[PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[Oleksandr Tyshchenko]

From: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>

If PSCI is available then most likely we are running on PSCI-enabled
U-Boot which, we assume, has already taken care of resetting CNTVOFF
and updating counter module before switching to non-secure mode
and we don't need to.

As the psci_smp_available() helper always returns false if CONFIG_SMP
is disabled, it can't be used safely as an indicator of PSCI usage.
For that reason, we check for the mandatory PSCI operation to be
available.

Please note, an extra check to prevent secure_cntvoff_init() from
being called for secondary CPUs in headsmp-apmu.S is not needed,
as SMP code for APMU based system is not executed if PSCI is in use.

Signed-off-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
CC: Julien Grall <julien.grall@arm.com>

---
   You can find previous discussions here:
   [v1]  https://lkml.org/lkml/2019/4/17/810
   [v2]  https://lkml.org/lkml/2019/5/3/338
   [v3]  https://lkml.org/lkml/2019/5/10/415
   [RFC] https://lkml.org/lkml/2019/5/10/473
   [v4]  https://lkml.org/lkml/2019/5/14/550

   Changes in v2:
      - Clarify patch subject/description
      - Don't use CONFIG_ARM_PSCI option, check whether the PSCI is available,
        by using psci_smp_available()
      - Check whether we are running on top of Xen, by using xen_domain()

   Changes in v3:
      - Don't check for the presence of Xen

   Changes in v4:
      - Don't use psci_smp_available() helper, check for psci_ops.cpu_on
        directly
      - Skip updating counter module if PSCI is available

   Changes in v5:
      - Check for psci_ops.cpu_on if CONFIG_ARM_PSCI_FW is defined
---
 arch/arm/mach-shmobile/setup-rcar-gen2.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/arch/arm/mach-shmobile/setup-rcar-gen2.c b/arch/arm/mach-shmobile/setup-rcar-gen2.c
index eea60b2..0eeca49b 100644
--- a/arch/arm/mach-shmobile/setup-rcar-gen2.c
+++ b/arch/arm/mach-shmobile/setup-rcar-gen2.c
@@ -17,6 +17,7 @@
 #include <linux/of.h>
 #include <linux/of_fdt.h>
 #include <linux/of_platform.h>
+#include <linux/psci.h>
 #include <asm/mach/arch.h>
 #include <asm/secure_cntvoff.h>
 #include "common.h"
@@ -62,6 +63,21 @@ void __init rcar_gen2_timer_init(void)
 {
 	void __iomem *base;
 	u32 freq;
+	bool need_update = true;
+
+	/*
+	 * If PSCI is available then most likely we are running on PSCI-enabled
+	 * U-Boot which, we assume, has already taken care of resetting CNTVOFF
+	 * and updating counter module before switching to non-secure mode
+	 * and we don't need to.
+	 */
+#if defined(CONFIG_ARM_PSCI_FW)
+	if (psci_ops.cpu_on)
+		need_update = false;
+#endif
+
+	if (need_update == false)
+		goto skip_update;
 
 	secure_cntvoff_init();
 
@@ -102,6 +118,7 @@ void __init rcar_gen2_timer_init(void)
 
 	iounmap(base);
 
+skip_update:
 	of_clk_init(NULL);
 	timer_probe();
 }
-- 
2.7.4

Re: [PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[Oleksandr]

Hello, all


Gentle reminder...

-- 
Regards,

Oleksandr Tyshchenko

Re: [PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[Geert Uytterhoeven]

Hi Oleksandr,

On Fri, May 17, 2019 at 11:58 AM Oleksandr Tyshchenko
<olekstysh@gmail.com> wrote:
> From: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
>
> If PSCI is available then most likely we are running on PSCI-enabled
> U-Boot which, we assume, has already taken care of resetting CNTVOFF
> and updating counter module before switching to non-secure mode
> and we don't need to.
>
> As the psci_smp_available() helper always returns false if CONFIG_SMP
> is disabled, it can't be used safely as an indicator of PSCI usage.
> For that reason, we check for the mandatory PSCI operation to be
> available.
>
> Please note, an extra check to prevent secure_cntvoff_init() from
> being called for secondary CPUs in headsmp-apmu.S is not needed,
> as SMP code for APMU based system is not executed if PSCI is in use.
>
> Signed-off-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>

>    Changes in v5:
>       - Check for psci_ops.cpu_on if CONFIG_ARM_PSCI_FW is defined

Thanks for the update!

Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>

Two cosmetic comments below. I'll leave it to Simon to ignore them for
applying ;-)

> @@ -62,6 +63,21 @@ void __init rcar_gen2_timer_init(void)
>  {
>         void __iomem *base;
>         u32 freq;
> +       bool need_update = true;

Some people like reverse Xmas tree declaration order...

> +
> +       /*
> +        * If PSCI is available then most likely we are running on PSCI-enabled
> +        * U-Boot which, we assume, has already taken care of resetting CNTVOFF
> +        * and updating counter module before switching to non-secure mode
> +        * and we don't need to.
> +        */
> +#if defined(CONFIG_ARM_PSCI_FW)

#ifdef CONFIG_ARM_PSCI_FW ?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[Oleksandr]

On 28.05.19 11:27, Geert Uytterhoeven wrote:
> Hi Oleksandr,

Hi Geert, all


> Thanks for the update!
>
> Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>

Thank you for review!


Just FYI:

Required support for Gen2 Stout board has been already added to Xen [1].

What needs to be done next is to add support for handling device-tree 
nodes with "interrupts-extended" property in Xen [2].
This is related to Gen3 as well.


[1] 
https://www.mail-archive.com/xen-devel@lists.xenproject.org/msg44386.html

[2] 
https://www.mail-archive.com/xen-devel@lists.xenproject.org/msg45951.html

>
> Two cosmetic comments below. I'll leave it to Simon to ignore them for
> applying ;-)
>
>> @@ -62,6 +63,21 @@ void __init rcar_gen2_timer_init(void)
>>   {
>>          void __iomem *base;
>>          u32 freq;
>> +       bool need_update = true;
> Some people like reverse Xmas tree declaration order...
>
>> +
>> +       /*
>> +        * If PSCI is available then most likely we are running on PSCI-enabled
>> +        * U-Boot which, we assume, has already taken care of resetting CNTVOFF
>> +        * and updating counter module before switching to non-secure mode
>> +        * and we don't need to.
>> +        */
>> +#if defined(CONFIG_ARM_PSCI_FW)
> #ifdef CONFIG_ARM_PSCI_FW ?
>
> Gr{oetje,eeting}s,
>
>                          Geert
>
-- 
Regards,

Oleksandr Tyshchenko

Re: [PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[Simon Horman]

On Tue, May 28, 2019 at 10:27:56AM +0200, Geert Uytterhoeven wrote:
> Hi Oleksandr,
> 
> On Fri, May 17, 2019 at 11:58 AM Oleksandr Tyshchenko
> <olekstysh@gmail.com> wrote:
> > From: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
> >
> > If PSCI is available then most likely we are running on PSCI-enabled
> > U-Boot which, we assume, has already taken care of resetting CNTVOFF
> > and updating counter module before switching to non-secure mode
> > and we don't need to.
> >
> > As the psci_smp_available() helper always returns false if CONFIG_SMP
> > is disabled, it can't be used safely as an indicator of PSCI usage.
> > For that reason, we check for the mandatory PSCI operation to be
> > available.
> >
> > Please note, an extra check to prevent secure_cntvoff_init() from
> > being called for secondary CPUs in headsmp-apmu.S is not needed,
> > as SMP code for APMU based system is not executed if PSCI is in use.
> >
> > Signed-off-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>
> 
> >    Changes in v5:
> >       - Check for psci_ops.cpu_on if CONFIG_ARM_PSCI_FW is defined
> 
> Thanks for the update!
> 
> Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
> 
> Two cosmetic comments below. I'll leave it to Simon to ignore them for
> applying ;-)

Oleksandr, could I trouble you to respin with the
changes suggested by Geert?

> 
> > @@ -62,6 +63,21 @@ void __init rcar_gen2_timer_init(void)
> >  {
> >         void __iomem *base;
> >         u32 freq;
> > +       bool need_update = true;
> 
> Some people like reverse Xmas tree declaration order...
> 
> > +
> > +       /*
> > +        * If PSCI is available then most likely we are running on PSCI-enabled
> > +        * U-Boot which, we assume, has already taken care of resetting CNTVOFF
> > +        * and updating counter module before switching to non-secure mode
> > +        * and we don't need to.
> > +        */
> > +#if defined(CONFIG_ARM_PSCI_FW)
> 
> #ifdef CONFIG_ARM_PSCI_FW ?
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> -- 
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
> 

Re: [PATCH V5] ARM: mach-shmobile: Don't init CNTVOFF/counter if PSCI is available

[Oleksandr]

> Oleksandr, could I trouble you to respin with the
> changes suggested by Geert?

Hi, Simon.

Sure, will send an updated patch tomorrow.


>
-- 
Regards,

Oleksandr Tyshchenko