[igt-dev] [PATCH i-g-t v2 4/5] runner: New test runner

Petri Latvala petri.latvala at intel.com
Wed Jun 13 10:50:54 UTC 2018



On 06/13/2018 01:20 PM, Chris Wilson wrote:
> Quoting Petri Latvala (2018-06-13 11:07:38)
>> +static int monitor_output(pid_t child,
>> +                          int outfd, int errfd, int kmsgfd, int sigfd,
>> +                          int *outputs,
>> +                          struct settings *settings)
> Why? Just integrate watchdog support into the test runner (igt).

You mean into the tests?

> No watching output required, if we have a long running test that
> produces no output, it can ping the watchdog itself. If the test itself
> should not use the watchdog as it only uses the kernel watchdog, again
> no issue.

If the test jams hard, it's unable to ping watchdogs, and we'd like to 
be able to kill the test in that case, rebooting (via watchdogs) only as 
the last resort.

If the test doesn't produce any output itself or in kernel messages, an 
interface could be added where tests could kill(getppid(), SIGUSR1) to 
tell the runner they're still alive.


Petri



More information about the igt-dev mailing list