On Tue, Mar 11, 2014 at 06:48:13PM -0500, Scott Wood wrote:

On Fri, 2014-03-07 at 12:58 +0800, Chenhui Zhao wrote:

quoted

diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index ac2621a..f3f4401 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c

@@ -405,8 +405,12 @@ void generic_cpu_die(unsigned int cpu)
 
 	for (i = 0; i < 100; i++) {
 		smp_rmb();
-		if (per_cpu(cpu_state, cpu) == CPU_DEAD)
+		if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
+#ifdef CONFIG_PPC64
+			paca[cpu].cpu_start = 0;
+#endif

Why wasn't this needed by previous ppc64 machines?

If not clear, cpu can't start in the case of cpu hotpolug.
The function pseries_cpu_die() in arch/powerpc/platforms/pseries/hotplug-cpu.c
also clears the flag.

quoted

diff --git a/arch/powerpc/platforms/85xx/smp.c b/arch/powerpc/platforms/85xx/smp.c
index 2e5911e..0047883 100644
--- a/arch/powerpc/platforms/85xx/smp.c
+++ b/arch/powerpc/platforms/85xx/smp.c

@@ -19,6 +19,7 @@
 #include <linux/kexec.h>
 #include <linux/highmem.h>
 #include <linux/cpu.h>
+#include <linux/smp.h>
 
 #include <asm/machdep.h>
 #include <asm/pgtable.h>

@@ -46,6 +47,17 @@ static u64 timebase;
 static int tb_req;
 static int tb_valid;
 
+#ifdef CONFIG_PPC_E500MC
+/* specify the cpu PM state when cpu dies, PH15/NAP is the default */
+int qoriq_cpu_die_state = E500_PM_PH15;
+#endif

static?  Is there any way to modify this other than modifying source
code?

BTW, QorIQ doesn't imply an e500mc derivative.

Will support e500, but for now these code support e500mc derivative in
advance.

Supposed qoriq_cpu_die_state can be changed by platform init code
if the default value is not proper for a specific platform.

quoted

@@ -125,6 +138,34 @@ static void mpc85xx_take_timebase(void)
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
+#ifdef CONFIG_PPC_E500MC
+static void qoriq_cpu_die(void)
+{
+	unsigned int cpu = smp_processor_id();
+
+	local_irq_disable();
+#ifdef CONFIG_PPC64
+	__hard_irq_disable();
+#endif

Why this instead of one call to hard_irq_disable() (no leading
underscores)?

-Scott

hard_irq_disable() will clear soft_enabled again. local_irq_disable()
has cleared it.

Will use hard_irq_disable() to replace these lines.

  local_irq_disable();
  #ifdef CONFIG_PPC64
  	__hard_irq_disable();
  #endif

-Chenhui