Re: [PATCH] KVM: arm/arm64: Assign pmc->idx before kvm_pmu_stop_counter()

From: Marc Zyngier <maz@kernel.org>
To: Julien Thierry <julien.thierry@arm.com>,
	Zenghui Yu <yuzenghui@huawei.com>,
	kvmarm@lists.cs.columbia.edu,
	linux-arm-kernel@lists.infradead.org
Cc: suzuki.poulose@arm.com, marc.zyngier@arm.com,
	linux-kernel@vger.kernel.org, james.morse@arm.com,
	andrew.murray@arm.com, wanghaibin.wang@huawei.com,
	julien.thierry.kdev@gmail.com
Subject: Re: [PATCH] KVM: arm/arm64: Assign pmc->idx before kvm_pmu_stop_counter()
Date: Wed, 17 Jul 2019 16:00:31 +0100	[thread overview]
Message-ID: <26b64d48-5ff9-7d62-bc44-601fdcc43223@kernel.org> (raw)
In-Reply-To: <01fa98c1-8274-445c-5e04-219372920ba2@arm.com>

On 17/07/2019 14:44, Julien Thierry wrote:
> Hi Zenghui,
> 
> On 17/07/2019 13:20, Zenghui Yu wrote:
>> We use "pmc->idx" and the "chained" bitmap to determine if the pmc is
>> chained, in kvm_pmu_pmc_is_chained().  But idx might be uninitialized
>> (and random) when we doing this decision, through a KVM_ARM_VCPU_INIT
>> ioctl -> kvm_pmu_vcpu_reset(). And the test_bit() against this random
>> idx will potentially hit a KASAN BUG [1].
>>
>> Fix it by moving the assignment of idx before kvm_pmu_stop_counter().
>>
>> [1] https://www.spinics.net/lists/kvm-arm/msg36700.html
>>
>> Fixes: 80f393a23be6 ("KVM: arm/arm64: Support chained PMU counters")
>> Suggested-by: Andrew Murray <andrew.murray@arm.com>
>> Cc: Marc Zyngier <maz@kernel.org>
>> Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>> ---
>>  virt/kvm/arm/pmu.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/virt/kvm/arm/pmu.c b/virt/kvm/arm/pmu.c
>> index 3dd8238..521bfdd 100644
>> --- a/virt/kvm/arm/pmu.c
>> +++ b/virt/kvm/arm/pmu.c
>> @@ -225,8 +225,8 @@ void kvm_pmu_vcpu_reset(struct kvm_vcpu *vcpu)
>>  	struct kvm_pmu *pmu = &vcpu->arch.pmu;
>>  
>>  	for (i = 0; i < ARMV8_PMU_MAX_COUNTERS; i++) {
>> -		kvm_pmu_stop_counter(vcpu, &pmu->pmc[i]);
>>  		pmu->pmc[i].idx = i;
> 
> Yes, this is kind of a static property that should really be part of a
> "kvm_pmu_vcpu_init()" or "kvm_pmu_vcpu_create()" and is not expected to
> be modified across resets...
> 
> There is no such function at the time and I'm unsure whether this
> warrants creating that separate function (I would still suggest creating
> it to make things clearer).

Yup, that's pretty bad, now that you mention it. I'd be all for the
introduction of kvm_pmu_vcpu_init(), given that we already have
kvm_pmu_vcpu_destroy().

> 
>> +		kvm_pmu_stop_counter(vcpu, &pmu->pmc[i]);
> 
> Whatever other opinions are on splitting pmu_vcpu_init/reset, that
> change makes sense and fixes the issue:
> 
> Acked-by: Julien Thierry <julien.thierry@arm.com>
> 
>>  	}
>>  
>>  	bitmap_zero(vcpu->arch.pmu.chained, ARMV8_PMU_MAX_COUNTER_PAIRS);
>>
> 
> Cheers,
> 

Zenghui, could you please update your patch to take the above into account?

Thanks,

	M.
-- 
Jazz is not dead, it just smells funny...

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel