Re: [PATCH v13 05/23] fsverity: improve flushing performance of fsverity_fill_zerohash

Christoph Hellwig <[email protected]> Fri, 24 Jul 2026 15:13:09 +0200
Newsgroups dev.linux.lists.fsverity,net.sourceforge.lists.linux-f2fs-devel,org.kernel.vger.linux-btrfs,org.kernel.vger.linux-ext4,org.kernel.vger.linux-fsdevel,org.kernel.vger.linux-xfs
Message-ID <[email protected]>
On Tue, Jul 21, 2026 at 08:40:42PM +0200, Andrey Albershteyn wrote:
> The current version calls flush_dcache_folio(), in memcpy_to_folio(), to
> flush whole folio on every digest (which is 128 for 4k). Open code folio
> mapping and flushing to copy all digests at once.

What is the point of this?  flush_dcache_folio is a no-op on sane
architectures, and this should not be a performance path, so
avoiding all the kmap logic does seem benefitial.  Or at least only
opencode it for !highmem to avoid the worst part.