[PATCH 5/6]physical CPU hot add

[PATCH 5/6]physical CPU hot add

[Li Shaohua]

Boot a CPU at runtime.

Signed-off-by: Li Shaohua<shaohua.li@intel.com>
---

 linux-2.6.11-root/arch/i386/kernel/smpboot.c |  112 ++++++++++++++++++++-------
 linux-2.6.11-root/drivers/base/cpu.c         |    8 +
 linux-2.6.11-root/include/asm-i386/smp.h     |    2 
 3 files changed, 93 insertions(+), 29 deletions(-)

diff -puN arch/i386/kernel/smpboot.c~warm_boot_cpu arch/i386/kernel/smpboot.c
--- linux-2.6.11/arch/i386/kernel/smpboot.c~warm_boot_cpu	2005-04-12 10:38:16.720411760 +0800
+++ linux-2.6.11-root/arch/i386/kernel/smpboot.c	2005-04-12 11:11:09.155556040 +0800
@@ -80,6 +80,12 @@ cpumask_t cpu_callin_map;
 cpumask_t cpu_callout_map;
 static cpumask_t smp_commenced_mask;
 
+/* TSC's upper 32 bits can't be written in eariler CPU (before prescott), there
+ * is no way to resync one AP against BP. TBD: for prescott and above, we
+ * should use IA64's algorithm
+ */
+static int __devinitdata tsc_sync_disabled;
+
 /* Per CPU bogomips and other parameters */
 struct cpuinfo_x86 cpu_data[NR_CPUS] __cacheline_aligned;
 
@@ -416,7 +422,7 @@ static void __devinit smp_callin(void)
 	/*
 	 *      Synchronize the TSC with the BP
 	 */
-	if (cpu_has_tsc && cpu_khz)
+	if (cpu_has_tsc && cpu_khz && !tsc_sync_disabled)
 		synchronize_tsc_ap();
 }
 
@@ -809,6 +815,31 @@ static inline int alloc_cpu_id(void)
 	return cpu;
 }
 
+#ifdef CONFIG_HOTPLUG_CPU
+static struct task_struct * __devinitdata cpu_idle_tasks[NR_CPUS];
+static inline struct task_struct * alloc_idle_task(int cpu)
+{
+	struct task_struct *idle;
+
+	if ((idle = cpu_idle_tasks[cpu]) != NULL) {
+		/* initialize thread_struct.  we really want to avoid destroy
+		 * idle tread
+		 */
+		idle->thread.esp = (unsigned long)(((struct pt_regs *)
+			(THREAD_SIZE + (unsigned long) idle->thread_info)) - 1);
+		init_idle(idle, cpu);
+		return idle;
+	}
+	idle = fork_idle(cpu);
+
+	if (!IS_ERR(idle))
+		cpu_idle_tasks[cpu] = idle;
+	return idle;
+}
+#else
+#define alloc_idle_task(cpu) fork_idle(cpu)
+#endif
+
 static int __devinit do_boot_cpu(int apicid, int cpu)
 /*
  * NOTE - on most systems this is a PHYSICAL apic ID, but on multiquad
@@ -828,7 +859,7 @@ static int __devinit do_boot_cpu(int api
 	 * We can't use kernel_thread since we must avoid to
 	 * reschedule the child.
 	 */
-	idle = fork_idle(cpu);
+	idle = alloc_idle_task(cpu);
 	if (IS_ERR(idle))
 		panic("failed fork for CPU %d", cpu);
 	idle->thread.eip = (unsigned long) start_secondary;
@@ -931,6 +962,55 @@ void cpu_exit_clear(void)
 	cpu_clear(cpu, smp_commenced_mask);
 	unmap_cpu_to_logical_apicid(cpu);
 }
+
+struct warm_boot_cpu_info {
+	struct completion *complete;
+	int apicid;
+	int cpu;
+};
+
+static void __devinit do_warm_boot_cpu(void *p)
+{
+	struct warm_boot_cpu_info *info = p;
+	do_boot_cpu(info->apicid, info->cpu);
+	complete(info->complete);
+}
+
+int __devinit smp_prepare_cpu(int cpu)
+{
+	DECLARE_COMPLETION(done);
+	struct warm_boot_cpu_info info;
+	struct work_struct task;
+	int	apicid, ret;
+
+	lock_cpu_hotplug();
+	apicid = x86_cpu_to_apicid[cpu];
+	if (apicid == BAD_APICID) {
+		ret = -ENODEV;
+		goto exit;
+	}
+
+	info.complete = &done;
+	info.apicid = apicid;
+	info.cpu = cpu;
+	INIT_WORK(&task, do_warm_boot_cpu, &info);
+
+	tsc_sync_disabled = 1;
+
+	/* init low mem mapping */
+	memcpy(swapper_pg_dir, swapper_pg_dir + USER_PGD_PTRS,
+			sizeof(swapper_pg_dir[0]) * KERNEL_PGD_PTRS);
+	flush_tlb_all();
+	schedule_work(&task);
+	wait_for_completion(&done);
+
+	tsc_sync_disabled = 0;
+	zap_low_mappings();
+	ret = 0;
+exit:
+	unlock_cpu_hotplug();
+	return ret;
+}
 #endif
 
 static void smp_tune_scheduling (void)
@@ -1169,24 +1249,6 @@ void __devinit smp_prepare_boot_cpu(void
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
-
-/* must be called with the cpucontrol mutex held */
-static int __devinit cpu_enable(unsigned int cpu)
-{
-	/* get the target out of its holding state */
-	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
-	wmb();
-
-	/* wait for the processor to ack it. timeout? */
-	while (!cpu_online(cpu))
-		cpu_relax();
-
-	fixup_irqs(cpu_online_map);
-	/* counter the disable in fixup_irqs() */
-	local_irq_enable();
-	return 0;
-}
-
 static void
 remove_siblinginfo(int cpu)
 {
@@ -1270,14 +1332,6 @@ int __devinit __cpu_up(unsigned int cpu)
 		return -EIO;
 	}
 
-#ifdef CONFIG_HOTPLUG_CPU
-	/* Already up, and in cpu_quiescent now? */
-	if (cpu_isset(cpu, smp_commenced_mask)) {
-		cpu_enable(cpu);
-		return 0;
-	}
-#endif
-
 	local_irq_enable();
 	/* Unleash the CPU! */
 	cpu_set(cpu, smp_commenced_mask);
@@ -1292,10 +1346,12 @@ void __init smp_cpus_done(unsigned int m
 	setup_ioapic_dest();
 #endif
 	zap_low_mappings();
+#ifndef CONFIG_HOTPLUG_CPU
 	/*
 	 * Disable executability of the SMP trampoline:
 	 */
 	set_kernel_exec((unsigned long)trampoline_base, trampoline_exec);
+#endif
 }
 
 void __init smp_intr_init(void)
diff -puN drivers/base/cpu.c~warm_boot_cpu drivers/base/cpu.c
--- linux-2.6.11/drivers/base/cpu.c~warm_boot_cpu	2005-04-12 10:38:16.721411608 +0800
+++ linux-2.6.11-root/drivers/base/cpu.c	2005-04-12 10:38:16.727410696 +0800
@@ -16,6 +16,11 @@ struct sysdev_class cpu_sysdev_class = {
 EXPORT_SYMBOL(cpu_sysdev_class);
 
 #ifdef CONFIG_HOTPLUG_CPU
+int __attribute__ ((weak)) smp_prepare_cpu(int cpu)
+{
+	return 0;
+}
+
 static ssize_t show_online(struct sys_device *dev, char *buf)
 {
 	struct cpu *cpu = container_of(dev, struct cpu, sysdev);
@@ -36,7 +41,8 @@ static ssize_t store_online(struct sys_d
 			kobject_hotplug(&dev->kobj, KOBJ_OFFLINE);
 		break;
 	case '1':
-		ret = cpu_up(cpu->sysdev.id);
+		if ((ret = smp_prepare_cpu(cpu->sysdev.id)) == 0)
+			ret = cpu_up(cpu->sysdev.id);
 		break;
 	default:
 		ret = -EINVAL;
diff -puN include/asm-i386/smp.h~warm_boot_cpu include/asm-i386/smp.h
--- linux-2.6.11/include/asm-i386/smp.h~warm_boot_cpu	2005-04-12 10:38:16.722411456 +0800
+++ linux-2.6.11-root/include/asm-i386/smp.h	2005-04-12 10:38:16.727410696 +0800
@@ -52,6 +52,8 @@ extern u8 x86_cpu_to_apicid[];
 #ifdef CONFIG_HOTPLUG_CPU
 extern void cpu_exit_clear(void);
 extern void cpu_uninit(void);
+
+extern int smp_prepare_cpu(int cpu);
 #endif
 
 /*
_

Re: [PATCH 5/6]physical CPU hot add

[Zwane Mwaikambo]

On Tue, 12 Apr 2005, Li Shaohua wrote:

>  #ifdef CONFIG_HOTPLUG_CPU
> +int __attribute__ ((weak)) smp_prepare_cpu(int cpu)
> +{
> +	return 0;
> +}
> +

Any way for you to avoid using weak attribute?

Thanks,
	Zwane

Re: [PATCH 5/6]physical CPU hot add

[Li Shaohua]

On Tue, 2005-04-12 at 20:17, Zwane Mwaikambo wrote:
> On Tue, 12 Apr 2005, Li Shaohua wrote:
> 
> >  #ifdef CONFIG_HOTPLUG_CPU
> > +int __attribute__ ((weak)) smp_prepare_cpu(int cpu)
> > +{
> > +	return 0;
> > +}
> > +
> 
> Any way for you to avoid using weak attribute?
Just want to avoid more 'ifdef' or 'define empty routine for other
archs' staffs. Someone prefer 'weak' attribute. Either way is ok to me,
but if you think the former is better, I'd change it.

Thanks,
Shaohua

Re: [PATCH 5/6]physical CPU hot add

[Zwane Mwaikambo]

On Tue, 12 Apr 2005, Li Shaohua wrote:

> On Tue, 2005-04-12 at 20:17, Zwane Mwaikambo wrote:
> > On Tue, 12 Apr 2005, Li Shaohua wrote:
> > 
> > >  #ifdef CONFIG_HOTPLUG_CPU
> > > +int __attribute__ ((weak)) smp_prepare_cpu(int cpu)
> > > +{
> > > +	return 0;
> > > +}
> > > +
> > 
> > Any way for you to avoid using weak attribute?
> Just want to avoid more 'ifdef' or 'define empty routine for other
> archs' staffs. Someone prefer 'weak' attribute. Either way is ok to me,
> but if you think the former is better, I'd change it.

The define method is fine and preferred.

Thanks!
	Zwane

Re: [PATCH 5/6]physical CPU hot add

[Li Shaohua]

On Tue, 2005-04-12 at 20:17, Zwane Mwaikambo wrote:
> On Tue, 12 Apr 2005, Li Shaohua wrote:
> 
> >  #ifdef CONFIG_HOTPLUG_CPU
> > +int __attribute__ ((weak)) smp_prepare_cpu(int cpu)
> > +{
> > +	return 0;
> > +}
> > +
> 
> Any way for you to avoid using weak attribute?
Replace weak attribute with define method as suggested.

Thanks,
Shaohua


---

 linux-2.6.11-root/arch/i386/kernel/smpboot.c |  112 ++++++++++++++++++++-------
 linux-2.6.11-root/drivers/base/cpu.c         |    7 +
 linux-2.6.11-root/include/asm-i386/smp.h     |    3 
 3 files changed, 93 insertions(+), 29 deletions(-)

diff -puN arch/i386/kernel/smpboot.c~warm_boot_cpu arch/i386/kernel/smpboot.c
--- linux-2.6.11/arch/i386/kernel/smpboot.c~warm_boot_cpu	2005-04-13 10:58:37.152081456 +0800
+++ linux-2.6.11-root/arch/i386/kernel/smpboot.c	2005-04-13 10:58:37.159080392 +0800
@@ -80,6 +80,12 @@ cpumask_t cpu_callin_map;
 cpumask_t cpu_callout_map;
 static cpumask_t smp_commenced_mask;
 
+/* TSC's upper 32 bits can't be written in eariler CPU (before prescott), there
+ * is no way to resync one AP against BP. TBD: for prescott and above, we
+ * should use IA64's algorithm
+ */
+static int __devinitdata tsc_sync_disabled;
+
 /* Per CPU bogomips and other parameters */
 struct cpuinfo_x86 cpu_data[NR_CPUS] __cacheline_aligned;
 
@@ -416,7 +422,7 @@ static void __devinit smp_callin(void)
 	/*
 	 *      Synchronize the TSC with the BP
 	 */
-	if (cpu_has_tsc && cpu_khz)
+	if (cpu_has_tsc && cpu_khz && !tsc_sync_disabled)
 		synchronize_tsc_ap();
 }
 
@@ -809,6 +815,31 @@ static inline int alloc_cpu_id(void)
 	return cpu;
 }
 
+#ifdef CONFIG_HOTPLUG_CPU
+static struct task_struct * __devinitdata cpu_idle_tasks[NR_CPUS];
+static inline struct task_struct * alloc_idle_task(int cpu)
+{
+	struct task_struct *idle;
+
+	if ((idle = cpu_idle_tasks[cpu]) != NULL) {
+		/* initialize thread_struct.  we really want to avoid destroy
+		 * idle tread
+		 */
+		idle->thread.esp = (unsigned long)(((struct pt_regs *)
+			(THREAD_SIZE + (unsigned long) idle->thread_info)) - 1);
+		init_idle(idle, cpu);
+		return idle;
+	}
+	idle = fork_idle(cpu);
+
+	if (!IS_ERR(idle))
+		cpu_idle_tasks[cpu] = idle;
+	return idle;
+}
+#else
+#define alloc_idle_task(cpu) fork_idle(cpu)
+#endif
+
 static int __devinit do_boot_cpu(int apicid, int cpu)
 /*
  * NOTE - on most systems this is a PHYSICAL apic ID, but on multiquad
@@ -828,7 +859,7 @@ static int __devinit do_boot_cpu(int api
 	 * We can't use kernel_thread since we must avoid to
 	 * reschedule the child.
 	 */
-	idle = fork_idle(cpu);
+	idle = alloc_idle_task(cpu);
 	if (IS_ERR(idle))
 		panic("failed fork for CPU %d", cpu);
 	idle->thread.eip = (unsigned long) start_secondary;
@@ -931,6 +962,55 @@ void cpu_exit_clear(void)
 	cpu_clear(cpu, smp_commenced_mask);
 	unmap_cpu_to_logical_apicid(cpu);
 }
+
+struct warm_boot_cpu_info {
+	struct completion *complete;
+	int apicid;
+	int cpu;
+};
+
+static void __devinit do_warm_boot_cpu(void *p)
+{
+	struct warm_boot_cpu_info *info = p;
+	do_boot_cpu(info->apicid, info->cpu);
+	complete(info->complete);
+}
+
+int __devinit smp_prepare_cpu(int cpu)
+{
+	DECLARE_COMPLETION(done);
+	struct warm_boot_cpu_info info;
+	struct work_struct task;
+	int	apicid, ret;
+
+	lock_cpu_hotplug();
+	apicid = x86_cpu_to_apicid[cpu];
+	if (apicid == BAD_APICID) {
+		ret = -ENODEV;
+		goto exit;
+	}
+
+	info.complete = &done;
+	info.apicid = apicid;
+	info.cpu = cpu;
+	INIT_WORK(&task, do_warm_boot_cpu, &info);
+
+	tsc_sync_disabled = 1;
+
+	/* init low mem mapping */
+	memcpy(swapper_pg_dir, swapper_pg_dir + USER_PGD_PTRS,
+			sizeof(swapper_pg_dir[0]) * KERNEL_PGD_PTRS);
+	flush_tlb_all();
+	schedule_work(&task);
+	wait_for_completion(&done);
+
+	tsc_sync_disabled = 0;
+	zap_low_mappings();
+	ret = 0;
+exit:
+	unlock_cpu_hotplug();
+	return ret;
+}
 #endif
 
 static void smp_tune_scheduling (void)
@@ -1169,24 +1249,6 @@ void __devinit smp_prepare_boot_cpu(void
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
-
-/* must be called with the cpucontrol mutex held */
-static int __devinit cpu_enable(unsigned int cpu)
-{
-	/* get the target out of its holding state */
-	per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
-	wmb();
-
-	/* wait for the processor to ack it. timeout? */
-	while (!cpu_online(cpu))
-		cpu_relax();
-
-	fixup_irqs(cpu_online_map);
-	/* counter the disable in fixup_irqs() */
-	local_irq_enable();
-	return 0;
-}
-
 static void
 remove_siblinginfo(int cpu)
 {
@@ -1270,14 +1332,6 @@ int __devinit __cpu_up(unsigned int cpu)
 		return -EIO;
 	}
 
-#ifdef CONFIG_HOTPLUG_CPU
-	/* Already up, and in cpu_quiescent now? */
-	if (cpu_isset(cpu, smp_commenced_mask)) {
-		cpu_enable(cpu);
-		return 0;
-	}
-#endif
-
 	local_irq_enable();
 	/* Unleash the CPU! */
 	cpu_set(cpu, smp_commenced_mask);
@@ -1292,10 +1346,12 @@ void __init smp_cpus_done(unsigned int m
 	setup_ioapic_dest();
 #endif
 	zap_low_mappings();
+#ifndef CONFIG_HOTPLUG_CPU
 	/*
 	 * Disable executability of the SMP trampoline:
 	 */
 	set_kernel_exec((unsigned long)trampoline_base, trampoline_exec);
+#endif
 }
 
 void __init smp_intr_init(void)
diff -puN drivers/base/cpu.c~warm_boot_cpu drivers/base/cpu.c
--- linux-2.6.11/drivers/base/cpu.c~warm_boot_cpu	2005-04-13 10:58:37.153081304 +0800
+++ linux-2.6.11-root/drivers/base/cpu.c	2005-04-13 10:58:37.159080392 +0800
@@ -16,6 +16,10 @@ struct sysdev_class cpu_sysdev_class = {
 EXPORT_SYMBOL(cpu_sysdev_class);
 
 #ifdef CONFIG_HOTPLUG_CPU
+#ifndef __HAVE_ARCH_SMP_PREPARE_CPU
+#define smp_prepare_cpu(cpu) (0)
+#endif
+
 static ssize_t show_online(struct sys_device *dev, char *buf)
 {
 	struct cpu *cpu = container_of(dev, struct cpu, sysdev);
@@ -36,7 +40,8 @@ static ssize_t store_online(struct sys_d
 			kobject_hotplug(&dev->kobj, KOBJ_OFFLINE);
 		break;
 	case '1':
-		ret = cpu_up(cpu->sysdev.id);
+		if ((ret = smp_prepare_cpu(cpu->sysdev.id)) == 0)
+			ret = cpu_up(cpu->sysdev.id);
 		break;
 	default:
 		ret = -EINVAL;
diff -puN include/asm-i386/smp.h~warm_boot_cpu include/asm-i386/smp.h
--- linux-2.6.11/include/asm-i386/smp.h~warm_boot_cpu	2005-04-13 10:58:37.155081000 +0800
+++ linux-2.6.11-root/include/asm-i386/smp.h	2005-04-13 11:00:26.353480320 +0800
@@ -52,6 +52,9 @@ extern u8 x86_cpu_to_apicid[];
 #ifdef CONFIG_HOTPLUG_CPU
 extern void cpu_exit_clear(void);
 extern void cpu_uninit(void);
+
+#define __HAVE_ARCH_SMP_PREPARE_CPU
+extern int smp_prepare_cpu(int cpu);
 #endif
 
 /*
_