2008-12-23 16:22:27

by Jaswinder Singh Rajput

Subject: [PATCH] x86: traps.c replace #if CONFIG_X86_32 with #ifdef CONFIG_X86_32

Impact: cleanup, avoid warning in X86_64

Fixes this warning in X86_64:
CC arch/x86/kernel/traps.o
arch/x86/kernel/traps.c:695:5: warning: "CONFIG_X86_32" is not defined

Signed-off-by: Jaswinder Singh <[email protected]>
---
arch/x86/kernel/traps.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index d327b53..2c00412 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -692,7 +692,7 @@ void math_error(void __user *ip)

err = swd & ~cwd & 0x3f;

-#if CONFIG_X86_32
+#ifdef CONFIG_X86_32
if (!err)
return;
#endif
--
1.5.5.1

2008-12-23 18:24:23

by H. Peter Anvin

Subject: Re: [PATCH] x86: traps.c replace #if CONFIG_X86_32 with #ifdef CONFIG_X86_32

Jaswinder Singh wrote:
> Impact: cleanup, avoid warning in X86_64
>
> Fixes this warning in X86_64:
> CC arch/x86/kernel/traps.o
> arch/x86/kernel/traps.c:695:5: warning: "CONFIG_X86_32" is not defined
>
> Signed-off-by: Jaswinder Singh <[email protected]>
> ---
> arch/x86/kernel/traps.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
> index d327b53..2c00412 100644
> --- a/arch/x86/kernel/traps.c
> +++ b/arch/x86/kernel/traps.c
> @@ -692,7 +692,7 @@ void math_error(void __user *ip)
>
> err = swd & ~cwd & 0x3f;
>
> -#if CONFIG_X86_32
> +#ifdef CONFIG_X86_32
> if (!err)
> return;
> #endif

Applied to tip:x86/fpu, thanks.

Andi: you might have a perspective on this: in math_error() we bail out
if there is no error on i386, but not on x86-64. I am rather assuming
that is because it simply can't happen on x86-64 (no IRQ 13 bullshit) so
it wasn't necessary?

Thanks,

-hpa

2008-12-23 18:35:52

by Andi Kleen

Subject: Re: [PATCH] x86: traps.c replace #if CONFIG_X86_32 with #ifdef CONFIG_X86_32

On Tue, Dec 23, 2008 at 10:20:37AM -0800, H. Peter Anvin wrote:
> > return;
> > #endif
>
> Applied to tip:x86/fpu, thanks.
>
> Andi: you might have a perspective on this: in math_error() we bail out
> if there is no error on i386, but not on x86-64. I am rather assuming
> that is because it simply can't happen on x86-64 (no IRQ 13 bullshit) so
> it wasn't necessary?

Probably yes because it was not needed. Don't remember the exact circumstances,
sorry.

But dropping the #ifdef should be fine obviously.

-Andi
--
[email protected]

2008-12-23 20:34:27

by Chuck Ebbert

Subject: Re: [PATCH] x86: traps.c replace #if CONFIG_X86_32 with #ifdef CONFIG_X86_32

On Tue, 23 Dec 2008 10:20:37 -0800
"H. Peter Anvin" <[email protected]> wrote:

> > --- a/arch/x86/kernel/traps.c
> > +++ b/arch/x86/kernel/traps.c
> > @@ -692,7 +692,7 @@ void math_error(void __user *ip)
> >
> > err = swd & ~cwd & 0x3f;
> >
> > -#if CONFIG_X86_32
> > +#ifdef CONFIG_X86_32
> > if (!err)
> > return;
> > #endif
>
> Applied to tip:x86/fpu, thanks.
>
> Andi: you might have a perspective on this: in math_error() we bail out
> if there is no error on i386, but not on x86-64. I am rather assuming
> that is because it simply can't happen on x86-64 (no IRQ 13 bullshit) so
> it wasn't necessary?
>

Some 32-bit x86 clones generate math error interrupts when there is no
actual error.