| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| hitech95 | Is this the right V4L2 or is just for DVB? I have a question about V4L2 CEC. I'm not sure if my question is more for the v4l2 community or DRM. I'm trying to implement the HDMI CEC HAL for Android. One of the requirement is the "isConnected" method thet return true if a device is connected to an HDMI. How can I know (from userspace) if an HDMI
display is connected? | [13:45] |
| pinchartl | hitech95: and HDMI output I assume ? that's more of a DRM topic. although I'm not sure if the CEC API also exposes that. that's a question for hverkuil | [13:48] |
| hverkuil | hitech95: I'm the CEC guy :-), so this is the right space.
CEC is both for inputs and outputs :-) Who needs to provide the isConnected method? Kernel or userspace? (I speak CEC, but not Android) Note: we are talking about the linux kernel CEC framework, right? If it is a vendor-specific implementation, then you are on your own... | [13:48] |
| hitech95 | hverkuil, Thanks forreply :) Over freedesktop they told me to use libdrm and call drmModeGetConnectorCurrent usign the connector ID received using the IOCTL CEC_ADAP_G_CONNECTOR_INFO. My HDMI are outputs so the connectors are managed by DRM.
I'm moving the my "second" steps after starting working on kernel simple modules (like i2c stuff) while also learning android HAL. So be kind :) "isConnected" is A method that i have to implement for android, they give me a port ID and I have to return true or false. | [13:52] |
| hverkuil | Ah, yes, that's my framework. | [13:53] |
| pinchartl | hverkuil: does the CEC UAPI provide HDMI connection status, or is that a task handled by DRM/KMS only ? | [13:53] |
| hitech95 | Yes, linux kernel CEC via V4L! (DRM and HDMI CEC has been implemented by the RPI fundadtion recently) | [13:54] |
| hverkuil | pinchartl: somewhat. It gives you the current physical address, which is similar, but not the same, as the connection status. If you see a physical address, then you are connected to an HDMI display. If not, then either you are disconnected, or the Display has no physical address defined in the EDID.
You can use the connector info to find the associated CEC device for a drm connector. hitech95: Do you have a pointer to documentation for that isConnected method? The devil is in the details... | [13:55] |
| hitech95 | hverkuil, yes: https://android.googlesource.com/platform/hardware/interfaces/+/master/tv/cec/1.0/IHdmiCec.hal#165 they say to monitor the 5V line buut AFIK sources have always 5V when on. And after looking into some other vendor implementation they return true or false based on HDMI state if the HDMI is an output. (Android TV can also be installed
on TVs that have HDMI inputs) The RPI adapter name is hardcoded: https://github.com/raspberrypi/linux/blob/rpi-5.10.y/drivers/gpu/drm/vc4/vc4_hdmi.c#L2371 Having two HDMIs this could couse some issues? | [14:02] |
| hverkuil | No, the connector info is filled here:
cec_fill_conn_info_from_drm(&conn_info, &vc4_hdmi->connector); cec_s_conn_info(vc4_hdmi->cec_adap, &conn_info); which is different for the two outputs. Yeah, the mention of 5V is weird. I think the Android CEC HAL is geared towards HDMI inputs, not outputs. Hmm, or perhaps they meant HPD instead of 5V? I suspect that's what they meant. | [14:05] |
| hitech95 | No idea (I'm just doing a DIY project during the pandemic) it can work also as output: https://android.googlesource.com/platform/hardware/interfaces/+/master/tv/cec/1.0/types.hal#164
To be precide only google know how the HAL must be implemented, Their description are garmage 90% of the time. BTW my only way to read HPD is via libDRM. | [14:07] |
| hverkuil | yes, drm knows the connection status.
got to go, need to do some shopping. | [14:12] |
| hitech95 | hverkuil, For what I've undertood you left out the connector status due tothe fact that an HDMI can also be an input and this case is not present in the frameworkand you left to userspace to check the hdmi status (if needed)
Thanks again for the help, happy shopping! | [14:14] |
| ...... (idle for 28mn) | ||
| hverkuil | hitech95: actually, I left out the connector status because it is 1) the job of drm to give you that, and 2) not even relevant for CEC since there are displays that pull down the HPD when in standby, but you can still use CEC to wake them up.
See the section "CEC Without HPD" here: https://hverkuil.home.xs4all.nl/cec-status.txt | [14:42] |
| hitech95 | hverkuil, a easier way would it be to see if a phyaddress is set. Is this a good way for an implementation? (looks like that you have to be drm master to read connector status :( and android HW composer is the master) | [14:44] |
| hverkuil | It depends on how isConnected is used. You could try to just return 1 there, since you can use CEC even if the HPD is low.
Alternatively, you can indeed use the presence of a valid physical address. But if CEC is only enabled if there is a PA, then you'll never be able to wake up displays that pull down the HPD when in standby. | [14:46] |
| hitech95 | I'll give it a try. Since it is also my first time writing stuff for android and using c++. (I'm mainly a java and javascript dev X) )
So a device could not have a phyAdd but you can still talk to it? Using broadcast address? So a stby display could have HPD set or not but CEC is available in any case. So DRM would fail on reporting in any case. by only checking the PhyADDr i could end up in a situation where the HDP is set but the address is not (STB) and so I cannot wakeup the display. Got that! | [14:58] |
| hverkuil | If there is not HPD then you send the Image View On message from address 15 ('Unregistered') to 0 (TV). | [15:00] |
| hitech95 | OH ok, got that! | [15:01] |
| ................... (idle for 1h31mn) | ||
| BWhitten | pinchartl, just saw your last message I would be very keen to know when you have a clean ISI driver for upstream, the #linux-imx upstreaming channel may be interested as well. May I PM you my email for CC on the series? | [16:32] |
| pinchartl | sure | [16:33] |
| ................................ (idle for 2h37mn) | ||
| hverkuil: I recall we had a discussion in the past about begin/commit operations for V4L2 control handlers, to allow implementing support for grouped writes on devices that support them (for instance, some sensors have a HOLD bit that can be set to freeze registers from being applied to the hardware, to allow configuring multiple parameters in an atomic way). I don't recall the outcome of that discussion
though :-S is it a feature you would accept ? | [19:10] | |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |