[PATCH] arm: fix a migrating irq bug when hotplug cpu

[PATCH] arm: fix a migrating irq bug when hotplug cpu

[Dietmar Eggemann]

Arm TC2 fails cpu hotplug stress test.

This issue was tracked down to a missing copy of the new affinity
cpumask for the vexpress-spc interrupt into struct
irq_common_data.affinity when the interrupt is migrated in
migrate_one_irq().

Fix it by replacing the arm specific hotplug cpu migration with the
generic irq code.

This is the counterpart implementation to commit 217d453d473c ("arm64:
fix a migrating irq bug when hotplug cpu").

Tested with cpu hotplug stress test on Arm TC2 (multi_v7_defconfig plus
CONFIG_ARM_BIG_LITTLE_CPUFREQ=y and CONFIG_ARM_VEXPRESS_SPC_CPUFREQ=y).
The vexpress-spc interrupt (irq=22) on this board is affine to CPU0.
Its affinity cpumask now changes correctly e.g. from 0 to 1-4 when
CPU0 is hotplugged out.

Suggested-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
---
 arch/arm/Kconfig           |  1 +
 arch/arm/include/asm/irq.h |  1 -
 arch/arm/kernel/irq.c      | 62 ----------------------------------------------
 arch/arm/kernel/smp.c      |  2 +-
 4 files changed, 2 insertions(+), 64 deletions(-)

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index 664e918e2624..26524b75970a 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -1400,6 +1400,7 @@ config NR_CPUS
 config HOTPLUG_CPU
 	bool "Support for hot-pluggable CPUs"
 	depends on SMP
+	select GENERIC_IRQ_MIGRATION
 	help
 	  Say Y here to experiment with turning CPUs off and on.  CPUs
 	  can be controlled through /sys/devices/system/cpu.
diff --git a/arch/arm/include/asm/irq.h b/arch/arm/include/asm/irq.h
index c883fcbe93b6..46d41140df27 100644
--- a/arch/arm/include/asm/irq.h
+++ b/arch/arm/include/asm/irq.h
@@ -25,7 +25,6 @@
 #ifndef __ASSEMBLY__
 struct irqaction;
 struct pt_regs;
-extern void migrate_irqs(void);
 
 extern void asm_do_IRQ(unsigned int, struct pt_regs *);
 void handle_IRQ(unsigned int, struct pt_regs *);
diff --git a/arch/arm/kernel/irq.c b/arch/arm/kernel/irq.c
index 9908dacf9229..844861368cd5 100644
--- a/arch/arm/kernel/irq.c
+++ b/arch/arm/kernel/irq.c
@@ -31,7 +31,6 @@
 #include <linux/smp.h>
 #include <linux/init.h>
 #include <linux/seq_file.h>
-#include <linux/ratelimit.h>
 #include <linux/errno.h>
 #include <linux/list.h>
 #include <linux/kallsyms.h>
@@ -109,64 +108,3 @@ int __init arch_probe_nr_irqs(void)
 	return nr_irqs;
 }
 #endif
-
-#ifdef CONFIG_HOTPLUG_CPU
-static bool migrate_one_irq(struct irq_desc *desc)
-{
-	struct irq_data *d = irq_desc_get_irq_data(desc);
-	const struct cpumask *affinity = irq_data_get_affinity_mask(d);
-	struct irq_chip *c;
-	bool ret = false;
-
-	/*
-	 * If this is a per-CPU interrupt, or the affinity does not
-	 * include this CPU, then we have nothing to do.
-	 */
-	if (irqd_is_per_cpu(d) || !cpumask_test_cpu(smp_processor_id(), affinity))
-		return false;
-
-	if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids) {
-		affinity = cpu_online_mask;
-		ret = true;
-	}
-
-	c = irq_data_get_irq_chip(d);
-	if (!c->irq_set_affinity)
-		pr_debug("IRQ%u: unable to set affinity\n", d->irq);
-	else if (c->irq_set_affinity(d, affinity, false) == IRQ_SET_MASK_OK && ret)
-		cpumask_copy(irq_data_get_affinity_mask(d), affinity);
-
-	return ret;
-}
-
-/*
- * The current CPU has been marked offline.  Migrate IRQs off this CPU.
- * If the affinity settings do not allow other CPUs, force them onto any
- * available CPU.
- *
- * Note: we must iterate over all IRQs, whether they have an attached
- * action structure or not, as we need to get chained interrupts too.
- */
-void migrate_irqs(void)
-{
-	unsigned int i;
-	struct irq_desc *desc;
-	unsigned long flags;
-
-	local_irq_save(flags);
-
-	for_each_irq_desc(i, desc) {
-		bool affinity_broken;
-
-		raw_spin_lock(&desc->lock);
-		affinity_broken = migrate_one_irq(desc);
-		raw_spin_unlock(&desc->lock);
-
-		if (affinity_broken)
-			pr_warn_ratelimited("IRQ%u no longer affine to CPU%u\n",
-				i, smp_processor_id());
-	}
-
-	local_irq_restore(flags);
-}
-#endif /* CONFIG_HOTPLUG_CPU */
diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c
index 3bf82232b1be..1d6f5ea522f4 100644
--- a/arch/arm/kernel/smp.c
+++ b/arch/arm/kernel/smp.c
@@ -254,7 +254,7 @@ int __cpu_disable(void)
 	/*
 	 * OK - migrate IRQs away from this CPU
 	 */
-	migrate_irqs();
+	irq_migrate_all_off_this_cpu();
 
 	/*
 	 * Flush user cache and TLB mappings, and then remove this CPU
-- 
2.11.0

Re: [PATCH] arm: fix a migrating irq bug when hotplug cpu

[Marc Zyngier]

On 10/01/2019 10:09, Dietmar Eggemann wrote:
> Arm TC2 fails cpu hotplug stress test.
> 
> This issue was tracked down to a missing copy of the new affinity
> cpumask for the vexpress-spc interrupt into struct
> irq_common_data.affinity when the interrupt is migrated in
> migrate_one_irq().
> 
> Fix it by replacing the arm specific hotplug cpu migration with the
> generic irq code.
> 
> This is the counterpart implementation to commit 217d453d473c ("arm64:
> fix a migrating irq bug when hotplug cpu").
> 
> Tested with cpu hotplug stress test on Arm TC2 (multi_v7_defconfig plus
> CONFIG_ARM_BIG_LITTLE_CPUFREQ=y and CONFIG_ARM_VEXPRESS_SPC_CPUFREQ=y).
> The vexpress-spc interrupt (irq=22) on this board is affine to CPU0.
> Its affinity cpumask now changes correctly e.g. from 0 to 1-4 when
> CPU0 is hotplugged out.
> 
> Suggested-by: Marc Zyngier <marc.zyngier@arm.com>
> Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>

Acked-by: Marc Zyngier <marc.zyngier@arm.com>

	M.
-- 
Jazz is not dead. It just smells funny...

Re: [PATCH] arm: fix a migrating irq bug when hotplug cpu

[Linus Walleij]

On Thu, Jan 10, 2019 at 11:09 AM Dietmar Eggemann
<dietmar.eggemann@arm.com> wrote:
>
> Arm TC2 fails cpu hotplug stress test.
>
> This issue was tracked down to a missing copy of the new affinity
> cpumask for the vexpress-spc interrupt into struct
> irq_common_data.affinity when the interrupt is migrated in
> migrate_one_irq().
>
> Fix it by replacing the arm specific hotplug cpu migration with the
> generic irq code.
>
> This is the counterpart implementation to commit 217d453d473c ("arm64:
> fix a migrating irq bug when hotplug cpu").
>
> Tested with cpu hotplug stress test on Arm TC2 (multi_v7_defconfig plus
> CONFIG_ARM_BIG_LITTLE_CPUFREQ=y and CONFIG_ARM_VEXPRESS_SPC_CPUFREQ=y).
> The vexpress-spc interrupt (irq=22) on this board is affine to CPU0.
> Its affinity cpumask now changes correctly e.g. from 0 to 1-4 when
> CPU0 is hotplugged out.
>
> Suggested-by: Marc Zyngier <marc.zyngier@arm.com>
> Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>

Looks solid to me(TM)

I would proceed to put it into Russell's patch tracker along with Marc's
ACK so Russell can process it, see:
http://www.arm.linux.org.uk/developer/patches/

Yours,
Linus Walleij

[PATCH] ARM: fix a migrating irq bug when hotplug cpu

[Yang Yingliang]

From: Yang Yingliang <yangyingliang@huawei.com>

When cpu is disabled, all irqs will be migratged to another cpu.
In some cases, a new affinity is different, it needed to be coppied
to irq's affinity. But if the type of irq is LPI, it's affinity will
not be coppied because of irq_set_affinity's return value.
So copy the affinity, when the return value is IRQ_SET_MASK_OK_DONE.

Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Russell King - ARM Linux <linux@arm.linux.org.uk>
---
  arch/arm/kernel/irq.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/arm/kernel/irq.c b/arch/arm/kernel/irq.c
index 350f188..32f2140 100644
--- a/arch/arm/kernel/irq.c
+++ b/arch/arm/kernel/irq.c
@@ -157,10 +157,13 @@ static bool migrate_one_irq(struct irq_desc *desc)
  	}

  	c = irq_data_get_irq_chip(d);
-	if (!c->irq_set_affinity)
+	if (!c->irq_set_affinity) {
  		pr_debug("IRQ%u: unable to set affinity\n", d->irq);
-	else if (c->irq_set_affinity(d, affinity, false) == IRQ_SET_MASK_OK && 
ret)
-		cpumask_copy(d->affinity, affinity);
+	} else {
+		int r = c->irq_set_affinity(d, affinity, false);
+		if ((r == IRQ_SET_MASK_OK || r == IRQ_SET_MASK_OK_DONE) && ret)
+			cpumask_copy(d->affinity, affinity);
+	}

  	return ret;
  }
-- 
2.5.0