[PATCH i-g-t] runner: Relax timeout reduction on soft lockup

[Kamil Konieczny]

Hi Janusz,
On 2025-07-04 at 15:30:35 +0200, Janusz Krzysztofik wrote:
> In case of soft lockups, it might be helpful from root cause analysis
> perspective to see if the test was still able to complete despite
> triggering the soft lockup warning, or if that soft lockup seems not
> recoverable without killing the test. For that to be possible, igt_runner
> should not kill the test too promptly if a soft lockup related kernel
> taint is detected.
> 
> On kernel taints, igt_runner now decreases per test and inactivity
> timeouts by a factor of 10.  Let it check if the taint is caused by a
> soft lockup and decrease the timeouts only by the factor of 2 in those
> cases.
> 
> Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik at linux.intel.com>
> ---
>  runner/executor.c | 13 +++++++++----
>  1 file changed, 9 insertions(+), 4 deletions(-)
> 
> diff --git a/runner/executor.c b/runner/executor.c
> index 13180a0a46..de9d29d28d 100644
> --- a/runner/executor.c
> +++ b/runner/executor.c
> @@ -871,10 +871,14 @@ static const char *need_to_timeout(struct settings *settings,
>  	if (settings->abort_mask & ABORT_TAINT &&
>  	    is_tainted(taints)) {
>  		/* list of timeouts that may postpone immediate kill on taint */
> -		if (settings->per_test_timeout || settings->inactivity_timeout)
> -			decrease = 10;
> -		else
> +		if (settings->per_test_timeout || settings->inactivity_timeout) {
> +			if (is_tainted(taints) == (1 << 9) && taints & (1 << 14))

Looks reasonable, imho there should be #define or constants
for those (1 << 9) and (1 << 14), at least for these and maybe
also other bits.

Regards,
Kamil

> +				decrease = 2;	/* only warn + soft lockup */
> +			else
> +				decrease = 10;
> +		} else {
>  			return "Killing the test because the kernel is tainted.\n";
> +		}
>  	}
>  
>  	if (settings->per_test_timeout != 0 &&
> @@ -1526,8 +1530,9 @@ static int monitor_output(pid_t child,
>  			sigfd = -1; /* we are dying, no signal handling for now */
>  		}
>  
> +		igt_kernel_tainted(&taints);
>  		timeout_reason = need_to_timeout(settings, killed,
> -						 igt_kernel_tainted(&taints),
> +						 taints,
>  						 igt_time_elapsed(&time_last_activity, &time_now),
>  						 igt_time_elapsed(&time_last_subtest, &time_now),
>  						 igt_time_elapsed(&time_killed, &time_now),
> -- 
> 2.50.0
>