 | u64 delalloc_end; | |
1297 | u64 found; | |
9655d298 | 1298 | struct extent_state *cached_state = NULL; |
c8b97818 CM |
1299 | int ret; |
1300 | int loops = 0; | |
1301 | ||
1302 | again: | |
1303 | /* step one, find a bunch of delalloc bytes starting at start */ | |
1304 | delalloc_start = *start; | |
1305 | delalloc_end = 0; | |
1306 | found = find_delalloc_range(tree, &delalloc_start, &delalloc_end, | |
c2a128d2 | 1307 | max_bytes, &cached_state); |
70b99e69 | 1308 | if (!found || delalloc_end <= *start) { |
c8b97818 CM |
1309 | *start = delalloc_start; |
1310 | *end = delalloc_end; | |
c2a128d2 | 1311 | free_extent_state(cached_state); |
c8b97818 CM |
1312 | return found; |
1313 | } | |
1314 | ||
70b99e69 CM |
1315 | /* |
1316 | * start comes from the offset of locked_page. We have to lock | |
1317 | * pages in order, so we can't process delalloc bytes before | |
1318 | * locked_page | |
1319 | */ | |
d397712b | 1320 | if (delalloc_start < *start) |
70b99e69 | 1321 | delalloc_start = *start; |
70b99e69 | 1322 | |
c8b97818 CM |
1323 | /* |
1324 | * make sure to limit the number of pages we try to lock down | |
1325 | * if we're looping. | |
1326 | */ | |
d397712b | 1327 | if (delalloc_end + 1 - delalloc_start > max_bytes && loops) |
771ed689 | 1328 | delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1; |
d397712b | 1329 | |
c8b97818 CM |
1330 | /* step two, lock all the pages after the page that has start */ |
1331 | ret = lock_delalloc_pages(inode, locked_page, | |
1332 | delalloc_start, delalloc_end); | |
1333 | if (ret == -EAGAIN) { | |
1334 | /* some of the pages are gone, lets avoid looping by | |
1335 | * shortening the size of the delalloc range we're searching | |
1336 | */ | |
9655d298 | 1337 | free_extent_state(cached_state); |
c8b97818 CM |
1338 | if (!loops) { |
1339 | unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1); | |
1340 | max_bytes = PAGE_CACHE_SIZE - offset; | |
1341 | loops = 1; | |
1342 | goto again; | |
1343 | } else { | |
1344 | found = 0; | |
1345 | goto out_failed; | |
1346 | } | |
1347 | } | |
1348 | BUG_ON(ret); | |
1349 | ||
1350 | /* step three, lock the state bits for the whole range */ | |
9655d298 CM |
1351 | lock_extent_bits(tree, delalloc_start, delalloc_end, |
1352 | 0, &cached_state, GFP_NOFS); | |
c8b97818 CM |
1353 | |
1354 | /* then test to make sure it is all still delalloc */ | |
1355 | ret = test_range_bit(tree, delalloc_start, delalloc_end, | |
9655d298 | 1356 | EXTENT_DELALLOC, 1, cached_state); |
c8b97818 | 1357 | if (!ret) { |
9655d298 CM |
1358 | unlock_extent_cached(tree, delalloc_start, delalloc_end, |
1359 | &cached_state, GFP_NOFS); | |
c8b97818 CM |
1360 | __unlock_for_delalloc(inode, locked_page, |
1361 | delalloc_start, delalloc_end); | |
1362 | cond_resched(); | |
1363 | goto again; | |
1364 | } | |
9655d298 | 1365 | free_extent_state(cached_state); |
c8b97818 CM |
1366 | *start = delalloc_start; |
1367 | *end = delalloc_end; | |
1368 | out_failed: | |
1369 | return found; | |
1370 | } | |
1371 | ||
1372 | int extent_clear_unlock_delalloc(struct inode *inode, | |
1373 | struct extent_io_tree *tree, | |
1374 | u64 start, u64 end, struct page *locked_page, | |
a791e35e | 1375 | unsigned long op) |
c8b97818 CM |
1376 | { |
1377 | int ret; | |
1378 | struct page *pages[16]; | |
1379 | unsigned long index = start >> PAGE_CACHE_SHIFT; | |
1380 | unsigned long end_index = end >> PAGE_CACHE_SHIFT; | |
1381 | unsigned long nr_pages = end_index - index + 1; | |
1382 | int i; | |
771ed689 | 1383 | int clear_bits = 0; |
c8b97818 | 1384 | |
a791e35e | 1385 | if (op & EXTENT_CLEAR_UNLOCK) |
771ed689 | 1386 | clear_bits |= EXTENT_LOCKED; |
a791e35e | 1387 | if (op & EXTENT_CLEAR_DIRTY) |
c8b97818 CM |
1388 | clear_bits |= EXTENT_DIRTY; |
1389 | ||
a791e35e | 1390 | if (op & EXTENT_CLEAR_DELALLOC) |
771ed689 CM |
1391 | clear_bits |= EXTENT_DELALLOC; |
1392 | ||
2c64c53d | 1393 | clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS); |
32c00aff JB |
1394 | if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY | |
1395 | EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK | | |
1396 | EXTENT_SET_PRIVATE2))) | |
771ed689 | 1397 | return 0; |
c8b97818 | 1398 | |
d397712b | 1399 | while (nr_pages > 0) { |
c8b97818 | 1400 | ret = find_get_pages_contig(inode->i_mapping, index, |
5b050f04 CM |
1401 | min_t(unsigned long, |
1402 | nr_pages, ARRAY_SIZE(pages)), pages); | |
c8b97818 | 1403 | for (i = 0; i < ret; i++) { |
8b62b72b | 1404 | |
a791e35e | 1405 | if (op & EXTENT_SET_PRIVATE2) |
8b62b72b CM |
1406 | SetPagePrivate2(pages[i]); |
1407 | ||
c8b97818 CM |
1408 | if (pages[i] == locked_page) { |
1409 | page_cache_release(pages[i]); | |
1410 | continue; | |
1411 | } | |
a791e35e | 1412 | if (op & EXTENT_CLEAR_DIRTY) |
c8b97818 | 1413 | clear_page_dirty_for_io(pages[i]); |
a791e35e | 1414 | if (op & EXTENT_SET_WRITEBACK) |
c8b97818 | 1415 | set_page_writeback(pages[i]); |
a791e35e | 1416 | if (op & EXTENT_END_WRITEBACK) |
c8b97818 | 1417 | end_page_writeback(pages[i]); |
a791e35e | 1418 | if (op & EXTENT_CLEAR_UNLOCK_PAGE) |
771ed689 | 1419 | unlock_page(pages[i]); |
c8b97818 CM |
1420 | page_cache_release(pages[i]); |
1421 | } | |
1422 | nr_pages -= ret; | |
1423 | index += ret; | |
1424 | cond_resched(); | |
1425 | } | |
1426 | return 0; | |
1427 | } | |
c8b97818 | 1428 | |
d352ac68 CM |
1429 | /* |
1430 | * count the number of bytes in the tree that have a given bit(s) | |
1431 | * set. This can be fairly slow, except for EXTENT_DIRTY which is | |
1432 | * cached. The total number found is returned. | |
1433 | */ | |
d1310b2e CM |
1434 | u64 count_range_bits(struct extent_io_tree *tree, |
1435 | u64 *start, u64 search_end, u64 max_bytes, | |
ec29ed5b | 1436 | unsigned long bits, int contig) |
d1310b2e CM |
1437 | { |
1438 | struct rb_node *node; | |
1439 | struct extent_state *state; | |
1440 | u64 cur_start = *start; | |
1441 | u64 total_bytes = 0; | |
ec29ed5b | 1442 | u64 last = 0; |
d1310b2e CM |
1443 | int found = 0; |
1444 | ||
1445 | if (search_end <= cur_start) { | |
d1310b2e CM |
1446 | WARN_ON(1); |
1447 | return 0; | |
1448 | } | |
1449 | ||
cad321ad | 1450 | spin_lock(&tree->lock); |
d1310b2e CM |
1451 | if (cur_start == 0 && bits == EXTENT_DIRTY) { |
1452 | total_bytes = tree->dirty_bytes; | |
1453 | goto out; | |
1454 | } | |
1455 | /* | |
1456 | * this search will find all the extents that end after | |
1457 | * our range starts. | |
1458 | */ | |
80ea96b1 | 1459 | node = tree_search(tree, cur_start); |
d397712b | 1460 | if (!node) |
d1310b2e | 1461 | goto out; |
d1310b2e | 1462 | |
d397712b | 1463 | while (1) { |
d1310b2e CM |
1464 | state = rb_entry(node, struct extent_state, rb_node); |
1465 | if (state->start > search_end) | |
1466 | break; | |
ec29ed5b CM |
1467 | if (contig && found && state->start > last + 1) |
1468 | break; | |
1469 | if (state->end >= cur_start && (state->state & bits) == bits) { | |
d1310b2e CM |
1470 | total_bytes += min(search_end, state->end) + 1 - |
1471 | max(cur_start, state->start); | |
1472 | if (total_bytes >= max_bytes) | |
1473 | break; | |
1474 | if (!found) { | |
1475 | *start = state->start; | |
1476 | found = 1; | |
1477 | } | |
ec29ed5b CM |
1478 | last = state->end; |
1479 | } else if (contig && found) { | |
1480 | break; | |
d1310b2e CM |
1481 | } |
1482 | node = rb_next(node); | |
1483 | if (!node) | |
1484 | break; | |
1485 | } | |
1486 | out: | |
cad321ad | 1487 | spin_unlock(&tree->lock); |
d1310b2e CM |
1488 | return total_bytes; |
1489 | } | |
b2950863 | 1490 | |
d352ac68 CM |
1491 | /* |
1492 | * set the private field for a given byte offset in the tree. If there isn't | |
1493 | * an extent_state there already, this does nothing. | |
1494 | */ | |
d1310b2e CM |
1495 | int set_state_private(struct extent_io_tree *tree, u64 start, u64 private) |
1496 | { | |
1497 | struct rb_node *node; | |
1498 | struct extent_state *state; | |
1499 | int ret = 0; | |
1500 | ||
cad321ad | 1501 | spin_lock(&tree->lock); |
d1310b2e CM |
1502 | /* |
1503 | * this search will find all the extents that end after | |
1504 | * our range starts. | |
1505 | */ | |
80ea96b1 | 1506 | node = tree_search(tree, start); |
2b114d1d | 1507 | if (!node) { |
d1310b2e CM |
1508 | ret = -ENOENT; |
1509 | goto out; | |
1510 | } | |
1511 | state = rb_entry(node, struct extent_state, rb_node); | |
1512 | if (state->start != start) { | |
1513 | ret = -ENOENT; | |
1514 | goto out; | |
1515 | } | |
1516 | state->private = private; | |
1517 | out: | |
cad321ad | 1518 | spin_unlock(&tree->lock); |
d1310b2e CM |
1519 | return ret; |
1520 | } | |
1521 | ||
1522 | int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private) | |
1523 | { | |
1524 | struct rb_node *node; | |
1525 | struct extent_state *state; | |
1526 | int ret = 0; | |
1527 | ||
cad321ad | 1528 | spin_lock(&tree->lock); |
d1310b2e CM |
1529 | /* |
1530 | * this search will find all the extents that end after | |
1531 | * our range starts. | |
1532 | */ | |
80ea96b1 | 1533 | node = tree_search(tree, start); |
2b114d1d | 1534 | if (!node) { |
d1310b2e CM |
1535 | ret = -ENOENT; |
1536 | goto out; | |
1537 | } | |
1538 | state = rb_entry(node, struct extent_state, rb_node); | |
1539 | if (state->start != start) { | |
1540 | ret = -ENOENT; | |
1541 | goto out; | |
1542 | } | |
1543 | *private = state->private; | |
1544 | out: | |
cad321ad | 1545 | spin_unlock(&tree->lock); |
d1310b2e CM |
1546 | return ret; |
1547 | } | |
1548 | ||
1549 | /* | |
1550 | * searches a range in the state tree for a given mask. | |
70dec807 | 1551 | * If 'filled' == 1, this returns 1 only if every extent in the tree |
d1310b2e CM |
1552 | * has the bits set. Otherwise, 1 is returned if any bit in the |
1553 | * range is found set. | |
1554 | */ | |
1555 | int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, | |
9655d298 | 1556 | int bits, int filled, struct extent_state *cached) |
d1310b2e CM |
1557 | { |
1558 | struct extent_state *state = NULL; | |
1559 | struct rb_node *node; | |
1560 | int bitset = 0; | |
d1310b2e | 1561 | |
cad321ad | 1562 | spin_lock(&tree->lock); |
9655d298 CM |
1563 | if (cached && cached->tree && cached->start == start) |
1564 | node = &cached->rb_node; | |
1565 | else | |
1566 | node = tree_search(tree, start); | |
d1310b2e CM |
1567 | while (node && start <= end) { |
1568 | state = rb_entry(node, struct extent_state, rb_node); | |
1569 | ||
1570 | if (filled && state->start > start) { | |
1571 | bitset = 0; | |
1572 | break; | |
1573 | } | |
1574 | ||
1575 | if (state->start > end) | |
1576 | break; | |
1577 | ||
1578 | if (state->state & bits) { | |
1579 | bitset = 1; | |
1580 | if (!filled) | |
1581 | break; | |
1582 | } else if (filled) { | |
1583 | bitset = 0; | |
1584 | break; | |
1585 | } | |
46562cec CM |
1586 | |
1587 | if (state->end == (u64)-1) | |
1588 | break; | |
1589 | ||
d1310b2e CM |
1590 | start = state->end + 1; |
1591 | if (start > end) | |
1592 | break; | |
1593 | node = rb_next(node); | |
1594 | if (!node) { | |
1595 | if (filled) | |
1596 | bitset = 0; | |
1597 | break; | |
1598 | } | |
1599 | } | |
cad321ad | 1600 | spin_unlock(&tree->lock); |
d1310b2e CM |
1601 | return bitset; |
1602 | } | |
d1310b2e CM |
1603 | |
1604 | /* | |
1605 | * helper function to set a given page up to date if all the | |
1606 | * extents in the tree for that page are up to date | |
1607 | */ | |
1608 | static int check_page_uptodate(struct extent_io_tree *tree, | |
1609 | struct page *page) | |
1610 | { | |
1611 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; | |
1612 | u64 end = start + PAGE_CACHE_SIZE - 1; | |
9655d298 | 1613 | if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL)) |
d1310b2e CM |
1614 | SetPageUptodate(page); |
1615 | return 0; | |
1616 | } | |
1617 | ||
1618 | /* | |
1619 | * helper function to unlock a page if all the extents in the tree | |
1620 | * for that page are unlocked | |
1621 | */ | |
1622 | static int check_page_locked(struct extent_io_tree *tree, | |
1623 | struct page *page) | |
1624 | { | |
1625 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; | |
1626 | u64 end = start + PAGE_CACHE_SIZE - 1; | |
9655d298 | 1627 | if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0, NULL)) |
d1310b2e CM |
1628 | unlock_page(page); |
1629 | return 0; | |
1630 | } | |
1631 | ||
1632 | /* | |
1633 | * helper function to end page writeback if all the extents | |
1634 | * in the tree for that page are done with writeback | |
1635 | */ | |
1636 | static int check_page_writeback(struct extent_io_tree *tree, | |
1637 | struct page *page) | |
1638 | { | |
1edbb734 | 1639 | end_page_writeback(page); |
d1310b2e CM |
1640 | return 0; |
1641 | } | |
1642 | ||
1643 | /* lots and lots of room for performance fixes in the end_bio funcs */ | |
1644 | ||
1645 | /* | |
1646 | * after a writepage IO is done, we need to: | |
1647 | * clear the uptodate bits on error | |
1648 | * clear the writeback bits in the extent tree for this IO | |
1649 | * end_page_writeback if the page has no more pending IO | |
1650 | * | |
1651 | * Scheduling is not allowed, so the extent state tree is expected | |
1652 | * to have one and only one object corresponding to this IO. | |
1653 | */ | |
d1310b2e | 1654 | static void end_bio_extent_writepage(struct bio *bio, int err) |
d1310b2e | 1655 | { |
1259ab75 | 1656 | int uptodate = err == 0; |
d1310b2e | 1657 | struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; |
902b22f3 | 1658 | struct extent_io_tree *tree; |
d1310b2e CM |
1659 | u64 start; |
1660 | u64 end; | |
1661 | int whole_page; | |
1259ab75 | 1662 | int ret; |
d1310b2e | 1663 | |
d1310b2e CM |
1664 | do { |
1665 | struct page *page = bvec->bv_page; | |
902b22f3 DW |
1666 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
1667 | ||
d1310b2e CM |
1668 | start = ((u64)page->index << PAGE_CACHE_SHIFT) + |
1669 | bvec->bv_offset; | |
1670 | end = start + bvec->bv_len - 1; | |
1671 | ||
1672 | if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) | |
1673 | whole_page = 1; | |
1674 | else | |
1675 | whole_page = 0; | |
1676 | ||
1677 | if (--bvec >= bio->bi_io_vec) | |
1678 | prefetchw(&bvec->bv_page->flags); | |
1259ab75 CM |
1679 | if (tree->ops && tree->ops->writepage_end_io_hook) { |
1680 | ret = tree->ops->writepage_end_io_hook(page, start, | |
902b22f3 | 1681 | end, NULL, uptodate); |
1259ab75 CM |
1682 | if (ret) |
1683 | uptodate = 0; | |
1684 | } | |
1685 | ||
1686 | if (!uptodate && tree->ops && | |
1687 | tree->ops->writepage_io_failed_hook) { | |
1688 | ret = tree->ops->writepage_io_failed_hook(bio, page, | |
902b22f3 | 1689 | start, end, NULL); |
1259ab75 | 1690 | if (ret == 0) { |
1259ab75 CM |
1691 | uptodate = (err == 0); |
1692 | continue; | |
1693 | } | |
1694 | } | |
1695 | ||
d1310b2e | 1696 | if (!uptodate) { |
2ac55d41 | 1697 | clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS); |
d1310b2e CM |
1698 | ClearPageUptodate(page); |
1699 | SetPageError(page); | |
1700 | } | |
70dec807 | 1701 | |
d1310b2e CM |
1702 | if (whole_page) |
1703 | end_page_writeback(page); | |
1704 | else | |
1705 | check_page_writeback(tree, page); | |
d1310b2e | 1706 | } while (bvec >= bio->bi_io_vec); |
2b1f55b0 | 1707 | |
d1310b2e | 1708 | bio_put(bio); |
d1310b2e CM |
1709 | } |
1710 | ||
1711 | /* | |
1712 | * after a readpage IO is done, we need to: | |
1713 | * clear the uptodate bits on error | |
1714 | * set the uptodate bits if things worked | |
1715 | * set the page up to date if all extents in the tree are uptodate | |
1716 | * clear the lock bit in the extent tree | |
1717 | * unlock the page if there are no other extents locked for it | |
1718 | * | |
1719 | * Scheduling is not allowed, so the extent state tree is expected | |
1720 | * to have one and only one object corresponding to this IO. | |
1721 | */ | |
d1310b2e | 1722 | static void end_bio_extent_readpage(struct bio *bio, int err) |
d1310b2e CM |
1723 | { |
1724 | int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); | |
4125bf76 CM |
1725 | struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1; |
1726 | struct bio_vec *bvec = bio->bi_io_vec; | |
902b22f3 | 1727 | struct extent_io_tree *tree; |
d1310b2e CM |
1728 | u64 start; |
1729 | u64 end; | |
1730 | int whole_page; | |
1731 | int ret; | |
1732 | ||
d20f7043 CM |
1733 | if (err) |
1734 | uptodate = 0; | |
1735 | ||
d1310b2e CM |
1736 | do { |
1737 | struct page *page = bvec->bv_page; | |
902b22f3 DW |
1738 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
1739 | ||
d1310b2e CM |
1740 | start = ((u64)page->index << PAGE_CACHE_SHIFT) + |
1741 | bvec->bv_offset; | |
1742 | end = start + bvec->bv_len - 1; | |
1743 | ||
1744 | if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE) | |
1745 | whole_page = 1; | |
1746 | else | |
1747 | whole_page = 0; | |
1748 | ||
4125bf76 | 1749 | if (++bvec <= bvec_end) |
d1310b2e CM |
1750 | prefetchw(&bvec->bv_page->flags); |
1751 | ||
1752 | if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) { | |
70dec807 | 1753 | ret = tree->ops->readpage_end_io_hook(page, start, end, |
902b22f3 | 1754 | NULL); |
d1310b2e CM |
1755 | if (ret) |
1756 | uptodate = 0; | |
1757 | } | |
7e38326f CM |
1758 | if (!uptodate && tree->ops && |
1759 | tree->ops->readpage_io_failed_hook) { | |
1760 | ret = tree->ops->readpage_io_failed_hook(bio, page, | |
902b22f3 | 1761 | start, end, NULL); |
7e38326f | 1762 | if (ret == 0) { |
3b951516 CM |
1763 | uptodate = |
1764 | test_bit(BIO_UPTODATE, &bio->bi_flags); | |
d20f7043 CM |
1765 | if (err) |
1766 | uptodate = 0; | |
7e38326f CM |
1767 | continue; |
1768 | } | |
1769 | } | |
d1310b2e | 1770 | |
771ed689 | 1771 | if (uptodate) { |
902b22f3 DW |
1772 | set_extent_uptodate(tree, start, end, |
1773 | GFP_ATOMIC); | |
771ed689 | 1774 | } |
902b22f3 | 1775 | unlock_extent(tree, start, end, GFP_ATOMIC); |
d1310b2e | 1776 | |
70dec807 CM |
1777 | if (whole_page) { |
1778 | if (uptodate) { | |
1779 | SetPageUptodate(page); | |
1780 | } else { | |
1781 | ClearPageUptodate(page); | |
1782 | SetPageError(page); | |
1783 | } | |
d1310b2e | 1784 | unlock_page(page); |
70dec807 CM |
1785 | } else { |
1786 | if (uptodate) { | |
1787 | check_page_uptodate(tree, page); | |
1788 | } else { | |
1789 | ClearPageUptodate(page); | |
1790 | SetPageError(page); | |
1791 | } | |
d1310b2e | 1792 | check_page_locked(tree, page); |
70dec807 | 1793 | } |
4125bf76 | 1794 | } while (bvec <= bvec_end); |
d1310b2e CM |
1795 | |
1796 | bio_put(bio); | |
d1310b2e CM |
1797 | } |
1798 | ||
1799 | /* | |
1800 | * IO done from prepare_write is pretty simple, we just unlock | |
1801 | * the structs in the extent tree when done, and set the uptodate bits | |
1802 | * as appropriate. | |
1803 | */ | |
d1310b2e | 1804 | static void end_bio_extent_preparewrite(struct bio *bio, int err) |
d1310b2e CM |
1805 | { |
1806 | const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags); | |
1807 | struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; | |
902b22f3 | 1808 | struct extent_io_tree *tree; |
d1310b2e CM |
1809 | u64 start; |
1810 | u64 end; | |
1811 | ||
d1310b2e CM |
1812 | do { |
1813 | struct page *page = bvec->bv_page; | |
902b22f3 DW |
1814 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
1815 | ||
d1310b2e CM |
1816 | start = ((u64)page->index << PAGE_CACHE_SHIFT) + |
1817 | bvec->bv_offset; | |
1818 | end = start + bvec->bv_len - 1; | |
1819 | ||
1820 | if (--bvec >= bio->bi_io_vec) | |
1821 | prefetchw(&bvec->bv_page->flags); | |
1822 | ||
1823 | if (uptodate) { | |
1824 | set_extent_uptodate(tree, start, end, GFP_ATOMIC); | |
1825 | } else { | |
1826 | ClearPageUptodate(page); | |
1827 | SetPageError(page); | |
1828 | } | |
1829 | ||
1830 | unlock_extent(tree, start, end, GFP_ATOMIC); | |
1831 | ||
1832 | } while (bvec >= bio->bi_io_vec); | |
1833 | ||
1834 | bio_put(bio); | |
d1310b2e CM |
1835 | } |
1836 | ||
88f794ed MX |
1837 | struct bio * |
1838 | btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, | |
1839 | gfp_t gfp_flags) | |
d1310b2e CM |
1840 | { |
1841 | struct bio *bio; | |
1842 | ||
1843 | bio = bio_alloc(gfp_flags, nr_vecs); | |
1844 | ||
1845 | if (bio == NULL && (current->flags & PF_MEMALLOC)) { | |
1846 | while (!bio && (nr_vecs /= 2)) | |
1847 | bio = bio_alloc(gfp_flags, nr_vecs); | |
1848 | } | |
1849 | ||
1850 | if (bio) { | |
e1c4b745 | 1851 | bio->bi_size = 0; |
d1310b2e CM |
1852 | bio->bi_bdev = bdev; |
1853 | bio->bi_sector = first_sector; | |
1854 | } | |
1855 | return bio; | |
1856 | } | |
1857 | ||
c8b97818 CM |
1858 | static int submit_one_bio(int rw, struct bio *bio, int mirror_num, |
1859 | unsigned long bio_flags) | |
d1310b2e | 1860 | { |
d1310b2e | 1861 | int ret = 0; |
70dec807 CM |
1862 | struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1; |
1863 | struct page *page = bvec->bv_page; | |
1864 | struct extent_io_tree *tree = bio->bi_private; | |
70dec807 | 1865 | u64 start; |
70dec807 CM |
1866 | |
1867 | start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset; | |
70dec807 | 1868 | |
902b22f3 | 1869 | bio->bi_private = NULL; |
d1310b2e CM |
1870 | |
1871 | bio_get(bio); | |
1872 | ||
065631f6 | 1873 | if (tree->ops && tree->ops->submit_bio_hook) |
6b82ce8d | 1874 | ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio, |
eaf25d93 | 1875 | mirror_num, bio_flags, start); |
0b86a832 CM |
1876 | else |
1877 | submit_bio(rw, bio); | |
d1310b2e CM |
1878 | if (bio_flagged(bio, BIO_EOPNOTSUPP)) |
1879 | ret = -EOPNOTSUPP; | |
1880 | bio_put(bio); | |
1881 | return ret; | |
1882 | } | |
1883 | ||
1884 | static int submit_extent_page(int rw, struct extent_io_tree *tree, | |
1885 | struct page *page, sector_t sector, | |
1886 | size_t size, unsigned long offset, | |
1887 | struct block_device *bdev, | |
1888 | struct bio **bio_ret, | |
1889 | unsigned long max_pages, | |
f188591e | 1890 | bio_end_io_t end_io_func, |
c8b97818 CM |
1891 | int mirror_num, |
1892 | unsigned long prev_bio_flags, | |
1893 | unsigned long bio_flags) | |
d1310b2e CM |
1894 | { |
1895 | int ret = 0; | |
1896 | struct bio *bio; | |
1897 | int nr; | |
c8b97818 CM |
1898 | int contig = 0; |
1899 | int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED; | |
1900 | int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED; | |
5b050f04 | 1901 | size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE); |
d1310b2e CM |
1902 | |
1903 | if (bio_ret && *bio_ret) { | |
1904 | bio = *bio_ret; | |
c8b97818 CM |
1905 | if (old_compressed) |
1906 | contig = bio->bi_sector == sector; | |
1907 | else | |
1908 | contig = bio->bi_sector + (bio->bi_size >> 9) == | |
1909 | sector; | |
1910 | ||
1911 | if (prev_bio_flags != bio_flags || !contig || | |
239b14b3 | 1912 | (tree->ops && tree->ops->merge_bio_hook && |
c8b97818 CM |
1913 | tree->ops->merge_bio_hook(page, offset, page_size, bio, |
1914 | bio_flags)) || | |
1915 | bio_add_page(bio, page, page_size, offset) < page_size) { | |
1916 | ret = submit_one_bio(rw, bio, mirror_num, | |
1917 | prev_bio_flags); | |
d1310b2e CM |
1918 | bio = NULL; |
1919 | } else { | |
1920 | return 0; | |
1921 | } | |
1922 | } | |
c8b97818 CM |
1923 | if (this_compressed) |
1924 | nr = BIO_MAX_PAGES; | |
1925 | else | |
1926 | nr = bio_get_nr_vecs(bdev); | |
1927 | ||
88f794ed | 1928 | bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); |
5df67083 TI |
1929 | if (!bio) |
1930 | return -ENOMEM; | |
70dec807 | 1931 | |
c8b97818 | 1932 | bio_add_page(bio, page, page_size, offset); |
d1310b2e CM |
1933 | bio->bi_end_io = end_io_func; |
1934 | bio->bi_private = tree; | |
70dec807 | 1935 | |
d397712b | 1936 | if (bio_ret) |
d1310b2e | 1937 | *bio_ret = bio; |
d397712b | 1938 | else |
c8b97818 | 1939 | ret = submit_one_bio(rw, bio, mirror_num, bio_flags); |
d1310b2e CM |
1940 | |
1941 | return ret; | |
1942 | } | |
1943 | ||
1944 | void set_page_extent_mapped(struct page *page) | |
1945 | { | |
1946 | if (!PagePrivate(page)) { | |
1947 | SetPagePrivate(page); | |
d1310b2e | 1948 | page_cache_get(page); |
6af118ce | 1949 | set_page_private(page, EXTENT_PAGE_PRIVATE); |
d1310b2e CM |
1950 | } |
1951 | } | |
1952 | ||
b2950863 | 1953 | static void set_page_extent_head(struct page *page, unsigned long len) |
d1310b2e | 1954 | { |
eb14ab8e | 1955 | WARN_ON(!PagePrivate(page)); |
d1310b2e CM |
1956 | set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2); |
1957 | } | |
1958 | ||
1959 | /* | |
1960 | * basic readpage implementation. Locked extent state structs are inserted | |
1961 | * into the tree that are removed when the IO is done (by the end_io | |
1962 | * handlers) | |
1963 | */ | |
1964 | static int __extent_read_full_page(struct extent_io_tree *tree, | |
1965 | struct page *page, | |
1966 | get_extent_t *get_extent, | |
c8b97818 CM |
1967 | struct bio **bio, int mirror_num, |
1968 | unsigned long *bio_flags) | |
d1310b2e CM |
1969 | { |
1970 | struct inode *inode = page->mapping->host; | |
1971 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; | |
1972 | u64 page_end = start + PAGE_CACHE_SIZE - 1; | |
1973 | u64 end; | |
1974 | u64 cur = start; | |
1975 | u64 extent_offset; | |
1976 | u64 last_byte = i_size_read(inode); | |
1977 | u64 block_start; | |
1978 | u64 cur_end; | |
1979 | sector_t sector; | |
1980 | struct extent_map *em; | |
1981 | struct block_device *bdev; | |
11c65dcc | 1982 | struct btrfs_ordered_extent *ordered; |
d1310b2e CM |
1983 | int ret; |
1984 | int nr = 0; | |
1985 | size_t page_offset = 0; | |
1986 | size_t iosize; | |
c8b97818 | 1987 | size_t disk_io_size; |
d1310b2e | 1988 | size_t blocksize = inode->i_sb->s_blocksize; |
c8b97818 | 1989 | unsigned long this_bio_flag = 0; |
d1310b2e CM |
1990 | |
1991 | set_page_extent_mapped(page); | |
1992 | ||
1993 | end = page_end; | |
11c65dcc JB |
1994 | while (1) { |
1995 | lock_extent(tree, start, end, GFP_NOFS); | |
1996 | ordered = btrfs_lookup_ordered_extent(inode, start); | |
1997 | if (!ordered) | |
1998 | break; | |
1999 | unlock_extent(tree, start, end, GFP_NOFS); | |
2000 | btrfs_start_ordered_extent(inode, ordered, 1); | |
2001 | btrfs_put_ordered_extent(ordered); | |
2002 | } | |
d1310b2e | 2003 | |
c8b97818 CM |
2004 | if (page->index == last_byte >> PAGE_CACHE_SHIFT) { |
2005 | char *userpage; | |
2006 | size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1); | |
2007 | ||
2008 | if (zero_offset) { | |
2009 | iosize = PAGE_CACHE_SIZE - zero_offset; | |
2010 | userpage = kmap_atomic(page, KM_USER0); | |
2011 | memset(userpage + zero_offset, 0, iosize); | |
2012 | flush_dcache_page(page); | |
2013 | kunmap_atomic(userpage, KM_USER0); | |
2014 | } | |
2015 | } | |
d1310b2e CM |
2016 | while (cur <= end) { |
2017 | if (cur >= last_byte) { | |
2018 | char *userpage; | |
2019 | iosize = PAGE_CACHE_SIZE - page_offset; | |
2020 | userpage = kmap_atomic(page, KM_USER0); | |
2021 | memset(userpage + page_offset, 0, iosize); | |
2022 | flush_dcache_page(page); | |
2023 | kunmap_atomic(userpage, KM_USER0); | |
2024 | set_extent_uptodate(tree, cur, cur + iosize - 1, | |
2025 | GFP_NOFS); | |
2026 | unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); | |
2027 | break; | |
2028 | } | |
2029 | em = get_extent(inode, page, page_offset, cur, | |
2030 | end - cur + 1, 0); | |
2031 | if (IS_ERR(em) || !em) { | |
2032 | SetPageError(page); | |
2033 | unlock_extent(tree, cur, end, GFP_NOFS); | |
2034 | break; | |
2035 | } | |
d1310b2e CM |
2036 | extent_offset = cur - em->start; |
2037 | BUG_ON(extent_map_end(em) <= cur); | |
2038 | BUG_ON(end < cur); | |
2039 | ||
261507a0 | 2040 | if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) { |
c8b97818 | 2041 | this_bio_flag = EXTENT_BIO_COMPRESSED; |
261507a0 LZ |
2042 | extent_set_compress_type(&this_bio_flag, |
2043 | em->compress_type); | |
2044 | } | |
c8b97818 | 2045 | |
d1310b2e CM |
2046 | iosize = min(extent_map_end(em) - cur, end - cur + 1); |
2047 | cur_end = min(extent_map_end(em) - 1, end); | |
2048 | iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); | |
c8b97818 CM |
2049 | if (this_bio_flag & EXTENT_BIO_COMPRESSED) { |
2050 | disk_io_size = em->block_len; | |
2051 | sector = em->block_start >> 9; | |
2052 | } else { | |
2053 | sector = (em->block_start + extent_offset) >> 9; | |
2054 | disk_io_size = iosize; | |
2055 | } | |
d1310b2e CM |
2056 | bdev = em->bdev; |
2057 | block_start = em->block_start; | |
d899e052 YZ |
2058 | if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) |
2059 | block_start = EXTENT_MAP_HOLE; | |
d1310b2e CM |
2060 | free_extent_map(em); |
2061 | em = NULL; | |
2062 | ||
2063 | /* we've found a hole, just zero and go on */ | |
2064 | if (block_start == EXTENT_MAP_HOLE) { | |
2065 | char *userpage; | |
2066 | userpage = kmap_atomic(page, KM_USER0); | |
2067 | memset(userpage + page_offset, 0, iosize); | |
2068 | flush_dcache_page(page); | |
2069 | kunmap_atomic(userpage, KM_USER0); | |
2070 | ||
2071 | set_extent_uptodate(tree, cur, cur + iosize - 1, | |
2072 | GFP_NOFS); | |
2073 | unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); | |
2074 | cur = cur + iosize; | |
2075 | page_offset += iosize; | |
2076 | continue; | |
2077 | } | |
2078 | /* the get_extent function already copied into the page */ | |
9655d298 CM |
2079 | if (test_range_bit(tree, cur, cur_end, |
2080 | EXTENT_UPTODATE, 1, NULL)) { | |
a1b32a59 | 2081 | check_page_uptodate(tree, page); |
d1310b2e CM |
2082 | unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); |
2083 | cur = cur + iosize; | |
2084 | page_offset += iosize; | |
2085 | continue; | |
2086 | } | |
70dec807 CM |
2087 | /* we have an inline extent but it didn't get marked up |
2088 | * to date. Error out | |
2089 | */ | |
2090 | if (block_start == EXTENT_MAP_INLINE) { | |
2091 | SetPageError(page); | |
2092 | unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS); | |
2093 | cur = cur + iosize; | |
2094 | page_offset += iosize; | |
2095 | continue; | |
2096 | } | |
d1310b2e CM |
2097 | |
2098 | ret = 0; | |
2099 | if (tree->ops && tree->ops->readpage_io_hook) { | |
2100 | ret = tree->ops->readpage_io_hook(page, cur, | |
2101 | cur + iosize - 1); | |
2102 | } | |
2103 | if (!ret) { | |
89642229 CM |
2104 | unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1; |
2105 | pnr -= page->index; | |
d1310b2e | 2106 | ret = submit_extent_page(READ, tree, page, |
c8b97818 | 2107 | sector, disk_io_size, page_offset, |
89642229 | 2108 | bdev, bio, pnr, |
c8b97818 CM |
2109 | end_bio_extent_readpage, mirror_num, |
2110 | *bio_flags, | |
2111 | this_bio_flag); | |
89642229 | 2112 | nr++; |
c8b97818 | 2113 | *bio_flags = this_bio_flag; |
d1310b2e CM |
2114 | } |
2115 | if (ret) | |
2116 | SetPageError(page); | |
2117 | cur = cur + iosize; | |
2118 | page_offset += iosize; | |
d1310b2e CM |
2119 | } |
2120 | if (!nr) { | |
2121 | if (!PageError(page)) | |
2122 | SetPageUptodate(page); | |
2123 | unlock_page(page); | |
2124 | } | |
2125 | return 0; | |
2126 | } | |
2127 | ||
2128 | int extent_read_full_page(struct extent_io_tree *tree, struct page *page, | |
2129 | get_extent_t *get_extent) | |
2130 | { | |
2131 | struct bio *bio = NULL; | |
c8b97818 | 2132 | unsigned long bio_flags = 0; |
d1310b2e CM |
2133 | int ret; |
2134 | ||
c8b97818 CM |
2135 | ret = __extent_read_full_page(tree, page, get_extent, &bio, 0, |
2136 | &bio_flags); | |
d1310b2e | 2137 | if (bio) |
6b82ce8d | 2138 | ret = submit_one_bio(READ, bio, 0, bio_flags); |
d1310b2e CM |
2139 | return ret; |
2140 | } | |
d1310b2e | 2141 | |
11c8349b CM |
2142 | static noinline void update_nr_written(struct page *page, |
2143 | struct writeback_control *wbc, | |
2144 | unsigned long nr_written) | |
2145 | { | |
2146 | wbc->nr_to_write -= nr_written; | |
2147 | if (wbc->range_cyclic || (wbc->nr_to_write > 0 && | |
2148 | wbc->range_start == 0 && wbc->range_end == LLONG_MAX)) | |
2149 | page->mapping->writeback_index = page->index + nr_written; | |
2150 | } | |
2151 | ||
d1310b2e CM |
2152 | /* |
2153 | * the writepage semantics are similar to regular writepage. extent | |
2154 | * records are inserted to lock ranges in the tree, and as dirty areas | |
2155 | * are found, they are marked writeback. Then the lock bits are removed | |
2156 | * and the end_io handler clears the writeback ranges | |
2157 | */ | |
2158 | static int __extent_writepage(struct page *page, struct writeback_control *wbc, | |
2159 | void *data) | |
2160 | { | |
2161 | struct inode *inode = page->mapping->host; | |
2162 | struct extent_page_data *epd = data; | |
2163 | struct extent_io_tree *tree = epd->tree; | |
2164 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2165 | u64 delalloc_start; | |
2166 | u64 page_end = start + PAGE_CACHE_SIZE - 1; | |
2167 | u64 end; | |
2168 | u64 cur = start; | |
2169 | u64 extent_offset; | |
2170 | u64 last_byte = i_size_read(inode); | |
2171 | u64 block_start; | |
2172 | u64 iosize; | |
2173 | sector_t sector; | |
2c64c53d | 2174 | struct extent_state *cached_state = NULL; |
d1310b2e CM |
2175 | struct extent_map *em; |
2176 | struct block_device *bdev; | |
2177 | int ret; | |
2178 | int nr = 0; | |
7f3c74fb | 2179 | size_t pg_offset = 0; |
d1310b2e CM |
2180 | size_t blocksize; |
2181 | loff_t i_size = i_size_read(inode); | |
2182 | unsigned long end_index = i_size >> PAGE_CACHE_SHIFT; | |
2183 | u64 nr_delalloc; | |
2184 | u64 delalloc_end; | |
c8b97818 CM |
2185 | int page_started; |
2186 | int compressed; | |
ffbd517d | 2187 | int write_flags; |
771ed689 | 2188 | unsigned long nr_written = 0; |
d1310b2e | 2189 | |
ffbd517d CM |
2190 | if (wbc->sync_mode == WB_SYNC_ALL) |
2191 | write_flags = WRITE_SYNC_PLUG; | |
2192 | else | |
2193 | write_flags = WRITE; | |
2194 | ||
d1310b2e | 2195 | WARN_ON(!PageLocked(page)); |
7f3c74fb | 2196 | pg_offset = i_size & (PAGE_CACHE_SIZE - 1); |
211c17f5 | 2197 | if (page->index > end_index || |
7f3c74fb | 2198 | (page->index == end_index && !pg_offset)) { |
39be25cd | 2199 | page->mapping->a_ops->invalidatepage(page, 0); |
d1310b2e CM |
2200 | unlock_page(page); |
2201 | return 0; | |
2202 | } | |
2203 | ||
2204 | if (page->index == end_index) { | |
2205 | char *userpage; | |
2206 | ||
d1310b2e | 2207 | userpage = kmap_atomic(page, KM_USER0); |
7f3c74fb CM |
2208 | memset(userpage + pg_offset, 0, |
2209 | PAGE_CACHE_SIZE - pg_offset); | |
d1310b2e | 2210 | kunmap_atomic(userpage, KM_USER0); |
211c17f5 | 2211 | flush_dcache_page(page); |
d1310b2e | 2212 | } |
7f3c74fb | 2213 | pg_offset = 0; |
d1310b2e CM |
2214 | |
2215 | set_page_extent_mapped(page); | |
2216 | ||
2217 | delalloc_start = start; | |
2218 | delalloc_end = 0; | |
c8b97818 | 2219 | page_started = 0; |
771ed689 | 2220 | if (!epd->extent_locked) { |
f85d7d6c | 2221 | u64 delalloc_to_write = 0; |
11c8349b CM |
2222 | /* |
2223 | * make sure the wbc mapping index is at least updated | |
2224 | * to this page. | |
2225 | */ | |
2226 | update_nr_written(page, wbc, 0); | |
2227 | ||
d397712b | 2228 | while (delalloc_end < page_end) { |
771ed689 | 2229 | nr_delalloc = find_lock_delalloc_range(inode, tree, |
c8b97818 CM |
2230 | page, |
2231 | &delalloc_start, | |
d1310b2e CM |
2232 | &delalloc_end, |
2233 | 128 * 1024 * 1024); | |
771ed689 CM |
2234 | if (nr_delalloc == 0) { |
2235 | delalloc_start = delalloc_end + 1; | |
2236 | continue; | |
2237 | } | |
2238 | tree->ops->fill_delalloc(inode, page, delalloc_start, | |
2239 | delalloc_end, &page_started, | |
2240 | &nr_written); | |
f85d7d6c CM |
2241 | /* |
2242 | * delalloc_end is already one less than the total | |
2243 | * length, so we don't subtract one from | |
2244 | * PAGE_CACHE_SIZE | |
2245 | */ | |
2246 | delalloc_to_write += (delalloc_end - delalloc_start + | |
2247 | PAGE_CACHE_SIZE) >> | |
2248 | PAGE_CACHE_SHIFT; | |
d1310b2e | 2249 | delalloc_start = delalloc_end + 1; |
d1310b2e | 2250 | } |
f85d7d6c CM |
2251 | if (wbc->nr_to_write < delalloc_to_write) { |
2252 | int thresh = 8192; | |
2253 | ||
2254 | if (delalloc_to_write < thresh * 2) | |
2255 | thresh = delalloc_to_write; | |
2256 | wbc->nr_to_write = min_t(u64, delalloc_to_write, | |
2257 | thresh); | |
2258 | } | |
c8b97818 | 2259 | |
771ed689 CM |
2260 | /* did the fill delalloc function already unlock and start |
2261 | * the IO? | |
2262 | */ | |
2263 | if (page_started) { | |
2264 | ret = 0; | |
11c8349b CM |
2265 | /* |
2266 | * we've unlocked the page, so we can't update | |
2267 | * the mapping's writeback index, just update | |
2268 | * nr_to_write. | |
2269 | */ | |
2270 | wbc->nr_to_write -= nr_written; | |
2271 | goto done_unlocked; | |
771ed689 | 2272 | } |
c8b97818 | 2273 | } |
247e743c | 2274 | if (tree->ops && tree->ops->writepage_start_hook) { |
c8b97818 CM |
2275 | ret = tree->ops->writepage_start_hook(page, start, |
2276 | page_end); | |
247e743c | 2277 | if (ret == -EAGAIN) { |
247e743c | 2278 | redirty_page_for_writepage(wbc, page); |
11c8349b | 2279 | update_nr_written(page, wbc, nr_written); |
247e743c | 2280 | unlock_page(page); |
771ed689 | 2281 | ret = 0; |
11c8349b | 2282 | goto done_unlocked; |
247e743c CM |
2283 | } |
2284 | } | |
2285 | ||
11c8349b CM |
2286 | /* |
2287 | * we don't want to touch the inode after unlocking the page, | |
2288 | * so we update the mapping writeback index now | |
2289 | */ | |
2290 | update_nr_written(page, wbc, nr_written + 1); | |
771ed689 | 2291 | |
d1310b2e | 2292 | end = page_end; |
d1310b2e | 2293 | if (last_byte <= start) { |
e6dcd2dc CM |
2294 | if (tree->ops && tree->ops->writepage_end_io_hook) |
2295 | tree->ops->writepage_end_io_hook(page, start, | |
2296 | page_end, NULL, 1); | |
d1310b2e CM |
2297 | goto done; |
2298 | } | |
2299 | ||
d1310b2e CM |
2300 | blocksize = inode->i_sb->s_blocksize; |
2301 | ||
2302 | while (cur <= end) { | |
2303 | if (cur >= last_byte) { | |
e6dcd2dc CM |
2304 | if (tree->ops && tree->ops->writepage_end_io_hook) |
2305 | tree->ops->writepage_end_io_hook(page, cur, | |
2306 | page_end, NULL, 1); | |
d1310b2e CM |
2307 | break; |
2308 | } | |
7f3c74fb | 2309 | em = epd->get_extent(inode, page, pg_offset, cur, |
d1310b2e CM |
2310 | end - cur + 1, 1); |
2311 | if (IS_ERR(em) || !em) { | |
2312 | SetPageError(page); | |
2313 | break; | |
2314 | } | |
2315 | ||
2316 | extent_offset = cur - em->start; | |
2317 | BUG_ON(extent_map_end(em) <= cur); | |
2318 | BUG_ON(end < cur); | |
2319 | iosize = min(extent_map_end(em) - cur, end - cur + 1); | |
2320 | iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1); | |
2321 | sector = (em->block_start + extent_offset) >> 9; | |
2322 | bdev = em->bdev; | |
2323 | block_start = em->block_start; | |
c8b97818 | 2324 | compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags); |
d1310b2e CM |
2325 | free_extent_map(em); |
2326 | em = NULL; | |
2327 | ||
c8b97818 CM |
2328 | /* |
2329 | * compressed and inline extents are written through other | |
2330 | * paths in the FS | |
2331 | */ | |
2332 | if (compressed || block_start == EXTENT_MAP_HOLE || | |
d1310b2e | 2333 | block_start == EXTENT_MAP_INLINE) { |
c8b97818 CM |
2334 | /* |
2335 | * end_io notification does not happen here for | |
2336 | * compressed extents | |
2337 | */ | |
2338 | if (!compressed && tree->ops && | |
2339 | tree->ops->writepage_end_io_hook) | |
e6dcd2dc CM |
2340 | tree->ops->writepage_end_io_hook(page, cur, |
2341 | cur + iosize - 1, | |
2342 | NULL, 1); | |
c8b97818 CM |
2343 | else if (compressed) { |
2344 | /* we don't want to end_page_writeback on | |
2345 | * a compressed extent. this happens | |
2346 | * elsewhere | |
2347 | */ | |
2348 | nr++; | |
2349 | } | |
2350 | ||
2351 | cur += iosize; | |
7f3c74fb | 2352 | pg_offset += iosize; |
d1310b2e CM |
2353 | continue; |
2354 | } | |
d1310b2e CM |
2355 | /* leave this out until we have a page_mkwrite call */ |
2356 | if (0 && !test_range_bit(tree, cur, cur + iosize - 1, | |
9655d298 | 2357 | EXTENT_DIRTY, 0, NULL)) { |
d1310b2e | 2358 | cur = cur + iosize; |
7f3c74fb | 2359 | pg_offset += iosize; |
d1310b2e CM |
2360 | continue; |
2361 | } | |
c8b97818 | 2362 | |
d1310b2e CM |
2363 | if (tree->ops && tree->ops->writepage_io_hook) { |
2364 | ret = tree->ops->writepage_io_hook(page, cur, | |
2365 | cur + iosize - 1); | |
2366 | } else { | |
2367 | ret = 0; | |
2368 | } | |
1259ab75 | 2369 | if (ret) { |
d1310b2e | 2370 | SetPageError(page); |
1259ab75 | 2371 | } else { |
d1310b2e | 2372 | unsigned long max_nr = end_index + 1; |
7f3c74fb | 2373 | |
d1310b2e CM |
2374 | set_range_writeback(tree, cur, cur + iosize - 1); |
2375 | if (!PageWriteback(page)) { | |
d397712b CM |
2376 | printk(KERN_ERR "btrfs warning page %lu not " |
2377 | "writeback, cur %llu end %llu\n", | |
2378 | page->index, (unsigned long long)cur, | |
d1310b2e CM |
2379 | (unsigned long long)end); |
2380 | } | |
2381 | ||
ffbd517d CM |
2382 | ret = submit_extent_page(write_flags, tree, page, |
2383 | sector, iosize, pg_offset, | |
2384 | bdev, &epd->bio, max_nr, | |
c8b97818 CM |
2385 | end_bio_extent_writepage, |
2386 | 0, 0, 0); | |
d1310b2e CM |
2387 | if (ret) |
2388 | SetPageError(page); | |
2389 | } | |
2390 | cur = cur + iosize; | |
7f3c74fb | 2391 | pg_offset += iosize; |
d1310b2e CM |
2392 | nr++; |
2393 | } | |
2394 | done: | |
2395 | if (nr == 0) { | |
2396 | /* make sure the mapping tag for page dirty gets cleared */ | |
2397 | set_page_writeback(page); | |
2398 | end_page_writeback(page); | |
2399 | } | |
d1310b2e | 2400 | unlock_page(page); |
771ed689 | 2401 | |
11c8349b CM |
2402 | done_unlocked: |
2403 | ||
2c64c53d CM |
2404 | /* drop our reference on any cached states */ |
2405 | free_extent_state(cached_state); | |
d1310b2e CM |
2406 | return 0; |
2407 | } | |
2408 | ||
d1310b2e | 2409 | /** |
4bef0848 | 2410 | * write_cache_pages - walk the list of dirty pages of the given address space and write all of them. |
d1310b2e CM |
2411 | * @mapping: address space structure to write |
2412 | * @wbc: subtract the number of written pages from *@wbc->nr_to_write | |
2413 | * @writepage: function called for each page | |
2414 | * @data: data passed to writepage function | |
2415 | * | |
2416 | * If a page is already under I/O, write_cache_pages() skips it, even | |
2417 | * if it's dirty. This is desirable behaviour for memory-cleaning writeback, | |
2418 | * but it is INCORRECT for data-integrity system calls such as fsync(). fsync() | |
2419 | * and msync() need to guarantee that all the data which was dirty at the time | |
2420 | * the call was made get new I/O started against them. If wbc->sync_mode is | |
2421 | * WB_SYNC_ALL then we were called for data integrity and we must wait for | |
2422 | * existing IO to complete. | |
2423 | */ | |
b2950863 | 2424 | static int extent_write_cache_pages(struct extent_io_tree *tree, |
4bef0848 CM |
2425 | struct address_space *mapping, |
2426 | struct writeback_control *wbc, | |
d2c3f4f6 CM |
2427 | writepage_t writepage, void *data, |
2428 | void (*flush_fn)(void *)) | |
d1310b2e | 2429 | { |
d1310b2e CM |
2430 | int ret = 0; |
2431 | int done = 0; | |
f85d7d6c | 2432 | int nr_to_write_done = 0; |
d1310b2e CM |
2433 | struct pagevec pvec; |
2434 | int nr_pages; | |
2435 | pgoff_t index; | |
2436 | pgoff_t end; /* Inclusive */ | |
2437 | int scanned = 0; | |
d1310b2e | 2438 | |
d1310b2e CM |
2439 | pagevec_init(&pvec, 0); |
2440 | if (wbc->range_cyclic) { | |
2441 | index = mapping->writeback_index; /* Start from prev offset */ | |
2442 | end = -1; | |
2443 | } else { | |
2444 | index = wbc->range_start >> PAGE_CACHE_SHIFT; | |
2445 | end = wbc->range_end >> PAGE_CACHE_SHIFT; | |
d1310b2e CM |
2446 | scanned = 1; |
2447 | } | |
2448 | retry: | |
f85d7d6c | 2449 | while (!done && !nr_to_write_done && (index <= end) && |
d1310b2e | 2450 | (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, |
d397712b CM |
2451 | PAGECACHE_TAG_DIRTY, min(end - index, |
2452 | (pgoff_t)PAGEVEC_SIZE-1) + 1))) { | |
d1310b2e CM |
2453 | unsigned i; |
2454 | ||
2455 | scanned = 1; | |
2456 | for (i = 0; i < nr_pages; i++) { | |
2457 | struct page *page = pvec.pages[i]; | |
2458 | ||
2459 | /* | |
2460 | * At this point we hold neither mapping->tree_lock nor | |
2461 | * lock on the page itself: the page may be truncated or | |
2462 | * invalidated (changing page->mapping to NULL), or even | |
2463 | * swizzled back from swapper_space to tmpfs file | |
2464 | * mapping | |
2465 | */ | |
4bef0848 CM |
2466 | if (tree->ops && tree->ops->write_cache_pages_lock_hook) |
2467 | tree->ops->write_cache_pages_lock_hook(page); | |
2468 | else | |
2469 | lock_page(page); | |
d1310b2e CM |
2470 | |
2471 | if (unlikely(page->mapping != mapping)) { | |
2472 | unlock_page(page); | |
2473 | continue; | |
2474 | } | |
2475 | ||
2476 | if (!wbc->range_cyclic && page->index > end) { | |
2477 | done = 1; | |
2478 | unlock_page(page); | |
2479 | continue; | |
2480 | } | |
2481 | ||
d2c3f4f6 | 2482 | if (wbc->sync_mode != WB_SYNC_NONE) { |
0e6bd956 CM |
2483 | if (PageWriteback(page)) |
2484 | flush_fn(data); | |
d1310b2e | 2485 | wait_on_page_writeback(page); |
d2c3f4f6 | 2486 | } |
d1310b2e CM |
2487 | |
2488 | if (PageWriteback(page) || | |
2489 | !clear_page_dirty_for_io(page)) { | |
2490 | unlock_page(page); | |
2491 | continue; | |
2492 | } | |
2493 | ||
2494 | ret = (*writepage)(page, wbc, data); | |
2495 | ||
2496 | if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) { | |
2497 | unlock_page(page); | |
2498 | ret = 0; | |
2499 | } | |
f85d7d6c | 2500 | if (ret) |
d1310b2e | 2501 | done = 1; |
f85d7d6c CM |
2502 | |
2503 | /* | |
2504 | * the filesystem may choose to bump up nr_to_write. | |
2505 | * We have to make sure to honor the new nr_to_write | |
2506 | * at any time | |
2507 | */ | |
2508 | nr_to_write_done = wbc->nr_to_write <= 0; | |
d1310b2e CM |
2509 | } |
2510 | pagevec_release(&pvec); | |
2511 | cond_resched(); | |
2512 | } | |
2513 | if (!scanned && !done) { | |
2514 | /* | |
2515 | * We hit the last page and there is more work to be done: wrap | |
2516 | * back to the start of the file | |
2517 | */ | |
2518 | scanned = 1; | |
2519 | index = 0; | |
2520 | goto retry; | |
2521 | } | |
d1310b2e CM |
2522 | return ret; |
2523 | } | |
d1310b2e | 2524 | |
ffbd517d | 2525 | static void flush_epd_write_bio(struct extent_page_data *epd) |
d2c3f4f6 | 2526 | { |
d2c3f4f6 | 2527 | if (epd->bio) { |
ffbd517d CM |
2528 | if (epd->sync_io) |
2529 | submit_one_bio(WRITE_SYNC, epd->bio, 0, 0); | |
2530 | else | |
2531 | submit_one_bio(WRITE, epd->bio, 0, 0); | |
d2c3f4f6 CM |
2532 | epd->bio = NULL; |
2533 | } | |
2534 | } | |
2535 | ||
ffbd517d CM |
2536 | static noinline void flush_write_bio(void *data) |
2537 | { | |
2538 | struct extent_page_data *epd = data; | |
2539 | flush_epd_write_bio(epd); | |
2540 | } | |
2541 | ||
d1310b2e CM |
2542 | int extent_write_full_page(struct extent_io_tree *tree, struct page *page, |
2543 | get_extent_t *get_extent, | |
2544 | struct writeback_control *wbc) | |
2545 | { | |
2546 | int ret; | |
2547 | struct address_space *mapping = page->mapping; | |
2548 | struct extent_page_data epd = { | |
2549 | .bio = NULL, | |
2550 | .tree = tree, | |
2551 | .get_extent = get_extent, | |
771ed689 | 2552 | .extent_locked = 0, |
ffbd517d | 2553 | .sync_io = wbc->sync_mode == WB_SYNC_ALL, |
d1310b2e CM |
2554 | }; |
2555 | struct writeback_control wbc_writepages = { | |
d313d7a3 | 2556 | .sync_mode = wbc->sync_mode, |
d1310b2e CM |
2557 | .older_than_this = NULL, |
2558 | .nr_to_write = 64, | |
2559 | .range_start = page_offset(page) + PAGE_CACHE_SIZE, | |
2560 | .range_end = (loff_t)-1, | |
2561 | }; | |
2562 | ||
d1310b2e CM |
2563 | ret = __extent_writepage(page, wbc, &epd); |
2564 | ||
4bef0848 | 2565 | extent_write_cache_pages(tree, mapping, &wbc_writepages, |
d2c3f4f6 | 2566 | __extent_writepage, &epd, flush_write_bio); |
ffbd517d | 2567 | flush_epd_write_bio(&epd); |
d1310b2e CM |
2568 | return ret; |
2569 | } | |
d1310b2e | 2570 | |
771ed689 CM |
2571 | int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode, |
2572 | u64 start, u64 end, get_extent_t *get_extent, | |
2573 | int mode) | |
2574 | { | |
2575 | int ret = 0; | |
2576 | struct address_space *mapping = inode->i_mapping; | |
2577 | struct page *page; | |
2578 | unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >> | |
2579 | PAGE_CACHE_SHIFT; | |
2580 | ||
2581 | struct extent_page_data epd = { | |
2582 | .bio = NULL, | |
2583 | .tree = tree, | |
2584 | .get_extent = get_extent, | |
2585 | .extent_locked = 1, | |
ffbd517d | 2586 | .sync_io = mode == WB_SYNC_ALL, |
771ed689 CM |
2587 | }; |
2588 | struct writeback_control wbc_writepages = { | |
771ed689 CM |
2589 | .sync_mode = mode, |
2590 | .older_than_this = NULL, | |
2591 | .nr_to_write = nr_pages * 2, | |
2592 | .range_start = start, | |
2593 | .range_end = end + 1, | |
2594 | }; | |
2595 | ||
d397712b | 2596 | while (start <= end) { |
771ed689 CM |
2597 | page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT); |
2598 | if (clear_page_dirty_for_io(page)) | |
2599 | ret = __extent_writepage(page, &wbc_writepages, &epd); | |
2600 | else { | |
2601 | if (tree->ops && tree->ops->writepage_end_io_hook) | |
2602 | tree->ops->writepage_end_io_hook(page, start, | |
2603 | start + PAGE_CACHE_SIZE - 1, | |
2604 | NULL, 1); | |
2605 | unlock_page(page); | |
2606 | } | |
2607 | page_cache_release(page); | |
2608 | start += PAGE_CACHE_SIZE; | |
2609 | } | |
2610 | ||
ffbd517d | 2611 | flush_epd_write_bio(&epd); |
771ed689 CM |
2612 | return ret; |
2613 | } | |
d1310b2e CM |
2614 | |
2615 | int extent_writepages(struct extent_io_tree *tree, | |
2616 | struct address_space *mapping, | |
2617 | get_extent_t *get_extent, | |
2618 | struct writeback_control *wbc) | |
2619 | { | |
2620 | int ret = 0; | |
2621 | struct extent_page_data epd = { | |
2622 | .bio = NULL, | |
2623 | .tree = tree, | |
2624 | .get_extent = get_extent, | |
771ed689 | 2625 | .extent_locked = 0, |
ffbd517d | 2626 | .sync_io = wbc->sync_mode == WB_SYNC_ALL, |
d1310b2e CM |
2627 | }; |
2628 | ||
4bef0848 | 2629 | ret = extent_write_cache_pages(tree, mapping, wbc, |
d2c3f4f6 CM |
2630 | __extent_writepage, &epd, |
2631 | flush_write_bio); | |
ffbd517d | 2632 | flush_epd_write_bio(&epd); |
d1310b2e CM |
2633 | return ret; |
2634 | } | |
d1310b2e CM |
2635 | |
2636 | int extent_readpages(struct extent_io_tree *tree, | |
2637 | struct address_space *mapping, | |
2638 | struct list_head *pages, unsigned nr_pages, | |
2639 | get_extent_t get_extent) | |
2640 | { | |
2641 | struct bio *bio = NULL; | |
2642 | unsigned page_idx; | |
c8b97818 | 2643 | unsigned long bio_flags = 0; |
d1310b2e | 2644 | |
d1310b2e CM |
2645 | for (page_idx = 0; page_idx < nr_pages; page_idx++) { |
2646 | struct page *page = list_entry(pages->prev, struct page, lru); | |
2647 | ||
2648 | prefetchw(&page->flags); | |
2649 | list_del(&page->lru); | |
28ecb609 | 2650 | if (!add_to_page_cache_lru(page, mapping, |
d1310b2e | 2651 | page->index, GFP_KERNEL)) { |
f188591e | 2652 | __extent_read_full_page(tree, page, get_extent, |
c8b97818 | 2653 | &bio, 0, &bio_flags); |
d1310b2e CM |
2654 | } |
2655 | page_cache_release(page); | |
2656 | } | |
d1310b2e CM |
2657 | BUG_ON(!list_empty(pages)); |
2658 | if (bio) | |
c8b97818 | 2659 | submit_one_bio(READ, bio, 0, bio_flags); |
d1310b2e CM |
2660 | return 0; |
2661 | } | |
d1310b2e CM |
2662 | |
2663 | /* | |
2664 | * basic invalidatepage code, this waits on any locked or writeback | |
2665 | * ranges corresponding to the page, and then deletes any extent state | |
2666 | * records from the tree | |
2667 | */ | |
2668 | int extent_invalidatepage(struct extent_io_tree *tree, | |
2669 | struct page *page, unsigned long offset) | |
2670 | { | |
2ac55d41 | 2671 | struct extent_state *cached_state = NULL; |
d1310b2e CM |
2672 | u64 start = ((u64)page->index << PAGE_CACHE_SHIFT); |
2673 | u64 end = start + PAGE_CACHE_SIZE - 1; | |
2674 | size_t blocksize = page->mapping->host->i_sb->s_blocksize; | |
2675 | ||
d397712b | 2676 | start += (offset + blocksize - 1) & ~(blocksize - 1); |
d1310b2e CM |
2677 | if (start > end) |
2678 | return 0; | |
2679 | ||
2ac55d41 | 2680 | lock_extent_bits(tree, start, end, 0, &cached_state, GFP_NOFS); |
1edbb734 | 2681 | wait_on_page_writeback(page); |
d1310b2e | 2682 | clear_extent_bit(tree, start, end, |
32c00aff JB |
2683 | EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC | |
2684 | EXTENT_DO_ACCOUNTING, | |
2ac55d41 | 2685 | 1, 1, &cached_state, GFP_NOFS); |
d1310b2e CM |
2686 | return 0; |
2687 | } | |
d1310b2e CM |
2688 | |
2689 | /* | |
2690 | * simple commit_write call, set_range_dirty is used to mark both | |
2691 | * the pages and the extent records as dirty | |
2692 | */ | |
2693 | int extent_commit_write(struct extent_io_tree *tree, | |
2694 | struct inode *inode, struct page *page, | |
2695 | unsigned from, unsigned to) | |
2696 | { | |
2697 | loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to; | |
2698 | ||
2699 | set_page_extent_mapped(page); | |
2700 | set_page_dirty(page); | |
2701 | ||
2702 | if (pos > inode->i_size) { | |
2703 | i_size_write(inode, pos); | |
2704 | mark_inode_dirty(inode); | |
2705 | } | |
2706 | return 0; | |
2707 | } | |
d1310b2e CM |
2708 | |
2709 | int extent_prepare_write(struct extent_io_tree *tree, | |
2710 | struct inode *inode, struct page *page, | |
2711 | unsigned from, unsigned to, get_extent_t *get_extent) | |
2712 | { | |
2713 | u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2714 | u64 page_end = page_start + PAGE_CACHE_SIZE - 1; | |
2715 | u64 block_start; | |
2716 | u64 orig_block_start; | |
2717 | u64 block_end; | |
2718 | u64 cur_end; | |
2719 | struct extent_map *em; | |
2720 | unsigned blocksize = 1 << inode->i_blkbits; | |
2721 | size_t page_offset = 0; | |
2722 | size_t block_off_start; | |
2723 | size_t block_off_end; | |
2724 | int err = 0; | |
2725 | int iocount = 0; | |
2726 | int ret = 0; | |
2727 | int isnew; | |
2728 | ||
2729 | set_page_extent_mapped(page); | |
2730 | ||
2731 | block_start = (page_start + from) & ~((u64)blocksize - 1); | |
2732 | block_end = (page_start + to - 1) | (blocksize - 1); | |
2733 | orig_block_start = block_start; | |
2734 | ||
2735 | lock_extent(tree, page_start, page_end, GFP_NOFS); | |
d397712b | 2736 | while (block_start <= block_end) { |
d1310b2e CM |
2737 | em = get_extent(inode, page, page_offset, block_start, |
2738 | block_end - block_start + 1, 1); | |
d397712b | 2739 | if (IS_ERR(em) || !em) |
d1310b2e | 2740 | goto err; |
d397712b | 2741 | |
d1310b2e CM |
2742 | cur_end = min(block_end, extent_map_end(em) - 1); |
2743 | block_off_start = block_start & (PAGE_CACHE_SIZE - 1); | |
2744 | block_off_end = block_off_start + blocksize; | |
2745 | isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS); | |
2746 | ||
2747 | if (!PageUptodate(page) && isnew && | |
2748 | (block_off_end > to || block_off_start < from)) { | |
2749 | void *kaddr; | |
2750 | ||
2751 | kaddr = kmap_atomic(page, KM_USER0); | |
2752 | if (block_off_end > to) | |
2753 | memset(kaddr + to, 0, block_off_end - to); | |
2754 | if (block_off_start < from) | |
2755 | memset(kaddr + block_off_start, 0, | |
2756 | from - block_off_start); | |
2757 | flush_dcache_page(page); | |
2758 | kunmap_atomic(kaddr, KM_USER0); | |
2759 | } | |
2760 | if ((em->block_start != EXTENT_MAP_HOLE && | |
2761 | em->block_start != EXTENT_MAP_INLINE) && | |
2762 | !isnew && !PageUptodate(page) && | |
2763 | (block_off_end > to || block_off_start < from) && | |
2764 | !test_range_bit(tree, block_start, cur_end, | |
9655d298 | 2765 | EXTENT_UPTODATE, 1, NULL)) { |
d1310b2e CM |
2766 | u64 sector; |
2767 | u64 extent_offset = block_start - em->start; | |
2768 | size_t iosize; | |
2769 | sector = (em->block_start + extent_offset) >> 9; | |
2770 | iosize = (cur_end - block_start + blocksize) & | |
2771 | ~((u64)blocksize - 1); | |
2772 | /* | |
2773 | * we've already got the extent locked, but we | |
2774 | * need to split the state such that our end_bio | |
2775 | * handler can clear the lock. | |
2776 | */ | |
2777 | set_extent_bit(tree, block_start, | |
2778 | block_start + iosize - 1, | |
2c64c53d | 2779 | EXTENT_LOCKED, 0, NULL, NULL, GFP_NOFS); |
d1310b2e CM |
2780 | ret = submit_extent_page(READ, tree, page, |
2781 | sector, iosize, page_offset, em->bdev, | |
2782 | NULL, 1, | |
c8b97818 CM |
2783 | end_bio_extent_preparewrite, 0, |
2784 | 0, 0); | |
411fc6bc AK |
2785 | if (ret && !err) |
2786 | err = ret; | |
d1310b2e CM |
2787 | iocount++; |
2788 | block_start = block_start + iosize; | |
2789 | } else { | |
2790 | set_extent_uptodate(tree, block_start, cur_end, | |
2791 | GFP_NOFS); | |
2792 | unlock_extent(tree, block_start, cur_end, GFP_NOFS); | |
2793 | block_start = cur_end + 1; | |
2794 | } | |
2795 | page_offset = block_start & (PAGE_CACHE_SIZE - 1); | |
2796 | free_extent_map(em); | |
2797 | } | |
2798 | if (iocount) { | |
2799 | wait_extent_bit(tree, orig_block_start, | |
2800 | block_end, EXTENT_LOCKED); | |
2801 | } | |
2802 | check_page_uptodate(tree, page); | |
2803 | err: | |
2804 | /* FIXME, zero out newly allocated blocks on error */ | |
2805 | return err; | |
2806 | } | |
d1310b2e | 2807 | |
7b13b7b1 CM |
2808 | /* |
2809 | * a helper for releasepage, this tests for areas of the page that | |
2810 | * are locked or under IO and drops the related state bits if it is safe | |
2811 | * to drop the page. | |
2812 | */ | |
2813 | int try_release_extent_state(struct extent_map_tree *map, | |
2814 | struct extent_io_tree *tree, struct page *page, | |
2815 | gfp_t mask) | |
2816 | { | |
2817 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2818 | u64 end = start + PAGE_CACHE_SIZE - 1; | |
2819 | int ret = 1; | |
2820 | ||
211f90e6 | 2821 | if (test_range_bit(tree, start, end, |
8b62b72b | 2822 | EXTENT_IOBITS, 0, NULL)) |
7b13b7b1 CM |
2823 | ret = 0; |
2824 | else { | |
2825 | if ((mask & GFP_NOFS) == GFP_NOFS) | |
2826 | mask = GFP_NOFS; | |
11ef160f CM |
2827 | /* |
2828 | * at this point we can safely clear everything except the | |
2829 | * locked bit and the nodatasum bit | |
2830 | */ | |
e3f24cc5 | 2831 | ret = clear_extent_bit(tree, start, end, |
11ef160f CM |
2832 | ~(EXTENT_LOCKED | EXTENT_NODATASUM), |
2833 | 0, 0, NULL, mask); | |
e3f24cc5 CM |
2834 | |
2835 | /* if clear_extent_bit failed for enomem reasons, | |
2836 | * we can't allow the release to continue. | |
2837 | */ | |
2838 | if (ret < 0) | |
2839 | ret = 0; | |
2840 | else | |
2841 | ret = 1; | |
7b13b7b1 CM |
2842 | } |
2843 | return ret; | |
2844 | } | |
7b13b7b1 | 2845 | |
d1310b2e CM |
2846 | /* |
2847 | * a helper for releasepage. As long as there are no locked extents | |
2848 | * in the range corresponding to the page, both state records and extent | |
2849 | * map records are removed | |
2850 | */ | |
2851 | int try_release_extent_mapping(struct extent_map_tree *map, | |
70dec807 CM |
2852 | struct extent_io_tree *tree, struct page *page, |
2853 | gfp_t mask) | |
d1310b2e CM |
2854 | { |
2855 | struct extent_map *em; | |
2856 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2857 | u64 end = start + PAGE_CACHE_SIZE - 1; | |
7b13b7b1 | 2858 | |
70dec807 CM |
2859 | if ((mask & __GFP_WAIT) && |
2860 | page->mapping->host->i_size > 16 * 1024 * 1024) { | |
39b5637f | 2861 | u64 len; |
70dec807 | 2862 | while (start <= end) { |
39b5637f | 2863 | len = end - start + 1; |
890871be | 2864 | write_lock(&map->lock); |
39b5637f | 2865 | em = lookup_extent_mapping(map, start, len); |
70dec807 | 2866 | if (!em || IS_ERR(em)) { |
890871be | 2867 | write_unlock(&map->lock); |
70dec807 CM |
2868 | break; |
2869 | } | |
7f3c74fb CM |
2870 | if (test_bit(EXTENT_FLAG_PINNED, &em->flags) || |
2871 | em->start != start) { | |
890871be | 2872 | write_unlock(&map->lock); |
70dec807 CM |
2873 | free_extent_map(em); |
2874 | break; | |
2875 | } | |
2876 | if (!test_range_bit(tree, em->start, | |
2877 | extent_map_end(em) - 1, | |
8b62b72b | 2878 | EXTENT_LOCKED | EXTENT_WRITEBACK, |
9655d298 | 2879 | 0, NULL)) { |
70dec807 CM |
2880 | remove_extent_mapping(map, em); |
2881 | /* once for the rb tree */ | |
2882 | free_extent_map(em); | |
2883 | } | |
2884 | start = extent_map_end(em); | |
890871be | 2885 | write_unlock(&map->lock); |
70dec807 CM |
2886 | |
2887 | /* once for us */ | |
d1310b2e CM |
2888 | free_extent_map(em); |
2889 | } | |
d1310b2e | 2890 | } |
7b13b7b1 | 2891 | return try_release_extent_state(map, tree, page, mask); |
d1310b2e | 2892 | } |
d1310b2e CM |
2893 | |
2894 | sector_t extent_bmap(struct address_space *mapping, sector_t iblock, | |
2895 | get_extent_t *get_extent) | |
2896 | { | |
2897 | struct inode *inode = mapping->host; | |
2ac55d41 | 2898 | struct extent_state *cached_state = NULL; |
d1310b2e CM |
2899 | u64 start = iblock << inode->i_blkbits; |
2900 | sector_t sector = 0; | |
d899e052 | 2901 | size_t blksize = (1 << inode->i_blkbits); |
d1310b2e CM |
2902 | struct extent_map *em; |
2903 | ||
2ac55d41 JB |
2904 | lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + blksize - 1, |
2905 | 0, &cached_state, GFP_NOFS); | |
d899e052 | 2906 | em = get_extent(inode, NULL, 0, start, blksize, 0); |
2ac55d41 JB |
2907 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, |
2908 | start + blksize - 1, &cached_state, GFP_NOFS); | |
d1310b2e CM |
2909 | if (!em || IS_ERR(em)) |
2910 | return 0; | |
2911 | ||
d899e052 | 2912 | if (em->block_start > EXTENT_MAP_LAST_BYTE) |
d1310b2e CM |
2913 | goto out; |
2914 | ||
2915 | sector = (em->block_start + start - em->start) >> inode->i_blkbits; | |
d1310b2e CM |
2916 | out: |
2917 | free_extent_map(em); | |
2918 | return sector; | |
2919 | } | |
2920 | ||
ec29ed5b CM |
2921 | /* |
2922 | * helper function for fiemap, which doesn't want to see any holes. | |
2923 | * This maps until we find something past 'last' | |
2924 | */ | |
2925 | static struct extent_map *get_extent_skip_holes(struct inode *inode, | |
2926 | u64 offset, | |
2927 | u64 last, | |
2928 | get_extent_t *get_extent) | |
2929 | { | |
2930 | u64 sectorsize = BTRFS_I(inode)->root->sectorsize; | |
2931 | struct extent_map *em; | |
2932 | u64 len; | |
2933 | ||
2934 | if (offset >= last) | |
2935 | return NULL; | |
2936 | ||
2937 | while(1) { | |
2938 | len = last - offset; | |
2939 | if (len == 0) | |
2940 | break; | |
2941 | len = (len + sectorsize - 1) & ~(sectorsize - 1); | |
2942 | em = get_extent(inode, NULL, 0, offset, len, 0); | |
2943 | if (!em || IS_ERR(em)) | |
2944 | return em; | |
2945 | ||
2946 | /* if this isn't a hole return it */ | |
2947 | if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags) && | |
2948 | em->block_start != EXTENT_MAP_HOLE) { | |
2949 | return em; | |
2950 | } | |
2951 | ||
2952 | /* this is a hole, advance to the next extent */ | |
2953 | offset = extent_map_end(em); | |
2954 | free_extent_map(em); | |
2955 | if (offset >= last) | |
2956 | break; | |
2957 | } | |
2958 | return NULL; | |
2959 | } | |
2960 | ||
1506fcc8 YS |
2961 | int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, |
2962 | __u64 start, __u64 len, get_extent_t *get_extent) | |
2963 | { | |
975f84fe | 2964 | int ret = 0; |
1506fcc8 YS |
2965 | u64 off = start; |
2966 | u64 max = start + len; | |
2967 | u32 flags = 0; | |
975f84fe JB |
2968 | u32 found_type; |
2969 | u64 last; | |
ec29ed5b | 2970 | u64 last_for_get_extent = 0; |
1506fcc8 | 2971 | u64 disko = 0; |
ec29ed5b | 2972 | u64 isize = i_size_read(inode); |
975f84fe | 2973 | struct btrfs_key found_key; |
1506fcc8 | 2974 | struct extent_map *em = NULL; |
2ac55d41 | 2975 | struct extent_state *cached_state = NULL; |
975f84fe JB |
2976 | struct btrfs_path *path; |
2977 | struct btrfs_file_extent_item *item; | |
1506fcc8 | 2978 | int end = 0; |
ec29ed5b CM |
2979 | u64 em_start = 0; |
2980 | u64 em_len = 0; | |
2981 | u64 em_end = 0; | |
1506fcc8 | 2982 | unsigned long emflags; |
1506fcc8 YS |
2983 | |
2984 | if (len == 0) | |
2985 | return -EINVAL; | |
2986 | ||
975f84fe JB |
2987 | path = btrfs_alloc_path(); |
2988 | if (!path) | |
2989 | return -ENOMEM; | |
2990 | path->leave_spinning = 1; | |
2991 | ||
ec29ed5b CM |
2992 | /* |
2993 | * lookup the last file extent. We're not using i_size here | |
2994 | * because there might be preallocation past i_size | |
2995 | */ | |
975f84fe JB |
2996 | ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root, |
2997 | path, inode->i_ino, -1, 0); | |
2998 | if (ret < 0) { | |
2999 | btrfs_free_path(path); | |
3000 | return ret; | |
3001 | } | |
3002 | WARN_ON(!ret); | |
3003 | path->slots[0]--; | |
3004 | item = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
3005 | struct btrfs_file_extent_item); | |
3006 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]); | |
3007 | found_type = btrfs_key_type(&found_key); | |
3008 | ||
ec29ed5b | 3009 | /* No extents, but there might be delalloc bits */ |
975f84fe JB |
3010 | if (found_key.objectid != inode->i_ino || |
3011 | found_type != BTRFS_EXTENT_DATA_KEY) { | |
ec29ed5b CM |
3012 | /* have to trust i_size as the end */ |
3013 | last = (u64)-1; | |
3014 | last_for_get_extent = isize; | |
3015 | } else { | |
3016 | /* | |
3017 | * remember the start of the last extent. There are a | |
3018 | * bunch of different factors that go into the length of the | |
3019 | * extent, so its much less complex to remember where it started | |
3020 | */ | |
3021 | last = found_key.offset; | |
3022 | last_for_get_extent = last + 1; | |
975f84fe | 3023 | } |
975f84fe JB |
3024 | btrfs_free_path(path); |
3025 | ||
ec29ed5b CM |
3026 | /* |
3027 | * we might have some extents allocated but more delalloc past those | |
3028 | * extents. so, we trust isize unless the start of the last extent is | |
3029 | * beyond isize | |
3030 | */ | |
3031 | if (last < isize) { | |
3032 | last = (u64)-1; | |
3033 | last_for_get_extent = isize; | |
3034 | } | |
3035 | ||
2ac55d41 JB |
3036 | lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0, |
3037 | &cached_state, GFP_NOFS); | |
ec29ed5b CM |
3038 | |
3039 | em = get_extent_skip_holes(inode, off, last_for_get_extent, | |
3040 | get_extent); | |
1506fcc8 YS |
3041 | if (!em) |
3042 | goto out; | |
3043 | if (IS_ERR(em)) { | |
3044 | ret = PTR_ERR(em); | |
3045 | goto out; | |
3046 | } | |
975f84fe | 3047 | |
1506fcc8 | 3048 | while (!end) { |
ea8efc74 CM |
3049 | u64 offset_in_extent; |
3050 | ||
3051 | /* break if the extent we found is outside the range */ | |
3052 | if (em->start >= max || extent_map_end(em) < off) | |
3053 | break; | |
3054 | ||
3055 | /* | |
3056 | * get_extent may return an extent that starts before our | |
3057 | * requested range. We have to make sure the ranges | |
3058 | * we return to fiemap always move forward and don't | |
3059 | * overlap, so adjust the offsets here | |
3060 | */ | |
3061 | em_start = max(em->start, off); | |
1506fcc8 | 3062 | |
ea8efc74 CM |
3063 | /* |
3064 | * record the offset from the start of the extent | |
3065 | * for adjusting the disk offset below | |
3066 | */ | |
3067 | offset_in_extent = em_start - em->start; | |
ec29ed5b | 3068 | em_end = extent_map_end(em); |
ea8efc74 | 3069 | em_len = em_end - em_start; |
ec29ed5b | 3070 | emflags = em->flags; |
1506fcc8 YS |
3071 | disko = 0; |
3072 | flags = 0; | |
3073 | ||
ea8efc74 CM |
3074 | /* |
3075 | * bump off for our next call to get_extent | |
3076 | */ | |
3077 | off = extent_map_end(em); | |
3078 | if (off >= max) | |
3079 | end = 1; | |
3080 | ||
93dbfad7 | 3081 | if (em->block_start == EXTENT_MAP_LAST_BYTE) { |
1506fcc8 YS |
3082 | end = 1; |
3083 | flags |= FIEMAP_EXTENT_LAST; | |
93dbfad7 | 3084 | } else if (em->block_start == EXTENT_MAP_INLINE) { |
1506fcc8 YS |
3085 | flags |= (FIEMAP_EXTENT_DATA_INLINE | |
3086 | FIEMAP_EXTENT_NOT_ALIGNED); | |
93dbfad7 | 3087 | } else if (em->block_start == EXTENT_MAP_DELALLOC) { |
1506fcc8 YS |
3088 | flags |= (FIEMAP_EXTENT_DELALLOC | |
3089 | FIEMAP_EXTENT_UNKNOWN); | |
93dbfad7 | 3090 | } else { |
ea8efc74 | 3091 | disko = em->block_start + offset_in_extent; |
1506fcc8 YS |
3092 | } |
3093 | if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) | |
3094 | flags |= FIEMAP_EXTENT_ENCODED; | |
3095 | ||
1506fcc8 YS |
3096 | free_extent_map(em); |
3097 | em = NULL; | |
ec29ed5b CM |
3098 | if ((em_start >= last) || em_len == (u64)-1 || |
3099 | (last == (u64)-1 && isize <= em_end)) { | |
1506fcc8 YS |
3100 | flags |= FIEMAP_EXTENT_LAST; |
3101 | end = 1; | |
3102 | } | |
3103 | ||
ec29ed5b CM |
3104 | /* now scan forward to see if this is really the last extent. */ |
3105 | em = get_extent_skip_holes(inode, off, last_for_get_extent, | |
3106 | get_extent); | |
3107 | if (IS_ERR(em)) { | |
3108 | ret = PTR_ERR(em); | |
3109 | goto out; | |
3110 | } | |
3111 | if (!em) { | |
975f84fe JB |
3112 | flags |= FIEMAP_EXTENT_LAST; |
3113 | end = 1; | |
3114 | } | |
ec29ed5b CM |
3115 | ret = fiemap_fill_next_extent(fieinfo, em_start, disko, |
3116 | em_len, flags); | |
3117 | if (ret) | |
3118 | goto out_free; | |
1506fcc8 YS |
3119 | } |
3120 | out_free: | |
3121 | free_extent_map(em); | |
3122 | out: | |
2ac55d41 JB |
3123 | unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len, |
3124 | &cached_state, GFP_NOFS); | |
1506fcc8 YS |
3125 | return ret; |
3126 | } | |
3127 | ||
d1310b2e CM |
3128 | static inline struct page *extent_buffer_page(struct extent_buffer *eb, |
3129 | unsigned long i) | |
3130 | { | |
3131 | struct page *p; | |
3132 | struct address_space *mapping; | |
3133 | ||
3134 | if (i == 0) | |
3135 | return eb->first_page; | |
3136 | i += eb->start >> PAGE_CACHE_SHIFT; | |
3137 | mapping = eb->first_page->mapping; | |
33958dc6 CM |
3138 | if (!mapping) |
3139 | return NULL; | |
0ee0fda0 SW |
3140 | |
3141 | /* | |
3142 | * extent_buffer_page is only called after pinning the page | |
3143 | * by increasing the reference count. So we know the page must | |
3144 | * be in the radix tree. | |
3145 | */ | |
0ee0fda0 | 3146 | rcu_read_lock(); |
d1310b2e | 3147 | p = radix_tree_lookup(&mapping->page_tree, i); |
0ee0fda0 | 3148 | rcu_read_unlock(); |
2b1f55b0 | 3149 | |
d1310b2e CM |
3150 | return p; |
3151 | } | |
3152 | ||
6af118ce | 3153 | static inline unsigned long num_extent_pages(u64 start, u64 len) |
728131d8 | 3154 | { |
6af118ce CM |
3155 | return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) - |
3156 | (start >> PAGE_CACHE_SHIFT); | |
728131d8 CM |
3157 | } |
3158 | ||
d1310b2e CM |
3159 | static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree, |
3160 | u64 start, | |
3161 | unsigned long len, | |
3162 | gfp_t mask) | |
3163 | { | |
3164 | struct extent_buffer *eb = NULL; | |
3935127c | 3165 | #if LEAK_DEBUG |
2d2ae547 | 3166 | unsigned long flags; |
4bef0848 | 3167 | #endif |
d1310b2e | 3168 | |
d1310b2e | 3169 | eb = kmem_cache_zalloc(extent_buffer_cache, mask); |
91ca338d TI |
3170 | if (eb == NULL) |
3171 | return NULL; | |
d1310b2e CM |
3172 | eb->start = start; |
3173 | eb->len = len; | |
b4ce94de CM |
3174 | spin_lock_init(&eb->lock); |
3175 | init_waitqueue_head(&eb->lock_wq); | |
3176 | ||
3935127c | 3177 | #if LEAK_DEBUG |
2d2ae547 CM |
3178 | spin_lock_irqsave(&leak_lock, flags); |
3179 | list_add(&eb->leak_list, &buffers); | |
3180 | spin_unlock_irqrestore(&leak_lock, flags); | |
4bef0848 | 3181 | #endif |
d1310b2e CM |
3182 | atomic_set(&eb->refs, 1); |
3183 | ||
3184 | return eb; | |
3185 | } | |
3186 | ||
3187 | static void __free_extent_buffer(struct extent_buffer *eb) | |
3188 | { | |
3935127c | 3189 | #if LEAK_DEBUG |
2d2ae547 CM |
3190 | unsigned long flags; |
3191 | spin_lock_irqsave(&leak_lock, flags); | |
3192 | list_del(&eb->leak_list); | |
3193 | spin_unlock_irqrestore(&leak_lock, flags); | |
4bef0848 | 3194 | #endif |
d1310b2e CM |
3195 | kmem_cache_free(extent_buffer_cache, eb); |
3196 | } | |
3197 | ||
897ca6e9 MX |
3198 | /* |
3199 | * Helper for releasing extent buffer page. | |
3200 | */ | |
3201 | static void btrfs_release_extent_buffer_page(struct extent_buffer *eb, | |
3202 | unsigned long start_idx) | |
3203 | { | |
3204 | unsigned long index; | |
3205 | struct page *page; | |
3206 | ||
3207 | if (!eb->first_page) | |
3208 | return; | |
3209 | ||
3210 | index = num_extent_pages(eb->start, eb->len); | |
3211 | if (start_idx >= index) | |
3212 | return; | |
3213 | ||
3214 | do { | |
3215 | index--; | |
3216 | page = extent_buffer_page(eb, index); | |
3217 | if (page) | |
3218 | page_cache_release(page); | |
3219 | } while (index != start_idx); | |
3220 | } | |
3221 | ||
3222 | /* | |
3223 | * Helper for releasing the extent buffer. | |
3224 | */ | |
3225 | static inline void btrfs_release_extent_buffer(struct extent_buffer *eb) | |
3226 | { | |
3227 | btrfs_release_extent_buffer_page(eb, 0); | |
3228 | __free_extent_buffer(eb); | |
3229 | } | |
3230 | ||
d1310b2e CM |
3231 | struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree, |
3232 | u64 start, unsigned long len, | |
3233 | struct page *page0, | |
3234 | gfp_t mask) | |
3235 | { | |
3236 | unsigned long num_pages = num_extent_pages(start, len); | |
3237 | unsigned long i; | |
3238 | unsigned long index = start >> PAGE_CACHE_SHIFT; | |
3239 | struct extent_buffer *eb; | |
6af118ce | 3240 | struct extent_buffer *exists = NULL; |
d1310b2e CM |
3241 | struct page *p; |
3242 | struct address_space *mapping = tree->mapping; | |
3243 | int uptodate = 1; | |
19fe0a8b | 3244 | int ret; |
d1310b2e | 3245 | |
19fe0a8b MX |
3246 | rcu_read_lock(); |
3247 | eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT); | |
3248 | if (eb && atomic_inc_not_zero(&eb->refs)) { | |
3249 | rcu_read_unlock(); | |
0f9dd46c | 3250 | mark_page_accessed(eb->first_page); |
6af118ce CM |
3251 | return eb; |
3252 | } | |
19fe0a8b | 3253 | rcu_read_unlock(); |
6af118ce | 3254 | |
d1310b2e | 3255 | eb = __alloc_extent_buffer(tree, start, len, mask); |
2b114d1d | 3256 | if (!eb) |
d1310b2e CM |
3257 | return NULL; |
3258 | ||
d1310b2e CM |
3259 | if (page0) { |
3260 | eb->first_page = page0; | |
3261 | i = 1; | |
3262 | index++; | |
3263 | page_cache_get(page0); | |
3264 | mark_page_accessed(page0); | |
3265 | set_page_extent_mapped(page0); | |
d1310b2e | 3266 | set_page_extent_head(page0, len); |
f188591e | 3267 | uptodate = PageUptodate(page0); |
d1310b2e CM |
3268 | } else { |
3269 | i = 0; | |
3270 | } | |
3271 | for (; i < num_pages; i++, index++) { | |
3272 | p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM); | |
3273 | if (!p) { | |
3274 | WARN_ON(1); | |
6af118ce | 3275 | goto free_eb; |
d1310b2e CM |
3276 | } |
3277 | set_page_extent_mapped(p); | |
3278 | mark_page_accessed(p); | |
3279 | if (i == 0) { | |
3280 | eb->first_page = p; | |
3281 | set_page_extent_head(p, len); | |
3282 | } else { | |
3283 | set_page_private(p, EXTENT_PAGE_PRIVATE); | |
3284 | } | |
3285 | if (!PageUptodate(p)) | |
3286 | uptodate = 0; | |
eb14ab8e CM |
3287 | |
3288 | /* | |
3289 | * see below about how we avoid a nasty race with release page | |
3290 | * and why we unlock later | |
3291 | */ | |
3292 | if (i != 0) | |
3293 | unlock_page(p); | |
d1310b2e CM |
3294 | } |
3295 | if (uptodate) | |
b4ce94de | 3296 | set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); |
d1310b2e | 3297 | |
19fe0a8b MX |
3298 | ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); |
3299 | if (ret) | |
3300 | goto free_eb; | |
3301 | ||
6af118ce | 3302 | spin_lock(&tree->buffer_lock); |
19fe0a8b MX |
3303 | ret = radix_tree_insert(&tree->buffer, start >> PAGE_CACHE_SHIFT, eb); |
3304 | if (ret == -EEXIST) { | |
3305 | exists = radix_tree_lookup(&tree->buffer, | |
3306 | start >> PAGE_CACHE_SHIFT); | |
6af118ce CM |
3307 | /* add one reference for the caller */ |
3308 | atomic_inc(&exists->refs); | |
3309 | spin_unlock(&tree->buffer_lock); | |
19fe0a8b | 3310 | radix_tree_preload_end(); |
6af118ce CM |
3311 | goto free_eb; |
3312 | } | |
6af118ce CM |
3313 | /* add one reference for the tree */ |
3314 | atomic_inc(&eb->refs); | |
f044ba78 | 3315 | spin_unlock(&tree->buffer_lock); |
19fe0a8b | 3316 | radix_tree_preload_end(); |
eb14ab8e CM |
3317 | |
3318 | /* | |
3319 | * there is a race where release page may have | |
3320 | * tried to find this extent buffer in the radix | |
3321 | * but failed. It will tell the VM it is safe to | |
3322 | * reclaim the, and it will clear the page private bit. | |
3323 | * We must make sure to set the page private bit properly | |
3324 | * after the extent buffer is in the radix tree so | |
3325 | * it doesn't get lost | |
3326 | */ | |
3327 | set_page_extent_mapped(eb->first_page); | |
3328 | set_page_extent_head(eb->first_page, eb->len); | |
3329 | if (!page0) | |
3330 | unlock_page(eb->first_page); | |
d1310b2e CM |
3331 | return eb; |
3332 | ||
6af118ce | 3333 | free_eb: |
eb14ab8e CM |
3334 | if (eb->first_page && !page0) |
3335 | unlock_page(eb->first_page); | |
3336 | ||
d1310b2e | 3337 | if (!atomic_dec_and_test(&eb->refs)) |
6af118ce | 3338 | return exists; |
897ca6e9 | 3339 | btrfs_release_extent_buffer(eb); |
6af118ce | 3340 | return exists; |
d1310b2e | 3341 | } |
d1310b2e CM |
3342 | |
3343 | struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree, | |
3344 | u64 start, unsigned long len, | |
3345 | gfp_t mask) | |
3346 | { | |
d1310b2e | 3347 | struct extent_buffer *eb; |
d1310b2e | 3348 | |
19fe0a8b MX |
3349 | rcu_read_lock(); |
3350 | eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT); | |
3351 | if (eb && atomic_inc_not_zero(&eb->refs)) { | |
3352 | rcu_read_unlock(); | |
0f9dd46c | 3353 | mark_page_accessed(eb->first_page); |
19fe0a8b MX |
3354 | return eb; |
3355 | } | |
3356 | rcu_read_unlock(); | |
0f9dd46c | 3357 | |
19fe0a8b | 3358 | return NULL; |
d1310b2e | 3359 | } |
d1310b2e CM |
3360 | |
3361 | void free_extent_buffer(struct extent_buffer *eb) | |
3362 | { | |
d1310b2e CM |
3363 | if (!eb) |
3364 | return; | |
3365 | ||
3366 | if (!atomic_dec_and_test(&eb->refs)) | |
3367 | return; | |
3368 | ||
6af118ce | 3369 | WARN_ON(1); |
d1310b2e | 3370 | } |
d1310b2e CM |
3371 | |
3372 | int clear_extent_buffer_dirty(struct extent_io_tree *tree, | |
3373 | struct extent_buffer *eb) | |
3374 | { | |
d1310b2e CM |
3375 | unsigned long i; |
3376 | unsigned long num_pages; | |
3377 | struct page *page; | |
3378 | ||
d1310b2e CM |
3379 | num_pages = num_extent_pages(eb->start, eb->len); |
3380 | ||
3381 | for (i = 0; i < num_pages; i++) { | |
3382 | page = extent_buffer_page(eb, i); | |
b9473439 | 3383 | if (!PageDirty(page)) |
d2c3f4f6 CM |
3384 | continue; |
3385 | ||
a61e6f29 | 3386 | lock_page(page); |
eb14ab8e CM |
3387 | WARN_ON(!PagePrivate(page)); |
3388 | ||
3389 | set_page_extent_mapped(page); | |
d1310b2e CM |
3390 | if (i == 0) |
3391 | set_page_extent_head(page, eb->len); | |
d1310b2e | 3392 | |
d1310b2e | 3393 | clear_page_dirty_for_io(page); |
0ee0fda0 | 3394 | spin_lock_irq(&page->mapping->tree_lock); |
d1310b2e CM |
3395 | if (!PageDirty(page)) { |
3396 | radix_tree_tag_clear(&page->mapping->page_tree, | |
3397 | page_index(page), | |
3398 | PAGECACHE_TAG_DIRTY); | |
3399 | } | |
0ee0fda0 | 3400 | spin_unlock_irq(&page->mapping->tree_lock); |
a61e6f29 | 3401 | unlock_page(page); |
d1310b2e CM |
3402 | } |
3403 | return 0; | |
3404 | } | |
d1310b2e CM |
3405 | |
3406 | int wait_on_extent_buffer_writeback(struct extent_io_tree *tree, | |
3407 | struct extent_buffer *eb) | |
3408 | { | |
3409 | return wait_on_extent_writeback(tree, eb->start, | |
3410 | eb->start + eb->len - 1); | |
3411 | } | |
d1310b2e CM |
3412 | |
3413 | int set_extent_buffer_dirty(struct extent_io_tree *tree, | |
3414 | struct extent_buffer *eb) | |
3415 | { | |
3416 | unsigned long i; | |
3417 | unsigned long num_pages; | |
b9473439 | 3418 | int was_dirty = 0; |
d1310b2e | 3419 | |
b9473439 | 3420 | was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); |
d1310b2e | 3421 | num_pages = num_extent_pages(eb->start, eb->len); |
b9473439 | 3422 | for (i = 0; i < num_pages; i++) |
d1310b2e | 3423 | __set_page_dirty_nobuffers(extent_buffer_page(eb, i)); |
b9473439 | 3424 | return was_dirty; |
d1310b2e | 3425 | } |
d1310b2e | 3426 | |
1259ab75 | 3427 | int clear_extent_buffer_uptodate(struct extent_io_tree *tree, |
2ac55d41 JB |
3428 | struct extent_buffer *eb, |
3429 | struct extent_state **cached_state) | |
1259ab75 CM |
3430 | { |
3431 | unsigned long i; | |
3432 | struct page *page; | |
3433 | unsigned long num_pages; | |
3434 | ||
3435 | num_pages = num_extent_pages(eb->start, eb->len); | |
b4ce94de | 3436 | clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); |
1259ab75 CM |
3437 | |
3438 | clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, | |
2ac55d41 | 3439 | cached_state, GFP_NOFS); |
1259ab75 CM |
3440 | for (i = 0; i < num_pages; i++) { |
3441 | page = extent_buffer_page(eb, i); | |
33958dc6 CM |
3442 | if (page) |
3443 | ClearPageUptodate(page); | |
1259ab75 CM |
3444 | } |
3445 | return 0; | |
3446 | } | |
3447 | ||
d1310b2e CM |
3448 | int set_extent_buffer_uptodate(struct extent_io_tree *tree, |
3449 | struct extent_buffer *eb) | |
3450 | { | |
3451 | unsigned long i; | |
3452 | struct page *page; | |
3453 | unsigned long num_pages; | |
3454 | ||
3455 | num_pages = num_extent_pages(eb->start, eb->len); | |
3456 | ||
3457 | set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1, | |
3458 | GFP_NOFS); | |
3459 | for (i = 0; i < num_pages; i++) { | |
3460 | page = extent_buffer_page(eb, i); | |
3461 | if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) || | |
3462 | ((i == num_pages - 1) && | |
3463 | ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) { | |
3464 | check_page_uptodate(tree, page); | |
3465 | continue; | |
3466 | } | |
3467 | SetPageUptodate(page); | |
3468 | } | |
3469 | return 0; | |
3470 | } | |
d1310b2e | 3471 | |
ce9adaa5 CM |
3472 | int extent_range_uptodate(struct extent_io_tree *tree, |
3473 | u64 start, u64 end) | |
3474 | { | |
3475 | struct page *page; | |
3476 | int ret; | |
3477 | int pg_uptodate = 1; | |
3478 | int uptodate; | |
3479 | unsigned long index; | |
3480 | ||
9655d298 | 3481 | ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL); |
ce9adaa5 CM |
3482 | if (ret) |
3483 | return 1; | |
d397712b | 3484 | while (start <= end) { |
ce9adaa5 CM |
3485 | index = start >> PAGE_CACHE_SHIFT; |
3486 | page = find_get_page(tree->mapping, index); | |
3487 | uptodate = PageUptodate(page); | |
3488 | page_cache_release(page); | |
3489 | if (!uptodate) { | |
3490 | pg_uptodate = 0; | |
3491 | break; | |
3492 | } | |
3493 | start += PAGE_CACHE_SIZE; | |
3494 | } | |
3495 | return pg_uptodate; | |
3496 | } | |
3497 | ||
d1310b2e | 3498 | int extent_buffer_uptodate(struct extent_io_tree *tree, |
2ac55d41 JB |
3499 | struct extent_buffer *eb, |
3500 | struct extent_state *cached_state) | |
d1310b2e | 3501 | { |
728131d8 | 3502 | int ret = 0; |
ce9adaa5 CM |
3503 | unsigned long num_pages; |
3504 | unsigned long i; | |
728131d8 CM |
3505 | struct page *page; |
3506 | int pg_uptodate = 1; | |
3507 | ||
b4ce94de | 3508 | if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags)) |
4235298e | 3509 | return 1; |
728131d8 | 3510 | |
4235298e | 3511 | ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1, |
2ac55d41 | 3512 | EXTENT_UPTODATE, 1, cached_state); |
4235298e CM |
3513 | if (ret) |
3514 | return ret; | |
728131d8 CM |
3515 | |
3516 | num_pages = num_extent_pages(eb->start, eb->len); | |
3517 | for (i = 0; i < num_pages; i++) { | |
3518 | page = extent_buffer_page(eb, i); | |
3519 | if (!PageUptodate(page)) { | |
3520 | pg_uptodate = 0; | |
3521 | break; | |
3522 | } | |
3523 | } | |
4235298e | 3524 | return pg_uptodate; |
d1310b2e | 3525 | } |
d1310b2e CM |
3526 | |
3527 | int read_extent_buffer_pages(struct extent_io_tree *tree, | |
3528 | struct extent_buffer *eb, | |
a86c12c7 | 3529 | u64 start, int wait, |
f188591e | 3530 | get_extent_t *get_extent, int mirror_num) |
d1310b2e CM |
3531 | { |
3532 | unsigned long i; | |
3533 | unsigned long start_i; | |
3534 | struct page *page; | |
3535 | int err; | |
3536 | int ret = 0; | |
ce9adaa5 CM |
3537 | int locked_pages = 0; |
3538 | int all_uptodate = 1; | |
3539 | int inc_all_pages = 0; | |
d1310b2e | 3540 | unsigned long num_pages; |
a86c12c7 | 3541 | struct bio *bio = NULL; |
c8b97818 | 3542 | unsigned long bio_flags = 0; |
a86c12c7 | 3543 | |
b4ce94de | 3544 | if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags)) |
d1310b2e CM |
3545 | return 0; |
3546 | ||
ce9adaa5 | 3547 | if (test_range_bit(tree, eb->start, eb->start + eb->len - 1, |
9655d298 | 3548 | EXTENT_UPTODATE, 1, NULL)) { |
d1310b2e CM |
3549 | return 0; |
3550 | } | |
3551 | ||
3552 | if (start) { | |
3553 | WARN_ON(start < eb->start); | |
3554 | start_i = (start >> PAGE_CACHE_SHIFT) - | |
3555 | (eb->start >> PAGE_CACHE_SHIFT); | |
3556 | } else { | |
3557 | start_i = 0; | |
3558 | } | |
3559 | ||
3560 | num_pages = num_extent_pages(eb->start, eb->len); | |
3561 | for (i = start_i; i < num_pages; i++) { | |
3562 | page = extent_buffer_page(eb, i); | |
d1310b2e | 3563 | if (!wait) { |
2db04966 | 3564 | if (!trylock_page(page)) |
ce9adaa5 | 3565 | goto unlock_exit; |
d1310b2e CM |
3566 | } else { |
3567 | lock_page(page); | |
3568 | } | |
ce9adaa5 | 3569 | locked_pages++; |
d397712b | 3570 | if (!PageUptodate(page)) |
ce9adaa5 | 3571 | all_uptodate = 0; |
ce9adaa5 CM |
3572 | } |
3573 | if (all_uptodate) { | |
3574 | if (start_i == 0) | |
b4ce94de | 3575 | set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); |
ce9adaa5 CM |
3576 | goto unlock_exit; |
3577 | } | |
3578 | ||
3579 | for (i = start_i; i < num_pages; i++) { | |
3580 | page = extent_buffer_page(eb, i); | |
eb14ab8e CM |
3581 | |
3582 | WARN_ON(!PagePrivate(page)); | |
3583 | ||
3584 | set_page_extent_mapped(page); | |
3585 | if (i == 0) | |
3586 | set_page_extent_head(page, eb->len); | |
3587 | ||
ce9adaa5 CM |
3588 | if (inc_all_pages) |
3589 | page_cache_get(page); | |
3590 | if (!PageUptodate(page)) { | |
3591 | if (start_i == 0) | |
3592 | inc_all_pages = 1; | |
f188591e | 3593 | ClearPageError(page); |
a86c12c7 | 3594 | err = __extent_read_full_page(tree, page, |
f188591e | 3595 | get_extent, &bio, |
c8b97818 | 3596 | mirror_num, &bio_flags); |
d397712b | 3597 | if (err) |
d1310b2e | 3598 | ret = err; |
d1310b2e CM |
3599 | } else { |
3600 | unlock_page(page); | |
3601 | } | |
3602 | } | |
3603 | ||
a86c12c7 | 3604 | if (bio) |
c8b97818 | 3605 | submit_one_bio(READ, bio, mirror_num, bio_flags); |
a86c12c7 | 3606 | |
d397712b | 3607 | if (ret || !wait) |
d1310b2e | 3608 | return ret; |
d397712b | 3609 | |
d1310b2e CM |
3610 | for (i = start_i; i < num_pages; i++) { |
3611 | page = extent_buffer_page(eb, i); | |
3612 | wait_on_page_locked(page); | |
d397712b | 3613 | if (!PageUptodate(page)) |
d1310b2e | 3614 | ret = -EIO; |
d1310b2e | 3615 | } |
d397712b | 3616 | |
d1310b2e | 3617 | if (!ret) |
b4ce94de | 3618 | set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); |
d1310b2e | 3619 | return ret; |
ce9adaa5 CM |
3620 | |
3621 | unlock_exit: | |
3622 | i = start_i; | |
d397712b | 3623 | while (locked_pages > 0) { |
ce9adaa5 CM |
3624 | page = extent_buffer_page(eb, i); |
3625 | i++; | |
3626 | unlock_page(page); | |
3627 | locked_pages--; | |
3628 | } | |
3629 | return ret; | |
d1310b2e | 3630 | } |
d1310b2e CM |
3631 | |
3632 | void read_extent_buffer(struct extent_buffer *eb, void *dstv, | |
3633 | unsigned long start, | |
3634 | unsigned long len) | |
3635 | { | |
3636 | size_t cur; | |
3637 | size_t offset; | |
3638 | struct page *page; | |
3639 | char *kaddr; | |
3640 | char *dst = (char *)dstv; | |
3641 | size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3642 | unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; | |
d1310b2e CM |
3643 | |
3644 | WARN_ON(start > eb->len); | |
3645 | WARN_ON(start + len > eb->start + eb->len); | |
3646 | ||
3647 | offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3648 | ||
d397712b | 3649 | while (len > 0) { |
d1310b2e | 3650 | page = extent_buffer_page(eb, i); |
d1310b2e CM |
3651 | |
3652 | cur = min(len, (PAGE_CACHE_SIZE - offset)); | |
3653 | kaddr = kmap_atomic(page, KM_USER1); | |
3654 | memcpy(dst, kaddr + offset, cur); | |
3655 | kunmap_atomic(kaddr, KM_USER1); | |
3656 | ||
3657 | dst += cur; | |
3658 | len -= cur; | |
3659 | offset = 0; | |
3660 | i++; | |
3661 | } | |
3662 | } | |
d1310b2e CM |
3663 | |
3664 | int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start, | |
3665 | unsigned long min_len, char **token, char **map, | |
3666 | unsigned long *map_start, | |
3667 | unsigned long *map_len, int km) | |
3668 | { | |
3669 | size_t offset = start & (PAGE_CACHE_SIZE - 1); | |
3670 | char *kaddr; | |
3671 | struct page *p; | |
3672 | size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3673 | unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; | |
3674 | unsigned long end_i = (start_offset + start + min_len - 1) >> | |
3675 | PAGE_CACHE_SHIFT; | |
3676 | ||
3677 | if (i != end_i) | |
3678 | return -EINVAL; | |
3679 | ||
3680 | if (i == 0) { | |
3681 | offset = start_offset; | |
3682 | *map_start = 0; | |
3683 | } else { | |
3684 | offset = 0; | |
3685 | *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset; | |
3686 | } | |
d397712b | 3687 | |
d1310b2e | 3688 | if (start + min_len > eb->len) { |
d397712b CM |
3689 | printk(KERN_ERR "btrfs bad mapping eb start %llu len %lu, " |
3690 | "wanted %lu %lu\n", (unsigned long long)eb->start, | |
3691 | eb->len, start, min_len); | |
d1310b2e | 3692 | WARN_ON(1); |
85026533 | 3693 | return -EINVAL; |
d1310b2e CM |
3694 | } |
3695 | ||
3696 | p = extent_buffer_page(eb, i); | |
d1310b2e CM |
3697 | kaddr = kmap_atomic(p, km); |
3698 | *token = kaddr; | |
3699 | *map = kaddr + offset; | |
3700 | *map_len = PAGE_CACHE_SIZE - offset; | |
3701 | return 0; | |
3702 | } | |
d1310b2e CM |
3703 | |
3704 | int map_extent_buffer(struct extent_buffer *eb, unsigned long start, | |
3705 | unsigned long min_len, | |
3706 | char **token, char **map, | |
3707 | unsigned long *map_start, | |
3708 | unsigned long *map_len, int km) | |
3709 | { | |
3710 | int err; | |
3711 | int save = 0; | |
3712 | if (eb->map_token) { | |
3713 | unmap_extent_buffer(eb, eb->map_token, km); | |
3714 | eb->map_token = NULL; | |
3715 | save = 1; | |
3716 | } | |
3717 | err = map_private_extent_buffer(eb, start, min_len, token, map, | |
3718 | map_start, map_len, km); | |
3719 | if (!err && save) { | |
3720 | eb->map_token = *token; | |
3721 | eb->kaddr = *map; | |
3722 | eb->map_start = *map_start; | |
3723 | eb->map_len = *map_len; | |
3724 | } | |
3725 | return err; | |
3726 | } | |
d1310b2e CM |
3727 | |
3728 | void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km) | |
3729 | { | |
3730 | kunmap_atomic(token, km); | |
3731 | } | |
d1310b2e CM |
3732 | |
3733 | int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv, | |
3734 | unsigned long start, | |
3735 | unsigned long len) | |
3736 | { | |
3737 | size_t cur; | |
3738 | size_t offset; | |
3739 | struct page *page; | |
3740 | char *kaddr; | |
3741 | char *ptr = (char *)ptrv; | |
3742 | size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3743 | unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; | |
3744 | int ret = 0; | |
3745 | ||
3746 | WARN_ON(start > eb->len); | |
3747 | WARN_ON(start + len > eb->start + eb->len); | |
3748 | ||
3749 | offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3750 | ||
d397712b | 3751 | while (len > 0) { |
d1310b2e | 3752 | page = extent_buffer_page(eb, i); |
d1310b2e CM |
3753 | |
3754 | cur = min(len, (PAGE_CACHE_SIZE - offset)); | |
3755 | ||
3756 | kaddr = kmap_atomic(page, KM_USER0); | |
3757 | ret = memcmp(ptr, kaddr + offset, cur); | |
3758 | kunmap_atomic(kaddr, KM_USER0); | |
3759 | if (ret) | |
3760 | break; | |
3761 | ||
3762 | ptr += cur; | |
3763 | len -= cur; | |
3764 | offset = 0; | |
3765 | i++; | |
3766 | } | |
3767 | return ret; | |
3768 | } | |
d1310b2e CM |
3769 | |
3770 | void write_extent_buffer(struct extent_buffer *eb, const void *srcv, | |
3771 | unsigned long start, unsigned long len) | |
3772 | { | |
3773 | size_t cur; | |
3774 | size_t offset; | |
3775 | struct page *page; | |
3776 | char *kaddr; | |
3777 | char *src = (char *)srcv; | |
3778 | size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3779 | unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; | |
3780 | ||
3781 | WARN_ON(start > eb->len); | |
3782 | WARN_ON(start + len > eb->start + eb->len); | |
3783 | ||
3784 | offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3785 | ||
d397712b | 3786 | while (len > 0) { |
d1310b2e CM |
3787 | page = extent_buffer_page(eb, i); |
3788 | WARN_ON(!PageUptodate(page)); | |
3789 | ||
3790 | cur = min(len, PAGE_CACHE_SIZE - offset); | |
3791 | kaddr = kmap_atomic(page, KM_USER1); | |
3792 | memcpy(kaddr + offset, src, cur); | |
3793 | kunmap_atomic(kaddr, KM_USER1); | |
3794 | ||
3795 | src += cur; | |
3796 | len -= cur; | |
3797 | offset = 0; | |
3798 | i++; | |
3799 | } | |
3800 | } | |
d1310b2e CM |
3801 | |
3802 | void memset_extent_buffer(struct extent_buffer *eb, char c, | |
3803 | unsigned long start, unsigned long len) | |
3804 | { | |
3805 | size_t cur; | |
3806 | size_t offset; | |
3807 | struct page *page; | |
3808 | char *kaddr; | |
3809 | size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3810 | unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT; | |
3811 | ||
3812 | WARN_ON(start > eb->len); | |
3813 | WARN_ON(start + len > eb->start + eb->len); | |
3814 | ||
3815 | offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3816 | ||
d397712b | 3817 | while (len > 0) { |
d1310b2e CM |
3818 | page = extent_buffer_page(eb, i); |
3819 | WARN_ON(!PageUptodate(page)); | |
3820 | ||
3821 | cur = min(len, PAGE_CACHE_SIZE - offset); | |
3822 | kaddr = kmap_atomic(page, KM_USER0); | |
3823 | memset(kaddr + offset, c, cur); | |
3824 | kunmap_atomic(kaddr, KM_USER0); | |
3825 | ||
3826 | len -= cur; | |
3827 | offset = 0; | |
3828 | i++; | |
3829 | } | |
3830 | } | |
d1310b2e CM |
3831 | |
3832 | void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src, | |
3833 | unsigned long dst_offset, unsigned long src_offset, | |
3834 | unsigned long len) | |
3835 | { | |
3836 | u64 dst_len = dst->len; | |
3837 | size_t cur; | |
3838 | size_t offset; | |
3839 | struct page *page; | |
3840 | char *kaddr; | |
3841 | size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3842 | unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; | |
3843 | ||
3844 | WARN_ON(src->len != dst_len); | |
3845 | ||
3846 | offset = (start_offset + dst_offset) & | |
3847 | ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3848 | ||
d397712b | 3849 | while (len > 0) { |
d1310b2e CM |
3850 | page = extent_buffer_page(dst, i); |
3851 | WARN_ON(!PageUptodate(page)); | |
3852 | ||
3853 | cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset)); | |
3854 | ||
3855 | kaddr = kmap_atomic(page, KM_USER0); | |
3856 | read_extent_buffer(src, kaddr + offset, src_offset, cur); | |
3857 | kunmap_atomic(kaddr, KM_USER0); | |
3858 | ||
3859 | src_offset += cur; | |
3860 | len -= cur; | |
3861 | offset = 0; | |
3862 | i++; | |
3863 | } | |
3864 | } | |
d1310b2e CM |
3865 | |
3866 | static void move_pages(struct page *dst_page, struct page *src_page, | |
3867 | unsigned long dst_off, unsigned long src_off, | |
3868 | unsigned long len) | |
3869 | { | |
3870 | char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); | |
3871 | if (dst_page == src_page) { | |
3872 | memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len); | |
3873 | } else { | |
3874 | char *src_kaddr = kmap_atomic(src_page, KM_USER1); | |
3875 | char *p = dst_kaddr + dst_off + len; | |
3876 | char *s = src_kaddr + src_off + len; | |
3877 | ||
3878 | while (len--) | |
3879 | *--p = *--s; | |
3880 | ||
3881 | kunmap_atomic(src_kaddr, KM_USER1); | |
3882 | } | |
3883 | kunmap_atomic(dst_kaddr, KM_USER0); | |
3884 | } | |
3885 | ||
3886 | static void copy_pages(struct page *dst_page, struct page *src_page, | |
3887 | unsigned long dst_off, unsigned long src_off, | |
3888 | unsigned long len) | |
3889 | { | |
3890 | char *dst_kaddr = kmap_atomic(dst_page, KM_USER0); | |
3891 | char *src_kaddr; | |
3892 | ||
3893 | if (dst_page != src_page) | |
3894 | src_kaddr = kmap_atomic(src_page, KM_USER1); | |
3895 | else | |
3896 | src_kaddr = dst_kaddr; | |
3897 | ||
3898 | memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len); | |
3899 | kunmap_atomic(dst_kaddr, KM_USER0); | |
3900 | if (dst_page != src_page) | |
3901 | kunmap_atomic(src_kaddr, KM_USER1); | |
3902 | } | |
3903 | ||
3904 | void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, | |
3905 | unsigned long src_offset, unsigned long len) | |
3906 | { | |
3907 | size_t cur; | |
3908 | size_t dst_off_in_page; | |
3909 | size_t src_off_in_page; | |
3910 | size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3911 | unsigned long dst_i; | |
3912 | unsigned long src_i; | |
3913 | ||
3914 | if (src_offset + len > dst->len) { | |
d397712b CM |
3915 | printk(KERN_ERR "btrfs memmove bogus src_offset %lu move " |
3916 | "len %lu dst len %lu\n", src_offset, len, dst->len); | |
d1310b2e CM |
3917 | BUG_ON(1); |
3918 | } | |
3919 | if (dst_offset + len > dst->len) { | |
d397712b CM |
3920 | printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move " |
3921 | "len %lu dst len %lu\n", dst_offset, len, dst->len); | |
d1310b2e CM |
3922 | BUG_ON(1); |
3923 | } | |
3924 | ||
d397712b | 3925 | while (len > 0) { |
d1310b2e CM |
3926 | dst_off_in_page = (start_offset + dst_offset) & |
3927 | ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3928 | src_off_in_page = (start_offset + src_offset) & | |
3929 | ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3930 | ||
3931 | dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT; | |
3932 | src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT; | |
3933 | ||
3934 | cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - | |
3935 | src_off_in_page)); | |
3936 | cur = min_t(unsigned long, cur, | |
3937 | (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page)); | |
3938 | ||
3939 | copy_pages(extent_buffer_page(dst, dst_i), | |
3940 | extent_buffer_page(dst, src_i), | |
3941 | dst_off_in_page, src_off_in_page, cur); | |
3942 | ||
3943 | src_offset += cur; | |
3944 | dst_offset += cur; | |
3945 | len -= cur; | |
3946 | } | |
3947 | } | |
d1310b2e CM |
3948 | |
3949 | void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset, | |
3950 | unsigned long src_offset, unsigned long len) | |
3951 | { | |
3952 | size_t cur; | |
3953 | size_t dst_off_in_page; | |
3954 | size_t src_off_in_page; | |
3955 | unsigned long dst_end = dst_offset + len - 1; | |
3956 | unsigned long src_end = src_offset + len - 1; | |
3957 | size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1); | |
3958 | unsigned long dst_i; | |
3959 | unsigned long src_i; | |
3960 | ||
3961 | if (src_offset + len > dst->len) { | |
d397712b CM |
3962 | printk(KERN_ERR "btrfs memmove bogus src_offset %lu move " |
3963 | "len %lu len %lu\n", src_offset, len, dst->len); | |
d1310b2e CM |
3964 | BUG_ON(1); |
3965 | } | |
3966 | if (dst_offset + len > dst->len) { | |
d397712b CM |
3967 | printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move " |
3968 | "len %lu len %lu\n", dst_offset, len, dst->len); | |
d1310b2e CM |
3969 | BUG_ON(1); |
3970 | } | |
3971 | if (dst_offset < src_offset) { | |
3972 | memcpy_extent_buffer(dst, dst_offset, src_offset, len); | |
3973 | return; | |
3974 | } | |
d397712b | 3975 | while (len > 0) { |
d1310b2e CM |
3976 | dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT; |
3977 | src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT; | |
3978 | ||
3979 | dst_off_in_page = (start_offset + dst_end) & | |
3980 | ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3981 | src_off_in_page = (start_offset + src_end) & | |
3982 | ((unsigned long)PAGE_CACHE_SIZE - 1); | |
3983 | ||
3984 | cur = min_t(unsigned long, len, src_off_in_page + 1); | |
3985 | cur = min(cur, dst_off_in_page + 1); | |
3986 | move_pages(extent_buffer_page(dst, dst_i), | |
3987 | extent_buffer_page(dst, src_i), | |
3988 | dst_off_in_page - cur + 1, | |
3989 | src_off_in_page - cur + 1, cur); | |
3990 | ||
3991 | dst_end -= cur; | |
3992 | src_end -= cur; | |
3993 | len -= cur; | |
3994 | } | |
3995 | } | |
6af118ce | 3996 | |
19fe0a8b MX |
3997 | static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head) |
3998 | { | |
3999 | struct extent_buffer *eb = | |
4000 | container_of(head, struct extent_buffer, rcu_head); | |
4001 | ||
4002 | btrfs_release_extent_buffer(eb); | |
4003 | } | |
4004 | ||
6af118ce CM |
4005 | int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page) |
4006 | { | |
4007 | u64 start = page_offset(page); | |
4008 | struct extent_buffer *eb; | |
4009 | int ret = 1; | |
6af118ce CM |
4010 | |
4011 | spin_lock(&tree->buffer_lock); | |
19fe0a8b | 4012 | eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT); |
45f49bce CM |
4013 | if (!eb) { |
4014 | spin_unlock(&tree->buffer_lock); | |
4015 | return ret; | |
4016 | } | |
6af118ce | 4017 | |
19fe0a8b | 4018 | if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { |
6af118ce CM |
4019 | ret = 0; |
4020 | goto out; | |
4021 | } | |
19fe0a8b MX |
4022 | |
4023 | /* | |
4024 | * set @eb->refs to 0 if it is already 1, and then release the @eb. | |
4025 | * Or go back. | |
4026 | */ | |
4027 | if (atomic_cmpxchg(&eb->refs, 1, 0) != 1) { | |
b9473439 CM |
4028 | ret = 0; |
4029 | goto out; | |
4030 | } | |
897ca6e9 | 4031 | |
19fe0a8b | 4032 | radix_tree_delete(&tree->buffer, start >> PAGE_CACHE_SHIFT); |
6af118ce CM |
4033 | out: |
4034 | spin_unlock(&tree->buffer_lock); | |
19fe0a8b MX |
4035 | |
4036 | /* at this point we can safely release the extent buffer */ | |
4037 | if (atomic_read(&eb->refs) == 0) | |
4038 | call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu); | |
6af118ce CM |
4039 | return ret; |
4040 | } |