[PATCH 0/3] KVM: s390: arm64: Fixes for arm64 KVM support

[PATCH 0/3] KVM: s390: arm64: Fixes for arm64 KVM support

[Andreas Grapentin]

Some fixups for the previous series that introduced arm64 KVM support on
s390. Forward-declare enum vcpu_sysreg because we are definitely going
to need that, then make the module naming consistent and iron out an
inconsistency in the definition of vcpu->arch.pause between s390 and
native arm64 KVM.

No functional changes intended on either arm64 or s390.

Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
---
Andreas Grapentin (2):
      KVM: s390: arm64: Declare enum vcpu_sysreg
      KVM: s390: arm64: Fix kvm_arm64 module naming consistency

Steffen Eiden (1):
      KVM: s390: arm64: Turn vcpu->arch.pause into a counter

 arch/s390/include/asm/kvm_host_arm64.h |  8 ++++----
 arch/s390/kvm/arm64/Makefile           | 12 ++++++------
 2 files changed, 10 insertions(+), 10 deletions(-)
---
base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4
change-id: 20261103-s390_arm64_kvm_fixups-b285654f214c

Best regards,
-- 
Andreas Grapentin <gra@linux.ibm.com>

[PATCH 1/3] KVM: s390: arm64: Turn vcpu->arch.pause into a counter

[Andreas Grapentin]

From: Steffen Eiden <seiden@linux.ibm.com>

The s390 arm64 vcpu struct mirrors arm64's kvm_vcpu_arch. When commit
e64309198eda2 ("KVM: arm64: Turn vcpu->arch.pause into a counter")
changed the type of kvm_vcpu_arch.pause from bool to atomic_t, the merge
into this tree missed the s390 copy, leaving it out of sync.

Nothing on the s390 side accesses vcpu->arch.pause yet, so this has
no functional effect. Align the type to keep the mirror identical to
its arm64 counterpart.

Fixes: e6c329180d1d ("KVM: s390: arm64: Introduce host definitions")
Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
---
 arch/s390/include/asm/kvm_host_arm64.h | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/arch/s390/include/asm/kvm_host_arm64.h b/arch/s390/include/asm/kvm_host_arm64.h
index 82343e723ebf..a3e40650d75d 100644
--- a/arch/s390/include/asm/kvm_host_arm64.h
+++ b/arch/s390/include/asm/kvm_host_arm64.h
@@ -63,11 +63,10 @@ struct kvm_vcpu_arch {
 	/*
 	 * Don't run the guest (internal implementation need).
 	 *
-	 * Contrary to the flags above, this is set/cleared outside of
-	 * a vcpu context, and thus cannot be mixed with the flags
-	 * themselves (or the flag accesses need to be made atomic).
+	 * Contrary to the flags above, this is updated outside of
+	 * a vcpu context, and thus cannot be mixed with the flags.
 	 */
-	bool pause;
+	atomic_t pause;
 
 	/* vcpu power state */
 	struct kvm_mp_state mp_state;

-- 
2.55.0

Re: [PATCH 1/3] KVM: s390: arm64: Turn vcpu->arch.pause into a counter

[Andreas Grapentin]

On Nov 03 26, Andreas Grapentin wrote:
> From: Steffen Eiden <seiden@linux.ibm.com>
> 
> The s390 arm64 vcpu struct mirrors arm64's kvm_vcpu_arch. When commit
> e64309198eda2 ("KVM: arm64: Turn vcpu->arch.pause into a counter")
> changed the type of kvm_vcpu_arch.pause from bool to atomic_t, the merge
> into this tree missed the s390 copy, leaving it out of sync.
> 
> Nothing on the s390 side accesses vcpu->arch.pause yet, so this has
> no functional effect. Align the type to keep the mirror identical to
> its arm64 counterpart.
> 
> Fixes: e6c329180d1d ("KVM: s390: arm64: Introduce host definitions")
> Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
> Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
> ---
>  arch/s390/include/asm/kvm_host_arm64.h | 7 +++----
>  1 file changed, 3 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/s390/include/asm/kvm_host_arm64.h b/arch/s390/include/asm/kvm_host_arm64.h
> index 82343e723ebf..a3e40650d75d 100644
> --- a/arch/s390/include/asm/kvm_host_arm64.h
> +++ b/arch/s390/include/asm/kvm_host_arm64.h
> @@ -63,11 +63,10 @@ struct kvm_vcpu_arch {
>  	/*
>  	 * Don't run the guest (internal implementation need).
>  	 *
> -	 * Contrary to the flags above, this is set/cleared outside of
> -	 * a vcpu context, and thus cannot be mixed with the flags
> -	 * themselves (or the flag accesses need to be made atomic).
> +	 * Contrary to the flags above, this is updated outside of
> +	 * a vcpu context, and thus cannot be mixed with the flags.
>  	 */
> -	bool pause;
> +	atomic_t pause;
>  
>  	/* vcpu power state */
>  	struct kvm_mp_state mp_state;
> 
> -- 
> 2.55.0
> 

Reviewed-by: Andreas Grapentin <gra@linux.ibm.com>

Best,
-A

[PATCH 2/3] KVM: s390: arm64: Declare enum vcpu_sysreg

[Andreas Grapentin]

Forward-declare enum vcpu_sysreg in kvm_host_arm64.h. This should have
been included in the first iteration of the header, since we will need
it to talk about system registers.

Since the definition of enum vcpu_sysreg is not shared yet, and not used
on the s390 side, this should not cause any functional changes.

Fixes: e6c329180d1d ("KVM: s390: arm64: Introduce host definitions")
Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
---
 arch/s390/include/asm/kvm_host_arm64.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/s390/include/asm/kvm_host_arm64.h b/arch/s390/include/asm/kvm_host_arm64.h
index a3e40650d75d..3e69724f17c9 100644
--- a/arch/s390/include/asm/kvm_host_arm64.h
+++ b/arch/s390/include/asm/kvm_host_arm64.h
@@ -8,6 +8,7 @@
 #include <asm/debug.h>
 
 #define vcpu_gp_regs(v)		((v)->arch.sae_block.gpr)
+enum vcpu_sysreg;
 
 #include <arm64/kvm_host.h>
 #include <arm64/ptrace.h>

-- 
2.55.0

Re: [PATCH 2/3] KVM: s390: arm64: Declare enum vcpu_sysreg

[Steffen Eiden]

On Tue, Nov 03, 2026 at 05:37:03PM +0100, Andreas Grapentin wrote:
> Forward-declare enum vcpu_sysreg in kvm_host_arm64.h. This should have
> been included in the first iteration of the header, since we will need
> it to talk about system registers.
> 
> Since the definition of enum vcpu_sysreg is not shared yet, and not used
> on the s390 side, this should not cause any functional changes.
> 
> Fixes: e6c329180d1d ("KVM: s390: arm64: Introduce host definitions")
> Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>

Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>

Re: [PATCH 2/3] KVM: s390: arm64: Declare enum vcpu_sysreg

[Christian Borntraeger]

Am 03.11.26 um 17:37 schrieb Andreas Grapentin:
> Forward-declare enum vcpu_sysreg in kvm_host_arm64.h. This should have
> been included in the first iteration of the header, since we will need
> it to talk about system registers.
> 
> Since the definition of enum vcpu_sysreg is not shared yet, and not used
> on the s390 side, this should not cause any functional changes.
> 
> Fixes: e6c329180d1d ("KVM: s390: arm64: Introduce host definitions")
> Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>

Acked-by: Christian Borntraeger <borntraeger@linux.ibm.com>


> ---
>   arch/s390/include/asm/kvm_host_arm64.h | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/arch/s390/include/asm/kvm_host_arm64.h b/arch/s390/include/asm/kvm_host_arm64.h
> index a3e40650d75d..3e69724f17c9 100644
> --- a/arch/s390/include/asm/kvm_host_arm64.h
> +++ b/arch/s390/include/asm/kvm_host_arm64.h
> @@ -8,6 +8,7 @@
>   #include <asm/debug.h>
>   
>   #define vcpu_gp_regs(v)		((v)->arch.sae_block.gpr)
> +enum vcpu_sysreg;
>   
>   #include <arm64/kvm_host.h>
>   #include <arm64/ptrace.h>
> 

[PATCH 3/3] KVM: s390: arm64: Fix kvm_arm64 module naming consistency

[Andreas Grapentin]

the module build for the kvm_arm64 module produced kvm-arm64.ko, fix the
Makefile to produce kvm_arm64.ko instead, so the output file name
matches the module name.

Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
---
 arch/s390/kvm/arm64/Makefile | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/s390/kvm/arm64/Makefile b/arch/s390/kvm/arm64/Makefile
index 3f0fe5c9f0a1..b48dc65b7aaa 100644
--- a/arch/s390/kvm/arm64/Makefile
+++ b/arch/s390/kvm/arm64/Makefile
@@ -29,7 +29,7 @@ $(obj)/%-arm64.o: $(src)/%.c FORCE
 	$(call if_changed_rule,cc_o_c)
 	$(call cmd,force_checksrc)
 
-# %-arm64.o stems don't match any kvm-arm64-y entry so Kbuild derives wrong
+# %-arm64.o stems don't match any kvm_arm64-y entry so Kbuild derives wrong
 # KBUILD_MODNAME/KBUILD_MODFILE. Force include a generated header to override.
 KVM_ARM64_MODNAME_H := $(obj)/generated/kvm_arm64_modname.h
 ccflags-y += -include $(KVM_ARM64_MODNAME_H)
@@ -48,7 +48,7 @@ define filechk_kvm_arm64_modname
 	echo '#undef __KBUILD_MODNAME';				\
 	echo '#define __KBUILD_MODNAME $(KVM_DEV_NAME)';	\
 	echo '#undef KBUILD_MODFILE';				\
-	echo '#define KBUILD_MODFILE "$(obj)/kvm-arm64"';	\
+	echo '#define KBUILD_MODFILE "$(obj)/kvm_arm64"';	\
 	echo '#endif /* __KVM_ARM64_MODNAME_H */'
 endef
 
@@ -60,7 +60,7 @@ $(addprefix $(obj)/,$(kvm-arm64-obj)): private part-of-builtin := y
 
 ifeq ($(CONFIG_KVM),m)
 
-kvm-arm64-y = $(kvm-arm64-obj)
+kvm_arm64-y = $(kvm-arm64-obj)
 
 else ifeq ($(CONFIG_KVM),y)
 
@@ -86,7 +86,7 @@ $(obj)/kvm_symbol_list: $(obj)/kvm-unnamespaced.o FORCE
 
 # cmd_oc_o_o carries -DKBUILD_MODFILE so generate_builtin_ranges.awk
 # can attribute kvm-namespaced.o to kvm_arm64 via its .cmd file.
-cmd_oc_o_o = $(cmd_objcopy) -DKBUILD_MODFILE='"$(obj)/kvm-arm64"'
+cmd_oc_o_o = $(cmd_objcopy) -DKBUILD_MODFILE='"$(obj)/kvm_arm64"'
 define rule_oc_o_o
 	$(call cmd,objcopy)
 	@printf '%s\n' 'savedcmd_$@ := $(call make-cmd,oc_o_o)' > $(dot-target).cmd
@@ -98,7 +98,7 @@ OBJCOPYFLAGS_kvm-namespaced.o := -O $(LD_BFD) --localize-symbols=$(obj)/kvm_symb
 $(obj)/kvm-namespaced.o: $(obj)/kvm-unnamespaced.o $(obj)/kvm_symbol_list FORCE
 	$(call if_changed_rule,oc_o_o)
 
-kvm-arm64-y = kvm-namespaced.o
+kvm_arm64-y = kvm-namespaced.o
 
 targets += kvm-unnamespaced.o kvm_symbol_list kvm-namespaced.o
 
@@ -106,5 +106,5 @@ endif
 
 clean-files += kvm-unnamespaced.o kvm_symbol_list kvm-namespaced.o
 
-obj-$(CONFIG_KVM) += kvm-arm64.o
+obj-$(CONFIG_KVM) += kvm_arm64.o
 

-- 
2.55.0

Re: [PATCH 3/3] KVM: s390: arm64: Fix kvm_arm64 module naming consistency

[Steffen Eiden]

On Tue, Nov 03, 2026 at 05:37:04PM +0100, Andreas Grapentin wrote:
> the module build for the kvm_arm64 module produced kvm-arm64.ko, fix the
Nit: start with an upper case T

> Makefile to produce kvm_arm64.ko instead, so the output file name
> matches the module name.
> 
> Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>

Reviewed-by: Steffen Eiden <seiden@linux.ibm.com>

Re: [PATCH 3/3] KVM: s390: arm64: Fix kvm_arm64 module naming consistency

[Christian Borntraeger]

Am 03.11.26 um 17:37 schrieb Andreas Grapentin:
> the module build for the kvm_arm64 module produced kvm-arm64.ko, fix the
> Makefile to produce kvm_arm64.ko instead, so the output file name
> matches the module name.
> 
> Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>

Its a stretch, but maybe add
Fixes: 91b87e6eea4d0 ("KVM: s390: arm64: Integrate arm on s390")
so that any backports will not miss this.

Acked-by: Christian Borntraeger <borntraeger@linux.ibm.com>

Re: [PATCH 0/3] KVM: s390: arm64: Fixes for arm64 KVM support

[Marc Zyngier]

On Tue, 03 Nov 2026 16:37:01 +0000,
Andreas Grapentin <gra@linux.ibm.com> wrote:
> 
> Some fixups for the previous series that introduced arm64 KVM support on
> s390. Forward-declare enum vcpu_sysreg because we are definitely going
> to need that, then make the module naming consistent and iron out an
> inconsistency in the definition of vcpu->arch.pause between s390 and
> native arm64 KVM.

It's an interesting one. Obviously, this wasn't caught by building the
upstream code. I guess this triggered with the rest of the patches?

> 
> No functional changes intended on either arm64 or s390.
> 
> Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
> ---
> Andreas Grapentin (2):
>       KVM: s390: arm64: Declare enum vcpu_sysreg
>       KVM: s390: arm64: Fix kvm_arm64 module naming consistency
> 
> Steffen Eiden (1):
>       KVM: s390: arm64: Turn vcpu->arch.pause into a counter
> 
>  arch/s390/include/asm/kvm_host_arm64.h |  8 ++++----
>  arch/s390/kvm/arm64/Makefile           | 12 ++++++------
>  2 files changed, 10 insertions(+), 10 deletions(-)

Looks all reasonable to me.

Acked-by: Marc Zyngier <maz@kernel.org>

	M.

-- 
Without deviation from the norm, progress is not possible.

Re: [PATCH 0/3] KVM: s390: arm64: Fixes for arm64 KVM support

[Steffen Eiden]

On Tue, Nov 03, 2026 at 06:31:31PM +0000, Marc Zyngier wrote:
> On Tue, 03 Nov 2026 16:37:01 +0000,
> Andreas Grapentin <gra@linux.ibm.com> wrote:
> > 
> > Some fixups for the previous series that introduced arm64 KVM support on
> > s390. Forward-declare enum vcpu_sysreg because we are definitely going
> > to need that, then make the module naming consistent and iron out an
> > inconsistency in the definition of vcpu->arch.pause between s390 and
> > native arm64 KVM.
> 
> It's an interesting one. Obviously, this wasn't caught by building the
> upstream code. I guess this triggered with the rest of the patches?

Yes, this was triggered by the rest of the patches. The field somehow
got in the first batch although it is not needed. We'll need it later
anyways so I did not delete it for this fixup.

> 
> > 
> > No functional changes intended on either arm64 or s390.
> > 
> > Signed-off-by: Andreas Grapentin <gra@linux.ibm.com>
> > ---
> > Andreas Grapentin (2):
> >       KVM: s390: arm64: Declare enum vcpu_sysreg
> >       KVM: s390: arm64: Fix kvm_arm64 module naming consistency
> > 
> > Steffen Eiden (1):
> >       KVM: s390: arm64: Turn vcpu->arch.pause into a counter
> > 
> >  arch/s390/include/asm/kvm_host_arm64.h |  8 ++++----
> >  arch/s390/kvm/arm64/Makefile           | 12 ++++++------
> >  2 files changed, 10 insertions(+), 10 deletions(-)
> 
> Looks all reasonable to me.
> 
> Acked-by: Marc Zyngier <maz@kernel.org>

Thanks.

	Steffen