Re: [PATCH 7/7] hw: Simplify using sysbus_init_irqs() [manual]

[Philippe Mathieu-Daudé]

On 1/6/23 07:59, Markus Armbruster wrote:
> Philippe Mathieu-Daudé <philmd@linaro.org> writes:
>
> > Audit the sysbus_init_irq() calls and manually convert
> > to sysbus_init_irqs() when a loop is involved.
> >
> > Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> > ---
> >   hw/intc/loongarch_extioi.c | 3 +--
> >   hw/intc/omap_intc.c        | 3 +--
> >   hw/pci-host/gpex.c         | 2 +-
> >   hw/timer/renesas_tmr.c     | 9 +++------
> >   4 files changed, 6 insertions(+), 11 deletions(-)
> >
> > diff --git a/hw/intc/loongarch_extioi.c b/hw/intc/loongarch_extioi.c
> > index db941de20e..c579636215 100644
> > --- a/hw/intc/loongarch_extioi.c
> > +++ b/hw/intc/loongarch_extioi.c
> > @@ -275,8 +275,7 @@ static void loongarch_extioi_instance_init(Object *obj)
> >       LoongArchExtIOI *s = LOONGARCH_EXTIOI(obj);
> >       int cpu, pin;
> >   
> > -    sysbus_init_irqs(SYS_BUS_DEVICE(dev), s->irq, EXTIOI_IRQS);
> > -
> > +    sysbus_init_irqs(dev, s->irq, EXTIOI_IRQS);
>
> Commit message claims "when a loop is involved".  No loop here.  That
> work was actually done in the previous patch:
>
>    diff --git a/hw/intc/loongarch_extioi.c b/hw/intc/loongarch_extioi.c
>    index 0e7a3e32f3..db941de20e 100644
>    --- a/hw/intc/loongarch_extioi.c
>    +++ b/hw/intc/loongarch_extioi.c
>    @@ -273,11 +273,9 @@ static void loongarch_extioi_instance_init(Object *obj)
>     {
>         SysBusDevice *dev = SYS_BUS_DEVICE(obj);
>         LoongArchExtIOI *s = LOONGARCH_EXTIOI(obj);
>    -    int i, cpu, pin;
>    +    int cpu, pin;
>
>    -    for (i = 0; i < EXTIOI_IRQS; i++) {
>    -        sysbus_init_irq(SYS_BUS_DEVICE(dev), &s->irq[i]);
>    -    }
>    +    sysbus_init_irqs(SYS_BUS_DEVICE(dev), s->irq, EXTIOI_IRQS);
>
>         qdev_init_gpio_in(DEVICE(obj), extioi_setirq, EXTIOI_IRQS);
>
> In this patch, you merely delete a superfluous type conversion that is
> present even before your series.

Right. I guess I did that automatically "why are we casting the same
type?" without even noticing.

> There are more of them in this function.  Please delete them all, and in
> a separate patch.

OK.

> Actually, there are more elsewhere.  Coccinelle script
>
>      @@
>      typedef SysBusDevice;
>      SysBusDevice *dev;
>      @@
>      -    SYS_BUS_DEVICE(dev)
>      +    dev
>
> finds some in hw/arm/xlnx-versal.c and hw/rx/rx62n.c, too.
>
> Would be nice to do this for every QOM type, but I don't know how
> without duplicating the semantic patch for each of them.  There are
> almost 150 uses os OBJECT_DECLARE_TYPE()...

Checking all QOM macros, I counted 1076 types...