[for-8.0,00/19] Convert most CPU classes to 3-phase reset

This patchset converts the TYPE_CPU base class and most subclasses
to use 3-phase reset. (The exception is s390, which is doing
something a bit odd with its reset, so the conversion there isn't
going to be simple like these others. So I'll do that one
separately.)

The rationale here is that we should be able to get rid of
all the remaining uses of device_class_set_parent_reset()
and remove/simplify some of the transitional code that's
currently bridging between "legacy" reset and 3-phase reset.

NB: even with this series, it's not possible to usefully do
anything requiring 3-phase reset of a CPU yet, because all
CPU objects get ad-hoc reset by some code somewhere doing
a cpu_reset() call on them, which will just do all 3 phases
in order. I would like to try to address that eventually,
but it's not trivial.

thanks
-- PMM

Peter Maydell (19):
  hw/core/cpu-common: Convert TYPE_CPU class to 3-phase reset
  target/arm: Convert to 3-phase reset
  target/avr: Convert to 3-phase reset
  target/cris: Convert to 3-phase reset
  target/hexagon: Convert to 3-phase reset
  target/i386: Convert to 3-phase reset
  target/loongarch: Convert to 3-phase reset
  target/m68k: Convert to 3-phase reset
  target/microblaze: Convert to 3-phase reset
  target/mips: Convert to 3-phase reset
  target/nios2: Convert to 3-phase reset
  target/openrisc: Convert to 3-phase reset
  target/ppc: Convert to 3-phase reset
  target/riscv: Convert to 3-phase reset
  target/rx: Convert to 3-phase reset
  target/sh4: Convert to 3-phase reset
  target/sparc: Convert to 3-phase reset
  target/tricore: Convert to 3-phase reset
  target/xtensa: Convert to 3-phase reset

 target/arm/cpu-qom.h        |  4 ++--
 target/avr/cpu-qom.h        |  4 ++--
 target/cris/cpu-qom.h       |  4 ++--
 target/hexagon/cpu.h        |  2 +-
 target/i386/cpu-qom.h       |  4 ++--
 target/loongarch/cpu.h      |  4 ++--
 target/m68k/cpu-qom.h       |  4 ++--
 target/microblaze/cpu-qom.h |  4 ++--
 target/mips/cpu-qom.h       |  4 ++--
 target/nios2/cpu.h          |  4 ++--
 target/openrisc/cpu.h       |  4 ++--
 target/ppc/cpu-qom.h        |  4 ++--
 target/riscv/cpu.h          |  4 ++--
 target/rx/cpu-qom.h         |  4 ++--
 target/sh4/cpu-qom.h        |  4 ++--
 target/sparc/cpu-qom.h      |  4 ++--
 target/tricore/cpu-qom.h    |  2 +-
 target/xtensa/cpu-qom.h     |  4 ++--
 hw/core/cpu-common.c        |  7 ++++---
 target/arm/cpu.c            | 13 +++++++++----
 target/avr/cpu.c            | 13 +++++++++----
 target/cris/cpu.c           | 12 ++++++++----
 target/hexagon/cpu.c        | 12 ++++++++----
 target/i386/cpu.c           | 12 ++++++++----
 target/loongarch/cpu.c      | 12 ++++++++----
 target/m68k/cpu.c           | 12 ++++++++----
 target/microblaze/cpu.c     | 12 ++++++++----
 target/mips/cpu.c           | 12 ++++++++----
 target/nios2/cpu.c          | 12 ++++++++----
 target/openrisc/cpu.c       | 12 ++++++++----
 target/ppc/cpu_init.c       | 12 ++++++++----
 target/riscv/cpu.c          | 12 ++++++++----
 target/rx/cpu.c             | 13 ++++++++-----
 target/sh4/cpu.c            | 12 ++++++++----
 target/sparc/cpu.c          | 12 ++++++++----
 target/tricore/cpu.c        | 12 ++++++++----
 target/xtensa/cpu.c         | 12 ++++++++----
 37 files changed, 184 insertions(+), 110 deletions(-)

On 24/11/22 12:50, Peter Maydell wrote:
> This patchset converts the TYPE_CPU base class and most subclasses
> to use 3-phase reset. (The exception is s390, which is doing
> something a bit odd with its reset, so the conversion there isn't
> going to be simple like these others. So I'll do that one
> separately.)
> 
> The rationale here is that we should be able to get rid of
> all the remaining uses of device_class_set_parent_reset()
> and remove/simplify some of the transitional code that's
> currently bridging between "legacy" reset and 3-phase reset.
> 
> NB: even with this series, it's not possible to usefully do
> anything requiring 3-phase reset of a CPU yet, because all
> CPU objects get ad-hoc reset by some code somewhere doing
> a cpu_reset() call on them, which will just do all 3 phases
> in order. I would like to try to address that eventually,
> but it's not trivial.
> 
> thanks
> -- PMM
> 
> Peter Maydell (19):
>    hw/core/cpu-common: Convert TYPE_CPU class to 3-phase reset
>    target/arm: Convert to 3-phase reset
>    target/avr: Convert to 3-phase reset
>    target/cris: Convert to 3-phase reset
>    target/hexagon: Convert to 3-phase reset
>    target/i386: Convert to 3-phase reset
>    target/loongarch: Convert to 3-phase reset
>    target/m68k: Convert to 3-phase reset
>    target/microblaze: Convert to 3-phase reset
>    target/mips: Convert to 3-phase reset
>    target/nios2: Convert to 3-phase reset
>    target/openrisc: Convert to 3-phase reset
>    target/ppc: Convert to 3-phase reset
>    target/riscv: Convert to 3-phase reset
>    target/rx: Convert to 3-phase reset
>    target/sh4: Convert to 3-phase reset
>    target/sparc: Convert to 3-phase reset
>    target/tricore: Convert to 3-phase reset
>    target/xtensa: Convert to 3-phase reset

Series:
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

On 11/24/22 03:50, Peter Maydell wrote:
> Peter Maydell (19):
>    hw/core/cpu-common: Convert TYPE_CPU class to 3-phase reset
>    target/arm: Convert to 3-phase reset
>    target/avr: Convert to 3-phase reset
>    target/cris: Convert to 3-phase reset
>    target/hexagon: Convert to 3-phase reset
>    target/i386: Convert to 3-phase reset
>    target/loongarch: Convert to 3-phase reset
>    target/m68k: Convert to 3-phase reset
>    target/microblaze: Convert to 3-phase reset
>    target/mips: Convert to 3-phase reset
>    target/nios2: Convert to 3-phase reset
>    target/openrisc: Convert to 3-phase reset
>    target/ppc: Convert to 3-phase reset
>    target/riscv: Convert to 3-phase reset
>    target/rx: Convert to 3-phase reset
>    target/sh4: Convert to 3-phase reset
>    target/sparc: Convert to 3-phase reset
>    target/tricore: Convert to 3-phase reset
>    target/xtensa: Convert to 3-phase reset

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

On 24/11/22 14:46, Philippe Mathieu-Daudé wrote:
> On 24/11/22 12:50, Peter Maydell wrote:
>> This patchset converts the TYPE_CPU base class and most subclasses
>> to use 3-phase reset. (The exception is s390, which is doing
>> something a bit odd with its reset, so the conversion there isn't
>> going to be simple like these others. So I'll do that one
>> separately.)
Note, we can then remove scripts/coccinelle/cpu-reset.cocci.

On Wed, 30 Nov 2022 at 10:51, Philippe Mathieu-Daudé <philmd@linaro.org> wrote:
>
> On 24/11/22 14:46, Philippe Mathieu-Daudé wrote:
> > On 24/11/22 12:50, Peter Maydell wrote:
> >> This patchset converts the TYPE_CPU base class and most subclasses
> >> to use 3-phase reset. (The exception is s390, which is doing
> >> something a bit odd with its reset, so the conversion there isn't
> >> going to be simple like these others. So I'll do that one
> >> separately.)
> Note, we can then remove scripts/coccinelle/cpu-reset.cocci.

What's our usual practice for out-of-date conversion coccinelle
scripts? That particular script was "we'll never need this again"
pretty much from the moment it was checked in, because we did
the conversion of all the targets in one go. But it's still
useful in some sense as a "this is an example of how to do
this kind of change with a coccinelle script"...

thanks
-- PMM

On 30/11/22 13:38, Peter Maydell wrote:
> On Wed, 30 Nov 2022 at 10:51, Philippe Mathieu-Daudé <philmd@linaro.org> wrote:
>>
>> On 24/11/22 14:46, Philippe Mathieu-Daudé wrote:
>>> On 24/11/22 12:50, Peter Maydell wrote:
>>>> This patchset converts the TYPE_CPU base class and most subclasses
>>>> to use 3-phase reset. (The exception is s390, which is doing
>>>> something a bit odd with its reset, so the conversion there isn't
>>>> going to be simple like these others. So I'll do that one
>>>> separately.)
>> Note, we can then remove scripts/coccinelle/cpu-reset.cocci.
> 
> What's our usual practice for out-of-date conversion coccinelle
> scripts? That particular script was "we'll never need this again"
> pretty much from the moment it was checked in, because we did
> the conversion of all the targets in one go. But it's still
> useful in some sense as a "this is an example of how to do
> this kind of change with a coccinelle script"...

The pattern I observed last years is:

  (a) If one-shot transformation, log the script in the commit
      description,

  (b) If the script can be used again in the future, commit it
      in scripts/coccinelle/,

  (c) Complex scripts are committed in scripts/coccinelle/ as
      example/reference.

I am fine keeping scripts/coccinelle/cpu-reset.cocci as an
example (c).

Regards,

Phil.

On Thu, 24 Nov 2022 at 11:50, Peter Maydell <peter.maydell@linaro.org> wrote:
>
> This patchset converts the TYPE_CPU base class and most subclasses
> to use 3-phase reset. (The exception is s390, which is doing
> something a bit odd with its reset, so the conversion there isn't
> going to be simple like these others. So I'll do that one
> separately.)

I plan to pick these up and send them in in a pullreq with
various other reset-related patches of mine.

thanks
-- PMM