genirq/exynos timers: Fix the CPU hotplug wreckage sanely

PATCH In review Thomas Gleixner 3 participants 11 messages Apr 16, 2014 – Apr 17, 2014

Subsystems (7)

Stable branch MIPS Arm Samsung Exynos irqchip drivers +3 more subsystems

lore

Share & tools

lore

Thomas Gleixner

author

2014-04-16 14:36 UTC #0 lore raw

cover genirq/exynos timers: Fix the CPU hotplug wreckage sanely

The current implementation of irq_set_affinity() refuses rightfully to route an interrupt to an offline cpu.

But there is a special case, where this is actually desired. Some of the ARM SoCs have per cpu timers which require setting the affinity during cpu startup where the cpu is not yet in the online mask.

If we can't do that, then the local timer interrupt for the about to become online cpu is routed to some random online cpu.

The developers of the affected machines tried to work around that issue, but that results in a massive mess in that timer code.

It's saner to provide a facility to force the affinity and make the affected machines use that.

The following series implements that logic w/o impact on any existing users.

The change to the genirq core code is not that bad:

 arch/mips/cavium-octeon/octeon-irq.c |    2 +-
 include/linux/interrupt.h            |   35 ++++++++++++++++++++++++++++++++++-
 include/linux/irq.h                  |    3 ++-
 kernel/irq/manage.c                  |   17 ++++++-----------
 4 files changed, 43 insertions(+), 14 deletions(-)

The resulting fixup for gic/exynos_mct is:

 clocksource/exynos_mct.c |   12 +++---------
 irqchip/irq-gic.c        |    8 ++++++--
 2 files changed, 9 insertions(+), 11 deletions(-)

Krzysztofs proposed workaround was slightly smaller than that, but I prefer having a clean solution for backporting to stable rather than a messy hack around which works.

@Krzysztof: Can you please retest the series? I've changed the core implementation versus the first attempt to make it less intrusive.

Thanks,

tglx

Thomas Gleixner

author

2014-04-16 14:36 UTC #1 lore raw

PATCH 1/4 genirq: Allow forcing cpu affinity of interrupts

The current implementation of irq_set_affinity() refuses rightfully to route an interrupt to an offline cpu.

If we can't do that, then the local timer interrupt for the about to become online cpu is routed to some random online cpu.

The developers of the affected machines tried to work around that issue, but that results in a massive mess in that timer code.

We have a yet unused argument in the set_affinity callbacks of the irq chips, which I added back then for a similar reason. It was never required so it got not used. But I'm happy that I never removed it.

That allows us to implement a sane handling of the above scenario. So the affected SoC drivers can add the required force handling to their interrupt chip, switch the timer code to irq_force_affinity() and things just work.

This does not affect any existing user of irq_set_affinity().

Signed-off-by Thomas Gleixner <[email protected]>

---
 arch/mips/cavium-octeon/octeon-irq.c |    2 +-
 include/linux/interrupt.h            |   35 ++++++++++++++++++++++++++++++++++-
 include/linux/irq.h                  |    3 ++-
 kernel/irq/manage.c                  |   17 ++++++-----------
 4 files changed, 43 insertions(+), 14 deletions(-)

arch/mips/cavium-octeon/octeon-irq.c +1 -1

Index: tip/arch/mips/cavium-octeon/octeon-irq.c
===================================================================
@@ -635,7 +635,7 @@ static void octeon_irq_cpu_offline_ciu(s
 		cpumask_clear(&new_affinity);
 		cpumask_set_cpu(cpumask_first(cpu_online_mask), &new_affinity);
 	}
-	__irq_set_affinity_locked(data, &new_affinity);
+	irq_set_affinity_locked(data, &new_affinity, false);
 }
 
 static int octeon_irq_ciu_set_affinity(struct irq_data *data,

include/linux/interrupt.h +34 -1

Index: tip/include/linux/interrupt.h
===================================================================
@@ -203,7 +203,40 @@ static inline int check_wakeup_irqs(void
 
 extern cpumask_var_t irq_default_affinity;
 
-extern int irq_set_affinity(unsigned int irq, const struct cpumask *cpumask);
+/* Internal implementation. Use the helpers below */
+extern int __irq_set_affinity(unsigned int irq, const struct cpumask *cpumask,
+			      bool force);
+
+/**
+ * irq_set_affinity - Set the irq affinity of a given irq
+ * @irq:	Interrupt to set affinity
+ * @mask:	cpumask
+ *
+ * Fails if cpumask does not contain an online CPU
+ */
+static inline int
+irq_set_affinity(unsigned int irq, const struct cpumask *cpumask)
+{
+	return __irq_set_affinity(irq, cpumask, false);
+}
+
+/**
+ * irq_force_affinity - Force the irq affinity of a given irq
+ * @irq:	Interrupt to set affinity
+ * @mask:	cpumask
+ *
+ * Same as irq_set_affinity, but without checking the mask against
+ * online cpus.
+ *
+ * Solely for low level cpu hotplug code, where we need to make per
+ * cpu interrupts affine before the cpu becomes online.
+ */
+static inline int
+irq_force_affinity(unsigned int irq, const struct cpumask *cpumask)
+{
+	return __irq_set_affinity(irq, cpumask, true);
+}
+
 extern int irq_can_set_affinity(unsigned int irq);
 extern int irq_select_affinity(unsigned int irq);

include/linux/irq.h +2 -1

Index: tip/include/linux/irq.h
===================================================================
@@ -394,7 +394,8 @@ extern void remove_percpu_irq(unsigned i
 
 extern void irq_cpu_online(void);
 extern void irq_cpu_offline(void);
-extern int __irq_set_affinity_locked(struct irq_data *data,  const struct cpumask *cpumask);
+extern int irq_set_affinity_locked(struct irq_data *data,
+				   const struct cpumask *cpumask, bool force);
 
 #if defined(CONFIG_SMP) && defined(CONFIG_GENERIC_PENDING_IRQ)
 void irq_move_irq(struct irq_data *data);

kernel/irq/manage.c +6 ~~-11~~

Index: tip/kernel/irq/manage.c
===================================================================
@@ -180,7 +180,7 @@ int irq_do_set_affinity(struct irq_data
 	struct irq_chip *chip = irq_data_get_irq_chip(data);
 	int ret;
 
-	ret = chip->irq_set_affinity(data, mask, false);
+	ret = chip->irq_set_affinity(data, mask, force);
 	switch (ret) {
 	case IRQ_SET_MASK_OK:
 		cpumask_copy(data->affinity, mask);

@@ -192,7 +192,8 @@ int irq_do_set_affinity(struct irq_data
 	return ret;
 }
 
-int __irq_set_affinity_locked(struct irq_data *data, const struct cpumask *mask)
+int irq_set_affinity_locked(struct irq_data *data, const struct cpumask *mask,
+			    bool force)
 {
 	struct irq_chip *chip = irq_data_get_irq_chip(data);
 	struct irq_desc *desc = irq_data_to_desc(data);

@@ -202,7 +203,7 @@ int __irq_set_affinity_locked(struct irq
 		return -EINVAL;
 
 	if (irq_can_move_pcntxt(data)) {
-		ret = irq_do_set_affinity(data, mask, false);
+		ret = irq_do_set_affinity(data, mask, force);
 	} else {
 		irqd_set_move_pending(data);
 		irq_copy_pending(desc, mask);

@@ -217,13 +218,7 @@ int __irq_set_affinity_locked(struct irq
 	return ret;
 }
 
-/**
- *	irq_set_affinity - Set the irq affinity of a given irq
- *	@irq:		Interrupt to set affinity
- *	@mask:		cpumask
- *
- */
-int irq_set_affinity(unsigned int irq, const struct cpumask *mask)
+int __irq_set_affinity(unsigned int irq, const struct cpumask *mask, bool force)
 {
 	struct irq_desc *desc = irq_to_desc(irq);
 	unsigned long flags;

@@ -233,7 +228,7 @@ int irq_set_affinity(unsigned int irq, c
 		return -EINVAL;
 
 	raw_spin_lock_irqsave(&desc->lock, flags);
-	ret =  __irq_set_affinity_locked(irq_desc_get_irq_data(desc), mask);
+	ret = irq_set_affinity_locked(irq_desc_get_irq_data(desc), mask, force);
 	raw_spin_unlock_irqrestore(&desc->lock, flags);
 	return ret;
 }

tip-bot for Thomas Gleixner

merge bot

2014-04-17 21:40 UTC #2 lore raw

[tip:irq/urgent] genirq: Allow forcing cpu affinity of interrupts

genirq: Allow forcing cpu affinity of interrupts

Commit-ID:  01f8fa4f01d8362358eb90e412bd7ae18a3ec1ad
Gitweb:     http://git.kernel.org/tip/01f8fa4f01d8362358eb90e412bd7ae18a3ec1ad
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Wed, 16 Apr 2014 14:36:44 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Thu, 17 Apr 2014 23:36:27 +0200

genirq: Allow forcing cpu affinity of interrupts

The current implementation of irq_set_affinity() refuses rightfully to route an interrupt to an offline cpu.

If we can't do that, then the local timer interrupt for the about to become online cpu is routed to some random online cpu.

The developers of the affected machines tried to work around that issue, but that results in a massive mess in that timer code.

This does not affect any existing user of irq_set_affinity().

Tagged for stable to allow a simple fix of the affected SoC clock event drivers.

Reported-and-tested-by Krzysztof Kozlowski <[email protected]>

Signed-off-by Thomas Gleixner <[email protected]>

Cc Kyungmin Park <[email protected]>

Cc Marek Szyprowski <[email protected]>

Cc Bartlomiej Zolnierkiewicz <[email protected]>

Cc Tomasz Figa <[email protected]>,

Cc Daniel Lezcano <[email protected]>,

Cc Kukjin Kim <[email protected]>

Cc [email protected],

Cc [email protected]

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Thomas Gleixner <[email protected]>

---
 arch/mips/cavium-octeon/octeon-irq.c |  2 +-
 include/linux/interrupt.h            | 35 ++++++++++++++++++++++++++++++++++-
 include/linux/irq.h                  |  3 ++-
 kernel/irq/manage.c                  | 17 ++++++-----------
 4 files changed, 43 insertions(+), 14 deletions(-)

arch/mips/cavium-octeon/octeon-irq.c +1 -1

index c2bb4f8..3aa5b46 100644
@@ -635,7 +635,7 @@ static void octeon_irq_cpu_offline_ciu(struct irq_data *data)
 		cpumask_clear(&new_affinity);
 		cpumask_set_cpu(cpumask_first(cpu_online_mask), &new_affinity);
 	}
-	__irq_set_affinity_locked(data, &new_affinity);
+	irq_set_affinity_locked(data, &new_affinity, false);
 }
 
 static int octeon_irq_ciu_set_affinity(struct irq_data *data,

include/linux/interrupt.h +34 -1

index c7bfac1..8834a7e 100644
@@ -203,7 +203,40 @@ static inline int check_wakeup_irqs(void) { return 0; }
 
 extern cpumask_var_t irq_default_affinity;
 
-extern int irq_set_affinity(unsigned int irq, const struct cpumask *cpumask);
+/* Internal implementation. Use the helpers below */
+extern int __irq_set_affinity(unsigned int irq, const struct cpumask *cpumask,
+			      bool force);
+
+/**
+ * irq_set_affinity - Set the irq affinity of a given irq
+ * @irq:	Interrupt to set affinity
+ * @mask:	cpumask
+ *
+ * Fails if cpumask does not contain an online CPU
+ */
+static inline int
+irq_set_affinity(unsigned int irq, const struct cpumask *cpumask)
+{
+	return __irq_set_affinity(irq, cpumask, false);
+}
+
+/**
+ * irq_force_affinity - Force the irq affinity of a given irq
+ * @irq:	Interrupt to set affinity
+ * @mask:	cpumask
+ *
+ * Same as irq_set_affinity, but without checking the mask against
+ * online cpus.
+ *
+ * Solely for low level cpu hotplug code, where we need to make per
+ * cpu interrupts affine before the cpu becomes online.
+ */
+static inline int
+irq_force_affinity(unsigned int irq, const struct cpumask *cpumask)
+{
+	return __irq_set_affinity(irq, cpumask, true);
+}
+
 extern int irq_can_set_affinity(unsigned int irq);
 extern int irq_select_affinity(unsigned int irq);

include/linux/irq.h +2 -1

index d278838..10a0b1a 100644
@@ -394,7 +394,8 @@ extern void remove_percpu_irq(unsigned int irq, struct irqaction *act);
 
 extern void irq_cpu_online(void);
 extern void irq_cpu_offline(void);
-extern int __irq_set_affinity_locked(struct irq_data *data,  const struct cpumask *cpumask);
+extern int irq_set_affinity_locked(struct irq_data *data,
+				   const struct cpumask *cpumask, bool force);
 
 #if defined(CONFIG_SMP) && defined(CONFIG_GENERIC_PENDING_IRQ)
 void irq_move_irq(struct irq_data *data);

kernel/irq/manage.c +6 ~~-11~~

index 2486a4c..d34131c 100644
@@ -180,7 +180,7 @@ int irq_do_set_affinity(struct irq_data *data, const struct cpumask *mask,
 	struct irq_chip *chip = irq_data_get_irq_chip(data);
 	int ret;
 
-	ret = chip->irq_set_affinity(data, mask, false);
+	ret = chip->irq_set_affinity(data, mask, force);
 	switch (ret) {
 	case IRQ_SET_MASK_OK:
 		cpumask_copy(data->affinity, mask);

@@ -192,7 +192,8 @@ int irq_do_set_affinity(struct irq_data *data, const struct cpumask *mask,
 	return ret;
 }
 
-int __irq_set_affinity_locked(struct irq_data *data, const struct cpumask *mask)
+int irq_set_affinity_locked(struct irq_data *data, const struct cpumask *mask,
+			    bool force)
 {
 	struct irq_chip *chip = irq_data_get_irq_chip(data);
 	struct irq_desc *desc = irq_data_to_desc(data);

@@ -202,7 +203,7 @@ int __irq_set_affinity_locked(struct irq_data *data, const struct cpumask *mask)
 		return -EINVAL;
 
 	if (irq_can_move_pcntxt(data)) {
-		ret = irq_do_set_affinity(data, mask, false);
+		ret = irq_do_set_affinity(data, mask, force);
 	} else {
 		irqd_set_move_pending(data);
 		irq_copy_pending(desc, mask);

@@ -217,13 +218,7 @@ int __irq_set_affinity_locked(struct irq_data *data, const struct cpumask *mask)
 	return ret;
 }
 
-/**
- *	irq_set_affinity - Set the irq affinity of a given irq
- *	@irq:		Interrupt to set affinity
- *	@mask:		cpumask
- *
- */
-int irq_set_affinity(unsigned int irq, const struct cpumask *mask)
+int __irq_set_affinity(unsigned int irq, const struct cpumask *mask, bool force)
 {
 	struct irq_desc *desc = irq_to_desc(irq);
 	unsigned long flags;

@@ -233,7 +228,7 @@ int irq_set_affinity(unsigned int irq, const struct cpumask *mask)
 		return -EINVAL;
 
 	raw_spin_lock_irqsave(&desc->lock, flags);
-	ret =  __irq_set_affinity_locked(irq_desc_get_irq_data(desc), mask);
+	ret = irq_set_affinity_locked(irq_desc_get_irq_data(desc), mask, force);
 	raw_spin_unlock_irqrestore(&desc->lock, flags);
 	return ret;
 }

Thomas Gleixner

author

2014-04-16 14:36 UTC #3 lore raw

PATCH 2/4 irqchip: gic: Support forced affinity setting

In reply to Thomas Gleixner

To support the affinity setting of per cpu timers in the early startup of a not yet online cpu, implement the force logic, which disables the cpu online check.

Tagged for stable to allow a simple fix of the affected SoC clock event drivers.

Signed-off-by Thomas Gleixner <[email protected]>

---
 drivers/irqchip/irq-gic.c |    8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

drivers/irqchip/irq-gic.c +6 -2

Index: tip/drivers/irqchip/irq-gic.c
===================================================================
@@ -246,10 +246,14 @@ static int gic_set_affinity(struct irq_d
 			    bool force)
 {
 	void __iomem *reg = gic_dist_base(d) + GIC_DIST_TARGET + (gic_irq(d) & ~3);
-	unsigned int shift = (gic_irq(d) % 4) * 8;
-	unsigned int cpu = cpumask_any_and(mask_val, cpu_online_mask);
+	unsigned int cpu, shift = (gic_irq(d) % 4) * 8;
 	u32 val, mask, bit;
 
+	if (!force)
+		cpu = cpumask_any_and(mask_val, cpu_online_mask);
+	else
+		cpu = cpumask_first(mask_val);
+
 	if (cpu >= NR_GIC_CPU_IF || cpu >= nr_cpu_ids)
 		return -EINVAL;

tip-bot for Thomas Gleixner

merge bot

2014-04-17 21:40 UTC #4 lore raw

[tip:irq/urgent] irqchip: Gic: Support forced affinity setting

irqchip: Gic: Support forced affinity setting

Commit-ID:  ffde1de64012c406dfdda8690918248b472f24e4
Gitweb:     http://git.kernel.org/tip/ffde1de64012c406dfdda8690918248b472f24e4
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Wed, 16 Apr 2014 14:36:44 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Thu, 17 Apr 2014 23:36:28 +0200

irqchip: Gic: Support forced affinity setting

To support the affinity setting of per cpu timers in the early startup of a not yet online cpu, implement the force logic, which disables the cpu online check.

Tagged for stable to allow a simple fix of the affected SoC clock event drivers.

Signed-off-by Thomas Gleixner <[email protected]>

Tested-by Krzysztof Kozlowski <[email protected]>

Cc Kyungmin Park <[email protected]>

Cc Marek Szyprowski <[email protected]>

Cc Bartlomiej Zolnierkiewicz <[email protected]>

Cc Tomasz Figa <[email protected]>,

Cc Daniel Lezcano <[email protected]>,

Cc Kukjin Kim <[email protected]>

Cc [email protected],

Cc [email protected]

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Thomas Gleixner <[email protected]>

---
 drivers/irqchip/irq-gic.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

drivers/irqchip/irq-gic.c +6 -2

index 4300b66..57d165e 100644
@@ -246,10 +246,14 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
 			    bool force)
 {
 	void __iomem *reg = gic_dist_base(d) + GIC_DIST_TARGET + (gic_irq(d) & ~3);
-	unsigned int shift = (gic_irq(d) % 4) * 8;
-	unsigned int cpu = cpumask_any_and(mask_val, cpu_online_mask);
+	unsigned int cpu, shift = (gic_irq(d) % 4) * 8;
 	u32 val, mask, bit;
 
+	if (!force)
+		cpu = cpumask_any_and(mask_val, cpu_online_mask);
+	else
+		cpu = cpumask_first(mask_val);
+
 	if (cpu >= NR_GIC_CPU_IF || cpu >= nr_cpu_ids)
 		return -EINVAL;

Thomas Gleixner

author

2014-04-16 14:36 UTC #5 lore raw

PATCH 3/4 clocksource: exynos_mct: Use irq_force_affinity() in cpu bringup

In reply to Thomas Gleixner

The starting cpu is not yet in the online mask so irq_set_affinity() fails which results in per cpu timers for this cpu ending up on some other online cpu, ususally cpu 0.

Use irq_force_affinity() which disables the online mask check and makes things work.

Signed-off-by Thomas Gleixner <[email protected]>

---
 drivers/clocksource/exynos_mct.c |    8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

drivers/clocksource/exynos_mct.c +1 -7

Index: tip/drivers/clocksource/exynos_mct.c
===================================================================
@@ -430,6 +430,7 @@ static int exynos4_local_timer_setup(str
 				evt->irq);
 			return -EIO;
 		}
+		irq_force_affinity(mct_irqs[MCT_L0_IRQ + cpu], cpumask_of(cpu));
 	} else {
 		enable_percpu_irq(mct_irqs[MCT_L0_IRQ], 0);
 	}

@@ -450,7 +451,6 @@ static int exynos4_mct_cpu_notify(struct
 					   unsigned long action, void *hcpu)
 {
 	struct mct_clock_event_device *mevt;
-	unsigned int cpu;
 
 	/*
 	 * Grab cpu pointer in each case to avoid spurious

@@ -461,12 +461,6 @@ static int exynos4_mct_cpu_notify(struct
 		mevt = this_cpu_ptr(&percpu_mct_tick);
 		exynos4_local_timer_setup(&mevt->evt);
 		break;
-	case CPU_ONLINE:
-		cpu = (unsigned long)hcpu;
-		if (mct_int_type == MCT_INT_SPI)
-			irq_set_affinity(mct_irqs[MCT_L0_IRQ + cpu],
-						cpumask_of(cpu));
-		break;
 	case CPU_DYING:
 		mevt = this_cpu_ptr(&percpu_mct_tick);
 		exynos4_local_timer_stop(&mevt->evt);

tip-bot for Thomas Gleixner

merge bot

2014-04-17 21:40 UTC #6 lore raw

[tip:irq/urgent] clocksource: Exynos_mct: Use irq_force_affinity( ) in cpu bringup

clocksource: Exynos_mct: Use irq_force_affinity() in cpu bringup

Commit-ID:  30ccf03b4a6a2102a2219058bdc6d779dc637dd7
Gitweb:     http://git.kernel.org/tip/30ccf03b4a6a2102a2219058bdc6d779dc637dd7
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Wed, 16 Apr 2014 14:36:45 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Thu, 17 Apr 2014 23:36:28 +0200

clocksource: Exynos_mct: Use irq_force_affinity() in cpu bringup

The starting cpu is not yet in the online mask so irq_set_affinity() fails which results in per cpu timers for this cpu ending up on some other online cpu, ususally cpu 0.

Use irq_force_affinity() which disables the online mask check and makes things work.

Signed-off-by Thomas Gleixner <[email protected]>

Tested-by Krzysztof Kozlowski <[email protected]>

Cc Kyungmin Park <[email protected]>

Cc Marek Szyprowski <[email protected]>

Cc Bartlomiej Zolnierkiewicz <[email protected]>

Cc Tomasz Figa <[email protected]>,

Cc Daniel Lezcano <[email protected]>,

Cc Kukjin Kim <[email protected]>

Cc [email protected],

Cc [email protected]

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Thomas Gleixner <[email protected]>

---
 drivers/clocksource/exynos_mct.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

drivers/clocksource/exynos_mct.c +1 -7

index a6ee6d7..b2d4163 100644
@@ -430,6 +430,7 @@ static int exynos4_local_timer_setup(struct clock_event_device *evt)
 				evt->irq);
 			return -EIO;
 		}
+		irq_force_affinity(mct_irqs[MCT_L0_IRQ + cpu], cpumask_of(cpu));
 	} else {
 		enable_percpu_irq(mct_irqs[MCT_L0_IRQ], 0);
 	}

@@ -450,7 +451,6 @@ static int exynos4_mct_cpu_notify(struct notifier_block *self,
 					   unsigned long action, void *hcpu)
 {
 	struct mct_clock_event_device *mevt;
-	unsigned int cpu;
 
 	/*
 	 * Grab cpu pointer in each case to avoid spurious

@@ -461,12 +461,6 @@ static int exynos4_mct_cpu_notify(struct notifier_block *self,
 		mevt = this_cpu_ptr(&percpu_mct_tick);
 		exynos4_local_timer_setup(&mevt->evt);
 		break;
-	case CPU_ONLINE:
-		cpu = (unsigned long)hcpu;
-		if (mct_int_type == MCT_INT_SPI)
-			irq_set_affinity(mct_irqs[MCT_L0_IRQ + cpu],
-						cpumask_of(cpu));
-		break;
 	case CPU_DYING:
 		mevt = this_cpu_ptr(&percpu_mct_tick);
 		exynos4_local_timer_stop(&mevt->evt);

Thomas Gleixner

author

2014-04-16 14:36 UTC #7 lore raw

PATCH 4/4 clocksource: exynos_mct: Register clock event after request_irq()

In reply to Thomas Gleixner

From: Krzysztof Kozlowski <[email protected]>

After hotplugging CPU1 the first call of interrupt handler for CPU1 oneshot timer was called on CPU0 because it fired before setting IRQ affinity. Affected are SoCs where Multi Core Timer interrupts are shared (SPI), e.g. Exynos 4210.

During setup of the MCT timers the clock event device should be registered after setting the affinity for interrupt. This will prevent starting the timer too early.

Signed-off-by Krzysztof Kozlowski <[email protected]>

Signed-off-by Thomas Gleixner <[email protected]>

---
 drivers/clocksource/exynos_mct.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/clocksource/exynos_mct.c +2 -2

Index: tip/drivers/clocksource/exynos_mct.c
===================================================================
@@ -416,8 +416,6 @@ static int exynos4_local_timer_setup(str
 	evt->set_mode = exynos4_tick_set_mode;
 	evt->features = CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT;
 	evt->rating = 450;
-	clockevents_config_and_register(evt, clk_rate / (TICK_BASE_CNT + 1),
-					0xf, 0x7fffffff);
 
 	exynos4_mct_write(TICK_BASE_CNT, mevt->base + MCT_L_TCNTB_OFFSET);

@@ -434,6 +432,8 @@ static int exynos4_local_timer_setup(str
 	} else {
 		enable_percpu_irq(mct_irqs[MCT_L0_IRQ], 0);
 	}
+	clockevents_config_and_register(evt, clk_rate / (TICK_BASE_CNT + 1),
+					0xf, 0x7fffffff);
 
 	return 0;
 }

tip-bot for Krzysztof Kozlowski

merge bot

2014-04-17 21:40 UTC #8 lore raw

[tip:irq/urgent] clocksource: Exynos_mct: Register clock event after request_irq()

clocksource: Exynos_mct: Register clock event after request_irq()

Commit-ID:  8db6e5104b77de5d0b7002b95069da0992a34be9
Gitweb:     http://git.kernel.org/tip/8db6e5104b77de5d0b7002b95069da0992a34be9
Author:     Krzysztof Kozlowski <[email protected]>
AuthorDate: Wed, 16 Apr 2014 14:36:45 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Thu, 17 Apr 2014 23:36:28 +0200

clocksource: Exynos_mct: Register clock event after request_irq()

During setup of the MCT timers the clock event device should be registered after setting the affinity for interrupt. This will prevent starting the timer too early.

Signed-off-by Krzysztof Kozlowski <[email protected]>

Signed-off-by Thomas Gleixner <[email protected]>

Cc Kyungmin Park <[email protected]>

Cc Marek Szyprowski <[email protected]>

Cc Bartlomiej Zolnierkiewicz <[email protected]>

Cc Tomasz Figa <[email protected]>,

Cc Daniel Lezcano <[email protected]>,

Cc Kukjin Kim <[email protected]>

Cc [email protected],

Cc [email protected]

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Thomas Gleixner <[email protected]>

---
 drivers/clocksource/exynos_mct.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/clocksource/exynos_mct.c +2 -2

index b2d4163..acf5a32 100644
@@ -416,8 +416,6 @@ static int exynos4_local_timer_setup(struct clock_event_device *evt)
 	evt->set_mode = exynos4_tick_set_mode;
 	evt->features = CLOCK_EVT_FEAT_PERIODIC | CLOCK_EVT_FEAT_ONESHOT;
 	evt->rating = 450;
-	clockevents_config_and_register(evt, clk_rate / (TICK_BASE_CNT + 1),
-					0xf, 0x7fffffff);
 
 	exynos4_mct_write(TICK_BASE_CNT, mevt->base + MCT_L_TCNTB_OFFSET);

@@ -434,6 +432,8 @@ static int exynos4_local_timer_setup(struct clock_event_device *evt)
 	} else {
 		enable_percpu_irq(mct_irqs[MCT_L0_IRQ], 0);
 	}
+	clockevents_config_and_register(evt, clk_rate / (TICK_BASE_CNT + 1),
+					0xf, 0x7fffffff);
 
 	return 0;
 }

Krzysztof Kozlowski

2014-04-16 15:59 UTC #9 lore raw

In reply to Thomas Gleixner

Quoted 33 lines · Thomas Gleixner

The current implementation of irq_set_affinity() refuses rightfully to route an interrupt to an offline cpu.
But there is a special case, where this is actually desired. Some of the ARM SoCs have per cpu timers which require setting the affinity during cpu startup where the cpu is not yet in the online mask.
If we can't do that, then the local timer interrupt for the about to become online cpu is routed to some random online cpu.
The developers of the affected machines tried to work around that issue, but that results in a massive mess in that timer code.
It's saner to provide a facility to force the affinity and make the affected machines use that.
The following series implements that logic w/o impact on any existing users.
The change to the genirq core code is not that bad:
 arch/mips/cavium-octeon/octeon-irq.c |    2 +-
 include/linux/interrupt.h            |   35 ++++++++++++++++++++++++++++++++++-
 include/linux/irq.h                  |    3 ++-
 kernel/irq/manage.c                  |   17 ++++++-----------
 4 files changed, 43 insertions(+), 14 deletions(-)
The resulting fixup for gic/exynos_mct is:
 clocksource/exynos_mct.c |   12 +++---------
 irqchip/irq-gic.c        |    8 ++++++--
 2 files changed, 9 insertions(+), 11 deletions(-)

Quoted 6 lines

Krzysztofs proposed workaround was slightly smaller than that, but I prefer having a clean solution for backporting to stable rather than a messy hack around which works.
@Krzysztof: Can you please retest the series? I've changed the core implementation versus the first attempt to make it less intrusive.

Works fine. Tested whole patchset on board with Exynos 3250.

Best regards,
Krzysztof

Thomas Gleixner

author

2014-04-16 21:16 UTC #10 lore raw

Quoted 5 lines · Krzysztof Kozlowski

Quoted 2 lines · Thomas Gleixner
@Krzysztof: Can you please retest the series? I've changed the core implementation versus the first attempt to make it less intrusive.
Works fine. Tested whole patchset on board with Exynos 3250.

Great! Now I just need to sell it :)