Re: [Qemu-devel] [PATCH v3 2/7] qom/cpu: move register_vmstate to common CPUClass.realizefn

[Igor Mammedov]

On Wed, 14 Jan 2015 15:27:25 +0800
Zhu Guihua <address@hidden> wrote:

> From: Gu Zheng <address@hidden>
> 
> Move cpu vmstate register from cpu_exec_init into cpu_common_realizefn,
> and use cc->get_arch_id as the instance id that suggested by Igor to
> fix the migration issue.
> 
> Signed-off-by: Gu Zheng <address@hidden>
> Signed-off-by: Zhu Guihua <address@hidden>
> ---
>  exec.c            | 32 +++++++++++++++++++-------------
>  include/qom/cpu.h |  2 ++
>  qom/cpu.c         |  2 ++
>  3 files changed, 23 insertions(+), 13 deletions(-)
> 
> diff --git a/exec.c b/exec.c
> index 081818e..c9ffda6 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -513,10 +513,28 @@ void tcg_cpu_address_space_init(CPUState *cpu, 
> AddressSpace *as)
>  }
>  #endif
>  
> +void cpu_vmstate_register(CPUState *cpu)
> +{
> +    CPUClass *cc = CPU_GET_CLASS(cpu);
> +    int cpu_index = cc->get_arch_id(cpu);
that probable would be source migration problems:
because cc->get_arch_id(cpu) depending on topology might
be not sequential, for example: sockets=4,core=3
that would create sparse APIC numbering.

as result migration from old qemu to one with this change
would be rejected due to vmsd id mismatch mismatch.

we need a better way to handle cross version migration
between old/new scheme.

> +
> +    if (qdev_get_vmsd(DEVICE(cpu)) == NULL) {
> +        vmstate_register(NULL, cpu_index, &vmstate_cpu_common, cpu);
> +    }
> +#if defined(CPU_SAVE_VERSION) && !defined(CONFIG_USER_ONLY)
that ifdef block affects only sparc/mips/cris and builds target specific code
while you are trying to call it from target independent qom/cpu.c

I'd suggest leave it where it was or better move into respective
targets realize_fns

> +    register_savevm(NULL, "cpu", cpu_index, CPU_SAVE_VERSION,
> +                    cpu_save, cpu_load, cpu->env_ptr);
> +    assert(cc->vmsd == NULL);
> +    assert(qdev_get_vmsd(DEVICE(cpu)) == NULL);
> +#endif
> +    if (cc->vmsd != NULL) {
> +        vmstate_register(NULL, cpu_index, cc->vmsd, cpu);
> +    }
> +}
> +
>  void cpu_exec_init(CPUArchState *env)
>  {
>      CPUState *cpu = ENV_GET_CPU(env);
> -    CPUClass *cc = CPU_GET_CLASS(cpu);
>      CPUState *some_cpu;
>      int cpu_index;
>  
> @@ -539,18 +557,6 @@ void cpu_exec_init(CPUArchState *env)
>  #if defined(CONFIG_USER_ONLY)
>      cpu_list_unlock();
>  #endif
> -    if (qdev_get_vmsd(DEVICE(cpu)) == NULL) {
> -        vmstate_register(NULL, cpu_index, &vmstate_cpu_common, cpu);
> -    }
> -#if defined(CPU_SAVE_VERSION) && !defined(CONFIG_USER_ONLY)
> -    register_savevm(NULL, "cpu", cpu_index, CPU_SAVE_VERSION,
> -                    cpu_save, cpu_load, env);
> -    assert(cc->vmsd == NULL);
> -    assert(qdev_get_vmsd(DEVICE(cpu)) == NULL);
> -#endif
> -    if (cc->vmsd != NULL) {
> -        vmstate_register(NULL, cpu_index, cc->vmsd, cpu);
> -    }
And in general do CONFIG_USER_ONLY targets actually need/use
migration code?

>  }
>  
>  #if defined(TARGET_HAS_ICE)
> diff --git a/include/qom/cpu.h b/include/qom/cpu.h
> index 2098f1c..936afcd 100644
> --- a/include/qom/cpu.h
> +++ b/include/qom/cpu.h
> @@ -562,6 +562,8 @@ void cpu_interrupt(CPUState *cpu, int mask);
>  
>  #endif /* USER_ONLY */
>  
> +void cpu_vmstate_register(CPUState *cpu);
> +
>  #ifdef CONFIG_SOFTMMU
>  static inline void cpu_unassigned_access(CPUState *cpu, hwaddr addr,
>                                           bool is_write, bool is_exec,
> diff --git a/qom/cpu.c b/qom/cpu.c
> index 9c68fa4..a639822 100644
> --- a/qom/cpu.c
> +++ b/qom/cpu.c
> @@ -302,6 +302,8 @@ static void cpu_common_realizefn(DeviceState *dev, Error 
> **errp)
>  {
>      CPUState *cpu = CPU(dev);
>  
> +    cpu_vmstate_register(cpu);
> +
>      if (dev->hotplugged) {
>          cpu_synchronize_post_init(cpu);
>          cpu_resume(cpu);