Blame view
fs/mbcache.c
16.4 KB
1da177e4c
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
/* * linux/fs/mbcache.c * (C) 2001-2002 Andreas Gruenbacher, <a.gruenbacher@computer.org> */ /* * Filesystem Meta Information Block Cache (mbcache) * * The mbcache caches blocks of block devices that need to be located * by their device/block number, as well as by other criteria (such * as the block's contents). * * There can only be one cache entry in a cache per device and block number. * Additional indexes need not be unique in this sense. The number of * additional indexes (=other criteria) can be hardwired at compile time * or specified at cache create time. * * Each cache entry is of fixed size. An entry may be `valid' or `invalid' * in the cache. A valid entry is in the main hash tables of the cache, * and may also be in the lru list. An invalid entry is not in any hashes * or lists. * * A valid cache entry is only in the lru list if no handles refer to it. * Invalid cache entries will be freed when the last handle to the cache * entry is released. Entries that cannot be freed immediately are put * back on the lru list. */ #include <linux/kernel.h> #include <linux/module.h> #include <linux/hash.h> #include <linux/fs.h> #include <linux/mm.h> #include <linux/slab.h> #include <linux/sched.h> #include <linux/init.h> #include <linux/mbcache.h> #ifdef MB_CACHE_DEBUG # define mb_debug(f...) do { \ printk(KERN_DEBUG f); \ printk(" "); \ } while (0) #define mb_assert(c) do { if (!(c)) \ printk(KERN_ERR "assertion " #c " failed "); \ } while(0) #else # define mb_debug(f...) do { } while(0) # define mb_assert(c) do { } while(0) #endif #define mb_error(f...) do { \ printk(KERN_ERR f); \ printk(" "); \ } while(0) #define MB_CACHE_WRITER ((unsigned short)~0U >> 1) |
75c96f858
|
62 |
static DECLARE_WAIT_QUEUE_HEAD(mb_cache_queue); |
1da177e4c
|
63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 |
MODULE_AUTHOR("Andreas Gruenbacher <a.gruenbacher@computer.org>"); MODULE_DESCRIPTION("Meta block cache (for extended attributes)"); MODULE_LICENSE("GPL"); EXPORT_SYMBOL(mb_cache_create); EXPORT_SYMBOL(mb_cache_shrink); EXPORT_SYMBOL(mb_cache_destroy); EXPORT_SYMBOL(mb_cache_entry_alloc); EXPORT_SYMBOL(mb_cache_entry_insert); EXPORT_SYMBOL(mb_cache_entry_release); EXPORT_SYMBOL(mb_cache_entry_free); EXPORT_SYMBOL(mb_cache_entry_get); #if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) EXPORT_SYMBOL(mb_cache_entry_find_first); EXPORT_SYMBOL(mb_cache_entry_find_next); #endif |
1da177e4c
|
80 81 82 83 84 85 86 87 88 |
/* * Global data: list of all mbcache's, lru list, and a spinlock for * accessing cache data structures on SMP machines. The lru list is * global across all mbcaches. */ static LIST_HEAD(mb_cache_list); static LIST_HEAD(mb_cache_lru_list); static DEFINE_SPINLOCK(mb_cache_spinlock); |
1da177e4c
|
89 |
|
1da177e4c
|
90 91 92 |
/* * What the mbcache registers as to get shrunk dynamically. */ |
1495f230f
|
93 94 |
static int mb_cache_shrink_fn(struct shrinker *shrink, struct shrink_control *sc); |
1da177e4c
|
95 |
|
8e1f936b7
|
96 97 98 99 |
static struct shrinker mb_cache_shrinker = { .shrink = mb_cache_shrink_fn, .seeks = DEFAULT_SEEKS, }; |
1da177e4c
|
100 101 102 103 104 105 |
static inline int __mb_cache_entry_is_hashed(struct mb_cache_entry *ce) { return !list_empty(&ce->e_block_list); } |
858119e15
|
106 |
static void |
1da177e4c
|
107 108 |
__mb_cache_entry_unhash(struct mb_cache_entry *ce) { |
1da177e4c
|
109 110 |
if (__mb_cache_entry_is_hashed(ce)) { list_del_init(&ce->e_block_list); |
2aec7c523
|
111 |
list_del(&ce->e_index.o_list); |
1da177e4c
|
112 113 |
} } |
858119e15
|
114 |
static void |
27496a8c6
|
115 |
__mb_cache_entry_forget(struct mb_cache_entry *ce, gfp_t gfp_mask) |
1da177e4c
|
116 117 118 119 |
{ struct mb_cache *cache = ce->e_cache; mb_assert(!(ce->e_used || ce->e_queued)); |
2aec7c523
|
120 121 |
kmem_cache_free(cache->c_entry_cache, ce); atomic_dec(&cache->c_entry_count); |
1da177e4c
|
122 |
} |
858119e15
|
123 |
static void |
1da177e4c
|
124 |
__mb_cache_entry_release_unlock(struct mb_cache_entry *ce) |
58f555e5f
|
125 |
__releases(mb_cache_spinlock) |
1da177e4c
|
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 |
{ /* Wake up all processes queuing for this cache entry. */ if (ce->e_queued) wake_up_all(&mb_cache_queue); if (ce->e_used >= MB_CACHE_WRITER) ce->e_used -= MB_CACHE_WRITER; ce->e_used--; if (!(ce->e_used || ce->e_queued)) { if (!__mb_cache_entry_is_hashed(ce)) goto forget; mb_assert(list_empty(&ce->e_lru_list)); list_add_tail(&ce->e_lru_list, &mb_cache_lru_list); } spin_unlock(&mb_cache_spinlock); return; forget: spin_unlock(&mb_cache_spinlock); __mb_cache_entry_forget(ce, GFP_KERNEL); } /* * mb_cache_shrink_fn() memory pressure callback * * This function is called by the kernel memory management when memory * gets low. * |
7f8275d0d
|
153 |
* @shrink: (ignored) |
1495f230f
|
154 |
* @sc: shrink_control passed from reclaim |
1da177e4c
|
155 156 157 158 |
* * Returns the number of objects which are present in the cache. */ static int |
1495f230f
|
159 |
mb_cache_shrink_fn(struct shrinker *shrink, struct shrink_control *sc) |
1da177e4c
|
160 161 |
{ LIST_HEAD(free_list); |
e566d48c9
|
162 163 |
struct mb_cache *cache; struct mb_cache_entry *entry, *tmp; |
1da177e4c
|
164 |
int count = 0; |
1495f230f
|
165 166 |
int nr_to_scan = sc->nr_to_scan; gfp_t gfp_mask = sc->gfp_mask; |
1da177e4c
|
167 |
|
1da177e4c
|
168 |
mb_debug("trying to free %d entries", nr_to_scan); |
e566d48c9
|
169 |
spin_lock(&mb_cache_spinlock); |
1da177e4c
|
170 171 172 173 174 175 176 |
while (nr_to_scan-- && !list_empty(&mb_cache_lru_list)) { struct mb_cache_entry *ce = list_entry(mb_cache_lru_list.next, struct mb_cache_entry, e_lru_list); list_move_tail(&ce->e_lru_list, &free_list); __mb_cache_entry_unhash(ce); } |
e566d48c9
|
177 178 179 180 181 |
list_for_each_entry(cache, &mb_cache_list, c_cache_list) { mb_debug("cache %s (%d)", cache->c_name, atomic_read(&cache->c_entry_count)); count += atomic_read(&cache->c_entry_count); } |
1da177e4c
|
182 |
spin_unlock(&mb_cache_spinlock); |
e566d48c9
|
183 184 |
list_for_each_entry_safe(entry, tmp, &free_list, e_lru_list) { __mb_cache_entry_forget(entry, gfp_mask); |
1da177e4c
|
185 |
} |
1da177e4c
|
186 187 188 189 190 191 192 193 194 195 196 197 198 |
return (count / 100) * sysctl_vfs_cache_pressure; } /* * mb_cache_create() create a new cache * * All entries in one cache are equal size. Cache entries may be from * multiple devices. If this is the first mbcache created, registers * the cache with kernel memory management. Returns NULL if no more * memory was available. * * @name: name of the cache (informal) |
1da177e4c
|
199 200 201 |
* @bucket_bits: log2(number of hash buckets) */ struct mb_cache * |
2aec7c523
|
202 |
mb_cache_create(const char *name, int bucket_bits) |
1da177e4c
|
203 |
{ |
2aec7c523
|
204 |
int n, bucket_count = 1 << bucket_bits; |
1da177e4c
|
205 |
struct mb_cache *cache = NULL; |
2aec7c523
|
206 |
cache = kmalloc(sizeof(struct mb_cache), GFP_KERNEL); |
1da177e4c
|
207 |
if (!cache) |
2aec7c523
|
208 |
return NULL; |
1da177e4c
|
209 |
cache->c_name = name; |
1da177e4c
|
210 211 |
atomic_set(&cache->c_entry_count, 0); cache->c_bucket_bits = bucket_bits; |
1da177e4c
|
212 213 214 215 216 217 |
cache->c_block_hash = kmalloc(bucket_count * sizeof(struct list_head), GFP_KERNEL); if (!cache->c_block_hash) goto fail; for (n=0; n<bucket_count; n++) INIT_LIST_HEAD(&cache->c_block_hash[n]); |
2aec7c523
|
218 219 220 221 222 223 224 225 |
cache->c_index_hash = kmalloc(bucket_count * sizeof(struct list_head), GFP_KERNEL); if (!cache->c_index_hash) goto fail; for (n=0; n<bucket_count; n++) INIT_LIST_HEAD(&cache->c_index_hash[n]); cache->c_entry_cache = kmem_cache_create(name, sizeof(struct mb_cache_entry), 0, |
20c2df83d
|
226 |
SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD, NULL); |
1da177e4c
|
227 |
if (!cache->c_entry_cache) |
2aec7c523
|
228 |
goto fail2; |
1da177e4c
|
229 |
|
3a48ee8a4
|
230 231 232 233 234 |
/* * Set an upper limit on the number of cache entries so that the hash * chains won't grow too long. */ cache->c_max_entries = bucket_count << 4; |
1da177e4c
|
235 236 237 238 |
spin_lock(&mb_cache_spinlock); list_add(&cache->c_cache_list, &mb_cache_list); spin_unlock(&mb_cache_spinlock); return cache; |
2aec7c523
|
239 240 |
fail2: kfree(cache->c_index_hash); |
1da177e4c
|
241 |
fail: |
2aec7c523
|
242 243 |
kfree(cache->c_block_hash); kfree(cache); |
1da177e4c
|
244 245 246 247 248 249 250 |
return NULL; } /* * mb_cache_shrink() * |
7f927fcc2
|
251 |
* Removes all cache entries of a device from the cache. All cache entries |
1da177e4c
|
252 253 254 |
* currently in use cannot be freed, and thus remain in the cache. All others * are freed. * |
1da177e4c
|
255 256 257 |
* @bdev: which device's cache entries to shrink */ void |
8c52ab42c
|
258 |
mb_cache_shrink(struct block_device *bdev) |
1da177e4c
|
259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 |
{ LIST_HEAD(free_list); struct list_head *l, *ltmp; spin_lock(&mb_cache_spinlock); list_for_each_safe(l, ltmp, &mb_cache_lru_list) { struct mb_cache_entry *ce = list_entry(l, struct mb_cache_entry, e_lru_list); if (ce->e_bdev == bdev) { list_move_tail(&ce->e_lru_list, &free_list); __mb_cache_entry_unhash(ce); } } spin_unlock(&mb_cache_spinlock); list_for_each_safe(l, ltmp, &free_list) { __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, e_lru_list), GFP_KERNEL); } } /* * mb_cache_destroy() * * Shrinks the cache to its minimum possible size (hopefully 0 entries), * and then destroys it. If this was the last mbcache, un-registers the * mbcache from kernel memory management. */ void mb_cache_destroy(struct mb_cache *cache) { LIST_HEAD(free_list); struct list_head *l, *ltmp; |
1da177e4c
|
292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 |
spin_lock(&mb_cache_spinlock); list_for_each_safe(l, ltmp, &mb_cache_lru_list) { struct mb_cache_entry *ce = list_entry(l, struct mb_cache_entry, e_lru_list); if (ce->e_cache == cache) { list_move_tail(&ce->e_lru_list, &free_list); __mb_cache_entry_unhash(ce); } } list_del(&cache->c_cache_list); spin_unlock(&mb_cache_spinlock); list_for_each_safe(l, ltmp, &free_list) { __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, e_lru_list), GFP_KERNEL); } if (atomic_read(&cache->c_entry_count) > 0) { mb_error("cache %s: %d orphaned entries", cache->c_name, atomic_read(&cache->c_entry_count)); } kmem_cache_destroy(cache->c_entry_cache); |
2aec7c523
|
317 |
kfree(cache->c_index_hash); |
1da177e4c
|
318 319 320 |
kfree(cache->c_block_hash); kfree(cache); } |
1da177e4c
|
321 322 323 324 325 326 327 328 329 |
/* * mb_cache_entry_alloc() * * Allocates a new cache entry. The new entry will not be valid initially, * and thus cannot be looked up yet. It should be filled with data, and * then inserted into the cache using mb_cache_entry_insert(). Returns NULL * if no more memory was available. */ struct mb_cache_entry * |
335e92e8a
|
330 |
mb_cache_entry_alloc(struct mb_cache *cache, gfp_t gfp_flags) |
1da177e4c
|
331 |
{ |
3a48ee8a4
|
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 |
struct mb_cache_entry *ce = NULL; if (atomic_read(&cache->c_entry_count) >= cache->c_max_entries) { spin_lock(&mb_cache_spinlock); if (!list_empty(&mb_cache_lru_list)) { ce = list_entry(mb_cache_lru_list.next, struct mb_cache_entry, e_lru_list); list_del_init(&ce->e_lru_list); __mb_cache_entry_unhash(ce); } spin_unlock(&mb_cache_spinlock); } if (!ce) { ce = kmem_cache_alloc(cache->c_entry_cache, gfp_flags); if (!ce) return NULL; |
f9e83489c
|
348 |
atomic_inc(&cache->c_entry_count); |
1da177e4c
|
349 350 351 |
INIT_LIST_HEAD(&ce->e_lru_list); INIT_LIST_HEAD(&ce->e_block_list); ce->e_cache = cache; |
1da177e4c
|
352 353 |
ce->e_queued = 0; } |
3a48ee8a4
|
354 |
ce->e_used = 1 + MB_CACHE_WRITER; |
1da177e4c
|
355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 |
return ce; } /* * mb_cache_entry_insert() * * Inserts an entry that was allocated using mb_cache_entry_alloc() into * the cache. After this, the cache entry can be looked up, but is not yet * in the lru list as the caller still holds a handle to it. Returns 0 on * success, or -EBUSY if a cache entry for that device + inode exists * already (this may happen after a failed lookup, but when another process * has inserted the same cache entry in the meantime). * * @bdev: device the cache entry belongs to * @block: block number |
2aec7c523
|
371 |
* @key: lookup key |
1da177e4c
|
372 373 374 |
*/ int mb_cache_entry_insert(struct mb_cache_entry *ce, struct block_device *bdev, |
2aec7c523
|
375 |
sector_t block, unsigned int key) |
1da177e4c
|
376 377 378 379 |
{ struct mb_cache *cache = ce->e_cache; unsigned int bucket; struct list_head *l; |
2aec7c523
|
380 |
int error = -EBUSY; |
1da177e4c
|
381 382 383 384 385 386 387 388 389 390 391 392 393 394 |
bucket = hash_long((unsigned long)bdev + (block & 0xffffffff), cache->c_bucket_bits); spin_lock(&mb_cache_spinlock); list_for_each_prev(l, &cache->c_block_hash[bucket]) { struct mb_cache_entry *ce = list_entry(l, struct mb_cache_entry, e_block_list); if (ce->e_bdev == bdev && ce->e_block == block) goto out; } __mb_cache_entry_unhash(ce); ce->e_bdev = bdev; ce->e_block = block; list_add(&ce->e_block_list, &cache->c_block_hash[bucket]); |
2aec7c523
|
395 396 397 |
ce->e_index.o_key = key; bucket = hash_long(key, cache->c_bucket_bits); list_add(&ce->e_index.o_list, &cache->c_index_hash[bucket]); |
1da177e4c
|
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 |
error = 0; out: spin_unlock(&mb_cache_spinlock); return error; } /* * mb_cache_entry_release() * * Release a handle to a cache entry. When the last handle to a cache entry * is released it is either freed (if it is invalid) or otherwise inserted * in to the lru list. */ void mb_cache_entry_release(struct mb_cache_entry *ce) { spin_lock(&mb_cache_spinlock); __mb_cache_entry_release_unlock(ce); } /* * mb_cache_entry_free() * * This is equivalent to the sequence mb_cache_entry_takeout() -- * mb_cache_entry_release(). */ void mb_cache_entry_free(struct mb_cache_entry *ce) { spin_lock(&mb_cache_spinlock); mb_assert(list_empty(&ce->e_lru_list)); __mb_cache_entry_unhash(ce); __mb_cache_entry_release_unlock(ce); } /* * mb_cache_entry_get() * * Get a cache entry by device / block number. (There can only be one entry * in the cache per device and block.) Returns NULL if no such cache entry * exists. The returned cache entry is locked for exclusive access ("single * writer"). */ struct mb_cache_entry * mb_cache_entry_get(struct mb_cache *cache, struct block_device *bdev, sector_t block) { unsigned int bucket; struct list_head *l; struct mb_cache_entry *ce; bucket = hash_long((unsigned long)bdev + (block & 0xffffffff), cache->c_bucket_bits); spin_lock(&mb_cache_spinlock); list_for_each(l, &cache->c_block_hash[bucket]) { ce = list_entry(l, struct mb_cache_entry, e_block_list); if (ce->e_bdev == bdev && ce->e_block == block) { DEFINE_WAIT(wait); if (!list_empty(&ce->e_lru_list)) list_del_init(&ce->e_lru_list); while (ce->e_used > 0) { ce->e_queued++; prepare_to_wait(&mb_cache_queue, &wait, TASK_UNINTERRUPTIBLE); spin_unlock(&mb_cache_spinlock); schedule(); spin_lock(&mb_cache_spinlock); ce->e_queued--; } finish_wait(&mb_cache_queue, &wait); ce->e_used += 1 + MB_CACHE_WRITER; if (!__mb_cache_entry_is_hashed(ce)) { __mb_cache_entry_release_unlock(ce); return NULL; } goto cleanup; } } ce = NULL; cleanup: spin_unlock(&mb_cache_spinlock); return ce; } #if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) static struct mb_cache_entry * __mb_cache_entry_find(struct list_head *l, struct list_head *head, |
2aec7c523
|
493 |
struct block_device *bdev, unsigned int key) |
1da177e4c
|
494 495 496 |
{ while (l != head) { struct mb_cache_entry *ce = |
2aec7c523
|
497 498 |
list_entry(l, struct mb_cache_entry, e_index.o_list); if (ce->e_bdev == bdev && ce->e_index.o_key == key) { |
1da177e4c
|
499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 |
DEFINE_WAIT(wait); if (!list_empty(&ce->e_lru_list)) list_del_init(&ce->e_lru_list); /* Incrementing before holding the lock gives readers priority over writers. */ ce->e_used++; while (ce->e_used >= MB_CACHE_WRITER) { ce->e_queued++; prepare_to_wait(&mb_cache_queue, &wait, TASK_UNINTERRUPTIBLE); spin_unlock(&mb_cache_spinlock); schedule(); spin_lock(&mb_cache_spinlock); ce->e_queued--; } finish_wait(&mb_cache_queue, &wait); if (!__mb_cache_entry_is_hashed(ce)) { __mb_cache_entry_release_unlock(ce); spin_lock(&mb_cache_spinlock); return ERR_PTR(-EAGAIN); } return ce; } l = l->next; } return NULL; } /* * mb_cache_entry_find_first() * * Find the first cache entry on a given device with a certain key in |
25985edce
|
535 |
* an additional index. Additional matches can be found with |
1da177e4c
|
536 537 538 539 |
* mb_cache_entry_find_next(). Returns NULL if no match was found. The * returned cache entry is locked for shared access ("multiple readers"). * * @cache: the cache to search |
1da177e4c
|
540 541 542 543 |
* @bdev: the device the cache entry should belong to * @key: the key in the index */ struct mb_cache_entry * |
2aec7c523
|
544 545 |
mb_cache_entry_find_first(struct mb_cache *cache, struct block_device *bdev, unsigned int key) |
1da177e4c
|
546 547 548 549 |
{ unsigned int bucket = hash_long(key, cache->c_bucket_bits); struct list_head *l; struct mb_cache_entry *ce; |
1da177e4c
|
550 |
spin_lock(&mb_cache_spinlock); |
2aec7c523
|
551 552 |
l = cache->c_index_hash[bucket].next; ce = __mb_cache_entry_find(l, &cache->c_index_hash[bucket], bdev, key); |
1da177e4c
|
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 |
spin_unlock(&mb_cache_spinlock); return ce; } /* * mb_cache_entry_find_next() * * Find the next cache entry on a given device with a certain key in an * additional index. Returns NULL if no match could be found. The previous * entry is atomatically released, so that mb_cache_entry_find_next() can * be called like this: * * entry = mb_cache_entry_find_first(); * while (entry) { * ... * entry = mb_cache_entry_find_next(entry, ...); * } * * @prev: The previous match |
1da177e4c
|
573 574 575 576 |
* @bdev: the device the cache entry should belong to * @key: the key in the index */ struct mb_cache_entry * |
2aec7c523
|
577 |
mb_cache_entry_find_next(struct mb_cache_entry *prev, |
1da177e4c
|
578 579 580 581 582 583 |
struct block_device *bdev, unsigned int key) { struct mb_cache *cache = prev->e_cache; unsigned int bucket = hash_long(key, cache->c_bucket_bits); struct list_head *l; struct mb_cache_entry *ce; |
1da177e4c
|
584 |
spin_lock(&mb_cache_spinlock); |
2aec7c523
|
585 586 |
l = prev->e_index.o_list.next; ce = __mb_cache_entry_find(l, &cache->c_index_hash[bucket], bdev, key); |
1da177e4c
|
587 588 589 590 591 592 593 594 |
__mb_cache_entry_release_unlock(prev); return ce; } #endif /* !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) */ static int __init init_mbcache(void) { |
8e1f936b7
|
595 |
register_shrinker(&mb_cache_shrinker); |
1da177e4c
|
596 597 598 599 600 |
return 0; } static void __exit exit_mbcache(void) { |
8e1f936b7
|
601 |
unregister_shrinker(&mb_cache_shrinker); |
1da177e4c
|
602 603 604 605 |
} module_init(init_mbcache) module_exit(exit_mbcache) |