Re: [RFC PATCH 4/6] x86/kvm: Add a host side support for virtual suspend time injection

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Monday, 2021-04-26 at 18:06:43 +09, Hikaru Nishida wrote:

> This patch implements virtual suspend time injection support for kvm
> hosts.
> If this functionality is enabled and the guest requests it, the host
> will stop all the clocks observed by the guest during the host's
> suspension and report the duration of suspend to the guest through
> struct kvm_host_suspend_time to give a chance to adjust CLOCK_BOOTTIME
> to the guest. This mechanism can be used to align the guest's clock
> behavior to the hosts' ones.
>
> Signed-off-by: Hikaru Nishida <hikalium@xxxxxxxxxxxx>
> ---
>
>  arch/x86/include/asm/kvm_host.h |  5 ++
>  arch/x86/kvm/cpuid.c            |  4 ++
>  arch/x86/kvm/x86.c              | 89 ++++++++++++++++++++++++++++++++-
>  include/linux/kvm_host.h        |  7 +++
>  kernel/time/timekeeping.c       |  3 ++
>  5 files changed, 107 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 3768819693e5..6584adaab3bf 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -994,6 +994,11 @@ struct kvm_arch {
>  
>  	gpa_t wall_clock;
>  
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +	struct gfn_to_hva_cache suspend_time;
> +	bool suspend_time_injection_enabled;
> +#endif /* KVM_VIRT_SUSPEND_TIMING */
> +
>  	bool mwait_in_guest;
>  	bool hlt_in_guest;
>  	bool pause_in_guest;
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index 6bd2f8b830e4..62224b7bd7f9 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -814,6 +814,10 @@ static inline int __do_cpuid_func(struct kvm_cpuid_array *array, u32 function)
>  			     (1 << KVM_FEATURE_PV_SCHED_YIELD) |
>  			     (1 << KVM_FEATURE_ASYNC_PF_INT);
>  
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +		entry->eax |= (1 << KVM_FEATURE_HOST_SUSPEND_TIME);
> +#endif
> +
>  		if (sched_info_on())
>  			entry->eax |= (1 << KVM_FEATURE_STEAL_TIME);
>  
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index eca63625aee4..d919f771ce31 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -1285,6 +1285,9 @@ static const u32 emulated_msrs_all[] = {
>  
>  	MSR_KVM_ASYNC_PF_EN, MSR_KVM_STEAL_TIME,
>  	MSR_KVM_PV_EOI_EN, MSR_KVM_ASYNC_PF_INT, MSR_KVM_ASYNC_PF_ACK,
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +	MSR_KVM_HOST_SUSPEND_TIME,
> +#endif
>  
>  	MSR_IA32_TSC_ADJUST,
>  	MSR_IA32_TSCDEADLINE,
> @@ -2020,6 +2023,20 @@ static void kvm_write_system_time(struct kvm_vcpu *vcpu, gpa_t system_time,
>  	return;
>  }
>  
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +static void kvm_write_suspend_time(struct kvm *kvm, bool updated)

The "updated" argument is not used.

> +{
> +	struct kvm_arch *ka = &kvm->arch;
> +	struct kvm_host_suspend_time st;
> +
> +	if (!ka->suspend_time_injection_enabled)
> +		return;
> +
> +	st.suspend_time_ns = kvm->suspend_time_ns;
> +	kvm_write_guest_cached(kvm, &ka->suspend_time, &st, sizeof(st));
> +}
> +#endif
> +
>  static uint32_t div_frac(uint32_t dividend, uint32_t divisor)
>  {
>  	do_shl32_div32(dividend, divisor);
> @@ -2653,6 +2670,9 @@ static void kvm_setup_pvclock_page(struct kvm_vcpu *v,
>  
>  	if (vcpu->pvclock_set_guest_stopped_request) {
>  		vcpu->hv_clock.flags |= PVCLOCK_GUEST_STOPPED;
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +		kvm_write_suspend_time(v->kvm, true);
> +#endif
>  		vcpu->pvclock_set_guest_stopped_request = false;
>  	}
>  
> @@ -3229,7 +3249,23 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>  
>  		vcpu->arch.msr_kvm_poll_control = data;
>  		break;
> -
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +	case MSR_KVM_HOST_SUSPEND_TIME:
> +		if (!guest_pv_has(vcpu, KVM_FEATURE_HOST_SUSPEND_TIME))
> +			return 1;
> +		if (!(data & KVM_MSR_ENABLED)) {
> +			vcpu->kvm->arch.suspend_time_injection_enabled = false;
> +			break;
> +		}
> +		if (kvm_gfn_to_hva_cache_init(vcpu->kvm,
> +		     &vcpu->kvm->arch.suspend_time, data & ~KVM_MSR_ENABLED,
> +		     sizeof(struct kvm_host_suspend_time))) {
> +			return 1;
> +		}
> +		vcpu->kvm->arch.suspend_time_injection_enabled = true;
> +		kvm_write_suspend_time(vcpu->kvm, false);
> +		break;
> +#endif
>  	case MSR_IA32_MCG_CTL:
>  	case MSR_IA32_MCG_STATUS:
>  	case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
> @@ -3535,6 +3571,15 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>  
>  		msr_info->data = vcpu->arch.msr_kvm_poll_control;
>  		break;
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +	case MSR_KVM_HOST_SUSPEND_TIME:
> +		if (!guest_pv_has(vcpu, KVM_FEATURE_HOST_SUSPEND_TIME))
> +			return 1;
> +		msr_info->data = vcpu->kvm->arch.suspend_time.gpa;
> +		if (vcpu->kvm->arch.suspend_time_injection_enabled)
> +			msr_info->data |= KVM_MSR_ENABLED;
> +		break;
> +#endif
>  	case MSR_IA32_P5_MC_ADDR:
>  	case MSR_IA32_P5_MC_TYPE:
>  	case MSR_IA32_MCG_CAP:
> @@ -11723,6 +11768,48 @@ int kvm_sev_es_string_io(struct kvm_vcpu *vcpu, unsigned int size,
>  }
>  EXPORT_SYMBOL_GPL(kvm_sev_es_string_io);
>  
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +void kvm_arch_timekeeping_inject_sleeptime(const struct timespec64 *delta)
> +{
> +	struct kvm_vcpu *vcpu;
> +	u64 suspend_time_ns;
> +	struct kvm *kvm;
> +	s64 adj;
> +	int i;
> +
> +	suspend_time_ns = timespec64_to_ns(delta);
> +	adj = tsc_khz * (suspend_time_ns / 1000000);
> +	/*
> +	 * Adjust TSCs on all vcpus as if they are stopped during a suspend.
> +	 * doing a similar thing in kvm_arch_hardware_enable().
> +	 */
> +	mutex_lock(&kvm_lock);
> +	list_for_each_entry(kvm, &vm_list, vm_list) {
> +		if (!kvm->arch.suspend_time_injection_enabled)
> +			continue;
> +
> +		kvm_for_each_vcpu(i, vcpu, kvm) {
> +			/*
> +			 * Adjustment here is always smaller than the gap
> +			 * observed by the guest, so subtracting the value
> +			 * here never rewinds the observed TSC in guests.
> +			 * This adjustment will be applied on the next
> +			 * kvm_arch_vcpu_load().
> +			 */
> +			vcpu->arch.tsc_offset_adjustment -= adj;
> +		}
> +		/*
> +		 * Move the offset of kvm_clock here as if it is stopped
> +		 * during the suspension.
> +		 */
> +		kvm->arch.kvmclock_offset -= suspend_time_ns;
> +		/* suspend_time is accumulated per VM. */
> +		kvm->suspend_time_ns += suspend_time_ns;
> +	}
> +	mutex_unlock(&kvm_lock);
> +}
> +#endif
> +
>  EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_entry);
>  EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_exit);
>  EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_fast_mmio);
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 1b65e7204344..e077b9a960fc 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -520,6 +520,9 @@ struct kvm {
>  	pid_t userspace_pid;
>  	unsigned int max_halt_poll_ns;
>  	u32 dirty_ring_size;
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +	u64 suspend_time_ns;
> +#endif
>  };
>  
>  #define kvm_err(fmt, ...) \
> @@ -1522,4 +1525,8 @@ static inline void kvm_handle_signal_exit(struct kvm_vcpu *vcpu)
>  /* Max number of entries allowed for each kvm dirty ring */
>  #define  KVM_DIRTY_RING_MAX_ENTRIES  65536
>  
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +void kvm_arch_timekeeping_inject_sleeptime(const struct timespec64 *delta);
> +#endif /* CONFIG_KVM_VIRT_SUSPEND_TIMING */
> +
>  #endif
> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
> index ff0304de7de9..342a032ad552 100644
> --- a/kernel/time/timekeeping.c
> +++ b/kernel/time/timekeeping.c
> @@ -1786,6 +1786,9 @@ void timekeeping_resume(void)
>  	if (inject_sleeptime) {
>  		suspend_timing_needed = false;
>  		__timekeeping_inject_sleeptime(tk, &ts_delta);
> +#ifdef CONFIG_KVM_VIRT_SUSPEND_TIMING
> +		kvm_arch_timekeeping_inject_sleeptime(&ts_delta);
> +#endif
>  	}
>  
>  	/* Re-base the last cycle value */
> -- 
> 2.31.1.498.g6c1eba8ee3d-goog

dme.
-- 
Stranded starfish have no place to hide.



[Index of Archives]     [KVM ARM]     [KVM ia64]     [KVM ppc]     [Virtualization Tools]     [Spice Development]     [Libvirt]     [Libvirt Users]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite Questions]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux