Skip to contentSkip to conversation

ext4: fix deadlock due to mbcache entry corruption

PATCH In review Jeremi Piotrowski 2 participants 2 messages Jan 5, 2023 Stable branch

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ext4: fix deadlock due to mbcache entry corruption

From: Jan Kara <[email protected]>

commit a44e84a9b7764c72896f7241a0ec9ac7e7ef38dd upstream.

When manipulating xattr blocks, we can deadlock infinitely looping inside ext4_xattr_block_set() where we constantly keep finding xattr block for reuse in mbcache but we are unable to reuse it because its reference count is too big. This happens because cache entry for the xattr block is marked as reusable (e_reusable set) although its reference count is too big. When this inconsistency happens, this inconsistent state is kept indefinitely and so ext4_xattr_block_set() keeps retrying indefinitely.

The inconsistent state is caused by non-atomic update of e_reusable bit. e_reusable is part of a bitfield and e_reusable update can race with update of e_referenced bit in the same bitfield resulting in loss of one of the updates. Fix the problem by using atomic bitops instead.

This bug has been around for many years, but it became *much* easier to hit after commit 65f8b80053a1 ("ext4: fix race when reusing xattr blocks").

A special backport to 5.15 was necessary due to changes to reference counting.

Fixes 6048c64b2609 ("mbcache: add reusable flag to cache entries")
Fixes 65f8b80053a1 ("ext4: fix race when reusing xattr blocks")
---
 fs/ext4/xattr.c         |  4 ++--
 fs/mbcache.c            | 14 ++++++++------
 include/linux/mbcache.h |  9 +++++++--
 3 files changed, 17 insertions(+), 10 deletions(-)
fs/ext4/xattr.c +2 -2
index 533216e80fa2..22700812a4d3 100644
@@ -1281,7 +1281,7 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode,
 				ce = mb_cache_entry_get(ea_block_cache, hash,
 							bh->b_blocknr);
 				if (ce) {
-					ce->e_reusable = 1;
+					set_bit(MBE_REUSABLE_B, &ce->e_flags);
 					mb_cache_entry_put(ea_block_cache, ce);
 				}
 			}
@@ -2042,7 +2042,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 				}
 				BHDR(new_bh)->h_refcount = cpu_to_le32(ref);
 				if (ref == EXT4_XATTR_REFCOUNT_MAX)
-					ce->e_reusable = 0;
+					clear_bit(MBE_REUSABLE_B, &ce->e_flags);
 				ea_bdebug(new_bh, "reusing; refcount now=%d",
 					  ref);
 				ext4_xattr_block_csum_set(inode, new_bh);
fs/mbcache.c +8 -6
index 2010bc80a3f2..ac07b50ea3df 100644
@@ -94,8 +94,9 @@ int mb_cache_entry_create(struct mb_cache *cache, gfp_t mask, u32 key,
 	atomic_set(&entry->e_refcnt, 1);
 	entry->e_key = key;
 	entry->e_value = value;
-	entry->e_reusable = reusable;
-	entry->e_referenced = 0;
+	entry->e_flags = 0;
+	if (reusable)
+		set_bit(MBE_REUSABLE_B, &entry->e_flags);
 	head = mb_cache_entry_head(cache, key);
 	hlist_bl_lock(head);
 	hlist_bl_for_each_entry(dup, dup_node, head, e_hash_list) {
@@ -155,7 +156,8 @@ static struct mb_cache_entry *__entry_find(struct mb_cache *cache,
 	while (node) {
 		entry = hlist_bl_entry(node, struct mb_cache_entry,
 				       e_hash_list);
-		if (entry->e_key == key && entry->e_reusable) {
+		if (entry->e_key == key &&
+		    test_bit(MBE_REUSABLE_B, &entry->e_flags)) {
 			atomic_inc(&entry->e_refcnt);
 			goto out;
 		}
@@ -325,7 +327,7 @@ EXPORT_SYMBOL(mb_cache_entry_delete_or_get);
 void mb_cache_entry_touch(struct mb_cache *cache,
 			  struct mb_cache_entry *entry)
 {
-	entry->e_referenced = 1;
+	set_bit(MBE_REFERENCED_B, &entry->e_flags);
 }
 EXPORT_SYMBOL(mb_cache_entry_touch);
 
@@ -350,8 +352,8 @@ static unsigned long mb_cache_shrink(struct mb_cache *cache,
 	while (nr_to_scan-- && !list_empty(&cache->c_list)) {
 		entry = list_first_entry(&cache->c_list,
 					 struct mb_cache_entry, e_list);
-		if (entry->e_referenced || atomic_read(&entry->e_refcnt) > 2) {
-			entry->e_referenced = 0;
+		if (test_bit(MBE_REFERENCED_B, &entry->e_flags) || atomic_read(&entry->e_refcnt) > 2) {
+			clear_bit(MBE_REFERENCED_B, &entry->e_flags);
 			list_move_tail(&entry->e_list, &cache->c_list);
 			continue;
 		}
include/linux/mbcache.h +7 -2
index 8eca7f25c432..62927f7e2588 100644
@@ -10,6 +10,12 @@
 
 struct mb_cache;
 
+/* Cache entry flags */
+enum {
+	MBE_REFERENCED_B = 0,
+	MBE_REUSABLE_B
+};
+
 struct mb_cache_entry {
 	/* List of entries in cache - protected by cache->c_list_lock */
 	struct list_head	e_list;
@@ -18,8 +24,7 @@ struct mb_cache_entry {
 	atomic_t		e_refcnt;
 	/* Key in hash - stable during lifetime of the entry */
 	u32			e_key;
-	u32			e_referenced:1;
-	u32			e_reusable:1;
+	unsigned long		e_flags;
 	/* User provided value - stable during lifetime of the entry */
 	u64			e_value;
 };
Signature · 1 line
2.25.1
Acked-by on v1 1/1
#1 lore raw
Quoted 24 lines · Jeremi Piotrowski

From: Jan Kara <[email protected]>

commit a44e84a9b7764c72896f7241a0ec9ac7e7ef38dd upstream.

When manipulating xattr blocks, we can deadlock infinitely looping inside ext4_xattr_block_set() where we constantly keep finding xattr block for reuse in mbcache but we are unable to reuse it because its reference count is too big. This happens because cache entry for the xattr block is marked as reusable (e_reusable set) although its reference count is too big. When this inconsistency happens, this inconsistent state is kept indefinitely and so ext4_xattr_block_set() keeps retrying indefinitely.

The inconsistent state is caused by non-atomic update of e_reusable bit. e_reusable is part of a bitfield and e_reusable update can race with update of e_referenced bit in the same bitfield resulting in loss of one of the updates. Fix the problem by using atomic bitops instead.

This bug has been around for many years, but it became *much* easier to hit after commit 65f8b80053a1 ("ext4: fix race when reusing xattr blocks").

A special backport to 5.15 was necessary due to changes to reference counting.

Quoted 11 lines
Fixes 6048c64b2609 ("mbcache: add reusable flag to cache entries")
Fixes 65f8b80053a1 ("ext4: fix race when reusing xattr blocks")

The backport looks good to me. Not sure if Greg tracks acks for stable trees but anyway Greg, feel free to consider this:

Honza

Quoted 104 lines
---
 fs/ext4/xattr.c         |  4 ++--
 fs/mbcache.c            | 14 ++++++++------
 include/linux/mbcache.h |  9 +++++++--
 3 files changed, 17 insertions(+), 10 deletions(-)
fs/ext4/xattr.c +2 -2
index 533216e80fa2..22700812a4d3 100644
@@ -1281,7 +1281,7 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode,
 				ce = mb_cache_entry_get(ea_block_cache, hash,
 							bh->b_blocknr);
 				if (ce) {
-					ce->e_reusable = 1;
+					set_bit(MBE_REUSABLE_B, &ce->e_flags);
 					mb_cache_entry_put(ea_block_cache, ce);
 				}
 			}
@@ -2042,7 +2042,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 				}
 				BHDR(new_bh)->h_refcount = cpu_to_le32(ref);
 				if (ref == EXT4_XATTR_REFCOUNT_MAX)
-					ce->e_reusable = 0;
+					clear_bit(MBE_REUSABLE_B, &ce->e_flags);
 				ea_bdebug(new_bh, "reusing; refcount now=%d",
 					  ref);
 				ext4_xattr_block_csum_set(inode, new_bh);
fs/mbcache.c +8 -6
index 2010bc80a3f2..ac07b50ea3df 100644
@@ -94,8 +94,9 @@ int mb_cache_entry_create(struct mb_cache *cache, gfp_t mask, u32 key,
 	atomic_set(&entry->e_refcnt, 1);
 	entry->e_key = key;
 	entry->e_value = value;
-	entry->e_reusable = reusable;
-	entry->e_referenced = 0;
+	entry->e_flags = 0;
+	if (reusable)
+		set_bit(MBE_REUSABLE_B, &entry->e_flags);
 	head = mb_cache_entry_head(cache, key);
 	hlist_bl_lock(head);
 	hlist_bl_for_each_entry(dup, dup_node, head, e_hash_list) {
@@ -155,7 +156,8 @@ static struct mb_cache_entry *__entry_find(struct mb_cache *cache,
 	while (node) {
 		entry = hlist_bl_entry(node, struct mb_cache_entry,
 				       e_hash_list);
-		if (entry->e_key == key && entry->e_reusable) {
+		if (entry->e_key == key &&
+		    test_bit(MBE_REUSABLE_B, &entry->e_flags)) {
 			atomic_inc(&entry->e_refcnt);
 			goto out;
 		}
@@ -325,7 +327,7 @@ EXPORT_SYMBOL(mb_cache_entry_delete_or_get);
 void mb_cache_entry_touch(struct mb_cache *cache,
 			  struct mb_cache_entry *entry)
 {
-	entry->e_referenced = 1;
+	set_bit(MBE_REFERENCED_B, &entry->e_flags);
 }
 EXPORT_SYMBOL(mb_cache_entry_touch);
 
@@ -350,8 +352,8 @@ static unsigned long mb_cache_shrink(struct mb_cache *cache,
 	while (nr_to_scan-- && !list_empty(&cache->c_list)) {
 		entry = list_first_entry(&cache->c_list,
 					 struct mb_cache_entry, e_list);
-		if (entry->e_referenced || atomic_read(&entry->e_refcnt) > 2) {
-			entry->e_referenced = 0;
+		if (test_bit(MBE_REFERENCED_B, &entry->e_flags) || atomic_read(&entry->e_refcnt) > 2) {
+			clear_bit(MBE_REFERENCED_B, &entry->e_flags);
 			list_move_tail(&entry->e_list, &cache->c_list);
 			continue;
 		}
include/linux/mbcache.h +7 -2
index 8eca7f25c432..62927f7e2588 100644
@@ -10,6 +10,12 @@
 
 struct mb_cache;
 
+/* Cache entry flags */
+enum {
+	MBE_REFERENCED_B = 0,
+	MBE_REUSABLE_B
+};
+
 struct mb_cache_entry {
 	/* List of entries in cache - protected by cache->c_list_lock */
 	struct list_head	e_list;
@@ -18,8 +24,7 @@ struct mb_cache_entry {
 	atomic_t		e_refcnt;
 	/* Key in hash - stable during lifetime of the entry */
 	u32			e_key;
-	u32			e_referenced:1;
-	u32			e_reusable:1;
+	unsigned long		e_flags;
 	/* User provided value - stable during lifetime of the entry */
 	u64			e_value;
 };
Signature · 1 line
2.25.1
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR