F2FS gains sub-page block support for 16KB Android kernels

F2FS can now mount and use filesystems whose block size is smaller than the kernel page size, removing a hard requirement that had blocked Android devices moving from 4KB-page to 16KB-page kernels.

Daeho Jeong of Google posted the 20-patch series on the Linux kernel list. Until now F2FS insisted that the on-disk block size equal PAGE_SIZE. A userdata partition formatted with 4KB blocks therefore could not be mounted once the device booted a 16KB-page kernel, even though the on-disk layout itself was unchanged. The series lifts that restriction while leaving the disk format alone.

The work builds on earlier large-folio support. When a single folio holds several filesystem blocks, F2FS tracks uptodate and dirty state per block rather than per page, and the read, write, garbage-collection, mmap, fallocate, swap, verity, atomic-write, and compression paths all operate in block units. When the block size still equals PAGE_SIZE the behavior is unchanged aside from incidental fixes that also help large folios.

The practical payoff is Android userdata migration: devices can keep their existing 4KB F2FS images while adopting 16KB pages for performance and memory-management gains. Compressed and atomic files needed extra care in subpage mode; the series brings those features along rather than leaving them disabled.