On Mon, Jul 12, 2021 at 12:24:11PM -0700, Linus Torvalds wrote: > I think architectures that have virtual caches might want to think > about this patch a bit more, but on the whole I can't argue against > the "it's badly documented and misused". > > No sane architecture will care, since dcache will be coherent (there > are more issues on the I$ side, but that's a different issue) Yeah. Once the arch maintainers look it it it might be worth to check if there is optimization potential for pages that are not in highmem and not in the page cache, as most architectures should be able to just do nothing in that case. Either way, I think getting patches 1-4 into 5.14 as bug fixes would be useful, 6 is a trivial cleanup and 5 is something we can chew on for a bit.
