From ef07b74e1be56f9eafda6aadebb9ebba0743c9f0 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Wed, 6 Aug 2025 12:11:30 +0100 Subject: btrfs: fix race between logging inode and checking if it was logged before There's a race between checking if an inode was logged before and logging an inode that can cause us to mark an inode as not logged just after it was logged by a concurrent task: 1) We have inode X which was not logged before neither in the current transaction not in past transaction since the inode was loaded into memory, so it's ->logged_trans value is 0; 2) We are at transaction N; 3) Task A calls inode_logged() against inode X, sees that ->logged_trans is 0 and there is a log tree and so it proceeds to search in the log tree for an inode item for inode X. It doesn't see any, but before it sets ->logged_trans to N - 1... 3) Task B calls btrfs_log_inode() against inode X, logs the inode and sets ->logged_trans to N; 4) Task A now sets ->logged_trans to N - 1; 5) At this point anyone calling inode_logged() gets 0 (inode not logged) since ->logged_trans is greater than 0 and less than N, but our inode was really logged. As a consequence operations like rename, unlink and link that happen afterwards in the current transaction end up not updating the log when they should. Fix this by ensuring inode_logged() only updates ->logged_trans in case the inode item is not found in the log tree if after tacking the inode's lock (spinlock struct btrfs_inode::lock) the ->logged_trans value is still zero, since the inode lock is what protects setting ->logged_trans at btrfs_log_inode(). Fixes: 0f8ce49821de ("btrfs: avoid inode logging during rename and link when possible") Reviewed-by: Boris Burkov Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/tree-log.c | 36 ++++++++++++++++++++++++++++++------ 1 file changed, 30 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/tree-log.c') diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 69e11557fd13..839dad701290 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -3340,6 +3340,31 @@ int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, return 0; } +static bool mark_inode_as_not_logged(const struct btrfs_trans_handle *trans, + struct btrfs_inode *inode) +{ + bool ret = false; + + /* + * Do this only if ->logged_trans is still 0 to prevent races with + * concurrent logging as we may see the inode not logged when + * inode_logged() is called but it gets logged after inode_logged() did + * not find it in the log tree and we end up setting ->logged_trans to a + * value less than trans->transid after the concurrent logging task has + * set it to trans->transid. As a consequence, subsequent rename, unlink + * and link operations may end up not logging new names and removing old + * names from the log. + */ + spin_lock(&inode->lock); + if (inode->logged_trans == 0) + inode->logged_trans = trans->transid - 1; + else if (inode->logged_trans == trans->transid) + ret = true; + spin_unlock(&inode->lock); + + return ret; +} + /* * Check if an inode was logged in the current transaction. This correctly deals * with the case where the inode was logged but has a logged_trans of 0, which @@ -3374,10 +3399,8 @@ static int inode_logged(const struct btrfs_trans_handle *trans, * transaction's ID, to avoid the search below in a future call in case * a log tree gets created after this. */ - if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &inode->root->state)) { - inode->logged_trans = trans->transid - 1; - return 0; - } + if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &inode->root->state)) + return mark_inode_as_not_logged(trans, inode); /* * We have a log tree and the inode's logged_trans is 0. We can't tell @@ -3431,8 +3454,7 @@ static int inode_logged(const struct btrfs_trans_handle *trans, * Set logged_trans to a value greater than 0 and less then the * current transaction to avoid doing the search in future calls. */ - inode->logged_trans = trans->transid - 1; - return 0; + return mark_inode_as_not_logged(trans, inode); } /* @@ -3440,7 +3462,9 @@ static int inode_logged(const struct btrfs_trans_handle *trans, * the current transacion's ID, to avoid future tree searches as long as * the inode is not evicted again. */ + spin_lock(&inode->lock); inode->logged_trans = trans->transid; + spin_unlock(&inode->lock); /* * If it's a directory, then we must set last_dir_index_offset to the -- cgit v1.2.3 From 59a0dd4ab98970086fd096281b1606c506ff2698 Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Wed, 6 Aug 2025 12:11:31 +0100 Subject: btrfs: fix race between setting last_dir_index_offset and inode logging At inode_logged() if we find that the inode was not logged before we update its ->last_dir_index_offset to (u64)-1 with the goal that the next directory log operation will see the (u64)-1 and then figure out it must check what was the index of the last logged dir index key and update ->last_dir_index_offset to that key's offset (this is done in update_last_dir_index_offset()). This however has a possibility for a time window where a race can happen and lead to directory logging skipping dir index keys that should be logged. The race happens like this: 1) Task A calls inode_logged(), sees ->logged_trans as 0 and then checks that the inode item was logged before, but before it sets the inode's ->last_dir_index_offset to (u64)-1... 2) Task B is at btrfs_log_inode() which calls inode_logged() early, and that has set ->last_dir_index_offset to (u64)-1; 3) Task B then enters log_directory_changes() which calls update_last_dir_index_offset(). There it sees ->last_dir_index_offset is (u64)-1 and that the inode was logged before (ctx->logged_before is true), and so it searches for the last logged dir index key in the log tree and it finds that it has an offset (index) value of N, so it sets ->last_dir_index_offset to N, so that we can skip index keys that are less than or equal to N (later at process_dir_items_leaf()); 4) Task A now sets ->last_dir_index_offset to (u64)-1, undoing the update that task B just did; 5) Task B will now skip every index key when it enters process_dir_items_leaf(), since ->last_dir_index_offset is (u64)-1. Fix this by making inode_logged() not touch ->last_dir_index_offset and initializing it to 0 when an inode is loaded (at btrfs_alloc_inode()) and then having update_last_dir_index_offset() treat a value of 0 as meaning we must check the log tree and update with the index of the last logged index key. This is fine since the minimum possible value for ->last_dir_index_offset is 1 (BTRFS_DIR_START_INDEX - 1 = 2 - 1 = 1). This also simplifies the management of ->last_dir_index_offset and now all accesses to it are done under the inode's log_mutex. Fixes: 0f8ce49821de ("btrfs: avoid inode logging during rename and link when possible") Reviewed-by: Boris Burkov Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/btrfs_inode.h | 2 +- fs/btrfs/inode.c | 1 + fs/btrfs/tree-log.c | 17 ++--------------- 3 files changed, 4 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/tree-log.c') diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index b99fb0273292..0387b9f43a52 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -248,7 +248,7 @@ struct btrfs_inode { u64 new_delalloc_bytes; /* * The offset of the last dir index key that was logged. - * This is used only for directories. + * This is used only for directories. Protected by 'log_mutex'. */ u64 last_dir_index_offset; }; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 321a46be0377..dd82dcc7b2b7 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -7829,6 +7829,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb) ei->last_sub_trans = 0; ei->logged_trans = 0; ei->delalloc_bytes = 0; + /* new_delalloc_bytes and last_dir_index_offset are in a union. */ ei->new_delalloc_bytes = 0; ei->defrag_bytes = 0; ei->disk_i_size = 0; diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 839dad701290..2a3d9e996e72 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -3466,19 +3466,6 @@ static int inode_logged(const struct btrfs_trans_handle *trans, inode->logged_trans = trans->transid; spin_unlock(&inode->lock); - /* - * If it's a directory, then we must set last_dir_index_offset to the - * maximum possible value, so that the next attempt to log the inode does - * not skip checking if dir index keys found in modified subvolume tree - * leaves have been logged before, otherwise it would result in attempts - * to insert duplicate dir index keys in the log tree. This must be done - * because last_dir_index_offset is an in-memory only field, not persisted - * in the inode item or any other on-disk structure, so its value is lost - * once the inode is evicted. - */ - if (S_ISDIR(inode->vfs_inode.i_mode)) - inode->last_dir_index_offset = (u64)-1; - return 1; } @@ -4069,7 +4056,7 @@ done: /* * If the inode was logged before and it was evicted, then its - * last_dir_index_offset is (u64)-1, so we don't the value of the last index + * last_dir_index_offset is 0, so we don't know the value of the last index * key offset. If that's the case, search for it and update the inode. This * is to avoid lookups in the log tree every time we try to insert a dir index * key from a leaf changed in the current transaction, and to allow us to always @@ -4085,7 +4072,7 @@ static int update_last_dir_index_offset(struct btrfs_inode *inode, lockdep_assert_held(&inode->log_mutex); - if (inode->last_dir_index_offset != (u64)-1) + if (inode->last_dir_index_offset != 0) return 0; if (!ctx->logged_before) { -- cgit v1.2.3 From 986bf6ed44dff7fbae7b43a0882757ee7f5ba21b Mon Sep 17 00:00:00 2001 From: Filipe Manana Date: Wed, 6 Aug 2025 12:11:32 +0100 Subject: btrfs: avoid load/store tearing races when checking if an inode was logged At inode_logged() we do a couple lockless checks for ->logged_trans, and these are generally safe except the second one in case we get a load or store tearing due to a concurrent call updating ->logged_trans (either at btrfs_log_inode() or later at inode_logged()). In the first case it's safe to compare to the current transaction ID since once ->logged_trans is set the current transaction, we never set it to a lower value. In the second case, where we check if it's greater than zero, we are prone to load/store tearing races, since we can have a concurrent task updating to the current transaction ID with store tearing for example, instead of updating with a single 64 bits write, to update with two 32 bits writes or four 16 bits writes. In that case the reading side at inode_logged() could see a positive value that does not match the current transaction and then return a false negative. Fix this by doing the second check while holding the inode's spinlock, add some comments about it too. Also add the data_race() annotation to the first check to avoid any reports from KCSAN (or similar tools) and comment about it. Fixes: 0f8ce49821de ("btrfs: avoid inode logging during rename and link when possible") Reviewed-by: Boris Burkov Signed-off-by: Filipe Manana Signed-off-by: David Sterba --- fs/btrfs/tree-log.c | 25 +++++++++++++++++++++---- 1 file changed, 21 insertions(+), 4 deletions(-) (limited to 'fs/btrfs/tree-log.c') diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 2a3d9e996e72..7d5d90845ca9 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -3382,15 +3382,32 @@ static int inode_logged(const struct btrfs_trans_handle *trans, struct btrfs_key key; int ret; - if (inode->logged_trans == trans->transid) + /* + * Quick lockless call, since once ->logged_trans is set to the current + * transaction, we never set it to a lower value anywhere else. + */ + if (data_race(inode->logged_trans) == trans->transid) return 1; /* - * If logged_trans is not 0, then we know the inode logged was not logged - * in this transaction, so we can return false right away. + * If logged_trans is not 0 and not trans->transid, then we know the + * inode was not logged in this transaction, so we can return false + * right away. We take the lock to avoid a race caused by load/store + * tearing with a concurrent btrfs_log_inode() call or a concurrent task + * in this function further below - an update to trans->transid can be + * teared into two 32 bits updates for example, in which case we could + * see a positive value that is not trans->transid and assume the inode + * was not logged when it was. */ - if (inode->logged_trans > 0) + spin_lock(&inode->lock); + if (inode->logged_trans == trans->transid) { + spin_unlock(&inode->lock); + return 1; + } else if (inode->logged_trans > 0) { + spin_unlock(&inode->lock); return 0; + } + spin_unlock(&inode->lock); /* * If no log tree was created for this root in this transaction, then -- cgit v1.2.3