Re: [Qemu-devel] [PATCH v11 5/6] arm: SoC model for Calxeda Highbank

[Mark Langsdorf]

On 01/19/2012 01:59 PM, Peter Maydell wrote:
> On 19 January 2012 19:58, Mark Langsdorf <address@hidden> wrote:
>> On 01/19/2012 01:44 PM, Peter Maydell wrote:
>>> I have a kernel now that seems to boot but then
>>> barfs with:
>>> Freeing init memory: 124K
>>> Kernel panic - not syncing: Attempted to kill init!
>>>
>>> but that's probably a misconfiguration in my kernel.
>>
>> You need to be passing -m 4089.
> 
> cam-vm-266:maverick:qemu-jeos$
> ~/linaro/qemu-from-laptop/qemu/arm-softmmu/qemu-system-arm -kernel
> zImage-with-dtb  -initrd build-arm/initramfs.img -M highbank -serial
> stdio -m 4089
> qemu: at most 2047 MB RAM can be simulated
> 
> I'd rather not have a model that doesn't work on 32 bit hosts
> if we can avoid it...

The memory size is determined by the dts:
http://lxr.linux.no/#linux+v3.2.1/arch/arm/boot/dts/highbank.dts#L60

So you should be able to set it to whatever you want as long as the
values match. I've tested a boot with -m 2047 and the memory range
going from 0 to 0x7ff00000. That required regenerating the dtb and
re-appending it to the kernel image, though. I don't know of an
easy to handle that on the command line.

Is that sufficient for me to resend the patch with the changes
covered below?

>> +    highbank_binfo.board_id = -1; /* provided by deviceTree */
>
> I'm still not sure about this.
>
> The linux Documentation/devicetree/booting-without-of.txt says
> there are two calling conventions for booting here, ATAGS and
> with-device-tree-blob, and that even with the device tree blob
> method it is still considered "good practise" to pass a valid
> machine number in r1. Can you point me to some documentation
> that says that -1 is OK now?

I can't, so I'll change the value. We only intend to support
device tree booting.

--Mark Langsdorf
Calxeda, Inc.