[RFC,4/7] qom: Introduce CPU class

It's abstract, derived directly from TYPE_OBJECT (to avoid dependency
on MODULE_INIT_DEVICE) and for now is empty.

Place it in hw/. Have user emulators pick it up via VPATH, building it
per target since they didn't use any qdev/QOM devices so far.

Introduce processor_init() for registering, and call module init as
needed.

Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Anthony Liguori <aliguori@us.ibm.com>
---
 Makefile.objs      |    1 +
 Makefile.target    |    9 ++++++---
 arch_init.c        |    1 +
 bsd-user/main.c    |    1 +
 darwin-user/main.c |    1 +
 hw/cpu.c           |   27 +++++++++++++++++++++++++++
 include/qemu/cpu.h |   27 +++++++++++++++++++++++++++
 linux-user/main.c  |    1 +
 module.h           |    2 ++
 9 files changed, 67 insertions(+), 3 deletions(-)
 create mode 100644 hw/cpu.c
 create mode 100644 include/qemu/cpu.h

On 01/29/2012 07:25 AM, Andreas Färber wrote:
> It's abstract, derived directly from TYPE_OBJECT (to avoid dependency
> on MODULE_INIT_DEVICE) and for now is empty.
>
> Place it in hw/. Have user emulators pick it up via VPATH, building it
> per target since they didn't use any qdev/QOM devices so far.
>
> Introduce processor_init() for registering, and call module init as
> needed.
>
> Signed-off-by: Andreas Färber<afaerber@suse.de>
> Cc: Anthony Liguori<aliguori@us.ibm.com>
> ---
>   Makefile.objs      |    1 +
>   Makefile.target    |    9 ++++++---
>   arch_init.c        |    1 +
>   bsd-user/main.c    |    1 +
>   darwin-user/main.c |    1 +
>   hw/cpu.c           |   27 +++++++++++++++++++++++++++
>   include/qemu/cpu.h |   27 +++++++++++++++++++++++++++
>   linux-user/main.c  |    1 +
>   module.h           |    2 ++
>   9 files changed, 67 insertions(+), 3 deletions(-)
>   create mode 100644 hw/cpu.c
>   create mode 100644 include/qemu/cpu.h
>
> diff --git a/Makefile.objs b/Makefile.objs
> index b942625..a4b20fa 100644
> --- a/Makefile.objs
> +++ b/Makefile.objs
> @@ -189,6 +189,7 @@ user-obj-y += $(trace-obj-y)
>
>   hw-obj-y =
>   hw-obj-y += vl.o loader.o
> +hw-obj-y += cpu.o
>   hw-obj-$(CONFIG_VIRTIO) += virtio-console.o
>   hw-obj-y += usb-libhw.o
>   hw-obj-$(CONFIG_VIRTIO_PCI) += virtio-pci.o
> diff --git a/Makefile.target b/Makefile.target
> index d1b7867..5d3470e 100644
> --- a/Makefile.target
> +++ b/Makefile.target
> @@ -107,7 +107,7 @@ signal.o: QEMU_CFLAGS += $(HELPER_CFLAGS)
>
>   ifdef CONFIG_LINUX_USER
>
> -$(call set-vpath, $(SRC_PATH)/linux-user:$(SRC_PATH)/linux-user/$(TARGET_ABI_DIR))
> +$(call set-vpath, $(SRC_PATH)/linux-user:$(SRC_PATH)/linux-user/$(TARGET_ABI_DIR):$(SRC_PATH)/hw)
>
>   QEMU_CFLAGS+=-I$(SRC_PATH)/linux-user/$(TARGET_ABI_DIR) -I$(SRC_PATH)/linux-user
>   obj-y = main.o syscall.o strace.o mmap.o signal.o thunk.o \
> @@ -130,6 +130,7 @@ obj-m68k-y += m68k-sim.o m68k-semi.o
>   $(obj-y) $(obj-$(TARGET_BASE_ARCH)-y): $(GENERATED_HEADERS)
>
>   obj-y += module.o
> +obj-y += cpu.o
>   obj-y += $(addprefix ../qom/, $(qom-y))
>   obj-y += $(addprefix ../libuser/, $(user-obj-y))
>   obj-y += $(addprefix ../libdis-user/, $(libdis-y))
> @@ -142,7 +143,7 @@ endif #CONFIG_LINUX_USER
>
>   ifdef CONFIG_DARWIN_USER
>
> -$(call set-vpath, $(SRC_PATH)/darwin-user)
> +$(call set-vpath, $(SRC_PATH)/darwin-user:$(SRC_PATH)/hw)
>
>   QEMU_CFLAGS+=-I$(SRC_PATH)/darwin-user -I$(SRC_PATH)/darwin-user/$(TARGET_ARCH)
>
> @@ -159,6 +160,7 @@ obj-i386-y += ioport-user.o
>   $(obj-y) $(obj-$(TARGET_BASE_ARCH)-y): $(GENERATED_HEADERS)
>
>   obj-y += module.o
> +obj-y += cpu.o
>   obj-y += $(addprefix ../qom/, $(qom-y))
>   obj-y += $(addprefix ../libuser/, $(user-obj-y))
>   obj-y += $(addprefix ../libdis-user/, $(libdis-y))
> @@ -171,7 +173,7 @@ endif #CONFIG_DARWIN_USER
>
>   ifdef CONFIG_BSD_USER
>
> -$(call set-vpath, $(SRC_PATH)/bsd-user)
> +$(call set-vpath, $(SRC_PATH)/bsd-user:$(SRC_PATH)/hw)
>
>   QEMU_CFLAGS+=-I$(SRC_PATH)/bsd-user -I$(SRC_PATH)/bsd-user/$(TARGET_ARCH)
>
> @@ -183,6 +185,7 @@ obj-i386-y += ioport-user.o
>   $(obj-y) $(obj-$(TARGET_BASE_ARCH)-y): $(GENERATED_HEADERS)
>
>   obj-y += module.o
> +obj-y += cpu.o
>   obj-y += $(addprefix ../qom/, $(qom-y))
>   obj-y += $(addprefix ../libuser/, $(user-obj-y))
>   obj-y += $(addprefix ../libdis-user/, $(libdis-y))
> diff --git a/arch_init.c b/arch_init.c
> index 2366511..c0d5f4f 100644
> --- a/arch_init.c
> +++ b/arch_init.c
> @@ -692,6 +692,7 @@ void do_smbios_option(const char *optarg)
>
>   void cpudef_init(void)
>   {
> +    module_call_init(MODULE_INIT_CPU);
>   #if defined(cpudef_setup)
>       cpudef_setup(); /* parse cpu definitions in target config file */
>   #endif
> diff --git a/bsd-user/main.c b/bsd-user/main.c
> index 2ff0361..70e1146 100644
> --- a/bsd-user/main.c
> +++ b/bsd-user/main.c
> @@ -761,6 +761,7 @@ int main(int argc, char **argv)
>       }
>
>       cpu_model = NULL;
> +    module_call_init(MODULE_INIT_CPU);
>   #if defined(cpudef_setup)
>       cpudef_setup(); /* parse cpu definitions in target config file (TBD) */
>   #endif
> diff --git a/darwin-user/main.c b/darwin-user/main.c
> index a4c630d..d065f00 100644
> --- a/darwin-user/main.c
> +++ b/darwin-user/main.c
> @@ -751,6 +751,7 @@ int main(int argc, char **argv)
>           usage();
>
>       module_call_init(MODULE_INIT_EARLY);
> +    module_call_init(MODULE_INIT_CPU);
>
>       optind = 1;
>       for(;;) {
> diff --git a/hw/cpu.c b/hw/cpu.c
> new file mode 100644
> index 0000000..c0e9cfa
> --- /dev/null
> +++ b/hw/cpu.c
> @@ -0,0 +1,27 @@
> +/*
> + * QEMU CPU model
> + *
> + * Copyright (c) 2012 SUSE LINUX Products GmbH
> + *
> + * Licensed under the terms of the GNU GPL version 2
> + * or (at your option) any later version.
> + */
> +
> +#include "qemu/object.h"
> +#include "qemu/cpu.h"
> +#include "qemu-common.h"
> +
> +static TypeInfo cpu_type_info = {
> +    .name = TYPE_CPU,
> +    .parent = TYPE_OBJECT,
> +    .instance_size = sizeof(CPU),


Probably want to do CPUState or something of that nature so that you can use 
CPU() as a dynamic_cast macro.

BTW, if the class_size == parent.class_size you don't need to define or specific 
the class.

Regards,

Anthony Liguori

> +    .abstract = true,
> +    .class_size = sizeof(CPUClass),
> +};
> +
> +static void cpu_register_types(void)
> +{
> +    type_register_static(&cpu_type_info);
> +}
> +
> +processor_init(cpu_register_types)
> diff --git a/include/qemu/cpu.h b/include/qemu/cpu.h
> new file mode 100644
> index 0000000..4b81f3b
> --- /dev/null
> +++ b/include/qemu/cpu.h
> @@ -0,0 +1,27 @@
> +/*
> + * QEMU CPU model
> + *
> + * Copyright (c) 2012 SUSE LINUX Products GmbH
> + *
> + * Licensed under the terms of the GNU GPL version 2
> + * or (at your option) any later version.
> + */
> +#ifndef QEMU_CPU_H
> +#define QEMU_CPU_H
> +
> +#include "qemu/object.h"
> +
> +#define TYPE_CPU "cpu"
> +
> +typedef struct CPUClass {
> +    ObjectClass parent_class;
> +} CPUClass;
> +
> +typedef struct CPU {
> +    Object parent_obj;
> +
> +    /* TODO Move common CPUState here */
> +} CPU;
> +
> +
> +#endif
> diff --git a/linux-user/main.c b/linux-user/main.c
> index d4368b6..e727e8d 100644
> --- a/linux-user/main.c
> +++ b/linux-user/main.c
> @@ -3304,6 +3304,7 @@ int main(int argc, char **argv, char **envp)
>       }
>
>       cpu_model = NULL;
> +    module_call_init(MODULE_INIT_CPU);
>   #if defined(cpudef_setup)
>       cpudef_setup(); /* parse cpu definitions in target config file (TBD) */
>   #endif
> diff --git a/module.h b/module.h
> index 567ff3a..512ba6c 100644
> --- a/module.h
> +++ b/module.h
> @@ -26,6 +26,7 @@ typedef enum {
>       MODULE_INIT_DEVICE,
>       MODULE_INIT_MACHINE,
>       MODULE_INIT_QAPI,
> +    MODULE_INIT_CPU,
>       MODULE_INIT_MAX
>   } module_init_type;
>
> @@ -34,6 +35,7 @@ typedef enum {
>   #define device_init(function) module_init(function, MODULE_INIT_DEVICE)
>   #define machine_init(function) module_init(function, MODULE_INIT_MACHINE)
>   #define qapi_init(function) module_init(function, MODULE_INIT_QAPI)
> +#define processor_init(function) module_init(function, MODULE_INIT_CPU)
>
>   void register_module_init(void (*fn)(void), module_init_type type);
>

Am 30.01.2012 03:14, schrieb Anthony Liguori:
> On 01/29/2012 07:25 AM, Andreas Färber wrote:
>> +static TypeInfo cpu_type_info = {
>> +    .name = TYPE_CPU,
>> +    .parent = TYPE_OBJECT,
>> +    .instance_size = sizeof(CPU),
> 
> 
> Probably want to do CPUState or something of that nature so that you can
> use CPU() as a dynamic_cast macro.

Yeah, I stumbled upon that but didn't need it so far. CPUState is
already taken by the old concept.
In an earlier version I had used CPUCore (CPU_CORE()) but then I thought
that doesn't sound right for HyperThreading (CPU -> cores -> threads).
Since QEMU so far uses the term "CPU" for all of those I went with that.
Any better suggestions?

> BTW, if the class_size == parent.class_size you don't need to define or
> specific the class.

I'm sorry? You mean we don't need CPUClass yet because it doesn't add
anything to ObjectClass yet? I was planning on having a reset function
defined there, to be called in place of current aliased cpu_reset(). For
that to work I am emptying cpu_arm_reset() step by step (cf. 9/7) until
it can defer to:
(*CPU_GET_CLASS(ENV_GET_OBJECT(env))->reset)(ENV_GET_OBJECT(env));

I've been thinking along the lines of: if something doesn't change at
runtime then it goes into a class.

Or did you mean something else?

Andreas

Am 30.01.2012 03:14, schrieb Anthony Liguori:
> On 01/29/2012 07:25 AM, Andreas Färber wrote:
>> +static TypeInfo cpu_type_info = {
>> +    .name = TYPE_CPU,
>> +    .parent = TYPE_OBJECT,
>> +    .instance_size = sizeof(CPU),
> 
> 
> Probably want to do CPUState or something of that nature so that you can
> use CPU() as a dynamic_cast macro.

My testing seems to indicate CPU and CPU(obj) can co-exist (GCC 4.6.2).

Considered name alternatives:

CPUState - already taken as #define for legacy CPU${arch}State
CPUCore - fits ARM but there's socket vs. core vs. thread on x86
CPUCommon - puts the emphasis on common, which is not a noun
CommonCPU - no other base class uses such a naming scheme
CPUModel - it's not just a model but the state of one instance
CPUFamily - same as for CPUModel

CPUState would be nicest IMO, but refactoring CPUState into something
else just to refactor it back to QOM CPUState feels like churn.

Andreas