diff options
author | Christoffer Dall | 2013-11-19 05:32:00 +0100 |
---|---|---|
committer | Peter Maydell | 2014-02-08 15:47:28 +0100 |
commit | 8d999995e45c1002aa11f269c98f2e93e6f8c42a (patch) | |
tree | e2df461084ecbcdebbe1a7e9c4e20b526dabb8e9 /hw/intc/gic_internal.h | |
parent | target-arm: Add support for AArch32 64bit VCVTB and VCVTT (diff) | |
download | qemu-8d999995e45c1002aa11f269c98f2e93e6f8c42a.tar.gz qemu-8d999995e45c1002aa11f269c98f2e93e6f8c42a.tar.xz qemu-8d999995e45c1002aa11f269c98f2e93e6f8c42a.zip |
arm_gic: Fix GIC pending behavior
The existing implementation of the pending behavior in gic_set_irq,
gic_complete_irq, and the distributor pending set/clear registers does
not follow the semantics of the GICv2.0 specs, but may implement the
11MPCore support. Therefore, maintain the existing semantics for
11MPCore and v7M NVIC and change the behavior to be in accordance with
the GICv2.0 specs for "generic implementations" (s->revision == 1 ||
s->revision == 2).
Generic implementations distinguish between setting a level-triggered
interrupt pending through writes to the GICD_ISPENDR and when hardware
raises the interrupt line. Writing to the GICD_ICPENDR will not cause
the interrupt to become non-pending if the line is still active, and
conversely, if the line is deactivated but the interrupt is marked as
pending through a write to GICD_ISPENDR, the interrupt remains pending.
Handle this situation in the GIC_TEST_PENDING (which now becomes a
static inline named gic_test_pending) and let the 'pending' field
correspond only to the latched state of the D-flip flop in the GICv2.0
specs Figure 4-10.
The following changes are added:
gic_test_pending:
Make this a static inline and split out the 11MPCore from the generic
behavior. For the generic behavior, consider interrupts pending if:
((s->irq_state[irq].pending & (cm) != 0) ||
(!GIC_TEST_EDGE_TRIGGER(irq) && GIC_TEST_LEVEL(irq, cm))
gic_set_irq:
Split out the 11MPCore from the generic behavior. For the generic
behavior, always GIC_SET_LEVEL() on positive level, but only
GIC_SET_PENDING for edge-triggered interrupts and always simply
GIC_CLEAR_LEVEL() on negative level.
gic_complete_irq:
Only resample the line for line-triggered interrupts on an 11MPCore.
Generic implementations will sample the line directly in
gic_test_pending().
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Diffstat (limited to 'hw/intc/gic_internal.h')
-rw-r--r-- | hw/intc/gic_internal.h | 16 |
1 files changed, 15 insertions, 1 deletions
diff --git a/hw/intc/gic_internal.h b/hw/intc/gic_internal.h index 8c02d5888c..92a6f7a3ff 100644 --- a/hw/intc/gic_internal.h +++ b/hw/intc/gic_internal.h @@ -34,7 +34,6 @@ #define GIC_TEST_ENABLED(irq, cm) ((s->irq_state[irq].enabled & (cm)) != 0) #define GIC_SET_PENDING(irq, cm) s->irq_state[irq].pending |= (cm) #define GIC_CLEAR_PENDING(irq, cm) s->irq_state[irq].pending &= ~(cm) -#define GIC_TEST_PENDING(irq, cm) ((s->irq_state[irq].pending & (cm)) != 0) #define GIC_SET_ACTIVE(irq, cm) s->irq_state[irq].active |= (cm) #define GIC_CLEAR_ACTIVE(irq, cm) s->irq_state[irq].active &= ~(cm) #define GIC_TEST_ACTIVE(irq, cm) ((s->irq_state[irq].active & (cm)) != 0) @@ -63,4 +62,19 @@ void gic_update(GICState *s); void gic_init_irqs_and_distributor(GICState *s, int num_irq); void gic_set_priority(GICState *s, int cpu, int irq, uint8_t val); +static inline bool gic_test_pending(GICState *s, int irq, int cm) +{ + if (s->revision == REV_NVIC || s->revision == REV_11MPCORE) { + return s->irq_state[irq].pending & cm; + } else { + /* Edge-triggered interrupts are marked pending on a rising edge, but + * level-triggered interrupts are either considered pending when the + * level is active or if software has explicitly written to + * GICD_ISPENDR to set the state pending. + */ + return (s->irq_state[irq].pending & cm) || + (!GIC_TEST_EDGE_TRIGGER(irq) && GIC_TEST_LEVEL(irq, cm)); + } +} + #endif /* !QEMU_ARM_GIC_INTERNAL_H */ |