[Lenny400] jz47xx Device Tree

Paul Boddie paul at boddie.org.uk
Sun Apr 5 21:12:54 CEST 2015


On Sunday 5. April 2015 19.34.29 Dr. H. Nikolaus Schaller wrote:
> After doing some more research, I think the status of MIPS DT is even more
> advanced than I had thought:
> 
> 1. 4.0-rc already supports a handful of DT based MIPS CPUs:
> 	
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/
> mips/boot/dts?id=refs/tags/v4.0-rc6

First of all, it's good to know that you are an expert in navigating the 
labyrinth of kernels, all (mostly) alike.

> 2. there are patches circulating on LKML to convert jz4740 to DT:
> 	https://lkml.org/lkml/2015/2/4/392
> 
> 	unfortunately they did end up - as far as I can see - in a social issue
> (within imgtec?) https://lkml.org/lkml/2015/2/4/463

I can imagine that this made for an awkward day at the office for someone. :-)

Having seen DT-related activity in the qi-kernel repository, I wonder how much 
of that stuff has been pulled into the imgtec repository.

> 	There has been some more activity here:
> 	https://patchwork.kernel.org/project/LKML/list/?q=jz4780
> 
> 	Some of these patches have already appeared in linux-next (so they will
> become part of 4.1-rc1 in ca. 4 weeks):
> https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/log/?id=r
> efs%2Ftags%2Fnext-20150402&qt=grep&q=jz4740

Well, that's quite recent, so it is encouraging.

> 	Driver for his activity is the JZ4780 CI20 board:
> http://www.elinux.org/CI20_upstream e.g.
> https://github.com/MIPS/CI20_linux/blob/ci20-v3.18/arch/mips/boot/dts/jz47
> 40.dtsi
> 
> 	The really good thing is that this is done more or less officially by
> Imgtec - so there is manpower behind.

My feeling is that they, as the effective owners of the MIPS platform, have to 
step up and make it a viable platform. It would be nice if their own PowerVR 
stuff wasn't highly proprietary, too, and maybe after enough criticism they 
will cooperate with people wanting to support that technology with Free 
Software drivers, especially since they will undoubtedly want to encourage 
their MIPS licencees to incorporate it into their products.

> Since we now know that the jz4730 also has some similarities as well, we
> should IMHO try to help to get the jz4740 DT patches upstream. Instead of
> tweaking the non-DT jz4740 code and #include like Paul’s patch does .
> Which might then lead soon (3-6 months?) to merge conflicts with
> linus/master.
> 
> And just add our own jz4730.dtsi and jz4730-minipc.dts and some jz4730
> specific drivers.
> 
> Seems to be easier and future-proof for me than starting to work on any
> non-DT code.

I agree with this. Maybe it will lead to something that actually works. ;-)

> But we should definitively use Paul’s patch as the basis for understanding
> what we have to change from jz4740 to jz4730.
> 
> What do you think?

I think Daniel and others with some actual experience with the jz4730 may also 
be able to provide insight into the mechanisms of the CPU that I haven't 
managed to grasp. Apart from the datasheet and attempting to understand the 
2.6 kernel code, and apart from looking at other documents about the other 
CPUs in the jz47xx series, I don't really have any special knowledge to share.

Paul


More information about the Lenny400 mailing list