Date:   Thu, 18 Nov 2021 18:02:17 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Ben Gardon <bgardon@google.com>, linux-kernel@vger.kernel.org,
        kvm@vger.kernel.org, Peter Xu <peterx@redhat.com>,
        Peter Shier <pshier@google.com>,
        David Matlack <dmatlack@google.com>,
        Mingwei Zhang <mizhang@google.com>,
        Yulei Zhang <yulei.kernel@gmail.com>,
        Wanpeng Li <kernellwp@gmail.com>,
        Xiao Guangrong <xiaoguangrong.eric@gmail.com>,
        Kai Huang <kai.huang@intel.com>,
        Keqian Zhu <zhukeqian1@huawei.com>,
        David Hildenbrand <david@redhat.com>
Subject: Re: [RFC 11/19] KVM: x86/mmu: Factor shadow_zero_check out of
 make_spte
Message-ID: <YZaVKcFoKR4lqDIZ@google.com>
References: <20211110223010.1392399-1-bgardon@google.com>
 <20211110223010.1392399-12-bgardon@google.com>
 <YZW02M0+YzAzBF/w@google.com>
 <YZXIqAHftH4d+B9Y@google.com>
 <YZaBSf+bPc69WR1R@google.com>
 <db8a2431-8a05-bd50-dd79-74c814c71edd@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <db8a2431-8a05-bd50-dd79-74c814c71edd@redhat.com>
Precedence: bulk

On Thu, Nov 18, 2021, Paolo Bonzini wrote:
> On 11/18/21 17:37, Sean Christopherson wrote:
> > > It's a bit ugly in that we'd pass both @kvm and @vcpu, so that needs some more
> > > thought, but at minimum it means there's no need to recalc the reserved bits.
> > 
> > Ok, I think my final vote is to have the reserved bits passed in, but with the
> > non-nested TDP reserved bits being computed at MMU init.
> 
> Yes, and that's also where I was getting with the idea of moving part of the
> "direct" MMU (man, naming these things is so hard) to struct kvm: split the
> per-vCPU state from the constant one and initialize the latter just once.
> Though perhaps I was putting the cart slightly before the horse.
> 
> On the topic of naming, we have a lot of things to name:
> 
> - the two MMU codebases: you Googlers are trying to grandfather "legacy" and
> "TDP" into upstream

Heh, I think that's like 99.9% me.

> but that's not a great name because the former is used also when shadowing
> EPT/NPT.  I'm thinking of standardizing on "shadow" and "TDP" (it's not
> perfect because of the 32-bit and tdp_mmu=0 cases, but it's a start).  Maybe
> even split parts of mmu.c out into shadow_mmu.c.

But shadow is flat out wrong until EPT and NPT support is ripped out of the "legacy"
MMU.

> - the two walkers (I'm quite convinced of splitting that part out of struct
> kvm_mmu and getting rid of walk_mmu/nested_mmu): that's easy, it can be
> walk01 and walk12 with "walk" pointing to one of them

I am all in favor of walk01 and walk12, the guest_mmu vs. nested_mmu confusion
is painful.

> - the two MMUs: with nested_mmu gone, root_mmu and guest_mmu are much less
> confusing and we can keep those names.

I would prefer root_mmu and nested_tdp_mmu.  guest_mmu is misleading because its
not used for all cases of sp->role.guest_mode=1, i.e. when L1 is not using TDP
then guest_mode=1 but KVM isn't using guest_mmu.