[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240920080012.74405-1-mankku@gmail.com>
Date: Fri, 20 Sep 2024 10:59:42 +0300
From: Markku Ahvenjärvi <mankku@...il.com>
To: Sean Christopherson <seanjc@...gle.com>,
Paolo Bonzini <pbonzini@...hat.com>,
Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...hat.com>,
Borislav Petkov <bp@...en8.de>,
Dave Hansen <dave.hansen@...ux.intel.com>,
x86@...nel.org,
"H. Peter Anvin" <hpa@...or.com>
Cc: mankku@...il.com,
janne.karhunen@...il.com,
kvm@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: [PATCH 0/1] KVM: nVMX: update VPPR on vmlaunch/vmresume
Hello,
We experience hanging of the guest when running a hypervisor under KVM on
VMX. The L1 guest hypervisor in this particular case is pKVM for Intel
Architecture [1]. The hang occurs when a second nested guest is launched (the
first being de-privileged host). We observed that external interrupt
vmexit would not be passed to L1, instead L0 would attempt to resume L2.
We isolated the problem to VPPR not being updated on nested vmlaunch/vmresume,
and that causes vmx_has_apicv_interrupt() in nested_vmx_enter_non_root_mode()
to miss interrupts. Updating VPPR in vmx_has_apicv_interrupt() ensures VPPR
to be up-to-date.
We don't fully understand why VPPR problem appears with pKVM-IA as L1, but not
with normal KVM as L1. On pKVM-IA some of the host functionality is moved from
vmx root to non-root, but I would appreciate if someone could clarify why
normal KVM as L1 is seemingly unaffected.
Thanks,
Markku
[1]: https://lore.kernel.org/kvm/20230312180048.1778187-1-jason.cj.chen@intel.com
Markku Ahvenjärvi (1):
KVM: nVMX: update VPPR on vmlaunch/vmresume
arch/x86/kvm/lapic.c | 9 +++++----
arch/x86/kvm/lapic.h | 1 +
arch/x86/kvm/vmx/nested.c | 5 +++--
3 files changed, 9 insertions(+), 6 deletions(-)
--
2.44.1
Powered by blists - more mailing lists