btrfs_setsize: Why do we truncate pages beyond newsize?

Hello all,

I am trying to fix failures of generic/029 and generic/030 (in subpagesize-blocksize scenario) using the pagecache_isize_extended() function provided by the commit "vfs: fix data corruption when blocksize < pagesize for mmaped data" (90a8020278c1598fafd071736a0846b38510309c).

In btrfs_setsize(), For the "newsize > oldsize" scenario, why do we have "truncate_pagecache(inode, newsize)" i.e. why do we truncate pages beyond the new size of the file (where pages normally shouldn't exist at this point of execution).

chandan