linux-kernel - Re: [RFC PATCH 2/3] kvm: x86: Use KVM_DEBUGREG_NEED_RELOAD instead of KVM_DEBUGREG_BP

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, 24 Apr 2020 21:21:57 +0800
From:   Xiaoyao Li <xiaoyao.li@...el.com>
To:     Sean Christopherson <sean.j.christopherson@...el.com>
Cc:     Paolo Bonzini <pbonzini@...hat.com>, kvm@...r.kernel.org,
        Nadav Amit <namit@...technion.ac.il>,
        Vitaly Kuznetsov <vkuznets@...hat.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 2/3] kvm: x86: Use KVM_DEBUGREG_NEED_RELOAD instead of
 KVM_DEBUGREG_BP_ENABLED

On 4/24/2020 3:29 AM, Sean Christopherson wrote:
> On Thu, Apr 16, 2020 at 06:15:08PM +0800, Xiaoyao Li wrote:
>> Once any #BP enabled in DR7, it will set KVM_DEBUGREG_BP_ENABLED, which
>> leads to reload DRn before every VM entry even if none of DRn changed.
>>
>> Drop KVM_DEBUGREG_BP_ENABLED flag and set KVM_DEBUGREG_NEED_RELOAD flag
>> for the cases that DRn need to be reloaded instead, to avoid unnecessary
>> DRn reload.
> 
> Loading DRs on every VM-Enter _is_ necessary if there are breakpoints
> enabled for the guest.  The hardware DR values are not "stable", e.g. they
> are loaded with the host's values immediately after saving the guest's
> value (if DR_EXITING is disabled) in vcpu_enter_guest(), notably iff the
> host has an active/enabled breakpoint.  

May bad, bbviously I didn't think about it.

> My bet is that DRs can be changed
> from interrupt context as well.
So set KVM_DEBUGREG_NEED_RELOAD in vcpu_load won't help.

> Loading DRs for the guest (not necessarily the same as the guest's DRs) is
> necessary if a breakpoint is enabled so that the #DB is actually hit in
> guest.  It's a similar concept to instructions that consume MSR values,
> e.g. SYSCALL, RDTSCP, etc..., even if KVM intercepts the MSR/DR, hardware
> still needs the correct value so that the guest behavior is correct.
> 
>> Signed-off-by: Xiaoyao Li <xiaoyao.li@...el.com>
>> ---
>>   arch/x86/include/asm/kvm_host.h | 3 +--
>>   arch/x86/kvm/x86.c              | 4 ++--
>>   2 files changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> index f465c76e6e5a..87e2d020351e 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -509,9 +509,8 @@ struct kvm_pmu {
>>   struct kvm_pmu_ops;
>>   
>>   enum {
>> -	KVM_DEBUGREG_BP_ENABLED = 1,
>> +	KVM_DEBUGREG_NEED_RELOAD = 1,
>>   	KVM_DEBUGREG_WONT_EXIT = 2,
>> -	KVM_DEBUGREG_NEED_RELOAD = 4,
>>   };
>>   
>>   struct kvm_mtrr_range {
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index cce926658d10..71264df64001 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -1086,9 +1086,8 @@ static void kvm_update_dr7(struct kvm_vcpu *vcpu)
>>   	else
>>   		dr7 = vcpu->arch.dr7;
>>   	kvm_x86_ops.set_dr7(vcpu, dr7);
>> -	vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_BP_ENABLED;
>>   	if (dr7 & DR7_BP_EN_MASK)
>> -		vcpu->arch.switch_db_regs |= KVM_DEBUGREG_BP_ENABLED;
>> +		vcpu->arch.switch_db_regs |= KVM_DEBUGREG_NEED_RELOAD;
>>   }
>>   
>>   static u64 kvm_dr6_fixed(struct kvm_vcpu *vcpu)
>> @@ -1128,6 +1127,7 @@ static int __kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val)
>>   		break;
>>   	}
>>   
>> +	vcpu->arch.switch_db_regs |= KVM_DEBUGREG_NEED_RELOAD;
>>   	return 0;
>>   }
>>   
>> -- 
>> 2.20.1
>>