Re: [PATCH 2/5] MdePkg/BaseLib: add ASSERT in ARM* SetJump implementations

["Ard Biesheuvel" <ard.biesheuvel@arm.com>]

On 10/1/20 10:55 PM, Leif Lindholm wrote:
> On Thu, Oct 01, 2020 at 22:49:05 +0200, Ard Biesheuvel wrote:
>> On 10/1/20 8:37 PM, Leif Lindholm wrote:
>>> The SetJump comment header states that:
>>>     If JumpBuffer is NULL, then ASSERT().
>>>
>>> However, this was not currently done.
>>> Add a call to InternalAssertJumpBuffer.
>>>
>>> Signed-off-by: Leif Lindholm <leif@nuviainc.com>
>>> ---
>>>    MdePkg/Library/BaseLib/AArch64/SetJumpLongJump.S   | 3 +++
>>>    MdePkg/Library/BaseLib/AArch64/SetJumpLongJump.asm | 3 +++
>>>    MdePkg/Library/BaseLib/Arm/SetJumpLongJump.S       | 3 +++
>>>    MdePkg/Library/BaseLib/Arm/SetJumpLongJump.asm     | 3 +++
>>>    4 files changed, 12 insertions(+)
>>>
>>> diff --git a/MdePkg/Library/BaseLib/AArch64/SetJumpLongJump.S b/MdePkg/Library/BaseLib/AArch64/SetJumpLongJump.S
>>> index 989736cee74c..34765a676430 100644
>>> --- a/MdePkg/Library/BaseLib/AArch64/SetJumpLongJump.S
>>> +++ b/MdePkg/Library/BaseLib/AArch64/SetJumpLongJump.S
>>> @@ -45,6 +45,9 @@ GCC_ASM_EXPORT(InternalLongJump)
>>>    #  );
>>>    #
>>>    ASM_PFX(SetJump):
>>> +        stp     x30, x0, [sp, #-16]!
>>> +        bl      InternalAssertJumpBuffer
>>> +        ldp     x30, x0, [sp], #16
>>
>> I think we should make this more idiomatic for AArch64 function calls, i.e.
>>
>>          stp     x29, x30, [sp, #-32]!
>>          mov     x29, sp
>>          str     x0, [sp, #16]
>>          bl      InternalAssertJumpBuffer
>>          ldr     x0, [sp, #16]
>>          ldp     x29, x30, [sp], #32
>>
>> That way, we'll have a well formed call stack with all the frame records
>> linked together, allowing the debugger to show you where SetJump() was
>> called from to begin with if you are stuck in the deadloop or hit the
>> breakpoint (depending on how the PCD was configured to begin with)
> 
> Mmm, you have a point.
> 
>> I wouldn't mind putting the whole thing inside #ifndef MDEPKG_NDEBUG /#endif
>> btw
> 
> Well...
> At that point we might as well go and un-bonkers-ify the interfaces and
> make the C function the wrapper that does the assert check before
> calling this function renamed to InternalSetJump - making it symmetric
> with the LongJump side of the equation.
> Which I was hoping to avoid, since that messes with all architectures.
> 
> Urgh, that is actually the sensible thing to do here, isn't it?
> 

Do rhetorical questions expect an answer?