[RFC PATCH 3/3] encoder-tpd12s015: keep the ls_oe_gpio on while the phys_addr is valid

Hans Verkuil hverkuil at xs4all.nl
Sat Apr 8 10:11:05 UTC 2017

Hi Tomi,

On 05/10/2016 01:36 PM, Tomi Valkeinen wrote:
> Hi Hans,
> On 29/04/16 12:39, Hans Verkuil wrote:
>> From: Hans Verkuil <hans.verkuil at cisco.com>
>> As long as there is a valid physical address in the EDID and the omap
>> CEC support is enabled, then we keep ls_oe_gpio on to ensure the CEC
>> signal is passed through the tpd12s015.
>> Signed-off-by: Hans Verkuil <hans.verkuil at cisco.com>
>> Suggested-by: Tomi Valkeinen <tomi.valkeinen at ti.com>
>> ---
>>  drivers/gpu/drm/omapdrm/displays/encoder-tpd12s015.c | 13 ++++++++++++-
>>  1 file changed, 12 insertions(+), 1 deletion(-)
>> diff --git a/drivers/gpu/drm/omapdrm/displays/encoder-tpd12s015.c b/drivers/gpu/drm/omapdrm/displays/encoder-tpd12s015.c
>> index 916a899..efbba23 100644
>> --- a/drivers/gpu/drm/omapdrm/displays/encoder-tpd12s015.c
>> +++ b/drivers/gpu/drm/omapdrm/displays/encoder-tpd12s015.c
>> @@ -16,6 +16,7 @@
>>  #include <linux/platform_device.h>
>>  #include <linux/gpio/consumer.h>
>> +#include <media/cec-edid.h>
>>  #include <video/omapdss.h>
>>  #include <video/omap-panel-data.h>
>> @@ -65,6 +66,7 @@ static void tpd_disconnect(struct omap_dss_device *dssdev,
>>  		return;
>>  	gpiod_set_value_cansleep(ddata->ct_cp_hpd_gpio, 0);
>> +	gpiod_set_value_cansleep(ddata->ls_oe_gpio, 0);
>>  	dst->src = NULL;
>>  	dssdev->dst = NULL;
>> @@ -142,6 +144,7 @@ static int tpd_read_edid(struct omap_dss_device *dssdev,
>>  {
>>  	struct panel_drv_data *ddata = to_panel_data(dssdev);
>>  	struct omap_dss_device *in = ddata->in;
>> +	bool valid_phys_addr = 0;
>>  	int r;
>>  	if (!gpiod_get_value_cansleep(ddata->hpd_gpio))
>> @@ -151,7 +154,15 @@ static int tpd_read_edid(struct omap_dss_device *dssdev,
>>  	r = in->ops.hdmi->read_edid(in, edid, len);
>> -	gpiod_set_value_cansleep(ddata->ls_oe_gpio, 0);
>> +	/*
>> +	 * In order to support CEC this pin should remain high
>> +	 * as long as the EDID has a valid physical address.
>> +	 */
>> +	valid_phys_addr =
>> +		cec_get_edid_phys_addr(edid, r, NULL) != CEC_PHYS_ADDR_INVALID;
>> +#endif
>> +	gpiod_set_value_cansleep(ddata->ls_oe_gpio, valid_phys_addr);
>>  	return r;
>>  }
> I think this works, but... Maybe it would be cleaner to have the LS_OE
> enabled if a cable is connected. That's actually what we had earlier,
> but I removed that due to a race issue:
> a87a6d6b09de3118e5679c2057b99b7791b7673b ("OMAPDSS: encoder-tpd12s015:
> Fix race issue with LS_OE"). Now, with CEC, there's need to have LS_OE
> enabled even after reading the EDID, so I think it's better to go back
> to the old model (after fixing the race issue, of course =).

So, this is a bit of a blast from the past since the omap4 CEC development
has been on hold for almost a year. But I am about to resume my work on this
now that the CEC framework was merged.

The latest code is here, if you are interested:


It's pretty much unchanged from the version I posted a year ago, just rebased.

But before I continue with this I have one question for you. First some

There is a special corner case (and I wasn't aware of that a year ago!) where
it is allowed to send a CEC message when there is *no HPD*.

The reason is that some displays turn off the hotplug detect pin when they go
into standby or when another input is active. The only way to communicate with
such displays is via CEC.

The problem is that without a HPD there is no EDID and basically no way for an
HDMI transmitter to detect that something is connected at all, unless you are
using CEC.

What this means is that if we want to implement this on the omap4 the CEC support
has to be on all the time.

We have seen modern displays that behave like this, so this is a real issue. And
this corner case is specifically allowed by the CEC specification: the Poll,
Image/Text View On and the Active Source messages can be sent to a TV even when
there is no HPD in order to turn on the display if it was in standby and to make
us the active input.

The CEC framework in the kernel supports this starting with 4.12 (this code is
in the panda-cec branch above).

If this *can't* be supported by the omap4, then I will likely have to add a CEC
capability to signal to the application that this specific corner case is not

I just did some tests with omap4 and I my impression is that this can't be
supported: when the HPD goes away it seems like most/all of the HDMI blocks are
all powered off and any attempt to even access the CEC registers will fail.

Changing this looks to be non-trivial if not impossible.

Can you confirm that that isn't possible? If you think this can be done, then
I'd appreciate if you can give me a few pointers.



More information about the dri-devel mailing list