On 8/6/23 16:41, Peter Maydell wrote:
On Wed, 31 May 2023 at 21:36, Philippe Mathieu-Daudé <phi...@linaro.org> wrote:
Release the IRQs allocated in sp804_realize() in the
corresponding sp804_unrealize() handler.
Signed-off-by: Philippe Mathieu-Daudé <phi...@linaro.org>
---
hw/timer/arm_timer.c | 10 ++++++++++
1 file changed, 10 insertions(+)
+static void sp804_unrealize(DeviceState *dev)
+{
+ SP804State *s = SP804(dev);
+
+ for (unsigned i = 0; i < ARRAY_SIZE(s->timer); i++) {
+ qemu_free_irq(s->timer[i]->irq);
+ }
+}
I don't really see the purpose in this. It doesn't actually
avoid a leak if we ever destroy an SP804, because
s->timer[i] itself is memory allocated by arm_timer_init()
which we don't clean up (the arm_timer_state not being
a qdev). If we did convert arm_timer_state to qdev
then these interrupts should turn into being sysbus irqs
and gpio inputs on the relevant devices. (In fact if you
were willing to take the migration-compat hit you
could just have the sp804 create a 2 input OR gate and
wire the arm_timer irqs up to it and use the output
of the OR gate as the sp804 outbound interrupt line.)
Thank for the suggestion, I didn't notice the OR.
