[PATCH 2/2] ARM: Exynos: Hook up power domains to generic power domain infrastructure
Thomas Abraham
thomas.abraham at linaro.org
Wed Dec 28 16:25:20 EST 2011
Hi Sylwester,
On 28 December 2011 04:44, Sylwester Nawrocki <snjw23 at gmail.com> wrote:
> Hi Thomas,
>
> On 12/12/2011 04:46 PM, Thomas Abraham wrote:
>> The generic power domain infrastructure is used to control the power domains
>> available on Exynos4. For non-dt platforms, the power domains are statically
>> instantiated. For dt platforms, the power domain nodes found in the device
>> tree are instantiated.
>>
>> Cc: Kukjin Kim <kgene.kim at samsung.com>
>> Cc: Rob Herring <rob.herring at calxeda.com>
>> Cc: Grant Likely <grant.likely at secretlab.ca>
>> Signed-off-by: Thomas Abraham <thomas.abraham at linaro.org>
>> ---
>> This patch is mainly derived from Mark Brown's work on generic power domain
>> support for s3c64xx platforms. The existing exynos4 power domain implementation
>> is not removed in this patch. The devices are not yet registered with the power
>> domains for non-dt platforms.
>>
>> arch/arm/mach-exynos/Kconfig | 1 +
>> arch/arm/mach-exynos/pm.c | 179 ++++++++++++++++++++++++++++++++++++++++++
>> 2 files changed, 180 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/arm/mach-exynos/Kconfig b/arch/arm/mach-exynos/Kconfig
>> index 0afcc3b..c1f77c7 100644
>> --- a/arch/arm/mach-exynos/Kconfig
>> +++ b/arch/arm/mach-exynos/Kconfig
>> @@ -29,6 +29,7 @@ config CPU_EXYNOS4210
>> default y
>> depends on ARCH_EXYNOS4
>> select SAMSUNG_DMADEV
>> + select PM_GENERIC_DOMAINS
>> select ARM_CPU_SUSPEND if PM
>> select S5P_PM if PM
>> select S5P_SLEEP if PM
>> diff --git a/arch/arm/mach-exynos/pm.c b/arch/arm/mach-exynos/pm.c
>> index 4093fea..a471ea6 100644
>> --- a/arch/arm/mach-exynos/pm.c
>> +++ b/arch/arm/mach-exynos/pm.c
>> @@ -20,6 +20,10 @@
>> #include <linux/io.h>
>> #include <linux/err.h>
>> #include <linux/clk.h>
>> +#include <linux/slab.h>
>> +#include <linux/pm_domain.h>
>> +#include <linux/delay.h>
>> +#include <linux/of_address.h>
>>
>> #include <asm/cacheflush.h>
>> #include <asm/hardware/cache-l2x0.h>
>> @@ -37,6 +41,13 @@
>> #include <mach/pm-core.h>
>> #include <mach/pmu.h>
>>
>> +struct exynos4_pm_domain {
>> + void __iomem *base;
>> + char const *name;
>> + bool is_off;
>> + struct generic_pm_domain pd;
>> +};
>> +
>> static struct sleep_save exynos4_set_clksrc[] = {
>> { .reg = S5P_CLKSRC_MASK_TOP , .val = 0x00000001, },
>> { .reg = S5P_CLKSRC_MASK_CAM , .val = 0x11111111, },
>> @@ -285,12 +296,172 @@ static struct sysdev_driver exynos4_pm_driver = {
>> .add = exynos4_pm_add,
>> };
>>
>> +static int exynos4_pd_power_on(struct generic_pm_domain *domain)
>> +{
>> + struct exynos4_pm_domain *pd;
>> + void __iomem *base;
>> + u32 timeout;
>> +
>> + pd = container_of(domain, struct exynos4_pm_domain, pd);
>> + base = pd->base;
>> +
>> + __raw_writel(S5P_INT_LOCAL_PWR_EN, base);
>> +
>> + /* Wait max 1ms */
>> + timeout = 10;
>> + while ((__raw_readl(base + 0x4) & S5P_INT_LOCAL_PWR_EN)
>> + != S5P_INT_LOCAL_PWR_EN) {
>> + if (!timeout) {
>> + pr_err("Power domain %s enable failed\n", domain->name);
>> + return -ETIMEDOUT;
>> + }
>> + timeout--;
>> + udelay(100);
>> + }
>> + return 0;
>> +}
>> +
>> +static int exynos4_pd_power_off(struct generic_pm_domain *domain)
>> +{
>> + struct exynos4_pm_domain *pd;
>> + void __iomem *base;
>> + u32 timeout;
>> +
>> + pd = container_of(domain, struct exynos4_pm_domain, pd);
>> + base = pd->base;
>> +
>> + __raw_writel(0, base);
>> +
>> + /* Wait max 1ms */
>> + timeout = 10;
>> + while (__raw_readl(base + 0x4) & S5P_INT_LOCAL_PWR_EN) {
>> + if (!timeout) {
>> + pr_err("Power domain %s disable failed\n", domain->name);
>> + return -ETIMEDOUT;
>> + }
>> + timeout--;
>> + udelay(100);
>> + }
>> + return 0;
>> +}
>> +
>> +static struct exynos4_pm_domain exynos4_pd_mfc = {
>> + .base = (void __iomem *)S5P_PMU_MFC_CONF,
>> + .name = "pd-mfc",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>> +
>> +static struct exynos4_pm_domain exynos4_pd_g3d = {
>> + .base = (void __iomem *)S5P_PMU_G3D_CONF,
>> + .name = "pd-g3d",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>> +
>> +static struct exynos4_pm_domain exynos4_pd_lcd0 = {
>> + .base = (void __iomem *)S5P_PMU_LCD0_CONF,
>> + .name = "pd-lcd0",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>> +
>> +static struct exynos4_pm_domain exynos4_pd_lcd1 = {
>> + .base = (void __iomem *)S5P_PMU_LCD1_CONF,
>> + .name = "pd-lcd1",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>> +
>> +static struct exynos4_pm_domain exynos4_pd_tv = {
>> + .base = (void __iomem *)S5P_PMU_TV_CONF,
>> + .name = "pd-tv",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>> +
>> +static struct exynos4_pm_domain exynos4_pd_cam = {
>> + .base = (void __iomem *)S5P_PMU_CAM_CONF,
>> + .name = "pd-cam",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>> +
>> +static struct exynos4_pm_domain exynos4_pd_gps = {
>> + .base = (void __iomem *)S5P_PMU_GPS_CONF,
>> + .name = "pd-gps",
>> + .pd = {
>> + .power_off = exynos4_pd_power_off,
>> + .power_on = exynos4_pd_power_on,
>> + },
>> +};
>
> I'm not sure if arch/arm/mach-exynos/pm.c is the right place to add this.
> IMHO it would be much better to put PD in separate file, pm-runtime.c
> or something like this.
Ok. I will move this to a new file pm-runtime.c.
>
> Can we assume various exynos versions will have same S5P_PMU_*_CONF
> addresses? For old s3c64xx SoCs such assumption could be valid but I'm
> a bit uncertain about Exynos series.
The addresses could be different, but it can be mapped to a
S5P_PMU_*_CONF. Maybe I did not understand your question here.
>
> IMHO the proper way to proceed would be to convert Samsung power domains
> to generic power domains and then add device tree support. Rather than
Yes, this patchset is intended to replace Samsung specific power
domains with generic power domains.
> doing a sort of work around. Much of the conversion to genpd work have
> been already done [1]. We would perhaps just need to drop the clock gating
> as there seem to be no agreement about this.
Ok. Thanks for the link. I think I missed looking at this patch.
Thanks,
Thomas.
>
> [1]. http://www.mail-archive.com/linux-samsung-soc@vger.kernel.org/msg07431.html
>
> --
>
> Thanks,
> Sylwester
More information about the devicetree-discuss
mailing list