mbcache.c 24.1 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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
/*
 * 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.
 */

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 62 63
/*
 * Lock descriptions and usage:
 *
 * Each hash chain of both the block and index hash tables now contains
 * a built-in lock used to serialize accesses to the hash chain.
 *
 * Accesses to global data structures mb_cache_list and mb_cache_lru_list
 * are serialized via the global spinlock mb_cache_spinlock.
 *
 * Each mb_cache_entry contains a spinlock, e_entry_lock, to serialize
 * accesses to its local data, such as e_used and e_queued.
 *
 * Lock ordering:
 *
 * Each block hash chain's lock has the highest lock order, followed by an
 * index hash chain's lock, mb_cache_bg_lock (used to implement mb_cache_entry's
 * lock), and mb_cach_spinlock, with the lowest order.  While holding
 * either a block or index hash chain lock, a thread can acquire an
 * mc_cache_bg_lock, which in turn can also acquire mb_cache_spinlock.
 *
 * Synchronization:
 *
 * Since both mb_cache_entry_get and mb_cache_entry_find scan the block and
 * index hash chian, it needs to lock the corresponding hash chain.  For each
 * mb_cache_entry within the chain, it needs to lock the mb_cache_entry to
 * prevent either any simultaneous release or free on the entry and also
 * to serialize accesses to either the e_used or e_queued member of the entry.
 *
 * To avoid having a dangling reference to an already freed
 * mb_cache_entry, an mb_cache_entry is only freed when it is not on a
 * block hash chain and also no longer being referenced, both e_used,
 * and e_queued are 0's.  When an mb_cache_entry is explicitly freed it is
 * first removed from a block hash chain.
 */

Linus Torvalds's avatar
Linus Torvalds committed
64 65 66 67 68 69 70 71
#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>
72
#include <linux/list_bl.h>
Linus Torvalds's avatar
Linus Torvalds committed
73
#include <linux/mbcache.h>
74
#include <linux/init.h>
75
#include <linux/blockgroup_lock.h>
76
#include <linux/log2.h>
Linus Torvalds's avatar
Linus Torvalds committed
77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96

#ifdef MB_CACHE_DEBUG
# define mb_debug(f...) do { \
		printk(KERN_DEBUG f); \
		printk("\n"); \
	} while (0)
#define mb_assert(c) do { if (!(c)) \
		printk(KERN_ERR "assertion " #c " failed\n"); \
	} 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("\n"); \
	} while(0)

#define MB_CACHE_WRITER ((unsigned short)~0U >> 1)

97
#define MB_CACHE_ENTRY_LOCK_BITS	ilog2(NR_BG_LOCKS)
98 99 100
#define	MB_CACHE_ENTRY_LOCK_INDEX(ce)			\
	(hash_long((unsigned long)ce, MB_CACHE_ENTRY_LOCK_BITS))

101
static DECLARE_WAIT_QUEUE_HEAD(mb_cache_queue);
102
static struct blockgroup_lock *mb_cache_bg_lock;
103
static struct kmem_cache *mb_cache_kmem_cache;
104

Linus Torvalds's avatar
Linus Torvalds committed
105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131
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

/*
 * 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);

132 133 134 135 136 137 138 139 140 141 142 143 144 145
static inline void
__spin_lock_mb_cache_entry(struct mb_cache_entry *ce)
{
	spin_lock(bgl_lock_ptr(mb_cache_bg_lock,
		MB_CACHE_ENTRY_LOCK_INDEX(ce)));
}

static inline void
__spin_unlock_mb_cache_entry(struct mb_cache_entry *ce)
{
	spin_unlock(bgl_lock_ptr(mb_cache_bg_lock,
		MB_CACHE_ENTRY_LOCK_INDEX(ce)));
}

Linus Torvalds's avatar
Linus Torvalds committed
146
static inline int
147
__mb_cache_entry_is_block_hashed(struct mb_cache_entry *ce)
Linus Torvalds's avatar
Linus Torvalds committed
148
{
149
	return !hlist_bl_unhashed(&ce->e_block_list);
Linus Torvalds's avatar
Linus Torvalds committed
150 151 152
}


153 154
static inline void
__mb_cache_entry_unhash_block(struct mb_cache_entry *ce)
Linus Torvalds's avatar
Linus Torvalds committed
155
{
156 157 158 159 160 161 162 163
	if (__mb_cache_entry_is_block_hashed(ce))
		hlist_bl_del_init(&ce->e_block_list);
}

static inline int
__mb_cache_entry_is_index_hashed(struct mb_cache_entry *ce)
{
	return !hlist_bl_unhashed(&ce->e_index.o_list);
Linus Torvalds's avatar
Linus Torvalds committed
164 165
}

166 167 168 169 170 171 172
static inline void
__mb_cache_entry_unhash_index(struct mb_cache_entry *ce)
{
	if (__mb_cache_entry_is_index_hashed(ce))
		hlist_bl_del_init(&ce->e_index.o_list);
}

173 174 175 176 177 178 179 180
/*
 * __mb_cache_entry_unhash_unlock()
 *
 * This function is called to unhash both the block and index hash
 * chain.
 * It assumes both the block and index hash chain is locked upon entry.
 * It also unlock both hash chains both exit
 */
181
static inline void
182
__mb_cache_entry_unhash_unlock(struct mb_cache_entry *ce)
183 184
{
	__mb_cache_entry_unhash_index(ce);
185
	hlist_bl_unlock(ce->e_index_hash_p);
186
	__mb_cache_entry_unhash_block(ce);
187
	hlist_bl_unlock(ce->e_block_hash_p);
188
}
Linus Torvalds's avatar
Linus Torvalds committed
189

190
static void
Al Viro's avatar
Al Viro committed
191
__mb_cache_entry_forget(struct mb_cache_entry *ce, gfp_t gfp_mask)
Linus Torvalds's avatar
Linus Torvalds committed
192 193 194
{
	struct mb_cache *cache = ce->e_cache;

195
	mb_assert(!(ce->e_used || ce->e_queued || atomic_read(&ce->e_refcnt)));
196 197
	kmem_cache_free(cache->c_entry_cache, ce);
	atomic_dec(&cache->c_entry_count);
Linus Torvalds's avatar
Linus Torvalds committed
198 199
}

200
static void
201
__mb_cache_entry_release(struct mb_cache_entry *ce)
Linus Torvalds's avatar
Linus Torvalds committed
202
{
203 204
	/* First lock the entry to serialize access to its local data. */
	__spin_lock_mb_cache_entry(ce);
Linus Torvalds's avatar
Linus Torvalds committed
205 206 207 208 209
	/* 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;
210 211 212 213
	/*
	 * Make sure that all cache entries on lru_list have
	 * both e_used and e_qued of 0s.
	 */
Linus Torvalds's avatar
Linus Torvalds committed
214
	ce->e_used--;
215 216 217
	if (!(ce->e_used || ce->e_queued || atomic_read(&ce->e_refcnt))) {
		if (!__mb_cache_entry_is_block_hashed(ce)) {
			__spin_unlock_mb_cache_entry(ce);
Linus Torvalds's avatar
Linus Torvalds committed
218
			goto forget;
219 220 221 222 223 224 225 226 227
		}
		/*
		 * Need access to lru list, first drop entry lock,
		 * then reacquire the lock in the proper order.
		 */
		spin_lock(&mb_cache_spinlock);
		if (list_empty(&ce->e_lru_list))
			list_add_tail(&ce->e_lru_list, &mb_cache_lru_list);
		spin_unlock(&mb_cache_spinlock);
Linus Torvalds's avatar
Linus Torvalds committed
228
	}
229
	__spin_unlock_mb_cache_entry(ce);
Linus Torvalds's avatar
Linus Torvalds committed
230 231
	return;
forget:
232
	mb_assert(list_empty(&ce->e_lru_list));
Linus Torvalds's avatar
Linus Torvalds committed
233 234 235 236
	__mb_cache_entry_forget(ce, GFP_KERNEL);
}

/*
237
 * mb_cache_shrink_scan()  memory pressure callback
Linus Torvalds's avatar
Linus Torvalds committed
238 239 240 241
 *
 * This function is called by the kernel memory management when memory
 * gets low.
 *
242
 * @shrink: (ignored)
243
 * @sc: shrink_control passed from reclaim
Linus Torvalds's avatar
Linus Torvalds committed
244
 *
245
 * Returns the number of objects freed.
Linus Torvalds's avatar
Linus Torvalds committed
246
 */
247 248
static unsigned long
mb_cache_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
Linus Torvalds's avatar
Linus Torvalds committed
249 250
{
	LIST_HEAD(free_list);
251
	struct mb_cache_entry *entry, *tmp;
252 253
	int nr_to_scan = sc->nr_to_scan;
	gfp_t gfp_mask = sc->gfp_mask;
254
	unsigned long freed = 0;
Linus Torvalds's avatar
Linus Torvalds committed
255 256

	mb_debug("trying to free %d entries", nr_to_scan);
257
	spin_lock(&mb_cache_spinlock);
258
	while ((nr_to_scan-- > 0) && !list_empty(&mb_cache_lru_list)) {
Linus Torvalds's avatar
Linus Torvalds committed
259 260
		struct mb_cache_entry *ce =
			list_entry(mb_cache_lru_list.next,
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
				struct mb_cache_entry, e_lru_list);
		list_del_init(&ce->e_lru_list);
		if (ce->e_used || ce->e_queued || atomic_read(&ce->e_refcnt))
			continue;
		spin_unlock(&mb_cache_spinlock);
		/* Prevent any find or get operation on the entry */
		hlist_bl_lock(ce->e_block_hash_p);
		hlist_bl_lock(ce->e_index_hash_p);
		/* Ignore if it is touched by a find/get */
		if (ce->e_used || ce->e_queued || atomic_read(&ce->e_refcnt) ||
			!list_empty(&ce->e_lru_list)) {
			hlist_bl_unlock(ce->e_index_hash_p);
			hlist_bl_unlock(ce->e_block_hash_p);
			spin_lock(&mb_cache_spinlock);
			continue;
		}
		__mb_cache_entry_unhash_unlock(ce);
		list_add_tail(&ce->e_lru_list, &free_list);
		spin_lock(&mb_cache_spinlock);
280 281
	}
	spin_unlock(&mb_cache_spinlock);
282

283 284
	list_for_each_entry_safe(entry, tmp, &free_list, e_lru_list) {
		__mb_cache_entry_forget(entry, gfp_mask);
285
		freed++;
Linus Torvalds's avatar
Linus Torvalds committed
286
	}
287 288 289 290 291 292 293 294 295 296
	return freed;
}

static unsigned long
mb_cache_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
	struct mb_cache *cache;
	unsigned long count = 0;

	spin_lock(&mb_cache_spinlock);
297 298 299 300 301
	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);
	}
Linus Torvalds's avatar
Linus Torvalds committed
302
	spin_unlock(&mb_cache_spinlock);
303

304
	return vfs_pressure_ratio(count);
Linus Torvalds's avatar
Linus Torvalds committed
305 306
}

307 308 309 310 311
static struct shrinker mb_cache_shrinker = {
	.count_objects = mb_cache_shrink_count,
	.scan_objects = mb_cache_shrink_scan,
	.seeks = DEFAULT_SEEKS,
};
Linus Torvalds's avatar
Linus Torvalds committed
312 313 314 315 316 317 318 319 320 321 322 323 324

/*
 * 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)
 * @bucket_bits: log2(number of hash buckets)
 */
struct mb_cache *
325
mb_cache_create(const char *name, int bucket_bits)
Linus Torvalds's avatar
Linus Torvalds committed
326
{
327
	int n, bucket_count = 1 << bucket_bits;
Linus Torvalds's avatar
Linus Torvalds committed
328 329
	struct mb_cache *cache = NULL;

330 331 332 333 334 335 336 337
	if (!mb_cache_bg_lock) {
		mb_cache_bg_lock = kmalloc(sizeof(struct blockgroup_lock),
			GFP_KERNEL);
		if (!mb_cache_bg_lock)
			return NULL;
		bgl_lock_init(mb_cache_bg_lock);
	}

338
	cache = kmalloc(sizeof(struct mb_cache), GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
339
	if (!cache)
340
		return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
341 342 343
	cache->c_name = name;
	atomic_set(&cache->c_entry_count, 0);
	cache->c_bucket_bits = bucket_bits;
344 345
	cache->c_block_hash = kmalloc(bucket_count *
		sizeof(struct hlist_bl_head), GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
346 347 348
	if (!cache->c_block_hash)
		goto fail;
	for (n=0; n<bucket_count; n++)
349 350 351
		INIT_HLIST_BL_HEAD(&cache->c_block_hash[n]);
	cache->c_index_hash = kmalloc(bucket_count *
		sizeof(struct hlist_bl_head), GFP_KERNEL);
352 353 354
	if (!cache->c_index_hash)
		goto fail;
	for (n=0; n<bucket_count; n++)
355
		INIT_HLIST_BL_HEAD(&cache->c_index_hash[n]);
356 357 358 359 360 361 362 363
	if (!mb_cache_kmem_cache) {
		mb_cache_kmem_cache = kmem_cache_create(name,
			sizeof(struct mb_cache_entry), 0,
			SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD, NULL);
		if (!mb_cache_kmem_cache)
			goto fail2;
	}
	cache->c_entry_cache = mb_cache_kmem_cache;
Linus Torvalds's avatar
Linus Torvalds committed
364

365 366 367 368 369 370
	/*
	 * 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;

Linus Torvalds's avatar
Linus Torvalds committed
371 372 373 374 375
	spin_lock(&mb_cache_spinlock);
	list_add(&cache->c_cache_list, &mb_cache_list);
	spin_unlock(&mb_cache_spinlock);
	return cache;

376 377 378
fail2:
	kfree(cache->c_index_hash);

Linus Torvalds's avatar
Linus Torvalds committed
379
fail:
380 381
	kfree(cache->c_block_hash);
	kfree(cache);
Linus Torvalds's avatar
Linus Torvalds committed
382 383 384 385 386 387 388
	return NULL;
}


/*
 * mb_cache_shrink()
 *
Alexey Dobriyan's avatar
Alexey Dobriyan committed
389
 * Removes all cache entries of a device from the cache. All cache entries
Linus Torvalds's avatar
Linus Torvalds committed
390 391 392 393 394 395
 * currently in use cannot be freed, and thus remain in the cache. All others
 * are freed.
 *
 * @bdev: which device's cache entries to shrink
 */
void
396
mb_cache_shrink(struct block_device *bdev)
Linus Torvalds's avatar
Linus Torvalds committed
397 398
{
	LIST_HEAD(free_list);
399 400
	struct list_head *l;
	struct mb_cache_entry *ce, *tmp;
Linus Torvalds's avatar
Linus Torvalds committed
401

402
	l = &mb_cache_lru_list;
Linus Torvalds's avatar
Linus Torvalds committed
403
	spin_lock(&mb_cache_spinlock);
404 405 406
	while (!list_is_last(l, &mb_cache_lru_list)) {
		l = l->next;
		ce = list_entry(l, struct mb_cache_entry, e_lru_list);
Linus Torvalds's avatar
Linus Torvalds committed
407
		if (ce->e_bdev == bdev) {
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
			list_del_init(&ce->e_lru_list);
			if (ce->e_used || ce->e_queued ||
				atomic_read(&ce->e_refcnt))
				continue;
			spin_unlock(&mb_cache_spinlock);
			/*
			 * Prevent any find or get operation on the entry.
			 */
			hlist_bl_lock(ce->e_block_hash_p);
			hlist_bl_lock(ce->e_index_hash_p);
			/* Ignore if it is touched by a find/get */
			if (ce->e_used || ce->e_queued ||
				atomic_read(&ce->e_refcnt) ||
				!list_empty(&ce->e_lru_list)) {
				hlist_bl_unlock(ce->e_index_hash_p);
				hlist_bl_unlock(ce->e_block_hash_p);
				l = &mb_cache_lru_list;
				spin_lock(&mb_cache_spinlock);
				continue;
			}
			__mb_cache_entry_unhash_unlock(ce);
			mb_assert(!(ce->e_used || ce->e_queued ||
				atomic_read(&ce->e_refcnt)));
			list_add_tail(&ce->e_lru_list, &free_list);
			l = &mb_cache_lru_list;
			spin_lock(&mb_cache_spinlock);
Linus Torvalds's avatar
Linus Torvalds committed
434 435 436
		}
	}
	spin_unlock(&mb_cache_spinlock);
437 438 439

	list_for_each_entry_safe(ce, tmp, &free_list, e_lru_list) {
		__mb_cache_entry_forget(ce, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
	}
}


/*
 * 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);
455
	struct mb_cache_entry *ce, *tmp;
Linus Torvalds's avatar
Linus Torvalds committed
456 457

	spin_lock(&mb_cache_spinlock);
458 459
	list_for_each_entry_safe(ce, tmp, &mb_cache_lru_list, e_lru_list) {
		if (ce->e_cache == cache)
Linus Torvalds's avatar
Linus Torvalds committed
460 461 462 463 464
			list_move_tail(&ce->e_lru_list, &free_list);
	}
	list_del(&cache->c_cache_list);
	spin_unlock(&mb_cache_spinlock);

465 466 467 468 469 470 471 472 473 474 475
	list_for_each_entry_safe(ce, tmp, &free_list, e_lru_list) {
		list_del_init(&ce->e_lru_list);
		/*
		 * Prevent any find or get operation on the entry.
		 */
		hlist_bl_lock(ce->e_block_hash_p);
		hlist_bl_lock(ce->e_index_hash_p);
		mb_assert(!(ce->e_used || ce->e_queued ||
			atomic_read(&ce->e_refcnt)));
		__mb_cache_entry_unhash_unlock(ce);
		__mb_cache_entry_forget(ce, GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
476 477 478 479 480 481 482 483
	}

	if (atomic_read(&cache->c_entry_count) > 0) {
		mb_error("cache %s: %d orphaned entries",
			  cache->c_name,
			  atomic_read(&cache->c_entry_count));
	}

484 485 486 487
	if (list_empty(&mb_cache_list)) {
		kmem_cache_destroy(mb_cache_kmem_cache);
		mb_cache_kmem_cache = NULL;
	}
488
	kfree(cache->c_index_hash);
Linus Torvalds's avatar
Linus Torvalds committed
489 490 491 492 493 494 495 496 497 498 499 500 501
	kfree(cache->c_block_hash);
	kfree(cache);
}

/*
 * 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 *
502
mb_cache_entry_alloc(struct mb_cache *cache, gfp_t gfp_flags)
Linus Torvalds's avatar
Linus Torvalds committed
503
{
504
	struct mb_cache_entry *ce;
505 506

	if (atomic_read(&cache->c_entry_count) >= cache->c_max_entries) {
507 508 509
		struct list_head *l;

		l = &mb_cache_lru_list;
510
		spin_lock(&mb_cache_spinlock);
511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
		while (!list_is_last(l, &mb_cache_lru_list)) {
			l = l->next;
			ce = list_entry(l, struct mb_cache_entry, e_lru_list);
			if (ce->e_cache == cache) {
				list_del_init(&ce->e_lru_list);
				if (ce->e_used || ce->e_queued ||
					atomic_read(&ce->e_refcnt))
					continue;
				spin_unlock(&mb_cache_spinlock);
				/*
				 * Prevent any find or get operation on the
				 * entry.
				 */
				hlist_bl_lock(ce->e_block_hash_p);
				hlist_bl_lock(ce->e_index_hash_p);
				/* Ignore if it is touched by a find/get */
				if (ce->e_used || ce->e_queued ||
					atomic_read(&ce->e_refcnt) ||
					!list_empty(&ce->e_lru_list)) {
					hlist_bl_unlock(ce->e_index_hash_p);
					hlist_bl_unlock(ce->e_block_hash_p);
					l = &mb_cache_lru_list;
					spin_lock(&mb_cache_spinlock);
					continue;
				}
				mb_assert(list_empty(&ce->e_lru_list));
				mb_assert(!(ce->e_used || ce->e_queued ||
					atomic_read(&ce->e_refcnt)));
				__mb_cache_entry_unhash_unlock(ce);
				goto found;
			}
542 543 544
		}
		spin_unlock(&mb_cache_spinlock);
	}
545 546 547 548 549 550 551 552 553 554 555 556

	ce = kmem_cache_alloc(cache->c_entry_cache, gfp_flags);
	if (!ce)
		return NULL;
	atomic_inc(&cache->c_entry_count);
	INIT_LIST_HEAD(&ce->e_lru_list);
	INIT_HLIST_BL_NODE(&ce->e_block_list);
	INIT_HLIST_BL_NODE(&ce->e_index.o_list);
	ce->e_cache = cache;
	ce->e_queued = 0;
	atomic_set(&ce->e_refcnt, 0);
found:
557 558
	ce->e_block_hash_p = &cache->c_block_hash[0];
	ce->e_index_hash_p = &cache->c_index_hash[0];
559
	ce->e_used = 1 + MB_CACHE_WRITER;
Linus Torvalds's avatar
Linus Torvalds committed
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
	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
576
 * @key: lookup key
Linus Torvalds's avatar
Linus Torvalds committed
577 578 579
 */
int
mb_cache_entry_insert(struct mb_cache_entry *ce, struct block_device *bdev,
580
		      sector_t block, unsigned int key)
Linus Torvalds's avatar
Linus Torvalds committed
581 582 583
{
	struct mb_cache *cache = ce->e_cache;
	unsigned int bucket;
584 585 586 587
	struct hlist_bl_node *l;
	struct hlist_bl_head *block_hash_p;
	struct hlist_bl_head *index_hash_p;
	struct mb_cache_entry *lce;
Linus Torvalds's avatar
Linus Torvalds committed
588

589
	mb_assert(ce);
Linus Torvalds's avatar
Linus Torvalds committed
590 591
	bucket = hash_long((unsigned long)bdev + (block & 0xffffffff), 
			   cache->c_bucket_bits);
592
	block_hash_p = &cache->c_block_hash[bucket];
593
	hlist_bl_lock(block_hash_p);
594
	hlist_bl_for_each_entry(lce, l, block_hash_p, e_block_list) {
595 596 597 598
		if (lce->e_bdev == bdev && lce->e_block == block) {
			hlist_bl_unlock(block_hash_p);
			return -EBUSY;
		}
Linus Torvalds's avatar
Linus Torvalds committed
599
	}
600
	mb_assert(!__mb_cache_entry_is_block_hashed(ce));
601 602
	__mb_cache_entry_unhash_block(ce);
	__mb_cache_entry_unhash_index(ce);
Linus Torvalds's avatar
Linus Torvalds committed
603 604
	ce->e_bdev = bdev;
	ce->e_block = block;
605
	ce->e_block_hash_p = block_hash_p;
606
	ce->e_index.o_key = key;
607 608
	hlist_bl_add_head(&ce->e_block_list, block_hash_p);
	hlist_bl_unlock(block_hash_p);
609
	bucket = hash_long(key, cache->c_bucket_bits);
610
	index_hash_p = &cache->c_index_hash[bucket];
611
	hlist_bl_lock(index_hash_p);
612 613
	ce->e_index_hash_p = index_hash_p;
	hlist_bl_add_head(&ce->e_index.o_list, index_hash_p);
614 615
	hlist_bl_unlock(index_hash_p);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
616 617 618 619 620 621 622 623 624 625 626 627 628
}


/*
 * 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)
{
629
	__mb_cache_entry_release(ce);
Linus Torvalds's avatar
Linus Torvalds committed
630 631 632 633 634 635 636 637 638 639
}


/*
 * mb_cache_entry_free()
 *
 */
void
mb_cache_entry_free(struct mb_cache_entry *ce)
{
640
	mb_assert(ce);
Linus Torvalds's avatar
Linus Torvalds committed
641
	mb_assert(list_empty(&ce->e_lru_list));
642 643 644 645 646 647 648
	hlist_bl_lock(ce->e_index_hash_p);
	__mb_cache_entry_unhash_index(ce);
	hlist_bl_unlock(ce->e_index_hash_p);
	hlist_bl_lock(ce->e_block_hash_p);
	__mb_cache_entry_unhash_block(ce);
	hlist_bl_unlock(ce->e_block_hash_p);
	__mb_cache_entry_release(ce);
Linus Torvalds's avatar
Linus Torvalds committed
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
}


/*
 * 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;
665
	struct hlist_bl_node *l;
Linus Torvalds's avatar
Linus Torvalds committed
666
	struct mb_cache_entry *ce;
667
	struct hlist_bl_head *block_hash_p;
Linus Torvalds's avatar
Linus Torvalds committed
668 669 670

	bucket = hash_long((unsigned long)bdev + (block & 0xffffffff),
			   cache->c_bucket_bits);
671
	block_hash_p = &cache->c_block_hash[bucket];
672 673
	/* First serialize access to the block corresponding hash chain. */
	hlist_bl_lock(block_hash_p);
674 675
	hlist_bl_for_each_entry(ce, l, block_hash_p, e_block_list) {
		mb_assert(ce->e_block_hash_p == block_hash_p);
Linus Torvalds's avatar
Linus Torvalds committed
676
		if (ce->e_bdev == bdev && ce->e_block == block) {
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698
			/*
			 * Prevent a free from removing the entry.
			 */
			atomic_inc(&ce->e_refcnt);
			hlist_bl_unlock(block_hash_p);
			__spin_lock_mb_cache_entry(ce);
			atomic_dec(&ce->e_refcnt);
			if (ce->e_used > 0) {
				DEFINE_WAIT(wait);
				while (ce->e_used > 0) {
					ce->e_queued++;
					prepare_to_wait(&mb_cache_queue, &wait,
							TASK_UNINTERRUPTIBLE);
					__spin_unlock_mb_cache_entry(ce);
					schedule();
					__spin_lock_mb_cache_entry(ce);
					ce->e_queued--;
				}
				finish_wait(&mb_cache_queue, &wait);
			}
			ce->e_used += 1 + MB_CACHE_WRITER;
			__spin_unlock_mb_cache_entry(ce);
Linus Torvalds's avatar
Linus Torvalds committed
699

700 701
			if (!list_empty(&ce->e_lru_list)) {
				spin_lock(&mb_cache_spinlock);
Linus Torvalds's avatar
Linus Torvalds committed
702 703 704
				list_del_init(&ce->e_lru_list);
				spin_unlock(&mb_cache_spinlock);
			}
705
			if (!__mb_cache_entry_is_block_hashed(ce)) {
706
				__mb_cache_entry_release(ce);
Linus Torvalds's avatar
Linus Torvalds committed
707 708
				return NULL;
			}
709
			return ce;
Linus Torvalds's avatar
Linus Torvalds committed
710 711
		}
	}
712 713
	hlist_bl_unlock(block_hash_p);
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
714 715 716 717 718
}

#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0)

static struct mb_cache_entry *
719
__mb_cache_entry_find(struct hlist_bl_node *l, struct hlist_bl_head *head,
720
		      struct block_device *bdev, unsigned int key)
Linus Torvalds's avatar
Linus Torvalds committed
721
{
722 723

	/* The index hash chain is alredy acquire by caller. */
724
	while (l != NULL) {
Linus Torvalds's avatar
Linus Torvalds committed
725
		struct mb_cache_entry *ce =
726 727 728
			hlist_bl_entry(l, struct mb_cache_entry,
				e_index.o_list);
		mb_assert(ce->e_index_hash_p == head);
729
		if (ce->e_bdev == bdev && ce->e_index.o_key == key) {
730 731 732 733 734 735 736 737
			/*
			 * Prevent a free from removing the entry.
			 */
			atomic_inc(&ce->e_refcnt);
			hlist_bl_unlock(head);
			__spin_lock_mb_cache_entry(ce);
			atomic_dec(&ce->e_refcnt);
			ce->e_used++;
Linus Torvalds's avatar
Linus Torvalds committed
738 739
			/* Incrementing before holding the lock gives readers
			   priority over writers. */
740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
			if (ce->e_used >= MB_CACHE_WRITER) {
				DEFINE_WAIT(wait);

				while (ce->e_used >= MB_CACHE_WRITER) {
					ce->e_queued++;
					prepare_to_wait(&mb_cache_queue, &wait,
							TASK_UNINTERRUPTIBLE);
					__spin_unlock_mb_cache_entry(ce);
					schedule();
					__spin_lock_mb_cache_entry(ce);
					ce->e_queued--;
				}
				finish_wait(&mb_cache_queue, &wait);
			}
			__spin_unlock_mb_cache_entry(ce);
			if (!list_empty(&ce->e_lru_list)) {
Linus Torvalds's avatar
Linus Torvalds committed
756
				spin_lock(&mb_cache_spinlock);
757 758
				list_del_init(&ce->e_lru_list);
				spin_unlock(&mb_cache_spinlock);
Linus Torvalds's avatar
Linus Torvalds committed
759
			}
760
			if (!__mb_cache_entry_is_block_hashed(ce)) {
761
				__mb_cache_entry_release(ce);
Linus Torvalds's avatar
Linus Torvalds committed
762 763 764 765 766 767
				return ERR_PTR(-EAGAIN);
			}
			return ce;
		}
		l = l->next;
	}
768
	hlist_bl_unlock(head);
Linus Torvalds's avatar
Linus Torvalds committed
769 770 771 772 773 774 775 776
	return NULL;
}


/*
 * mb_cache_entry_find_first()
 *
 * Find the first cache entry on a given device with a certain key in
Lucas De Marchi's avatar
Lucas De Marchi committed
777
 * an additional index. Additional matches can be found with
Linus Torvalds's avatar
Linus Torvalds committed
778 779 780 781 782 783 784 785
 * 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
 * @bdev: the device the cache entry should belong to
 * @key: the key in the index
 */
struct mb_cache_entry *
786 787
mb_cache_entry_find_first(struct mb_cache *cache, struct block_device *bdev,
			  unsigned int key)
Linus Torvalds's avatar
Linus Torvalds committed
788 789
{
	unsigned int bucket = hash_long(key, cache->c_bucket_bits);
790 791 792
	struct hlist_bl_node *l;
	struct mb_cache_entry *ce = NULL;
	struct hlist_bl_head *index_hash_p;
Linus Torvalds's avatar
Linus Torvalds committed
793

794
	index_hash_p = &cache->c_index_hash[bucket];
795
	hlist_bl_lock(index_hash_p);
796 797 798
	if (!hlist_bl_empty(index_hash_p)) {
		l = hlist_bl_first(index_hash_p);
		ce = __mb_cache_entry_find(l, index_hash_p, bdev, key);
799 800
	} else
		hlist_bl_unlock(index_hash_p);
Linus Torvalds's avatar
Linus Torvalds committed
801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823
	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
 * @bdev: the device the cache entry should belong to
 * @key: the key in the index
 */
struct mb_cache_entry *
824
mb_cache_entry_find_next(struct mb_cache_entry *prev,
Linus Torvalds's avatar
Linus Torvalds committed
825 826 827 828
			 struct block_device *bdev, unsigned int key)
{
	struct mb_cache *cache = prev->e_cache;
	unsigned int bucket = hash_long(key, cache->c_bucket_bits);
829
	struct hlist_bl_node *l;
Linus Torvalds's avatar
Linus Torvalds committed
830
	struct mb_cache_entry *ce;
831
	struct hlist_bl_head *index_hash_p;
Linus Torvalds's avatar
Linus Torvalds committed
832

833 834
	index_hash_p = &cache->c_index_hash[bucket];
	mb_assert(prev->e_index_hash_p == index_hash_p);
835
	hlist_bl_lock(index_hash_p);
836
	mb_assert(!hlist_bl_empty(index_hash_p));
837
	l = prev->e_index.o_list.next;
838
	ce = __mb_cache_entry_find(l, index_hash_p, bdev, key);
839
	__mb_cache_entry_release(prev);
Linus Torvalds's avatar
Linus Torvalds committed
840 841 842 843 844 845 846
	return ce;
}

#endif  /* !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) */

static int __init init_mbcache(void)
{
847
	register_shrinker(&mb_cache_shrinker);
Linus Torvalds's avatar
Linus Torvalds committed
848 849 850 851 852
	return 0;
}

static void __exit exit_mbcache(void)
{
853
	unregister_shrinker(&mb_cache_shrinker);
Linus Torvalds's avatar
Linus Torvalds committed
854 855 856 857 858
}

module_init(init_mbcache)
module_exit(exit_mbcache)