mm/filemap: don't revert iter on -EIOCBQUEUED

Currently, if I/O is enqueued for async execution direct paths of
generic_file_{read,write}_iter() will always revert the iter.  There are
no users expecting that, and that is also costly.  Leave iterators as is
on -EIOCBQUEUED.

Link: https://lkml.kernel.org/r/f5247b60e7abbd2ff850cd108491f53a2e0c501a.1610207781.git.asml.silence@gmail.com
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap.c

@@ -2615,7 +2615,8 @@ generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
 			iocb->ki_pos += retval;
 			count -= retval;
 		}
-		iov_iter_revert(iter, count - iov_iter_count(iter));
+		if (retval != -EIOCBQUEUED)
+			iov_iter_revert(iter, count - iov_iter_count(iter));
 
 		/*
 		 * Btrfs can have a short DIO read if we encounter
@@ -3426,7 +3427,8 @@ generic_file_direct_write(struct kiocb *iocb, struct iov_iter *from)
 		}
 		iocb->ki_pos = pos;
 	}
-	iov_iter_revert(from, write_len - iov_iter_count(from));
+	if (written != -EIOCBQUEUED)
+		iov_iter_revert(from, write_len - iov_iter_count(from));
 out:
 	return written;
 }