linux-kernel - Re: [PART1 RFC 5/9] svm: Add VMEXIT handlers for AVIC

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 for Android: free password hash cracker in your pocket

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

Message-ID: <56C5FCCF.5010802@redhat.com>
Date:	Thu, 18 Feb 2016 18:18:07 +0100
From:	Paolo Bonzini <pbonzini@...hat.com>
To:	Radim Krčmář <rkrcmar@...hat.com>
Cc:	Suravee Suthikulpanit <Suravee.Suthikulpanit@....com>,
	joro@...tes.org, alex.williamson@...hat.com, gleb@...nel.org,
	kvm@...r.kernel.org, linux-kernel@...r.kernel.org, wei@...hat.com,
	sherry.hurwitz@....com, "Wu, Feng" <feng.wu@...el.com>
Subject: Re: [PART1 RFC 5/9] svm: Add VMEXIT handlers for AVIC



On 18/02/2016 17:27, Radim Krčmář wrote:
> 2016-02-18 16:53+0100, Paolo Bonzini:
>> Patch 9 is okay, but it is also necessary to clear IsRunning in
>> kvm_arch_vcpu_blocking and set it in kvm_arch_vcpu_unblocking.  In
>> addition, vcpu_put/vcpu_load should not modify IsRunning between
>> kvm_arch_vcpu_blocking and kvm_arch_vcpu_unblocking.  Do you agree?
> 
> Yes.
> 
> I think we don't need to clear IsRunning on preemption, which would
> simplify the protection.  (I haven't thought much about userspace exit,
> so maybe we could skip that one as well, but we don't need to now.)
> 
> The reason for that is that KVM knows that the VCPU was scheduled out,
> so it couldn't do much in the AVIC VMEXIT.
> (KVM could force scheduler to pritioritize the VCPU, but our kick
>  doesn't do that now and it seems like a bad idea.)
> 
> Does it seem reasonable?

Yes, and in fact it wouldn't need to clear and set IsRunning on
vcpu_put/vcpu_load; only on vcpu_blocking/vcpu_unblocking.

The IsRunning flag is more of a IsNotHalted flag, in the end.

Paolo