z3fold.c 30.6 KB
Newer Older
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
/*
 * z3fold.c
 *
 * Author: Vitaly Wool <vitaly.wool@konsulko.com>
 * Copyright (C) 2016, Sony Mobile Communications Inc.
 *
 * This implementation is based on zbud written by Seth Jennings.
 *
 * z3fold is an special purpose allocator for storing compressed pages. It
 * can store up to three compressed pages per page which improves the
 * compression ratio of zbud while retaining its main concepts (e. g. always
 * storing an integral number of objects per page) and simplicity.
 * It still has simple and deterministic reclaim properties that make it
 * preferable to a higher density approach (with no requirement on integral
 * number of object per page) when reclaim is used.
 *
 * As in zbud, pages are divided into "chunks".  The size of the chunks is
 * fixed at compile time and is determined by NCHUNKS_ORDER below.
 *
 * z3fold doesn't export any API and is meant to be used via zpool API.
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/atomic.h>
26
#include <linux/sched.h>
27 28 29
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/module.h>
30
#include <linux/percpu.h>
31
#include <linux/preempt.h>
32
#include <linux/workqueue.h>
33 34 35 36 37 38 39
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/zpool.h>

/*****************
 * Structures
*****************/
40 41 42 43 44 45 46 47 48 49 50 51 52 53
struct z3fold_pool;
struct z3fold_ops {
	int (*evict)(struct z3fold_pool *pool, unsigned long handle);
};

enum buddy {
	HEADLESS = 0,
	FIRST,
	MIDDLE,
	LAST,
	BUDDIES_MAX
};

/*
54
 * struct z3fold_header - z3fold page metadata occupying first chunks of each
55
 *			z3fold page, except for HEADLESS pages
56 57
 * @buddy:		links the z3fold page into the relevant list in the
 *			pool
58
 * @page_lock:		per-page lock
59 60 61 62
 * @refcount:		reference count for the z3fold page
 * @work:		work_struct for page layout optimization
 * @pool:		pointer to the pool which this page belongs to
 * @cpu:		CPU which this page "belongs" to
63 64 65 66 67 68 69
 * @first_chunks:	the size of the first buddy in chunks, 0 if free
 * @middle_chunks:	the size of the middle buddy in chunks, 0 if free
 * @last_chunks:	the size of the last buddy in chunks, 0 if free
 * @first_num:		the starting number (for the first handle)
 */
struct z3fold_header {
	struct list_head buddy;
70
	spinlock_t page_lock;
71
	struct kref refcount;
72 73 74
	struct work_struct work;
	struct z3fold_pool *pool;
	short cpu;
75 76 77 78 79 80 81
	unsigned short first_chunks;
	unsigned short middle_chunks;
	unsigned short last_chunks;
	unsigned short start_middle;
	unsigned short first_num:2;
};

82 83 84 85
/*
 * NCHUNKS_ORDER determines the internal allocation granularity, effectively
 * adjusting internal fragmentation.  It also determines the number of
 * freelists maintained in each pool. NCHUNKS_ORDER of 6 means that the
86 87 88 89 90
 * allocation granularity will be in chunks of size PAGE_SIZE/64. Some chunks
 * in the beginning of an allocated page are occupied by z3fold header, so
 * NCHUNKS will be calculated to 63 (or 62 in case CONFIG_DEBUG_SPINLOCK=y),
 * which shows the max number of free chunks in z3fold page, also there will
 * be 63, or 62, respectively, freelists per pool.
91 92 93 94 95
 */
#define NCHUNKS_ORDER	6

#define CHUNK_SHIFT	(PAGE_SHIFT - NCHUNKS_ORDER)
#define CHUNK_SIZE	(1 << CHUNK_SHIFT)
96 97 98
#define ZHDR_SIZE_ALIGNED round_up(sizeof(struct z3fold_header), CHUNK_SIZE)
#define ZHDR_CHUNKS	(ZHDR_SIZE_ALIGNED >> CHUNK_SHIFT)
#define TOTAL_CHUNKS	(PAGE_SIZE >> CHUNK_SHIFT)
99 100
#define NCHUNKS		((PAGE_SIZE - ZHDR_SIZE_ALIGNED) >> CHUNK_SHIFT)

101
#define BUDDY_MASK	(0x3)
102
#define BUDDY_SHIFT	2
103 104 105

/**
 * struct z3fold_pool - stores metadata for each z3fold pool
106 107 108 109 110 111
 * @name:	pool name
 * @lock:	protects pool unbuddied/lru lists
 * @stale_lock:	protects pool stale page list
 * @unbuddied:	per-cpu array of lists tracking z3fold pages that contain 2-
 *		buddies; the list each z3fold page is added to depends on
 *		the size of its free region.
112 113
 * @lru:	list tracking the z3fold pages in LRU order by most recently
 *		added buddy.
114
 * @stale:	list of pages marked for freeing
115 116 117
 * @pages_nr:	number of z3fold pages in the pool.
 * @ops:	pointer to a structure of user defined operations specified at
 *		pool creation time.
118 119 120
 * @compact_wq:	workqueue for page layout background optimization
 * @release_wq:	workqueue for safe page release
 * @work:	work_struct for safe page release
121 122 123 124 125
 *
 * This structure is allocated at pool creation time and maintains metadata
 * pertaining to a particular z3fold pool.
 */
struct z3fold_pool {
126
	const char *name;
127
	spinlock_t lock;
128 129
	spinlock_t stale_lock;
	struct list_head *unbuddied;
130
	struct list_head lru;
131
	struct list_head stale;
132
	atomic64_t pages_nr;
133 134 135
	const struct z3fold_ops *ops;
	struct zpool *zpool;
	const struct zpool_ops *zpool_ops;
136 137 138
	struct workqueue_struct *compact_wq;
	struct workqueue_struct *release_wq;
	struct work_struct work;
139 140 141 142 143 144
};

/*
 * Internal z3fold page flags
 */
enum z3fold_page_flags {
145
	PAGE_HEADLESS = 0,
146
	MIDDLE_CHUNK_MAPPED,
147
	NEEDS_COMPACTING,
148
	PAGE_STALE,
149
	PAGE_CLAIMED, /* by either reclaim or free */
150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
};

/*****************
 * Helpers
*****************/

/* Converts an allocation size in bytes to size in z3fold chunks */
static int size_to_chunks(size_t size)
{
	return (size + CHUNK_SIZE - 1) >> CHUNK_SHIFT;
}

#define for_each_unbuddied_list(_iter, _begin) \
	for ((_iter) = (_begin); (_iter) < NCHUNKS; (_iter)++)

165 166
static void compact_page_work(struct work_struct *w);

167
/* Initializes the z3fold header of a newly allocated z3fold page */
168 169
static struct z3fold_header *init_z3fold_page(struct page *page,
					struct z3fold_pool *pool)
170 171 172 173 174 175
{
	struct z3fold_header *zhdr = page_address(page);

	INIT_LIST_HEAD(&page->lru);
	clear_bit(PAGE_HEADLESS, &page->private);
	clear_bit(MIDDLE_CHUNK_MAPPED, &page->private);
176 177
	clear_bit(NEEDS_COMPACTING, &page->private);
	clear_bit(PAGE_STALE, &page->private);
178
	clear_bit(PAGE_CLAIMED, &page->private);
179

180
	spin_lock_init(&zhdr->page_lock);
181
	kref_init(&zhdr->refcount);
182 183 184 185 186
	zhdr->first_chunks = 0;
	zhdr->middle_chunks = 0;
	zhdr->last_chunks = 0;
	zhdr->first_num = 0;
	zhdr->start_middle = 0;
187 188
	zhdr->cpu = -1;
	zhdr->pool = pool;
189
	INIT_LIST_HEAD(&zhdr->buddy);
190
	INIT_WORK(&zhdr->work, compact_page_work);
191 192 193 194
	return zhdr;
}

/* Resets the struct page fields and frees the page */
195
static void free_z3fold_page(struct page *page)
196
{
197 198 199
	__free_page(page);
}

200 201 202 203 204 205
/* Lock a z3fold page */
static inline void z3fold_page_lock(struct z3fold_header *zhdr)
{
	spin_lock(&zhdr->page_lock);
}

206 207 208 209 210 211
/* Try to lock a z3fold page */
static inline int z3fold_page_trylock(struct z3fold_header *zhdr)
{
	return spin_trylock(&zhdr->page_lock);
}

212 213 214 215 216 217
/* Unlock a z3fold page */
static inline void z3fold_page_unlock(struct z3fold_header *zhdr)
{
	spin_unlock(&zhdr->page_lock);
}

218 219 220 221 222 223 224 225 226
/*
 * Encodes the handle of a particular buddy within a z3fold page
 * Pool lock should be held as this function accesses first_num
 */
static unsigned long encode_handle(struct z3fold_header *zhdr, enum buddy bud)
{
	unsigned long handle;

	handle = (unsigned long)zhdr;
227 228 229 230 231
	if (bud != HEADLESS) {
		handle |= (bud + zhdr->first_num) & BUDDY_MASK;
		if (bud == LAST)
			handle |= (zhdr->last_chunks << BUDDY_SHIFT);
	}
232 233 234 235 236 237 238 239 240
	return handle;
}

/* Returns the z3fold page where a given handle is stored */
static struct z3fold_header *handle_to_z3fold_header(unsigned long handle)
{
	return (struct z3fold_header *)(handle & PAGE_MASK);
}

241 242 243 244 245 246
/* only for LAST bud, returns zero otherwise */
static unsigned short handle_to_chunks(unsigned long handle)
{
	return (handle & ~PAGE_MASK) >> BUDDY_SHIFT;
}

247 248 249 250 251
/*
 * (handle & BUDDY_MASK) < zhdr->first_num is possible in encode_handle
 *  but that doesn't matter. because the masking will result in the
 *  correct buddy number.
 */
252 253 254 255 256 257
static enum buddy handle_to_buddy(unsigned long handle)
{
	struct z3fold_header *zhdr = handle_to_z3fold_header(handle);
	return (handle - zhdr->first_num) & BUDDY_MASK;
}

258 259 260 261 262 263 264
static void __release_z3fold_page(struct z3fold_header *zhdr, bool locked)
{
	struct page *page = virt_to_page(zhdr);
	struct z3fold_pool *pool = zhdr->pool;

	WARN_ON(!list_empty(&zhdr->buddy));
	set_bit(PAGE_STALE, &page->private);
265
	clear_bit(NEEDS_COMPACTING, &page->private);
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327
	spin_lock(&pool->lock);
	if (!list_empty(&page->lru))
		list_del(&page->lru);
	spin_unlock(&pool->lock);
	if (locked)
		z3fold_page_unlock(zhdr);
	spin_lock(&pool->stale_lock);
	list_add(&zhdr->buddy, &pool->stale);
	queue_work(pool->release_wq, &pool->work);
	spin_unlock(&pool->stale_lock);
}

static void __attribute__((__unused__))
			release_z3fold_page(struct kref *ref)
{
	struct z3fold_header *zhdr = container_of(ref, struct z3fold_header,
						refcount);
	__release_z3fold_page(zhdr, false);
}

static void release_z3fold_page_locked(struct kref *ref)
{
	struct z3fold_header *zhdr = container_of(ref, struct z3fold_header,
						refcount);
	WARN_ON(z3fold_page_trylock(zhdr));
	__release_z3fold_page(zhdr, true);
}

static void release_z3fold_page_locked_list(struct kref *ref)
{
	struct z3fold_header *zhdr = container_of(ref, struct z3fold_header,
					       refcount);
	spin_lock(&zhdr->pool->lock);
	list_del_init(&zhdr->buddy);
	spin_unlock(&zhdr->pool->lock);

	WARN_ON(z3fold_page_trylock(zhdr));
	__release_z3fold_page(zhdr, true);
}

static void free_pages_work(struct work_struct *w)
{
	struct z3fold_pool *pool = container_of(w, struct z3fold_pool, work);

	spin_lock(&pool->stale_lock);
	while (!list_empty(&pool->stale)) {
		struct z3fold_header *zhdr = list_first_entry(&pool->stale,
						struct z3fold_header, buddy);
		struct page *page = virt_to_page(zhdr);

		list_del(&zhdr->buddy);
		if (WARN_ON(!test_bit(PAGE_STALE, &page->private)))
			continue;
		spin_unlock(&pool->stale_lock);
		cancel_work_sync(&zhdr->work);
		free_z3fold_page(page);
		cond_resched();
		spin_lock(&pool->stale_lock);
	}
	spin_unlock(&pool->stale_lock);
}

328 329 330 331 332 333 334 335 336 337 338 339 340 341
/*
 * Returns the number of free chunks in a z3fold page.
 * NB: can't be used with HEADLESS pages.
 */
static int num_free_chunks(struct z3fold_header *zhdr)
{
	int nfree;
	/*
	 * If there is a middle object, pick up the bigger free space
	 * either before or after it. Otherwise just subtract the number
	 * of chunks occupied by the first and the last objects.
	 */
	if (zhdr->middle_chunks != 0) {
		int nfree_before = zhdr->first_chunks ?
342
			0 : zhdr->start_middle - ZHDR_CHUNKS;
343
		int nfree_after = zhdr->last_chunks ?
344 345
			0 : TOTAL_CHUNKS -
				(zhdr->start_middle + zhdr->middle_chunks);
346 347 348 349 350 351
		nfree = max(nfree_before, nfree_after);
	} else
		nfree = NCHUNKS - zhdr->first_chunks - zhdr->last_chunks;
	return nfree;
}

352 353 354 355 356 357 358 359 360
static inline void *mchunk_memmove(struct z3fold_header *zhdr,
				unsigned short dst_chunk)
{
	void *beg = zhdr;
	return memmove(beg + (dst_chunk << CHUNK_SHIFT),
		       beg + (zhdr->start_middle << CHUNK_SHIFT),
		       zhdr->middle_chunks << CHUNK_SHIFT);
}

361
#define BIG_CHUNK_GAP	3
362 363 364 365 366
/* Has to be called with lock held */
static int z3fold_compact_page(struct z3fold_header *zhdr)
{
	struct page *page = virt_to_page(zhdr);

367 368
	if (test_bit(MIDDLE_CHUNK_MAPPED, &page->private))
		return 0; /* can't move middle chunk, it's used */
369

370 371 372 373 374 375
	if (zhdr->middle_chunks == 0)
		return 0; /* nothing to compact */

	if (zhdr->first_chunks == 0 && zhdr->last_chunks == 0) {
		/* move to the beginning */
		mchunk_memmove(zhdr, ZHDR_CHUNKS);
376 377 378 379
		zhdr->first_chunks = zhdr->middle_chunks;
		zhdr->middle_chunks = 0;
		zhdr->start_middle = 0;
		zhdr->first_num++;
380
		return 1;
381
	}
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404

	/*
	 * moving data is expensive, so let's only do that if
	 * there's substantial gain (at least BIG_CHUNK_GAP chunks)
	 */
	if (zhdr->first_chunks != 0 && zhdr->last_chunks == 0 &&
	    zhdr->start_middle - (zhdr->first_chunks + ZHDR_CHUNKS) >=
			BIG_CHUNK_GAP) {
		mchunk_memmove(zhdr, zhdr->first_chunks + ZHDR_CHUNKS);
		zhdr->start_middle = zhdr->first_chunks + ZHDR_CHUNKS;
		return 1;
	} else if (zhdr->last_chunks != 0 && zhdr->first_chunks == 0 &&
		   TOTAL_CHUNKS - (zhdr->last_chunks + zhdr->start_middle
					+ zhdr->middle_chunks) >=
			BIG_CHUNK_GAP) {
		unsigned short new_start = TOTAL_CHUNKS - zhdr->last_chunks -
			zhdr->middle_chunks;
		mchunk_memmove(zhdr, new_start);
		zhdr->start_middle = new_start;
		return 1;
	}

	return 0;
405 406
}

407 408 409 410 411 412 413 414 415 416 417 418
static void do_compact_page(struct z3fold_header *zhdr, bool locked)
{
	struct z3fold_pool *pool = zhdr->pool;
	struct page *page;
	struct list_head *unbuddied;
	int fchunks;

	page = virt_to_page(zhdr);
	if (locked)
		WARN_ON(z3fold_page_trylock(zhdr));
	else
		z3fold_page_lock(zhdr);
419
	if (WARN_ON(!test_and_clear_bit(NEEDS_COMPACTING, &page->private))) {
420 421 422 423 424 425 426
		z3fold_page_unlock(zhdr);
		return;
	}
	spin_lock(&pool->lock);
	list_del_init(&zhdr->buddy);
	spin_unlock(&pool->lock);

427 428 429 430 431
	if (kref_put(&zhdr->refcount, release_z3fold_page_locked)) {
		atomic64_dec(&pool->pages_nr);
		return;
	}

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
	z3fold_compact_page(zhdr);
	unbuddied = get_cpu_ptr(pool->unbuddied);
	fchunks = num_free_chunks(zhdr);
	if (fchunks < NCHUNKS &&
	    (!zhdr->first_chunks || !zhdr->middle_chunks ||
			!zhdr->last_chunks)) {
		/* the page's not completely free and it's unbuddied */
		spin_lock(&pool->lock);
		list_add(&zhdr->buddy, &unbuddied[fchunks]);
		spin_unlock(&pool->lock);
		zhdr->cpu = smp_processor_id();
	}
	put_cpu_ptr(pool->unbuddied);
	z3fold_page_unlock(zhdr);
}

static void compact_page_work(struct work_struct *w)
{
	struct z3fold_header *zhdr = container_of(w, struct z3fold_header,
						work);

	do_compact_page(zhdr, false);
}


/*
 * API Functions
 */

/**
 * z3fold_create_pool() - create a new z3fold pool
 * @name:	pool name
 * @gfp:	gfp flags when allocating the z3fold pool structure
 * @ops:	user-defined operations for the z3fold pool
 *
 * Return: pointer to the new z3fold pool or NULL if the metadata allocation
 * failed.
 */
static struct z3fold_pool *z3fold_create_pool(const char *name, gfp_t gfp,
		const struct z3fold_ops *ops)
{
	struct z3fold_pool *pool = NULL;
	int i, cpu;

	pool = kzalloc(sizeof(struct z3fold_pool), gfp);
	if (!pool)
		goto out;
	spin_lock_init(&pool->lock);
	spin_lock_init(&pool->stale_lock);
	pool->unbuddied = __alloc_percpu(sizeof(struct list_head)*NCHUNKS, 2);
Xidong Wang's avatar
Xidong Wang committed
482 483
	if (!pool->unbuddied)
		goto out_pool;
484 485 486 487 488 489 490 491 492 493 494 495
	for_each_possible_cpu(cpu) {
		struct list_head *unbuddied =
				per_cpu_ptr(pool->unbuddied, cpu);
		for_each_unbuddied_list(i, 0)
			INIT_LIST_HEAD(&unbuddied[i]);
	}
	INIT_LIST_HEAD(&pool->lru);
	INIT_LIST_HEAD(&pool->stale);
	atomic64_set(&pool->pages_nr, 0);
	pool->name = name;
	pool->compact_wq = create_singlethread_workqueue(pool->name);
	if (!pool->compact_wq)
Xidong Wang's avatar
Xidong Wang committed
496
		goto out_unbuddied;
497 498 499 500 501 502 503 504 505
	pool->release_wq = create_singlethread_workqueue(pool->name);
	if (!pool->release_wq)
		goto out_wq;
	INIT_WORK(&pool->work, free_pages_work);
	pool->ops = ops;
	return pool;

out_wq:
	destroy_workqueue(pool->compact_wq);
Xidong Wang's avatar
Xidong Wang committed
506 507 508
out_unbuddied:
	free_percpu(pool->unbuddied);
out_pool:
509
	kfree(pool);
Xidong Wang's avatar
Xidong Wang committed
510
out:
511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
	return NULL;
}

/**
 * z3fold_destroy_pool() - destroys an existing z3fold pool
 * @pool:	the z3fold pool to be destroyed
 *
 * The pool should be emptied before this function is called.
 */
static void z3fold_destroy_pool(struct z3fold_pool *pool)
{
	destroy_workqueue(pool->release_wq);
	destroy_workqueue(pool->compact_wq);
	kfree(pool);
}

527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
/**
 * z3fold_alloc() - allocates a region of a given size
 * @pool:	z3fold pool from which to allocate
 * @size:	size in bytes of the desired allocation
 * @gfp:	gfp flags used if the pool needs to grow
 * @handle:	handle of the new allocation
 *
 * This function will attempt to find a free region in the pool large enough to
 * satisfy the allocation request.  A search of the unbuddied lists is
 * performed first. If no suitable free region is found, then a new page is
 * allocated and added to the pool to satisfy the request.
 *
 * gfp should not set __GFP_HIGHMEM as highmem pages cannot be used
 * as z3fold pool pages.
 *
 * Return: 0 if success and handle is set, otherwise -EINVAL if the size or
 * gfp arguments are invalid or -ENOMEM if the pool was unable to allocate
 * a new page.
 */
static int z3fold_alloc(struct z3fold_pool *pool, size_t size, gfp_t gfp,
			unsigned long *handle)
{
	int chunks = 0, i, freechunks;
	struct z3fold_header *zhdr = NULL;
551
	struct page *page = NULL;
552
	enum buddy bud;
553
	bool can_sleep = (gfp & __GFP_RECLAIM) == __GFP_RECLAIM;
554 555 556 557 558 559 560 561 562 563

	if (!size || (gfp & __GFP_HIGHMEM))
		return -EINVAL;

	if (size > PAGE_SIZE)
		return -ENOSPC;

	if (size > PAGE_SIZE - ZHDR_SIZE_ALIGNED - CHUNK_SIZE)
		bud = HEADLESS;
	else {
564
		struct list_head *unbuddied;
565 566
		chunks = size_to_chunks(size);

567
lookup:
568
		/* First, try to find an unbuddied z3fold page. */
569
		unbuddied = get_cpu_ptr(pool->unbuddied);
570
		for_each_unbuddied_list(i, chunks) {
571 572 573
			struct list_head *l = &unbuddied[i];

			zhdr = list_first_entry_or_null(READ_ONCE(l),
574
						struct z3fold_header, buddy);
575 576

			if (!zhdr)
577
				continue;
578 579 580 581 582 583 584 585 586 587

			/* Re-check under lock. */
			spin_lock(&pool->lock);
			l = &unbuddied[i];
			if (unlikely(zhdr != list_first_entry(READ_ONCE(l),
					struct z3fold_header, buddy)) ||
			    !z3fold_page_trylock(zhdr)) {
				spin_unlock(&pool->lock);
				put_cpu_ptr(pool->unbuddied);
				goto lookup;
588 589
			}
			list_del_init(&zhdr->buddy);
590
			zhdr->cpu = -1;
591 592 593
			spin_unlock(&pool->lock);

			page = virt_to_page(zhdr);
594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614
			if (test_bit(NEEDS_COMPACTING, &page->private)) {
				z3fold_page_unlock(zhdr);
				zhdr = NULL;
				put_cpu_ptr(pool->unbuddied);
				if (can_sleep)
					cond_resched();
				goto lookup;
			}

			/*
			 * this page could not be removed from its unbuddied
			 * list while pool lock was held, and then we've taken
			 * page lock so kref_put could not be called before
			 * we got here, so it's safe to just call kref_get()
			 */
			kref_get(&zhdr->refcount);
			break;
		}
		put_cpu_ptr(pool->unbuddied);

		if (zhdr) {
615 616 617
			if (zhdr->first_chunks == 0) {
				if (zhdr->middle_chunks != 0 &&
				    chunks >= zhdr->start_middle)
618
					bud = LAST;
619 620 621 622 623 624 625
				else
					bud = FIRST;
			} else if (zhdr->last_chunks == 0)
				bud = LAST;
			else if (zhdr->middle_chunks == 0)
				bud = MIDDLE;
			else {
626
				if (kref_put(&zhdr->refcount,
627
					     release_z3fold_page_locked))
628
					atomic64_dec(&pool->pages_nr);
629 630
				else
					z3fold_page_unlock(zhdr);
631 632
				pr_err("No free chunks in unbuddied\n");
				WARN_ON(1);
633
				goto lookup;
634
			}
635
			goto found;
636 637 638 639
		}
		bud = FIRST;
	}

640 641 642 643 644 645 646 647
	spin_lock(&pool->stale_lock);
	zhdr = list_first_entry_or_null(&pool->stale,
					struct z3fold_header, buddy);
	/*
	 * Before allocating a page, let's see if we can take one from the
	 * stale pages list. cancel_work_sync() can sleep so we must make
	 * sure it won't be called in case we're in atomic context.
	 */
648
	if (zhdr && (can_sleep || !work_pending(&zhdr->work))) {
649 650 651 652 653 654 655 656 657 658
		list_del(&zhdr->buddy);
		spin_unlock(&pool->stale_lock);
		if (can_sleep)
			cancel_work_sync(&zhdr->work);
		page = virt_to_page(zhdr);
	} else {
		spin_unlock(&pool->stale_lock);
		page = alloc_page(gfp);
	}

659 660
	if (!page)
		return -ENOMEM;
661

662
	atomic64_inc(&pool->pages_nr);
663
	zhdr = init_z3fold_page(page, pool);
664 665 666 667 668

	if (bud == HEADLESS) {
		set_bit(PAGE_HEADLESS, &page->private);
		goto headless;
	}
669
	z3fold_page_lock(zhdr);
670 671 672 673 674 675 676 677

found:
	if (bud == FIRST)
		zhdr->first_chunks = chunks;
	else if (bud == LAST)
		zhdr->last_chunks = chunks;
	else {
		zhdr->middle_chunks = chunks;
678
		zhdr->start_middle = zhdr->first_chunks + ZHDR_CHUNKS;
679 680 681 682
	}

	if (zhdr->first_chunks == 0 || zhdr->last_chunks == 0 ||
			zhdr->middle_chunks == 0) {
683 684
		struct list_head *unbuddied = get_cpu_ptr(pool->unbuddied);

685 686
		/* Add to unbuddied list */
		freechunks = num_free_chunks(zhdr);
687 688 689 690 691
		spin_lock(&pool->lock);
		list_add(&zhdr->buddy, &unbuddied[freechunks]);
		spin_unlock(&pool->lock);
		zhdr->cpu = smp_processor_id();
		put_cpu_ptr(pool->unbuddied);
692 693 694
	}

headless:
695
	spin_lock(&pool->lock);
696 697 698 699 700 701 702 703
	/* Add/move z3fold page to beginning of LRU */
	if (!list_empty(&page->lru))
		list_del(&page->lru);

	list_add(&page->lru, &pool->lru);

	*handle = encode_handle(zhdr, bud);
	spin_unlock(&pool->lock);
704 705
	if (bud != HEADLESS)
		z3fold_page_unlock(zhdr);
706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729

	return 0;
}

/**
 * z3fold_free() - frees the allocation associated with the given handle
 * @pool:	pool in which the allocation resided
 * @handle:	handle associated with the allocation returned by z3fold_alloc()
 *
 * In the case that the z3fold page in which the allocation resides is under
 * reclaim, as indicated by the PG_reclaim flag being set, this function
 * only sets the first|last_chunks to 0.  The page is actually freed
 * once both buddies are evicted (see z3fold_reclaim_page() below).
 */
static void z3fold_free(struct z3fold_pool *pool, unsigned long handle)
{
	struct z3fold_header *zhdr;
	struct page *page;
	enum buddy bud;

	zhdr = handle_to_z3fold_header(handle);
	page = virt_to_page(zhdr);

	if (test_bit(PAGE_HEADLESS, &page->private)) {
730 731 732 733 734 735 736 737 738 739 740
		/* if a headless page is under reclaim, just leave.
		 * NB: we use test_and_set_bit for a reason: if the bit
		 * has not been set before, we release this page
		 * immediately so we don't care about its value any more.
		 */
		if (!test_and_set_bit(PAGE_CLAIMED, &page->private)) {
			spin_lock(&pool->lock);
			list_del(&page->lru);
			spin_unlock(&pool->lock);
			free_z3fold_page(page);
			atomic64_dec(&pool->pages_nr);
741
		}
742
		return;
743 744
	}

745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
	/* Non-headless case */
	z3fold_page_lock(zhdr);
	bud = handle_to_buddy(handle);

	switch (bud) {
	case FIRST:
		zhdr->first_chunks = 0;
		break;
	case MIDDLE:
		zhdr->middle_chunks = 0;
		break;
	case LAST:
		zhdr->last_chunks = 0;
		break;
	default:
		pr_err("%s: unknown bud %d\n", __func__, bud);
		WARN_ON(1);
		z3fold_page_unlock(zhdr);
763 764 765 766 767 768 769
		return;
	}

	if (kref_put(&zhdr->refcount, release_z3fold_page_locked_list)) {
		atomic64_dec(&pool->pages_nr);
		return;
	}
770
	if (test_bit(PAGE_CLAIMED, &page->private)) {
771 772 773
		z3fold_page_unlock(zhdr);
		return;
	}
774
	if (test_and_set_bit(NEEDS_COMPACTING, &page->private)) {
775
		z3fold_page_unlock(zhdr);
776 777 778
		return;
	}
	if (zhdr->cpu < 0 || !cpu_online(zhdr->cpu)) {
779
		spin_lock(&pool->lock);
780
		list_del_init(&zhdr->buddy);
781
		spin_unlock(&pool->lock);
782
		zhdr->cpu = -1;
783
		kref_get(&zhdr->refcount);
784 785
		do_compact_page(zhdr, true);
		return;
786
	}
787
	kref_get(&zhdr->refcount);
788 789
	queue_work_on(zhdr->cpu, pool->compact_wq, &zhdr->work);
	z3fold_page_unlock(zhdr);
790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
}

/**
 * z3fold_reclaim_page() - evicts allocations from a pool page and frees it
 * @pool:	pool from which a page will attempt to be evicted
 * @retires:	number of pages on the LRU list for which eviction will
 *		be attempted before failing
 *
 * z3fold reclaim is different from normal system reclaim in that it is done
 * from the bottom, up. This is because only the bottom layer, z3fold, has
 * information on how the allocations are organized within each z3fold page.
 * This has the potential to create interesting locking situations between
 * z3fold and the user, however.
 *
 * To avoid these, this is how z3fold_reclaim_page() should be called:

 * The user detects a page should be reclaimed and calls z3fold_reclaim_page().
 * z3fold_reclaim_page() will remove a z3fold page from the pool LRU list and
 * call the user-defined eviction handler with the pool and handle as
 * arguments.
 *
 * If the handle can not be evicted, the eviction handler should return
 * non-zero. z3fold_reclaim_page() will add the z3fold page back to the
 * appropriate list and try the next z3fold page on the LRU up to
 * a user defined number of retries.
 *
 * If the handle is successfully evicted, the eviction handler should
 * return 0 _and_ should have called z3fold_free() on the handle. z3fold_free()
 * contains logic to delay freeing the page if the page is under reclaim,
 * as indicated by the setting of the PG_reclaim flag on the underlying page.
 *
 * If all buddies in the z3fold page are successfully evicted, then the
 * z3fold page can be freed.
 *
 * Returns: 0 if page is successfully freed, otherwise -EINVAL if there are
 * no pages to evict or an eviction handler is not registered, -EAGAIN if
 * the retry limit was hit.
 */
static int z3fold_reclaim_page(struct z3fold_pool *pool, unsigned int retries)
{
830 831 832 833
	int i, ret = 0;
	struct z3fold_header *zhdr = NULL;
	struct page *page = NULL;
	struct list_head *pos;
834 835 836
	unsigned long first_handle = 0, middle_handle = 0, last_handle = 0;

	spin_lock(&pool->lock);
837
	if (!pool->ops || !pool->ops->evict || retries == 0) {
838 839 840 841
		spin_unlock(&pool->lock);
		return -EINVAL;
	}
	for (i = 0; i < retries; i++) {
842 843 844 845
		if (list_empty(&pool->lru)) {
			spin_unlock(&pool->lock);
			return -EINVAL;
		}
846 847
		list_for_each_prev(pos, &pool->lru) {
			page = list_entry(pos, struct page, lru);
848 849 850 851 852 853 854 855

			/* this bit could have been set by free, in which case
			 * we pass over to the next page in the pool.
			 */
			if (test_and_set_bit(PAGE_CLAIMED, &page->private))
				continue;

			zhdr = page_address(page);
856 857 858
			if (test_bit(PAGE_HEADLESS, &page->private))
				break;

859 860
			if (!z3fold_page_trylock(zhdr)) {
				zhdr = NULL;
861
				continue; /* can't evict at this point */
862
			}
863 864 865
			kref_get(&zhdr->refcount);
			list_del_init(&zhdr->buddy);
			zhdr->cpu = -1;
866
			break;
867 868
		}

869 870 871
		if (!zhdr)
			break;

872
		list_del_init(&page->lru);
873
		spin_unlock(&pool->lock);
874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889

		if (!test_bit(PAGE_HEADLESS, &page->private)) {
			/*
			 * We need encode the handles before unlocking, since
			 * we can race with free that will set
			 * (first|last)_chunks to 0
			 */
			first_handle = 0;
			last_handle = 0;
			middle_handle = 0;
			if (zhdr->first_chunks)
				first_handle = encode_handle(zhdr, FIRST);
			if (zhdr->middle_chunks)
				middle_handle = encode_handle(zhdr, MIDDLE);
			if (zhdr->last_chunks)
				last_handle = encode_handle(zhdr, LAST);
890 891 892 893
			/*
			 * it's safe to unlock here because we hold a
			 * reference to this page
			 */
894
			z3fold_page_unlock(zhdr);
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916
		} else {
			first_handle = encode_handle(zhdr, HEADLESS);
			last_handle = middle_handle = 0;
		}

		/* Issue the eviction callback(s) */
		if (middle_handle) {
			ret = pool->ops->evict(pool, middle_handle);
			if (ret)
				goto next;
		}
		if (first_handle) {
			ret = pool->ops->evict(pool, first_handle);
			if (ret)
				goto next;
		}
		if (last_handle) {
			ret = pool->ops->evict(pool, last_handle);
			if (ret)
				goto next;
		}
next:
917 918 919
		if (test_bit(PAGE_HEADLESS, &page->private)) {
			if (ret == 0) {
				free_z3fold_page(page);
920
				atomic64_dec(&pool->pages_nr);
921 922
				return 0;
			}
923 924 925 926 927
			spin_lock(&pool->lock);
			list_add(&page->lru, &pool->lru);
			spin_unlock(&pool->lock);
		} else {
			z3fold_page_lock(zhdr);
928
			clear_bit(PAGE_CLAIMED, &page->private);
929 930 931 932 933 934 935 936 937 938 939 940
			if (kref_put(&zhdr->refcount,
					release_z3fold_page_locked)) {
				atomic64_dec(&pool->pages_nr);
				return 0;
			}
			/*
			 * if we are here, the page is still not completely
			 * free. Take the global pool lock then to be able
			 * to add it back to the lru list
			 */
			spin_lock(&pool->lock);
			list_add(&page->lru, &pool->lru);
941
			spin_unlock(&pool->lock);
942
			z3fold_page_unlock(zhdr);
943
		}
944

945 946
		/* We started off locked to we need to lock the pool back */
		spin_lock(&pool->lock);
947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
	}
	spin_unlock(&pool->lock);
	return -EAGAIN;
}

/**
 * z3fold_map() - maps the allocation associated with the given handle
 * @pool:	pool in which the allocation resides
 * @handle:	handle associated with the allocation to be mapped
 *
 * Extracts the buddy number from handle and constructs the pointer to the
 * correct starting chunk within the page.
 *
 * Returns: a pointer to the mapped allocation
 */
static void *z3fold_map(struct z3fold_pool *pool, unsigned long handle)
{
	struct z3fold_header *zhdr;
	struct page *page;
	void *addr;
	enum buddy buddy;

	zhdr = handle_to_z3fold_header(handle);
	addr = zhdr;
	page = virt_to_page(zhdr);

	if (test_bit(PAGE_HEADLESS, &page->private))
		goto out;

976
	z3fold_page_lock(zhdr);
977 978 979 980 981 982 983 984 985 986
	buddy = handle_to_buddy(handle);
	switch (buddy) {
	case FIRST:
		addr += ZHDR_SIZE_ALIGNED;
		break;
	case MIDDLE:
		addr += zhdr->start_middle << CHUNK_SHIFT;
		set_bit(MIDDLE_CHUNK_MAPPED, &page->private);
		break;
	case LAST:
987
		addr += PAGE_SIZE - (handle_to_chunks(handle) << CHUNK_SHIFT);
988 989 990 991 992 993 994
		break;
	default:
		pr_err("unknown buddy id %d\n", buddy);
		WARN_ON(1);
		addr = NULL;
		break;
	}
995 996

	z3fold_page_unlock(zhdr);
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
out:
	return addr;
}

/**
 * z3fold_unmap() - unmaps the allocation associated with the given handle
 * @pool:	pool in which the allocation resides
 * @handle:	handle associated with the allocation to be unmapped
 */
static void z3fold_unmap(struct z3fold_pool *pool, unsigned long handle)
{
	struct z3fold_header *zhdr;
	struct page *page;
	enum buddy buddy;

	zhdr = handle_to_z3fold_header(handle);
	page = virt_to_page(zhdr);

1015
	if (test_bit(PAGE_HEADLESS, &page->private))
1016 1017
		return;

1018
	z3fold_page_lock(zhdr);
1019 1020 1021
	buddy = handle_to_buddy(handle);
	if (buddy == MIDDLE)
		clear_bit(MIDDLE_CHUNK_MAPPED, &page->private);
1022
	z3fold_page_unlock(zhdr);
1023 1024 1025 1026 1027 1028
}

/**
 * z3fold_get_pool_size() - gets the z3fold pool size in pages
 * @pool:	pool whose size is being queried
 *
1029
 * Returns: size in pages of the given pool.
1030 1031 1032
 */
static u64 z3fold_get_pool_size(struct z3fold_pool *pool)
{
1033
	return atomic64_read(&pool->pages_nr);
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057
}

/*****************
 * zpool
 ****************/

static int z3fold_zpool_evict(struct z3fold_pool *pool, unsigned long handle)
{
	if (pool->zpool && pool->zpool_ops && pool->zpool_ops->evict)
		return pool->zpool_ops->evict(pool->zpool, handle);
	else
		return -ENOENT;
}

static const struct z3fold_ops z3fold_zpool_ops = {
	.evict =	z3fold_zpool_evict
};

static void *z3fold_zpool_create(const char *name, gfp_t gfp,
			       const struct zpool_ops *zpool_ops,
			       struct zpool *zpool)
{
	struct z3fold_pool *pool;

1058 1059
	pool = z3fold_create_pool(name, gfp,
				zpool_ops ? &z3fold_zpool_ops : NULL);
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
	if (pool) {
		pool->zpool = zpool;
		pool->zpool_ops = zpool_ops;
	}
	return pool;
}

static void z3fold_zpool_destroy(void *pool)
{
	z3fold_destroy_pool(pool);
}

static int z3fold_zpool_malloc(void *pool, size_t size, gfp_t gfp,
			unsigned long *handle)
{
	return z3fold_alloc(pool, size, gfp, handle);
}
static void z3fold_zpool_free(void *pool, unsigned long handle)
{
	z3fold_free(pool, handle);
}

static int z3fold_zpool_shrink(void *pool, unsigned int pages,
			unsigned int *reclaimed)
{
	unsigned int total = 0;
	int ret = -EINVAL;

	while (total < pages) {
		ret = z3fold_reclaim_page(pool, 8);
		if (ret < 0)
			break;
		total++;
	}

	if (reclaimed)
		*reclaimed = total;

	return ret;
}

static void *z3fold_zpool_map(void *pool, unsigned long handle,
			enum zpool_mapmode mm)
{
	return z3fold_map(pool, handle);
}
static void z3fold_zpool_unmap(void *pool, unsigned long handle)
{
	z3fold_unmap(pool, handle);
}

static u64 z3fold_zpool_total_size(void *pool)
{
	return z3fold_get_pool_size(pool) * PAGE_SIZE;
}

static struct zpool_driver z3fold_zpool_driver = {
	.type =		"z3fold",
	.owner =	THIS_MODULE,
	.create =	z3fold_zpool_create,
	.destroy =	z3fold_zpool_destroy,
	.malloc =	z3fold_zpool_malloc,
	.free =		z3fold_zpool_free,
	.shrink =	z3fold_zpool_shrink,
	.map =		z3fold_zpool_map,
	.unmap =	z3fold_zpool_unmap,
	.total_size =	z3fold_zpool_total_size,
};

MODULE_ALIAS("zpool-z3fold");

static int __init init_z3fold(void)
{
1133 1134
	/* Make sure the z3fold header is not larger than the page size */
	BUILD_BUG_ON(ZHDR_SIZE_ALIGNED > PAGE_SIZE);
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
	zpool_register_driver(&z3fold_zpool_driver);

	return 0;
}

static void __exit exit_z3fold(void)
{
	zpool_unregister_driver(&z3fold_zpool_driver);
}

module_init(init_z3fold);
module_exit(exit_z3fold);

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Vitaly Wool <vitalywool@gmail.com>");
MODULE_DESCRIPTION("3-Fold Allocator for Compressed Pages");