[LSF/MM/BPF TOPIC] File system checksum offload

hch at infradead.org hch at infradead.org
Mon Feb 3 00:06:45 PST 2025


On Mon, Feb 03, 2025 at 08:04:42AM +0000, Johannes Thumshirn wrote:
> Well for the WAF part, it'll save us 32 Bytes per FS sector (typically 
> 4k) in the btrfs case, that's ~0.8% of the space.

It saves you from the cascading btree updates.  With nocow this could
actually allow btrfs to write with WAF=1 for pure overwrites.  But even
with data cow it will reduce a lot of churn, especially for O_SYNC
writes.




More information about the Linux-nvme mailing list