Re: [PATCH] drm/i915/dp: Ignore HPD when in DPLL enable/disable cycle

[Jani Nikula <jani.nikula@linux.intel.com>]

On Fri, 17 Apr 2026, Jani Nikula <jani.nikula@linux.intel.com> wrote:
> On Fri, 17 Apr 2026, Suraj Kandpal <suraj.kandpal@intel.com> wrote:
>> When we are enable/disable DPLL cycle there are chances the connected
>> monitors is still sending us HPD signals. This causes us to handle them
>> which require DPCD read. These sometimes race with the DPLL getting
>> disabled in between causing AUX failures via Timeout.
>> Introduce atomic variable link_teardown which is used to track if
>> we are in DPLL enable/disable cycle. We ignore HPDs during this time.
>> Re-enable after DPLL is up so that we can avoid populating logs
>> with expected logs AUX timeout failures.
>
> How is this not racy?

Oh, please don't cross-post trybot list with other lists.

>
>>
>> Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
>> ---
>>  drivers/gpu/drm/i915/display/intel_ddi.c           | 5 +++++
>>  drivers/gpu/drm/i915/display/intel_display_types.h | 1 +
>>  drivers/gpu/drm/i915/display/intel_dp.c            | 5 +++++
>>  3 files changed, 11 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c b/drivers/gpu/drm/i915/display/intel_ddi.c
>> index 178074316a2c..4a523eb56bc4 100644
>> --- a/drivers/gpu/drm/i915/display/intel_ddi.c
>> +++ b/drivers/gpu/drm/i915/display/intel_ddi.c
>> @@ -2086,8 +2086,12 @@ static struct intel_dpll *hsw_ddi_get_pll(struct intel_encoder *encoder)
>>  void intel_ddi_enable_clock(struct intel_encoder *encoder,
>>  			    const struct intel_crtc_state *crtc_state)
>>  {
>> +	struct intel_digital_port *dig_port = enc_to_dig_port(encoder);
>> +
>>  	if (encoder->enable_clock)
>>  		encoder->enable_clock(encoder, crtc_state);
>> +
>> +	atomic_set(&dig_port->link_teardown, 0);
>>  }
>>  
>>  void intel_ddi_disable_clock(struct intel_encoder *encoder)
>> @@ -3181,6 +3185,7 @@ static void intel_ddi_post_disable_dp(struct intel_atomic_state *state,
>>  					dig_port->ddi_io_power_domain,
>>  					wakeref);
>>  
>> +	atomic_set(&dig_port->link_teardown, 1);
>>  	intel_ddi_disable_clock(encoder);
>>  
>>  	/* De-select Thunderbolt */
>> diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h b/drivers/gpu/drm/i915/display/intel_display_types.h
>> index c81916761850..f59bbb2fb260 100644
>> --- a/drivers/gpu/drm/i915/display/intel_display_types.h
>> +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
>> @@ -1985,6 +1985,7 @@ struct intel_digital_port {
>>  	enum intel_display_power_domain ddi_io_power_domain;
>>  	struct ref_tracker *ddi_io_wakeref;
>>  	struct ref_tracker *aux_wakeref;
>> +	atomic_t link_teardown;
>>  
>>  	struct intel_tc_port *tc;
>>  
>> diff --git a/drivers/gpu/drm/i915/display/intel_dp.c b/drivers/gpu/drm/i915/display/intel_dp.c
>> index 35b8fb5740aa..9177fe9b3e84 100644
>> --- a/drivers/gpu/drm/i915/display/intel_dp.c
>> +++ b/drivers/gpu/drm/i915/display/intel_dp.c
>> @@ -6889,6 +6889,11 @@ intel_dp_hpd_pulse(struct intel_digital_port *dig_port, bool long_hpd)
>>  	struct intel_dp *intel_dp = &dig_port->dp;
>>  	u8 dpcd[DP_RECEIVER_CAP_SIZE];
>>  
>> +	if (atomic_read(&dig_port->link_teardown)) {
>> +		drm_dbg_kms("Ignoring HPD since DPLL is getting disabled\n");
>> +		return IRQ_NONE;
>> +	}
>> +
>>  	if (dig_port->base.type == INTEL_OUTPUT_EDP &&
>>  	    (long_hpd ||
>>  	     intel_display_rpm_suspended(display) ||

-- 
Jani Nikula, Intel