Re: multiboot2 structs and packed attribute

["Vladimir 'phcoder' Serbinenko" <phcoder@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1864 bytes --]

On Tuesday, March 8, 2016, Daniel Kiper <daniel.kiper@oracle.com> wrote:

> On Mon, Mar 07, 2016 at 08:42:38PM +0000, Vladimir 'phcoder' Serbinenko
> wrote:
> > Le lun. 7 mars 2016 14:35, Daniel Kiper <daniel.kiper@oracle.com
> <javascript:;>> a écrit :
> >
> > > Hi,
> > >
> > > Does anybody know why only multiboot2 multiboot_mmap_entry struct
> > > has packed attribute? Is it by design or by mistake? I think that
> > > we should use packed attribute for every multiboot2 protocol related
> > > struct. If not then we should probably remove this attribute from
> > > multiboot_mmap_entry struct too. What do you think about that?
> > >
> > We put packed only when it changes relative offsets of members or size of
> > the structure. Is it the case for structures you have in mind?
>
> I do not have anything specific. However, it looks like stray stuff.
>
> Hmmm... I took a look at multiboot_mmap_entry once again and it seems
> that this struct really does not need packed attribute. All members are
> in right order and even compiler should not add padding. Am I right?
>
Yes, good catch.

>
> Anyway, maybe we should add somewhere in the header comment saying that
> all/most of multiboot2 structures are build in a way that fulfill alignment
> requirements out of the box and compiler does not need to do any
> rearrangements.
> However, we should warn that new extensions should be added carefully and
> some structs may require GRUB_PACKED attribute in the future to properly
> represent added interface.
>
This probably should go to spec rather than the header file.

>
> Or we can add GRUB_PACKED to every struct. However, this seems redundant
> here.
> There is no chance that somebody will rearrange, especially blindly,
> existing structs.
>
> Daniel
>


-- 
Regards
Vladimir 'phcoder' Serbinenko

[-- Attachment #2: Type: text/html, Size: 2508 bytes --]