2008-01-15 22:22:04

by Frank Rowand

[permalink] [raw]
Subject: [PATCH 3/4] RT: remove finish_arch_switch

From: Frank Rowand <[email protected]>

This is probably just a temporary workaround for one procssor - the MIPS
community will most likely want to architect a solution to this issue.

Make of preempt kernel barfs in kernel/sched.c ifdef finish_arch_switch.
Remove the finish_arch_switch() for boards with TX49xx MIPS processor.

Signed-off-by: Frank Rowand <[email protected]>
---
include/asm-mips/mach-tx49xx/cpu-feature-overrides.h | 7 7 + 0 - 0 !
include/asm-mips/system.h | 3 3 + 0 - 0 !
2 files changed, 10 insertions(+)

Index: linux-2.6.24-rc7/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
===================================================================
--- linux-2.6.24-rc7.orig/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
+++ linux-2.6.24-rc7/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
@@ -1,6 +1,13 @@
#ifndef __ASM_MACH_TX49XX_CPU_FEATURE_OVERRIDES_H
#define __ASM_MACH_TX49XX_CPU_FEATURE_OVERRIDES_H

+/* finish_arch_switch_empty is defined if we know finish_arch_switch() will
+ * be empty, based on the lack of features defined in this file. This is
+ * needed because config preempt will barf in kernel/sched.c ifdef
+ * finish_arch_switch
+ */
+#define finish_arch_switch_empty
+
#define cpu_has_llsc 1
#define cpu_has_64bits 1
#define cpu_has_inclusive_pcaches 0
Index: linux-2.6.24-rc7/include/asm-mips/system.h
===================================================================
--- linux-2.6.24-rc7.orig/include/asm-mips/system.h
+++ linux-2.6.24-rc7/include/asm-mips/system.h
@@ -70,6 +70,8 @@ do { \
(last) = resume(prev, next, task_thread_info(next)); \
} while (0)

+/* preempt kernel barfs in kernel/sched.c ifdef finish_arch_switch */
+#ifndef finish_arch_switch_empty
#define finish_arch_switch(prev) \
do { \
if (cpu_has_dsp) \
@@ -77,6 +79,7 @@ do { \
if (cpu_has_userlocal) \
write_c0_userlocal(current_thread_info()->tp_value); \
} while (0)
+#endif

static inline unsigned long __xchg_u32(volatile int * m, unsigned int val)
{



2008-01-16 02:15:08

by Steven Rostedt

[permalink] [raw]
Subject: Re: [PATCH 3/4] RT: remove finish_arch_switch

On Tue, Jan 15, 2008 at 02:20:46PM -0800, Frank Rowand wrote:
> From: Frank Rowand <[email protected]>
>
>
> Index: linux-2.6.24-rc7/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
> ===================================================================
> --- linux-2.6.24-rc7.orig/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
> +++ linux-2.6.24-rc7/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
> @@ -1,6 +1,13 @@
> #ifndef __ASM_MACH_TX49XX_CPU_FEATURE_OVERRIDES_H
> #define __ASM_MACH_TX49XX_CPU_FEATURE_OVERRIDES_H
>
> +/* finish_arch_switch_empty is defined if we know finish_arch_switch() will
> + * be empty, based on the lack of features defined in this file. This is
> + * needed because config preempt will barf in kernel/sched.c ifdef
> + * finish_arch_switch
> + */
> +#define finish_arch_switch_empty
> +
> #define cpu_has_llsc 1
> #define cpu_has_64bits 1
> #define cpu_has_inclusive_pcaches 0
> Index: linux-2.6.24-rc7/include/asm-mips/system.h
> ===================================================================
> --- linux-2.6.24-rc7.orig/include/asm-mips/system.h
> +++ linux-2.6.24-rc7/include/asm-mips/system.h
> @@ -70,6 +70,8 @@ do { \
> (last) = resume(prev, next, task_thread_info(next)); \
> } while (0)
>
> +/* preempt kernel barfs in kernel/sched.c ifdef finish_arch_switch */
> +#ifndef finish_arch_switch_empty

I'll take this patch for now, but currently it looks like a hack. I know
you said that, but I'm hoping someone will come up with a better
solution.

Thanks,

-- Steve

> #define finish_arch_switch(prev) \
> do { \
> if (cpu_has_dsp) \
> @@ -77,6 +79,7 @@ do { \
> if (cpu_has_userlocal) \
> write_c0_userlocal(current_thread_info()->tp_value); \
> } while (0)
> +#endif
>
> static inline unsigned long __xchg_u32(volatile int * m, unsigned int val)
> {
>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/

2008-01-16 02:19:36

by Rowand, Frank

[permalink] [raw]
Subject: RE: [PATCH 3/4] RT: remove finish_arch_switch

Steve,

Thanks, I'll bring this up over on the linux-mips list to see how they
really want to handle it.

-Frank


-----Original Message-----
From: Steven Rostedt [mailto:[email protected]]
Sent: Tue 1/15/2008 6:14 PM
To: Rowand, Frank
Cc: [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH 3/4] RT: remove finish_arch_switch

On Tue, Jan 15, 2008 at 02:20:46PM -0800, Frank Rowand wrote:
> From: Frank Rowand <[email protected]>
>
>
> Index: linux-2.6.24-rc7/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
> ===================================================================
> --- linux-2.6.24-rc7.orig/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
> +++ linux-2.6.24-rc7/include/asm-mips/mach-tx49xx/cpu-feature-overrides.h
> @@ -1,6 +1,13 @@
> #ifndef __ASM_MACH_TX49XX_CPU_FEATURE_OVERRIDES_H
> #define __ASM_MACH_TX49XX_CPU_FEATURE_OVERRIDES_H
>
> +/* finish_arch_switch_empty is defined if we know finish_arch_switch() will
> + * be empty, based on the lack of features defined in this file. This is
> + * needed because config preempt will barf in kernel/sched.c ifdef
> + * finish_arch_switch
> + */
> +#define finish_arch_switch_empty
> +
> #define cpu_has_llsc 1
> #define cpu_has_64bits 1
> #define cpu_has_inclusive_pcaches 0
> Index: linux-2.6.24-rc7/include/asm-mips/system.h
> ===================================================================
> --- linux-2.6.24-rc7.orig/include/asm-mips/system.h
> +++ linux-2.6.24-rc7/include/asm-mips/system.h
> @@ -70,6 +70,8 @@ do { \
> (last) = resume(prev, next, task_thread_info(next)); \
> } while (0)
>
> +/* preempt kernel barfs in kernel/sched.c ifdef finish_arch_switch */
> +#ifndef finish_arch_switch_empty

I'll take this patch for now, but currently it looks like a hack. I know
you said that, but I'm hoping someone will come up with a better
solution.

Thanks,

-- Steve

> #define finish_arch_switch(prev) \
> do { \
> if (cpu_has_dsp) \
> @@ -77,6 +79,7 @@ do { \
> if (cpu_has_userlocal) \
> write_c0_userlocal(current_thread_info()->tp_value); \
> } while (0)
> +#endif
>
> static inline unsigned long __xchg_u32(volatile int * m, unsigned int val)
> {
>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/