Commit | Line | Data |
---|---|---|
306b0c95 | 1 | /* |
f1e3cfff | 2 | * Compressed RAM block device |
306b0c95 | 3 | * |
1130ebba | 4 | * Copyright (C) 2008, 2009, 2010 Nitin Gupta |
306b0c95 NG |
5 | * |
6 | * This code is released using a dual license strategy: BSD/GPL | |
7 | * You can choose the licence that better fits your requirements. | |
8 | * | |
9 | * Released under the terms of 3-clause BSD License | |
10 | * Released under the terms of GNU General Public License Version 2.0 | |
11 | * | |
12 | * Project home: http://compcache.googlecode.com | |
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/lzo.h> |
306b0c95 | 33 | #include <linux/string.h> |
306b0c95 | 34 | #include <linux/vmalloc.h> |
306b0c95 | 35 | |
16a4bfb9 | 36 | #include "zram_drv.h" |
306b0c95 NG |
37 | |
38 | /* Globals */ | |
f1e3cfff | 39 | static int zram_major; |
43801f6e | 40 | struct zram *zram_devices; |
306b0c95 | 41 | |
306b0c95 | 42 | /* Module params (documentation at end) */ |
5fa5a901 | 43 | static unsigned int num_devices; |
33863c21 NG |
44 | |
45 | static void zram_stat_inc(u32 *v) | |
46 | { | |
47 | *v = *v + 1; | |
48 | } | |
49 | ||
50 | static void zram_stat_dec(u32 *v) | |
51 | { | |
52 | *v = *v - 1; | |
53 | } | |
54 | ||
55 | static void zram_stat64_add(struct zram *zram, u64 *v, u64 inc) | |
56 | { | |
57 | spin_lock(&zram->stat64_lock); | |
58 | *v = *v + inc; | |
59 | spin_unlock(&zram->stat64_lock); | |
60 | } | |
61 | ||
62 | static void zram_stat64_sub(struct zram *zram, u64 *v, u64 dec) | |
63 | { | |
64 | spin_lock(&zram->stat64_lock); | |
65 | *v = *v - dec; | |
66 | spin_unlock(&zram->stat64_lock); | |
67 | } | |
68 | ||
69 | static void zram_stat64_inc(struct zram *zram, u64 *v) | |
70 | { | |
71 | zram_stat64_add(zram, v, 1); | |
72 | } | |
306b0c95 | 73 | |
f1e3cfff NG |
74 | static int zram_test_flag(struct zram *zram, u32 index, |
75 | enum zram_pageflags flag) | |
306b0c95 | 76 | { |
f1e3cfff | 77 | return zram->table[index].flags & BIT(flag); |
306b0c95 NG |
78 | } |
79 | ||
f1e3cfff NG |
80 | static void zram_set_flag(struct zram *zram, u32 index, |
81 | enum zram_pageflags flag) | |
306b0c95 | 82 | { |
f1e3cfff | 83 | zram->table[index].flags |= BIT(flag); |
306b0c95 NG |
84 | } |
85 | ||
f1e3cfff NG |
86 | static void zram_clear_flag(struct zram *zram, u32 index, |
87 | enum zram_pageflags flag) | |
306b0c95 | 88 | { |
f1e3cfff | 89 | zram->table[index].flags &= ~BIT(flag); |
306b0c95 NG |
90 | } |
91 | ||
92 | static int page_zero_filled(void *ptr) | |
93 | { | |
94 | unsigned int pos; | |
95 | unsigned long *page; | |
96 | ||
97 | page = (unsigned long *)ptr; | |
98 | ||
99 | for (pos = 0; pos != PAGE_SIZE / sizeof(*page); pos++) { | |
100 | if (page[pos]) | |
101 | return 0; | |
102 | } | |
103 | ||
104 | return 1; | |
105 | } | |
106 | ||
f1e3cfff | 107 | static void zram_set_disksize(struct zram *zram, size_t totalram_bytes) |
306b0c95 | 108 | { |
f1e3cfff | 109 | if (!zram->disksize) { |
306b0c95 NG |
110 | pr_info( |
111 | "disk size not provided. You can use disksize_kb module " | |
112 | "param to specify size.\nUsing default: (%u%% of RAM).\n", | |
113 | default_disksize_perc_ram | |
114 | ); | |
f1e3cfff | 115 | zram->disksize = default_disksize_perc_ram * |
306b0c95 NG |
116 | (totalram_bytes / 100); |
117 | } | |
118 | ||
f1e3cfff | 119 | if (zram->disksize > 2 * (totalram_bytes)) { |
306b0c95 | 120 | pr_info( |
f1e3cfff | 121 | "There is little point creating a zram of greater than " |
306b0c95 | 122 | "twice the size of memory since we expect a 2:1 compression " |
f1e3cfff NG |
123 | "ratio. Note that zram uses about 0.1%% of the size of " |
124 | "the disk when not in use so a huge zram is " | |
306b0c95 NG |
125 | "wasteful.\n" |
126 | "\tMemory Size: %zu kB\n" | |
33863c21 | 127 | "\tSize you selected: %llu kB\n" |
306b0c95 | 128 | "Continuing anyway ...\n", |
f1e3cfff | 129 | totalram_bytes >> 10, zram->disksize |
306b0c95 NG |
130 | ); |
131 | } | |
132 | ||
f1e3cfff | 133 | zram->disksize &= PAGE_MASK; |
306b0c95 NG |
134 | } |
135 | ||
f1e3cfff | 136 | static void zram_free_page(struct zram *zram, size_t index) |
306b0c95 | 137 | { |
c2344348 | 138 | unsigned long handle = zram->table[index].handle; |
130f315a | 139 | u16 size = zram->table[index].size; |
306b0c95 | 140 | |
fd1a30de | 141 | if (unlikely(!handle)) { |
2e882281 NG |
142 | /* |
143 | * No memory is allocated for zero filled pages. | |
144 | * Simply clear zero page flag. | |
145 | */ | |
f1e3cfff NG |
146 | if (zram_test_flag(zram, index, ZRAM_ZERO)) { |
147 | zram_clear_flag(zram, index, ZRAM_ZERO); | |
148 | zram_stat_dec(&zram->stats.pages_zero); | |
306b0c95 NG |
149 | } |
150 | return; | |
151 | } | |
152 | ||
130f315a MK |
153 | if (unlikely(size > max_zpage_size)) |
154 | zram_stat_dec(&zram->stats.bad_compress); | |
306b0c95 | 155 | |
fd1a30de | 156 | zs_free(zram->mem_pool, handle); |
306b0c95 | 157 | |
130f315a | 158 | if (size <= PAGE_SIZE / 2) |
f1e3cfff | 159 | zram_stat_dec(&zram->stats.good_compress); |
306b0c95 | 160 | |
fd1a30de NG |
161 | zram_stat64_sub(zram, &zram->stats.compr_size, |
162 | zram->table[index].size); | |
f1e3cfff | 163 | zram_stat_dec(&zram->stats.pages_stored); |
306b0c95 | 164 | |
c2344348 | 165 | zram->table[index].handle = 0; |
fd1a30de | 166 | zram->table[index].size = 0; |
306b0c95 NG |
167 | } |
168 | ||
924bd88d | 169 | static void handle_zero_page(struct bio_vec *bvec) |
306b0c95 | 170 | { |
924bd88d | 171 | struct page *page = bvec->bv_page; |
306b0c95 | 172 | void *user_mem; |
306b0c95 | 173 | |
ba82fe2e | 174 | user_mem = kmap_atomic(page); |
924bd88d | 175 | memset(user_mem + bvec->bv_offset, 0, bvec->bv_len); |
ba82fe2e | 176 | kunmap_atomic(user_mem); |
306b0c95 | 177 | |
30fb8a71 | 178 | flush_dcache_page(page); |
306b0c95 NG |
179 | } |
180 | ||
924bd88d JM |
181 | static inline int is_partial_io(struct bio_vec *bvec) |
182 | { | |
183 | return bvec->bv_len != PAGE_SIZE; | |
184 | } | |
185 | ||
8c921b2b | 186 | static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec, |
924bd88d | 187 | u32 index, int offset, struct bio *bio) |
306b0c95 | 188 | { |
8c921b2b JM |
189 | int ret; |
190 | size_t clen; | |
191 | struct page *page; | |
924bd88d | 192 | unsigned char *user_mem, *cmem, *uncmem = NULL; |
a1dd52af | 193 | |
8c921b2b | 194 | page = bvec->bv_page; |
306b0c95 | 195 | |
8c921b2b | 196 | if (zram_test_flag(zram, index, ZRAM_ZERO)) { |
924bd88d | 197 | handle_zero_page(bvec); |
8c921b2b JM |
198 | return 0; |
199 | } | |
306b0c95 | 200 | |
8c921b2b | 201 | /* Requested page is not present in compressed area */ |
fd1a30de | 202 | if (unlikely(!zram->table[index].handle)) { |
8c921b2b JM |
203 | pr_debug("Read before write: sector=%lu, size=%u", |
204 | (ulong)(bio->bi_sector), bio->bi_size); | |
924bd88d | 205 | handle_zero_page(bvec); |
8c921b2b JM |
206 | return 0; |
207 | } | |
306b0c95 | 208 | |
924bd88d JM |
209 | if (is_partial_io(bvec)) { |
210 | /* Use a temporary buffer to decompress the page */ | |
211 | uncmem = kmalloc(PAGE_SIZE, GFP_KERNEL); | |
212 | if (!uncmem) { | |
213 | pr_info("Error allocating temp memory!\n"); | |
214 | return -ENOMEM; | |
215 | } | |
216 | } | |
217 | ||
ba82fe2e | 218 | user_mem = kmap_atomic(page); |
924bd88d JM |
219 | if (!is_partial_io(bvec)) |
220 | uncmem = user_mem; | |
8c921b2b | 221 | clen = PAGE_SIZE; |
306b0c95 | 222 | |
b7418510 SJ |
223 | cmem = zs_map_object(zram->mem_pool, zram->table[index].handle, |
224 | ZS_MM_RO); | |
306b0c95 | 225 | |
c8f2f0db NG |
226 | if (zram->table[index].size == PAGE_SIZE) { |
227 | memcpy(uncmem, cmem, PAGE_SIZE); | |
228 | ret = LZO_E_OK; | |
229 | } else { | |
230 | ret = lzo1x_decompress_safe(cmem, zram->table[index].size, | |
924bd88d | 231 | uncmem, &clen); |
c8f2f0db | 232 | } |
924bd88d JM |
233 | |
234 | if (is_partial_io(bvec)) { | |
235 | memcpy(user_mem + bvec->bv_offset, uncmem + offset, | |
236 | bvec->bv_len); | |
237 | kfree(uncmem); | |
238 | } | |
306b0c95 | 239 | |
fd1a30de | 240 | zs_unmap_object(zram->mem_pool, zram->table[index].handle); |
ba82fe2e | 241 | kunmap_atomic(user_mem); |
a1dd52af | 242 | |
8c921b2b JM |
243 | /* Should NEVER happen. Return bio error if it does. */ |
244 | if (unlikely(ret != LZO_E_OK)) { | |
245 | pr_err("Decompression failed! err=%d, page=%u\n", ret, index); | |
246 | zram_stat64_inc(zram, &zram->stats.failed_reads); | |
247 | return ret; | |
a1dd52af | 248 | } |
306b0c95 | 249 | |
8c921b2b | 250 | flush_dcache_page(page); |
306b0c95 | 251 | |
8c921b2b | 252 | return 0; |
306b0c95 NG |
253 | } |
254 | ||
924bd88d JM |
255 | static int zram_read_before_write(struct zram *zram, char *mem, u32 index) |
256 | { | |
257 | int ret; | |
258 | size_t clen = PAGE_SIZE; | |
924bd88d | 259 | unsigned char *cmem; |
374a6919 | 260 | unsigned long handle = zram->table[index].handle; |
924bd88d | 261 | |
374a6919 | 262 | if (zram_test_flag(zram, index, ZRAM_ZERO) || !handle) { |
924bd88d JM |
263 | memset(mem, 0, PAGE_SIZE); |
264 | return 0; | |
265 | } | |
266 | ||
b7418510 | 267 | cmem = zs_map_object(zram->mem_pool, handle, ZS_MM_RO); |
130f315a | 268 | ret = lzo1x_decompress_safe(cmem, zram->table[index].size, |
924bd88d | 269 | mem, &clen); |
374a6919 | 270 | zs_unmap_object(zram->mem_pool, handle); |
924bd88d JM |
271 | |
272 | /* Should NEVER happen. Return bio error if it does. */ | |
273 | if (unlikely(ret != LZO_E_OK)) { | |
274 | pr_err("Decompression failed! err=%d, page=%u\n", ret, index); | |
275 | zram_stat64_inc(zram, &zram->stats.failed_reads); | |
276 | return ret; | |
277 | } | |
278 | ||
279 | return 0; | |
280 | } | |
281 | ||
282 | static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index, | |
283 | int offset) | |
306b0c95 | 284 | { |
8c921b2b | 285 | int ret; |
8c921b2b | 286 | size_t clen; |
c2344348 | 287 | unsigned long handle; |
130f315a | 288 | struct page *page; |
924bd88d | 289 | unsigned char *user_mem, *cmem, *src, *uncmem = NULL; |
306b0c95 | 290 | |
8c921b2b JM |
291 | page = bvec->bv_page; |
292 | src = zram->compress_buffer; | |
306b0c95 | 293 | |
924bd88d JM |
294 | if (is_partial_io(bvec)) { |
295 | /* | |
296 | * This is a partial IO. We need to read the full page | |
297 | * before to write the changes. | |
298 | */ | |
299 | uncmem = kmalloc(PAGE_SIZE, GFP_KERNEL); | |
300 | if (!uncmem) { | |
301 | pr_info("Error allocating temp memory!\n"); | |
302 | ret = -ENOMEM; | |
303 | goto out; | |
304 | } | |
305 | ret = zram_read_before_write(zram, uncmem, index); | |
306 | if (ret) { | |
307 | kfree(uncmem); | |
308 | goto out; | |
309 | } | |
310 | } | |
311 | ||
8c921b2b JM |
312 | /* |
313 | * System overwrites unused sectors. Free memory associated | |
314 | * with this sector now. | |
315 | */ | |
fd1a30de | 316 | if (zram->table[index].handle || |
8c921b2b JM |
317 | zram_test_flag(zram, index, ZRAM_ZERO)) |
318 | zram_free_page(zram, index); | |
306b0c95 | 319 | |
ba82fe2e | 320 | user_mem = kmap_atomic(page); |
924bd88d JM |
321 | |
322 | if (is_partial_io(bvec)) | |
323 | memcpy(uncmem + offset, user_mem + bvec->bv_offset, | |
324 | bvec->bv_len); | |
325 | else | |
326 | uncmem = user_mem; | |
327 | ||
328 | if (page_zero_filled(uncmem)) { | |
ba82fe2e | 329 | kunmap_atomic(user_mem); |
924bd88d JM |
330 | if (is_partial_io(bvec)) |
331 | kfree(uncmem); | |
8c921b2b JM |
332 | zram_stat_inc(&zram->stats.pages_zero); |
333 | zram_set_flag(zram, index, ZRAM_ZERO); | |
924bd88d JM |
334 | ret = 0; |
335 | goto out; | |
8c921b2b | 336 | } |
306b0c95 | 337 | |
924bd88d | 338 | ret = lzo1x_1_compress(uncmem, PAGE_SIZE, src, &clen, |
8c921b2b | 339 | zram->compress_workmem); |
306b0c95 | 340 | |
ba82fe2e | 341 | kunmap_atomic(user_mem); |
924bd88d JM |
342 | if (is_partial_io(bvec)) |
343 | kfree(uncmem); | |
306b0c95 | 344 | |
8c921b2b | 345 | if (unlikely(ret != LZO_E_OK)) { |
8c921b2b | 346 | pr_err("Compression failed! err=%d\n", ret); |
924bd88d | 347 | goto out; |
8c921b2b | 348 | } |
306b0c95 | 349 | |
c8f2f0db | 350 | if (unlikely(clen > max_zpage_size)) { |
130f315a | 351 | zram_stat_inc(&zram->stats.bad_compress); |
c8f2f0db NG |
352 | src = uncmem; |
353 | clen = PAGE_SIZE; | |
354 | } | |
a1dd52af | 355 | |
130f315a | 356 | handle = zs_malloc(zram->mem_pool, clen); |
fd1a30de | 357 | if (!handle) { |
8c921b2b JM |
358 | pr_info("Error allocating memory for compressed " |
359 | "page: %u, size=%zu\n", index, clen); | |
924bd88d JM |
360 | ret = -ENOMEM; |
361 | goto out; | |
8c921b2b | 362 | } |
b7418510 | 363 | cmem = zs_map_object(zram->mem_pool, handle, ZS_MM_WO); |
306b0c95 | 364 | |
8c921b2b | 365 | memcpy(cmem, src, clen); |
306b0c95 | 366 | |
130f315a | 367 | zs_unmap_object(zram->mem_pool, handle); |
fd1a30de NG |
368 | |
369 | zram->table[index].handle = handle; | |
370 | zram->table[index].size = clen; | |
306b0c95 | 371 | |
8c921b2b JM |
372 | /* Update stats */ |
373 | zram_stat64_add(zram, &zram->stats.compr_size, clen); | |
374 | zram_stat_inc(&zram->stats.pages_stored); | |
375 | if (clen <= PAGE_SIZE / 2) | |
376 | zram_stat_inc(&zram->stats.good_compress); | |
306b0c95 | 377 | |
8c921b2b | 378 | return 0; |
924bd88d JM |
379 | |
380 | out: | |
381 | if (ret) | |
382 | zram_stat64_inc(zram, &zram->stats.failed_writes); | |
383 | return ret; | |
8c921b2b JM |
384 | } |
385 | ||
386 | static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index, | |
924bd88d | 387 | int offset, struct bio *bio, int rw) |
8c921b2b | 388 | { |
c5bde238 | 389 | int ret; |
8c921b2b | 390 | |
c5bde238 JM |
391 | if (rw == READ) { |
392 | down_read(&zram->lock); | |
393 | ret = zram_bvec_read(zram, bvec, index, offset, bio); | |
394 | up_read(&zram->lock); | |
395 | } else { | |
396 | down_write(&zram->lock); | |
397 | ret = zram_bvec_write(zram, bvec, index, offset); | |
398 | up_write(&zram->lock); | |
399 | } | |
400 | ||
401 | return ret; | |
924bd88d JM |
402 | } |
403 | ||
404 | static void update_position(u32 *index, int *offset, struct bio_vec *bvec) | |
405 | { | |
406 | if (*offset + bvec->bv_len >= PAGE_SIZE) | |
407 | (*index)++; | |
408 | *offset = (*offset + bvec->bv_len) % PAGE_SIZE; | |
8c921b2b JM |
409 | } |
410 | ||
411 | static void __zram_make_request(struct zram *zram, struct bio *bio, int rw) | |
412 | { | |
924bd88d | 413 | int i, offset; |
8c921b2b JM |
414 | u32 index; |
415 | struct bio_vec *bvec; | |
416 | ||
417 | switch (rw) { | |
418 | case READ: | |
419 | zram_stat64_inc(zram, &zram->stats.num_reads); | |
420 | break; | |
421 | case WRITE: | |
422 | zram_stat64_inc(zram, &zram->stats.num_writes); | |
423 | break; | |
424 | } | |
425 | ||
426 | index = bio->bi_sector >> SECTORS_PER_PAGE_SHIFT; | |
924bd88d | 427 | offset = (bio->bi_sector & (SECTORS_PER_PAGE - 1)) << SECTOR_SHIFT; |
8c921b2b JM |
428 | |
429 | bio_for_each_segment(bvec, bio, i) { | |
924bd88d JM |
430 | int max_transfer_size = PAGE_SIZE - offset; |
431 | ||
432 | if (bvec->bv_len > max_transfer_size) { | |
433 | /* | |
434 | * zram_bvec_rw() can only make operation on a single | |
435 | * zram page. Split the bio vector. | |
436 | */ | |
437 | struct bio_vec bv; | |
438 | ||
439 | bv.bv_page = bvec->bv_page; | |
440 | bv.bv_len = max_transfer_size; | |
441 | bv.bv_offset = bvec->bv_offset; | |
442 | ||
443 | if (zram_bvec_rw(zram, &bv, index, offset, bio, rw) < 0) | |
444 | goto out; | |
445 | ||
446 | bv.bv_len = bvec->bv_len - max_transfer_size; | |
447 | bv.bv_offset += max_transfer_size; | |
448 | if (zram_bvec_rw(zram, &bv, index+1, 0, bio, rw) < 0) | |
449 | goto out; | |
450 | } else | |
451 | if (zram_bvec_rw(zram, bvec, index, offset, bio, rw) | |
452 | < 0) | |
453 | goto out; | |
454 | ||
455 | update_position(&index, &offset, bvec); | |
a1dd52af | 456 | } |
306b0c95 NG |
457 | |
458 | set_bit(BIO_UPTODATE, &bio->bi_flags); | |
459 | bio_endio(bio, 0); | |
7d7854b4 | 460 | return; |
306b0c95 NG |
461 | |
462 | out: | |
306b0c95 | 463 | bio_io_error(bio); |
306b0c95 NG |
464 | } |
465 | ||
306b0c95 | 466 | /* |
924bd88d | 467 | * Check if request is within bounds and aligned on zram logical blocks. |
306b0c95 | 468 | */ |
f1e3cfff | 469 | static inline int valid_io_request(struct zram *zram, struct bio *bio) |
306b0c95 NG |
470 | { |
471 | if (unlikely( | |
f1e3cfff | 472 | (bio->bi_sector >= (zram->disksize >> SECTOR_SHIFT)) || |
924bd88d JM |
473 | (bio->bi_sector & (ZRAM_SECTOR_PER_LOGICAL_BLOCK - 1)) || |
474 | (bio->bi_size & (ZRAM_LOGICAL_BLOCK_SIZE - 1)))) { | |
306b0c95 NG |
475 | |
476 | return 0; | |
477 | } | |
478 | ||
a1dd52af | 479 | /* I/O request is valid */ |
306b0c95 NG |
480 | return 1; |
481 | } | |
482 | ||
483 | /* | |
f1e3cfff | 484 | * Handler function for all zram I/O requests. |
306b0c95 | 485 | */ |
5a7bbad2 | 486 | static void zram_make_request(struct request_queue *queue, struct bio *bio) |
306b0c95 | 487 | { |
f1e3cfff | 488 | struct zram *zram = queue->queuedata; |
306b0c95 | 489 | |
0900beae JM |
490 | if (unlikely(!zram->init_done) && zram_init_device(zram)) |
491 | goto error; | |
492 | ||
493 | down_read(&zram->init_lock); | |
494 | if (unlikely(!zram->init_done)) | |
495 | goto error_unlock; | |
496 | ||
f1e3cfff NG |
497 | if (!valid_io_request(zram, bio)) { |
498 | zram_stat64_inc(zram, &zram->stats.invalid_io); | |
0900beae | 499 | goto error_unlock; |
6642a67c JM |
500 | } |
501 | ||
8c921b2b | 502 | __zram_make_request(zram, bio, bio_data_dir(bio)); |
0900beae | 503 | up_read(&zram->init_lock); |
306b0c95 | 504 | |
b4fdcb02 | 505 | return; |
0900beae JM |
506 | |
507 | error_unlock: | |
508 | up_read(&zram->init_lock); | |
509 | error: | |
510 | bio_io_error(bio); | |
306b0c95 NG |
511 | } |
512 | ||
0900beae | 513 | void __zram_reset_device(struct zram *zram) |
306b0c95 | 514 | { |
97a06382 | 515 | size_t index; |
306b0c95 | 516 | |
f1e3cfff | 517 | zram->init_done = 0; |
7eef7533 | 518 | |
306b0c95 | 519 | /* Free various per-device buffers */ |
f1e3cfff NG |
520 | kfree(zram->compress_workmem); |
521 | free_pages((unsigned long)zram->compress_buffer, 1); | |
306b0c95 | 522 | |
f1e3cfff NG |
523 | zram->compress_workmem = NULL; |
524 | zram->compress_buffer = NULL; | |
306b0c95 | 525 | |
f1e3cfff NG |
526 | /* Free all pages that are still in this zram device */ |
527 | for (index = 0; index < zram->disksize >> PAGE_SHIFT; index++) { | |
c2344348 | 528 | unsigned long handle = zram->table[index].handle; |
fd1a30de | 529 | if (!handle) |
306b0c95 NG |
530 | continue; |
531 | ||
130f315a | 532 | zs_free(zram->mem_pool, handle); |
306b0c95 NG |
533 | } |
534 | ||
f1e3cfff NG |
535 | vfree(zram->table); |
536 | zram->table = NULL; | |
306b0c95 | 537 | |
fd1a30de | 538 | zs_destroy_pool(zram->mem_pool); |
f1e3cfff | 539 | zram->mem_pool = NULL; |
306b0c95 | 540 | |
306b0c95 | 541 | /* Reset stats */ |
f1e3cfff | 542 | memset(&zram->stats, 0, sizeof(zram->stats)); |
306b0c95 | 543 | |
f1e3cfff | 544 | zram->disksize = 0; |
0900beae JM |
545 | } |
546 | ||
547 | void zram_reset_device(struct zram *zram) | |
548 | { | |
549 | down_write(&zram->init_lock); | |
550 | __zram_reset_device(zram); | |
551 | up_write(&zram->init_lock); | |
306b0c95 NG |
552 | } |
553 | ||
33863c21 | 554 | int zram_init_device(struct zram *zram) |
306b0c95 NG |
555 | { |
556 | int ret; | |
557 | size_t num_pages; | |
306b0c95 | 558 | |
0900beae | 559 | down_write(&zram->init_lock); |
484875ad | 560 | |
f1e3cfff | 561 | if (zram->init_done) { |
0900beae | 562 | up_write(&zram->init_lock); |
484875ad | 563 | return 0; |
306b0c95 NG |
564 | } |
565 | ||
f1e3cfff | 566 | zram_set_disksize(zram, totalram_pages << PAGE_SHIFT); |
306b0c95 | 567 | |
f1e3cfff NG |
568 | zram->compress_workmem = kzalloc(LZO1X_MEM_COMPRESS, GFP_KERNEL); |
569 | if (!zram->compress_workmem) { | |
306b0c95 NG |
570 | pr_err("Error allocating compressor working memory!\n"); |
571 | ret = -ENOMEM; | |
5a18c531 | 572 | goto fail_no_table; |
306b0c95 NG |
573 | } |
574 | ||
fb927284 JM |
575 | zram->compress_buffer = |
576 | (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 1); | |
f1e3cfff | 577 | if (!zram->compress_buffer) { |
306b0c95 NG |
578 | pr_err("Error allocating compressor buffer space\n"); |
579 | ret = -ENOMEM; | |
5a18c531 | 580 | goto fail_no_table; |
306b0c95 NG |
581 | } |
582 | ||
f1e3cfff | 583 | num_pages = zram->disksize >> PAGE_SHIFT; |
5b84cc78 | 584 | zram->table = vzalloc(num_pages * sizeof(*zram->table)); |
f1e3cfff NG |
585 | if (!zram->table) { |
586 | pr_err("Error allocating zram address table\n"); | |
306b0c95 | 587 | ret = -ENOMEM; |
5a18c531 | 588 | goto fail_no_table; |
306b0c95 | 589 | } |
306b0c95 | 590 | |
f1e3cfff | 591 | set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT); |
306b0c95 | 592 | |
f1e3cfff NG |
593 | /* zram devices sort of resembles non-rotational disks */ |
594 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue); | |
306b0c95 | 595 | |
fd1a30de | 596 | zram->mem_pool = zs_create_pool("zram", GFP_NOIO | __GFP_HIGHMEM); |
f1e3cfff | 597 | if (!zram->mem_pool) { |
306b0c95 NG |
598 | pr_err("Error creating memory pool\n"); |
599 | ret = -ENOMEM; | |
600 | goto fail; | |
601 | } | |
602 | ||
f1e3cfff | 603 | zram->init_done = 1; |
0900beae | 604 | up_write(&zram->init_lock); |
306b0c95 NG |
605 | |
606 | pr_debug("Initialization done!\n"); | |
607 | return 0; | |
608 | ||
5a18c531 JM |
609 | fail_no_table: |
610 | /* To prevent accessing table entries during cleanup */ | |
611 | zram->disksize = 0; | |
306b0c95 | 612 | fail: |
0900beae JM |
613 | __zram_reset_device(zram); |
614 | up_write(&zram->init_lock); | |
306b0c95 NG |
615 | pr_err("Initialization failed: err=%d\n", ret); |
616 | return ret; | |
617 | } | |
618 | ||
2ccbec05 NG |
619 | static void zram_slot_free_notify(struct block_device *bdev, |
620 | unsigned long index) | |
107c161b | 621 | { |
f1e3cfff | 622 | struct zram *zram; |
107c161b | 623 | |
f1e3cfff NG |
624 | zram = bdev->bd_disk->private_data; |
625 | zram_free_page(zram, index); | |
626 | zram_stat64_inc(zram, &zram->stats.notify_free); | |
107c161b NG |
627 | } |
628 | ||
f1e3cfff | 629 | static const struct block_device_operations zram_devops = { |
f1e3cfff | 630 | .swap_slot_free_notify = zram_slot_free_notify, |
107c161b | 631 | .owner = THIS_MODULE |
306b0c95 NG |
632 | }; |
633 | ||
f1e3cfff | 634 | static int create_device(struct zram *zram, int device_id) |
306b0c95 | 635 | { |
de1a21a0 NG |
636 | int ret = 0; |
637 | ||
c5bde238 | 638 | init_rwsem(&zram->lock); |
0900beae | 639 | init_rwsem(&zram->init_lock); |
f1e3cfff | 640 | spin_lock_init(&zram->stat64_lock); |
306b0c95 | 641 | |
f1e3cfff NG |
642 | zram->queue = blk_alloc_queue(GFP_KERNEL); |
643 | if (!zram->queue) { | |
306b0c95 NG |
644 | pr_err("Error allocating disk queue for device %d\n", |
645 | device_id); | |
de1a21a0 NG |
646 | ret = -ENOMEM; |
647 | goto out; | |
306b0c95 NG |
648 | } |
649 | ||
f1e3cfff NG |
650 | blk_queue_make_request(zram->queue, zram_make_request); |
651 | zram->queue->queuedata = zram; | |
306b0c95 NG |
652 | |
653 | /* gendisk structure */ | |
f1e3cfff NG |
654 | zram->disk = alloc_disk(1); |
655 | if (!zram->disk) { | |
656 | blk_cleanup_queue(zram->queue); | |
94b8435f | 657 | pr_warn("Error allocating disk structure for device %d\n", |
306b0c95 | 658 | device_id); |
de1a21a0 NG |
659 | ret = -ENOMEM; |
660 | goto out; | |
306b0c95 NG |
661 | } |
662 | ||
f1e3cfff NG |
663 | zram->disk->major = zram_major; |
664 | zram->disk->first_minor = device_id; | |
665 | zram->disk->fops = &zram_devops; | |
666 | zram->disk->queue = zram->queue; | |
667 | zram->disk->private_data = zram; | |
668 | snprintf(zram->disk->disk_name, 16, "zram%d", device_id); | |
306b0c95 | 669 | |
33863c21 | 670 | /* Actual capacity set using syfs (/sys/block/zram<id>/disksize */ |
f1e3cfff | 671 | set_capacity(zram->disk, 0); |
5d83d5a0 | 672 | |
a1dd52af NG |
673 | /* |
674 | * To ensure that we always get PAGE_SIZE aligned | |
675 | * and n*PAGE_SIZED sized I/O requests. | |
676 | */ | |
f1e3cfff | 677 | blk_queue_physical_block_size(zram->disk->queue, PAGE_SIZE); |
7b19b8d4 RJ |
678 | blk_queue_logical_block_size(zram->disk->queue, |
679 | ZRAM_LOGICAL_BLOCK_SIZE); | |
f1e3cfff NG |
680 | blk_queue_io_min(zram->disk->queue, PAGE_SIZE); |
681 | blk_queue_io_opt(zram->disk->queue, PAGE_SIZE); | |
5d83d5a0 | 682 | |
f1e3cfff | 683 | add_disk(zram->disk); |
306b0c95 | 684 | |
33863c21 NG |
685 | ret = sysfs_create_group(&disk_to_dev(zram->disk)->kobj, |
686 | &zram_disk_attr_group); | |
687 | if (ret < 0) { | |
94b8435f | 688 | pr_warn("Error creating sysfs group"); |
33863c21 NG |
689 | goto out; |
690 | } | |
33863c21 | 691 | |
f1e3cfff | 692 | zram->init_done = 0; |
de1a21a0 NG |
693 | |
694 | out: | |
695 | return ret; | |
306b0c95 NG |
696 | } |
697 | ||
f1e3cfff | 698 | static void destroy_device(struct zram *zram) |
306b0c95 | 699 | { |
33863c21 NG |
700 | sysfs_remove_group(&disk_to_dev(zram->disk)->kobj, |
701 | &zram_disk_attr_group); | |
33863c21 | 702 | |
f1e3cfff NG |
703 | if (zram->disk) { |
704 | del_gendisk(zram->disk); | |
705 | put_disk(zram->disk); | |
306b0c95 NG |
706 | } |
707 | ||
f1e3cfff NG |
708 | if (zram->queue) |
709 | blk_cleanup_queue(zram->queue); | |
306b0c95 NG |
710 | } |
711 | ||
5fa5a901 NG |
712 | unsigned int zram_get_num_devices(void) |
713 | { | |
714 | return num_devices; | |
715 | } | |
716 | ||
f1e3cfff | 717 | static int __init zram_init(void) |
306b0c95 | 718 | { |
de1a21a0 | 719 | int ret, dev_id; |
306b0c95 | 720 | |
5fa5a901 | 721 | if (num_devices > max_num_devices) { |
94b8435f | 722 | pr_warn("Invalid value for num_devices: %u\n", |
5fa5a901 | 723 | num_devices); |
de1a21a0 NG |
724 | ret = -EINVAL; |
725 | goto out; | |
306b0c95 NG |
726 | } |
727 | ||
f1e3cfff NG |
728 | zram_major = register_blkdev(0, "zram"); |
729 | if (zram_major <= 0) { | |
94b8435f | 730 | pr_warn("Unable to get major number\n"); |
de1a21a0 NG |
731 | ret = -EBUSY; |
732 | goto out; | |
306b0c95 NG |
733 | } |
734 | ||
5fa5a901 | 735 | if (!num_devices) { |
306b0c95 | 736 | pr_info("num_devices not specified. Using default: 1\n"); |
5fa5a901 | 737 | num_devices = 1; |
306b0c95 NG |
738 | } |
739 | ||
740 | /* Allocate the device array and initialize each one */ | |
5fa5a901 NG |
741 | pr_info("Creating %u devices ...\n", num_devices); |
742 | zram_devices = kzalloc(num_devices * sizeof(struct zram), GFP_KERNEL); | |
43801f6e | 743 | if (!zram_devices) { |
de1a21a0 NG |
744 | ret = -ENOMEM; |
745 | goto unregister; | |
746 | } | |
306b0c95 | 747 | |
5fa5a901 | 748 | for (dev_id = 0; dev_id < num_devices; dev_id++) { |
43801f6e | 749 | ret = create_device(&zram_devices[dev_id], dev_id); |
de1a21a0 | 750 | if (ret) |
3bf040c7 | 751 | goto free_devices; |
de1a21a0 NG |
752 | } |
753 | ||
306b0c95 | 754 | return 0; |
de1a21a0 | 755 | |
3bf040c7 | 756 | free_devices: |
de1a21a0 | 757 | while (dev_id) |
43801f6e NW |
758 | destroy_device(&zram_devices[--dev_id]); |
759 | kfree(zram_devices); | |
de1a21a0 | 760 | unregister: |
f1e3cfff | 761 | unregister_blkdev(zram_major, "zram"); |
de1a21a0 | 762 | out: |
306b0c95 NG |
763 | return ret; |
764 | } | |
765 | ||
f1e3cfff | 766 | static void __exit zram_exit(void) |
306b0c95 NG |
767 | { |
768 | int i; | |
f1e3cfff | 769 | struct zram *zram; |
306b0c95 | 770 | |
5fa5a901 | 771 | for (i = 0; i < num_devices; i++) { |
43801f6e | 772 | zram = &zram_devices[i]; |
306b0c95 | 773 | |
f1e3cfff NG |
774 | destroy_device(zram); |
775 | if (zram->init_done) | |
33863c21 | 776 | zram_reset_device(zram); |
306b0c95 NG |
777 | } |
778 | ||
f1e3cfff | 779 | unregister_blkdev(zram_major, "zram"); |
306b0c95 | 780 | |
43801f6e | 781 | kfree(zram_devices); |
306b0c95 NG |
782 | pr_debug("Cleanup done!\n"); |
783 | } | |
784 | ||
5fa5a901 NG |
785 | module_param(num_devices, uint, 0); |
786 | MODULE_PARM_DESC(num_devices, "Number of zram devices"); | |
306b0c95 | 787 | |
f1e3cfff NG |
788 | module_init(zram_init); |
789 | module_exit(zram_exit); | |
306b0c95 NG |
790 | |
791 | MODULE_LICENSE("Dual BSD/GPL"); | |
792 | MODULE_AUTHOR("Nitin Gupta <ngupta@vflare.org>"); | |
f1e3cfff | 793 | MODULE_DESCRIPTION("Compressed RAM Block Device"); |