2002-02-11 13:30:04

by Jeff Garzik

[permalink] [raw]
Subject: RFC: scheduler, and per-arch switch_to

Do we really care about the third arg to the switch_to() macro?

IMHO it would be nice to define the architecture context switch
interface like

void switch_to(struct thread_info *from, struct thread_info *to);

because we don't really seem to do much with the third arg, AFAICS.

--
Jeff Garzik | "I went through my candy like hot oatmeal
Building 1024 | through an internally-buttered weasel."
MandrakeSoft | - goats.com


2002-02-11 14:08:28

by Thomas Capricelli

[permalink] [raw]
Subject: Re: RFC: scheduler, and per-arch switch_to



I agree. While porting to zeta (see previous mail), i've had a look at some
different arch/ implementation of switch_to(), and i have not been able to
understand why the third arg was given.
Anyway, I don't use it, and I'm happy with this new interface.

Thomas



On Monday 11 February 2002 14:29, Jeff Garzik wrote:
> Do we really care about the third arg to the switch_to() macro?
>
> IMHO it would be nice to define the architecture context switch
> interface like
>
> void switch_to(struct thread_info *from, struct thread_info *to);
>
> because we don't really seem to do much with the third arg, AFAICS.

2002-02-11 14:17:35

by Russell King

[permalink] [raw]
Subject: Re: RFC: scheduler, and per-arch switch_to

On Mon, Feb 11, 2002 at 08:29:36AM -0500, Jeff Garzik wrote:
> Do we really care about the third arg to the switch_to() macro?
>
> IMHO it would be nice to define the architecture context switch
> interface like
>
> void switch_to(struct thread_info *from, struct thread_info *to);
>
> because we don't really seem to do much with the third arg, AFAICS.

It used to be a method to get the previously running task struct so some
cleanup could be done after the actual switch. Before you think about
"oh, that's prev anyway" think about what happens when "prev" is in some
random compiler defined CPU register, and your switch_to function saves
and restores all CPU registers.

In our current implementation, it looks like the third arg is no longer
necessary - Ingo?

--
Russell King ([email protected]) The developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html

2002-02-11 21:49:56

by Ingo Molnar

[permalink] [raw]
Subject: Re: RFC: scheduler, and per-arch switch_to


On Mon, 11 Feb 2002, Jeff Garzik wrote:

> Do we really care about the third arg to the switch_to() macro?

we dont need it anymore, i'll remove it in my next patchset (if you dont
beat me at it).

Ingo