[patch 1/1] x86-64: forgot asmlinkage on sys_mmap

[patch 1/1] x86-64: forgot asmlinkage on sys_mmap

[blaisorblade]

CC: Andi Kleen <ak@suse.de>

I think it should be there, please check better.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
---

 linux-2.6.11-paolo/arch/x86_64/kernel/sys_x86_64.c |    2 +-
 1 files changed, 1 insertion(+), 1 deletion(-)

diff -puN arch/x86_64/kernel/sys_x86_64.c~x86-64-asmlinkage-forgot arch/x86_64/kernel/sys_x86_64.c
--- linux-2.6.11/arch/x86_64/kernel/sys_x86_64.c~x86-64-asmlinkage-forgot	2005-03-05 19:58:38.339212568 +0100
+++ linux-2.6.11-paolo/arch/x86_64/kernel/sys_x86_64.c	2005-03-05 19:58:51.062278368 +0100
@@ -38,7 +38,7 @@ asmlinkage long sys_pipe(int __user *fil
 	return error;
 }
 
-long sys_mmap(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags,
+asmlinkage long sys_mmap(unsigned long addr, unsigned long len, unsigned long prot, unsigned long flags,
 	unsigned long fd, unsigned long off)
 {
 	long error;
_

Re: [patch 1/1] x86-64: forgot asmlinkage on sys_mmap

[Andi Kleen]

blaisorblade@yahoo.it writes:

> CC: Andi Kleen <ak@suse.de>
>
> I think it should be there, please check better.

It doesn't matter. asmlinkage is a nop on x86-64.

-Andi

Re: [patch 1/1] x86-64: forgot asmlinkage on sys_mmap

[Blaisorblade]

On Wednesday 09 March 2005 18:24, Andi Kleen wrote:
> blaisorblade@yahoo.it writes:
> > CC: Andi Kleen <ak@suse.de>
> >
> > I think it should be there, please check better.
>
> It doesn't matter. asmlinkage is a nop on x86-64.

Yes, otherwise nothing would work on x86-64 with mmap broken, but for 
cleanness and for the case this change it should be there (otherwise why 
asmlinkage is used in the rest of the file).

And for i386 asmlinkage acquired significance only recently.
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade

Re: [patch 1/1] x86-64: forgot asmlinkage on sys_mmap

[Andi Kleen]

On Wed, Mar 09, 2005 at 07:24:00PM +0100, Blaisorblade wrote:
> On Wednesday 09 March 2005 18:24, Andi Kleen wrote:
> > blaisorblade@yahoo.it writes:
> > > CC: Andi Kleen <ak@suse.de>
> > >
> > > I think it should be there, please check better.
> >
> > It doesn't matter. asmlinkage is a nop on x86-64.
> 
> Yes, otherwise nothing would work on x86-64 with mmap broken, but for 
> cleanness and for the case this change it should be there (otherwise why 
> asmlinkage is used in the rest of the file).

Only because it was cut'n'pasted from i386 originally.

> 
> And for i386 asmlinkage acquired significance only recently.

Actually it doesn't neither on i386. That's because entry.S happens to put the 
arguments both into registers and the stack in the right order, so both 
register and stack argument calling conventions work.

But it is slightly safer to have it. When you use the stack arguments
the C code is allowed to modify it, and when the system call is restarted
later you could see garbage. In practice that's not a big issue because
only very few system calls are restartable.

ptrace also could see corrupted state, but that's in general a non issue.

-Andi

Re: [patch 1/1] x86-64: forgot asmlinkage on sys_mmap

[Blaisorblade]

On Wednesday 09 March 2005 20:34, Andi Kleen wrote:
> On Wed, Mar 09, 2005 at 07:24:00PM +0100, Blaisorblade wrote:
> > On Wednesday 09 March 2005 18:24, Andi Kleen wrote:
> > > blaisorblade@yahoo.it writes:
> > > > CC: Andi Kleen <ak@suse.de>
> > > >
> > > > I think it should be there, please check better.
> > >
> > > It doesn't matter. asmlinkage is a nop on x86-64.
> >
> > Yes, otherwise nothing would work on x86-64 with mmap broken, but for
> > cleanness and for the case this change it should be there (otherwise why
> > asmlinkage is used in the rest of the file).
>
> Only because it was cut'n'pasted from i386 originally.
>
> > And for i386 asmlinkage acquired significance only recently.
>
> Actually it doesn't neither on i386. That's because entry.S happens to put
> the arguments both into registers and the stack in the right order, so both
> register and stack argument calling conventions work.
>
> But it is slightly safer to have it. When you use the stack arguments
> the C code is allowed to modify it, and when the system call is restarted
> later you could see garbage. In practice that's not a big issue because
> only very few system calls are restartable.
>
> ptrace also could see corrupted state, but that's in general a non issue.
Ok, thanks for the info, I hope it's applied anyway.
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade