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 | ||
beca3ec7 SS |
111 | static ssize_t max_comp_streams_show(struct device *dev, |
112 | struct device_attribute *attr, char *buf) | |
113 | { | |
114 | int val; | |
115 | struct zram *zram = dev_to_zram(dev); | |
116 | ||
117 | down_read(&zram->init_lock); | |
118 | val = zram->max_comp_streams; | |
119 | up_read(&zram->init_lock); | |
120 | ||
121 | return sprintf(buf, "%d\n", val); | |
122 | } | |
123 | ||
124 | static ssize_t max_comp_streams_store(struct device *dev, | |
125 | struct device_attribute *attr, const char *buf, size_t len) | |
126 | { | |
127 | int num; | |
128 | struct zram *zram = dev_to_zram(dev); | |
129 | ||
130 | if (kstrtoint(buf, 0, &num)) | |
131 | return -EINVAL; | |
132 | if (num < 1) | |
133 | return -EINVAL; | |
134 | down_write(&zram->init_lock); | |
135 | if (init_done(zram)) { | |
fe8eb122 SS |
136 | if (zcomp_set_max_streams(zram->comp, num)) |
137 | pr_info("Cannot change max compression streams\n"); | |
beca3ec7 SS |
138 | } |
139 | zram->max_comp_streams = num; | |
140 | up_write(&zram->init_lock); | |
141 | return len; | |
142 | } | |
143 | ||
e46b8a03 SS |
144 | static ssize_t comp_algorithm_show(struct device *dev, |
145 | struct device_attribute *attr, char *buf) | |
146 | { | |
147 | size_t sz; | |
148 | struct zram *zram = dev_to_zram(dev); | |
149 | ||
150 | down_read(&zram->init_lock); | |
151 | sz = zcomp_available_show(zram->compressor, buf); | |
152 | up_read(&zram->init_lock); | |
153 | ||
154 | return sz; | |
155 | } | |
156 | ||
157 | static ssize_t comp_algorithm_store(struct device *dev, | |
158 | struct device_attribute *attr, const char *buf, size_t len) | |
159 | { | |
160 | struct zram *zram = dev_to_zram(dev); | |
161 | down_write(&zram->init_lock); | |
162 | if (init_done(zram)) { | |
163 | up_write(&zram->init_lock); | |
164 | pr_info("Can't change algorithm for initialized device\n"); | |
165 | return -EBUSY; | |
166 | } | |
167 | strlcpy(zram->compressor, buf, sizeof(zram->compressor)); | |
168 | up_write(&zram->init_lock); | |
169 | return len; | |
170 | } | |
171 | ||
92967471 | 172 | /* flag operations needs meta->tb_lock */ |
8b3cc3ed | 173 | static int zram_test_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 174 | enum zram_pageflags flag) |
306b0c95 | 175 | { |
8b3cc3ed | 176 | return meta->table[index].flags & BIT(flag); |
306b0c95 NG |
177 | } |
178 | ||
8b3cc3ed | 179 | static void zram_set_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 180 | enum zram_pageflags flag) |
306b0c95 | 181 | { |
8b3cc3ed | 182 | meta->table[index].flags |= BIT(flag); |
306b0c95 NG |
183 | } |
184 | ||
8b3cc3ed | 185 | static void zram_clear_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 186 | enum zram_pageflags flag) |
306b0c95 | 187 | { |
8b3cc3ed | 188 | meta->table[index].flags &= ~BIT(flag); |
306b0c95 NG |
189 | } |
190 | ||
9b3bb7ab SS |
191 | static inline int is_partial_io(struct bio_vec *bvec) |
192 | { | |
193 | return bvec->bv_len != PAGE_SIZE; | |
194 | } | |
195 | ||
196 | /* | |
197 | * Check if request is within bounds and aligned on zram logical blocks. | |
198 | */ | |
199 | static inline int valid_io_request(struct zram *zram, struct bio *bio) | |
200 | { | |
201 | u64 start, end, bound; | |
a539c72a | 202 | |
9b3bb7ab | 203 | /* unaligned request */ |
4f024f37 KO |
204 | if (unlikely(bio->bi_iter.bi_sector & |
205 | (ZRAM_SECTOR_PER_LOGICAL_BLOCK - 1))) | |
9b3bb7ab | 206 | return 0; |
4f024f37 | 207 | if (unlikely(bio->bi_iter.bi_size & (ZRAM_LOGICAL_BLOCK_SIZE - 1))) |
9b3bb7ab SS |
208 | return 0; |
209 | ||
4f024f37 KO |
210 | start = bio->bi_iter.bi_sector; |
211 | end = start + (bio->bi_iter.bi_size >> SECTOR_SHIFT); | |
9b3bb7ab SS |
212 | bound = zram->disksize >> SECTOR_SHIFT; |
213 | /* out of range range */ | |
75c7caf5 | 214 | if (unlikely(start >= bound || end > bound || start > end)) |
9b3bb7ab SS |
215 | return 0; |
216 | ||
217 | /* I/O request is valid */ | |
218 | return 1; | |
219 | } | |
220 | ||
221 | static void zram_meta_free(struct zram_meta *meta) | |
222 | { | |
223 | zs_destroy_pool(meta->mem_pool); | |
9b3bb7ab SS |
224 | vfree(meta->table); |
225 | kfree(meta); | |
226 | } | |
227 | ||
228 | static struct zram_meta *zram_meta_alloc(u64 disksize) | |
229 | { | |
230 | size_t num_pages; | |
231 | struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL); | |
232 | if (!meta) | |
233 | goto out; | |
234 | ||
9b3bb7ab SS |
235 | num_pages = disksize >> PAGE_SHIFT; |
236 | meta->table = vzalloc(num_pages * sizeof(*meta->table)); | |
237 | if (!meta->table) { | |
238 | pr_err("Error allocating zram address table\n"); | |
b7ca232e | 239 | goto free_meta; |
9b3bb7ab SS |
240 | } |
241 | ||
242 | meta->mem_pool = zs_create_pool(GFP_NOIO | __GFP_HIGHMEM); | |
243 | if (!meta->mem_pool) { | |
244 | pr_err("Error creating memory pool\n"); | |
245 | goto free_table; | |
246 | } | |
247 | ||
92967471 | 248 | rwlock_init(&meta->tb_lock); |
9b3bb7ab SS |
249 | return meta; |
250 | ||
251 | free_table: | |
252 | vfree(meta->table); | |
9b3bb7ab SS |
253 | free_meta: |
254 | kfree(meta); | |
255 | meta = NULL; | |
256 | out: | |
257 | return meta; | |
258 | } | |
259 | ||
260 | static void update_position(u32 *index, int *offset, struct bio_vec *bvec) | |
261 | { | |
262 | if (*offset + bvec->bv_len >= PAGE_SIZE) | |
263 | (*index)++; | |
264 | *offset = (*offset + bvec->bv_len) % PAGE_SIZE; | |
265 | } | |
266 | ||
306b0c95 NG |
267 | static int page_zero_filled(void *ptr) |
268 | { | |
269 | unsigned int pos; | |
270 | unsigned long *page; | |
271 | ||
272 | page = (unsigned long *)ptr; | |
273 | ||
274 | for (pos = 0; pos != PAGE_SIZE / sizeof(*page); pos++) { | |
275 | if (page[pos]) | |
276 | return 0; | |
277 | } | |
278 | ||
279 | return 1; | |
280 | } | |
281 | ||
9b3bb7ab SS |
282 | static void handle_zero_page(struct bio_vec *bvec) |
283 | { | |
284 | struct page *page = bvec->bv_page; | |
285 | void *user_mem; | |
286 | ||
287 | user_mem = kmap_atomic(page); | |
288 | if (is_partial_io(bvec)) | |
289 | memset(user_mem + bvec->bv_offset, 0, bvec->bv_len); | |
290 | else | |
291 | clear_page(user_mem); | |
292 | kunmap_atomic(user_mem); | |
293 | ||
294 | flush_dcache_page(page); | |
295 | } | |
296 | ||
92967471 | 297 | /* NOTE: caller should hold meta->tb_lock with write-side */ |
f1e3cfff | 298 | static void zram_free_page(struct zram *zram, size_t index) |
306b0c95 | 299 | { |
8b3cc3ed MK |
300 | struct zram_meta *meta = zram->meta; |
301 | unsigned long handle = meta->table[index].handle; | |
306b0c95 | 302 | |
fd1a30de | 303 | if (unlikely(!handle)) { |
2e882281 NG |
304 | /* |
305 | * No memory is allocated for zero filled pages. | |
306 | * Simply clear zero page flag. | |
307 | */ | |
8b3cc3ed MK |
308 | if (zram_test_flag(meta, index, ZRAM_ZERO)) { |
309 | zram_clear_flag(meta, index, ZRAM_ZERO); | |
90a7806e | 310 | atomic64_dec(&zram->stats.zero_pages); |
306b0c95 NG |
311 | } |
312 | return; | |
313 | } | |
314 | ||
8b3cc3ed | 315 | zs_free(meta->mem_pool, handle); |
306b0c95 | 316 | |
90a7806e SS |
317 | atomic64_sub(meta->table[index].size, &zram->stats.compr_data_size); |
318 | atomic64_dec(&zram->stats.pages_stored); | |
306b0c95 | 319 | |
8b3cc3ed MK |
320 | meta->table[index].handle = 0; |
321 | meta->table[index].size = 0; | |
306b0c95 NG |
322 | } |
323 | ||
37b51fdd | 324 | static int zram_decompress_page(struct zram *zram, char *mem, u32 index) |
306b0c95 | 325 | { |
b7ca232e | 326 | int ret = 0; |
37b51fdd | 327 | unsigned char *cmem; |
8b3cc3ed | 328 | struct zram_meta *meta = zram->meta; |
92967471 MK |
329 | unsigned long handle; |
330 | u16 size; | |
331 | ||
332 | read_lock(&meta->tb_lock); | |
333 | handle = meta->table[index].handle; | |
334 | size = meta->table[index].size; | |
306b0c95 | 335 | |
8b3cc3ed | 336 | if (!handle || zram_test_flag(meta, index, ZRAM_ZERO)) { |
92967471 | 337 | read_unlock(&meta->tb_lock); |
42e99bd9 | 338 | clear_page(mem); |
8c921b2b JM |
339 | return 0; |
340 | } | |
306b0c95 | 341 | |
8b3cc3ed | 342 | cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_RO); |
92967471 | 343 | if (size == PAGE_SIZE) |
42e99bd9 | 344 | copy_page(mem, cmem); |
37b51fdd | 345 | else |
b7ca232e | 346 | ret = zcomp_decompress(zram->comp, cmem, size, mem); |
8b3cc3ed | 347 | zs_unmap_object(meta->mem_pool, handle); |
92967471 | 348 | read_unlock(&meta->tb_lock); |
a1dd52af | 349 | |
8c921b2b | 350 | /* Should NEVER happen. Return bio error if it does. */ |
b7ca232e | 351 | if (unlikely(ret)) { |
8c921b2b | 352 | pr_err("Decompression failed! err=%d, page=%u\n", ret, index); |
da5cc7d3 | 353 | atomic64_inc(&zram->stats.failed_reads); |
8c921b2b | 354 | return ret; |
a1dd52af | 355 | } |
306b0c95 | 356 | |
8c921b2b | 357 | return 0; |
306b0c95 NG |
358 | } |
359 | ||
37b51fdd SS |
360 | static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec, |
361 | u32 index, int offset, struct bio *bio) | |
924bd88d JM |
362 | { |
363 | int ret; | |
37b51fdd SS |
364 | struct page *page; |
365 | unsigned char *user_mem, *uncmem = NULL; | |
8b3cc3ed | 366 | struct zram_meta *meta = zram->meta; |
37b51fdd SS |
367 | page = bvec->bv_page; |
368 | ||
92967471 | 369 | read_lock(&meta->tb_lock); |
8b3cc3ed MK |
370 | if (unlikely(!meta->table[index].handle) || |
371 | zram_test_flag(meta, index, ZRAM_ZERO)) { | |
92967471 | 372 | read_unlock(&meta->tb_lock); |
37b51fdd | 373 | handle_zero_page(bvec); |
924bd88d JM |
374 | return 0; |
375 | } | |
92967471 | 376 | read_unlock(&meta->tb_lock); |
924bd88d | 377 | |
37b51fdd SS |
378 | if (is_partial_io(bvec)) |
379 | /* Use a temporary buffer to decompress the page */ | |
7e5a5104 MK |
380 | uncmem = kmalloc(PAGE_SIZE, GFP_NOIO); |
381 | ||
382 | user_mem = kmap_atomic(page); | |
383 | if (!is_partial_io(bvec)) | |
37b51fdd SS |
384 | uncmem = user_mem; |
385 | ||
386 | if (!uncmem) { | |
387 | pr_info("Unable to allocate temp memory\n"); | |
388 | ret = -ENOMEM; | |
389 | goto out_cleanup; | |
390 | } | |
924bd88d | 391 | |
37b51fdd | 392 | ret = zram_decompress_page(zram, uncmem, index); |
924bd88d | 393 | /* Should NEVER happen. Return bio error if it does. */ |
b7ca232e | 394 | if (unlikely(ret)) |
37b51fdd | 395 | goto out_cleanup; |
924bd88d | 396 | |
37b51fdd SS |
397 | if (is_partial_io(bvec)) |
398 | memcpy(user_mem + bvec->bv_offset, uncmem + offset, | |
399 | bvec->bv_len); | |
400 | ||
401 | flush_dcache_page(page); | |
402 | ret = 0; | |
403 | out_cleanup: | |
404 | kunmap_atomic(user_mem); | |
405 | if (is_partial_io(bvec)) | |
406 | kfree(uncmem); | |
407 | return ret; | |
924bd88d JM |
408 | } |
409 | ||
410 | static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index, | |
411 | int offset) | |
306b0c95 | 412 | { |
397c6066 | 413 | int ret = 0; |
8c921b2b | 414 | size_t clen; |
c2344348 | 415 | unsigned long handle; |
130f315a | 416 | struct page *page; |
924bd88d | 417 | unsigned char *user_mem, *cmem, *src, *uncmem = NULL; |
8b3cc3ed | 418 | struct zram_meta *meta = zram->meta; |
b7ca232e | 419 | struct zcomp_strm *zstrm; |
e46e3315 | 420 | bool locked = false; |
306b0c95 | 421 | |
8c921b2b | 422 | page = bvec->bv_page; |
924bd88d JM |
423 | if (is_partial_io(bvec)) { |
424 | /* | |
425 | * This is a partial IO. We need to read the full page | |
426 | * before to write the changes. | |
427 | */ | |
7e5a5104 | 428 | uncmem = kmalloc(PAGE_SIZE, GFP_NOIO); |
924bd88d | 429 | if (!uncmem) { |
924bd88d JM |
430 | ret = -ENOMEM; |
431 | goto out; | |
432 | } | |
37b51fdd | 433 | ret = zram_decompress_page(zram, uncmem, index); |
397c6066 | 434 | if (ret) |
924bd88d | 435 | goto out; |
924bd88d JM |
436 | } |
437 | ||
b7ca232e | 438 | zstrm = zcomp_strm_find(zram->comp); |
e46e3315 | 439 | locked = true; |
ba82fe2e | 440 | user_mem = kmap_atomic(page); |
924bd88d | 441 | |
397c6066 | 442 | if (is_partial_io(bvec)) { |
924bd88d JM |
443 | memcpy(uncmem + offset, user_mem + bvec->bv_offset, |
444 | bvec->bv_len); | |
397c6066 NG |
445 | kunmap_atomic(user_mem); |
446 | user_mem = NULL; | |
447 | } else { | |
924bd88d | 448 | uncmem = user_mem; |
397c6066 | 449 | } |
924bd88d JM |
450 | |
451 | if (page_zero_filled(uncmem)) { | |
ba82fe2e | 452 | kunmap_atomic(user_mem); |
f40ac2ae | 453 | /* Free memory associated with this sector now. */ |
92967471 | 454 | write_lock(&zram->meta->tb_lock); |
f40ac2ae | 455 | zram_free_page(zram, index); |
92967471 MK |
456 | zram_set_flag(meta, index, ZRAM_ZERO); |
457 | write_unlock(&zram->meta->tb_lock); | |
f40ac2ae | 458 | |
90a7806e | 459 | atomic64_inc(&zram->stats.zero_pages); |
924bd88d JM |
460 | ret = 0; |
461 | goto out; | |
8c921b2b | 462 | } |
306b0c95 | 463 | |
b7ca232e | 464 | ret = zcomp_compress(zram->comp, zstrm, uncmem, &clen); |
397c6066 NG |
465 | if (!is_partial_io(bvec)) { |
466 | kunmap_atomic(user_mem); | |
467 | user_mem = NULL; | |
468 | uncmem = NULL; | |
469 | } | |
306b0c95 | 470 | |
b7ca232e | 471 | if (unlikely(ret)) { |
8c921b2b | 472 | pr_err("Compression failed! err=%d\n", ret); |
924bd88d | 473 | goto out; |
8c921b2b | 474 | } |
b7ca232e | 475 | src = zstrm->buffer; |
c8f2f0db | 476 | if (unlikely(clen > max_zpage_size)) { |
c8f2f0db | 477 | clen = PAGE_SIZE; |
397c6066 NG |
478 | if (is_partial_io(bvec)) |
479 | src = uncmem; | |
c8f2f0db | 480 | } |
a1dd52af | 481 | |
8b3cc3ed | 482 | handle = zs_malloc(meta->mem_pool, clen); |
fd1a30de | 483 | if (!handle) { |
596b3dd4 MR |
484 | pr_info("Error allocating memory for compressed page: %u, size=%zu\n", |
485 | index, clen); | |
924bd88d JM |
486 | ret = -ENOMEM; |
487 | goto out; | |
8c921b2b | 488 | } |
8b3cc3ed | 489 | cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO); |
306b0c95 | 490 | |
42e99bd9 | 491 | if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) { |
397c6066 | 492 | src = kmap_atomic(page); |
42e99bd9 | 493 | copy_page(cmem, src); |
397c6066 | 494 | kunmap_atomic(src); |
42e99bd9 JL |
495 | } else { |
496 | memcpy(cmem, src, clen); | |
497 | } | |
306b0c95 | 498 | |
b7ca232e SS |
499 | zcomp_strm_release(zram->comp, zstrm); |
500 | locked = false; | |
8b3cc3ed | 501 | zs_unmap_object(meta->mem_pool, handle); |
fd1a30de | 502 | |
f40ac2ae SS |
503 | /* |
504 | * Free memory associated with this sector | |
505 | * before overwriting unused sectors. | |
506 | */ | |
92967471 | 507 | write_lock(&zram->meta->tb_lock); |
f40ac2ae SS |
508 | zram_free_page(zram, index); |
509 | ||
8b3cc3ed MK |
510 | meta->table[index].handle = handle; |
511 | meta->table[index].size = clen; | |
92967471 | 512 | write_unlock(&zram->meta->tb_lock); |
306b0c95 | 513 | |
8c921b2b | 514 | /* Update stats */ |
90a7806e SS |
515 | atomic64_add(clen, &zram->stats.compr_data_size); |
516 | atomic64_inc(&zram->stats.pages_stored); | |
924bd88d | 517 | out: |
e46e3315 | 518 | if (locked) |
b7ca232e | 519 | zcomp_strm_release(zram->comp, zstrm); |
397c6066 NG |
520 | if (is_partial_io(bvec)) |
521 | kfree(uncmem); | |
924bd88d | 522 | if (ret) |
da5cc7d3 | 523 | atomic64_inc(&zram->stats.failed_writes); |
924bd88d | 524 | return ret; |
8c921b2b JM |
525 | } |
526 | ||
527 | static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index, | |
be257c61 | 528 | int offset, struct bio *bio) |
8c921b2b | 529 | { |
c5bde238 | 530 | int ret; |
be257c61 | 531 | int rw = bio_data_dir(bio); |
8c921b2b | 532 | |
be257c61 SS |
533 | if (rw == READ) { |
534 | atomic64_inc(&zram->stats.num_reads); | |
c5bde238 | 535 | ret = zram_bvec_read(zram, bvec, index, offset, bio); |
be257c61 SS |
536 | } else { |
537 | atomic64_inc(&zram->stats.num_writes); | |
c5bde238 | 538 | ret = zram_bvec_write(zram, bvec, index, offset); |
be257c61 | 539 | } |
c5bde238 JM |
540 | |
541 | return ret; | |
924bd88d JM |
542 | } |
543 | ||
2b86ab9c | 544 | static void zram_reset_device(struct zram *zram, bool reset_capacity) |
924bd88d | 545 | { |
9b3bb7ab SS |
546 | size_t index; |
547 | struct zram_meta *meta; | |
548 | ||
644d4787 | 549 | down_write(&zram->init_lock); |
be2d1d56 | 550 | if (!init_done(zram)) { |
644d4787 | 551 | up_write(&zram->init_lock); |
9b3bb7ab | 552 | return; |
644d4787 | 553 | } |
9b3bb7ab SS |
554 | |
555 | meta = zram->meta; | |
9b3bb7ab SS |
556 | /* Free all pages that are still in this zram device */ |
557 | for (index = 0; index < zram->disksize >> PAGE_SHIFT; index++) { | |
558 | unsigned long handle = meta->table[index].handle; | |
559 | if (!handle) | |
560 | continue; | |
561 | ||
562 | zs_free(meta->mem_pool, handle); | |
563 | } | |
564 | ||
b7ca232e | 565 | zcomp_destroy(zram->comp); |
beca3ec7 SS |
566 | zram->max_comp_streams = 1; |
567 | ||
9b3bb7ab SS |
568 | zram_meta_free(zram->meta); |
569 | zram->meta = NULL; | |
570 | /* Reset stats */ | |
571 | memset(&zram->stats, 0, sizeof(zram->stats)); | |
572 | ||
573 | zram->disksize = 0; | |
2b86ab9c MK |
574 | if (reset_capacity) |
575 | set_capacity(zram->disk, 0); | |
644d4787 | 576 | up_write(&zram->init_lock); |
9b3bb7ab SS |
577 | } |
578 | ||
9b3bb7ab SS |
579 | static ssize_t disksize_store(struct device *dev, |
580 | struct device_attribute *attr, const char *buf, size_t len) | |
581 | { | |
582 | u64 disksize; | |
d61f98c7 | 583 | struct zcomp *comp; |
9b3bb7ab SS |
584 | struct zram_meta *meta; |
585 | struct zram *zram = dev_to_zram(dev); | |
d61f98c7 | 586 | int err = -EINVAL; |
9b3bb7ab SS |
587 | |
588 | disksize = memparse(buf, NULL); | |
589 | if (!disksize) | |
590 | return -EINVAL; | |
591 | ||
592 | disksize = PAGE_ALIGN(disksize); | |
593 | meta = zram_meta_alloc(disksize); | |
db5d711e MK |
594 | if (!meta) |
595 | return -ENOMEM; | |
b67d1ec1 | 596 | |
d61f98c7 SS |
597 | comp = zcomp_create(zram->compressor, zram->max_comp_streams); |
598 | if (!comp) { | |
599 | pr_info("Cannot initialise %s compressing backend\n", | |
600 | zram->compressor); | |
601 | goto out_cleanup; | |
602 | } | |
603 | ||
9b3bb7ab | 604 | down_write(&zram->init_lock); |
be2d1d56 | 605 | if (init_done(zram)) { |
d61f98c7 | 606 | up_write(&zram->init_lock); |
9b3bb7ab | 607 | pr_info("Cannot change disksize for initialized device\n"); |
b7ca232e | 608 | err = -EBUSY; |
d61f98c7 | 609 | goto out_cleanup; |
9b3bb7ab SS |
610 | } |
611 | ||
b67d1ec1 | 612 | zram->meta = meta; |
d61f98c7 | 613 | zram->comp = comp; |
9b3bb7ab SS |
614 | zram->disksize = disksize; |
615 | set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT); | |
9b3bb7ab SS |
616 | up_write(&zram->init_lock); |
617 | ||
618 | return len; | |
b7ca232e | 619 | |
d61f98c7 SS |
620 | out_cleanup: |
621 | if (comp) | |
622 | zcomp_destroy(comp); | |
b7ca232e SS |
623 | zram_meta_free(meta); |
624 | return err; | |
9b3bb7ab SS |
625 | } |
626 | ||
627 | static ssize_t reset_store(struct device *dev, | |
628 | struct device_attribute *attr, const char *buf, size_t len) | |
629 | { | |
630 | int ret; | |
631 | unsigned short do_reset; | |
632 | struct zram *zram; | |
633 | struct block_device *bdev; | |
634 | ||
635 | zram = dev_to_zram(dev); | |
636 | bdev = bdget_disk(zram->disk, 0); | |
637 | ||
46a51c80 RK |
638 | if (!bdev) |
639 | return -ENOMEM; | |
640 | ||
9b3bb7ab | 641 | /* Do not reset an active device! */ |
1b672224 RK |
642 | if (bdev->bd_holders) { |
643 | ret = -EBUSY; | |
644 | goto out; | |
645 | } | |
9b3bb7ab SS |
646 | |
647 | ret = kstrtou16(buf, 10, &do_reset); | |
648 | if (ret) | |
1b672224 | 649 | goto out; |
9b3bb7ab | 650 | |
1b672224 RK |
651 | if (!do_reset) { |
652 | ret = -EINVAL; | |
653 | goto out; | |
654 | } | |
9b3bb7ab SS |
655 | |
656 | /* Make sure all pending I/O is finished */ | |
46a51c80 | 657 | fsync_bdev(bdev); |
1b672224 | 658 | bdput(bdev); |
9b3bb7ab | 659 | |
2b86ab9c | 660 | zram_reset_device(zram, true); |
9b3bb7ab | 661 | return len; |
1b672224 RK |
662 | |
663 | out: | |
664 | bdput(bdev); | |
665 | return ret; | |
8c921b2b JM |
666 | } |
667 | ||
be257c61 | 668 | static void __zram_make_request(struct zram *zram, struct bio *bio) |
8c921b2b | 669 | { |
7988613b | 670 | int offset; |
8c921b2b | 671 | u32 index; |
7988613b KO |
672 | struct bio_vec bvec; |
673 | struct bvec_iter iter; | |
8c921b2b | 674 | |
4f024f37 KO |
675 | index = bio->bi_iter.bi_sector >> SECTORS_PER_PAGE_SHIFT; |
676 | offset = (bio->bi_iter.bi_sector & | |
677 | (SECTORS_PER_PAGE - 1)) << SECTOR_SHIFT; | |
8c921b2b | 678 | |
7988613b | 679 | bio_for_each_segment(bvec, bio, iter) { |
924bd88d JM |
680 | int max_transfer_size = PAGE_SIZE - offset; |
681 | ||
7988613b | 682 | if (bvec.bv_len > max_transfer_size) { |
924bd88d JM |
683 | /* |
684 | * zram_bvec_rw() can only make operation on a single | |
685 | * zram page. Split the bio vector. | |
686 | */ | |
687 | struct bio_vec bv; | |
688 | ||
7988613b | 689 | bv.bv_page = bvec.bv_page; |
924bd88d | 690 | bv.bv_len = max_transfer_size; |
7988613b | 691 | bv.bv_offset = bvec.bv_offset; |
924bd88d | 692 | |
be257c61 | 693 | if (zram_bvec_rw(zram, &bv, index, offset, bio) < 0) |
924bd88d JM |
694 | goto out; |
695 | ||
7988613b | 696 | bv.bv_len = bvec.bv_len - max_transfer_size; |
924bd88d | 697 | bv.bv_offset += max_transfer_size; |
be257c61 | 698 | if (zram_bvec_rw(zram, &bv, index + 1, 0, bio) < 0) |
924bd88d JM |
699 | goto out; |
700 | } else | |
be257c61 | 701 | if (zram_bvec_rw(zram, &bvec, index, offset, bio) < 0) |
924bd88d JM |
702 | goto out; |
703 | ||
7988613b | 704 | update_position(&index, &offset, &bvec); |
a1dd52af | 705 | } |
306b0c95 NG |
706 | |
707 | set_bit(BIO_UPTODATE, &bio->bi_flags); | |
708 | bio_endio(bio, 0); | |
7d7854b4 | 709 | return; |
306b0c95 NG |
710 | |
711 | out: | |
306b0c95 | 712 | bio_io_error(bio); |
306b0c95 NG |
713 | } |
714 | ||
306b0c95 | 715 | /* |
f1e3cfff | 716 | * Handler function for all zram I/O requests. |
306b0c95 | 717 | */ |
5a7bbad2 | 718 | static void zram_make_request(struct request_queue *queue, struct bio *bio) |
306b0c95 | 719 | { |
f1e3cfff | 720 | struct zram *zram = queue->queuedata; |
306b0c95 | 721 | |
0900beae | 722 | down_read(&zram->init_lock); |
be2d1d56 | 723 | if (unlikely(!init_done(zram))) |
3de738cd | 724 | goto error; |
0900beae | 725 | |
f1e3cfff | 726 | if (!valid_io_request(zram, bio)) { |
da5cc7d3 | 727 | atomic64_inc(&zram->stats.invalid_io); |
3de738cd | 728 | goto error; |
6642a67c JM |
729 | } |
730 | ||
be257c61 | 731 | __zram_make_request(zram, bio); |
0900beae | 732 | up_read(&zram->init_lock); |
306b0c95 | 733 | |
b4fdcb02 | 734 | return; |
0900beae | 735 | |
0900beae | 736 | error: |
3de738cd | 737 | up_read(&zram->init_lock); |
0900beae | 738 | bio_io_error(bio); |
306b0c95 NG |
739 | } |
740 | ||
2ccbec05 NG |
741 | static void zram_slot_free_notify(struct block_device *bdev, |
742 | unsigned long index) | |
107c161b | 743 | { |
f1e3cfff | 744 | struct zram *zram; |
f614a9f4 | 745 | struct zram_meta *meta; |
107c161b | 746 | |
f1e3cfff | 747 | zram = bdev->bd_disk->private_data; |
f614a9f4 | 748 | meta = zram->meta; |
a0c516cb | 749 | |
f614a9f4 MK |
750 | write_lock(&meta->tb_lock); |
751 | zram_free_page(zram, index); | |
752 | write_unlock(&meta->tb_lock); | |
753 | atomic64_inc(&zram->stats.notify_free); | |
107c161b NG |
754 | } |
755 | ||
f1e3cfff | 756 | static const struct block_device_operations zram_devops = { |
f1e3cfff | 757 | .swap_slot_free_notify = zram_slot_free_notify, |
107c161b | 758 | .owner = THIS_MODULE |
306b0c95 NG |
759 | }; |
760 | ||
9b3bb7ab SS |
761 | static DEVICE_ATTR(disksize, S_IRUGO | S_IWUSR, |
762 | disksize_show, disksize_store); | |
763 | static DEVICE_ATTR(initstate, S_IRUGO, initstate_show, NULL); | |
764 | static DEVICE_ATTR(reset, S_IWUSR, NULL, reset_store); | |
9b3bb7ab | 765 | static DEVICE_ATTR(orig_data_size, S_IRUGO, orig_data_size_show, NULL); |
9b3bb7ab | 766 | static DEVICE_ATTR(mem_used_total, S_IRUGO, mem_used_total_show, NULL); |
beca3ec7 SS |
767 | static DEVICE_ATTR(max_comp_streams, S_IRUGO | S_IWUSR, |
768 | max_comp_streams_show, max_comp_streams_store); | |
e46b8a03 SS |
769 | static DEVICE_ATTR(comp_algorithm, S_IRUGO | S_IWUSR, |
770 | comp_algorithm_show, comp_algorithm_store); | |
9b3bb7ab | 771 | |
a68eb3b6 SS |
772 | ZRAM_ATTR_RO(num_reads); |
773 | ZRAM_ATTR_RO(num_writes); | |
64447249 SS |
774 | ZRAM_ATTR_RO(failed_reads); |
775 | ZRAM_ATTR_RO(failed_writes); | |
a68eb3b6 SS |
776 | ZRAM_ATTR_RO(invalid_io); |
777 | ZRAM_ATTR_RO(notify_free); | |
778 | ZRAM_ATTR_RO(zero_pages); | |
779 | ZRAM_ATTR_RO(compr_data_size); | |
780 | ||
9b3bb7ab SS |
781 | static struct attribute *zram_disk_attrs[] = { |
782 | &dev_attr_disksize.attr, | |
783 | &dev_attr_initstate.attr, | |
784 | &dev_attr_reset.attr, | |
785 | &dev_attr_num_reads.attr, | |
786 | &dev_attr_num_writes.attr, | |
64447249 SS |
787 | &dev_attr_failed_reads.attr, |
788 | &dev_attr_failed_writes.attr, | |
9b3bb7ab SS |
789 | &dev_attr_invalid_io.attr, |
790 | &dev_attr_notify_free.attr, | |
791 | &dev_attr_zero_pages.attr, | |
792 | &dev_attr_orig_data_size.attr, | |
793 | &dev_attr_compr_data_size.attr, | |
794 | &dev_attr_mem_used_total.attr, | |
beca3ec7 | 795 | &dev_attr_max_comp_streams.attr, |
e46b8a03 | 796 | &dev_attr_comp_algorithm.attr, |
9b3bb7ab SS |
797 | NULL, |
798 | }; | |
799 | ||
800 | static struct attribute_group zram_disk_attr_group = { | |
801 | .attrs = zram_disk_attrs, | |
802 | }; | |
803 | ||
f1e3cfff | 804 | static int create_device(struct zram *zram, int device_id) |
306b0c95 | 805 | { |
39a9b8ac | 806 | int ret = -ENOMEM; |
de1a21a0 | 807 | |
0900beae | 808 | init_rwsem(&zram->init_lock); |
306b0c95 | 809 | |
f1e3cfff NG |
810 | zram->queue = blk_alloc_queue(GFP_KERNEL); |
811 | if (!zram->queue) { | |
306b0c95 NG |
812 | pr_err("Error allocating disk queue for device %d\n", |
813 | device_id); | |
de1a21a0 | 814 | goto out; |
306b0c95 NG |
815 | } |
816 | ||
f1e3cfff NG |
817 | blk_queue_make_request(zram->queue, zram_make_request); |
818 | zram->queue->queuedata = zram; | |
306b0c95 NG |
819 | |
820 | /* gendisk structure */ | |
f1e3cfff NG |
821 | zram->disk = alloc_disk(1); |
822 | if (!zram->disk) { | |
94b8435f | 823 | pr_warn("Error allocating disk structure for device %d\n", |
306b0c95 | 824 | device_id); |
39a9b8ac | 825 | goto out_free_queue; |
306b0c95 NG |
826 | } |
827 | ||
f1e3cfff NG |
828 | zram->disk->major = zram_major; |
829 | zram->disk->first_minor = device_id; | |
830 | zram->disk->fops = &zram_devops; | |
831 | zram->disk->queue = zram->queue; | |
832 | zram->disk->private_data = zram; | |
833 | snprintf(zram->disk->disk_name, 16, "zram%d", device_id); | |
306b0c95 | 834 | |
33863c21 | 835 | /* Actual capacity set using syfs (/sys/block/zram<id>/disksize */ |
f1e3cfff | 836 | set_capacity(zram->disk, 0); |
b67d1ec1 SS |
837 | /* zram devices sort of resembles non-rotational disks */ |
838 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue); | |
a1dd52af NG |
839 | /* |
840 | * To ensure that we always get PAGE_SIZE aligned | |
841 | * and n*PAGE_SIZED sized I/O requests. | |
842 | */ | |
f1e3cfff | 843 | blk_queue_physical_block_size(zram->disk->queue, PAGE_SIZE); |
7b19b8d4 RJ |
844 | blk_queue_logical_block_size(zram->disk->queue, |
845 | ZRAM_LOGICAL_BLOCK_SIZE); | |
f1e3cfff NG |
846 | blk_queue_io_min(zram->disk->queue, PAGE_SIZE); |
847 | blk_queue_io_opt(zram->disk->queue, PAGE_SIZE); | |
5d83d5a0 | 848 | |
f1e3cfff | 849 | add_disk(zram->disk); |
306b0c95 | 850 | |
33863c21 NG |
851 | ret = sysfs_create_group(&disk_to_dev(zram->disk)->kobj, |
852 | &zram_disk_attr_group); | |
853 | if (ret < 0) { | |
94b8435f | 854 | pr_warn("Error creating sysfs group"); |
39a9b8ac | 855 | goto out_free_disk; |
33863c21 | 856 | } |
e46b8a03 | 857 | strlcpy(zram->compressor, default_compressor, sizeof(zram->compressor)); |
be2d1d56 | 858 | zram->meta = NULL; |
beca3ec7 | 859 | zram->max_comp_streams = 1; |
39a9b8ac | 860 | return 0; |
de1a21a0 | 861 | |
39a9b8ac JL |
862 | out_free_disk: |
863 | del_gendisk(zram->disk); | |
864 | put_disk(zram->disk); | |
865 | out_free_queue: | |
866 | blk_cleanup_queue(zram->queue); | |
de1a21a0 NG |
867 | out: |
868 | return ret; | |
306b0c95 NG |
869 | } |
870 | ||
f1e3cfff | 871 | static void destroy_device(struct zram *zram) |
306b0c95 | 872 | { |
33863c21 NG |
873 | sysfs_remove_group(&disk_to_dev(zram->disk)->kobj, |
874 | &zram_disk_attr_group); | |
33863c21 | 875 | |
59d3fe54 RK |
876 | del_gendisk(zram->disk); |
877 | put_disk(zram->disk); | |
306b0c95 | 878 | |
59d3fe54 | 879 | blk_cleanup_queue(zram->queue); |
306b0c95 NG |
880 | } |
881 | ||
f1e3cfff | 882 | static int __init zram_init(void) |
306b0c95 | 883 | { |
de1a21a0 | 884 | int ret, dev_id; |
306b0c95 | 885 | |
5fa5a901 | 886 | if (num_devices > max_num_devices) { |
94b8435f | 887 | pr_warn("Invalid value for num_devices: %u\n", |
5fa5a901 | 888 | num_devices); |
de1a21a0 NG |
889 | ret = -EINVAL; |
890 | goto out; | |
306b0c95 NG |
891 | } |
892 | ||
f1e3cfff NG |
893 | zram_major = register_blkdev(0, "zram"); |
894 | if (zram_major <= 0) { | |
94b8435f | 895 | pr_warn("Unable to get major number\n"); |
de1a21a0 NG |
896 | ret = -EBUSY; |
897 | goto out; | |
306b0c95 NG |
898 | } |
899 | ||
306b0c95 | 900 | /* Allocate the device array and initialize each one */ |
5fa5a901 | 901 | zram_devices = kzalloc(num_devices * sizeof(struct zram), GFP_KERNEL); |
43801f6e | 902 | if (!zram_devices) { |
de1a21a0 NG |
903 | ret = -ENOMEM; |
904 | goto unregister; | |
905 | } | |
306b0c95 | 906 | |
5fa5a901 | 907 | for (dev_id = 0; dev_id < num_devices; dev_id++) { |
43801f6e | 908 | ret = create_device(&zram_devices[dev_id], dev_id); |
de1a21a0 | 909 | if (ret) |
3bf040c7 | 910 | goto free_devices; |
de1a21a0 NG |
911 | } |
912 | ||
ca3d70bd DB |
913 | pr_info("Created %u device(s) ...\n", num_devices); |
914 | ||
306b0c95 | 915 | return 0; |
de1a21a0 | 916 | |
3bf040c7 | 917 | free_devices: |
de1a21a0 | 918 | while (dev_id) |
43801f6e NW |
919 | destroy_device(&zram_devices[--dev_id]); |
920 | kfree(zram_devices); | |
de1a21a0 | 921 | unregister: |
f1e3cfff | 922 | unregister_blkdev(zram_major, "zram"); |
de1a21a0 | 923 | out: |
306b0c95 NG |
924 | return ret; |
925 | } | |
926 | ||
f1e3cfff | 927 | static void __exit zram_exit(void) |
306b0c95 NG |
928 | { |
929 | int i; | |
f1e3cfff | 930 | struct zram *zram; |
306b0c95 | 931 | |
5fa5a901 | 932 | for (i = 0; i < num_devices; i++) { |
43801f6e | 933 | zram = &zram_devices[i]; |
306b0c95 | 934 | |
f1e3cfff | 935 | destroy_device(zram); |
2b86ab9c MK |
936 | /* |
937 | * Shouldn't access zram->disk after destroy_device | |
938 | * because destroy_device already released zram->disk. | |
939 | */ | |
940 | zram_reset_device(zram, false); | |
306b0c95 NG |
941 | } |
942 | ||
f1e3cfff | 943 | unregister_blkdev(zram_major, "zram"); |
306b0c95 | 944 | |
43801f6e | 945 | kfree(zram_devices); |
306b0c95 NG |
946 | pr_debug("Cleanup done!\n"); |
947 | } | |
948 | ||
f1e3cfff NG |
949 | module_init(zram_init); |
950 | module_exit(zram_exit); | |
306b0c95 | 951 | |
9b3bb7ab SS |
952 | module_param(num_devices, uint, 0); |
953 | MODULE_PARM_DESC(num_devices, "Number of zram devices"); | |
954 | ||
306b0c95 NG |
955 | MODULE_LICENSE("Dual BSD/GPL"); |
956 | MODULE_AUTHOR("Nitin Gupta <ngupta@vflare.org>"); | |
f1e3cfff | 957 | MODULE_DESCRIPTION("Compressed RAM Block Device"); |