Tegra SoCs with 64-bit ARM support don't currently support deep CPU
low-power states in mainline Linux. When this support is added in the
future, it will probably look rather different from the existing
32-bit ARM support, since the ARM64 maintainers' strong preference is
to use PSCI to implement it.
So, for the time being, prevent the CPU suspend-related code and data
in the Tegra PMC driver from compiling on ARM64.
Signed-off-by: Paul Walmsley <[email protected]>
Signed-off-by: Paul Walmsley <[email protected]>
Cc: Thierry Reding <[email protected]>
Cc: Allen Martin <[email protected]>
Cc: Stephen Warren <[email protected]>
Cc: Alexandre Courbot <[email protected]>
---
Applies on next-20141209.
Intended for v3.20.
Boot-tested on Tegra124 Jetson TK1 on next-20141209.
Also boot-tested on Tegra132 Norrin FFD on next-20141209 + some unrelated
patches.
drivers/soc/tegra/pmc.c | 7 +++++--
include/soc/tegra/pm.h | 2 +-
2 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/soc/tegra/pmc.c b/drivers/soc/tegra/pmc.c
index a2c0ceb95f8f..4bdc654bd747 100644
--- a/drivers/soc/tegra/pmc.c
+++ b/drivers/soc/tegra/pmc.c
@@ -739,7 +739,7 @@ static int tegra_pmc_probe(struct platform_device *pdev)
return 0;
}
-#ifdef CONFIG_PM_SLEEP
+#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
static int tegra_pmc_suspend(struct device *dev)
{
tegra_pmc_writel(virt_to_phys(tegra_resume), PMC_SCRATCH41);
@@ -753,10 +753,11 @@ static int tegra_pmc_resume(struct device *dev)
return 0;
}
-#endif
static SIMPLE_DEV_PM_OPS(tegra_pmc_pm_ops, tegra_pmc_suspend, tegra_pmc_resume);
+#endif
+
static const char * const tegra20_powergates[] = {
[TEGRA_POWERGATE_CPU] = "cpu",
[TEGRA_POWERGATE_3D] = "3d",
@@ -894,7 +895,9 @@ static struct platform_driver tegra_pmc_driver = {
.name = "tegra-pmc",
.suppress_bind_attrs = true,
.of_match_table = tegra_pmc_match,
+#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
.pm = &tegra_pmc_pm_ops,
+#endif
},
.probe = tegra_pmc_probe,
};
diff --git a/include/soc/tegra/pm.h b/include/soc/tegra/pm.h
index 30fe2078a547..03909101d4e7 100644
--- a/include/soc/tegra/pm.h
+++ b/include/soc/tegra/pm.h
@@ -17,7 +17,7 @@ enum tegra_suspend_mode {
TEGRA_MAX_SUSPEND_MODE,
};
-#ifdef CONFIG_PM_SLEEP
+#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
enum tegra_suspend_mode
tegra_pm_validate_suspend_mode(enum tegra_suspend_mode mode);
--
2.1.3
On Tue, Dec 09, 2014 at 10:36:50PM +0000, Paul Walmsley wrote:
>
> Tegra SoCs with 64-bit ARM support don't currently support deep CPU
> low-power states in mainline Linux. When this support is added in the
> future, it will probably look rather different from the existing
> 32-bit ARM support, since the ARM64 maintainers' strong preference is
> to use PSCI to implement it.
>
> So, for the time being, prevent the CPU suspend-related code and data
> in the Tegra PMC driver from compiling on ARM64.
>
> Signed-off-by: Paul Walmsley <[email protected]>
> Signed-off-by: Paul Walmsley <[email protected]>
> Cc: Thierry Reding <[email protected]>
> Cc: Allen Martin <[email protected]>
> Cc: Stephen Warren <[email protected]>
> Cc: Alexandre Courbot <[email protected]>
> ---
> Applies on next-20141209.
> Intended for v3.20.
> Boot-tested on Tegra124 Jetson TK1 on next-20141209.
> Also boot-tested on Tegra132 Norrin FFD on next-20141209 + some unrelated
> patches.
>
> drivers/soc/tegra/pmc.c | 7 +++++--
> include/soc/tegra/pm.h | 2 +-
> 2 files changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/soc/tegra/pmc.c b/drivers/soc/tegra/pmc.c
> index a2c0ceb95f8f..4bdc654bd747 100644
> --- a/drivers/soc/tegra/pmc.c
> +++ b/drivers/soc/tegra/pmc.c
> @@ -739,7 +739,7 @@ static int tegra_pmc_probe(struct platform_device *pdev)
> return 0;
> }
>
> -#ifdef CONFIG_PM_SLEEP
> +#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
I would slightly prefer having this as separate #ifdef lines to make it
clearer that we're special-casing 32-bit ARM.
> static int tegra_pmc_suspend(struct device *dev)
> {
> tegra_pmc_writel(virt_to_phys(tegra_resume), PMC_SCRATCH41);
> @@ -753,10 +753,11 @@ static int tegra_pmc_resume(struct device *dev)
>
> return 0;
> }
> -#endif
>
> static SIMPLE_DEV_PM_OPS(tegra_pmc_pm_ops, tegra_pmc_suspend, tegra_pmc_resume);
>
> +#endif
I don't think this is necessary, SIMPLE_DEV_PM_OPS will not use the
supend or resume hooks if PM_SLEEP is disabled. The result will be an
empty dev_pm_ops structure. While that's somewhat useless, at least
it'll allow...
> +
> static const char * const tegra20_powergates[] = {
> [TEGRA_POWERGATE_CPU] = "cpu",
> [TEGRA_POWERGATE_3D] = "3d",
> @@ -894,7 +895,9 @@ static struct platform_driver tegra_pmc_driver = {
> .name = "tegra-pmc",
> .suppress_bind_attrs = true,
> .of_match_table = tegra_pmc_match,
> +#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
> .pm = &tegra_pmc_pm_ops,
> +#endif
... this #ifdef to be avoided.
> },
> .probe = tegra_pmc_probe,
> };
> diff --git a/include/soc/tegra/pm.h b/include/soc/tegra/pm.h
> index 30fe2078a547..03909101d4e7 100644
> --- a/include/soc/tegra/pm.h
> +++ b/include/soc/tegra/pm.h
> @@ -17,7 +17,7 @@ enum tegra_suspend_mode {
> TEGRA_MAX_SUSPEND_MODE,
> };
>
> -#ifdef CONFIG_PM_SLEEP
> +#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
Similarly to the above, I'd prefer this to be two separate lines. That
also has the advantage that if we do add code for suspend/resume on 64
bit ARM later on we don't need to untangle the conditional again.
Thierry
Hello Thierry
On Wed, 10 Dec 2014, Thierry Reding wrote:
> On Tue, Dec 09, 2014 at 10:36:50PM +0000, Paul Walmsley wrote:
> >
> > Tegra SoCs with 64-bit ARM support don't currently support deep CPU
> > low-power states in mainline Linux. When this support is added in the
> > future, it will probably look rather different from the existing
> > 32-bit ARM support, since the ARM64 maintainers' strong preference is
> > to use PSCI to implement it.
> >
> > So, for the time being, prevent the CPU suspend-related code and data
> > in the Tegra PMC driver from compiling on ARM64.
> >
> > Signed-off-by: Paul Walmsley <[email protected]>
> > Signed-off-by: Paul Walmsley <[email protected]>
> > Cc: Thierry Reding <[email protected]>
> > Cc: Allen Martin <[email protected]>
> > Cc: Stephen Warren <[email protected]>
> > Cc: Alexandre Courbot <[email protected]>
> > ---
> > Applies on next-20141209.
> > Intended for v3.20.
> > Boot-tested on Tegra124 Jetson TK1 on next-20141209.
> > Also boot-tested on Tegra132 Norrin FFD on next-20141209 + some unrelated
> > patches.
> >
> > drivers/soc/tegra/pmc.c | 7 +++++--
> > include/soc/tegra/pm.h | 2 +-
> > 2 files changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/soc/tegra/pmc.c b/drivers/soc/tegra/pmc.c
> > index a2c0ceb95f8f..4bdc654bd747 100644
> > --- a/drivers/soc/tegra/pmc.c
> > +++ b/drivers/soc/tegra/pmc.c
> > @@ -739,7 +739,7 @@ static int tegra_pmc_probe(struct platform_device *pdev)
> > return 0;
> > }
> >
> > -#ifdef CONFIG_PM_SLEEP
> > +#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
>
> I would slightly prefer having this as separate #ifdef lines to make it
> clearer that we're special-casing 32-bit ARM.
That's fine, I'll update that when the patch is reposted against
v3.19-rc1.
> > static int tegra_pmc_suspend(struct device *dev)
> > {
> > tegra_pmc_writel(virt_to_phys(tegra_resume), PMC_SCRATCH41);
> > @@ -753,10 +753,11 @@ static int tegra_pmc_resume(struct device *dev)
> >
> > return 0;
> > }
> > -#endif
> >
> > static SIMPLE_DEV_PM_OPS(tegra_pmc_pm_ops, tegra_pmc_suspend, tegra_pmc_resume);
> >
> > +#endif
>
> I don't think this is necessary, SIMPLE_DEV_PM_OPS will not use the
> supend or resume hooks if PM_SLEEP is disabled.
PM_SLEEP is enabled by default for the ARM64 defconfig, and I don't plan
to patch that out.
> > diff --git a/include/soc/tegra/pm.h b/include/soc/tegra/pm.h
> > index 30fe2078a547..03909101d4e7 100644
> > --- a/include/soc/tegra/pm.h
> > +++ b/include/soc/tegra/pm.h
> > @@ -17,7 +17,7 @@ enum tegra_suspend_mode {
> > TEGRA_MAX_SUSPEND_MODE,
> > };
> >
> > -#ifdef CONFIG_PM_SLEEP
> > +#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_ARM)
>
> Similarly to the above, I'd prefer this to be two separate lines. That
> also has the advantage that if we do add code for suspend/resume on 64
> bit ARM later on we don't need to untangle the conditional again.
I'm fine to change it, although it seems rather unlikely that this will be
a significant issue.
- Paul