splice: Fix filemap_splice_read() to use the correct inode

Fix filemap_splice_read() to use file->f_mapping->host, not file->f_inode,
as the source of the file size because in the case of a block device,
file->f_inode points to the block-special file (which is typically 0
length) and not the backing store.

Fixes: 07073eb ("splice: Add a func to do a splice from a buffered file without ITER_PIPE")
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Christian Brauner <brauner@kernel.org>
cc: Steve French <stfrench@microsoft.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Al Viro <viro@zeniv.linux.org.uk>
cc: David Hildenbrand <david@redhat.com>
cc: John Hubbard <jhubbard@nvidia.com>
cc: linux-mm@kvack.org
cc: linux-block@vger.kernel.org
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/20230522135018.2742245-2-dhowells@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

mm/filemap.c

@@ -2900,7 +2900,7 @@ ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
 	do {
 		cond_resched();
 
-		if (*ppos >= i_size_read(file_inode(in)))
+		if (*ppos >= i_size_read(in->f_mapping->host))
 			break;
 
 		iocb.ki_pos = *ppos;
@@ -2916,7 +2916,7 @@ ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
 		 * part of the page is not copied back to userspace (unless
 		 * another truncate extends the file - this is desired though).
 		 */
-		isize = i_size_read(file_inode(in));
+		isize = i_size_read(in->f_mapping->host);
 		if (unlikely(*ppos >= isize))
 			break;
 		end_offset = min_t(loff_t, isize, *ppos + len);