Received-SPF: pass (google.com: domain of linux-kernel+bounces-94546-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Precedence: bulk
MIME-Version: 1.0
References: <20240306195031.490994-1-quic_abhinavk@quicinc.com>
 <CAA8EJpq=5=L5RdVZRkf=e2wyjQufnSzEC+=19FjCCF9S6SSEJA@mail.gmail.com> <2295092e-355b-4ebf-f630-14623cf7d9a3@quicinc.com>
In-Reply-To: <2295092e-355b-4ebf-f630-14623cf7d9a3@quicinc.com>
From: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Date: Wed, 6 Mar 2024 22:05:50 +0200
Message-ID: <CAA8EJpovNX31rWS_TdBTQiDmk0+kGBPdax6JncZupbXL0iUZhQ@mail.gmail.com>
Subject: Re: [RFC PATCH] drm/msm/dp: move link_ready out of HPD event thread
To: Abhinav Kumar <quic_abhinavk@quicinc.com>
Cc: freedreno@lists.freedesktop.org, Rob Clark <robdclark@gmail.com>, 
	Sean Paul <sean@poorly.run>, Marijn Suijten <marijn.suijten@somainline.org>, 
	David Airlie <airlied@gmail.com>, Daniel Vetter <daniel@ffwll.ch>, dri-devel@lists.freedesktop.org, 
	swboyd@chromium.org, quic_jesszhan@quicinc.com, quic_parellan@quicinc.com, 
	quic_khsieh@quicinc.com, linux-arm-msm@vger.kernel.org, 
	linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"

On Wed, 6 Mar 2024 at 21:59, Abhinav Kumar <quic_abhinavk@quicinc.com> wrote:
>
>
>
> On 3/6/2024 11:52 AM, Dmitry Baryshkov wrote:
> > On Wed, 6 Mar 2024 at 21:50, Abhinav Kumar <quic_abhinavk@quicinc.com> wrote:
> >>
> >> There are cases where the userspace might still send another
> >> frame after the HPD disconnect causing a modeset cycle after
> >> a disconnect. This messes the internal state machine of MSM DP driver
> >> and can lead to a crash as there can be an imbalance between
> >> bridge_disable() and bridge_enable().
> >>
> >> This was also previously reported on [1] for which [2] was posted
> >> and helped resolve the issue by rejecting commits if the DP is not
> >> in connected state.
> >>
> >> The change resolved the bug but there can also be another race condition.
> >> If hpd_event_thread does not pick up the EV_USER_NOTIFICATION and process it
> >> link_ready will also not be set to false allowing the frame to sneak in.
> >>
> >> Lets move setting link_ready outside of hpd_event_thread() processing to
> >> eliminate a window of race condition.
> >>
> >> [1] : https://gitlab.freedesktop.org/drm/msm/-/issues/17
> >> [2] : https://lore.kernel.org/all/1664408211-25314-1-git-send-email-quic_khsieh@quicinc.com/
> >>
> >> Signed-off-by: Abhinav Kumar <quic_abhinavk@quicinc.com>
> >> ---
> >>   drivers/gpu/drm/msm/dp/dp_display.c | 7 +++++--
> >>   1 file changed, 5 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/gpu/drm/msm/dp/dp_display.c b/drivers/gpu/drm/msm/dp/dp_display.c
> >> index 068d44eeaa07..e00092904ccc 100644
> >> --- a/drivers/gpu/drm/msm/dp/dp_display.c
> >> +++ b/drivers/gpu/drm/msm/dp/dp_display.c
> >> @@ -345,8 +345,6 @@ static int dp_display_send_hpd_notification(struct dp_display_private *dp,
> >>                                                           dp->panel->downstream_ports);
> >>          }
> >>
> >> -       dp->dp_display.link_ready = hpd;
> >> -
> >>          drm_dbg_dp(dp->drm_dev, "type=%d hpd=%d\n",
> >>                          dp->dp_display.connector_type, hpd);
> >>          drm_bridge_hpd_notify(bridge, dp->dp_display.link_ready);
> >> @@ -399,6 +397,8 @@ static int dp_display_process_hpd_high(struct dp_display_private *dp)
> >>                  goto end;
> >>          }
> >>
> >> +       dp->dp_display.link_ready = true;
> >
> > Do we need any kind of locking now?
> >
>
> hmm ... correct me if I have missed any flows but I think all paths
> where we will set link_ready are already protected by event_mutex?

I didn't check the source code, that's why I was asking. If it's
protected by event_mutex, then it should be fine.

>
> >> +
> >>          dp_add_event(dp, EV_USER_NOTIFICATION, true, 0);
> >>
> >>   end:
> >> @@ -466,6 +466,8 @@ static int dp_display_notify_disconnect(struct device *dev)
> >>   {
> >>          struct dp_display_private *dp = dev_get_dp_display_private(dev);
> >>
> >> +       dp->dp_display.link_ready = false;
> >> +
> >>          dp_add_event(dp, EV_USER_NOTIFICATION, false, 0);
> >>
> >>          return 0;
> >> @@ -487,6 +489,7 @@ static int dp_display_handle_port_status_changed(struct dp_display_private *dp)
> >>                  drm_dbg_dp(dp->drm_dev, "sink count is zero, nothing to do\n");
> >>                  if (dp->hpd_state != ST_DISCONNECTED) {
> >>                          dp->hpd_state = ST_DISCONNECT_PENDING;
> >> +                       dp->dp_display.link_ready = false;
> >>                          dp_add_event(dp, EV_USER_NOTIFICATION, false, 0);
> >>                  }
> >>          } else {
> >> --
> >> 2.34.1
> >>
> >
> >


-- 
With best wishes
Dmitry