Re: [PATCH v4] drm/xe/cri: Implement Wa_14028586678

[Daniel Charles <daniel.charles@intel.com>]

On 10/30/2026 10:49 AM, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [Medium] drm/xe: xe_wa_apply_tile_workarounds() runs before xe_pcode_ready() on resume
> --
>
> commit 3d1f950d0d5e813299acdbeedeb799005ea2bb7c
> Author: Fei Yang <fei.yang@intel.com>
>
> drm/xe/cri: Implement Wa_14028586678
>
> This commit implements hardware workaround Wa_14028586678 on Crescent
> Island by updating DWNSTRM_ACCESS_MODE_GAMREQ_MERTSS in
> xe_wa_apply_tile_workarounds() and calling xe_wa_apply_tile_workarounds()
> on runtime D3cold resume.
>
>> diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c
>> index 5a562d8daeb94..d4f93de368fc4 100644
>> --- a/drivers/gpu/drm/xe/xe_pm.c
>> +++ b/drivers/gpu/drm/xe/xe_pm.c
>> @@ -687,6 +687,7 @@ int xe_pm_runtime_suspend(struct xe_device *xe)
>>    */
>>   int xe_pm_runtime_resume(struct xe_device *xe)
>>   {
>> +	struct xe_tile *tile;
>>   	struct xe_gt *gt;
>>   	u8 id;
>>   	int err = 0;
>> @@ -702,6 +703,9 @@ int xe_pm_runtime_resume(struct xe_device *xe)
>>   		if (err)
>>   			goto out;
>>   
>> +		for_each_tile(tile, xe, id)
>> +			xe_wa_apply_tile_workarounds(tile);
>> +
> [Severity: Medium]
> In xe_pm_runtime_resume(), xe_wa_apply_tile_workarounds() is called after
> xe_pcode_ready(xe, true).
>
> However, in the system resume path in drivers/gpu/drm/xe/xe_pm.c,
> xe_pm_resume() calls xe_wa_apply_tile_workarounds() before
> xe_pcode_ready(xe, true):
>
> drivers/gpu/drm/xe/xe_pm.c:xe_pm_resume() {
> 	...
> 	for_each_tile(tile, xe, id)
> 		xe_wa_apply_tile_workarounds(tile);
>
> 	err = xe_pcode_ready(xe, true);
> 	if (err)
> 		return err;
> 	...
> }
>
> As noted in xe_device_probe_early() in drivers/gpu/drm/xe/xe_device.c:
>
> 	/*
> 	 * Check for pcode uncore_init status to confirm if the SoC
> 	 * initialization is complete. Until done, any MMIO or lmem access from
> 	 * the driver will be blocked
> 	 */
> 	err = xe_pcode_probe_early(xe);
>
> When resuming from system suspend, can the MMIO read-modify-write to
> DWNSTRM_ACCESS_MODE_GAMREQ_MERTSS in xe_wa_apply_tile_workarounds() run
> before PCODE SoC initialization has completed and while MMIO access is
> still blocked, leaving Wa_14028586678 unapplied or corrupting other fields
> of the register?
>
> Should xe_pm_resume() also wait for xe_pcode_ready(xe, true) before calling
> xe_wa_apply_tile_workarounds()?
I'll move it after xe_pcode_ready(xe, true)
>