]>
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> |
fcfa8d95 | 34 | #include <linux/err.h> |
306b0c95 | 35 | |
16a4bfb9 | 36 | #include "zram_drv.h" |
306b0c95 NG |
37 | |
38 | /* Globals */ | |
f1e3cfff | 39 | static int zram_major; |
0f0e3ba3 | 40 | static struct zram *zram_devices; |
b7ca232e | 41 | static const char *default_compressor = "lzo"; |
306b0c95 | 42 | |
306b0c95 | 43 | /* Module params (documentation at end) */ |
ca3d70bd | 44 | static unsigned int num_devices = 1; |
33863c21 | 45 | |
a68eb3b6 SS |
46 | #define ZRAM_ATTR_RO(name) \ |
47 | static ssize_t zram_attr_##name##_show(struct device *d, \ | |
48 | struct device_attribute *attr, char *b) \ | |
49 | { \ | |
50 | struct zram *zram = dev_to_zram(d); \ | |
56b4e8cb | 51 | return scnprintf(b, PAGE_SIZE, "%llu\n", \ |
a68eb3b6 SS |
52 | (u64)atomic64_read(&zram->stats.name)); \ |
53 | } \ | |
54 | static struct device_attribute dev_attr_##name = \ | |
55 | __ATTR(name, S_IRUGO, zram_attr_##name##_show, NULL); | |
56 | ||
be2d1d56 SS |
57 | static inline int init_done(struct zram *zram) |
58 | { | |
59 | return zram->meta != NULL; | |
60 | } | |
61 | ||
9b3bb7ab SS |
62 | static inline struct zram *dev_to_zram(struct device *dev) |
63 | { | |
64 | return (struct zram *)dev_to_disk(dev)->private_data; | |
65 | } | |
66 | ||
67 | static ssize_t disksize_show(struct device *dev, | |
68 | struct device_attribute *attr, char *buf) | |
69 | { | |
70 | struct zram *zram = dev_to_zram(dev); | |
71 | ||
56b4e8cb | 72 | return scnprintf(buf, PAGE_SIZE, "%llu\n", zram->disksize); |
9b3bb7ab SS |
73 | } |
74 | ||
75 | static ssize_t initstate_show(struct device *dev, | |
76 | struct device_attribute *attr, char *buf) | |
77 | { | |
a68eb3b6 | 78 | u32 val; |
9b3bb7ab SS |
79 | struct zram *zram = dev_to_zram(dev); |
80 | ||
a68eb3b6 SS |
81 | down_read(&zram->init_lock); |
82 | val = init_done(zram); | |
83 | up_read(&zram->init_lock); | |
9b3bb7ab | 84 | |
56b4e8cb | 85 | return scnprintf(buf, PAGE_SIZE, "%u\n", val); |
9b3bb7ab SS |
86 | } |
87 | ||
88 | static ssize_t orig_data_size_show(struct device *dev, | |
89 | struct device_attribute *attr, char *buf) | |
90 | { | |
91 | struct zram *zram = dev_to_zram(dev); | |
92 | ||
56b4e8cb | 93 | return scnprintf(buf, PAGE_SIZE, "%llu\n", |
90a7806e | 94 | (u64)(atomic64_read(&zram->stats.pages_stored)) << PAGE_SHIFT); |
9b3bb7ab SS |
95 | } |
96 | ||
9b3bb7ab SS |
97 | static ssize_t mem_used_total_show(struct device *dev, |
98 | struct device_attribute *attr, char *buf) | |
99 | { | |
100 | u64 val = 0; | |
101 | struct zram *zram = dev_to_zram(dev); | |
102 | struct zram_meta *meta = zram->meta; | |
103 | ||
104 | down_read(&zram->init_lock); | |
be2d1d56 | 105 | if (init_done(zram)) |
722cdc17 | 106 | val = zs_get_total_pages(meta->mem_pool); |
9b3bb7ab SS |
107 | up_read(&zram->init_lock); |
108 | ||
722cdc17 | 109 | return scnprintf(buf, PAGE_SIZE, "%llu\n", val << PAGE_SHIFT); |
9b3bb7ab SS |
110 | } |
111 | ||
beca3ec7 SS |
112 | static ssize_t max_comp_streams_show(struct device *dev, |
113 | struct device_attribute *attr, char *buf) | |
114 | { | |
115 | int val; | |
116 | struct zram *zram = dev_to_zram(dev); | |
117 | ||
118 | down_read(&zram->init_lock); | |
119 | val = zram->max_comp_streams; | |
120 | up_read(&zram->init_lock); | |
121 | ||
56b4e8cb | 122 | return scnprintf(buf, PAGE_SIZE, "%d\n", val); |
beca3ec7 SS |
123 | } |
124 | ||
9ada9da9 MK |
125 | static ssize_t mem_limit_show(struct device *dev, |
126 | struct device_attribute *attr, char *buf) | |
127 | { | |
128 | u64 val; | |
129 | struct zram *zram = dev_to_zram(dev); | |
130 | ||
131 | down_read(&zram->init_lock); | |
132 | val = zram->limit_pages; | |
133 | up_read(&zram->init_lock); | |
134 | ||
135 | return scnprintf(buf, PAGE_SIZE, "%llu\n", val << PAGE_SHIFT); | |
136 | } | |
137 | ||
138 | static ssize_t mem_limit_store(struct device *dev, | |
139 | struct device_attribute *attr, const char *buf, size_t len) | |
140 | { | |
141 | u64 limit; | |
142 | char *tmp; | |
143 | struct zram *zram = dev_to_zram(dev); | |
144 | ||
145 | limit = memparse(buf, &tmp); | |
146 | if (buf == tmp) /* no chars parsed, invalid input */ | |
147 | return -EINVAL; | |
148 | ||
149 | down_write(&zram->init_lock); | |
150 | zram->limit_pages = PAGE_ALIGN(limit) >> PAGE_SHIFT; | |
151 | up_write(&zram->init_lock); | |
152 | ||
153 | return len; | |
154 | } | |
155 | ||
461a8eee MK |
156 | static ssize_t mem_used_max_show(struct device *dev, |
157 | struct device_attribute *attr, char *buf) | |
158 | { | |
159 | u64 val = 0; | |
160 | struct zram *zram = dev_to_zram(dev); | |
161 | ||
162 | down_read(&zram->init_lock); | |
163 | if (init_done(zram)) | |
164 | val = atomic_long_read(&zram->stats.max_used_pages); | |
165 | up_read(&zram->init_lock); | |
166 | ||
167 | return scnprintf(buf, PAGE_SIZE, "%llu\n", val << PAGE_SHIFT); | |
168 | } | |
169 | ||
170 | static ssize_t mem_used_max_store(struct device *dev, | |
171 | struct device_attribute *attr, const char *buf, size_t len) | |
172 | { | |
173 | int err; | |
174 | unsigned long val; | |
175 | struct zram *zram = dev_to_zram(dev); | |
176 | struct zram_meta *meta = zram->meta; | |
177 | ||
178 | err = kstrtoul(buf, 10, &val); | |
179 | if (err || val != 0) | |
180 | return -EINVAL; | |
181 | ||
182 | down_read(&zram->init_lock); | |
183 | if (init_done(zram)) | |
184 | atomic_long_set(&zram->stats.max_used_pages, | |
185 | zs_get_total_pages(meta->mem_pool)); | |
186 | up_read(&zram->init_lock); | |
187 | ||
188 | return len; | |
189 | } | |
190 | ||
beca3ec7 SS |
191 | static ssize_t max_comp_streams_store(struct device *dev, |
192 | struct device_attribute *attr, const char *buf, size_t len) | |
193 | { | |
194 | int num; | |
195 | struct zram *zram = dev_to_zram(dev); | |
60a726e3 | 196 | int ret; |
beca3ec7 | 197 | |
60a726e3 MK |
198 | ret = kstrtoint(buf, 0, &num); |
199 | if (ret < 0) | |
200 | return ret; | |
beca3ec7 SS |
201 | if (num < 1) |
202 | return -EINVAL; | |
60a726e3 | 203 | |
beca3ec7 SS |
204 | down_write(&zram->init_lock); |
205 | if (init_done(zram)) { | |
60a726e3 | 206 | if (!zcomp_set_max_streams(zram->comp, num)) { |
fe8eb122 | 207 | pr_info("Cannot change max compression streams\n"); |
60a726e3 MK |
208 | ret = -EINVAL; |
209 | goto out; | |
210 | } | |
beca3ec7 | 211 | } |
60a726e3 | 212 | |
beca3ec7 | 213 | zram->max_comp_streams = num; |
60a726e3 MK |
214 | ret = len; |
215 | out: | |
beca3ec7 | 216 | up_write(&zram->init_lock); |
60a726e3 | 217 | return ret; |
beca3ec7 SS |
218 | } |
219 | ||
e46b8a03 SS |
220 | static ssize_t comp_algorithm_show(struct device *dev, |
221 | struct device_attribute *attr, char *buf) | |
222 | { | |
223 | size_t sz; | |
224 | struct zram *zram = dev_to_zram(dev); | |
225 | ||
226 | down_read(&zram->init_lock); | |
227 | sz = zcomp_available_show(zram->compressor, buf); | |
228 | up_read(&zram->init_lock); | |
229 | ||
230 | return sz; | |
231 | } | |
232 | ||
233 | static ssize_t comp_algorithm_store(struct device *dev, | |
234 | struct device_attribute *attr, const char *buf, size_t len) | |
235 | { | |
236 | struct zram *zram = dev_to_zram(dev); | |
237 | down_write(&zram->init_lock); | |
238 | if (init_done(zram)) { | |
239 | up_write(&zram->init_lock); | |
240 | pr_info("Can't change algorithm for initialized device\n"); | |
241 | return -EBUSY; | |
242 | } | |
243 | strlcpy(zram->compressor, buf, sizeof(zram->compressor)); | |
244 | up_write(&zram->init_lock); | |
245 | return len; | |
246 | } | |
247 | ||
92967471 | 248 | /* flag operations needs meta->tb_lock */ |
8b3cc3ed | 249 | static int zram_test_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 250 | enum zram_pageflags flag) |
306b0c95 | 251 | { |
d2d5e762 | 252 | return meta->table[index].value & BIT(flag); |
306b0c95 NG |
253 | } |
254 | ||
8b3cc3ed | 255 | static void zram_set_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 256 | enum zram_pageflags flag) |
306b0c95 | 257 | { |
d2d5e762 | 258 | meta->table[index].value |= BIT(flag); |
306b0c95 NG |
259 | } |
260 | ||
8b3cc3ed | 261 | static void zram_clear_flag(struct zram_meta *meta, u32 index, |
f1e3cfff | 262 | enum zram_pageflags flag) |
306b0c95 | 263 | { |
d2d5e762 WY |
264 | meta->table[index].value &= ~BIT(flag); |
265 | } | |
266 | ||
267 | static size_t zram_get_obj_size(struct zram_meta *meta, u32 index) | |
268 | { | |
269 | return meta->table[index].value & (BIT(ZRAM_FLAG_SHIFT) - 1); | |
270 | } | |
271 | ||
272 | static void zram_set_obj_size(struct zram_meta *meta, | |
273 | u32 index, size_t size) | |
274 | { | |
275 | unsigned long flags = meta->table[index].value >> ZRAM_FLAG_SHIFT; | |
276 | ||
277 | meta->table[index].value = (flags << ZRAM_FLAG_SHIFT) | size; | |
306b0c95 NG |
278 | } |
279 | ||
9b3bb7ab SS |
280 | static inline int is_partial_io(struct bio_vec *bvec) |
281 | { | |
282 | return bvec->bv_len != PAGE_SIZE; | |
283 | } | |
284 | ||
285 | /* | |
286 | * Check if request is within bounds and aligned on zram logical blocks. | |
287 | */ | |
288 | static inline int valid_io_request(struct zram *zram, struct bio *bio) | |
289 | { | |
290 | u64 start, end, bound; | |
a539c72a | 291 | |
9b3bb7ab | 292 | /* unaligned request */ |
4f024f37 KO |
293 | if (unlikely(bio->bi_iter.bi_sector & |
294 | (ZRAM_SECTOR_PER_LOGICAL_BLOCK - 1))) | |
9b3bb7ab | 295 | return 0; |
4f024f37 | 296 | if (unlikely(bio->bi_iter.bi_size & (ZRAM_LOGICAL_BLOCK_SIZE - 1))) |
9b3bb7ab SS |
297 | return 0; |
298 | ||
4f024f37 KO |
299 | start = bio->bi_iter.bi_sector; |
300 | end = start + (bio->bi_iter.bi_size >> SECTOR_SHIFT); | |
9b3bb7ab SS |
301 | bound = zram->disksize >> SECTOR_SHIFT; |
302 | /* out of range range */ | |
75c7caf5 | 303 | if (unlikely(start >= bound || end > bound || start > end)) |
9b3bb7ab SS |
304 | return 0; |
305 | ||
306 | /* I/O request is valid */ | |
307 | return 1; | |
308 | } | |
309 | ||
310 | static void zram_meta_free(struct zram_meta *meta) | |
311 | { | |
312 | zs_destroy_pool(meta->mem_pool); | |
9b3bb7ab SS |
313 | vfree(meta->table); |
314 | kfree(meta); | |
315 | } | |
316 | ||
317 | static struct zram_meta *zram_meta_alloc(u64 disksize) | |
318 | { | |
319 | size_t num_pages; | |
320 | struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL); | |
321 | if (!meta) | |
322 | goto out; | |
323 | ||
9b3bb7ab SS |
324 | num_pages = disksize >> PAGE_SHIFT; |
325 | meta->table = vzalloc(num_pages * sizeof(*meta->table)); | |
326 | if (!meta->table) { | |
327 | pr_err("Error allocating zram address table\n"); | |
b7ca232e | 328 | goto free_meta; |
9b3bb7ab SS |
329 | } |
330 | ||
331 | meta->mem_pool = zs_create_pool(GFP_NOIO | __GFP_HIGHMEM); | |
332 | if (!meta->mem_pool) { | |
333 | pr_err("Error creating memory pool\n"); | |
334 | goto free_table; | |
335 | } | |
336 | ||
337 | return meta; | |
338 | ||
339 | free_table: | |
340 | vfree(meta->table); | |
9b3bb7ab SS |
341 | free_meta: |
342 | kfree(meta); | |
343 | meta = NULL; | |
344 | out: | |
345 | return meta; | |
346 | } | |
347 | ||
348 | static void update_position(u32 *index, int *offset, struct bio_vec *bvec) | |
349 | { | |
350 | if (*offset + bvec->bv_len >= PAGE_SIZE) | |
351 | (*index)++; | |
352 | *offset = (*offset + bvec->bv_len) % PAGE_SIZE; | |
353 | } | |
354 | ||
306b0c95 NG |
355 | static int page_zero_filled(void *ptr) |
356 | { | |
357 | unsigned int pos; | |
358 | unsigned long *page; | |
359 | ||
360 | page = (unsigned long *)ptr; | |
361 | ||
362 | for (pos = 0; pos != PAGE_SIZE / sizeof(*page); pos++) { | |
363 | if (page[pos]) | |
364 | return 0; | |
365 | } | |
366 | ||
367 | return 1; | |
368 | } | |
369 | ||
9b3bb7ab SS |
370 | static void handle_zero_page(struct bio_vec *bvec) |
371 | { | |
372 | struct page *page = bvec->bv_page; | |
373 | void *user_mem; | |
374 | ||
375 | user_mem = kmap_atomic(page); | |
376 | if (is_partial_io(bvec)) | |
377 | memset(user_mem + bvec->bv_offset, 0, bvec->bv_len); | |
378 | else | |
379 | clear_page(user_mem); | |
380 | kunmap_atomic(user_mem); | |
381 | ||
382 | flush_dcache_page(page); | |
383 | } | |
384 | ||
d2d5e762 WY |
385 | |
386 | /* | |
387 | * To protect concurrent access to the same index entry, | |
388 | * caller should hold this table index entry's bit_spinlock to | |
389 | * indicate this index entry is accessing. | |
390 | */ | |
f1e3cfff | 391 | static void zram_free_page(struct zram *zram, size_t index) |
306b0c95 | 392 | { |
8b3cc3ed MK |
393 | struct zram_meta *meta = zram->meta; |
394 | unsigned long handle = meta->table[index].handle; | |
306b0c95 | 395 | |
fd1a30de | 396 | if (unlikely(!handle)) { |
2e882281 NG |
397 | /* |
398 | * No memory is allocated for zero filled pages. | |
399 | * Simply clear zero page flag. | |
400 | */ | |
8b3cc3ed MK |
401 | if (zram_test_flag(meta, index, ZRAM_ZERO)) { |
402 | zram_clear_flag(meta, index, ZRAM_ZERO); | |
90a7806e | 403 | atomic64_dec(&zram->stats.zero_pages); |
306b0c95 NG |
404 | } |
405 | return; | |
406 | } | |
407 | ||
8b3cc3ed | 408 | zs_free(meta->mem_pool, handle); |
306b0c95 | 409 | |
d2d5e762 WY |
410 | atomic64_sub(zram_get_obj_size(meta, index), |
411 | &zram->stats.compr_data_size); | |
90a7806e | 412 | atomic64_dec(&zram->stats.pages_stored); |
306b0c95 | 413 | |
8b3cc3ed | 414 | meta->table[index].handle = 0; |
d2d5e762 | 415 | zram_set_obj_size(meta, index, 0); |
306b0c95 NG |
416 | } |
417 | ||
37b51fdd | 418 | static int zram_decompress_page(struct zram *zram, char *mem, u32 index) |
306b0c95 | 419 | { |
b7ca232e | 420 | int ret = 0; |
37b51fdd | 421 | unsigned char *cmem; |
8b3cc3ed | 422 | struct zram_meta *meta = zram->meta; |
92967471 | 423 | unsigned long handle; |
023b409f | 424 | size_t size; |
92967471 | 425 | |
d2d5e762 | 426 | bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value); |
92967471 | 427 | handle = meta->table[index].handle; |
d2d5e762 | 428 | size = zram_get_obj_size(meta, index); |
306b0c95 | 429 | |
8b3cc3ed | 430 | if (!handle || zram_test_flag(meta, index, ZRAM_ZERO)) { |
d2d5e762 | 431 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
42e99bd9 | 432 | clear_page(mem); |
8c921b2b JM |
433 | return 0; |
434 | } | |
306b0c95 | 435 | |
8b3cc3ed | 436 | cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_RO); |
92967471 | 437 | if (size == PAGE_SIZE) |
42e99bd9 | 438 | copy_page(mem, cmem); |
37b51fdd | 439 | else |
b7ca232e | 440 | ret = zcomp_decompress(zram->comp, cmem, size, mem); |
8b3cc3ed | 441 | zs_unmap_object(meta->mem_pool, handle); |
d2d5e762 | 442 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
a1dd52af | 443 | |
8c921b2b | 444 | /* Should NEVER happen. Return bio error if it does. */ |
b7ca232e | 445 | if (unlikely(ret)) { |
8c921b2b | 446 | pr_err("Decompression failed! err=%d, page=%u\n", ret, index); |
8c921b2b | 447 | return ret; |
a1dd52af | 448 | } |
306b0c95 | 449 | |
8c921b2b | 450 | return 0; |
306b0c95 NG |
451 | } |
452 | ||
37b51fdd SS |
453 | static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec, |
454 | u32 index, int offset, struct bio *bio) | |
924bd88d JM |
455 | { |
456 | int ret; | |
37b51fdd SS |
457 | struct page *page; |
458 | unsigned char *user_mem, *uncmem = NULL; | |
8b3cc3ed | 459 | struct zram_meta *meta = zram->meta; |
37b51fdd SS |
460 | page = bvec->bv_page; |
461 | ||
d2d5e762 | 462 | bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value); |
8b3cc3ed MK |
463 | if (unlikely(!meta->table[index].handle) || |
464 | zram_test_flag(meta, index, ZRAM_ZERO)) { | |
d2d5e762 | 465 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
37b51fdd | 466 | handle_zero_page(bvec); |
924bd88d JM |
467 | return 0; |
468 | } | |
d2d5e762 | 469 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
924bd88d | 470 | |
37b51fdd SS |
471 | if (is_partial_io(bvec)) |
472 | /* Use a temporary buffer to decompress the page */ | |
7e5a5104 MK |
473 | uncmem = kmalloc(PAGE_SIZE, GFP_NOIO); |
474 | ||
475 | user_mem = kmap_atomic(page); | |
476 | if (!is_partial_io(bvec)) | |
37b51fdd SS |
477 | uncmem = user_mem; |
478 | ||
479 | if (!uncmem) { | |
480 | pr_info("Unable to allocate temp memory\n"); | |
481 | ret = -ENOMEM; | |
482 | goto out_cleanup; | |
483 | } | |
924bd88d | 484 | |
37b51fdd | 485 | ret = zram_decompress_page(zram, uncmem, index); |
924bd88d | 486 | /* Should NEVER happen. Return bio error if it does. */ |
b7ca232e | 487 | if (unlikely(ret)) |
37b51fdd | 488 | goto out_cleanup; |
924bd88d | 489 | |
37b51fdd SS |
490 | if (is_partial_io(bvec)) |
491 | memcpy(user_mem + bvec->bv_offset, uncmem + offset, | |
492 | bvec->bv_len); | |
493 | ||
494 | flush_dcache_page(page); | |
495 | ret = 0; | |
496 | out_cleanup: | |
497 | kunmap_atomic(user_mem); | |
498 | if (is_partial_io(bvec)) | |
499 | kfree(uncmem); | |
500 | return ret; | |
924bd88d JM |
501 | } |
502 | ||
461a8eee MK |
503 | static inline void update_used_max(struct zram *zram, |
504 | const unsigned long pages) | |
505 | { | |
506 | int old_max, cur_max; | |
507 | ||
508 | old_max = atomic_long_read(&zram->stats.max_used_pages); | |
509 | ||
510 | do { | |
511 | cur_max = old_max; | |
512 | if (pages > cur_max) | |
513 | old_max = atomic_long_cmpxchg( | |
514 | &zram->stats.max_used_pages, cur_max, pages); | |
515 | } while (old_max != cur_max); | |
516 | } | |
517 | ||
924bd88d JM |
518 | static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index, |
519 | int offset) | |
306b0c95 | 520 | { |
397c6066 | 521 | int ret = 0; |
8c921b2b | 522 | size_t clen; |
c2344348 | 523 | unsigned long handle; |
130f315a | 524 | struct page *page; |
924bd88d | 525 | unsigned char *user_mem, *cmem, *src, *uncmem = NULL; |
8b3cc3ed | 526 | struct zram_meta *meta = zram->meta; |
b7ca232e | 527 | struct zcomp_strm *zstrm; |
e46e3315 | 528 | bool locked = false; |
461a8eee | 529 | unsigned long alloced_pages; |
306b0c95 | 530 | |
8c921b2b | 531 | page = bvec->bv_page; |
924bd88d JM |
532 | if (is_partial_io(bvec)) { |
533 | /* | |
534 | * This is a partial IO. We need to read the full page | |
535 | * before to write the changes. | |
536 | */ | |
7e5a5104 | 537 | uncmem = kmalloc(PAGE_SIZE, GFP_NOIO); |
924bd88d | 538 | if (!uncmem) { |
924bd88d JM |
539 | ret = -ENOMEM; |
540 | goto out; | |
541 | } | |
37b51fdd | 542 | ret = zram_decompress_page(zram, uncmem, index); |
397c6066 | 543 | if (ret) |
924bd88d | 544 | goto out; |
924bd88d JM |
545 | } |
546 | ||
b7ca232e | 547 | zstrm = zcomp_strm_find(zram->comp); |
e46e3315 | 548 | locked = true; |
ba82fe2e | 549 | user_mem = kmap_atomic(page); |
924bd88d | 550 | |
397c6066 | 551 | if (is_partial_io(bvec)) { |
924bd88d JM |
552 | memcpy(uncmem + offset, user_mem + bvec->bv_offset, |
553 | bvec->bv_len); | |
397c6066 NG |
554 | kunmap_atomic(user_mem); |
555 | user_mem = NULL; | |
556 | } else { | |
924bd88d | 557 | uncmem = user_mem; |
397c6066 | 558 | } |
924bd88d JM |
559 | |
560 | if (page_zero_filled(uncmem)) { | |
ba82fe2e | 561 | kunmap_atomic(user_mem); |
f40ac2ae | 562 | /* Free memory associated with this sector now. */ |
d2d5e762 | 563 | bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value); |
f40ac2ae | 564 | zram_free_page(zram, index); |
92967471 | 565 | zram_set_flag(meta, index, ZRAM_ZERO); |
d2d5e762 | 566 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
f40ac2ae | 567 | |
90a7806e | 568 | atomic64_inc(&zram->stats.zero_pages); |
924bd88d JM |
569 | ret = 0; |
570 | goto out; | |
8c921b2b | 571 | } |
306b0c95 | 572 | |
b7ca232e | 573 | ret = zcomp_compress(zram->comp, zstrm, uncmem, &clen); |
397c6066 NG |
574 | if (!is_partial_io(bvec)) { |
575 | kunmap_atomic(user_mem); | |
576 | user_mem = NULL; | |
577 | uncmem = NULL; | |
578 | } | |
306b0c95 | 579 | |
b7ca232e | 580 | if (unlikely(ret)) { |
8c921b2b | 581 | pr_err("Compression failed! err=%d\n", ret); |
924bd88d | 582 | goto out; |
8c921b2b | 583 | } |
b7ca232e | 584 | src = zstrm->buffer; |
c8f2f0db | 585 | if (unlikely(clen > max_zpage_size)) { |
c8f2f0db | 586 | clen = PAGE_SIZE; |
397c6066 NG |
587 | if (is_partial_io(bvec)) |
588 | src = uncmem; | |
c8f2f0db | 589 | } |
a1dd52af | 590 | |
8b3cc3ed | 591 | handle = zs_malloc(meta->mem_pool, clen); |
fd1a30de | 592 | if (!handle) { |
596b3dd4 MR |
593 | pr_info("Error allocating memory for compressed page: %u, size=%zu\n", |
594 | index, clen); | |
924bd88d JM |
595 | ret = -ENOMEM; |
596 | goto out; | |
8c921b2b | 597 | } |
9ada9da9 | 598 | |
461a8eee MK |
599 | alloced_pages = zs_get_total_pages(meta->mem_pool); |
600 | if (zram->limit_pages && alloced_pages > zram->limit_pages) { | |
9ada9da9 MK |
601 | zs_free(meta->mem_pool, handle); |
602 | ret = -ENOMEM; | |
603 | goto out; | |
604 | } | |
605 | ||
461a8eee MK |
606 | update_used_max(zram, alloced_pages); |
607 | ||
8b3cc3ed | 608 | cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO); |
306b0c95 | 609 | |
42e99bd9 | 610 | if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) { |
397c6066 | 611 | src = kmap_atomic(page); |
42e99bd9 | 612 | copy_page(cmem, src); |
397c6066 | 613 | kunmap_atomic(src); |
42e99bd9 JL |
614 | } else { |
615 | memcpy(cmem, src, clen); | |
616 | } | |
306b0c95 | 617 | |
b7ca232e SS |
618 | zcomp_strm_release(zram->comp, zstrm); |
619 | locked = false; | |
8b3cc3ed | 620 | zs_unmap_object(meta->mem_pool, handle); |
fd1a30de | 621 | |
f40ac2ae SS |
622 | /* |
623 | * Free memory associated with this sector | |
624 | * before overwriting unused sectors. | |
625 | */ | |
d2d5e762 | 626 | bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value); |
f40ac2ae SS |
627 | zram_free_page(zram, index); |
628 | ||
8b3cc3ed | 629 | meta->table[index].handle = handle; |
d2d5e762 WY |
630 | zram_set_obj_size(meta, index, clen); |
631 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); | |
306b0c95 | 632 | |
8c921b2b | 633 | /* Update stats */ |
90a7806e SS |
634 | atomic64_add(clen, &zram->stats.compr_data_size); |
635 | atomic64_inc(&zram->stats.pages_stored); | |
924bd88d | 636 | out: |
e46e3315 | 637 | if (locked) |
b7ca232e | 638 | zcomp_strm_release(zram->comp, zstrm); |
397c6066 NG |
639 | if (is_partial_io(bvec)) |
640 | kfree(uncmem); | |
924bd88d | 641 | return ret; |
8c921b2b JM |
642 | } |
643 | ||
644 | static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index, | |
be257c61 | 645 | int offset, struct bio *bio) |
8c921b2b | 646 | { |
c5bde238 | 647 | int ret; |
be257c61 | 648 | int rw = bio_data_dir(bio); |
8c921b2b | 649 | |
be257c61 SS |
650 | if (rw == READ) { |
651 | atomic64_inc(&zram->stats.num_reads); | |
c5bde238 | 652 | ret = zram_bvec_read(zram, bvec, index, offset, bio); |
be257c61 SS |
653 | } else { |
654 | atomic64_inc(&zram->stats.num_writes); | |
c5bde238 | 655 | ret = zram_bvec_write(zram, bvec, index, offset); |
be257c61 | 656 | } |
c5bde238 | 657 | |
0cf1e9d6 CY |
658 | if (unlikely(ret)) { |
659 | if (rw == READ) | |
660 | atomic64_inc(&zram->stats.failed_reads); | |
661 | else | |
662 | atomic64_inc(&zram->stats.failed_writes); | |
663 | } | |
664 | ||
c5bde238 | 665 | return ret; |
924bd88d JM |
666 | } |
667 | ||
f4659d8e JK |
668 | /* |
669 | * zram_bio_discard - handler on discard request | |
670 | * @index: physical block index in PAGE_SIZE units | |
671 | * @offset: byte offset within physical block | |
672 | */ | |
673 | static void zram_bio_discard(struct zram *zram, u32 index, | |
674 | int offset, struct bio *bio) | |
675 | { | |
676 | size_t n = bio->bi_iter.bi_size; | |
d2d5e762 | 677 | struct zram_meta *meta = zram->meta; |
f4659d8e JK |
678 | |
679 | /* | |
680 | * zram manages data in physical block size units. Because logical block | |
681 | * size isn't identical with physical block size on some arch, we | |
682 | * could get a discard request pointing to a specific offset within a | |
683 | * certain physical block. Although we can handle this request by | |
684 | * reading that physiclal block and decompressing and partially zeroing | |
685 | * and re-compressing and then re-storing it, this isn't reasonable | |
686 | * because our intent with a discard request is to save memory. So | |
687 | * skipping this logical block is appropriate here. | |
688 | */ | |
689 | if (offset) { | |
38515c73 | 690 | if (n <= (PAGE_SIZE - offset)) |
f4659d8e JK |
691 | return; |
692 | ||
38515c73 | 693 | n -= (PAGE_SIZE - offset); |
f4659d8e JK |
694 | index++; |
695 | } | |
696 | ||
697 | while (n >= PAGE_SIZE) { | |
d2d5e762 | 698 | bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value); |
f4659d8e | 699 | zram_free_page(zram, index); |
d2d5e762 | 700 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
015254da | 701 | atomic64_inc(&zram->stats.notify_free); |
f4659d8e JK |
702 | index++; |
703 | n -= PAGE_SIZE; | |
704 | } | |
705 | } | |
706 | ||
2b86ab9c | 707 | static void zram_reset_device(struct zram *zram, bool reset_capacity) |
924bd88d | 708 | { |
9b3bb7ab SS |
709 | size_t index; |
710 | struct zram_meta *meta; | |
711 | ||
644d4787 | 712 | down_write(&zram->init_lock); |
9ada9da9 MK |
713 | |
714 | zram->limit_pages = 0; | |
715 | ||
be2d1d56 | 716 | if (!init_done(zram)) { |
644d4787 | 717 | up_write(&zram->init_lock); |
9b3bb7ab | 718 | return; |
644d4787 | 719 | } |
9b3bb7ab SS |
720 | |
721 | meta = zram->meta; | |
9b3bb7ab SS |
722 | /* Free all pages that are still in this zram device */ |
723 | for (index = 0; index < zram->disksize >> PAGE_SHIFT; index++) { | |
724 | unsigned long handle = meta->table[index].handle; | |
725 | if (!handle) | |
726 | continue; | |
727 | ||
728 | zs_free(meta->mem_pool, handle); | |
729 | } | |
730 | ||
b7ca232e | 731 | zcomp_destroy(zram->comp); |
beca3ec7 SS |
732 | zram->max_comp_streams = 1; |
733 | ||
9b3bb7ab SS |
734 | zram_meta_free(zram->meta); |
735 | zram->meta = NULL; | |
736 | /* Reset stats */ | |
737 | memset(&zram->stats, 0, sizeof(zram->stats)); | |
738 | ||
739 | zram->disksize = 0; | |
b4c5c609 | 740 | if (reset_capacity) |
2b86ab9c | 741 | set_capacity(zram->disk, 0); |
b4c5c609 | 742 | |
644d4787 | 743 | up_write(&zram->init_lock); |
b4c5c609 MK |
744 | |
745 | /* | |
746 | * Revalidate disk out of the init_lock to avoid lockdep splat. | |
747 | * It's okay because disk's capacity is protected by init_lock | |
748 | * so that revalidate_disk always sees up-to-date capacity. | |
749 | */ | |
750 | if (reset_capacity) | |
751 | revalidate_disk(zram->disk); | |
9b3bb7ab SS |
752 | } |
753 | ||
9b3bb7ab SS |
754 | static ssize_t disksize_store(struct device *dev, |
755 | struct device_attribute *attr, const char *buf, size_t len) | |
756 | { | |
757 | u64 disksize; | |
d61f98c7 | 758 | struct zcomp *comp; |
9b3bb7ab SS |
759 | struct zram_meta *meta; |
760 | struct zram *zram = dev_to_zram(dev); | |
fcfa8d95 | 761 | int err; |
9b3bb7ab SS |
762 | |
763 | disksize = memparse(buf, NULL); | |
764 | if (!disksize) | |
765 | return -EINVAL; | |
766 | ||
767 | disksize = PAGE_ALIGN(disksize); | |
768 | meta = zram_meta_alloc(disksize); | |
db5d711e MK |
769 | if (!meta) |
770 | return -ENOMEM; | |
b67d1ec1 | 771 | |
d61f98c7 | 772 | comp = zcomp_create(zram->compressor, zram->max_comp_streams); |
fcfa8d95 | 773 | if (IS_ERR(comp)) { |
d61f98c7 SS |
774 | pr_info("Cannot initialise %s compressing backend\n", |
775 | zram->compressor); | |
fcfa8d95 SS |
776 | err = PTR_ERR(comp); |
777 | goto out_free_meta; | |
d61f98c7 SS |
778 | } |
779 | ||
9b3bb7ab | 780 | down_write(&zram->init_lock); |
be2d1d56 | 781 | if (init_done(zram)) { |
9b3bb7ab | 782 | pr_info("Cannot change disksize for initialized device\n"); |
b7ca232e | 783 | err = -EBUSY; |
fcfa8d95 | 784 | goto out_destroy_comp; |
9b3bb7ab SS |
785 | } |
786 | ||
b67d1ec1 | 787 | zram->meta = meta; |
d61f98c7 | 788 | zram->comp = comp; |
9b3bb7ab SS |
789 | zram->disksize = disksize; |
790 | set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT); | |
9b3bb7ab | 791 | up_write(&zram->init_lock); |
b4c5c609 MK |
792 | |
793 | /* | |
794 | * Revalidate disk out of the init_lock to avoid lockdep splat. | |
795 | * It's okay because disk's capacity is protected by init_lock | |
796 | * so that revalidate_disk always sees up-to-date capacity. | |
797 | */ | |
798 | revalidate_disk(zram->disk); | |
799 | ||
9b3bb7ab | 800 | return len; |
b7ca232e | 801 | |
fcfa8d95 SS |
802 | out_destroy_comp: |
803 | up_write(&zram->init_lock); | |
804 | zcomp_destroy(comp); | |
805 | out_free_meta: | |
b7ca232e SS |
806 | zram_meta_free(meta); |
807 | return err; | |
9b3bb7ab SS |
808 | } |
809 | ||
810 | static ssize_t reset_store(struct device *dev, | |
811 | struct device_attribute *attr, const char *buf, size_t len) | |
812 | { | |
813 | int ret; | |
814 | unsigned short do_reset; | |
815 | struct zram *zram; | |
816 | struct block_device *bdev; | |
817 | ||
818 | zram = dev_to_zram(dev); | |
819 | bdev = bdget_disk(zram->disk, 0); | |
820 | ||
46a51c80 RK |
821 | if (!bdev) |
822 | return -ENOMEM; | |
823 | ||
9b3bb7ab | 824 | /* Do not reset an active device! */ |
1b672224 RK |
825 | if (bdev->bd_holders) { |
826 | ret = -EBUSY; | |
827 | goto out; | |
828 | } | |
9b3bb7ab SS |
829 | |
830 | ret = kstrtou16(buf, 10, &do_reset); | |
831 | if (ret) | |
1b672224 | 832 | goto out; |
9b3bb7ab | 833 | |
1b672224 RK |
834 | if (!do_reset) { |
835 | ret = -EINVAL; | |
836 | goto out; | |
837 | } | |
9b3bb7ab SS |
838 | |
839 | /* Make sure all pending I/O is finished */ | |
46a51c80 | 840 | fsync_bdev(bdev); |
1b672224 | 841 | bdput(bdev); |
9b3bb7ab | 842 | |
2b86ab9c | 843 | zram_reset_device(zram, true); |
9b3bb7ab | 844 | return len; |
1b672224 RK |
845 | |
846 | out: | |
847 | bdput(bdev); | |
848 | return ret; | |
8c921b2b JM |
849 | } |
850 | ||
be257c61 | 851 | static void __zram_make_request(struct zram *zram, struct bio *bio) |
8c921b2b | 852 | { |
7988613b | 853 | int offset; |
8c921b2b | 854 | u32 index; |
7988613b KO |
855 | struct bio_vec bvec; |
856 | struct bvec_iter iter; | |
8c921b2b | 857 | |
4f024f37 KO |
858 | index = bio->bi_iter.bi_sector >> SECTORS_PER_PAGE_SHIFT; |
859 | offset = (bio->bi_iter.bi_sector & | |
860 | (SECTORS_PER_PAGE - 1)) << SECTOR_SHIFT; | |
8c921b2b | 861 | |
f4659d8e JK |
862 | if (unlikely(bio->bi_rw & REQ_DISCARD)) { |
863 | zram_bio_discard(zram, index, offset, bio); | |
864 | bio_endio(bio, 0); | |
865 | return; | |
866 | } | |
867 | ||
7988613b | 868 | bio_for_each_segment(bvec, bio, iter) { |
924bd88d JM |
869 | int max_transfer_size = PAGE_SIZE - offset; |
870 | ||
7988613b | 871 | if (bvec.bv_len > max_transfer_size) { |
924bd88d JM |
872 | /* |
873 | * zram_bvec_rw() can only make operation on a single | |
874 | * zram page. Split the bio vector. | |
875 | */ | |
876 | struct bio_vec bv; | |
877 | ||
7988613b | 878 | bv.bv_page = bvec.bv_page; |
924bd88d | 879 | bv.bv_len = max_transfer_size; |
7988613b | 880 | bv.bv_offset = bvec.bv_offset; |
924bd88d | 881 | |
be257c61 | 882 | if (zram_bvec_rw(zram, &bv, index, offset, bio) < 0) |
924bd88d JM |
883 | goto out; |
884 | ||
7988613b | 885 | bv.bv_len = bvec.bv_len - max_transfer_size; |
924bd88d | 886 | bv.bv_offset += max_transfer_size; |
be257c61 | 887 | if (zram_bvec_rw(zram, &bv, index + 1, 0, bio) < 0) |
924bd88d JM |
888 | goto out; |
889 | } else | |
be257c61 | 890 | if (zram_bvec_rw(zram, &bvec, index, offset, bio) < 0) |
924bd88d JM |
891 | goto out; |
892 | ||
7988613b | 893 | update_position(&index, &offset, &bvec); |
a1dd52af | 894 | } |
306b0c95 NG |
895 | |
896 | set_bit(BIO_UPTODATE, &bio->bi_flags); | |
897 | bio_endio(bio, 0); | |
7d7854b4 | 898 | return; |
306b0c95 NG |
899 | |
900 | out: | |
306b0c95 | 901 | bio_io_error(bio); |
306b0c95 NG |
902 | } |
903 | ||
306b0c95 | 904 | /* |
f1e3cfff | 905 | * Handler function for all zram I/O requests. |
306b0c95 | 906 | */ |
5a7bbad2 | 907 | static void zram_make_request(struct request_queue *queue, struct bio *bio) |
306b0c95 | 908 | { |
f1e3cfff | 909 | struct zram *zram = queue->queuedata; |
306b0c95 | 910 | |
0900beae | 911 | down_read(&zram->init_lock); |
be2d1d56 | 912 | if (unlikely(!init_done(zram))) |
3de738cd | 913 | goto error; |
0900beae | 914 | |
f1e3cfff | 915 | if (!valid_io_request(zram, bio)) { |
da5cc7d3 | 916 | atomic64_inc(&zram->stats.invalid_io); |
3de738cd | 917 | goto error; |
6642a67c JM |
918 | } |
919 | ||
be257c61 | 920 | __zram_make_request(zram, bio); |
0900beae | 921 | up_read(&zram->init_lock); |
306b0c95 | 922 | |
b4fdcb02 | 923 | return; |
0900beae | 924 | |
0900beae | 925 | error: |
3de738cd | 926 | up_read(&zram->init_lock); |
0900beae | 927 | bio_io_error(bio); |
306b0c95 NG |
928 | } |
929 | ||
2ccbec05 NG |
930 | static void zram_slot_free_notify(struct block_device *bdev, |
931 | unsigned long index) | |
107c161b | 932 | { |
f1e3cfff | 933 | struct zram *zram; |
f614a9f4 | 934 | struct zram_meta *meta; |
107c161b | 935 | |
f1e3cfff | 936 | zram = bdev->bd_disk->private_data; |
f614a9f4 | 937 | meta = zram->meta; |
a0c516cb | 938 | |
d2d5e762 | 939 | bit_spin_lock(ZRAM_ACCESS, &meta->table[index].value); |
f614a9f4 | 940 | zram_free_page(zram, index); |
d2d5e762 | 941 | bit_spin_unlock(ZRAM_ACCESS, &meta->table[index].value); |
f614a9f4 | 942 | atomic64_inc(&zram->stats.notify_free); |
107c161b NG |
943 | } |
944 | ||
f1e3cfff | 945 | static const struct block_device_operations zram_devops = { |
f1e3cfff | 946 | .swap_slot_free_notify = zram_slot_free_notify, |
107c161b | 947 | .owner = THIS_MODULE |
306b0c95 NG |
948 | }; |
949 | ||
9b3bb7ab SS |
950 | static DEVICE_ATTR(disksize, S_IRUGO | S_IWUSR, |
951 | disksize_show, disksize_store); | |
952 | static DEVICE_ATTR(initstate, S_IRUGO, initstate_show, NULL); | |
953 | static DEVICE_ATTR(reset, S_IWUSR, NULL, reset_store); | |
9b3bb7ab | 954 | static DEVICE_ATTR(orig_data_size, S_IRUGO, orig_data_size_show, NULL); |
9b3bb7ab | 955 | static DEVICE_ATTR(mem_used_total, S_IRUGO, mem_used_total_show, NULL); |
9ada9da9 MK |
956 | static DEVICE_ATTR(mem_limit, S_IRUGO | S_IWUSR, mem_limit_show, |
957 | mem_limit_store); | |
461a8eee MK |
958 | static DEVICE_ATTR(mem_used_max, S_IRUGO | S_IWUSR, mem_used_max_show, |
959 | mem_used_max_store); | |
beca3ec7 SS |
960 | static DEVICE_ATTR(max_comp_streams, S_IRUGO | S_IWUSR, |
961 | max_comp_streams_show, max_comp_streams_store); | |
e46b8a03 SS |
962 | static DEVICE_ATTR(comp_algorithm, S_IRUGO | S_IWUSR, |
963 | comp_algorithm_show, comp_algorithm_store); | |
9b3bb7ab | 964 | |
a68eb3b6 SS |
965 | ZRAM_ATTR_RO(num_reads); |
966 | ZRAM_ATTR_RO(num_writes); | |
64447249 SS |
967 | ZRAM_ATTR_RO(failed_reads); |
968 | ZRAM_ATTR_RO(failed_writes); | |
a68eb3b6 SS |
969 | ZRAM_ATTR_RO(invalid_io); |
970 | ZRAM_ATTR_RO(notify_free); | |
971 | ZRAM_ATTR_RO(zero_pages); | |
972 | ZRAM_ATTR_RO(compr_data_size); | |
973 | ||
9b3bb7ab SS |
974 | static struct attribute *zram_disk_attrs[] = { |
975 | &dev_attr_disksize.attr, | |
976 | &dev_attr_initstate.attr, | |
977 | &dev_attr_reset.attr, | |
978 | &dev_attr_num_reads.attr, | |
979 | &dev_attr_num_writes.attr, | |
64447249 SS |
980 | &dev_attr_failed_reads.attr, |
981 | &dev_attr_failed_writes.attr, | |
9b3bb7ab SS |
982 | &dev_attr_invalid_io.attr, |
983 | &dev_attr_notify_free.attr, | |
984 | &dev_attr_zero_pages.attr, | |
985 | &dev_attr_orig_data_size.attr, | |
986 | &dev_attr_compr_data_size.attr, | |
987 | &dev_attr_mem_used_total.attr, | |
9ada9da9 | 988 | &dev_attr_mem_limit.attr, |
461a8eee | 989 | &dev_attr_mem_used_max.attr, |
beca3ec7 | 990 | &dev_attr_max_comp_streams.attr, |
e46b8a03 | 991 | &dev_attr_comp_algorithm.attr, |
9b3bb7ab SS |
992 | NULL, |
993 | }; | |
994 | ||
995 | static struct attribute_group zram_disk_attr_group = { | |
996 | .attrs = zram_disk_attrs, | |
997 | }; | |
998 | ||
f1e3cfff | 999 | static int create_device(struct zram *zram, int device_id) |
306b0c95 | 1000 | { |
39a9b8ac | 1001 | int ret = -ENOMEM; |
de1a21a0 | 1002 | |
0900beae | 1003 | init_rwsem(&zram->init_lock); |
306b0c95 | 1004 | |
f1e3cfff NG |
1005 | zram->queue = blk_alloc_queue(GFP_KERNEL); |
1006 | if (!zram->queue) { | |
306b0c95 NG |
1007 | pr_err("Error allocating disk queue for device %d\n", |
1008 | device_id); | |
de1a21a0 | 1009 | goto out; |
306b0c95 NG |
1010 | } |
1011 | ||
f1e3cfff NG |
1012 | blk_queue_make_request(zram->queue, zram_make_request); |
1013 | zram->queue->queuedata = zram; | |
306b0c95 NG |
1014 | |
1015 | /* gendisk structure */ | |
f1e3cfff NG |
1016 | zram->disk = alloc_disk(1); |
1017 | if (!zram->disk) { | |
94b8435f | 1018 | pr_warn("Error allocating disk structure for device %d\n", |
306b0c95 | 1019 | device_id); |
39a9b8ac | 1020 | goto out_free_queue; |
306b0c95 NG |
1021 | } |
1022 | ||
f1e3cfff NG |
1023 | zram->disk->major = zram_major; |
1024 | zram->disk->first_minor = device_id; | |
1025 | zram->disk->fops = &zram_devops; | |
1026 | zram->disk->queue = zram->queue; | |
1027 | zram->disk->private_data = zram; | |
1028 | snprintf(zram->disk->disk_name, 16, "zram%d", device_id); | |
306b0c95 | 1029 | |
33863c21 | 1030 | /* Actual capacity set using syfs (/sys/block/zram<id>/disksize */ |
f1e3cfff | 1031 | set_capacity(zram->disk, 0); |
b67d1ec1 SS |
1032 | /* zram devices sort of resembles non-rotational disks */ |
1033 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue); | |
b277da0a | 1034 | queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zram->disk->queue); |
a1dd52af NG |
1035 | /* |
1036 | * To ensure that we always get PAGE_SIZE aligned | |
1037 | * and n*PAGE_SIZED sized I/O requests. | |
1038 | */ | |
f1e3cfff | 1039 | blk_queue_physical_block_size(zram->disk->queue, PAGE_SIZE); |
7b19b8d4 RJ |
1040 | blk_queue_logical_block_size(zram->disk->queue, |
1041 | ZRAM_LOGICAL_BLOCK_SIZE); | |
f1e3cfff NG |
1042 | blk_queue_io_min(zram->disk->queue, PAGE_SIZE); |
1043 | blk_queue_io_opt(zram->disk->queue, PAGE_SIZE); | |
f4659d8e JK |
1044 | zram->disk->queue->limits.discard_granularity = PAGE_SIZE; |
1045 | zram->disk->queue->limits.max_discard_sectors = UINT_MAX; | |
1046 | /* | |
1047 | * zram_bio_discard() will clear all logical blocks if logical block | |
1048 | * size is identical with physical block size(PAGE_SIZE). But if it is | |
1049 | * different, we will skip discarding some parts of logical blocks in | |
1050 | * the part of the request range which isn't aligned to physical block | |
1051 | * size. So we can't ensure that all discarded logical blocks are | |
1052 | * zeroed. | |
1053 | */ | |
1054 | if (ZRAM_LOGICAL_BLOCK_SIZE == PAGE_SIZE) | |
1055 | zram->disk->queue->limits.discard_zeroes_data = 1; | |
1056 | else | |
1057 | zram->disk->queue->limits.discard_zeroes_data = 0; | |
1058 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zram->disk->queue); | |
5d83d5a0 | 1059 | |
f1e3cfff | 1060 | add_disk(zram->disk); |
306b0c95 | 1061 | |
33863c21 NG |
1062 | ret = sysfs_create_group(&disk_to_dev(zram->disk)->kobj, |
1063 | &zram_disk_attr_group); | |
1064 | if (ret < 0) { | |
94b8435f | 1065 | pr_warn("Error creating sysfs group"); |
39a9b8ac | 1066 | goto out_free_disk; |
33863c21 | 1067 | } |
e46b8a03 | 1068 | strlcpy(zram->compressor, default_compressor, sizeof(zram->compressor)); |
be2d1d56 | 1069 | zram->meta = NULL; |
beca3ec7 | 1070 | zram->max_comp_streams = 1; |
39a9b8ac | 1071 | return 0; |
de1a21a0 | 1072 | |
39a9b8ac JL |
1073 | out_free_disk: |
1074 | del_gendisk(zram->disk); | |
1075 | put_disk(zram->disk); | |
1076 | out_free_queue: | |
1077 | blk_cleanup_queue(zram->queue); | |
de1a21a0 NG |
1078 | out: |
1079 | return ret; | |
306b0c95 NG |
1080 | } |
1081 | ||
f1e3cfff | 1082 | static void destroy_device(struct zram *zram) |
306b0c95 | 1083 | { |
33863c21 NG |
1084 | sysfs_remove_group(&disk_to_dev(zram->disk)->kobj, |
1085 | &zram_disk_attr_group); | |
33863c21 | 1086 | |
59d3fe54 RK |
1087 | del_gendisk(zram->disk); |
1088 | put_disk(zram->disk); | |
306b0c95 | 1089 | |
59d3fe54 | 1090 | blk_cleanup_queue(zram->queue); |
306b0c95 NG |
1091 | } |
1092 | ||
f1e3cfff | 1093 | static int __init zram_init(void) |
306b0c95 | 1094 | { |
de1a21a0 | 1095 | int ret, dev_id; |
306b0c95 | 1096 | |
5fa5a901 | 1097 | if (num_devices > max_num_devices) { |
94b8435f | 1098 | pr_warn("Invalid value for num_devices: %u\n", |
5fa5a901 | 1099 | num_devices); |
de1a21a0 NG |
1100 | ret = -EINVAL; |
1101 | goto out; | |
306b0c95 NG |
1102 | } |
1103 | ||
f1e3cfff NG |
1104 | zram_major = register_blkdev(0, "zram"); |
1105 | if (zram_major <= 0) { | |
94b8435f | 1106 | pr_warn("Unable to get major number\n"); |
de1a21a0 NG |
1107 | ret = -EBUSY; |
1108 | goto out; | |
306b0c95 NG |
1109 | } |
1110 | ||
306b0c95 | 1111 | /* Allocate the device array and initialize each one */ |
5fa5a901 | 1112 | zram_devices = kzalloc(num_devices * sizeof(struct zram), GFP_KERNEL); |
43801f6e | 1113 | if (!zram_devices) { |
de1a21a0 NG |
1114 | ret = -ENOMEM; |
1115 | goto unregister; | |
1116 | } | |
306b0c95 | 1117 | |
5fa5a901 | 1118 | for (dev_id = 0; dev_id < num_devices; dev_id++) { |
43801f6e | 1119 | ret = create_device(&zram_devices[dev_id], dev_id); |
de1a21a0 | 1120 | if (ret) |
3bf040c7 | 1121 | goto free_devices; |
de1a21a0 NG |
1122 | } |
1123 | ||
ca3d70bd DB |
1124 | pr_info("Created %u device(s) ...\n", num_devices); |
1125 | ||
306b0c95 | 1126 | return 0; |
de1a21a0 | 1127 | |
3bf040c7 | 1128 | free_devices: |
de1a21a0 | 1129 | while (dev_id) |
43801f6e NW |
1130 | destroy_device(&zram_devices[--dev_id]); |
1131 | kfree(zram_devices); | |
de1a21a0 | 1132 | unregister: |
f1e3cfff | 1133 | unregister_blkdev(zram_major, "zram"); |
de1a21a0 | 1134 | out: |
306b0c95 NG |
1135 | return ret; |
1136 | } | |
1137 | ||
f1e3cfff | 1138 | static void __exit zram_exit(void) |
306b0c95 NG |
1139 | { |
1140 | int i; | |
f1e3cfff | 1141 | struct zram *zram; |
306b0c95 | 1142 | |
5fa5a901 | 1143 | for (i = 0; i < num_devices; i++) { |
43801f6e | 1144 | zram = &zram_devices[i]; |
306b0c95 | 1145 | |
f1e3cfff | 1146 | destroy_device(zram); |
2b86ab9c MK |
1147 | /* |
1148 | * Shouldn't access zram->disk after destroy_device | |
1149 | * because destroy_device already released zram->disk. | |
1150 | */ | |
1151 | zram_reset_device(zram, false); | |
306b0c95 NG |
1152 | } |
1153 | ||
f1e3cfff | 1154 | unregister_blkdev(zram_major, "zram"); |
306b0c95 | 1155 | |
43801f6e | 1156 | kfree(zram_devices); |
306b0c95 NG |
1157 | pr_debug("Cleanup done!\n"); |
1158 | } | |
1159 | ||
f1e3cfff NG |
1160 | module_init(zram_init); |
1161 | module_exit(zram_exit); | |
306b0c95 | 1162 | |
9b3bb7ab SS |
1163 | module_param(num_devices, uint, 0); |
1164 | MODULE_PARM_DESC(num_devices, "Number of zram devices"); | |
1165 | ||
306b0c95 NG |
1166 | MODULE_LICENSE("Dual BSD/GPL"); |
1167 | MODULE_AUTHOR("Nitin Gupta <ngupta@vflare.org>"); | |
f1e3cfff | 1168 | MODULE_DESCRIPTION("Compressed RAM Block Device"); |