Re: [regression] suspend stress test stalls within 30 minutes

[Dave Hansen]

On 5/16/24 07:25, Pawan Gupta wrote:
> On Thu, May 16, 2024 at 12:03:22AM -0700, Pawan Gupta wrote:
>> I am running the suspend test now and will update in the morning if I
>> could reproduce the hang.
> Completed 500 suspend iterations, but the hang is not reproduced 🙁
> I have restarted the test.
> 
> BTW, could you please share your /proc/cmdline? Also, was there any
> workload running with the suspend test? If I could not reproduce it in the
> next run, I will run stress-ng with the suspend test.

I'd suggest two things:

  * Run everything in tools/testing/selftests/x86 in a loop during the
    suspend test.
  * Run perf or something else to generate some NMIs.

Those will ensure the dark corners of the entry code are being
sufficiently prodded.

I also have sneaking suspicion that microcode updates during resume are
the aggravating factor.