From: Chao Gao <chao.gao@xxxxxxxxx> At init time, KVM does compatibility checks to ensure that all online CPUs support hardware virtualization and a common set of features. But KVM uses hotplugged CPUs without such compatibility checks. On Intel CPUs, this leads to #GP if the hotplugged CPU doesn't support VMX or vmentry failure if the hotplugged CPU doesn't meet minimal feature requirements. Do compatibility checks when onlining a CPU and abort the online process if the hotplugged CPU is incompatible with online CPUs. CPU hotplug is disabled during hardware_enable_all() to prevent the corner case as shown below. A hotplugged CPU marks itself online in cpu_online_mask (1) and enables interrupt (2) before invoking callbacks registered in ONLINE section (3). So, if hardware_enable_all() is invoked on another CPU right after (2), then on_each_cpu() in hardware_enable_all() invokes hardware_enable_nolock() on the hotplugged CPU before kvm_online_cpu() is called. This makes the CPU escape from compatibility checks, which is risky. start_secondary { ... set_cpu_online(smp_processor_id(), true); <- 1 ... local_irq_enable(); <- 2 ... cpu_startup_entry(CPUHP_AP_ONLINE_IDLE); <- 3 } Keep compatibility checks at KVM init time. It can help to find incompatibility issues earlier and refuse to load arch KVM module (e.g., kvm-intel). Loosen the WARN_ON in kvm_arch_check_processor_compat so that it can be invoked from KVM's CPU hotplug callback (i.e., kvm_online_cpu). Other arch doesn't depends on prohibiting of preemption because powerpc has "strcmp(cur_cpu_spec->cpu_name, "model name")" and other arch has "return 0". Only x86 kvm_arch_check_processor_compat() has issue. Opportunistically, add a pr_err() for setup_vmcs_config() path in vmx_check_processor_compatibility() so that each possible error path has its own error message. Convert printk(KERN_ERR ... to pr_err to please checkpatch.pl Signed-off-by: Chao Gao <chao.gao@xxxxxxxxx> Reviewed-by: Sean Christopherson <seanjc@xxxxxxxxxx> Link: https://lore.kernel.org/r/20220216031528.92558-7-chao.gao@xxxxxxxxx Signed-off-by: Isaku Yamahata <isaku.yamahata@xxxxxxxxx> Reviewed-by: Yuan Yao <yuan.yao@xxxxxxxxx> --- arch/x86/kvm/vmx/vmx.c | 10 ++++++---- arch/x86/kvm/x86.c | 11 +++++++++-- virt/kvm/kvm_main.c | 18 +++++++++++++++++- 3 files changed, 32 insertions(+), 7 deletions(-) diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c index 3cf7f18a4115..2a1ab6495299 100644 --- a/arch/x86/kvm/vmx/vmx.c +++ b/arch/x86/kvm/vmx/vmx.c @@ -7421,20 +7421,22 @@ static int vmx_check_processor_compatibility(void) { struct vmcs_config vmcs_conf; struct vmx_capability vmx_cap; + int cpu = smp_processor_id(); if (!this_cpu_has(X86_FEATURE_MSR_IA32_FEAT_CTL) || !this_cpu_has(X86_FEATURE_VMX)) { - pr_err("kvm: VMX is disabled on CPU %d\n", smp_processor_id()); + pr_err("kvm: VMX is disabled on CPU %d\n", cpu); return -EIO; } - if (setup_vmcs_config(&vmcs_conf, &vmx_cap) < 0) + if (setup_vmcs_config(&vmcs_conf, &vmx_cap) < 0) { + pr_err("kvm: failed to setup vmcs config on CPU %d\n", cpu); return -EIO; + } if (nested) nested_vmx_setup_ctls_msrs(&vmcs_conf.nested, vmx_cap.ept); if (memcmp(&vmcs_config, &vmcs_conf, sizeof(struct vmcs_config)) != 0) { - printk(KERN_ERR "kvm: CPU %d feature inconsistency!\n", - smp_processor_id()); + pr_err("kvm: CPU %d feature inconsistency!\n", cpu); return -EIO; } return 0; diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index f6f014e78ab9..0ab8db07fa0e 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -12000,9 +12000,16 @@ void kvm_arch_hardware_unsetup(void) int kvm_arch_check_processor_compat(void) { - struct cpuinfo_x86 *c = &cpu_data(smp_processor_id()); + int cpu = smp_processor_id(); + struct cpuinfo_x86 *c = &cpu_data(cpu); - WARN_ON(!irqs_disabled()); + /* + * Compatibility checks are done when loading KVM or in KVM's CPU + * hotplug callback. It ensures all online CPUs are compatible to run + * vCPUs. For other cases, compatibility checks are unnecessary or + * even problematic. Try to detect improper usages here. + */ + WARN_ON(!irqs_disabled() && cpu_active(cpu)); if (__cr4_reserved_bits(cpu_has, c) != __cr4_reserved_bits(cpu_has, &boot_cpu_data)) diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index db1303e2abc9..0ac00c711384 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -5013,7 +5013,11 @@ static void hardware_enable_nolock(void *caller_name) static int kvm_online_cpu(unsigned int cpu) { - int ret = 0; + int ret; + + ret = kvm_arch_check_processor_compat(); + if (ret) + return ret; raw_spin_lock(&kvm_count_lock); /* @@ -5073,6 +5077,17 @@ static int hardware_enable_all(void) { int r = 0; + /* + * During onlining a CPU, cpu_online_mask is set before kvm_online_cpu() + * is called. on_each_cpu() between them includes the CPU. As a result, + * hardware_enable_nolock() may get invoked before kvm_online_cpu(). + * This would enable hardware virtualization on that cpu without + * compatibility checks, which can potentially crash system or break + * running VMs. + * + * Disable CPU hotplug to prevent this case from happening. + */ + cpus_read_lock(); raw_spin_lock(&kvm_count_lock); kvm_usage_count++; @@ -5087,6 +5102,7 @@ static int hardware_enable_all(void) } raw_spin_unlock(&kvm_count_lock); + cpus_read_unlock(); return r; } -- 2.25.1