Re: [PATCH 2/3] drm/xe: Do not wake device during a GT reset

[Matthew Auld <matthew.auld@intel.com>]

On 22/10/2025 01:55, Matthew Brost wrote:
> Waking the device during a GT reset can lead to unintended memory
> allocation, which is not allowed since GT resets occur in the reclaim
> path. Prevent this by holding a PM reference while a reset is in flight.
> 
> Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
> Cc: stable@vger.kernel.org
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>

Reviewed-by: Matthew Auld <matthew.auld@intel.com>


> ---
>   drivers/gpu/drm/xe/xe_gt.c | 19 ++++++++++++-------
>   1 file changed, 12 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
> index d8e94fb8b9bd..89808b33d0a8 100644
> --- a/drivers/gpu/drm/xe/xe_gt.c
> +++ b/drivers/gpu/drm/xe/xe_gt.c
> @@ -818,17 +818,19 @@ static int gt_reset(struct xe_gt *gt)
>   	unsigned int fw_ref;
>   	int err;
>   
> -	if (xe_device_wedged(gt_to_xe(gt)))
> -		return -ECANCELED;
> +	if (xe_device_wedged(gt_to_xe(gt))) {
> +		err = -ECANCELED;
> +		goto err_pm_put;
> +	}
>   
>   	/* We only support GT resets with GuC submission */
> -	if (!xe_device_uc_enabled(gt_to_xe(gt)))
> -		return -ENODEV;
> +	if (!xe_device_uc_enabled(gt_to_xe(gt))) {
> +		err = -ENODEV;
> +		goto err_pm_put;
> +	}
>   
>   	xe_gt_info(gt, "reset started\n");
>   
> -	xe_pm_runtime_get(gt_to_xe(gt));
> -
>   	if (xe_fault_inject_gt_reset()) {
>   		err = -ECANCELED;
>   		goto err_fail;
> @@ -875,6 +877,7 @@ static int gt_reset(struct xe_gt *gt)
>   	xe_gt_err(gt, "reset failed (%pe)\n", ERR_PTR(err));
>   
>   	xe_device_declare_wedged(gt_to_xe(gt));
> +err_pm_put:
>   	xe_pm_runtime_put(gt_to_xe(gt));
>   
>   	return err;
> @@ -896,7 +899,9 @@ void xe_gt_reset_async(struct xe_gt *gt)
>   		return;
>   
>   	xe_gt_info(gt, "reset queued\n");
> -	queue_work(gt->ordered_wq, &gt->reset.worker);
> +	xe_pm_runtime_get_noresume(gt_to_xe(gt));
> +	if (!queue_work(gt->ordered_wq, &gt->reset.worker))
> +		xe_pm_runtime_put(gt_to_xe(gt));
>   }
>   
>   void xe_gt_suspend_prepare(struct xe_gt *gt)