[GIT PULL] KVM/arm64 updates for Linux 7.4

[Marc Zyngier]

On Wed, 14 Oct 2026 17:30:50 +0100,
Paolo Bonzini <pbonzini at redhat.com> wrote:
> 
> On Wed, Oct 14, 2026 at 6:21 PM Marc Zyngier <maz at kernel.org> wrote:
> >
> > Paolo,
> >
> > Here's the set of KVM/arm64 updates for 7.4. After a couple of
> > releases mostly focused on mopping the AI wave, we're back to merging
> > new code, and there is a lot of it. Given the size and the conflicts
> > that this generates (see below), I've decided to send the pull request
> > early.
> >
> > Details are as usual in the tag below, but there is a few things to
> > call out:
> >
> > - a lot of pKVM updates, finally delivering on the isolation promises,
> >   thanks to Fuad's work on upstreaming this stuff.
> >
> > - the first stab at sharing code with s390... More fun to come, but
> >   that's a start
> >
> > - More GICv5 code. As always with the GIC, it is huge... But for once,
> >   the code is much simpler than the previous iterations, and also
> >   comes with a bunch of selftests from the beginning.
> >
> > As a consequence, the diffstat is a gold-plated pig.
> >
> > Things to note:
> >
> > - this drags the kvmarm-fixes-7.3-1 tag as a way to resolve bad
> >   conflicts with some of the patches
> >
> > - there is some additional conflicts with 7.3-rc6, due to
> >   kvmarm-fixes-7.3-2 landing a bit late. I have stashed a resolution
> >   here [1], which you can either apply if you're basing the merge on
> >   -rc6, or mention to Linus as you send your own PR. For everybody to
> >   enjoy, I also add the resolution at the end of this email.
> 
> I've done the unspeakable and pulled kvmarm-fixes-7.3-2 into kvm/next
> so that this PR will go on top; that conflict is too much to send to
> Linus. Please check in kvm/queue that I copied your resolution
> correctly. As soon as you confirm I'll move it to kvm/next.

Looks pretty good to me:

$ git diff kvmarm/next-7.4+fixes-7.3-2  kvm/queue arch/arm64/
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 1c6632216e327..5c9843338c510 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -32,6 +32,9 @@
 #include <asm/vncr_mapping.h>
 
 #define __KVM_HAVE_ARCH_INTC_INITIALIZED
+#define __KVM_HAVE_MSI
+#define __KVM_HAVE_READONLY_MEM
+#define __KVM_HAVE_VCPU_RUN_PID_CHANGE
 
 #define KVM_HALT_POLL_NS_DEFAULT 500000
 
diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
index 71233068b7cb6..48d531e302d33 100644
--- a/arch/arm64/kvm/Kconfig
+++ b/arch/arm64/kvm/Kconfig
@@ -22,18 +22,14 @@ menuconfig KVM
        select KVM_COMMON
        select KVM_GENERIC_HARDWARE_ENABLING
        select HAVE_KVM_CPU_RELAX_INTERCEPT
-       select KVM_MMIO
        select KVM_GENERIC_DIRTYLOG_READ_PROTECT
        select VIRT_XFER_TO_GUEST_WORK
        select KVM_VFIO
        select HAVE_KVM_DIRTY_RING_ACQ_REL
        select NEED_KVM_DIRTY_RING_WITH_BITMAP
-       select HAVE_KVM_MSI
        select HAVE_KVM_IRQCHIP
        select HAVE_KVM_IRQ_ROUTING
        select HAVE_KVM_IRQ_BYPASS
-       select HAVE_KVM_READONLY_MEM
-       select HAVE_KVM_VCPU_RUN_PID_CHANGE
        select SCHED_INFO
        select GUEST_PERF_EVENTS if PERF_EVENTS
        select KVM_GUEST_MEMFD

which is what I'd expected. It even builds, so it is obviously
perfect. Go ahead!

Thanks,

	M.

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