[igt-dev] [PATCH i-g-t 3/3] tests/i915_hangman: add multigpu basic error state subtest

[Mauro Carvalho Chehab]

On Thu, 20 Apr 2023 20:24:40 +0200
Kamil Konieczny <kamil.konieczny at linux.intel.com> wrote:

> Verify that basic error capture works in multiGPU hardware
> configuration for two or more GPUs.
> 
> Signed-off-by: Kamil Konieczny <kamil.konieczny at linux.intel.com>

LGTM.

Reviewed-by: Mauro Carvalho Chehab <mchehab at kernel.org>
> ---
>  tests/i915/i915_hangman.c | 63 ++++++++++++++++++++++++++++++++++-----
>  1 file changed, 55 insertions(+), 8 deletions(-)
> 
> diff --git a/tests/i915/i915_hangman.c b/tests/i915/i915_hangman.c
> index d76d0a037..580b3a739 100644
> --- a/tests/i915/i915_hangman.c
> +++ b/tests/i915/i915_hangman.c
> @@ -39,6 +39,7 @@
>  #include "igt_store.h"
>  #include "igt_sysfs.h"
>  #include "igt_debugfs.h"
> +#include "igt_device_scan.h"
>  #include "sw_sync.h"
>  
>  #ifndef I915_PARAM_CMD_PARSER_VERSION
> @@ -514,6 +515,21 @@ static void do_tests(int device, int sysfs, const char *name, const char *prefix
>  	}
>  }
>  
> +static int set_engines_prop(int device, struct gem_engine_properties *params)
> +{
> +	const struct intel_execution_engine2 *e;
> +	int num = 0;
> +
> +	for_each_physical_engine(device, e) {
> +		params[num].engine = e;
> +		params[num].preempt_timeout = 500;
> +		params[num].heartbeat_interval = 1000;
> +		gem_engine_properties_configure(device, params + num++);
> +	}
> +
> +	return num;
> +}
> +
>  igt_main
>  {
>  	const intel_ctx_t *ctx;
> @@ -522,10 +538,9 @@ igt_main
>  	int num_engines = 0;
>  	int sysfs_fd = -1;
>  	int i915 = -1;
> +	int gpu_count;
>  
>  	igt_fixture {
> -		const struct intel_execution_engine2 *e;
> -
>  		i915 = drm_open_driver(DRIVER_INTEL);
>  		igt_require_gem(i915);
>  
> @@ -540,12 +555,8 @@ igt_main
>  
>  		gem_require_mmap_device_coherent(i915);
>  
> -		for_each_physical_engine(i915, e) {
> -			saved_params[num_engines].engine = e;
> -			saved_params[num_engines].preempt_timeout = 500;
> -			saved_params[num_engines].heartbeat_interval = 1000;
> -			gem_engine_properties_configure(i915, saved_params + num_engines++);
> -		}
> +		num_engines = set_engines_prop(i915, &saved_params[0]);
> +		gpu_count = igt_device_filter_count();
>  	}
>  
>  	igt_describe("Basic error capture");
> @@ -576,6 +587,42 @@ igt_main
>  
>  	do_tests(i915, sysfs_fd, "engine", "engine", ctx);
>  
> +	igt_describe("Basic multigpu error capture");
> +	igt_subtest("multigpu-error-basic") {
> +		igt_require(gpu_count > 1);
> +
> +		igt_multi_fork(child, gpu_count) {
> +			int gpu_fd, gpu_sysfs;
> +
> +			if (child)
> +				gpu_fd = __drm_open_driver_another(child, DRIVER_INTEL);
> +			else
> +				gpu_fd = gem_reopen_driver(i915);
> +
> +			igt_assert(gpu_fd > 0);
> +			igt_require_gem(gpu_fd);
> +			ctx = intel_ctx_create_all_physical(gpu_fd);
> +			hang = igt_allow_hang(gpu_fd, ctx->id, HANG_ALLOW_CAPTURE | HANG_WANT_ENGINE_RESET);
> +			gpu_sysfs = igt_sysfs_open(gpu_fd);
> +			igt_assert(gpu_sysfs != -1);
> +			igt_require(has_error_state(gpu_sysfs));
> +			gem_require_mmap_device_coherent(gpu_sysfs);
> +			num_engines = set_engines_prop(gpu_fd, &saved_params[0]);
> +
> +			test_error_state_basic(gpu_fd, gpu_sysfs);
> +
> +			igt_disallow_hang(gpu_fd, hang);
> +			intel_ctx_destroy(gpu_fd, ctx);
> +			if (child)
> +				for (int i = 0; i < num_engines; i++)
> +					gem_engine_properties_restore(i915, saved_params + i);
> +			close(gpu_fd);
> +			close(gpu_sysfs);
> +		}
> +
> +		igt_waitchildren();
> +	}
> +
>  	igt_fixture {
>  		int i;
>