[PATCHES] powerpc: Cleanup and fix CPU hotplug

[PATCHES] powerpc: Cleanup and fix CPU hotplug

[Benjamin Herrenschmidt]

This series contain a number of cleanups and fixes to our CPU hotplug
code, including fixing crashes on G5 machines, properly re-initializing
idle threads etc...

Tested on PowerMac G5 and pSeries machines.

[PATCH 01/18] powerpc: Make decrementer interrupt robust against offlined CPUs

[Benjamin Herrenschmidt]

With some implementations, it is possible that a timer interrupt
occurs every few seconds on an offline CPU. In this case, just
re-arm the decrementer and return immediately

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/kernel/time.c |   15 +++++++++++----
 1 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index 09d31db..ecebc7c 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -577,14 +577,21 @@ void timer_interrupt(struct pt_regs * regs)
 	struct clock_event_device *evt = &decrementer->event;
 	u64 now;
 
+	/* Ensure a positive value is written to the decrementer, or else
+	 * some CPUs will continuue to take decrementer exceptions
+	 */
+	set_dec(DECREMENTER_MAX);
+
+	/* Some implementations of hotplug will get timer interrupts while
+	 * offline, just ignore these
+	 */
+	if (!cpu_online(smp_processor_id()))
+		return;
+
 	trace_timer_interrupt_entry(regs);
 
 	__get_cpu_var(irq_stat).timer_irqs++;
 
-	/* Ensure a positive value is written to the decrementer, or else
-	 * some CPUs will continuue to take decrementer exceptions */
-	set_dec(DECREMENTER_MAX);
-
 #if defined(CONFIG_PPC32) && defined(CONFIG_PMAC)
 	if (atomic_read(&ppc_n_lost_interrupts) != 0)
 		do_IRQ(regs);
-- 
1.7.1

[PATCH 02/18] powerpc/smp: soft-replugged CPUs must go back to start_secondary

[Benjamin Herrenschmidt]

Various thing are torn down when a CPU is hot-unplugged. That CPU
is expected to go back to start_secondary when re-plugged to re
initialize everything, such as clock sources, maps, ...

Some implementations just return from cpu_die() callback
in the idle loop when the CPU is "re-plugged". This is not enough.

We fix it using a little asm trampoline which resets the stack
and calls back into start_secondary as if we were all fresh from
boot. The trampoline already existed on ppc64, but we add it for
ppc32

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/smp.h                  |    1 +
 arch/powerpc/kernel/head_32.S                   |    9 +++++++++
 arch/powerpc/kernel/smp.c                       |    9 +++++++--
 arch/powerpc/platforms/pseries/offline_states.h |    2 --
 4 files changed, 17 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
index 66e237b..1de0e97 100644
--- a/arch/powerpc/include/asm/smp.h
+++ b/arch/powerpc/include/asm/smp.h
@@ -36,6 +36,7 @@ extern void cpu_die(void);
 
 extern void smp_send_debugger_break(int cpu);
 extern void smp_message_recv(int);
+extern void start_secondary_resume(void);
 
 DECLARE_PER_CPU(unsigned int, cpu_pvr);
 
diff --git a/arch/powerpc/kernel/head_32.S b/arch/powerpc/kernel/head_32.S
index 98c4b29..c5c24be 100644
--- a/arch/powerpc/kernel/head_32.S
+++ b/arch/powerpc/kernel/head_32.S
@@ -890,6 +890,15 @@ __secondary_start:
 	mtspr	SPRN_SRR1,r4
 	SYNC
 	RFI
+
+_GLOBAL(start_secondary_resume)
+	/* Reset stack */
+	rlwinm	r1,r1,0,0,(31-THREAD_SHIFT)	/* current_thread_info() */
+	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
+	li	r3,0
+	std	r3,0(r1)		/* Zero the stack frame pointer	*/
+	bl	start_secondary
+	b	.
 #endif /* CONFIG_SMP */
 
 #ifdef CONFIG_KVM_BOOK3S_HANDLER
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 9813605..1c9956c 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -502,7 +502,7 @@ static struct device_node *cpu_to_l2cache(int cpu)
 }
 
 /* Activate a secondary processor. */
-int __devinit start_secondary(void *unused)
+void __devinit start_secondary(void *unused)
 {
 	unsigned int cpu = smp_processor_id();
 	struct device_node *l2_cache;
@@ -558,7 +558,8 @@ int __devinit start_secondary(void *unused)
 	local_irq_enable();
 
 	cpu_idle();
-	return 0;
+
+	BUG();
 }
 
 int setup_profiling_timer(unsigned int multiplier)
@@ -660,5 +661,9 @@ void cpu_die(void)
 {
 	if (ppc_md.cpu_die)
 		ppc_md.cpu_die();
+
+	/* If we return, we re-enter start_secondary */
+	start_secondary_resume();
 }
+
 #endif
diff --git a/arch/powerpc/platforms/pseries/offline_states.h b/arch/powerpc/platforms/pseries/offline_states.h
index 75a6f48..08672d9 100644
--- a/arch/powerpc/platforms/pseries/offline_states.h
+++ b/arch/powerpc/platforms/pseries/offline_states.h
@@ -34,6 +34,4 @@ static inline void set_default_offline_state(int cpu)
 #endif
 
 extern enum cpu_state_vals get_preferred_offline_state(int cpu);
-extern int start_secondary(void);
-extern void start_secondary_resume(void);
 #endif
-- 
1.7.1

[PATCH 03/18] powerpc/smp: Fix generic_mach_cpu_die()

[Benjamin Herrenschmidt]

This is used by some "soft" hotplug implementations. I needs to
call idle_task_exit() when the CPU is going away, and we remove
the now no-longer needed set_cpu_online() and local_irq_enable()
which are handled by the return to start_secondary

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/kernel/smp.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 1c9956c..3c0fab5 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -362,14 +362,13 @@ void generic_mach_cpu_die(void)
 	unsigned int cpu;
 
 	local_irq_disable();
+	idle_task_exit();
 	cpu = smp_processor_id();
 	printk(KERN_DEBUG "CPU%d offline\n", cpu);
 	__get_cpu_var(cpu_state) = CPU_DEAD;
 	smp_wmb();
 	while (__get_cpu_var(cpu_state) != CPU_UP_PREPARE)
 		cpu_relax();
-	set_cpu_online(cpu, true);
-	local_irq_enable();
 }
 #endif
 
-- 
1.7.1

[PATCH 04/18] powerpc/smp: Remove unused generic_cpu_enable()

[Benjamin Herrenschmidt]

Nobody uses it, besides we should always use the normal __cpu_up
path anyways

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/smp.h |    1 -
 arch/powerpc/kernel/smp.c      |   22 ----------------------
 2 files changed, 0 insertions(+), 23 deletions(-)

diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
index 1de0e97..a629b6f 100644
--- a/arch/powerpc/include/asm/smp.h
+++ b/arch/powerpc/include/asm/smp.h
@@ -43,7 +43,6 @@ DECLARE_PER_CPU(unsigned int, cpu_pvr);
 #ifdef CONFIG_HOTPLUG_CPU
 extern void fixup_irqs(const struct cpumask *map);
 int generic_cpu_disable(void);
-int generic_cpu_enable(unsigned int cpu);
 void generic_cpu_die(unsigned int cpu);
 void generic_mach_cpu_die(void);
 #endif
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 3c0fab5..19d0c25 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -322,28 +322,6 @@ int generic_cpu_disable(void)
 	return 0;
 }
 
-int generic_cpu_enable(unsigned int cpu)
-{
-	/* Do the normal bootup if we haven't
-	 * already bootstrapped. */
-	if (system_state != SYSTEM_RUNNING)
-		return -ENOSYS;
-
-	/* get the target out of it's holding state */
-	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
-	smp_wmb();
-
-	while (!cpu_online(cpu))
-		cpu_relax();
-
-#ifdef CONFIG_PPC64
-	fixup_irqs(cpu_online_mask);
-	/* counter the irq disable in fixup_irqs */
-	local_irq_enable();
-#endif
-	return 0;
-}
-
 void generic_cpu_die(unsigned int cpu)
 {
 	int i;
-- 
1.7.1

[PATCH 05/18] powerpc/smp: Remove unused smp_ops->cpu_enable()

[Benjamin Herrenschmidt]

Remove the last remnants of cpu_enable(), everybody uses the normal
__cpu_up() path now

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/machdep.h    |    1 -
 arch/powerpc/kernel/smp.c             |   10 ----------
 arch/powerpc/platforms/powermac/smp.c |    2 --
 3 files changed, 0 insertions(+), 13 deletions(-)

diff --git a/arch/powerpc/include/asm/machdep.h b/arch/powerpc/include/asm/machdep.h
index fe56a23..bcfc0da 100644
--- a/arch/powerpc/include/asm/machdep.h
+++ b/arch/powerpc/include/asm/machdep.h
@@ -37,7 +37,6 @@ struct smp_ops_t {
 	void  (*setup_cpu)(int nr);
 	void  (*take_timebase)(void);
 	void  (*give_timebase)(void);
-	int   (*cpu_enable)(unsigned int nr);
 	int   (*cpu_disable)(void);
 	void  (*cpu_die)(unsigned int nr);
 	int   (*cpu_bootable)(unsigned int nr);
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 19d0c25..be7d728 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -350,21 +350,11 @@ void generic_mach_cpu_die(void)
 }
 #endif
 
-static int __devinit cpu_enable(unsigned int cpu)
-{
-	if (smp_ops && smp_ops->cpu_enable)
-		return smp_ops->cpu_enable(cpu);
-
-	return -ENOSYS;
-}
-
 int __cpuinit __cpu_up(unsigned int cpu)
 {
 	int c;
 
 	secondary_ti = current_set[cpu];
-	if (!cpu_enable(cpu))
-		return 0;
 
 	if (smp_ops == NULL ||
 	    (smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu)))
diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index c95215f..ebd2b7e 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -923,8 +923,6 @@ struct smp_ops_t core99_smp_ops = {
 # if defined(CONFIG_PPC64)
 	.cpu_disable	= generic_cpu_disable,
 	.cpu_die	= generic_cpu_die,
-	/* intentionally do *NOT* assign cpu_enable,
-	 * the generic code will use kick_cpu then! */
 # endif
 #endif
 };
-- 
1.7.1

[PATCH 06/18] powerpc/pmac/smp: Fix 32-bit PowerMac cpu_die

[Benjamin Herrenschmidt]

Use generic cpu_state, call idle_task_exit() properly, and
remove smp_core99_cpu_die() which isn't useful, the generic
function does the job just fine.
---
 arch/powerpc/include/asm/smp.h        |    1 +
 arch/powerpc/platforms/powermac/smp.c |   25 +++++--------------------
 2 files changed, 6 insertions(+), 20 deletions(-)

diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
index a629b6f..9fe5597 100644
--- a/arch/powerpc/include/asm/smp.h
+++ b/arch/powerpc/include/asm/smp.h
@@ -45,6 +45,7 @@ extern void fixup_irqs(const struct cpumask *map);
 int generic_cpu_disable(void);
 void generic_cpu_die(unsigned int cpu);
 void generic_mach_cpu_die(void);
+DECLARE_PER_CPU(int, cpu_state);
 #endif
 
 #ifdef CONFIG_PPC64
diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index ebd2b7e..e0ac7bb 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -880,31 +880,17 @@ int smp_core99_cpu_disable(void)
 	return 0;
 }
 
-static int cpu_dead[NR_CPUS];
-
 void pmac32_cpu_die(void)
 {
 	local_irq_disable();
-	cpu_dead[smp_processor_id()] = 1;
+	idle_task_exit();
+	printk(KERN_DEBUG "CPU%d offline\n", smp_processor_id());
+	__get_cpu_var(cpu_state) = CPU_DEAD;
+	smp_wmb();
 	mb();
 	low_cpu_die();
 }
 
-void smp_core99_cpu_die(unsigned int cpu)
-{
-	int timeout;
-
-	timeout = 1000;
-	while (!cpu_dead[cpu]) {
-		if (--timeout == 0) {
-			printk("CPU %u refused to die!\n", cpu);
-			break;
-		}
-		msleep(1);
-	}
-	cpu_dead[cpu] = 0;
-}
-
 #endif /* CONFIG_HOTPLUG_CPU && CONFIG_PP32 */
 
 /* Core99 Macs (dual G4s and G5s) */
@@ -918,12 +904,11 @@ struct smp_ops_t core99_smp_ops = {
 #if defined(CONFIG_HOTPLUG_CPU)
 # if defined(CONFIG_PPC32)
 	.cpu_disable	= smp_core99_cpu_disable,
-	.cpu_die	= smp_core99_cpu_die,
 # endif
 # if defined(CONFIG_PPC64)
 	.cpu_disable	= generic_cpu_disable,
-	.cpu_die	= generic_cpu_die,
 # endif
+	.cpu_die	= generic_cpu_die,
 #endif
 };
 
-- 
1.7.1

[PATCH 07/18] powerpc/pmac/smp: Rename fixup_irqs() to migrate_irqs() and use it on ppc32

[Benjamin Herrenschmidt]

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/smp.h |    2 +-
 arch/powerpc/kernel/irq.c      |    3 ++-
 arch/powerpc/kernel/smp.c      |    2 +-
 3 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
index 9fe5597..7e99771 100644
--- a/arch/powerpc/include/asm/smp.h
+++ b/arch/powerpc/include/asm/smp.h
@@ -41,7 +41,7 @@ extern void start_secondary_resume(void);
 DECLARE_PER_CPU(unsigned int, cpu_pvr);
 
 #ifdef CONFIG_HOTPLUG_CPU
-extern void fixup_irqs(const struct cpumask *map);
+extern void migrate_irqs(void);
 int generic_cpu_disable(void);
 void generic_cpu_die(unsigned int cpu);
 void generic_mach_cpu_die(void);
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index ce557f6..cab8654 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -303,12 +303,13 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
-void fixup_irqs(const struct cpumask *map)
+void migrate_irqs(void)
 {
 	struct irq_desc *desc;
 	unsigned int irq;
 	static int warned;
 	cpumask_var_t mask;
+	const struct cpumask *map = cpu_online_mask;
 
 	alloc_cpumask_var(&mask, GFP_KERNEL);
 
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index be7d728..f6cc5c1 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -317,8 +317,8 @@ int generic_cpu_disable(void)
 	set_cpu_online(cpu, false);
 #ifdef CONFIG_PPC64
 	vdso_data->processorCount--;
-	fixup_irqs(cpu_online_mask);
 #endif
+	migrate_irqs();
 	return 0;
 }
 
-- 
1.7.1

[PATCH 08/18] powerpc/pmac/smp: Fixup smp_core99_cpu_disable() and use it on 64-bit

[Benjamin Herrenschmidt]

Use the generic code, just add the MPIC priority setting,

I don't see any use in mucking around with the decrementer,
as 32-bit will have EE off all along, and 64-bit will be able
to deal with it.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/platforms/powermac/smp.c |   17 +++++------------
 1 files changed, 5 insertions(+), 12 deletions(-)

diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index e0ac7bb..eda4709 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -867,16 +867,14 @@ static void __devinit smp_core99_setup_cpu(int cpu_nr)
 
 #if defined(CONFIG_HOTPLUG_CPU) && defined(CONFIG_PPC32)
 
-int smp_core99_cpu_disable(void)
+static int smp_core99_cpu_disable(void)
 {
-	set_cpu_online(smp_processor_id(), false);
+	int rc = generic_cpu_disable();
+	if (rc)
+		return rc;
 
-	/* XXX reset cpu affinity here */
 	mpic_cpu_set_priority(0xf);
-	asm volatile("mtdec %0" : : "r" (0x7fffffff));
-	mb();
-	udelay(20);
-	asm volatile("mtdec %0" : : "r" (0x7fffffff));
+
 	return 0;
 }
 
@@ -902,12 +900,7 @@ struct smp_ops_t core99_smp_ops = {
 	.give_timebase	= smp_core99_give_timebase,
 	.take_timebase	= smp_core99_take_timebase,
 #if defined(CONFIG_HOTPLUG_CPU)
-# if defined(CONFIG_PPC32)
 	.cpu_disable	= smp_core99_cpu_disable,
-# endif
-# if defined(CONFIG_PPC64)
-	.cpu_disable	= generic_cpu_disable,
-# endif
 	.cpu_die	= generic_cpu_die,
 #endif
 };
-- 
1.7.1

[PATCH 09/18] powerpc/pmac/smp: Consolidate 32-bit and 64-bit PowerMac cpu_die in one file

[Benjamin Herrenschmidt]

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/platforms/powermac/pmac.h  |    1 -
 arch/powerpc/platforms/powermac/setup.c |   56 -----------------------------
 arch/powerpc/platforms/powermac/smp.c   |   58 +++++++++++++++++++++++++++++--
 3 files changed, 55 insertions(+), 60 deletions(-)

diff --git a/arch/powerpc/platforms/powermac/pmac.h b/arch/powerpc/platforms/powermac/pmac.h
index f0bc08f..20468f4 100644
--- a/arch/powerpc/platforms/powermac/pmac.h
+++ b/arch/powerpc/platforms/powermac/pmac.h
@@ -33,7 +33,6 @@ extern void pmac_setup_pci_dma(void);
 extern void pmac_check_ht_link(void);
 
 extern void pmac_setup_smp(void);
-extern void pmac32_cpu_die(void);
 extern void low_cpu_die(void) __attribute__((noreturn));
 
 extern int pmac_nvram_init(void);
diff --git a/arch/powerpc/platforms/powermac/setup.c b/arch/powerpc/platforms/powermac/setup.c
index d5aceb7..aa45281 100644
--- a/arch/powerpc/platforms/powermac/setup.c
+++ b/arch/powerpc/platforms/powermac/setup.c
@@ -650,51 +650,6 @@ static int pmac_pci_probe_mode(struct pci_bus *bus)
 		return PCI_PROBE_NORMAL;
 	return PCI_PROBE_DEVTREE;
 }
-
-#ifdef CONFIG_HOTPLUG_CPU
-/* access per cpu vars from generic smp.c */
-DECLARE_PER_CPU(int, cpu_state);
-
-static void pmac64_cpu_die(void)
-{
-	/*
-	 * turn off as much as possible, we'll be
-	 * kicked out as this will only be invoked
-	 * on core99 platforms for now ...
-	 */
-
-	printk(KERN_INFO "CPU#%d offline\n", smp_processor_id());
-	__get_cpu_var(cpu_state) = CPU_DEAD;
-	smp_wmb();
-
-	/*
-	 * during the path that leads here preemption is disabled,
-	 * reenable it now so that when coming up preempt count is
-	 * zero correctly
-	 */
-	preempt_enable();
-
-	/*
-	 * hard-disable interrupts for the non-NAP case, the NAP code
-	 * needs to re-enable interrupts (but soft-disables them)
-	 */
-	hard_irq_disable();
-
-	while (1) {
-		/* let's not take timer interrupts too often ... */
-		set_dec(0x7fffffff);
-
-		/* should always be true at this point */
-		if (cpu_has_feature(CPU_FTR_CAN_NAP))
-			power4_cpu_offline_powersave();
-		else {
-			HMT_low();
-			HMT_very_low();
-		}
-	}
-}
-#endif /* CONFIG_HOTPLUG_CPU */
-
 #endif /* CONFIG_PPC64 */
 
 define_machine(powermac) {
@@ -726,15 +681,4 @@ define_machine(powermac) {
 	.pcibios_after_init	= pmac_pcibios_after_init,
 	.phys_mem_access_prot	= pci_phys_mem_access_prot,
 #endif
-#ifdef CONFIG_HOTPLUG_CPU
-#ifdef CONFIG_PPC64
-	.cpu_die		= pmac64_cpu_die,
-#endif
-#ifdef CONFIG_PPC32
-	.cpu_die		= pmac32_cpu_die,
-#endif
-#endif
-#if defined(CONFIG_HOTPLUG_CPU) && defined(CONFIG_PPC32)
-	.cpu_die		= generic_mach_cpu_die,
-#endif
 };
diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index eda4709..50e15b4 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -865,7 +865,7 @@ static void __devinit smp_core99_setup_cpu(int cpu_nr)
 }
 
 
-#if defined(CONFIG_HOTPLUG_CPU) && defined(CONFIG_PPC32)
+#ifdef CONFIG_HOTPLUG_CPU
 
 static int smp_core99_cpu_disable(void)
 {
@@ -878,7 +878,9 @@ static int smp_core99_cpu_disable(void)
 	return 0;
 }
 
-void pmac32_cpu_die(void)
+#ifdef CONFIG_PPC32
+
+static void pmac_cpu_die(void)
 {
 	local_irq_disable();
 	idle_task_exit();
@@ -889,7 +891,52 @@ void pmac32_cpu_die(void)
 	low_cpu_die();
 }
 
-#endif /* CONFIG_HOTPLUG_CPU && CONFIG_PP32 */
+#else /* CONFIG_PPC32 */
+
+static void pmac_cpu_die(void)
+{
+	local_irq_disable();
+	idle_task_exit();
+
+	/*
+	 * turn off as much as possible, we'll be
+	 * kicked out as this will only be invoked
+	 * on core99 platforms for now ...
+	 */
+
+	printk(KERN_INFO "CPU#%d offline\n", smp_processor_id());
+	__get_cpu_var(cpu_state) = CPU_DEAD;
+	smp_wmb();
+
+	/*
+	 * during the path that leads here preemption is disabled,
+	 * reenable it now so that when coming up preempt count is
+	 * zero correctly
+	 */
+	preempt_enable();
+
+	/*
+	 * hard-disable interrupts for the non-NAP case, the NAP code
+	 * needs to re-enable interrupts (but soft-disables them)
+	 */
+	hard_irq_disable();
+
+	while (1) {
+		/* let's not take timer interrupts too often ... */
+		set_dec(0x7fffffff);
+
+		/* should always be true at this point */
+		if (cpu_has_feature(CPU_FTR_CAN_NAP))
+			power4_cpu_offline_powersave();
+		else {
+			HMT_low();
+			HMT_very_low();
+		}
+	}
+}
+
+#endif /* else CONFIG_PPC32 */
+#endif /* CONFIG_HOTPLUG_CPU */
 
 /* Core99 Macs (dual G4s and G5s) */
 struct smp_ops_t core99_smp_ops = {
@@ -933,5 +980,10 @@ void __init pmac_setup_smp(void)
 		smp_ops = &psurge_smp_ops;
 	}
 #endif /* CONFIG_PPC32 */
+
+#ifdef CONFIG_HOTPLUG_CPU
+	ppc_md.cpu_die = pmac_cpu_die;
+#endif
 }
 
+
-- 
1.7.1

[PATCH 10/18] powerpc/pmac/smp: Remove HMT changes for PowerMac offline code

[Benjamin Herrenschmidt]

Those instructions do nothing on non-threaded processors such
as 970's used on those machines.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/platforms/powermac/smp.c |    4 ----
 1 files changed, 0 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index 50e15b4..53bee66 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -928,10 +928,6 @@ static void pmac_cpu_die(void)
 		/* should always be true at this point */
 		if (cpu_has_feature(CPU_FTR_CAN_NAP))
 			power4_cpu_offline_powersave();
-		else {
-			HMT_low();
-			HMT_very_low();
-		}
 	}
 }
 
-- 
1.7.1

[PATCH 11/18] powerpc/pmac/smp: Properly NAP offlined CPU on G5

[Benjamin Herrenschmidt]

The current code soft-disables, and then goes to NAP mode which
turns interrupts on. That means that if an interrupt occurs, we
will hit the masked interrupt code path which isn't what we want,
as it will return with EE off, which will either get us out of
NAP mode, or fail to enter it (according to spec).

Instead, let's just rely on the fact that it is safe to take
decrementer interrupts on an offline CPU and leave interrupts
enabled. We can also get rid of the special case in asm for
power4_cpu_offline_powersave() and just use power4_idle().

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/machdep.h    |    1 -
 arch/powerpc/kernel/head_64.S         |    7 +++++++
 arch/powerpc/kernel/idle_power4.S     |   21 ---------------------
 arch/powerpc/platforms/powermac/smp.c |   14 ++++++++------
 4 files changed, 15 insertions(+), 28 deletions(-)

diff --git a/arch/powerpc/include/asm/machdep.h b/arch/powerpc/include/asm/machdep.h
index bcfc0da..578d330 100644
--- a/arch/powerpc/include/asm/machdep.h
+++ b/arch/powerpc/include/asm/machdep.h
@@ -266,7 +266,6 @@ struct machdep_calls {
 
 extern void e500_idle(void);
 extern void power4_idle(void);
-extern void power4_cpu_offline_powersave(void);
 extern void ppc6xx_idle(void);
 extern void book3e_idle(void);
 
diff --git a/arch/powerpc/kernel/head_64.S b/arch/powerpc/kernel/head_64.S
index 782f23d..271140b 100644
--- a/arch/powerpc/kernel/head_64.S
+++ b/arch/powerpc/kernel/head_64.S
@@ -536,6 +536,13 @@ _GLOBAL(pmac_secondary_start)
 	add	r13,r13,r4		/* for this processor.		*/
 	mtspr	SPRN_SPRG_PACA,r13	/* Save vaddr of paca in an SPRG*/
 
+	/* Mark interrupts soft and hard disabled (they might be enabled
+	 * in the PACA when doing hotplug)
+	 */
+	li	r0,0
+	stb	r0,PACASOFTIRQEN(r13)
+	stb	r0,PACAHARDIRQEN(r13)
+
 	/* Create a temp kernel stack for use before relocation is on.	*/
 	ld	r1,PACAEMERGSP(r13)
 	subi	r1,r1,STACK_FRAME_OVERHEAD
diff --git a/arch/powerpc/kernel/idle_power4.S b/arch/powerpc/kernel/idle_power4.S
index 5328709..ba31954 100644
--- a/arch/powerpc/kernel/idle_power4.S
+++ b/arch/powerpc/kernel/idle_power4.S
@@ -53,24 +53,3 @@ END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
 	isync
 	b	1b
 
-_GLOBAL(power4_cpu_offline_powersave)
-	/* Go to NAP now */
-	mfmsr	r7
-	rldicl	r0,r7,48,1
-	rotldi	r0,r0,16
-	mtmsrd	r0,1			/* hard-disable interrupts */
-	li	r0,1
-	li	r6,0
-	stb	r0,PACAHARDIRQEN(r13)	/* we'll hard-enable shortly */
-	stb	r6,PACASOFTIRQEN(r13)	/* soft-disable irqs */
-BEGIN_FTR_SECTION
-	DSSALL
-	sync
-END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
-	ori	r7,r7,MSR_EE
-	oris	r7,r7,MSR_POW@h
-	sync
-	isync
-	mtmsrd	r7
-	isync
-	blr
diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index 53bee66..837989e 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -916,18 +916,20 @@ static void pmac_cpu_die(void)
 	preempt_enable();
 
 	/*
-	 * hard-disable interrupts for the non-NAP case, the NAP code
-	 * needs to re-enable interrupts (but soft-disables them)
+	 * Re-enable interrupts. The NAP code needs to enable them
+	 * anyways, do it now so we deal with the case where one already
+	 * happened while soft-disabled.
+	 * We shouldn't get any external interrupts, only decrementer, and the
+	 * decrementer handler is safe for use on offline CPUs
 	 */
-	hard_irq_disable();
+	local_irq_enable();
 
 	while (1) {
 		/* let's not take timer interrupts too often ... */
 		set_dec(0x7fffffff);
 
-		/* should always be true at this point */
-		if (cpu_has_feature(CPU_FTR_CAN_NAP))
-			power4_cpu_offline_powersave();
+		/* Enter NAP mode */
+		power4_idle();
 	}
 }
 
-- 
1.7.1

[PATCH 12/18] powerpc/pmac: Rename cpu_state in therm_pm72 to avoid collision

[Benjamin Herrenschmidt]

This collides with the cpu_state in our SMP code, use processor_state
instead.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 drivers/macintosh/therm_pm72.c |   60 ++++++++++++++++++++--------------------
 1 files changed, 30 insertions(+), 30 deletions(-)

diff --git a/drivers/macintosh/therm_pm72.c b/drivers/macintosh/therm_pm72.c
index f3a29f2..02b4f13 100644
--- a/drivers/macintosh/therm_pm72.c
+++ b/drivers/macintosh/therm_pm72.c
@@ -153,7 +153,7 @@ static struct i2c_adapter *		u3_0;
 static struct i2c_adapter *		u3_1;
 static struct i2c_adapter *		k2;
 static struct i2c_client *		fcu;
-static struct cpu_pid_state		cpu_state[2];
+static struct cpu_pid_state		processor_state[2];
 static struct basckside_pid_params	backside_params;
 static struct backside_pid_state	backside_state;
 static struct drives_pid_state		drives_state;
@@ -664,8 +664,8 @@ static int read_eeprom(int cpu, struct mpu_data *out)
 
 static void fetch_cpu_pumps_minmax(void)
 {
-	struct cpu_pid_state *state0 = &cpu_state[0];
-	struct cpu_pid_state *state1 = &cpu_state[1];
+	struct cpu_pid_state *state0 = &processor_state[0];
+	struct cpu_pid_state *state1 = &processor_state[1];
 	u16 pump_min = 0, pump_max = 0xffff;
 	u16 tmp[4];
 
@@ -717,17 +717,17 @@ static ssize_t show_##name(struct device *dev, struct device_attribute *attr, ch
 	return sprintf(buf, "%d", data);			\
 }
 
-BUILD_SHOW_FUNC_FIX(cpu0_temperature, cpu_state[0].last_temp)
-BUILD_SHOW_FUNC_FIX(cpu0_voltage, cpu_state[0].voltage)
-BUILD_SHOW_FUNC_FIX(cpu0_current, cpu_state[0].current_a)
-BUILD_SHOW_FUNC_INT(cpu0_exhaust_fan_rpm, cpu_state[0].rpm)
-BUILD_SHOW_FUNC_INT(cpu0_intake_fan_rpm, cpu_state[0].intake_rpm)
+BUILD_SHOW_FUNC_FIX(cpu0_temperature, processor_state[0].last_temp)
+BUILD_SHOW_FUNC_FIX(cpu0_voltage, processor_state[0].voltage)
+BUILD_SHOW_FUNC_FIX(cpu0_current, processor_state[0].current_a)
+BUILD_SHOW_FUNC_INT(cpu0_exhaust_fan_rpm, processor_state[0].rpm)
+BUILD_SHOW_FUNC_INT(cpu0_intake_fan_rpm, processor_state[0].intake_rpm)
 
-BUILD_SHOW_FUNC_FIX(cpu1_temperature, cpu_state[1].last_temp)
-BUILD_SHOW_FUNC_FIX(cpu1_voltage, cpu_state[1].voltage)
-BUILD_SHOW_FUNC_FIX(cpu1_current, cpu_state[1].current_a)
-BUILD_SHOW_FUNC_INT(cpu1_exhaust_fan_rpm, cpu_state[1].rpm)
-BUILD_SHOW_FUNC_INT(cpu1_intake_fan_rpm, cpu_state[1].intake_rpm)
+BUILD_SHOW_FUNC_FIX(cpu1_temperature, processor_state[1].last_temp)
+BUILD_SHOW_FUNC_FIX(cpu1_voltage, processor_state[1].voltage)
+BUILD_SHOW_FUNC_FIX(cpu1_current, processor_state[1].current_a)
+BUILD_SHOW_FUNC_INT(cpu1_exhaust_fan_rpm, processor_state[1].rpm)
+BUILD_SHOW_FUNC_INT(cpu1_intake_fan_rpm, processor_state[1].intake_rpm)
 
 BUILD_SHOW_FUNC_FIX(backside_temperature, backside_state.last_temp)
 BUILD_SHOW_FUNC_INT(backside_fan_pwm, backside_state.pwm)
@@ -919,8 +919,8 @@ static void do_cpu_pid(struct cpu_pid_state *state, s32 temp, s32 power)
 
 static void do_monitor_cpu_combined(void)
 {
-	struct cpu_pid_state *state0 = &cpu_state[0];
-	struct cpu_pid_state *state1 = &cpu_state[1];
+	struct cpu_pid_state *state0 = &processor_state[0];
+	struct cpu_pid_state *state1 = &processor_state[1];
 	s32 temp0, power0, temp1, power1;
 	s32 temp_combi, power_combi;
 	int rc, intake, pump;
@@ -1150,7 +1150,7 @@ static void do_monitor_cpu_rack(struct cpu_pid_state *state)
 /*
  * Initialize the state structure for one CPU control loop
  */
-static int init_cpu_state(struct cpu_pid_state *state, int index)
+static int init_processor_state(struct cpu_pid_state *state, int index)
 {
 	int err;
 
@@ -1205,7 +1205,7 @@ static int init_cpu_state(struct cpu_pid_state *state, int index)
 /*
  * Dispose of the state data for one CPU control loop
  */
-static void dispose_cpu_state(struct cpu_pid_state *state)
+static void dispose_processor_state(struct cpu_pid_state *state)
 {
 	if (state->monitor == NULL)
 		return;
@@ -1804,9 +1804,9 @@ static int main_control_loop(void *x)
 		set_pwm_fan(SLOTS_FAN_PWM_INDEX, SLOTS_FAN_DEFAULT_PWM);
 
 	/* Initialize ADCs */
-	initialize_adc(&cpu_state[0]);
-	if (cpu_state[1].monitor != NULL)
-		initialize_adc(&cpu_state[1]);
+	initialize_adc(&processor_state[0]);
+	if (processor_state[1].monitor != NULL)
+		initialize_adc(&processor_state[1]);
 
 	fcu_tickle_ticks = FCU_TICKLE_TICKS;
 
@@ -1833,14 +1833,14 @@ static int main_control_loop(void *x)
 		if (cpu_pid_type == CPU_PID_TYPE_COMBINED)
 			do_monitor_cpu_combined();
 		else if (cpu_pid_type == CPU_PID_TYPE_RACKMAC) {
-			do_monitor_cpu_rack(&cpu_state[0]);
-			if (cpu_state[1].monitor != NULL)
-				do_monitor_cpu_rack(&cpu_state[1]);
+			do_monitor_cpu_rack(&processor_state[0]);
+			if (processor_state[1].monitor != NULL)
+				do_monitor_cpu_rack(&processor_state[1]);
 			// better deal with UP
 		} else {
-			do_monitor_cpu_split(&cpu_state[0]);
-			if (cpu_state[1].monitor != NULL)
-				do_monitor_cpu_split(&cpu_state[1]);
+			do_monitor_cpu_split(&processor_state[0]);
+			if (processor_state[1].monitor != NULL)
+				do_monitor_cpu_split(&processor_state[1]);
 			// better deal with UP
 		}
 		/* Then, the rest */
@@ -1885,8 +1885,8 @@ static int main_control_loop(void *x)
  */
 static void dispose_control_loops(void)
 {
-	dispose_cpu_state(&cpu_state[0]);
-	dispose_cpu_state(&cpu_state[1]);
+	dispose_processor_state(&processor_state[0]);
+	dispose_processor_state(&processor_state[1]);
 	dispose_backside_state(&backside_state);
 	dispose_drives_state(&drives_state);
 	dispose_slots_state(&slots_state);
@@ -1928,12 +1928,12 @@ static int create_control_loops(void)
 	/* Create control loops for everything. If any fail, everything
 	 * fails
 	 */
-	if (init_cpu_state(&cpu_state[0], 0))
+	if (init_processor_state(&processor_state[0], 0))
 		goto fail;
 	if (cpu_pid_type == CPU_PID_TYPE_COMBINED)
 		fetch_cpu_pumps_minmax();
 
-	if (cpu_count > 1 && init_cpu_state(&cpu_state[1], 1))
+	if (cpu_count > 1 && init_processor_state(&processor_state[1], 1))
 		goto fail;
 	if (init_backside_state(&backside_state))
 		goto fail;
-- 
1.7.1

[PATCH 13/18] powerpc/smp: Add a smp_ops->bringup_up() done callback

[Benjamin Herrenschmidt]

This allows us to stop abusing smp_ops->setup_cpu() for cleanup
tasks that have to take place after the initial boot time CPU
bringup.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/machdep.h |    1 +
 arch/powerpc/kernel/smp.c          |    4 ++++
 2 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/include/asm/machdep.h b/arch/powerpc/include/asm/machdep.h
index 578d330..e4f0191 100644
--- a/arch/powerpc/include/asm/machdep.h
+++ b/arch/powerpc/include/asm/machdep.h
@@ -35,6 +35,7 @@ struct smp_ops_t {
 	int   (*probe)(void);
 	void  (*kick_cpu)(int nr);
 	void  (*setup_cpu)(int nr);
+	void  (*bringup_done)(void);
 	void  (*take_timebase)(void);
 	void  (*give_timebase)(void);
 	int   (*cpu_disable)(void);
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index f6cc5c1..df37397 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -553,7 +553,11 @@ void __init smp_cpus_done(unsigned int max_cpus)
 
 	free_cpumask_var(old_mask);
 
+	if (smp_ops && smp_ops->bringup_done)
+		smp_ops->bringup_done();
+
 	dump_numa_cpu_topology();
+
 }
 
 int arch_sd_sibling_asym_packing(void)
-- 
1.7.1

[PATCH 14/18] powerpc/pmac/smp: Fix CPU hotplug crashes on some machines

[Benjamin Herrenschmidt]

On some machines that use i2c to synchronize the timebases (such
as PowerMac7,2/7,3 G5 machines), hotplug CPU would crash when
putting back a new CPU online due to the underlying i2c bus being
closed.

This uses the newly added bringup_done() callback to move the close
along with other housekeeping calls, and adds a CPU notifier to
re-open the i2c bus around subsequent hotplug operations

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/platforms/powermac/smp.c |   71 +++++++++++++++++++++++++-------
 1 files changed, 55 insertions(+), 16 deletions(-)

diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index 837989e..74a43c6 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -840,30 +840,68 @@ static void __devinit smp_core99_setup_cpu(int cpu_nr)
 
 	/* Setup openpic */
 	mpic_setup_this_cpu();
+}
 
-	if (cpu_nr == 0) {
-#ifdef CONFIG_PPC64
-		extern void g5_phy_disable_cpu1(void);
+#ifdef CONFIG_HOTPLUG_CPU
+static int smp_core99_cpu_notify(struct notifier_block *self,
+				 unsigned long action, void *hcpu)
+{
+	int rc;
 
-		/* Close i2c bus if it was used for tb sync */
+	switch(action) {
+	case CPU_UP_PREPARE:
+	case CPU_UP_PREPARE_FROZEN:
+		/* Open i2c bus if it was used for tb sync */
 		if (pmac_tb_clock_chip_host) {
-			pmac_i2c_close(pmac_tb_clock_chip_host);
-			pmac_tb_clock_chip_host	= NULL;
+			rc = pmac_i2c_open(pmac_tb_clock_chip_host, 1);
+			if (rc) {
+				pr_err("Failed to open i2c bus for time sync\n");
+				return notifier_from_errno(rc);
+			}
 		}
+		break;
+	case CPU_ONLINE:
+	case CPU_UP_CANCELED:
+		/* Close i2c bus if it was used for tb sync */
+		if (pmac_tb_clock_chip_host)
+			pmac_i2c_close(pmac_tb_clock_chip_host);
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_OK;
+}
 
-		/* If we didn't start the second CPU, we must take
-		 * it off the bus
-		 */
-		if (of_machine_is_compatible("MacRISC4") &&
-		    num_online_cpus() < 2)		
-			g5_phy_disable_cpu1();
-#endif /* CONFIG_PPC64 */
+static struct notifier_block __cpuinitdata smp_core99_cpu_nb = {
+	.notifier_call	= smp_core99_cpu_notify,
+};
+#endif /* CONFIG_HOTPLUG_CPU */
 
-		if (ppc_md.progress)
-			ppc_md.progress("core99_setup_cpu 0 done", 0x349);
+static void __init smp_core99_bringup_done(void)
+{
+#ifdef CONFIG_PPC64
+	extern void g5_phy_disable_cpu1(void);
+
+	/* Close i2c bus if it was used for tb sync */
+	if (pmac_tb_clock_chip_host)
+		pmac_i2c_close(pmac_tb_clock_chip_host);
+
+	/* If we didn't start the second CPU, we must take
+	 * it off the bus.
+	 */
+	if (of_machine_is_compatible("MacRISC4") &&
+	    num_online_cpus() < 2) {
+		set_cpu_present(1, false);
+		g5_phy_disable_cpu1();
 	}
-}
+#endif /* CONFIG_PPC64 */
 
+#ifdef CONFIG_HOTPLUG_CPU
+	register_cpu_notifier(&smp_core99_cpu_nb);
+#endif
+	if (ppc_md.progress)
+		ppc_md.progress("smp_core99_bringup_done", 0x349);
+}
 
 #ifdef CONFIG_HOTPLUG_CPU
 
@@ -940,6 +978,7 @@ static void pmac_cpu_die(void)
 struct smp_ops_t core99_smp_ops = {
 	.message_pass	= smp_mpic_message_pass,
 	.probe		= smp_core99_probe,
+	.bringup_done	= smp_core99_bringup_done,
 	.kick_cpu	= smp_core99_kick_cpu,
 	.setup_cpu	= smp_core99_setup_cpu,
 	.give_timebase	= smp_core99_give_timebase,
-- 
1.7.1

[PATCH 15/18] powerpc/smp: Don't expose per-cpu "cpu_state" array

[Benjamin Herrenschmidt]

Instead, keep it static, expose an accessor and use that from
the PowerMac code. Avoids easy namespace collisions and will
make it easier to consolidate with other implementations.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/include/asm/smp.h        |    2 +-
 arch/powerpc/kernel/smp.c             |    7 ++++++-
 arch/powerpc/platforms/powermac/smp.c |   12 ++++++++----
 3 files changed, 15 insertions(+), 6 deletions(-)

diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
index 7e99771..a902a0d 100644
--- a/arch/powerpc/include/asm/smp.h
+++ b/arch/powerpc/include/asm/smp.h
@@ -45,7 +45,7 @@ extern void migrate_irqs(void);
 int generic_cpu_disable(void);
 void generic_cpu_die(unsigned int cpu);
 void generic_mach_cpu_die(void);
-DECLARE_PER_CPU(int, cpu_state);
+void generic_set_cpu_dead(unsigned int cpu);
 #endif
 
 #ifdef CONFIG_PPC64
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index df37397..7beb001 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -305,7 +305,7 @@ void __devinit smp_prepare_boot_cpu(void)
 
 #ifdef CONFIG_HOTPLUG_CPU
 /* State of each CPU during hotplug phases */
-DEFINE_PER_CPU(int, cpu_state) = { 0 };
+static DEFINE_PER_CPU(int, cpu_state) = { 0 };
 
 int generic_cpu_disable(void)
 {
@@ -348,6 +348,11 @@ void generic_mach_cpu_die(void)
 	while (__get_cpu_var(cpu_state) != CPU_UP_PREPARE)
 		cpu_relax();
 }
+
+void generic_set_cpu_dead(unsigned int cpu)
+{
+	__get_cpu_var(cpu_state) = CPU_DEAD;
+}
 #endif
 
 int __cpuinit __cpu_up(unsigned int cpu)
diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index 74a43c6..ce5b4f5 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -920,10 +920,12 @@ static int smp_core99_cpu_disable(void)
 
 static void pmac_cpu_die(void)
 {
+	int cpu = smp_processor_id();
+
 	local_irq_disable();
 	idle_task_exit();
-	printk(KERN_DEBUG "CPU%d offline\n", smp_processor_id());
-	__get_cpu_var(cpu_state) = CPU_DEAD;
+	pr_debug("CPU%d offline\n", cpu);
+	generic_set_cpu_dead(cpu);
 	smp_wmb();
 	mb();
 	low_cpu_die();
@@ -933,6 +935,8 @@ static void pmac_cpu_die(void)
 
 static void pmac_cpu_die(void)
 {
+	int cpu = smp_processor_id();
+
 	local_irq_disable();
 	idle_task_exit();
 
@@ -942,8 +946,8 @@ static void pmac_cpu_die(void)
 	 * on core99 platforms for now ...
 	 */
 
-	printk(KERN_INFO "CPU#%d offline\n", smp_processor_id());
-	__get_cpu_var(cpu_state) = CPU_DEAD;
+	printk(KERN_INFO "CPU#%d offline\n", cpu);
+	generic_set_cpu_dead(cpu);
 	smp_wmb();
 
 	/*
-- 
1.7.1

[PATCH 16/18] powerpc/smp: Create idle threads on demand and properly reset them

[Benjamin Herrenschmidt]

Instead of creating idle threads at boot for all possible CPUs, we
create them on demand, like x86 or ARM, and we properly call init_idle
to re-initialize an idle thread when a CPU was unplugged and is now
re-plugged.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/kernel/smp.c |  100 +++++++++++++++++++++++++++++++++++----------
 1 files changed, 78 insertions(+), 22 deletions(-)

diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 7beb001..e337073 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -57,6 +57,25 @@
 #define DBG(fmt...)
 #endif
 
+
+/* Store all idle threads, this can be reused instead of creating
+* a new thread. Also avoids complicated thread destroy functionality
+* for idle threads.
+*/
+#ifdef CONFIG_HOTPLUG_CPU
+/*
+ * Needed only for CONFIG_HOTPLUG_CPU because __cpuinitdata is
+ * removed after init for !CONFIG_HOTPLUG_CPU.
+ */
+static DEFINE_PER_CPU(struct task_struct *, idle_thread_array);
+#define get_idle_for_cpu(x)      (per_cpu(idle_thread_array, x))
+#define set_idle_for_cpu(x, p)   (per_cpu(idle_thread_array, x) = (p))
+#else
+static struct task_struct *idle_thread_array[NR_CPUS] __cpuinitdata ;
+#define get_idle_for_cpu(x)      (idle_thread_array[(x)])
+#define set_idle_for_cpu(x, p)   (idle_thread_array[(x)] = (p))
+#endif
+
 struct thread_info *secondary_ti;
 
 DEFINE_PER_CPU(cpumask_var_t, cpu_sibling_map);
@@ -238,23 +257,6 @@ static void __devinit smp_store_cpu_info(int id)
 	per_cpu(cpu_pvr, id) = mfspr(SPRN_PVR);
 }
 
-static void __init smp_create_idle(unsigned int cpu)
-{
-	struct task_struct *p;
-
-	/* create a process for the processor */
-	p = fork_idle(cpu);
-	if (IS_ERR(p))
-		panic("failed fork for CPU %u: %li", cpu, PTR_ERR(p));
-#ifdef CONFIG_PPC64
-	paca[cpu].__current = p;
-	paca[cpu].kstack = (unsigned long) task_thread_info(p)
-		+ THREAD_SIZE - STACK_FRAME_OVERHEAD;
-#endif
-	current_set[cpu] = task_thread_info(p);
-	task_thread_info(p)->cpu = cpu;
-}
-
 void __init smp_prepare_cpus(unsigned int max_cpus)
 {
 	unsigned int cpu;
@@ -288,10 +290,6 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 			max_cpus = NR_CPUS;
 	else
 		max_cpus = 1;
-
-	for_each_possible_cpu(cpu)
-		if (cpu != boot_cpuid)
-			smp_create_idle(cpu);
 }
 
 void __devinit smp_prepare_boot_cpu(void)
@@ -355,9 +353,62 @@ void generic_set_cpu_dead(unsigned int cpu)
 }
 #endif
 
+struct create_idle {
+	struct work_struct work;
+	struct task_struct *idle;
+	struct completion done;
+	int cpu;
+};
+
+static void __cpuinit do_fork_idle(struct work_struct *work)
+{
+	struct create_idle *c_idle =
+		container_of(work, struct create_idle, work);
+
+	c_idle->idle = fork_idle(c_idle->cpu);
+	complete(&c_idle->done);
+}
+
+static int __cpuinit create_idle(unsigned int cpu)
+{
+	struct thread_info *ti;
+	struct create_idle c_idle = {
+		.cpu	= cpu,
+		.done	= COMPLETION_INITIALIZER_ONSTACK(c_idle.done),
+	};
+	INIT_WORK_ONSTACK(&c_idle.work, do_fork_idle);
+
+	c_idle.idle = get_idle_for_cpu(cpu);
+
+	/* We can't use kernel_thread since we must avoid to
+	 * reschedule the child. We use a workqueue because
+	 * we want to fork from a kernel thread, not whatever
+	 * userspace process happens to be trying to online us.
+	 */
+	if (!c_idle.idle) {
+		schedule_work(&c_idle.work);
+		wait_for_completion(&c_idle.done);
+	} else
+		init_idle(c_idle.idle, cpu);
+	if (IS_ERR(c_idle.idle)) {		
+		pr_err("Failed fork for CPU %u: %li", cpu, PTR_ERR(c_idle.idle));
+		return PTR_ERR(c_idle.idle);
+	}
+	ti = task_thread_info(c_idle.idle);
+
+#ifdef CONFIG_PPC64
+	paca[cpu].__current = c_idle.idle;
+	paca[cpu].kstack = (unsigned long)ti + THREAD_SIZE - STACK_FRAME_OVERHEAD;
+#endif
+	ti->cpu = cpu;
+	current_set[cpu] = ti;
+
+	return 0;
+}
+
 int __cpuinit __cpu_up(unsigned int cpu)
 {
-	int c;
+	int rc, c;
 
 	secondary_ti = current_set[cpu];
 
@@ -365,6 +416,11 @@ int __cpuinit __cpu_up(unsigned int cpu)
 	    (smp_ops->cpu_bootable && !smp_ops->cpu_bootable(cpu)))
 		return -EINVAL;
 
+	/* Make sure we have an idle thread */
+	rc = create_idle(cpu);
+	if (rc)
+		return rc;
+
 	/* Make sure callin-map entry is 0 (can be leftover a CPU
 	 * hotplug
 	 */
-- 
1.7.1

[PATCH 17/18] powerpc/smp: Increase vdso_data->processorCount, not just decrease it

[Benjamin Herrenschmidt]

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/kernel/smp.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index e337073..6023395 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -551,6 +551,10 @@ void __devinit start_secondary(void *unused)
 
 	secondary_cpu_time_init();
 
+#ifdef CONFIG_PPC64
+	if (system_state == SYSTEM_RUNNING)
+		vdso_data->processorCount++;
+#endif
 	ipi_call_lock();
 	notify_cpu_starting(cpu);
 	set_cpu_online(cpu, true);
-- 
1.7.1

[PATCH 18/18] powerpc/pmac/smp: Remove no-longer needed preempt workaround

[Benjamin Herrenschmidt]

The generic code properly re-initializes the preempt count in the
idle thread now

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---
 arch/powerpc/platforms/powermac/smp.c |    7 -------
 1 files changed, 0 insertions(+), 7 deletions(-)

diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index ce5b4f5..a830c5e 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -951,13 +951,6 @@ static void pmac_cpu_die(void)
 	smp_wmb();
 
 	/*
-	 * during the path that leads here preemption is disabled,
-	 * reenable it now so that when coming up preempt count is
-	 * zero correctly
-	 */
-	preempt_enable();
-
-	/*
 	 * Re-enable interrupts. The NAP code needs to enable them
 	 * anyways, do it now so we deal with the case where one already
 	 * happened while soft-disabled.
-- 
1.7.1

Re: [PATCH 17/18] powerpc/smp: Increase vdso_data->processorCount, not just decrease it

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 1032 bytes --]

On Tue, 2011-03-08 at 17:37 +1100, Benjamin Herrenschmidt wrote:
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
>  arch/powerpc/kernel/smp.c |    4 ++++
>  1 files changed, 4 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
> index e337073..6023395 100644
> --- a/arch/powerpc/kernel/smp.c
> +++ b/arch/powerpc/kernel/smp.c
> @@ -551,6 +551,10 @@ void __devinit start_secondary(void *unused)
>  
>  	secondary_cpu_time_init();
>  
> +#ifdef CONFIG_PPC64
> +	if (system_state == SYSTEM_RUNNING)
> +		vdso_data->processorCount++;
> +#endif

So the SYSTEM_RUNNING check is to avoid clashing with the logic in
smp_setup_cpu_maps() I presume:

arch/powerpc/kernel/setup-common.c:     vdso_data->processorCount = num_present_cpus();


But why not remove that, and let the increment in start_secondary() do
all the work?

With the current code if a cpu is present but fails to come up the count
will be wrong I think.

cheers


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [PATCH 17/18] powerpc/smp: Increase vdso_data->processorCount, not just decrease it

[Benjamin Herrenschmidt]

On Thu, 2011-03-10 at 14:41 +1100, Michael Ellerman wrote:
> 
> So the SYSTEM_RUNNING check is to avoid clashing with the logic in
> smp_setup_cpu_maps() I presume:
> 
> arch/powerpc/kernel/setup-common.c:     vdso_data->processorCount = num_present_cpus();
> 
> 
> But why not remove that, and let the increment in start_secondary() do
> all the work?
> 
> With the current code if a cpu is present but fails to come up the count
> will be wrong I think. 

I agree, on the other hand I'm worried about the way that things it
used in lparcfg.c (which looks bogus regardless) and it's not clear
to me what userspace uses it for (I bet some "licencing" stuff :-)

The lparcfg bit looks like something that should turn into counting
possible processors I suppose but it's hard to tell.

Cheers,
Ben.

Re: [PATCH 01/18] powerpc: Make decrementer interrupt robust against offlined CPUs

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 1009 bytes --]

On Tue, 2011-03-08 at 17:37 +1100, Benjamin Herrenschmidt wrote:
> With some implementations, it is possible that a timer interrupt
> occurs every few seconds on an offline CPU. In this case, just
> re-arm the decrementer and return immediately
> 
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
>  arch/powerpc/kernel/time.c |   15 +++++++++++----
>  1 files changed, 11 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
> index 09d31db..ecebc7c 100644
> --- a/arch/powerpc/kernel/time.c
> +++ b/arch/powerpc/kernel/time.c
> @@ -577,14 +577,21 @@ void timer_interrupt(struct pt_regs * regs)
>  	struct clock_event_device *evt = &decrementer->event;
>  	u64 now;
>  
> +	/* Ensure a positive value is written to the decrementer, or else
> +	 * some CPUs will continuue to take decrementer exceptions
                                  ^

Seeing as you're moving it anyway, and maybe a full-stop.

cheers


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [PATCH 15/18] powerpc/smp: Don't expose per-cpu "cpu_state" array

[Michael Ellerman]

[-- Attachment #1: Type: text/plain, Size: 1179 bytes --]

On Tue, 2011-03-08 at 17:37 +1100, Benjamin Herrenschmidt wrote:
> Instead, keep it static, expose an accessor and use that from
> the PowerMac code. Avoids easy namespace collisions and will
> make it easier to consolidate with other implementations.
> 
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
>  arch/powerpc/include/asm/smp.h        |    2 +-
>  arch/powerpc/kernel/smp.c             |    7 ++++++-
>  arch/powerpc/platforms/powermac/smp.c |   12 ++++++++----
>  3 files changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/smp.h b/arch/powerpc/include/asm/smp.h
> index 7e99771..a902a0d 100644
> --- a/arch/powerpc/include/asm/smp.h
> +++ b/arch/powerpc/include/asm/smp.h
> @@ -45,7 +45,7 @@ extern void migrate_irqs(void);
>  int generic_cpu_disable(void);
>  void generic_cpu_die(unsigned int cpu);
>  void generic_mach_cpu_die(void);
> -DECLARE_PER_CPU(int, cpu_state);
> +void generic_set_cpu_dead(unsigned int cpu);

Takes a cpu parameter ..

> +void generic_set_cpu_dead(unsigned int cpu)
> +{
> +	__get_cpu_var(cpu_state) = CPU_DEAD;
> +}

But doesn't use it.

cheers


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [PATCH 15/18] powerpc/smp: Don't expose per-cpu "cpu_state" array

[Benjamin Herrenschmidt]

On Tue, 2011-03-29 at 16:37 +1100, Michael Ellerman wrote:
> Takes a cpu parameter ..
> 
> > +void generic_set_cpu_dead(unsigned int cpu)
> > +{
> > +     __get_cpu_var(cpu_state) = CPU_DEAD;
> > +}
> 
> But doesn't use it.

Yeah that's a bug. Fortunately it's always called on the current CPU :-)
I'll fix it tho.

Thanks,
Ben.