Re: [PATCH 3/4] KVM: Update gfn_to_pfn_cache khva when it moves within the same page

[Sean Christopherson <seanjc@google.com>]

On Tue, Nov 22, 2022, David Woodhouse wrote:
> On Tue, 2022-11-22 at 16:49 +0000, Paul Durrant wrote:
> > > Tags need your real name, not just your email address, i.e. this should be:
> > >     Reviewed-by: Paul Durrant <paul@xen.org>
> > 
> > Yes indeed it should. Don't know how I managed to screw that up... It's 
> > not like haven't type that properly hundreds of times on Xen patch reviews.
> 
> All sorted in the tree I'm curating
> https://git.infradead.org/users/dwmw2/linux.git/shortlog/refs/heads/gpc-fixes
> 
> Of those, three are marked as Cc:stable and want to go into the 6.1 release:
> 
>       KVM: x86/xen: Validate port number in SCHEDOP_poll
>       KVM: x86/xen: Only do in-kernel acceleration of hypercalls for guest CPL0
>       KVM: Update gfn_to_pfn_cache khva when it moves within the same page
> 
> The rest (including the runstate compatibility fixes) are less
> critical.
> 
> Sean, given that this now includes your patch series which in turn you
> took over from Michal, how would you prefer me to proceed?

If you're ok with a slight delay for fixing the runstate stuff, I think it makes
sense to push out everything else to fixes out to 6.3.  At the very least, I want
to explore using a gfn_to_hva_cache instead of two gfn_to_pfn_cache structures
for the runstate, which at this point would be a bit rushed for 6.2.

Pushing out to 6.3 will also give us time to do the more aggressive cleanup of
adding kvm_gpc_lock(), which I think will yield APIs that both of us are happy with,
i.e. the gpa+len of the cache can be immutable, but users can still validate their
individual usage.

> David Woodhouse (7):
>       KVM: x86/xen: Validate port number in SCHEDOP_poll
>       KVM: x86/xen: Only do in-kernel acceleration of hypercalls for guest CPL0
>       KVM: x86/xen: Add CPL to Xen hypercall tracepoint
>       MAINTAINERS: Add KVM x86/xen maintainer list
>       KVM: x86/xen: Compatibility fixes for shared runstate area
>       KVM: Update gfn_to_pfn_cache khva when it moves within the same page
>       KVM: x86/xen: Add runstate tests for 32-bit mode and crossing page boundary
> 
> Michal Luczaj (6):
>       KVM: Shorten gfn_to_pfn_cache function names
>       KVM: x86: Remove unused argument in gpc_unmap_khva()
>       KVM: Store immutable gfn_to_pfn_cache properties
>       KVM: Use gfn_to_pfn_cache's immutable "kvm" in kvm_gpc_check()
>       KVM: Clean up hva_to_pfn_retry()
>       KVM: Use gfn_to_pfn_cache's immutable "kvm" in kvm_gpc_refresh()
> 
> Sean Christopherson (4):
>       KVM: Drop KVM's API to allow temporarily unmapping gfn=>pfn cache
>       KVM: Do not partially reinitialize gfn=>pfn cache during activation
>       KVM: Drop @gpa from exported gfn=>pfn cache check() and refresh() helpers
>       KVM: Skip unnecessary "unmap" if gpc is already valid during refresh
> 
> We can reinstate the 'immutable length' thing on top, if we pick one of
> the discussed options for coping with the fact that for the runstate
> area, it *isn't* immutable. I'm slightly leaning towards just setting
> the length to '1' despite it being a lie.