Re: [PATCH 2/2] sysemu/kvm: Reduce target-specific declarations

[Philippe Mathieu-Daudé]

On 17/12/22 01:28, Richard Henderson wrote:
> On 12/16/22 14:07, Philippe Mathieu-Daudé wrote:
> > Only the declarations using the target_ulong type are
> > target specific.
> >
> > Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
> > ---
> >   include/sysemu/kvm.h | 25 ++++++++++++-------------
> >   1 file changed, 12 insertions(+), 13 deletions(-)
> >
> > diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
> > index c8281c07a7..a53d6dab49 100644
> > --- a/include/sysemu/kvm.h
> > +++ b/include/sysemu/kvm.h
> > @@ -242,9 +242,6 @@ bool kvm_arm_supports_user_irq(void);
> >   int kvm_on_sigbus_vcpu(CPUState *cpu, int code, void *addr);
> >   int kvm_on_sigbus(int code, void *addr);
> > -#ifdef NEED_CPU_H
> > -#include "cpu.h"
> > -
> >   void kvm_flush_coalesced_mmio_buffer(void);
> >   /**
> > @@ -410,6 +407,9 @@ void kvm_get_apic_state(DeviceState *d, struct 
> > kvm_lapic_state *kapic);
> >   struct kvm_guest_debug;
> >   struct kvm_debug_exit_arch;
> > +#ifdef NEED_CPU_H
> > +#include "cpu.h"
> > +
> >   struct kvm_sw_breakpoint {
> >       target_ulong pc;
> >       target_ulong saved_insn;
> > @@ -436,6 +436,15 @@ void kvm_arch_update_guest_debug(CPUState *cpu, 
> > struct kvm_guest_debug *dbg);
> >   bool kvm_arch_stop_on_emulation_error(CPUState *cpu);
> > +uint32_t kvm_arch_get_supported_cpuid(KVMState *env, uint32_t function,
> > +                                      uint32_t index, int reg);
> > +uint64_t kvm_arch_get_supported_msr_feature(KVMState *s, uint32_t 
> > index);
> > +
> > +int kvm_physical_memory_addr_from_host(KVMState *s, void *ram_addr,
> > +                                       hwaddr *phys_addr);
>
> Why did these need to move?

kvm_arch_get_XXX() don't need to move, but they are only called from
target-specific code, so there is no point in declaring them for
target-agnostic part; besides that helps catching unnecessary
target-specific stuff built within target-agnostic code.


Normally kvm_enabled() shouldn't be use by user-mode code; we could
poison it. In practice we have few common files calling it, so this
header is included, declaring kvm_physical_memory_addr_from_host()
which uses hwaddr. I'm trying to not define hwaddr in user-mode.

I thought this patch would be trivial :/ I'll split it with tiny /
better justified ones.

Thanks,

Phil.