Re: [PULL 15/32] tests/functional/aarch64: add tests for FEAT_RME

[Alex Bennée]

Pierrick Bouvier <pierrick.bouvier@linaro.org> writes:

> Hi Thomas,
>
> On 1/24/25 05:29, Thomas Huth wrote:
>> On 10/01/2025 14.17, Alex Bennée wrote:
>>> From: Pierrick Bouvier <pierrick.bouvier@linaro.org>
>>>
>>> This boot an OP-TEE environment, and launch a nested guest VM inside it
>>> using the Realms feature. We do it for virt and sbsa-ref platforms.
>>    Hi,
>> FWIW, I just saw this test_aarch64_rme_virt test failing once.
>> Looking at
>> the console.log, it seems like the guest crashed somewhere at the end of the
>> firmware or very early in the kernel:
>> EFI stub: Booting Linux Kernel...
>> 2025-01-24 13:25:28,226:
>> 2025-01-24 13:25:28,226:
>> 2025-01-24 13:25:28,227: Synchronous Exception at 0x00000000BF526498
>> 2025-01-24 13:25:28,227:
>> 2025-01-24 13:25:28,227:
>> 2025-01-24 13:25:28,228: Synchronous Exception at 0x00000000BF526498
>> I wasn't able to reproduce it afterwards anymore, but in case
>> somebody runs
>> into this later again and wonders if it is a later regression: It happened
>> for me with commit cf86770c7aa31ebd.
>>    Thomas
>> 
>
> I met this kind of error before (not on rme), and if I remember well,
> it's coming from EDK2.
>
> We might have a race condition here.

I'm measuring a 1-3% hit rates:

  retry.py -n 400 -c -- ./pyvenv/bin/meson test --setup thorough --suite 
func-thorough func-aarch64-aarch64_rme_virt func-aarch64-aarch64_rme_sbsaref -v

Gave:

  Results summary:
  0: 388 times (97.00%), avg time 196.924 (7.99 varience/2.83 deviation)
  1: 12 times (3.00%), avg time 1205.900 (0.00 varience/0.01 deviation)
  Ran command 400 times, 388 passes

It might be worth trying it under rr --chaos mode and seeing if that
shakes anything out.

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro