Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190521105203.154043-1-darekm@google.com> <16889839-b4e9-9984-2e36-5f07ceb7d7f2@xs4all.nl>
In-Reply-To: <16889839-b4e9-9984-2e36-5f07ceb7d7f2@xs4all.nl>
From:   Dariusz Marcinkiewicz <darekm@google.com>
Date:   Tue, 28 May 2019 09:53:23 +0200
Message-ID: <CALFZZQH7-mzRZbji1w-UwRLdw6c8+CThn0e96oG8E2Eq1ivANQ@mail.gmail.com>
Subject: Re: [PATCH v7 1/3] media: cec: expose HDMI connector to CEC dev mapping
To:     Hans Verkuil <hverkuil@xs4all.nl>
Cc:     linux-media@vger.kernel.org, hans.verkuil@cisco.com,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, May 24, 2019 at 11:21 AM Hans Verkuil <hverkuil@xs4all.nl> wrote:
>
> Hi Dariusz,
>
> I did some more testing with the Khadas VIM2 and found another problem,
> something that will, unfortunately, require some redesign.
>
...
>
> The other problem is in the CEC driver: it creates the CEC device as
> soon as the HDMI device is found (cec_notifier_parse_hdmi_phandle).
>
> But that doesn't mean that the HDMI device also had registered itself
> as a CEC notifier.
>
> Until now that never mattered: as long as the HDMI device was found
> the CEC adapter would function fine, it would just have no physical
> address until so notified by the HDMI device once it registered its
> CEC notifier.
>
> But if we want to have valid connector info during the lifetime of
> the CEC adapter, then this no longer works.
>
> I'm not entirely sure how to handle this.
>
> Another issue here is that when the HDMI driver removes the notifier,
> then it should also zero the connector info. Remember that both the
> HDMI and the CEC drivers can be loaded and unloaded independently from
> one another.
>
Given all of the above, what do you think about coming back to the v1
of the patch, where a connector info could be set on an adapter at any
time and an event was used to notify userland when that happened? That
approach seems to cover all the scenarios mentioned above.

Thank you for testing the patches!

Best regards.