Writing to the EOIR register before masking the HW mapped timer interrupt
can cause taking another timer interrupt immediatly after exception return.
This doesn't happen all the time, because KVM reevaluates the state of
pending HW mapped level sensitive interrupts on each guest exit. If a guest
exit occurs after masking the timer interrupt, but before the ERET, when
the extra interrupt is pending, then KVM will remove it.
Move the write after the IMASK bit has been set to prevent this from
happening.
Signed-off-by: Alexandru Elisei <alexandru.elisei@xxxxxxx>
---
arm/timer.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/arm/timer.c b/arm/timer.c
index 78f0dd870993..7ae169bd687e 100644
--- a/arm/timer.c
+++ b/arm/timer.c
@@ -149,8 +149,8 @@ static void irq_handler(struct pt_regs *regs)
u32 irqstat = gic_read_iar();
u32 irqnr = gic_iar_irqnr(irqstat);
- if (irqnr != GICC_INT_SPURIOUS)
- gic_write_eoir(irqstat);
+ if (irqnr == GICC_INT_SPURIOUS)
+ return;
if (irqnr == PPI(vtimer_info.irq)) {
info = &vtimer_info;
@@ -162,7 +162,11 @@ static void irq_handler(struct pt_regs *regs)
}
info->write_ctl(ARCH_TIMER_CTL_IMASK | ARCH_TIMER_CTL_ENABLE);
+ isb();
+
info->irq_received = true;
+
+ gic_write_eoir(irqstat);
}
static bool gic_timer_pending(struct timer_info *info)
--
2.20.1
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
