diff mbox series

KVM: LAPIC: Also cancel preemption timer during SET_LAPIC

Message ID 1643102220-35667-1-git-send-email-wanpengli@tencent.com (mailing list archive)
State New, archived
Headers show
Series KVM: LAPIC: Also cancel preemption timer during SET_LAPIC | expand

Commit Message

Wanpeng Li Jan. 25, 2022, 9:17 a.m. UTC
From: Wanpeng Li <wanpengli@tencent.com>

The below warning is splatting during guest reboot.

  ------------[ cut here ]------------
  WARNING: CPU: 0 PID: 1931 at arch/x86/kvm/x86.c:10322 kvm_arch_vcpu_ioctl_run+0x874/0x880 [kvm]
  CPU: 0 PID: 1931 Comm: qemu-system-x86 Tainted: G          I       5.17.0-rc1+ #5
  RIP: 0010:kvm_arch_vcpu_ioctl_run+0x874/0x880 [kvm]
  Call Trace:
   <TASK>
   kvm_vcpu_ioctl+0x279/0x710 [kvm]
   __x64_sys_ioctl+0x83/0xb0
   do_syscall_64+0x3b/0xc0
   entry_SYSCALL_64_after_hwframe+0x44/0xae
  RIP: 0033:0x7fd39797350b

This can be triggered by not exposing tsc-deadline mode and doing a reboot in
the guest. The lapic_shutdown() function which is called in sys_reboot path
will not disarm the flying timer, it just masks LVTT. lapic_shutdown() clears
APIC state w/ LVT_MASKED and timer-mode bit is 0, this can trigger timer-mode
switch between tsc-deadline and oneshot/periodic, which can result in preemption
timer be cancelled in apic_update_lvtt(). However, We can't depend on this when 
not exposing tsc-deadline mode and oneshot/periodic modes emulated by preemption 
timer. Qemu will synchronise states around reset, let's cancel preemption timer 
under KVM_SET_LAPIC.

Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
---
 arch/x86/kvm/lapic.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Paolo Bonzini Jan. 25, 2022, 11:04 a.m. UTC | #1
On 1/25/22 10:17, Wanpeng Li wrote:
> From: Wanpeng Li <wanpengli@tencent.com>
> 
> The below warning is splatting during guest reboot.
> 
>    ------------[ cut here ]------------
>    WARNING: CPU: 0 PID: 1931 at arch/x86/kvm/x86.c:10322 kvm_arch_vcpu_ioctl_run+0x874/0x880 [kvm]
>    CPU: 0 PID: 1931 Comm: qemu-system-x86 Tainted: G          I       5.17.0-rc1+ #5
>    RIP: 0010:kvm_arch_vcpu_ioctl_run+0x874/0x880 [kvm]
>    Call Trace:
>     <TASK>
>     kvm_vcpu_ioctl+0x279/0x710 [kvm]
>     __x64_sys_ioctl+0x83/0xb0
>     do_syscall_64+0x3b/0xc0
>     entry_SYSCALL_64_after_hwframe+0x44/0xae
>    RIP: 0033:0x7fd39797350b
> 
> This can be triggered by not exposing tsc-deadline mode and doing a reboot in
> the guest. The lapic_shutdown() function which is called in sys_reboot path
> will not disarm the flying timer, it just masks LVTT. lapic_shutdown() clears
> APIC state w/ LVT_MASKED and timer-mode bit is 0, this can trigger timer-mode
> switch between tsc-deadline and oneshot/periodic, which can result in preemption
> timer be cancelled in apic_update_lvtt(). However, We can't depend on this when
> not exposing tsc-deadline mode and oneshot/periodic modes emulated by preemption
> timer. Qemu will synchronise states around reset, let's cancel preemption timer
> under KVM_SET_LAPIC.
> 
> Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
> ---
>   arch/x86/kvm/lapic.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> index baca9fa37a91..4662469240bc 100644
> --- a/arch/x86/kvm/lapic.c
> +++ b/arch/x86/kvm/lapic.c
> @@ -2629,7 +2629,7 @@ int kvm_apic_set_state(struct kvm_vcpu *vcpu, struct kvm_lapic_state *s)
>   	kvm_apic_set_version(vcpu);
>   
>   	apic_update_ppr(apic);
> -	hrtimer_cancel(&apic->lapic_timer.timer);
> +	cancel_apic_timer(apic);
>   	apic->lapic_timer.expired_tscdeadline = 0;
>   	apic_update_lvtt(apic);
>   	apic_manage_nmi_watchdog(apic, kvm_lapic_get_reg(apic, APIC_LVT0));

Queued, with Cc to stable.

Paolo
diff mbox series

Patch

diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index baca9fa37a91..4662469240bc 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -2629,7 +2629,7 @@  int kvm_apic_set_state(struct kvm_vcpu *vcpu, struct kvm_lapic_state *s)
 	kvm_apic_set_version(vcpu);
 
 	apic_update_ppr(apic);
-	hrtimer_cancel(&apic->lapic_timer.timer);
+	cancel_apic_timer(apic);
 	apic->lapic_timer.expired_tscdeadline = 0;
 	apic_update_lvtt(apic);
 	apic_manage_nmi_watchdog(apic, kvm_lapic_get_reg(apic, APIC_LVT0));