Re: [Intel-gfx] [PATCH 01/14] drm/i915/tc: Abort DP AUX transfer on a disconnected TC port

[Andrzej Hajda <andrzej.hajda@intel.com>]

On 16.03.2023 14:17, Imre Deak wrote:
> On TC ports the 4ms AUX timeout combined with the 5 * 32 retry
> attempts during DPCD accesses adds a 640ms delay to each access if the
> sink is disconnected. This in turn slows down a modeset during which the
> sink is disconnected (for instance a disabling modeset).
> 
> Prevent the above delay by aborting AUX transfers on a TC port with a
> disconnected sink.
> 
> The DP 1.4a link CTS (4.2.1.5 Source Device Inactive HPD / Inactive AUX
> Test") also requires not to initiate AUX transfers on disconnected DP
> ports in general, however this patch doesn't change the behavior on
> non-TC ports, leaving that for a follow-up.
> 
> Reported-and-tested-by: Chris Chiu <chris.chiu@canonical.com>
> Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8279
> Signed-off-by: Imre Deak <imre.deak@intel.com>
> ---
>   drivers/gpu/drm/i915/display/intel_dp_aux.c | 15 +++++++++++++--
>   drivers/gpu/drm/i915/display/intel_tc.c     | 15 +++++++++++----
>   drivers/gpu/drm/i915/display/intel_tc.h     |  1 +
>   3 files changed, 25 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_dp_aux.c b/drivers/gpu/drm/i915/display/intel_dp_aux.c
> index 96967e21c94c2..eb07dc5d87099 100644
> --- a/drivers/gpu/drm/i915/display/intel_dp_aux.c
> +++ b/drivers/gpu/drm/i915/display/intel_dp_aux.c
> @@ -205,8 +205,19 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp,
>   	for (i = 0; i < ARRAY_SIZE(ch_data); i++)
>   		ch_data[i] = intel_dp->aux_ch_data_reg(intel_dp, i);
>   
> -	if (is_tc_port)
> +	if (is_tc_port) {
>   		intel_tc_port_lock(dig_port);
> +		/*
> +		 * Abort transfers on a disconnected port as required by
> +		 * DP 1.4a link CTS 4.2.1.5, also avoiding the long AUX
> +		 * timeouts that would otherwise happen.
> +		 * TODO: abort the transfer on non-TC ports as well.

Yes, for all ports this is definitely desirable, for example grep shows
about 15000 timeouts per one CI dmesg[1].

At least I hope, this cures the plague.

Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>

[1]: 
http://gfx-ci.igk.intel.com/tree/drm-tip/CI_DRM_12884/fi-elk-e7500/dmesg0.txt

Regards
Andrzej


> +		 */
> +		if (!intel_tc_port_connected_locked(&dig_port->base)) {
> +			ret = -ENXIO;
> +			goto out_unlock;
> +		}
> +	}
>   
>   	aux_domain = intel_aux_power_domain(dig_port);
>   
> @@ -367,7 +378,7 @@ intel_dp_aux_xfer(struct intel_dp *intel_dp,
>   
>   	intel_pps_unlock(intel_dp, pps_wakeref);
>   	intel_display_power_put_async(i915, aux_domain, aux_wakeref);
> -
> +out_unlock:
>   	if (is_tc_port)
>   		intel_tc_port_unlock(dig_port);
>   
> diff --git a/drivers/gpu/drm/i915/display/intel_tc.c b/drivers/gpu/drm/i915/display/intel_tc.c
> index f45328712bff1..050f998284592 100644
> --- a/drivers/gpu/drm/i915/display/intel_tc.c
> +++ b/drivers/gpu/drm/i915/display/intel_tc.c
> @@ -768,16 +768,23 @@ void intel_tc_port_sanitize_mode(struct intel_digital_port *dig_port)
>    * connected ports are usable, and avoids exposing to the users objects they
>    * can't really use.
>    */
> +bool intel_tc_port_connected_locked(struct intel_encoder *encoder)
> +{
> +	struct intel_digital_port *dig_port = enc_to_dig_port(encoder);
> +	struct drm_i915_private *i915 = to_i915(dig_port->base.base.dev);
> +
> +	drm_WARN_ON(&i915->drm, !intel_tc_port_ref_held(dig_port));
> +
> +	return tc_port_live_status_mask(dig_port) & BIT(dig_port->tc_mode);
> +}
> +
>   bool intel_tc_port_connected(struct intel_encoder *encoder)
>   {
>   	struct intel_digital_port *dig_port = enc_to_dig_port(encoder);
>   	bool is_connected;
>   
>   	intel_tc_port_lock(dig_port);
> -
> -	is_connected = tc_port_live_status_mask(dig_port) &
> -		       BIT(dig_port->tc_mode);
> -
> +	is_connected = intel_tc_port_connected_locked(encoder);
>   	intel_tc_port_unlock(dig_port);
>   
>   	return is_connected;
> diff --git a/drivers/gpu/drm/i915/display/intel_tc.h b/drivers/gpu/drm/i915/display/intel_tc.h
> index d54082e2d5e8d..93813056043a5 100644
> --- a/drivers/gpu/drm/i915/display/intel_tc.h
> +++ b/drivers/gpu/drm/i915/display/intel_tc.h
> @@ -17,6 +17,7 @@ bool intel_tc_port_in_dp_alt_mode(struct intel_digital_port *dig_port);
>   bool intel_tc_port_in_legacy_mode(struct intel_digital_port *dig_port);
>   
>   bool intel_tc_port_connected(struct intel_encoder *encoder);
> +bool intel_tc_port_connected_locked(struct intel_encoder *encoder);
>   
>   u32 intel_tc_port_get_lane_mask(struct intel_digital_port *dig_port);
>   u32 intel_tc_port_get_pin_assignment_mask(struct intel_digital_port *dig_port);