Re: [PATCH i-g-t] tests/intel/i915_module_load: Ignore missing PCI reset

Hi Janusz,
On 2026-08-20 at 20:41:43 +0200, Janusz Krzysztofik wrote:
> Subtest fault-injection now performs PCI device reset before every driver
> bind attempt and fails if that reset doesn't succeed.  That was OK if all
> devices being tested supported a PCI reset method.  However, machines
> with DG1 SKU 0x4905 used in CI suffer from device breakage and system
> freezes after any PCI reset attempts, triggering CI job aborts.  That's
> going to change soon with an upcoming kernel patch that disables PCI
> reset capabilities on that card.
> 
> Attempt PCI reset and fail if not successful only if supported, otherwise
> submit a debug message and continue the exercise.  With that in place,
> the subtest should no longer abort CI jobs on DG1 as soon as the above
> mentioned kernel fix is in place.
> 
> Signed-off-by: Janusz Krzysztofik <[email protected]>
LGTM
Reviewed-by: Kamil Konieczny <[email protected]>

Regards,
Kamil

> ---
>  tests/intel/i915_module_load.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/intel/i915_module_load.c b/tests/intel/i915_module_load.c
> index 26e30a1000..68e2386465 100644
> --- a/tests/intel/i915_module_load.c
> +++ b/tests/intel/i915_module_load.c
> @@ -463,7 +463,8 @@ inject_fault_probe(int fd, const char pci_slot[], const char function_name[],
>  	injection_list_add(function_name);
>  	set_retval(function_name, inject_error);
>  
> -	igt_assert(igt_sysfs_set(devicefd, "reset", "1"));
> +	if (!igt_debug_on(!igt_sysfs_has_attr(devicefd, "reset")))
> +		igt_assert(igt_sysfs_set(devicefd, "reset", "1"));
>  	igt_kmod_bind("i915", pci_slot);
>  
>  	err = -errno;
> -- 
> 2.54.0
>