Re: [PATCH 18/18] x86: add multiboot2 protocol support for EFI platforms

[Jan Beulich]

>>> On 27.03.15 at 15:57, <address@hidden> wrote:
> On Fri, Mar 27, 2015 at 02:34:19PM +0000, Jan Beulich wrote:
>> >>> On 27.03.15 at 15:26, <address@hidden> wrote:
>> > On Fri, Mar 27, 2015 at 01:36:32PM +0000, Jan Beulich wrote:
>> >> >>> On 27.03.15 at 14:06, <address@hidden> wrote:
>> >> > On Tue, Mar 17, 2015 at 10:32:01AM +0000, Jan Beulich wrote:
>> >> >> >>> On 30.01.15 at 18:54, <address@hidden> wrote:
>> >> >> > +        /* Skip Multiboot2 information fixed part */
>> >> >> > +        lea     MB2_fixed_sizeof(%ebx),%ecx
>> >> >>
>> >> >> Let's please not add more assumptions than necessary about stuff
>> >> >> being below 4G.
>> >> >
>> >> > I am not sure what do you mean by that.
>> >>
>> >> See the 32-bit register used for addressing here (and in many more
>> >> places)?
>> >
>> > This is what I expected but I was confused because you were referring only
>> > here to this problem. Anyway, is it possible to do this in different way?
>> > Should we care if image is always loaded at 0x100000 right now? Even with
>> > Xen early boot code being relocatable loader could not load image higher
>> > than 0xffffffff - 14 MiB.
>>
>> I don't understand what you're alluding to. Just use 64-bit registers
>> for memory accesses and LEAs, and be done. This will result in smaller
>> code as a benefit.
> 
> Well... How can I do that here if processor is in 32-bit mode? Maybe,
> we could that things after switching to 64-bit mode. However, I think
> this requires separate patch to do these changes.

No, if the processor is in 32-bit mode, then using 32-bit registers is
fine of course. But I'm pretty certain I spotted at least some cases
where it looked like you used 32-bit registers in 64-bit mode.

Jan