1 files changed, 136 insertions, 66 deletions

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 855935f6671a..611b66d73e80 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1845,8 +1845,10 @@ static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
 	int ret;
 
 	ret = btrfs_map_bio(root, rw, bio, mirror_num, 1);
-	if (ret)
-		bio_endio(bio, ret);
+	if (ret) {
+		bio->bi_error = ret;
+		bio_endio(bio);
+	}
 	return ret;
 }
 
@@ -1906,8 +1908,10 @@ mapit:
 	ret = btrfs_map_bio(root, rw, bio, mirror_num, 0);
 
 out:
-	if (ret < 0)
-		bio_endio(bio, ret);
+	if (ret < 0) {
+		bio->bi_error = ret;
+		bio_endio(bio);
+	}
 	return ret;
 }
 
@@ -3654,6 +3658,35 @@ cache_index:
 		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
 			&BTRFS_I(inode)->runtime_flags);
 
+	/*
+	 * We don't persist the id of the transaction where an unlink operation
+	 * against the inode was last made. So here we assume the inode might
+	 * have been evicted, and therefore the exact value of last_unlink_trans
+	 * lost, and set it to last_trans to avoid metadata inconsistencies
+	 * between the inode and its parent if the inode is fsync'ed and the log
+	 * replayed. For example, in the scenario:
+	 *
+	 * touch mydir/foo
+	 * ln mydir/foo mydir/bar
+	 * sync
+	 * unlink mydir/bar
+	 * echo 2 > /proc/sys/vm/drop_caches   # evicts inode
+	 * xfs_io -c fsync mydir/foo
+	 * <power failure>
+	 * mount fs, triggers fsync log replay
+	 *
+	 * We must make sure that when we fsync our inode foo we also log its
+	 * parent inode, otherwise after log replay the parent still has the
+	 * dentry with the "bar" name but our inode foo has a link count of 1
+	 * and doesn't have an inode ref with the name "bar" anymore.
+	 *
+	 * Setting last_unlink_trans to last_trans is a pessimistic approach,
+	 * but it guarantees correctness at the expense of ocassional full
+	 * transaction commits on fsync if our inode is a directory, or if our
+	 * inode is not a directory, logging its parent unnecessarily.
+	 */
+	BTRFS_I(inode)->last_unlink_trans = BTRFS_I(inode)->last_trans;
+
 	path->slots[0]++;
 	if (inode->i_nlink != 1 ||
 	    path->slots[0] >= btrfs_header_nritems(leaf))
@@ -4209,7 +4242,7 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 	u64 extent_num_bytes = 0;
 	u64 extent_offset = 0;
 	u64 item_end = 0;
-	u64 last_size = (u64)-1;
+	u64 last_size = new_size;
 	u32 found_type = (u8)-1;
 	int found_extent;
 	int del_item;
@@ -4493,8 +4526,7 @@ out:
 			btrfs_abort_transaction(trans, root, ret);
 	}
 error:
-	if (last_size != (u64)-1 &&
-	    root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
+	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
 		btrfs_ordered_update_i_size(inode, last_size, NULL);
 
 	btrfs_free_path(path);
@@ -4989,8 +5021,9 @@ static void evict_inode_truncate_pages(struct inode *inode)
 	/*
 	 * Keep looping until we have no more ranges in the io tree.
 	 * We can have ongoing bios started by readpages (called from readahead)
-	 * that didn't get their end io callbacks called yet or they are still
-	 * in progress ((extent_io.c:end_bio_extent_readpage()). This means some
+	 * that have their endio callback (extent_io.c:end_bio_extent_readpage)
+	 * still in progress (unlocked the pages in the bio but did not yet
+	 * unlocked the ranges in the io tree). Therefore this means some
 	 * ranges can still be locked and eviction started because before
 	 * submitting those bios, which are executed by a separate task (work
 	 * queue kthread), inode references (inode->i_count) were not taken
@@ -5051,7 +5084,8 @@ void btrfs_evict_inode(struct inode *inode)
 		goto no_delete;
 	}
 	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
-	btrfs_wait_ordered_range(inode, 0, (u64)-1);
+	if (!special_file(inode->i_mode))
+		btrfs_wait_ordered_range(inode, 0, (u64)-1);
 
 	btrfs_free_io_failure_record(inode, 0, (u64)-1);
 
@@ -6876,8 +6910,7 @@ out:
 
 	trace_btrfs_get_extent(root, em);
 
-	if (path)
-		btrfs_free_path(path);
+	btrfs_free_path(path);
 	if (trans) {
 		ret = btrfs_end_transaction(trans, root);
 		if (!err)
@@ -7376,6 +7409,10 @@ static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
 	return em;
 }
 
+struct btrfs_dio_data {
+	u64 outstanding_extents;
+	u64 reserve;
+};
 
 static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
 				   struct buffer_head *bh_result, int create)
@@ -7383,10 +7420,10 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
 	struct extent_map *em;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_state *cached_state = NULL;
+	struct btrfs_dio_data *dio_data = NULL;
 	u64 start = iblock << inode->i_blkbits;
 	u64 lockstart, lockend;
 	u64 len = bh_result->b_size;
-	u64 *outstanding_extents = NULL;
 	int unlock_bits = EXTENT_LOCKED;
 	int ret = 0;
 
@@ -7404,7 +7441,7 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
 		 * that anything that needs to check if there's a transction doesn't get
 		 * confused.
 		 */
-		outstanding_extents = current->journal_info;
+		dio_data = current->journal_info;
 		current->journal_info = NULL;
 	}
 
@@ -7536,16 +7573,18 @@ unlock:
 		 * within our reservation, otherwise we need to adjust our inode
 		 * counter appropriately.
 		 */
-		if (*outstanding_extents) {
-			(*outstanding_extents)--;
+		if (dio_data->outstanding_extents) {
+			(dio_data->outstanding_extents)--;
 		} else {
 			spin_lock(&BTRFS_I(inode)->lock);
 			BTRFS_I(inode)->outstanding_extents++;
 			spin_unlock(&BTRFS_I(inode)->lock);
 		}
 
-		current->journal_info = outstanding_extents;
 		btrfs_free_reserved_data_space(inode, len);
+		WARN_ON(dio_data->reserve < len);
+		dio_data->reserve -= len;
+		current->journal_info = dio_data;
 	}
 
 	/*
@@ -7568,8 +7607,8 @@ unlock:
 unlock_err:
 	clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
 			 unlock_bits, 1, 0, &cached_state, GFP_NOFS);
-	if (outstanding_extents)
-		current->journal_info = outstanding_extents;
+	if (dio_data)
+		current->journal_info = dio_data;
 	return ret;
 }
 
@@ -7687,13 +7726,13 @@ struct btrfs_retry_complete {
 	int uptodate;
 };
 
-static void btrfs_retry_endio_nocsum(struct bio *bio, int err)
+static void btrfs_retry_endio_nocsum(struct bio *bio)
 {
 	struct btrfs_retry_complete *done = bio->bi_private;
 	struct bio_vec *bvec;
 	int i;
 
-	if (err)
+	if (bio->bi_error)
 		goto end;
 
 	done->uptodate = 1;
@@ -7742,7 +7781,7 @@ try_again:
 	return 0;
 }
 
-static void btrfs_retry_endio(struct bio *bio, int err)
+static void btrfs_retry_endio(struct bio *bio)
 {
 	struct btrfs_retry_complete *done = bio->bi_private;
 	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
@@ -7751,7 +7790,7 @@ static void btrfs_retry_endio(struct bio *bio, int err)
 	int ret;
 	int i;
 
-	if (err)
+	if (bio->bi_error)
 		goto end;
 
 	uptodate = 1;
@@ -7834,12 +7873,13 @@ static int btrfs_subio_endio_read(struct inode *inode,
 	}
 }
 
-static void btrfs_endio_direct_read(struct bio *bio, int err)
+static void btrfs_endio_direct_read(struct bio *bio)
 {
 	struct btrfs_dio_private *dip = bio->bi_private;
 	struct inode *inode = dip->inode;
 	struct bio *dio_bio;
 	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
+	int err = bio->bi_error;
 
 	if (dip->flags & BTRFS_DIO_ORIG_BIO_SUBMITTED)
 		err = btrfs_subio_endio_read(inode, io_bio, err);
@@ -7850,17 +7890,14 @@ static void btrfs_endio_direct_read(struct bio *bio, int err)
 
 	kfree(dip);
 
-	/* If we had a csum failure make sure to clear the uptodate flag */
-	if (err)
-		clear_bit(BIO_UPTODATE, &dio_bio->bi_flags);
-	dio_end_io(dio_bio, err);
+	dio_end_io(dio_bio, bio->bi_error);
 
 	if (io_bio->end_io)
 		io_bio->end_io(io_bio, err);
 	bio_put(bio);
 }
 
-static void btrfs_endio_direct_write(struct bio *bio, int err)
+static void btrfs_endio_direct_write(struct bio *bio)
 {
 	struct btrfs_dio_private *dip = bio->bi_private;
 	struct inode *inode = dip->inode;
@@ -7871,12 +7908,11 @@ static void btrfs_endio_direct_write(struct bio *bio, int err)
 	struct bio *dio_bio;
 	int ret;
 
-	if (err)
-		goto out_done;
 again:
 	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
 						   &ordered_offset,
-						   ordered_bytes, !err);
+						   ordered_bytes,
+						   !bio->bi_error);
 	if (!ret)
 		goto out_test;
 
@@ -7895,15 +7931,11 @@ out_test:
 		ordered = NULL;
 		goto again;
 	}
-out_done:
 	dio_bio = dip->dio_bio;
 
 	kfree(dip);
 
-	/* If we had an error make sure to clear the uptodate flag */
-	if (err)
-		clear_bit(BIO_UPTODATE, &dio_bio->bi_flags);
-	dio_end_io(dio_bio, err);
+	dio_end_io(dio_bio, bio->bi_error);
 	bio_put(bio);
 }
 
@@ -7918,9 +7950,10 @@ static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw,
 	return 0;
 }
 
-static void btrfs_end_dio_bio(struct bio *bio, int err)
+static void btrfs_end_dio_bio(struct bio *bio)
 {
 	struct btrfs_dio_private *dip = bio->bi_private;
+	int err = bio->bi_error;
 
 	if (err)
 		btrfs_warn(BTRFS_I(dip->inode)->root->fs_info,
@@ -7949,8 +7982,8 @@ static void btrfs_end_dio_bio(struct bio *bio, int err)
 	if (dip->errors) {
 		bio_io_error(dip->orig_bio);
 	} else {
-		set_bit(BIO_UPTODATE, &dip->dio_bio->bi_flags);
-		bio_endio(dip->orig_bio, 0);
+		dip->dio_bio->bi_error = 0;
+		bio_endio(dip->orig_bio);
 	}
 out:
 	bio_put(bio);
@@ -7959,8 +7992,11 @@ out:
 static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev,
 				       u64 first_sector, gfp_t gfp_flags)
 {
-	int nr_vecs = bio_get_nr_vecs(bdev);
-	return btrfs_bio_alloc(bdev, first_sector, nr_vecs, gfp_flags);
+	struct bio *bio;
+	bio = btrfs_bio_alloc(bdev, first_sector, BIO_MAX_PAGES, gfp_flags);
+	if (bio)
+		bio_associate_current(bio);
+	return bio;
 }
 
 static inline int btrfs_lookup_and_bind_dio_csum(struct btrfs_root *root,
@@ -8163,9 +8199,8 @@ out_err:
 static void btrfs_submit_direct(int rw, struct bio *dio_bio,
 				struct inode *inode, loff_t file_offset)
 {
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct btrfs_dio_private *dip;
-	struct bio *io_bio;
+	struct btrfs_dio_private *dip = NULL;
+	struct bio *io_bio = NULL;
 	struct btrfs_io_bio *btrfs_bio;
 	int skip_sum;
 	int write = rw & REQ_WRITE;
@@ -8182,7 +8217,7 @@ static void btrfs_submit_direct(int rw, struct bio *dio_bio,
 	dip = kzalloc(sizeof(*dip), GFP_NOFS);
 	if (!dip) {
 		ret = -ENOMEM;
-		goto free_io_bio;
+		goto free_ordered;
 	}
 
 	dip->private = dio_bio->bi_private;
@@ -8210,25 +8245,56 @@ static void btrfs_submit_direct(int rw, struct bio *dio_bio,
 
 	if (btrfs_bio->end_io)
 		btrfs_bio->end_io(btrfs_bio, ret);
-free_io_bio:
-	bio_put(io_bio);
 
 free_ordered:
 	/*
-	 * If this is a write, we need to clean up the reserved space and kill
-	 * the ordered extent.
+	 * If we arrived here it means either we failed to submit the dip
+	 * or we either failed to clone the dio_bio or failed to allocate the
+	 * dip. If we cloned the dio_bio and allocated the dip, we can just
+	 * call bio_endio against our io_bio so that we get proper resource
+	 * cleanup if we fail to submit the dip, otherwise, we must do the
+	 * same as btrfs_endio_direct_[write|read] because we can't call these
+	 * callbacks - they require an allocated dip and a clone of dio_bio.
 	 */
-	if (write) {
-		struct btrfs_ordered_extent *ordered;
-		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
-		if (!test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags) &&
-		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags))
-			btrfs_free_reserved_extent(root, ordered->start,
-						   ordered->disk_len, 1);
-		btrfs_put_ordered_extent(ordered);
-		btrfs_put_ordered_extent(ordered);
+	if (io_bio && dip) {
+		io_bio->bi_error = -EIO;
+		bio_endio(io_bio);
+		/*
+		 * The end io callbacks free our dip, do the final put on io_bio
+		 * and all the cleanup and final put for dio_bio (through
+		 * dio_end_io()).
+		 */
+		dip = NULL;
+		io_bio = NULL;
+	} else {
+		if (write) {
+			struct btrfs_ordered_extent *ordered;
+
+			ordered = btrfs_lookup_ordered_extent(inode,
+							      file_offset);
+			set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
+			/*
+			 * Decrements our ref on the ordered extent and removes
+			 * the ordered extent from the inode's ordered tree,
+			 * doing all the proper resource cleanup such as for the
+			 * reserved space and waking up any waiters for this
+			 * ordered extent (through btrfs_remove_ordered_extent).
+			 */
+			btrfs_finish_ordered_io(ordered);
+		} else {
+			unlock_extent(&BTRFS_I(inode)->io_tree, file_offset,
+			      file_offset + dio_bio->bi_iter.bi_size - 1);
+		}
+		dio_bio->bi_error = -EIO;
+		/*
+		 * Releases and cleans up our dio_bio, no need to bio_put()
+		 * nor bio_endio()/bio_io_error() against dio_bio.
+		 */
+		dio_end_io(dio_bio, ret);
 	}
-	bio_endio(dio_bio, ret);
+	if (io_bio)
+		bio_put(io_bio);
+	kfree(dip);
 }
 
 static ssize_t check_direct_IO(struct btrfs_root *root, struct kiocb *iocb,
@@ -8269,7 +8335,8 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 {
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_mapping->host;
-	u64 outstanding_extents = 0;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_dio_data dio_data = { 0 };
 	size_t count = 0;
 	int flags = 0;
 	bool wakeup = true;
@@ -8307,7 +8374,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 		ret = btrfs_delalloc_reserve_space(inode, count);
 		if (ret)
 			goto out;
-		outstanding_extents = div64_u64(count +
+		dio_data.outstanding_extents = div64_u64(count +
 						BTRFS_MAX_EXTENT_SIZE - 1,
 						BTRFS_MAX_EXTENT_SIZE);
 
@@ -8316,7 +8383,8 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 		 * do the accounting properly if we go over the number we
 		 * originally calculated.  Abuse current->journal_info for this.
 		 */
-		current->journal_info = &outstanding_extents;
+		dio_data.reserve = round_up(count, root->sectorsize);
+		current->journal_info = &dio_data;
 	} else if (test_bit(BTRFS_INODE_READDIO_NEED_LOCK,
 				     &BTRFS_I(inode)->runtime_flags)) {
 		inode_dio_end(inode);
@@ -8330,9 +8398,11 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 				   btrfs_submit_direct, flags);
 	if (iov_iter_rw(iter) == WRITE) {
 		current->journal_info = NULL;
-		if (ret < 0 && ret != -EIOCBQUEUED)
-			btrfs_delalloc_release_space(inode, count);
-		else if (ret >= 0 && (size_t)ret < count)
+		if (ret < 0 && ret != -EIOCBQUEUED) {
+			if (dio_data.reserve)
+				btrfs_delalloc_release_space(inode,
+							dio_data.reserve);
+		} else if (ret >= 0 && (size_t)ret < count)
 			btrfs_delalloc_release_space(inode,
 						     count - (size_t)ret);
 	}