linux-kernel - Re: [PATCH 09/12] KVM: MMU: look for a cached PGD when going from 32-bit to 64-bit

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 11 Feb 2022 12:45:20 +0100
From:   Paolo Bonzini <pbonzini@...hat.com>
To:     Sean Christopherson <seanjc@...gle.com>
Cc:     linux-kernel@...r.kernel.org, kvm@...r.kernel.org,
        vkuznets@...hat.com, mlevitsk@...hat.com, dmatlack@...gle.com
Subject: Re: [PATCH 09/12] KVM: MMU: look for a cached PGD when going from
 32-bit to 64-bit

On 2/11/22 02:32, Sean Christopherson wrote:
> Maybe cached_root_find_and_rotate() or cached_root_find_and_age()?

I'll go for cached_root_find_and_keep_current() and 
cached_root_find_without_current(), respectively.

> 
> Hmm, while we're refactoring this, I'd really prefer we not grab vcpu->arch.mmu
> way down in the helpers.  @vcpu is needed only for the request, so what about
> doing this?
> 
> 	if (!fast_pgd_switch(vcpu, new_pgd, new_role)) {
> 		/*
> 		 * <whatever kvm_mmu_reload() becomes> will set up a new root
> 		 * prior to the next VM-Enter.  Free the current root if it's
> 		 * valid, i.e. if a valid root was evicted from the cache.
> 		 */
> 		if (VALID_PAGE(vcpu->arch.mmu->root.hpa))
> 			kvm_mmu_free_roots(vcpu, vcpu->arch.mmu, KVM_MMU_ROOT_CURRENT);
> 		return;
> 	}

I tried, but it's much easier to describe the cache functions if their 
common postcondition is "vcpu->arch.mmu->root.hpa is never stale"; which 
requires not a struct kvm_vcpu* but at least a struct kvm*, for the MMU 
lock.

I could change kvm_mmu_free_roots and cached_root_* to take a struct 
kvm* plus a struct kvm_mmu*.  Does that sound better?

Paolo