Re: [PATCH 1/6] video: tegra: Add nvhost driver

["Terje Bergström" <tbergstrom-DDmLM1+adcrQT0dZR+AlfA@public.gmane.org>]

I'll reply to the rest later, but here some initial comments.

On 24.11.2012 21:04, Thierry Reding wrote:
> The usual way to do this is to adapt downstream kernels to upstream
> changes, not the other way around.

Yep, we're doing both ways - leveraging downstream and upstream code.

> Some more documentation about host1x would certainly be very welcome. If
> you write something up, maybe you can push to get it included in the TRM
> as well. The host1x chapter in the TRM is a bit sketchy and something
> more like a programming guide would be really useful.

Got it. I hope user space code would help here, as it would be a
definite explanation of how the thing is supposed to be used.

> Funny enough, the chapter about the 2D engine says that it is merely
> there to document the registers and explicitly not as a programming
> guide because it is expected that NVIDIA supplied drivers will always be
> used. I had a good laugh when I read that. =)

Ouch. :-)

> The benefit is that smaller changes that add a single particular feature
> are a lot easier to review and verify.

Yep, I'm working on this. I think I found a way.

> But that's precisely my point. If the hardware is called host1x, why not
> call the driver host1x as well. We do the same for all other hardware
> blocks. The I2C controller driver is named i2c-tegra just as the RTC
> driver is named rtc-tegra. Why should the driver for host1x not be named
> host1x?

Ok, I got it.

Terje