Re: [Xen-devel] [PATCH 00/18] x86: multiboot2 protocol support

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 04/02/15 09:51, Jan Beulich wrote:
>>>> On 04.02.15 at 10:04, <andrew.cooper3@xxxxxxxxxx> wrote:
>> On 03/02/2015 17:14, Daniel Kiper wrote:
>>> On Mon, Feb 02, 2015 at 09:28:49AM +0000, Jan Beulich wrote:
>>>>>>> On 30.01.15 at 18:54, <daniel.kiper@xxxxxxxxxx> wrote:
>>>>>   - xen.efi build will not so strongly depend
>>>>>     on a given GCC and binutils version.
>>>> While I can see the possibility of making the binutils version
>>>> dependency go away (by manually creating the PE header), I can't
>>>> see how you'd overcome the gcc one: The MS calling convention
>>>> support is still going to be needed (not having looked at the patches
>>> Right, I forgot about that one.
>>>
>>>> themselves yet, I can't see myself accepting the introduction of
>>>> stubs to convert between calling conventions).
>> How about __attribute__((ms_abi)) ?  It would appear to exist for this
>> purpose.
> But that's the point: Older compilers don't support it. And with
> compilers supporting it we need no stubs.

If the use of __attribute__((ms_abi)) was suitably contained within a
#ifdef CONFIG_EFI, I don't see an problem.  CONFIG_EFI could derive
primarily from a compiler version check if we don't wish to force a
minimum newer version of gcc.

One way or another, a newer set of tools is needed to build EFI support,
and a binary capable of both legacy and efi boot is quite desirable to have.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel