Re: [PATCH] x86/bhi: BHI mitigation can trigger warning in #DB handler

[Alexandre Chartre <alexandre.chartre@oracle.com>]

On 5/23/24 20:27, Alexandre Chartre wrote:
> 
> 
> On 5/23/24 19:53, Andrew Cooper wrote:
>> On 23/05/2024 6:03 pm, Alexandre Chartre wrote:
>>>
>>> On 5/23/24 17:36, Dave Hansen wrote:
>>>> On 5/23/24 07:52, Alexandre Chartre wrote:
>>>>>> Should we wrap up this gem and put it with the other entry selftests?
>>>>>
>>>>> It looks like tools/testing/selftests/x86/single_step_syscall.c tests
>>>>> sysenter with TF set but it doesn't check if the kernel issues any
>>>>> warning.
>>>>
>>>> Does it actually trip the warning though? I'm a bit surprised that
>>>> nobody reported it if so.
>>>
>>> single_step_syscall does trigger the warning:
>>>
>>> $ ./single_step_syscall
>>> [RUN]    Set TF and check nop
>>> [OK]    Survived with TF set and 26 traps
>>> [RUN]    Set TF and check syscall-less opportunistic sysret
>>> [OK]    Survived with TF set and 30 traps
>>> [RUN]    Set TF and check a fast syscall
>>> [OK]    Survived with TF set and 40 traps
>>> [RUN]    Fast syscall with TF cleared
>>> [OK]    Nothing unexpected happened
>>> [RUN]    Set TF and check SYSENTER
>>>      Got SIGSEGV with RIP=ed7fe579, TF=256
>>> [RUN]    Fast syscall with TF cleared
>>> [OK]    Nothing unexpected happened
>>
>> :-/
>>
>> What about the exit code?
>>
>> I find the absence of a [FAIL] concerning...
>>
> 
> $ ./single_step_syscall
> [RUN]    Set TF and check nop
> [OK]    Survived with TF set and 26 traps
> [RUN]    Set TF and check syscall-less opportunistic sysret
> [OK]    Survived with TF set and 30 traps
> [RUN]    Set TF and check a fast syscall
> [OK]    Survived with TF set and 40 traps
> [RUN]    Fast syscall with TF cleared
> [OK]    Nothing unexpected happened
> [RUN]    Set TF and check SYSENTER
>      Got SIGSEGV with RIP=bb44b579, TF=256
> [RUN]    Fast syscall with TF cleared
> [OK]    Nothing unexpected happened
> 
> $ echo $?
> 0
> 
> The program runs as expected (but it doesn't expect much than a SIGSEGV).
> It triggers a warning from the kernel but it doesn't check if a warning
> was produced.
> 

Actually it checks that TF isn't cleared after the sysenter, but that's all.

alex.