| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAA8EJprp_yx0zEBtrUwxXooTDDYWnsU2++VOz=3AgV8zDbMvWg@mail.gmail.com>
Date: Wed, 4 Oct 2023 01:36:56 +0300
From: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
To: Kuogee Hsieh <quic_khsieh@...cinc.com>
Cc: dri-devel@...ts.freedesktop.org, robdclark@...il.com,
sean@...rly.run, swboyd@...omium.org, dianders@...omium.org,
vkoul@...nel.org, daniel@...ll.ch, airlied@...il.com,
agross@...nel.org, andersson@...nel.org, quic_abhinavk@...cinc.com,
quic_jesszhan@...cinc.com, quic_sbillaka@...cinc.com,
marijn.suijten@...ainline.org, freedreno@...ts.freedesktop.org,
linux-arm-msm@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v4 7/8] drm/msm/dp: add pm_runtime_force_suspend()/resume()
On Wed, 4 Oct 2023 at 01:12, Kuogee Hsieh <quic_khsieh@...cinc.com> wrote:
>
>
> On 10/3/2023 10:53 AM, Dmitry Baryshkov wrote:
> > On Tue, 3 Oct 2023 at 19:44, Kuogee Hsieh <quic_khsieh@...cinc.com> wrote:
> >>
> >> On 9/27/2023 3:00 PM, Dmitry Baryshkov wrote:
> >>> On Wed, 27 Sept 2023 at 23:54, Kuogee Hsieh <quic_khsieh@...cinc.com> wrote:
> >>>> After incorporated pm_runtime framework into eDP/DP driver, the
> >>> incorporating
> >>>
> >>>
> >>>> original dp_pm_suspend() to handle power off both DP phy and
> >>>> controller during suspend and dp_pm_resume() to handle power on
> >>>> both DP phy and controller during resume are not necessary since
> >>>> those function are replaced by dp_pm_runtime_suspend() and
> >>>> dp_pm_runtime_resume() through pm runtime framework.
> >>>> Therefore add pm framework provides functions,
> >>>> pm_runtime_force_suspend()/resume() to complete incorporating pm
> >>>> runtime framework into DP driver.
> >>>>
> >>>> Changes in v4:
> >>>> -- drop both dp_pm_prepare() and dp_pm_compete() from this change
> >>>> -- delete ST_SUSPENDED state
> >>>> -- rewording commit text to add more details regrading the purpose
> >>>> of this change
> >>>>
> >>>> Changes in v3:
> >>>> -- replace dp_pm_suspend() with pm_runtime_force_suspend()
> >>>> -- replace dp_pm_resume() with pm_runtime_force_resume()
> >>>>
> >>>> Signed-off-by: Kuogee Hsieh <quic_khsieh@...cinc.com>
> >>>> ---
> >>>> drivers/gpu/drm/msm/dp/dp_display.c | 113 ++----------------------------------
> >>>> 1 file changed, 5 insertions(+), 108 deletions(-)
> >>>>
> >>>> diff --git a/drivers/gpu/drm/msm/dp/dp_display.c b/drivers/gpu/drm/msm/dp/dp_display.c
> >>>> index 9158a2c..711d262 100644
> >>>> --- a/drivers/gpu/drm/msm/dp/dp_display.c
> >>>> +++ b/drivers/gpu/drm/msm/dp/dp_display.c
> >>>> @@ -49,7 +49,6 @@ enum {
> >>>> ST_CONNECTED,
> >>>> ST_DISCONNECT_PENDING,
> >>>> ST_DISPLAY_OFF,
> >>>> - ST_SUSPENDED,
> >>>> };
> >>>>
> >>>> enum {
> >>>> @@ -560,7 +559,7 @@ static int dp_hpd_plug_handle(struct dp_display_private *dp, u32 data)
> >>>> drm_dbg_dp(dp->drm_dev, "Before, type=%d hpd_state=%d\n",
> >>>> dp->dp_display.connector_type, state);
> >>>>
> >>>> - if (state == ST_DISPLAY_OFF || state == ST_SUSPENDED) {
> >>>> + if (state == ST_DISPLAY_OFF) {
> >>>> mutex_unlock(&dp->event_mutex);
> >>>> return 0;
> >>>> }
> >>>> @@ -674,7 +673,7 @@ static int dp_irq_hpd_handle(struct dp_display_private *dp, u32 data)
> >>>> drm_dbg_dp(dp->drm_dev, "Before, type=%d hpd_state=%d\n",
> >>>> dp->dp_display.connector_type, state);
> >>>>
> >>>> - if (state == ST_DISPLAY_OFF || state == ST_SUSPENDED) {
> >>>> + if (state == ST_DISPLAY_OFF) {
> >>>> mutex_unlock(&dp->event_mutex);
> >>>> return 0;
> >>>> }
> >>>> @@ -1321,110 +1320,10 @@ static int dp_pm_runtime_resume(struct device *dev)
> >>>> return 0;
> >>>> }
> >>>>
> >>>> -static int dp_pm_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;
> >>>> - int sink_count = 0;
> >>>> -
> >>>> - dp = container_of(dp_display, struct dp_display_private, dp_display);
> >>>> -
> >>>> - mutex_lock(&dp->event_mutex);
> >>>> -
> >>>> - drm_dbg_dp(dp->drm_dev,
> >>>> - "Before, type=%d core_inited=%d phy_inited=%d power_on=%d\n",
> >>>> - dp->dp_display.connector_type, dp->core_initialized,
> >>>> - dp->phy_initialized, dp_display->power_on);
> >>>> -
> >>>> - /* start from disconnected state */
> >>>> - dp->hpd_state = ST_DISCONNECTED;
> >>>> -
> >>>> - /* turn on dp ctrl/phy */
> >>>> - dp_display_host_init(dp);
> >>>> -
> >>>> - if (dp_display->is_edp)
> >>>> - dp_catalog_ctrl_hpd_enable(dp->catalog);
> >>>> -
> >>>> - if (dp_catalog_link_is_connected(dp->catalog)) {
> >>>> - /*
> >>>> - * set sink to normal operation mode -- D0
> >>>> - * before dpcd read
> >>>> - */
> >>>> - dp_display_host_phy_init(dp);
> >>>> - dp_link_psm_config(dp->link, &dp->panel->link_info, false);
> >>>> - sink_count = drm_dp_read_sink_count(dp->aux);
> >>>> - if (sink_count < 0)
> >>>> - sink_count = 0;
> >>>> -
> >>>> - dp_display_host_phy_exit(dp);
> >>>> - }
> >>>> -
> >>>> - dp->link->sink_count = sink_count;
> >>>> - /*
> >>>> - * can not declared display is connected unless
> >>>> - * HDMI cable is plugged in and sink_count of
> >>>> - * dongle become 1
> >>>> - * also only signal audio when disconnected
> >>>> - */
> >>>> - if (dp->link->sink_count) {
> >>>> - dp->dp_display.link_ready = true;
> >>>> - } else {
> >>>> - dp->dp_display.link_ready = false;
> >>>> - dp_display_handle_plugged_change(dp_display, false);
> >>>> - }
> >>>> -
> >>>> - drm_dbg_dp(dp->drm_dev,
> >>>> - "After, type=%d sink=%d conn=%d core_init=%d phy_init=%d power=%d\n",
> >>>> - dp->dp_display.connector_type, dp->link->sink_count,
> >>>> - dp->dp_display.link_ready, dp->core_initialized,
> >>>> - dp->phy_initialized, dp_display->power_on);
> >>>> -
> >>>> - mutex_unlock(&dp->event_mutex);
> >>>> -
> >>>> - return 0;
> >>>> -}
> >>>> -
> >>>> -static int dp_pm_suspend(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);
> >>>> -
> >>>> - mutex_lock(&dp->event_mutex);
> >>>> -
> >>>> - drm_dbg_dp(dp->drm_dev,
> >>>> - "Before, type=%d core_inited=%d phy_inited=%d power_on=%d\n",
> >>>> - dp->dp_display.connector_type, dp->core_initialized,
> >>>> - dp->phy_initialized, dp_display->power_on);
> >>>> -
> >>>> - /* mainlink enabled */
> >>>> - if (dp_power_clk_status(dp->power, DP_CTRL_PM))
> >>>> - dp_ctrl_off_link_stream(dp->ctrl);
> >>>> -
> >>>> - dp_display_host_phy_exit(dp);
> >>> I was under the impression that dp_pm_runtime_suspend / _resume
> >>> functions perform phy init/exit only in eDP cases. Can we really drop
> >>> the main suspend/resume functions?
> >> yes on eDP case since it is embedded.
> > Let me ask the same question in a different way:
> >
> > dp_pm_suspend() / dp_pm_resume() functions contain several calls to DP
> > functions. Why can we drop them now? Maybe they had to be dropped in
> > one of the previous patches, when you have added proper runtime PM
> > support?
> >
> > Could you please confirm that after each patch the DP driver is
> > working, that there are no hidden dependencies between patches?
>
> patch #5 ==> drm/msm/dp: incorporate pm_runtime framework into DP driver
>
> patch #6 ==> drm/msm/dp: delete EV_HPD_INIT_SETUP
>
> patch #7 ==> drm/msm/dp: add pm_runtime_force_suspend()/resume() <==
> both dp_pm_suspend() and dp_pm_resume() are dropped here
>
>
> Patch #5 is this patch and dp_pm_suspend() and dp_pm_resume() still kept.
>
> patch #7 drop both dp_pm_suspend() and dp_pm_resume().
This makes me wonder if squashing #5 and #7 would have resulted in a
cleaner and easier to understand patch. This way we'd be able to
observer code moved from suspend/resume to runtime_suspend /
runtime_resume
>
> In order to keep every patch work for suspend/resume test, I drop
> dp_pm_susend() and dp_pm_resuem() at patch #7.
>
> yes, i confirm each patch DP driver is working.
Ack, thank you.
>
>
> >
> >> for external DP case, there are two steps
> >>
> >> step 1: enable DP controller's hpd block and start waiting for hpd
> >> interrupts at dp_display_hpd_enable()
> The step number I mentioned here is for hpd_internal == true case.
> > Step 1 should be optional. DP should be functional even if the
> > .hpd_enable was not called. Have you tested this usecase?
>
> yes, for hpd_internal == false, step #1 is not required.
>
> however I do not have device to test it.
You have. Just add a dp-connector with hpd-gpios to any of your
DP-enabled devices and change the pinctrl for hpd pin to use the gpio
function instead of HPD.
>
> But i think it should work since pm_runtime_resume_and_get() and
> pm_runtime_put_sync() to dp_hpd_plug_handle() and dp_hpd_unplug_handle()
> respectively.
Ack.
> >> step 2: at plugin interrupts, dp_display_host_phy_init()
> >>
> >> step 3: at unplug interrupt: dp_bridge_atomic_post_disable()
> >> dp_display_host_phy_exi()
> >>
> >> at runtime, there is loop between step 2 and step 3
> >>
> >> step 4: disable DP controller's hpd block
> >>
> >>>> -
> >>>> - /* host_init will be called at pm_resume */
> >>>> - dp_display_host_deinit(dp);
> >>>> -
> >>>> - dp->hpd_state = ST_SUSPENDED;
> >>>> -
> >>>> - drm_dbg_dp(dp->drm_dev,
> >>>> - "After, type=%d core_inited=%d phy_inited=%d power_on=%d\n",
> >>>> - dp->dp_display.connector_type, dp->core_initialized,
> >>>> - dp->phy_initialized, dp_display->power_on);
> >>>> -
> >>>> - mutex_unlock(&dp->event_mutex);
> >>>> -
> >>>> - return 0;
> >>>> -}
> >>>> -
> >>>> static const struct dev_pm_ops dp_pm_ops = {
> >>>> SET_RUNTIME_PM_OPS(dp_pm_runtime_suspend, dp_pm_runtime_resume, NULL)
> >>>> - .suspend = dp_pm_suspend,
> >>>> - .resume = dp_pm_resume,
> >>>> + SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
> >>>> + pm_runtime_force_resume)
> >>>> };
> >>>>
> >>>> static struct platform_driver dp_display_driver = {
> >>>> @@ -1658,9 +1557,6 @@ void dp_bridge_atomic_post_disable(struct drm_bridge *drm_bridge,
> >>>>
> >>>> dp_display = container_of(dp, struct dp_display_private, dp_display);
> >>>>
> >>>> - if (dp->is_edp)
> >>>> - dp_hpd_unplug_handle(dp_display, 0);
> >>> Why?
> >> dp_hpd_unplug_handle() does not tear down phy.
> >>
> >> Therefore eDP does not need to call unplug handle.
> > I don't fully understand your argument here. Could you please
> > describe, why this function call was necessary beforehand and what is
> > being changed now, so that it becomes unnecessary?
>
> dp_hpd_unplug_handle() is not necessary for eDP from very beginning
> since dp_bridge_atomic_enable() do it all (tear down link and phy).
Excuse me? Where does dp_bridge_atomic_enable() tear down the link?
>
> I think it was added long time ago mistakenly just like to be
> compatible with external DP since external DP always have
> dp_hpd_unplug_handle() called from irq_handle().
>
> i can restore it back if you insist it.
I insist on having clean documented patches. If function call was not
necessary beforehand, please drop it in the separate patch with proper
description of why and what.
>
>
> >>
> >>
> >>>> -
> >>>> mutex_lock(&dp_display->event_mutex);
> >>>>
> >>>> state = dp_display->hpd_state;
> >>>> @@ -1748,6 +1644,7 @@ void dp_bridge_hpd_disable(struct drm_bridge *bridge)
> >>>> dp_catalog_ctrl_hpd_disable(dp->catalog);
> >>>>
> >>>> dp_display->internal_hpd = false;
> >>>> + dp->hpd_state = ST_DISCONNECTED;
> >>> Why? We have only disabled sending of the HPD events. The dongle might
> >>> still be connected.
> >> dp_bridge_hpd_disable() disable dp controller hpd block (no more hpd
> >> interrupt will be received).
> >>
> >> dp_bridge_hpd_disable() should happen after DP main link had been teared
> >> down already.
> > No, this assumption is incorrect. hpd_disable can happen at any point
> > during runtime.
> > It merely disables HPD interrupt generation, it has nothing to do with
> > the DP block being enabled or not.
> >
> >> Therefore hpd_state need to be in default state so that next plugin
> >> handle will be start with correct state.
> >>
> >>
> >>>> pm_runtime_mark_last_busy(&dp->pdev->dev);
> >>>> pm_runtime_put_autosuspend(&dp->pdev->dev);
> >>>> --
> >>>> 2.7.4
> >>>>
> >>> --
> >>> With best wishes
> >>>
> >>> Dmitry
> >
> >
--
With best wishes
Dmitry
Powered by blists - more mailing lists