Re: Zynq FDTisation [was Re: Cortex-A9 boot problems with d-cache on Zynq]

[Nick Hudson <nick.hudson%gmx.co.uk@localhost>]

On 13/06/2019 01:44, Robert Swindells wrote:
> Alexander Nasonov <alnsn%yandex.ru@localhost> wrote:
> > Nick Hudson wrote:
> > > To get the ball rolling I've added early Zynq-7000 support to GENERIC
> > >
> > > http://mail-index.netbsd.org/source-changes/2019/06/11/msg106540.html
> > >
> > > A clock driver is the next step.
> >
> > It panics on Zybo Z7-20:
>
> Works as expected for me on my Parallella, it panics much later in the
> boot sequence.
>
> BUT, I'm using a config file that only includes the drivers for the Zynq
> and different link commands.
>
> > Zynq> fatload mmc 0 100000 netbsd.ub
>
> Any *.ub file will have a small header prepended to it.
>
> If the image is linked at 0x100000 and you load the .ub file at that
> address it won't work.

u-boot isn't the problem here. The arm start code makes alignment
assumptions which break for Alexander's load address.


> That is why I prefer to use the netbsd.gz.ub file and load it at an
> address a long way above the kernel link address.
>
> U-boot will decompress it to 0x100000 before running it as part of the
> 'bootm' operation.

arm uses kernel_noload now which doesn't work (for some reason) with
gzipped images.


> I'll put my config files into CVS tomorrow.

There's no need, thanks.

Nick