[igt-dev] [PATCH i-g-t] i915/gem_exec_create: Added test description for test case

[Chris Wilson]

Quoting sai.gowtham.ch at intel.com (2020-06-30 07:41:10)
> From: sai gowtham <sai.gowtham.ch at intel.com>
> 
> Added test description for test and to all 3 subtests.
> 
> Signed-off-by: sai gowtham <sai.gowtham.ch at intel.com>
> ---
>  tests/i915/gem_exec_create.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/i915/gem_exec_create.c b/tests/i915/gem_exec_create.c
> index ee7d4e5d..5236d477 100644
> --- a/tests/i915/gem_exec_create.c
> +++ b/tests/i915/gem_exec_create.c
> @@ -42,6 +42,10 @@
>  #include "i915/gem.h"
>  #include "igt.h"
>  
> +IGT_TEST_DESCRIPTION("\n This test runs an uninterrupted loop within the specified timeouts"
> +               " with different flags & ncpus combinations and checks if there are any"
> +               " missed interrupts reported from the driver.");

No. The intent of the test is to flood the driver with transient active
objects and check we don't kill the system under the mempressure. Some
of the symptoms we look for include mysterious hangs.

> +
>  #define ENGINE_FLAGS  (I915_EXEC_RING_MASK | I915_EXEC_BSD_MASK)
>  
>  static double elapsed(const struct timespec *start, const struct timespec *end)
> @@ -133,12 +137,16 @@ igt_main
>  
>                 igt_fork_hang_detector(device);
>         }
> -

Lost visual whitespacing.

> +       igt_describe("Runs single uninterrupted loop for 2s.");
>         igt_subtest("basic")
>                 all(device, 0, 2, 1);
> +
> +       igt_describe("Runs single uninterrupted loop for 20s.");
>         igt_subtest("forked")
>                 all(device, 0, 20, ncpus);
>  
> +       igt_describe("Runs single uninterrupted loop with"
> +                       " different flag combination for 20s.");

That is abysmal. We already know the above and more from the test setup
itself. The difference in the test intentions should be discernible from
the different path coverage in the driver...

Quick sanitycheck; concurrent execution; forced reclaim, behaving like a
bad application leaking its bo cache.

>         igt_subtest("madvise")
>                 all(device, LEAK, 20, 1);