On Sun, 03 Oct 2004 10:17:42 +0800, Bao zhao <paozhaokeats@xxxxxxxxxxx> wrote:
> hi, I'm sorry for misleading you because of my poor english. I'm talking
> about switch_to function,
> this function's first argument and third argument always are same ,so I
> think switch_to 's arguments is redundant.
>
> >From: manish regmi <regmi.manish@xxxxxxxxx>
> >Reply-To: manish regmi <regmi.manish@xxxxxxxxx>
> >To: Bao zhao <paozhaokeats@xxxxxxxxxxx>
> >Subject: Re: why does switch_to need three arguments?
> >Date: Sun, 3 Oct 2004 07:54:49 +0545
>
>
> >
> >On Sun, 03 Oct 2004 08:41:29 +0800, Bao zhao <paozhaokeats@xxxxxxxxxxx>
> >wrote:
> > > In 2.6.0 kernel,
> > >
> > > context_switch function is the only one which calls switch_to, and the
> >first
> > > argument and the third argument are same. Is it not redundant?
> > >
> > > static inline task_t * context_switch(runqueue_t *rq, task_t *prev,
> >task_t
> > > *next)
> > > {
> > > ....
> > > switch_to(prev, next, prev);
> > > ....
> > > }
> > >
> > > Thanks in advance !
> > >
> >
> >hi,
> > In 2.6 kernel, there is a per cpu runqueue. So, the first argument
> >is the runqueue, the second and third are the tasks to switch.
> >
> >regards Manish
>
Sorry,
That was my mistake.
As you see in switch_to, It call __switch_to which is a c function in
arch/asm/process.c. __switch_to returns the pointer to prev task after
changing the various stuffs.
No as you know in i386 the returned value in C is stored in EAX
register. So it is stored in last.
Now during times like ret_from_fork which is called by fresh forked
task. it requires the pointer to prev task on eax. And call to
schedule tail to finish fork .
pushl %eax
call schedule_tail
GET_THREAD_INFO(%ebp)
popl %eax
I think it is for times like this.
regards manish
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
