Re: [PATCH v4 17/37] mips: inline serial_init()

[Marc-André Lureau]

Hi

On Mon, Nov 25, 2019 at 5:33 PM Aleksandar Markovic
<address@hidden> wrote:
>
>
>
> On Monday, November 25, 2019, Marc-André Lureau <address@hidden> wrote:
>>
>> Hi
>>
>> On Mon, Nov 25, 2019 at 2:12 PM Aleksandar Markovic
>> <address@hidden> wrote:
>> >
>> >
>> >
>> > On Wednesday, November 20, 2019, Marc-André Lureau <address@hidden> wrote:
>> >>
>> >> The function is specific to mipssim, let's inline it.
>> >>
>> >> Signed-off-by: Marc-André Lureau <address@hidden>
>> >> ---
>> >>  hw/char/serial.c         | 16 ----------------
>> >>  hw/mips/mips_mipssim.c   | 15 ++++++++++++---
>> >>  include/hw/char/serial.h |  2 --
>> >>  3 files changed, 12 insertions(+), 21 deletions(-)
>> >>
>> >> diff --git a/hw/char/serial.c b/hw/char/serial.c
>> >> index 164146ede8..23f0b02516 100644
>> >> --- a/hw/char/serial.c
>> >> +++ b/hw/char/serial.c
>> >> @@ -1023,22 +1023,6 @@ static const TypeInfo serial_io_info = {
>> >>      .class_init = serial_io_class_init,
>> >>  };
>> >>
>> >> -SerialIO *serial_init(int base, qemu_irq irq, int baudbase,
>> >> -                         Chardev *chr, MemoryRegion *system_io)
>> >> -{
>> >> -    SerialIO *sio = SERIAL_IO(qdev_create(NULL, TYPE_SERIAL_IO));
>> >> -
>> >> -    qdev_prop_set_uint32(DEVICE(sio), "baudbase", baudbase);
>> >> -    qdev_prop_set_chr(DEVICE(sio), "chardev", chr);
>> >> -    qdev_set_legacy_instance_id(DEVICE(sio), base, 2);
>> >> -    qdev_init_nofail(DEVICE(sio));
>> >> -
>> >> -    sysbus_connect_irq(SYS_BUS_DEVICE(sio), 0, irq);
>> >> -    memory_region_add_subregion(system_io, base, &sio->serial.io);
>> >> -
>> >> -    return sio;
>> >> -}
>> >> -
>> >>  static Property serial_properties[] = {
>> >>      DEFINE_PROP_CHR("chardev", SerialState, chr),
>> >>      DEFINE_PROP_UINT32("baudbase", SerialState, baudbase, 115200),
>> >> diff --git a/hw/mips/mips_mipssim.c b/hw/mips/mips_mipssim.c
>> >> index 282bbecb24..bfafa4d7e9 100644
>> >> --- a/hw/mips/mips_mipssim.c
>> >> +++ b/hw/mips/mips_mipssim.c
>> >> @@ -40,6 +40,7 @@
>> >>  #include "hw/loader.h"
>> >>  #include "elf.h"
>> >>  #include "hw/sysbus.h"
>> >> +#include "hw/qdev-properties.h"
>> >>  #include "exec/address-spaces.h"
>> >>  #include "qemu/error-report.h"
>> >>  #include "sysemu/qtest.h"
>> >> @@ -219,9 +220,17 @@ mips_mipssim_init(MachineState *machine)
>> >>       * A single 16450 sits at offset 0x3f8. It is attached to
>> >>       * MIPS CPU INT2, which is interrupt 4.
>> >>       */
>> >> -    if (serial_hd(0)) a
>> >> -        serial_init(0x3f8, env->irq[4], 115200, serial_hd(0),
>> >> -                    get_system_io());
>> >> +    if (serial_hd(0)) {
>> >> +        DeviceState *dev = qdev_create(NULL, TYPE_SERIAL_IO);
>> >> +
>> >> +        qdev_prop_set_uint32(DEVICE(dev), "baudbase", 115200);
>> >> +        qdev_prop_set_chr(dev, "chardev", serial_hd(0));
>> >> +        qdev_set_legacy_instance_id(dev, 0x3f8, 2);
>> >> +        qdev_init_nofail(dev);
>> >> +        sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, env->irq[4]);
>> >> +        memory_region_add_subregion(get_system_io(), 0x3f8,
>> >> +                                    &SERIAL_IO(dev)->serial.io);
>> >> +    }
>> >
>> >
>> >
>> > Please explain why the code in the deleted function and the new function 
>> > are not identical. Why is the new code better?
>>
>> What differences do you spot?
>>
>> serial_init() looks very generic, but it is not, and this is
>> confusing. QDev users should call qdev_create().
>>
>> If there is a single user, it is better to inline the code where it is
>> being used, or move the function there.
>>
>
> Marc-Andre, you misunderstood me: I am not complaining about inlining at all, 
> I would do the same if I were you.
>
> I just see you expressed the same thing from former function in a slightly 
> different way in the inlined version, and they look functionally equivalent 
> to me. But I suppose the new way is better, and I want to know why. Is it a 
> more standard way, or something even more? Please understand that I don't 
> come from device emulation background - I want to learn a little bit about 
> device-related QEMU customs and conventions, that is all. :)
>
> For example,
>
> in old version:
>
> >> -SerialIO *serial_init(int base, qemu_irq irq, int baudbase,
> >> -                         Chardev *chr, MemoryRegion *system_io)
> >> -{
> >> -    SerialIO *sio = SERIAL_IO(qdev_create(NULL, TYPE_SERIAL_IO));
> >> -
> >> -    qdev_prop_set_uint32(DEVICE(sio), "baudbase", baudbase);
> >> -    qdev_prop_set_chr(DEVICE(sio), "chardev", chr);
> >> -    qdev_set_legacy_instance_id(DEVICE(sio), base, 2);
> >> -    qdev_init_nofail(DEVICE(sio));
> >> -
> >> -    sysbus_connect_irq(SYS_BUS_DEVICE(sio), 0, irq);
> >> -    memory_region_add_subregion(system_io, base, &sio->serial.io);
> >> -
> >> -    return sio;
> >> -}
>
> in new version:
>
>  >> +    if (serial_hd(0)) {
> >> +        DeviceState *dev = qdev_create(NULL, TYPE_SERIAL_IO);
> >> +
> >> +        qdev_prop_set_uint32(DEVICE(dev), "baudbase", 115200);
> >> +        qdev_prop_set_chr(dev, "chardev", serial_hd(0));   <-- you didn't 
> >> use here DEVICE(), while it is used in the previous line, and also in the 
> >> old version, and similar questions
> >> +        qdev_set_legacy_instance_id(dev, 0x3f8, 2);
> >> +        qdev_init_nofail(dev);
> >> +        sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, env->irq[4]);
> >> +        memory_region_add_subregion(get_system_io(), 0x3f8,
> >> +                                    &SERIAL_IO(dev)->serial.io);
> >> +    }
>
> For you all these details are probably perfectly natural and obvois, but not 
> for me.

Oh, there is no big difference between a variable using the base type
DeviceState * or the instance type SerialIO *.

Basically, you need a variable to hold the instance, and then you cast
it with DEVICE() or SERIAL_IO() or event SYS_BUS_DEVICE() as
necessary, depending on what type the function you call expect.

I probably picked DeviceState* because it is the most commonly used,
and it avoids having plenty of DEVICE() casts.