[Linux-ia64] Re: PAL entry point

[Linux-ia64] Re: PAL entry point

[David Mosberger]

>>>>> On Wed, 17 Apr 2002 15:09:51 +0200, Christian Hildner <christian.hildner@hob.de> said:

  Christian> Stephane, David, do you agree making the symbol
  Christian> pal_entry_point global (included in pal.S)? I would need
  Christian> access to the PAL routines directly from my program
  Christian> without using the provided functions.

I'm not sure that's a good idea.  PAL has very particular constraints
on the environment in which it can be called in.  Why aren't the
normal entry points (ia64_pal_call_static(), ia64_pal_call_stacked(),
ia64_pal_call_phys_static(), and ia64_pal_call_phys_stacked())
sufficient?

	--david

Re: [Linux-ia64] Re: PAL entry point

[Stephane Eranian]

Christian,

On Wed, Apr 17, 2002 at 07:54:14AM -0700, David Mosberger wrote:
> >>>>> On Wed, 17 Apr 2002 15:09:51 +0200, Christian Hildner <christian.hildner@hob.de> said:
> 
>   Christian> Stephane, David, do you agree making the symbol
>   Christian> pal_entry_point global (included in pal.S)? I would need
>   Christian> access to the PAL routines directly from my program
>   Christian> without using the provided functions.
> 
> I'm not sure that's a good idea.  PAL has very particular constraints
> on the environment in which it can be called in.  Why aren't the
> normal entry points (ia64_pal_call_static(), ia64_pal_call_stacked(),
> ia64_pal_call_phys_static(), and ia64_pal_call_phys_stacked())
> sufficient?

I don't think it is a good idea because by itself knowing the entry
point value does not allow you to simply call it. If you look
at pal.S, you see that there is much more involved in calling
a PAL procedure, especially for those labeled physical/static.
So I agree with David, that we should instead export the 4 call
stubs that will allow you (with less effort) to call PAL.

-- 
-Stephane

[Linux-ia64] Re: PAL entry point

[Christian Hildner]

David Mosberger schrieb:

> >>>>> On Wed, 17 Apr 2002 15:09:51 +0200, Christian Hildner <christian.hildner@hob.de> said:
>
>   Christian> Stephane, David, do you agree making the symbol
>   Christian> pal_entry_point global (included in pal.S)? I would need
>   Christian> access to the PAL routines directly from my program
>   Christian> without using the provided functions.
>
> I'm not sure that's a good idea.  PAL has very particular constraints
> on the environment in which it can be called in.  Why aren't the
> normal entry points (ia64_pal_call_static(), ia64_pal_call_stacked(),
> ia64_pal_call_phys_static(), and ia64_pal_call_phys_stacked())
> sufficient?
>
>         --david

The problem is that in the context from which I want to call PAL there is no register stack
available and so there is no possibility to call a regular kernel function. I want to call the
PAL in physical/static mode and I know that it's not the easiest way to do this but I think I
know what I am doing.

Christian

[Linux-ia64] Re: PAL entry point

[Christian Hildner]

David Mosberger schrieb:

> >>>>> On Tue, 23 Apr 2002 07:28:35 +0200, Christian Hildner <christian.hildner@hob.de> said:
>
>   Christian> The supplied stubs use register stack and
>   Christian> certain registers that I am using too. Furthermore I
>   Christian> cannot rely on this routines beeing unchanged in the
>   Christian> future. So I want to code my own mode switch stub.
>
> I see.  So why don't you grab the PAL entry point from the SAL system table?
> The address of that table is stored in the "efi" table (see asm-ia64/efi.h).
>
>
>         --david

Yes, that's it.

Christian