Re: [PATCH] arm64: cpufeature: make sure cpu_mitigations_off then kpti off

["liusong" <liusong@linux.alibaba.com>]

>On Thu, Jun 16, 2022 at 08:35:38PM +0800, Liu Song wrote:
>> From: Liu Song <liusong@linux.alibaba.com>
>>
>> If kaslr is enabled, kpti cannot be turned off even if "mitigations=off",
>> adjust the code order to ensure that kpti is off when "mitigations=off".
>>
>> Signed-off-by: Liu Song <liusong@linux.alibaba.com>
>> ---
>>  arch/arm64/kernel/cpufeature.c | 10 +++++-----
>>  1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
>> index 42ea2bd..27d4850 100644
>> --- a/arch/arm64/kernel/cpufeature.c
>> +++ b/arch/arm64/kernel/cpufeature.c
>> @@ -1616,6 +1616,11 @@ static bool unmap_kernel_at_el0(const struct arm64_cpu_capabilities *entry,
>>    __kpti_forced = -1;
>>   }
>> 
>> + if (cpu_mitigations_off() && !__kpti_forced) {
>> +  str = "mitigations=off";
>> +  __kpti_forced = -1;
>> + }
>> +
>>   /* Useful for KASLR robustness */
>>   if (kaslr_requires_kpti()) {
>>    if (!__kpti_forced) {
>> @@ -1624,11 +1629,6 @@ static bool unmap_kernel_at_el0(const struct arm64_cpu_capabilities *entry,
>>    }
>>   }
>> 
>> - if (cpu_mitigations_off() && !__kpti_forced) {
>> -  str = "mitigations=off";
>> -  __kpti_forced = -1;
>> - }
>
>The current behaviour is intentional: KASLR is trivial to bypass when kpti
>is disabled.
>
>Will

Hi

According to the description of mitigaions=off in kernel-parameters.txt,
it is equivalent to kpti=0 under ARM64, so the description here will mislead
users. I have proposed a patch to fix this.

Thanks