Date:   Mon, 27 Feb 2023 15:36:41 -0800
From:   Bjorn Andersson <quic_bjorande@quicinc.com>
To:     Sankeerth Billakanti <quic_sbillaka@quicinc.com>
CC:     <dri-devel@lists.freedesktop.org>, <linux-arm-msm@vger.kernel.org>,
        <freedreno@lists.freedesktop.org>, <linux-kernel@vger.kernel.org>,
        <robdclark@gmail.com>, <seanpaul@chromium.org>,
        <swboyd@chromium.org>, <quic_kalyant@quicinc.com>,
        <quic_abhinavk@quicinc.com>, <dianders@chromium.org>,
        <quic_khsieh@quicinc.com>, <sean@poorly.run>, <airlied@linux.ie>,
        <daniel@ffwll.ch>, <dmitry.baryshkov@linaro.org>,
        <quic_vproddut@quicinc.com>
Subject: Re: [RFC PATCH 2/2] drm/msm/dp: enable pm_runtime support for dp
 driver
Message-ID: <20230227233641.GC4062527@hu-bjorande-lv.qualcomm.com>
References: <20230223135635.30659-1-quic_sbillaka@quicinc.com>
 <20230223135635.30659-3-quic_sbillaka@quicinc.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Disposition: inline
In-Reply-To: <20230223135635.30659-3-quic_sbillaka@quicinc.com>
Precedence: bulk

On Thu, Feb 23, 2023 at 07:26:35PM +0530, Sankeerth Billakanti wrote:
> diff --git a/drivers/gpu/drm/msm/dp/dp_aux.c b/drivers/gpu/drm/msm/dp/dp_aux.c
[..]
> +static int dp_runtime_resume(struct device *dev)
> +{
> +	struct platform_device *pdev = to_platform_device(dev);
> +	struct msm_dp *dp_display = platform_get_drvdata(pdev);
> +	struct dp_display_private *dp;
> +
> +	dp = container_of(dp_display, struct dp_display_private, dp_display);
> +	dp_display_host_init(dp);
> +
> +	if (dp->dp_display.is_edp) {
> +		dp_display_host_phy_init(dp);
> +	} else {
> +		dp_catalog_hpd_config_intr(dp->catalog,
> +				DP_DP_HPD_PLUG_INT_MASK |
> +				DP_DP_HPD_UNPLUG_INT_MASK,
> +				true);

I believe this is backwards.

Only in the event that there's no "downstream" HPD handler should we use
the internal HPD. This is signalled by the DRM framework by a call to
dp_bridge_hpd_enable(). So we should use that to enable/disable the
internal HPD handler.

When this happens, we have a reason for keeping power on; i.e. call
pm_runtime_get(). Once we have power/clocking, we'd call
dp_catalog_hpd_config_intr(), from dp_bridge_hpd_enable().


In the case that the internal HPD handling is not use,
dp_bridge_hpd_enable() will not be called, instead once the downstream
hpd handler switches state dp_bridge_hpd_notify() will be invoked.

In this case, we need the DP controller to be powered/clocked between
connector_status_connected and connector_status_disconnected.


I believe this should allow the DP controller(s) to stay powered down in
the case where we have external HPD handling (e.g. USB Type-C or
gpio-based dp-connector).

Regards,
Bjorn