On 21/03/2019 00:58, Andrey Smirnov wrote: > Regardless of how it's going to be implemented in the end, there'd > have to be a way to specify which HPD input is being used. Which means
True. > a either a new DT binding or re-using already existing to be agreed on > by DT folks. It just seems to me that there exists a much stronger > case to solve this using existing "language" of GPIO references as > opposed to introducing some vendor specific binding serving just this > single purpose. If DT is supposed to be used to describe the HW, then, > IMHO, it might be the other way around, TC358767 is also a GPIO > expander and has to be modeled/implemented as such, whether anyone > would ever use it in such capacity fully isn't that significant. Yep. But few points: - TC358767 node will expose gpios and then it uses them itself. It does look slightly silly in the DT data =). It's not often when you create a reference from a node to itself. - This also needs irqchip implemention to support HPD irq. I have never written one, but I presume it's not too complex, but not trivial either. - All this adds quite a big amount of code, compared to only few lines of code if this is done internally. >> Then we have two cases 1) HPD connected to TC358767, 2) HPD goes >> directly to the SoC, or worded differently, HPD is handled by something >> else than TC358767. >> >> 1) was implemented in this current patch, and there's no real benefit >> with the gpiochip. It's somewhat confusing that the driver provides a >> gpiochip which the same driver then uses, for its internal functionality. >> >> 2) should actually not involve TC358767 driver at all as it's totally >> outside TC358767. >> > > There's already precedent for such usage in ti-tfp410.c, analogix/ and > andanalogix-anx78xx.c, so it's not unheard of. Yes, but I believe the direction has been to move away from that. >> If the HPD goes from the DP connector to the SoC, we should have the DP >> connector driver handle it. Currently that connector is in the TC358767 >> driver, but it should really be separated. >> > > Sure, there's definitely more than one way to solve this. > >> So... Obviously what's missing from the current patch is that we need to >> be able to say which of the two GPIOs are used for the HPD (if any). But >> I'm debating with myself whether gpiochip here is a sane choice or not. > > Yeah, maybe it'd be best to submit a patch to DT mailing list and see > what they have to say? Yep. I'll write the irchip support too, out of interest, and see what it looks like. But this has the feel of over-engineering. Tomi -- Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel