On 2016/4/22 21:26, Andrew Jones wrote:
>> + core->flags = cpu_to_le32(1);
>> > + }
>> > + g_free(cpu_node);
>> > +
>> > + mem_base = guest_info->memmap[VIRT_MEM].base;
>> > + for (i = 0; i < nb_numa_nodes; ++i) {
>> > + mem_len = numa_info[i].node_mem;
>> > + numamem = acpi_data_push(table_data, sizeof *numamem);
>> > + numamem->type = ACPI_SRAT_MEMORY;
>> > + numamem->length = sizeof(*numamem);
>> > + memset(numamem->proximity, 0, 4);
>> > + numamem->proximity[0] = i;
> This is weird (but I see x86 does it too). The spec says proximity is
> "Integer that represents the proximity domain to which the processor
> belongs", and its 4 bytes. So why doesn't the structure define it as
> a uint32_t and then we'd just do
>
> numamem->proximity = cpu_to_le32(i);
>
> (adding Igor)
>
>> > + numamem->flags = cpu_to_le32(1);
>> > + numamem->base_addr = cpu_to_le64(mem_base);
>> > + numamem->range_length = cpu_to_le64(mem_len);
> How about moving acpi_build_srat_memory from hw/i386/acpi-build.c to
> somewhere in hw/acpi and reusing it?
>
Good point! Will do that.
Thanks,
--
Shannon
