linux-kernel - Re: [PATCH v5 5/9] drm/msm/hdmi: turn mode_set into atomic

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <8dd4a43e-d83c-1f36-21ff-61e13ff751e7@quicinc.com>
Date: Thu, 20 Jun 2024 15:05:16 -0700
From: Abhinav Kumar <quic_abhinavk@...cinc.com>
To: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
CC: Andrzej Hajda <andrzej.hajda@...el.com>,
        Neil Armstrong
	<neil.armstrong@...aro.org>,
        Robert Foss <rfoss@...nel.org>,
        Laurent Pinchart
	<Laurent.pinchart@...asonboard.com>,
        Jonas Karlman <jonas@...boo.se>,
        Jernej
 Skrabec <jernej.skrabec@...il.com>,
        Maarten Lankhorst
	<maarten.lankhorst@...ux.intel.com>,
        Maxime Ripard <mripard@...nel.org>,
        Thomas Zimmermann <tzimmermann@...e.de>,
        David Airlie <airlied@...il.com>, Daniel Vetter <daniel@...ll.ch>,
        Rob Clark <robdclark@...il.com>, Sean Paul
	<sean@...rly.run>,
        Marijn Suijten <marijn.suijten@...ainline.org>,
        <dri-devel@...ts.freedesktop.org>, <linux-arm-msm@...r.kernel.org>,
        <freedreno@...ts.freedesktop.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v5 5/9] drm/msm/hdmi: turn mode_set into atomic_enable



On 6/20/2024 2:24 PM, Dmitry Baryshkov wrote:
> On Thu, Jun 20, 2024 at 01:49:33PM GMT, Abhinav Kumar wrote:
>>
>>
>> On 6/20/2024 1:32 PM, Dmitry Baryshkov wrote:
>>> On Thu, Jun 20, 2024 at 01:27:15PM GMT, Abhinav Kumar wrote:
>>>>
>>>>
>>>> On 6/7/2024 6:23 AM, Dmitry Baryshkov wrote:
>>>>> The mode_set callback is deprecated, it doesn't get the
>>>>> drm_bridge_state, just mode-related argumetns. Turn it into the
>>>>> atomic_enable callback as suggested by the documentation.
>>>>>
>>>>
>>>> mode_set is deprecated but atomic_mode_set is not.
>>>
>>> There is no atomic_mode_set() in drm_bridge_funcs. Also:
>>>
>>
>> Please excuse me. I thought since encoder has atomic_mode_set(), bridge has
>> one too.
>>
>>> * This is deprecated, do not use!
>>> * New drivers shall set their mode in the
>>> * &drm_bridge_funcs.atomic_enable operation.
>>>
>>
>> Yes I saw this note but it also says "new drivers" and not really enforcing
>> migrating existing ones which are using modeset to atomic_enable.
> 
> Well, deprecated functions are supposed to be migrated.
> 

Along with rest of the pieces of the driver :)

>> My concern is that today the timing engine setup happens in encoder's
>> enable() and the hdmi's timing is programmed in mode_set().
>>
>> Ideally, we should program hdmi's timing registers first before the
>> encoder's timing.
>>
>> Although timing engine is not enabled yet, till post_kickoff, this is
>> changing the sequence.
>>
>> If this really required for rest of this series?
> 
> No, it was just worth doing it as I was doing conversion to atomic_*
> anyway. I can delay this patch for now.
> 
> Two questions:
> 
> 1) Are you sure regarding the HDMI timing registers vs INTF order? I
> observe the underrun issues while changing modes on db820c.
> 

Yes this is the order I see in the docs:

1) setup HDMI PHY and PLL
2) setup HDMI video path correctly (HDMI timing registers)
3) setup timing generator to match the HDMI video in (2)
4) Enable timing engine

This change is mixing up the order of (2) and (3).

> 2) What should be the order between programming of the HDMI timing
> engine and HDMI PHY?
> 

Mentioned above.

> What would be your opinion on moving HDMI timing programming to
> atomic_pre_enable? (the exact place depends on the answer on the second
> question).
> 

So

-> bridge->pre_enable() : Do HDMI timing programming
-> encoder->atomic_enable() : Do timing engine programming
-> post_kickoff() does timing engine enable

This matches the steps I wrote above.

Hence this is fine from my PoV.

>>
>>>>
>>>> I would rather use atomic_mode_set because moving to atomic_enable() would
>>>> be incorrect.
>>>>
>>>> That would be called after encoder's enable and hence changes the sequence.
>>>> That was not the intention of this patch.
>>>>
>>>> NAK.
>>>>
>>>>> Acked-by: Maxime Ripard <mripard@...nel.org>
>>>>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
>>>>> ---
>>>>>     drivers/gpu/drm/msm/hdmi/hdmi_bridge.c | 33 ++++++++++++++++++++++++++-------
>>>>>     1 file changed, 26 insertions(+), 7 deletions(-)
>>>
>>>
>