[PATCH V5 2/6] Add shared update_fdt() function for ARM/ARM64

Roy Franz roy.franz at linaro.org
Mon Dec 2 18:09:26 EST 2013


On Mon, Dec 2, 2013 at 6:56 AM, Mark Salter <msalter at redhat.com> wrote:
> On Fri, 2013-11-29 at 11:30 +0000, Matt Fleming wrote:
>> >       /*
>> >        * Add an additional efi_memory_desc_t because we're doing an
>> >        * allocation which may be in a new descriptor region.
>> > +      * We allocate as EFI_RUNTIME_SERVICES_DATA since this is what
>> > +      * we want for when we pass the memory map to the kernel.  This
>> > +      * function is also used to get the memory map for other uses,
>> > +      * but is always freed by the stub so the allocation type
>> > +      * doesn't matter.
>> >        */
>> >       *map_size += sizeof(*m);
>> >       status = efi_call_phys3(sys_table_arg->boottime->allocate_pool,
>> > -                             EFI_LOADER_DATA, *map_size, (void **)&m);
>> > +                             EFI_RUNTIME_SERVICES_DATA, *map_size,
>> > +                             (void **)&m);
>> >       if (status != EFI_SUCCESS)
>> >               goto fail;
>>
>>
>> OK, this needs a stronger justification. Presumably the reason for this
>> change is that you want the allocation to hang around once the kernel is
>> running? We have this problem on x86 and it's solved by reserving the
>> memory early in the kernel, e.g. memblock_reserve().
>> EFI_RUNTIME_SERVICES_DATA is for firmware use, not for kernel data.
>>
>
> Yeah, arm64 reserves it and is okay with it being EFI_LOADER_DATA.
>
> --Mark

I have discussed this with Leif and will change this back to
EFI_LOADER_DATA, and Leif will update
his runtime services patch to reserve this as done in your arm64 implementation.

Thanks,
Roy



More information about the linux-arm-kernel mailing list