3 files changed, 63 insertions, 7 deletions

diff --git a/arch/arm/mach-shmobile/smp-emev2.c b/arch/arm/mach-shmobile/smp-emev2.c
index 535426c306b..f6745628628 100644
--- a/arch/arm/mach-shmobile/smp-emev2.c
+++ b/arch/arm/mach-shmobile/smp-emev2.c
@@ -32,8 +32,24 @@
 
 #define EMEV2_SCU_BASE 0x1e000000
 
+static DEFINE_SPINLOCK(scu_lock);
 static void __iomem *scu_base;
 
+static void modify_scu_cpu_psr(unsigned long set, unsigned long clr)
+{
+	unsigned long tmp;
+
+	/* we assume this code is running on a different cpu
+	 * than the one that is changing coherency setting */
+	spin_lock(&scu_lock);
+	tmp = readl(scu_base + 8);
+	tmp &= ~clr;
+	tmp |= set;
+	writel(tmp, scu_base + 8);
+	spin_unlock(&scu_lock);
+
+}
+
 static unsigned int __init emev2_get_core_count(void)
 {
 	if (!scu_base) {
@@ -79,7 +95,7 @@ static int __cpuinit emev2_boot_secondary(unsigned int cpu, struct task_struct *
 	cpu = cpu_logical_map(cpu);
 
 	/* enable cache coherency */
-	scu_power_mode(scu_base, 0);
+	modify_scu_cpu_psr(0, 3 << (cpu * 8));
 
 	/* Tell ROM loader about our vector (in headsmp.S) */
 	emev2_set_boot_vector(__pa(shmobile_secondary_vector));
@@ -90,10 +106,12 @@ static int __cpuinit emev2_boot_secondary(unsigned int cpu, struct task_struct *
 
 static void __init emev2_smp_prepare_cpus(unsigned int max_cpus)
 {
+	int cpu = cpu_logical_map(0);
+
 	scu_enable(scu_base);
 
 	/* enable cache coherency on CPU0 */
-	scu_power_mode(scu_base, 0);
+	modify_scu_cpu_psr(0, 3 << (cpu * 8));
 }
 
 static void __init emev2_smp_init_cpus(void)
diff --git a/arch/arm/mach-shmobile/smp-r8a7779.c b/arch/arm/mach-shmobile/smp-r8a7779.c
index 9def0f22bf2..2ce6af9a6a3 100644
--- a/arch/arm/mach-shmobile/smp-r8a7779.c
+++ b/arch/arm/mach-shmobile/smp-r8a7779.c
@@ -61,6 +61,9 @@ static void __iomem *scu_base_addr(void)
 	return (void __iomem *)0xf0000000;
 }
 
+static DEFINE_SPINLOCK(scu_lock);
+static unsigned long tmp;
+
 #ifdef CONFIG_HAVE_ARM_TWD
 static DEFINE_TWD_LOCAL_TIMER(twd_local_timer, 0xf0000600, 29);
 
@@ -70,6 +73,20 @@ void __init r8a7779_register_twd(void)
 }
 #endif
 
+static void modify_scu_cpu_psr(unsigned long set, unsigned long clr)
+{
+	void __iomem *scu_base = scu_base_addr();
+
+	spin_lock(&scu_lock);
+	tmp = __raw_readl(scu_base + 8);
+	tmp &= ~clr;
+	tmp |= set;
+	spin_unlock(&scu_lock);
+
+	/* disable cache coherency after releasing the lock */
+	__raw_writel(tmp, scu_base + 8);
+}
+
 static unsigned int __init r8a7779_get_core_count(void)
 {
 	void __iomem *scu_base = scu_base_addr();
@@ -85,7 +102,7 @@ static int r8a7779_platform_cpu_kill(unsigned int cpu)
 	cpu = cpu_logical_map(cpu);
 
 	/* disable cache coherency */
-	scu_power_mode(scu_base_addr(), 3);
+	modify_scu_cpu_psr(3 << (cpu * 8), 0);
 
 	if (cpu < ARRAY_SIZE(r8a7779_ch_cpu))
 		ch = r8a7779_ch_cpu[cpu];
@@ -128,7 +145,7 @@ static int __cpuinit r8a7779_boot_secondary(unsigned int cpu, struct task_struct
 	cpu = cpu_logical_map(cpu);
 
 	/* enable cache coherency */
-	scu_power_mode(scu_base_addr(), 0);
+	modify_scu_cpu_psr(0, 3 << (cpu * 8));
 
 	if (cpu < ARRAY_SIZE(r8a7779_ch_cpu))
 		ch = r8a7779_ch_cpu[cpu];
@@ -141,13 +158,15 @@ static int __cpuinit r8a7779_boot_secondary(unsigned int cpu, struct task_struct
 
 static void __init r8a7779_smp_prepare_cpus(unsigned int max_cpus)
 {
+	int cpu = cpu_logical_map(0);
+
 	scu_enable(scu_base_addr());
 
 	/* Map the reset vector (in headsmp.S) */
 	__raw_writel(__pa(shmobile_secondary_vector), AVECR);
 
 	/* enable cache coherency on CPU0 */
-	scu_power_mode(scu_base_addr(), 0);
+	modify_scu_cpu_psr(0, 3 << (cpu * 8));
 
 	r8a7779_pm_init();
 
diff --git a/arch/arm/mach-shmobile/smp-sh73a0.c b/arch/arm/mach-shmobile/smp-sh73a0.c
index 96ddb97babb..624f00f70ab 100644
--- a/arch/arm/mach-shmobile/smp-sh73a0.c
+++ b/arch/arm/mach-shmobile/smp-sh73a0.c
@@ -41,6 +41,9 @@ static void __iomem *scu_base_addr(void)
 	return (void __iomem *)0xf0000000;
 }
 
+static DEFINE_SPINLOCK(scu_lock);
+static unsigned long tmp;
+
 #ifdef CONFIG_HAVE_ARM_TWD
 static DEFINE_TWD_LOCAL_TIMER(twd_local_timer, 0xf0000600, 29);
 void __init sh73a0_register_twd(void)
@@ -49,6 +52,20 @@ void __init sh73a0_register_twd(void)
 }
 #endif
 
+static void modify_scu_cpu_psr(unsigned long set, unsigned long clr)
+{
+	void __iomem *scu_base = scu_base_addr();
+
+	spin_lock(&scu_lock);
+	tmp = __raw_readl(scu_base + 8);
+	tmp &= ~clr;
+	tmp |= set;
+	spin_unlock(&scu_lock);
+
+	/* disable cache coherency after releasing the lock */
+	__raw_writel(tmp, scu_base + 8);
+}
+
 static unsigned int __init sh73a0_get_core_count(void)
 {
 	void __iomem *scu_base = scu_base_addr();
@@ -66,7 +83,7 @@ static int __cpuinit sh73a0_boot_secondary(unsigned int cpu, struct task_struct
 	cpu = cpu_logical_map(cpu);
 
 	/* enable cache coherency */
-	scu_power_mode(scu_base_addr(), 0);
+	modify_scu_cpu_psr(0, 3 << (cpu * 8));
 
 	if (((__raw_readl(PSTR) >> (4 * cpu)) & 3) == 3)
 		__raw_writel(1 << cpu, WUPCR);	/* wake up */
@@ -78,6 +95,8 @@ static int __cpuinit sh73a0_boot_secondary(unsigned int cpu, struct task_struct
 
 static void __init sh73a0_smp_prepare_cpus(unsigned int max_cpus)
 {
+	int cpu = cpu_logical_map(0);
+
 	scu_enable(scu_base_addr());
 
 	/* Map the reset vector (in headsmp.S) */
@@ -85,7 +104,7 @@ static void __init sh73a0_smp_prepare_cpus(unsigned int max_cpus)
 	__raw_writel(__pa(shmobile_secondary_vector), SBAR);
 
 	/* enable cache coherency on CPU0 */
-	scu_power_mode(scu_base_addr(), 0);
+	modify_scu_cpu_psr(0, 3 << (cpu * 8));
 }
 
 static void __init sh73a0_smp_init_cpus(void)