ARM: MCPM: Add provision for platform specific wfi alternative
diff mbox series

Message ID 20190210231212.20470-1-stuart.menefy@mathembedded.com
State New
Headers show
Series
  • ARM: MCPM: Add provision for platform specific wfi alternative
Related show

Commit Message

Stuart Menefy Feb. 10, 2019, 11:12 p.m. UTC
On some platforms (in particular the Exynos5260) the wfi is not
executed directly by the operating system but by trapping into the
secure monitor. This allows us to replicate the behaviour of the
vendor supplied kernel using the MCPM framework.

Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com>
---
 arch/arm/common/mcpm_entry.c | 8 ++++++--
 arch/arm/include/asm/mcpm.h  | 1 +
 2 files changed, 7 insertions(+), 2 deletions(-)

Comments

Stuart Menefy Feb. 11, 2019, 1:53 p.m. UTC | #1
Folks

On Sun, 10 Feb 2019 at 23:12, Stuart Menefy
<stuart.menefy@mathembedded.com> wrote:
>
> On some platforms (in particular the Exynos5260) the wfi is not
> executed directly by the operating system but by trapping into the
> secure monitor. This allows us to replicate the behaviour of the
> vendor supplied kernel using the MCPM framework.
>
> Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com>

I forgot to add, this is a part of a series of patches which updates
the Exynos 5260
support, allowing use of the kernel on secure parts and adding suspend
functionality,
Version 1 of the patches has been posted to the samsung-soc mailing list:
https://www.spinics.net/lists/linux-samsung-soc/#65035
and I'm just about to post version 2. However as this is the only patch which
touches generic ARM files I though I'd better check the approach is acceptable
to a wider audience.

Stuart
Russell King - ARM Linux admin Feb. 11, 2019, 2:12 p.m. UTC | #2
On Mon, Feb 11, 2019 at 01:53:14PM +0000, Stuart Menefy wrote:
> Folks
> 
> On Sun, 10 Feb 2019 at 23:12, Stuart Menefy
> <stuart.menefy@mathembedded.com> wrote:
> >
> > On some platforms (in particular the Exynos5260) the wfi is not
> > executed directly by the operating system but by trapping into the
> > secure monitor. This allows us to replicate the behaviour of the
> > vendor supplied kernel using the MCPM framework.
> >
> > Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com>
> 
> I forgot to add, this is a part of a series of patches which updates
> the Exynos 5260
> support, allowing use of the kernel on secure parts and adding suspend
> functionality,
> Version 1 of the patches has been posted to the samsung-soc mailing list:
> https://www.spinics.net/lists/linux-samsung-soc/#65035
> and I'm just about to post version 2. However as this is the only patch which
> touches generic ARM files I though I'd better check the approach is acceptable
> to a wider audience.

Please don't limit patches to just the SoC specific mailing list - copy
linux-arm-kernel and linux-kernel as well.  Thanks.
Nicolas Pitre Feb. 11, 2019, 3:57 p.m. UTC | #3
On Mon, 11 Feb 2019, Stuart Menefy wrote:

> Folks
> 
> On Sun, 10 Feb 2019 at 23:12, Stuart Menefy
> <stuart.menefy@mathembedded.com> wrote:
> >
> > On some platforms (in particular the Exynos5260) the wfi is not
> > executed directly by the operating system but by trapping into the
> > secure monitor. This allows us to replicate the behaviour of the
> > vendor supplied kernel using the MCPM framework.
> >
> > Signed-off-by: Stuart Menefy <stuart.menefy@mathembedded.com>
> 
> I forgot to add, this is a part of a series of patches which updates
> the Exynos 5260
> support, allowing use of the kernel on secure parts and adding suspend
> functionality,
> Version 1 of the patches has been posted to the samsung-soc mailing list:
> https://www.spinics.net/lists/linux-samsung-soc/#65035
> and I'm just about to post version 2. However as this is the only patch which
> touches generic ARM files I though I'd better check the approach is acceptable
> to a wider audience.

I was about to ask that you also post those patches making use of this 
change.  It is not possible to evaluate if the approach makes sense 
without seeing how it is used.


Nicolas

Patch
diff mbox series

diff --git a/arch/arm/common/mcpm_entry.c b/arch/arm/common/mcpm_entry.c
index ad574d20415c..524be627c87e 100644
--- a/arch/arm/common/mcpm_entry.c
+++ b/arch/arm/common/mcpm_entry.c
@@ -289,8 +289,12 @@  void mcpm_cpu_power_down(void)
 	__mcpm_cpu_down(cpu, cluster);
 
 	/* Now we are prepared for power-down, do it: */
-	if (cpu_going_down)
-		wfi();
+	if (cpu_going_down) {
+		if (platform_ops->wfi_alternative)
+			platform_ops->wfi_alternative(last_man);
+		else
+			wfi();
+	}
 
 	/*
 	 * It is possible for a power_up request to happen concurrently
diff --git a/arch/arm/include/asm/mcpm.h b/arch/arm/include/asm/mcpm.h
index acd4983d9b1f..96fdb41ac626 100644
--- a/arch/arm/include/asm/mcpm.h
+++ b/arch/arm/include/asm/mcpm.h
@@ -223,6 +223,7 @@  struct mcpm_platform_ops {
 	int (*cluster_powerup)(unsigned int cluster);
 	void (*cpu_suspend_prepare)(unsigned int cpu, unsigned int cluster);
 	void (*cpu_powerdown_prepare)(unsigned int cpu, unsigned int cluster);
+	void (*wfi_alternative)(bool last_man);
 	void (*cluster_powerdown_prepare)(unsigned int cluster);
 	void (*cpu_cache_disable)(void);
 	void (*cluster_cache_disable)(void);