Re: [PATCH 14/23] make section names compatible with -ffunction-sections -fdata-sections: parisc

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Wed, 2008-07-02 at 02:00 +0200, Denys Vlasenko wrote:
> On Wednesday 02 July 2008 01:41, James Bottomley wrote:
> > On Wed, 2008-07-02 at 02:39 +0200, Denys Vlasenko wrote:
> > > The purpose of this patch is to make kernel buildable
> > > with "gcc -ffunction-sections -fdata-sections".
> > > This patch fixes parisc architecture.
> > > 
> > > Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
> > 
> > Um ... if you look at the Makefile you'll see we already build parisc
> > with -ffunction-sections; we have to: our relative jumps are too small
> > to guarantee finding the stubs in large files.
> > 
> > Since our text is -ffunction-sections compatible already, I question the
> > need for transformations like this:
> > 
> > > -               *(.text.do_softirq)
> > > -               *(.text.sys_exit)
> > > -               *(.text.do_sigaltstack)
> > > -               *(.text.do_fork)
> > > +               *(.do_softirq.text)
> > > +               *(.sys_exit.text)
> > > +               *(.do_sigaltstack.text)
> > > +               *(.do_fork.text)
> 
> arch/parisc/kernel/vmlinux.lds.S contains these lines:
> 
>                TEXT_TEXT
>                SCHED_TEXT
>                LOCK_TEXT
>                *(.text.do_softirq)
>                *(.text.sys_exit)
>                *(.text.do_sigaltstack)
>                *(.text.do_fork)
> 
> which suggested to me that for parisc it is important to have
> these sections in that place (after LOCK_TEXT) and order.

Yes ... moderately.

> If you use -ffunction-sections, any function with the name
> do_fork (say, a static function somewhere) will end up in
> .text.do_fork function, and will be "mixed up" with
> global do_fork(). For parisc it is maybe not a problem
> (I am not an expert) but in other places/arches people
> clearly would not want this kind of things to happen.

Erm, but we're trying to name text sections of the -ffunction-sections.
We're doing this because we are trying to do a semblance of a reasonable
arrangement for the relative jumps (and avoid stubs).  Your patch is
explicitly breaking all of this.

> In order to handle these situations uniformly, in these patches
> I decided to _never_ use .text.XXXX names for sections,
> effectively leaving them "reserved for gcc's use".

We need to use the names gcc outputs, not some random name.

> Did I understand you right that in this chunk I need to
> leave .text.FUNC_NAME as it was before?

Yes.

> > And thus by the same token the data transformations.
> 
> It would be easiest for me if you will reply to the parisc patch
> and indicate all parts where I should NOT do name change.

That would be the piece above.

James