On 3/6/23 20:07, Mark Cave-Ayland wrote:
On 31/05/2023 21:35, Philippe Mathieu-Daudé wrote:
Introduce the ARM_TIMER sysbus device.
arm_timer_new() is converted as QOM instance init()/finalize()
handlers. Note in arm_timer_finalize() we release a ptimer handle
which was previously leaked.
ArmTimerState is directly embedded into SP804State/IcpPitState,
and is initialized as a QOM child.
Since the timer frequency belongs to ARM_TIMER, have it hold the
QOM property. SP804State/IcpPitState directly access it.
Similarly the SP804State/IcpPitState input IRQ becomes the
ARM_TIMER sysbus output IRQ.
Signed-off-by: Philippe Mathieu-Daudé <phi...@linaro.org>
---
hw/timer/arm_timer.c | 109 +++++++++++++++++++++++++++----------------
1 file changed, 70 insertions(+), 39 deletions(-)
-static void arm_timer_reset(ArmTimerState *s)
+static void arm_timer_reset(DeviceState *dev)
{
+ ArmTimerState *s = ARM_TIMER(dev);
+
s->control = TIMER_CTRL_IE;
}
If you're currently set up to test the ARM timers with these changes, is
it worth considering converting this to use the Resettable interface at
the same time?
Good point. Then ARM_TIMER doesn't need to inherit from SysBus: if the
parent device resets it explicitly, it can be a plan QDev.
Peter, what do you think?
Even generically, I wonder if a QDev could resets all its QOM children,
propagating each ResetType.
