From 9dbb029b9c44a84968317d462eaa5a748317d8fb Mon Sep 17 00:00:00 2001 From: Michal Luczaj Date: Tue, 15 Aug 2023 00:08:35 +0200 Subject: [PATCH 01/18] KVM: x86: Remove redundant vcpu->arch.cr0 assignments Drop the vcpu->arch.cr0 assignment after static_call(kvm_x86_set_cr0). CR0 was already set by {vmx,svm}_set_cr0(). Signed-off-by: Michal Luczaj Link: https://lore.kernel.org/r/20230814222358.707877-2-mhal@rbox.co Signed-off-by: Sean Christopherson --- arch/x86/kvm/smm.c | 1 - arch/x86/kvm/x86.c | 1 - 2 files changed, 2 deletions(-) diff --git a/arch/x86/kvm/smm.c b/arch/x86/kvm/smm.c index b42111a24cc2..dc3d95fdca7d 100644 --- a/arch/x86/kvm/smm.c +++ b/arch/x86/kvm/smm.c @@ -324,7 +324,6 @@ void enter_smm(struct kvm_vcpu *vcpu) cr0 = vcpu->arch.cr0 & ~(X86_CR0_PE | X86_CR0_EM | X86_CR0_TS | X86_CR0_PG); static_call(kvm_x86_set_cr0)(vcpu, cr0); - vcpu->arch.cr0 = cr0; static_call(kvm_x86_set_cr4)(vcpu, 0); diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 9f18b06bbda6..d960d4094a0b 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -11521,7 +11521,6 @@ static int __set_sregs_common(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs, *mmu_reset_needed |= kvm_read_cr0(vcpu) != sregs->cr0; static_call(kvm_x86_set_cr0)(vcpu, sregs->cr0); - vcpu->arch.cr0 = sregs->cr0; *mmu_reset_needed |= kvm_read_cr4(vcpu) != sregs->cr4; static_call(kvm_x86_set_cr4)(vcpu, sregs->cr4); From 4346db6e6e7a63c25a2f6dd93f4613d6a17dbd4e Mon Sep 17 00:00:00 2001 From: Michal Luczaj Date: Tue, 15 Aug 2023 00:08:36 +0200 Subject: [PATCH 02/18] KVM: x86: Force TLB flush on userspace changes to special registers Userspace can directly modify the content of vCPU's CR0, CR3, and CR4 via KVM_SYNC_X86_SREGS and KVM_SET_SREGS{,2}. Make sure that KVM flushes guest TLB entries and paging-structure caches if a (partial) guest TLB flush is architecturally required based on the CRn changes. To keep things simple, flush whenever KVM resets the MMU context, i.e. if any bits in CR0, CR3, CR4, or EFER are modified. This is extreme overkill, but stuffing state from userspace is not such a hot path that preserving guest TLB state is a priority. Suggested-by: Paolo Bonzini Signed-off-by: Michal Luczaj Link: https://lore.kernel.org/r/20230814222358.707877-3-mhal@rbox.co [sean: call out that the flushing on MMU context resets is for simplicity] Signed-off-by: Sean Christopherson --- arch/x86/kvm/x86.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index d960d4094a0b..ec728aef9acc 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -11564,8 +11564,10 @@ static int __set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs) if (ret) return ret; - if (mmu_reset_needed) + if (mmu_reset_needed) { kvm_mmu_reset_context(vcpu); + kvm_make_request(KVM_REQ_TLB_FLUSH_GUEST, vcpu); + } max_bits = KVM_NR_INTERRUPTS; pending_vec = find_first_bit( @@ -11606,8 +11608,10 @@ static int __set_sregs2(struct kvm_vcpu *vcpu, struct kvm_sregs2 *sregs2) mmu_reset_needed = 1; vcpu->arch.pdptrs_from_userspace = true; } - if (mmu_reset_needed) + if (mmu_reset_needed) { kvm_mmu_reset_context(vcpu); + kvm_make_request(KVM_REQ_TLB_FLUSH_GUEST, vcpu); + } return 0; } From f10a570b093e60c6bd3f210ae909f014f421352a Mon Sep 17 00:00:00 2001 From: Kyle Meyer Date: Thu, 24 Aug 2023 16:52:46 -0500 Subject: [PATCH 03/18] KVM: x86: Add CONFIG_KVM_MAX_NR_VCPUS to allow up to 4096 vCPUs Add a Kconfig entry to set the maximum number of vCPUs per KVM guest and set the default value to 4096 when MAXSMP is enabled, as there are use cases that want to create more than the currently allowed 1024 vCPUs and are more than happy to eat the memory overhead. The Hyper-V TLFS doesn't allow more than 64 sparse banks, i.e. allows a maximum of 4096 virtual CPUs. Cap KVM's maximum number of virtual CPUs to 4096 to avoid exceeding Hyper-V's limit as KVM support for Hyper-V is unconditional, and alternatives like dynamically disabling Hyper-V enlightenments that rely on sparse banks would require non-trivial code changes. Suggested-by: Sean Christopherson Signed-off-by: Kyle Meyer Link: https://lore.kernel.org/r/20230824215244.3897419-1-kyle.meyer@hpe.com [sean: massage changelog with --verbose, document #ifdef mess] Signed-off-by: Sean Christopherson --- arch/x86/include/asm/kvm_host.h | 8 ++++++++ arch/x86/kvm/Kconfig | 11 +++++++++++ 2 files changed, 19 insertions(+) diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h index 17715cb8731d..c356aa27fbee 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -39,7 +39,15 @@ #define __KVM_HAVE_ARCH_VCPU_DEBUGFS +/* + * CONFIG_KVM_MAX_NR_VCPUS is defined iff CONFIG_KVM!=n, provide a dummy max if + * KVM is disabled (arbitrarily use the default from CONFIG_KVM_MAX_NR_VCPUS). + */ +#ifdef CONFIG_KVM_MAX_NR_VCPUS +#define KVM_MAX_VCPUS CONFIG_KVM_MAX_NR_VCPUS +#else #define KVM_MAX_VCPUS 1024 +#endif /* * In x86, the VCPU ID corresponds to the APIC ID, and APIC IDs diff --git a/arch/x86/kvm/Kconfig b/arch/x86/kvm/Kconfig index ed90f148140d..950c12868d30 100644 --- a/arch/x86/kvm/Kconfig +++ b/arch/x86/kvm/Kconfig @@ -154,4 +154,15 @@ config KVM_PROVE_MMU config KVM_EXTERNAL_WRITE_TRACKING bool +config KVM_MAX_NR_VCPUS + int "Maximum number of vCPUs per KVM guest" + depends on KVM + range 1024 4096 + default 4096 if MAXSMP + default 1024 + help + Set the maximum number of vCPUs per KVM guest. Larger values will increase + the memory footprint of each KVM guest, regardless of how many vCPUs are + created for a given VM. + endif # VIRTUALIZATION From 6f0f23ef76be11bc6ace10df5e807f3542da8772 Mon Sep 17 00:00:00 2001 From: Josh Poimboeuf Date: Fri, 25 Aug 2023 00:01:35 -0700 Subject: [PATCH 04/18] KVM: x86: Add IBPB_BRTYPE support Add support for the IBPB_BRTYPE CPUID flag, which indicates that IBPB includes branch type prediction flushing. Note, like SRSO_NO, advertise support for IBPB_BRTYPE even if it's not enumerated by in the raw CPUID, i.e. bypass the cpuid_count() in __kvm_cpu_cap_mask(). Some CPUs that gained support via a uCode patch don't report IBPB_BRTYPE via CPUID (the kernel forces the flag). Opportunistically use kvm_cpu_cap_check_and_set() for SRSO_NO instead of manually querying host support (cpu_feature_enabled() and boot_cpu_has() yield the same end result in this case). Signed-off-by: Josh Poimboeuf Link: https://lore.kernel.org/r/79d5f5914fb42c2c62418ffbcd78f138645ded21.1692919072.git.jpoimboe@kernel.org Signed-off-by: Sean Christopherson --- arch/x86/kvm/cpuid.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c index 0544e30b4946..8f26a929d510 100644 --- a/arch/x86/kvm/cpuid.c +++ b/arch/x86/kvm/cpuid.c @@ -764,8 +764,8 @@ void kvm_set_cpu_caps(void) F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */ ); - if (cpu_feature_enabled(X86_FEATURE_SRSO_NO)) - kvm_cpu_cap_set(X86_FEATURE_SRSO_NO); + kvm_cpu_cap_check_and_set(X86_FEATURE_IBPB_BRTYPE); + kvm_cpu_cap_check_and_set(X86_FEATURE_SRSO_NO); kvm_cpu_cap_init_kvm_defined(CPUID_8000_0022_EAX, F(PERFMON_V2) From e47d86083c66525b89c7fc66cdd64d5937725563 Mon Sep 17 00:00:00 2001 From: Josh Poimboeuf Date: Fri, 25 Aug 2023 00:01:36 -0700 Subject: [PATCH 05/18] KVM: x86: Add SBPB support Add support for the AMD Selective Branch Predictor Barrier (SBPB) by advertising the CPUID bit and handling PRED_CMD writes accordingly. Note, like SRSO_NO and IBPB_BRTYPE before it, advertise support for SBPB even if it's not enumerated by in the raw CPUID. Some CPUs that gained support via a uCode patch don't report SBPB via CPUID (the kernel forces the flag). Signed-off-by: Josh Poimboeuf Link: https://lore.kernel.org/r/a4ab1e7fe50096d50fde33e739ed2da40b41ea6a.1692919072.git.jpoimboe@kernel.org Co-developed-by: Sean Christopherson Signed-off-by: Sean Christopherson --- arch/x86/kvm/cpuid.c | 1 + arch/x86/kvm/cpuid.h | 3 ++- arch/x86/kvm/x86.c | 29 ++++++++++++++++++++++++----- 3 files changed, 27 insertions(+), 6 deletions(-) diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c index 8f26a929d510..552dc4f3899b 100644 --- a/arch/x86/kvm/cpuid.c +++ b/arch/x86/kvm/cpuid.c @@ -764,6 +764,7 @@ void kvm_set_cpu_caps(void) F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */ ); + kvm_cpu_cap_check_and_set(X86_FEATURE_SBPB); kvm_cpu_cap_check_and_set(X86_FEATURE_IBPB_BRTYPE); kvm_cpu_cap_check_and_set(X86_FEATURE_SRSO_NO); diff --git a/arch/x86/kvm/cpuid.h b/arch/x86/kvm/cpuid.h index 284fa4704553..0b90532b6e26 100644 --- a/arch/x86/kvm/cpuid.h +++ b/arch/x86/kvm/cpuid.h @@ -174,7 +174,8 @@ static inline bool guest_has_spec_ctrl_msr(struct kvm_vcpu *vcpu) static inline bool guest_has_pred_cmd_msr(struct kvm_vcpu *vcpu) { return (guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL) || - guest_cpuid_has(vcpu, X86_FEATURE_AMD_IBPB)); + guest_cpuid_has(vcpu, X86_FEATURE_AMD_IBPB) || + guest_cpuid_has(vcpu, X86_FEATURE_SBPB)); } static inline bool supports_cpuid_fault(struct kvm_vcpu *vcpu) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index ec728aef9acc..2b5b325e19f2 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3670,17 +3670,36 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) vcpu->arch.perf_capabilities = data; kvm_pmu_refresh(vcpu); break; - case MSR_IA32_PRED_CMD: - if (!msr_info->host_initiated && !guest_has_pred_cmd_msr(vcpu)) + case MSR_IA32_PRED_CMD: { + u64 reserved_bits = ~(PRED_CMD_IBPB | PRED_CMD_SBPB); + + if (!msr_info->host_initiated) { + if ((!guest_has_pred_cmd_msr(vcpu))) + return 1; + + if (!guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL) && + !guest_cpuid_has(vcpu, X86_FEATURE_AMD_IBPB)) + reserved_bits |= PRED_CMD_IBPB; + + if (!guest_cpuid_has(vcpu, X86_FEATURE_SBPB)) + reserved_bits |= PRED_CMD_SBPB; + } + + if (!boot_cpu_has(X86_FEATURE_IBPB)) + reserved_bits |= PRED_CMD_IBPB; + + if (!boot_cpu_has(X86_FEATURE_SBPB)) + reserved_bits |= PRED_CMD_SBPB; + + if (data & reserved_bits) return 1; - if (!boot_cpu_has(X86_FEATURE_IBPB) || (data & ~PRED_CMD_IBPB)) - return 1; if (!data) break; - wrmsrl(MSR_IA32_PRED_CMD, PRED_CMD_IBPB); + wrmsrl(MSR_IA32_PRED_CMD, data); break; + } case MSR_IA32_FLUSH_CMD: if (!msr_info->host_initiated && !guest_cpuid_has(vcpu, X86_FEATURE_FLUSH_L1D)) From 5d6d6a7d7e66a98bb3432478d226c68c219aaba3 Mon Sep 17 00:00:00 2001 From: David Woodhouse Date: Thu, 5 Oct 2023 10:16:10 +0100 Subject: [PATCH 06/18] KVM: x86: Refine calculation of guest wall clock to use a single TSC read MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When populating the guest's PV wall clock information, KVM currently does a simple 'kvm_get_real_ns() - get_kvmclock_ns(kvm)'. This is an antipattern which should be avoided; when working with the relationship between two clocks, it's never correct to obtain one of them "now" and then the other at a slightly different "now" after an unspecified period of preemption (which might not even be under the control of the kernel, if this is an L1 hosting an L2 guest under nested virtualization). Add a kvm_get_wall_clock_epoch() function to return the guest wall clock epoch in nanoseconds using the same method as __get_kvmclock() — by using kvm_get_walltime_and_clockread() to calculate both the wall clock and KVM clock time from a *single* TSC reading. The condition using get_cpu_tsc_khz() is equivalent to the version in __get_kvmclock() which separately checks for the CONSTANT_TSC feature or the per-CPU cpu_tsc_khz. Which is what get_cpu_tsc_khz() does anyway. Signed-off-by: David Woodhouse Link: https://lore.kernel.org/r/bfc6d3d7cfb88c47481eabbf5a30a264c58c7789.camel@infradead.org Signed-off-by: Sean Christopherson --- arch/x86/kvm/x86.c | 85 ++++++++++++++++++++++++++++++++++++++++++---- arch/x86/kvm/x86.h | 1 + arch/x86/kvm/xen.c | 4 +-- 3 files changed, 81 insertions(+), 9 deletions(-) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 2b5b325e19f2..42b33ef997a5 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -2331,14 +2331,9 @@ static void kvm_write_wall_clock(struct kvm *kvm, gpa_t wall_clock, int sec_hi_o if (kvm_write_guest(kvm, wall_clock, &version, sizeof(version))) return; - /* - * The guest calculates current wall clock time by adding - * system time (updated by kvm_guest_time_update below) to the - * wall clock specified here. We do the reverse here. - */ - wall_nsec = ktime_get_real_ns() - get_kvmclock_ns(kvm); + wall_nsec = kvm_get_wall_clock_epoch(kvm); - wc.nsec = do_div(wall_nsec, 1000000000); + wc.nsec = do_div(wall_nsec, NSEC_PER_SEC); wc.sec = (u32)wall_nsec; /* overflow in 2106 guest time */ wc.version = version; @@ -3241,6 +3236,82 @@ static int kvm_guest_time_update(struct kvm_vcpu *v) return 0; } +/* + * The pvclock_wall_clock ABI tells the guest the wall clock time at + * which it started (i.e. its epoch, when its kvmclock was zero). + * + * In fact those clocks are subtly different; wall clock frequency is + * adjusted by NTP and has leap seconds, while the kvmclock is a + * simple function of the TSC without any such adjustment. + * + * Perhaps the ABI should have exposed CLOCK_TAI and a ratio between + * that and kvmclock, but even that would be subject to change over + * time. + * + * Attempt to calculate the epoch at a given moment using the *same* + * TSC reading via kvm_get_walltime_and_clockread() to obtain both + * wallclock and kvmclock times, and subtracting one from the other. + * + * Fall back to using their values at slightly different moments by + * calling ktime_get_real_ns() and get_kvmclock_ns() separately. + */ +uint64_t kvm_get_wall_clock_epoch(struct kvm *kvm) +{ +#ifdef CONFIG_X86_64 + struct pvclock_vcpu_time_info hv_clock; + struct kvm_arch *ka = &kvm->arch; + unsigned long seq, local_tsc_khz; + struct timespec64 ts; + uint64_t host_tsc; + + do { + seq = read_seqcount_begin(&ka->pvclock_sc); + + local_tsc_khz = 0; + if (!ka->use_master_clock) + break; + + /* + * The TSC read and the call to get_cpu_tsc_khz() must happen + * on the same CPU. + */ + get_cpu(); + + local_tsc_khz = get_cpu_tsc_khz(); + + if (local_tsc_khz && + !kvm_get_walltime_and_clockread(&ts, &host_tsc)) + local_tsc_khz = 0; /* Fall back to old method */ + + put_cpu(); + + /* + * These values must be snapshotted within the seqcount loop. + * After that, it's just mathematics which can happen on any + * CPU at any time. + */ + hv_clock.tsc_timestamp = ka->master_cycle_now; + hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset; + + } while (read_seqcount_retry(&ka->pvclock_sc, seq)); + + /* + * If the conditions were right, and obtaining the wallclock+TSC was + * successful, calculate the KVM clock at the corresponding time and + * subtract one from the other to get the guest's epoch in nanoseconds + * since 1970-01-01. + */ + if (local_tsc_khz) { + kvm_get_time_scale(NSEC_PER_SEC, local_tsc_khz * NSEC_PER_USEC, + &hv_clock.tsc_shift, + &hv_clock.tsc_to_system_mul); + return ts.tv_nsec + NSEC_PER_SEC * ts.tv_sec - + __pvclock_read_cycles(&hv_clock, host_tsc); + } +#endif + return ktime_get_real_ns() - get_kvmclock_ns(kvm); +} + /* * kvmclock updates which are isolated to a given vcpu, such as * vcpu->cpu migration, should not allow system_timestamp from diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h index 1e7be1f6ab29..5184fde1dc54 100644 --- a/arch/x86/kvm/x86.h +++ b/arch/x86/kvm/x86.h @@ -293,6 +293,7 @@ static inline bool kvm_check_has_quirk(struct kvm *kvm, u64 quirk) void kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq, int inc_eip); u64 get_kvmclock_ns(struct kvm *kvm); +uint64_t kvm_get_wall_clock_epoch(struct kvm *kvm); int kvm_read_guest_virt(struct kvm_vcpu *vcpu, gva_t addr, void *val, unsigned int bytes, diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c index 40edf4d1974c..b946d9f28030 100644 --- a/arch/x86/kvm/xen.c +++ b/arch/x86/kvm/xen.c @@ -59,7 +59,7 @@ static int kvm_xen_shared_info_init(struct kvm *kvm, gfn_t gfn) * This code mirrors kvm_write_wall_clock() except that it writes * directly through the pfn cache and doesn't mark the page dirty. */ - wall_nsec = ktime_get_real_ns() - get_kvmclock_ns(kvm); + wall_nsec = kvm_get_wall_clock_epoch(kvm); /* It could be invalid again already, so we need to check */ read_lock_irq(&gpc->lock); @@ -98,7 +98,7 @@ static int kvm_xen_shared_info_init(struct kvm *kvm, gfn_t gfn) wc_version = wc->version = (wc->version + 1) | 1; smp_wmb(); - wc->nsec = do_div(wall_nsec, 1000000000); + wc->nsec = do_div(wall_nsec, NSEC_PER_SEC); wc->sec = (u32)wall_nsec; *wc_sec_hi = wall_nsec >> 32; smp_wmb(); From 598a790fc20f06e5582c939a4c5864ff1105c477 Mon Sep 17 00:00:00 2001 From: Jim Mattson Date: Fri, 29 Sep 2023 16:02:44 -0700 Subject: [PATCH 07/18] KVM: x86: Allow HWCR.McStatusWrEn to be cleared once set When HWCR is set to 0, store 0 in vcpu->arch.msr_hwcr. Fixes: 191c8137a939 ("x86/kvm: Implement HWCR support") Signed-off-by: Jim Mattson Link: https://lore.kernel.org/r/20230929230246.1954854-2-jmattson@google.com Signed-off-by: Sean Christopherson --- arch/x86/kvm/x86.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 42b33ef997a5..6f4534046803 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3791,12 +3791,11 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) data &= ~(u64)0x8; /* ignore TLB cache disable */ /* Handle McStatusWrEn */ - if (data == BIT_ULL(18)) { - vcpu->arch.msr_hwcr = data; - } else if (data != 0) { + if (data & ~BIT_ULL(18)) { kvm_pr_unimpl_wrmsr(vcpu, msr, data); return 1; } + vcpu->arch.msr_hwcr = data; break; case MSR_FAM10H_MMIO_CONF_BASE: if (data != 0) { From 8b0e00fba93449ecdda2c641e90c9b1f25f46669 Mon Sep 17 00:00:00 2001 From: Jim Mattson Date: Fri, 29 Sep 2023 16:02:45 -0700 Subject: [PATCH 08/18] KVM: x86: Virtualize HWCR.TscFreqSel[bit 24] MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit On certain CPUs, Linux guests expect HWCR.TscFreqSel[bit 24] to be set. If it isn't set, they complain: [Firmware Bug]: TSC doesn't count with P0 frequency! Allow userspace (and the guest) to set this bit in the virtual HWCR to eliminate the above complaint. Allow the guest to write the bit even though its is R/O on *some* CPUs. Like many bits in HWRC, TscFreqSel is not architectural at all. On Family 10h[1], it was R/W and powered on as 0. In Family 15h, one of the "changes relative to Family 10H Revision D processors[2] was: • MSRC001_0015 [Hardware Configuration (HWCR)]: • Dropped TscFreqSel; TSC can no longer be selected to run at NB P0-state. Despite the "Dropped" above, that same document later describes HWCR[bit 24] as follows: TscFreqSel: TSC frequency select. Read-only. Reset: 1. 1=The TSC increments at the P0 frequency If the guest clears the bit, the worst case scenario is the guest will be no worse off than it is today, e.g. the whining may return after a guest clears the bit and kexec()'s into a new kernel. [1] https://www.amd.com/content/dam/amd/en/documents/archived-tech-docs/programmer-references/31116.pdf [2] https://www.amd.com/content/dam/amd/en/documents/archived-tech-docs/programmer-references/42301_15h_Mod_00h-0Fh_BKDG.pdf, Signed-off-by: Jim Mattson Link: https://lore.kernel.org/r/20230929230246.1954854-3-jmattson@google.com [sean: elaborate on why the bit is writable by the guest] Signed-off-by: Sean Christopherson --- arch/x86/kvm/x86.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 6f4534046803..6c3485215994 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3790,8 +3790,12 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) data &= ~(u64)0x100; /* ignore ignne emulation enable */ data &= ~(u64)0x8; /* ignore TLB cache disable */ - /* Handle McStatusWrEn */ - if (data & ~BIT_ULL(18)) { + /* + * Allow McStatusWrEn and TscFreqSel. (Linux guests from v3.2 + * through at least v6.6 whine if TscFreqSel is clear, + * depending on F/M/S. + */ + if (data & ~(BIT_ULL(18) | BIT_ULL(24))) { kvm_pr_unimpl_wrmsr(vcpu, msr, data); return 1; } From 591455325a79dbe3be429006f8156f727b2a52dd Mon Sep 17 00:00:00 2001 From: Jim Mattson Date: Fri, 29 Sep 2023 16:02:46 -0700 Subject: [PATCH 09/18] KVM: selftests: Test behavior of HWCR, a.k.a. MSR_K7_HWCR Verify the following behavior holds true for writes and reads of HWCR from host userspace: * Attempts to set bits 3, 6, or 8 are ignored * Bits 18 and 24 are the only bits that can be set * Any bit that can be set can also be cleared Signed-off-by: Jim Mattson Link: https://lore.kernel.org/r/20230929230246.1954854-4-jmattson@google.com Co-developed-by: Sean Christopherson Signed-off-by: Sean Christopherson --- tools/testing/selftests/kvm/Makefile | 1 + .../selftests/kvm/x86_64/hwcr_msr_test.c | 47 +++++++++++++++++++ 2 files changed, 48 insertions(+) create mode 100644 tools/testing/selftests/kvm/x86_64/hwcr_msr_test.c diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile index a3bb36fb3cfc..fb01c3f8d3da 100644 --- a/tools/testing/selftests/kvm/Makefile +++ b/tools/testing/selftests/kvm/Makefile @@ -66,6 +66,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/dirty_log_page_splitting_test TEST_GEN_PROGS_x86_64 += x86_64/get_msr_index_features TEST_GEN_PROGS_x86_64 += x86_64/exit_on_emulation_failure_test TEST_GEN_PROGS_x86_64 += x86_64/fix_hypercall_test +TEST_GEN_PROGS_x86_64 += x86_64/hwcr_msr_test TEST_GEN_PROGS_x86_64 += x86_64/hyperv_clock TEST_GEN_PROGS_x86_64 += x86_64/hyperv_cpuid TEST_GEN_PROGS_x86_64 += x86_64/hyperv_evmcs diff --git a/tools/testing/selftests/kvm/x86_64/hwcr_msr_test.c b/tools/testing/selftests/kvm/x86_64/hwcr_msr_test.c new file mode 100644 index 000000000000..df351ae17029 --- /dev/null +++ b/tools/testing/selftests/kvm/x86_64/hwcr_msr_test.c @@ -0,0 +1,47 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2023, Google LLC. + */ + +#define _GNU_SOURCE /* for program_invocation_short_name */ +#include + +#include "test_util.h" +#include "kvm_util.h" +#include "vmx.h" + +void test_hwcr_bit(struct kvm_vcpu *vcpu, unsigned int bit) +{ + const uint64_t ignored = BIT_ULL(3) | BIT_ULL(6) | BIT_ULL(8); + const uint64_t valid = BIT_ULL(18) | BIT_ULL(24); + const uint64_t legal = ignored | valid; + uint64_t val = BIT_ULL(bit); + uint64_t actual; + int r; + + r = _vcpu_set_msr(vcpu, MSR_K7_HWCR, val); + TEST_ASSERT(val & ~legal ? !r : r == 1, + "Expected KVM_SET_MSRS(MSR_K7_HWCR) = 0x%lx to %s", + val, val & ~legal ? "fail" : "succeed"); + + actual = vcpu_get_msr(vcpu, MSR_K7_HWCR); + TEST_ASSERT(actual == (val & valid), + "Bit %u: unexpected HWCR 0x%lx; expected 0x%lx", + bit, actual, (val & valid)); + + vcpu_set_msr(vcpu, MSR_K7_HWCR, 0); +} + +int main(int argc, char *argv[]) +{ + struct kvm_vm *vm; + struct kvm_vcpu *vcpu; + unsigned int bit; + + vm = vm_create_with_one_vcpu(&vcpu, NULL); + + for (bit = 0; bit < BITS_PER_LONG; bit++) + test_hwcr_bit(vcpu, bit); + + kvm_vm_free(vm); +} From bf328e22e47242695445a33da6887ba2ef9aec19 Mon Sep 17 00:00:00 2001 From: Like Xu Date: Sun, 8 Oct 2023 10:53:35 +0800 Subject: [PATCH 10/18] KVM: x86: Don't sync user-written TSC against startup values The legacy API for setting the TSC is fundamentally broken, and only allows userspace to set a TSC "now", without any way to account for time lost between the calculation of the value, and the kernel eventually handling the ioctl. To work around this, KVM has a hack which, if a TSC is set with a value which is within a second's worth of the last TSC "written" to any vCPU in the VM, assumes that userspace actually intended the two TSC values to be in sync and adjusts the newly-written TSC value accordingly. Thus, when a VMM restores a guest after suspend or migration using the legacy API, the TSCs aren't necessarily *right*, but at least they're in sync. This trick falls down when restoring a guest which genuinely has been running for less time than the 1 second of imprecision KVM allows for in in the legacy API. On *creation*, the first vCPU starts its TSC counting from zero, and the subsequent vCPUs synchronize to that. But then when the VMM tries to restore a vCPU's intended TSC, because the VM has been alive for less than 1 second and KVM's default TSC value for new vCPU's is '0', the intended TSC is within a second of the last "written" TSC and KVM incorrectly adjusts the intended TSC in an attempt to synchronize. But further hacks can be piled onto KVM's existing hackish ABI, and declare that the *first* value written by *userspace* (on any vCPU) should not be subject to this "correction", i.e. KVM can assume that the first write from userspace is not an attempt to sync up with TSC values that only come from the kernel's default vCPU creation. To that end: Add a flag, kvm->arch.user_set_tsc, protected by kvm->arch.tsc_write_lock, to record that a TSC for at least one vCPU in the VM *has* been set by userspace, and make the 1-second slop hack only trigger if user_set_tsc is already set. Note that userspace can explicitly request a *synchronization* of the TSC by writing zero. For the purpose of user_set_tsc, an explicit synchronization counts as "setting" the TSC, i.e. if userspace then subsequently writes an explicit non-zero value which happens to be within 1 second of the previous value, the new value will be "corrected". This behavior is deliberate, as treating explicit synchronization as "setting" the TSC preserves KVM's existing behaviour inasmuch as possible (KVM always applied the 1-second "correction" regardless of whether the write came from userspace vs. the kernel). Reported-by: Yong He Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217423 Suggested-by: Oliver Upton Original-by: Oliver Upton Original-by: Sean Christopherson Signed-off-by: Like Xu Tested-by: Yong He Reviewed-by: Maxim Levitsky Link: https://lore.kernel.org/r/20231008025335.7419-1-likexu@tencent.com Signed-off-by: Sean Christopherson --- arch/x86/include/asm/kvm_host.h | 1 + arch/x86/kvm/x86.c | 34 +++++++++++++++++++++++---------- 2 files changed, 25 insertions(+), 10 deletions(-) diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h index c356aa27fbee..da497cc56e84 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -1332,6 +1332,7 @@ struct kvm_arch { int nr_vcpus_matched_tsc; u32 default_tsc_khz; + bool user_set_tsc; seqcount_raw_spinlock_t pvclock_sc; bool use_master_clock; diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 6c3485215994..756bdc1127ff 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -2709,8 +2709,9 @@ static void __kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 offset, u64 tsc, kvm_track_tsc_matching(vcpu); } -static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data) +static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 *user_value) { + u64 data = user_value ? *user_value : 0; struct kvm *kvm = vcpu->kvm; u64 offset, ns, elapsed; unsigned long flags; @@ -2725,25 +2726,37 @@ static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data) if (vcpu->arch.virtual_tsc_khz) { if (data == 0) { /* - * detection of vcpu initialization -- need to sync - * with other vCPUs. This particularly helps to keep - * kvm_clock stable after CPU hotplug + * Force synchronization when creating a vCPU, or when + * userspace explicitly writes a zero value. */ synchronizing = true; - } else { + } else if (kvm->arch.user_set_tsc) { u64 tsc_exp = kvm->arch.last_tsc_write + nsec_to_cycles(vcpu, elapsed); u64 tsc_hz = vcpu->arch.virtual_tsc_khz * 1000LL; /* - * Special case: TSC write with a small delta (1 second) - * of virtual cycle time against real time is - * interpreted as an attempt to synchronize the CPU. + * Here lies UAPI baggage: when a user-initiated TSC write has + * a small delta (1 second) of virtual cycle time against the + * previously set vCPU, we assume that they were intended to be + * in sync and the delta was only due to the racy nature of the + * legacy API. + * + * This trick falls down when restoring a guest which genuinely + * has been running for less time than the 1 second of imprecision + * which we allow for in the legacy API. In this case, the first + * value written by userspace (on any vCPU) should not be subject + * to this 'correction' to make it sync up with values that only + * come from the kernel's default vCPU creation. Make the 1-second + * slop hack only trigger if the user_set_tsc flag is already set. */ synchronizing = data < tsc_exp + tsc_hz && data + tsc_hz > tsc_exp; } } + if (user_value) + kvm->arch.user_set_tsc = true; + /* * For a reliable TSC, we can match TSC offsets, and for an unstable * TSC, we add elapsed time in this computation. We could let the @@ -3870,7 +3883,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) break; case MSR_IA32_TSC: if (msr_info->host_initiated) { - kvm_synchronize_tsc(vcpu, data); + kvm_synchronize_tsc(vcpu, &data); } else { u64 adj = kvm_compute_l1_tsc_offset(vcpu, data) - vcpu->arch.l1_tsc_offset; adjust_tsc_offset_guest(vcpu, adj); @@ -5629,6 +5642,7 @@ static int kvm_arch_tsc_set_attr(struct kvm_vcpu *vcpu, tsc = kvm_scale_tsc(rdtsc(), vcpu->arch.l1_tsc_scaling_ratio) + offset; ns = get_kvmclock_base_ns(); + kvm->arch.user_set_tsc = true; __kvm_synchronize_tsc(vcpu, offset, tsc, ns, matched); raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags); @@ -12055,7 +12069,7 @@ void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu) if (mutex_lock_killable(&vcpu->mutex)) return; vcpu_load(vcpu); - kvm_synchronize_tsc(vcpu, 0); + kvm_synchronize_tsc(vcpu, NULL); vcpu_put(vcpu); /* poll control enabled by default */ From 26951ec8623e915823985e86d2c428213f110659 Mon Sep 17 00:00:00 2001 From: Peng Hao Date: Fri, 13 Oct 2023 19:30:20 +0800 Subject: [PATCH 11/18] KVM: x86: Use octal for file permission Convert all module params to octal permissions to improve code readability and to make checkpatch happy: WARNING: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'. Signed-off-by: Peng Hao Link: https://lore.kernel.org/r/20231013113020.77523-1-flyingpeng@tencent.com Signed-off-by: Sean Christopherson --- arch/x86/kvm/svm/svm.c | 2 +- arch/x86/kvm/vmx/vmx.c | 20 ++++++++++---------- arch/x86/kvm/x86.c | 18 +++++++++--------- 3 files changed, 20 insertions(+), 20 deletions(-) diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c index 9507df93f410..3872c4f3689c 100644 --- a/arch/x86/kvm/svm/svm.c +++ b/arch/x86/kvm/svm/svm.c @@ -199,7 +199,7 @@ module_param_named(npt, npt_enabled, bool, 0444); /* allow nested virtualization in KVM/SVM */ static int nested = true; -module_param(nested, int, S_IRUGO); +module_param(nested, int, 0444); /* enable/disable Next RIP Save */ int nrips = true; diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c index 72e3943f3693..f17cb5f84f4e 100644 --- a/arch/x86/kvm/vmx/vmx.c +++ b/arch/x86/kvm/vmx/vmx.c @@ -82,28 +82,28 @@ bool __read_mostly enable_vpid = 1; module_param_named(vpid, enable_vpid, bool, 0444); static bool __read_mostly enable_vnmi = 1; -module_param_named(vnmi, enable_vnmi, bool, S_IRUGO); +module_param_named(vnmi, enable_vnmi, bool, 0444); bool __read_mostly flexpriority_enabled = 1; -module_param_named(flexpriority, flexpriority_enabled, bool, S_IRUGO); +module_param_named(flexpriority, flexpriority_enabled, bool, 0444); bool __read_mostly enable_ept = 1; -module_param_named(ept, enable_ept, bool, S_IRUGO); +module_param_named(ept, enable_ept, bool, 0444); bool __read_mostly enable_unrestricted_guest = 1; module_param_named(unrestricted_guest, - enable_unrestricted_guest, bool, S_IRUGO); + enable_unrestricted_guest, bool, 0444); bool __read_mostly enable_ept_ad_bits = 1; -module_param_named(eptad, enable_ept_ad_bits, bool, S_IRUGO); +module_param_named(eptad, enable_ept_ad_bits, bool, 0444); static bool __read_mostly emulate_invalid_guest_state = true; -module_param(emulate_invalid_guest_state, bool, S_IRUGO); +module_param(emulate_invalid_guest_state, bool, 0444); static bool __read_mostly fasteoi = 1; -module_param(fasteoi, bool, S_IRUGO); +module_param(fasteoi, bool, 0444); -module_param(enable_apicv, bool, S_IRUGO); +module_param(enable_apicv, bool, 0444); bool __read_mostly enable_ipiv = true; module_param(enable_ipiv, bool, 0444); @@ -114,10 +114,10 @@ module_param(enable_ipiv, bool, 0444); * use VMX instructions. */ static bool __read_mostly nested = 1; -module_param(nested, bool, S_IRUGO); +module_param(nested, bool, 0444); bool __read_mostly enable_pml = 1; -module_param_named(pml, enable_pml, bool, S_IRUGO); +module_param_named(pml, enable_pml, bool, 0444); static bool __read_mostly error_on_inconsistent_vmcs_config = true; module_param(error_on_inconsistent_vmcs_config, bool, 0444); diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 756bdc1127ff..94ccad95f7e0 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -145,21 +145,21 @@ EXPORT_STATIC_CALL_GPL(kvm_x86_get_cs_db_l_bits); EXPORT_STATIC_CALL_GPL(kvm_x86_cache_reg); static bool __read_mostly ignore_msrs = 0; -module_param(ignore_msrs, bool, S_IRUGO | S_IWUSR); +module_param(ignore_msrs, bool, 0644); bool __read_mostly report_ignored_msrs = true; -module_param(report_ignored_msrs, bool, S_IRUGO | S_IWUSR); +module_param(report_ignored_msrs, bool, 0644); EXPORT_SYMBOL_GPL(report_ignored_msrs); unsigned int min_timer_period_us = 200; -module_param(min_timer_period_us, uint, S_IRUGO | S_IWUSR); +module_param(min_timer_period_us, uint, 0644); static bool __read_mostly kvmclock_periodic_sync = true; -module_param(kvmclock_periodic_sync, bool, S_IRUGO); +module_param(kvmclock_periodic_sync, bool, 0444); /* tsc tolerance in parts per million - default to 1/2 of the NTP threshold */ static u32 __read_mostly tsc_tolerance_ppm = 250; -module_param(tsc_tolerance_ppm, uint, S_IRUGO | S_IWUSR); +module_param(tsc_tolerance_ppm, uint, 0644); /* * lapic timer advance (tscdeadline mode only) in nanoseconds. '-1' enables @@ -168,13 +168,13 @@ module_param(tsc_tolerance_ppm, uint, S_IRUGO | S_IWUSR); * tuning, i.e. allows privileged userspace to set an exact advancement time. */ static int __read_mostly lapic_timer_advance_ns = -1; -module_param(lapic_timer_advance_ns, int, S_IRUGO | S_IWUSR); +module_param(lapic_timer_advance_ns, int, 0644); static bool __read_mostly vector_hashing = true; -module_param(vector_hashing, bool, S_IRUGO); +module_param(vector_hashing, bool, 0444); bool __read_mostly enable_vmware_backdoor = false; -module_param(enable_vmware_backdoor, bool, S_IRUGO); +module_param(enable_vmware_backdoor, bool, 0444); EXPORT_SYMBOL_GPL(enable_vmware_backdoor); /* @@ -186,7 +186,7 @@ static int __read_mostly force_emulation_prefix; module_param(force_emulation_prefix, int, 0644); int __read_mostly pi_inject_timer = -1; -module_param(pi_inject_timer, bint, S_IRUGO | S_IWUSR); +module_param(pi_inject_timer, bint, 0644); /* Enable/disable PMU virtualization */ bool __read_mostly enable_pmu = true; From 3d30bfcbdc268b41a7d9253a9cb3082281223b19 Mon Sep 17 00:00:00 2001 From: David Matlack Date: Mon, 16 Oct 2023 15:12:28 -0700 Subject: [PATCH 12/18] KVM: x86/mmu: Stop kicking vCPUs to sync the dirty log when PML is disabled Stop kicking vCPUs in kvm_arch_sync_dirty_log() when PML is disabled. Kicking vCPUs when PML is disabled serves no purpose and could negatively impact guest performance. This restores KVM's behavior to prior to 5.12 commit a018eba53870 ("KVM: x86: Move MMU's PML logic to common code"), which replaced a static_call_cond(kvm_x86_flush_log_dirty) with unconditional calls to kvm_vcpu_kick(). Fixes: a018eba53870 ("KVM: x86: Move MMU's PML logic to common code") Signed-off-by: David Matlack Link: https://lore.kernel.org/r/20231016221228.1348318-1-dmatlack@google.com Signed-off-by: Sean Christopherson --- arch/x86/kvm/x86.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 94ccad95f7e0..051ca1978442 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -6355,6 +6355,9 @@ void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot) struct kvm_vcpu *vcpu; unsigned long i; + if (!kvm_x86_ops.cpu_dirty_log_size) + return; + kvm_for_each_vcpu(i, vcpu, kvm) kvm_vcpu_kick(vcpu); } From 5a989bbead4cc9cd67c3911226ad857d293643d8 Mon Sep 17 00:00:00 2001 From: Mingwei Zhang Date: Tue, 17 Oct 2023 23:26:10 +0000 Subject: [PATCH 13/18] KVM: x86: Update the variable naming in kvm_x86_ops.sched_in() Update the variable with name 'kvm' in kvm_x86_ops.sched_in() to 'vcpu' to avoid confusions. Variable naming in KVM has a clear convention that 'kvm' refers to pointer of type 'struct kvm *', while 'vcpu' refers to pointer of type 'struct kvm_vcpu *'. Fix this 9-year old naming issue for fun. Signed-off-by: Mingwei Zhang Link: https://lore.kernel.org/r/20231017232610.4008690-1-mizhang@google.com Signed-off-by: Sean Christopherson --- arch/x86/include/asm/kvm_host.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h index da497cc56e84..63e08b17a7b7 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -1701,7 +1701,7 @@ struct kvm_x86_ops { void (*request_immediate_exit)(struct kvm_vcpu *vcpu); - void (*sched_in)(struct kvm_vcpu *kvm, int cpu); + void (*sched_in)(struct kvm_vcpu *vcpu, int cpu); /* * Size of the CPU's dirty log buffer, i.e. VMX's PML buffer. A zero From d6800af51c76b6dae20e6023bbdc9b3da3ab5121 Mon Sep 17 00:00:00 2001 From: Nicolas Saenz Julienne Date: Tue, 17 Oct 2023 15:51:02 +0000 Subject: [PATCH 14/18] KVM: x86: hyper-v: Don't auto-enable stimer on write from user-space Don't apply the stimer's counter side effects when modifying its value from user-space, as this may trigger spurious interrupts. For example: - The stimer is configured in auto-enable mode. - The stimer's count is set and the timer enabled. - The stimer expires, an interrupt is injected. - The VM is live migrated. - The stimer config and count are deserialized, auto-enable is ON, the stimer is re-enabled. - The stimer expires right away, and injects an unwarranted interrupt. Cc: stable@vger.kernel.org Fixes: 1f4b34f825e8 ("kvm/x86: Hyper-V SynIC timers") Signed-off-by: Nicolas Saenz Julienne Reviewed-by: Vitaly Kuznetsov Link: https://lore.kernel.org/r/20231017155101.40677-1-nsaenz@amazon.com Signed-off-by: Sean Christopherson --- arch/x86/kvm/hyperv.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c index 7c2dac6824e2..238afd7335e4 100644 --- a/arch/x86/kvm/hyperv.c +++ b/arch/x86/kvm/hyperv.c @@ -727,10 +727,12 @@ static int stimer_set_count(struct kvm_vcpu_hv_stimer *stimer, u64 count, stimer_cleanup(stimer); stimer->count = count; - if (stimer->count == 0) - stimer->config.enable = 0; - else if (stimer->config.auto_enable) - stimer->config.enable = 1; + if (!host) { + if (stimer->count == 0) + stimer->config.enable = 0; + else if (stimer->config.auto_enable) + stimer->config.enable = 1; + } if (stimer->config.enable) stimer_mark_pending(stimer, false); From 2081a8450ef803e2e1136a80975eebc8682879b5 Mon Sep 17 00:00:00 2001 From: Dongli Zhang Date: Sun, 1 Oct 2023 14:36:37 -0700 Subject: [PATCH 15/18] KVM: x86: remove always-false condition in kvmclock_sync_fn The 'kvmclock_periodic_sync' is a readonly param that cannot change after bootup. The kvm_arch_vcpu_postcreate() is not going to schedule the kvmclock_sync_work if kvmclock_periodic_sync == false. As a result, the "if (!kvmclock_periodic_sync)" can never be true if the kvmclock_sync_work = kvmclock_sync_fn() is scheduled. Link: https://lore.kernel.org/kvm/a461bf3f-c17e-9c3f-56aa-726225e8391d@oracle.com Signed-off-by: Dongli Zhang Link: https://lore.kernel.org/r/20231001213637.76686-1-dongli.zhang@oracle.com Signed-off-by: Sean Christopherson --- arch/x86/kvm/x86.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 051ca1978442..2b1d8e22502b 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3374,9 +3374,6 @@ static void kvmclock_sync_fn(struct work_struct *work) kvmclock_sync_work); struct kvm *kvm = container_of(ka, struct kvm, arch); - if (!kvmclock_periodic_sync) - return; - schedule_delayed_work(&kvm->arch.kvmclock_update_work, 0); schedule_delayed_work(&kvm->arch.kvmclock_sync_work, KVMCLOCK_SYNC_PERIOD); From 329369caeccb1a0cb0dc5f864d6a671ed8190c11 Mon Sep 17 00:00:00 2001 From: Jim Mattson Date: Wed, 4 Oct 2023 20:12:37 -0700 Subject: [PATCH 16/18] x86: KVM: Add feature flag for CPUID.80000021H:EAX[bit 1] Define an X86_FEATURE_* flag for CPUID.80000021H:EAX.[bit 1], and advertise the feature to userspace via KVM_GET_SUPPORTED_CPUID. Per AMD's "Processor Programming Reference (PPR) for AMD Family 19h Model 61h, Revision B1 Processors (56713-B1-PUB)," this CPUID bit indicates that a WRMSR to MSR_FS_BASE, MSR_GS_BASE, or MSR_KERNEL_GS_BASE is non-serializing. This is a change in previously architected behavior. Effectively, this CPUID bit is a "defeature" bit, or a reverse polarity feature bit. When this CPUID bit is clear, the feature (serialization on WRMSR to any of these three MSRs) is available. When this CPUID bit is set, the feature is not available. KVM_GET_SUPPORTED_CPUID must pass this bit through from the underlying hardware, if it is set. Leaving the bit clear claims that WRMSR to these three MSRs will be serializing in a guest running under KVM. That isn't true. Though KVM could emulate the feature by intercepting writes to the specified MSRs, it does not do so today. The guest is allowed direct read/write access to these MSRs without interception, so the innate hardware behavior is preserved under KVM. Signed-off-by: Jim Mattson Reviewed-by: Maxim Levitsky Reviewed-by: Borislav Petkov (AMD) Link: https://lore.kernel.org/r/20231005031237.1652871-1-jmattson@google.com Signed-off-by: Sean Christopherson --- arch/x86/include/asm/cpufeatures.h | 1 + arch/x86/kvm/cpuid.c | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h index 58cb9495e40f..4af140cf5719 100644 --- a/arch/x86/include/asm/cpufeatures.h +++ b/arch/x86/include/asm/cpufeatures.h @@ -443,6 +443,7 @@ /* AMD-defined Extended Feature 2 EAX, CPUID level 0x80000021 (EAX), word 20 */ #define X86_FEATURE_NO_NESTED_DATA_BP (20*32+ 0) /* "" No Nested Data Breakpoints */ +#define X86_FEATURE_WRMSR_XX_BASE_NS (20*32+ 1) /* "" WRMSR to {FS,GS,KERNEL_GS}_BASE is non-serializing */ #define X86_FEATURE_LFENCE_RDTSC (20*32+ 2) /* "" LFENCE always serializing / synchronizes RDTSC */ #define X86_FEATURE_NULL_SEL_CLR_BASE (20*32+ 6) /* "" Null Selector Clears Base */ #define X86_FEATURE_AUTOIBRS (20*32+ 8) /* "" Automatic IBRS */ diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c index 552dc4f3899b..9b2d65d62d78 100644 --- a/arch/x86/kvm/cpuid.c +++ b/arch/x86/kvm/cpuid.c @@ -761,7 +761,8 @@ void kvm_set_cpu_caps(void) kvm_cpu_cap_mask(CPUID_8000_0021_EAX, F(NO_NESTED_DATA_BP) | F(LFENCE_RDTSC) | 0 /* SmmPgCfgLock */ | - F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */ + F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */ | + F(WRMSR_XX_BASE_NS) ); kvm_cpu_cap_check_and_set(X86_FEATURE_SBPB); From 122ae01c5159eb1584e07907aec89e2470622baa Mon Sep 17 00:00:00 2001 From: Liang Chen Date: Thu, 19 Oct 2023 12:33:36 +0800 Subject: [PATCH 17/18] KVM: x86: remove the unused assigned_dev_head from kvm_arch Legacy device assignment was dropped years ago. This field is not used anymore. Signed-off-by: Liang Chen Link: https://lore.kernel.org/r/20231019043336.8998-1-liangchen.linux@gmail.com Signed-off-by: Sean Christopherson --- arch/x86/include/asm/kvm_host.h | 1 - arch/x86/kvm/x86.c | 1 - 2 files changed, 2 deletions(-) diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h index 63e08b17a7b7..f980a2a9dd7e 100644 --- a/arch/x86/include/asm/kvm_host.h +++ b/arch/x86/include/asm/kvm_host.h @@ -1284,7 +1284,6 @@ struct kvm_arch { */ spinlock_t mmu_unsync_pages_lock; - struct list_head assigned_dev_head; struct iommu_domain *iommu_domain; bool iommu_noncoherent; #define __KVM_HAVE_ARCH_NONCOHERENT_DMA diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 2b1d8e22502b..923eeca1c971 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -12425,7 +12425,6 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) goto out_uninit_mmu; INIT_HLIST_HEAD(&kvm->arch.mask_notifier_list); - INIT_LIST_HEAD(&kvm->arch.assigned_dev_head); atomic_set(&kvm->arch.noncoherent_dma_count, 0); /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */ From 2770d4722036d6bd24bcb78e9cd7f6e572077d03 Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Thu, 19 Oct 2023 18:06:57 +0200 Subject: [PATCH 18/18] KVM: x86: Ignore MSR_AMD64_TW_CFG access Hyper-V enabled Windows Server 2022 KVM VM cannot be started on Zen1 Ryzen since it crashes at boot with SYSTEM_THREAD_EXCEPTION_NOT_HANDLED + STATUS_PRIVILEGED_INSTRUCTION (in other words, because of an unexpected #GP in the guest kernel). This is because Windows tries to set bit 8 in MSR_AMD64_TW_CFG and can't handle receiving a #GP when doing so. Give this MSR the same treatment that commit 2e32b7190641 ("x86, kvm: Add MSR_AMD64_BU_CFG2 to the list of ignored MSRs") gave MSR_AMD64_BU_CFG2 under justification that this MSR is baremetal-relevant only. Although apparently it was then needed for Linux guests, not Windows as in this case. With this change, the aforementioned guest setup is able to finish booting successfully. This issue can be reproduced either on a Summit Ridge Ryzen (with just "-cpu host") or on a Naples EPYC (with "-cpu host,stepping=1" since EPYC is ordinarily stepping 2). Alternatively, userspace could solve the problem by using MSR filters, but forcing every userspace to define a filter isn't very friendly and doesn't add much, if any, value. The only potential hiccup is if one of these "baremetal-only" MSRs ever requires actual emulation and/or has F/M/S specific behavior. But if that happens, then KVM can still punt *that* handling to userspace since userspace MSR filters "win" over KVM's default handling. Signed-off-by: Maciej S. Szmigiero Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/1ce85d9c7c9e9632393816cf19c902e0a3f411f1.1697731406.git.maciej.szmigiero@oracle.com [sean: call out MSR filtering alternative] Signed-off-by: Sean Christopherson --- arch/x86/include/asm/msr-index.h | 1 + arch/x86/kvm/x86.c | 2 ++ 2 files changed, 3 insertions(+) diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h index 1d111350197f..8bcbebb56b8f 100644 --- a/arch/x86/include/asm/msr-index.h +++ b/arch/x86/include/asm/msr-index.h @@ -553,6 +553,7 @@ #define MSR_AMD64_CPUID_FN_1 0xc0011004 #define MSR_AMD64_LS_CFG 0xc0011020 #define MSR_AMD64_DC_CFG 0xc0011022 +#define MSR_AMD64_TW_CFG 0xc0011023 #define MSR_AMD64_DE_CFG 0xc0011029 #define MSR_AMD64_DE_CFG_LFENCE_SERIALIZE_BIT 1 diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 923eeca1c971..725df1f2e2b7 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -3722,6 +3722,7 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) case MSR_AMD64_PATCH_LOADER: case MSR_AMD64_BU_CFG2: case MSR_AMD64_DC_CFG: + case MSR_AMD64_TW_CFG: case MSR_F15H_EX_CFG: break; @@ -4168,6 +4169,7 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info) case MSR_AMD64_BU_CFG2: case MSR_IA32_PERF_CTL: case MSR_AMD64_DC_CFG: + case MSR_AMD64_TW_CFG: case MSR_F15H_EX_CFG: /* * Intel Sandy Bridge CPUs must support the RAPL (running average power