Re: [PATCH v2 3/6] kvm: x86: Emulate split-lock access as a write

[Xiaoyao Li <xiaoyao.li@intel.com>]

On 2/4/2020 4:54 AM, Sean Christopherson wrote:
> On Mon, Feb 03, 2020 at 11:16:05PM +0800, Xiaoyao Li wrote:
>> If split lock detect is enabled (warn/fatal), #AC handler calls die()
>> when split lock happens in kernel.
>>
>> A sane guest should never tigger emulation on a split-lock access, but
>> it cannot prevent malicous guest from doing this. So just emulating the
>> access as a write if it's a split-lock access to avoid malicous guest
>> polluting the kernel log.
>>
>> More detail analysis can be found:
>> https://lkml.kernel.org/r/20200131200134.GD18946@linux.intel.com
>>
>> Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
>> ---
>>   arch/x86/kvm/x86.c | 11 +++++++++++
>>   1 file changed, 11 insertions(+)
>>
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index 2d3be7f3ad67..821b7404c0fd 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -5847,6 +5847,13 @@ static int emulator_write_emulated(struct x86_emulate_ctxt *ctxt,
>>   	(cmpxchg64((u64 *)(ptr), *(u64 *)(old), *(u64 *)(new)) == *(u64 *)(old))
>>   #endif
>>   
>> +static inline bool across_cache_line_access(gpa_t gpa, unsigned int bytes)
> 
> s/across/split so as not to introduce another name.
> 
>> +{
>> +	unsigned int cache_line_size = cache_line_size();
>> +
>> +	return (gpa & (cache_line_size - 1)) + bytes > cache_line_size;
> 
> I'd prefer to use the same logic as the page-split to avoid having to
> reason about the correctness of two different algorithms.
> 
>> +}
>> +
>>   static int emulator_cmpxchg_emulated(struct x86_emulate_ctxt *ctxt,
>>   				     unsigned long addr,
>>   				     const void *old,
>> @@ -5873,6 +5880,10 @@ static int emulator_cmpxchg_emulated(struct x86_emulate_ctxt *ctxt,
>>   	if (((gpa + bytes - 1) & PAGE_MASK) != (gpa & PAGE_MASK))
>>   		goto emul_write;
>>   
>> +	if (get_split_lock_detect_state() != sld_off &&
>> +	    across_cache_line_access(gpa, bytes))
>> +		goto emul_write;
> 
> As an alternative to the above, the page/line splits can be handled in a
> single check, e.g.
> 
> 	page_line_mask = PAGE_MASK;
> 	if (is_split_lock_detect_enabled())
> 		page_line_mask = cache_line_size() - 1;
> 	if (((gpa + bytes - 1) & page_line_mask) != (gpa & page_line_mask))
> 		goto emul_write;

It's better, will use your suggestion.
Thanks!

>> +
>>   	if (kvm_vcpu_map(vcpu, gpa_to_gfn(gpa), &map))
>>   		goto emul_write;
>>   
>> -- 
>> 2.23.0
>>