[Bionic][PATCH 1/1] powerpc/64s: Fix lost pending interrupt due to race causing lost update to irq_happened

Joseph Salisbury joseph.salisbury at canonical.com
Tue Apr 3 17:20:14 UTC 2018

From: Nicholas Piggin <npiggin at gmail.com>

BugLink: http://bugs.launchpad.net/bugs/1757497

force_external_irq_replay() can be called in the do_IRQ path with
interrupts hard enabled and soft disabled if may_hard_irq_enable() set
MSR[EE]=1. It updates local_paca->irq_happened with a load, modify,
store sequence. If a maskable interrupt hits during this sequence, it
will go to the masked handler to be marked pending in irq_happened.
This update will be lost when the interrupt returns and the store
instruction executes. This can result in unpredictable latencies,
timeouts, lockups, etc.

Fix this by ensuring hard interrupts are disabled before modifying

This could cause any maskable asynchronous interrupt to get lost, but
it was noticed on P9 SMP system doing RDMA NVMe target over 100GbE,
so very high external interrupt rate and high IPI rate. The hang was
bisected down to enabling doorbell interrupts for IPIs. These provided
an interrupt type that could run at high rates in the do_IRQ path,
stressing the race.

Fixes: 1d607bb3bd60 ("powerpc/irq: Add mechanism to force a replay of interrupts")
Cc: stable at vger.kernel.org # v4.8+
Reported-by: Carol L. Soto <clsoto at us.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin at gmail.com>
Signed-off-by: Michael Ellerman <mpe at ellerman.id.au>
(cherry picked from commit ff6781fd1bb404d8a551c02c35c70cec1da17ff1)
Signed-off-by: Joseph Salisbury <joseph.salisbury at canonical.com>
 arch/powerpc/kernel/irq.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index b7a8452..c933632 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -475,6 +475,14 @@ void force_external_irq_replay(void)
+	/*
+	 * Interrupts must always be hard disabled before irq_happened is
+	 * modified (to prevent lost update in case of interrupt between
+	 * load and store).
+	 */
+	__hard_irq_disable();
+	local_paca->irq_happened |= PACA_IRQ_HARD_DIS;
 	/* Indicate in the PACA that we have an interrupt to replay */
 	local_paca->irq_happened |= PACA_IRQ_EE;

More information about the kernel-team mailing list