Re: [Qemu-devel] Linux kernel configurations for various Microblaze systems

[Guenter Roeck <linux@roeck-us.net>]

On 06/27/2018 09:12 PM, Edgar E. Iglesias wrote:
> On Wed, Jun 27, 2018 at 05:34:15PM -0700, Guenter Roeck wrote:
>> Hi Edgar,
>>
>> On 06/27/2018 03:47 PM, Edgar E. Iglesias wrote:
>>
>>>
>>> Hi Guenter,
>>>
>>> Re: ml605
>>> The thing to watch out for with kernel configs for microblaze systems is the
>>> physmem address setting. Unfortunately the kernel does not auto-detect this.
>>>
>>> I'm attaching a kernelconfig for the ml605 that works for me and a qemu commandline.
>>>
>>> qemu-system-microblazeel -M petalogix-ml605 -m 256 -serial mon:stdio -dtb petalogix-ml605.dtb -kernel arch/microblaze/boot/linux.bin -append rdinit=/bin/sh
>>>
>>
>> Can you also send me the dtb file or even better its source ? I don't think
>> the one in qemu is correct. It lists the serial port at 0x83e00000, but qemu
>> instantiates it at 0x83e01000.
> 
> I used the one from QEMU. IIRC, there's some offsetting going on there
> that differs between the Xilinx version and the PC version of these
> UARTs...

Yes, I noticed after I sent the previous e-mail.

> 
> Can you try again and see how it goes?
> If it doesn't work, I can share my kernel Image and we can compare from there.
> 

I tried lots of things, nothing worked. Until I got your e-mail, tried again,
and suddenly it works just great. Go figure. Turns out I had used a big endian
toolchain to build a little endian target. Shameface.

I do get this odd gcc warning:

cc1: warning: -mxl-multiply-high can be used only with -mcpu=v6.00.a or greater

Actual parameters are "-mcpu=v10.0 -mxl-multiply-high", so that doesn't really make
sense to me. I tried various toolchains, all with the same result. Any idea what
that is about ?

Thanks,
Guenter