[PATCH v3] drm/bridge:anx7625: Update HDCP status at atomic_enable()

Tue Dec 17 12:30:04 UTC 2024

On Tue, Dec 17, 2024 at 01:50:00AM +0000, Xin Ji wrote:
> > -----Original Message-----
> > From: Pin-yen Lin <treapking at chromium.org>
> > Sent: Monday, December 16, 2024 8:05 PM
> > To: Dmitry Baryshkov <dmitry.baryshkov at linaro.org>
> > Cc: Xin Ji <xji at analogixsemi.com>; Andrzej Hajda <andrzej.hajda at intel.com>;
> > Neil Armstrong <neil.armstrong at linaro.org>; Robert Foss <rfoss at kernel.org>;
> > Laurent Pinchart <Laurent.pinchart at ideasonboard.com>; Jonas Karlman
> > <jonas at kwiboo.se>; Jernej Skrabec <jernej.skrabec at gmail.com>; Maarten
> > Lankhorst <maarten.lankhorst at linux.intel.com>; Maxime Ripard
> > <mripard at kernel.org>; Thomas Zimmermann <tzimmermann at suse.de>; David
> > Airlie <airlied at gmail.com>; Simona Vetter <simona at ffwll.ch>; Bernie Liang
> > <bliang at analogixsemi.com>; Qilin Wen <qwen at analogixsemi.com>;
> > treapking at google.com; dri-devel at lists.freedesktop.org; linux-
> > kernel at vger.kernel.org
> > Subject: Re: [PATCH v3] drm/bridge:anx7625: Update HDCP status at
> > atomic_enable()
> > 
> > 
> > Hi Dmitry,
> > 
> > On Mon, Dec 16, 2024 at 7:59 PM Dmitry Baryshkov
> > <dmitry.baryshkov at linaro.org> wrote:
> > >
> > > On Mon, Dec 16, 2024 at 08:33:23AM +0000, Xin Ji wrote:
> > > > > -----Original Message-----
> > > > > From: Dmitry Baryshkov <dmitry.baryshkov at linaro.org>
> > > > > Sent: Friday, December 13, 2024 9:17 PM
> > > > > To: Xin Ji <xji at analogixsemi.com>
> > > > > Cc: Andrzej Hajda <andrzej.hajda at intel.com>; Neil Armstrong
> > > > > <neil.armstrong at linaro.org>; Robert Foss <rfoss at kernel.org>;
> > > > > Laurent Pinchart <Laurent.pinchart at ideasonboard.com>; Jonas
> > > > > Karlman <jonas at kwiboo.se>; Jernej Skrabec
> > > > > <jernej.skrabec at gmail.com>; Maarten Lankhorst
> > > > > <maarten.lankhorst at linux.intel.com>; Maxime Ripard
> > > > > <mripard at kernel.org>; Thomas Zimmermann <tzimmermann at suse.de>;
> > > > > David Airlie <airlied at gmail.com>; Simona Vetter <simona at ffwll.ch>;
> > > > > Bernie Liang <bliang at analogixsemi.com>; Qilin Wen
> > > > > <qwen at analogixsemi.com>; treapking at google.com;
> > > > > dri-devel at lists.freedesktop.org; linux- kernel at vger.kernel.org
> > > > > Subject: Re: [PATCH v3] drm/bridge:anx7625: Update HDCP status at
> > > > > atomic_enable()
> > > > >
> > > > > CAUTION: This email originated from outside of the organization.
> > > > > Please do not click links or open attachments unless you recognize
> > > > > the sender, and know the content is safe.
> > > > >
> > > > >
> > > > > On Fri, 13 Dec 2024 at 13:00, Xin Ji <xji at analogixsemi.com> wrote:
> > > > > >
> > > > > > Hi Dmitry, sorry, I didn't clear describe the reason.
> > > > >
> > > > > Please. Do not top-post. Please paste your answer under the
> > > > > question, not somewhere at the top of the email. This allows us to
> > > > > have a more constructive dialog. Additional bonus if you can fix
> > > > > your email client to insert sensible quoting information instead of dumping
> > the headers of the original email.
> > > > Hi Dmitry, OK, sorry about it. Currently, we have problem to fetch
> > > > email from Microsoft on Ubuntu. I'll try to fix it later.
> > > > >
> > > > > >
> > > > > > Anx7625 implement DSI to DP convert behind USB Type-C port, when
> > > > > > user plug into USB Type-C Dongle with DP monitor, the user space
> > > > > > will enable HDCP feature, then kernel do HDCP and output display
> > > > > > and set HDCP content to ENABLE, but the issue happened if user
> > > > > > manually change the
> > > > > monitor's resolution later.
> > > > > >
> > > > > > Each time user change the resolution, kernel will call bridge
> > > > > > interface .atomic_disable() and .atomic_enable(), the old driver
> > > > > > will keep HDCP state to ENABLE, this is a BUG, when user change
> > > > > > the resolution, kernel must change HDCP content too (mustn't
> > > > > > keep to ENABLE),
> > > > >
> > > > > Why? Could you please point me to the corresponding documentation
> > > > > or a code path in the other driver? Preferably i915, AMD or Nouveau.
> > > > As
> > https://elixir.bootlin.com/linux/v6.12.5/source/drivers/gpu/drm/drm_connecto
> > r.c#L1423:
> > > >         - ENABLED -> DESIRED (termination of authentication) As
> > > > there is no other interface to tell anx7625 bridge driver, so the I
> > > > think best place to handle ENABLE -> DESIRED in .atomic_disable().
> > >
> > > I was looking for something like cdns_mhdp_connector_atomic_check(),
> > > which switches to UNDESIRED if there is no new CRTC. Likewise i915
> > > driver performs this in intel_hdcp_atomic_check() if there is a need
> > > for modeset.
> Hi Dmitry, the bridge driver is different with i915, anx7625 bridge driver
> only implement some passive callback interface. There is no HDCP status
> checking while do resolution switch.

Does that mean that you _have_ to reestablish HDCP after mode switching
in order for it to work?

> > 
> > I believe you mean "DESIRED" here.
> > >
> > > For the "termination of authentication" case see
> > > cdns_mhdp_hdcp_check_link(), which detects if the HDCP got disabled by
> > > HW and then updates the status accordingly.
> > >
> > > >
> > > > >
> > > > > > as DRM doc said, kernel cannot change from ENABLE to UNDESIRE,
> > > > > > so next patch, I'll change it to DESIRE in .atomic_disable().
> > >
> > > This e.g. will result in HDCP being restarted for all modesets. Is
> > > this an expected behaviour?
> > 
> > The bridge could be powered off between .atomic_disable()
> > and .atomic_enable(), though I'm not sure if this would be a concern for
> > anx7625. I'll let Xin to comment on this.
> The flow of resolution switch is:
>     -> .atomic_disable().
>         -> .atomic_enable().
>             -> HDCP status check and enable.
> As resolution switch is triggered by user space, at this moment, the HDCP
> works fine(cannot get disabled event from HW).
> So, I think, .atomic_disable() is
> the best place to do termination of authentication (turn HDCP status to DESIRED).
> This is the only entrance which can do termination of authentication, I cannot
> find others.

I'm probably still missing a reference to the standard (please excuse
me, I don't know it by heart). Is it a standard requirement to
reauthenticate on the mode switch?

BTW, while reading the anx7625 code, I noticed that there is a possible
race between .atomic_enable() / .atomic_disable() which set
ctx->connector and the hdcp_check_work_func() which reads ctx->connector
without any protection.

-- 
With best wishes
Dmitry