# F2FS adds multi-grain timestamps and metadata cache for Linux 7.4

> **Key takeaway:** F2FS gains opt-in multi-grain timestamp support and a dedicated metadata cache infrastructure that removes the page-size-equals-block-size assumption, enabling future large folio support.

**Published:** 2026-10-29T16:52:52+00:00 
**Source:** Phoronix Benchmarks & Drivers 
**Category:** kernel-os 
**Canonical URL:** https://fosswire.org/news/f2fs-adds-multi-grain-timestamps-and-metadata-cache-for-linu.html 

## Summary
The Flash Friendly File System (F2FS) has received a substantial update for the upcoming Linux 7.4 kernel, headlined by opt-in multi-grain timestamp support and a new dedicated metadata cache infrastructure. Multi-grain timestamps allow F2FS to mix coarse and fine-grained timestamps, matching capabilities already present in EXT4, Btrfs, and XFS, with practical benefits for NFS cache coherency without performance penalties. The metadata cache work decouples metadata pages from the VFS page cache, moving meta, node, and compress blocks out of the page cache and parameterizing on-disk block layouts dynamically.

This architectural change removes the longstanding assumption that page size must equal block size, laying groundwork for large folio support and variable page-to-block mappings in future kernel releases. The update also includes bio entry-count throttling for asynchronous write completion, sysfs exposure of the resizable_tail feature, and numerous stability fixes addressing race conditions in compression, memory-mapped writes, checkpointing, and sequential zoned devices.

## Storage architecture and metadata changes in F2FS
- **Dedicated metadata cache:** Meta, node, and compress blocks are moved out of the VFS page cache, eliminating the page == block size assumption for metadata and enabling future large folio support.
- **Dynamic block layout parameterization:** On-disk block layouts and geometry macros are now dynamically parameterized, paving the way for page != block configurations.
- **Multi-grain timestamps:** F2FS now supports opt-in multigrain timestamps for mixing coarse and fine-grained timestamp precision, improving NFS cache coherency without performance degradation.
- **Asynchronous write throttling:** Bio entry-count throttling is added for write_end_io completion paths to manage asynchronous write completion behavior.
- **Resizable tail exposure:** The resizable_tail feature is now configurable via sysfs, and critical bug fixes address race conditions across compression, mmap faults, checkpoint/shutdown, filesystem freeze, and sequential zoned device handling.

## Why it matters
For Linux distributions shipping F2FS on flash storage, embedded systems, and mobile devices, these changes improve both immediate stability and future-proof the filesystem against evolving kernel memory management. The metadata cache rework aligns F2FS with broader Linux filesystem modernization efforts seen in EXT4 and Btrfs, while multi-grain timestamp parity simplifies NFS deployments that rely on F2FS-backed storage. System administrators and distribution maintainers should note that these improvements will arrive with Linux 7.4, with backporting decisions left to individual distribution kernel teams.