Merge: dm: sync outstanding bugfixes to v6.15 from upstream

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/6900

JIRA: https://issues.redhat.com/browse/RHEL-92761
JIRA: https://issues.redhat.com/browse/RHEL-56106
Tested: lvm2-testsuite
Upstream Status: kernel/git/torvalds/linux.git

Pull in outstanding dm bugfiles from upstream

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>

Approved-by: Matthew Sakai <msakai@redhat.com>
Approved-by: Mikuláš Patočka <mpatocka@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: Jan Stancek <jstancek@redhat.com>

Author: jstancek

drivers/md/dm-bufio.c

@@ -68,6 +68,8 @@
 #define LIST_DIRTY	1
 #define LIST_SIZE	2
 
+#define SCAN_RESCHED_CYCLE	16
+
 /*--------------------------------------------------------------*/
 
 /*
@@ -2426,7 +2428,12 @@ static void __scan(struct dm_bufio_client *c)
 
 			atomic_long_dec(&c->need_shrink);
 			freed++;
-			cond_resched();
+
+			if (unlikely(freed % SCAN_RESCHED_CYCLE == 0)) {
+				dm_bufio_unlock(c);
+				cond_resched();
+				dm_bufio_lock(c);
+			}
 		}
 	}
 }

drivers/md/dm-cache-target.c

@@ -406,6 +406,12 @@ struct cache {
 	mempool_t migration_pool;
 
 	struct bio_set bs;
+
+	/*
+	 * Cache_size entries. Set bits indicate blocks mapped beyond the
+	 * target length, which are marked for invalidation.
+	 */
+	unsigned long *invalid_bitset;
 };
 
 struct per_bio_data {
@@ -1922,6 +1928,9 @@ static void __destroy(struct cache *cache)
 	if (cache->discard_bitset)
 		free_bitset(cache->discard_bitset);
 
+	if (cache->invalid_bitset)
+		free_bitset(cache->invalid_bitset);
+
 	if (cache->copier)
 		dm_kcopyd_client_destroy(cache->copier);
 
@@ -2510,6 +2519,13 @@ static int cache_create(struct cache_args *ca, struct cache **result)
 	}
 	clear_bitset(cache->discard_bitset, from_dblock(cache->discard_nr_blocks));
 
+	cache->invalid_bitset = alloc_bitset(from_cblock(cache->cache_size));
+	if (!cache->invalid_bitset) {
+		*error = "could not allocate bitset for invalid blocks";
+		goto bad;
+	}
+	clear_bitset(cache->invalid_bitset, from_cblock(cache->cache_size));
+
 	cache->copier = dm_kcopyd_client_create(&dm_kcopyd_throttle);
 	if (IS_ERR(cache->copier)) {
 		*error = "could not create kcopyd client";
@@ -2808,6 +2824,24 @@ static int load_mapping(void *context, dm_oblock_t oblock, dm_cblock_t cblock,
 	return policy_load_mapping(cache->policy, oblock, cblock, dirty, hint, hint_valid);
 }
 
+static int load_filtered_mapping(void *context, dm_oblock_t oblock, dm_cblock_t cblock,
+				 bool dirty, uint32_t hint, bool hint_valid)
+{
+	struct cache *cache = context;
+
+	if (from_oblock(oblock) >= from_oblock(cache->origin_blocks)) {
+		if (dirty) {
+			DMERR("%s: unable to shrink origin; cache block %u is dirty",
+			      cache_device_name(cache), from_cblock(cblock));
+			return -EFBIG;
+		}
+		set_bit(from_cblock(cblock), cache->invalid_bitset);
+		return 0;
+	}
+
+	return load_mapping(context, oblock, cblock, dirty, hint, hint_valid);
+}
+
 /*
  * The discard block size in the on disk metadata is not
  * necessarily the same as we're currently using.  So we have to
@@ -2899,6 +2933,27 @@ static dm_cblock_t get_cache_dev_size(struct cache *cache)
 	return to_cblock(size);
 }
 
+static bool can_resume(struct cache *cache)
+{
+	/*
+	 * Disallow retrying the resume operation for devices that failed the
+	 * first resume attempt, as the failure leaves the policy object partially
+	 * initialized. Retrying could trigger BUG_ON when loading cache mappings
+	 * into the incomplete policy object.
+	 */
+	if (cache->sized && !cache->loaded_mappings) {
+		if (get_cache_mode(cache) != CM_WRITE)
+			DMERR("%s: unable to resume a failed-loaded cache, please check metadata.",
+			      cache_device_name(cache));
+		else
+			DMERR("%s: unable to resume cache due to missing proper cache table reload",
+			      cache_device_name(cache));
+		return false;
+	}
+
+	return true;
+}
+
 static bool can_resize(struct cache *cache, dm_cblock_t new_size)
 {
 	if (from_cblock(new_size) > from_cblock(cache->cache_size)) {
@@ -2941,12 +2996,33 @@ static int resize_cache_dev(struct cache *cache, dm_cblock_t new_size)
 	return 0;
 }
 
+static int truncate_oblocks(struct cache *cache)
+{
+	uint32_t nr_blocks = from_cblock(cache->cache_size);
+	uint32_t i;
+	int r;
+
+	for_each_set_bit(i, cache->invalid_bitset, nr_blocks) {
+		r = dm_cache_remove_mapping(cache->cmd, to_cblock(i));
+		if (r) {
+			DMERR_LIMIT("%s: invalidation failed; couldn't update on disk metadata",
+				    cache_device_name(cache));
+			return r;
+		}
+	}
+
+	return 0;
+}
+
 static int cache_preresume(struct dm_target *ti)
 {
 	int r = 0;
 	struct cache *cache = ti->private;
 	dm_cblock_t csize = get_cache_dev_size(cache);
 
+	if (!can_resume(cache))
+		return -EINVAL;
+
 	/*
 	 * Check to see if the cache has resized.
 	 */
@@ -2962,11 +3038,25 @@ static int cache_preresume(struct dm_target *ti)
 	}
 
 	if (!cache->loaded_mappings) {
+		/*
+		 * The fast device could have been resized since the last
+		 * failed preresume attempt.  To be safe we start by a blank
+		 * bitset for cache blocks.
+		 */
+		clear_bitset(cache->invalid_bitset, from_cblock(cache->cache_size));
+
 		r = dm_cache_load_mappings(cache->cmd, cache->policy,
-					   load_mapping, cache);
+					   load_filtered_mapping, cache);
 		if (r) {
 			DMERR("%s: could not load cache mappings", cache_device_name(cache));
-			metadata_operation_failed(cache, "dm_cache_load_mappings", r);
+			if (r != -EFBIG)
+				metadata_operation_failed(cache, "dm_cache_load_mappings", r);
+			return r;
+		}
+
+		r = truncate_oblocks(cache);
+		if (r) {
+			metadata_operation_failed(cache, "dm_cache_remove_mapping", r);
 			return r;
 		}
 
@@ -3426,7 +3516,7 @@ static void cache_io_hints(struct dm_target *ti, struct queue_limits *limits)
 
 static struct target_type cache_target = {
 	.name = "cache",
-	.version = {2, 2, 0},
+	.version = {2, 3, 0},
 	.module = THIS_MODULE,
 	.ctr = cache_ctr,
 	.dtr = cache_dtr,

drivers/md/dm-ebs-target.c

@@ -390,6 +390,12 @@ static int ebs_map(struct dm_target *ti, struct bio *bio)
 	return DM_MAPIO_REMAPPED;
 }
 
+static void ebs_postsuspend(struct dm_target *ti)
+{
+	struct ebs_c *ec = ti->private;
+	dm_bufio_client_reset(ec->bufio);
+}
+
 static void ebs_status(struct dm_target *ti, status_type_t type,
 		       unsigned int status_flags, char *result, unsigned int maxlen)
 {
@@ -447,6 +453,7 @@ static struct target_type ebs_target = {
 	.ctr		 = ebs_ctr,
 	.dtr		 = ebs_dtr,
 	.map		 = ebs_map,
+	.postsuspend	 = ebs_postsuspend,
 	.status		 = ebs_status,
 	.io_hints	 = ebs_io_hints,
 	.prepare_ioctl	 = ebs_prepare_ioctl,

drivers/md/dm-integrity.c

@@ -21,6 +21,7 @@
 #include <linux/reboot.h>
 #include <crypto/hash.h>
 #include <crypto/skcipher.h>
+#include <crypto/algapi.h>
 #include <linux/async_tx.h>
 #include <linux/dm-bufio.h>
 
@@ -516,7 +517,7 @@ static int sb_mac(struct dm_integrity_c *ic, bool wr)
 			dm_integrity_io_error(ic, "crypto_shash_digest", r);
 			return r;
 		}
-		if (memcmp(mac, actual_mac, mac_size)) {
+		if (crypto_memneq(mac, actual_mac, mac_size)) {
 			dm_integrity_io_error(ic, "superblock mac", -EILSEQ);
 			dm_audit_log_target(DM_MSG_PREFIX, "mac-superblock", ic->ti, 0);
 			return -EILSEQ;
@@ -859,7 +860,7 @@ static void rw_section_mac(struct dm_integrity_c *ic, unsigned int section, bool
 		if (likely(wr))
 			memcpy(&js->mac, result + (j * JOURNAL_MAC_PER_SECTOR), JOURNAL_MAC_PER_SECTOR);
 		else {
-			if (memcmp(&js->mac, result + (j * JOURNAL_MAC_PER_SECTOR), JOURNAL_MAC_PER_SECTOR)) {
+			if (crypto_memneq(&js->mac, result + (j * JOURNAL_MAC_PER_SECTOR), JOURNAL_MAC_PER_SECTOR)) {
 				dm_integrity_io_error(ic, "journal mac", -EILSEQ);
 				dm_audit_log_target(DM_MSG_PREFIX, "mac-journal", ic->ti, 0);
 			}
@@ -1401,10 +1402,9 @@ static bool find_newer_committed_node(struct dm_integrity_c *ic, struct journal_
 static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, sector_t *metadata_block,
 			       unsigned int *metadata_offset, unsigned int total_size, int op)
 {
-#define MAY_BE_FILLER		1
-#define MAY_BE_HASH		2
 	unsigned int hash_offset = 0;
-	unsigned int may_be = MAY_BE_HASH | (ic->discard ? MAY_BE_FILLER : 0);
+	unsigned char mismatch_hash = 0;
+	unsigned char mismatch_filler = !ic->discard;
 
 	do {
 		unsigned char *data, *dp;
@@ -1425,37 +1425,38 @@ static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, se
 		if (op == TAG_READ) {
 			memcpy(tag, dp, to_copy);
 		} else if (op == TAG_WRITE) {
-			if (memcmp(dp, tag, to_copy)) {
+			if (crypto_memneq(dp, tag, to_copy)) {
 				memcpy(dp, tag, to_copy);
 				dm_bufio_mark_partial_buffer_dirty(b, *metadata_offset, *metadata_offset + to_copy);
 			}
 		} else {
 			/* e.g.: op == TAG_CMP */
 
 			if (likely(is_power_of_2(ic->tag_size))) {
-				if (unlikely(memcmp(dp, tag, to_copy)))
-					if (unlikely(!ic->discard) ||
-					    unlikely(memchr_inv(dp, DISCARD_FILLER, to_copy) != NULL)) {
-						goto thorough_test;
-				}
+				if (unlikely(crypto_memneq(dp, tag, to_copy)))
+					goto thorough_test;
 			} else {
 				unsigned int i, ts;
 thorough_test:
 				ts = total_size;
 
 				for (i = 0; i < to_copy; i++, ts--) {
-					if (unlikely(dp[i] != tag[i]))
-						may_be &= ~MAY_BE_HASH;
-					if (likely(dp[i] != DISCARD_FILLER))
-						may_be &= ~MAY_BE_FILLER;
+					/*
+					 * Warning: the control flow must not be
+					 * dependent on match/mismatch of
+					 * individual bytes.
+					 */
+					mismatch_hash |= dp[i] ^ tag[i];
+					mismatch_filler |= dp[i] ^ DISCARD_FILLER;
 					hash_offset++;
 					if (unlikely(hash_offset == ic->tag_size)) {
-						if (unlikely(!may_be)) {
+						if (unlikely(mismatch_hash) && unlikely(mismatch_filler)) {
 							dm_bufio_release(b);
 							return ts;
 						}
 						hash_offset = 0;
-						may_be = MAY_BE_HASH | (ic->discard ? MAY_BE_FILLER : 0);
+						mismatch_hash = 0;
+						mismatch_filler = !ic->discard;
 					}
 				}
 			}
@@ -1476,8 +1477,6 @@ static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, se
 	} while (unlikely(total_size));
 
 	return 0;
-#undef MAY_BE_FILLER
-#undef MAY_BE_HASH
 }
 
 struct flush_request {
@@ -2076,7 +2075,7 @@ static bool __journal_read_write(struct dm_integrity_io *dio, struct bio *bio,
 					char checksums_onstack[MAX_T(size_t, HASH_MAX_DIGESTSIZE, MAX_TAG_SIZE)];
 
 					integrity_sector_checksum(ic, logical_sector, mem + bv.bv_offset, checksums_onstack);
-					if (unlikely(memcmp(checksums_onstack, journal_entry_tag(ic, je), ic->tag_size))) {
+					if (unlikely(crypto_memneq(checksums_onstack, journal_entry_tag(ic, je), ic->tag_size))) {
 						DMERR_LIMIT("Checksum failed when reading from journal, at sector 0x%llx",
 							    logical_sector);
 						dm_audit_log_bio(DM_MSG_PREFIX, "journal-checksum",
@@ -2595,7 +2594,7 @@ static void dm_integrity_inline_recheck(struct work_struct *w)
 		bio_put(outgoing_bio);
 
 		integrity_sector_checksum(ic, dio->bio_details.bi_iter.bi_sector, outgoing_data, digest);
-		if (unlikely(memcmp(digest, dio->integrity_payload, min(crypto_shash_digestsize(ic->internal_hash), ic->tag_size)))) {
+		if (unlikely(crypto_memneq(digest, dio->integrity_payload, min(crypto_shash_digestsize(ic->internal_hash), ic->tag_size)))) {
 			DMERR_LIMIT("%pg: Checksum failed at sector 0x%llx",
 				ic->dev->bdev, dio->bio_details.bi_iter.bi_sector);
 			atomic64_inc(&ic->number_of_mismatches);
@@ -2634,7 +2633,7 @@ static int dm_integrity_end_io(struct dm_target *ti, struct bio *bio, blk_status
 				char *mem = bvec_kmap_local(&bv);
 				//memset(mem, 0xff, ic->sectors_per_block << SECTOR_SHIFT);
 				integrity_sector_checksum(ic, dio->bio_details.bi_iter.bi_sector, mem, digest);
-				if (unlikely(memcmp(digest, dio->integrity_payload + pos,
+				if (unlikely(crypto_memneq(digest, dio->integrity_payload + pos,
 						min(crypto_shash_digestsize(ic->internal_hash), ic->tag_size)))) {
 					kunmap_local(mem);
 					dm_integrity_free_payload(dio);
@@ -2911,7 +2910,7 @@ static void do_journal_write(struct dm_integrity_c *ic, unsigned int write_start
 
 					integrity_sector_checksum(ic, sec + ((l - j) << ic->sb->log2_sectors_per_block),
 								  (char *)access_journal_data(ic, i, l), test_tag);
-					if (unlikely(memcmp(test_tag, journal_entry_tag(ic, je2), ic->tag_size))) {
+					if (unlikely(crypto_memneq(test_tag, journal_entry_tag(ic, je2), ic->tag_size))) {
 						dm_integrity_io_error(ic, "tag mismatch when replaying journal", -EILSEQ);
 						dm_audit_log_target(DM_MSG_PREFIX, "integrity-replay-journal", ic->ti, 0);
 					}
@@ -5084,16 +5083,19 @@ static int dm_integrity_ctr(struct dm_target *ti, unsigned int argc, char **argv
 
 		ic->recalc_bitmap = dm_integrity_alloc_page_list(n_bitmap_pages);
 		if (!ic->recalc_bitmap) {
+			ti->error = "Could not allocate memory for bitmap";
 			r = -ENOMEM;
 			goto bad;
 		}
 		ic->may_write_bitmap = dm_integrity_alloc_page_list(n_bitmap_pages);
 		if (!ic->may_write_bitmap) {
+			ti->error = "Could not allocate memory for bitmap";
 			r = -ENOMEM;
 			goto bad;
 		}
 		ic->bbs = kvmalloc_array(ic->n_bitmap_blocks, sizeof(struct bitmap_block_status), GFP_KERNEL);
 		if (!ic->bbs) {
+			ti->error = "Could not allocate memory for bitmap";
 			r = -ENOMEM;
 			goto bad;
 		}
@@ -5174,7 +5176,7 @@ static void dm_integrity_dtr(struct dm_target *ti)
 	BUG_ON(!RB_EMPTY_ROOT(&ic->in_progress));
 	BUG_ON(!list_empty(&ic->wait_list));
 
-	if (ic->mode == 'B')
+	if (ic->mode == 'B' && ic->bitmap_flush_work.work.func)
 		cancel_delayed_work_sync(&ic->bitmap_flush_work);
 	if (ic->metadata_wq)
 		destroy_workqueue(ic->metadata_wq);

drivers/md/dm-table.c

@@ -523,8 +523,9 @@ static char **realloc_argv(unsigned int *size, char **old_argv)
 		gfp = GFP_NOIO;
 	}
 	argv = kmalloc_array(new_size, sizeof(*argv), gfp);
-	if (argv && old_argv) {
-		memcpy(argv, old_argv, *size * sizeof(*argv));
+	if (argv) {
+		if (old_argv)
+			memcpy(argv, old_argv, *size * sizeof(*argv));
 		*size = new_size;
 	}
 
@@ -1177,7 +1178,7 @@ static int dm_keyslot_evict(struct blk_crypto_profile *profile,
 
 	t = dm_get_live_table(md, &srcu_idx);
 	if (!t)
-		return 0;
+		goto put_live_table;
 
 	for (unsigned int i = 0; i < t->num_targets; i++) {
 		struct dm_target *ti = dm_table_get_target(t, i);
@@ -1188,6 +1189,7 @@ static int dm_keyslot_evict(struct blk_crypto_profile *profile,
 					  (void *)key);
 	}
 
+put_live_table:
 	dm_put_live_table(md, srcu_idx);
 	return 0;
 }