Re: [PATCH 01/16] ARM: scu: Provide support for parsing SCU device node to enable SCU

[Jisheng Zhang]

Hi Pankaj,

On Mon, 14 Nov 2016 14:54:59 +0800 Jisheng Zhang wrote:

> On Mon, 14 Nov 2016 14:12:51 +0800 Jisheng Zhang wrote:
> 
> > Hi Pankaj,
> > 
> > On Mon, 14 Nov 2016 10:31:56 +0530 Pankaj Dubey wrote:
> >   
> > > Many platforms are duplicating code for enabling SCU, lets add
> > > common code to enable SCU by parsing SCU device node so the duplication
> > > in each platform can be avoided.
> > > 
> > > CC: Krzysztof Kozlowski <krzk@xxxxxxxxxx>
> > > CC: Jisheng Zhang <jszhang@xxxxxxxxxxx>
> > > CC: Russell King <linux@xxxxxxxxxxxxxxx>
> > > CC: Dinh Nguyen <dinguyen@xxxxxxxxxxxxxxxxxxxxx>
> > > CC: Patrice Chotard <patrice.chotard@xxxxxx>
> > > CC: Linus Walleij <linus.walleij@xxxxxxxxxx>
> > > CC: Liviu Dudau <liviu.dudau@xxxxxxx>
> > > CC: Ray Jui <rjui@xxxxxxxxxxxx>
> > > CC: Stephen Warren <swarren@xxxxxxxxxxxxx>
> > > CC: Heiko Stuebner <heiko@xxxxxxxxx>
> > > CC: Shawn Guo <shawnguo@xxxxxxxxxx>
> > > CC: Michal Simek <michal.simek@xxxxxxxxxx>
> > > CC: Wei Xu <xuwei5@xxxxxxxxxxxxx>
> > > CC: Andrew Lunn <andrew@xxxxxxx>
> > > CC: Jun Nie <jun.nie@xxxxxxxxxx> 
> > > Suggested-by: Arnd Bergmann <arnd@xxxxxxxx>
> > > Signed-off-by: Pankaj Dubey <pankaj.dubey@xxxxxxxxxxx>
> > > ---
> > >  arch/arm/include/asm/smp_scu.h |  4 +++
> > >  arch/arm/kernel/smp_scu.c      | 56 ++++++++++++++++++++++++++++++++++++++++++
> > >  2 files changed, 60 insertions(+)
> > > 
> > > diff --git a/arch/arm/include/asm/smp_scu.h b/arch/arm/include/asm/smp_scu.h
> > > index bfe163c..fdeec07 100644
> > > --- a/arch/arm/include/asm/smp_scu.h
> > > +++ b/arch/arm/include/asm/smp_scu.h
> > > @@ -39,8 +39,12 @@ static inline int scu_power_mode(void __iomem *scu_base, unsigned int mode)
> > >  
> > >  #if defined(CONFIG_SMP) && defined(CONFIG_HAVE_ARM_SCU)
> > >  void scu_enable(void __iomem *scu_base);
> > > +void __iomem *of_scu_get_base(void);
> > > +int of_scu_enable(void);
> > >  #else
> > >  static inline void scu_enable(void __iomem *scu_base) {}
> > > +static inline void __iomem *of_scu_get_base(void) {return NULL; }
> > > +static inline int of_scu_enable(void) {return 0; }
> > >  #endif
> > >  
> > >  #endif
> > > diff --git a/arch/arm/kernel/smp_scu.c b/arch/arm/kernel/smp_scu.c
> > > index 72f9241..d0ac3ed 100644
> > > --- a/arch/arm/kernel/smp_scu.c
> > > +++ b/arch/arm/kernel/smp_scu.c
> > > @@ -10,6 +10,7 @@
> > >   */
> > >  #include <linux/init.h>
> > >  #include <linux/io.h>
> > > +#include <linux/of_address.h>
> > >  
> > >  #include <asm/smp_plat.h>
> > >  #include <asm/smp_scu.h>
> > > @@ -70,6 +71,61 @@ void scu_enable(void __iomem *scu_base)
> > >  	 */
> > >  	flush_cache_all();
> > >  }
> > > +
> > > +static const struct of_device_id scu_match[] = {
> > > +	{ .compatible = "arm,cortex-a9-scu", },
> > > +	{ .compatible = "arm,cortex-a5-scu", },
> > > +	{ }
> > > +};
> > > +
> > > +/*
> > > + * Helper API to get SCU base address
> > > + * In case platform DT do not have SCU node, or iomap fails
> > > + * this call will fallback and will try to map via call to
> > > + * scu_a9_get_base.
> > > + * This will return ownership of scu_base to the caller
> > > + */
> > > +void __iomem *of_scu_get_base(void)
> > > +{
> > > +	unsigned long base = 0;
> > > +	struct device_node *np;
> > > +	void __iomem *scu_base;
> > > +
> > > +	np = of_find_matching_node(NULL, scu_match);    
> > 
> > could we check np before calling of_iomap()?
> >   
> > > +	scu_base = of_iomap(np, 0);
> > > +	of_node_put(np);
> > > +	if (!scu_base) {
> > > +		pr_err("%s failed to map scu_base via DT\n", __func__);    
> > 
> > For non-ca5, non-ca9 based SoCs, we'll see this error msg. We understand
> > what does it mean, but it may confuse normal users. In current version,
> > berlin doesn't complain like this for non-ca9 SoCs  
> 
> oops, I just realized that the non-ca9 berlin arm SoC version isn't upstreamed.
> Below is the draft version I planed. Basically speaking, the code tries to
> find "arm,cortex-a9-scu" node from DT, if can't, we think we don't need to
> worry about SCU. Is there any elegant solution for my situation?

I just realized that (another realized :D) we uses PSCI enable method for
non-ca9 base SoCs, so "marvell,berlin-smp" enable method is only for the SoCs
which have CA9 compatible SCU. Sorry for the noise, your patch looks good to me.