On Sat, 18 Aug 2012, Peter Maydell wrote:
On 18 August 2012 03:55, Brendan Fennell <bfenn...@skynet.ie> wrote:Signed-off-by: Brendan Fennell <bfenn...@skynet.ie> --- hw/pl190.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/hw/pl190.c b/hw/pl190.c index cb50afb..d69d5be 100644 --- a/hw/pl190.c +++ b/hw/pl190.c @@ -133,7 +133,7 @@ static uint64_t pl190_read(void *opaque, target_phys_addr_t offset, s->priority = i; pl190_update(s); } - return s->vect_addr[s->priority]; + return s->vect_addr[s->priority - 1]; case 13: /* DEFVECTADDR */ return s->vect_addr[16]; default:This doesn't look right -- if s->priority is zero then we'll read off the beginning of the array. What's the actual bug you're trying to fix here?
The bug is that when, for example, interrupt 4 triggers the VECTADDR of interrupt 5 is returned by pl190_read().
Each s->prio_mask entry contains the interrupt mask for all *higher* priority interrupts, see pl190_update_vectors(). This means that s->prio_mask[0] is always zero (as zero is the highest priority), s->priority can never be zero as ((s->level | s->soft_level) & s->prio_mask[0]) is always zero.
Therefore after the for loop in pl190_read() i is the index of the current highest priority interrupt + 1. Brendan.
-- PMM
