SEV-ES TSC_AUX virtualization fix and optimization
v2 cover SEV-ES TSC_AUX virtualization fix and optimization
This patch series provides fixes to the TSC_AUX virtualization support and an optimization to reduce the number of WRMSRs to TSC_AUX when it is virtualized.
---
Changes since v1:
- Move TSC_AUX virtualization support out of init_vmcb_after_set_cpuid() path and into the vcpu_after_set_cpuid() path
- Add an additional patch to properly set or clear intercepts based on TSC_AUX virtualization requirements
- Simplify the TSC_AUX virtualization optimization to set the host save area TSC_AUX value once during svm_hardware_enable().
- Since the TSC_AUX virtualization can't be disabled for an SEV-ES guest, eliminate the "v_tsc_aux" flag and check against the host feature and type of guest, directly.
Patches based on https://git.kernel.org/pub/scm/virt/kvm/kvm.git master
and commit:
7c7cce2cf7ee ("Merge tag 'kvmarm-fixes-6.6-1' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD")
Tom Lendacky (3):
KVM: SVM: Fix TSC_AUX virtualization setup
KVM: SVM: Fix TSC_AUX virtualization intercept update logic
KVM: SVM: Do not use user return MSR support for virtualized TSC_AUX
arch/x86/kvm/svm/sev.c | 34 +++++++++++++++++++++++++-------- arch/x86/kvm/svm/svm.c | 43 ++++++++++++++++++++++++++++++++++-------- arch/x86/kvm/svm/svm.h | 1 + 3 files changed, 62 insertions(+), 16 deletions(-)
Signature · 1 line
2.41.0
PATCH v2 3/3 KVM: SVM: Do not use user return MSR support for virtualized TSC_AUX
When the TSC_AUX MSR is virtualized, the TSC_AUX value is swap type "B" within the VMSA. This means that the guest value is loaded on VMRUN and the host value is restored from the host save area on #VMEXIT.
Since the value is restored on #VMEXIT, the KVM user return MSR support for TSC_AUX can be replaced by populating the host save area with the current host value of TSC_AUX. And, since TSC_AUX is not changed by Linux post-boot, the host save area can be set once in svm_hardware_enable(). This eliminates the two WRMSR instructions associated with the user return MSR support.
--- arch/x86/kvm/svm/svm.c | 34 +++++++++++++++++++++++++++++++++- 1 file changed, 33 insertions(+), 1 deletion(-)
arch/x86/kvm/svm/svm.c +33 -1
@@ -683,6 +683,21 @@ static int svm_hardware_enable(void)amd_pmu_enable_virt (); +/* +* If TSC_AUX virtualization is supported, TSC_AUX becomes a swap type +* "B" field (see sev_es_prepare_switch_to_guest()) for SEV-ES guests. +* Since Linux does not change the value of TSC_AUX once set, prime the +* TSC_AUX field now to avoid a RDMSR on every vCPU run. +*/ +if (boot_cpu_has (X86_FEATURE_V_TSC_AUX )) { +struct sev_es_save_area *hostsa; +u32 msr_hi; + + hostsa = (struct sev_es_save_area *)(page_address (sd->save_area) +0x400 ); + +rdmsr (MSR_TSC_AUX , hostsa->tsc_aux, msr_hi); + } +return 0 ; }
@@ -1532,7 +1547,14 @@ static void svm_prepare_switch_to_guest(struct kvm_vcpu *vcpu)if (tsc_scaling)__svm_write_tsc_multiplier (vcpu->arch.tsc_scaling_ratio); -if (likely (tsc_aux_uret_slot >=0 )) +/* +* TSC_AUX is always virtualized for SEV-ES guests when the feature is +* available. The user return MSR support is not required in this case +* because TSC_AUX is restored on #VMEXIT from the host save area +* (which has been initialized in svm_hardware_enable()). +*/ +if (likely (tsc_aux_uret_slot >=0 ) && + (!boot_cpu_has (X86_FEATURE_V_TSC_AUX ) || !sev_es_guest (vcpu->kvm)))kvm_set_user_return_msr (tsc_aux_uret_slot, svm->tsc_aux,-1ull ); svm->guest_state_loaded = true;
@@ -3086,6 +3108,16 @@ static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr) svm->sysenter_esp_hi =guest_cpuid_is_intel (vcpu) ? (data >>32 ) :0 ;break ;case MSR_TSC_AUX : +/* +* TSC_AUX is always virtualized for SEV-ES guests when the +* feature is available. The user return MSR support is not +* required in this case because TSC_AUX is restored on #VMEXIT +* from the host save area (which has been initialized in +* svm_hardware_enable()). +*/ +if (boot_cpu_has (X86_FEATURE_V_TSC_AUX ) &&sev_es_guest (vcpu->kvm)) +break ; +/* * TSC_AUX is usually changed only during boot and never read * directly. Intercept TSC_AUX instead of exposing it to the
Signature · 1 line
2.41.0
PATCH v2 2/3 KVM: SVM: Fix TSC_AUX virtualization intercept update logic
In reply to Tom LendackyWith the TSC_AUX virtualization support now in the vcpu_set_after_cpuid() path, the intercepts must be either cleared or set based on the guest CPUID input. Currently the support only clears the intercepts.
Also, vcpu_set_after_cpuid() calls svm_recalc_instruction_intercepts() as part of the processing, so the setting or clearing of the RDTSCP intercept can be dropped from the TSC_AUX virtualization support.
Update the support to always set or clear the TSC_AUX MSR intercept based on the virtualization requirements.
--- arch/x86/kvm/svm/sev.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)
arch/x86/kvm/svm/sev.c +5 -6
@@ -2966,12 +2966,11 @@ static void sev_es_vcpu_after_set_cpuid(struct vcpu_svm *svm) {struct kvm_vcpu *vcpu = &svm->vcpu; -if (boot_cpu_has (X86_FEATURE_V_TSC_AUX ) && - (guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP ) || -guest_cpuid_has (vcpu,X86_FEATURE_RDPID ))) { -set_msr_interception (vcpu, svm->msrpm,MSR_TSC_AUX ,1 ,1 ); -if (guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP )) -svm_clr_intercept (svm,INTERCEPT_RDTSCP ); +if (boot_cpu_has (X86_FEATURE_V_TSC_AUX )) { +bool v_tsc_aux =guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP ) || +guest_cpuid_has (vcpu,X86_FEATURE_RDPID ); + +set_msr_interception (vcpu, svm->msrpm,MSR_TSC_AUX , v_tsc_aux, v_tsc_aux); } }
Signature · 1 line
2.41.0
PATCH v2 1/3 KVM: SVM: Fix TSC_AUX virtualization setup
In reply to Tom LendackyThe checks for virtualizing TSC_AUX occur during the vCPU reset processing path. However, at the time of initial vCPU reset processing, when the vCPU is first created, not all of the guest CPUID information has been set. In this case the RDTSCP and RDPID feature support for the guest is not in place and so TSC_AUX virtualization is not established.
This continues for each vCPU created for the guest. On the first boot of an AP, vCPU reset processing is executed as a result of an APIC INIT event, this time with all of the guest CPUID information set, resulting in TSC_AUX virtualization being enabled, but only for the APs. The BSP always sees a TSC_AUX value of 0 which probably went unnoticed because, at least for Linux, the BSP TSC_AUX value is 0.
Move the TSC_AUX virtualization enablement out of the init_vmcb() path and into the vcpu_after_set_cpuid() path to allow for proper initialization of the support after the guest CPUID information has been set.
--- arch/x86/kvm/svm/sev.c | 35 +++++++++++++++++++++++++++-------- arch/x86/kvm/svm/svm.c | 9 ++------- arch/x86/kvm/svm/svm.h | 1 + 3 files changed, 30 insertions(+), 15 deletions(-)
arch/x86/kvm/svm/sev.c +27 -8
@@ -2962,6 +2962,33 @@ int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in) count, in); } +static void sev_es_vcpu_after_set_cpuid (struct vcpu_svm *svm) +{ +struct kvm_vcpu *vcpu = &svm->vcpu; + +if (boot_cpu_has (X86_FEATURE_V_TSC_AUX ) && + (guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP ) || +guest_cpuid_has (vcpu,X86_FEATURE_RDPID ))) { +set_msr_interception (vcpu, svm->msrpm,MSR_TSC_AUX ,1 ,1 ); +if (guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP )) +svm_clr_intercept (svm,INTERCEPT_RDTSCP ); + } +} + +void sev_vcpu_after_set_cpuid (struct vcpu_svm *svm) +{ +struct kvm_vcpu *vcpu = &svm->vcpu; +struct kvm_cpuid_entry2 *best; + +/* For sev guests, the memory encryption bit is not reserved in CR3. */ + best =kvm_find_cpuid_entry (vcpu,0x8000001F ); +if (best) + vcpu->arch.reserved_gpa_bits &= ~(1UL << (best->ebx &0x3f )); + +if (sev_es_guest (svm->vcpu.kvm)) +sev_es_vcpu_after_set_cpuid (svm); +} +static void sev_es_init_vmcb (struct vcpu_svm *svm) {struct vmcb *vmcb = svm->vmcb01.ptr;
@@ -3024,14 +3051,6 @@ static void sev_es_init_vmcb(struct vcpu_svm *svm)set_msr_interception (vcpu, svm->msrpm,MSR_IA32_LASTBRANCHTOIP ,1 ,1 );set_msr_interception (vcpu, svm->msrpm,MSR_IA32_LASTINTFROMIP ,1 ,1 );set_msr_interception (vcpu, svm->msrpm,MSR_IA32_LASTINTTOIP ,1 ,1 ); - -if (boot_cpu_has (X86_FEATURE_V_TSC_AUX ) && - (guest_cpuid_has (&svm->vcpu,X86_FEATURE_RDTSCP ) || -guest_cpuid_has (&svm->vcpu,X86_FEATURE_RDPID ))) { -set_msr_interception (vcpu, svm->msrpm,MSR_TSC_AUX ,1 ,1 ); -if (guest_cpuid_has (&svm->vcpu,X86_FEATURE_RDTSCP )) -svm_clr_intercept (svm,INTERCEPT_RDTSCP ); - } }void sev_init_vmcb (struct vcpu_svm *svm)
arch/x86/kvm/svm/svm.c +2 -7
@@ -4284,7 +4284,6 @@ static bool svm_has_emulated_msr(struct kvm *kvm, u32 index)static void svm_vcpu_after_set_cpuid (struct kvm_vcpu *vcpu) {struct vcpu_svm *svm =to_svm (vcpu); -struct kvm_cpuid_entry2 *best;/* * SVM doesn't provide a way to disable just XSAVES in the guest, KVM
@@ -4328,12 +4327,8 @@ static void svm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)set_msr_interception (vcpu, svm->msrpm,MSR_IA32_FLUSH_CMD ,0 , !!guest_cpuid_has (vcpu,X86_FEATURE_FLUSH_L1D )); -/* For sev guests, the memory encryption bit is not reserved in CR3. */ -if (sev_guest (vcpu->kvm)) { - best =kvm_find_cpuid_entry (vcpu,0x8000001F ); -if (best) - vcpu->arch.reserved_gpa_bits &= ~(1UL << (best->ebx &0x3f )); - } +if (sev_guest (vcpu->kvm)) +sev_vcpu_after_set_cpuid (svm);init_vmcb_after_set_cpuid (vcpu); }
arch/x86/kvm/svm/svm.h +1 -0
@@ -684,6 +684,7 @@ void __init sev_hardware_setup(void);void sev_hardware_unsetup (void );int sev_cpu_init (struct svm_cpu_data *sd);void sev_init_vmcb (struct vcpu_svm *svm); +void sev_vcpu_after_set_cpuid (struct vcpu_svm *svm);void sev_free_vcpu (struct kvm_vcpu *vcpu);int sev_handle_vmgexit (struct kvm_vcpu *vcpu);int sev_es_string_io (struct vcpu_svm *svm,int size,unsigned int port,int in);
Signature · 1 line
2.41.0
Re: [PATCH v2 2/3] KVM: SVM: Fix TSC_AUX virtualization intercept update logic
In reply to Tom LendackyQuoted 8 lines · Tom Lendacky
With the
TSC_AUXvirtualization support now in thevcpu_set_after_cpuid()path, the intercepts must be either cleared or set based on the guest CPUID input. Currently the support only clears the intercepts.Also,
vcpu_set_after_cpuid()callssvm_recalc_instruction_intercepts()as part of the processing, so the setting or clearing of the RDTSCP intercept can be dropped from theTSC_AUXvirtualization support.
Quoted 5 lines
Update the support to always set or clear the
TSC_AUXMSR intercept based on the virtualization requirements.Fixes 296d5a17e793 ("KVM: SEV-ES: Use V_TSC_AUX if available instead of RDTSC/MSR_TSC_AUX intercepts")Signed-off-by Tom Lendacky <[email protected]>
In the end, I was really on the fence about this being a separate patch given the common Fixes: tag. Your call, feel free to squash with patch #1 and massage the commit message if you'd like (or I can send a v3 with them squashed if you prefer).
Thanks,
Tom
Quoted 26 lines
--- arch/x86/kvm/svm/sev.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)arch/x86/kvm/svm/sev.c +5
-6@@ -2966,12 +2966,11 @@ static void sev_es_vcpu_after_set_cpuid(struct vcpu_svm *svm) {struct kvm_vcpu *vcpu = &svm->vcpu; -if (boot_cpu_has (X86_FEATURE_V_TSC_AUX ) && - (guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP ) || -guest_cpuid_has (vcpu,X86_FEATURE_RDPID ))) { -set_msr_interception (vcpu, svm->msrpm,MSR_TSC_AUX ,1 ,1 ); -if (guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP )) -svm_clr_intercept (svm,INTERCEPT_RDTSCP ); +if (boot_cpu_has (X86_FEATURE_V_TSC_AUX )) { +bool v_tsc_aux =guest_cpuid_has (vcpu,X86_FEATURE_RDTSCP ) || +guest_cpuid_has (vcpu,X86_FEATURE_RDPID ); + +set_msr_interception (vcpu, svm->msrpm,MSR_TSC_AUX , v_tsc_aux, v_tsc_aux); } }
Paolo Bonzini
maintainerQueued, thanks. The part that stood out in patch 2 is the removal of svm_clr_intercept(), which also applies when the initialization is done in the wrong place. Either way, svm_clr_intercept() is always going to be called by svm_recalc_instruction_intercepts() if guest has the RDTSC bit in its CPUID.
So I extracted that into a separate patch and squashed the rest of patch 2 into patch 1.
Paolo
Quoted 8 lines · Paolo Bonzini
Queued, thanks. The part that stood out in patch 2 is the removal of
svm_clr_intercept(), which also applies when the initialization is done in the wrong place. Either way,svm_clr_intercept()is always going to be called bysvm_recalc_instruction_intercepts()if guest has the RDTSC bit in its CPUID.So I extracted that into a separate patch and squashed the rest of patch 2 into patch 1.
Works for me. Thanks, Paolo!
Quoted 4 lines
Paolo