Now partially booting on our CN58XX eval board

[Patrick Mahan]

M. Warner Losh wrote:
> In message: <4C77EB9F.4020705 at adaranet.com>
>             Patrick Mahan <pmahan at adaranet.com> writes:
> : We have it now starting to boot on our eval board with the 1.8.1 SDK
> : bootloader.
> :
> : It panic's on a bad address during a device attach, perhaps in the
> : ethernet driver setup?  I am attaching the output from the console
> : below.
> :
> : I do see in the console output
> :
> :   'Octeon SDK: Cavium Networks Octeon SDK version 1.9.0, build 312'
> :
> : Which I am assuming is the cavium SDK pieces being included as part of
> : the mips kernel.
> 
> That's right.  This is the cnusers.org SDK that we've included in the
> kernel.
> 
> : Warner, is this what you would see on happen on under 1.8.1?
> 
> No.  I have older uboot systems, and they refuse to properly load the
> kernel.  see below...
> 
> : I am attaching (again) my kernel config.  Is anything else needed?
> :
> : Patrick
> : Adara Networks
> 
> [trimmed]
> : octebus0: [FILTER]
> : Interface 1 has 4 ports (RGMII)
> : panic: cvmx_phys_to_ptr: mapping high address (0x2090a880) not yet
> : supported.
> 
> I am seeing this when I boot a 32-bit kernel on two of the three
> boards I have access to (both of them, btw, have older uboot loaders
> on them).
> 
> The problem here is that cvmx_phys_to_ptr only supports direct
> mapping, so we can only use the first 512MB of memory.  Juli and I
> have talked about clamping the available memory to using only the
> first 512MB on 32-bit kernels, but we don't like this idea and are
> casting about for others.
> 
> The 64-bit kernel doesn't suffer this limitation.
> 

Okay, however, we are currently stuck at 32-bits :-(

I have an engineer that thinks this would be fun to resolve, so I am letting
him run with this for now.  Is this an address coming from the mii layer?

Thanks,

Patrick
Adara Networks