Re: [Qemu-devel] [PATCH V15 4/5] i386: add a Virtual Machine Generation ID device

[Gal Hammer <ghammer@redhat.com>]

On 03/06/2015 19:37, Paolo Bonzini wrote:
>
> Since Michael said he'd merge this version, there's just one thing I
> would change.  You can access the region only one byte at a time, and
> let the memory core fix endianness here:

I've applied the change offered below. Thanks.

There are two unanswered questions:

1. You wrote that I should not use the memory_region_init_io() function. 
Should I use memory_region_init_ram() instead?

2. Is it possible to create a sysbus device using the "-device" command 
line argument? I vaguely recall that it is not possible to do it and 
that's the reason that I specifically add the device in the pc init.

Thanks,

     Gal.

> On 27/04/2015 13:19, Gal Hammer wrote:
>> +static uint64_t vmgenid_ram_read(void *opaque, hwaddr addr,
>> +                                 unsigned size)
>> +{
>> +    VmGenIdState *s = VMGENID(opaque);
>> +    uint64_t value;
>> +
>> +    memcpy(&value, s->guid + addr, size);
>
> value = s->guid[addr];
>
>> +    return value;
>> +}
>> +
>> +static const MemoryRegionOps vmgenid_ram_ops = {
>> +    .read = vmgenid_ram_read,
>> +    .valid = {
>> +        .min_access_size = 4,
>> +        .max_access_size = 4,
>> +    },
>
> And instead of this .valid declaration, use this:
>
>      .valid.min_access_size = 1,
>      .valid.max_access_size = 4,
>      .impl.min_access_size = 1,
>      .impl.max_access_size = 1,
>
>
>> +    .endianness = DEVICE_NATIVE_ENDIAN,
>
>      .endianness = DEVICE_LITTLE_ENDIAN
>
>
>> +};
>
> Thanks,
>
> Paolo
>