Re: [PATCH v4 1/6] xen: introduce SIMPLE_DECL_SECTION

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Fri, Sep 27, 2024 at 11:07:58AM +0200, oleksii.kurochko@xxxxxxxxx wrote:
> On Fri, 2024-09-27 at 09:58 +0200, Roger Pau Monné wrote:
> > On Thu, Sep 26, 2024 at 06:54:20PM +0200, Oleksii Kurochko wrote:
> > > Introduce SIMPLE_DECL_SECTION to cover the case when
> > > an architecture wants to declare a section without specifying
> > > of load address for the section.
> > > 
> > > Update x86/xen.lds.S to use SIMPLE_DECL_SECTION.
> > 
> > No strong opinion, but I feel SIMPLE is not very descriptive.  It
> > might be better to do it the other way around: introduce a define for
> > when the DECL_SECTION macro should specify a load address:
> > DECL_SECTION_WITH_LADDR for example.
> In the next patch, two sections are introduced: dt_dev_info and
> acpi_dev_info. The definition of these sections has been made common
> and moved to xen.lds.h, and it looks like this:
>    +#define DT_DEV_INFO(secname)    \
>    +  . = ALIGN(POINTER_ALIGN);     \
>    +  DECL_SECTION(secname) {       \
>    +       _sdevice = .;           \
>    +       *(secname)              \
>    +       _edevice = .;           \
>    +  } :text
> (A similar approach is used for ACPI, please refer to the next patch in
> this series.)
> 
> For PPC, DECL_SECTION should specify a load address, whereas for Arm
> and RISC-V, it should not.
> 
> With this generalization, the name of DECL_SECTION should have the same
> name in both cases, whether a load address needs to be specified or not

Oh, sorry, I think you misunderstood my suggestion.

I'm not suggesting to introduce a new macro named
DECL_SECTION_WITH_LADDR(), but rather to use DECL_SECTION_WITH_LADDR
instead of SIMPLE_DECL_SECTION in order to signal whether
DECL_SECTION() should specify a load address or not, iow:

#ifdef DECL_SECTION_WITH_LADDR
# define DECL_SECTION(x) x : AT(ADDR(x) - __XEN_VIRT_START)
#else
# define DECL_SECTION(x) x :
#endif

Thanks, Roger.