Re: [RFC PATCH v3 11/11] tests: add plugin asserting correctness of discon event's to_pc

[Pierrick Bouvier <pierrick.bouvier@linaro.org>]

On 12/5/24 13:22, Julian Ganz wrote:
> Hi Pierrick,
> 
> December 5, 2024 at 6:30 PM, "Pierrick Bouvier" wrote:
>> On 12/5/24 05:10, Julian Ganz wrote:
>>>   December 5, 2024 at 12:33 AM, "Pierrick Bouvier" wrote:
>>>> Trap target PC mismatch
>>>>   Expected: 23faf3a80
>>>>   Encountered: 23faf3a84
>>>>
>>>>   From what I understand, it means that the next_pc we have is incorrect.
>>>>
>>>   Yes, this is indeed incorrect, and also a perfect example why this test
>>>   plugin exists. There are likely other errors lurking in target specific
>>>   code. Did you happen to also log interrupts? Do you remember what image
>>>   you used?
>>>
>> I used exactly this:
>>
>> ./build/qemu-system-aarch64 -plugin ./build/tests/tcg/plugins/libdiscons.so -smp 4 -M virt -d plugin -m 8G -device virtio-blk-pci,drive=root -drive if=none,id=root,file=./debianaarch64.img -M virt -cpu max,pauth=off -drive if=pflash,readonly=on,file=/usr/share/AAVMF/AAVMF_CODE.fd -drive if=pflash,file=./AAVMF_VARS.fd
>>
>> The arm64 image is a vanilla debian stable I installed.
>> AAVMF* files come from qemu-efi-aarch64 debian package.
> 
> Thanks! I will have a closer look and include a fix in the next iteration.
> 
>>> Btw: this also illustrates another issue I have with from_pc: we can
>>>   test the behavior for to_pc, but doing this meaningfully for from_pc
>>>   via a plugin is next to impossible because the instruction it points to
>>>   is not observable via an exec callback. At least not in the general
>>>   case, even not if we only consider a single type of event.
>>>
>> We can store the next_expected pc for each instruction (from current_instruction + insn_length), and we should be able to compare that with the expected from_pc.
>> This is mostly what contrib/plugins/cflow.c does.
>>
>> With that, we can test from_pc.
> 
> I'm not confident that this will work reliably for branch, jump and
> other "interesting" instructions. But I can have a closer look at the
> cflow plugin and try to figure out how that plugin handles those cases.
> 

It won't work for latest instructions in a tb (because we don't know 
what will be the next one), but should be good for all the others cases.

> Regards,
> Julian Ganz