[PATCH] ARM: shmobile: r8a7790 SMP prototype v5

[PATCH] ARM: shmobile: r8a7790 SMP prototype v5

[Magnus Damm]

From: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>

Add SMP prototype support for r8a7790 by enabling
one cluster of either 4 x Cortex-A7 or 4 x Cortex-A15.

This patch only adds support for booting, at this point
no CPU Hotplug is included. In v5 both the big and the
LITTLE clusters are working with either architected timers
(in physical mode only for CA7) or CMT and broadcast timer.

On r8a7790 the MD6 pin control boot processor, and on the
Lager board SW8.7 can be used to select big or LITTLE.

In big boot mode the CPU cores will be configured as:
 - CPU0, CPU1, CPU2, CPU3: enabled (CA15)
 - CPU4, CPU5, CPU6, CPU7: disabled (CA7)

In LITTLE boot mode the CPU cores will be configured as:
 - CPU0, CPU5, CPU6, CPU7: enabled (CA7)
 - CPU1, CPU2, CPU3, CPU4: disabled (CA15)

Base addresses for RST and APMU are still hard coded
but DT will be used for them in future versions.

Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
Signed-off-by: Magnus Damm <damm@opensource.se>
---

 Changes since v4:
 - Rolled in the following patches from Kuribayashi-san, thanks!
   ARM: shmobile: smp-r8a7790: remove the obsolete RESCNT register
   ARM: shmobile: smp-r8a7790: rename s/carescnt/rescnt/g
   ARM: shmobile: smp-r8a7790: rename s/cabar/bar/g as well
   ARM: shmobile: smp-r8a7790: power up a CPU core at .smp_boot_secondary stage

 Written against kernel.org renesas git tag renesas-devel-20130717

 arch/arm/boot/dts/r8a7790.dtsi                 |   49 ++++++++
 arch/arm/mach-shmobile/Makefile                |    1 
 arch/arm/mach-shmobile/board-lager-reference.c |    1 
 arch/arm/mach-shmobile/board-lager.c           |    1 
 arch/arm/mach-shmobile/include/mach/common.h   |    1 
 arch/arm/mach-shmobile/include/mach/r8a7790.h  |    1 
 arch/arm/mach-shmobile/setup-r8a7790.c         |    1 
 arch/arm/mach-shmobile/smp-r8a7790.c           |  143 ++++++++++++++++++++++++
 8 files changed, 198 insertions(+)

--- 0001/arch/arm/boot/dts/r8a7790.dtsi
+++ work/arch/arm/boot/dts/r8a7790.dtsi	2013-07-18 05:35:38.000000000 +0900
@@ -24,6 +24,55 @@
 			reg = <0>;
 			clock-frequency = <1300000000>;
 		};
+
+		cpu1: cpu at 1 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a15";
+			reg = <1>;
+			clock-frequency = <1300000000>;
+		};
+
+		cpu2: cpu at 2 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a15";
+			reg = <2>;
+			clock-frequency = <1300000000>;
+		};
+
+		cpu3: cpu at 3 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a15";
+			reg = <3>;
+			clock-frequency = <1300000000>;
+		};
+
+		cpu4: cpu at 4 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a7";
+			reg = <0x100>;
+			clock-frequency = <780000000>;
+		};
+
+		cpu5: cpu at 5 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a7";
+			reg = <0x101>;
+			clock-frequency = <780000000>;
+		};
+
+		cpu6: cpu at 6 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a7";
+			reg = <0x102>;
+			clock-frequency = <780000000>;
+		};
+
+		cpu7: cpu at 7 {
+			device_type = "cpu";
+			compatible = "arm,cortex-a7";
+			reg = <0x103>;
+			clock-frequency = <780000000>;
+		};
 	};
 
 	gic: interrupt-controller at f1001000 {
--- 0001/arch/arm/mach-shmobile/Makefile
+++ work/arch/arm/mach-shmobile/Makefile	2013-07-18 05:35:38.000000000 +0900
@@ -19,6 +19,7 @@ obj-$(CONFIG_ARCH_EMEV2)	+= setup-emev2.
 smp-y				:= platsmp.o headsmp.o
 smp-$(CONFIG_ARCH_SH73A0)	+= smp-sh73a0.o headsmp-scu.o
 smp-$(CONFIG_ARCH_R8A7779)	+= smp-r8a7779.o headsmp-scu.o
+smp-$(CONFIG_ARCH_R8A7790)	+= smp-r8a7790.o
 smp-$(CONFIG_ARCH_EMEV2)	+= smp-emev2.o headsmp-scu.o
 
 # IRQ objects
--- 0001/arch/arm/mach-shmobile/board-lager-reference.c
+++ work/arch/arm/mach-shmobile/board-lager-reference.c	2013-07-18 05:35:38.000000000 +0900
@@ -38,6 +38,7 @@ static const char *lager_boards_compat_d
 };
 
 DT_MACHINE_START(LAGER_DT, "lager")
+	.smp		= smp_ops(r8a7790_smp_ops),
 	.init_early	= r8a7790_init_delay,
 	.init_machine	= lager_add_standard_devices,
 	.init_time	= r8a7790_timer_init,
--- 0001/arch/arm/mach-shmobile/board-lager.c
+++ work/arch/arm/mach-shmobile/board-lager.c	2013-07-18 05:35:38.000000000 +0900
@@ -133,6 +133,7 @@ static const char *lager_boards_compat_d
 };
 
 DT_MACHINE_START(LAGER_DT, "lager")
+	.smp		= smp_ops(r8a7790_smp_ops),
 	.init_early	= r8a7790_init_delay,
 	.init_time	= r8a7790_timer_init,
 	.init_machine	= lager_add_standard_devices,
--- 0001/arch/arm/mach-shmobile/include/mach/common.h
+++ work/arch/arm/mach-shmobile/include/mach/common.h	2013-07-18 05:35:38.000000000 +0900
@@ -11,6 +11,7 @@ extern void shmobile_boot_vector(void);
 extern unsigned long shmobile_boot_fn;
 extern unsigned long shmobile_boot_arg;
 extern void shmobile_boot_scu(void);
+extern void shmobile_invalidate_start(void);
 struct clk;
 extern int shmobile_clk_init(void);
 extern void shmobile_handle_irq_intc(struct pt_regs *);
--- 0001/arch/arm/mach-shmobile/include/mach/r8a7790.h
+++ work/arch/arm/mach-shmobile/include/mach/r8a7790.h	2013-07-18 05:35:38.000000000 +0900
@@ -7,6 +7,7 @@ void r8a7790_clock_init(void);
 void r8a7790_pinmux_init(void);
 void r8a7790_init_delay(void);
 void r8a7790_timer_init(void);
+extern struct smp_operations r8a7790_smp_ops;
 
 #define MD(nr) BIT(nr)
 u32 r8a7790_read_mode_pins(void);
--- 0001/arch/arm/mach-shmobile/setup-r8a7790.c
+++ work/arch/arm/mach-shmobile/setup-r8a7790.c	2013-07-18 05:35:38.000000000 +0900
@@ -281,6 +281,7 @@ static const char *r8a7790_boards_compat
 };
 
 DT_MACHINE_START(R8A7790_DT, "Generic R8A7790 (Flattened Device Tree)")
+	.smp		= smp_ops(r8a7790_smp_ops),
 	.init_early	= r8a7790_init_delay,
 	.init_time	= r8a7790_timer_init,
 	.dt_compat	= r8a7790_boards_compat_dt,
--- /dev/null
+++ work/arch/arm/mach-shmobile/smp-r8a7790.c	2013-07-18 05:40:04.000000000 +0900
@@ -0,0 +1,143 @@
+/*
+ * SMP support for r8a7790
+ *
+ * Copyright (C) 2012-2013 Renesas Solutions Corp.
+ * Copyright (C) 2012 Takashi Yoshii <takashi.yoshii.ze@renesas.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 2 of the License.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/smp.h>
+#include <linux/io.h>
+#include <asm/smp_plat.h>
+#include <mach/common.h>
+
+#define RST	0xe6160000
+#define CA15BAR	0x0020
+#define CA7BAR	0x0030
+#define CA15RESCNT 0x0040
+#define CA7RESCNT 0x0044
+
+#define APMU	0xe6150000
+#define CA15WUPCR 0x2010
+#define CA7WUPCR 0x1010
+
+#define MERAM	0xe8080000
+
+enum { R8A7790_CLST_CA15, R8A7790_CLST_CA7, R8A7790_CLST_NR };
+
+static struct {
+	unsigned int bar;
+	unsigned int rescnt;
+	unsigned int rescnt_magic;
+} r8a7790_clst[R8A7790_CLST_NR] = {
+	[R8A7790_CLST_CA15] = {
+		.bar = CA15BAR,
+		.rescnt = CA15RESCNT,
+		.rescnt_magic = 0xa5a50000,
+	},
+	[R8A7790_CLST_CA7] = {
+		.bar = CA7BAR,
+		.rescnt = CA7RESCNT,
+		.rescnt_magic = 0x5a5a0000,
+	},
+};
+
+#define r8a7790_clst_id(cpu) (cpu_logical_map((cpu)) >> 8)
+#define r8a7790_cpu_id(cpu) (cpu_logical_map((cpu)) & 0xff)
+
+static void r8a7790_deassert_reset(unsigned int cpu)
+{
+	void __iomem *p, *rescnt;
+	u32 bar, mask, magic;
+	unsigned int clst_id = r8a7790_clst_id(cpu);
+
+	/* setup reset vectors */
+	p = ioremap_nocache(RST, 0x63);
+	bar = (MERAM >> 8) & 0xfffffc00;
+	writel_relaxed(bar, p + r8a7790_clst[clst_id].bar);
+	writel_relaxed(bar | 0x10, p + r8a7790_clst[clst_id].bar);
+
+	/* enable per-core clocks */
+	mask = BIT(3 - r8a7790_cpu_id(cpu));
+	magic = r8a7790_clst[clst_id].rescnt_magic;
+	rescnt = p + r8a7790_clst[clst_id].rescnt;
+	writel_relaxed((readl_relaxed(rescnt) & ~mask) | magic, rescnt);
+
+	iounmap(p);
+}
+
+static void r8a7790_assert_reset(unsigned int cpu)
+{
+	void __iomem *p, *rescnt;
+	u32 mask, magic;
+	unsigned int clst_id = r8a7790_clst_id(cpu);
+
+	p = ioremap_nocache(RST, 0x63);
+
+	/* disable per-core clocks */
+	mask = BIT(3 - r8a7790_cpu_id(cpu));
+	magic = r8a7790_clst[clst_id].rescnt_magic;
+	rescnt = p + r8a7790_clst[clst_id].rescnt;
+	writel_relaxed((readl_relaxed(rescnt) | mask) | magic, rescnt);
+
+	iounmap(p);
+}
+
+static void r8a7790_power_on(unsigned int cpu)
+{
+	void __iomem *p, *cawupcr;
+
+	/* wake up CPU core via APMU */
+	p = ioremap_nocache(APMU, 0x3000);
+	cawupcr = p + (r8a7790_clst_id(cpu) ? CA7WUPCR : CA15WUPCR);
+	writel_relaxed(BIT(r8a7790_cpu_id(cpu)), cawupcr);
+
+	/* wait for APMU to finish */
+	while (readl_relaxed(cawupcr) != 0)
+		;
+
+	iounmap(p);
+}
+
+static void __init r8a7790_smp_prepare_cpus(unsigned int max_cpus)
+{
+	void __iomem *p;
+	unsigned int k;
+
+	shmobile_boot_fn = virt_to_phys(shmobile_invalidate_start);
+
+	/* MERAM for jump stub, because BAR requires 256KB aligned address */
+	p = ioremap_nocache(MERAM, 16);
+	memcpy_toio(p, shmobile_boot_vector, 16);
+	iounmap(p);
+
+	/* keep secondary CPU cores in reset */
+	for (k = 1; k < 8; k++)
+		r8a7790_assert_reset(k);
+}
+
+static int __cpuinit r8a7790_boot_secondary(unsigned int cpu,
+					    struct task_struct *idle)
+{
+	/* only allow a single cluster for now */
+	if (r8a7790_clst_id(cpu) != r8a7790_clst_id(0))
+		return -ENOTSUPP;
+
+	r8a7790_power_on(cpu);
+	r8a7790_deassert_reset(cpu);
+	return 0;
+}
+
+struct smp_operations r8a7790_smp_ops __initdata = {
+	.smp_prepare_cpus	= r8a7790_smp_prepare_cpus,
+	.smp_boot_secondary	= r8a7790_boot_secondary,
+};

[PATCH] ARM: shmobile: r8a7790 SMP prototype v5

[Shinya Kuribayashi]

On 7/18/2013 5:49 AM, Magnus Damm wrote:
>  Changes since v4:
>  - Rolled in the following patches from Kuribayashi-san, thanks!
>    ARM: shmobile: smp-r8a7790: remove the obsolete RESCNT register
>    ARM: shmobile: smp-r8a7790: rename s/carescnt/rescnt/g
>    ARM: shmobile: smp-r8a7790: rename s/cabar/bar/g as well
>    ARM: shmobile: smp-r8a7790: power up a CPU core at .smp_boot_secondary stage
> 
>  Written against kernel.org renesas git tag renesas-devel-20130717

Thanks for the update, looks good to me.
--
Shinya Kuribayashi
Renesas Electronics

[PATCH] ARM: shmobile: r8a7790 SMP prototype v5

[Simon Horman]

On Thu, Jul 18, 2013 at 09:32:26AM +0900, Shinya Kuribayashi wrote:
> On 7/18/2013 5:49 AM, Magnus Damm wrote:
> >  Changes since v4:
> >  - Rolled in the following patches from Kuribayashi-san, thanks!
> >    ARM: shmobile: smp-r8a7790: remove the obsolete RESCNT register
> >    ARM: shmobile: smp-r8a7790: rename s/carescnt/rescnt/g
> >    ARM: shmobile: smp-r8a7790: rename s/cabar/bar/g as well
> >    ARM: shmobile: smp-r8a7790: power up a CPU core at .smp_boot_secondary stage
> > 
> >  Written against kernel.org renesas git tag renesas-devel-20130717
> 
> Thanks for the update, looks good to me.

Thanks.

This seems like it is ready to queue-up.
Magnus, are there any outstanding reviews or issues?

[PATCH] ARM: shmobile: r8a7790 SMP prototype v5

[Simon Horman]

On Fri, Jul 19, 2013 at 11:29:04AM +0900, Simon Horman wrote:
> On Thu, Jul 18, 2013 at 09:32:26AM +0900, Shinya Kuribayashi wrote:
> > On 7/18/2013 5:49 AM, Magnus Damm wrote:
> > >  Changes since v4:
> > >  - Rolled in the following patches from Kuribayashi-san, thanks!
> > >    ARM: shmobile: smp-r8a7790: remove the obsolete RESCNT register
> > >    ARM: shmobile: smp-r8a7790: rename s/carescnt/rescnt/g
> > >    ARM: shmobile: smp-r8a7790: rename s/cabar/bar/g as well
> > >    ARM: shmobile: smp-r8a7790: power up a CPU core at .smp_boot_secondary stage
> > > 
> > >  Written against kernel.org renesas git tag renesas-devel-20130717
> > 
> > Thanks for the update, looks good to me.
> 
> Thanks.
> 
> This seems like it is ready to queue-up.
> Magnus, are there any outstanding reviews or issues?

I have placed this in topic/lager-smp-v5 in the renesas tree
to aid review. It is not yet queued up for mainline.

topic/lager-smp still contains v3 of this patch.
Let me know if I can delete this branch.