projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branches 'acpi-bus' and 'acpi-video'
[linux-block.git] / drivers / md / dm-integrity.c
CommitLineData
3bd94003 1// SPDX-License-Identifier: GPL-2.0-only
7eada909
MP		 2/*
3 * Copyright (C) 2016-2017 Red Hat, Inc. All rights reserved.
4 * Copyright (C) 2016-2017 Milan Broz
5 * Copyright (C) 2016-2017 Mikulas Patocka
6 *
7 * This file is released under the GPL.
8 */
9
248aa264
MS		 10#include "dm-bio-record.h"
11
d3e632f0 12#include <linux/compiler.h>
7eada909
MP		 13#include <linux/module.h>
14#include <linux/device-mapper.h>
15#include <linux/dm-io.h>
16#include <linux/vmalloc.h>
17#include <linux/sort.h>
18#include <linux/rbtree.h>
19#include <linux/delay.h>
20#include <linux/random.h>
1f5a7759 21#include <linux/reboot.h>
7eada909
MP		 22#include <crypto/hash.h>
23#include <crypto/skcipher.h>
24#include <linux/async_tx.h>
afa53df8 25#include <linux/dm-bufio.h>
7eada909 26
82bb8599
MW		 27#include "dm-audit.h"
28
7eada909
MP		 29#define DM_MSG_PREFIX "integrity"
30
31#define DEFAULT_INTERLEAVE_SECTORS 32768
32#define DEFAULT_JOURNAL_SIZE_FACTOR 7
468dfca3 33#define DEFAULT_SECTORS_PER_BITMAP_BIT 32768
7eada909
MP		 34#define DEFAULT_BUFFER_SECTORS 128
35#define DEFAULT_JOURNAL_WATERMARK 50
36#define DEFAULT_SYNC_MSEC 10000
6d50eb47 37#define DEFAULT_MAX_JOURNAL_SECTORS (IS_ENABLED(CONFIG_64BIT) ? 131072 : 8192)
56b67a4f
MP		 38#define MIN_LOG2_INTERLEAVE_SECTORS 3
39#define MAX_LOG2_INTERLEAVE_SECTORS 31
7eada909 40#define METADATA_WORKQUEUE_MAX_ACTIVE 16
6d50eb47 41#define RECALC_SECTORS (IS_ENABLED(CONFIG_64BIT) ? 32768 : 2048)
a3fcf725 42#define RECALC_WRITE_SUPER 16
468dfca3
MP		 43#define BITMAP_BLOCK_SIZE 4096 /* don't change it */
44#define BITMAP_FLUSH_INTERVAL (10 * HZ)
84597a44 45#define DISCARD_FILLER 0xf6
09d85f8d 46#define SALT_SIZE 16
7eada909
MP		 47
48/*
49 * Warning - DEBUG_PRINT prints security-sensitive data to the log,
50 * so it should not be enabled in the official kernel
51 */
52//#define DEBUG_PRINT
53//#define INTERNAL_VERIFY
54
55/*
56 * On disk structures
57 */
58
59#define SB_MAGIC "integrt"
1f9fc0b8
MP		 60#define SB_VERSION_1 1
61#define SB_VERSION_2 2
468dfca3 62#define SB_VERSION_3 3
d537858a 63#define SB_VERSION_4 4
09d85f8d 64#define SB_VERSION_5 5
7eada909 65#define SB_SECTORS 8
9d609f85 66#define MAX_SECTORS_PER_BLOCK 8
7eada909
MP		 67
68struct superblock {
69 __u8 magic[8];
70 __u8 version;
71 __u8 log2_interleave_sectors;
bc8f3d46
MP		 72 __le16 integrity_tag_size;
73 __le32 journal_sections;
74 __le64 provided_data_sectors; /* userspace uses this value */
75 __le32 flags;
9d609f85 76 __u8 log2_sectors_per_block;
468dfca3
MP		 77 __u8 log2_blocks_per_bitmap_bit;
78 __u8 pad[2];
bc8f3d46 79 __le64 recalc_sector;
09d85f8d
MP		 80 __u8 pad2[8];
81 __u8 salt[SALT_SIZE];
7eada909
MP		 82};
83
84#define SB_FLAG_HAVE_JOURNAL_MAC 0x1
a3fcf725 85#define SB_FLAG_RECALCULATING 0x2
468dfca3 86#define SB_FLAG_DIRTY_BITMAP 0x4
d537858a 87#define SB_FLAG_FIXED_PADDING 0x8
09d85f8d 88#define SB_FLAG_FIXED_HMAC 0x10
7eada909
MP		 89
90#define JOURNAL_ENTRY_ROUNDUP 8
91
bc8f3d46 92typedef __le64 commit_id_t;
7eada909
MP		 93#define JOURNAL_MAC_PER_SECTOR 8
94
95struct journal_entry {
96 union {
97 struct {
bc8f3d46
MP		 98 __le32 sector_lo;
99 __le32 sector_hi;
7eada909 100 } s;
bc8f3d46 101 __le64 sector;
7eada909 102 } u;
b18ae8dd 103 commit_id_t last_bytes[];
9d609f85 104 /* __u8 tag[0]; */
7eada909
MP		 105};
106
9d609f85
MP		 107#define journal_entry_tag(ic, je) ((__u8 *)&(je)->last_bytes[(ic)->sectors_per_block])
108
7eada909 109#if BITS_PER_LONG == 64
d3e632f0 110#define journal_entry_set_sector(je, x) do { smp_wmb(); WRITE_ONCE((je)->u.sector, cpu_to_le64(x)); } while (0)
7eada909 111#else
72deb455 112#define journal_entry_set_sector(je, x) do { (je)->u.s.sector_lo = cpu_to_le32(x); smp_wmb(); WRITE_ONCE((je)->u.s.sector_hi, cpu_to_le32((x) >> 32)); } while (0)
7eada909 113#endif
72deb455 114#define journal_entry_get_sector(je) le64_to_cpu((je)->u.sector)
7eada909 115#define journal_entry_is_unused(je) ((je)->u.s.sector_hi == cpu_to_le32(-1))
6cc435fa 116#define journal_entry_set_unused(je) ((je)->u.s.sector_hi = cpu_to_le32(-1))
7eada909 117#define journal_entry_is_inprogress(je) ((je)->u.s.sector_hi == cpu_to_le32(-2))
6cc435fa 118#define journal_entry_set_inprogress(je) ((je)->u.s.sector_hi = cpu_to_le32(-2))
7eada909
MP		 119
120#define JOURNAL_BLOCK_SECTORS 8
121#define JOURNAL_SECTOR_DATA ((1 << SECTOR_SHIFT) - sizeof(commit_id_t))
122#define JOURNAL_MAC_SIZE (JOURNAL_MAC_PER_SECTOR * JOURNAL_BLOCK_SECTORS)
123
124struct journal_sector {
f069c7ab
KC		 125 struct_group(sectors,
126 __u8 entries[JOURNAL_SECTOR_DATA - JOURNAL_MAC_PER_SECTOR];
127 __u8 mac[JOURNAL_MAC_PER_SECTOR];
128 );
7eada909
MP		 129 commit_id_t commit_id;
130};
131
9d609f85 132#define MAX_TAG_SIZE (JOURNAL_SECTOR_DATA - JOURNAL_MAC_PER_SECTOR - offsetof(struct journal_entry, last_bytes[MAX_SECTORS_PER_BLOCK]))
7eada909
MP		 133
134#define METADATA_PADDING_SECTORS 8
135
136#define N_COMMIT_IDS 4
137
138static unsigned char prev_commit_seq(unsigned char seq)
139{
140 return (seq + N_COMMIT_IDS - 1) % N_COMMIT_IDS;
141}
142
143static unsigned char next_commit_seq(unsigned char seq)
144{
145 return (seq + 1) % N_COMMIT_IDS;
146}
147
148/*
149 * In-memory structures
150 */
151
152struct journal_node {
153 struct rb_node node;
154 sector_t sector;
155};
156
157struct alg_spec {
158 char *alg_string;
159 char *key_string;
160 __u8 *key;
86a3238c 161 unsigned int key_size;
7eada909
MP		 162};
163
164struct dm_integrity_c {
165 struct dm_dev *dev;
356d9d52 166 struct dm_dev *meta_dev;
86a3238c 167 unsigned int tag_size;
7eada909
MP		 168 __s8 log2_tag_size;
169 sector_t start;
6f1c819c 170 mempool_t journal_io_mempool;
7eada909
MP		 171 struct dm_io_client *io;
172 struct dm_bufio_client *bufio;
173 struct workqueue_struct *metadata_wq;
174 struct superblock *sb;
86a3238c
HM		 175 unsigned int journal_pages;
176 unsigned int n_bitmap_blocks;
468dfca3 177
7eada909
MP		 178 struct page_list *journal;
179 struct page_list *journal_io;
180 struct page_list *journal_xor;
468dfca3
MP		 181 struct page_list *recalc_bitmap;
182 struct page_list *may_write_bitmap;
183 struct bitmap_block_status *bbs;
86a3238c 184 unsigned int bitmap_flush_interval;
48271493
MP		 185 int synchronous_mode;
186 struct bio_list synchronous_bios;
468dfca3 187 struct delayed_work bitmap_flush_work;
7eada909
MP		 188
189 struct crypto_skcipher *journal_crypt;
190 struct scatterlist **journal_scatterlist;
191 struct scatterlist **journal_io_scatterlist;
192 struct skcipher_request **sk_requests;
193
194 struct crypto_shash *journal_mac;
195
196 struct journal_node *journal_tree;
197 struct rb_root journal_tree_root;
198
199 sector_t provided_data_sectors;
200
201 unsigned short journal_entry_size;
202 unsigned char journal_entries_per_sector;
203 unsigned char journal_section_entries;
9d609f85 204 unsigned short journal_section_sectors;
86a3238c
HM		 205 unsigned int journal_sections;
206 unsigned int journal_entries;
356d9d52
MP		 207 sector_t data_device_sectors;
208 sector_t meta_device_sectors;
86a3238c
HM		 209 unsigned int initial_sectors;
210 unsigned int metadata_run;
7eada909
MP		 211 __s8 log2_metadata_run;
212 __u8 log2_buffer_sectors;
9d609f85 213 __u8 sectors_per_block;
468dfca3 214 __u8 log2_blocks_per_bitmap_bit;
7eada909
MP		 215
216 unsigned char mode;
7eada909
MP		 217
218 int failed;
219
220 struct crypto_shash *internal_hash;
221
adc0daad
MP		 222 struct dm_target *ti;
223
7eada909
MP		 224 /* these variables are locked with endio_wait.lock */
225 struct rb_root in_progress;
724376a0 226 struct list_head wait_list;
7eada909
MP		 227 wait_queue_head_t endio_wait;
228 struct workqueue_struct *wait_wq;
53770f0e 229 struct workqueue_struct *offload_wq;
7eada909
MP		 230
231 unsigned char commit_seq;
232 commit_id_t commit_ids[N_COMMIT_IDS];
233
86a3238c
HM		 234 unsigned int committed_section;
235 unsigned int n_committed_sections;
7eada909 236
86a3238c
HM		 237 unsigned int uncommitted_section;
238 unsigned int n_uncommitted_sections;
7eada909 239
86a3238c 240 unsigned int free_section;
7eada909 241 unsigned char free_section_entry;
86a3238c 242 unsigned int free_sectors;
7eada909 243
86a3238c 244 unsigned int free_sectors_threshold;
7eada909
MP		 245
246 struct workqueue_struct *commit_wq;
247 struct work_struct commit_work;
248
249 struct workqueue_struct *writer_wq;
250 struct work_struct writer_work;
251
a3fcf725
MP		 252 struct workqueue_struct *recalc_wq;
253 struct work_struct recalc_work;
a3fcf725 254
7eada909
MP		 255 struct bio_list flush_bio_list;
256
257 unsigned long autocommit_jiffies;
258 struct timer_list autocommit_timer;
86a3238c 259 unsigned int autocommit_msec;
7eada909
MP		 260
261 wait_queue_head_t copy_to_journal_wait;
262
263 struct completion crypto_backoff;
264
984bf2cc 265 bool wrote_to_journal;
7eada909
MP		 266 bool journal_uptodate;
267 bool just_formatted;
468dfca3 268 bool recalculate_flag;
db7b93e3 269 bool reset_recalculate_flag;
84597a44 270 bool discard;
5c024064 271 bool fix_padding;
09d85f8d 272 bool fix_hmac;
5c024064 273 bool legacy_recalculate;
7eada909
MP		 274
275 struct alg_spec internal_hash_alg;
276 struct alg_spec journal_crypt_alg;
277 struct alg_spec journal_mac_alg;
3f2e5393
MP		 278
279 atomic64_t number_of_mismatches;
1f5a7759
MP		 280
281 struct notifier_block reboot_notifier;
7eada909
MP		 282};
283
284struct dm_integrity_range {
285 sector_t logical_sector;
4f43446d 286 sector_t n_sectors;
724376a0
MP		 287 bool waiting;
288 union {
289 struct rb_node node;
290 struct {
291 struct task_struct *task;
292 struct list_head wait_entry;
293 };
294 };
7eada909
MP		 295};
296
297struct dm_integrity_io {
298 struct work_struct work;
299
300 struct dm_integrity_c *ic;
ff07a02e 301 enum req_op op;
7eada909
MP		 302 bool fua;
303
304 struct dm_integrity_range range;
305
306 sector_t metadata_block;
86a3238c 307 unsigned int metadata_offset;
7eada909
MP		 308
309 atomic_t in_flight;
4e4cbee9 310 blk_status_t bi_status;
7eada909
MP		 311
312 struct completion *completion;
313
248aa264 314 struct dm_bio_details bio_details;
7eada909
MP		 315};
316
317struct journal_completion {
318 struct dm_integrity_c *ic;
319 atomic_t in_flight;
320 struct completion comp;
321};
322
323struct journal_io {
324 struct dm_integrity_range range;
325 struct journal_completion *comp;
326};
327
468dfca3
MP		 328struct bitmap_block_status {
329 struct work_struct work;
330 struct dm_integrity_c *ic;
86a3238c 331 unsigned int idx;
468dfca3
MP		 332 unsigned long *bitmap;
333 struct bio_list bio_queue;
334 spinlock_t bio_queue_lock;
335
336};
337
7eada909
MP		 338static struct kmem_cache *journal_io_cache;
339
340#define JOURNAL_IO_MEMPOOL 32
341
342#ifdef DEBUG_PRINT
25c9a4ab
AS		 343#define DEBUG_print(x, ...) printk(KERN_DEBUG x, ##__VA_ARGS__)
344#define DEBUG_bytes(bytes, len, msg, ...) printk(KERN_DEBUG msg "%s%*ph\n", ##__VA_ARGS__, \
345 len ? ": " : "", len, bytes)
7eada909
MP		 346#else
347#define DEBUG_print(x, ...) do { } while (0)
348#define DEBUG_bytes(bytes, len, msg, ...) do { } while (0)
349#endif
350
54d4e6ab
MG		 351static void dm_integrity_prepare(struct request *rq)
352{
353}
354
355static void dm_integrity_complete(struct request *rq, unsigned int nr_bytes)
356{
357}
358
7eada909
MP		 359/*
360 * DM Integrity profile, protection is performed layer above (dm-crypt)
361 */
7c373d66 362static const struct blk_integrity_profile dm_integrity_profile = {
7eada909
MP		 363 .name = "DM-DIF-EXT-TAG",
364 .generate_fn = NULL,
365 .verify_fn = NULL,
54d4e6ab
MG		 366 .prepare_fn = dm_integrity_prepare,
367 .complete_fn = dm_integrity_complete,
7eada909
MP		 368};
369
370static void dm_integrity_map_continue(struct dm_integrity_io *dio, bool from_map);
371static void integrity_bio_wait(struct work_struct *w);
372static void dm_integrity_dtr(struct dm_target *ti);
373
374static void dm_integrity_io_error(struct dm_integrity_c *ic, const char *msg, int err)
375{
3f2e5393
MP		 376 if (err == -EILSEQ)
377 atomic64_inc(&ic->number_of_mismatches);
7eada909
MP		 378 if (!cmpxchg(&ic->failed, 0, err))
379 DMERR("Error on %s: %d", msg, err);
380}
381
382static int dm_integrity_failed(struct dm_integrity_c *ic)
383{
d3e632f0 384 return READ_ONCE(ic->failed);
7eada909
MP		 385}
386
5c024064
MP		 387static bool dm_integrity_disable_recalculate(struct dm_integrity_c *ic)
388{
09d85f8d
MP		 389 if (ic->legacy_recalculate)
390 return false;
391 if (!(ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) ?
392 ic->internal_hash_alg.key || ic->journal_mac_alg.key :
393 ic->internal_hash_alg.key && !ic->journal_mac_alg.key)
5c024064
MP		 394 return true;
395 return false;
396}
397
86a3238c
HM		 398static commit_id_t dm_integrity_commit_id(struct dm_integrity_c *ic, unsigned int i,
399 unsigned int j, unsigned char seq)
7eada909
MP		 400{
401 /*
402 * Xor the number with section and sector, so that if a piece of
403 * journal is written at wrong place, it is detected.
404 */
405 return ic->commit_ids[seq] ^ cpu_to_le64(((__u64)i << 32) ^ j);
406}
407
408static void get_area_and_offset(struct dm_integrity_c *ic, sector_t data_sector,
409 sector_t *area, sector_t *offset)
410{
356d9d52
MP		 411 if (!ic->meta_dev) {
412 __u8 log2_interleave_sectors = ic->sb->log2_interleave_sectors;
413 *area = data_sector >> log2_interleave_sectors;
86a3238c 414 *offset = (unsigned int)data_sector & ((1U << log2_interleave_sectors) - 1);
356d9d52
MP		 415 } else {
416 *area = 0;
417 *offset = data_sector;
418 }
7eada909
MP		 419}
420
9d609f85
MP		 421#define sector_to_block(ic, n) \
422do { \
86a3238c 423 BUG_ON((n) & (unsigned int)((ic)->sectors_per_block - 1)); \
9d609f85
MP		 424 (n) >>= (ic)->sb->log2_sectors_per_block; \
425} while (0)
426
7eada909 427static __u64 get_metadata_sector_and_offset(struct dm_integrity_c *ic, sector_t area,
86a3238c 428 sector_t offset, unsigned int *metadata_offset)
7eada909
MP		 429{
430 __u64 ms;
86a3238c 431 unsigned int mo;
7eada909
MP		 432
433 ms = area << ic->sb->log2_interleave_sectors;
434 if (likely(ic->log2_metadata_run >= 0))
435 ms += area << ic->log2_metadata_run;
436 else
437 ms += area * ic->metadata_run;
438 ms >>= ic->log2_buffer_sectors;
439
9d609f85
MP		 440 sector_to_block(ic, offset);
441
7eada909
MP		 442 if (likely(ic->log2_tag_size >= 0)) {
443 ms += offset >> (SECTOR_SHIFT + ic->log2_buffer_sectors - ic->log2_tag_size);
444 mo = (offset << ic->log2_tag_size) & ((1U << SECTOR_SHIFT << ic->log2_buffer_sectors) - 1);
445 } else {
446 ms += (__u64)offset * ic->tag_size >> (SECTOR_SHIFT + ic->log2_buffer_sectors);
447 mo = (offset * ic->tag_size) & ((1U << SECTOR_SHIFT << ic->log2_buffer_sectors) - 1);
448 }
449 *metadata_offset = mo;
450 return ms;
451}
452
453static sector_t get_data_sector(struct dm_integrity_c *ic, sector_t area, sector_t offset)
454{
455 sector_t result;
456
356d9d52
MP		 457 if (ic->meta_dev)
458 return offset;
459
7eada909
MP		 460 result = area << ic->sb->log2_interleave_sectors;
461 if (likely(ic->log2_metadata_run >= 0))
462 result += (area + 1) << ic->log2_metadata_run;
463 else
464 result += (area + 1) * ic->metadata_run;
465
466 result += (sector_t)ic->initial_sectors + offset;
71e9ddbc
MP		 467 result += ic->start;
468
7eada909
MP		 469 return result;
470}
471
86a3238c 472static void wraparound_section(struct dm_integrity_c *ic, unsigned int *sec_ptr)
7eada909
MP		 473{
474 if (unlikely(*sec_ptr >= ic->journal_sections))
475 *sec_ptr -= ic->journal_sections;
476}
477
1f9fc0b8
MP		 478static void sb_set_version(struct dm_integrity_c *ic)
479{
09d85f8d
MP		 480 if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC))
481 ic->sb->version = SB_VERSION_5;
482 else if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING))
d537858a
MP		 483 ic->sb->version = SB_VERSION_4;
484 else if (ic->mode == 'B' || ic->sb->flags & cpu_to_le32(SB_FLAG_DIRTY_BITMAP))
468dfca3
MP		 485 ic->sb->version = SB_VERSION_3;
486 else if (ic->meta_dev || ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))
1f9fc0b8
MP		 487 ic->sb->version = SB_VERSION_2;
488 else
489 ic->sb->version = SB_VERSION_1;
490}
491
09d85f8d
MP		 492static int sb_mac(struct dm_integrity_c *ic, bool wr)
493{
494 SHASH_DESC_ON_STACK(desc, ic->journal_mac);
495 int r;
86a3238c 496 unsigned int size = crypto_shash_digestsize(ic->journal_mac);
09d85f8d
MP		 497
498 if (sizeof(struct superblock) + size > 1 << SECTOR_SHIFT) {
499 dm_integrity_io_error(ic, "digest is too long", -EINVAL);
500 return -EINVAL;
501 }
502
503 desc->tfm = ic->journal_mac;
504
505 r = crypto_shash_init(desc);
506 if (unlikely(r < 0)) {
507 dm_integrity_io_error(ic, "crypto_shash_init", r);
508 return r;
509 }
510
511 r = crypto_shash_update(desc, (__u8 *)ic->sb, (1 << SECTOR_SHIFT) - size);
512 if (unlikely(r < 0)) {
513 dm_integrity_io_error(ic, "crypto_shash_update", r);
514 return r;
515 }
516
517 if (likely(wr)) {
518 r = crypto_shash_final(desc, (__u8 *)ic->sb + (1 << SECTOR_SHIFT) - size);
519 if (unlikely(r < 0)) {
520 dm_integrity_io_error(ic, "crypto_shash_final", r);
521 return r;
522 }
523 } else {
524 __u8 result[HASH_MAX_DIGESTSIZE];
0ef0b471 525
09d85f8d
MP		 526 r = crypto_shash_final(desc, result);
527 if (unlikely(r < 0)) {
528 dm_integrity_io_error(ic, "crypto_shash_final", r);
529 return r;
530 }
531 if (memcmp((__u8 *)ic->sb + (1 << SECTOR_SHIFT) - size, result, size)) {
532 dm_integrity_io_error(ic, "superblock mac", -EILSEQ);
82bb8599 533 dm_audit_log_target(DM_MSG_PREFIX, "mac-superblock", ic->ti, 0);
09d85f8d
MP		 534 return -EILSEQ;
535 }
536 }
537
538 return 0;
539}
540
c9154a4c 541static int sync_rw_sb(struct dm_integrity_c *ic, blk_opf_t opf)
7eada909
MP		 542{
543 struct dm_io_request io_req;
544 struct dm_io_region io_loc;
c9154a4c 545 const enum req_op op = opf & REQ_OP_MASK;
09d85f8d 546 int r;
7eada909 547
c9154a4c 548 io_req.bi_opf = opf;
7eada909
MP		 549 io_req.mem.type = DM_IO_KMEM;
550 io_req.mem.ptr.addr = ic->sb;
551 io_req.notify.fn = NULL;
552 io_req.client = ic->io;
356d9d52 553 io_loc.bdev = ic->meta_dev ? ic->meta_dev->bdev : ic->dev->bdev;
7eada909
MP		 554 io_loc.sector = ic->start;
555 io_loc.count = SB_SECTORS;
556
09d85f8d 557 if (op == REQ_OP_WRITE) {
5f1c56b3 558 sb_set_version(ic);
09d85f8d
MP		 559 if (ic->journal_mac && ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
560 r = sb_mac(ic, true);
561 if (unlikely(r))
562 return r;
563 }
564 }
5f1c56b3 565
09d85f8d
MP		 566 r = dm_io(&io_req, 1, &io_loc, NULL);
567 if (unlikely(r))
568 return r;
569
570 if (op == REQ_OP_READ) {
571 if (ic->mode != 'R' && ic->journal_mac && ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
572 r = sb_mac(ic, false);
573 if (unlikely(r))
574 return r;
575 }
576 }
577
578 return 0;
7eada909
MP		 579}
580
468dfca3
MP		 581#define BITMAP_OP_TEST_ALL_SET 0
582#define BITMAP_OP_TEST_ALL_CLEAR 1
583#define BITMAP_OP_SET 2
584#define BITMAP_OP_CLEAR 3
585
05d6909e
MS		 586static bool block_bitmap_op(struct dm_integrity_c *ic, struct page_list *bitmap,
587 sector_t sector, sector_t n_sectors, int mode)
468dfca3
MP		 588{
589 unsigned long bit, end_bit, this_end_bit, page, end_page;
590 unsigned long *data;
591
592 if (unlikely(((sector | n_sectors) & ((1 << ic->sb->log2_sectors_per_block) - 1)) != 0)) {
05d6909e 593 DMCRIT("invalid bitmap access (%llx,%llx,%d,%d,%d)",
7649194a
MP		 594 sector,
595 n_sectors,
468dfca3
MP		 596 ic->sb->log2_sectors_per_block,
597 ic->log2_blocks_per_bitmap_bit,
598 mode);
599 BUG();
600 }
601
602 if (unlikely(!n_sectors))
603 return true;
604
605 bit = sector >> (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
05d6909e
MS		 606 end_bit = (sector + n_sectors - 1) >>
607 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
468dfca3
MP		 608
609 page = bit / (PAGE_SIZE * 8);
610 bit %= PAGE_SIZE * 8;
611
612 end_page = end_bit / (PAGE_SIZE * 8);
613 end_bit %= PAGE_SIZE * 8;
614
615repeat:
2d0f25cb 616 if (page < end_page)
468dfca3 617 this_end_bit = PAGE_SIZE * 8 - 1;
2d0f25cb 618 else
468dfca3 619 this_end_bit = end_bit;
468dfca3
MP		 620
621 data = lowmem_page_address(bitmap[page].page);
622
623 if (mode == BITMAP_OP_TEST_ALL_SET) {
624 while (bit <= this_end_bit) {
625 if (!(bit % BITS_PER_LONG) && this_end_bit >= bit + BITS_PER_LONG - 1) {
626 do {
627 if (data[bit / BITS_PER_LONG] != -1)
628 return false;
629 bit += BITS_PER_LONG;
630 } while (this_end_bit >= bit + BITS_PER_LONG - 1);
631 continue;
632 }
633 if (!test_bit(bit, data))
634 return false;
635 bit++;
636 }
637 } else if (mode == BITMAP_OP_TEST_ALL_CLEAR) {
638 while (bit <= this_end_bit) {
639 if (!(bit % BITS_PER_LONG) && this_end_bit >= bit + BITS_PER_LONG - 1) {
640 do {
641 if (data[bit / BITS_PER_LONG] != 0)
642 return false;
643 bit += BITS_PER_LONG;
644 } while (this_end_bit >= bit + BITS_PER_LONG - 1);
645 continue;
646 }
647 if (test_bit(bit, data))
648 return false;
649 bit++;
650 }
651 } else if (mode == BITMAP_OP_SET) {
652 while (bit <= this_end_bit) {
653 if (!(bit % BITS_PER_LONG) && this_end_bit >= bit + BITS_PER_LONG - 1) {
654 do {
655 data[bit / BITS_PER_LONG] = -1;
656 bit += BITS_PER_LONG;
657 } while (this_end_bit >= bit + BITS_PER_LONG - 1);
658 continue;
659 }
660 __set_bit(bit, data);
661 bit++;
662 }
663 } else if (mode == BITMAP_OP_CLEAR) {
664 if (!bit && this_end_bit == PAGE_SIZE * 8 - 1)
665 clear_page(data);
03b18887
HM		 666 else {
667 while (bit <= this_end_bit) {
668 if (!(bit % BITS_PER_LONG) && this_end_bit >= bit + BITS_PER_LONG - 1) {
669 do {
670 data[bit / BITS_PER_LONG] = 0;
671 bit += BITS_PER_LONG;
672 } while (this_end_bit >= bit + BITS_PER_LONG - 1);
673 continue;
674 }
675 __clear_bit(bit, data);
676 bit++;
468dfca3 677 }
468dfca3
MP		 678 }
679 } else {
680 BUG();
681 }
682
683 if (unlikely(page < end_page)) {
684 bit = 0;
685 page++;
686 goto repeat;
687 }
688
689 return true;
690}
691
692static void block_bitmap_copy(struct dm_integrity_c *ic, struct page_list *dst, struct page_list *src)
693{
86a3238c
HM		 694 unsigned int n_bitmap_pages = DIV_ROUND_UP(ic->n_bitmap_blocks, PAGE_SIZE / BITMAP_BLOCK_SIZE);
695 unsigned int i;
468dfca3
MP		 696
697 for (i = 0; i < n_bitmap_pages; i++) {
698 unsigned long *dst_data = lowmem_page_address(dst[i].page);
699 unsigned long *src_data = lowmem_page_address(src[i].page);
0ef0b471 700
468dfca3
MP		 701 copy_page(dst_data, src_data);
702 }
703}
704
705static struct bitmap_block_status *sector_to_bitmap_block(struct dm_integrity_c *ic, sector_t sector)
706{
86a3238c
HM		 707 unsigned int bit = sector >> (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
708 unsigned int bitmap_block = bit / (BITMAP_BLOCK_SIZE * 8);
468dfca3
MP		 709
710 BUG_ON(bitmap_block >= ic->n_bitmap_blocks);
711 return &ic->bbs[bitmap_block];
712}
713
86a3238c 714static void access_journal_check(struct dm_integrity_c *ic, unsigned int section, unsigned int offset,
7eada909
MP		 715 bool e, const char *function)
716{
717#if defined(CONFIG_DM_DEBUG) || defined(INTERNAL_VERIFY)
86a3238c 718 unsigned int limit = e ? ic->journal_section_entries : ic->journal_section_sectors;
7eada909
MP		 719
720 if (unlikely(section >= ic->journal_sections) ||
721 unlikely(offset >= limit)) {
05d6909e
MS		 722 DMCRIT("%s: invalid access at (%u,%u), limit (%u,%u)",
723 function, section, offset, ic->journal_sections, limit);
7eada909
MP		 724 BUG();
725 }
726#endif
727}
728
86a3238c
HM		 729static void page_list_location(struct dm_integrity_c *ic, unsigned int section, unsigned int offset,
730 unsigned int *pl_index, unsigned int *pl_offset)
7eada909 731{
86a3238c 732 unsigned int sector;
7eada909 733
56b67a4f 734 access_journal_check(ic, section, offset, false, "page_list_location");
7eada909
MP		 735
736 sector = section * ic->journal_section_sectors + offset;
737
738 *pl_index = sector >> (PAGE_SHIFT - SECTOR_SHIFT);
739 *pl_offset = (sector << SECTOR_SHIFT) & (PAGE_SIZE - 1);
740}
741
742static struct journal_sector *access_page_list(struct dm_integrity_c *ic, struct page_list *pl,
86a3238c 743 unsigned int section, unsigned int offset, unsigned int *n_sectors)
7eada909 744{
86a3238c 745 unsigned int pl_index, pl_offset;
7eada909
MP		 746 char *va;
747
748 page_list_location(ic, section, offset, &pl_index, &pl_offset);
749
750 if (n_sectors)
751 *n_sectors = (PAGE_SIZE - pl_offset) >> SECTOR_SHIFT;
752
753 va = lowmem_page_address(pl[pl_index].page);
754
755 return (struct journal_sector *)(va + pl_offset);
756}
757
86a3238c 758static struct journal_sector *access_journal(struct dm_integrity_c *ic, unsigned int section, unsigned int offset)
7eada909
MP		 759{
760 return access_page_list(ic, ic->journal, section, offset, NULL);
761}
762
86a3238c 763static struct journal_entry *access_journal_entry(struct dm_integrity_c *ic, unsigned int section, unsigned int n)
7eada909 764{
86a3238c 765 unsigned int rel_sector, offset;
7eada909
MP		 766 struct journal_sector *js;
767
768 access_journal_check(ic, section, n, true, "access_journal_entry");
769
770 rel_sector = n % JOURNAL_BLOCK_SECTORS;
771 offset = n / JOURNAL_BLOCK_SECTORS;
772
773 js = access_journal(ic, section, rel_sector);
774 return (struct journal_entry *)((char *)js + offset * ic->journal_entry_size);
775}
776
86a3238c 777static struct journal_sector *access_journal_data(struct dm_integrity_c *ic, unsigned int section, unsigned int n)
7eada909 778{
9d609f85 779 n <<= ic->sb->log2_sectors_per_block;
7eada909 780
9d609f85
MP		 781 n += JOURNAL_BLOCK_SECTORS;
782
783 access_journal_check(ic, section, n, false, "access_journal_data");
784
785 return access_journal(ic, section, n);
7eada909
MP		 786}
787
86a3238c 788static void section_mac(struct dm_integrity_c *ic, unsigned int section, __u8 result[JOURNAL_MAC_SIZE])
7eada909
MP		 789{
790 SHASH_DESC_ON_STACK(desc, ic->journal_mac);
791 int r;
86a3238c 792 unsigned int j, size;
7eada909
MP		 793
794 desc->tfm = ic->journal_mac;
7eada909
MP		 795
796 r = crypto_shash_init(desc);
09d85f8d 797 if (unlikely(r < 0)) {
7eada909
MP		 798 dm_integrity_io_error(ic, "crypto_shash_init", r);
799 goto err;
800 }
801
09d85f8d 802 if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
bc8f3d46 803 __le64 section_le;
09d85f8d
MP		 804
805 r = crypto_shash_update(desc, (__u8 *)&ic->sb->salt, SALT_SIZE);
806 if (unlikely(r < 0)) {
807 dm_integrity_io_error(ic, "crypto_shash_update", r);
808 goto err;
809 }
810
811 section_le = cpu_to_le64(section);
8d1058fb 812 r = crypto_shash_update(desc, (__u8 *)&section_le, sizeof(section_le));
09d85f8d
MP		 813 if (unlikely(r < 0)) {
814 dm_integrity_io_error(ic, "crypto_shash_update", r);
815 goto err;
816 }
817 }
818
7eada909
MP		 819 for (j = 0; j < ic->journal_section_entries; j++) {
820 struct journal_entry *je = access_journal_entry(ic, section, j);
0ef0b471 821
8d1058fb 822 r = crypto_shash_update(desc, (__u8 *)&je->u.sector, sizeof(je->u.sector));
09d85f8d 823 if (unlikely(r < 0)) {
7eada909
MP		 824 dm_integrity_io_error(ic, "crypto_shash_update", r);
825 goto err;
826 }
827 }
828
829 size = crypto_shash_digestsize(ic->journal_mac);
830
831 if (likely(size <= JOURNAL_MAC_SIZE)) {
832 r = crypto_shash_final(desc, result);
09d85f8d 833 if (unlikely(r < 0)) {
7eada909
MP		 834 dm_integrity_io_error(ic, "crypto_shash_final", r);
835 goto err;
836 }
837 memset(result + size, 0, JOURNAL_MAC_SIZE - size);
838 } else {
6d39a124
KC		 839 __u8 digest[HASH_MAX_DIGESTSIZE];
840
841 if (WARN_ON(size > sizeof(digest))) {
842 dm_integrity_io_error(ic, "digest_size", -EINVAL);
843 goto err;
844 }
7eada909 845 r = crypto_shash_final(desc, digest);
09d85f8d 846 if (unlikely(r < 0)) {
7eada909
MP		 847 dm_integrity_io_error(ic, "crypto_shash_final", r);
848 goto err;
849 }
850 memcpy(result, digest, JOURNAL_MAC_SIZE);
851 }
852
853 return;
854err:
855 memset(result, 0, JOURNAL_MAC_SIZE);
856}
857
86a3238c 858static void rw_section_mac(struct dm_integrity_c *ic, unsigned int section, bool wr)
7eada909
MP		 859{
860 __u8 result[JOURNAL_MAC_SIZE];
86a3238c 861 unsigned int j;
7eada909
MP		 862
863 if (!ic->journal_mac)
864 return;
865
866 section_mac(ic, section, result);
867
868 for (j = 0; j < JOURNAL_BLOCK_SECTORS; j++) {
869 struct journal_sector *js = access_journal(ic, section, j);
870
871 if (likely(wr))
872 memcpy(&js->mac, result + (j * JOURNAL_MAC_PER_SECTOR), JOURNAL_MAC_PER_SECTOR);
873 else {
82bb8599 874 if (memcmp(&js->mac, result + (j * JOURNAL_MAC_PER_SECTOR), JOURNAL_MAC_PER_SECTOR)) {
7eada909 875 dm_integrity_io_error(ic, "journal mac", -EILSEQ);
82bb8599
MW		 876 dm_audit_log_target(DM_MSG_PREFIX, "mac-journal", ic->ti, 0);
877 }
7eada909
MP		 878 }
879 }
880}
881
882static void complete_journal_op(void *context)
883{
884 struct journal_completion *comp = context;
0ef0b471 885
7eada909
MP		 886 BUG_ON(!atomic_read(&comp->in_flight));
887 if (likely(atomic_dec_and_test(&comp->in_flight)))
888 complete(&comp->comp);
889}
890
86a3238c
HM		 891static void xor_journal(struct dm_integrity_c *ic, bool encrypt, unsigned int section,
892 unsigned int n_sections, struct journal_completion *comp)
7eada909
MP		 893{
894 struct async_submit_ctl submit;
895 size_t n_bytes = (size_t)(n_sections * ic->journal_section_sectors) << SECTOR_SHIFT;
86a3238c 896 unsigned int pl_index, pl_offset, section_index;
7eada909
MP		 897 struct page_list *source_pl, *target_pl;
898
899 if (likely(encrypt)) {
900 source_pl = ic->journal;
901 target_pl = ic->journal_io;
902 } else {
903 source_pl = ic->journal_io;
904 target_pl = ic->journal;
905 }
906
907 page_list_location(ic, section, 0, &pl_index, &pl_offset);
908
909 atomic_add(roundup(pl_offset + n_bytes, PAGE_SIZE) >> PAGE_SHIFT, &comp->in_flight);
910
911 init_async_submit(&submit, ASYNC_TX_XOR_ZERO_DST, NULL, complete_journal_op, comp, NULL);
912
913 section_index = pl_index;
914
915 do {
916 size_t this_step;
917 struct page *src_pages[2];
918 struct page *dst_page;
919
920 while (unlikely(pl_index == section_index)) {
86a3238c 921 unsigned int dummy;
0ef0b471 922
7eada909
MP		 923 if (likely(encrypt))
924 rw_section_mac(ic, section, true);
925 section++;
926 n_sections--;
927 if (!n_sections)
928 break;
929 page_list_location(ic, section, 0, &section_index, &dummy);
930 }
931
932 this_step = min(n_bytes, (size_t)PAGE_SIZE - pl_offset);
933 dst_page = target_pl[pl_index].page;
934 src_pages[0] = source_pl[pl_index].page;
935 src_pages[1] = ic->journal_xor[pl_index].page;
936
937 async_xor(dst_page, src_pages, pl_offset, 2, this_step, &submit);
938
939 pl_index++;
940 pl_offset = 0;
941 n_bytes -= this_step;
942 } while (n_bytes);
943
944 BUG_ON(n_sections);
945
946 async_tx_issue_pending_all();
947}
948
dcfe653d 949static void complete_journal_encrypt(void *data, int err)
7eada909 950{
dcfe653d 951 struct journal_completion *comp = data;
0ef0b471 952
7eada909
MP		 953 if (unlikely(err)) {
954 if (likely(err == -EINPROGRESS)) {
955 complete(&comp->ic->crypto_backoff);
956 return;
957 }
958 dm_integrity_io_error(comp->ic, "asynchronous encrypt", err);
959 }
960 complete_journal_op(comp);
961}
962
963static bool do_crypt(bool encrypt, struct skcipher_request *req, struct journal_completion *comp)
964{
965 int r;
0ef0b471 966
432061b3 967 skcipher_request_set_callback(req, CRYPTO_TFM_REQ_MAY_BACKLOG,
7eada909
MP		 968 complete_journal_encrypt, comp);
969 if (likely(encrypt))
970 r = crypto_skcipher_encrypt(req);
971 else
972 r = crypto_skcipher_decrypt(req);
973 if (likely(!r))
974 return false;
975 if (likely(r == -EINPROGRESS))
976 return true;
977 if (likely(r == -EBUSY)) {
978 wait_for_completion(&comp->ic->crypto_backoff);
979 reinit_completion(&comp->ic->crypto_backoff);
980 return true;
981 }
982 dm_integrity_io_error(comp->ic, "encrypt", r);
983 return false;
984}
985
86a3238c
HM		 986static void crypt_journal(struct dm_integrity_c *ic, bool encrypt, unsigned int section,
987 unsigned int n_sections, struct journal_completion *comp)
7eada909
MP		 988{
989 struct scatterlist **source_sg;
990 struct scatterlist **target_sg;
991
992 atomic_add(2, &comp->in_flight);
993
994 if (likely(encrypt)) {
995 source_sg = ic->journal_scatterlist;
996 target_sg = ic->journal_io_scatterlist;
997 } else {
998 source_sg = ic->journal_io_scatterlist;
999 target_sg = ic->journal_scatterlist;
1000 }
1001
1002 do {
1003 struct skcipher_request *req;
86a3238c 1004 unsigned int ivsize;
7eada909
MP		 1005 char *iv;
1006
1007 if (likely(encrypt))
1008 rw_section_mac(ic, section, true);
1009
1010 req = ic->sk_requests[section];
1011 ivsize = crypto_skcipher_ivsize(ic->journal_crypt);
1012 iv = req->iv;
1013
1014 memcpy(iv, iv + ivsize, ivsize);
1015
1016 req->src = source_sg[section];
1017 req->dst = target_sg[section];
1018
1019 if (unlikely(do_crypt(encrypt, req, comp)))
1020 atomic_inc(&comp->in_flight);
1021
1022 section++;
1023 n_sections--;
1024 } while (n_sections);
1025
1026 atomic_dec(&comp->in_flight);
1027 complete_journal_op(comp);
1028}
1029
86a3238c
HM		 1030static void encrypt_journal(struct dm_integrity_c *ic, bool encrypt, unsigned int section,
1031 unsigned int n_sections, struct journal_completion *comp)
7eada909
MP		 1032{
1033 if (ic->journal_xor)
1034 return xor_journal(ic, encrypt, section, n_sections, comp);
1035 else
1036 return crypt_journal(ic, encrypt, section, n_sections, comp);
1037}
1038
1039static void complete_journal_io(unsigned long error, void *context)
1040{
1041 struct journal_completion *comp = context;
0ef0b471 1042
7eada909
MP		 1043 if (unlikely(error != 0))
1044 dm_integrity_io_error(comp->ic, "writing journal", -EIO);
1045 complete_journal_op(comp);
1046}
1047
c9154a4c 1048static void rw_journal_sectors(struct dm_integrity_c *ic, blk_opf_t opf,
86a3238c 1049 unsigned int sector, unsigned int n_sectors,
c9154a4c 1050 struct journal_completion *comp)
7eada909
MP		 1051{
1052 struct dm_io_request io_req;
1053 struct dm_io_region io_loc;
86a3238c 1054 unsigned int pl_index, pl_offset;
7eada909
MP		 1055 int r;
1056
1057 if (unlikely(dm_integrity_failed(ic))) {
1058 if (comp)
1059 complete_journal_io(-1UL, comp);
1060 return;
1061 }
1062
7eada909
MP		 1063 pl_index = sector >> (PAGE_SHIFT - SECTOR_SHIFT);
1064 pl_offset = (sector << SECTOR_SHIFT) & (PAGE_SIZE - 1);
1065
c9154a4c 1066 io_req.bi_opf = opf;
7eada909
MP		 1067 io_req.mem.type = DM_IO_PAGE_LIST;
1068 if (ic->journal_io)
1069 io_req.mem.ptr.pl = &ic->journal_io[pl_index];
1070 else
1071 io_req.mem.ptr.pl = &ic->journal[pl_index];
1072 io_req.mem.offset = pl_offset;
1073 if (likely(comp != NULL)) {
1074 io_req.notify.fn = complete_journal_io;
1075 io_req.notify.context = comp;
1076 } else {
1077 io_req.notify.fn = NULL;
1078 }
1079 io_req.client = ic->io;
356d9d52 1080 io_loc.bdev = ic->meta_dev ? ic->meta_dev->bdev : ic->dev->bdev;
7eada909
MP		 1081 io_loc.sector = ic->start + SB_SECTORS + sector;
1082 io_loc.count = n_sectors;
1083
1084 r = dm_io(&io_req, 1, &io_loc, NULL);
1085 if (unlikely(r)) {
c9154a4c
BVA		 1086 dm_integrity_io_error(ic, (opf & REQ_OP_MASK) == REQ_OP_READ ?
1087 "reading journal" : "writing journal", r);
7eada909
MP		 1088 if (comp) {
1089 WARN_ONCE(1, "asynchronous dm_io failed: %d", r);
1090 complete_journal_io(-1UL, comp);
1091 }
1092 }
1093}
1094
c9154a4c 1095static void rw_journal(struct dm_integrity_c *ic, blk_opf_t opf,
86a3238c 1096 unsigned int section, unsigned int n_sections,
c9154a4c 1097 struct journal_completion *comp)
981e8a98 1098{
86a3238c 1099 unsigned int sector, n_sectors;
981e8a98
MP		 1100
1101 sector = section * ic->journal_section_sectors;
1102 n_sectors = n_sections * ic->journal_section_sectors;
1103
c9154a4c 1104 rw_journal_sectors(ic, opf, sector, n_sectors, comp);
981e8a98
MP		 1105}
1106
86a3238c 1107static void write_journal(struct dm_integrity_c *ic, unsigned int commit_start, unsigned int commit_sections)
7eada909
MP		 1108{
1109 struct journal_completion io_comp;
1110 struct journal_completion crypt_comp_1;
1111 struct journal_completion crypt_comp_2;
86a3238c 1112 unsigned int i;
7eada909
MP		 1113
1114 io_comp.ic = ic;
b5e8ad92 1115 init_completion(&io_comp.comp);
7eada909
MP		 1116
1117 if (commit_start + commit_sections <= ic->journal_sections) {
1118 io_comp.in_flight = (atomic_t)ATOMIC_INIT(1);
1119 if (ic->journal_io) {
1120 crypt_comp_1.ic = ic;
b5e8ad92 1121 init_completion(&crypt_comp_1.comp);
7eada909
MP		 1122 crypt_comp_1.in_flight = (atomic_t)ATOMIC_INIT(0);
1123 encrypt_journal(ic, true, commit_start, commit_sections, &crypt_comp_1);
1124 wait_for_completion_io(&crypt_comp_1.comp);
1125 } else {
1126 for (i = 0; i < commit_sections; i++)
1127 rw_section_mac(ic, commit_start + i, true);
1128 }
c9154a4c 1129 rw_journal(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, commit_start,
ff0361b3 1130 commit_sections, &io_comp);
7eada909 1131 } else {
86a3238c 1132 unsigned int to_end;
0ef0b471 1133
7eada909
MP		 1134 io_comp.in_flight = (atomic_t)ATOMIC_INIT(2);
1135 to_end = ic->journal_sections - commit_start;
1136 if (ic->journal_io) {
1137 crypt_comp_1.ic = ic;
b5e8ad92 1138 init_completion(&crypt_comp_1.comp);
7eada909
MP		 1139 crypt_comp_1.in_flight = (atomic_t)ATOMIC_INIT(0);
1140 encrypt_journal(ic, true, commit_start, to_end, &crypt_comp_1);
1141 if (try_wait_for_completion(&crypt_comp_1.comp)) {
c9154a4c
BVA		 1142 rw_journal(ic, REQ_OP_WRITE | REQ_FUA,
1143 commit_start, to_end, &io_comp);
b5e8ad92 1144 reinit_completion(&crypt_comp_1.comp);
7eada909
MP		 1145 crypt_comp_1.in_flight = (atomic_t)ATOMIC_INIT(0);
1146 encrypt_journal(ic, true, 0, commit_sections - to_end, &crypt_comp_1);
1147 wait_for_completion_io(&crypt_comp_1.comp);
1148 } else {
1149 crypt_comp_2.ic = ic;
b5e8ad92 1150 init_completion(&crypt_comp_2.comp);
7eada909
MP		 1151 crypt_comp_2.in_flight = (atomic_t)ATOMIC_INIT(0);
1152 encrypt_journal(ic, true, 0, commit_sections - to_end, &crypt_comp_2);
1153 wait_for_completion_io(&crypt_comp_1.comp);
c9154a4c 1154 rw_journal(ic, REQ_OP_WRITE | REQ_FUA, commit_start, to_end, &io_comp);
7eada909
MP		 1155 wait_for_completion_io(&crypt_comp_2.comp);
1156 }
1157 } else {
1158 for (i = 0; i < to_end; i++)
1159 rw_section_mac(ic, commit_start + i, true);
c9154a4c 1160 rw_journal(ic, REQ_OP_WRITE | REQ_FUA, commit_start, to_end, &io_comp);
7eada909
MP		 1161 for (i = 0; i < commit_sections - to_end; i++)
1162 rw_section_mac(ic, i, true);
1163 }
c9154a4c 1164 rw_journal(ic, REQ_OP_WRITE | REQ_FUA, 0, commit_sections - to_end, &io_comp);
7eada909
MP		 1165 }
1166
1167 wait_for_completion_io(&io_comp.comp);
1168}
1169
86a3238c
HM		 1170static void copy_from_journal(struct dm_integrity_c *ic, unsigned int section, unsigned int offset,
1171 unsigned int n_sectors, sector_t target, io_notify_fn fn, void *data)
7eada909
MP		 1172{
1173 struct dm_io_request io_req;
1174 struct dm_io_region io_loc;
1175 int r;
86a3238c 1176 unsigned int sector, pl_index, pl_offset;
7eada909 1177
86a3238c 1178 BUG_ON((target | n_sectors | offset) & (unsigned int)(ic->sectors_per_block - 1));
9d609f85 1179
7eada909
MP		 1180 if (unlikely(dm_integrity_failed(ic))) {
1181 fn(-1UL, data);
1182 return;
1183 }
1184
1185 sector = section * ic->journal_section_sectors + JOURNAL_BLOCK_SECTORS + offset;
1186
1187 pl_index = sector >> (PAGE_SHIFT - SECTOR_SHIFT);
1188 pl_offset = (sector << SECTOR_SHIFT) & (PAGE_SIZE - 1);
1189
581075e4 1190 io_req.bi_opf = REQ_OP_WRITE;
7eada909
MP		 1191 io_req.mem.type = DM_IO_PAGE_LIST;
1192 io_req.mem.ptr.pl = &ic->journal[pl_index];
1193 io_req.mem.offset = pl_offset;
1194 io_req.notify.fn = fn;
1195 io_req.notify.context = data;
1196 io_req.client = ic->io;
1197 io_loc.bdev = ic->dev->bdev;
71e9ddbc 1198 io_loc.sector = target;
7eada909
MP		 1199 io_loc.count = n_sectors;
1200
1201 r = dm_io(&io_req, 1, &io_loc, NULL);
1202 if (unlikely(r)) {
1203 WARN_ONCE(1, "asynchronous dm_io failed: %d", r);
1204 fn(-1UL, data);
1205 }
1206}
1207
724376a0
MP		 1208static bool ranges_overlap(struct dm_integrity_range *range1, struct dm_integrity_range *range2)
1209{
1210 return range1->logical_sector < range2->logical_sector + range2->n_sectors &&
4ed319c6 1211 range1->logical_sector + range1->n_sectors > range2->logical_sector;
724376a0
MP		 1212}
1213
1214static bool add_new_range(struct dm_integrity_c *ic, struct dm_integrity_range *new_range, bool check_waiting)
7eada909
MP		 1215{
1216 struct rb_node **n = &ic->in_progress.rb_node;
1217 struct rb_node *parent;
1218
86a3238c 1219 BUG_ON((new_range->logical_sector | new_range->n_sectors) & (unsigned int)(ic->sectors_per_block - 1));
9d609f85 1220
724376a0
MP		 1221 if (likely(check_waiting)) {
1222 struct dm_integrity_range *range;
0ef0b471 1223
724376a0
MP		 1224 list_for_each_entry(range, &ic->wait_list, wait_entry) {
1225 if (unlikely(ranges_overlap(range, new_range)))
1226 return false;
1227 }
1228 }
1229
7eada909
MP		 1230 parent = NULL;
1231
1232 while (*n) {
1233 struct dm_integrity_range *range = container_of(*n, struct dm_integrity_range, node);
1234
1235 parent = *n;
2d0f25cb 1236 if (new_range->logical_sector + new_range->n_sectors <= range->logical_sector)
7eada909 1237 n = &range->node.rb_left;
2d0f25cb 1238 else if (new_range->logical_sector >= range->logical_sector + range->n_sectors)
7eada909 1239 n = &range->node.rb_right;
2d0f25cb 1240 else
7eada909 1241 return false;
7eada909
MP		 1242 }
1243
1244 rb_link_node(&new_range->node, parent, n);
1245 rb_insert_color(&new_range->node, &ic->in_progress);
1246
1247 return true;
1248}
1249
1250static void remove_range_unlocked(struct dm_integrity_c *ic, struct dm_integrity_range *range)
1251{
1252 rb_erase(&range->node, &ic->in_progress);
724376a0
MP		 1253 while (unlikely(!list_empty(&ic->wait_list))) {
1254 struct dm_integrity_range *last_range =
1255 list_first_entry(&ic->wait_list, struct dm_integrity_range, wait_entry);
1256 struct task_struct *last_range_task;
0ef0b471 1257
724376a0
MP		 1258 last_range_task = last_range->task;
1259 list_del(&last_range->wait_entry);
1260 if (!add_new_range(ic, last_range, false)) {
1261 last_range->task = last_range_task;
1262 list_add(&last_range->wait_entry, &ic->wait_list);
1263 break;
1264 }
1265 last_range->waiting = false;
1266 wake_up_process(last_range_task);
1267 }
7eada909
MP		 1268}
1269
1270static void remove_range(struct dm_integrity_c *ic, struct dm_integrity_range *range)
1271{
1272 unsigned long flags;
1273
1274 spin_lock_irqsave(&ic->endio_wait.lock, flags);
1275 remove_range_unlocked(ic, range);
1276 spin_unlock_irqrestore(&ic->endio_wait.lock, flags);
1277}
1278
724376a0
MP		 1279static void wait_and_add_new_range(struct dm_integrity_c *ic, struct dm_integrity_range *new_range)
1280{
1281 new_range->waiting = true;
1282 list_add_tail(&new_range->wait_entry, &ic->wait_list);
1283 new_range->task = current;
1284 do {
1285 __set_current_state(TASK_UNINTERRUPTIBLE);
1286 spin_unlock_irq(&ic->endio_wait.lock);
1287 io_schedule();
1288 spin_lock_irq(&ic->endio_wait.lock);
1289 } while (unlikely(new_range->waiting));
1290}
1291
8b3bbd49
MP		 1292static void add_new_range_and_wait(struct dm_integrity_c *ic, struct dm_integrity_range *new_range)
1293{
1294 if (unlikely(!add_new_range(ic, new_range, true)))
1295 wait_and_add_new_range(ic, new_range);
1296}
1297
7eada909
MP		 1298static void init_journal_node(struct journal_node *node)
1299{
1300 RB_CLEAR_NODE(&node->node);
1301 node->sector = (sector_t)-1;
1302}
1303
1304static void add_journal_node(struct dm_integrity_c *ic, struct journal_node *node, sector_t sector)
1305{
1306 struct rb_node **link;
1307 struct rb_node *parent;
1308
1309 node->sector = sector;
1310 BUG_ON(!RB_EMPTY_NODE(&node->node));
1311
1312 link = &ic->journal_tree_root.rb_node;
1313 parent = NULL;
1314
1315 while (*link) {
1316 struct journal_node *j;
0ef0b471 1317
7eada909
MP		 1318 parent = *link;
1319 j = container_of(parent, struct journal_node, node);
1320 if (sector < j->sector)
1321 link = &j->node.rb_left;
1322 else
1323 link = &j->node.rb_right;
1324 }
1325
1326 rb_link_node(&node->node, parent, link);
1327 rb_insert_color(&node->node, &ic->journal_tree_root);
1328}
1329
1330static void remove_journal_node(struct dm_integrity_c *ic, struct journal_node *node)
1331{
1332 BUG_ON(RB_EMPTY_NODE(&node->node));
1333 rb_erase(&node->node, &ic->journal_tree_root);
1334 init_journal_node(node);
1335}
1336
1337#define NOT_FOUND (-1U)
1338
86a3238c 1339static unsigned int find_journal_node(struct dm_integrity_c *ic, sector_t sector, sector_t *next_sector)
7eada909
MP		 1340{
1341 struct rb_node *n = ic->journal_tree_root.rb_node;
86a3238c 1342 unsigned int found = NOT_FOUND;
0ef0b471 1343
7eada909
MP		 1344 *next_sector = (sector_t)-1;
1345 while (n) {
1346 struct journal_node *j = container_of(n, struct journal_node, node);
0ef0b471 1347
2d0f25cb 1348 if (sector == j->sector)
7eada909 1349 found = j - ic->journal_tree;
2d0f25cb 1350
7eada909
MP		 1351 if (sector < j->sector) {
1352 *next_sector = j->sector;
1353 n = j->node.rb_left;
2d0f25cb 1354 } else
7eada909 1355 n = j->node.rb_right;
7eada909
MP		 1356 }
1357
1358 return found;
1359}
1360
86a3238c 1361static bool test_journal_node(struct dm_integrity_c *ic, unsigned int pos, sector_t sector)
7eada909
MP		 1362{
1363 struct journal_node *node, *next_node;
1364 struct rb_node *next;
1365
1366 if (unlikely(pos >= ic->journal_entries))
1367 return false;
1368 node = &ic->journal_tree[pos];
1369 if (unlikely(RB_EMPTY_NODE(&node->node)))
1370 return false;
1371 if (unlikely(node->sector != sector))
1372 return false;
1373
1374 next = rb_next(&node->node);
1375 if (unlikely(!next))
1376 return true;
1377
1378 next_node = container_of(next, struct journal_node, node);
1379 return next_node->sector != sector;
1380}
1381
1382static bool find_newer_committed_node(struct dm_integrity_c *ic, struct journal_node *node)
1383{
1384 struct rb_node *next;
1385 struct journal_node *next_node;
86a3238c 1386 unsigned int next_section;
7eada909
MP		 1387
1388 BUG_ON(RB_EMPTY_NODE(&node->node));
1389
1390 next = rb_next(&node->node);
1391 if (unlikely(!next))
1392 return false;
1393
1394 next_node = container_of(next, struct journal_node, node);
1395
1396 if (next_node->sector != node->sector)
1397 return false;
1398
86a3238c 1399 next_section = (unsigned int)(next_node - ic->journal_tree) / ic->journal_section_entries;
7eada909
MP		 1400 if (next_section >= ic->committed_section &&
1401 next_section < ic->committed_section + ic->n_committed_sections)
1402 return true;
1403 if (next_section + ic->journal_sections < ic->committed_section + ic->n_committed_sections)
1404 return true;
1405
1406 return false;
1407}
1408
1409#define TAG_READ 0
1410#define TAG_WRITE 1
1411#define TAG_CMP 2
1412
1413static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, sector_t *metadata_block,
86a3238c 1414 unsigned int *metadata_offset, unsigned int total_size, int op)
7eada909 1415{
84597a44
MP		 1416#define MAY_BE_FILLER 1
1417#define MAY_BE_HASH 2
86a3238c
HM		 1418 unsigned int hash_offset = 0;
1419 unsigned int may_be = MAY_BE_HASH | (ic->discard ? MAY_BE_FILLER : 0);
84597a44 1420
7eada909
MP		 1421 do {
1422 unsigned char *data, *dp;
1423 struct dm_buffer *b;
86a3238c 1424 unsigned int to_copy;
7eada909
MP		 1425 int r;
1426
1427 r = dm_integrity_failed(ic);
1428 if (unlikely(r))
1429 return r;
1430
1431 data = dm_bufio_read(ic->bufio, *metadata_block, &b);
5e3d0e37 1432 if (IS_ERR(data))
7eada909
MP		 1433 return PTR_ERR(data);
1434
1435 to_copy = min((1U << SECTOR_SHIFT << ic->log2_buffer_sectors) - *metadata_offset, total_size);
1436 dp = data + *metadata_offset;
1437 if (op == TAG_READ) {
1438 memcpy(tag, dp, to_copy);
1439 } else if (op == TAG_WRITE) {
a9c0fda4
MP		 1440 if (memcmp(dp, tag, to_copy)) {
1441 memcpy(dp, tag, to_copy);
1442 dm_bufio_mark_partial_buffer_dirty(b, *metadata_offset, *metadata_offset + to_copy);
1443 }
84597a44 1444 } else {
7eada909 1445 /* e.g.: op == TAG_CMP */
7eada909 1446
84597a44
MP		 1447 if (likely(is_power_of_2(ic->tag_size))) {
1448 if (unlikely(memcmp(dp, tag, to_copy)))
1449 if (unlikely(!ic->discard) ||
8267d8fb 1450 unlikely(memchr_inv(dp, DISCARD_FILLER, to_copy) != NULL)) {
84597a44
MP		 1451 goto thorough_test;
1452 }
1453 } else {
86a3238c 1454 unsigned int i, ts;
84597a44
MP		 1455thorough_test:
1456 ts = total_size;
1457
1458 for (i = 0; i < to_copy; i++, ts--) {
1459 if (unlikely(dp[i] != tag[i]))
1460 may_be &= ~MAY_BE_HASH;
1461 if (likely(dp[i] != DISCARD_FILLER))
1462 may_be &= ~MAY_BE_FILLER;
1463 hash_offset++;
1464 if (unlikely(hash_offset == ic->tag_size)) {
1465 if (unlikely(!may_be)) {
1466 dm_bufio_release(b);
1467 return ts;
1468 }
1469 hash_offset = 0;
1470 may_be = MAY_BE_HASH | (ic->discard ? MAY_BE_FILLER : 0);
1471 }
7eada909 1472 }
7eada909
MP		 1473 }
1474 }
1475 dm_bufio_release(b);
1476
1477 tag += to_copy;
1478 *metadata_offset += to_copy;
1479 if (unlikely(*metadata_offset == 1U << SECTOR_SHIFT << ic->log2_buffer_sectors)) {
1480 (*metadata_block)++;
1481 *metadata_offset = 0;
1482 }
84597a44 1483
2d0f25cb 1484 if (unlikely(!is_power_of_2(ic->tag_size)))
84597a44 1485 hash_offset = (hash_offset + to_copy) % ic->tag_size;
84597a44 1486
7eada909
MP		 1487 total_size -= to_copy;
1488 } while (unlikely(total_size));
1489
1490 return 0;
84597a44
MP		 1491#undef MAY_BE_FILLER
1492#undef MAY_BE_HASH
7eada909
MP		 1493}
1494
9b594826
MP		 1495struct flush_request {
1496 struct dm_io_request io_req;
1497 struct dm_io_region io_reg;
1498 struct dm_integrity_c *ic;
1499 struct completion comp;
1500};
1501
1502static void flush_notify(unsigned long error, void *fr_)
1503{
1504 struct flush_request *fr = fr_;
0ef0b471 1505
9b594826 1506 if (unlikely(error != 0))
23c4ecbc 1507 dm_integrity_io_error(fr->ic, "flushing disk cache", -EIO);
9b594826
MP		 1508 complete(&fr->comp);
1509}
1510
1511static void dm_integrity_flush_buffers(struct dm_integrity_c *ic, bool flush_data)
7eada909
MP		 1512{
1513 int r;
9b594826
MP		 1514 struct flush_request fr;
1515
1516 if (!ic->meta_dev)
1517 flush_data = false;
1518 if (flush_data) {
581075e4 1519 fr.io_req.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC,
9b594826
MP		 1520 fr.io_req.mem.type = DM_IO_KMEM,
1521 fr.io_req.mem.ptr.addr = NULL,
1522 fr.io_req.notify.fn = flush_notify,
1523 fr.io_req.notify.context = &fr;
1524 fr.io_req.client = dm_bufio_get_dm_io_client(ic->bufio),
1525 fr.io_reg.bdev = ic->dev->bdev,
1526 fr.io_reg.sector = 0,
1527 fr.io_reg.count = 0,
1528 fr.ic = ic;
1529 init_completion(&fr.comp);
1530 r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL);
1531 BUG_ON(r);
1532 }
1533
7eada909
MP		 1534 r = dm_bufio_write_dirty_buffers(ic->bufio);
1535 if (unlikely(r))
1536 dm_integrity_io_error(ic, "writing tags", r);
9b594826
MP		 1537
1538 if (flush_data)
1539 wait_for_completion(&fr.comp);
7eada909
MP		 1540}
1541
1542static void sleep_on_endio_wait(struct dm_integrity_c *ic)
1543{
1544 DECLARE_WAITQUEUE(wait, current);
0ef0b471 1545
7eada909
MP		 1546 __add_wait_queue(&ic->endio_wait, &wait);
1547 __set_current_state(TASK_UNINTERRUPTIBLE);
1548 spin_unlock_irq(&ic->endio_wait.lock);
1549 io_schedule();
1550 spin_lock_irq(&ic->endio_wait.lock);
1551 __remove_wait_queue(&ic->endio_wait, &wait);
1552}
1553
8376d3c1 1554static void autocommit_fn(struct timer_list *t)
7eada909 1555{
8376d3c1 1556 struct dm_integrity_c *ic = from_timer(ic, t, autocommit_timer);
7eada909
MP		 1557
1558 if (likely(!dm_integrity_failed(ic)))
1559 queue_work(ic->commit_wq, &ic->commit_work);
1560}
1561
1562static void schedule_autocommit(struct dm_integrity_c *ic)
1563{
1564 if (!timer_pending(&ic->autocommit_timer))
1565 mod_timer(&ic->autocommit_timer, jiffies + ic->autocommit_jiffies);
1566}
1567
1568static void submit_flush_bio(struct dm_integrity_c *ic, struct dm_integrity_io *dio)
1569{
1570 struct bio *bio;
7def52b7
MS		 1571 unsigned long flags;
1572
1573 spin_lock_irqsave(&ic->endio_wait.lock, flags);
7eada909
MP		 1574 bio = dm_bio_from_per_bio_data(dio, sizeof(struct dm_integrity_io));
1575 bio_list_add(&ic->flush_bio_list, bio);
7def52b7
MS		 1576 spin_unlock_irqrestore(&ic->endio_wait.lock, flags);
1577
7eada909
MP		 1578 queue_work(ic->commit_wq, &ic->commit_work);
1579}
1580
1581static void do_endio(struct dm_integrity_c *ic, struct bio *bio)
1582{
0ef0b471
HM		 1583 int r;
1584
1585 r = dm_integrity_failed(ic);
4e4cbee9
CH		 1586 if (unlikely(r) && !bio->bi_status)
1587 bio->bi_status = errno_to_blk_status(r);
48271493
MP		 1588 if (unlikely(ic->synchronous_mode) && bio_op(bio) == REQ_OP_WRITE) {
1589 unsigned long flags;
0ef0b471 1590
48271493
MP		 1591 spin_lock_irqsave(&ic->endio_wait.lock, flags);
1592 bio_list_add(&ic->synchronous_bios, bio);
1593 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, 0);
1594 spin_unlock_irqrestore(&ic->endio_wait.lock, flags);
1595 return;
1596 }
7eada909
MP		 1597 bio_endio(bio);
1598}
1599
1600static void do_endio_flush(struct dm_integrity_c *ic, struct dm_integrity_io *dio)
1601{
1602 struct bio *bio = dm_bio_from_per_bio_data(dio, sizeof(struct dm_integrity_io));
1603
4e4cbee9 1604 if (unlikely(dio->fua) && likely(!bio->bi_status) && likely(!dm_integrity_failed(ic)))
7eada909
MP		 1605 submit_flush_bio(ic, dio);
1606 else
1607 do_endio(ic, bio);
1608}
1609
1610static void dec_in_flight(struct dm_integrity_io *dio)
1611{
1612 if (atomic_dec_and_test(&dio->in_flight)) {
1613 struct dm_integrity_c *ic = dio->ic;
1614 struct bio *bio;
1615
1616 remove_range(ic, &dio->range);
1617
84597a44 1618 if (dio->op == REQ_OP_WRITE || unlikely(dio->op == REQ_OP_DISCARD))
7eada909
MP		 1619 schedule_autocommit(ic);
1620
1621 bio = dm_bio_from_per_bio_data(dio, sizeof(struct dm_integrity_io));
4e4cbee9
CH		 1622 if (unlikely(dio->bi_status) && !bio->bi_status)
1623 bio->bi_status = dio->bi_status;
1624 if (likely(!bio->bi_status) && unlikely(bio_sectors(bio) != dio->range.n_sectors)) {
7eada909
MP		 1625 dio->range.logical_sector += dio->range.n_sectors;
1626 bio_advance(bio, dio->range.n_sectors << SECTOR_SHIFT);
1627 INIT_WORK(&dio->work, integrity_bio_wait);
53770f0e 1628 queue_work(ic->offload_wq, &dio->work);
7eada909
MP		 1629 return;
1630 }
1631 do_endio_flush(ic, dio);
1632 }
1633}
1634
1635static void integrity_end_io(struct bio *bio)
1636{
1637 struct dm_integrity_io *dio = dm_per_bio_data(bio, sizeof(struct dm_integrity_io));
1638
248aa264
MS		 1639 dm_bio_restore(&dio->bio_details, bio);
1640 if (bio->bi_integrity)
7eada909 1641 bio->bi_opf |= REQ_INTEGRITY;
7eada909
MP		 1642
1643 if (dio->completion)
1644 complete(dio->completion);
1645
1646 dec_in_flight(dio);
1647}
1648
1649static void integrity_sector_checksum(struct dm_integrity_c *ic, sector_t sector,
1650 const char *data, char *result)
1651{
bc8f3d46 1652 __le64 sector_le = cpu_to_le64(sector);
7eada909
MP		 1653 SHASH_DESC_ON_STACK(req, ic->internal_hash);
1654 int r;
86a3238c 1655 unsigned int digest_size;
7eada909
MP		 1656
1657 req->tfm = ic->internal_hash;
7eada909
MP		 1658
1659 r = crypto_shash_init(req);
1660 if (unlikely(r < 0)) {
1661 dm_integrity_io_error(ic, "crypto_shash_init", r);
1662 goto failed;
1663 }
1664
09d85f8d
MP		 1665 if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
1666 r = crypto_shash_update(req, (__u8 *)&ic->sb->salt, SALT_SIZE);
1667 if (unlikely(r < 0)) {
1668 dm_integrity_io_error(ic, "crypto_shash_update", r);
1669 goto failed;
1670 }
1671 }
1672
8d1058fb 1673 r = crypto_shash_update(req, (const __u8 *)&sector_le, sizeof(sector_le));
7eada909
MP		 1674 if (unlikely(r < 0)) {
1675 dm_integrity_io_error(ic, "crypto_shash_update", r);
1676 goto failed;
1677 }
1678
9d609f85 1679 r = crypto_shash_update(req, data, ic->sectors_per_block << SECTOR_SHIFT);
7eada909
MP		 1680 if (unlikely(r < 0)) {
1681 dm_integrity_io_error(ic, "crypto_shash_update", r);
1682 goto failed;
1683 }
1684
1685 r = crypto_shash_final(req, result);
1686 if (unlikely(r < 0)) {
1687 dm_integrity_io_error(ic, "crypto_shash_final", r);
1688 goto failed;
1689 }
1690
1691 digest_size = crypto_shash_digestsize(ic->internal_hash);
1692 if (unlikely(digest_size < ic->tag_size))
1693 memset(result + digest_size, 0, ic->tag_size - digest_size);
1694
1695 return;
1696
1697failed:
1698 /* this shouldn't happen anyway, the hash functions have no reason to fail */
1699 get_random_bytes(result, ic->tag_size);
1700}
1701
1702static void integrity_metadata(struct work_struct *w)
1703{
1704 struct dm_integrity_io *dio = container_of(w, struct dm_integrity_io, work);
1705 struct dm_integrity_c *ic = dio->ic;
1706
1707 int r;
1708
1709 if (ic->internal_hash) {
1710 struct bvec_iter iter;
1711 struct bio_vec bv;
86a3238c 1712 unsigned int digest_size = crypto_shash_digestsize(ic->internal_hash);
7eada909
MP		 1713 struct bio *bio = dm_bio_from_per_bio_data(dio, sizeof(struct dm_integrity_io));
1714 char *checksums;
86a3238c 1715 unsigned int extra_space = unlikely(digest_size > ic->tag_size) ? digest_size - ic->tag_size : 0;
ba287d7c 1716 char checksums_onstack[max_t(size_t, HASH_MAX_DIGESTSIZE, MAX_TAG_SIZE)];
84597a44 1717 sector_t sector;
86a3238c 1718 unsigned int sectors_to_process;
7eada909 1719
c2bcb2b7
MP		 1720 if (unlikely(ic->mode == 'R'))
1721 goto skip_io;
1722
84597a44
MP		 1723 if (likely(dio->op != REQ_OP_DISCARD))
1724 checksums = kmalloc((PAGE_SIZE >> SECTOR_SHIFT >> ic->sb->log2_sectors_per_block) * ic->tag_size + extra_space,
1725 GFP_NOIO | __GFP_NORETRY | __GFP_NOWARN);
1726 else
1727 checksums = kmalloc(PAGE_SIZE, GFP_NOIO | __GFP_NORETRY | __GFP_NOWARN);
6d39a124 1728 if (!checksums) {
7eada909 1729 checksums = checksums_onstack;
6d39a124
KC		 1730 if (WARN_ON(extra_space &&
1731 digest_size > sizeof(checksums_onstack))) {
1732 r = -EINVAL;
1733 goto error;
1734 }
1735 }
7eada909 1736
84597a44 1737 if (unlikely(dio->op == REQ_OP_DISCARD)) {
86a3238c
HM		 1738 unsigned int bi_size = dio->bio_details.bi_iter.bi_size;
1739 unsigned int max_size = likely(checksums != checksums_onstack) ? PAGE_SIZE : HASH_MAX_DIGESTSIZE;
1740 unsigned int max_blocks = max_size / ic->tag_size;
0ef0b471 1741
84597a44
MP		 1742 memset(checksums, DISCARD_FILLER, max_size);
1743
1744 while (bi_size) {
86a3238c 1745 unsigned int this_step_blocks = bi_size >> (SECTOR_SHIFT + ic->sb->log2_sectors_per_block);
0ef0b471 1746
84597a44
MP		 1747 this_step_blocks = min(this_step_blocks, max_blocks);
1748 r = dm_integrity_rw_tag(ic, checksums, &dio->metadata_block, &dio->metadata_offset,
1749 this_step_blocks * ic->tag_size, TAG_WRITE);
1750 if (unlikely(r)) {
1751 if (likely(checksums != checksums_onstack))
1752 kfree(checksums);
1753 goto error;
1754 }
1755
84597a44 1756 bi_size -= this_step_blocks << (SECTOR_SHIFT + ic->sb->log2_sectors_per_block);
84597a44
MP		 1757 }
1758
1759 if (likely(checksums != checksums_onstack))
1760 kfree(checksums);
1761 goto skip_io;
1762 }
1763
84597a44
MP		 1764 sector = dio->range.logical_sector;
1765 sectors_to_process = dio->range.n_sectors;
1766
248aa264 1767 __bio_for_each_segment(bv, bio, iter, dio->bio_details.bi_iter) {
86a3238c 1768 unsigned int pos;
7eada909
MP		 1769 char *mem, *checksums_ptr;
1770
1771again:
c12d205d 1772 mem = bvec_kmap_local(&bv);
7eada909
MP		 1773 pos = 0;
1774 checksums_ptr = checksums;
1775 do {
1776 integrity_sector_checksum(ic, sector, mem + pos, checksums_ptr);
1777 checksums_ptr += ic->tag_size;
9d609f85
MP		 1778 sectors_to_process -= ic->sectors_per_block;
1779 pos += ic->sectors_per_block << SECTOR_SHIFT;
1780 sector += ic->sectors_per_block;
7eada909 1781 } while (pos < bv.bv_len && sectors_to_process && checksums != checksums_onstack);
c12d205d 1782 kunmap_local(mem);
7eada909
MP		 1783
1784 r = dm_integrity_rw_tag(ic, checksums, &dio->metadata_block, &dio->metadata_offset,
84597a44 1785 checksums_ptr - checksums, dio->op == REQ_OP_READ ? TAG_CMP : TAG_WRITE);
7eada909
MP		 1786 if (unlikely(r)) {
1787 if (r > 0) {
82bb8599
MW		 1788 sector_t s;
1789
1790 s = sector - ((r + ic->tag_size - 1) / ic->tag_size);
0a806cfd
CH		 1791 DMERR_LIMIT("%pg: Checksum failed at sector 0x%llx",
1792 bio->bi_bdev, s);
7eada909 1793 r = -EILSEQ;
3f2e5393 1794 atomic64_inc(&ic->number_of_mismatches);
82bb8599
MW		 1795 dm_audit_log_bio(DM_MSG_PREFIX, "integrity-checksum",
1796 bio, s, 0);
7eada909
MP		 1797 }
1798 if (likely(checksums != checksums_onstack))
1799 kfree(checksums);
1800 goto error;
1801 }
1802
1803 if (!sectors_to_process)
1804 break;
1805
1806 if (unlikely(pos < bv.bv_len)) {
1807 bv.bv_offset += pos;
1808 bv.bv_len -= pos;
1809 goto again;
1810 }
1811 }
1812
1813 if (likely(checksums != checksums_onstack))
1814 kfree(checksums);
1815 } else {
248aa264 1816 struct bio_integrity_payload *bip = dio->bio_details.bi_integrity;
7eada909
MP		 1817
1818 if (bip) {
1819 struct bio_vec biv;
1820 struct bvec_iter iter;
86a3238c 1821 unsigned int data_to_process = dio->range.n_sectors;
0ef0b471 1822
9d609f85
MP		 1823 sector_to_block(ic, data_to_process);
1824 data_to_process *= ic->tag_size;
7eada909
MP		 1825
1826 bip_for_each_vec(biv, bip, iter) {
1827 unsigned char *tag;
86a3238c 1828 unsigned int this_len;
7eada909
MP		 1829
1830 BUG_ON(PageHighMem(biv.bv_page));
964cacfd 1831 tag = bvec_virt(&biv);
7eada909
MP		 1832 this_len = min(biv.bv_len, data_to_process);
1833 r = dm_integrity_rw_tag(ic, tag, &dio->metadata_block, &dio->metadata_offset,
84597a44 1834 this_len, dio->op == REQ_OP_READ ? TAG_READ : TAG_WRITE);
7eada909
MP		 1835 if (unlikely(r))
1836 goto error;
1837 data_to_process -= this_len;
1838 if (!data_to_process)
1839 break;
1840 }
1841 }
1842 }
c2bcb2b7 1843skip_io:
7eada909
MP		 1844 dec_in_flight(dio);
1845 return;
1846error:
4e4cbee9 1847 dio->bi_status = errno_to_blk_status(r);
7eada909
MP		 1848 dec_in_flight(dio);
1849}
1850
1851static int dm_integrity_map(struct dm_target *ti, struct bio *bio)
1852{
1853 struct dm_integrity_c *ic = ti->private;
1854 struct dm_integrity_io *dio = dm_per_bio_data(bio, sizeof(struct dm_integrity_io));
9d609f85 1855 struct bio_integrity_payload *bip;
7eada909
MP		 1856
1857 sector_t area, offset;
1858
1859 dio->ic = ic;
4e4cbee9 1860 dio->bi_status = 0;
84597a44
MP		 1861 dio->op = bio_op(bio);
1862
1863 if (unlikely(dio->op == REQ_OP_DISCARD)) {
1864 if (ti->max_io_len) {
1865 sector_t sec = dm_target_offset(ti, bio->bi_iter.bi_sector);
86a3238c 1866 unsigned int log2_max_io_len = __fls(ti->max_io_len);
84597a44
MP		 1867 sector_t start_boundary = sec >> log2_max_io_len;
1868 sector_t end_boundary = (sec + bio_sectors(bio) - 1) >> log2_max_io_len;
0ef0b471 1869
84597a44
MP		 1870 if (start_boundary < end_boundary) {
1871 sector_t len = ti->max_io_len - (sec & (ti->max_io_len - 1));
0ef0b471 1872
84597a44
MP		 1873 dm_accept_partial_bio(bio, len);
1874 }
1875 }
1876 }
7eada909
MP		 1877
1878 if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
1879 submit_flush_bio(ic, dio);
1880 return DM_MAPIO_SUBMITTED;
1881 }
1882
1883 dio->range.logical_sector = dm_target_offset(ti, bio->bi_iter.bi_sector);
84597a44 1884 dio->fua = dio->op == REQ_OP_WRITE && bio->bi_opf & REQ_FUA;
7eada909
MP		 1885 if (unlikely(dio->fua)) {
1886 /*
1887 * Don't pass down the FUA flag because we have to flush
1888 * disk cache anyway.
1889 */
1890 bio->bi_opf &= ~REQ_FUA;
1891 }
1892 if (unlikely(dio->range.logical_sector + bio_sectors(bio) > ic->provided_data_sectors)) {
1893 DMERR("Too big sector number: 0x%llx + 0x%x > 0x%llx",
7649194a
MP		 1894 dio->range.logical_sector, bio_sectors(bio),
1895 ic->provided_data_sectors);
846785e6 1896 return DM_MAPIO_KILL;
7eada909 1897 }
86a3238c 1898 if (unlikely((dio->range.logical_sector | bio_sectors(bio)) & (unsigned int)(ic->sectors_per_block - 1))) {
9d609f85
MP		 1899 DMERR("Bio not aligned on %u sectors: 0x%llx, 0x%x",
1900 ic->sectors_per_block,
7649194a 1901 dio->range.logical_sector, bio_sectors(bio));
846785e6 1902 return DM_MAPIO_KILL;
9d609f85
MP		 1903 }
1904
84597a44 1905 if (ic->sectors_per_block > 1 && likely(dio->op != REQ_OP_DISCARD)) {
9d609f85
MP		 1906 struct bvec_iter iter;
1907 struct bio_vec bv;
0ef0b471 1908
9d609f85 1909 bio_for_each_segment(bv, bio, iter) {
95b1369a 1910 if (unlikely(bv.bv_len & ((ic->sectors_per_block << SECTOR_SHIFT) - 1))) {
9d609f85
MP		 1911 DMERR("Bio vector (%u,%u) is not aligned on %u-sector boundary",
1912 bv.bv_offset, bv.bv_len, ic->sectors_per_block);
846785e6 1913 return DM_MAPIO_KILL;
9d609f85
MP		 1914 }
1915 }
1916 }
1917
1918 bip = bio_integrity(bio);
1919 if (!ic->internal_hash) {
1920 if (bip) {
86a3238c 1921 unsigned int wanted_tag_size = bio_sectors(bio) >> ic->sb->log2_sectors_per_block;
0ef0b471 1922
9d609f85
MP		 1923 if (ic->log2_tag_size >= 0)
1924 wanted_tag_size <<= ic->log2_tag_size;
1925 else
1926 wanted_tag_size *= ic->tag_size;
1927 if (unlikely(wanted_tag_size != bip->bip_iter.bi_size)) {
05d6909e
MS		 1928 DMERR("Invalid integrity data size %u, expected %u",
1929 bip->bip_iter.bi_size, wanted_tag_size);
846785e6 1930 return DM_MAPIO_KILL;
9d609f85
MP		 1931 }
1932 }
1933 } else {
1934 if (unlikely(bip != NULL)) {
1935 DMERR("Unexpected integrity data when using internal hash");
846785e6 1936 return DM_MAPIO_KILL;
9d609f85
MP		 1937 }
1938 }
7eada909 1939
84597a44 1940 if (unlikely(ic->mode == 'R') && unlikely(dio->op != REQ_OP_READ))
846785e6 1941 return DM_MAPIO_KILL;
c2bcb2b7 1942
7eada909
MP		 1943 get_area_and_offset(ic, dio->range.logical_sector, &area, &offset);
1944 dio->metadata_block = get_metadata_sector_and_offset(ic, area, offset, &dio->metadata_offset);
1945 bio->bi_iter.bi_sector = get_data_sector(ic, area, offset);
1946
1947 dm_integrity_map_continue(dio, true);
1948 return DM_MAPIO_SUBMITTED;
1949}
1950
1951static bool __journal_read_write(struct dm_integrity_io *dio, struct bio *bio,
86a3238c 1952 unsigned int journal_section, unsigned int journal_entry)
7eada909
MP		 1953{
1954 struct dm_integrity_c *ic = dio->ic;
1955 sector_t logical_sector;
86a3238c 1956 unsigned int n_sectors;
7eada909
MP		 1957
1958 logical_sector = dio->range.logical_sector;
1959 n_sectors = dio->range.n_sectors;
1960 do {
1961 struct bio_vec bv = bio_iovec(bio);
1962 char *mem;
1963
1964 if (unlikely(bv.bv_len >> SECTOR_SHIFT > n_sectors))
1965 bv.bv_len = n_sectors << SECTOR_SHIFT;
1966 n_sectors -= bv.bv_len >> SECTOR_SHIFT;
1967 bio_advance_iter(bio, &bio->bi_iter, bv.bv_len);
1968retry_kmap:
1cef171a 1969 mem = kmap_local_page(bv.bv_page);
84597a44 1970 if (likely(dio->op == REQ_OP_WRITE))
7eada909
MP		 1971 flush_dcache_page(bv.bv_page);
1972
1973 do {
1974 struct journal_entry *je = access_journal_entry(ic, journal_section, journal_entry);
1975
84597a44 1976 if (unlikely(dio->op == REQ_OP_READ)) {
7eada909 1977 struct journal_sector *js;
9d609f85 1978 char *mem_ptr;
86a3238c 1979 unsigned int s;
7eada909
MP		 1980
1981 if (unlikely(journal_entry_is_inprogress(je))) {
1982 flush_dcache_page(bv.bv_page);
25058d1c 1983 kunmap_local(mem);
7eada909
MP		 1984
1985 __io_wait_event(ic->copy_to_journal_wait, !journal_entry_is_inprogress(je));
1986 goto retry_kmap;
1987 }
1988 smp_rmb();
1989 BUG_ON(journal_entry_get_sector(je) != logical_sector);
1990 js = access_journal_data(ic, journal_section, journal_entry);
9d609f85
MP		 1991 mem_ptr = mem + bv.bv_offset;
1992 s = 0;
1993 do {
1994 memcpy(mem_ptr, js, JOURNAL_SECTOR_DATA);
1995 *(commit_id_t *)(mem_ptr + JOURNAL_SECTOR_DATA) = je->last_bytes[s];
1996 js++;
1997 mem_ptr += 1 << SECTOR_SHIFT;
1998 } while (++s < ic->sectors_per_block);
7eada909
MP		 1999#ifdef INTERNAL_VERIFY
2000 if (ic->internal_hash) {
ba287d7c 2001 char checksums_onstack[max_t(size_t, HASH_MAX_DIGESTSIZE, MAX_TAG_SIZE)];
7eada909
MP		 2002
2003 integrity_sector_checksum(ic, logical_sector, mem + bv.bv_offset, checksums_onstack);
9d609f85 2004 if (unlikely(memcmp(checksums_onstack, journal_entry_tag(ic, je), ic->tag_size))) {
22555744 2005 DMERR_LIMIT("Checksum failed when reading from journal, at sector 0x%llx",
7649194a 2006 logical_sector);
82bb8599
MW		 2007 dm_audit_log_bio(DM_MSG_PREFIX, "journal-checksum",
2008 bio, logical_sector, 0);
7eada909
MP		 2009 }
2010 }
2011#endif
2012 }
2013
2014 if (!ic->internal_hash) {
2015 struct bio_integrity_payload *bip = bio_integrity(bio);
86a3238c 2016 unsigned int tag_todo = ic->tag_size;
9d609f85 2017 char *tag_ptr = journal_entry_tag(ic, je);
7eada909 2018
03b18887
HM		 2019 if (bip) {
2020 do {
2021 struct bio_vec biv = bvec_iter_bvec(bip->bip_vec, bip->bip_iter);
2022 unsigned int tag_now = min(biv.bv_len, tag_todo);
2023 char *tag_addr;
0ef0b471 2024
03b18887
HM		 2025 BUG_ON(PageHighMem(biv.bv_page));
2026 tag_addr = bvec_virt(&biv);
2027 if (likely(dio->op == REQ_OP_WRITE))
2028 memcpy(tag_ptr, tag_addr, tag_now);
2029 else
2030 memcpy(tag_addr, tag_ptr, tag_now);
2031 bvec_iter_advance(bip->bip_vec, &bip->bip_iter, tag_now);
2032 tag_ptr += tag_now;
2033 tag_todo -= tag_now;
2034 } while (unlikely(tag_todo));
2035 } else if (likely(dio->op == REQ_OP_WRITE))
2036 memset(tag_ptr, 0, tag_todo);
7eada909
MP		 2037 }
2038
84597a44 2039 if (likely(dio->op == REQ_OP_WRITE)) {
7eada909 2040 struct journal_sector *js;
86a3238c 2041 unsigned int s;
7eada909
MP		 2042
2043 js = access_journal_data(ic, journal_section, journal_entry);
9d609f85
MP		 2044 memcpy(js, mem + bv.bv_offset, ic->sectors_per_block << SECTOR_SHIFT);
2045
2046 s = 0;
2047 do {
2048 je->last_bytes[s] = js[s].commit_id;
2049 } while (++s < ic->sectors_per_block);
7eada909
MP		 2050
2051 if (ic->internal_hash) {
86a3238c 2052 unsigned int digest_size = crypto_shash_digestsize(ic->internal_hash);
0ef0b471 2053
7eada909 2054 if (unlikely(digest_size > ic->tag_size)) {
6d39a124 2055 char checksums_onstack[HASH_MAX_DIGESTSIZE];
0ef0b471 2056
7eada909 2057 integrity_sector_checksum(ic, logical_sector, (char *)js, checksums_onstack);
9d609f85 2058 memcpy(journal_entry_tag(ic, je), checksums_onstack, ic->tag_size);
7eada909 2059 } else
9d609f85 2060 integrity_sector_checksum(ic, logical_sector, (char *)js, journal_entry_tag(ic, je));
7eada909
MP		 2061 }
2062
2063 journal_entry_set_sector(je, logical_sector);
2064 }
9d609f85 2065 logical_sector += ic->sectors_per_block;
7eada909
MP		 2066
2067 journal_entry++;
2068 if (unlikely(journal_entry == ic->journal_section_entries)) {
2069 journal_entry = 0;
2070 journal_section++;
2071 wraparound_section(ic, &journal_section);
2072 }
2073
9d609f85
MP		 2074 bv.bv_offset += ic->sectors_per_block << SECTOR_SHIFT;
2075 } while (bv.bv_len -= ic->sectors_per_block << SECTOR_SHIFT);
7eada909 2076
84597a44 2077 if (unlikely(dio->op == REQ_OP_READ))
7eada909 2078 flush_dcache_page(bv.bv_page);
25058d1c 2079 kunmap_local(mem);
7eada909
MP		 2080 } while (n_sectors);
2081
84597a44 2082 if (likely(dio->op == REQ_OP_WRITE)) {
7eada909
MP		 2083 smp_mb();
2084 if (unlikely(waitqueue_active(&ic->copy_to_journal_wait)))
2085 wake_up(&ic->copy_to_journal_wait);
2d0f25cb 2086 if (READ_ONCE(ic->free_sectors) <= ic->free_sectors_threshold)
7eada909 2087 queue_work(ic->commit_wq, &ic->commit_work);
2d0f25cb 2088 else
7eada909 2089 schedule_autocommit(ic);
2d0f25cb 2090 } else
7eada909 2091 remove_range(ic, &dio->range);
7eada909
MP		 2092
2093 if (unlikely(bio->bi_iter.bi_size)) {
2094 sector_t area, offset;
2095
2096 dio->range.logical_sector = logical_sector;
2097 get_area_and_offset(ic, dio->range.logical_sector, &area, &offset);
2098 dio->metadata_block = get_metadata_sector_and_offset(ic, area, offset, &dio->metadata_offset);
2099 return true;
2100 }
2101
2102 return false;
2103}
2104
2105static void dm_integrity_map_continue(struct dm_integrity_io *dio, bool from_map)
2106{
2107 struct dm_integrity_c *ic = dio->ic;
2108 struct bio *bio = dm_bio_from_per_bio_data(dio, sizeof(struct dm_integrity_io));
86a3238c
HM		 2109 unsigned int journal_section, journal_entry;
2110 unsigned int journal_read_pos;
7eada909 2111 struct completion read_comp;
31843eda 2112 bool discard_retried = false;
84597a44 2113 bool need_sync_io = ic->internal_hash && dio->op == REQ_OP_READ;
0ef0b471 2114
84597a44
MP		 2115 if (unlikely(dio->op == REQ_OP_DISCARD) && ic->mode != 'D')
2116 need_sync_io = true;
7eada909
MP		 2117
2118 if (need_sync_io && from_map) {
2119 INIT_WORK(&dio->work, integrity_bio_wait);
53770f0e 2120 queue_work(ic->offload_wq, &dio->work);
7eada909
MP		 2121 return;
2122 }
2123
2124lock_retry:
2125 spin_lock_irq(&ic->endio_wait.lock);
2126retry:
2127 if (unlikely(dm_integrity_failed(ic))) {
2128 spin_unlock_irq(&ic->endio_wait.lock);
2129 do_endio(ic, bio);
2130 return;
2131 }
2132 dio->range.n_sectors = bio_sectors(bio);
2133 journal_read_pos = NOT_FOUND;
84597a44
MP		 2134 if (ic->mode == 'J' && likely(dio->op != REQ_OP_DISCARD)) {
2135 if (dio->op == REQ_OP_WRITE) {
86a3238c
HM		 2136 unsigned int next_entry, i, pos;
2137 unsigned int ws, we, range_sectors;
7eada909 2138
9dd59727 2139 dio->range.n_sectors = min(dio->range.n_sectors,
4f43446d 2140 (sector_t)ic->free_sectors << ic->sb->log2_sectors_per_block);
518748b1
MP		 2141 if (unlikely(!dio->range.n_sectors)) {
2142 if (from_map)
2143 goto offload_to_thread;
2144 sleep_on_endio_wait(ic);
2145 goto retry;
2146 }
9dd59727
MP		 2147 range_sectors = dio->range.n_sectors >> ic->sb->log2_sectors_per_block;
2148 ic->free_sectors -= range_sectors;
7eada909
MP		 2149 journal_section = ic->free_section;
2150 journal_entry = ic->free_section_entry;
2151
9dd59727 2152 next_entry = ic->free_section_entry + range_sectors;
7eada909
MP		 2153 ic->free_section_entry = next_entry % ic->journal_section_entries;
2154 ic->free_section += next_entry / ic->journal_section_entries;
2155 ic->n_uncommitted_sections += next_entry / ic->journal_section_entries;
2156 wraparound_section(ic, &ic->free_section);
2157
2158 pos = journal_section * ic->journal_section_entries + journal_entry;
2159 ws = journal_section;
2160 we = journal_entry;
9d609f85
MP		 2161 i = 0;
2162 do {
7eada909
MP		 2163 struct journal_entry *je;
2164
2165 add_journal_node(ic, &ic->journal_tree[pos], dio->range.logical_sector + i);
2166 pos++;
2167 if (unlikely(pos >= ic->journal_entries))
2168 pos = 0;
2169
2170 je = access_journal_entry(ic, ws, we);
2171 BUG_ON(!journal_entry_is_unused(je));
2172 journal_entry_set_inprogress(je);
2173 we++;
2174 if (unlikely(we == ic->journal_section_entries)) {
2175 we = 0;
2176 ws++;
2177 wraparound_section(ic, &ws);
2178 }
9d609f85 2179 } while ((i += ic->sectors_per_block) < dio->range.n_sectors);
7eada909
MP		 2180
2181 spin_unlock_irq(&ic->endio_wait.lock);
2182 goto journal_read_write;
2183 } else {
2184 sector_t next_sector;
0ef0b471 2185
7eada909
MP		 2186 journal_read_pos = find_journal_node(ic, dio->range.logical_sector, &next_sector);
2187 if (likely(journal_read_pos == NOT_FOUND)) {
2188 if (unlikely(dio->range.n_sectors > next_sector - dio->range.logical_sector))
2189 dio->range.n_sectors = next_sector - dio->range.logical_sector;
2190 } else {
86a3238c
HM		 2191 unsigned int i;
2192 unsigned int jp = journal_read_pos + 1;
0ef0b471 2193
9d609f85
MP		 2194 for (i = ic->sectors_per_block; i < dio->range.n_sectors; i += ic->sectors_per_block, jp++) {
2195 if (!test_journal_node(ic, jp, dio->range.logical_sector + i))
7eada909
MP		 2196 break;
2197 }
2198 dio->range.n_sectors = i;
2199 }
2200 }
2201 }
724376a0 2202 if (unlikely(!add_new_range(ic, &dio->range, true))) {
7eada909
MP		 2203 /*
2204 * We must not sleep in the request routine because it could
2205 * stall bios on current->bio_list.
2206 * So, we offload the bio to a workqueue if we have to sleep.
2207 */
7eada909 2208 if (from_map) {
518748b1 2209offload_to_thread:
7eada909
MP		 2210 spin_unlock_irq(&ic->endio_wait.lock);
2211 INIT_WORK(&dio->work, integrity_bio_wait);
2212 queue_work(ic->wait_wq, &dio->work);
2213 return;
7eada909 2214 }
5729b6e5
MP		 2215 if (journal_read_pos != NOT_FOUND)
2216 dio->range.n_sectors = ic->sectors_per_block;
724376a0 2217 wait_and_add_new_range(ic, &dio->range);
5729b6e5
MP		 2218 /*
2219 * wait_and_add_new_range drops the spinlock, so the journal
2220 * may have been changed arbitrarily. We need to recheck.
2221 * To simplify the code, we restrict I/O size to just one block.
2222 */
2223 if (journal_read_pos != NOT_FOUND) {
2224 sector_t next_sector;
0ef0b471
HM		 2225 unsigned int new_pos;
2226
2227 new_pos = find_journal_node(ic, dio->range.logical_sector, &next_sector);
5729b6e5
MP		 2228 if (unlikely(new_pos != journal_read_pos)) {
2229 remove_range_unlocked(ic, &dio->range);
2230 goto retry;
2231 }
2232 }
7eada909 2233 }
31843eda
MP		 2234 if (ic->mode == 'J' && likely(dio->op == REQ_OP_DISCARD) && !discard_retried) {
2235 sector_t next_sector;
0ef0b471
HM		 2236 unsigned int new_pos;
2237
2238 new_pos = find_journal_node(ic, dio->range.logical_sector, &next_sector);
31843eda
MP		 2239 if (unlikely(new_pos != NOT_FOUND) ||
2240 unlikely(next_sector < dio->range.logical_sector - dio->range.n_sectors)) {
2241 remove_range_unlocked(ic, &dio->range);
2242 spin_unlock_irq(&ic->endio_wait.lock);
2243 queue_work(ic->commit_wq, &ic->commit_work);
2244 flush_workqueue(ic->commit_wq);
2245 queue_work(ic->writer_wq, &ic->writer_work);
2246 flush_workqueue(ic->writer_wq);
2247 discard_retried = true;
2248 goto lock_retry;
2249 }
2250 }
7eada909
MP		 2251 spin_unlock_irq(&ic->endio_wait.lock);
2252
2253 if (unlikely(journal_read_pos != NOT_FOUND)) {
2254 journal_section = journal_read_pos / ic->journal_section_entries;
2255 journal_entry = journal_read_pos % ic->journal_section_entries;
2256 goto journal_read_write;
2257 }
2258
84597a44 2259 if (ic->mode == 'B' && (dio->op == REQ_OP_WRITE || unlikely(dio->op == REQ_OP_DISCARD))) {
05d6909e
MS		 2260 if (!block_bitmap_op(ic, ic->may_write_bitmap, dio->range.logical_sector,
2261 dio->range.n_sectors, BITMAP_OP_TEST_ALL_SET)) {
2262 struct bitmap_block_status *bbs;
468dfca3 2263
05d6909e 2264 bbs = sector_to_bitmap_block(ic, dio->range.logical_sector);
468dfca3
MP		 2265 spin_lock(&bbs->bio_queue_lock);
2266 bio_list_add(&bbs->bio_queue, bio);
2267 spin_unlock(&bbs->bio_queue_lock);
468dfca3 2268 queue_work(ic->writer_wq, &bbs->work);
468dfca3
MP		 2269 return;
2270 }
2271 }
2272
7eada909
MP		 2273 dio->in_flight = (atomic_t)ATOMIC_INIT(2);
2274
2275 if (need_sync_io) {
b5e8ad92 2276 init_completion(&read_comp);
7eada909
MP		 2277 dio->completion = &read_comp;
2278 } else
2279 dio->completion = NULL;
2280
248aa264 2281 dm_bio_record(&dio->bio_details, bio);
74d46992 2282 bio_set_dev(bio, ic->dev->bdev);
7eada909
MP		 2283 bio->bi_integrity = NULL;
2284 bio->bi_opf &= ~REQ_INTEGRITY;
7eada909 2285 bio->bi_end_io = integrity_end_io;
7eada909 2286 bio->bi_iter.bi_size = dio->range.n_sectors << SECTOR_SHIFT;
248aa264 2287
84597a44
MP		 2288 if (unlikely(dio->op == REQ_OP_DISCARD) && likely(ic->mode != 'D')) {
2289 integrity_metadata(&dio->work);
9b594826 2290 dm_integrity_flush_buffers(ic, false);
84597a44
MP		 2291
2292 dio->in_flight = (atomic_t)ATOMIC_INIT(1);
2293 dio->completion = NULL;
2294
ed00aabd 2295 submit_bio_noacct(bio);
84597a44
MP		 2296
2297 return;
2298 }
2299
ed00aabd 2300 submit_bio_noacct(bio);
7eada909
MP		 2301
2302 if (need_sync_io) {
2303 wait_for_completion_io(&read_comp);
468dfca3 2304 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING) &&
a3fcf725
MP		 2305 dio->range.logical_sector + dio->range.n_sectors > le64_to_cpu(ic->sb->recalc_sector))
2306 goto skip_check;
468dfca3 2307 if (ic->mode == 'B') {
05d6909e
MS		 2308 if (!block_bitmap_op(ic, ic->recalc_bitmap, dio->range.logical_sector,
2309 dio->range.n_sectors, BITMAP_OP_TEST_ALL_CLEAR))
468dfca3
MP		 2310 goto skip_check;
2311 }
2312
b7e326f7
HL		 2313 if (likely(!bio->bi_status))
2314 integrity_metadata(&dio->work);
2315 else
a3fcf725 2316skip_check:
b7e326f7 2317 dec_in_flight(dio);
7eada909
MP		 2318 } else {
2319 INIT_WORK(&dio->work, integrity_metadata);
2320 queue_work(ic->metadata_wq, &dio->work);
2321 }
2322
2323 return;
2324
2325journal_read_write:
2326 if (unlikely(__journal_read_write(dio, bio, journal_section, journal_entry)))
2327 goto lock_retry;
2328
2329 do_endio_flush(ic, dio);
2330}
2331
2332
2333static void integrity_bio_wait(struct work_struct *w)
2334{
2335 struct dm_integrity_io *dio = container_of(w, struct dm_integrity_io, work);
2336
2337 dm_integrity_map_continue(dio, false);
2338}
2339
2340static void pad_uncommitted(struct dm_integrity_c *ic)
2341{
2342 if (ic->free_section_entry) {
2343 ic->free_sectors -= ic->journal_section_entries - ic->free_section_entry;
2344 ic->free_section_entry = 0;
2345 ic->free_section++;
2346 wraparound_section(ic, &ic->free_section);
2347 ic->n_uncommitted_sections++;
2348 }
468dfca3 2349 if (WARN_ON(ic->journal_sections * ic->journal_section_entries !=
05d6909e
MS		 2350 (ic->n_uncommitted_sections + ic->n_committed_sections) *
2351 ic->journal_section_entries + ic->free_sectors)) {
2352 DMCRIT("journal_sections %u, journal_section_entries %u, "
2353 "n_uncommitted_sections %u, n_committed_sections %u, "
2354 "journal_section_entries %u, free_sectors %u",
2355 ic->journal_sections, ic->journal_section_entries,
2356 ic->n_uncommitted_sections, ic->n_committed_sections,
2357 ic->journal_section_entries, ic->free_sectors);
468dfca3 2358 }
7eada909
MP		 2359}
2360
2361static void integrity_commit(struct work_struct *w)
2362{
2363 struct dm_integrity_c *ic = container_of(w, struct dm_integrity_c, commit_work);
86a3238c
HM		 2364 unsigned int commit_start, commit_sections;
2365 unsigned int i, j, n;
7eada909
MP		 2366 struct bio *flushes;
2367
2368 del_timer(&ic->autocommit_timer);
2369
2370 spin_lock_irq(&ic->endio_wait.lock);
2371 flushes = bio_list_get(&ic->flush_bio_list);
2372 if (unlikely(ic->mode != 'J')) {
2373 spin_unlock_irq(&ic->endio_wait.lock);
9b594826 2374 dm_integrity_flush_buffers(ic, true);
7eada909
MP		 2375 goto release_flush_bios;
2376 }
2377
2378 pad_uncommitted(ic);
2379 commit_start = ic->uncommitted_section;
2380 commit_sections = ic->n_uncommitted_sections;
2381 spin_unlock_irq(&ic->endio_wait.lock);
2382
2383 if (!commit_sections)
2384 goto release_flush_bios;
2385
984bf2cc
MP		 2386 ic->wrote_to_journal = true;
2387
7eada909
MP		 2388 i = commit_start;
2389 for (n = 0; n < commit_sections; n++) {
2390 for (j = 0; j < ic->journal_section_entries; j++) {
2391 struct journal_entry *je;
0ef0b471 2392
7eada909
MP		 2393 je = access_journal_entry(ic, i, j);
2394 io_wait_event(ic->copy_to_journal_wait, !journal_entry_is_inprogress(je));
2395 }
2396 for (j = 0; j < ic->journal_section_sectors; j++) {
2397 struct journal_sector *js;
0ef0b471 2398
7eada909
MP		 2399 js = access_journal(ic, i, j);
2400 js->commit_id = dm_integrity_commit_id(ic, i, j, ic->commit_seq);
2401 }
2402 i++;
2403 if (unlikely(i >= ic->journal_sections))
2404 ic->commit_seq = next_commit_seq(ic->commit_seq);
2405 wraparound_section(ic, &i);
2406 }
2407 smp_rmb();
2408
2409 write_journal(ic, commit_start, commit_sections);
2410
2411 spin_lock_irq(&ic->endio_wait.lock);
2412 ic->uncommitted_section += commit_sections;
2413 wraparound_section(ic, &ic->uncommitted_section);
2414 ic->n_uncommitted_sections -= commit_sections;
2415 ic->n_committed_sections += commit_sections;
2416 spin_unlock_irq(&ic->endio_wait.lock);
2417
d3e632f0 2418 if (READ_ONCE(ic->free_sectors) <= ic->free_sectors_threshold)
7eada909
MP		 2419 queue_work(ic->writer_wq, &ic->writer_work);
2420
2421release_flush_bios:
2422 while (flushes) {
2423 struct bio *next = flushes->bi_next;
0ef0b471 2424
7eada909
MP		 2425 flushes->bi_next = NULL;
2426 do_endio(ic, flushes);
2427 flushes = next;
2428 }
2429}
2430
2431static void complete_copy_from_journal(unsigned long error, void *context)
2432{
2433 struct journal_io *io = context;
2434 struct journal_completion *comp = io->comp;
2435 struct dm_integrity_c *ic = comp->ic;
0ef0b471 2436
7eada909 2437 remove_range(ic, &io->range);
6f1c819c 2438 mempool_free(io, &ic->journal_io_mempool);
7eada909
MP		 2439 if (unlikely(error != 0))
2440 dm_integrity_io_error(ic, "copying from journal", -EIO);
2441 complete_journal_op(comp);
2442}
2443
9d609f85
MP		 2444static void restore_last_bytes(struct dm_integrity_c *ic, struct journal_sector *js,
2445 struct journal_entry *je)
2446{
86a3238c 2447 unsigned int s = 0;
0ef0b471 2448
9d609f85
MP		 2449 do {
2450 js->commit_id = je->last_bytes[s];
2451 js++;
2452 } while (++s < ic->sectors_per_block);
2453}
2454
86a3238c
HM		 2455static void do_journal_write(struct dm_integrity_c *ic, unsigned int write_start,
2456 unsigned int write_sections, bool from_replay)
7eada909 2457{
86a3238c 2458 unsigned int i, j, n;
7eada909 2459 struct journal_completion comp;
a7c3e62b
MP		 2460 struct blk_plug plug;
2461
2462 blk_start_plug(&plug);
7eada909
MP		 2463
2464 comp.ic = ic;
2465 comp.in_flight = (atomic_t)ATOMIC_INIT(1);
b5e8ad92 2466 init_completion(&comp.comp);
7eada909
MP		 2467
2468 i = write_start;
2469 for (n = 0; n < write_sections; n++, i++, wraparound_section(ic, &i)) {
2470#ifndef INTERNAL_VERIFY
2471 if (unlikely(from_replay))
2472#endif
2473 rw_section_mac(ic, i, false);
2474 for (j = 0; j < ic->journal_section_entries; j++) {
2475 struct journal_entry *je = access_journal_entry(ic, i, j);
2476 sector_t sec, area, offset;
86a3238c 2477 unsigned int k, l, next_loop;
7eada909 2478 sector_t metadata_block;
86a3238c 2479 unsigned int metadata_offset;
7eada909
MP		 2480 struct journal_io *io;
2481
2482 if (journal_entry_is_unused(je))
2483 continue;
2484 BUG_ON(unlikely(journal_entry_is_inprogress(je)) && !from_replay);
2485 sec = journal_entry_get_sector(je);
9d609f85 2486 if (unlikely(from_replay)) {
86a3238c 2487 if (unlikely(sec & (unsigned int)(ic->sectors_per_block - 1))) {
9d609f85
MP		 2488 dm_integrity_io_error(ic, "invalid sector in journal", -EIO);
2489 sec &= ~(sector_t)(ic->sectors_per_block - 1);
2490 }
cc09e8a9
MP		 2491 if (unlikely(sec >= ic->provided_data_sectors)) {
2492 journal_entry_set_unused(je);
2493 continue;
2494 }
9d609f85 2495 }
7eada909 2496 get_area_and_offset(ic, sec, &area, &offset);
9d609f85 2497 restore_last_bytes(ic, access_journal_data(ic, i, j), je);
7eada909
MP		 2498 for (k = j + 1; k < ic->journal_section_entries; k++) {
2499 struct journal_entry *je2 = access_journal_entry(ic, i, k);
2500 sector_t sec2, area2, offset2;
0ef0b471 2501
7eada909
MP		 2502 if (journal_entry_is_unused(je2))
2503 break;
2504 BUG_ON(unlikely(journal_entry_is_inprogress(je2)) && !from_replay);
2505 sec2 = journal_entry_get_sector(je2);
f6f72f32
MP		 2506 if (unlikely(sec2 >= ic->provided_data_sectors))
2507 break;
7eada909 2508 get_area_and_offset(ic, sec2, &area2, &offset2);
9d609f85 2509 if (area2 != area || offset2 != offset + ((k - j) << ic->sb->log2_sectors_per_block))
7eada909 2510 break;
9d609f85 2511 restore_last_bytes(ic, access_journal_data(ic, i, k), je2);
7eada909
MP		 2512 }
2513 next_loop = k - 1;
2514
6f1c819c 2515 io = mempool_alloc(&ic->journal_io_mempool, GFP_NOIO);
7eada909
MP		 2516 io->comp = &comp;
2517 io->range.logical_sector = sec;
9d609f85 2518 io->range.n_sectors = (k - j) << ic->sb->log2_sectors_per_block;
7eada909
MP		 2519
2520 spin_lock_irq(&ic->endio_wait.lock);
8b3bbd49 2521 add_new_range_and_wait(ic, &io->range);
7eada909
MP		 2522
2523 if (likely(!from_replay)) {
2524 struct journal_node *section_node = &ic->journal_tree[i * ic->journal_section_entries];
2525
2526 /* don't write if there is newer committed sector */
2527 while (j < k && find_newer_committed_node(ic, &section_node[j])) {
2528 struct journal_entry *je2 = access_journal_entry(ic, i, j);
2529
2530 journal_entry_set_unused(je2);
2531 remove_journal_node(ic, &section_node[j]);
2532 j++;
9d609f85
MP		 2533 sec += ic->sectors_per_block;
2534 offset += ic->sectors_per_block;
7eada909
MP		 2535 }
2536 while (j < k && find_newer_committed_node(ic, &section_node[k - 1])) {
2537 struct journal_entry *je2 = access_journal_entry(ic, i, k - 1);
2538
2539 journal_entry_set_unused(je2);
2540 remove_journal_node(ic, &section_node[k - 1]);
2541 k--;
2542 }
2543 if (j == k) {
2544 remove_range_unlocked(ic, &io->range);
2545 spin_unlock_irq(&ic->endio_wait.lock);
6f1c819c 2546 mempool_free(io, &ic->journal_io_mempool);
7eada909
MP		 2547 goto skip_io;
2548 }
2d0f25cb 2549 for (l = j; l < k; l++)
7eada909 2550 remove_journal_node(ic, &section_node[l]);
7eada909
MP		 2551 }
2552 spin_unlock_irq(&ic->endio_wait.lock);
2553
2554 metadata_block = get_metadata_sector_and_offset(ic, area, offset, &metadata_offset);
2555 for (l = j; l < k; l++) {
2556 int r;
2557 struct journal_entry *je2 = access_journal_entry(ic, i, l);
2558
2559 if (
2560#ifndef INTERNAL_VERIFY
2561 unlikely(from_replay) &&
2562#endif
2563 ic->internal_hash) {
6d39a124 2564 char test_tag[max_t(size_t, HASH_MAX_DIGESTSIZE, MAX_TAG_SIZE)];
7eada909 2565
9d609f85 2566 integrity_sector_checksum(ic, sec + ((l - j) << ic->sb->log2_sectors_per_block),
7eada909 2567 (char *)access_journal_data(ic, i, l), test_tag);
82bb8599 2568 if (unlikely(memcmp(test_tag, journal_entry_tag(ic, je2), ic->tag_size))) {
7eada909 2569 dm_integrity_io_error(ic, "tag mismatch when replaying journal", -EILSEQ);
82bb8599
MW		 2570 dm_audit_log_target(DM_MSG_PREFIX, "integrity-replay-journal", ic->ti, 0);
2571 }
7eada909
MP		 2572 }
2573
2574 journal_entry_set_unused(je2);
9d609f85 2575 r = dm_integrity_rw_tag(ic, journal_entry_tag(ic, je2), &metadata_block, &metadata_offset,
7eada909 2576 ic->tag_size, TAG_WRITE);
2d0f25cb 2577 if (unlikely(r))
7eada909 2578 dm_integrity_io_error(ic, "reading tags", r);
7eada909
MP		 2579 }
2580
2581 atomic_inc(&comp.in_flight);
9d609f85
MP		 2582 copy_from_journal(ic, i, j << ic->sb->log2_sectors_per_block,
2583 (k - j) << ic->sb->log2_sectors_per_block,
2584 get_data_sector(ic, area, offset),
7eada909
MP		 2585 complete_copy_from_journal, io);
2586skip_io:
2587 j = next_loop;
2588 }
2589 }
2590
2591 dm_bufio_write_dirty_buffers_async(ic->bufio);
2592
a7c3e62b
MP		 2593 blk_finish_plug(&plug);
2594
7eada909
MP		 2595 complete_journal_op(&comp);
2596 wait_for_completion_io(&comp.comp);
2597
9b594826 2598 dm_integrity_flush_buffers(ic, true);
7eada909
MP		 2599}
2600
2601static void integrity_writer(struct work_struct *w)
2602{
2603 struct dm_integrity_c *ic = container_of(w, struct dm_integrity_c, writer_work);
86a3238c 2604 unsigned int write_start, write_sections;
86a3238c 2605 unsigned int prev_free_sectors;
7eada909 2606
7eada909
MP		 2607 spin_lock_irq(&ic->endio_wait.lock);
2608 write_start = ic->committed_section;
2609 write_sections = ic->n_committed_sections;
2610 spin_unlock_irq(&ic->endio_wait.lock);
2611
2612 if (!write_sections)
2613 return;
2614
2615 do_journal_write(ic, write_start, write_sections, false);
2616
2617 spin_lock_irq(&ic->endio_wait.lock);
2618
2619 ic->committed_section += write_sections;
2620 wraparound_section(ic, &ic->committed_section);
2621 ic->n_committed_sections -= write_sections;
2622
2623 prev_free_sectors = ic->free_sectors;
2624 ic->free_sectors += write_sections * ic->journal_section_entries;
2625 if (unlikely(!prev_free_sectors))
2626 wake_up_locked(&ic->endio_wait);
2627
2628 spin_unlock_irq(&ic->endio_wait.lock);
2629}
2630
a3fcf725
MP		 2631static void recalc_write_super(struct dm_integrity_c *ic)
2632{
2633 int r;
2634
9b594826 2635 dm_integrity_flush_buffers(ic, false);
a3fcf725
MP		 2636 if (dm_integrity_failed(ic))
2637 return;
2638
c9154a4c 2639 r = sync_rw_sb(ic, REQ_OP_WRITE);
a3fcf725
MP		 2640 if (unlikely(r))
2641 dm_integrity_io_error(ic, "writing superblock", r);
2642}
2643
2644static void integrity_recalc(struct work_struct *w)
2645{
2646 struct dm_integrity_c *ic = container_of(w, struct dm_integrity_c, recalc_work);
da8b4fc1
MP		 2647 size_t recalc_tags_size;
2648 u8 *recalc_buffer = NULL;
2649 u8 *recalc_tags = NULL;
a3fcf725
MP		 2650 struct dm_integrity_range range;
2651 struct dm_io_request io_req;
2652 struct dm_io_region io_loc;
2653 sector_t area, offset;
2654 sector_t metadata_block;
86a3238c 2655 unsigned int metadata_offset;
468dfca3 2656 sector_t logical_sector, n_sectors;
a3fcf725 2657 __u8 *t;
86a3238c 2658 unsigned int i;
a3fcf725 2659 int r;
86a3238c 2660 unsigned int super_counter = 0;
3be16228 2661 unsigned recalc_sectors = RECALC_SECTORS;
a3fcf725 2662
3be16228
MP		 2663retry:
2664 recalc_buffer = __vmalloc(recalc_sectors << SECTOR_SHIFT, GFP_NOIO);
da8b4fc1 2665 if (!recalc_buffer) {
3be16228
MP		 2666oom:
2667 recalc_sectors >>= 1;
2668 if (recalc_sectors >= 1U << ic->sb->log2_sectors_per_block)
2669 goto retry;
da8b4fc1
MP		 2670 DMCRIT("out of memory for recalculate buffer - recalculation disabled");
2671 goto free_ret;
2672 }
3be16228 2673 recalc_tags_size = (recalc_sectors >> ic->sb->log2_sectors_per_block) * ic->tag_size;
da8b4fc1
MP		 2674 if (crypto_shash_digestsize(ic->internal_hash) > ic->tag_size)
2675 recalc_tags_size += crypto_shash_digestsize(ic->internal_hash) - ic->tag_size;
2676 recalc_tags = kvmalloc(recalc_tags_size, GFP_NOIO);
2677 if (!recalc_tags) {
3be16228 2678 vfree(recalc_buffer);
d4a3806b 2679 recalc_buffer = NULL;
3be16228 2680 goto oom;
da8b4fc1 2681 }
a3fcf725 2682
468dfca3
MP		 2683 DEBUG_print("start recalculation... (position %llx)\n", le64_to_cpu(ic->sb->recalc_sector));
2684
a3fcf725
MP		 2685 spin_lock_irq(&ic->endio_wait.lock);
2686
2687next_chunk:
2688
5df96f2b 2689 if (unlikely(dm_post_suspending(ic->ti)))
a3fcf725
MP		 2690 goto unlock_ret;
2691
2692 range.logical_sector = le64_to_cpu(ic->sb->recalc_sector);
468dfca3
MP		 2693 if (unlikely(range.logical_sector >= ic->provided_data_sectors)) {
2694 if (ic->mode == 'B') {
e27fec66 2695 block_bitmap_op(ic, ic->recalc_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
468dfca3
MP		 2696 DEBUG_print("queue_delayed_work: bitmap_flush_work\n");
2697 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, 0);
2698 }
a3fcf725 2699 goto unlock_ret;
468dfca3 2700 }
a3fcf725
MP		 2701
2702 get_area_and_offset(ic, range.logical_sector, &area, &offset);
3be16228 2703 range.n_sectors = min((sector_t)recalc_sectors, ic->provided_data_sectors - range.logical_sector);
a3fcf725 2704 if (!ic->meta_dev)
86a3238c 2705 range.n_sectors = min(range.n_sectors, ((sector_t)1U << ic->sb->log2_interleave_sectors) - (unsigned int)offset);
a3fcf725 2706
8b3bbd49 2707 add_new_range_and_wait(ic, &range);
a3fcf725 2708 spin_unlock_irq(&ic->endio_wait.lock);
468dfca3
MP		 2709 logical_sector = range.logical_sector;
2710 n_sectors = range.n_sectors;
2711
2712 if (ic->mode == 'B') {
2d0f25cb 2713 if (block_bitmap_op(ic, ic->recalc_bitmap, logical_sector, n_sectors, BITMAP_OP_TEST_ALL_CLEAR))
468dfca3 2714 goto advance_and_next;
2d0f25cb 2715
05d6909e
MS		 2716 while (block_bitmap_op(ic, ic->recalc_bitmap, logical_sector,
2717 ic->sectors_per_block, BITMAP_OP_TEST_ALL_CLEAR)) {
468dfca3
MP		 2718 logical_sector += ic->sectors_per_block;
2719 n_sectors -= ic->sectors_per_block;
2720 cond_resched();
2721 }
05d6909e
MS		 2722 while (block_bitmap_op(ic, ic->recalc_bitmap, logical_sector + n_sectors - ic->sectors_per_block,
2723 ic->sectors_per_block, BITMAP_OP_TEST_ALL_CLEAR)) {
468dfca3
MP		 2724 n_sectors -= ic->sectors_per_block;
2725 cond_resched();
2726 }
2727 get_area_and_offset(ic, logical_sector, &area, &offset);
2728 }
2729
7649194a 2730 DEBUG_print("recalculating: %llx, %llx\n", logical_sector, n_sectors);
a3fcf725
MP		 2731
2732 if (unlikely(++super_counter == RECALC_WRITE_SUPER)) {
2733 recalc_write_super(ic);
2d0f25cb 2734 if (ic->mode == 'B')
468dfca3 2735 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, ic->bitmap_flush_interval);
2d0f25cb 2736
a3fcf725
MP		 2737 super_counter = 0;
2738 }
2739
2740 if (unlikely(dm_integrity_failed(ic)))
2741 goto err;
2742
581075e4 2743 io_req.bi_opf = REQ_OP_READ;
dbae70d4 2744 io_req.mem.type = DM_IO_VMA;
da8b4fc1 2745 io_req.mem.ptr.addr = recalc_buffer;
dbae70d4
MP		 2746 io_req.notify.fn = NULL;
2747 io_req.client = ic->io;
2748 io_loc.bdev = ic->dev->bdev;
2749 io_loc.sector = get_data_sector(ic, area, offset);
2750 io_loc.count = n_sectors;
a3fcf725 2751
dbae70d4
MP		 2752 r = dm_io(&io_req, 1, &io_loc, NULL);
2753 if (unlikely(r)) {
2754 dm_integrity_io_error(ic, "reading data", r);
2755 goto err;
2756 }
a3fcf725 2757
da8b4fc1 2758 t = recalc_tags;
dbae70d4 2759 for (i = 0; i < n_sectors; i += ic->sectors_per_block) {
da8b4fc1 2760 integrity_sector_checksum(ic, logical_sector + i, recalc_buffer + (i << SECTOR_SHIFT), t);
dbae70d4 2761 t += ic->tag_size;
a3fcf725
MP		 2762 }
2763
2764 metadata_block = get_metadata_sector_and_offset(ic, area, offset, &metadata_offset);
2765
da8b4fc1 2766 r = dm_integrity_rw_tag(ic, recalc_tags, &metadata_block, &metadata_offset, t - recalc_tags, TAG_WRITE);
a3fcf725
MP		 2767 if (unlikely(r)) {
2768 dm_integrity_io_error(ic, "writing tags", r);
2769 goto err;
2770 }
2771
e27fec66
MP		 2772 if (ic->mode == 'B') {
2773 sector_t start, end;
0ef0b471 2774
e27fec66
MP		 2775 start = (range.logical_sector >>
2776 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit)) <<
2777 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
2778 end = ((range.logical_sector + range.n_sectors) >>
2779 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit)) <<
2780 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
2781 block_bitmap_op(ic, ic->recalc_bitmap, start, end - start, BITMAP_OP_CLEAR);
2782 }
2783
468dfca3
MP		 2784advance_and_next:
2785 cond_resched();
2786
a3fcf725
MP		 2787 spin_lock_irq(&ic->endio_wait.lock);
2788 remove_range_unlocked(ic, &range);
2789 ic->sb->recalc_sector = cpu_to_le64(range.logical_sector + range.n_sectors);
2790 goto next_chunk;
2791
2792err:
2793 remove_range(ic, &range);
da8b4fc1 2794 goto free_ret;
a3fcf725
MP		 2795
2796unlock_ret:
2797 spin_unlock_irq(&ic->endio_wait.lock);
2798
2799 recalc_write_super(ic);
da8b4fc1
MP		 2800
2801free_ret:
2802 vfree(recalc_buffer);
2803 kvfree(recalc_tags);
a3fcf725
MP		 2804}
2805
468dfca3
MP		 2806static void bitmap_block_work(struct work_struct *w)
2807{
2808 struct bitmap_block_status *bbs = container_of(w, struct bitmap_block_status, work);
2809 struct dm_integrity_c *ic = bbs->ic;
2810 struct bio *bio;
2811 struct bio_list bio_queue;
2812 struct bio_list waiting;
2813
2814 bio_list_init(&waiting);
2815
2816 spin_lock(&bbs->bio_queue_lock);
2817 bio_queue = bbs->bio_queue;
2818 bio_list_init(&bbs->bio_queue);
2819 spin_unlock(&bbs->bio_queue_lock);
2820
2821 while ((bio = bio_list_pop(&bio_queue))) {
2822 struct dm_integrity_io *dio;
2823
2824 dio = dm_per_bio_data(bio, sizeof(struct dm_integrity_io));
2825
05d6909e
MS		 2826 if (block_bitmap_op(ic, ic->may_write_bitmap, dio->range.logical_sector,
2827 dio->range.n_sectors, BITMAP_OP_TEST_ALL_SET)) {
468dfca3
MP		 2828 remove_range(ic, &dio->range);
2829 INIT_WORK(&dio->work, integrity_bio_wait);
53770f0e 2830 queue_work(ic->offload_wq, &dio->work);
468dfca3 2831 } else {
05d6909e
MS		 2832 block_bitmap_op(ic, ic->journal, dio->range.logical_sector,
2833 dio->range.n_sectors, BITMAP_OP_SET);
468dfca3
MP		 2834 bio_list_add(&waiting, bio);
2835 }
2836 }
2837
2838 if (bio_list_empty(&waiting))
2839 return;
2840
c9154a4c 2841 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC,
05d6909e
MS		 2842 bbs->idx * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT),
2843 BITMAP_BLOCK_SIZE >> SECTOR_SHIFT, NULL);
468dfca3
MP		 2844
2845 while ((bio = bio_list_pop(&waiting))) {
2846 struct dm_integrity_io *dio = dm_per_bio_data(bio, sizeof(struct dm_integrity_io));
2847
05d6909e
MS		 2848 block_bitmap_op(ic, ic->may_write_bitmap, dio->range.logical_sector,
2849 dio->range.n_sectors, BITMAP_OP_SET);
468dfca3
MP		 2850
2851 remove_range(ic, &dio->range);
2852 INIT_WORK(&dio->work, integrity_bio_wait);
53770f0e 2853 queue_work(ic->offload_wq, &dio->work);
468dfca3
MP		 2854 }
2855
2856 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, ic->bitmap_flush_interval);
2857}
2858
2859static void bitmap_flush_work(struct work_struct *work)
2860{
2861 struct dm_integrity_c *ic = container_of(work, struct dm_integrity_c, bitmap_flush_work.work);
2862 struct dm_integrity_range range;
2863 unsigned long limit;
48271493 2864 struct bio *bio;
468dfca3 2865
9b594826 2866 dm_integrity_flush_buffers(ic, false);
468dfca3
MP		 2867
2868 range.logical_sector = 0;
2869 range.n_sectors = ic->provided_data_sectors;
2870
2871 spin_lock_irq(&ic->endio_wait.lock);
2872 add_new_range_and_wait(ic, &range);
2873 spin_unlock_irq(&ic->endio_wait.lock);
2874
9b594826 2875 dm_integrity_flush_buffers(ic, true);
468dfca3
MP		 2876
2877 limit = ic->provided_data_sectors;
2878 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) {
2879 limit = le64_to_cpu(ic->sb->recalc_sector)
2880 >> (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit)
2881 << (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
2882 }
48271493 2883 /*DEBUG_print("zeroing journal\n");*/
468dfca3
MP		 2884 block_bitmap_op(ic, ic->journal, 0, limit, BITMAP_OP_CLEAR);
2885 block_bitmap_op(ic, ic->may_write_bitmap, 0, limit, BITMAP_OP_CLEAR);
2886
c9154a4c 2887 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
05d6909e 2888 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
468dfca3 2889
48271493
MP		 2890 spin_lock_irq(&ic->endio_wait.lock);
2891 remove_range_unlocked(ic, &range);
2892 while (unlikely((bio = bio_list_pop(&ic->synchronous_bios)) != NULL)) {
2893 bio_endio(bio);
2894 spin_unlock_irq(&ic->endio_wait.lock);
2895 spin_lock_irq(&ic->endio_wait.lock);
2896 }
2897 spin_unlock_irq(&ic->endio_wait.lock);
468dfca3
MP		 2898}
2899
2900
86a3238c
HM		 2901static void init_journal(struct dm_integrity_c *ic, unsigned int start_section,
2902 unsigned int n_sections, unsigned char commit_seq)
7eada909 2903{
86a3238c 2904 unsigned int i, j, n;
7eada909
MP		 2905
2906 if (!n_sections)
2907 return;
2908
2909 for (n = 0; n < n_sections; n++) {
2910 i = start_section + n;
2911 wraparound_section(ic, &i);
2912 for (j = 0; j < ic->journal_section_sectors; j++) {
2913 struct journal_sector *js = access_journal(ic, i, j);
0ef0b471 2914
f069c7ab
KC		 2915 BUILD_BUG_ON(sizeof(js->sectors) != JOURNAL_SECTOR_DATA);
2916 memset(&js->sectors, 0, sizeof(js->sectors));
7eada909
MP		 2917 js->commit_id = dm_integrity_commit_id(ic, i, j, commit_seq);
2918 }
2919 for (j = 0; j < ic->journal_section_entries; j++) {
2920 struct journal_entry *je = access_journal_entry(ic, i, j);
0ef0b471 2921
7eada909
MP		 2922 journal_entry_set_unused(je);
2923 }
2924 }
2925
2926 write_journal(ic, start_section, n_sections);
2927}
2928
86a3238c 2929static int find_commit_seq(struct dm_integrity_c *ic, unsigned int i, unsigned int j, commit_id_t id)
7eada909
MP		 2930{
2931 unsigned char k;
0ef0b471 2932
7eada909
MP		 2933 for (k = 0; k < N_COMMIT_IDS; k++) {
2934 if (dm_integrity_commit_id(ic, i, j, k) == id)
2935 return k;
2936 }
2937 dm_integrity_io_error(ic, "journal commit id", -EIO);
2938 return -EIO;
2939}
2940
2941static void replay_journal(struct dm_integrity_c *ic)
2942{
86a3238c 2943 unsigned int i, j;
7eada909 2944 bool used_commit_ids[N_COMMIT_IDS];
86a3238c
HM		 2945 unsigned int max_commit_id_sections[N_COMMIT_IDS];
2946 unsigned int write_start, write_sections;
2947 unsigned int continue_section;
7eada909
MP		 2948 bool journal_empty;
2949 unsigned char unused, last_used, want_commit_seq;
2950
c2bcb2b7
MP		 2951 if (ic->mode == 'R')
2952 return;
2953
7eada909
MP		 2954 if (ic->journal_uptodate)
2955 return;
2956
2957 last_used = 0;
2958 write_start = 0;
2959
2960 if (!ic->just_formatted) {
2961 DEBUG_print("reading journal\n");
c9154a4c 2962 rw_journal(ic, REQ_OP_READ, 0, ic->journal_sections, NULL);
7eada909
MP		 2963 if (ic->journal_io)
2964 DEBUG_bytes(lowmem_page_address(ic->journal_io[0].page), 64, "read journal");
2965 if (ic->journal_io) {
2966 struct journal_completion crypt_comp;
0ef0b471 2967
7eada909 2968 crypt_comp.ic = ic;
b5e8ad92 2969 init_completion(&crypt_comp.comp);
7eada909
MP		 2970 crypt_comp.in_flight = (atomic_t)ATOMIC_INIT(0);
2971 encrypt_journal(ic, false, 0, ic->journal_sections, &crypt_comp);
2972 wait_for_completion(&crypt_comp.comp);
2973 }
2974 DEBUG_bytes(lowmem_page_address(ic->journal[0].page), 64, "decrypted journal");
2975 }
2976
2977 if (dm_integrity_failed(ic))
2978 goto clear_journal;
2979
2980 journal_empty = true;
8d1058fb
HM		 2981 memset(used_commit_ids, 0, sizeof(used_commit_ids));
2982 memset(max_commit_id_sections, 0, sizeof(max_commit_id_sections));
7eada909
MP		 2983 for (i = 0; i < ic->journal_sections; i++) {
2984 for (j = 0; j < ic->journal_section_sectors; j++) {
2985 int k;
2986 struct journal_sector *js = access_journal(ic, i, j);
0ef0b471 2987
7eada909
MP		 2988 k = find_commit_seq(ic, i, j, js->commit_id);
2989 if (k < 0)
2990 goto clear_journal;
2991 used_commit_ids[k] = true;
2992 max_commit_id_sections[k] = i;
2993 }
2994 if (journal_empty) {
2995 for (j = 0; j < ic->journal_section_entries; j++) {
2996 struct journal_entry *je = access_journal_entry(ic, i, j);
0ef0b471 2997
7eada909
MP		 2998 if (!journal_entry_is_unused(je)) {
2999 journal_empty = false;
3000 break;
3001 }
3002 }
3003 }
3004 }
3005
3006 if (!used_commit_ids[N_COMMIT_IDS - 1]) {
3007 unused = N_COMMIT_IDS - 1;
3008 while (unused && !used_commit_ids[unused - 1])
3009 unused--;
3010 } else {
3011 for (unused = 0; unused < N_COMMIT_IDS; unused++)
3012 if (!used_commit_ids[unused])
3013 break;
3014 if (unused == N_COMMIT_IDS) {
3015 dm_integrity_io_error(ic, "journal commit ids", -EIO);
3016 goto clear_journal;
3017 }
3018 }
3019 DEBUG_print("first unused commit seq %d [%d,%d,%d,%d]\n",
3020 unused, used_commit_ids[0], used_commit_ids[1],
3021 used_commit_ids[2], used_commit_ids[3]);
3022
3023 last_used = prev_commit_seq(unused);
3024 want_commit_seq = prev_commit_seq(last_used);
3025
3026 if (!used_commit_ids[want_commit_seq] && used_commit_ids[prev_commit_seq(want_commit_seq)])
3027 journal_empty = true;
3028
3029 write_start = max_commit_id_sections[last_used] + 1;
3030 if (unlikely(write_start >= ic->journal_sections))
3031 want_commit_seq = next_commit_seq(want_commit_seq);
3032 wraparound_section(ic, &write_start);
3033
3034 i = write_start;
3035 for (write_sections = 0; write_sections < ic->journal_sections; write_sections++) {
3036 for (j = 0; j < ic->journal_section_sectors; j++) {
3037 struct journal_sector *js = access_journal(ic, i, j);
3038
3039 if (js->commit_id != dm_integrity_commit_id(ic, i, j, want_commit_seq)) {
3040 /*
3041 * This could be caused by crash during writing.
3042 * We won't replay the inconsistent part of the
3043 * journal.
3044 */
3045 DEBUG_print("commit id mismatch at position (%u, %u): %d != %d\n",
3046 i, j, find_commit_seq(ic, i, j, js->commit_id), want_commit_seq);
3047 goto brk;
3048 }
3049 }
3050 i++;
3051 if (unlikely(i >= ic->journal_sections))
3052 want_commit_seq = next_commit_seq(want_commit_seq);
3053 wraparound_section(ic, &i);
3054 }
3055brk:
3056
3057 if (!journal_empty) {
3058 DEBUG_print("replaying %u sections, starting at %u, commit seq %d\n",
3059 write_sections, write_start, want_commit_seq);
3060 do_journal_write(ic, write_start, write_sections, true);
3061 }
3062
3063 if (write_sections == ic->journal_sections && (ic->mode == 'J' || journal_empty)) {
3064 continue_section = write_start;
3065 ic->commit_seq = want_commit_seq;
3066 DEBUG_print("continuing from section %u, commit seq %d\n", write_start, ic->commit_seq);
3067 } else {
86a3238c 3068 unsigned int s;
7eada909 3069 unsigned char erase_seq;
0ef0b471 3070
7eada909
MP		 3071clear_journal:
3072 DEBUG_print("clearing journal\n");
3073
3074 erase_seq = prev_commit_seq(prev_commit_seq(last_used));
3075 s = write_start;
3076 init_journal(ic, s, 1, erase_seq);
3077 s++;
3078 wraparound_section(ic, &s);
3079 if (ic->journal_sections >= 2) {
3080 init_journal(ic, s, ic->journal_sections - 2, erase_seq);
3081 s += ic->journal_sections - 2;
3082 wraparound_section(ic, &s);
3083 init_journal(ic, s, 1, erase_seq);
3084 }
3085
3086 continue_section = 0;
3087 ic->commit_seq = next_commit_seq(erase_seq);
3088 }
3089
3090 ic->committed_section = continue_section;
3091 ic->n_committed_sections = 0;
3092
3093 ic->uncommitted_section = continue_section;
3094 ic->n_uncommitted_sections = 0;
3095
3096 ic->free_section = continue_section;
3097 ic->free_section_entry = 0;
3098 ic->free_sectors = ic->journal_entries;
3099
3100 ic->journal_tree_root = RB_ROOT;
3101 for (i = 0; i < ic->journal_entries; i++)
3102 init_journal_node(&ic->journal_tree[i]);
3103}
3104
48271493 3105static void dm_integrity_enter_synchronous_mode(struct dm_integrity_c *ic)
1f5a7759 3106{
1c131886 3107 DEBUG_print("%s\n", __func__);
1f5a7759
MP		 3108
3109 if (ic->mode == 'B') {
48271493
MP		 3110 ic->bitmap_flush_interval = msecs_to_jiffies(10) + 1;
3111 ic->synchronous_mode = 1;
3112
1f5a7759
MP		 3113 cancel_delayed_work_sync(&ic->bitmap_flush_work);
3114 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, 0);
3115 flush_workqueue(ic->commit_wq);
3116 }
48271493
MP		 3117}
3118
3119static int dm_integrity_reboot(struct notifier_block *n, unsigned long code, void *x)
3120{
3121 struct dm_integrity_c *ic = container_of(n, struct dm_integrity_c, reboot_notifier);
3122
1c131886 3123 DEBUG_print("%s\n", __func__);
48271493
MP		 3124
3125 dm_integrity_enter_synchronous_mode(ic);
1f5a7759
MP		 3126
3127 return NOTIFY_DONE;
3128}
3129
7eada909
MP		 3130static void dm_integrity_postsuspend(struct dm_target *ti)
3131{
26cb62a2 3132 struct dm_integrity_c *ic = ti->private;
468dfca3 3133 int r;
7eada909 3134
1f5a7759 3135 WARN_ON(unregister_reboot_notifier(&ic->reboot_notifier));
7eada909
MP		 3136
3137 del_timer_sync(&ic->autocommit_timer);
3138
a3fcf725
MP		 3139 if (ic->recalc_wq)
3140 drain_workqueue(ic->recalc_wq);
3141
468dfca3
MP		 3142 if (ic->mode == 'B')
3143 cancel_delayed_work_sync(&ic->bitmap_flush_work);
3144
7eada909
MP		 3145 queue_work(ic->commit_wq, &ic->commit_work);
3146 drain_workqueue(ic->commit_wq);
3147
3148 if (ic->mode == 'J') {
5e5dab5e 3149 queue_work(ic->writer_wq, &ic->writer_work);
7eada909 3150 drain_workqueue(ic->writer_wq);
9b594826 3151 dm_integrity_flush_buffers(ic, true);
984bf2cc
MP		 3152 if (ic->wrote_to_journal) {
3153 init_journal(ic, ic->free_section,
3154 ic->journal_sections - ic->free_section, ic->commit_seq);
3155 if (ic->free_section) {
3156 init_journal(ic, 0, ic->free_section,
3157 next_commit_seq(ic->commit_seq));
3158 }
3159 }
7eada909
MP		 3160 }
3161
468dfca3 3162 if (ic->mode == 'B') {
9b594826 3163 dm_integrity_flush_buffers(ic, true);
468dfca3 3164#if 1
05d6909e 3165 /* set to 0 to test bitmap replay code */
468dfca3
MP		 3166 init_journal(ic, 0, ic->journal_sections, 0);
3167 ic->sb->flags &= ~cpu_to_le32(SB_FLAG_DIRTY_BITMAP);
c9154a4c 3168 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
468dfca3
MP		 3169 if (unlikely(r))
3170 dm_integrity_io_error(ic, "writing superblock", r);
3171#endif
3172 }
3173
7eada909
MP		 3174 BUG_ON(!RB_EMPTY_ROOT(&ic->in_progress));
3175
3176 ic->journal_uptodate = true;
3177}
3178
3179static void dm_integrity_resume(struct dm_target *ti)
3180{
26cb62a2 3181 struct dm_integrity_c *ic = ti->private;
1ac2c15a 3182 __u64 old_provided_data_sectors = le64_to_cpu(ic->sb->provided_data_sectors);
468dfca3 3183 int r;
1ac2c15a 3184
468dfca3
MP		 3185 DEBUG_print("resume\n");
3186
984bf2cc
MP		 3187 ic->wrote_to_journal = false;
3188
1ac2c15a
MP		 3189 if (ic->provided_data_sectors != old_provided_data_sectors) {
3190 if (ic->provided_data_sectors > old_provided_data_sectors &&
3191 ic->mode == 'B' &&
3192 ic->sb->log2_blocks_per_bitmap_bit == ic->log2_blocks_per_bitmap_bit) {
c9154a4c 3193 rw_journal_sectors(ic, REQ_OP_READ, 0,
1ac2c15a
MP		 3194 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3195 block_bitmap_op(ic, ic->journal, old_provided_data_sectors,
3196 ic->provided_data_sectors - old_provided_data_sectors, BITMAP_OP_SET);
c9154a4c 3197 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
1ac2c15a
MP		 3198 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3199 }
3200
3201 ic->sb->provided_data_sectors = cpu_to_le64(ic->provided_data_sectors);
c9154a4c 3202 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
1ac2c15a
MP		 3203 if (unlikely(r))
3204 dm_integrity_io_error(ic, "writing superblock", r);
3205 }
3206
468dfca3
MP		 3207 if (ic->sb->flags & cpu_to_le32(SB_FLAG_DIRTY_BITMAP)) {
3208 DEBUG_print("resume dirty_bitmap\n");
c9154a4c 3209 rw_journal_sectors(ic, REQ_OP_READ, 0,
05d6909e 3210 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
468dfca3 3211 if (ic->mode == 'B') {
db7b93e3
MP		 3212 if (ic->sb->log2_blocks_per_bitmap_bit == ic->log2_blocks_per_bitmap_bit &&
3213 !ic->reset_recalculate_flag) {
468dfca3
MP		 3214 block_bitmap_copy(ic, ic->recalc_bitmap, ic->journal);
3215 block_bitmap_copy(ic, ic->may_write_bitmap, ic->journal);
05d6909e
MS		 3216 if (!block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors,
3217 BITMAP_OP_TEST_ALL_CLEAR)) {
468dfca3
MP		 3218 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3219 ic->sb->recalc_sector = cpu_to_le64(0);
3220 }
3221 } else {
05d6909e
MS		 3222 DEBUG_print("non-matching blocks_per_bitmap_bit: %u, %u\n",
3223 ic->sb->log2_blocks_per_bitmap_bit, ic->log2_blocks_per_bitmap_bit);
468dfca3
MP		 3224 ic->sb->log2_blocks_per_bitmap_bit = ic->log2_blocks_per_bitmap_bit;
3225 block_bitmap_op(ic, ic->recalc_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_SET);
3226 block_bitmap_op(ic, ic->may_write_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_SET);
3227 block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors, BITMAP_OP_SET);
c9154a4c 3228 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
05d6909e 3229 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
468dfca3
MP		 3230 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3231 ic->sb->recalc_sector = cpu_to_le64(0);
3232 }
3233 } else {
3234 if (!(ic->sb->log2_blocks_per_bitmap_bit == ic->log2_blocks_per_bitmap_bit &&
db7b93e3
MP		 3235 block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors, BITMAP_OP_TEST_ALL_CLEAR)) ||
3236 ic->reset_recalculate_flag) {
468dfca3
MP		 3237 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3238 ic->sb->recalc_sector = cpu_to_le64(0);
3239 }
3240 init_journal(ic, 0, ic->journal_sections, 0);
3241 replay_journal(ic);
3242 ic->sb->flags &= ~cpu_to_le32(SB_FLAG_DIRTY_BITMAP);
3243 }
c9154a4c 3244 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
468dfca3
MP		 3245 if (unlikely(r))
3246 dm_integrity_io_error(ic, "writing superblock", r);
3247 } else {
3248 replay_journal(ic);
db7b93e3
MP		 3249 if (ic->reset_recalculate_flag) {
3250 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3251 ic->sb->recalc_sector = cpu_to_le64(0);
3252 }
468dfca3 3253 if (ic->mode == 'B') {
468dfca3
MP		 3254 ic->sb->flags |= cpu_to_le32(SB_FLAG_DIRTY_BITMAP);
3255 ic->sb->log2_blocks_per_bitmap_bit = ic->log2_blocks_per_bitmap_bit;
c9154a4c 3256 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
468dfca3
MP		 3257 if (unlikely(r))
3258 dm_integrity_io_error(ic, "writing superblock", r);
7eada909 3259
d5bdf661
MP		 3260 block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
3261 block_bitmap_op(ic, ic->recalc_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
3262 block_bitmap_op(ic, ic->may_write_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
3263 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING) &&
3264 le64_to_cpu(ic->sb->recalc_sector) < ic->provided_data_sectors) {
3265 block_bitmap_op(ic, ic->journal, le64_to_cpu(ic->sb->recalc_sector),
3266 ic->provided_data_sectors - le64_to_cpu(ic->sb->recalc_sector), BITMAP_OP_SET);
3267 block_bitmap_op(ic, ic->recalc_bitmap, le64_to_cpu(ic->sb->recalc_sector),
3268 ic->provided_data_sectors - le64_to_cpu(ic->sb->recalc_sector), BITMAP_OP_SET);
3269 block_bitmap_op(ic, ic->may_write_bitmap, le64_to_cpu(ic->sb->recalc_sector),
3270 ic->provided_data_sectors - le64_to_cpu(ic->sb->recalc_sector), BITMAP_OP_SET);
3271 }
c9154a4c 3272 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
05d6909e 3273 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
468dfca3
MP		 3274 }
3275 }
7eada909 3276
468dfca3
MP		 3277 DEBUG_print("testing recalc: %x\n", ic->sb->flags);
3278 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) {
a3fcf725 3279 __u64 recalc_pos = le64_to_cpu(ic->sb->recalc_sector);
0ef0b471 3280
7649194a 3281 DEBUG_print("recalc pos: %llx / %llx\n", recalc_pos, ic->provided_data_sectors);
a3fcf725
MP		 3282 if (recalc_pos < ic->provided_data_sectors) {
3283 queue_work(ic->recalc_wq, &ic->recalc_work);
3284 } else if (recalc_pos > ic->provided_data_sectors) {
3285 ic->sb->recalc_sector = cpu_to_le64(ic->provided_data_sectors);
3286 recalc_write_super(ic);
3287 }
3288 }
1f5a7759
MP		 3289
3290 ic->reboot_notifier.notifier_call = dm_integrity_reboot;
3291 ic->reboot_notifier.next = NULL;
3292 ic->reboot_notifier.priority = INT_MAX - 1; /* be notified after md and before hardware drivers */
3293 WARN_ON(register_reboot_notifier(&ic->reboot_notifier));
48271493
MP		 3294
3295#if 0
05d6909e 3296 /* set to 1 to stress test synchronous mode */
48271493
MP		 3297 dm_integrity_enter_synchronous_mode(ic);
3298#endif
7eada909
MP		 3299}
3300
3301static void dm_integrity_status(struct dm_target *ti, status_type_t type,
86a3238c 3302 unsigned int status_flags, char *result, unsigned int maxlen)
7eada909 3303{
26cb62a2 3304 struct dm_integrity_c *ic = ti->private;
86a3238c 3305 unsigned int arg_count;
7eada909
MP		 3306 size_t sz = 0;
3307
3308 switch (type) {
3309 case STATUSTYPE_INFO:
f84fd2c9 3310 DMEMIT("%llu %llu",
e7fc1e57 3311 (unsigned long long)atomic64_read(&ic->number_of_mismatches),
7649194a 3312 ic->provided_data_sectors);
a3fcf725 3313 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))
7649194a 3314 DMEMIT(" %llu", le64_to_cpu(ic->sb->recalc_sector));
a3fcf725
MP		 3315 else
3316 DMEMIT(" -");
7eada909
MP		 3317 break;
3318
3319 case STATUSTYPE_TABLE: {
3320 __u64 watermark_percentage = (__u64)(ic->journal_entries - ic->free_sectors_threshold) * 100;
0ef0b471 3321
7eada909
MP		 3322 watermark_percentage += ic->journal_entries / 2;
3323 do_div(watermark_percentage, ic->journal_entries);
893e3c39 3324 arg_count = 3;
356d9d52 3325 arg_count += !!ic->meta_dev;
9d609f85 3326 arg_count += ic->sectors_per_block != 1;
a3fcf725 3327 arg_count += !!(ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING));
db7b93e3 3328 arg_count += ic->reset_recalculate_flag;
84597a44 3329 arg_count += ic->discard;
893e3c39
MP		 3330 arg_count += ic->mode == 'J';
3331 arg_count += ic->mode == 'J';
468dfca3
MP		 3332 arg_count += ic->mode == 'B';
3333 arg_count += ic->mode == 'B';
7eada909
MP		 3334 arg_count += !!ic->internal_hash_alg.alg_string;
3335 arg_count += !!ic->journal_crypt_alg.alg_string;
3336 arg_count += !!ic->journal_mac_alg.alg_string;
d537858a 3337 arg_count += (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING)) != 0;
09d85f8d 3338 arg_count += (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) != 0;
5c024064 3339 arg_count += ic->legacy_recalculate;
7649194a 3340 DMEMIT("%s %llu %u %c %u", ic->dev->name, ic->start,
7eada909 3341 ic->tag_size, ic->mode, arg_count);
356d9d52
MP		 3342 if (ic->meta_dev)
3343 DMEMIT(" meta_device:%s", ic->meta_dev->name);
a3fcf725
MP		 3344 if (ic->sectors_per_block != 1)
3345 DMEMIT(" block_size:%u", ic->sectors_per_block << SECTOR_SHIFT);
7fc2e47f 3346 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))
a3fcf725 3347 DMEMIT(" recalculate");
db7b93e3
MP		 3348 if (ic->reset_recalculate_flag)
3349 DMEMIT(" reset_recalculate");
84597a44
MP		 3350 if (ic->discard)
3351 DMEMIT(" allow_discards");
56b67a4f
MP		 3352 DMEMIT(" journal_sectors:%u", ic->initial_sectors - SB_SECTORS);
3353 DMEMIT(" interleave_sectors:%u", 1U << ic->sb->log2_interleave_sectors);
3354 DMEMIT(" buffer_sectors:%u", 1U << ic->log2_buffer_sectors);
893e3c39 3355 if (ic->mode == 'J') {
86a3238c 3356 DMEMIT(" journal_watermark:%u", (unsigned int)watermark_percentage);
893e3c39
MP		 3357 DMEMIT(" commit_time:%u", ic->autocommit_msec);
3358 }
468dfca3 3359 if (ic->mode == 'B') {
7649194a 3360 DMEMIT(" sectors_per_bit:%llu", (sector_t)ic->sectors_per_block << ic->log2_blocks_per_bitmap_bit);
468dfca3
MP		 3361 DMEMIT(" bitmap_flush_interval:%u", jiffies_to_msecs(ic->bitmap_flush_interval));
3362 }
d537858a
MP		 3363 if ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING)) != 0)
3364 DMEMIT(" fix_padding");
09d85f8d
MP		 3365 if ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) != 0)
3366 DMEMIT(" fix_hmac");
5c024064
MP		 3367 if (ic->legacy_recalculate)
3368 DMEMIT(" legacy_recalculate");
7eada909
MP		 3369
3370#define EMIT_ALG(a, n) \
3371 do { \
3372 if (ic->a.alg_string) { \
3373 DMEMIT(" %s:%s", n, ic->a.alg_string); \
3374 if (ic->a.key_string) \
3375 DMEMIT(":%s", ic->a.key_string);\
3376 } \
3377 } while (0)
56b67a4f
MP		 3378 EMIT_ALG(internal_hash_alg, "internal_hash");
3379 EMIT_ALG(journal_crypt_alg, "journal_crypt");
3380 EMIT_ALG(journal_mac_alg, "journal_mac");
7eada909
MP		 3381 break;
3382 }
8ec45662
TS		 3383 case STATUSTYPE_IMA:
3384 DMEMIT_TARGET_NAME_VERSION(ti->type);
3385 DMEMIT(",dev_name=%s,start=%llu,tag_size=%u,mode=%c",
3386 ic->dev->name, ic->start, ic->tag_size, ic->mode);
3387
3388 if (ic->meta_dev)
3389 DMEMIT(",meta_device=%s", ic->meta_dev->name);
3390 if (ic->sectors_per_block != 1)
3391 DMEMIT(",block_size=%u", ic->sectors_per_block << SECTOR_SHIFT);
3392
3393 DMEMIT(",recalculate=%c", (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) ?
3394 'y' : 'n');
3395 DMEMIT(",allow_discards=%c", ic->discard ? 'y' : 'n');
3396 DMEMIT(",fix_padding=%c",
3397 ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING)) != 0) ? 'y' : 'n');
3398 DMEMIT(",fix_hmac=%c",
3399 ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) != 0) ? 'y' : 'n');
3400 DMEMIT(",legacy_recalculate=%c", ic->legacy_recalculate ? 'y' : 'n');
3401
3402 DMEMIT(",journal_sectors=%u", ic->initial_sectors - SB_SECTORS);
3403 DMEMIT(",interleave_sectors=%u", 1U << ic->sb->log2_interleave_sectors);
3404 DMEMIT(",buffer_sectors=%u", 1U << ic->log2_buffer_sectors);
8ec45662
TS		 3405 DMEMIT(";");
3406 break;
7eada909
MP		 3407 }
3408}
3409
3410static int dm_integrity_iterate_devices(struct dm_target *ti,
3411 iterate_devices_callout_fn fn, void *data)
3412{
3413 struct dm_integrity_c *ic = ti->private;
3414
356d9d52
MP		 3415 if (!ic->meta_dev)
3416 return fn(ti, ic->dev, ic->start + ic->initial_sectors + ic->metadata_run, ti->len, data);
3417 else
3418 return fn(ti, ic->dev, 0, ti->len, data);
7eada909
MP		 3419}
3420
9d609f85
MP		 3421static void dm_integrity_io_hints(struct dm_target *ti, struct queue_limits *limits)
3422{
3423 struct dm_integrity_c *ic = ti->private;
3424
3425 if (ic->sectors_per_block > 1) {
3426 limits->logical_block_size = ic->sectors_per_block << SECTOR_SHIFT;
3427 limits->physical_block_size = ic->sectors_per_block << SECTOR_SHIFT;
3428 blk_limits_io_min(limits, ic->sectors_per_block << SECTOR_SHIFT);
29aa778b 3429 limits->dma_alignment = limits->logical_block_size - 1;
9d609f85
MP		 3430 }
3431}
3432
7eada909
MP		 3433static void calculate_journal_section_size(struct dm_integrity_c *ic)
3434{
86a3238c 3435 unsigned int sector_space = JOURNAL_SECTOR_DATA;
7eada909
MP		 3436
3437 ic->journal_sections = le32_to_cpu(ic->sb->journal_sections);
9d609f85 3438 ic->journal_entry_size = roundup(offsetof(struct journal_entry, last_bytes[ic->sectors_per_block]) + ic->tag_size,
7eada909
MP		 3439 JOURNAL_ENTRY_ROUNDUP);
3440
3441 if (ic->sb->flags & cpu_to_le32(SB_FLAG_HAVE_JOURNAL_MAC))
3442 sector_space -= JOURNAL_MAC_PER_SECTOR;
3443 ic->journal_entries_per_sector = sector_space / ic->journal_entry_size;
3444 ic->journal_section_entries = ic->journal_entries_per_sector * JOURNAL_BLOCK_SECTORS;
9d609f85 3445 ic->journal_section_sectors = (ic->journal_section_entries << ic->sb->log2_sectors_per_block) + JOURNAL_BLOCK_SECTORS;
7eada909
MP		 3446 ic->journal_entries = ic->journal_section_entries * ic->journal_sections;
3447}
3448
3449static int calculate_device_limits(struct dm_integrity_c *ic)
3450{
3451 __u64 initial_sectors;
7eada909
MP		 3452
3453 calculate_journal_section_size(ic);
3454 initial_sectors = SB_SECTORS + (__u64)ic->journal_section_sectors * ic->journal_sections;
356d9d52 3455 if (initial_sectors + METADATA_PADDING_SECTORS >= ic->meta_device_sectors || initial_sectors > UINT_MAX)
7eada909
MP		 3456 return -EINVAL;
3457 ic->initial_sectors = initial_sectors;
3458
356d9d52
MP		 3459 if (!ic->meta_dev) {
3460 sector_t last_sector, last_area, last_offset;
7eada909 3461
d537858a
MP		 3462 /* we have to maintain excessive padding for compatibility with existing volumes */
3463 __u64 metadata_run_padding =
3464 ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING) ?
3465 (__u64)(METADATA_PADDING_SECTORS << SECTOR_SHIFT) :
3466 (__u64)(1 << SECTOR_SHIFT << METADATA_PADDING_SECTORS);
3467
3468 ic->metadata_run = round_up((__u64)ic->tag_size << (ic->sb->log2_interleave_sectors - ic->sb->log2_sectors_per_block),
3469 metadata_run_padding) >> SECTOR_SHIFT;
356d9d52
MP		 3470 if (!(ic->metadata_run & (ic->metadata_run - 1)))
3471 ic->log2_metadata_run = __ffs(ic->metadata_run);
3472 else
3473 ic->log2_metadata_run = -1;
7eada909 3474
356d9d52
MP		 3475 get_area_and_offset(ic, ic->provided_data_sectors - 1, &last_area, &last_offset);
3476 last_sector = get_data_sector(ic, last_area, last_offset);
3477 if (last_sector < ic->start || last_sector >= ic->meta_device_sectors)
3478 return -EINVAL;
3479 } else {
30bba430 3480 __u64 meta_size = (ic->provided_data_sectors >> ic->sb->log2_sectors_per_block) * ic->tag_size;
0ef0b471 3481
356d9d52
MP		 3482 meta_size = (meta_size + ((1U << (ic->log2_buffer_sectors + SECTOR_SHIFT)) - 1))
3483 >> (ic->log2_buffer_sectors + SECTOR_SHIFT);
3484 meta_size <<= ic->log2_buffer_sectors;
3485 if (ic->initial_sectors + meta_size < ic->initial_sectors ||
3486 ic->initial_sectors + meta_size > ic->meta_device_sectors)
3487 return -EINVAL;
3488 ic->metadata_run = 1;
3489 ic->log2_metadata_run = 0;
3490 }
7eada909
MP		 3491
3492 return 0;
3493}
3494
87fb177b
MP		 3495static void get_provided_data_sectors(struct dm_integrity_c *ic)
3496{
3497 if (!ic->meta_dev) {
3498 int test_bit;
0ef0b471 3499
87fb177b
MP		 3500 ic->provided_data_sectors = 0;
3501 for (test_bit = fls64(ic->meta_device_sectors) - 1; test_bit >= 3; test_bit--) {
3502 __u64 prev_data_sectors = ic->provided_data_sectors;
3503
3504 ic->provided_data_sectors |= (sector_t)1 << test_bit;
3505 if (calculate_device_limits(ic))
3506 ic->provided_data_sectors = prev_data_sectors;
3507 }
3508 } else {
3509 ic->provided_data_sectors = ic->data_device_sectors;
3510 ic->provided_data_sectors &= ~(sector_t)(ic->sectors_per_block - 1);
3511 }
3512}
3513
86a3238c
HM		 3514static int initialize_superblock(struct dm_integrity_c *ic,
3515 unsigned int journal_sectors, unsigned int interleave_sectors)
7eada909 3516{
86a3238c 3517 unsigned int journal_sections;
7eada909
MP		 3518 int test_bit;
3519
56b67a4f 3520 memset(ic->sb, 0, SB_SECTORS << SECTOR_SHIFT);
7eada909 3521 memcpy(ic->sb->magic, SB_MAGIC, 8);
7eada909 3522 ic->sb->integrity_tag_size = cpu_to_le16(ic->tag_size);
9d609f85 3523 ic->sb->log2_sectors_per_block = __ffs(ic->sectors_per_block);
7eada909
MP		 3524 if (ic->journal_mac_alg.alg_string)
3525 ic->sb->flags |= cpu_to_le32(SB_FLAG_HAVE_JOURNAL_MAC);
3526
3527 calculate_journal_section_size(ic);
3528 journal_sections = journal_sectors / ic->journal_section_sectors;
3529 if (!journal_sections)
3530 journal_sections = 1;
7eada909 3531
09d85f8d
MP		 3532 if (ic->fix_hmac && (ic->internal_hash_alg.alg_string || ic->journal_mac_alg.alg_string)) {
3533 ic->sb->flags |= cpu_to_le32(SB_FLAG_FIXED_HMAC);
3534 get_random_bytes(ic->sb->salt, SALT_SIZE);
3535 }
3536
356d9d52 3537 if (!ic->meta_dev) {
d537858a
MP		 3538 if (ic->fix_padding)
3539 ic->sb->flags |= cpu_to_le32(SB_FLAG_FIXED_PADDING);
356d9d52
MP		 3540 ic->sb->journal_sections = cpu_to_le32(journal_sections);
3541 if (!interleave_sectors)
3542 interleave_sectors = DEFAULT_INTERLEAVE_SECTORS;
3543 ic->sb->log2_interleave_sectors = __fls(interleave_sectors);
ba287d7c
HM		 3544 ic->sb->log2_interleave_sectors = max_t(__u8, MIN_LOG2_INTERLEAVE_SECTORS, ic->sb->log2_interleave_sectors);
3545 ic->sb->log2_interleave_sectors = min_t(__u8, MAX_LOG2_INTERLEAVE_SECTORS, ic->sb->log2_interleave_sectors);
356d9d52 3546
87fb177b 3547 get_provided_data_sectors(ic);
356d9d52
MP		 3548 if (!ic->provided_data_sectors)
3549 return -EINVAL;
3550 } else {
3551 ic->sb->log2_interleave_sectors = 0;
87fb177b
MP		 3552
3553 get_provided_data_sectors(ic);
3554 if (!ic->provided_data_sectors)
3555 return -EINVAL;
356d9d52
MP		 3556
3557try_smaller_buffer:
3558 ic->sb->journal_sections = cpu_to_le32(0);
3559 for (test_bit = fls(journal_sections) - 1; test_bit >= 0; test_bit--) {
3560 __u32 prev_journal_sections = le32_to_cpu(ic->sb->journal_sections);
3561 __u32 test_journal_sections = prev_journal_sections | (1U << test_bit);
0ef0b471 3562
356d9d52
MP		 3563 if (test_journal_sections > journal_sections)
3564 continue;
3565 ic->sb->journal_sections = cpu_to_le32(test_journal_sections);
3566 if (calculate_device_limits(ic))
3567 ic->sb->journal_sections = cpu_to_le32(prev_journal_sections);
7eada909 3568
356d9d52
MP		 3569 }
3570 if (!le32_to_cpu(ic->sb->journal_sections)) {
3571 if (ic->log2_buffer_sectors > 3) {
3572 ic->log2_buffer_sectors--;
3573 goto try_smaller_buffer;
3574 }
3575 return -EINVAL;
3576 }
7eada909
MP		 3577 }
3578
7eada909
MP		 3579 ic->sb->provided_data_sectors = cpu_to_le64(ic->provided_data_sectors);
3580
1f9fc0b8
MP		 3581 sb_set_version(ic);
3582
7eada909
MP		 3583 return 0;
3584}
3585
3586static void dm_integrity_set(struct dm_target *ti, struct dm_integrity_c *ic)
3587{
3588 struct gendisk *disk = dm_disk(dm_table_get_md(ti->table));
3589 struct blk_integrity bi;
3590
3591 memset(&bi, 0, sizeof(bi));
3592 bi.profile = &dm_integrity_profile;
9d609f85
MP		 3593 bi.tuple_size = ic->tag_size;
3594 bi.tag_size = bi.tuple_size;
84ff1bcc 3595 bi.interval_exp = ic->sb->log2_sectors_per_block + SECTOR_SHIFT;
7eada909
MP		 3596
3597 blk_integrity_register(disk, &bi);
3598 blk_queue_max_integrity_segments(disk->queue, UINT_MAX);
3599}
3600
d5027e03 3601static void dm_integrity_free_page_list(struct page_list *pl)
7eada909 3602{
86a3238c 3603 unsigned int i;
7eada909
MP		 3604
3605 if (!pl)
3606 return;
d5027e03
MP		 3607 for (i = 0; pl[i].page; i++)
3608 __free_page(pl[i].page);
7eada909
MP		 3609 kvfree(pl);
3610}
3611
86a3238c 3612static struct page_list *dm_integrity_alloc_page_list(unsigned int n_pages)
7eada909 3613{
7eada909 3614 struct page_list *pl;
86a3238c 3615 unsigned int i;
7eada909 3616
d5027e03 3617 pl = kvmalloc_array(n_pages + 1, sizeof(struct page_list), GFP_KERNEL | __GFP_ZERO);
7eada909
MP		 3618 if (!pl)
3619 return NULL;
3620
d5027e03 3621 for (i = 0; i < n_pages; i++) {
7eada909
MP		 3622 pl[i].page = alloc_page(GFP_KERNEL);
3623 if (!pl[i].page) {
d5027e03 3624 dm_integrity_free_page_list(pl);
7eada909
MP		 3625 return NULL;
3626 }
3627 if (i)
3628 pl[i - 1].next = &pl[i];
3629 }
d5027e03
MP		 3630 pl[i].page = NULL;
3631 pl[i].next = NULL;
7eada909
MP		 3632
3633 return pl;
3634}
3635
3636static void dm_integrity_free_journal_scatterlist(struct dm_integrity_c *ic, struct scatterlist **sl)
3637{
86a3238c 3638 unsigned int i;
0ef0b471 3639
7eada909
MP		 3640 for (i = 0; i < ic->journal_sections; i++)
3641 kvfree(sl[i]);
fc8cec11 3642 kvfree(sl);
7eada909
MP		 3643}
3644
05d6909e
MS		 3645static struct scatterlist **dm_integrity_alloc_journal_scatterlist(struct dm_integrity_c *ic,
3646 struct page_list *pl)
7eada909
MP		 3647{
3648 struct scatterlist **sl;
86a3238c 3649 unsigned int i;
7eada909 3650
344476e1
KC		 3651 sl = kvmalloc_array(ic->journal_sections,
3652 sizeof(struct scatterlist *),
3653 GFP_KERNEL | __GFP_ZERO);
7eada909
MP		 3654 if (!sl)
3655 return NULL;
3656
3657 for (i = 0; i < ic->journal_sections; i++) {
3658 struct scatterlist *s;
86a3238c
HM		 3659 unsigned int start_index, start_offset;
3660 unsigned int end_index, end_offset;
3661 unsigned int n_pages;
3662 unsigned int idx;
7eada909
MP		 3663
3664 page_list_location(ic, i, 0, &start_index, &start_offset);
05d6909e
MS		 3665 page_list_location(ic, i, ic->journal_section_sectors - 1,
3666 &end_index, &end_offset);
7eada909
MP		 3667
3668 n_pages = (end_index - start_index + 1);
3669
344476e1
KC		 3670 s = kvmalloc_array(n_pages, sizeof(struct scatterlist),
3671 GFP_KERNEL);
7eada909
MP		 3672 if (!s) {
3673 dm_integrity_free_journal_scatterlist(ic, sl);
3674 return NULL;
3675 }
3676
3677 sg_init_table(s, n_pages);
3678 for (idx = start_index; idx <= end_index; idx++) {
3679 char *va = lowmem_page_address(pl[idx].page);
86a3238c 3680 unsigned int start = 0, end = PAGE_SIZE;
0ef0b471 3681
7eada909
MP		 3682 if (idx == start_index)
3683 start = start_offset;
3684 if (idx == end_index)
3685 end = end_offset + (1 << SECTOR_SHIFT);
3686 sg_set_buf(&s[idx - start_index], va + start, end - start);
3687 }
3688
3689 sl[i] = s;
3690 }
3691
3692 return sl;
3693}
3694
3695static void free_alg(struct alg_spec *a)
3696{
453431a5
WL		 3697 kfree_sensitive(a->alg_string);
3698 kfree_sensitive(a->key);
8d1058fb 3699 memset(a, 0, sizeof(*a));
7eada909
MP		 3700}
3701
3702static int get_alg_and_key(const char *arg, struct alg_spec *a, char **error, char *error_inval)
3703{
3704 char *k;
3705
3706 free_alg(a);
3707
3708 a->alg_string = kstrdup(strchr(arg, ':') + 1, GFP_KERNEL);
3709 if (!a->alg_string)
3710 goto nomem;
3711
3712 k = strchr(a->alg_string, ':');
3713 if (k) {
7eada909
MP		 3714 *k = 0;
3715 a->key_string = k + 1;
3716 if (strlen(a->key_string) & 1)
3717 goto inval;
3718
3719 a->key_size = strlen(a->key_string) / 2;
3720 a->key = kmalloc(a->key_size, GFP_KERNEL);
3721 if (!a->key)
3722 goto nomem;
6625d903
MP		 3723 if (hex2bin(a->key, a->key_string, a->key_size))
3724 goto inval;
7eada909
MP		 3725 }
3726
3727 return 0;
3728inval:
3729 *error = error_inval;
3730 return -EINVAL;
3731nomem:
3732 *error = "Out of memory for an argument";
3733 return -ENOMEM;
3734}
3735
3736static int get_mac(struct crypto_shash **hash, struct alg_spec *a, char **error,
3737 char *error_alg, char *error_key)
3738{
3739 int r;
3740
3741 if (a->alg_string) {
a7a10bce 3742 *hash = crypto_alloc_shash(a->alg_string, 0, CRYPTO_ALG_ALLOCATES_MEMORY);
7eada909
MP		 3743 if (IS_ERR(*hash)) {
3744 *error = error_alg;
3745 r = PTR_ERR(*hash);
3746 *hash = NULL;
3747 return r;
3748 }
3749
3750 if (a->key) {
3751 r = crypto_shash_setkey(*hash, a->key, a->key_size);
3752 if (r) {
3753 *error = error_key;
3754 return r;
3755 }
e16b4f99
MB		 3756 } else if (crypto_shash_get_flags(*hash) & CRYPTO_TFM_NEED_KEY) {
3757 *error = error_key;
3758 return -ENOKEY;
7eada909
MP		 3759 }
3760 }
3761
3762 return 0;
3763}
3764
1aa0efd4
MS		 3765static int create_journal(struct dm_integrity_c *ic, char **error)
3766{
3767 int r = 0;
86a3238c 3768 unsigned int i;
1aa0efd4 3769 __u64 journal_pages, journal_desc_size, journal_tree_size;
717f4b1c
MP		 3770 unsigned char *crypt_data = NULL, *crypt_iv = NULL;
3771 struct skcipher_request *req = NULL;
56b67a4f
MP		 3772
3773 ic->commit_ids[0] = cpu_to_le64(0x1111111111111111ULL);
3774 ic->commit_ids[1] = cpu_to_le64(0x2222222222222222ULL);
3775 ic->commit_ids[2] = cpu_to_le64(0x3333333333333333ULL);
3776 ic->commit_ids[3] = cpu_to_le64(0x4444444444444444ULL);
1aa0efd4
MS		 3777
3778 journal_pages = roundup((__u64)ic->journal_sections * ic->journal_section_sectors,
3779 PAGE_SIZE >> SECTOR_SHIFT) >> (PAGE_SHIFT - SECTOR_SHIFT);
3780 journal_desc_size = journal_pages * sizeof(struct page_list);
ca79b0c2 3781 if (journal_pages >= totalram_pages() - totalhigh_pages() || journal_desc_size > ULONG_MAX) {
1aa0efd4
MS		 3782 *error = "Journal doesn't fit into memory";
3783 r = -ENOMEM;
3784 goto bad;
3785 }
3786 ic->journal_pages = journal_pages;
3787
d5027e03 3788 ic->journal = dm_integrity_alloc_page_list(ic->journal_pages);
1aa0efd4
MS		 3789 if (!ic->journal) {
3790 *error = "Could not allocate memory for journal";
3791 r = -ENOMEM;
3792 goto bad;
3793 }
3794 if (ic->journal_crypt_alg.alg_string) {
86a3238c 3795 unsigned int ivsize, blocksize;
1aa0efd4
MS		 3796 struct journal_completion comp;
3797
3798 comp.ic = ic;
a7a10bce 3799 ic->journal_crypt = crypto_alloc_skcipher(ic->journal_crypt_alg.alg_string, 0, CRYPTO_ALG_ALLOCATES_MEMORY);
1aa0efd4
MS		 3800 if (IS_ERR(ic->journal_crypt)) {
3801 *error = "Invalid journal cipher";
3802 r = PTR_ERR(ic->journal_crypt);
3803 ic->journal_crypt = NULL;
3804 goto bad;
3805 }
3806 ivsize = crypto_skcipher_ivsize(ic->journal_crypt);
3807 blocksize = crypto_skcipher_blocksize(ic->journal_crypt);
3808
3809 if (ic->journal_crypt_alg.key) {
3810 r = crypto_skcipher_setkey(ic->journal_crypt, ic->journal_crypt_alg.key,
3811 ic->journal_crypt_alg.key_size);
3812 if (r) {
3813 *error = "Error setting encryption key";
3814 goto bad;
3815 }
3816 }
3817 DEBUG_print("cipher %s, block size %u iv size %u\n",
3818 ic->journal_crypt_alg.alg_string, blocksize, ivsize);
3819
d5027e03 3820 ic->journal_io = dm_integrity_alloc_page_list(ic->journal_pages);
1aa0efd4
MS		 3821 if (!ic->journal_io) {
3822 *error = "Could not allocate memory for journal io";
3823 r = -ENOMEM;
3824 goto bad;
3825 }
3826
3827 if (blocksize == 1) {
3828 struct scatterlist *sg;
717f4b1c
MP		 3829
3830 req = skcipher_request_alloc(ic->journal_crypt, GFP_KERNEL);
3831 if (!req) {
3832 *error = "Could not allocate crypt request";
3833 r = -ENOMEM;
3834 goto bad;
3835 }
3836
131670c2 3837 crypt_iv = kzalloc(ivsize, GFP_KERNEL);
717f4b1c
MP		 3838 if (!crypt_iv) {
3839 *error = "Could not allocate iv";
3840 r = -ENOMEM;
3841 goto bad;
3842 }
1aa0efd4 3843
d5027e03 3844 ic->journal_xor = dm_integrity_alloc_page_list(ic->journal_pages);
1aa0efd4
MS		 3845 if (!ic->journal_xor) {
3846 *error = "Could not allocate memory for journal xor";
3847 r = -ENOMEM;
3848 goto bad;
3849 }
3850
344476e1
KC		 3851 sg = kvmalloc_array(ic->journal_pages + 1,
3852 sizeof(struct scatterlist),
3853 GFP_KERNEL);
1aa0efd4
MS		 3854 if (!sg) {
3855 *error = "Unable to allocate sg list";
3856 r = -ENOMEM;
3857 goto bad;
3858 }
3859 sg_init_table(sg, ic->journal_pages + 1);
3860 for (i = 0; i < ic->journal_pages; i++) {
3861 char *va = lowmem_page_address(ic->journal_xor[i].page);
0ef0b471 3862
1aa0efd4
MS		 3863 clear_page(va);
3864 sg_set_buf(&sg[i], va, PAGE_SIZE);
3865 }
8d1058fb 3866 sg_set_buf(&sg[i], &ic->commit_ids, sizeof(ic->commit_ids));
1aa0efd4 3867
05d6909e 3868 skcipher_request_set_crypt(req, sg, sg,
8d1058fb 3869 PAGE_SIZE * ic->journal_pages + sizeof(ic->commit_ids), crypt_iv);
b5e8ad92 3870 init_completion(&comp.comp);
1aa0efd4
MS		 3871 comp.in_flight = (atomic_t)ATOMIC_INIT(1);
3872 if (do_crypt(true, req, &comp))
3873 wait_for_completion(&comp.comp);
3874 kvfree(sg);
3875 r = dm_integrity_failed(ic);
3876 if (r) {
3877 *error = "Unable to encrypt journal";
3878 goto bad;
3879 }
3880 DEBUG_bytes(lowmem_page_address(ic->journal_xor[0].page), 64, "xor data");
3881
3882 crypto_free_skcipher(ic->journal_crypt);
3883 ic->journal_crypt = NULL;
3884 } else {
86a3238c 3885 unsigned int crypt_len = roundup(ivsize, blocksize);
56b67a4f 3886
717f4b1c
MP		 3887 req = skcipher_request_alloc(ic->journal_crypt, GFP_KERNEL);
3888 if (!req) {
3889 *error = "Could not allocate crypt request";
3890 r = -ENOMEM;
3891 goto bad;
3892 }
3893
3894 crypt_iv = kmalloc(ivsize, GFP_KERNEL);
3895 if (!crypt_iv) {
3896 *error = "Could not allocate iv";
3897 r = -ENOMEM;
3898 goto bad;
3899 }
3900
56b67a4f
MP		 3901 crypt_data = kmalloc(crypt_len, GFP_KERNEL);
3902 if (!crypt_data) {
3903 *error = "Unable to allocate crypt data";
3904 r = -ENOMEM;
3905 goto bad;
3906 }
1aa0efd4 3907
1aa0efd4
MS		 3908 ic->journal_scatterlist = dm_integrity_alloc_journal_scatterlist(ic, ic->journal);
3909 if (!ic->journal_scatterlist) {
3910 *error = "Unable to allocate sg list";
3911 r = -ENOMEM;
3912 goto bad;
3913 }
3914 ic->journal_io_scatterlist = dm_integrity_alloc_journal_scatterlist(ic, ic->journal_io);
3915 if (!ic->journal_io_scatterlist) {
3916 *error = "Unable to allocate sg list";
3917 r = -ENOMEM;
3918 goto bad;
3919 }
344476e1
KC		 3920 ic->sk_requests = kvmalloc_array(ic->journal_sections,
3921 sizeof(struct skcipher_request *),
3922 GFP_KERNEL | __GFP_ZERO);
1aa0efd4
MS		 3923 if (!ic->sk_requests) {
3924 *error = "Unable to allocate sk requests";
3925 r = -ENOMEM;
3926 goto bad;
3927 }
3928 for (i = 0; i < ic->journal_sections; i++) {
3929 struct scatterlist sg;
3930 struct skcipher_request *section_req;
bc8f3d46 3931 __le32 section_le = cpu_to_le32(i);
1aa0efd4 3932
717f4b1c 3933 memset(crypt_iv, 0x00, ivsize);
1aa0efd4 3934 memset(crypt_data, 0x00, crypt_len);
ba287d7c 3935 memcpy(crypt_data, &section_le, min_t(size_t, crypt_len, sizeof(section_le)));
1aa0efd4
MS		 3936
3937 sg_init_one(&sg, crypt_data, crypt_len);
717f4b1c 3938 skcipher_request_set_crypt(req, &sg, &sg, crypt_len, crypt_iv);
b5e8ad92 3939 init_completion(&comp.comp);
1aa0efd4
MS		 3940 comp.in_flight = (atomic_t)ATOMIC_INIT(1);
3941 if (do_crypt(true, req, &comp))
3942 wait_for_completion(&comp.comp);
3943
3944 r = dm_integrity_failed(ic);
3945 if (r) {
3946 *error = "Unable to generate iv";
3947 goto bad;
3948 }
3949
3950 section_req = skcipher_request_alloc(ic->journal_crypt, GFP_KERNEL);
3951 if (!section_req) {
3952 *error = "Unable to allocate crypt request";
3953 r = -ENOMEM;
3954 goto bad;
3955 }
6da2ec56
KC		 3956 section_req->iv = kmalloc_array(ivsize, 2,
3957 GFP_KERNEL);
1aa0efd4
MS		 3958 if (!section_req->iv) {
3959 skcipher_request_free(section_req);
3960 *error = "Unable to allocate iv";
3961 r = -ENOMEM;
3962 goto bad;
3963 }
3964 memcpy(section_req->iv + ivsize, crypt_data, ivsize);
3965 section_req->cryptlen = (size_t)ic->journal_section_sectors << SECTOR_SHIFT;
3966 ic->sk_requests[i] = section_req;
3967 DEBUG_bytes(crypt_data, ivsize, "iv(%u)", i);
3968 }
3969 }
3970 }
3971
3972 for (i = 0; i < N_COMMIT_IDS; i++) {
86a3238c 3973 unsigned int j;
0ef0b471 3974
1aa0efd4
MS		 3975retest_commit_id:
3976 for (j = 0; j < i; j++) {
3977 if (ic->commit_ids[j] == ic->commit_ids[i]) {
3978 ic->commit_ids[i] = cpu_to_le64(le64_to_cpu(ic->commit_ids[i]) + 1);
3979 goto retest_commit_id;
3980 }
3981 }
3982 DEBUG_print("commit id %u: %016llx\n", i, ic->commit_ids[i]);
3983 }
3984
3985 journal_tree_size = (__u64)ic->journal_entries * sizeof(struct journal_node);
3986 if (journal_tree_size > ULONG_MAX) {
3987 *error = "Journal doesn't fit into memory";
3988 r = -ENOMEM;
3989 goto bad;
3990 }
702a6204 3991 ic->journal_tree = kvmalloc(journal_tree_size, GFP_KERNEL);
1aa0efd4
MS		 3992 if (!ic->journal_tree) {
3993 *error = "Could not allocate memory for journal tree";
3994 r = -ENOMEM;
3995 }
3996bad:
56b67a4f 3997 kfree(crypt_data);
717f4b1c
MP		 3998 kfree(crypt_iv);
3999 skcipher_request_free(req);
4000
1aa0efd4
MS		 4001 return r;
4002}
4003
7eada909 4004/*
56b67a4f 4005 * Construct a integrity mapping
7eada909
MP		 4006 *
4007 * Arguments:
4008 * device
4009 * offset from the start of the device
4010 * tag size
468dfca3 4011 * D - direct writes, J - journal writes, B - bitmap mode, R - recovery mode
7eada909
MP		 4012 * number of optional arguments
4013 * optional arguments:
56b67a4f
MP		 4014 * journal_sectors
4015 * interleave_sectors
4016 * buffer_sectors
4017 * journal_watermark
4018 * commit_time
88ad5d1e
MP		 4019 * meta_device
4020 * block_size
468dfca3
MP		 4021 * sectors_per_bit
4022 * bitmap_flush_interval
56b67a4f
MP		 4023 * internal_hash
4024 * journal_crypt
4025 * journal_mac
88ad5d1e 4026 * recalculate
7eada909 4027 */
86a3238c 4028static int dm_integrity_ctr(struct dm_target *ti, unsigned int argc, char **argv)
7eada909
MP		 4029{
4030 struct dm_integrity_c *ic;
4031 char dummy;
4032 int r;
86a3238c 4033 unsigned int extra_args;
7eada909 4034 struct dm_arg_set as;
5916a22b 4035 static const struct dm_arg _args[] = {
db7b93e3 4036 {0, 18, "Invalid number of feature args"},
7eada909 4037 };
86a3238c 4038 unsigned int journal_sectors, interleave_sectors, buffer_sectors, journal_watermark, sync_msec;
7eada909 4039 bool should_write_sb;
7eada909
MP		 4040 __u64 threshold;
4041 unsigned long long start;
468dfca3
MP		 4042 __s8 log2_sectors_per_bitmap_bit = -1;
4043 __s8 log2_blocks_per_bitmap_bit;
4044 __u64 bits_in_journal;
4045 __u64 n_bitmap_bits;
7eada909
MP		 4046
4047#define DIRECT_ARGUMENTS 4
4048
4049 if (argc <= DIRECT_ARGUMENTS) {
4050 ti->error = "Invalid argument count";
4051 return -EINVAL;
4052 }
4053
4054 ic = kzalloc(sizeof(struct dm_integrity_c), GFP_KERNEL);
4055 if (!ic) {
4056 ti->error = "Cannot allocate integrity context";
4057 return -ENOMEM;
4058 }
4059 ti->private = ic;
4060 ti->per_io_data_size = sizeof(struct dm_integrity_io);
adc0daad 4061 ic->ti = ti;
7eada909 4062
7eada909 4063 ic->in_progress = RB_ROOT;
724376a0 4064 INIT_LIST_HEAD(&ic->wait_list);
7eada909
MP		 4065 init_waitqueue_head(&ic->endio_wait);
4066 bio_list_init(&ic->flush_bio_list);
4067 init_waitqueue_head(&ic->copy_to_journal_wait);
4068 init_completion(&ic->crypto_backoff);
3f2e5393 4069 atomic64_set(&ic->number_of_mismatches, 0);
468dfca3 4070 ic->bitmap_flush_interval = BITMAP_FLUSH_INTERVAL;
7eada909
MP		 4071
4072 r = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &ic->dev);
4073 if (r) {
4074 ti->error = "Device lookup failed";
4075 goto bad;
4076 }
4077
4078 if (sscanf(argv[1], "%llu%c", &start, &dummy) != 1 || start != (sector_t)start) {
4079 ti->error = "Invalid starting offset";
4080 r = -EINVAL;
4081 goto bad;
4082 }
4083 ic->start = start;
4084
4085 if (strcmp(argv[2], "-")) {
4086 if (sscanf(argv[2], "%u%c", &ic->tag_size, &dummy) != 1 || !ic->tag_size) {
4087 ti->error = "Invalid tag size";
4088 r = -EINVAL;
4089 goto bad;
4090 }
4091 }
4092
05d6909e
MS		 4093 if (!strcmp(argv[3], "J") || !strcmp(argv[3], "B") ||
4094 !strcmp(argv[3], "D") || !strcmp(argv[3], "R")) {
7eada909 4095 ic->mode = argv[3][0];
468dfca3
MP		 4096 } else {
4097 ti->error = "Invalid mode (expecting J, B, D, R)";
7eada909
MP		 4098 r = -EINVAL;
4099 goto bad;
4100 }
4101
356d9d52 4102 journal_sectors = 0;
7eada909
MP		 4103 interleave_sectors = DEFAULT_INTERLEAVE_SECTORS;
4104 buffer_sectors = DEFAULT_BUFFER_SECTORS;
4105 journal_watermark = DEFAULT_JOURNAL_WATERMARK;
4106 sync_msec = DEFAULT_SYNC_MSEC;
9d609f85 4107 ic->sectors_per_block = 1;
7eada909
MP		 4108
4109 as.argc = argc - DIRECT_ARGUMENTS;
4110 as.argv = argv + DIRECT_ARGUMENTS;
4111 r = dm_read_arg_group(_args, &as, &extra_args, &ti->error);
4112 if (r)
4113 goto bad;
4114
4115 while (extra_args--) {
4116 const char *opt_string;
86a3238c 4117 unsigned int val;
468dfca3 4118 unsigned long long llval;
0ef0b471 4119
7eada909
MP		 4120 opt_string = dm_shift_arg(&as);
4121 if (!opt_string) {
4122 r = -EINVAL;
4123 ti->error = "Not enough feature arguments";
4124 goto bad;
4125 }
56b67a4f 4126 if (sscanf(opt_string, "journal_sectors:%u%c", &val, &dummy) == 1)
356d9d52 4127 journal_sectors = val ? val : 1;
56b67a4f 4128 else if (sscanf(opt_string, "interleave_sectors:%u%c", &val, &dummy) == 1)
7eada909 4129 interleave_sectors = val;
56b67a4f 4130 else if (sscanf(opt_string, "buffer_sectors:%u%c", &val, &dummy) == 1)
7eada909 4131 buffer_sectors = val;
56b67a4f 4132 else if (sscanf(opt_string, "journal_watermark:%u%c", &val, &dummy) == 1 && val <= 100)
7eada909 4133 journal_watermark = val;
56b67a4f 4134 else if (sscanf(opt_string, "commit_time:%u%c", &val, &dummy) == 1)
7eada909 4135 sync_msec = val;
0d74e6a3 4136 else if (!strncmp(opt_string, "meta_device:", strlen("meta_device:"))) {
356d9d52
MP		 4137 if (ic->meta_dev) {
4138 dm_put_device(ti, ic->meta_dev);
4139 ic->meta_dev = NULL;
4140 }
05d6909e
MS		 4141 r = dm_get_device(ti, strchr(opt_string, ':') + 1,
4142 dm_table_get_mode(ti->table), &ic->meta_dev);
356d9d52
MP		 4143 if (r) {
4144 ti->error = "Device lookup failed";
4145 goto bad;
4146 }
4147 } else if (sscanf(opt_string, "block_size:%u%c", &val, &dummy) == 1) {
9d609f85
MP		 4148 if (val < 1 << SECTOR_SHIFT ||
4149 val > MAX_SECTORS_PER_BLOCK << SECTOR_SHIFT ||
255e2646 4150 (val & (val - 1))) {
9d609f85
MP		 4151 r = -EINVAL;
4152 ti->error = "Invalid block_size argument";
4153 goto bad;
4154 }
4155 ic->sectors_per_block = val >> SECTOR_SHIFT;
468dfca3
MP		 4156 } else if (sscanf(opt_string, "sectors_per_bit:%llu%c", &llval, &dummy) == 1) {
4157 log2_sectors_per_bitmap_bit = !llval ? 0 : __ilog2_u64(llval);
4158 } else if (sscanf(opt_string, "bitmap_flush_interval:%u%c", &val, &dummy) == 1) {
4159 if (val >= (uint64_t)UINT_MAX * 1000 / HZ) {
4160 r = -EINVAL;
4161 ti->error = "Invalid bitmap_flush_interval argument";
17e9e134 4162 goto bad;
468dfca3
MP		 4163 }
4164 ic->bitmap_flush_interval = msecs_to_jiffies(val);
0d74e6a3 4165 } else if (!strncmp(opt_string, "internal_hash:", strlen("internal_hash:"))) {
7eada909 4166 r = get_alg_and_key(opt_string, &ic->internal_hash_alg, &ti->error,
56b67a4f 4167 "Invalid internal_hash argument");
7eada909
MP		 4168 if (r)
4169 goto bad;
0d74e6a3 4170 } else if (!strncmp(opt_string, "journal_crypt:", strlen("journal_crypt:"))) {
7eada909 4171 r = get_alg_and_key(opt_string, &ic->journal_crypt_alg, &ti->error,
56b67a4f 4172 "Invalid journal_crypt argument");
7eada909
MP		 4173 if (r)
4174 goto bad;
0d74e6a3 4175 } else if (!strncmp(opt_string, "journal_mac:", strlen("journal_mac:"))) {
09d85f8d 4176 r = get_alg_and_key(opt_string, &ic->journal_mac_alg, &ti->error,
56b67a4f 4177 "Invalid journal_mac argument");
7eada909
MP		 4178 if (r)
4179 goto bad;
a3fcf725 4180 } else if (!strcmp(opt_string, "recalculate")) {
468dfca3 4181 ic->recalculate_flag = true;
db7b93e3
MP		 4182 } else if (!strcmp(opt_string, "reset_recalculate")) {
4183 ic->recalculate_flag = true;
4184 ic->reset_recalculate_flag = true;
84597a44
MP		 4185 } else if (!strcmp(opt_string, "allow_discards")) {
4186 ic->discard = true;
d537858a
MP		 4187 } else if (!strcmp(opt_string, "fix_padding")) {
4188 ic->fix_padding = true;
09d85f8d
MP		 4189 } else if (!strcmp(opt_string, "fix_hmac")) {
4190 ic->fix_hmac = true;
5c024064
MP		 4191 } else if (!strcmp(opt_string, "legacy_recalculate")) {
4192 ic->legacy_recalculate = true;
7eada909
MP		 4193 } else {
4194 r = -EINVAL;
4195 ti->error = "Invalid argument";
4196 goto bad;
4197 }
4198 }
4199
6dcbb52c 4200 ic->data_device_sectors = bdev_nr_sectors(ic->dev->bdev);
356d9d52
MP		 4201 if (!ic->meta_dev)
4202 ic->meta_device_sectors = ic->data_device_sectors;
4203 else
6dcbb52c 4204 ic->meta_device_sectors = bdev_nr_sectors(ic->meta_dev->bdev);
356d9d52
MP		 4205
4206 if (!journal_sectors) {
4207 journal_sectors = min((sector_t)DEFAULT_MAX_JOURNAL_SECTORS,
05d6909e 4208 ic->data_device_sectors >> DEFAULT_JOURNAL_SIZE_FACTOR);
356d9d52
MP		 4209 }
4210
4211 if (!buffer_sectors)
4212 buffer_sectors = 1;
4213 ic->log2_buffer_sectors = min((int)__fls(buffer_sectors), 31 - SECTOR_SHIFT);
4214
7eada909
MP		 4215 r = get_mac(&ic->internal_hash, &ic->internal_hash_alg, &ti->error,
4216 "Invalid internal hash", "Error setting internal hash key");
4217 if (r)
4218 goto bad;
4219
4220 r = get_mac(&ic->journal_mac, &ic->journal_mac_alg, &ti->error,
4221 "Invalid journal mac", "Error setting journal mac key");
4222 if (r)
4223 goto bad;
4224
4225 if (!ic->tag_size) {
4226 if (!ic->internal_hash) {
4227 ti->error = "Unknown tag size";
4228 r = -EINVAL;
4229 goto bad;
4230 }
4231 ic->tag_size = crypto_shash_digestsize(ic->internal_hash);
4232 }
4233 if (ic->tag_size > MAX_TAG_SIZE) {
4234 ti->error = "Too big tag size";
4235 r = -EINVAL;
4236 goto bad;
4237 }
4238 if (!(ic->tag_size & (ic->tag_size - 1)))
4239 ic->log2_tag_size = __ffs(ic->tag_size);
4240 else
4241 ic->log2_tag_size = -1;
4242
468dfca3
MP		 4243 if (ic->mode == 'B' && !ic->internal_hash) {
4244 r = -EINVAL;
4245 ti->error = "Bitmap mode can be only used with internal hash";
4246 goto bad;
4247 }
4248
84597a44
MP		 4249 if (ic->discard && !ic->internal_hash) {
4250 r = -EINVAL;
4251 ti->error = "Discard can be only used with internal hash";
4252 goto bad;
4253 }
4254
7eada909
MP		 4255 ic->autocommit_jiffies = msecs_to_jiffies(sync_msec);
4256 ic->autocommit_msec = sync_msec;
8376d3c1 4257 timer_setup(&ic->autocommit_timer, autocommit_fn, 0);
7eada909
MP		 4258
4259 ic->io = dm_io_client_create();
4260 if (IS_ERR(ic->io)) {
4261 r = PTR_ERR(ic->io);
4262 ic->io = NULL;
4263 ti->error = "Cannot allocate dm io";
4264 goto bad;
4265 }
4266
6f1c819c
KO		 4267 r = mempool_init_slab_pool(&ic->journal_io_mempool, JOURNAL_IO_MEMPOOL, journal_io_cache);
4268 if (r) {
7eada909
MP		 4269 ti->error = "Cannot allocate mempool";
4270 goto bad;
4271 }
4272
4273 ic->metadata_wq = alloc_workqueue("dm-integrity-metadata",
4274 WQ_MEM_RECLAIM, METADATA_WORKQUEUE_MAX_ACTIVE);
4275 if (!ic->metadata_wq) {
4276 ti->error = "Cannot allocate workqueue";
4277 r = -ENOMEM;
4278 goto bad;
4279 }
4280
4281 /*
57bbf99c 4282 * If this workqueue weren't ordered, it would cause bio reordering
7eada909
MP		 4283 * and reduced performance.
4284 */
57bbf99c 4285 ic->wait_wq = alloc_ordered_workqueue("dm-integrity-wait", WQ_MEM_RECLAIM);
7eada909
MP		 4286 if (!ic->wait_wq) {
4287 ti->error = "Cannot allocate workqueue";
4288 r = -ENOMEM;
4289 goto bad;
4290 }
4291
53770f0e
MP		 4292 ic->offload_wq = alloc_workqueue("dm-integrity-offload", WQ_MEM_RECLAIM,
4293 METADATA_WORKQUEUE_MAX_ACTIVE);
4294 if (!ic->offload_wq) {
4295 ti->error = "Cannot allocate workqueue";
4296 r = -ENOMEM;
4297 goto bad;
4298 }
4299
7eada909
MP		 4300 ic->commit_wq = alloc_workqueue("dm-integrity-commit", WQ_MEM_RECLAIM, 1);
4301 if (!ic->commit_wq) {
4302 ti->error = "Cannot allocate workqueue";
4303 r = -ENOMEM;
4304 goto bad;
4305 }
4306 INIT_WORK(&ic->commit_work, integrity_commit);
4307
468dfca3 4308 if (ic->mode == 'J' || ic->mode == 'B') {
7eada909
MP		 4309 ic->writer_wq = alloc_workqueue("dm-integrity-writer", WQ_MEM_RECLAIM, 1);
4310 if (!ic->writer_wq) {
4311 ti->error = "Cannot allocate workqueue";
4312 r = -ENOMEM;
4313 goto bad;
4314 }
4315 INIT_WORK(&ic->writer_work, integrity_writer);
4316 }
4317
4318 ic->sb = alloc_pages_exact(SB_SECTORS << SECTOR_SHIFT, GFP_KERNEL);
4319 if (!ic->sb) {
4320 r = -ENOMEM;
4321 ti->error = "Cannot allocate superblock area";
4322 goto bad;
4323 }
4324
c9154a4c 4325 r = sync_rw_sb(ic, REQ_OP_READ);
7eada909
MP		 4326 if (r) {
4327 ti->error = "Error reading superblock";
4328 goto bad;
4329 }
c2bcb2b7
MP		 4330 should_write_sb = false;
4331 if (memcmp(ic->sb->magic, SB_MAGIC, 8)) {
4332 if (ic->mode != 'R') {
56b67a4f
MP		 4333 if (memchr_inv(ic->sb, 0, SB_SECTORS << SECTOR_SHIFT)) {
4334 r = -EINVAL;
4335 ti->error = "The device is not initialized";
4336 goto bad;
7eada909
MP		 4337 }
4338 }
4339
4340 r = initialize_superblock(ic, journal_sectors, interleave_sectors);
4341 if (r) {
4342 ti->error = "Could not initialize superblock";
4343 goto bad;
4344 }
c2bcb2b7
MP		 4345 if (ic->mode != 'R')
4346 should_write_sb = true;
7eada909
MP		 4347 }
4348
09d85f8d 4349 if (!ic->sb->version || ic->sb->version > SB_VERSION_5) {
7eada909
MP		 4350 r = -EINVAL;
4351 ti->error = "Unknown version";
4352 goto bad;
4353 }
4354 if (le16_to_cpu(ic->sb->integrity_tag_size) != ic->tag_size) {
4355 r = -EINVAL;
9d609f85
MP		 4356 ti->error = "Tag size doesn't match the information in superblock";
4357 goto bad;
4358 }
4359 if (ic->sb->log2_sectors_per_block != __ffs(ic->sectors_per_block)) {
4360 r = -EINVAL;
4361 ti->error = "Block size doesn't match the information in superblock";
7eada909
MP		 4362 goto bad;
4363 }
bc86a41e
MP		 4364 if (!le32_to_cpu(ic->sb->journal_sections)) {
4365 r = -EINVAL;
4366 ti->error = "Corrupted superblock, journal_sections is 0";
4367 goto bad;
4368 }
7eada909 4369 /* make sure that ti->max_io_len doesn't overflow */
356d9d52
MP		 4370 if (!ic->meta_dev) {
4371 if (ic->sb->log2_interleave_sectors < MIN_LOG2_INTERLEAVE_SECTORS ||
4372 ic->sb->log2_interleave_sectors > MAX_LOG2_INTERLEAVE_SECTORS) {
4373 r = -EINVAL;
4374 ti->error = "Invalid interleave_sectors in the superblock";
4375 goto bad;
4376 }
4377 } else {
4378 if (ic->sb->log2_interleave_sectors) {
4379 r = -EINVAL;
4380 ti->error = "Invalid interleave_sectors in the superblock";
4381 goto bad;
4382 }
7eada909 4383 }
1ac2c15a 4384 if (!!(ic->sb->flags & cpu_to_le32(SB_FLAG_HAVE_JOURNAL_MAC)) != !!ic->journal_mac_alg.alg_string) {
7eada909 4385 r = -EINVAL;
1ac2c15a 4386 ti->error = "Journal mac mismatch";
7eada909
MP		 4387 goto bad;
4388 }
1ac2c15a
MP		 4389
4390 get_provided_data_sectors(ic);
4391 if (!ic->provided_data_sectors) {
7eada909 4392 r = -EINVAL;
1ac2c15a 4393 ti->error = "The device is too small";
7eada909
MP		 4394 goto bad;
4395 }
356d9d52
MP		 4396
4397try_smaller_buffer:
7eada909
MP		 4398 r = calculate_device_limits(ic);
4399 if (r) {
356d9d52
MP		 4400 if (ic->meta_dev) {
4401 if (ic->log2_buffer_sectors > 3) {
4402 ic->log2_buffer_sectors--;
4403 goto try_smaller_buffer;
4404 }
4405 }
7eada909
MP		 4406 ti->error = "The device is too small";
4407 goto bad;
4408 }
468dfca3
MP		 4409
4410 if (log2_sectors_per_bitmap_bit < 0)
4411 log2_sectors_per_bitmap_bit = __fls(DEFAULT_SECTORS_PER_BITMAP_BIT);
4412 if (log2_sectors_per_bitmap_bit < ic->sb->log2_sectors_per_block)
4413 log2_sectors_per_bitmap_bit = ic->sb->log2_sectors_per_block;
4414
4415 bits_in_journal = ((__u64)ic->journal_section_sectors * ic->journal_sections) << (SECTOR_SHIFT + 3);
4416 if (bits_in_journal > UINT_MAX)
4417 bits_in_journal = UINT_MAX;
4418 while (bits_in_journal < (ic->provided_data_sectors + ((sector_t)1 << log2_sectors_per_bitmap_bit) - 1) >> log2_sectors_per_bitmap_bit)
4419 log2_sectors_per_bitmap_bit++;
4420
4421 log2_blocks_per_bitmap_bit = log2_sectors_per_bitmap_bit - ic->sb->log2_sectors_per_block;
4422 ic->log2_blocks_per_bitmap_bit = log2_blocks_per_bitmap_bit;
2d0f25cb 4423 if (should_write_sb)
468dfca3 4424 ic->sb->log2_blocks_per_bitmap_bit = log2_blocks_per_bitmap_bit;
2d0f25cb 4425
468dfca3
MP		 4426 n_bitmap_bits = ((ic->provided_data_sectors >> ic->sb->log2_sectors_per_block)
4427 + (((sector_t)1 << log2_blocks_per_bitmap_bit) - 1)) >> log2_blocks_per_bitmap_bit;
4428 ic->n_bitmap_blocks = DIV_ROUND_UP(n_bitmap_bits, BITMAP_BLOCK_SIZE * 8);
4429
356d9d52
MP		 4430 if (!ic->meta_dev)
4431 ic->log2_buffer_sectors = min(ic->log2_buffer_sectors, (__u8)__ffs(ic->metadata_run));
4432
2ad50606
OM		 4433 if (ti->len > ic->provided_data_sectors) {
4434 r = -EINVAL;
4435 ti->error = "Not enough provided sectors for requested mapping size";
4436 goto bad;
4437 }
7eada909 4438
7eada909
MP		 4439
4440 threshold = (__u64)ic->journal_entries * (100 - journal_watermark);
4441 threshold += 50;
4442 do_div(threshold, 100);
4443 ic->free_sectors_threshold = threshold;
4444
4445 DEBUG_print("initialized:\n");
4446 DEBUG_print(" integrity_tag_size %u\n", le16_to_cpu(ic->sb->integrity_tag_size));
4447 DEBUG_print(" journal_entry_size %u\n", ic->journal_entry_size);
4448 DEBUG_print(" journal_entries_per_sector %u\n", ic->journal_entries_per_sector);
4449 DEBUG_print(" journal_section_entries %u\n", ic->journal_section_entries);
4450 DEBUG_print(" journal_section_sectors %u\n", ic->journal_section_sectors);
86a3238c 4451 DEBUG_print(" journal_sections %u\n", (unsigned int)le32_to_cpu(ic->sb->journal_sections));
7eada909
MP		 4452 DEBUG_print(" journal_entries %u\n", ic->journal_entries);
4453 DEBUG_print(" log2_interleave_sectors %d\n", ic->sb->log2_interleave_sectors);
6dcbb52c 4454 DEBUG_print(" data_device_sectors 0x%llx\n", bdev_nr_sectors(ic->dev->bdev));
7eada909
MP		 4455 DEBUG_print(" initial_sectors 0x%x\n", ic->initial_sectors);
4456 DEBUG_print(" metadata_run 0x%x\n", ic->metadata_run);
4457 DEBUG_print(" log2_metadata_run %d\n", ic->log2_metadata_run);
7649194a 4458 DEBUG_print(" provided_data_sectors 0x%llx (%llu)\n", ic->provided_data_sectors, ic->provided_data_sectors);
7eada909 4459 DEBUG_print(" log2_buffer_sectors %u\n", ic->log2_buffer_sectors);
7649194a 4460 DEBUG_print(" bits_in_journal %llu\n", bits_in_journal);
7eada909 4461
468dfca3 4462 if (ic->recalculate_flag && !(ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))) {
a3fcf725
MP		 4463 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
4464 ic->sb->recalc_sector = cpu_to_le64(0);
4465 }
4466
468dfca3 4467 if (ic->internal_hash) {
e8c2566f 4468 ic->recalc_wq = alloc_workqueue("dm-integrity-recalc", WQ_MEM_RECLAIM, 1);
255e2646 4469 if (!ic->recalc_wq) {
a3fcf725
MP		 4470 ti->error = "Cannot allocate workqueue";
4471 r = -ENOMEM;
4472 goto bad;
4473 }
4474 INIT_WORK(&ic->recalc_work, integrity_recalc);
2d06dfec
MP		 4475 } else {
4476 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) {
4477 ti->error = "Recalculate can only be specified with internal_hash";
4478 r = -EINVAL;
4479 goto bad;
4480 }
a3fcf725
MP		 4481 }
4482
5c024064
MP		 4483 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING) &&
4484 le64_to_cpu(ic->sb->recalc_sector) < ic->provided_data_sectors &&
4485 dm_integrity_disable_recalculate(ic)) {
4486 ti->error = "Recalculating with HMAC is disabled for security reasons - if you really need it, use the argument \"legacy_recalculate\"";
4487 r = -EOPNOTSUPP;
4488 goto bad;
4489 }
4490
356d9d52 4491 ic->bufio = dm_bufio_client_create(ic->meta_dev ? ic->meta_dev->bdev : ic->dev->bdev,
0fcb100d 4492 1U << (SECTOR_SHIFT + ic->log2_buffer_sectors), 1, 0, NULL, NULL, 0);
7eada909
MP		 4493 if (IS_ERR(ic->bufio)) {
4494 r = PTR_ERR(ic->bufio);
4495 ti->error = "Cannot initialize dm-bufio";
4496 ic->bufio = NULL;
4497 goto bad;
4498 }
4499 dm_bufio_set_sector_offset(ic->bufio, ic->start + ic->initial_sectors);
4500
c2bcb2b7
MP		 4501 if (ic->mode != 'R') {
4502 r = create_journal(ic, &ti->error);
4503 if (r)
4504 goto bad;
468dfca3
MP		 4505
4506 }
4507
4508 if (ic->mode == 'B') {
86a3238c
HM		 4509 unsigned int i;
4510 unsigned int n_bitmap_pages = DIV_ROUND_UP(ic->n_bitmap_blocks, PAGE_SIZE / BITMAP_BLOCK_SIZE);
468dfca3
MP		 4511
4512 ic->recalc_bitmap = dm_integrity_alloc_page_list(n_bitmap_pages);
4513 if (!ic->recalc_bitmap) {
4514 r = -ENOMEM;
4515 goto bad;
4516 }
4517 ic->may_write_bitmap = dm_integrity_alloc_page_list(n_bitmap_pages);
4518 if (!ic->may_write_bitmap) {
4519 r = -ENOMEM;
4520 goto bad;
4521 }
4522 ic->bbs = kvmalloc_array(ic->n_bitmap_blocks, sizeof(struct bitmap_block_status), GFP_KERNEL);
4523 if (!ic->bbs) {
4524 r = -ENOMEM;
4525 goto bad;
4526 }
4527 INIT_DELAYED_WORK(&ic->bitmap_flush_work, bitmap_flush_work);
4528 for (i = 0; i < ic->n_bitmap_blocks; i++) {
4529 struct bitmap_block_status *bbs = &ic->bbs[i];
86a3238c 4530 unsigned int sector, pl_index, pl_offset;
468dfca3
MP		 4531
4532 INIT_WORK(&bbs->work, bitmap_block_work);
4533 bbs->ic = ic;
4534 bbs->idx = i;
4535 bio_list_init(&bbs->bio_queue);
4536 spin_lock_init(&bbs->bio_queue_lock);
4537
4538 sector = i * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT);
4539 pl_index = sector >> (PAGE_SHIFT - SECTOR_SHIFT);
4540 pl_offset = (sector << SECTOR_SHIFT) & (PAGE_SIZE - 1);
4541
4542 bbs->bitmap = lowmem_page_address(ic->journal[pl_index].page) + pl_offset;
4543 }
c2bcb2b7 4544 }
7eada909
MP		 4545
4546 if (should_write_sb) {
7eada909
MP		 4547 init_journal(ic, 0, ic->journal_sections, 0);
4548 r = dm_integrity_failed(ic);
4549 if (unlikely(r)) {
4550 ti->error = "Error initializing journal";
4551 goto bad;
4552 }
c9154a4c 4553 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
7eada909
MP		 4554 if (r) {
4555 ti->error = "Error initializing superblock";
4556 goto bad;
4557 }
4558 ic->just_formatted = true;
4559 }
4560
356d9d52
MP		 4561 if (!ic->meta_dev) {
4562 r = dm_set_target_max_io_len(ti, 1U << ic->sb->log2_interleave_sectors);
4563 if (r)
4564 goto bad;
4565 }
468dfca3 4566 if (ic->mode == 'B') {
0ef0b471
HM		 4567 unsigned int max_io_len;
4568
4569 max_io_len = ((sector_t)ic->sectors_per_block << ic->log2_blocks_per_bitmap_bit) * (BITMAP_BLOCK_SIZE * 8);
468dfca3
MP		 4570 if (!max_io_len)
4571 max_io_len = 1U << 31;
4572 DEBUG_print("max_io_len: old %u, new %u\n", ti->max_io_len, max_io_len);
4573 if (!ti->max_io_len || ti->max_io_len > max_io_len) {
4574 r = dm_set_target_max_io_len(ti, max_io_len);
4575 if (r)
4576 goto bad;
4577 }
4578 }
7eada909
MP		 4579
4580 if (!ic->internal_hash)
4581 dm_integrity_set(ti, ic);
4582
4583 ti->num_flush_bios = 1;
4584 ti->flush_supported = true;
84597a44
MP		 4585 if (ic->discard)
4586 ti->num_discard_bios = 1;
7eada909 4587
82bb8599 4588 dm_audit_log_ctr(DM_MSG_PREFIX, ti, 1);
7eada909 4589 return 0;
468dfca3 4590
7eada909 4591bad:
82bb8599 4592 dm_audit_log_ctr(DM_MSG_PREFIX, ti, 0);
7eada909
MP		 4593 dm_integrity_dtr(ti);
4594 return r;
4595}
4596
4597static void dm_integrity_dtr(struct dm_target *ti)
4598{
4599 struct dm_integrity_c *ic = ti->private;
4600
4601 BUG_ON(!RB_EMPTY_ROOT(&ic->in_progress));
724376a0 4602 BUG_ON(!list_empty(&ic->wait_list));
7eada909 4603
f50cb2cb
LM		 4604 if (ic->mode == 'B')
4605 cancel_delayed_work_sync(&ic->bitmap_flush_work);
7eada909
MP		 4606 if (ic->metadata_wq)
4607 destroy_workqueue(ic->metadata_wq);
4608 if (ic->wait_wq)
4609 destroy_workqueue(ic->wait_wq);
53770f0e
MP		 4610 if (ic->offload_wq)
4611 destroy_workqueue(ic->offload_wq);
7eada909
MP		 4612 if (ic->commit_wq)
4613 destroy_workqueue(ic->commit_wq);
4614 if (ic->writer_wq)
4615 destroy_workqueue(ic->writer_wq);
a3fcf725
MP		 4616 if (ic->recalc_wq)
4617 destroy_workqueue(ic->recalc_wq);
468dfca3 4618 kvfree(ic->bbs);
7eada909
MP		 4619 if (ic->bufio)
4620 dm_bufio_client_destroy(ic->bufio);
6f1c819c 4621 mempool_exit(&ic->journal_io_mempool);
7eada909
MP		 4622 if (ic->io)
4623 dm_io_client_destroy(ic->io);
4624 if (ic->dev)
4625 dm_put_device(ti, ic->dev);
356d9d52
MP		 4626 if (ic->meta_dev)
4627 dm_put_device(ti, ic->meta_dev);
d5027e03
MP		 4628 dm_integrity_free_page_list(ic->journal);
4629 dm_integrity_free_page_list(ic->journal_io);
4630 dm_integrity_free_page_list(ic->journal_xor);
468dfca3
MP		 4631 dm_integrity_free_page_list(ic->recalc_bitmap);
4632 dm_integrity_free_page_list(ic->may_write_bitmap);
7eada909
MP		 4633 if (ic->journal_scatterlist)
4634 dm_integrity_free_journal_scatterlist(ic, ic->journal_scatterlist);
4635 if (ic->journal_io_scatterlist)
4636 dm_integrity_free_journal_scatterlist(ic, ic->journal_io_scatterlist);
4637 if (ic->sk_requests) {
86a3238c 4638 unsigned int i;
7eada909
MP		 4639
4640 for (i = 0; i < ic->journal_sections; i++) {
0ef0b471
HM		 4641 struct skcipher_request *req;
4642
4643 req = ic->sk_requests[i];
7eada909 4644 if (req) {
453431a5 4645 kfree_sensitive(req->iv);
7eada909
MP		 4646 skcipher_request_free(req);
4647 }
4648 }
4649 kvfree(ic->sk_requests);
4650 }
4651 kvfree(ic->journal_tree);
4652 if (ic->sb)
4653 free_pages_exact(ic->sb, SB_SECTORS << SECTOR_SHIFT);
4654
4655 if (ic->internal_hash)
4656 crypto_free_shash(ic->internal_hash);
4657 free_alg(&ic->internal_hash_alg);
4658
4659 if (ic->journal_crypt)
4660 crypto_free_skcipher(ic->journal_crypt);
4661 free_alg(&ic->journal_crypt_alg);
4662
4663 if (ic->journal_mac)
4664 crypto_free_shash(ic->journal_mac);
4665 free_alg(&ic->journal_mac_alg);
4666
4667 kfree(ic);
82bb8599 4668 dm_audit_log_dtr(DM_MSG_PREFIX, ti, 1);
7eada909
MP		 4669}
4670
4671static struct target_type integrity_target = {
4672 .name = "integrity",
dbae70d4 4673 .version = {1, 10, 0},
7eada909
MP		 4674 .module = THIS_MODULE,
4675 .features = DM_TARGET_SINGLETON | DM_TARGET_INTEGRITY,
4676 .ctr = dm_integrity_ctr,
4677 .dtr = dm_integrity_dtr,
4678 .map = dm_integrity_map,
4679 .postsuspend = dm_integrity_postsuspend,
4680 .resume = dm_integrity_resume,
4681 .status = dm_integrity_status,
4682 .iterate_devices = dm_integrity_iterate_devices,
9d609f85 4683 .io_hints = dm_integrity_io_hints,
7eada909
MP		 4684};
4685
5efedc9b 4686static int __init dm_integrity_init(void)
7eada909
MP		 4687{
4688 int r;
4689
4690 journal_io_cache = kmem_cache_create("integrity_journal_io",
4691 sizeof(struct journal_io), 0, 0, NULL);
4692 if (!journal_io_cache) {
4693 DMERR("can't allocate journal io cache");
4694 return -ENOMEM;
4695 }
4696
4697 r = dm_register_target(&integrity_target);
6b79a428 4698 if (r < 0) {
6b79a428
MS		 4699 kmem_cache_destroy(journal_io_cache);
4700 return r;
4701 }
7eada909 4702
6b79a428 4703 return 0;
7eada909
MP		 4704}
4705
5efedc9b 4706static void __exit dm_integrity_exit(void)
7eada909
MP		 4707{
4708 dm_unregister_target(&integrity_target);
4709 kmem_cache_destroy(journal_io_cache);
4710}
4711
4712module_init(dm_integrity_init);
4713module_exit(dm_integrity_exit);
4714
4715MODULE_AUTHOR("Milan Broz");
4716MODULE_AUTHOR("Mikulas Patocka");
4717MODULE_DESCRIPTION(DM_NAME " target for integrity tags extension");
4718MODULE_LICENSE("GPL");
Linux 6.x block layer and io_uring tree(s)