On 01/04/2020 14:33, Laurent Pinchart wrote:
Hi Tomi,
On Wed, Apr 01, 2020 at 02:30:25PM +0300, Tomi Valkeinen wrote:
On 25/02/2020 17:01, Laurent Pinchart wrote:
On Tue, Feb 25, 2020 at 12:20:40AM +0100, Sebastian Reichel wrote:
This drops the virtual channel logic. Afterwards DSI clients request their channel number and get the virtual channel with the same number or -EBUSY if already in use.
I wonder why this level of indirection was used, allocating "virtual VCs". A single virtual indirection should be enough :-) I may be missing some context though, I'll defer that to Tomi, but for me,
I haven't reviewed the code yet, and it's been a long time since I wrote this code. But maybe this explains at least some:
(I hope I remember this right)
DSI packets have virtual channel IDs (VCID). That's number 0-3 that needs to be in the packets.
DSI IP has virtual channel "blocks" (VC), with associated registers. So 4 VC register blocks. These are not related to DSI virtual channel IDs in any way.
To do DSI transactions, you choose a VC, and program it. A VC can send data via video pipeline, or transmit and receive data messages created with CPU. And in both cases, you need to include the VCID in the transmissions, of course.
So, I think a normal use case could be a single panel, with VCID 0. To send video data and control messages, you would use VC0 and VC1. VC0 would be configured for video data, and VC1 would be configured for control messages.
And if I recall right, currently you first request a free VC from the IP with request_vc(). Then you use set_vc_id(channel, id) to set the VCID, used when doing transactions with that VC.
So the virtual channel naming is pretty confusing in the DSI IP, in my opinion.
I wasn't aware of those details, thank you for the explanation. It's quite confusing indeed, let's try to document the architecture in a comment block at the beginning of the dsi.c file for later reference.
But also, I think there's much room for cleanups and improvements. I don't think we have ever really supported multiple DSI peripherals, even in theory. So just one peripheral, with VCID always 0.
Even if we need two VCs to manage that single peripheral (I think that's often the case, we want one VC for video, one for control), we could fully hide that detail into the driver. This won't work with more than 2 DSI peripherals, but I think we can just say the driver supports a single peripheral, and that's it.
But with a quick browsing of this patch, I don't think it does it right, as it looks to me that the patch makes VCID == VC.
Tomi