| Message ID | 20230915124840.474888-2-kristina.martsenko@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | KVM: arm64: Support for Arm v8.8 memcpy instructions in KVM guests | expand |
Hi Kristina, On Fri, 15 Sep 2023 13:48:38 +0100, Kristina Martsenko <kristina.martsenko@arm.com> wrote: > > At the moment the HCRX_EL2 system register is always initialized to > HCRX_GUEST_FLAGS when running a guest. Instead, choose the configuration > at vcpu reset time and save it in the vcpu struct, similarly to how > HCR_EL2 is set up. This will be needed in a subsequent change to > configure the register based on CPU features detected at runtime. > > Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com> > --- > arch/arm64/include/asm/kvm_emulate.h | 5 +++++ > arch/arm64/include/asm/kvm_host.h | 1 + > arch/arm64/kvm/arm.c | 1 + > arch/arm64/kvm/hyp/include/hyp/switch.h | 2 +- > arch/arm64/kvm/hyp/nvhe/hyp-main.c | 1 + > arch/arm64/kvm/hyp/nvhe/pkvm.c | 1 + > 6 files changed, 10 insertions(+), 1 deletion(-) > > diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h > index 3d6725ff0bf6..64ea27e6deb1 100644 > --- a/arch/arm64/include/asm/kvm_emulate.h > +++ b/arch/arm64/include/asm/kvm_emulate.h > @@ -134,6 +134,11 @@ static inline void vcpu_ptrauth_disable(struct kvm_vcpu *vcpu) > vcpu->arch.hcr_el2 &= ~(HCR_API | HCR_APK); > } > > +static inline void vcpu_reset_hcrx(struct kvm_vcpu *vcpu) > +{ > + vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS; > +} > + > static inline unsigned long vcpu_get_vsesr(struct kvm_vcpu *vcpu) > { > return vcpu->arch.vsesr_el2; > diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h > index af06ccb7ee34..2764748756a7 100644 > --- a/arch/arm64/include/asm/kvm_host.h > +++ b/arch/arm64/include/asm/kvm_host.h > @@ -487,6 +487,7 @@ struct kvm_vcpu_arch { > > /* Values of trap registers for the guest. */ > u64 hcr_el2; > + u64 hcrx_el2; Do we really need this extra field? Yes, this is only an extra 64bit, but they tend to accumulate... Looking at patch #3, the change is related to this: vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS; + + if (cpus_have_final_cap(ARM64_HAS_MOPS)) { + vcpu->arch.hcrx_el2 |= HCRX_EL2_MSCEn; + vcpu->arch.hcrx_el2 |= HCRX_EL2_MCE2; + } meaning that this is a constant value for a given boot of the host. At this stage, I'd rather you define HCRX_GUEST_FLAGS as: #define HCRX_GUEST_FLAGS \ (HCRX_EL2_SMPME | HCRX_EL2_TCR2En | \ cpus_have_final_cap(ARM64_HAS_MOPS) ? \ (HCRX_EL2_MSCEn | HCRX_EL2_MCE2) : 0) and drop the new field altogether, until we have something that requires dynamic flipping of an HCRX_EL2 field. Thanks, M.
On 18/09/2023 12:10, Marc Zyngier wrote: > Hi Kristina, Hi Marc, > On Fri, 15 Sep 2023 13:48:38 +0100, > Kristina Martsenko <kristina.martsenko@arm.com> wrote: >> >> At the moment the HCRX_EL2 system register is always initialized to >> HCRX_GUEST_FLAGS when running a guest. Instead, choose the configuration >> at vcpu reset time and save it in the vcpu struct, similarly to how >> HCR_EL2 is set up. This will be needed in a subsequent change to >> configure the register based on CPU features detected at runtime. >> >> Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com> >> --- >> arch/arm64/include/asm/kvm_emulate.h | 5 +++++ >> arch/arm64/include/asm/kvm_host.h | 1 + >> arch/arm64/kvm/arm.c | 1 + >> arch/arm64/kvm/hyp/include/hyp/switch.h | 2 +- >> arch/arm64/kvm/hyp/nvhe/hyp-main.c | 1 + >> arch/arm64/kvm/hyp/nvhe/pkvm.c | 1 + >> 6 files changed, 10 insertions(+), 1 deletion(-) >> >> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h >> index 3d6725ff0bf6..64ea27e6deb1 100644 >> --- a/arch/arm64/include/asm/kvm_emulate.h >> +++ b/arch/arm64/include/asm/kvm_emulate.h >> @@ -134,6 +134,11 @@ static inline void vcpu_ptrauth_disable(struct kvm_vcpu *vcpu) >> vcpu->arch.hcr_el2 &= ~(HCR_API | HCR_APK); >> } >> >> +static inline void vcpu_reset_hcrx(struct kvm_vcpu *vcpu) >> +{ >> + vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS; >> +} >> + >> static inline unsigned long vcpu_get_vsesr(struct kvm_vcpu *vcpu) >> { >> return vcpu->arch.vsesr_el2; >> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h >> index af06ccb7ee34..2764748756a7 100644 >> --- a/arch/arm64/include/asm/kvm_host.h >> +++ b/arch/arm64/include/asm/kvm_host.h >> @@ -487,6 +487,7 @@ struct kvm_vcpu_arch { >> >> /* Values of trap registers for the guest. */ >> u64 hcr_el2; >> + u64 hcrx_el2; > > Do we really need this extra field? Yes, this is only an extra 64bit, > but they tend to accumulate... > > Looking at patch #3, the change is related to this: > > vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS; > + > + if (cpus_have_final_cap(ARM64_HAS_MOPS)) { > + vcpu->arch.hcrx_el2 |= HCRX_EL2_MSCEn; > + vcpu->arch.hcrx_el2 |= HCRX_EL2_MCE2; > + } > > meaning that this is a constant value for a given boot of the host. > > At this stage, I'd rather you define HCRX_GUEST_FLAGS as: > > #define HCRX_GUEST_FLAGS \ > (HCRX_EL2_SMPME | HCRX_EL2_TCR2En | \ > cpus_have_final_cap(ARM64_HAS_MOPS) ? \ > (HCRX_EL2_MSCEn | HCRX_EL2_MCE2) : 0) > > and drop the new field altogether, until we have something that > requires dynamic flipping of an HCRX_EL2 field. Makes sense, the field isn't strictly required yet, I'll drop it in v2. Thanks, Kristina
diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h index 3d6725ff0bf6..64ea27e6deb1 100644 --- a/arch/arm64/include/asm/kvm_emulate.h +++ b/arch/arm64/include/asm/kvm_emulate.h @@ -134,6 +134,11 @@ static inline void vcpu_ptrauth_disable(struct kvm_vcpu *vcpu) vcpu->arch.hcr_el2 &= ~(HCR_API | HCR_APK); } +static inline void vcpu_reset_hcrx(struct kvm_vcpu *vcpu) +{ + vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS; +} + static inline unsigned long vcpu_get_vsesr(struct kvm_vcpu *vcpu) { return vcpu->arch.vsesr_el2; diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h index af06ccb7ee34..2764748756a7 100644 --- a/arch/arm64/include/asm/kvm_host.h +++ b/arch/arm64/include/asm/kvm_host.h @@ -487,6 +487,7 @@ struct kvm_vcpu_arch { /* Values of trap registers for the guest. */ u64 hcr_el2; + u64 hcrx_el2; u64 mdcr_el2; u64 cptr_el2; diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c index 4866b3f7b4ea..fa359fd6fca9 100644 --- a/arch/arm64/kvm/arm.c +++ b/arch/arm64/kvm/arm.c @@ -1318,6 +1318,7 @@ static int kvm_arch_vcpu_ioctl_vcpu_init(struct kvm_vcpu *vcpu, } vcpu_reset_hcr(vcpu); + vcpu_reset_hcrx(vcpu); vcpu->arch.cptr_el2 = kvm_get_reset_cptr_el2(vcpu); /* diff --git a/arch/arm64/kvm/hyp/include/hyp/switch.h b/arch/arm64/kvm/hyp/include/hyp/switch.h index 9cfe6bd1dbe4..119b75344505 100644 --- a/arch/arm64/kvm/hyp/include/hyp/switch.h +++ b/arch/arm64/kvm/hyp/include/hyp/switch.h @@ -198,7 +198,7 @@ static inline void __activate_traps_common(struct kvm_vcpu *vcpu) write_sysreg(vcpu->arch.mdcr_el2, mdcr_el2); if (cpus_have_final_cap(ARM64_HAS_HCX)) { - u64 hcrx = HCRX_GUEST_FLAGS; + u64 hcrx = vcpu->arch.hcrx_el2; if (vcpu_has_nv(vcpu) && !is_hyp_ctxt(vcpu)) { u64 clr = 0, set = 0; diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-main.c b/arch/arm64/kvm/hyp/nvhe/hyp-main.c index 857d9bc04fd4..26692083b80b 100644 --- a/arch/arm64/kvm/hyp/nvhe/hyp-main.c +++ b/arch/arm64/kvm/hyp/nvhe/hyp-main.c @@ -35,6 +35,7 @@ static void flush_hyp_vcpu(struct pkvm_hyp_vcpu *hyp_vcpu) hyp_vcpu->vcpu.arch.hw_mmu = host_vcpu->arch.hw_mmu; hyp_vcpu->vcpu.arch.hcr_el2 = host_vcpu->arch.hcr_el2; + hyp_vcpu->vcpu.arch.hcrx_el2 = host_vcpu->arch.hcrx_el2; hyp_vcpu->vcpu.arch.mdcr_el2 = host_vcpu->arch.mdcr_el2; hyp_vcpu->vcpu.arch.cptr_el2 = host_vcpu->arch.cptr_el2; diff --git a/arch/arm64/kvm/hyp/nvhe/pkvm.c b/arch/arm64/kvm/hyp/nvhe/pkvm.c index 8033ef353a5d..8c8b37525dce 100644 --- a/arch/arm64/kvm/hyp/nvhe/pkvm.c +++ b/arch/arm64/kvm/hyp/nvhe/pkvm.c @@ -188,6 +188,7 @@ static void pvm_init_trap_regs(struct kvm_vcpu *vcpu) /* Clear res0 and set res1 bits to trap potential new features. */ vcpu->arch.hcr_el2 &= ~(HCR_RES0); + vcpu->arch.hcrx_el2 &= ~(HCRX_EL2_RES0); vcpu->arch.mdcr_el2 &= ~(MDCR_EL2_RES0); if (!has_hvhe()) { vcpu->arch.cptr_el2 |= CPTR_NVHE_EL2_RES1;
At the moment the HCRX_EL2 system register is always initialized to HCRX_GUEST_FLAGS when running a guest. Instead, choose the configuration at vcpu reset time and save it in the vcpu struct, similarly to how HCR_EL2 is set up. This will be needed in a subsequent change to configure the register based on CPU features detected at runtime. Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com> --- arch/arm64/include/asm/kvm_emulate.h | 5 +++++ arch/arm64/include/asm/kvm_host.h | 1 + arch/arm64/kvm/arm.c | 1 + arch/arm64/kvm/hyp/include/hyp/switch.h | 2 +- arch/arm64/kvm/hyp/nvhe/hyp-main.c | 1 + arch/arm64/kvm/hyp/nvhe/pkvm.c | 1 + 6 files changed, 10 insertions(+), 1 deletion(-)