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

[Peter Maydell]

On 6 January 2012 16:58, Mark Langsdorf <address@hidden> wrote:
> On 01/06/2012 10:29 AM, Peter Maydell wrote:
>>> +    /* Override default RAM size */
>>> +    if (ram_size == 0x8000000) {
>>> +        if (sizeof(long) == 8) {
>>> +            ram_size = 0xff900000;
>>> +        } else {
>>> +            ram_size = 0x80000000;
>>> +        }
>>
>> Yuck. Model behaviour shouldn't depend on properties of
>> the host system like sizeof(long).
>
> The board is populated with 4G of DRAM, which we'd like
> to support if the host can. Is there a better way to do
> that?

Don't mess with the default, have the user specify a RAM size that
makes sense for their host machine. It's not fantastic but it's what
QEMU has at the moment.

I once tried to suggest a patchset which would allow boards to
specify their min/max/default RAM sizes but it got shot down.

>>> +    sysmem = get_system_memory();
>>> +    dram = g_new(MemoryRegion, 1);
>>> +    memory_region_init_ram(dram, "highbank.dram", ram_size);
>>> +    /* SDRAM at address zero.  */
>>> +    memory_region_add_subregion(sysmem, 0, dram);
>>> +
>>> +    sysram = g_new(MemoryRegion, 1);
>>> +    memory_region_init_ram(sysram, "highbank.sysram", 0x8000);
>>> +    memory_region_add_subregion(sysmem, 0xfff88000, sysram);
>>> +    if (load_image_targphys("sysram.bin", 0xfff88000, 0x8000) < 0) {
>>> +            fprintf(stderr, "Unable to load sysram.bin\n");
>>> +    }
>>
>> Is this for some sort of BIOS-image equivalent?
>
> Yes, uboot and the like. It isn't necessary to boot the system,
> but it models the actual board a bit better.

If it's not necessary, should we be failing if there isn't an image file
present?

-- PMM