Re: [PATCH RESEND v4 2/2] arm/arm64: KVM: Add KVM_GET/SET_VCPU_EVENTS

[gengdongjiu]

> 
> > >>>>> --- a/arch/arm/include/uapi/asm/kvm.h
> > >>>>> +++ b/arch/arm/include/uapi/asm/kvm.h
> > >>>>> @@ -124,6 +124,18 @@ struct kvm_sync_regs {  struct
> > >>>>> kvm_arch_memory_slot {  };
> > >>>>>
> > >>>>> +/* for KVM_GET/SET_VCPU_EVENTS */ struct kvm_vcpu_events {
> > >>>>> +	struct {
> > >>>>> +		__u8 serror_pending;
> > >>>>> +		__u8 serror_has_esr;
> > >>>>> +		/* Align it to 8 bytes */
> > >>>>> +		__u8 pad[6];
> > >>>>> +		__u64 serror_esr;
> > >>>>> +	} exception;
> > >>>>> +	__u32 reserved[12];
> > >>>>> +};
> > >>>>> +
> > >>>>
> > >>>> You haven't defined __KVM_HAVE_VCPU_EVENTS for 32bit, so
> > >>>> presumably this struct will never be used. Why is it here?
> > >>
> > >>>   if not add it for 32 bits. the 32 arm platform will build Fail, whether you have good
> > >>>    idea to avoid this Failure if not add this struct for the 32 bit?
> > >>
> > >> How does this 32bit code build without this patch?
> > >> If do you provide the struct, how will that code build with older headers?
> > >>
> > >> As far as I can see, this is what the __KVM_HAVE_VCPU_EVENTS define is for.
> > >>
> > >> This should be both, or neither. Having just the struct is useless.
> > > It because the caller of kvm_arm_vcpu_get/set_events() is in "virt/kvm/arm/arm.c".
> > > the virt/kvm/arm/arm.c will used by both arm64 and arm.
> > > so It needs to add kvm_arm_vcpu_get/set_events() for the 32 bits,
> > > however, kvm_arm_vcpu_get/set_events() will directly return,
> >
> > So you are adding a uapi struct that user-space can't actually use, to avoid a kernel build-error. Fine, it just looks really strange.
> >
> > 32bit user-space shouldn't try to call this as check-extension reports
> > it as not present. If it does, it gets -EINVAL back, which is also the default for kvm_arch_vcpu_ioctl().

Just think out a method to avoid adding the structure for the arm32.

Using below methods:

#Ifdef __KVM_HAVE_VCPU_EVENTS
	xxxxxxxxxxxxxxxxxxxxx
#else
	Xxxxxxxxxxxxxxxx
#endif

> 
> It indeed looks strange. Because "virt/kvm/arm/arm.c" is shared by arm64 and arm32, if not adding, it will build-error. Do you have a better
> idea to avoid adding in arm32?
> I still not think out a good method. 32 user-space will not call this KVM_GET/SET_VCPU_EVENTS IOCTL as check-extension reports it as not
> present.
> 
> >
> >
> > Thanks,
> >
> > James