Skip to contentSkip to conversation

KVM: x86: Properly handle APF vs disabled LAPIC situation

PATCH In review Guoqing Jiang 2 participants 2 messages May 24, 2022 – May 26, 2022 Stable branch

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: x86: Properly handle APF vs disabled LAPIC situation

From: Vitaly Kuznetsov <[email protected]>

Backport of commit 2f15d027c05fac406decdb5eceb9ec0902b68f53 upstream.

Async PF 'page ready' event may happen when LAPIC is (temporary) disabled. In particular, Sebastien reports that when Linux kernel is directly booted by Cloud Hypervisor, LAPIC is 'software disabled' when APF mechanism is initialized. On initialization KVM tries to inject 'wakeup all' event and puts the corresponding token to the slot. It is, however, failing to inject an interrupt (kvm_apic_set_irq() -> __apic_accept_irq() -> !apic_enabled()) so the guest never gets notified and the whole APF mechanism gets stuck. The same issue is likely to happen if the guest temporary disables LAPIC and a previously unavailable page becomes available.

Do two things to resolve the issue:
- Avoid dequeuing 'page ready' events from APF queue when LAPIC is disabled.
- Trigger an attempt to deliver pending 'page ready' events when LAPIC becomes enabled (SPIV or MSR_IA32_APICBASE).

Message-Id: <[email protected]>

[Guoqing: backport to 5.10-stable ]

---
Hi,

We encountered below task hang issue with 5.10.113 stable kernel.

[  246.845061] INFO: task rpmbuild:2303 blocked for more than 122 seconds.
[  246.846269]       Not tainted 5.10.113-1.1.se2-default #1
[  246.847103] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[  246.848248] task:rpmbuild        state:D stack:    0 pid: 2303 ppid:  2302 flags:0x00000000
[  246.848252] Call Trace:
[  246.848266]  __schedule+0x3f6/0x7c0
[  246.848289]  ? __handle_mm_fault+0x3dd/0x6d0
[  246.848291]  schedule+0x46/0xb0
[  246.848295]  kvm_async_pf_task_wait_schedule+0x4b/0x90
[  246.848297]  ? handle_mm_fault+0xbc/0x280
[  246.848300]  __kvm_handle_async_pf+0x4f/0xb0
[  246.848303]  exc_page_fault+0x204/0x540
[  246.848305]  ? asm_exc_page_fault+0x8/0x30
[  246.848307]  asm_exc_page_fault+0x1e/0x30
[  246.848310] RIP: 0033:0x7f122fbdfc90

And after investigating, this patch resolve the issue. 5.12 stable kernel has already merged it by commit 36825931c607.

Thanks,
Guoqing

 arch/x86/kvm/lapic.c | 6 ++++++
 arch/x86/kvm/x86.c   | 2 +-
 2 files changed, 7 insertions(+), 1 deletion(-)
arch/x86/kvm/lapic.c +6 -0
index a3ef793fce5f..6ed6b090be94 100644
@@ -297,6 +297,10 @@ static inline void apic_set_spiv(struct kvm_lapic *apic, u32 val)
 
 		atomic_set_release(&apic->vcpu->kvm->arch.apic_map_dirty, DIRTY);
 	}
+
+	/* Check if there are APF page ready requests pending */
+	if (enabled)
+		kvm_make_request(KVM_REQ_APF_READY, apic->vcpu);
 }
 
 static inline void kvm_apic_set_xapic_id(struct kvm_lapic *apic, u8 id)
@@ -2260,6 +2264,8 @@ void kvm_lapic_set_base(struct kvm_vcpu *vcpu, u64 value)
 		if (value & MSR_IA32_APICBASE_ENABLE) {
 			kvm_apic_set_xapic_id(apic, vcpu->vcpu_id);
 			static_key_slow_dec_deferred(&apic_hw_disabled);
+			/* Check if there are APF page ready requests pending */
+			kvm_make_request(KVM_REQ_APF_READY, vcpu);
 		} else {
 			static_key_slow_inc(&apic_hw_disabled.key);
 			atomic_set_release(&apic->vcpu->kvm->arch.apic_map_dirty, DIRTY);
arch/x86/kvm/x86.c +1 -1
index 4588f73bf59a..ae18062c26a6 100644
@@ -11146,7 +11146,7 @@ bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu)
 	if (!kvm_pv_async_pf_enabled(vcpu))
 		return true;
 	else
-		return apf_pageready_slot_free(vcpu);
+		return kvm_lapic_enabled(vcpu) && apf_pageready_slot_free(vcpu);
 }
 
 void kvm_arch_start_assignment(struct kvm *kvm)
Signature · 1 line
2.34.1
maintainer
#1 lore raw
Quoted 48 lines · Guoqing Jiang

From: Vitaly Kuznetsov <[email protected]>

Backport of commit 2f15d027c05fac406decdb5eceb9ec0902b68f53 upstream.

Async PF 'page ready' event may happen when LAPIC is (temporary) disabled. In particular, Sebastien reports that when Linux kernel is directly booted by Cloud Hypervisor, LAPIC is 'software disabled' when APF mechanism is initialized. On initialization KVM tries to inject 'wakeup all' event and puts the corresponding token to the slot. It is, however, failing to inject an interrupt (kvm_apic_set_irq() -> __apic_accept_irq() -> !apic_enabled()) so the guest never gets notified and the whole APF mechanism gets stuck. The same issue is likely to happen if the guest temporary disables LAPIC and a previously unavailable page becomes available.

Do two things to resolve the issue:
- Avoid dequeuing 'page ready' events from APF queue when LAPIC is disabled.
- Trigger an attempt to deliver pending 'page ready' events when LAPIC becomes enabled (SPIV or MSR_IA32_APICBASE).

Message-Id: <[email protected]>

[Guoqing: backport to 5.10-stable ]

---
Hi,

We encountered below task hang issue with 5.10.113 stable kernel.

[  246.845061] INFO: task rpmbuild:2303 blocked for more than 122 seconds.
[  246.846269]       Not tainted 5.10.113-1.1.se2-default #1
[  246.847103] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[  246.848248] task:rpmbuild        state:D stack:    0 pid: 2303 ppid:  2302 flags:0x00000000
[  246.848252] Call Trace:
[  246.848266]  __schedule+0x3f6/0x7c0
[  246.848289]  ? __handle_mm_fault+0x3dd/0x6d0
[  246.848291]  schedule+0x46/0xb0
[  246.848295]  kvm_async_pf_task_wait_schedule+0x4b/0x90
[  246.848297]  ? handle_mm_fault+0xbc/0x280
[  246.848300]  __kvm_handle_async_pf+0x4f/0xb0
[  246.848303]  exc_page_fault+0x204/0x540
[  246.848305]  ? asm_exc_page_fault+0x8/0x30
[  246.848307]  asm_exc_page_fault+0x1e/0x30
[  246.848310] RIP: 0033:0x7f122fbdfc90
Quoted 2 lines

And after investigating, this patch resolve the issue. 5.12 stable kernel has already merged it by commit 36825931c607.

Now queued up, thanks.

greg k-h