Re: [Qemu-devel] [PATCH v1 4/5] hw/arm: Add the STM32F4xx SoC

[Alistair Francis <alistair23@gmail.com>]

On Tue, Apr 30, 2019 at 8:59 AM Peter Maydell <peter.maydell@linaro.org> wrote:
>
> On Mon, 29 Apr 2019 at 06:38, Alistair Francis <alistair@alistair23.me> wrote:
> >
> > Signed-off-by: Alistair Francis <alistair@alistair23.me>
> > ---
> >  MAINTAINERS                     |   8 +
> >  default-configs/arm-softmmu.mak |   1 +
> >  hw/arm/Kconfig                  |   3 +
> >  hw/arm/Makefile.objs            |   1 +
> >  hw/arm/stm32f405_soc.c          | 292 ++++++++++++++++++++++++++++++++
> >  include/hw/arm/stm32f405_soc.h  |  70 ++++++++
> >  6 files changed, 375 insertions(+)
> >  create mode 100644 hw/arm/stm32f405_soc.c
> >  create mode 100644 include/hw/arm/stm32f405_soc.h
>
> Looks good; a few minor things below.
>
> > +static void stm32f405_soc_initfn(Object *obj)
> > +{
> > +    STM32F405State *s = STM32F405_SOC(obj);
> > +    int i;
> > +
> > +    sysbus_init_child_obj(obj, "armv7m", &s->armv7m, sizeof(s->armv7m),
> > +                          TYPE_ARMV7M);
> > +
> > +    sysbus_init_child_obj(obj, "syscfg", &s->syscfg, sizeof(s->syscfg),
> > +                          TYPE_STM32F4XX_SYSCFG);
> > +
> > +    for (i = 0; i < STM_NUM_USARTS; i++) {
> > +        sysbus_init_child_obj(obj, "usart[*]", &s->usart[i],
> > +                              sizeof(s->usart[i]), TYPE_STM32F2XX_USART);
> > +    }
> > +
> > +    for (i = 0; i < STM_NUM_TIMERS; i++) {
> > +        sysbus_init_child_obj(obj, "timer[*]", &s->timer[i],
> > +                              sizeof(s->timer[i]), TYPE_STM32F2XX_TIMER);
> > +    }
> > +
> > +    s->adc_irqs = OR_IRQ(object_new(TYPE_OR_IRQ));
>
> It would be more in keeping with the style of the rest of this
> device to have the device be inline in the STM32F405State
> struct and initialized with object_initialize_child() rather
> than allocated separately with object_new(). (hw/arm/armsse.c
> has an example of doing this with a TYPE_OR_IRQ object.)

I have addressed all your comments.

Alistair

>
> > +
> > +    for (i = 0; i < STM_NUM_ADCS; i++) {
> > +        sysbus_init_child_obj(obj, "adc[*]", &s->adc[i], sizeof(s->adc[i]),
> > +                              TYPE_STM32F2XX_ADC);
> > +    }
> > +
> > +    for (i = 0; i < STM_NUM_SPIS; i++) {
> > +        sysbus_init_child_obj(obj, "spi[*]", &s->spi[i], sizeof(s->spi[i]),
> > +                              TYPE_STM32F2XX_SPI);
> > +    }
> > +
> > +    sysbus_init_child_obj(obj, "exti", &s->exti, sizeof(s->exti),
> > +                          TYPE_STM32F4XX_EXTI);
> > +}
> > +
> > +static void stm32f405_soc_realize(DeviceState *dev_soc, Error **errp)
> > +{
> > +    STM32F405State *s = STM32F405_SOC(dev_soc);
> > +    DeviceState *dev, *armv7m;
> > +    SysBusDevice *busdev;
> > +    Error *err = NULL;
> > +    int i;
> > +
> > +    MemoryRegion *system_memory = get_system_memory();
> > +    MemoryRegion *sram = g_new(MemoryRegion, 1);
> > +    MemoryRegion *flash = g_new(MemoryRegion, 1);
> > +    MemoryRegion *flash_alias = g_new(MemoryRegion, 1);
>
> I would prefer to have these MemoryRegions be in the STM32F405State
> struct rather than separately allocated.
>
> > +
> > +    memory_region_init_ram(flash, NULL, "STM32F405.flash", FLASH_SIZE,
> > +                           &error_fatal);
>
> Better to pass the error back up via errp rather than use error_fatal
> in a realize function.
>
> > +    memory_region_init_alias(flash_alias, NULL, "STM32F405.flash.alias",
> > +                             flash, 0, FLASH_SIZE);
> > +
> > +    memory_region_set_readonly(flash, true);
> > +    memory_region_set_readonly(flash_alias, true);
> > +
> > +    memory_region_add_subregion(system_memory, FLASH_BASE_ADDRESS, flash);
> > +    memory_region_add_subregion(system_memory, 0, flash_alias);
> > +
> > +    memory_region_init_ram(sram, NULL, "STM32F405.sram", SRAM_SIZE,
> > +                           &error_fatal);
> > +    memory_region_add_subregion(system_memory, SRAM_BASE_ADDRESS, sram);
> > +
> > +    armv7m = DEVICE(&s->armv7m);
> > +    qdev_prop_set_uint32(armv7m, "num-irq", 96);
> > +    qdev_prop_set_string(armv7m, "cpu-type", s->cpu_type);
> > +    qdev_prop_set_bit(armv7m, "enable-bitband", true);
> > +    object_property_set_link(OBJECT(&s->armv7m), OBJECT(get_system_memory()),
>
> You could use OBJECT(system_memory) rather than calling
> get_system_memory() again.
>
> > +static Property stm32f405_soc_properties[] = {
> > +    DEFINE_PROP_STRING("cpu-type", STM32F405State, cpu_type),
> > +    DEFINE_PROP_END_OF_LIST(),
> > +};
> > +
> > +static void stm32f405_soc_class_init(ObjectClass *klass, void *data)
> > +{
> > +    DeviceClass *dc = DEVICE_CLASS(klass);
> > +
> > +    dc->realize = stm32f405_soc_realize;
> > +    dc->props = stm32f405_soc_properties;
>
> A comment here "No vmstate or reset required: device has no internal state"
> would help indicate that dc->vmsd and dc->reset have not merely
> been forgotten.
>
> (Eventually I might actually write a patch to let us express
> in code "dc->vmsd = device_has_no_state;"...)
>
> > +}
>
> thanks
> -- PMM