for-6.17-rc4-tag

-----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEE8rQSAMVO+zA4DBdWxWXV+ddtWDsFAmi3Od8ACgkQxWXV+ddt
 WDvZrA/7BOm51t1iNjw46NFviYq5u9/uTjL9hnYJoaXIH/DUdK5PyTACwlDJsrKk
 aFYTSH04OfFhmvy2Fje1YxmkuCx+VQ49qUdUBZdbx5wD7Rj+FthXcSLxrIATfQ7W
 qrLyHr65pxpVI9zWdZh+E2Ls5beCDqi/Rcdidii+NqTdnJJ/l21LXNsLPw6O09rh
 dRIJlb4fr+3ioEkAMDfC5p5lLUb/76lAeB+3CHc2cB4m3hPRMZeET/bd8pXo4g2D
 C9vjOp2asRzawM9bg33jb6LTQlHW+Yug1N13MN0tx01pTvnRbs/mGq7qlnCPtDXo
 SbFUTDqMIXnkq9ohdaOXJH9mPksG/vq+2GoS5bCORvsl+cDbFtzQsjWoC7wkcXa+
 e8JfcKrjTUKGSaCOacbz+CHEzAhiv5yCdIloofUQSQJfw+R0rsQbGlZuCP9dlzD7
 GGlnQKD23duaLs0u9wA+KC6T+Ifz7z42GRDBlzo/9ZXQcSunJJvK5hsE6S5wejka
 aZqSPfyb95gvYb+LnXepHCrRVi3lCSfV3W6m6xg2fDEgwBvxe8iEZnzSUmcMMcvu
 nY+sXS3ezpBLym+3NRYVWMlbk+Jre28TkeUCrTMTvd6nPCK9yIOeygx6XR8jKZRb
 tJG0nhs3daqcG/EUmEKwQOar5lF7dpr8kAI+LbK+Wi8a92tTxL4=
 =fL6z
 -----END PGP SIGNATURE-----

Merge tag 'for-6.17-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - fix a few races related to inode link count

 - fix inode leak on failure to add link to inode

 - move transaction aborts closer to where they happen

* tag 'for-6.17-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: avoid load/store tearing races when checking if an inode was logged
  btrfs: fix race between setting last_dir_index_offset and inode logging
  btrfs: fix race between logging inode and checking if it was logged before
  btrfs: simplify error handling logic for btrfs_link()
  btrfs: fix inode leak on failure to add link to inode
  btrfs: abort transaction on failure to add link to inode

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;
 	};

fs/btrfs/inode.c

@@ -6805,7 +6805,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	struct fscrypt_name fname;
 	u64 index;
 	int ret;
-	int drop_inode = 0;
 
 	/* do not allow sys_link's with other subvols of the same device */
 	if (btrfs_root_id(root) != btrfs_root_id(BTRFS_I(inode)->root))
@@ -6837,44 +6836,44 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 
 	/* There are several dir indexes for this inode, clear the cache. */
 	BTRFS_I(inode)->dir_index = 0ULL;
-	inc_nlink(inode);
 	inode_inc_iversion(inode);
 	inode_set_ctime_current(inode);
-	ihold(inode);
 	set_bit(BTRFS_INODE_COPY_EVERYTHING, &BTRFS_I(inode)->runtime_flags);
 
 	ret = btrfs_add_link(trans, BTRFS_I(dir), BTRFS_I(inode),
 			     &fname.disk_name, 1, index);
-
-	if (ret) {
-		drop_inode = 1;
-	} else {
-		struct dentry *parent = dentry->d_parent;
-
-		ret = btrfs_update_inode(trans, BTRFS_I(inode));
 	if (ret)
 		goto fail;
+
+	/* Link added now we update the inode item with the new link count. */
+	inc_nlink(inode);
+	ret = btrfs_update_inode(trans, BTRFS_I(inode));
+	if (ret) {
+		btrfs_abort_transaction(trans, ret);
+		goto fail;
+	}
+
 	if (inode->i_nlink == 1) {
 		/*
-			 * If new hard link count is 1, it's a file created
+		 * If the new hard link count is 1, it's a file created with the
-			 * with open(2) O_TMPFILE flag.
+		 * open(2) O_TMPFILE flag.
 		 */
 		ret = btrfs_orphan_del(trans, BTRFS_I(inode));
-			if (ret)
+		if (ret) {
+			btrfs_abort_transaction(trans, ret);
 			goto fail;
 		}
-		d_instantiate(dentry, inode);
-		btrfs_log_new_name(trans, old_dentry, NULL, 0, parent);
 	}
 
+	/* Grab reference for the new dentry passed to d_instantiate(). */
+	ihold(inode);
+	d_instantiate(dentry, inode);
+	btrfs_log_new_name(trans, old_dentry, NULL, 0, dentry->d_parent);
+
 fail:
 	fscrypt_free_filename(&fname);
 	if (trans)
 		btrfs_end_transaction(trans);
-	if (drop_inode) {
-		inode_dec_link_count(inode);
-		iput(inode);
-	}
 	btrfs_btree_balance_dirty(fs_info);
 	return ret;
 }
@@ -7830,6 +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;

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
@@ -3357,15 +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
+	 * If logged_trans is not 0 and not trans->transid, then we know the
-	 * in this transaction, so we can return false right away.
+	 * 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
@@ -3374,10 +3416,8 @@ 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)) {
+	if (!test_bit(BTRFS_ROOT_HAS_LOG_TREE, &inode->root->state))
-		inode->logged_trans = trans->transid - 1;
+		return mark_inode_as_not_logged(trans, inode);
-		return 0;
-	}
 
 	/*
 	 * We have a log tree and the inode's logged_trans is 0. We can't tell
@@ -3431,8 +3471,7 @@ 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 mark_inode_as_not_logged(trans, inode);
-		return 0;
 	}
 
 	/*
@@ -3440,20 +3479,9 @@ 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
-	 * 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;
 }
@@ -4045,7 +4073,7 @@ 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
@@ -4061,7 +4089,7 @@ 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) {