(subset) [PATCH v9 00/24] KVM: s390: Introduce arm64 KVM

[Steffen Eiden]

On Thu, Oct 01, 2026 at 04:12:38PM +0100, Marc Zyngier wrote:
> On Thu, 01 Oct 2026 14:27:15 +0100,
> Steffen Eiden <seiden at linux.ibm.com> wrote:
> > 
> > On Thu, Oct 01, 2026 at 12:04:51PM +0100, Marc Zyngier wrote:
> > > On Wed, 30 Sep 2026 19:34:15 +0200, Steffen Eiden wrote:
> > > > Introduce arm-on-s390. Enable KVM-accelerated ARM CPU virtualisation on s390.
> > > > 
> > > > The new KVM functionally depends on KVM/vfio: Use file-based reference
> > > > counting for KVM [5] The series itself should apply w/o those.
> > > > 
> > > > For the original, now partially outdated, cover letter, have a look at versions
> > > > one to three [1].
> > > > 
> > > > [...]
> > > 
> > > Applied to next, thanks!
> > > 
> > > [05/24] arm64: ptrace: Use constants for compat register numbers
> > >         commit: 295d3df9b1e425099a4eb37954a02c4a08fdb496
> > > [06/24] arm64: sysreg: Convert SPSR_ELx to automatic register generation
> > >         commit: d3d78520068779779e1ed7aa0dd8a3ce63e1ca43
> > > [07/24] KVM: arm64: Access elements of vcpu_gp_regs individually
> > >         commit: 29d92585ab987bb80b1e78d811aaea2a8db86e62
> > > [08/24] KVM: arm64: Use accessor functions for core regs
> > >         commit: 89ed62818f10e38e64fea46d43ee6455a1beb647
> > > [09/24] arm64: Prepare sharing arm64 headers with s390
> > >         commit: df6a70516adec8b7f5fa49e1755d8ee33d980d9c
> > > [10/24] arm64: Share arm64 headers with s390
> > >         commit: c7e18ba462536e870b0ff4deca581babcac9135b
> > > [11/24] KVM: arm64: Share arm64 code with s390
> > >         commit: dcb776bfcccc9662001a75b99dbfbef55f46eb31
> > > 
> > > Note that there were a few conflicts. Please check that I resolved
> > > them correctly.
> > 
> > 
> > After rebasing on those the arm-on-s390 part does not build anymore :( 
> 
> Well, that's sort of expected. The kvmarm tree doesn't have the S390
> stuff yet, and contains tons of new stuff.
>

yes, but sometimes I like to dream 

> > We are missing the definition of
> > static __always_inline bool esr_abt_is_s1ptw(unsigned long esr)
> > 
> > 
> > quick fix: 
> > 
> > on top of [10/24]  arm64: Share arm64 headers with s390
> > 
> > diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> > index bf35cbe8ddcb..0708f792c631 100644
> > --- a/arch/arm64/include/asm/kvm_emulate.h
> > +++ b/arch/arm64/include/asm/kvm_emulate.h
> > @@ -367,10 +367,13 @@ static __always_inline u64 kvm_vcpu_get_esr(const struct kvm_vcpu *vcpu)
> >         return vcpu->arch.fault.esr_el2;
> >  }
> > 
> > +#if ARM64_S390_COMMON == 1
> >  static __always_inline bool esr_abt_is_s1ptw(unsigned long esr)
> >  {
> >         return esr & ESR_ELx_S1PTW;
> >  }
> > +
> > +#endif /* ARM64_S390_COMMON == 1 */
> > 
> >  static __always_inline bool esr_abt_is_exec_fault(unsigned long esr)
> >  {
> > 
> > 
> > 
> > A cleaner solution it to move this defenition just below  
> > void kvm_vcpu_wfi(struct kvm_vcpu *vcpu) in 
> > [09/24] arm64: Prepare sharing arm64 headers with s390
> > and expand the shared region in the following patch.
> 
> I'm redoing the merge and fixing it as follows (note that the
> ARM64_S390_COMMON guards are already present):
> 
> @@ -336,16 +367,6 @@ static __always_inline u64 kvm_vcpu_get_esr(const struct kvm_vcpu *vcpu)
>  	return vcpu->arch.fault.esr_el2;
>  }
>  
> -static __always_inline bool esr_abt_is_s1ptw(unsigned long esr)
> -{
> -	return esr & ESR_ELx_S1PTW;
> -}
> -
> -static __always_inline bool esr_abt_is_exec_fault(unsigned long esr)
> -{
> -	return esr_trap_is_iabt(esr) && !esr_abt_is_s1ptw(esr);
> -}
> -
>  static inline bool guest_hyp_wfx_traps_enabled(const struct kvm_vcpu *vcpu)
>  {
>  	u64 esr = kvm_vcpu_get_esr(vcpu);
> @@ -389,6 +410,12 @@ static inline u64 kvm_vcpu_get_disr(const struct kvm_vcpu *vcpu)
>  	return vcpu->arch.fault.disr_el1;
>  }
>  
> +#if ARM64_S390_COMMON == 1
> +static __always_inline bool esr_abt_is_s1ptw(unsigned long esr)
> +{
> +	return esr & ESR_ELx_S1PTW;
> +}
> +
>  static inline u32 kvm_vcpu_hvc_get_imm(const struct kvm_vcpu *vcpu)
>  {
>  	return kvm_vcpu_get_esr(vcpu) & ESR_ELx_xVC_IMM_MASK;
> @@ -456,6 +483,13 @@ static __always_inline u8 kvm_vcpu_trap_get_fault(const struct kvm_vcpu *vcpu)
>  	return kvm_vcpu_get_esr(vcpu) & ESR_ELx_FSC;
>  }
>  
> +#endif /* ARM64_S390_COMMON == 1 */
> +
> +static __always_inline bool esr_abt_is_exec_fault(unsigned long esr)
> +{
> +	return esr_trap_is_iabt(esr) && !esr_abt_is_s1ptw(esr);
> +}
> +
>  static inline
>  bool kvm_vcpu_trap_is_permission_fault(const struct kvm_vcpu *vcpu)
>  {
>
> 
> Does this work for you?
>

Yes the positioning of the functions look good. It builds on top of
kvmarm/next + this (sightly relocated (line numbes are different) patch
+ arm on s390 stuff.

I'll work on a verification tool for you guys to make this process
easier.

Thank you.

	Steffen