Lines Matching refs:ic
107 #define journal_entry_tag(ic, je) ((__u8 *)&(je)->last_bytes[(ic)->sectors_per_block])
302 struct dm_integrity_c *ic;
320 struct dm_integrity_c *ic;
332 struct dm_integrity_c *ic;
376 static void dm_integrity_io_error(struct dm_integrity_c *ic, const char *msg, int err)
379 atomic64_inc(&ic->number_of_mismatches);
380 if (!cmpxchg(&ic->failed, 0, err))
384 static int dm_integrity_failed(struct dm_integrity_c *ic)
386 return READ_ONCE(ic->failed);
389 static bool dm_integrity_disable_recalculate(struct dm_integrity_c *ic)
391 if (ic->legacy_recalculate)
393 if (!(ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) ?
394 ic->internal_hash_alg.key || ic->journal_mac_alg.key :
395 ic->internal_hash_alg.key && !ic->journal_mac_alg.key)
400 static commit_id_t dm_integrity_commit_id(struct dm_integrity_c *ic, unsigned int i,
407 return ic->commit_ids[seq] ^ cpu_to_le64(((__u64)i << 32) ^ j);
410 static void get_area_and_offset(struct dm_integrity_c *ic, sector_t data_sector,
413 if (!ic->meta_dev) {
414 __u8 log2_interleave_sectors = ic->sb->log2_interleave_sectors;
423 #define sector_to_block(ic, n) \
425 BUG_ON((n) & (unsigned int)((ic)->sectors_per_block - 1)); \
426 (n) >>= (ic)->sb->log2_sectors_per_block; \
429 static __u64 get_metadata_sector_and_offset(struct dm_integrity_c *ic, sector_t area,
435 ms = area << ic->sb->log2_interleave_sectors;
436 if (likely(ic->log2_metadata_run >= 0))
437 ms += area << ic->log2_metadata_run;
439 ms += area * ic->metadata_run;
440 ms >>= ic->log2_buffer_sectors;
442 sector_to_block(ic, offset);
444 if (likely(ic->log2_tag_size >= 0)) {
445 ms += offset >> (SECTOR_SHIFT + ic->log2_buffer_sectors - ic->log2_tag_size);
446 mo = (offset << ic->log2_tag_size) & ((1U << SECTOR_SHIFT << ic->log2_buffer_sectors) - 1);
448 ms += (__u64)offset * ic->tag_size >> (SECTOR_SHIFT + ic->log2_buffer_sectors);
449 mo = (offset * ic->tag_size) & ((1U << SECTOR_SHIFT << ic->log2_buffer_sectors) - 1);
455 static sector_t get_data_sector(struct dm_integrity_c *ic, sector_t area, sector_t offset)
459 if (ic->meta_dev)
462 result = area << ic->sb->log2_interleave_sectors;
463 if (likely(ic->log2_metadata_run >= 0))
464 result += (area + 1) << ic->log2_metadata_run;
466 result += (area + 1) * ic->metadata_run;
468 result += (sector_t)ic->initial_sectors + offset;
469 result += ic->start;
474 static void wraparound_section(struct dm_integrity_c *ic, unsigned int *sec_ptr)
476 if (unlikely(*sec_ptr >= ic->journal_sections))
477 *sec_ptr -= ic->journal_sections;
480 static void sb_set_version(struct dm_integrity_c *ic)
482 if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC))
483 ic->sb->version = SB_VERSION_5;
484 else if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING))
485 ic->sb->version = SB_VERSION_4;
486 else if (ic->mode == 'B' || ic->sb->flags & cpu_to_le32(SB_FLAG_DIRTY_BITMAP))
487 ic->sb->version = SB_VERSION_3;
488 else if (ic->meta_dev || ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))
489 ic->sb->version = SB_VERSION_2;
491 ic->sb->version = SB_VERSION_1;
494 static int sb_mac(struct dm_integrity_c *ic, bool wr)
496 SHASH_DESC_ON_STACK(desc, ic->journal_mac);
498 unsigned int size = crypto_shash_digestsize(ic->journal_mac);
501 dm_integrity_io_error(ic, "digest is too long", -EINVAL);
505 desc->tfm = ic->journal_mac;
509 dm_integrity_io_error(ic, "crypto_shash_init", r);
513 r = crypto_shash_update(desc, (__u8 *)ic->sb, (1 << SECTOR_SHIFT) - size);
515 dm_integrity_io_error(ic, "crypto_shash_update", r);
520 r = crypto_shash_final(desc, (__u8 *)ic->sb + (1 << SECTOR_SHIFT) - size);
522 dm_integrity_io_error(ic, "crypto_shash_final", r);
530 dm_integrity_io_error(ic, "crypto_shash_final", r);
533 if (memcmp((__u8 *)ic->sb + (1 << SECTOR_SHIFT) - size, result, size)) {
534 dm_integrity_io_error(ic, "superblock mac", -EILSEQ);
535 dm_audit_log_target(DM_MSG_PREFIX, "mac-superblock", ic->ti, 0);
543 static int sync_rw_sb(struct dm_integrity_c *ic, blk_opf_t opf)
552 io_req.mem.ptr.addr = ic->sb;
554 io_req.client = ic->io;
555 io_loc.bdev = ic->meta_dev ? ic->meta_dev->bdev : ic->dev->bdev;
556 io_loc.sector = ic->start;
560 sb_set_version(ic);
561 if (ic->journal_mac && ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
562 r = sb_mac(ic, true);
573 if (ic->mode != 'R' && ic->journal_mac && ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
574 r = sb_mac(ic, false);
588 static bool block_bitmap_op(struct dm_integrity_c *ic, struct page_list *bitmap,
594 if (unlikely(((sector | n_sectors) & ((1 << ic->sb->log2_sectors_per_block) - 1)) != 0)) {
598 ic->sb->log2_sectors_per_block,
599 ic->log2_blocks_per_bitmap_bit,
607 bit = sector >> (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
609 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
694 static void block_bitmap_copy(struct dm_integrity_c *ic, struct page_list *dst, struct page_list *src)
696 unsigned int n_bitmap_pages = DIV_ROUND_UP(ic->n_bitmap_blocks, PAGE_SIZE / BITMAP_BLOCK_SIZE);
707 static struct bitmap_block_status *sector_to_bitmap_block(struct dm_integrity_c *ic, sector_t sector)
709 unsigned int bit = sector >> (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
712 BUG_ON(bitmap_block >= ic->n_bitmap_blocks);
713 return &ic->bbs[bitmap_block];
716 static void access_journal_check(struct dm_integrity_c *ic, unsigned int section, unsigned int offset,
720 unsigned int limit = e ? ic->journal_section_entries : ic->journal_section_sectors;
722 if (unlikely(section >= ic->journal_sections) ||
725 function, section, offset, ic->journal_sections, limit);
731 static void page_list_location(struct dm_integrity_c *ic, unsigned int section, unsigned int offset,
736 access_journal_check(ic, section, offset, false, "page_list_location");
738 sector = section * ic->journal_section_sectors + offset;
744 static struct journal_sector *access_page_list(struct dm_integrity_c *ic, struct page_list *pl,
750 page_list_location(ic, section, offset, &pl_index, &pl_offset);
760 static struct journal_sector *access_journal(struct dm_integrity_c *ic, unsigned int section, unsigned int offset)
762 return access_page_list(ic, ic->journal, section, offset, NULL);
765 static struct journal_entry *access_journal_entry(struct dm_integrity_c *ic, unsigned int section, unsigned int n)
770 access_journal_check(ic, section, n, true, "access_journal_entry");
775 js = access_journal(ic, section, rel_sector);
776 return (struct journal_entry *)((char *)js + offset * ic->journal_entry_size);
779 static struct journal_sector *access_journal_data(struct dm_integrity_c *ic, unsigned int section, unsigned int n)
781 n <<= ic->sb->log2_sectors_per_block;
785 access_journal_check(ic, section, n, false, "access_journal_data");
787 return access_journal(ic, section, n);
790 static void section_mac(struct dm_integrity_c *ic, unsigned int section, __u8 result[JOURNAL_MAC_SIZE])
792 SHASH_DESC_ON_STACK(desc, ic->journal_mac);
796 desc->tfm = ic->journal_mac;
800 dm_integrity_io_error(ic, "crypto_shash_init", r);
804 if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
807 r = crypto_shash_update(desc, (__u8 *)&ic->sb->salt, SALT_SIZE);
809 dm_integrity_io_error(ic, "crypto_shash_update", r);
816 dm_integrity_io_error(ic, "crypto_shash_update", r);
821 for (j = 0; j < ic->journal_section_entries; j++) {
822 struct journal_entry *je = access_journal_entry(ic, section, j);
826 dm_integrity_io_error(ic, "crypto_shash_update", r);
831 size = crypto_shash_digestsize(ic->journal_mac);
836 dm_integrity_io_error(ic, "crypto_shash_final", r);
844 dm_integrity_io_error(ic, "digest_size", -EINVAL);
849 dm_integrity_io_error(ic, "crypto_shash_final", r);
860 static void rw_section_mac(struct dm_integrity_c *ic, unsigned int section, bool wr)
865 if (!ic->journal_mac)
868 section_mac(ic, section, result);
871 struct journal_sector *js = access_journal(ic, section, j);
877 dm_integrity_io_error(ic, "journal mac", -EILSEQ);
878 dm_audit_log_target(DM_MSG_PREFIX, "mac-journal", ic->ti, 0);
893 static void xor_journal(struct dm_integrity_c *ic, bool encrypt, unsigned int section,
897 size_t n_bytes = (size_t)(n_sections * ic->journal_section_sectors) << SECTOR_SHIFT;
902 source_pl = ic->journal;
903 target_pl = ic->journal_io;
905 source_pl = ic->journal_io;
906 target_pl = ic->journal;
909 page_list_location(ic, section, 0, &pl_index, &pl_offset);
926 rw_section_mac(ic, section, true);
931 page_list_location(ic, section, 0, §ion_index, &dummy);
937 src_pages[1] = ic->journal_xor[pl_index].page;
957 complete(&comp->ic->crypto_backoff);
960 dm_integrity_io_error(comp->ic, "asynchronous encrypt", err);
980 wait_for_completion(&comp->ic->crypto_backoff);
981 reinit_completion(&comp->ic->crypto_backoff);
984 dm_integrity_io_error(comp->ic, "encrypt", r);
988 static void crypt_journal(struct dm_integrity_c *ic, bool encrypt, unsigned int section,
997 source_sg = ic->journal_scatterlist;
998 target_sg = ic->journal_io_scatterlist;
1000 source_sg = ic->journal_io_scatterlist;
1001 target_sg = ic->journal_scatterlist;
1010 rw_section_mac(ic, section, true);
1012 req = ic->sk_requests[section];
1013 ivsize = crypto_skcipher_ivsize(ic->journal_crypt);
1032 static void encrypt_journal(struct dm_integrity_c *ic, bool encrypt, unsigned int section,
1035 if (ic->journal_xor)
1036 return xor_journal(ic, encrypt, section, n_sections, comp);
1038 return crypt_journal(ic, encrypt, section, n_sections, comp);
1046 dm_integrity_io_error(comp->ic, "writing journal", -EIO);
1050 static void rw_journal_sectors(struct dm_integrity_c *ic, blk_opf_t opf,
1059 if (unlikely(dm_integrity_failed(ic))) {
1070 if (ic->journal_io)
1071 io_req.mem.ptr.pl = &ic->journal_io[pl_index];
1073 io_req.mem.ptr.pl = &ic->journal[pl_index];
1081 io_req.client = ic->io;
1082 io_loc.bdev = ic->meta_dev ? ic->meta_dev->bdev : ic->dev->bdev;
1083 io_loc.sector = ic->start + SB_SECTORS + sector;
1088 dm_integrity_io_error(ic, (opf & REQ_OP_MASK) == REQ_OP_READ ?
1097 static void rw_journal(struct dm_integrity_c *ic, blk_opf_t opf,
1103 sector = section * ic->journal_section_sectors;
1104 n_sectors = n_sections * ic->journal_section_sectors;
1106 rw_journal_sectors(ic, opf, sector, n_sectors, comp);
1109 static void write_journal(struct dm_integrity_c *ic, unsigned int commit_start, unsigned int commit_sections)
1116 io_comp.ic = ic;
1119 if (commit_start + commit_sections <= ic->journal_sections) {
1121 if (ic->journal_io) {
1122 crypt_comp_1.ic = ic;
1125 encrypt_journal(ic, true, commit_start, commit_sections, &crypt_comp_1);
1129 rw_section_mac(ic, commit_start + i, true);
1131 rw_journal(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, commit_start,
1137 to_end = ic->journal_sections - commit_start;
1138 if (ic->journal_io) {
1139 crypt_comp_1.ic = ic;
1142 encrypt_journal(ic, true, commit_start, to_end, &crypt_comp_1);
1144 rw_journal(ic, REQ_OP_WRITE | REQ_FUA,
1148 encrypt_journal(ic, true, 0, commit_sections - to_end, &crypt_comp_1);
1151 crypt_comp_2.ic = ic;
1154 encrypt_journal(ic, true, 0, commit_sections - to_end, &crypt_comp_2);
1156 rw_journal(ic, REQ_OP_WRITE | REQ_FUA, commit_start, to_end, &io_comp);
1161 rw_section_mac(ic, commit_start + i, true);
1162 rw_journal(ic, REQ_OP_WRITE | REQ_FUA, commit_start, to_end, &io_comp);
1164 rw_section_mac(ic, i, true);
1166 rw_journal(ic, REQ_OP_WRITE | REQ_FUA, 0, commit_sections - to_end, &io_comp);
1172 static void copy_from_journal(struct dm_integrity_c *ic, unsigned int section, unsigned int offset,
1180 BUG_ON((target | n_sectors | offset) & (unsigned int)(ic->sectors_per_block - 1));
1182 if (unlikely(dm_integrity_failed(ic))) {
1187 sector = section * ic->journal_section_sectors + JOURNAL_BLOCK_SECTORS + offset;
1194 io_req.mem.ptr.pl = &ic->journal[pl_index];
1198 io_req.client = ic->io;
1199 io_loc.bdev = ic->dev->bdev;
1216 static bool add_new_range(struct dm_integrity_c *ic, struct dm_integrity_range *new_range, bool check_waiting)
1218 struct rb_node **n = &ic->in_progress.rb_node;
1221 BUG_ON((new_range->logical_sector | new_range->n_sectors) & (unsigned int)(ic->sectors_per_block - 1));
1226 list_for_each_entry(range, &ic->wait_list, wait_entry) {
1247 rb_insert_color(&new_range->node, &ic->in_progress);
1252 static void remove_range_unlocked(struct dm_integrity_c *ic, struct dm_integrity_range *range)
1254 rb_erase(&range->node, &ic->in_progress);
1255 while (unlikely(!list_empty(&ic->wait_list))) {
1257 list_first_entry(&ic->wait_list, struct dm_integrity_range, wait_entry);
1262 if (!add_new_range(ic, last_range, false)) {
1264 list_add(&last_range->wait_entry, &ic->wait_list);
1272 static void remove_range(struct dm_integrity_c *ic, struct dm_integrity_range *range)
1276 spin_lock_irqsave(&ic->endio_wait.lock, flags);
1277 remove_range_unlocked(ic, range);
1278 spin_unlock_irqrestore(&ic->endio_wait.lock, flags);
1281 static void wait_and_add_new_range(struct dm_integrity_c *ic, struct dm_integrity_range *new_range)
1284 list_add_tail(&new_range->wait_entry, &ic->wait_list);
1288 spin_unlock_irq(&ic->endio_wait.lock);
1290 spin_lock_irq(&ic->endio_wait.lock);
1294 static void add_new_range_and_wait(struct dm_integrity_c *ic, struct dm_integrity_range *new_range)
1296 if (unlikely(!add_new_range(ic, new_range, true)))
1297 wait_and_add_new_range(ic, new_range);
1306 static void add_journal_node(struct dm_integrity_c *ic, struct journal_node *node, sector_t sector)
1314 link = &ic->journal_tree_root.rb_node;
1329 rb_insert_color(&node->node, &ic->journal_tree_root);
1332 static void remove_journal_node(struct dm_integrity_c *ic, struct journal_node *node)
1335 rb_erase(&node->node, &ic->journal_tree_root);
1341 static unsigned int find_journal_node(struct dm_integrity_c *ic, sector_t sector, sector_t *next_sector)
1343 struct rb_node *n = ic->journal_tree_root.rb_node;
1351 found = j - ic->journal_tree;
1363 static bool test_journal_node(struct dm_integrity_c *ic, unsigned int pos, sector_t sector)
1368 if (unlikely(pos >= ic->journal_entries))
1370 node = &ic->journal_tree[pos];
1384 static bool find_newer_committed_node(struct dm_integrity_c *ic, struct journal_node *node)
1401 next_section = (unsigned int)(next_node - ic->journal_tree) / ic->journal_section_entries;
1402 if (next_section >= ic->committed_section &&
1403 next_section < ic->committed_section + ic->n_committed_sections)
1405 if (next_section + ic->journal_sections < ic->committed_section + ic->n_committed_sections)
1415 static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, sector_t *metadata_block,
1421 unsigned int may_be = MAY_BE_HASH | (ic->discard ? MAY_BE_FILLER : 0);
1429 r = dm_integrity_failed(ic);
1433 data = dm_bufio_read(ic->bufio, *metadata_block, &b);
1437 to_copy = min((1U << SECTOR_SHIFT << ic->log2_buffer_sectors) - *metadata_offset, total_size);
1449 if (likely(is_power_of_2(ic->tag_size))) {
1451 if (unlikely(!ic->discard) ||
1466 if (unlikely(hash_offset == ic->tag_size)) {
1472 may_be = MAY_BE_HASH | (ic->discard ? MAY_BE_FILLER : 0);
1481 if (unlikely(*metadata_offset == 1U << SECTOR_SHIFT << ic->log2_buffer_sectors)) {
1486 if (unlikely(!is_power_of_2(ic->tag_size)))
1487 hash_offset = (hash_offset + to_copy) % ic->tag_size;
1500 struct dm_integrity_c *ic;
1509 dm_integrity_io_error(fr->ic, "flushing disk cache", -EIO);
1513 static void dm_integrity_flush_buffers(struct dm_integrity_c *ic, bool flush_data)
1518 if (!ic->meta_dev)
1526 fr.io_req.client = dm_bufio_get_dm_io_client(ic->bufio),
1527 fr.io_reg.bdev = ic->dev->bdev,
1530 fr.ic = ic;
1536 r = dm_bufio_write_dirty_buffers(ic->bufio);
1538 dm_integrity_io_error(ic, "writing tags", r);
1544 static void sleep_on_endio_wait(struct dm_integrity_c *ic)
1548 __add_wait_queue(&ic->endio_wait, &wait);
1550 spin_unlock_irq(&ic->endio_wait.lock);
1552 spin_lock_irq(&ic->endio_wait.lock);
1553 __remove_wait_queue(&ic->endio_wait, &wait);
1558 struct dm_integrity_c *ic = from_timer(ic, t, autocommit_timer);
1560 if (likely(!dm_integrity_failed(ic)))
1561 queue_work(ic->commit_wq, &ic->commit_work);
1564 static void schedule_autocommit(struct dm_integrity_c *ic)
1566 if (!timer_pending(&ic->autocommit_timer))
1567 mod_timer(&ic->autocommit_timer, jiffies + ic->autocommit_jiffies);
1570 static void submit_flush_bio(struct dm_integrity_c *ic, struct dm_integrity_io *dio)
1575 spin_lock_irqsave(&ic->endio_wait.lock, flags);
1577 bio_list_add(&ic->flush_bio_list, bio);
1578 spin_unlock_irqrestore(&ic->endio_wait.lock, flags);
1580 queue_work(ic->commit_wq, &ic->commit_work);
1583 static void do_endio(struct dm_integrity_c *ic, struct bio *bio)
1587 r = dm_integrity_failed(ic);
1590 if (unlikely(ic->synchronous_mode) && bio_op(bio) == REQ_OP_WRITE) {
1593 spin_lock_irqsave(&ic->endio_wait.lock, flags);
1594 bio_list_add(&ic->synchronous_bios, bio);
1595 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, 0);
1596 spin_unlock_irqrestore(&ic->endio_wait.lock, flags);
1602 static void do_endio_flush(struct dm_integrity_c *ic, struct dm_integrity_io *dio)
1606 if (unlikely(dio->fua) && likely(!bio->bi_status) && likely(!dm_integrity_failed(ic)))
1607 submit_flush_bio(ic, dio);
1609 do_endio(ic, bio);
1615 struct dm_integrity_c *ic = dio->ic;
1618 remove_range(ic, &dio->range);
1621 schedule_autocommit(ic);
1630 queue_work(ic->offload_wq, &dio->work);
1633 do_endio_flush(ic, dio);
1651 static void integrity_sector_checksum(struct dm_integrity_c *ic, sector_t sector,
1655 SHASH_DESC_ON_STACK(req, ic->internal_hash);
1659 req->tfm = ic->internal_hash;
1663 dm_integrity_io_error(ic, "crypto_shash_init", r);
1667 if (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) {
1668 r = crypto_shash_update(req, (__u8 *)&ic->sb->salt, SALT_SIZE);
1670 dm_integrity_io_error(ic, "crypto_shash_update", r);
1677 dm_integrity_io_error(ic, "crypto_shash_update", r);
1681 r = crypto_shash_update(req, data, ic->sectors_per_block << SECTOR_SHIFT);
1683 dm_integrity_io_error(ic, "crypto_shash_update", r);
1689 dm_integrity_io_error(ic, "crypto_shash_final", r);
1693 digest_size = crypto_shash_digestsize(ic->internal_hash);
1694 if (unlikely(digest_size < ic->tag_size))
1695 memset(result + digest_size, 0, ic->tag_size - digest_size);
1701 get_random_bytes(result, ic->tag_size);
1707 struct dm_integrity_c *ic = dio->ic;
1713 get_area_and_offset(ic, dio->range.logical_sector, &area, &offset);
1714 dio->metadata_block = get_metadata_sector_and_offset(ic, area, offset,
1716 sector = get_data_sector(ic, area, offset);
1719 page = mempool_alloc(&ic->recheck_pool, GFP_NOIO);
1735 io_req.client = ic->io;
1736 io_loc.bdev = ic->dev->bdev;
1738 io_loc.count = ic->sectors_per_block;
1754 integrity_sector_checksum(ic, logical_sector, buffer, checksum);
1755 r = dm_integrity_rw_tag(ic, checksum, &dio->metadata_block,
1756 &dio->metadata_offset, ic->tag_size, TAG_CMP);
1761 atomic64_inc(&ic->number_of_mismatches);
1771 memcpy(mem + pos, buffer, ic->sectors_per_block << SECTOR_SHIFT);
1774 pos += ic->sectors_per_block << SECTOR_SHIFT;
1775 sector += ic->sectors_per_block;
1776 logical_sector += ic->sectors_per_block;
1780 mempool_free(page, &ic->recheck_pool);
1786 struct dm_integrity_c *ic = dio->ic;
1790 if (ic->internal_hash) {
1793 unsigned int digest_size = crypto_shash_digestsize(ic->internal_hash);
1796 unsigned int extra_space = unlikely(digest_size > ic->tag_size) ? digest_size - ic->tag_size : 0;
1801 if (unlikely(ic->mode == 'R'))
1805 checksums = kmalloc((PAGE_SIZE >> SECTOR_SHIFT >> ic->sb->log2_sectors_per_block) * ic->tag_size + extra_space,
1821 unsigned int max_blocks = max_size / ic->tag_size;
1826 unsigned int this_step_blocks = bi_size >> (SECTOR_SHIFT + ic->sb->log2_sectors_per_block);
1829 r = dm_integrity_rw_tag(ic, checksums, &dio->metadata_block, &dio->metadata_offset,
1830 this_step_blocks * ic->tag_size, TAG_WRITE);
1837 bi_size -= this_step_blocks << (SECTOR_SHIFT + ic->sb->log2_sectors_per_block);
1858 integrity_sector_checksum(ic, sector, mem + pos, checksums_ptr);
1859 checksums_ptr += ic->tag_size;
1860 sectors_to_process -= ic->sectors_per_block;
1861 pos += ic->sectors_per_block << SECTOR_SHIFT;
1862 sector += ic->sectors_per_block;
1866 r = dm_integrity_rw_tag(ic, checksums, &dio->metadata_block, &dio->metadata_offset,
1898 sector_to_block(ic, data_to_process);
1899 data_to_process *= ic->tag_size;
1908 r = dm_integrity_rw_tag(ic, tag, &dio->metadata_block, &dio->metadata_offset,
1928 struct dm_integrity_c *ic = ti->private;
1934 dio->ic = ic;
1954 submit_flush_bio(ic, dio);
1967 if (unlikely(dio->range.logical_sector + bio_sectors(bio) > ic->provided_data_sectors)) {
1970 ic->provided_data_sectors);
1973 if (unlikely((dio->range.logical_sector | bio_sectors(bio)) & (unsigned int)(ic->sectors_per_block - 1))) {
1975 ic->sectors_per_block,
1980 if (ic->sectors_per_block > 1 && likely(dio->op != REQ_OP_DISCARD)) {
1985 if (unlikely(bv.bv_len & ((ic->sectors_per_block << SECTOR_SHIFT) - 1))) {
1987 bv.bv_offset, bv.bv_len, ic->sectors_per_block);
1994 if (!ic->internal_hash) {
1996 unsigned int wanted_tag_size = bio_sectors(bio) >> ic->sb->log2_sectors_per_block;
1998 if (ic->log2_tag_size >= 0)
1999 wanted_tag_size <<= ic->log2_tag_size;
2001 wanted_tag_size *= ic->tag_size;
2015 if (unlikely(ic->mode == 'R') && unlikely(dio->op != REQ_OP_READ))
2018 get_area_and_offset(ic, dio->range.logical_sector, &area, &offset);
2019 dio->metadata_block = get_metadata_sector_and_offset(ic, area, offset, &dio->metadata_offset);
2020 bio->bi_iter.bi_sector = get_data_sector(ic, area, offset);
2029 struct dm_integrity_c *ic = dio->ic;
2049 struct journal_entry *je = access_journal_entry(ic, journal_section, journal_entry);
2060 __io_wait_event(ic->copy_to_journal_wait, !journal_entry_is_inprogress(je));
2065 js = access_journal_data(ic, journal_section, journal_entry);
2073 } while (++s < ic->sectors_per_block);
2075 if (ic->internal_hash) {
2078 integrity_sector_checksum(ic, logical_sector, mem + bv.bv_offset, checksums_onstack);
2079 if (unlikely(memcmp(checksums_onstack, journal_entry_tag(ic, je), ic->tag_size))) {
2089 if (!ic->internal_hash) {
2091 unsigned int tag_todo = ic->tag_size;
2092 char *tag_ptr = journal_entry_tag(ic, je);
2118 js = access_journal_data(ic, journal_section, journal_entry);
2119 memcpy(js, mem + bv.bv_offset, ic->sectors_per_block << SECTOR_SHIFT);
2124 } while (++s < ic->sectors_per_block);
2126 if (ic->internal_hash) {
2127 unsigned int digest_size = crypto_shash_digestsize(ic->internal_hash);
2129 if (unlikely(digest_size > ic->tag_size)) {
2132 integrity_sector_checksum(ic, logical_sector, (char *)js, checksums_onstack);
2133 memcpy(journal_entry_tag(ic, je), checksums_onstack, ic->tag_size);
2135 integrity_sector_checksum(ic, logical_sector, (char *)js, journal_entry_tag(ic, je));
2140 logical_sector += ic->sectors_per_block;
2143 if (unlikely(journal_entry == ic->journal_section_entries)) {
2146 wraparound_section(ic, &journal_section);
2149 bv.bv_offset += ic->sectors_per_block << SECTOR_SHIFT;
2150 } while (bv.bv_len -= ic->sectors_per_block << SECTOR_SHIFT);
2159 if (unlikely(waitqueue_active(&ic->copy_to_journal_wait)))
2160 wake_up(&ic->copy_to_journal_wait);
2161 if (READ_ONCE(ic->free_sectors) <= ic->free_sectors_threshold)
2162 queue_work(ic->commit_wq, &ic->commit_work);
2164 schedule_autocommit(ic);
2166 remove_range(ic, &dio->range);
2172 get_area_and_offset(ic, dio->range.logical_sector, &area, &offset);
2173 dio->metadata_block = get_metadata_sector_and_offset(ic, area, offset, &dio->metadata_offset);
2182 struct dm_integrity_c *ic = dio->ic;
2188 bool need_sync_io = ic->internal_hash && dio->op == REQ_OP_READ;
2190 if (unlikely(dio->op == REQ_OP_DISCARD) && ic->mode != 'D')
2195 queue_work(ic->offload_wq, &dio->work);
2200 spin_lock_irq(&ic->endio_wait.lock);
2202 if (unlikely(dm_integrity_failed(ic))) {
2203 spin_unlock_irq(&ic->endio_wait.lock);
2204 do_endio(ic, bio);
2209 if (ic->mode == 'J' && likely(dio->op != REQ_OP_DISCARD)) {
2215 (sector_t)ic->free_sectors << ic->sb->log2_sectors_per_block);
2219 sleep_on_endio_wait(ic);
2222 range_sectors = dio->range.n_sectors >> ic->sb->log2_sectors_per_block;
2223 ic->free_sectors -= range_sectors;
2224 journal_section = ic->free_section;
2225 journal_entry = ic->free_section_entry;
2227 next_entry = ic->free_section_entry + range_sectors;
2228 ic->free_section_entry = next_entry % ic->journal_section_entries;
2229 ic->free_section += next_entry / ic->journal_section_entries;
2230 ic->n_uncommitted_sections += next_entry / ic->journal_section_entries;
2231 wraparound_section(ic, &ic->free_section);
2233 pos = journal_section * ic->journal_section_entries + journal_entry;
2240 add_journal_node(ic, &ic->journal_tree[pos], dio->range.logical_sector + i);
2242 if (unlikely(pos >= ic->journal_entries))
2245 je = access_journal_entry(ic, ws, we);
2249 if (unlikely(we == ic->journal_section_entries)) {
2252 wraparound_section(ic, &ws);
2254 } while ((i += ic->sectors_per_block) < dio->range.n_sectors);
2256 spin_unlock_irq(&ic->endio_wait.lock);
2261 journal_read_pos = find_journal_node(ic, dio->range.logical_sector, &next_sector);
2269 for (i = ic->sectors_per_block; i < dio->range.n_sectors; i += ic->sectors_per_block, jp++) {
2270 if (!test_journal_node(ic, jp, dio->range.logical_sector + i))
2277 if (unlikely(!add_new_range(ic, &dio->range, true))) {
2285 spin_unlock_irq(&ic->endio_wait.lock);
2287 queue_work(ic->wait_wq, &dio->work);
2291 dio->range.n_sectors = ic->sectors_per_block;
2292 wait_and_add_new_range(ic, &dio->range);
2302 new_pos = find_journal_node(ic, dio->range.logical_sector, &next_sector);
2304 remove_range_unlocked(ic, &dio->range);
2309 if (ic->mode == 'J' && likely(dio->op == REQ_OP_DISCARD) && !discard_retried) {
2313 new_pos = find_journal_node(ic, dio->range.logical_sector, &next_sector);
2316 remove_range_unlocked(ic, &dio->range);
2317 spin_unlock_irq(&ic->endio_wait.lock);
2318 queue_work(ic->commit_wq, &ic->commit_work);
2319 flush_workqueue(ic->commit_wq);
2320 queue_work(ic->writer_wq, &ic->writer_work);
2321 flush_workqueue(ic->writer_wq);
2326 spin_unlock_irq(&ic->endio_wait.lock);
2329 journal_section = journal_read_pos / ic->journal_section_entries;
2330 journal_entry = journal_read_pos % ic->journal_section_entries;
2334 if (ic->mode == 'B' && (dio->op == REQ_OP_WRITE || unlikely(dio->op == REQ_OP_DISCARD))) {
2335 if (!block_bitmap_op(ic, ic->may_write_bitmap, dio->range.logical_sector,
2339 bbs = sector_to_bitmap_block(ic, dio->range.logical_sector);
2343 queue_work(ic->writer_wq, &bbs->work);
2357 bio_set_dev(bio, ic->dev->bdev);
2363 if (unlikely(dio->op == REQ_OP_DISCARD) && likely(ic->mode != 'D')) {
2365 dm_integrity_flush_buffers(ic, false);
2379 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING) &&
2380 dio->range.logical_sector + dio->range.n_sectors > le64_to_cpu(ic->sb->recalc_sector))
2382 if (ic->mode == 'B') {
2383 if (!block_bitmap_op(ic, ic->recalc_bitmap, dio->range.logical_sector,
2395 queue_work(ic->metadata_wq, &dio->work);
2404 do_endio_flush(ic, dio);
2415 static void pad_uncommitted(struct dm_integrity_c *ic)
2417 if (ic->free_section_entry) {
2418 ic->free_sectors -= ic->journal_section_entries - ic->free_section_entry;
2419 ic->free_section_entry = 0;
2420 ic->free_section++;
2421 wraparound_section(ic, &ic->free_section);
2422 ic->n_uncommitted_sections++;
2424 if (WARN_ON(ic->journal_sections * ic->journal_section_entries !=
2425 (ic->n_uncommitted_sections + ic->n_committed_sections) *
2426 ic->journal_section_entries + ic->free_sectors)) {
2430 ic->journal_sections, ic->journal_section_entries,
2431 ic->n_uncommitted_sections, ic->n_committed_sections,
2432 ic->journal_section_entries, ic->free_sectors);
2438 struct dm_integrity_c *ic = container_of(w, struct dm_integrity_c, commit_work);
2443 del_timer(&ic->autocommit_timer);
2445 spin_lock_irq(&ic->endio_wait.lock);
2446 flushes = bio_list_get(&ic->flush_bio_list);
2447 if (unlikely(ic->mode != 'J')) {
2448 spin_unlock_irq(&ic->endio_wait.lock);
2449 dm_integrity_flush_buffers(ic, true);
2453 pad_uncommitted(ic);
2454 commit_start = ic->uncommitted_section;
2455 commit_sections = ic->n_uncommitted_sections;
2456 spin_unlock_irq(&ic->endio_wait.lock);
2461 ic->wrote_to_journal = true;
2465 for (j = 0; j < ic->journal_section_entries; j++) {
2468 je = access_journal_entry(ic, i, j);
2469 io_wait_event(ic->copy_to_journal_wait, !journal_entry_is_inprogress(je));
2471 for (j = 0; j < ic->journal_section_sectors; j++) {
2474 js = access_journal(ic, i, j);
2475 js->commit_id = dm_integrity_commit_id(ic, i, j, ic->commit_seq);
2478 if (unlikely(i >= ic->journal_sections))
2479 ic->commit_seq = next_commit_seq(ic->commit_seq);
2480 wraparound_section(ic, &i);
2484 write_journal(ic, commit_start, commit_sections);
2486 spin_lock_irq(&ic->endio_wait.lock);
2487 ic->uncommitted_section += commit_sections;
2488 wraparound_section(ic, &ic->uncommitted_section);
2489 ic->n_uncommitted_sections -= commit_sections;
2490 ic->n_committed_sections += commit_sections;
2491 spin_unlock_irq(&ic->endio_wait.lock);
2493 if (READ_ONCE(ic->free_sectors) <= ic->free_sectors_threshold)
2494 queue_work(ic->writer_wq, &ic->writer_work);
2501 do_endio(ic, flushes);
2510 struct dm_integrity_c *ic = comp->ic;
2512 remove_range(ic, &io->range);
2513 mempool_free(io, &ic->journal_io_mempool);
2515 dm_integrity_io_error(ic, "copying from journal", -EIO);
2519 static void restore_last_bytes(struct dm_integrity_c *ic, struct journal_sector *js,
2527 } while (++s < ic->sectors_per_block);
2530 static void do_journal_write(struct dm_integrity_c *ic, unsigned int write_start,
2539 comp.ic = ic;
2544 for (n = 0; n < write_sections; n++, i++, wraparound_section(ic, &i)) {
2548 rw_section_mac(ic, i, false);
2549 for (j = 0; j < ic->journal_section_entries; j++) {
2550 struct journal_entry *je = access_journal_entry(ic, i, j);
2562 if (unlikely(sec & (unsigned int)(ic->sectors_per_block - 1))) {
2563 dm_integrity_io_error(ic, "invalid sector in journal", -EIO);
2564 sec &= ~(sector_t)(ic->sectors_per_block - 1);
2566 if (unlikely(sec >= ic->provided_data_sectors)) {
2571 get_area_and_offset(ic, sec, &area, &offset);
2572 restore_last_bytes(ic, access_journal_data(ic, i, j), je);
2573 for (k = j + 1; k < ic->journal_section_entries; k++) {
2574 struct journal_entry *je2 = access_journal_entry(ic, i, k);
2581 if (unlikely(sec2 >= ic->provided_data_sectors))
2583 get_area_and_offset(ic, sec2, &area2, &offset2);
2584 if (area2 != area || offset2 != offset + ((k - j) << ic->sb->log2_sectors_per_block))
2586 restore_last_bytes(ic, access_journal_data(ic, i, k), je2);
2590 io = mempool_alloc(&ic->journal_io_mempool, GFP_NOIO);
2593 io->range.n_sectors = (k - j) << ic->sb->log2_sectors_per_block;
2595 spin_lock_irq(&ic->endio_wait.lock);
2596 add_new_range_and_wait(ic, &io->range);
2599 struct journal_node *section_node = &ic->journal_tree[i * ic->journal_section_entries];
2602 while (j < k && find_newer_committed_node(ic, §ion_node[j])) {
2603 struct journal_entry *je2 = access_journal_entry(ic, i, j);
2606 remove_journal_node(ic, §ion_node[j]);
2608 sec += ic->sectors_per_block;
2609 offset += ic->sectors_per_block;
2611 while (j < k && find_newer_committed_node(ic, §ion_node[k - 1])) {
2612 struct journal_entry *je2 = access_journal_entry(ic, i, k - 1);
2615 remove_journal_node(ic, §ion_node[k - 1]);
2619 remove_range_unlocked(ic, &io->range);
2620 spin_unlock_irq(&ic->endio_wait.lock);
2621 mempool_free(io, &ic->journal_io_mempool);
2625 remove_journal_node(ic, §ion_node[l]);
2627 spin_unlock_irq(&ic->endio_wait.lock);
2629 metadata_block = get_metadata_sector_and_offset(ic, area, offset, &metadata_offset);
2632 struct journal_entry *je2 = access_journal_entry(ic, i, l);
2638 ic->internal_hash) {
2641 integrity_sector_checksum(ic, sec + ((l - j) << ic->sb->log2_sectors_per_block),
2642 (char *)access_journal_data(ic, i, l), test_tag);
2643 if (unlikely(memcmp(test_tag, journal_entry_tag(ic, je2), ic->tag_size))) {
2644 dm_integrity_io_error(ic, "tag mismatch when replaying journal", -EILSEQ);
2645 dm_audit_log_target(DM_MSG_PREFIX, "integrity-replay-journal", ic->ti, 0);
2650 r = dm_integrity_rw_tag(ic, journal_entry_tag(ic, je2), &metadata_block, &metadata_offset,
2651 ic->tag_size, TAG_WRITE);
2653 dm_integrity_io_error(ic, "reading tags", r);
2657 copy_from_journal(ic, i, j << ic->sb->log2_sectors_per_block,
2658 (k - j) << ic->sb->log2_sectors_per_block,
2659 get_data_sector(ic, area, offset),
2666 dm_bufio_write_dirty_buffers_async(ic->bufio);
2673 dm_integrity_flush_buffers(ic, true);
2678 struct dm_integrity_c *ic = container_of(w, struct dm_integrity_c, writer_work);
2682 spin_lock_irq(&ic->endio_wait.lock);
2683 write_start = ic->committed_section;
2684 write_sections = ic->n_committed_sections;
2685 spin_unlock_irq(&ic->endio_wait.lock);
2690 do_journal_write(ic, write_start, write_sections, false);
2692 spin_lock_irq(&ic->endio_wait.lock);
2694 ic->committed_section += write_sections;
2695 wraparound_section(ic, &ic->committed_section);
2696 ic->n_committed_sections -= write_sections;
2698 prev_free_sectors = ic->free_sectors;
2699 ic->free_sectors += write_sections * ic->journal_section_entries;
2701 wake_up_locked(&ic->endio_wait);
2703 spin_unlock_irq(&ic->endio_wait.lock);
2706 static void recalc_write_super(struct dm_integrity_c *ic)
2710 dm_integrity_flush_buffers(ic, false);
2711 if (dm_integrity_failed(ic))
2714 r = sync_rw_sb(ic, REQ_OP_WRITE);
2716 dm_integrity_io_error(ic, "writing superblock", r);
2721 struct dm_integrity_c *ic = container_of(w, struct dm_integrity_c, recalc_work);
2743 if (recalc_sectors >= 1U << ic->sb->log2_sectors_per_block)
2748 recalc_tags_size = (recalc_sectors >> ic->sb->log2_sectors_per_block) * ic->tag_size;
2749 if (crypto_shash_digestsize(ic->internal_hash) > ic->tag_size)
2750 recalc_tags_size += crypto_shash_digestsize(ic->internal_hash) - ic->tag_size;
2758 DEBUG_print("start recalculation... (position %llx)\n", le64_to_cpu(ic->sb->recalc_sector));
2760 spin_lock_irq(&ic->endio_wait.lock);
2764 if (unlikely(dm_post_suspending(ic->ti)))
2767 range.logical_sector = le64_to_cpu(ic->sb->recalc_sector);
2768 if (unlikely(range.logical_sector >= ic->provided_data_sectors)) {
2769 if (ic->mode == 'B') {
2770 block_bitmap_op(ic, ic->recalc_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
2772 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, 0);
2777 get_area_and_offset(ic, range.logical_sector, &area, &offset);
2778 range.n_sectors = min((sector_t)recalc_sectors, ic->provided_data_sectors - range.logical_sector);
2779 if (!ic->meta_dev)
2780 range.n_sectors = min(range.n_sectors, ((sector_t)1U << ic->sb->log2_interleave_sectors) - (unsigned int)offset);
2782 add_new_range_and_wait(ic, &range);
2783 spin_unlock_irq(&ic->endio_wait.lock);
2787 if (ic->mode == 'B') {
2788 if (block_bitmap_op(ic, ic->recalc_bitmap, logical_sector, n_sectors, BITMAP_OP_TEST_ALL_CLEAR))
2791 while (block_bitmap_op(ic, ic->recalc_bitmap, logical_sector,
2792 ic->sectors_per_block, BITMAP_OP_TEST_ALL_CLEAR)) {
2793 logical_sector += ic->sectors_per_block;
2794 n_sectors -= ic->sectors_per_block;
2797 while (block_bitmap_op(ic, ic->recalc_bitmap, logical_sector + n_sectors - ic->sectors_per_block,
2798 ic->sectors_per_block, BITMAP_OP_TEST_ALL_CLEAR)) {
2799 n_sectors -= ic->sectors_per_block;
2802 get_area_and_offset(ic, logical_sector, &area, &offset);
2808 recalc_write_super(ic);
2809 if (ic->mode == 'B')
2810 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, ic->bitmap_flush_interval);
2815 if (unlikely(dm_integrity_failed(ic)))
2822 io_req.client = ic->io;
2823 io_loc.bdev = ic->dev->bdev;
2824 io_loc.sector = get_data_sector(ic, area, offset);
2829 dm_integrity_io_error(ic, "reading data", r);
2834 for (i = 0; i < n_sectors; i += ic->sectors_per_block) {
2835 integrity_sector_checksum(ic, logical_sector + i, recalc_buffer + (i << SECTOR_SHIFT), t);
2836 t += ic->tag_size;
2839 metadata_block = get_metadata_sector_and_offset(ic, area, offset, &metadata_offset);
2841 r = dm_integrity_rw_tag(ic, recalc_tags, &metadata_block, &metadata_offset, t - recalc_tags, TAG_WRITE);
2843 dm_integrity_io_error(ic, "writing tags", r);
2847 if (ic->mode == 'B') {
2851 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit)) <<
2852 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
2854 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit)) <<
2855 (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
2856 block_bitmap_op(ic, ic->recalc_bitmap, start, end - start, BITMAP_OP_CLEAR);
2862 spin_lock_irq(&ic->endio_wait.lock);
2863 remove_range_unlocked(ic, &range);
2864 ic->sb->recalc_sector = cpu_to_le64(range.logical_sector + range.n_sectors);
2868 remove_range(ic, &range);
2872 spin_unlock_irq(&ic->endio_wait.lock);
2874 recalc_write_super(ic);
2884 struct dm_integrity_c *ic = bbs->ic;
2901 if (block_bitmap_op(ic, ic->may_write_bitmap, dio->range.logical_sector,
2903 remove_range(ic, &dio->range);
2905 queue_work(ic->offload_wq, &dio->work);
2907 block_bitmap_op(ic, ic->journal, dio->range.logical_sector,
2916 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC,
2923 block_bitmap_op(ic, ic->may_write_bitmap, dio->range.logical_sector,
2926 remove_range(ic, &dio->range);
2928 queue_work(ic->offload_wq, &dio->work);
2931 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, ic->bitmap_flush_interval);
2936 struct dm_integrity_c *ic = container_of(work, struct dm_integrity_c, bitmap_flush_work.work);
2941 dm_integrity_flush_buffers(ic, false);
2944 range.n_sectors = ic->provided_data_sectors;
2946 spin_lock_irq(&ic->endio_wait.lock);
2947 add_new_range_and_wait(ic, &range);
2948 spin_unlock_irq(&ic->endio_wait.lock);
2950 dm_integrity_flush_buffers(ic, true);
2952 limit = ic->provided_data_sectors;
2953 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) {
2954 limit = le64_to_cpu(ic->sb->recalc_sector)
2955 >> (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit)
2956 << (ic->sb->log2_sectors_per_block + ic->log2_blocks_per_bitmap_bit);
2959 block_bitmap_op(ic, ic->journal, 0, limit, BITMAP_OP_CLEAR);
2960 block_bitmap_op(ic, ic->may_write_bitmap, 0, limit, BITMAP_OP_CLEAR);
2962 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
2963 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
2965 spin_lock_irq(&ic->endio_wait.lock);
2966 remove_range_unlocked(ic, &range);
2967 while (unlikely((bio = bio_list_pop(&ic->synchronous_bios)) != NULL)) {
2969 spin_unlock_irq(&ic->endio_wait.lock);
2970 spin_lock_irq(&ic->endio_wait.lock);
2972 spin_unlock_irq(&ic->endio_wait.lock);
2976 static void init_journal(struct dm_integrity_c *ic, unsigned int start_section,
2986 wraparound_section(ic, &i);
2987 for (j = 0; j < ic->journal_section_sectors; j++) {
2988 struct journal_sector *js = access_journal(ic, i, j);
2992 js->commit_id = dm_integrity_commit_id(ic, i, j, commit_seq);
2994 for (j = 0; j < ic->journal_section_entries; j++) {
2995 struct journal_entry *je = access_journal_entry(ic, i, j);
3001 write_journal(ic, start_section, n_sections);
3004 static int find_commit_seq(struct dm_integrity_c *ic, unsigned int i, unsigned int j, commit_id_t id)
3009 if (dm_integrity_commit_id(ic, i, j, k) == id)
3012 dm_integrity_io_error(ic, "journal commit id", -EIO);
3016 static void replay_journal(struct dm_integrity_c *ic)
3026 if (ic->mode == 'R')
3029 if (ic->journal_uptodate)
3035 if (!ic->just_formatted) {
3037 rw_journal(ic, REQ_OP_READ, 0, ic->journal_sections, NULL);
3038 if (ic->journal_io)
3039 DEBUG_bytes(lowmem_page_address(ic->journal_io[0].page), 64, "read journal");
3040 if (ic->journal_io) {
3043 crypt_comp.ic = ic;
3046 encrypt_journal(ic, false, 0, ic->journal_sections, &crypt_comp);
3049 DEBUG_bytes(lowmem_page_address(ic->journal[0].page), 64, "decrypted journal");
3052 if (dm_integrity_failed(ic))
3058 for (i = 0; i < ic->journal_sections; i++) {
3059 for (j = 0; j < ic->journal_section_sectors; j++) {
3061 struct journal_sector *js = access_journal(ic, i, j);
3063 k = find_commit_seq(ic, i, j, js->commit_id);
3070 for (j = 0; j < ic->journal_section_entries; j++) {
3071 struct journal_entry *je = access_journal_entry(ic, i, j);
3090 dm_integrity_io_error(ic, "journal commit ids", -EIO);
3105 if (unlikely(write_start >= ic->journal_sections))
3107 wraparound_section(ic, &write_start);
3110 for (write_sections = 0; write_sections < ic->journal_sections; write_sections++) {
3111 for (j = 0; j < ic->journal_section_sectors; j++) {
3112 struct journal_sector *js = access_journal(ic, i, j);
3114 if (js->commit_id != dm_integrity_commit_id(ic, i, j, want_commit_seq)) {
3121 i, j, find_commit_seq(ic, i, j, js->commit_id), want_commit_seq);
3126 if (unlikely(i >= ic->journal_sections))
3128 wraparound_section(ic, &i);
3135 do_journal_write(ic, write_start, write_sections, true);
3138 if (write_sections == ic->journal_sections && (ic->mode == 'J' || journal_empty)) {
3140 ic->commit_seq = want_commit_seq;
3141 DEBUG_print("continuing from section %u, commit seq %d\n", write_start, ic->commit_seq);
3151 init_journal(ic, s, 1, erase_seq);
3153 wraparound_section(ic, &s);
3154 if (ic->journal_sections >= 2) {
3155 init_journal(ic, s, ic->journal_sections - 2, erase_seq);
3156 s += ic->journal_sections - 2;
3157 wraparound_section(ic, &s);
3158 init_journal(ic, s, 1, erase_seq);
3162 ic->commit_seq = next_commit_seq(erase_seq);
3165 ic->committed_section = continue_section;
3166 ic->n_committed_sections = 0;
3168 ic->uncommitted_section = continue_section;
3169 ic->n_uncommitted_sections = 0;
3171 ic->free_section = continue_section;
3172 ic->free_section_entry = 0;
3173 ic->free_sectors = ic->journal_entries;
3175 ic->journal_tree_root = RB_ROOT;
3176 for (i = 0; i < ic->journal_entries; i++)
3177 init_journal_node(&ic->journal_tree[i]);
3180 static void dm_integrity_enter_synchronous_mode(struct dm_integrity_c *ic)
3184 if (ic->mode == 'B') {
3185 ic->bitmap_flush_interval = msecs_to_jiffies(10) + 1;
3186 ic->synchronous_mode = 1;
3188 cancel_delayed_work_sync(&ic->bitmap_flush_work);
3189 queue_delayed_work(ic->commit_wq, &ic->bitmap_flush_work, 0);
3190 flush_workqueue(ic->commit_wq);
3196 struct dm_integrity_c *ic = container_of(n, struct dm_integrity_c, reboot_notifier);
3200 dm_integrity_enter_synchronous_mode(ic);
3207 struct dm_integrity_c *ic = ti->private;
3210 WARN_ON(unregister_reboot_notifier(&ic->reboot_notifier));
3212 del_timer_sync(&ic->autocommit_timer);
3214 if (ic->recalc_wq)
3215 drain_workqueue(ic->recalc_wq);
3217 if (ic->mode == 'B')
3218 cancel_delayed_work_sync(&ic->bitmap_flush_work);
3220 queue_work(ic->commit_wq, &ic->commit_work);
3221 drain_workqueue(ic->commit_wq);
3223 if (ic->mode == 'J') {
3224 queue_work(ic->writer_wq, &ic->writer_work);
3225 drain_workqueue(ic->writer_wq);
3226 dm_integrity_flush_buffers(ic, true);
3227 if (ic->wrote_to_journal) {
3228 init_journal(ic, ic->free_section,
3229 ic->journal_sections - ic->free_section, ic->commit_seq);
3230 if (ic->free_section) {
3231 init_journal(ic, 0, ic->free_section,
3232 next_commit_seq(ic->commit_seq));
3237 if (ic->mode == 'B') {
3238 dm_integrity_flush_buffers(ic, true);
3241 init_journal(ic, 0, ic->journal_sections, 0);
3242 ic->sb->flags &= ~cpu_to_le32(SB_FLAG_DIRTY_BITMAP);
3243 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
3245 dm_integrity_io_error(ic, "writing superblock", r);
3249 BUG_ON(!RB_EMPTY_ROOT(&ic->in_progress));
3251 ic->journal_uptodate = true;
3256 struct dm_integrity_c *ic = ti->private;
3257 __u64 old_provided_data_sectors = le64_to_cpu(ic->sb->provided_data_sectors);
3262 ic->wrote_to_journal = false;
3264 if (ic->provided_data_sectors != old_provided_data_sectors) {
3265 if (ic->provided_data_sectors > old_provided_data_sectors &&
3266 ic->mode == 'B' &&
3267 ic->sb->log2_blocks_per_bitmap_bit == ic->log2_blocks_per_bitmap_bit) {
3268 rw_journal_sectors(ic, REQ_OP_READ, 0,
3269 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3270 block_bitmap_op(ic, ic->journal, old_provided_data_sectors,
3271 ic->provided_data_sectors - old_provided_data_sectors, BITMAP_OP_SET);
3272 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
3273 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3276 ic->sb->provided_data_sectors = cpu_to_le64(ic->provided_data_sectors);
3277 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
3279 dm_integrity_io_error(ic, "writing superblock", r);
3282 if (ic->sb->flags & cpu_to_le32(SB_FLAG_DIRTY_BITMAP)) {
3284 rw_journal_sectors(ic, REQ_OP_READ, 0,
3285 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3286 if (ic->mode == 'B') {
3287 if (ic->sb->log2_blocks_per_bitmap_bit == ic->log2_blocks_per_bitmap_bit &&
3288 !ic->reset_recalculate_flag) {
3289 block_bitmap_copy(ic, ic->recalc_bitmap, ic->journal);
3290 block_bitmap_copy(ic, ic->may_write_bitmap, ic->journal);
3291 if (!block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors,
3293 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3294 ic->sb->recalc_sector = cpu_to_le64(0);
3298 ic->sb->log2_blocks_per_bitmap_bit, ic->log2_blocks_per_bitmap_bit);
3299 ic->sb->log2_blocks_per_bitmap_bit = ic->log2_blocks_per_bitmap_bit;
3300 block_bitmap_op(ic, ic->recalc_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_SET);
3301 block_bitmap_op(ic, ic->may_write_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_SET);
3302 block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors, BITMAP_OP_SET);
3303 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
3304 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3305 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3306 ic->sb->recalc_sector = cpu_to_le64(0);
3309 if (!(ic->sb->log2_blocks_per_bitmap_bit == ic->log2_blocks_per_bitmap_bit &&
3310 block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors, BITMAP_OP_TEST_ALL_CLEAR)) ||
3311 ic->reset_recalculate_flag) {
3312 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3313 ic->sb->recalc_sector = cpu_to_le64(0);
3315 init_journal(ic, 0, ic->journal_sections, 0);
3316 replay_journal(ic);
3317 ic->sb->flags &= ~cpu_to_le32(SB_FLAG_DIRTY_BITMAP);
3319 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
3321 dm_integrity_io_error(ic, "writing superblock", r);
3323 replay_journal(ic);
3324 if (ic->reset_recalculate_flag) {
3325 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
3326 ic->sb->recalc_sector = cpu_to_le64(0);
3328 if (ic->mode == 'B') {
3329 ic->sb->flags |= cpu_to_le32(SB_FLAG_DIRTY_BITMAP);
3330 ic->sb->log2_blocks_per_bitmap_bit = ic->log2_blocks_per_bitmap_bit;
3331 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
3333 dm_integrity_io_error(ic, "writing superblock", r);
3335 block_bitmap_op(ic, ic->journal, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
3336 block_bitmap_op(ic, ic->recalc_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
3337 block_bitmap_op(ic, ic->may_write_bitmap, 0, ic->provided_data_sectors, BITMAP_OP_CLEAR);
3338 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING) &&
3339 le64_to_cpu(ic->sb->recalc_sector) < ic->provided_data_sectors) {
3340 block_bitmap_op(ic, ic->journal, le64_to_cpu(ic->sb->recalc_sector),
3341 ic->provided_data_sectors - le64_to_cpu(ic->sb->recalc_sector), BITMAP_OP_SET);
3342 block_bitmap_op(ic, ic->recalc_bitmap, le64_to_cpu(ic->sb->recalc_sector),
3343 ic->provided_data_sectors - le64_to_cpu(ic->sb->recalc_sector), BITMAP_OP_SET);
3344 block_bitmap_op(ic, ic->may_write_bitmap, le64_to_cpu(ic->sb->recalc_sector),
3345 ic->provided_data_sectors - le64_to_cpu(ic->sb->recalc_sector), BITMAP_OP_SET);
3347 rw_journal_sectors(ic, REQ_OP_WRITE | REQ_FUA | REQ_SYNC, 0,
3348 ic->n_bitmap_blocks * (BITMAP_BLOCK_SIZE >> SECTOR_SHIFT), NULL);
3352 DEBUG_print("testing recalc: %x\n", ic->sb->flags);
3353 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) {
3354 __u64 recalc_pos = le64_to_cpu(ic->sb->recalc_sector);
3356 DEBUG_print("recalc pos: %llx / %llx\n", recalc_pos, ic->provided_data_sectors);
3357 if (recalc_pos < ic->provided_data_sectors) {
3358 queue_work(ic->recalc_wq, &ic->recalc_work);
3359 } else if (recalc_pos > ic->provided_data_sectors) {
3360 ic->sb->recalc_sector = cpu_to_le64(ic->provided_data_sectors);
3361 recalc_write_super(ic);
3365 ic->reboot_notifier.notifier_call = dm_integrity_reboot;
3366 ic->reboot_notifier.next = NULL;
3367 ic->reboot_notifier.priority = INT_MAX - 1; /* be notified after md and before hardware drivers */
3368 WARN_ON(register_reboot_notifier(&ic->reboot_notifier));
3372 dm_integrity_enter_synchronous_mode(ic);
3379 struct dm_integrity_c *ic = ti->private;
3386 (unsigned long long)atomic64_read(&ic->number_of_mismatches),
3387 ic->provided_data_sectors);
3388 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))
3389 DMEMIT(" %llu", le64_to_cpu(ic->sb->recalc_sector));
3395 __u64 watermark_percentage = (__u64)(ic->journal_entries - ic->free_sectors_threshold) * 100;
3397 watermark_percentage += ic->journal_entries / 2;
3398 do_div(watermark_percentage, ic->journal_entries);
3400 arg_count += !!ic->meta_dev;
3401 arg_count += ic->sectors_per_block != 1;
3402 arg_count += !!(ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING));
3403 arg_count += ic->reset_recalculate_flag;
3404 arg_count += ic->discard;
3405 arg_count += ic->mode == 'J';
3406 arg_count += ic->mode == 'J';
3407 arg_count += ic->mode == 'B';
3408 arg_count += ic->mode == 'B';
3409 arg_count += !!ic->internal_hash_alg.alg_string;
3410 arg_count += !!ic->journal_crypt_alg.alg_string;
3411 arg_count += !!ic->journal_mac_alg.alg_string;
3412 arg_count += (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING)) != 0;
3413 arg_count += (ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) != 0;
3414 arg_count += ic->legacy_recalculate;
3415 DMEMIT("%s %llu %u %c %u", ic->dev->name, ic->start,
3416 ic->tag_size, ic->mode, arg_count);
3417 if (ic->meta_dev)
3418 DMEMIT(" meta_device:%s", ic->meta_dev->name);
3419 if (ic->sectors_per_block != 1)
3420 DMEMIT(" block_size:%u", ic->sectors_per_block << SECTOR_SHIFT);
3421 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))
3423 if (ic->reset_recalculate_flag)
3425 if (ic->discard)
3427 DMEMIT(" journal_sectors:%u", ic->initial_sectors - SB_SECTORS);
3428 DMEMIT(" interleave_sectors:%u", 1U << ic->sb->log2_interleave_sectors);
3429 DMEMIT(" buffer_sectors:%u", 1U << ic->log2_buffer_sectors);
3430 if (ic->mode == 'J') {
3432 DMEMIT(" commit_time:%u", ic->autocommit_msec);
3434 if (ic->mode == 'B') {
3435 DMEMIT(" sectors_per_bit:%llu", (sector_t)ic->sectors_per_block << ic->log2_blocks_per_bitmap_bit);
3436 DMEMIT(" bitmap_flush_interval:%u", jiffies_to_msecs(ic->bitmap_flush_interval));
3438 if ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING)) != 0)
3440 if ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) != 0)
3442 if (ic->legacy_recalculate)
3447 if (ic->a.alg_string) { \
3448 DMEMIT(" %s:%s", n, ic->a.alg_string); \
3449 if (ic->a.key_string) \
3450 DMEMIT(":%s", ic->a.key_string);\
3461 ic->dev->name, ic->start, ic->tag_size, ic->mode);
3463 if (ic->meta_dev)
3464 DMEMIT(",meta_device=%s", ic->meta_dev->name);
3465 if (ic->sectors_per_block != 1)
3466 DMEMIT(",block_size=%u", ic->sectors_per_block << SECTOR_SHIFT);
3468 DMEMIT(",recalculate=%c", (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) ?
3470 DMEMIT(",allow_discards=%c", ic->discard ? 'y' : 'n');
3472 ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING)) != 0) ? 'y' : 'n');
3474 ((ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_HMAC)) != 0) ? 'y' : 'n');
3475 DMEMIT(",legacy_recalculate=%c", ic->legacy_recalculate ? 'y' : 'n');
3477 DMEMIT(",journal_sectors=%u", ic->initial_sectors - SB_SECTORS);
3478 DMEMIT(",interleave_sectors=%u", 1U << ic->sb->log2_interleave_sectors);
3479 DMEMIT(",buffer_sectors=%u", 1U << ic->log2_buffer_sectors);
3488 struct dm_integrity_c *ic = ti->private;
3490 if (!ic->meta_dev)
3491 return fn(ti, ic->dev, ic->start + ic->initial_sectors + ic->metadata_run, ti->len, data);
3493 return fn(ti, ic->dev, 0, ti->len, data);
3498 struct dm_integrity_c *ic = ti->private;
3500 if (ic->sectors_per_block > 1) {
3501 limits->logical_block_size = ic->sectors_per_block << SECTOR_SHIFT;
3502 limits->physical_block_size = ic->sectors_per_block << SECTOR_SHIFT;
3503 blk_limits_io_min(limits, ic->sectors_per_block << SECTOR_SHIFT);
3508 static void calculate_journal_section_size(struct dm_integrity_c *ic)
3512 ic->journal_sections = le32_to_cpu(ic->sb->journal_sections);
3513 ic->journal_entry_size = roundup(offsetof(struct journal_entry, last_bytes[ic->sectors_per_block]) + ic->tag_size,
3516 if (ic->sb->flags & cpu_to_le32(SB_FLAG_HAVE_JOURNAL_MAC))
3518 ic->journal_entries_per_sector = sector_space / ic->journal_entry_size;
3519 ic->journal_section_entries = ic->journal_entries_per_sector * JOURNAL_BLOCK_SECTORS;
3520 ic->journal_section_sectors = (ic->journal_section_entries << ic->sb->log2_sectors_per_block) + JOURNAL_BLOCK_SECTORS;
3521 ic->journal_entries = ic->journal_section_entries * ic->journal_sections;
3524 static int calculate_device_limits(struct dm_integrity_c *ic)
3528 calculate_journal_section_size(ic);
3529 initial_sectors = SB_SECTORS + (__u64)ic->journal_section_sectors * ic->journal_sections;
3530 if (initial_sectors + METADATA_PADDING_SECTORS >= ic->meta_device_sectors || initial_sectors > UINT_MAX)
3532 ic->initial_sectors = initial_sectors;
3534 if (!ic->meta_dev) {
3539 ic->sb->flags & cpu_to_le32(SB_FLAG_FIXED_PADDING) ?
3543 ic->metadata_run = round_up((__u64)ic->tag_size << (ic->sb->log2_interleave_sectors - ic->sb->log2_sectors_per_block),
3545 if (!(ic->metadata_run & (ic->metadata_run - 1)))
3546 ic->log2_metadata_run = __ffs(ic->metadata_run);
3548 ic->log2_metadata_run = -1;
3550 get_area_and_offset(ic, ic->provided_data_sectors - 1, &last_area, &last_offset);
3551 last_sector = get_data_sector(ic, last_area, last_offset);
3552 if (last_sector < ic->start || last_sector >= ic->meta_device_sectors)
3555 __u64 meta_size = (ic->provided_data_sectors >> ic->sb->log2_sectors_per_block) * ic->tag_size;
3557 meta_size = (meta_size + ((1U << (ic->log2_buffer_sectors + SECTOR_SHIFT)) - 1))
3558 >> (ic->log2_buffer_sectors + SECTOR_SHIFT);
3559 meta_size <<= ic->log2_buffer_sectors;
3560 if (ic->initial_sectors + meta_size < ic->initial_sectors ||
3561 ic->initial_sectors + meta_size > ic->meta_device_sectors)
3563 ic->metadata_run = 1;
3564 ic->log2_metadata_run = 0;
3570 static void get_provided_data_sectors(struct dm_integrity_c *ic)
3572 if (!ic->meta_dev) {
3575 ic->provided_data_sectors = 0;
3576 for (test_bit = fls64(ic->meta_device_sectors) - 1; test_bit >= 3; test_bit--) {
3577 __u64 prev_data_sectors = ic->provided_data_sectors;
3579 ic->provided_data_sectors |= (sector_t)1 << test_bit;
3580 if (calculate_device_limits(ic))
3581 ic->provided_data_sectors = prev_data_sectors;
3584 ic->provided_data_sectors = ic->data_device_sectors;
3585 ic->provided_data_sectors &= ~(sector_t)(ic->sectors_per_block - 1);
3589 static int initialize_superblock(struct dm_integrity_c *ic,
3595 memset(ic->sb, 0, SB_SECTORS << SECTOR_SHIFT);
3596 memcpy(ic->sb->magic, SB_MAGIC, 8);
3597 ic->sb->integrity_tag_size = cpu_to_le16(ic->tag_size);
3598 ic->sb->log2_sectors_per_block = __ffs(ic->sectors_per_block);
3599 if (ic->journal_mac_alg.alg_string)
3600 ic->sb->flags |= cpu_to_le32(SB_FLAG_HAVE_JOURNAL_MAC);
3602 calculate_journal_section_size(ic);
3603 journal_sections = journal_sectors / ic->journal_section_sectors;
3607 if (ic->fix_hmac && (ic->internal_hash_alg.alg_string || ic->journal_mac_alg.alg_string)) {
3608 ic->sb->flags |= cpu_to_le32(SB_FLAG_FIXED_HMAC);
3609 get_random_bytes(ic->sb->salt, SALT_SIZE);
3612 if (!ic->meta_dev) {
3613 if (ic->fix_padding)
3614 ic->sb->flags |= cpu_to_le32(SB_FLAG_FIXED_PADDING);
3615 ic->sb->journal_sections = cpu_to_le32(journal_sections);
3618 ic->sb->log2_interleave_sectors = __fls(interleave_sectors);
3619 ic->sb->log2_interleave_sectors = max_t(__u8, MIN_LOG2_INTERLEAVE_SECTORS, ic->sb->log2_interleave_sectors);
3620 ic->sb->log2_interleave_sectors = min_t(__u8, MAX_LOG2_INTERLEAVE_SECTORS, ic->sb->log2_interleave_sectors);
3622 get_provided_data_sectors(ic);
3623 if (!ic->provided_data_sectors)
3626 ic->sb->log2_interleave_sectors = 0;
3628 get_provided_data_sectors(ic);
3629 if (!ic->provided_data_sectors)
3633 ic->sb->journal_sections = cpu_to_le32(0);
3635 __u32 prev_journal_sections = le32_to_cpu(ic->sb->journal_sections);
3640 ic->sb->journal_sections = cpu_to_le32(test_journal_sections);
3641 if (calculate_device_limits(ic))
3642 ic->sb->journal_sections = cpu_to_le32(prev_journal_sections);
3645 if (!le32_to_cpu(ic->sb->journal_sections)) {
3646 if (ic->log2_buffer_sectors > 3) {
3647 ic->log2_buffer_sectors--;
3654 ic->sb->provided_data_sectors = cpu_to_le64(ic->provided_data_sectors);
3656 sb_set_version(ic);
3661 static void dm_integrity_set(struct dm_target *ti, struct dm_integrity_c *ic)
3668 bi.tuple_size = ic->tag_size;
3670 bi.interval_exp = ic->sb->log2_sectors_per_block + SECTOR_SHIFT;
3711 static void dm_integrity_free_journal_scatterlist(struct dm_integrity_c *ic, struct scatterlist **sl)
3715 for (i = 0; i < ic->journal_sections; i++)
3720 static struct scatterlist **dm_integrity_alloc_journal_scatterlist(struct dm_integrity_c *ic,
3726 sl = kvmalloc_array(ic->journal_sections,
3732 for (i = 0; i < ic->journal_sections; i++) {
3739 page_list_location(ic, i, 0, &start_index, &start_offset);
3740 page_list_location(ic, i, ic->journal_section_sectors - 1,
3748 dm_integrity_free_journal_scatterlist(ic, sl);
3840 static int create_journal(struct dm_integrity_c *ic, char **error)
3848 ic->commit_ids[0] = cpu_to_le64(0x1111111111111111ULL);
3849 ic->commit_ids[1] = cpu_to_le64(0x2222222222222222ULL);
3850 ic->commit_ids[2] = cpu_to_le64(0x3333333333333333ULL);
3851 ic->commit_ids[3] = cpu_to_le64(0x4444444444444444ULL);
3853 journal_pages = roundup((__u64)ic->journal_sections * ic->journal_section_sectors,
3861 ic->journal_pages = journal_pages;
3863 ic->journal = dm_integrity_alloc_page_list(ic->journal_pages);
3864 if (!ic->journal) {
3869 if (ic->journal_crypt_alg.alg_string) {
3873 comp.ic = ic;
3874 ic->journal_crypt = crypto_alloc_skcipher(ic->journal_crypt_alg.alg_string, 0, CRYPTO_ALG_ALLOCATES_MEMORY);
3875 if (IS_ERR(ic->journal_crypt)) {
3877 r = PTR_ERR(ic->journal_crypt);
3878 ic->journal_crypt = NULL;
3881 ivsize = crypto_skcipher_ivsize(ic->journal_crypt);
3882 blocksize = crypto_skcipher_blocksize(ic->journal_crypt);
3884 if (ic->journal_crypt_alg.key) {
3885 r = crypto_skcipher_setkey(ic->journal_crypt, ic->journal_crypt_alg.key,
3886 ic->journal_crypt_alg.key_size);
3893 ic->journal_crypt_alg.alg_string, blocksize, ivsize);
3895 ic->journal_io = dm_integrity_alloc_page_list(ic->journal_pages);
3896 if (!ic->journal_io) {
3905 req = skcipher_request_alloc(ic->journal_crypt, GFP_KERNEL);
3919 ic->journal_xor = dm_integrity_alloc_page_list(ic->journal_pages);
3920 if (!ic->journal_xor) {
3926 sg = kvmalloc_array(ic->journal_pages + 1,
3934 sg_init_table(sg, ic->journal_pages + 1);
3935 for (i = 0; i < ic->journal_pages; i++) {
3936 char *va = lowmem_page_address(ic->journal_xor[i].page);
3941 sg_set_buf(&sg[i], &ic->commit_ids, sizeof(ic->commit_ids));
3944 PAGE_SIZE * ic->journal_pages + sizeof(ic->commit_ids), crypt_iv);
3950 r = dm_integrity_failed(ic);
3955 DEBUG_bytes(lowmem_page_address(ic->journal_xor[0].page), 64, "xor data");
3957 crypto_free_skcipher(ic->journal_crypt);
3958 ic->journal_crypt = NULL;
3962 req = skcipher_request_alloc(ic->journal_crypt, GFP_KERNEL);
3983 ic->journal_scatterlist = dm_integrity_alloc_journal_scatterlist(ic, ic->journal);
3984 if (!ic->journal_scatterlist) {
3989 ic->journal_io_scatterlist = dm_integrity_alloc_journal_scatterlist(ic, ic->journal_io);
3990 if (!ic->journal_io_scatterlist) {
3995 ic->sk_requests = kvmalloc_array(ic->journal_sections,
3998 if (!ic->sk_requests) {
4003 for (i = 0; i < ic->journal_sections; i++) {
4019 r = dm_integrity_failed(ic);
4025 section_req = skcipher_request_alloc(ic->journal_crypt, GFP_KERNEL);
4040 section_req->cryptlen = (size_t)ic->journal_section_sectors << SECTOR_SHIFT;
4041 ic->sk_requests[i] = section_req;
4052 if (ic->commit_ids[j] == ic->commit_ids[i]) {
4053 ic->commit_ids[i] = cpu_to_le64(le64_to_cpu(ic->commit_ids[i]) + 1);
4057 DEBUG_print("commit id %u: %016llx\n", i, ic->commit_ids[i]);
4060 journal_tree_size = (__u64)ic->journal_entries * sizeof(struct journal_node);
4066 ic->journal_tree = kvmalloc(journal_tree_size, GFP_KERNEL);
4067 if (!ic->journal_tree) {
4105 struct dm_integrity_c *ic;
4129 ic = kzalloc(sizeof(struct dm_integrity_c), GFP_KERNEL);
4130 if (!ic) {
4134 ti->private = ic;
4136 ic->ti = ti;
4138 ic->in_progress = RB_ROOT;
4139 INIT_LIST_HEAD(&ic->wait_list);
4140 init_waitqueue_head(&ic->endio_wait);
4141 bio_list_init(&ic->flush_bio_list);
4142 init_waitqueue_head(&ic->copy_to_journal_wait);
4143 init_completion(&ic->crypto_backoff);
4144 atomic64_set(&ic->number_of_mismatches, 0);
4145 ic->bitmap_flush_interval = BITMAP_FLUSH_INTERVAL;
4147 r = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &ic->dev);
4158 ic->start = start;
4161 if (sscanf(argv[2], "%u%c", &ic->tag_size, &dummy) != 1 || !ic->tag_size) {
4170 ic->mode = argv[3][0];
4182 ic->sectors_per_block = 1;
4212 if (ic->meta_dev) {
4213 dm_put_device(ti, ic->meta_dev);
4214 ic->meta_dev = NULL;
4217 dm_table_get_mode(ti->table), &ic->meta_dev);
4230 ic->sectors_per_block = val >> SECTOR_SHIFT;
4239 ic->bitmap_flush_interval = msecs_to_jiffies(val);
4241 r = get_alg_and_key(opt_string, &ic->internal_hash_alg, &ti->error,
4246 r = get_alg_and_key(opt_string, &ic->journal_crypt_alg, &ti->error,
4251 r = get_alg_and_key(opt_string, &ic->journal_mac_alg, &ti->error,
4256 ic->recalculate_flag = true;
4258 ic->recalculate_flag = true;
4259 ic->reset_recalculate_flag = true;
4261 ic->discard = true;
4263 ic->fix_padding = true;
4265 ic->fix_hmac = true;
4267 ic->legacy_recalculate = true;
4275 ic->data_device_sectors = bdev_nr_sectors(ic->dev->bdev);
4276 if (!ic->meta_dev)
4277 ic->meta_device_sectors = ic->data_device_sectors;
4279 ic->meta_device_sectors = bdev_nr_sectors(ic->meta_dev->bdev);
4283 ic->data_device_sectors >> DEFAULT_JOURNAL_SIZE_FACTOR);
4288 ic->log2_buffer_sectors = min((int)__fls(buffer_sectors), 31 - SECTOR_SHIFT);
4290 r = get_mac(&ic->internal_hash, &ic->internal_hash_alg, &ti->error,
4295 r = get_mac(&ic->journal_mac, &ic->journal_mac_alg, &ti->error,
4300 if (!ic->tag_size) {
4301 if (!ic->internal_hash) {
4306 ic->tag_size = crypto_shash_digestsize(ic->internal_hash);
4308 if (ic->tag_size > MAX_TAG_SIZE) {
4313 if (!(ic->tag_size & (ic->tag_size - 1)))
4314 ic->log2_tag_size = __ffs(ic->tag_size);
4316 ic->log2_tag_size = -1;
4318 if (ic->mode == 'B' && !ic->internal_hash) {
4324 if (ic->discard && !ic->internal_hash) {
4330 ic->autocommit_jiffies = msecs_to_jiffies(sync_msec);
4331 ic->autocommit_msec = sync_msec;
4332 timer_setup(&ic->autocommit_timer, autocommit_fn, 0);
4334 ic->io = dm_io_client_create();
4335 if (IS_ERR(ic->io)) {
4336 r = PTR_ERR(ic->io);
4337 ic->io = NULL;
4342 r = mempool_init_slab_pool(&ic->journal_io_mempool, JOURNAL_IO_MEMPOOL, journal_io_cache);
4348 r = mempool_init_page_pool(&ic->recheck_pool, 1, 0);
4354 ic->metadata_wq = alloc_workqueue("dm-integrity-metadata",
4356 if (!ic->metadata_wq) {
4366 ic->wait_wq = alloc_ordered_workqueue("dm-integrity-wait", WQ_MEM_RECLAIM);
4367 if (!ic->wait_wq) {
4373 ic->offload_wq = alloc_workqueue("dm-integrity-offload", WQ_MEM_RECLAIM,
4375 if (!ic->offload_wq) {
4381 ic->commit_wq = alloc_workqueue("dm-integrity-commit", WQ_MEM_RECLAIM, 1);
4382 if (!ic->commit_wq) {
4387 INIT_WORK(&ic->commit_work, integrity_commit);
4389 if (ic->mode == 'J' || ic->mode == 'B') {
4390 ic->writer_wq = alloc_workqueue("dm-integrity-writer", WQ_MEM_RECLAIM, 1);
4391 if (!ic->writer_wq) {
4396 INIT_WORK(&ic->writer_work, integrity_writer);
4399 ic->sb = alloc_pages_exact(SB_SECTORS << SECTOR_SHIFT, GFP_KERNEL);
4400 if (!ic->sb) {
4406 r = sync_rw_sb(ic, REQ_OP_READ);
4412 if (memcmp(ic->sb->magic, SB_MAGIC, 8)) {
4413 if (ic->mode != 'R') {
4414 if (memchr_inv(ic->sb, 0, SB_SECTORS << SECTOR_SHIFT)) {
4421 r = initialize_superblock(ic, journal_sectors, interleave_sectors);
4426 if (ic->mode != 'R')
4430 if (!ic->sb->version || ic->sb->version > SB_VERSION_5) {
4435 if (le16_to_cpu(ic->sb->integrity_tag_size) != ic->tag_size) {
4440 if (ic->sb->log2_sectors_per_block != __ffs(ic->sectors_per_block)) {
4445 if (!le32_to_cpu(ic->sb->journal_sections)) {
4451 if (!ic->meta_dev) {
4452 if (ic->sb->log2_interleave_sectors < MIN_LOG2_INTERLEAVE_SECTORS ||
4453 ic->sb->log2_interleave_sectors > MAX_LOG2_INTERLEAVE_SECTORS) {
4459 if (ic->sb->log2_interleave_sectors) {
4465 if (!!(ic->sb->flags & cpu_to_le32(SB_FLAG_HAVE_JOURNAL_MAC)) != !!ic->journal_mac_alg.alg_string) {
4471 get_provided_data_sectors(ic);
4472 if (!ic->provided_data_sectors) {
4479 r = calculate_device_limits(ic);
4481 if (ic->meta_dev) {
4482 if (ic->log2_buffer_sectors > 3) {
4483 ic->log2_buffer_sectors--;
4493 if (log2_sectors_per_bitmap_bit < ic->sb->log2_sectors_per_block)
4494 log2_sectors_per_bitmap_bit = ic->sb->log2_sectors_per_block;
4496 bits_in_journal = ((__u64)ic->journal_section_sectors * ic->journal_sections) << (SECTOR_SHIFT + 3);
4499 while (bits_in_journal < (ic->provided_data_sectors + ((sector_t)1 << log2_sectors_per_bitmap_bit) - 1) >> log2_sectors_per_bitmap_bit)
4502 log2_blocks_per_bitmap_bit = log2_sectors_per_bitmap_bit - ic->sb->log2_sectors_per_block;
4503 ic->log2_blocks_per_bitmap_bit = log2_blocks_per_bitmap_bit;
4505 ic->sb->log2_blocks_per_bitmap_bit = log2_blocks_per_bitmap_bit;
4507 n_bitmap_bits = ((ic->provided_data_sectors >> ic->sb->log2_sectors_per_block)
4509 ic->n_bitmap_blocks = DIV_ROUND_UP(n_bitmap_bits, BITMAP_BLOCK_SIZE * 8);
4511 if (!ic->meta_dev)
4512 ic->log2_buffer_sectors = min(ic->log2_buffer_sectors, (__u8)__ffs(ic->metadata_run));
4514 if (ti->len > ic->provided_data_sectors) {
4521 threshold = (__u64)ic->journal_entries * (100 - journal_watermark);
4524 ic->free_sectors_threshold = threshold;
4527 DEBUG_print(" integrity_tag_size %u\n", le16_to_cpu(ic->sb->integrity_tag_size));
4528 DEBUG_print(" journal_entry_size %u\n", ic->journal_entry_size);
4529 DEBUG_print(" journal_entries_per_sector %u\n", ic->journal_entries_per_sector);
4530 DEBUG_print(" journal_section_entries %u\n", ic->journal_section_entries);
4531 DEBUG_print(" journal_section_sectors %u\n", ic->journal_section_sectors);
4532 DEBUG_print(" journal_sections %u\n", (unsigned int)le32_to_cpu(ic->sb->journal_sections));
4533 DEBUG_print(" journal_entries %u\n", ic->journal_entries);
4534 DEBUG_print(" log2_interleave_sectors %d\n", ic->sb->log2_interleave_sectors);
4535 DEBUG_print(" data_device_sectors 0x%llx\n", bdev_nr_sectors(ic->dev->bdev));
4536 DEBUG_print(" initial_sectors 0x%x\n", ic->initial_sectors);
4537 DEBUG_print(" metadata_run 0x%x\n", ic->metadata_run);
4538 DEBUG_print(" log2_metadata_run %d\n", ic->log2_metadata_run);
4539 DEBUG_print(" provided_data_sectors 0x%llx (%llu)\n", ic->provided_data_sectors, ic->provided_data_sectors);
4540 DEBUG_print(" log2_buffer_sectors %u\n", ic->log2_buffer_sectors);
4543 if (ic->recalculate_flag && !(ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING))) {
4544 ic->sb->flags |= cpu_to_le32(SB_FLAG_RECALCULATING);
4545 ic->sb->recalc_sector = cpu_to_le64(0);
4548 if (ic->internal_hash) {
4549 ic->recalc_wq = alloc_workqueue("dm-integrity-recalc", WQ_MEM_RECLAIM, 1);
4550 if (!ic->recalc_wq) {
4555 INIT_WORK(&ic->recalc_work, integrity_recalc);
4557 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING)) {
4564 if (ic->sb->flags & cpu_to_le32(SB_FLAG_RECALCULATING) &&
4565 le64_to_cpu(ic->sb->recalc_sector) < ic->provided_data_sectors &&
4566 dm_integrity_disable_recalculate(ic)) {
4572 ic->bufio = dm_bufio_client_create(ic->meta_dev ? ic->meta_dev->bdev : ic->dev->bdev,
4573 1U << (SECTOR_SHIFT + ic->log2_buffer_sectors), 1, 0, NULL, NULL, 0);
4574 if (IS_ERR(ic->bufio)) {
4575 r = PTR_ERR(ic->bufio);
4577 ic->bufio = NULL;
4580 dm_bufio_set_sector_offset(ic->bufio, ic->start + ic->initial_sectors);
4582 if (ic->mode != 'R') {
4583 r = create_journal(ic, &ti->error);
4589 if (ic->mode == 'B') {
4591 unsigned int n_bitmap_pages = DIV_ROUND_UP(ic->n_bitmap_blocks, PAGE_SIZE / BITMAP_BLOCK_SIZE);
4593 ic->recalc_bitmap = dm_integrity_alloc_page_list(n_bitmap_pages);
4594 if (!ic->recalc_bitmap) {
4598 ic->may_write_bitmap = dm_integrity_alloc_page_list(n_bitmap_pages);
4599 if (!ic->may_write_bitmap) {
4603 ic->bbs = kvmalloc_array(ic->n_bitmap_blocks, sizeof(struct bitmap_block_status), GFP_KERNEL);
4604 if (!ic->bbs) {
4608 INIT_DELAYED_WORK(&ic->bitmap_flush_work, bitmap_flush_work);
4609 for (i = 0; i < ic->n_bitmap_blocks; i++) {
4610 struct bitmap_block_status *bbs = &ic->bbs[i];
4614 bbs->ic = ic;
4623 bbs->bitmap = lowmem_page_address(ic->journal[pl_index].page) + pl_offset;
4628 init_journal(ic, 0, ic->journal_sections, 0);
4629 r = dm_integrity_failed(ic);
4634 r = sync_rw_sb(ic, REQ_OP_WRITE | REQ_FUA);
4639 ic->just_formatted = true;
4642 if (!ic->meta_dev) {
4643 r = dm_set_target_max_io_len(ti, 1U << ic->sb->log2_interleave_sectors);
4647 if (ic->mode == 'B') {
4650 max_io_len = ((sector_t)ic->sectors_per_block << ic->log2_blocks_per_bitmap_bit) * (BITMAP_BLOCK_SIZE * 8);
4661 if (!ic->internal_hash)
4662 dm_integrity_set(ti, ic);
4666 if (ic->discard)
4680 struct dm_integrity_c *ic = ti->private;
4682 BUG_ON(!RB_EMPTY_ROOT(&ic->in_progress));
4683 BUG_ON(!list_empty(&ic->wait_list));
4685 if (ic->mode == 'B')
4686 cancel_delayed_work_sync(&ic->bitmap_flush_work);
4687 if (ic->metadata_wq)
4688 destroy_workqueue(ic->metadata_wq);
4689 if (ic->wait_wq)
4690 destroy_workqueue(ic->wait_wq);
4691 if (ic->offload_wq)
4692 destroy_workqueue(ic->offload_wq);
4693 if (ic->commit_wq)
4694 destroy_workqueue(ic->commit_wq);
4695 if (ic->writer_wq)
4696 destroy_workqueue(ic->writer_wq);
4697 if (ic->recalc_wq)
4698 destroy_workqueue(ic->recalc_wq);
4699 kvfree(ic->bbs);
4700 if (ic->bufio)
4701 dm_bufio_client_destroy(ic->bufio);
4702 mempool_exit(&ic->recheck_pool);
4703 mempool_exit(&ic->journal_io_mempool);
4704 if (ic->io)
4705 dm_io_client_destroy(ic->io);
4706 if (ic->dev)
4707 dm_put_device(ti, ic->dev);
4708 if (ic->meta_dev)
4709 dm_put_device(ti, ic->meta_dev);
4710 dm_integrity_free_page_list(ic->journal);
4711 dm_integrity_free_page_list(ic->journal_io);
4712 dm_integrity_free_page_list(ic->journal_xor);
4713 dm_integrity_free_page_list(ic->recalc_bitmap);
4714 dm_integrity_free_page_list(ic->may_write_bitmap);
4715 if (ic->journal_scatterlist)
4716 dm_integrity_free_journal_scatterlist(ic, ic->journal_scatterlist);
4717 if (ic->journal_io_scatterlist)
4718 dm_integrity_free_journal_scatterlist(ic, ic->journal_io_scatterlist);
4719 if (ic->sk_requests) {
4722 for (i = 0; i < ic->journal_sections; i++) {
4725 req = ic->sk_requests[i];
4731 kvfree(ic->sk_requests);
4733 kvfree(ic->journal_tree);
4734 if (ic->sb)
4735 free_pages_exact(ic->sb, SB_SECTORS << SECTOR_SHIFT);
4737 if (ic->internal_hash)
4738 crypto_free_shash(ic->internal_hash);
4739 free_alg(&ic->internal_hash_alg);
4741 if (ic->journal_crypt)
4742 crypto_free_skcipher(ic->journal_crypt);
4743 free_alg(&ic->journal_crypt_alg);
4745 if (ic->journal_mac)
4746 crypto_free_shash(ic->journal_mac);
4747 free_alg(&ic->journal_mac_alg);
4749 kfree(ic);