]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/staging/zram/zram_drv.c
block: Convert bio_for_each_segment() to bvec_iter
[mirror_ubuntu-bionic-kernel.git] / drivers / staging / zram / zram_drv.c
CommitLineData
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 39static int zram_major;
0f0e3ba3 40static struct zram *zram_devices;
306b0c95 41
306b0c95 42/* Module params (documentation at end) */
ca3d70bd 43static unsigned int num_devices = 1;
33863c21 44
9b3bb7ab
SS
45static inline struct zram *dev_to_zram(struct device *dev)
46{
47 return (struct zram *)dev_to_disk(dev)->private_data;
48}
49
50static ssize_t disksize_show(struct device *dev,
51 struct device_attribute *attr, char *buf)
52{
53 struct zram *zram = dev_to_zram(dev);
54
55 return sprintf(buf, "%llu\n", zram->disksize);
56}
57
58static ssize_t initstate_show(struct device *dev,
59 struct device_attribute *attr, char *buf)
60{
61 struct zram *zram = dev_to_zram(dev);
62
63 return sprintf(buf, "%u\n", zram->init_done);
64}
65
66static ssize_t num_reads_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",
72 (u64)atomic64_read(&zram->stats.num_reads));
73}
74
75static ssize_t num_writes_show(struct device *dev,
76 struct device_attribute *attr, char *buf)
77{
78 struct zram *zram = dev_to_zram(dev);
79
80 return sprintf(buf, "%llu\n",
81 (u64)atomic64_read(&zram->stats.num_writes));
82}
83
84static ssize_t invalid_io_show(struct device *dev,
85 struct device_attribute *attr, char *buf)
86{
87 struct zram *zram = dev_to_zram(dev);
88
89 return sprintf(buf, "%llu\n",
90 (u64)atomic64_read(&zram->stats.invalid_io));
91}
92
93static ssize_t notify_free_show(struct device *dev,
94 struct device_attribute *attr, char *buf)
95{
96 struct zram *zram = dev_to_zram(dev);
97
98 return sprintf(buf, "%llu\n",
99 (u64)atomic64_read(&zram->stats.notify_free));
100}
101
102static ssize_t zero_pages_show(struct device *dev,
103 struct device_attribute *attr, char *buf)
104{
105 struct zram *zram = dev_to_zram(dev);
106
107 return sprintf(buf, "%u\n", zram->stats.pages_zero);
108}
109
110static ssize_t orig_data_size_show(struct device *dev,
111 struct device_attribute *attr, char *buf)
112{
113 struct zram *zram = dev_to_zram(dev);
114
115 return sprintf(buf, "%llu\n",
116 (u64)(zram->stats.pages_stored) << PAGE_SHIFT);
117}
118
119static ssize_t compr_data_size_show(struct device *dev,
120 struct device_attribute *attr, char *buf)
121{
122 struct zram *zram = dev_to_zram(dev);
123
124 return sprintf(buf, "%llu\n",
125 (u64)atomic64_read(&zram->stats.compr_size));
126}
127
128static ssize_t mem_used_total_show(struct device *dev,
129 struct device_attribute *attr, char *buf)
130{
131 u64 val = 0;
132 struct zram *zram = dev_to_zram(dev);
133 struct zram_meta *meta = zram->meta;
134
135 down_read(&zram->init_lock);
136 if (zram->init_done)
137 val = zs_get_total_size_bytes(meta->mem_pool);
138 up_read(&zram->init_lock);
139
140 return sprintf(buf, "%llu\n", val);
141}
142
8b3cc3ed 143static int zram_test_flag(struct zram_meta *meta, u32 index,
f1e3cfff 144 enum zram_pageflags flag)
306b0c95 145{
8b3cc3ed 146 return meta->table[index].flags & BIT(flag);
306b0c95
NG
147}
148
8b3cc3ed 149static void zram_set_flag(struct zram_meta *meta, u32 index,
f1e3cfff 150 enum zram_pageflags flag)
306b0c95 151{
8b3cc3ed 152 meta->table[index].flags |= BIT(flag);
306b0c95
NG
153}
154
8b3cc3ed 155static void zram_clear_flag(struct zram_meta *meta, u32 index,
f1e3cfff 156 enum zram_pageflags flag)
306b0c95 157{
8b3cc3ed 158 meta->table[index].flags &= ~BIT(flag);
306b0c95
NG
159}
160
9b3bb7ab
SS
161static inline int is_partial_io(struct bio_vec *bvec)
162{
163 return bvec->bv_len != PAGE_SIZE;
164}
165
166/*
167 * Check if request is within bounds and aligned on zram logical blocks.
168 */
169static inline int valid_io_request(struct zram *zram, struct bio *bio)
170{
171 u64 start, end, bound;
a539c72a 172
9b3bb7ab 173 /* unaligned request */
4f024f37
KO
174 if (unlikely(bio->bi_iter.bi_sector &
175 (ZRAM_SECTOR_PER_LOGICAL_BLOCK - 1)))
9b3bb7ab 176 return 0;
4f024f37 177 if (unlikely(bio->bi_iter.bi_size & (ZRAM_LOGICAL_BLOCK_SIZE - 1)))
9b3bb7ab
SS
178 return 0;
179
4f024f37
KO
180 start = bio->bi_iter.bi_sector;
181 end = start + (bio->bi_iter.bi_size >> SECTOR_SHIFT);
9b3bb7ab
SS
182 bound = zram->disksize >> SECTOR_SHIFT;
183 /* out of range range */
75c7caf5 184 if (unlikely(start >= bound || end > bound || start > end))
9b3bb7ab
SS
185 return 0;
186
187 /* I/O request is valid */
188 return 1;
189}
190
191static void zram_meta_free(struct zram_meta *meta)
192{
193 zs_destroy_pool(meta->mem_pool);
194 kfree(meta->compress_workmem);
195 free_pages((unsigned long)meta->compress_buffer, 1);
196 vfree(meta->table);
197 kfree(meta);
198}
199
200static struct zram_meta *zram_meta_alloc(u64 disksize)
201{
202 size_t num_pages;
203 struct zram_meta *meta = kmalloc(sizeof(*meta), GFP_KERNEL);
204 if (!meta)
205 goto out;
206
207 meta->compress_workmem = kzalloc(LZO1X_MEM_COMPRESS, GFP_KERNEL);
208 if (!meta->compress_workmem)
209 goto free_meta;
210
211 meta->compress_buffer =
212 (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 1);
213 if (!meta->compress_buffer) {
214 pr_err("Error allocating compressor buffer space\n");
215 goto free_workmem;
216 }
217
218 num_pages = disksize >> PAGE_SHIFT;
219 meta->table = vzalloc(num_pages * sizeof(*meta->table));
220 if (!meta->table) {
221 pr_err("Error allocating zram address table\n");
222 goto free_buffer;
223 }
224
225 meta->mem_pool = zs_create_pool(GFP_NOIO | __GFP_HIGHMEM);
226 if (!meta->mem_pool) {
227 pr_err("Error creating memory pool\n");
228 goto free_table;
229 }
230
231 return meta;
232
233free_table:
234 vfree(meta->table);
235free_buffer:
236 free_pages((unsigned long)meta->compress_buffer, 1);
237free_workmem:
238 kfree(meta->compress_workmem);
239free_meta:
240 kfree(meta);
241 meta = NULL;
242out:
243 return meta;
244}
245
246static void update_position(u32 *index, int *offset, struct bio_vec *bvec)
247{
248 if (*offset + bvec->bv_len >= PAGE_SIZE)
249 (*index)++;
250 *offset = (*offset + bvec->bv_len) % PAGE_SIZE;
251}
252
306b0c95
NG
253static int page_zero_filled(void *ptr)
254{
255 unsigned int pos;
256 unsigned long *page;
257
258 page = (unsigned long *)ptr;
259
260 for (pos = 0; pos != PAGE_SIZE / sizeof(*page); pos++) {
261 if (page[pos])
262 return 0;
263 }
264
265 return 1;
266}
267
9b3bb7ab
SS
268static void handle_zero_page(struct bio_vec *bvec)
269{
270 struct page *page = bvec->bv_page;
271 void *user_mem;
272
273 user_mem = kmap_atomic(page);
274 if (is_partial_io(bvec))
275 memset(user_mem + bvec->bv_offset, 0, bvec->bv_len);
276 else
277 clear_page(user_mem);
278 kunmap_atomic(user_mem);
279
280 flush_dcache_page(page);
281}
282
f1e3cfff 283static void zram_free_page(struct zram *zram, size_t index)
306b0c95 284{
8b3cc3ed
MK
285 struct zram_meta *meta = zram->meta;
286 unsigned long handle = meta->table[index].handle;
287 u16 size = meta->table[index].size;
306b0c95 288
fd1a30de 289 if (unlikely(!handle)) {
2e882281
NG
290 /*
291 * No memory is allocated for zero filled pages.
292 * Simply clear zero page flag.
293 */
8b3cc3ed
MK
294 if (zram_test_flag(meta, index, ZRAM_ZERO)) {
295 zram_clear_flag(meta, index, ZRAM_ZERO);
d178a07c 296 zram->stats.pages_zero--;
306b0c95
NG
297 }
298 return;
299 }
300
130f315a 301 if (unlikely(size > max_zpage_size))
d178a07c 302 zram->stats.bad_compress--;
306b0c95 303
8b3cc3ed 304 zs_free(meta->mem_pool, handle);
306b0c95 305
130f315a 306 if (size <= PAGE_SIZE / 2)
d178a07c 307 zram->stats.good_compress--;
306b0c95 308
da5cc7d3 309 atomic64_sub(meta->table[index].size, &zram->stats.compr_size);
d178a07c 310 zram->stats.pages_stored--;
306b0c95 311
8b3cc3ed
MK
312 meta->table[index].handle = 0;
313 meta->table[index].size = 0;
306b0c95
NG
314}
315
37b51fdd 316static int zram_decompress_page(struct zram *zram, char *mem, u32 index)
306b0c95 317{
37b51fdd
SS
318 int ret = LZO_E_OK;
319 size_t clen = PAGE_SIZE;
320 unsigned char *cmem;
8b3cc3ed
MK
321 struct zram_meta *meta = zram->meta;
322 unsigned long handle = meta->table[index].handle;
306b0c95 323
8b3cc3ed 324 if (!handle || zram_test_flag(meta, index, ZRAM_ZERO)) {
42e99bd9 325 clear_page(mem);
8c921b2b
JM
326 return 0;
327 }
306b0c95 328
8b3cc3ed
MK
329 cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_RO);
330 if (meta->table[index].size == PAGE_SIZE)
42e99bd9 331 copy_page(mem, cmem);
37b51fdd 332 else
8b3cc3ed 333 ret = lzo1x_decompress_safe(cmem, meta->table[index].size,
37b51fdd 334 mem, &clen);
8b3cc3ed 335 zs_unmap_object(meta->mem_pool, handle);
a1dd52af 336
8c921b2b
JM
337 /* Should NEVER happen. Return bio error if it does. */
338 if (unlikely(ret != LZO_E_OK)) {
339 pr_err("Decompression failed! err=%d, page=%u\n", ret, index);
da5cc7d3 340 atomic64_inc(&zram->stats.failed_reads);
8c921b2b 341 return ret;
a1dd52af 342 }
306b0c95 343
8c921b2b 344 return 0;
306b0c95
NG
345}
346
37b51fdd
SS
347static int zram_bvec_read(struct zram *zram, struct bio_vec *bvec,
348 u32 index, int offset, struct bio *bio)
924bd88d
JM
349{
350 int ret;
37b51fdd
SS
351 struct page *page;
352 unsigned char *user_mem, *uncmem = NULL;
8b3cc3ed 353 struct zram_meta *meta = zram->meta;
37b51fdd
SS
354 page = bvec->bv_page;
355
8b3cc3ed
MK
356 if (unlikely(!meta->table[index].handle) ||
357 zram_test_flag(meta, index, ZRAM_ZERO)) {
37b51fdd 358 handle_zero_page(bvec);
924bd88d
JM
359 return 0;
360 }
361
37b51fdd
SS
362 if (is_partial_io(bvec))
363 /* Use a temporary buffer to decompress the page */
7e5a5104
MK
364 uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
365
366 user_mem = kmap_atomic(page);
367 if (!is_partial_io(bvec))
37b51fdd
SS
368 uncmem = user_mem;
369
370 if (!uncmem) {
371 pr_info("Unable to allocate temp memory\n");
372 ret = -ENOMEM;
373 goto out_cleanup;
374 }
924bd88d 375
37b51fdd 376 ret = zram_decompress_page(zram, uncmem, index);
924bd88d 377 /* Should NEVER happen. Return bio error if it does. */
25eeb667 378 if (unlikely(ret != LZO_E_OK))
37b51fdd 379 goto out_cleanup;
924bd88d 380
37b51fdd
SS
381 if (is_partial_io(bvec))
382 memcpy(user_mem + bvec->bv_offset, uncmem + offset,
383 bvec->bv_len);
384
385 flush_dcache_page(page);
386 ret = 0;
387out_cleanup:
388 kunmap_atomic(user_mem);
389 if (is_partial_io(bvec))
390 kfree(uncmem);
391 return ret;
924bd88d
JM
392}
393
394static int zram_bvec_write(struct zram *zram, struct bio_vec *bvec, u32 index,
395 int offset)
306b0c95 396{
397c6066 397 int ret = 0;
8c921b2b 398 size_t clen;
c2344348 399 unsigned long handle;
130f315a 400 struct page *page;
924bd88d 401 unsigned char *user_mem, *cmem, *src, *uncmem = NULL;
8b3cc3ed 402 struct zram_meta *meta = zram->meta;
306b0c95 403
8c921b2b 404 page = bvec->bv_page;
8b3cc3ed 405 src = meta->compress_buffer;
306b0c95 406
924bd88d
JM
407 if (is_partial_io(bvec)) {
408 /*
409 * This is a partial IO. We need to read the full page
410 * before to write the changes.
411 */
7e5a5104 412 uncmem = kmalloc(PAGE_SIZE, GFP_NOIO);
924bd88d 413 if (!uncmem) {
924bd88d
JM
414 ret = -ENOMEM;
415 goto out;
416 }
37b51fdd 417 ret = zram_decompress_page(zram, uncmem, index);
397c6066 418 if (ret)
924bd88d 419 goto out;
924bd88d
JM
420 }
421
ba82fe2e 422 user_mem = kmap_atomic(page);
924bd88d 423
397c6066 424 if (is_partial_io(bvec)) {
924bd88d
JM
425 memcpy(uncmem + offset, user_mem + bvec->bv_offset,
426 bvec->bv_len);
397c6066
NG
427 kunmap_atomic(user_mem);
428 user_mem = NULL;
429 } else {
924bd88d 430 uncmem = user_mem;
397c6066 431 }
924bd88d
JM
432
433 if (page_zero_filled(uncmem)) {
ba82fe2e 434 kunmap_atomic(user_mem);
f40ac2ae
SS
435 /* Free memory associated with this sector now. */
436 zram_free_page(zram, index);
437
d178a07c 438 zram->stats.pages_zero++;
8b3cc3ed 439 zram_set_flag(meta, index, ZRAM_ZERO);
924bd88d
JM
440 ret = 0;
441 goto out;
8c921b2b 442 }
306b0c95 443
a0c516cb
MK
444 /*
445 * zram_slot_free_notify could miss free so that let's
446 * double check.
447 */
448 if (unlikely(meta->table[index].handle ||
449 zram_test_flag(meta, index, ZRAM_ZERO)))
450 zram_free_page(zram, index);
451
924bd88d 452 ret = lzo1x_1_compress(uncmem, PAGE_SIZE, src, &clen,
8b3cc3ed 453 meta->compress_workmem);
306b0c95 454
397c6066
NG
455 if (!is_partial_io(bvec)) {
456 kunmap_atomic(user_mem);
457 user_mem = NULL;
458 uncmem = NULL;
459 }
306b0c95 460
8c921b2b 461 if (unlikely(ret != LZO_E_OK)) {
8c921b2b 462 pr_err("Compression failed! err=%d\n", ret);
924bd88d 463 goto out;
8c921b2b 464 }
306b0c95 465
c8f2f0db 466 if (unlikely(clen > max_zpage_size)) {
d178a07c 467 zram->stats.bad_compress++;
c8f2f0db 468 clen = PAGE_SIZE;
397c6066
NG
469 src = NULL;
470 if (is_partial_io(bvec))
471 src = uncmem;
c8f2f0db 472 }
a1dd52af 473
8b3cc3ed 474 handle = zs_malloc(meta->mem_pool, clen);
fd1a30de 475 if (!handle) {
596b3dd4
MR
476 pr_info("Error allocating memory for compressed page: %u, size=%zu\n",
477 index, clen);
924bd88d
JM
478 ret = -ENOMEM;
479 goto out;
8c921b2b 480 }
8b3cc3ed 481 cmem = zs_map_object(meta->mem_pool, handle, ZS_MM_WO);
306b0c95 482
42e99bd9 483 if ((clen == PAGE_SIZE) && !is_partial_io(bvec)) {
397c6066 484 src = kmap_atomic(page);
42e99bd9 485 copy_page(cmem, src);
397c6066 486 kunmap_atomic(src);
42e99bd9
JL
487 } else {
488 memcpy(cmem, src, clen);
489 }
306b0c95 490
8b3cc3ed 491 zs_unmap_object(meta->mem_pool, handle);
fd1a30de 492
f40ac2ae
SS
493 /*
494 * Free memory associated with this sector
495 * before overwriting unused sectors.
496 */
497 zram_free_page(zram, index);
498
8b3cc3ed
MK
499 meta->table[index].handle = handle;
500 meta->table[index].size = clen;
306b0c95 501
8c921b2b 502 /* Update stats */
da5cc7d3 503 atomic64_add(clen, &zram->stats.compr_size);
d178a07c 504 zram->stats.pages_stored++;
8c921b2b 505 if (clen <= PAGE_SIZE / 2)
d178a07c 506 zram->stats.good_compress++;
306b0c95 507
924bd88d 508out:
397c6066
NG
509 if (is_partial_io(bvec))
510 kfree(uncmem);
511
924bd88d 512 if (ret)
da5cc7d3 513 atomic64_inc(&zram->stats.failed_writes);
924bd88d 514 return ret;
8c921b2b
JM
515}
516
a0c516cb
MK
517static void handle_pending_slot_free(struct zram *zram)
518{
519 struct zram_slot_free *free_rq;
520
521 spin_lock(&zram->slot_free_lock);
522 while (zram->slot_free_rq) {
523 free_rq = zram->slot_free_rq;
524 zram->slot_free_rq = free_rq->next;
525 zram_free_page(zram, free_rq->index);
526 kfree(free_rq);
527 }
528 spin_unlock(&zram->slot_free_lock);
529}
530
8c921b2b 531static int zram_bvec_rw(struct zram *zram, struct bio_vec *bvec, u32 index,
924bd88d 532 int offset, struct bio *bio, int rw)
8c921b2b 533{
c5bde238 534 int ret;
8c921b2b 535
c5bde238
JM
536 if (rw == READ) {
537 down_read(&zram->lock);
a0c516cb 538 handle_pending_slot_free(zram);
c5bde238
JM
539 ret = zram_bvec_read(zram, bvec, index, offset, bio);
540 up_read(&zram->lock);
541 } else {
542 down_write(&zram->lock);
a0c516cb 543 handle_pending_slot_free(zram);
c5bde238
JM
544 ret = zram_bvec_write(zram, bvec, index, offset);
545 up_write(&zram->lock);
546 }
547
548 return ret;
924bd88d
JM
549}
550
2b86ab9c 551static void zram_reset_device(struct zram *zram, bool reset_capacity)
924bd88d 552{
9b3bb7ab
SS
553 size_t index;
554 struct zram_meta *meta;
555
a0c516cb
MK
556 flush_work(&zram->free_work);
557
644d4787
SS
558 down_write(&zram->init_lock);
559 if (!zram->init_done) {
560 up_write(&zram->init_lock);
9b3bb7ab 561 return;
644d4787 562 }
9b3bb7ab
SS
563
564 meta = zram->meta;
565 zram->init_done = 0;
566
567 /* Free all pages that are still in this zram device */
568 for (index = 0; index < zram->disksize >> PAGE_SHIFT; index++) {
569 unsigned long handle = meta->table[index].handle;
570 if (!handle)
571 continue;
572
573 zs_free(meta->mem_pool, handle);
574 }
575
576 zram_meta_free(zram->meta);
577 zram->meta = NULL;
578 /* Reset stats */
579 memset(&zram->stats, 0, sizeof(zram->stats));
580
581 zram->disksize = 0;
2b86ab9c
MK
582 if (reset_capacity)
583 set_capacity(zram->disk, 0);
644d4787 584 up_write(&zram->init_lock);
9b3bb7ab
SS
585}
586
587static void zram_init_device(struct zram *zram, struct zram_meta *meta)
588{
589 if (zram->disksize > 2 * (totalram_pages << PAGE_SHIFT)) {
590 pr_info(
591 "There is little point creating a zram of greater than "
592 "twice the size of memory since we expect a 2:1 compression "
593 "ratio. Note that zram uses about 0.1%% of the size of "
594 "the disk when not in use so a huge zram is "
595 "wasteful.\n"
596 "\tMemory Size: %lu kB\n"
597 "\tSize you selected: %llu kB\n"
598 "Continuing anyway ...\n",
599 (totalram_pages << PAGE_SHIFT) >> 10, zram->disksize >> 10
600 );
601 }
602
603 /* zram devices sort of resembles non-rotational disks */
604 queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue);
605
606 zram->meta = meta;
607 zram->init_done = 1;
608
609 pr_debug("Initialization done!\n");
610}
611
612static ssize_t disksize_store(struct device *dev,
613 struct device_attribute *attr, const char *buf, size_t len)
614{
615 u64 disksize;
616 struct zram_meta *meta;
617 struct zram *zram = dev_to_zram(dev);
618
619 disksize = memparse(buf, NULL);
620 if (!disksize)
621 return -EINVAL;
622
623 disksize = PAGE_ALIGN(disksize);
624 meta = zram_meta_alloc(disksize);
625 down_write(&zram->init_lock);
626 if (zram->init_done) {
627 up_write(&zram->init_lock);
628 zram_meta_free(meta);
629 pr_info("Cannot change disksize for initialized device\n");
630 return -EBUSY;
631 }
632
633 zram->disksize = disksize;
634 set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT);
635 zram_init_device(zram, meta);
636 up_write(&zram->init_lock);
637
638 return len;
639}
640
641static ssize_t reset_store(struct device *dev,
642 struct device_attribute *attr, const char *buf, size_t len)
643{
644 int ret;
645 unsigned short do_reset;
646 struct zram *zram;
647 struct block_device *bdev;
648
649 zram = dev_to_zram(dev);
650 bdev = bdget_disk(zram->disk, 0);
651
46a51c80
RK
652 if (!bdev)
653 return -ENOMEM;
654
9b3bb7ab
SS
655 /* Do not reset an active device! */
656 if (bdev->bd_holders)
657 return -EBUSY;
658
659 ret = kstrtou16(buf, 10, &do_reset);
660 if (ret)
661 return ret;
662
663 if (!do_reset)
664 return -EINVAL;
665
666 /* Make sure all pending I/O is finished */
46a51c80 667 fsync_bdev(bdev);
9b3bb7ab 668
2b86ab9c 669 zram_reset_device(zram, true);
9b3bb7ab 670 return len;
8c921b2b
JM
671}
672
673static void __zram_make_request(struct zram *zram, struct bio *bio, int rw)
674{
7988613b 675 int offset;
8c921b2b 676 u32 index;
7988613b
KO
677 struct bio_vec bvec;
678 struct bvec_iter iter;
8c921b2b
JM
679
680 switch (rw) {
681 case READ:
da5cc7d3 682 atomic64_inc(&zram->stats.num_reads);
8c921b2b
JM
683 break;
684 case WRITE:
da5cc7d3 685 atomic64_inc(&zram->stats.num_writes);
8c921b2b
JM
686 break;
687 }
688
4f024f37
KO
689 index = bio->bi_iter.bi_sector >> SECTORS_PER_PAGE_SHIFT;
690 offset = (bio->bi_iter.bi_sector &
691 (SECTORS_PER_PAGE - 1)) << SECTOR_SHIFT;
8c921b2b 692
7988613b 693 bio_for_each_segment(bvec, bio, iter) {
924bd88d
JM
694 int max_transfer_size = PAGE_SIZE - offset;
695
7988613b 696 if (bvec.bv_len > max_transfer_size) {
924bd88d
JM
697 /*
698 * zram_bvec_rw() can only make operation on a single
699 * zram page. Split the bio vector.
700 */
701 struct bio_vec bv;
702
7988613b 703 bv.bv_page = bvec.bv_page;
924bd88d 704 bv.bv_len = max_transfer_size;
7988613b 705 bv.bv_offset = bvec.bv_offset;
924bd88d
JM
706
707 if (zram_bvec_rw(zram, &bv, index, offset, bio, rw) < 0)
708 goto out;
709
7988613b 710 bv.bv_len = bvec.bv_len - max_transfer_size;
924bd88d
JM
711 bv.bv_offset += max_transfer_size;
712 if (zram_bvec_rw(zram, &bv, index+1, 0, bio, rw) < 0)
713 goto out;
714 } else
7988613b 715 if (zram_bvec_rw(zram, &bvec, index, offset, bio, rw)
924bd88d
JM
716 < 0)
717 goto out;
718
7988613b 719 update_position(&index, &offset, &bvec);
a1dd52af 720 }
306b0c95
NG
721
722 set_bit(BIO_UPTODATE, &bio->bi_flags);
723 bio_endio(bio, 0);
7d7854b4 724 return;
306b0c95
NG
725
726out:
306b0c95 727 bio_io_error(bio);
306b0c95
NG
728}
729
306b0c95 730/*
f1e3cfff 731 * Handler function for all zram I/O requests.
306b0c95 732 */
5a7bbad2 733static void zram_make_request(struct request_queue *queue, struct bio *bio)
306b0c95 734{
f1e3cfff 735 struct zram *zram = queue->queuedata;
306b0c95 736
0900beae
JM
737 down_read(&zram->init_lock);
738 if (unlikely(!zram->init_done))
3de738cd 739 goto error;
0900beae 740
f1e3cfff 741 if (!valid_io_request(zram, bio)) {
da5cc7d3 742 atomic64_inc(&zram->stats.invalid_io);
3de738cd 743 goto error;
6642a67c
JM
744 }
745
8c921b2b 746 __zram_make_request(zram, bio, bio_data_dir(bio));
0900beae 747 up_read(&zram->init_lock);
306b0c95 748
b4fdcb02 749 return;
0900beae 750
0900beae 751error:
3de738cd 752 up_read(&zram->init_lock);
0900beae 753 bio_io_error(bio);
306b0c95
NG
754}
755
a0c516cb
MK
756static void zram_slot_free(struct work_struct *work)
757{
758 struct zram *zram;
759
760 zram = container_of(work, struct zram, free_work);
761 down_write(&zram->lock);
762 handle_pending_slot_free(zram);
763 up_write(&zram->lock);
764}
765
766static void add_slot_free(struct zram *zram, struct zram_slot_free *free_rq)
767{
768 spin_lock(&zram->slot_free_lock);
769 free_rq->next = zram->slot_free_rq;
770 zram->slot_free_rq = free_rq;
771 spin_unlock(&zram->slot_free_lock);
772}
773
2ccbec05
NG
774static void zram_slot_free_notify(struct block_device *bdev,
775 unsigned long index)
107c161b 776{
f1e3cfff 777 struct zram *zram;
a0c516cb 778 struct zram_slot_free *free_rq;
107c161b 779
f1e3cfff 780 zram = bdev->bd_disk->private_data;
da5cc7d3 781 atomic64_inc(&zram->stats.notify_free);
a0c516cb
MK
782
783 free_rq = kmalloc(sizeof(struct zram_slot_free), GFP_ATOMIC);
784 if (!free_rq)
785 return;
786
787 free_rq->index = index;
788 add_slot_free(zram, free_rq);
789 schedule_work(&zram->free_work);
107c161b
NG
790}
791
f1e3cfff 792static const struct block_device_operations zram_devops = {
f1e3cfff 793 .swap_slot_free_notify = zram_slot_free_notify,
107c161b 794 .owner = THIS_MODULE
306b0c95
NG
795};
796
9b3bb7ab
SS
797static DEVICE_ATTR(disksize, S_IRUGO | S_IWUSR,
798 disksize_show, disksize_store);
799static DEVICE_ATTR(initstate, S_IRUGO, initstate_show, NULL);
800static DEVICE_ATTR(reset, S_IWUSR, NULL, reset_store);
801static DEVICE_ATTR(num_reads, S_IRUGO, num_reads_show, NULL);
802static DEVICE_ATTR(num_writes, S_IRUGO, num_writes_show, NULL);
803static DEVICE_ATTR(invalid_io, S_IRUGO, invalid_io_show, NULL);
804static DEVICE_ATTR(notify_free, S_IRUGO, notify_free_show, NULL);
805static DEVICE_ATTR(zero_pages, S_IRUGO, zero_pages_show, NULL);
806static DEVICE_ATTR(orig_data_size, S_IRUGO, orig_data_size_show, NULL);
807static DEVICE_ATTR(compr_data_size, S_IRUGO, compr_data_size_show, NULL);
808static DEVICE_ATTR(mem_used_total, S_IRUGO, mem_used_total_show, NULL);
809
810static struct attribute *zram_disk_attrs[] = {
811 &dev_attr_disksize.attr,
812 &dev_attr_initstate.attr,
813 &dev_attr_reset.attr,
814 &dev_attr_num_reads.attr,
815 &dev_attr_num_writes.attr,
816 &dev_attr_invalid_io.attr,
817 &dev_attr_notify_free.attr,
818 &dev_attr_zero_pages.attr,
819 &dev_attr_orig_data_size.attr,
820 &dev_attr_compr_data_size.attr,
821 &dev_attr_mem_used_total.attr,
822 NULL,
823};
824
825static struct attribute_group zram_disk_attr_group = {
826 .attrs = zram_disk_attrs,
827};
828
f1e3cfff 829static int create_device(struct zram *zram, int device_id)
306b0c95 830{
39a9b8ac 831 int ret = -ENOMEM;
de1a21a0 832
c5bde238 833 init_rwsem(&zram->lock);
0900beae 834 init_rwsem(&zram->init_lock);
306b0c95 835
a0c516cb
MK
836 INIT_WORK(&zram->free_work, zram_slot_free);
837 spin_lock_init(&zram->slot_free_lock);
838 zram->slot_free_rq = NULL;
839
f1e3cfff
NG
840 zram->queue = blk_alloc_queue(GFP_KERNEL);
841 if (!zram->queue) {
306b0c95
NG
842 pr_err("Error allocating disk queue for device %d\n",
843 device_id);
de1a21a0 844 goto out;
306b0c95
NG
845 }
846
f1e3cfff
NG
847 blk_queue_make_request(zram->queue, zram_make_request);
848 zram->queue->queuedata = zram;
306b0c95
NG
849
850 /* gendisk structure */
f1e3cfff
NG
851 zram->disk = alloc_disk(1);
852 if (!zram->disk) {
94b8435f 853 pr_warn("Error allocating disk structure for device %d\n",
306b0c95 854 device_id);
39a9b8ac 855 goto out_free_queue;
306b0c95
NG
856 }
857
f1e3cfff
NG
858 zram->disk->major = zram_major;
859 zram->disk->first_minor = device_id;
860 zram->disk->fops = &zram_devops;
861 zram->disk->queue = zram->queue;
862 zram->disk->private_data = zram;
863 snprintf(zram->disk->disk_name, 16, "zram%d", device_id);
306b0c95 864
33863c21 865 /* Actual capacity set using syfs (/sys/block/zram<id>/disksize */
f1e3cfff 866 set_capacity(zram->disk, 0);
5d83d5a0 867
a1dd52af
NG
868 /*
869 * To ensure that we always get PAGE_SIZE aligned
870 * and n*PAGE_SIZED sized I/O requests.
871 */
f1e3cfff 872 blk_queue_physical_block_size(zram->disk->queue, PAGE_SIZE);
7b19b8d4
RJ
873 blk_queue_logical_block_size(zram->disk->queue,
874 ZRAM_LOGICAL_BLOCK_SIZE);
f1e3cfff
NG
875 blk_queue_io_min(zram->disk->queue, PAGE_SIZE);
876 blk_queue_io_opt(zram->disk->queue, PAGE_SIZE);
5d83d5a0 877
f1e3cfff 878 add_disk(zram->disk);
306b0c95 879
33863c21
NG
880 ret = sysfs_create_group(&disk_to_dev(zram->disk)->kobj,
881 &zram_disk_attr_group);
882 if (ret < 0) {
94b8435f 883 pr_warn("Error creating sysfs group");
39a9b8ac 884 goto out_free_disk;
33863c21 885 }
33863c21 886
f1e3cfff 887 zram->init_done = 0;
39a9b8ac 888 return 0;
de1a21a0 889
39a9b8ac
JL
890out_free_disk:
891 del_gendisk(zram->disk);
892 put_disk(zram->disk);
893out_free_queue:
894 blk_cleanup_queue(zram->queue);
de1a21a0
NG
895out:
896 return ret;
306b0c95
NG
897}
898
f1e3cfff 899static void destroy_device(struct zram *zram)
306b0c95 900{
33863c21
NG
901 sysfs_remove_group(&disk_to_dev(zram->disk)->kobj,
902 &zram_disk_attr_group);
33863c21 903
59d3fe54
RK
904 del_gendisk(zram->disk);
905 put_disk(zram->disk);
306b0c95 906
59d3fe54 907 blk_cleanup_queue(zram->queue);
306b0c95
NG
908}
909
f1e3cfff 910static int __init zram_init(void)
306b0c95 911{
de1a21a0 912 int ret, dev_id;
306b0c95 913
5fa5a901 914 if (num_devices > max_num_devices) {
94b8435f 915 pr_warn("Invalid value for num_devices: %u\n",
5fa5a901 916 num_devices);
de1a21a0
NG
917 ret = -EINVAL;
918 goto out;
306b0c95
NG
919 }
920
f1e3cfff
NG
921 zram_major = register_blkdev(0, "zram");
922 if (zram_major <= 0) {
94b8435f 923 pr_warn("Unable to get major number\n");
de1a21a0
NG
924 ret = -EBUSY;
925 goto out;
306b0c95
NG
926 }
927
306b0c95 928 /* Allocate the device array and initialize each one */
5fa5a901 929 zram_devices = kzalloc(num_devices * sizeof(struct zram), GFP_KERNEL);
43801f6e 930 if (!zram_devices) {
de1a21a0
NG
931 ret = -ENOMEM;
932 goto unregister;
933 }
306b0c95 934
5fa5a901 935 for (dev_id = 0; dev_id < num_devices; dev_id++) {
43801f6e 936 ret = create_device(&zram_devices[dev_id], dev_id);
de1a21a0 937 if (ret)
3bf040c7 938 goto free_devices;
de1a21a0
NG
939 }
940
ca3d70bd
DB
941 pr_info("Created %u device(s) ...\n", num_devices);
942
306b0c95 943 return 0;
de1a21a0 944
3bf040c7 945free_devices:
de1a21a0 946 while (dev_id)
43801f6e
NW
947 destroy_device(&zram_devices[--dev_id]);
948 kfree(zram_devices);
de1a21a0 949unregister:
f1e3cfff 950 unregister_blkdev(zram_major, "zram");
de1a21a0 951out:
306b0c95
NG
952 return ret;
953}
954
f1e3cfff 955static void __exit zram_exit(void)
306b0c95
NG
956{
957 int i;
f1e3cfff 958 struct zram *zram;
306b0c95 959
5fa5a901 960 for (i = 0; i < num_devices; i++) {
43801f6e 961 zram = &zram_devices[i];
306b0c95 962
f1e3cfff 963 destroy_device(zram);
2b86ab9c
MK
964 /*
965 * Shouldn't access zram->disk after destroy_device
966 * because destroy_device already released zram->disk.
967 */
968 zram_reset_device(zram, false);
306b0c95
NG
969 }
970
f1e3cfff 971 unregister_blkdev(zram_major, "zram");
306b0c95 972
43801f6e 973 kfree(zram_devices);
306b0c95
NG
974 pr_debug("Cleanup done!\n");
975}
976
f1e3cfff
NG
977module_init(zram_init);
978module_exit(zram_exit);
306b0c95 979
9b3bb7ab
SS
980module_param(num_devices, uint, 0);
981MODULE_PARM_DESC(num_devices, "Number of zram devices");
982
306b0c95
NG
983MODULE_LICENSE("Dual BSD/GPL");
984MODULE_AUTHOR("Nitin Gupta <ngupta@vflare.org>");
f1e3cfff 985MODULE_DESCRIPTION("Compressed RAM Block Device");