Mark Cave-Ayland <mark.cave-ayl...@ilande.co.uk> writes: > On 01/06/2023 10:00, Markus Armbruster wrote: > >> Mark Cave-Ayland <mark.cave-ayl...@ilande.co.uk> writes: >> >>> On 31/05/2023 16:00, Markus Armbruster wrote: >>> >>>> Philippe Mathieu-Daudé <phi...@linaro.org> writes: >>>> >>>>> On 31/5/23 14:53, Mark Cave-Ayland wrote: >>>>>> Also change the instantiation of the CPU to use object_initialize_child() >>>>>> followed by a separate realisation. >>>>>> Signed-off-by: Mark Cave-Ayland <mark.cave-ayl...@ilande.co.uk> >>>>>> --- >>>>>> hw/m68k/q800.c | 13 ++++++++----- >>>>>> include/hw/m68k/q800.h | 2 ++ >>>>>> 2 files changed, 10 insertions(+), 5 deletions(-) >>>>>> diff --git a/hw/m68k/q800.c b/hw/m68k/q800.c >>>>>> index 3730b30dd1..c34b2548ca 100644 >>>>>> --- a/hw/m68k/q800.c >>>>>> +++ b/hw/m68k/q800.c >>>>>> @@ -364,7 +364,7 @@ static uint8_t fake_mac_rom[] = { >>>>>> static void q800_machine_init(MachineState *machine) >>>>>> { >>>>>> - M68kCPU *cpu = NULL; >>>>>> + Q800MachineState *m = Q800_MACHINE(machine); >>>>>> int linux_boot; >>>>>> int32_t kernel_size; >>>>>> uint64_t elf_entry; >>>>>> @@ -407,8 +407,10 @@ static void q800_machine_init(MachineState *machine) >>>>>> } >>>>>> /* init CPUs */ >>>>>> - cpu = M68K_CPU(cpu_create(machine->cpu_type)); >>>>>> - qemu_register_reset(main_cpu_reset, cpu); >>>>>> + object_initialize_child(OBJECT(machine), "cpu", &m->cpu, >>>>>> + M68K_CPU_TYPE_NAME("m68040")); >>>>>> + object_property_set_bool(OBJECT(&m->cpu), "realized", true, >>>>>> &error_fatal); >>>>> >>>>> CPUs are QDev-based, shouldn't we use qdev_realize()? >>>> >>>> Yes, we should. >>>> [...] >>> >>> Interesting. I remember thinking that CPUs were different, so I'm fairly >>> sure I borrowed this from some similar code in hw/arm :) >>> >>> Shouldn't the above be directly equivalent to qdev_realize(dev, NULL, >>> &error_fatal) given that the CPU doesn't connect to a bus? >> >> It's been a while since I worked on this... >> >> Commit ce189ab230b (qdev: Convert bus-less devices to qdev_realize() >> with Coccinelle) looks like you're right. > > Thanks for the confirmation! Given that this matches existing code that > doesn't use cpu_create(), I'm inclined to keep this as-is to avoid creating > another pattern for instantiating CPUs.
Wherever you *can* use qdev_realize(), you should. The less we access property "realized" outside qdev core, the better. I few accesses have crept in since I converted the tree to qdev_realize() & friends. Another conversion pass would be in order.