Re: [PATCH v2] x86: don't ever patch back to UP if we unplug cpus.

[Rusty Russell <rusty@rustcorp.com.au>]

On Wed, 22 Aug 2012 11:41:53 +0200, Ingo Molnar <mingo@kernel.org> wrote:
> 
> * Rusty Russell <rusty@rustcorp.com.au> wrote:
> 
> > We still patch SMP instructions to UP variants if we boot with a
> > single CPU, but not at any other time.  In particular, not if we
> > unplug CPUs to return to a single cpu.
> > 
> > Paul McKenney points out:
> > 
> >  mean offline overhead is 6251/48=130.2 milliseconds.
> > 
> >  If I remove the alternatives_smp_switch() from the offline
> >  path [...] the mean offline overhead is 550/42=13.1 milliseconds
> > 
> > Basically, we're never going to get those 120ms back, and the code is
> > pretty messy.
> > 
> > We get rid of:
> > 1) The "smp-alt-once" boot option.  It's actually "smp-alt-boot", the
> >    documentation is wrong.  It's now the default.
> > 2) The skip_smp_alternatives flag used by suspend.
> > 3) arch_disable_nonboot_cpus_begin() and arch_disable_nonboot_cpus_end()
> >    which were only used to set this one flag.
> > 
> > Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
> > ---
> >  Documentation/kernel-parameters.txt |    3 -
> >  arch/x86/include/asm/alternative.h  |    4 -
> >  arch/x86/kernel/alternative.c       |  104 +++++++-----------------------------
> >  arch/x86/kernel/smpboot.c           |   20 ------
> >  kernel/cpu.c                        |   11 ---
> >  5 files changed, 27 insertions(+), 115 deletions(-)
> 
> and this version breaks the build on allyesconfig-x86-64:
> 
> arch/x86/xen/smp.c:380:3: error: implicit declaration of function ‘alternatives_smp_switch’ [-Werror=implicit-function-declaration]

Thanks, grepped harder.  Xen fixed now, thanks.

Cheers,
Rusty.

Subject: x86/smp: Don't ever patch back to UP if we unplug cpus

We still patch SMP instructions to UP variants if we boot with a
single CPU, but not at any other time.  In particular, not if we
unplug CPUs to return to a single cpu.

Paul McKenney points out:

 mean offline overhead is 6251/48=130.2 milliseconds.

 If I remove the alternatives_smp_switch() from the offline
 path [...] the mean offline overhead is 550/42=13.1 milliseconds

Basically, we're never going to get those 120ms back, and the
code is pretty messy.

We get rid of:

 1) The "smp-alt-once" boot option. It's actually "smp-alt-boot", the
    documentation is wrong. It's now the default.

 2) The skip_smp_alternatives flag used by suspend.

 3) arch_disable_nonboot_cpus_begin() and arch_disable_nonboot_cpus_end()
    which were only used to set this one flag.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Paul McKenney <paul.mckenney@us.ibm.com>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index ad7e2e5..7aef334 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -2638,9 +2638,6 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
 	smart2=		[HW]
 			Format: <io1>[,<io2>[,...,<io8>]]
 
-	smp-alt-once	[X86-32,SMP] On a hotplug CPU system, only
-			attempt to substitute SMP alternatives once at boot.
-
 	smsc-ircc2.nopnp	[HW] Don't use PNP to discover SMC devices
 	smsc-ircc2.ircc_cfg=	[HW] Device configuration I/O port
 	smsc-ircc2.ircc_sir=	[HW] SIR base I/O port
diff --git a/arch/x86/include/asm/alternative.h b/arch/x86/include/asm/alternative.h
index 7078068..444704c 100644
--- a/arch/x86/include/asm/alternative.h
+++ b/arch/x86/include/asm/alternative.h
@@ -60,7 +60,7 @@ extern void alternatives_smp_module_add(struct module *mod, char *name,
 					void *locks, void *locks_end,
 					void *text, void *text_end);
 extern void alternatives_smp_module_del(struct module *mod);
-extern void alternatives_smp_switch(int smp);
+extern void alternatives_enable_smp(void);
 extern int alternatives_text_reserved(void *start, void *end);
 extern bool skip_smp_alternatives;
 #else
@@ -68,7 +68,7 @@ static inline void alternatives_smp_module_add(struct module *mod, char *name,
 					       void *locks, void *locks_end,
 					       void *text, void *text_end) {}
 static inline void alternatives_smp_module_del(struct module *mod) {}
-static inline void alternatives_smp_switch(int smp) {}
+static inline void alternatives_enable_smp(void) {}
 static inline int alternatives_text_reserved(void *start, void *end)
 {
 	return 0;
diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index ced4534..357475a 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -23,19 +23,6 @@
 
 #define MAX_PATCH_LEN (255-1)
 
-#ifdef CONFIG_HOTPLUG_CPU
-static int smp_alt_once;
-
-static int __init bootonly(char *str)
-{
-	smp_alt_once = 1;
-	return 1;
-}
-__setup("smp-alt-boot", bootonly);
-#else
-#define smp_alt_once 1
-#endif
-
 static int __initdata_or_module debug_alternative;
 
 static int __init debug_alt(char *str)
@@ -326,9 +313,6 @@ static void alternatives_smp_unlock(const s32 *start, const s32 *end,
 {
 	const s32 *poff;
 
-	if (noreplace_smp)
-		return;
-
 	mutex_lock(&text_mutex);
 	for (poff = start; poff < end; poff++) {
 		u8 *ptr = (u8 *)poff + *poff;
@@ -359,7 +343,7 @@ struct smp_alt_module {
 };
 static LIST_HEAD(smp_alt_modules);
 static DEFINE_MUTEX(smp_alt);
-static int smp_mode = 1;	/* protected by smp_alt */
+static bool uniproc_patched = false;	/* protected by smp_alt */
 
 void __init_or_module alternatives_smp_module_add(struct module *mod,
 						  char *name,
@@ -368,19 +352,18 @@ void __init_or_module alternatives_smp_module_add(struct module *mod,
 {
 	struct smp_alt_module *smp;
 
-	if (noreplace_smp)
-		return;
+	mutex_lock(&smp_alt);
+	if (!uniproc_patched)
+		goto unlock;
 
-	if (smp_alt_once) {
-		if (boot_cpu_has(X86_FEATURE_UP))
-			alternatives_smp_unlock(locks, locks_end,
-						text, text_end);
-		return;
-	}
+	if (num_possible_cpus() == 1)
+		/* Don't bother remembering, we'll never have to undo it. */
+		goto smp_unlock;
 
 	smp = kzalloc(sizeof(*smp), GFP_KERNEL);
 	if (NULL == smp)
-		return; /* we'll run the (safe but slow) SMP code then ... */
+		/* we'll run the (safe but slow) SMP code then ... */
+		goto unlock;
 
 	smp->mod	= mod;
 	smp->name	= name;
@@ -392,11 +375,10 @@ void __init_or_module alternatives_smp_module_add(struct module *mod,
 		__func__, smp->locks, smp->locks_end,
 		smp->text, smp->text_end, smp->name);
 
-	mutex_lock(&smp_alt);
 	list_add_tail(&smp->next, &smp_alt_modules);
-	if (boot_cpu_has(X86_FEATURE_UP))
-		alternatives_smp_unlock(smp->locks, smp->locks_end,
-					smp->text, smp->text_end);
+smp_unlock:
+	alternatives_smp_unlock(locks, locks_end, text, text_end);
+unlock:
 	mutex_unlock(&smp_alt);
 }
 
@@ -404,24 +386,18 @@ void __init_or_module alternatives_smp_module_del(struct module *mod)
 {
 	struct smp_alt_module *item;
 
-	if (smp_alt_once || noreplace_smp)
-		return;
-
 	mutex_lock(&smp_alt);
 	list_for_each_entry(item, &smp_alt_modules, next) {
 		if (mod != item->mod)
 			continue;
 		list_del(&item->next);
-		mutex_unlock(&smp_alt);
-		DPRINTK("%s: %s\n", __func__, item->name);
 		kfree(item);
-		return;
+		break;
 	}
 	mutex_unlock(&smp_alt);
 }
 
-bool skip_smp_alternatives;
-void alternatives_smp_switch(int smp)
+void alternatives_enable_smp(void)
 {
 	struct smp_alt_module *mod;
 
@@ -436,34 +412,21 @@ void alternatives_smp_switch(int smp)
 	pr_info("lockdep: fixing up alternatives\n");
 #endif
 
-	if (noreplace_smp || smp_alt_once || skip_smp_alternatives)
-		return;
-	BUG_ON(!smp && (num_online_cpus() > 1));
+	/* Why bother if there are no other CPUs? */
+	BUG_ON(num_possible_cpus() == 1);
 
 	mutex_lock(&smp_alt);
 
-	/*
-	 * Avoid unnecessary switches because it forces JIT based VMs to
-	 * throw away all cached translations, which can be quite costly.
-	 */
-	if (smp == smp_mode) {
-		/* nothing */
-	} else if (smp) {
+	if (uniproc_patched) {
 		pr_info("switching to SMP code\n");
+		BUG_ON(num_online_cpus() != 1);
 		clear_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
 		clear_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
 		list_for_each_entry(mod, &smp_alt_modules, next)
 			alternatives_smp_lock(mod->locks, mod->locks_end,
 					      mod->text, mod->text_end);
-	} else {
-		pr_info("switching to UP code\n");
-		set_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
-		set_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
-		list_for_each_entry(mod, &smp_alt_modules, next)
-			alternatives_smp_unlock(mod->locks, mod->locks_end,
-						mod->text, mod->text_end);
+		uniproc_patched = false;
 	}
-	smp_mode = smp;
 	mutex_unlock(&smp_alt);
 }
 
@@ -540,40 +503,22 @@ void __init alternative_instructions(void)
 
 	apply_alternatives(__alt_instructions, __alt_instructions_end);
 
-	/* switch to patch-once-at-boottime-only mode and free the
-	 * tables in case we know the number of CPUs will never ever
-	 * change */
-#ifdef CONFIG_HOTPLUG_CPU
-	if (num_possible_cpus() < 2)
-		smp_alt_once = 1;
-#endif
-
 #ifdef CONFIG_SMP
-	if (smp_alt_once) {
-		if (1 == num_possible_cpus()) {
-			pr_info("switching to UP code\n");
-			set_cpu_cap(&boot_cpu_data, X86_FEATURE_UP);
-			set_cpu_cap(&cpu_data(0), X86_FEATURE_UP);
-
-			alternatives_smp_unlock(__smp_locks, __smp_locks_end,
-						_text, _etext);
-		}
-	} else {
+	/* Patch to UP if other cpus not imminent. */
+	if (!noreplace_smp && (num_present_cpus() == 1 || setup_max_cpus <= 1)) {
+		uniproc_patched = true;
 		alternatives_smp_module_add(NULL, "core kernel",
 					    __smp_locks, __smp_locks_end,
 					    _text, _etext);
-
-		/* Only switch to UP mode if we don't immediately boot others */
-		if (num_present_cpus() == 1 || setup_max_cpus <= 1)
-			alternatives_smp_switch(0);
 	}
-#endif
- 	apply_paravirt(__parainstructions, __parainstructions_end);
 
-	if (smp_alt_once)
+	if (!uniproc_patched || num_possible_cpus() == 1)
 		free_init_pages("SMP alternatives",
 				(unsigned long)__smp_locks,
 				(unsigned long)__smp_locks_end);
+#endif
+
+	apply_paravirt(__parainstructions, __parainstructions_end);
 
 	restart_nmi();
 }
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 7c5a8c3..c80a33b 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -665,7 +665,8 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu, struct task_struct *idle)
 	unsigned long boot_error = 0;
 	int timeout;
 
-	alternatives_smp_switch(1);
+	/* Just in case we booted with a single CPU. */
+	alternatives_enable_smp();
 
 	idle->thread.sp = (unsigned long) (((struct pt_regs *)
 			  (THREAD_SIZE +  task_stack_page(idle))) - 1);
@@ -1053,20 +1054,6 @@ out:
 	preempt_enable();
 }
 
-void arch_disable_nonboot_cpus_begin(void)
-{
-	/*
-	 * Avoid the smp alternatives switch during the disable_nonboot_cpus().
-	 * In the suspend path, we will be back in the SMP mode shortly anyways.
-	 */
-	skip_smp_alternatives = true;
-}
-
-void arch_disable_nonboot_cpus_end(void)
-{
-	skip_smp_alternatives = false;
-}
-
 void arch_enable_nonboot_cpus_begin(void)
 {
 	set_mtrr_aps_delayed_init();
@@ -1256,9 +1243,6 @@ void native_cpu_die(unsigned int cpu)
 		if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
 			if (system_state == SYSTEM_RUNNING)
 				pr_info("CPU %u is now offline\n", cpu);
-
-			if (1 == num_online_cpus())
-				alternatives_smp_switch(0);
 			return;
 		}
 		msleep(100);
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index f58dca7..353c50f 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -377,7 +377,8 @@ static int __cpuinit xen_cpu_up(unsigned int cpu, struct task_struct *idle)
 		return rc;
 
 	if (num_online_cpus() == 1)
-		alternatives_smp_switch(1);
+		/* Just in case we booted with a single CPU. */
+		alternatives_enable_smp();
 
 	rc = xen_smp_intr_init(cpu);
 	if (rc)
@@ -424,9 +425,6 @@ static void xen_cpu_die(unsigned int cpu)
 	unbind_from_irqhandler(per_cpu(xen_irq_work, cpu), NULL);
 	xen_uninit_lock_cpu(cpu);
 	xen_teardown_timer(cpu);
-
-	if (num_online_cpus() == 1)
-		alternatives_smp_switch(0);
 }
 
 static void __cpuinit xen_play_dead(void) /* used only with HOTPLUG_CPU */
diff --git a/kernel/cpu.c b/kernel/cpu.c
index e615dfb..f560598 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -447,14 +447,6 @@ EXPORT_SYMBOL_GPL(cpu_up);
 #ifdef CONFIG_PM_SLEEP_SMP
 static cpumask_var_t frozen_cpus;
 
-void __weak arch_disable_nonboot_cpus_begin(void)
-{
-}
-
-void __weak arch_disable_nonboot_cpus_end(void)
-{
-}
-
 int disable_nonboot_cpus(void)
 {
 	int cpu, first_cpu, error = 0;
@@ -466,7 +458,6 @@ int disable_nonboot_cpus(void)
 	 * with the userspace trying to use the CPU hotplug at the same time
 	 */
 	cpumask_clear(frozen_cpus);
-	arch_disable_nonboot_cpus_begin();
 
 	printk("Disabling non-boot CPUs ...\n");
 	for_each_online_cpu(cpu) {
@@ -482,8 +473,6 @@ int disable_nonboot_cpus(void)
 		}
 	}
 
-	arch_disable_nonboot_cpus_end();
-
 	if (!error) {
 		BUG_ON(num_online_cpus() > 1);
 		/* Make sure the CPUs won't be enabled by someone else */