Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH v3 4/7] arm64: add sysfs vulnerability show for meltdown
To:     Jeremy Linton <jeremy.linton@arm.com>,
        linux-arm-kernel@lists.infradead.org
Cc:     catalin.marinas@arm.com, will.deacon@arm.com, marc.zyngier@arm.com,
        suzuki.poulose@arm.com, dave.martin@arm.com,
        shankerd@codeaurora.org, linux-kernel@vger.kernel.org,
        ykaukab@suse.de, mlangsdo@redhat.com, steven.price@arm.com,
        stefan.wahren@i2se.com
References: <20190109235544.2992426-1-jeremy.linton@arm.com>
 <20190109235544.2992426-5-jeremy.linton@arm.com>
 <8c8b564a-1d65-bc18-73fb-58b349a47800@arm.com>
 <7e843245-56c0-d7c1-38ba-27ff231a500a@arm.com>
From:   Julien Thierry <julien.thierry@arm.com>
Message-ID: <fbcfae96-6cca-2a7a-a91b-8ffcc1a4c803@arm.com>
Date:   Thu, 10 Jan 2019 14:16:01 +0000
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.2.1
MIME-Version: 1.0
In-Reply-To: <7e843245-56c0-d7c1-38ba-27ff231a500a@arm.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 10/01/2019 14:10, Jeremy Linton wrote:
> Hi Julien,
> 
> On 01/10/2019 03:23 AM, Julien Thierry wrote:
>> Hi Jeremy,
>>
>> On 09/01/2019 23:55, Jeremy Linton wrote:
>>> Display the mitigation status if active, otherwise
>>> assume the cpu is safe unless it doesn't have CSV3
>>> and isn't in our whitelist.
>>>
>>> Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
>>> ---
>>>   arch/arm64/kernel/cpufeature.c | 32 +++++++++++++++++++++++++++-----
>>>   1 file changed, 27 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/arch/arm64/kernel/cpufeature.c
>>> b/arch/arm64/kernel/cpufeature.c
>>> index ab784d7a0083..ef7bbc49ef78 100644
>>> --- a/arch/arm64/kernel/cpufeature.c
>>> +++ b/arch/arm64/kernel/cpufeature.c
>>> @@ -944,8 +944,12 @@ has_useable_cnp(const struct
>>> arm64_cpu_capabilities *entry, int scope)
>>>       return has_cpuid_feature(entry, scope);
>>>   }
>>>   +/* default value is invalid until unmap_kernel_at_el0() runs */
>>> +static bool __meltdown_safe = true;
>>> +
>>>   #ifdef CONFIG_UNMAP_KERNEL_AT_EL0
>>>   static int __kpti_forced; /* 0: not forced, >0: forced on, <0:
>>> forced off */
>>> +extern uint arm64_requested_vuln_attrs;
>>>     static bool is_cpu_meltdown_safe(void)
>>>   {
>>> @@ -972,6 +976,14 @@ static bool unmap_kernel_at_el0(const struct
>>> arm64_cpu_capabilities *entry,
>>>   {
>>>       char const *str = "command line option";
>>>   +    bool meltdown_safe = is_cpu_meltdown_safe() ||
>>> +        has_cpuid_feature(entry, scope);
>>> +
>>> +    if (!meltdown_safe)
>>> +        __meltdown_safe = false;
>>> +
>>> +    arm64_requested_vuln_attrs |= VULN_MELTDOWN;
>>> +
>>>       /*
>>>        * For reasons that aren't entirely clear, enabling KPTI on Cavium
>>>        * ThunderX leads to apparent I-cache corruption of kernel
>>> text, which
>>> @@ -993,11 +1005,7 @@ static bool unmap_kernel_at_el0(const struct
>>> arm64_cpu_capabilities *entry,
>>>       if (IS_ENABLED(CONFIG_RANDOMIZE_BASE))
>>>           return true;
>>>   -    if (is_cpu_meltdown_safe())
>>> -        return false;
>>> -
>>> -    /* Defer to CPU feature registers */
>>> -    return !has_cpuid_feature(entry, scope);
>>> +    return !meltdown_safe;
>>>   }
>>>     static void
>>> @@ -2065,3 +2073,17 @@ static int __init enable_mrs_emulation(void)
>>>   }
>>>     core_initcall(enable_mrs_emulation);
>>> +
>>> +#ifdef CONFIG_GENERIC_CPU_VULNERABILITIES
>>> +ssize_t cpu_show_meltdown(struct device *dev, struct
>>> device_attribute *attr,
>>> +        char *buf)
>>> +{
>>> +    if (arm64_kernel_unmapped_at_el0())
>>> +        return sprintf(buf, "Mitigation: KPTI\n");
>>> +
>>> +    if (__meltdown_safe)
>>> +        return sprintf(buf, "Not affected\n");
>>
>> An issue I see is that we don't even bother to check it that CPUs are
>> meltdown safe if CONFIG_UNMAP_KERNEL_AT_EL0 is not defined but here
>> we'll advertise that the system is meltdown safe.
> 
> That check isn't necessary anymore because the sysfs attribute is only
> populated if unmap_kernel_at_el0() runs (assuming I haven't messed
> something up). That was Dave/Will's suggestions in the last thread about
> how to handle this case.
> 

Oh right, I missed that bit. Sorry for the noise.

>>
>> I think that checking whether we know that CPUs are meltdown safe should
>> be separated from whether mitigation is applied.
>>
>> Someone who knows thinks their CPUs are in the white list might want to
>> compile out code that does the kpti, but it would be good to give them a
>> proper diagnostic whether they were wrong or not.
>>
>> Cheers,
>>
> 

-- 
Julien Thierry