xref: /openbmc/linux/mm/zbud.c (revision d7955ce4)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * zbud.c
4  *
5  * Copyright (C) 2013, Seth Jennings, IBM
6  *
7  * Concepts based on zcache internal zbud allocator by Dan Magenheimer.
8  *
9  * zbud is an special purpose allocator for storing compressed pages.  Contrary
10  * to what its name may suggest, zbud is not a buddy allocator, but rather an
11  * allocator that "buddies" two compressed pages together in a single memory
12  * page.
13  *
14  * While this design limits storage density, it has simple and deterministic
15  * reclaim properties that make it preferable to a higher density approach when
16  * reclaim will be used.
17  *
18  * zbud works by storing compressed pages, or "zpages", together in pairs in a
19  * single memory page called a "zbud page".  The first buddy is "left
20  * justified" at the beginning of the zbud page, and the last buddy is "right
21  * justified" at the end of the zbud page.  The benefit is that if either
22  * buddy is freed, the freed buddy space, coalesced with whatever slack space
23  * that existed between the buddies, results in the largest possible free region
24  * within the zbud page.
25  *
26  * zbud also provides an attractive lower bound on density. The ratio of zpages
27  * to zbud pages can not be less than 1.  This ensures that zbud can never "do
28  * harm" by using more pages to store zpages than the uncompressed zpages would
29  * have used on their own.
30  *
31  * zbud pages are divided into "chunks".  The size of the chunks is fixed at
32  * compile time and determined by NCHUNKS_ORDER below.  Dividing zbud pages
33  * into chunks allows organizing unbuddied zbud pages into a manageable number
34  * of unbuddied lists according to the number of free chunks available in the
35  * zbud page.
36  *
37  * The zbud API differs from that of conventional allocators in that the
38  * allocation function, zbud_alloc(), returns an opaque handle to the user,
39  * not a dereferenceable pointer.  The user must map the handle using
40  * zbud_map() in order to get a usable pointer by which to access the
41  * allocation data and unmap the handle with zbud_unmap() when operations
42  * on the allocation data are complete.
43  */
44 
45 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
46 
47 #include <linux/atomic.h>
48 #include <linux/list.h>
49 #include <linux/mm.h>
50 #include <linux/module.h>
51 #include <linux/preempt.h>
52 #include <linux/slab.h>
53 #include <linux/spinlock.h>
54 #include <linux/zpool.h>
55 
56 /*****************
57  * Structures
58 *****************/
59 /*
60  * NCHUNKS_ORDER determines the internal allocation granularity, effectively
61  * adjusting internal fragmentation.  It also determines the number of
62  * freelists maintained in each pool. NCHUNKS_ORDER of 6 means that the
63  * allocation granularity will be in chunks of size PAGE_SIZE/64. As one chunk
64  * in allocated page is occupied by zbud header, NCHUNKS will be calculated to
65  * 63 which shows the max number of free chunks in zbud page, also there will be
66  * 63 freelists per pool.
67  */
68 #define NCHUNKS_ORDER	6
69 
70 #define CHUNK_SHIFT	(PAGE_SHIFT - NCHUNKS_ORDER)
71 #define CHUNK_SIZE	(1 << CHUNK_SHIFT)
72 #define ZHDR_SIZE_ALIGNED CHUNK_SIZE
73 #define NCHUNKS		((PAGE_SIZE - ZHDR_SIZE_ALIGNED) >> CHUNK_SHIFT)
74 
75 struct zbud_pool;
76 
77 /**
78  * struct zbud_pool - stores metadata for each zbud pool
79  * @lock:	protects all pool fields and first|last_chunk fields of any
80  *		zbud page in the pool
81  * @unbuddied:	array of lists tracking zbud pages that only contain one buddy;
82  *		the lists each zbud page is added to depends on the size of
83  *		its free region.
84  * @buddied:	list tracking the zbud pages that contain two buddies;
85  *		these zbud pages are full
86  * @pages_nr:	number of zbud pages in the pool.
87  *
88  * This structure is allocated at pool creation time and maintains metadata
89  * pertaining to a particular zbud pool.
90  */
91 struct zbud_pool {
92 	spinlock_t lock;
93 	union {
94 		/*
95 		 * Reuse unbuddied[0] as buddied on the ground that
96 		 * unbuddied[0] is unused.
97 		 */
98 		struct list_head buddied;
99 		struct list_head unbuddied[NCHUNKS];
100 	};
101 	u64 pages_nr;
102 };
103 
104 /*
105  * struct zbud_header - zbud page metadata occupying the first chunk of each
106  *			zbud page.
107  * @buddy:	links the zbud page into the unbuddied/buddied lists in the pool
108  * @first_chunks:	the size of the first buddy in chunks, 0 if free
109  * @last_chunks:	the size of the last buddy in chunks, 0 if free
110  */
111 struct zbud_header {
112 	struct list_head buddy;
113 	unsigned int first_chunks;
114 	unsigned int last_chunks;
115 };
116 
117 /*****************
118  * Helpers
119 *****************/
120 /* Just to make the code easier to read */
121 enum buddy {
122 	FIRST,
123 	LAST
124 };
125 
126 /* Converts an allocation size in bytes to size in zbud chunks */
127 static int size_to_chunks(size_t size)
128 {
129 	return (size + CHUNK_SIZE - 1) >> CHUNK_SHIFT;
130 }
131 
132 #define for_each_unbuddied_list(_iter, _begin) \
133 	for ((_iter) = (_begin); (_iter) < NCHUNKS; (_iter)++)
134 
135 /* Initializes the zbud header of a newly allocated zbud page */
136 static struct zbud_header *init_zbud_page(struct page *page)
137 {
138 	struct zbud_header *zhdr = page_address(page);
139 	zhdr->first_chunks = 0;
140 	zhdr->last_chunks = 0;
141 	INIT_LIST_HEAD(&zhdr->buddy);
142 	return zhdr;
143 }
144 
145 /* Resets the struct page fields and frees the page */
146 static void free_zbud_page(struct zbud_header *zhdr)
147 {
148 	__free_page(virt_to_page(zhdr));
149 }
150 
151 /*
152  * Encodes the handle of a particular buddy within a zbud page
153  * Pool lock should be held as this function accesses first|last_chunks
154  */
155 static unsigned long encode_handle(struct zbud_header *zhdr, enum buddy bud)
156 {
157 	unsigned long handle;
158 
159 	/*
160 	 * For now, the encoded handle is actually just the pointer to the data
161 	 * but this might not always be the case.  A little information hiding.
162 	 * Add CHUNK_SIZE to the handle if it is the first allocation to jump
163 	 * over the zbud header in the first chunk.
164 	 */
165 	handle = (unsigned long)zhdr;
166 	if (bud == FIRST)
167 		/* skip over zbud header */
168 		handle += ZHDR_SIZE_ALIGNED;
169 	else /* bud == LAST */
170 		handle += PAGE_SIZE - (zhdr->last_chunks  << CHUNK_SHIFT);
171 	return handle;
172 }
173 
174 /* Returns the zbud page where a given handle is stored */
175 static struct zbud_header *handle_to_zbud_header(unsigned long handle)
176 {
177 	return (struct zbud_header *)(handle & PAGE_MASK);
178 }
179 
180 /* Returns the number of free chunks in a zbud page */
181 static int num_free_chunks(struct zbud_header *zhdr)
182 {
183 	/*
184 	 * Rather than branch for different situations, just use the fact that
185 	 * free buddies have a length of zero to simplify everything.
186 	 */
187 	return NCHUNKS - zhdr->first_chunks - zhdr->last_chunks;
188 }
189 
190 /*****************
191  * API Functions
192 *****************/
193 /**
194  * zbud_create_pool() - create a new zbud pool
195  * @gfp:	gfp flags when allocating the zbud pool structure
196  *
197  * Return: pointer to the new zbud pool or NULL if the metadata allocation
198  * failed.
199  */
200 static struct zbud_pool *zbud_create_pool(gfp_t gfp)
201 {
202 	struct zbud_pool *pool;
203 	int i;
204 
205 	pool = kzalloc(sizeof(struct zbud_pool), gfp);
206 	if (!pool)
207 		return NULL;
208 	spin_lock_init(&pool->lock);
209 	for_each_unbuddied_list(i, 0)
210 		INIT_LIST_HEAD(&pool->unbuddied[i]);
211 	INIT_LIST_HEAD(&pool->buddied);
212 	pool->pages_nr = 0;
213 	return pool;
214 }
215 
216 /**
217  * zbud_destroy_pool() - destroys an existing zbud pool
218  * @pool:	the zbud pool to be destroyed
219  *
220  * The pool should be emptied before this function is called.
221  */
222 static void zbud_destroy_pool(struct zbud_pool *pool)
223 {
224 	kfree(pool);
225 }
226 
227 /**
228  * zbud_alloc() - allocates a region of a given size
229  * @pool:	zbud pool from which to allocate
230  * @size:	size in bytes of the desired allocation
231  * @gfp:	gfp flags used if the pool needs to grow
232  * @handle:	handle of the new allocation
233  *
234  * This function will attempt to find a free region in the pool large enough to
235  * satisfy the allocation request.  A search of the unbuddied lists is
236  * performed first. If no suitable free region is found, then a new page is
237  * allocated and added to the pool to satisfy the request.
238  *
239  * gfp should not set __GFP_HIGHMEM as highmem pages cannot be used
240  * as zbud pool pages.
241  *
242  * Return: 0 if success and handle is set, otherwise -EINVAL if the size or
243  * gfp arguments are invalid or -ENOMEM if the pool was unable to allocate
244  * a new page.
245  */
246 static int zbud_alloc(struct zbud_pool *pool, size_t size, gfp_t gfp,
247 			unsigned long *handle)
248 {
249 	int chunks, i, freechunks;
250 	struct zbud_header *zhdr = NULL;
251 	enum buddy bud;
252 	struct page *page;
253 
254 	if (!size || (gfp & __GFP_HIGHMEM))
255 		return -EINVAL;
256 	if (size > PAGE_SIZE - ZHDR_SIZE_ALIGNED - CHUNK_SIZE)
257 		return -ENOSPC;
258 	chunks = size_to_chunks(size);
259 	spin_lock(&pool->lock);
260 
261 	/* First, try to find an unbuddied zbud page. */
262 	for_each_unbuddied_list(i, chunks) {
263 		if (!list_empty(&pool->unbuddied[i])) {
264 			zhdr = list_first_entry(&pool->unbuddied[i],
265 					struct zbud_header, buddy);
266 			list_del(&zhdr->buddy);
267 			if (zhdr->first_chunks == 0)
268 				bud = FIRST;
269 			else
270 				bud = LAST;
271 			goto found;
272 		}
273 	}
274 
275 	/* Couldn't find unbuddied zbud page, create new one */
276 	spin_unlock(&pool->lock);
277 	page = alloc_page(gfp);
278 	if (!page)
279 		return -ENOMEM;
280 	spin_lock(&pool->lock);
281 	pool->pages_nr++;
282 	zhdr = init_zbud_page(page);
283 	bud = FIRST;
284 
285 found:
286 	if (bud == FIRST)
287 		zhdr->first_chunks = chunks;
288 	else
289 		zhdr->last_chunks = chunks;
290 
291 	if (zhdr->first_chunks == 0 || zhdr->last_chunks == 0) {
292 		/* Add to unbuddied list */
293 		freechunks = num_free_chunks(zhdr);
294 		list_add(&zhdr->buddy, &pool->unbuddied[freechunks]);
295 	} else {
296 		/* Add to buddied list */
297 		list_add(&zhdr->buddy, &pool->buddied);
298 	}
299 
300 	*handle = encode_handle(zhdr, bud);
301 	spin_unlock(&pool->lock);
302 
303 	return 0;
304 }
305 
306 /**
307  * zbud_free() - frees the allocation associated with the given handle
308  * @pool:	pool in which the allocation resided
309  * @handle:	handle associated with the allocation returned by zbud_alloc()
310  */
311 static void zbud_free(struct zbud_pool *pool, unsigned long handle)
312 {
313 	struct zbud_header *zhdr;
314 	int freechunks;
315 
316 	spin_lock(&pool->lock);
317 	zhdr = handle_to_zbud_header(handle);
318 
319 	/* If first buddy, handle will be page aligned */
320 	if ((handle - ZHDR_SIZE_ALIGNED) & ~PAGE_MASK)
321 		zhdr->last_chunks = 0;
322 	else
323 		zhdr->first_chunks = 0;
324 
325 	/* Remove from existing buddy list */
326 	list_del(&zhdr->buddy);
327 
328 	if (zhdr->first_chunks == 0 && zhdr->last_chunks == 0) {
329 		/* zbud page is empty, free */
330 		free_zbud_page(zhdr);
331 		pool->pages_nr--;
332 	} else {
333 		/* Add to unbuddied list */
334 		freechunks = num_free_chunks(zhdr);
335 		list_add(&zhdr->buddy, &pool->unbuddied[freechunks]);
336 	}
337 
338 	spin_unlock(&pool->lock);
339 }
340 
341 /**
342  * zbud_map() - maps the allocation associated with the given handle
343  * @pool:	pool in which the allocation resides
344  * @handle:	handle associated with the allocation to be mapped
345  *
346  * While trivial for zbud, the mapping functions for others allocators
347  * implementing this allocation API could have more complex information encoded
348  * in the handle and could create temporary mappings to make the data
349  * accessible to the user.
350  *
351  * Returns: a pointer to the mapped allocation
352  */
353 static void *zbud_map(struct zbud_pool *pool, unsigned long handle)
354 {
355 	return (void *)(handle);
356 }
357 
358 /**
359  * zbud_unmap() - maps the allocation associated with the given handle
360  * @pool:	pool in which the allocation resides
361  * @handle:	handle associated with the allocation to be unmapped
362  */
363 static void zbud_unmap(struct zbud_pool *pool, unsigned long handle)
364 {
365 }
366 
367 /**
368  * zbud_get_pool_size() - gets the zbud pool size in pages
369  * @pool:	pool whose size is being queried
370  *
371  * Returns: size in pages of the given pool.  The pool lock need not be
372  * taken to access pages_nr.
373  */
374 static u64 zbud_get_pool_size(struct zbud_pool *pool)
375 {
376 	return pool->pages_nr;
377 }
378 
379 /*****************
380  * zpool
381  ****************/
382 
383 static void *zbud_zpool_create(const char *name, gfp_t gfp)
384 {
385 	return zbud_create_pool(gfp);
386 }
387 
388 static void zbud_zpool_destroy(void *pool)
389 {
390 	zbud_destroy_pool(pool);
391 }
392 
393 static int zbud_zpool_malloc(void *pool, size_t size, gfp_t gfp,
394 			unsigned long *handle)
395 {
396 	return zbud_alloc(pool, size, gfp, handle);
397 }
398 static void zbud_zpool_free(void *pool, unsigned long handle)
399 {
400 	zbud_free(pool, handle);
401 }
402 
403 static void *zbud_zpool_map(void *pool, unsigned long handle,
404 			enum zpool_mapmode mm)
405 {
406 	return zbud_map(pool, handle);
407 }
408 static void zbud_zpool_unmap(void *pool, unsigned long handle)
409 {
410 	zbud_unmap(pool, handle);
411 }
412 
413 static u64 zbud_zpool_total_size(void *pool)
414 {
415 	return zbud_get_pool_size(pool) * PAGE_SIZE;
416 }
417 
418 static struct zpool_driver zbud_zpool_driver = {
419 	.type =		"zbud",
420 	.sleep_mapped = true,
421 	.owner =	THIS_MODULE,
422 	.create =	zbud_zpool_create,
423 	.destroy =	zbud_zpool_destroy,
424 	.malloc =	zbud_zpool_malloc,
425 	.free =		zbud_zpool_free,
426 	.map =		zbud_zpool_map,
427 	.unmap =	zbud_zpool_unmap,
428 	.total_size =	zbud_zpool_total_size,
429 };
430 
431 MODULE_ALIAS("zpool-zbud");
432 
433 static int __init init_zbud(void)
434 {
435 	/* Make sure the zbud header will fit in one chunk */
436 	BUILD_BUG_ON(sizeof(struct zbud_header) > ZHDR_SIZE_ALIGNED);
437 	pr_info("loaded\n");
438 
439 	zpool_register_driver(&zbud_zpool_driver);
440 
441 	return 0;
442 }
443 
444 static void __exit exit_zbud(void)
445 {
446 	zpool_unregister_driver(&zbud_zpool_driver);
447 	pr_info("unloaded\n");
448 }
449 
450 module_init(init_zbud);
451 module_exit(exit_zbud);
452 
453 MODULE_LICENSE("GPL");
454 MODULE_AUTHOR("Seth Jennings <sjennings@variantweb.net>");
455 MODULE_DESCRIPTION("Buddy Allocator for Compressed Pages");
456