On Sat, 13 Nov 2021 at 06:16, Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > > On Thu, Nov 11, 2021 at 7:25 PM Dave Airlie <airlied@xxxxxxxxx> wrote: > > > > I missed a drm-misc-next pull for the main pull last week. It wasn't > > that major and isn't the bulk of this at all. This has a bunch of > > fixes all over, a lot for amdgpu and i915. > > Ugh. > > The i915 conflict was trivial, but made me aware of that absolutely > disgusting "wbinvd_on_all_cpus()" hack. > > And that thing is much too ugly to survive. I made my merge resolution > remove that disgusting thing. > > That driver is x86-only anyway, so it all seemed completely bogus in > the first place. > > And if there is some actual non-x86 work in progress for i915, then > that wbinvd_on_all_cpus() needs to be replaced with something proper > and architecture-neutral anyway, most definitely involving a name > change, and almost certainly also involving a range for the cache > writeback. > > Because that "create broken macro on other architectures" thing is > *NOT* acceptable. > > And I sincerely hope to the gods that no cache-incoherent i915 mess > ever makes it out of the x86 world. Incoherent IO was always a > historical mistake and should never ever happen again, so we should > not spread that horrific pattern around. i915 will no longer be x86-64 only in theory, since Intel now produces PCIe graphics cards using the same hw designs. These shouldn't AFAIK require the same incoherent architecture, though PCIe unsnooped transactions are a thing in the real world. The thing is the same driver needs to build/work for the integrated and discrete cards, hence this hack, but I'm sure someone can Intel can do better. I'll leave it to Daniel to figure out who/how. Dave.
