Re: + fully-honor-vdso_enabled.patch added to -mm tree

[Oleg Nesterov <oleg@tv-sign.ru>]

On 03/05, Paul Mundt wrote:
>
> On Sat, Mar 03, 2007 at 01:18:54AM +0300, Oleg Nesterov wrote:
> > On 03/02, John Reiser wrote:
> > > Paul Mundt has commented on setup_rt_frame() and provided a patch which
> > > bullet-proofs that area.  I will include that patch into the next revision.
> > 
> > Confused. I still think his patch incomplete. Don't we need the same check
> > in setup_rt_frame() ?
> > 
> Yes, you're right. The patch is missing the proper restorer logic in the
> setup_rt_frame() case. This should handle both of them, restoring the
> previous logic when the VDSO is disabled.
> 
> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
> 
> --
> 
>  arch/i386/kernel/signal.c |   15 +++++++++------
>  1 file changed, 9 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/i386/kernel/signal.c b/arch/i386/kernel/signal.c
> index 4f99e87..26c2497 100644
> --- a/arch/i386/kernel/signal.c
> +++ b/arch/i386/kernel/signal.c
> @@ -350,7 +350,7 @@ static int setup_frame(int sig, struct k_sigaction *ka,
>  			goto give_sigsegv;
>  	}
>  
> -	if (current->binfmt->hasvdso)
> +	if (current->binfmt->hasvdso && current->mm->context.vdso)

I think this is correct, but a bit strange.

The "->context.vdso != NULL" check relies on the fact that .vdso == NULL
after mm_alloc (because arch_setup_additional_pages() doesn' initialize
it when vdso_enabled == 0, and it has to be != NULL otherwise).

This means that binfmt->hasvdso in essence is not used, at least for i386.
Isn't it better to kill ->hasvdso and just use ->context.vdso ? Every usage
of ->hasvdso should also check ->context.vdso anyway.

Oleg.