]>
Commit | Line | Data |
---|---|---|
306b0c95 | 1 | /* |
f1e3cfff | 2 | * Compressed RAM block device |
306b0c95 | 3 | * |
1130ebba | 4 | * Copyright (C) 2008, 2009, 2010 Nitin Gupta |
7bfb3de8 | 5 | * 2012, 2013 Minchan Kim |
306b0c95 NG |
6 | * |
7 | * This code is released using a dual license strategy: BSD/GPL | |
8 | * You can choose the licence that better fits your requirements. | |
9 | * | |
10 | * Released under the terms of 3-clause BSD License | |
11 | * Released under the terms of GNU General Public License Version 2.0 | |
12 | * | |
306b0c95 NG |
13 | */ |
14 | ||
f1e3cfff | 15 | #define KMSG_COMPONENT "zram" |
306b0c95 NG |
16 | #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt |
17 | ||
b1f5b81e RJ |
18 | #ifdef CONFIG_ZRAM_DEBUG |
19 | #define DEBUG | |
20 | #endif | |
21 | ||
306b0c95 NG |
22 | #include <linux/module.h> |
23 | #include <linux/kernel.h> | |
8946a086 | 24 | #include <linux/bio.h> |
306b0c95 NG |
25 | #include <linux/bitops.h> |
26 | #include <linux/blkdev.h> | |
27 | #include <linux/buffer_head.h> | |
28 | #include <linux/device.h> | |
29 | #include <linux/genhd.h> | |
30 | #include <linux/highmem.h> | |
5a0e3ad6 | 31 | #include <linux/slab.h> |
306b0c95 | 32 | #include <linux/string.h> |
306b0c95 | 33 | #include <linux/vmalloc.h> |
306b0c95 | 34 | |
16a4bfb9 | 35 | #include "zram_drv.h" |
306b0c95 NG |
36 | |
37 | /* Globals */ | |
f1e3cfff | 38 | static int zram_major; |
0f0e3ba3 | 39 | static struct zram *zram_devices; |
b7ca232e | 40 | static const char *default_compressor = "lzo"; |
306b0c95 | 41 | |
306b0c95 | 42 | /* Module params (documentation at end) */ |
ca3d70bd | 43 | static unsigned int num_devices = 1; |
33863c21 | 44 | |
a68eb3b6 SS |
45 | #define ZRAM_ATTR_RO(name) \ |
46 | static ssize_t zram_attr_##name##_show(struct device *d, \ | |
47 | struct device_attribute *attr, char *b) \ | |
48 | { \ | |
49 | struct zram *zram = dev_to_zram(d); \ | |
50 | return sprintf(b, "%llu\n", \ | |
51 | (u64)atomic64_read(&zram->stats.name)); \ | |
52 | } \ | |
53 | static struct device_attribute dev_attr_##name = \ | |
54 | __ATTR(name, S_IRUGO, zram_attr_##name##_show, NULL); | |
55 | ||
be2d1d56 SS |
56 | static inline int init_done(struct zram *zram) |
57 | { | |
58 | return zram->meta != NULL; | |
59 | } | |
60 | ||
9b3bb7ab SS |
61 | static inline struct zram *dev_to_zram(struct device *dev) |
62 | { | |
63 | return (struct zram *)dev_to_disk(dev)->private_data; | |
64 | } | |
65 | ||
66 | static ssize_t disksize_show(struct device *dev, | |
67 | struct device_attribute *attr, char *buf) | |
68 | { | |
69 | struct zram *zram = dev_to_zram(dev); | |
70 | ||
71 | return sprintf(buf, "%llu\n", zram->disksize); | |
72 | } | |
73 | ||
74 | static ssize_t initstate_show(struct device *dev, | |
75 | struct device_attribute *attr, char *buf) | |
76 | { | |
a68eb3b6 | 77 | u32 val; |
9b3bb7ab SS |
78 | struct zram *zram = dev_to_zram(dev); |
79 | ||
a68eb3b6 SS |
80 | down_read(&zram->init_lock); |
81 | val = init_done(zram); | |
82 | up_read(&zram->init_lock); | |
9b3bb7ab | 83 | |
a68eb3b6 | 84 | return sprintf(buf, "%u\n", val); |
9b3bb7ab SS |
85 | } |
86 | ||
87 | static ssize_t orig_data_size_show(struct device *dev, | |
88 | struct device_attribute *attr, char *buf) | |
89 | { | |
90 | struct zram *zram = dev_to_zram(dev); | |
91 | ||
92 | return sprintf(buf, "%llu\n", | |
90a7806e | 93 | (u64)(atomic64_read(&zram->stats.pages_stored)) << PAGE_SHIFT); |
9b3bb7ab SS |
94 | } |
95 | ||
9b3bb7ab SS |
96 | static ssize_t mem_used_total_show(struct device *dev, |
97 | struct device_attribute *attr, char *buf) | |
98 | { | |
99 | u64 val = 0; | |
100 | struct zram *zram = dev_to_zram(dev); | |
101 | struct zram_meta *meta = zram->meta; | |
102 | ||
103 | down_read(&zram->init_lock); | |
be2d1d56 | 104 | if (init_done(zram)) |
9b3bb7ab SS |
105 | val = zs_get_total_size_bytes(meta->mem_pool); |
106 | up_read(&zram->init_lock); | |
107 | ||
108 | return sprintf(buf, "%llu\n", val); | |
109 | } | |
110 | ||
92967471 | 111 | /* flag operations needs meta->tb_lock */ |
8b3cc3ed | 112 | static int zram_test_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 113 | enum zram_pageflags flag) |
306b0c95 | 114 | { |
8b3cc3ed | 115 | return meta->table[index].flags & BIT(flag); |
306b0c95 NG |
116 | } |
117 | ||
8b3cc3ed | 118 | static void zram_set_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 119 | enum zram_pageflags flag) |
306b0c95 | 120 | { |
8b3cc3ed | 121 | meta->table[index].flags |= BIT(flag); |
306b0c95 NG |
122 | } |
123 | ||
8b3cc3ed | 124 | static void zram_clear_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 125 | enum zram_pageflags flag) |
306b0c95 | 126 | { |
8b3cc3ed | 127 | meta->table[index].flags &= ~BIT(flag); |
306b0c95 NG |
128 | } |
129 | ||
9b3bb7ab SS |
130 | static inline int is_partial_io(struct bio_vec *bvec) |
131 | { | |
132 | return bvec->bv_len != PAGE_SIZE; | |
133 | } | |
134 | ||
135 | /* | |
136 | * Check if request is within bounds and aligned on zram logical blocks. | |
137 | */ | |
138 | static inline int valid_io_request(struct zram *zram, struct bio *bio) | |
139 | { | |
140 | u64 start, end, bound; | |
a539c72a | 141 | |
9b3bb7ab | 142 | /* unaligned request */ |
4f024f37 KO |
143 | if (unlikely(bio->bi_iter.bi_sector & |
144 | (ZRAM_SECTOR_PER_LOGICAL_BLOCK - 1))) | |
9b3bb7ab | 145 | return 0; |
4f024f37 | 146 | if (unlikely(bio->bi_iter.bi_size & (ZRAM_LOGICAL_BLOCK_SIZE - 1))) |
9b3bb7ab SS |
147 | return 0; |
148 | ||
4f024f37 KO |
149 | start = bio->bi_iter.bi_sector; |
150 | end = start + (bio->bi_iter.bi_size >> SECTOR_SHIFT); | |
9b3bb7ab SS |
151 | bound = zram->disksize >> SECTOR_SHIFT; |
152 | /* out of range range */ | |
75c7caf5 | 153 | if (unlikely(start >= bound || end > bound || start > end)) |
9b3bb7ab SS |
154 | return 0; |
155 | ||
156 | /* I/O request is valid */ | |
157 | return 1; | |
158 | } | |
159 | ||
160 | static void zram_meta_free(struct zram_meta *meta) | |
161 | { | |
162 | zs_destroy_pool(meta->mem_pool); | |
9b3bb7ab SS |
163 | vfree(meta->table); |
164 | kfree(meta); | |
165 | } | |
166 | ||
167 | static struct zram_meta *zram_meta_alloc(u64 disksize) | |
168 | { | |
169 | size_t num_pages; | |
170 | struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL); | |
171 | if (!meta) | |
172 | goto out; | |
173 | ||
9b3bb7ab SS |
174 | num_pages = disksize >> PAGE_SHIFT; |
175 | meta->table = vzalloc(num_pages * sizeof(*meta->table)); | |
176 | if (!meta->table) { | |
177 | pr_err("Error allocating zram address table\n"); | |
b7ca232e | 178 | goto free_meta; |
9b3bb7ab SS |
179 | } |
180 | ||
181 | meta->mem_pool = zs_create_pool(GFP_NOIO | __GFP_HIGHMEM); | |
182 | if (!meta->mem_pool) { | |
183 | pr_err("Error creating memory pool\n"); | |
184 | goto free_table; | |
185 | } | |
186 | ||
92967471 | 187 | rwlock_init(&meta->tb_lock); |
9b3bb7ab SS |
188 | return meta; |
189 | ||
190 | free_table: | |
191 | vfree(meta->table); | |
9b3bb7ab SS |
192 | free_meta: |
193 | kfree(meta); | |
194 | meta = NULL; | |
195 | out: | |
196 | return meta; | |
197 | } | |
198 | ||
199 | static void update_position(u32 *index, int *offset, struct bio_vec *bvec) | |
200 | { | |
201 | if (*offset + bvec->bv_len >= PAGE_SIZE) | |
202 | (*index)++; | |
203 | *offset = (*offset + bvec->bv_len) % PAGE_SIZE; | |
204 | } | |
205 | ||
306b0c95 NG |
206 | static int page_zero_filled(void *ptr) |
207 | { | |
208 | unsigned int pos; | |
209 | unsigned long *page; | |
210 | ||
211 | page = (unsigned long *)ptr; | |
212 | ||
213 | for (pos = 0; pos != PAGE_SIZE / sizeof(*page); pos++) { | |
214 | if (page[pos]) | |
215 | return 0; | |
216 | } | |
217 | ||
218 | return 1; | |
219 | } | |
220 | ||
9b3bb7ab SS |
221 | static void handle_zero_page(struct bio_vec *bvec) |
222 | { | |
223 | struct page *page = bvec->bv_page; | |
224 | void *user_mem; | |
225 | ||
226 | user_mem = kmap_atomic(page); | |
227 | if (is_partial_io(bvec)) | |
228 | memset(user_mem + bvec->bv_offset, 0, bvec->bv_len); | |
229 | else | |
230 | clear_page(user_mem); | |
231 | kunmap_atomic(user_mem); | |
232 | ||
233 | flush_dcache_page(page); | |
234 | } | |
235 | ||
92967471 | 236 | /* NOTE: caller should hold meta->tb_lock with write-side */ |
f1e3cfff | 237 | static void zram_free_page(struct zram *zram, size_t index) |
306b0c95 | 238 | { |
8b3cc3ed MK |
239 | struct zram_meta *meta = zram->meta; |
240 | unsigned long handle = meta->table[index].handle; | |
306b0c95 | 241 | |
fd1a30de | 242 | if (unlikely(!handle)) { |
2e882281 NG |
243 | /* |
244 | * No memory is allocated for zero filled pages. | |
245 | * Simply clear zero page flag. | |
246 | */ | |
8b3cc3ed MK |
247 | if (zram_test_flag(meta, index, ZRAM_ZERO)) { |
248 | zram_clear_flag(meta, index, ZRAM_ZERO); | |
90a7806e | 249 | atomic64_dec(&zram->stats.zero_pages); |
306b0c95 NG |
250 | } |
251 | return; | |
252 | } | |
253 | ||
8b3cc3ed | 254 | zs_free(meta->mem_pool, handle); |
306b0c95 | 255 | |
90a7806e SS |
256 | atomic64_sub(meta->table[index].size, &zram->stats.compr_data_size); |
257 | atomic64_dec(&zram->stats.pages_stored); | |
306b0c95 | 258 | |
8b3cc3ed MK |
259 | meta->table[index].handle = 0; |
260 | meta->table[index].size = 0; | |
306b0c95 NG |
261 | } |
262 | ||
37b51fdd | 263 | static int zram_decompress_page(struct zram *zram, char *mem, u32 index) |
306b0c95 | 264 | { |
b7ca232e | 265 | int ret = 0; |
37b51fdd | 266 | unsigned char *cmem; |
8b3cc3ed | 267 | struct zram_meta *meta = zram->meta; |
92967471 MK |
268 | unsigned long handle; |
269 | u16 size; | |
270 | ||
271 | read_lock(&meta->tb_lock); | |
272 | handle = meta->table[index].handle; | |
273 | size = meta->table[index].size; | |
306b0c95 | 274 | |
8b3cc3ed | 275 | if (!handle || zram_test_flag(meta, index, ZRAM_ZERO)) { |
92967471 | 276 | read_unlock(&meta->tb_lock); |
42e99bd9 | 277 | clear_page(mem); |
8c921b2b JM |
278 | return 0; |
279 | } | |
306b0c95 | 280 | |
8b3cc3ed | 281 | cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_RO); |
92967471 | 282 | if (size == PAGE_SIZE) |
42e99bd9 | 283 | copy_page(mem, cmem); |
37b51fdd | 284 | else |
b7ca232e | 285 | ret = zcomp_decompress(zram->comp, cmem, size, mem); |
8b3cc3ed | 286 | zs_unmap_object(meta->mem_pool, handle); |
92967471 | 287 | read_unlock(&meta->tb_lock); |
a1dd52af | 288 | |
8c921b2b | 289 | /* Should NEVER happen. Return bio error if it does. */ |
b7ca232e | 290 | if (unlikely(ret)) { |
8c921b2b | 291 | pr_err("Decompression failed! err=%d, page=%u\n", ret, index); |
da5cc7d3 | 292 | atomic64_inc(&zram->stats.failed_reads); |
8c921b2b | 293 | return ret; |
a1dd52af | 294 | } |
306b0c95 | 295 | |
8c921b2b | 296 | return 0; |
306b0c95 NG |
297 | } |
298 | ||
37b51fdd SS |
299 | static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec, |
300 | u32 index, int offset, struct bio *bio) | |
924bd88d JM |
301 | { |
302 | int ret; | |
37b51fdd SS |
303 | struct page *page; |
304 | unsigned char *user_mem, *uncmem = NULL; | |
8b3cc3ed | 305 | struct zram_meta *meta = zram->meta; |
37b51fdd SS |
306 | page = bvec->bv_page; |
307 | ||
92967471 | 308 | read_lock(&meta->tb_lock); |
8b3cc3ed MK |
309 | if (unlikely(!meta->table[index].handle) || |
310 | zram_test_flag(meta, index, ZRAM_ZERO)) { | |
92967471 | 311 | read_unlock(&meta->tb_lock); |
37b51fdd | 312 | handle_zero_page(bvec); |
924bd88d JM |
313 | return 0; |
314 | } | |
92967471 | 315 | read_unlock(&meta->tb_lock); |
924bd88d | 316 | |
37b51fdd SS |
317 | if (is_partial_io(bvec)) |
318 | /* Use a temporary buffer to decompress the page */ | |
7e5a5104 MK |
319 | uncmem = kmalloc(PAGE_SIZE, GFP_NOIO); |
320 | ||
321 | user_mem = kmap_atomic(page); | |
322 | if (!is_partial_io(bvec)) | |
37b51fdd SS |
323 | uncmem = user_mem; |
324 | ||
325 | if (!uncmem) { | |
326 | pr_info("Unable to allocate temp memory\n"); | |
327 | ret = -ENOMEM; | |
328 | goto out_cleanup; | |
329 | } | |
924bd88d | 330 | |
37b51fdd | 331 | ret = zram_decompress_page(zram, uncmem, index); |
924bd88d | 332 | /* Should NEVER happen. Return bio error if it does. */ |
b7ca232e | 333 | if (unlikely(ret)) |
37b51fdd | 334 | goto out_cleanup; |
924bd88d | 335 | |
37b51fdd SS |
336 | if (is_partial_io(bvec)) |
337 | memcpy(user_mem + bvec->bv_offset, uncmem + offset, | |
338 | bvec->bv_len); | |
339 | ||
340 | flush_dcache_page(page); | |
341 | ret = 0; | |
342 | out_cleanup: | |
343 | kunmap_atomic(user_mem); | |
344 | if (is_partial_io(bvec)) | |
345 | kfree(uncmem); | |
346 | return ret; | |
924bd88d JM |
347 | } |
348 | ||
349 | static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index, | |
350 | int offset) | |
306b0c95 | 351 | { |
397c6066 | 352 | int ret = 0; |
8c921b2b | 353 | size_t clen; |
c2344348 | 354 | unsigned long handle; |
130f315a | 355 | struct page *page; |
924bd88d | 356 | unsigned char *user_mem, *cmem, *src, *uncmem = NULL; |
8b3cc3ed | 357 | struct zram_meta *meta = zram->meta; |
b7ca232e | 358 | struct zcomp_strm *zstrm; |
e46e3315 | 359 | bool locked = false; |
306b0c95 | 360 | |
8c921b2b | 361 | page = bvec->bv_page; |
924bd88d JM |
362 | if (is_partial_io(bvec)) { |
363 | /* | |
364 | * This is a partial IO. We need to read the full page | |
365 | * before to write the changes. | |
366 | */ | |
7e5a5104 | 367 | uncmem = kmalloc(PAGE_SIZE, GFP_NOIO); |
924bd88d | 368 | if (!uncmem) { |
924bd88d JM |
369 | ret = -ENOMEM; |
370 | goto out; | |
371 | } | |
37b51fdd | 372 | ret = zram_decompress_page(zram, uncmem, index); |
397c6066 | 373 | if (ret) |
924bd88d | 374 | goto out; |
924bd88d JM |
375 | } |
376 | ||
b7ca232e | 377 | zstrm = zcomp_strm_find(zram->comp); |
e46e3315 | 378 | locked = true; |
ba82fe2e | 379 | user_mem = kmap_atomic(page); |
924bd88d | 380 | |
397c6066 | 381 | if (is_partial_io(bvec)) { |
924bd88d JM |
382 | memcpy(uncmem + offset, user_mem + bvec->bv_offset, |
383 | bvec->bv_len); | |
397c6066 NG |
384 | kunmap_atomic(user_mem); |
385 | user_mem = NULL; | |
386 | } else { | |
924bd88d | 387 | uncmem = user_mem; |
397c6066 | 388 | } |
924bd88d JM |
389 | |
390 | if (page_zero_filled(uncmem)) { | |
ba82fe2e | 391 | kunmap_atomic(user_mem); |
f40ac2ae | 392 | /* Free memory associated with this sector now. */ |
92967471 | 393 | write_lock(&zram->meta->tb_lock); |
f40ac2ae | 394 | zram_free_page(zram, index); |
92967471 MK |
395 | zram_set_flag(meta, index, ZRAM_ZERO); |
396 | write_unlock(&zram->meta->tb_lock); | |
f40ac2ae | 397 | |
90a7806e | 398 | atomic64_inc(&zram->stats.zero_pages); |
924bd88d JM |
399 | ret = 0; |
400 | goto out; | |
8c921b2b | 401 | } |
306b0c95 | 402 | |
b7ca232e | 403 | ret = zcomp_compress(zram->comp, zstrm, uncmem, &clen); |
397c6066 NG |
404 | if (!is_partial_io(bvec)) { |
405 | kunmap_atomic(user_mem); | |
406 | user_mem = NULL; | |
407 | uncmem = NULL; | |
408 | } | |
306b0c95 | 409 | |
b7ca232e | 410 | if (unlikely(ret)) { |
8c921b2b | 411 | pr_err("Compression failed! err=%d\n", ret); |
924bd88d | 412 | goto out; |
8c921b2b | 413 | } |
b7ca232e | 414 | src = zstrm->buffer; |
c8f2f0db | 415 | if (unlikely(clen > max_zpage_size)) { |
c8f2f0db | 416 | clen = PAGE_SIZE; |
397c6066 NG |
417 | if (is_partial_io(bvec)) |
418 | src = uncmem; | |
c8f2f0db | 419 | } |
a1dd52af | 420 | |
8b3cc3ed | 421 | handle = zs_malloc(meta->mem_pool, clen); |
fd1a30de | 422 | if (!handle) { |
596b3dd4 MR |
423 | pr_info("Error allocating memory for compressed page: %u, size=%zu\n", |
424 | index, clen); | |
924bd88d JM |
425 | ret = -ENOMEM; |
426 | goto out; | |
8c921b2b | 427 | } |
8b3cc3ed | 428 | cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO); |
306b0c95 | 429 | |
42e99bd9 | 430 | if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) { |
397c6066 | 431 | src = kmap_atomic(page); |
42e99bd9 | 432 | copy_page(cmem, src); |
397c6066 | 433 | kunmap_atomic(src); |
42e99bd9 JL |
434 | } else { |
435 | memcpy(cmem, src, clen); | |
436 | } | |
306b0c95 | 437 | |
b7ca232e SS |
438 | zcomp_strm_release(zram->comp, zstrm); |
439 | locked = false; | |
8b3cc3ed | 440 | zs_unmap_object(meta->mem_pool, handle); |
fd1a30de | 441 | |
f40ac2ae SS |
442 | /* |
443 | * Free memory associated with this sector | |
444 | * before overwriting unused sectors. | |
445 | */ | |
92967471 | 446 | write_lock(&zram->meta->tb_lock); |
f40ac2ae SS |
447 | zram_free_page(zram, index); |
448 | ||
8b3cc3ed MK |
449 | meta->table[index].handle = handle; |
450 | meta->table[index].size = clen; | |
92967471 | 451 | write_unlock(&zram->meta->tb_lock); |
306b0c95 | 452 | |
8c921b2b | 453 | /* Update stats */ |
90a7806e SS |
454 | atomic64_add(clen, &zram->stats.compr_data_size); |
455 | atomic64_inc(&zram->stats.pages_stored); | |
924bd88d | 456 | out: |
e46e3315 | 457 | if (locked) |
b7ca232e | 458 | zcomp_strm_release(zram->comp, zstrm); |
397c6066 NG |
459 | if (is_partial_io(bvec)) |
460 | kfree(uncmem); | |
924bd88d | 461 | if (ret) |
da5cc7d3 | 462 | atomic64_inc(&zram->stats.failed_writes); |
924bd88d | 463 | return ret; |
8c921b2b JM |
464 | } |
465 | ||
466 | static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index, | |
be257c61 | 467 | int offset, struct bio *bio) |
8c921b2b | 468 | { |
c5bde238 | 469 | int ret; |
be257c61 | 470 | int rw = bio_data_dir(bio); |
8c921b2b | 471 | |
be257c61 SS |
472 | if (rw == READ) { |
473 | atomic64_inc(&zram->stats.num_reads); | |
c5bde238 | 474 | ret = zram_bvec_read(zram, bvec, index, offset, bio); |
be257c61 SS |
475 | } else { |
476 | atomic64_inc(&zram->stats.num_writes); | |
c5bde238 | 477 | ret = zram_bvec_write(zram, bvec, index, offset); |
be257c61 | 478 | } |
c5bde238 JM |
479 | |
480 | return ret; | |
924bd88d JM |
481 | } |
482 | ||
2b86ab9c | 483 | static void zram_reset_device(struct zram *zram, bool reset_capacity) |
924bd88d | 484 | { |
9b3bb7ab SS |
485 | size_t index; |
486 | struct zram_meta *meta; | |
487 | ||
644d4787 | 488 | down_write(&zram->init_lock); |
be2d1d56 | 489 | if (!init_done(zram)) { |
644d4787 | 490 | up_write(&zram->init_lock); |
9b3bb7ab | 491 | return; |
644d4787 | 492 | } |
9b3bb7ab SS |
493 | |
494 | meta = zram->meta; | |
9b3bb7ab SS |
495 | /* Free all pages that are still in this zram device */ |
496 | for (index = 0; index < zram->disksize >> PAGE_SHIFT; index++) { | |
497 | unsigned long handle = meta->table[index].handle; | |
498 | if (!handle) | |
499 | continue; | |
500 | ||
501 | zs_free(meta->mem_pool, handle); | |
502 | } | |
503 | ||
b7ca232e | 504 | zcomp_destroy(zram->comp); |
9b3bb7ab SS |
505 | zram_meta_free(zram->meta); |
506 | zram->meta = NULL; | |
507 | /* Reset stats */ | |
508 | memset(&zram->stats, 0, sizeof(zram->stats)); | |
509 | ||
510 | zram->disksize = 0; | |
2b86ab9c MK |
511 | if (reset_capacity) |
512 | set_capacity(zram->disk, 0); | |
644d4787 | 513 | up_write(&zram->init_lock); |
9b3bb7ab SS |
514 | } |
515 | ||
9b3bb7ab SS |
516 | static ssize_t disksize_store(struct device *dev, |
517 | struct device_attribute *attr, const char *buf, size_t len) | |
518 | { | |
519 | u64 disksize; | |
520 | struct zram_meta *meta; | |
521 | struct zram *zram = dev_to_zram(dev); | |
b7ca232e | 522 | int err; |
9b3bb7ab SS |
523 | |
524 | disksize = memparse(buf, NULL); | |
525 | if (!disksize) | |
526 | return -EINVAL; | |
527 | ||
528 | disksize = PAGE_ALIGN(disksize); | |
529 | meta = zram_meta_alloc(disksize); | |
db5d711e MK |
530 | if (!meta) |
531 | return -ENOMEM; | |
b67d1ec1 | 532 | |
9b3bb7ab | 533 | down_write(&zram->init_lock); |
be2d1d56 | 534 | if (init_done(zram)) { |
9b3bb7ab | 535 | pr_info("Cannot change disksize for initialized device\n"); |
b7ca232e SS |
536 | err = -EBUSY; |
537 | goto out_free_meta; | |
538 | } | |
539 | ||
540 | zram->comp = zcomp_create(default_compressor); | |
541 | if (!zram->comp) { | |
542 | pr_info("Cannot initialise %s compressing backend\n", | |
543 | default_compressor); | |
544 | err = -EINVAL; | |
545 | goto out_free_meta; | |
9b3bb7ab SS |
546 | } |
547 | ||
b67d1ec1 | 548 | zram->meta = meta; |
9b3bb7ab SS |
549 | zram->disksize = disksize; |
550 | set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT); | |
9b3bb7ab SS |
551 | up_write(&zram->init_lock); |
552 | ||
553 | return len; | |
b7ca232e SS |
554 | |
555 | out_free_meta: | |
556 | up_write(&zram->init_lock); | |
557 | zram_meta_free(meta); | |
558 | return err; | |
9b3bb7ab SS |
559 | } |
560 | ||
561 | static ssize_t reset_store(struct device *dev, | |
562 | struct device_attribute *attr, const char *buf, size_t len) | |
563 | { | |
564 | int ret; | |
565 | unsigned short do_reset; | |
566 | struct zram *zram; | |
567 | struct block_device *bdev; | |
568 | ||
569 | zram = dev_to_zram(dev); | |
570 | bdev = bdget_disk(zram->disk, 0); | |
571 | ||
46a51c80 RK |
572 | if (!bdev) |
573 | return -ENOMEM; | |
574 | ||
9b3bb7ab | 575 | /* Do not reset an active device! */ |
1b672224 RK |
576 | if (bdev->bd_holders) { |
577 | ret = -EBUSY; | |
578 | goto out; | |
579 | } | |
9b3bb7ab SS |
580 | |
581 | ret = kstrtou16(buf, 10, &do_reset); | |
582 | if (ret) | |
1b672224 | 583 | goto out; |
9b3bb7ab | 584 | |
1b672224 RK |
585 | if (!do_reset) { |
586 | ret = -EINVAL; | |
587 | goto out; | |
588 | } | |
9b3bb7ab SS |
589 | |
590 | /* Make sure all pending I/O is finished */ | |
46a51c80 | 591 | fsync_bdev(bdev); |
1b672224 | 592 | bdput(bdev); |
9b3bb7ab | 593 | |
2b86ab9c | 594 | zram_reset_device(zram, true); |
9b3bb7ab | 595 | return len; |
1b672224 RK |
596 | |
597 | out: | |
598 | bdput(bdev); | |
599 | return ret; | |
8c921b2b JM |
600 | } |
601 | ||
be257c61 | 602 | static void __zram_make_request(struct zram *zram, struct bio *bio) |
8c921b2b | 603 | { |
7988613b | 604 | int offset; |
8c921b2b | 605 | u32 index; |
7988613b KO |
606 | struct bio_vec bvec; |
607 | struct bvec_iter iter; | |
8c921b2b | 608 | |
4f024f37 KO |
609 | index = bio->bi_iter.bi_sector >> SECTORS_PER_PAGE_SHIFT; |
610 | offset = (bio->bi_iter.bi_sector & | |
611 | (SECTORS_PER_PAGE - 1)) << SECTOR_SHIFT; | |
8c921b2b | 612 | |
7988613b | 613 | bio_for_each_segment(bvec, bio, iter) { |
924bd88d JM |
614 | int max_transfer_size = PAGE_SIZE - offset; |
615 | ||
7988613b | 616 | if (bvec.bv_len > max_transfer_size) { |
924bd88d JM |
617 | /* |
618 | * zram_bvec_rw() can only make operation on a single | |
619 | * zram page. Split the bio vector. | |
620 | */ | |
621 | struct bio_vec bv; | |
622 | ||
7988613b | 623 | bv.bv_page = bvec.bv_page; |
924bd88d | 624 | bv.bv_len = max_transfer_size; |
7988613b | 625 | bv.bv_offset = bvec.bv_offset; |
924bd88d | 626 | |
be257c61 | 627 | if (zram_bvec_rw(zram, &bv, index, offset, bio) < 0) |
924bd88d JM |
628 | goto out; |
629 | ||
7988613b | 630 | bv.bv_len = bvec.bv_len - max_transfer_size; |
924bd88d | 631 | bv.bv_offset += max_transfer_size; |
be257c61 | 632 | if (zram_bvec_rw(zram, &bv, index + 1, 0, bio) < 0) |
924bd88d JM |
633 | goto out; |
634 | } else | |
be257c61 | 635 | if (zram_bvec_rw(zram, &bvec, index, offset, bio) < 0) |
924bd88d JM |
636 | goto out; |
637 | ||
7988613b | 638 | update_position(&index, &offset, &bvec); |
a1dd52af | 639 | } |
306b0c95 NG |
640 | |
641 | set_bit(BIO_UPTODATE, &bio->bi_flags); | |
642 | bio_endio(bio, 0); | |
7d7854b4 | 643 | return; |
306b0c95 NG |
644 | |
645 | out: | |
306b0c95 | 646 | bio_io_error(bio); |
306b0c95 NG |
647 | } |
648 | ||
306b0c95 | 649 | /* |
f1e3cfff | 650 | * Handler function for all zram I/O requests. |
306b0c95 | 651 | */ |
5a7bbad2 | 652 | static void zram_make_request(struct request_queue *queue, struct bio *bio) |
306b0c95 | 653 | { |
f1e3cfff | 654 | struct zram *zram = queue->queuedata; |
306b0c95 | 655 | |
0900beae | 656 | down_read(&zram->init_lock); |
be2d1d56 | 657 | if (unlikely(!init_done(zram))) |
3de738cd | 658 | goto error; |
0900beae | 659 | |
f1e3cfff | 660 | if (!valid_io_request(zram, bio)) { |
da5cc7d3 | 661 | atomic64_inc(&zram->stats.invalid_io); |
3de738cd | 662 | goto error; |
6642a67c JM |
663 | } |
664 | ||
be257c61 | 665 | __zram_make_request(zram, bio); |
0900beae | 666 | up_read(&zram->init_lock); |
306b0c95 | 667 | |
b4fdcb02 | 668 | return; |
0900beae | 669 | |
0900beae | 670 | error: |
3de738cd | 671 | up_read(&zram->init_lock); |
0900beae | 672 | bio_io_error(bio); |
306b0c95 NG |
673 | } |
674 | ||
2ccbec05 NG |
675 | static void zram_slot_free_notify(struct block_device *bdev, |
676 | unsigned long index) | |
107c161b | 677 | { |
f1e3cfff | 678 | struct zram *zram; |
f614a9f4 | 679 | struct zram_meta *meta; |
107c161b | 680 | |
f1e3cfff | 681 | zram = bdev->bd_disk->private_data; |
f614a9f4 | 682 | meta = zram->meta; |
a0c516cb | 683 | |
f614a9f4 MK |
684 | write_lock(&meta->tb_lock); |
685 | zram_free_page(zram, index); | |
686 | write_unlock(&meta->tb_lock); | |
687 | atomic64_inc(&zram->stats.notify_free); | |
107c161b NG |
688 | } |
689 | ||
f1e3cfff | 690 | static const struct block_device_operations zram_devops = { |
f1e3cfff | 691 | .swap_slot_free_notify = zram_slot_free_notify, |
107c161b | 692 | .owner = THIS_MODULE |
306b0c95 NG |
693 | }; |
694 | ||
9b3bb7ab SS |
695 | static DEVICE_ATTR(disksize, S_IRUGO | S_IWUSR, |
696 | disksize_show, disksize_store); | |
697 | static DEVICE_ATTR(initstate, S_IRUGO, initstate_show, NULL); | |
698 | static DEVICE_ATTR(reset, S_IWUSR, NULL, reset_store); | |
9b3bb7ab | 699 | static DEVICE_ATTR(orig_data_size, S_IRUGO, orig_data_size_show, NULL); |
9b3bb7ab SS |
700 | static DEVICE_ATTR(mem_used_total, S_IRUGO, mem_used_total_show, NULL); |
701 | ||
a68eb3b6 SS |
702 | ZRAM_ATTR_RO(num_reads); |
703 | ZRAM_ATTR_RO(num_writes); | |
64447249 SS |
704 | ZRAM_ATTR_RO(failed_reads); |
705 | ZRAM_ATTR_RO(failed_writes); | |
a68eb3b6 SS |
706 | ZRAM_ATTR_RO(invalid_io); |
707 | ZRAM_ATTR_RO(notify_free); | |
708 | ZRAM_ATTR_RO(zero_pages); | |
709 | ZRAM_ATTR_RO(compr_data_size); | |
710 | ||
9b3bb7ab SS |
711 | static struct attribute *zram_disk_attrs[] = { |
712 | &dev_attr_disksize.attr, | |
713 | &dev_attr_initstate.attr, | |
714 | &dev_attr_reset.attr, | |
715 | &dev_attr_num_reads.attr, | |
716 | &dev_attr_num_writes.attr, | |
64447249 SS |
717 | &dev_attr_failed_reads.attr, |
718 | &dev_attr_failed_writes.attr, | |
9b3bb7ab SS |
719 | &dev_attr_invalid_io.attr, |
720 | &dev_attr_notify_free.attr, | |
721 | &dev_attr_zero_pages.attr, | |
722 | &dev_attr_orig_data_size.attr, | |
723 | &dev_attr_compr_data_size.attr, | |
724 | &dev_attr_mem_used_total.attr, | |
725 | NULL, | |
726 | }; | |
727 | ||
728 | static struct attribute_group zram_disk_attr_group = { | |
729 | .attrs = zram_disk_attrs, | |
730 | }; | |
731 | ||
f1e3cfff | 732 | static int create_device(struct zram *zram, int device_id) |
306b0c95 | 733 | { |
39a9b8ac | 734 | int ret = -ENOMEM; |
de1a21a0 | 735 | |
0900beae | 736 | init_rwsem(&zram->init_lock); |
306b0c95 | 737 | |
f1e3cfff NG |
738 | zram->queue = blk_alloc_queue(GFP_KERNEL); |
739 | if (!zram->queue) { | |
306b0c95 NG |
740 | pr_err("Error allocating disk queue for device %d\n", |
741 | device_id); | |
de1a21a0 | 742 | goto out; |
306b0c95 NG |
743 | } |
744 | ||
f1e3cfff NG |
745 | blk_queue_make_request(zram->queue, zram_make_request); |
746 | zram->queue->queuedata = zram; | |
306b0c95 NG |
747 | |
748 | /* gendisk structure */ | |
f1e3cfff NG |
749 | zram->disk = alloc_disk(1); |
750 | if (!zram->disk) { | |
94b8435f | 751 | pr_warn("Error allocating disk structure for device %d\n", |
306b0c95 | 752 | device_id); |
39a9b8ac | 753 | goto out_free_queue; |
306b0c95 NG |
754 | } |
755 | ||
f1e3cfff NG |
756 | zram->disk->major = zram_major; |
757 | zram->disk->first_minor = device_id; | |
758 | zram->disk->fops = &zram_devops; | |
759 | zram->disk->queue = zram->queue; | |
760 | zram->disk->private_data = zram; | |
761 | snprintf(zram->disk->disk_name, 16, "zram%d", device_id); | |
306b0c95 | 762 | |
33863c21 | 763 | /* Actual capacity set using syfs (/sys/block/zram<id>/disksize */ |
f1e3cfff | 764 | set_capacity(zram->disk, 0); |
b67d1ec1 SS |
765 | /* zram devices sort of resembles non-rotational disks */ |
766 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue); | |
a1dd52af NG |
767 | /* |
768 | * To ensure that we always get PAGE_SIZE aligned | |
769 | * and n*PAGE_SIZED sized I/O requests. | |
770 | */ | |
f1e3cfff | 771 | blk_queue_physical_block_size(zram->disk->queue, PAGE_SIZE); |
7b19b8d4 RJ |
772 | blk_queue_logical_block_size(zram->disk->queue, |
773 | ZRAM_LOGICAL_BLOCK_SIZE); | |
f1e3cfff NG |
774 | blk_queue_io_min(zram->disk->queue, PAGE_SIZE); |
775 | blk_queue_io_opt(zram->disk->queue, PAGE_SIZE); | |
5d83d5a0 | 776 | |
f1e3cfff | 777 | add_disk(zram->disk); |
306b0c95 | 778 | |
33863c21 NG |
779 | ret = sysfs_create_group(&disk_to_dev(zram->disk)->kobj, |
780 | &zram_disk_attr_group); | |
781 | if (ret < 0) { | |
94b8435f | 782 | pr_warn("Error creating sysfs group"); |
39a9b8ac | 783 | goto out_free_disk; |
33863c21 | 784 | } |
33863c21 | 785 | |
be2d1d56 | 786 | zram->meta = NULL; |
39a9b8ac | 787 | return 0; |
de1a21a0 | 788 | |
39a9b8ac JL |
789 | out_free_disk: |
790 | del_gendisk(zram->disk); | |
791 | put_disk(zram->disk); | |
792 | out_free_queue: | |
793 | blk_cleanup_queue(zram->queue); | |
de1a21a0 NG |
794 | out: |
795 | return ret; | |
306b0c95 NG |
796 | } |
797 | ||
f1e3cfff | 798 | static void destroy_device(struct zram *zram) |
306b0c95 | 799 | { |
33863c21 NG |
800 | sysfs_remove_group(&disk_to_dev(zram->disk)->kobj, |
801 | &zram_disk_attr_group); | |
33863c21 | 802 | |
59d3fe54 RK |
803 | del_gendisk(zram->disk); |
804 | put_disk(zram->disk); | |
306b0c95 | 805 | |
59d3fe54 | 806 | blk_cleanup_queue(zram->queue); |
306b0c95 NG |
807 | } |
808 | ||
f1e3cfff | 809 | static int __init zram_init(void) |
306b0c95 | 810 | { |
de1a21a0 | 811 | int ret, dev_id; |
306b0c95 | 812 | |
5fa5a901 | 813 | if (num_devices > max_num_devices) { |
94b8435f | 814 | pr_warn("Invalid value for num_devices: %u\n", |
5fa5a901 | 815 | num_devices); |
de1a21a0 NG |
816 | ret = -EINVAL; |
817 | goto out; | |
306b0c95 NG |
818 | } |
819 | ||
f1e3cfff NG |
820 | zram_major = register_blkdev(0, "zram"); |
821 | if (zram_major <= 0) { | |
94b8435f | 822 | pr_warn("Unable to get major number\n"); |
de1a21a0 NG |
823 | ret = -EBUSY; |
824 | goto out; | |
306b0c95 NG |
825 | } |
826 | ||
306b0c95 | 827 | /* Allocate the device array and initialize each one */ |
5fa5a901 | 828 | zram_devices = kzalloc(num_devices * sizeof(struct zram), GFP_KERNEL); |
43801f6e | 829 | if (!zram_devices) { |
de1a21a0 NG |
830 | ret = -ENOMEM; |
831 | goto unregister; | |
832 | } | |
306b0c95 | 833 | |
5fa5a901 | 834 | for (dev_id = 0; dev_id < num_devices; dev_id++) { |
43801f6e | 835 | ret = create_device(&zram_devices[dev_id], dev_id); |
de1a21a0 | 836 | if (ret) |
3bf040c7 | 837 | goto free_devices; |
de1a21a0 NG |
838 | } |
839 | ||
ca3d70bd DB |
840 | pr_info("Created %u device(s) ...\n", num_devices); |
841 | ||
306b0c95 | 842 | return 0; |
de1a21a0 | 843 | |
3bf040c7 | 844 | free_devices: |
de1a21a0 | 845 | while (dev_id) |
43801f6e NW |
846 | destroy_device(&zram_devices[--dev_id]); |
847 | kfree(zram_devices); | |
de1a21a0 | 848 | unregister: |
f1e3cfff | 849 | unregister_blkdev(zram_major, "zram"); |
de1a21a0 | 850 | out: |
306b0c95 NG |
851 | return ret; |
852 | } | |
853 | ||
f1e3cfff | 854 | static void __exit zram_exit(void) |
306b0c95 NG |
855 | { |
856 | int i; | |
f1e3cfff | 857 | struct zram *zram; |
306b0c95 | 858 | |
5fa5a901 | 859 | for (i = 0; i < num_devices; i++) { |
43801f6e | 860 | zram = &zram_devices[i]; |
306b0c95 | 861 | |
f1e3cfff | 862 | destroy_device(zram); |
2b86ab9c MK |
863 | /* |
864 | * Shouldn't access zram->disk after destroy_device | |
865 | * because destroy_device already released zram->disk. | |
866 | */ | |
867 | zram_reset_device(zram, false); | |
306b0c95 NG |
868 | } |
869 | ||
f1e3cfff | 870 | unregister_blkdev(zram_major, "zram"); |
306b0c95 | 871 | |
43801f6e | 872 | kfree(zram_devices); |
306b0c95 NG |
873 | pr_debug("Cleanup done!\n"); |
874 | } | |
875 | ||
f1e3cfff NG |
876 | module_init(zram_init); |
877 | module_exit(zram_exit); | |
306b0c95 | 878 | |
9b3bb7ab SS |
879 | module_param(num_devices, uint, 0); |
880 | MODULE_PARM_DESC(num_devices, "Number of zram devices"); | |
881 | ||
306b0c95 NG |
882 | MODULE_LICENSE("Dual BSD/GPL"); |
883 | MODULE_AUTHOR("Nitin Gupta <ngupta@vflare.org>"); | |
f1e3cfff | 884 | MODULE_DESCRIPTION("Compressed RAM Block Device"); |