| Message ID | 1399990378-7023-1-git-send-email-k.kozlowski@samsung.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On wto, 2014-05-13 at 16:12 +0200, Krzysztof Kozlowski wrote: > On Exynos4212 USE_DELAYED_RESET_ASSERTION must be set in > ARM_CORE1_OPTION register during CPU power down. This is the proper way > of powering down CPU. Additionally without this the CPU clock down won't > work after powering down CPU1 and CPU will work at full frequency chosen > by CPUfreq governor. > > Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> I am humbly reminding myself - any comments on this patch? Best regards, Krzysztof > --- > arch/arm/mach-exynos/hotplug.c | 28 +++++++++++++++++++++++++--- > arch/arm/mach-exynos/regs-pmu.h | 2 ++ > 2 files changed, 27 insertions(+), 3 deletions(-) > > diff --git a/arch/arm/mach-exynos/hotplug.c b/arch/arm/mach-exynos/hotplug.c > index 5eead530c6f8..59b813e74558 100644 > --- a/arch/arm/mach-exynos/hotplug.c > +++ b/arch/arm/mach-exynos/hotplug.c > @@ -74,7 +74,7 @@ static inline void cpu_enter_lowpower_a15(void) > dsb(); > } > > -static inline void cpu_leave_lowpower(void) > +static inline void cpu_leave_lowpower(unsigned int cpu) > { > unsigned int v; > > @@ -88,6 +88,14 @@ static inline void cpu_leave_lowpower(void) > : "=&r" (v) > : "Ir" (CR_C), "Ir" (0x40) > : "cc"); > + > + if (cpu == 1 && soc_is_exynos4212()) { > + unsigned int tmp; > + > + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); > + tmp &= ~(S5P_USE_DELAYED_RESET_ASSERTION); > + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); > + } > } > > static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > @@ -95,8 +103,22 @@ static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > for (;;) { > > /* make cpu1 to be turned off at next WFI command */ > - if (cpu == 1) > + if (cpu == 1) { > + if (soc_is_exynos4212()) { > + unsigned int tmp; > + > + /* > + * Exynos 4212 requires setting > + * USE_DELAYED_RESET_ASSERTION so the CPU idle > + * clock down feature could properly detect > + * global idle state when CPU1 is off. > + */ > + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); > + tmp |= S5P_USE_DELAYED_RESET_ASSERTION; > + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); > + } > __raw_writel(0, S5P_ARM_CORE1_CONFIGURATION); > + } > > /* > * here's the WFI > @@ -152,7 +174,7 @@ void __ref exynos_cpu_die(unsigned int cpu) > * bring this CPU back into the world of cache > * coherency, and then restore interrupts > */ > - cpu_leave_lowpower(); > + cpu_leave_lowpower(cpu); > > if (spurious) > pr_warn("CPU%u: %u spurious wakeup calls\n", cpu, spurious); > diff --git a/arch/arm/mach-exynos/regs-pmu.h b/arch/arm/mach-exynos/regs-pmu.h > index 4f6a2560d022..1a3da4ef0e22 100644 > --- a/arch/arm/mach-exynos/regs-pmu.h > +++ b/arch/arm/mach-exynos/regs-pmu.h > @@ -25,6 +25,7 @@ > > #define S5P_USE_STANDBY_WFI0 (1 << 16) > #define S5P_USE_STANDBY_WFE0 (1 << 24) > +#define S5P_USE_DELAYED_RESET_ASSERTION BIT(12) > > #define EXYNOS_SWRESET S5P_PMUREG(0x0400) > #define EXYNOS5440_SWRESET S5P_PMUREG(0x00C4) > @@ -107,6 +108,7 @@ > > #define S5P_ARM_CORE1_CONFIGURATION S5P_PMUREG(0x2080) > #define S5P_ARM_CORE1_STATUS S5P_PMUREG(0x2084) > +#define S5P_ARM_CORE1_OPTION S5P_PMUREG(0x2088) > > #define S5P_PAD_RET_MAUDIO_OPTION S5P_PMUREG(0x3028) > #define S5P_PAD_RET_GPIO_OPTION S5P_PMUREG(0x3108)
Hi Krzysztof, On 13.05.2014 16:12, Krzysztof Kozlowski wrote: > On Exynos4212 USE_DELAYED_RESET_ASSERTION must be set in > ARM_CORE1_OPTION register during CPU power down. This is the proper way > of powering down CPU. Additionally without this the CPU clock down won't > work after powering down CPU1 and CPU will work at full frequency chosen > by CPUfreq governor. > > Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> > --- > arch/arm/mach-exynos/hotplug.c | 28 +++++++++++++++++++++++++--- > arch/arm/mach-exynos/regs-pmu.h | 2 ++ > 2 files changed, 27 insertions(+), 3 deletions(-) > > diff --git a/arch/arm/mach-exynos/hotplug.c b/arch/arm/mach-exynos/hotplug.c > index 5eead530c6f8..59b813e74558 100644 > --- a/arch/arm/mach-exynos/hotplug.c > +++ b/arch/arm/mach-exynos/hotplug.c > @@ -74,7 +74,7 @@ static inline void cpu_enter_lowpower_a15(void) > dsb(); > } > > -static inline void cpu_leave_lowpower(void) > +static inline void cpu_leave_lowpower(unsigned int cpu) > { > unsigned int v; > > @@ -88,6 +88,14 @@ static inline void cpu_leave_lowpower(void) > : "=&r" (v) > : "Ir" (CR_C), "Ir" (0x40) > : "cc"); > + > + if (cpu == 1 && soc_is_exynos4212()) { AFAIK this is needed not only on Exynos4x12, but at least also on 4412. Also I don't think the (cpu == 1) condition is necessary. See below. > + unsigned int tmp; > + > + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); > + tmp &= ~(S5P_USE_DELAYED_RESET_ASSERTION); > + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); > + } > } > > static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > @@ -95,8 +103,22 @@ static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > for (;;) { > > /* make cpu1 to be turned off at next WFI command */ > - if (cpu == 1) > + if (cpu == 1) { I think this patch should be rebased on top of my "[PATCH] ARM: EXYNOS: Fix core ID used by platsmp and hotplug code". Then instead of if (cpu == 1) you would just use EXYNOS_ARM_CORE_OPTION(core_id). Also same comment as above, I believe this is also needed on Exynos4412. Best regards, Tomasz
On pi?, 2014-05-30 at 13:50 +0200, Tomasz Figa wrote: > Hi Krzysztof, > > On 13.05.2014 16:12, Krzysztof Kozlowski wrote: > > On Exynos4212 USE_DELAYED_RESET_ASSERTION must be set in > > ARM_CORE1_OPTION register during CPU power down. This is the proper way > > of powering down CPU. Additionally without this the CPU clock down won't > > work after powering down CPU1 and CPU will work at full frequency chosen > > by CPUfreq governor. > > > > Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> > > --- > > arch/arm/mach-exynos/hotplug.c | 28 +++++++++++++++++++++++++--- > > arch/arm/mach-exynos/regs-pmu.h | 2 ++ > > 2 files changed, 27 insertions(+), 3 deletions(-) > > > > diff --git a/arch/arm/mach-exynos/hotplug.c b/arch/arm/mach-exynos/hotplug.c > > index 5eead530c6f8..59b813e74558 100644 > > --- a/arch/arm/mach-exynos/hotplug.c > > +++ b/arch/arm/mach-exynos/hotplug.c > > @@ -74,7 +74,7 @@ static inline void cpu_enter_lowpower_a15(void) > > dsb(); > > } > > > > -static inline void cpu_leave_lowpower(void) > > +static inline void cpu_leave_lowpower(unsigned int cpu) > > { > > unsigned int v; > > > > @@ -88,6 +88,14 @@ static inline void cpu_leave_lowpower(void) > > : "=&r" (v) > > : "Ir" (CR_C), "Ir" (0x40) > > : "cc"); > > + > > + if (cpu == 1 && soc_is_exynos4212()) { > > AFAIK this is needed not only on Exynos4x12, but at least also on 4412. > Also I don't think the (cpu == 1) condition is necessary. See below. I'll test it on Exynos4412. > > > + unsigned int tmp; > > + > > + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); > > + tmp &= ~(S5P_USE_DELAYED_RESET_ASSERTION); > > + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); > > + } > > } > > > > static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > > @@ -95,8 +103,22 @@ static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > > for (;;) { > > > > /* make cpu1 to be turned off at next WFI command */ > > - if (cpu == 1) > > + if (cpu == 1) { > > I think this patch should be rebased on top of my "[PATCH] ARM: EXYNOS: > Fix core ID used by platsmp and hotplug code". Then instead of if (cpu > == 1) you would just use EXYNOS_ARM_CORE_OPTION(core_id). > > Also same comment as above, I believe this is also needed on Exynos4412. Probably your patch needs rebasing too. I couldn't apply (cleanly) on current Linus' tree and on Kukjin's for-next. What are the requirements for it? Best regards, Krzysztof
On 30.05.2014 14:46, Krzysztof Kozlowski wrote: > On pi?, 2014-05-30 at 13:50 +0200, Tomasz Figa wrote: >> Hi Krzysztof, >> >> On 13.05.2014 16:12, Krzysztof Kozlowski wrote: >>> On Exynos4212 USE_DELAYED_RESET_ASSERTION must be set in >>> ARM_CORE1_OPTION register during CPU power down. This is the proper way >>> of powering down CPU. Additionally without this the CPU clock down won't >>> work after powering down CPU1 and CPU will work at full frequency chosen >>> by CPUfreq governor. >>> >>> Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> >>> --- >>> arch/arm/mach-exynos/hotplug.c | 28 +++++++++++++++++++++++++--- >>> arch/arm/mach-exynos/regs-pmu.h | 2 ++ >>> 2 files changed, 27 insertions(+), 3 deletions(-) >>> >>> diff --git a/arch/arm/mach-exynos/hotplug.c b/arch/arm/mach-exynos/hotplug.c >>> index 5eead530c6f8..59b813e74558 100644 >>> --- a/arch/arm/mach-exynos/hotplug.c >>> +++ b/arch/arm/mach-exynos/hotplug.c >>> @@ -74,7 +74,7 @@ static inline void cpu_enter_lowpower_a15(void) >>> dsb(); >>> } >>> >>> -static inline void cpu_leave_lowpower(void) >>> +static inline void cpu_leave_lowpower(unsigned int cpu) >>> { >>> unsigned int v; >>> >>> @@ -88,6 +88,14 @@ static inline void cpu_leave_lowpower(void) >>> : "=&r" (v) >>> : "Ir" (CR_C), "Ir" (0x40) >>> : "cc"); >>> + >>> + if (cpu == 1 && soc_is_exynos4212()) { >> >> AFAIK this is needed not only on Exynos4x12, but at least also on 4412. >> Also I don't think the (cpu == 1) condition is necessary. See below. > > I'll test it on Exynos4412. > >> >>> + unsigned int tmp; >>> + >>> + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); >>> + tmp &= ~(S5P_USE_DELAYED_RESET_ASSERTION); >>> + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); >>> + } >>> } >>> >>> static inline void platform_do_lowpower(unsigned int cpu, int *spurious) >>> @@ -95,8 +103,22 @@ static inline void platform_do_lowpower(unsigned int cpu, int *spurious) >>> for (;;) { >>> >>> /* make cpu1 to be turned off at next WFI command */ >>> - if (cpu == 1) >>> + if (cpu == 1) { >> >> I think this patch should be rebased on top of my "[PATCH] ARM: EXYNOS: >> Fix core ID used by platsmp and hotplug code". Then instead of if (cpu >> == 1) you would just use EXYNOS_ARM_CORE_OPTION(core_id). >> >> Also same comment as above, I believe this is also needed on Exynos4412. > > Probably your patch needs rebasing too. I couldn't apply (cleanly) on > current Linus' tree and on Kukjin's for-next. What are the requirements > for it? Hmm, I based it on Kukjin's for-next from the day I sent the patch, 23.05 if I remember correctly. AFAICT everything needed by this patch is already merged in Kukjin's tree (and I believe already was when I sent that patch). To make sure, I'm talking about this patch: http://www.spinics.net/lists/linux-samsung-soc/msg31604.html Best regards, Tomasz
On pi?, 2014-05-30 at 14:56 +0200, Tomasz Figa wrote: > On 30.05.2014 14:46, Krzysztof Kozlowski wrote: > > On pi?, 2014-05-30 at 13:50 +0200, Tomasz Figa wrote: > >> Hi Krzysztof, > >> > >> On 13.05.2014 16:12, Krzysztof Kozlowski wrote: > >>> On Exynos4212 USE_DELAYED_RESET_ASSERTION must be set in > >>> ARM_CORE1_OPTION register during CPU power down. This is the proper way > >>> of powering down CPU. Additionally without this the CPU clock down won't > >>> work after powering down CPU1 and CPU will work at full frequency chosen > >>> by CPUfreq governor. > >>> > >>> Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> > >>> --- > >>> arch/arm/mach-exynos/hotplug.c | 28 +++++++++++++++++++++++++--- > >>> arch/arm/mach-exynos/regs-pmu.h | 2 ++ > >>> 2 files changed, 27 insertions(+), 3 deletions(-) > >>> > >>> diff --git a/arch/arm/mach-exynos/hotplug.c b/arch/arm/mach-exynos/hotplug.c > >>> index 5eead530c6f8..59b813e74558 100644 > >>> --- a/arch/arm/mach-exynos/hotplug.c > >>> +++ b/arch/arm/mach-exynos/hotplug.c > >>> @@ -74,7 +74,7 @@ static inline void cpu_enter_lowpower_a15(void) > >>> dsb(); > >>> } > >>> > >>> -static inline void cpu_leave_lowpower(void) > >>> +static inline void cpu_leave_lowpower(unsigned int cpu) > >>> { > >>> unsigned int v; > >>> > >>> @@ -88,6 +88,14 @@ static inline void cpu_leave_lowpower(void) > >>> : "=&r" (v) > >>> : "Ir" (CR_C), "Ir" (0x40) > >>> : "cc"); > >>> + > >>> + if (cpu == 1 && soc_is_exynos4212()) { > >> > >> AFAIK this is needed not only on Exynos4x12, but at least also on 4412. > >> Also I don't think the (cpu == 1) condition is necessary. See below. > > > > I'll test it on Exynos4412. > > > >> > >>> + unsigned int tmp; > >>> + > >>> + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); > >>> + tmp &= ~(S5P_USE_DELAYED_RESET_ASSERTION); > >>> + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); > >>> + } > >>> } > >>> > >>> static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > >>> @@ -95,8 +103,22 @@ static inline void platform_do_lowpower(unsigned int cpu, int *spurious) > >>> for (;;) { > >>> > >>> /* make cpu1 to be turned off at next WFI command */ > >>> - if (cpu == 1) > >>> + if (cpu == 1) { > >> > >> I think this patch should be rebased on top of my "[PATCH] ARM: EXYNOS: > >> Fix core ID used by platsmp and hotplug code". Then instead of if (cpu > >> == 1) you would just use EXYNOS_ARM_CORE_OPTION(core_id). > >> > >> Also same comment as above, I believe this is also needed on Exynos4412. > > > > Probably your patch needs rebasing too. I couldn't apply (cleanly) on > > current Linus' tree and on Kukjin's for-next. What are the requirements > > for it? > > Hmm, I based it on Kukjin's for-next from the day I sent the patch, > 23.05 if I remember correctly. AFAICT everything needed by this patch is > already merged in Kukjin's tree (and I believe already was when I sent > that patch). > > To make sure, I'm talking about this patch: > http://www.spinics.net/lists/linux-samsung-soc/msg31604.html Yep, that's the patch but still it does not apply cleanly :). Anyway I'll send next version of my patch reflecting your comments: 1. Limit to all Exynos4 SoCs. 2. Rebase on your patch to utilize EXYNOS_ARM_CORE_OPTION(core_id). Best regards, Krzysztof
diff --git a/arch/arm/mach-exynos/hotplug.c b/arch/arm/mach-exynos/hotplug.c index 5eead530c6f8..59b813e74558 100644 --- a/arch/arm/mach-exynos/hotplug.c +++ b/arch/arm/mach-exynos/hotplug.c @@ -74,7 +74,7 @@ static inline void cpu_enter_lowpower_a15(void) dsb(); } -static inline void cpu_leave_lowpower(void) +static inline void cpu_leave_lowpower(unsigned int cpu) { unsigned int v; @@ -88,6 +88,14 @@ static inline void cpu_leave_lowpower(void) : "=&r" (v) : "Ir" (CR_C), "Ir" (0x40) : "cc"); + + if (cpu == 1 && soc_is_exynos4212()) { + unsigned int tmp; + + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); + tmp &= ~(S5P_USE_DELAYED_RESET_ASSERTION); + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); + } } static inline void platform_do_lowpower(unsigned int cpu, int *spurious) @@ -95,8 +103,22 @@ static inline void platform_do_lowpower(unsigned int cpu, int *spurious) for (;;) { /* make cpu1 to be turned off at next WFI command */ - if (cpu == 1) + if (cpu == 1) { + if (soc_is_exynos4212()) { + unsigned int tmp; + + /* + * Exynos 4212 requires setting + * USE_DELAYED_RESET_ASSERTION so the CPU idle + * clock down feature could properly detect + * global idle state when CPU1 is off. + */ + tmp = __raw_readl(S5P_ARM_CORE1_OPTION); + tmp |= S5P_USE_DELAYED_RESET_ASSERTION; + __raw_writel(tmp, S5P_ARM_CORE1_OPTION); + } __raw_writel(0, S5P_ARM_CORE1_CONFIGURATION); + } /* * here's the WFI @@ -152,7 +174,7 @@ void __ref exynos_cpu_die(unsigned int cpu) * bring this CPU back into the world of cache * coherency, and then restore interrupts */ - cpu_leave_lowpower(); + cpu_leave_lowpower(cpu); if (spurious) pr_warn("CPU%u: %u spurious wakeup calls\n", cpu, spurious); diff --git a/arch/arm/mach-exynos/regs-pmu.h b/arch/arm/mach-exynos/regs-pmu.h index 4f6a2560d022..1a3da4ef0e22 100644 --- a/arch/arm/mach-exynos/regs-pmu.h +++ b/arch/arm/mach-exynos/regs-pmu.h @@ -25,6 +25,7 @@ #define S5P_USE_STANDBY_WFI0 (1 << 16) #define S5P_USE_STANDBY_WFE0 (1 << 24) +#define S5P_USE_DELAYED_RESET_ASSERTION BIT(12) #define EXYNOS_SWRESET S5P_PMUREG(0x0400) #define EXYNOS5440_SWRESET S5P_PMUREG(0x00C4) @@ -107,6 +108,7 @@ #define S5P_ARM_CORE1_CONFIGURATION S5P_PMUREG(0x2080) #define S5P_ARM_CORE1_STATUS S5P_PMUREG(0x2084) +#define S5P_ARM_CORE1_OPTION S5P_PMUREG(0x2088) #define S5P_PAD_RET_MAUDIO_OPTION S5P_PMUREG(0x3028) #define S5P_PAD_RET_GPIO_OPTION S5P_PMUREG(0x3108)
On Exynos4212 USE_DELAYED_RESET_ASSERTION must be set in ARM_CORE1_OPTION register during CPU power down. This is the proper way of powering down CPU. Additionally without this the CPU clock down won't work after powering down CPU1 and CPU will work at full frequency chosen by CPUfreq governor. Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com> --- arch/arm/mach-exynos/hotplug.c | 28 +++++++++++++++++++++++++--- arch/arm/mach-exynos/regs-pmu.h | 2 ++ 2 files changed, 27 insertions(+), 3 deletions(-)