Re: [Intel-xe] [PATCH v5 2/2] drm/xe: Introduce fault injection for gt reset

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Tue, Jul 18, 2023 at 07:02:16PM +0530, Himal Prasad Ghimiray wrote:
> To trigger gt reset failure:
>  echo 100 >  /sys/kernel/debug/dri/<cardX>/fail_gt_reset/probability
>  echo 2 >  /sys/kernel/debug/dri/<cardX>/fail_gt_reset/times

why 2 and not 1?

anyway, neat solution!

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

> 
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Cc: Lucas De Marchi <lucas.demarchi@intel.com>
> Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_debugfs.c | 10 ++++++++++
>  drivers/gpu/drm/xe/xe_gt.c      |  8 +++++++-
>  drivers/gpu/drm/xe/xe_gt.h      | 14 ++++++++++++++
>  3 files changed, 31 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_debugfs.c b/drivers/gpu/drm/xe/xe_debugfs.c
> index 7827a785b020..08d5bdf4cf61 100644
> --- a/drivers/gpu/drm/xe/xe_debugfs.c
> +++ b/drivers/gpu/drm/xe/xe_debugfs.c
> @@ -5,6 +5,7 @@
>  
>  #include "xe_debugfs.h"
>  
> +#include <linux/fault-inject.h>
>  #include <linux/string_helpers.h>
>  
>  #include <drm/drm_debugfs.h>
> @@ -20,6 +21,10 @@
>  #include "xe_vm.h"
>  #endif
>  
> +#ifdef CONFIG_FAULT_INJECTION
> +DECLARE_FAULT_ATTR(gt_reset_failure);
> +#endif
> +
>  static struct xe_device *node_to_xe(struct drm_info_node *node)
>  {
>  	return to_xe_device(node->minor->dev);
> @@ -131,4 +136,9 @@ void xe_debugfs_register(struct xe_device *xe)
>  
>  	for_each_gt(gt, xe, id)
>  		xe_gt_debugfs_register(gt);
> +
> +#ifdef CONFIG_FAULT_INJECTION
> +	fault_create_debugfs_attr("fail_gt_reset", root, &gt_reset_failure);
> +#endif
> +
>  }
> diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
> index 1db4d610f2fd..370d4b96e616 100644
> --- a/drivers/gpu/drm/xe/xe_gt.c
> +++ b/drivers/gpu/drm/xe/xe_gt.c
> @@ -525,6 +525,11 @@ static int gt_reset(struct xe_gt *gt)
>  
>  	xe_gt_info(gt, "reset started\n");
>  
> +	if (xe_fault_inject_gt_reset()) {
> +		err = -ECANCELED;
> +		goto err_fail;
> +	}
> +
>  	xe_gt_sanitize(gt);
>  
>  	xe_device_mem_access_get(gt_to_xe(gt));
> @@ -562,6 +567,7 @@ static int gt_reset(struct xe_gt *gt)
>  err_msg:
>  	XE_WARN_ON(xe_uc_start(&gt->uc));
>  	xe_device_mem_access_put(gt_to_xe(gt));
> +err_fail:
>  	xe_gt_err(gt, "reset failed (%pe)\n", ERR_PTR(err));
>  
>  	/* Notify userspace about gt reset failure */
> @@ -582,7 +588,7 @@ void xe_gt_reset_async(struct xe_gt *gt)
>  	xe_gt_info(gt, "trying reset\n");
>  
>  	/* Don't do a reset while one is already in flight */
> -	if (xe_uc_reset_prepare(&gt->uc))
> +	if (!xe_fault_inject_gt_reset() && xe_uc_reset_prepare(&gt->uc))
>  		return;
>  
>  	xe_gt_info(gt, "reset queued\n");
> diff --git a/drivers/gpu/drm/xe/xe_gt.h b/drivers/gpu/drm/xe/xe_gt.h
> index 7298653a73de..caded203a8a0 100644
> --- a/drivers/gpu/drm/xe/xe_gt.h
> +++ b/drivers/gpu/drm/xe/xe_gt.h
> @@ -7,6 +7,7 @@
>  #define _XE_GT_H_
>  
>  #include <drm/drm_util.h>
> +#include <linux/fault-inject.h>
>  
>  #include "xe_device_types.h"
>  #include "xe_hw_engine.h"
> @@ -16,6 +17,19 @@
>  		for_each_if(((hwe__) = (gt__)->hw_engines + (id__)) && \
>  			  xe_hw_engine_is_valid((hwe__)))
>  
> +#ifdef CONFIG_FAULT_INJECTION
> +extern struct fault_attr gt_reset_failure;
> +static inline bool xe_fault_inject_gt_reset(void)
> +{
> +	return should_fail(&gt_reset_failure, 1);
> +}
> +#else
> +static inline bool xe_fault_inject_gt_reset(void)
> +{
> +	return false;
> +}
> +#endif
> +
>  struct xe_gt *xe_gt_alloc(struct xe_tile *tile);
>  int xe_gt_init_early(struct xe_gt *gt);
>  int xe_gt_init(struct xe_gt *gt);
> -- 
> 2.25.1
>