Message ID | 1542262461-29024-4-git-send-email-firoz.khan@linaro.org |
---|---|
State | Superseded |
Headers | show |
Series | mips: system call table generation support | expand |
Hi Firoz, On Thu, Nov 15, 2018 at 11:44:19AM +0530, Firoz Khan wrote: > diff --git a/arch/mips/kernel/scall32-o32.S b/arch/mips/kernel/scall32-o32.S > index a9b895f..4eee437 100644 > --- a/arch/mips/kernel/scall32-o32.S > +++ b/arch/mips/kernel/scall32-o32.S > @@ -208,6 +208,18 @@ einval: li v0, -ENOSYS > jr ra > END(sys_syscall) > > +#ifdef CONFIG_MIPS_MT_FPAFF > + /* > + * For FPU affinity scheduling on MIPS MT processors, we need to > + * intercept sys_sched_xxxaffinity() calls until we get a proper hook > + * in kernel/sched/core.c. Considered only temporary we only support > + * these hooks for the 32-bit kernel - there is no MIPS64 MT processor > + * atm. > + */ > +#define mipsmt_sys_sched_setaffinity sys_sched_setaffinity > +#define mipsmt_sys_sched_getaffinity sys_sched_getaffinity Is this backwards? ie. should it be: #define sys_sched_setaffinity mipsmt_sys_sched_setaffinity #define sys_sched_getaffinity mipsmt_sys_sched_getaffinity ? I don't see how the mipsmt_* functions will ever be used after this patch. Thanks, Paul > +#endif /* CONFIG_MIPS_MT_FPAFF */ > + > .align 2 > .type sys_call_table, @object > EXPORT(sys_call_table) > @@ -450,20 +462,8 @@ EXPORT(sys_call_table) > PTR sys_tkill > PTR sys_sendfile64 > PTR sys_futex > -#ifdef CONFIG_MIPS_MT_FPAFF > - /* > - * For FPU affinity scheduling on MIPS MT processors, we need to > - * intercept sys_sched_xxxaffinity() calls until we get a proper hook > - * in kernel/sched/core.c. Considered only temporary we only support > - * these hooks for the 32-bit kernel - there is no MIPS64 MT processor > - * atm. > - */ > - PTR mipsmt_sys_sched_setaffinity > - PTR mipsmt_sys_sched_getaffinity > -#else > PTR sys_sched_setaffinity > PTR sys_sched_getaffinity /* 4240 */ > -#endif /* CONFIG_MIPS_MT_FPAFF */ > PTR sys_io_setup > PTR sys_io_destroy > PTR sys_io_getevents > -- > 1.9.1 >
On Mon, Nov 19, 2018 at 11:29 PM Paul Burton <paul.burton@mips.com> wrote: > > Hi Firoz, > > On Thu, Nov 15, 2018 at 11:44:19AM +0530, Firoz Khan wrote: > > diff --git a/arch/mips/kernel/scall32-o32.S b/arch/mips/kernel/scall32-o32.S > > index a9b895f..4eee437 100644 > > --- a/arch/mips/kernel/scall32-o32.S > > +++ b/arch/mips/kernel/scall32-o32.S > > @@ -208,6 +208,18 @@ einval: li v0, -ENOSYS > > jr ra > > END(sys_syscall) > > > > +#ifdef CONFIG_MIPS_MT_FPAFF > > + /* > > + * For FPU affinity scheduling on MIPS MT processors, we need to > > + * intercept sys_sched_xxxaffinity() calls until we get a proper hook > > + * in kernel/sched/core.c. Considered only temporary we only support > > + * these hooks for the 32-bit kernel - there is no MIPS64 MT processor > > + * atm. > > + */ > > +#define mipsmt_sys_sched_setaffinity sys_sched_setaffinity > > +#define mipsmt_sys_sched_getaffinity sys_sched_getaffinity > > Is this backwards? ie. should it be: > > #define sys_sched_setaffinity mipsmt_sys_sched_setaffinity > #define sys_sched_getaffinity mipsmt_sys_sched_getaffinity > > ? > > I don't see how the mipsmt_* functions will ever be used after this > patch. Good catch! > > - * atm. > > - */ > > - PTR mipsmt_sys_sched_setaffinity > > - PTR mipsmt_sys_sched_getaffinity > > -#else > > PTR sys_sched_setaffinity > > PTR sys_sched_getaffinity /* 4240 */ > > -#endif /* CONFIG_MIPS_MT_FPAFF */ > > PTR sys_io_setup My guess would be that he removed the wrong lines here instead, and the first half was intentional. Arnd
diff --git a/arch/mips/kernel/scall32-o32.S b/arch/mips/kernel/scall32-o32.S index a9b895f..4eee437 100644 --- a/arch/mips/kernel/scall32-o32.S +++ b/arch/mips/kernel/scall32-o32.S @@ -208,6 +208,18 @@ einval: li v0, -ENOSYS jr ra END(sys_syscall) +#ifdef CONFIG_MIPS_MT_FPAFF + /* + * For FPU affinity scheduling on MIPS MT processors, we need to + * intercept sys_sched_xxxaffinity() calls until we get a proper hook + * in kernel/sched/core.c. Considered only temporary we only support + * these hooks for the 32-bit kernel - there is no MIPS64 MT processor + * atm. + */ +#define mipsmt_sys_sched_setaffinity sys_sched_setaffinity +#define mipsmt_sys_sched_getaffinity sys_sched_getaffinity +#endif /* CONFIG_MIPS_MT_FPAFF */ + .align 2 .type sys_call_table, @object EXPORT(sys_call_table) @@ -450,20 +462,8 @@ EXPORT(sys_call_table) PTR sys_tkill PTR sys_sendfile64 PTR sys_futex -#ifdef CONFIG_MIPS_MT_FPAFF - /* - * For FPU affinity scheduling on MIPS MT processors, we need to - * intercept sys_sched_xxxaffinity() calls until we get a proper hook - * in kernel/sched/core.c. Considered only temporary we only support - * these hooks for the 32-bit kernel - there is no MIPS64 MT processor - * atm. - */ - PTR mipsmt_sys_sched_setaffinity - PTR mipsmt_sys_sched_getaffinity -#else PTR sys_sched_setaffinity PTR sys_sched_getaffinity /* 4240 */ -#endif /* CONFIG_MIPS_MT_FPAFF */ PTR sys_io_setup PTR sys_io_destroy PTR sys_io_getevents
The config flag - CONFIG_MIPS_MT_FPAFF uses to check whether which syscall entries need to be used in scall32-o32.S file. One of the patch in this patch series will generate syscall table file. But CONFIG_MIPS_MT_FPAFF flag will add more complexity in the script to generate the syscall table file. In order to come up with a common implementation across all archit- ecture, we need to remove mipsmt_sys_sched_setaffinity and mipsmt- _sys_sched_getaffinity from the table and define it in other way. Signed-off-by: Firoz Khan <firoz.khan@linaro.org> --- arch/mips/kernel/scall32-o32.S | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) -- 1.9.1