Re: [PATCH v7 7/10] KVM: arm/arm64: context-switch ptrauth registers

[James Morse <james.morse@arm.com>]

Hi Amit, Kristina,

On 27/03/2019 03:21, Amit Daniel Kachhap wrote:
> On 3/26/19 11:31 PM, Kristina Martsenko wrote:
>> On 26/03/2019 04:03, Amit Daniel Kachhap wrote:
>>> On 3/26/19 1:34 AM, Kristina Martsenko wrote:
>>>> On 19/03/2019 08:30, Amit Daniel Kachhap wrote:
>>>>> From: Mark Rutland <mark.rutland@arm.com>
>>>>>
>>>>> When pointer authentication is supported, a guest may wish to use it.
>>>>> This patch adds the necessary KVM infrastructure for this to work, with
>>>>> a semi-lazy context switch of the pointer auth state.

>>>>> +    if (test_bit(KVM_ARM_VCPU_PTRAUTH_ADDRESS, vcpu->arch.features) ||
>>>>> +        test_bit(KVM_ARM_VCPU_PTRAUTH_GENERIC, vcpu->arch.features)) {>>>>> +        /* Verify that KVM startup matches the conditions for ptrauth */
>>>>> +        if (WARN_ON(!vcpu_has_ptrauth(vcpu)))
>>>>> +            return -EINVAL;
>>>>> +    }

>> One more thing - I think the WARN_ON() here should be removed. Otherwise
>> if panic_on_warn is set then userspace can panic the kernel. I think
>> WARN_ON is only for internal kernel errors (see comment in
>> include/asm-generic/bug.h).
> 
> The documentation makes sense so in this case a pr_err like message will suffice.

(could it be a kvm_debug() at most?)

Do we need to print anything at all? User-space asked us for something we can't do.
Filling up the kernel log with user-space's mistakes makes it harder to debug the kernel
when something goes wrong.

kvm_arm_pmu_v3_init() returns -ENODEV if you ask if for the PMU and the platform can't
support it. Isn't the returned error enough?


> Btw
> there is one WARN in the function kvm_set_ipa_limit in the same file.

That is called once via kvm_arch_init(), it can't be triggered repeatedly from user-space.


Thanks,

James