[PATCH 1/2] powerpc/ps3: Fix lost SMP IPIs

[PATCH 1/2] powerpc/ps3: Fix lost SMP IPIs

[Geoff Levand]

Fixes the PS3 bootup hang introduced in 3.0-rc1 by:

  commit 317f394160e9beb97d19a84c39b7e5eb3d7815a
  sched: Move the second half of ttwu() to the remote cpu

Move the PS3's LV1 EOI call lv1_end_of_interrupt_ext() from ps3_chip_eoi()
to ps3_get_irq() for IPI messages.

If lv1_send_event_locally() is called between a previous call to
lv1_send_event_locally() and the coresponding call to
lv1_end_of_interrupt_ext() the second event will not be delivered to the
target cpu.

The PS3's SMP IPIs are implemented using lv1_send_event_locally(), so if two
IPI messages of the same type are sent to the same target in a relatively
short period of time the second IPI event can become lost when
lv1_end_of_interrupt_ext() is called from ps3_chip_eoi().

CC: stable@kernel.org
Signed-off-by: Geoff Levand <geoff@infradead.org>
---
 arch/powerpc/platforms/ps3/interrupt.c |   23 ++++++++++++++++++++++-
 arch/powerpc/platforms/ps3/platform.h  |    1 +
 arch/powerpc/platforms/ps3/smp.c       |    2 ++
 3 files changed, 25 insertions(+), 1 deletions(-)

diff --git a/arch/powerpc/platforms/ps3/interrupt.c b/arch/powerpc/platforms/ps3/interrupt.c
index 404bc52..1d6f4f4 100644
--- a/arch/powerpc/platforms/ps3/interrupt.c
+++ b/arch/powerpc/platforms/ps3/interrupt.c
@@ -88,6 +88,7 @@ struct ps3_private {
 	struct ps3_bmp bmp __attribute__ ((aligned (PS3_BMP_MINALIGN)));
 	u64 ppe_id;
 	u64 thread_id;
+	unsigned long ipi_mask;
 };
 
 static DEFINE_PER_CPU(struct ps3_private, ps3_private);
@@ -144,7 +145,11 @@ static void ps3_chip_unmask(struct irq_data *d)
 static void ps3_chip_eoi(struct irq_data *d)
 {
 	const struct ps3_private *pd = irq_data_get_irq_chip_data(d);
-	lv1_end_of_interrupt_ext(pd->ppe_id, pd->thread_id, d->irq);
+
+	/* non-IPIs are EOIed here. */
+
+	if (!test_bit(63 - d->irq, &pd->ipi_mask))
+		lv1_end_of_interrupt_ext(pd->ppe_id, pd->thread_id, d->irq);
 }
 
 /**
@@ -691,6 +696,16 @@ void __init ps3_register_ipi_debug_brk(unsigned int cpu, unsigned int virq)
 		cpu, virq, pd->bmp.ipi_debug_brk_mask);
 }
 
+void __init ps3_register_ipi_irq(unsigned int cpu, unsigned int virq)
+{
+	struct ps3_private *pd = &per_cpu(ps3_private, cpu);
+
+	set_bit(63 - virq, &pd->ipi_mask);
+
+	DBG("%s:%d: cpu %u, virq %u, ipi_mask %lxh\n", __func__, __LINE__,
+		cpu, virq, pd->ipi_mask);
+}
+
 static unsigned int ps3_get_irq(void)
 {
 	struct ps3_private *pd = &__get_cpu_var(ps3_private);
@@ -720,6 +735,12 @@ static unsigned int ps3_get_irq(void)
 		BUG();
 	}
 #endif
+
+	/* IPIs are EOIed here. */
+
+	if (test_bit(63 - plug, &pd->ipi_mask))
+		lv1_end_of_interrupt_ext(pd->ppe_id, pd->thread_id, plug);
+
 	return plug;
 }
 
diff --git a/arch/powerpc/platforms/ps3/platform.h b/arch/powerpc/platforms/ps3/platform.h
index 9a196a8..1a633ed 100644
--- a/arch/powerpc/platforms/ps3/platform.h
+++ b/arch/powerpc/platforms/ps3/platform.h
@@ -43,6 +43,7 @@ void ps3_mm_shutdown(void);
 void ps3_init_IRQ(void);
 void ps3_shutdown_IRQ(int cpu);
 void __init ps3_register_ipi_debug_brk(unsigned int cpu, unsigned int virq);
+void __init ps3_register_ipi_irq(unsigned int cpu, unsigned int virq);
 
 /* smp */
 
diff --git a/arch/powerpc/platforms/ps3/smp.c b/arch/powerpc/platforms/ps3/smp.c
index 4c44794..f609345 100644
--- a/arch/powerpc/platforms/ps3/smp.c
+++ b/arch/powerpc/platforms/ps3/smp.c
@@ -94,6 +94,8 @@ static void __init ps3_smp_setup_cpu(int cpu)
 
 		if (result)
 			virqs[i] = NO_IRQ;
+		else
+			ps3_register_ipi_irq(cpu, virqs[i]);
 	}
 
 	ps3_register_ipi_debug_brk(cpu, virqs[PPC_MSG_DEBUGGER_BREAK]);
-- 
1.7.0.4

Re: [PATCH 1/2] powerpc/ps3: Fix lost SMP IPIs

[Peter Zijlstra]

On Tue, 2011-11-08 at 14:37 -0800, Geoff Levand wrote:
> Fixes the PS3 bootup hang introduced in 3.0-rc1 by:
>=20
>   commit 317f394160e9beb97d19a84c39b7e5eb3d7815a
>   sched: Move the second half of ttwu() to the remote cpu
>=20
> Move the PS3's LV1 EOI call lv1_end_of_interrupt_ext() from ps3_chip_eoi(=
)
> to ps3_get_irq() for IPI messages.
>=20
> If lv1_send_event_locally() is called between a previous call to
> lv1_send_event_locally() and the coresponding call to
> lv1_end_of_interrupt_ext() the second event will not be delivered to the
> target cpu.
>=20
> The PS3's SMP IPIs are implemented using lv1_send_event_locally(), so if =
two
> IPI messages of the same type are sent to the same target in a relatively
> short period of time the second IPI event can become lost when
> lv1_end_of_interrupt_ext() is called from ps3_chip_eoi().=20

Ah glad you found it! I'm not going to ack it since PPC guts are way
beyond me.