On Mon, Dec 11, 2023 at 05:52:17PM +0100, Jan Kara wrote:

quoted

+void bdev_associated_mapping(struct block_device *bdev,
+			     struct address_space *mapping)
+{
+	mapping->host = bdev->bd_inode;
+}

Here I'm not sure - is the helper really a win? It seems a bit obscure to
me. This initialization of another mapping for a bdev looks really special.

If we want to hide bd_inode we'll something like this helper even if
I don't particularly like it either.

But it might be a good idea to move out of this series and into the
follow on removing bd_inode, as it's rather pointless without that
context.