Virtual machines can exploit Intel ISA characterstics to cause
functional denial of service to the VMM. This series introduces a new
feature named Notify VM exit, which can help mitigate such kind of
attack.
Patch 1: An extension of KVM_SET_VCPU_EVENTS ioctl to inject a
synthesized shutdown event from user space. This is also a fix for other
synthesized triple fault, e.g. the RSM patch or nested_vmx_abort(),
which could get lost when exit to userspace to do migrate.
Patch 2: The main patch to enable Notify VM exit.
Patch 3: Add document for the new KVM capability and KVM exit reason.
---
Change logs:
v4 -> v5
- rename KVM_VCPUEVENTS_SHUTDOWN to KVM_VCPUEVENTS_TRIPLE_FAULT. Make it
bidirection and add it to get_vcpu_events. (Sean)
- v4: https://lore.kernel.org/all/[email protected]/
v3 -> v4
- Change this feature to per-VM scope. (Jim)
- Once VM_CONTEXT_INVALID set in exit_qualification, exit to user space
notify this fatal case, especially the notify VM exit happens in L2.
(Jim)
- extend KVM_SET_VCPU_EVENTS to allow user space to inject a shutdown
event. (Jim)
- A minor code changes.
- Add document for the new KVM capability.
- v3: https://lore.kernel.org/lkml/[email protected]/
v2 -> v3
- add a vcpu state notify_window_exits to record the number of
occurence as well as a pr_warn output. (Sean)
- Add the handling in nested VM to prevent L1 bypassing the restriction
through launching a L2. (Sean)
- Only kill L2 when L2 VM is context invalid, synthesize a
EXIT_REASON_TRIPLE_FAULT to L1 (Sean)
- To ease the current implementation, make module parameter
notify_window read-only. (Sean)
- Disable notify window exit by default.
- v2: https://lore.kernel.org/lkml/[email protected]/
v1 -> v2
- Default set notify window to 0, less than 0 to disable.
- Add more description in commit message.
---
Chenyi Qiang (2):
KVM: X86: Save&restore the triple fault request
KVM: Add document for KVM_CAP_X86_NOTIFY_VMEXIT and KVM_EXIT_NOTIFY
Tao Xu (1):
KVM: VMX: Enable Notify VM exit
Documentation/virt/kvm/api.rst | 45 ++++++++++++++++++++++++++++++
arch/x86/include/asm/kvm_host.h | 5 ++++
arch/x86/include/asm/vmx.h | 7 +++++
arch/x86/include/asm/vmxfeatures.h | 1 +
arch/x86/include/uapi/asm/kvm.h | 1 +
arch/x86/include/uapi/asm/vmx.h | 4 ++-
arch/x86/kvm/vmx/capabilities.h | 6 ++++
arch/x86/kvm/vmx/nested.c | 17 ++++++++++-
arch/x86/kvm/vmx/vmx.c | 44 +++++++++++++++++++++++++++--
arch/x86/kvm/x86.c | 28 +++++++++++++++++--
arch/x86/kvm/x86.h | 5 ++++
include/uapi/linux/kvm.h | 7 +++++
12 files changed, 162 insertions(+), 8 deletions(-)
--
2.17.1
From: Tao Xu <[email protected]>
There are cases that malicious virtual machines can cause CPU stuck (due
to event windows don't open up), e.g., infinite loop in microcode when
nested #AC (CVE-2015-5307). No event window means no event (NMI, SMI and
IRQ) can be delivered. It leads the CPU to be unavailable to host or
other VMs.
VMM can enable notify VM exit that a VM exit generated if no event
window occurs in VM non-root mode for a specified amount of time (notify
window).
Feature enabling:
- The new vmcs field SECONDARY_EXEC_NOTIFY_VM_EXITING is introduced to
enable this feature. VMM can set NOTIFY_WINDOW vmcs field to adjust
the expected notify window.
- Add a new KVM capability KVM_CAP_X86_NOTIFY_VMEXIT so that user space
can query and enable this feature in per-VM scope. Notify window is
also provided from user space during this process.
- if notify_window < 0, feature disabled;
- if notify_window >= 0, feature enabled;
- There's a possibility, however small, that a notify VM exit happens
with VM_CONTEXT_INVALID set in exit qualification, which means VM
context is corrupted. To avoid the false positive and a well-behaved
guest gets killed, set notify window as -1 to disable this feature by
default.
- It's safe to even set notify window to zero since an internal
hardware threshold is added to vmcs.notifiy_window.
VM exit handling:
- Introduce a vcpu state notify_window_exits to records the count of
notify VM exits and expose it through the debugfs.
- Warn the notify vm exit in kernel log since host can a) get an
indication that a guest is potentially malicious and b) rule out (or
confirm) notify VM exits as the source of degraded guest performance.
- Notify VM exit can happen incident to delivery of a vector event.
Allow it in KVM.
- Once VM_CONTEXT_INVALID bit is set, exit to user space for further
handling.
Nested handling
- Nested notify VM exits are not supported yet. Keep the same notify
window control in vmcs02 as vmcs01, so that L1 can't escape the
restriction of notify VM exits through launching L2 VM.
- When L2 VM is context invalid and user space should synthesize a shutdown
event to a vcpu. KVM makes KVM_REQ_TRIPLE_FAULT request accordingly
and it would synthesize a nested triple fault exit to L1 hypervisor to
kill L2.
Notify VM exit is defined in latest Intel Architecture Instruction Set
Extensions Programming Reference, chapter 9.2.
Co-developed-by: Xiaoyao Li <[email protected]>
Signed-off-by: Xiaoyao Li <[email protected]>
Signed-off-by: Tao Xu <[email protected]>
Co-developed-by: Chenyi Qiang <[email protected]>
Signed-off-by: Chenyi Qiang <[email protected]>
---
arch/x86/include/asm/kvm_host.h | 5 ++++
arch/x86/include/asm/vmx.h | 7 +++++
arch/x86/include/asm/vmxfeatures.h | 1 +
arch/x86/include/uapi/asm/vmx.h | 4 ++-
arch/x86/kvm/vmx/capabilities.h | 6 ++++
arch/x86/kvm/vmx/nested.c | 17 +++++++++++-
arch/x86/kvm/vmx/vmx.c | 44 ++++++++++++++++++++++++++++--
arch/x86/kvm/x86.c | 19 +++++++++++--
arch/x86/kvm/x86.h | 5 ++++
include/uapi/linux/kvm.h | 7 +++++
10 files changed, 108 insertions(+), 7 deletions(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 3a2c855f04e3..807562bb64ef 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1149,6 +1149,8 @@ struct kvm_arch {
bool bus_lock_detection_enabled;
bool enable_pmu;
+
+ int notify_window;
/*
* If exit_on_emulation_error is set, and the in-kernel instruction
* emulator fails to emulate an instruction, allow userspace
@@ -1285,6 +1287,7 @@ struct kvm_vcpu_stat {
u64 directed_yield_attempted;
u64 directed_yield_successful;
u64 guest_mode;
+ u64 notify_window_exits;
};
struct x86_instruction_info;
@@ -1641,6 +1644,8 @@ extern u64 kvm_max_tsc_scaling_ratio;
extern u64 kvm_default_tsc_scaling_ratio;
/* bus lock detection supported? */
extern bool kvm_has_bus_lock_exit;
+/* notify VM exit supported? */
+extern bool kvm_has_notify_vmexit;
extern u64 kvm_mce_cap_supported;
diff --git a/arch/x86/include/asm/vmx.h b/arch/x86/include/asm/vmx.h
index 0ffaa3156a4e..9104c85a973f 100644
--- a/arch/x86/include/asm/vmx.h
+++ b/arch/x86/include/asm/vmx.h
@@ -74,6 +74,7 @@
#define SECONDARY_EXEC_TSC_SCALING VMCS_CONTROL_BIT(TSC_SCALING)
#define SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE VMCS_CONTROL_BIT(USR_WAIT_PAUSE)
#define SECONDARY_EXEC_BUS_LOCK_DETECTION VMCS_CONTROL_BIT(BUS_LOCK_DETECTION)
+#define SECONDARY_EXEC_NOTIFY_VM_EXITING VMCS_CONTROL_BIT(NOTIFY_VM_EXITING)
#define PIN_BASED_EXT_INTR_MASK VMCS_CONTROL_BIT(INTR_EXITING)
#define PIN_BASED_NMI_EXITING VMCS_CONTROL_BIT(NMI_EXITING)
@@ -269,6 +270,7 @@ enum vmcs_field {
SECONDARY_VM_EXEC_CONTROL = 0x0000401e,
PLE_GAP = 0x00004020,
PLE_WINDOW = 0x00004022,
+ NOTIFY_WINDOW = 0x00004024,
VM_INSTRUCTION_ERROR = 0x00004400,
VM_EXIT_REASON = 0x00004402,
VM_EXIT_INTR_INFO = 0x00004404,
@@ -555,6 +557,11 @@ enum vm_entry_failure_code {
#define EPT_VIOLATION_EXECUTABLE (1 << EPT_VIOLATION_EXECUTABLE_BIT)
#define EPT_VIOLATION_GVA_TRANSLATED (1 << EPT_VIOLATION_GVA_TRANSLATED_BIT)
+/*
+ * Exit Qualifications for NOTIFY VM EXIT
+ */
+#define NOTIFY_VM_CONTEXT_INVALID BIT(0)
+
/*
* VM-instruction error numbers
*/
diff --git a/arch/x86/include/asm/vmxfeatures.h b/arch/x86/include/asm/vmxfeatures.h
index d9a74681a77d..15f0f2ab4f95 100644
--- a/arch/x86/include/asm/vmxfeatures.h
+++ b/arch/x86/include/asm/vmxfeatures.h
@@ -84,5 +84,6 @@
#define VMX_FEATURE_USR_WAIT_PAUSE ( 2*32+ 26) /* Enable TPAUSE, UMONITOR, UMWAIT in guest */
#define VMX_FEATURE_ENCLV_EXITING ( 2*32+ 28) /* "" VM-Exit on ENCLV (leaf dependent) */
#define VMX_FEATURE_BUS_LOCK_DETECTION ( 2*32+ 30) /* "" VM-Exit when bus lock caused */
+#define VMX_FEATURE_NOTIFY_VM_EXITING ( 2*32+ 31) /* VM-Exit when no event windows after notify window */
#endif /* _ASM_X86_VMXFEATURES_H */
diff --git a/arch/x86/include/uapi/asm/vmx.h b/arch/x86/include/uapi/asm/vmx.h
index 946d761adbd3..ef4c80f6553e 100644
--- a/arch/x86/include/uapi/asm/vmx.h
+++ b/arch/x86/include/uapi/asm/vmx.h
@@ -91,6 +91,7 @@
#define EXIT_REASON_UMWAIT 67
#define EXIT_REASON_TPAUSE 68
#define EXIT_REASON_BUS_LOCK 74
+#define EXIT_REASON_NOTIFY 75
#define VMX_EXIT_REASONS \
{ EXIT_REASON_EXCEPTION_NMI, "EXCEPTION_NMI" }, \
@@ -153,7 +154,8 @@
{ EXIT_REASON_XRSTORS, "XRSTORS" }, \
{ EXIT_REASON_UMWAIT, "UMWAIT" }, \
{ EXIT_REASON_TPAUSE, "TPAUSE" }, \
- { EXIT_REASON_BUS_LOCK, "BUS_LOCK" }
+ { EXIT_REASON_BUS_LOCK, "BUS_LOCK" }, \
+ { EXIT_REASON_NOTIFY, "NOTIFY"}
#define VMX_EXIT_REASON_FLAGS \
{ VMX_EXIT_REASONS_FAILED_VMENTRY, "FAILED_VMENTRY" }
diff --git a/arch/x86/kvm/vmx/capabilities.h b/arch/x86/kvm/vmx/capabilities.h
index 3f430e218375..0102a6e8a194 100644
--- a/arch/x86/kvm/vmx/capabilities.h
+++ b/arch/x86/kvm/vmx/capabilities.h
@@ -417,4 +417,10 @@ static inline u64 vmx_supported_debugctl(void)
return debugctl;
}
+static inline bool cpu_has_notify_vmexit(void)
+{
+ return vmcs_config.cpu_based_2nd_exec_ctrl &
+ SECONDARY_EXEC_NOTIFY_VM_EXITING;
+}
+
#endif /* __KVM_X86_VMX_CAPS_H */
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index f18744f7ff82..1bcf086d2ed4 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -2134,6 +2134,8 @@ static u64 nested_vmx_calc_efer(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
static void prepare_vmcs02_constant_state(struct vcpu_vmx *vmx)
{
+ struct kvm *kvm = vmx->vcpu.kvm;
+
/*
* If vmcs02 hasn't been initialized, set the constant vmcs02 state
* according to L0's settings (vmcs12 is irrelevant here). Host
@@ -2176,6 +2178,9 @@ static void prepare_vmcs02_constant_state(struct vcpu_vmx *vmx)
if (cpu_has_vmx_encls_vmexit())
vmcs_write64(ENCLS_EXITING_BITMAP, INVALID_GPA);
+ if (kvm_notify_vmexit_enabled(kvm))
+ vmcs_write32(NOTIFY_WINDOW, kvm->arch.notify_window);
+
/*
* Set the MSR load/store lists to match L0's settings. Only the
* addresses are constant (for vmcs02), the counts can change based
@@ -4218,8 +4223,15 @@ static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
/*
* Transfer the event that L0 or L1 may wanted to inject into
* L2 to IDT_VECTORING_INFO_FIELD.
+ *
+ * Skip this if the exit is due to a NOTIFY_VM_CONTEXT_INVALID
+ * exit; in that case, L0 will synthesize a nested TRIPLE_FAULT
+ * vmexit to kill L2. No IDT vectoring info is recorded for
+ * triple faults, and __vmx_handle_exit does not expect it.
*/
- vmcs12_save_pending_event(vcpu, vmcs12);
+ if (!(to_vmx(vcpu)->exit_reason.basic == EXIT_REASON_NOTIFY) &&
+ kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu))
+ vmcs12_save_pending_event(vcpu, vmcs12);
/*
* According to spec, there's no need to store the guest's
@@ -6085,6 +6097,9 @@ static bool nested_vmx_l1_wants_exit(struct kvm_vcpu *vcpu,
SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE);
case EXIT_REASON_ENCLS:
return nested_vmx_exit_handled_encls(vcpu, vmcs12);
+ case EXIT_REASON_NOTIFY:
+ /* Notify VM exit is not exposed to L1 */
+ return false;
default:
return true;
}
diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
index e8963f5af618..bd4f117c468c 100644
--- a/arch/x86/kvm/vmx/vmx.c
+++ b/arch/x86/kvm/vmx/vmx.c
@@ -2472,7 +2472,8 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf,
SECONDARY_EXEC_PT_USE_GPA |
SECONDARY_EXEC_PT_CONCEAL_VMX |
SECONDARY_EXEC_ENABLE_VMFUNC |
- SECONDARY_EXEC_BUS_LOCK_DETECTION;
+ SECONDARY_EXEC_BUS_LOCK_DETECTION |
+ SECONDARY_EXEC_NOTIFY_VM_EXITING;
if (cpu_has_sgx())
opt2 |= SECONDARY_EXEC_ENCLS_EXITING;
if (adjust_vmx_controls(min2, opt2,
@@ -4362,6 +4363,9 @@ static u32 vmx_secondary_exec_control(struct vcpu_vmx *vmx)
if (!vcpu->kvm->arch.bus_lock_detection_enabled)
exec_control &= ~SECONDARY_EXEC_BUS_LOCK_DETECTION;
+ if (!kvm_notify_vmexit_enabled(vcpu->kvm))
+ exec_control &= ~SECONDARY_EXEC_NOTIFY_VM_EXITING;
+
return exec_control;
}
@@ -4369,6 +4373,8 @@ static u32 vmx_secondary_exec_control(struct vcpu_vmx *vmx)
static void init_vmcs(struct vcpu_vmx *vmx)
{
+ struct kvm *kvm = vmx->vcpu.kvm;
+
if (nested)
nested_vmx_set_vmcs_shadowing_bitmap();
@@ -4397,12 +4403,15 @@ static void init_vmcs(struct vcpu_vmx *vmx)
vmcs_write64(POSTED_INTR_DESC_ADDR, __pa((&vmx->pi_desc)));
}
- if (!kvm_pause_in_guest(vmx->vcpu.kvm)) {
+ if (!kvm_pause_in_guest(kvm)) {
vmcs_write32(PLE_GAP, ple_gap);
vmx->ple_window = ple_window;
vmx->ple_window_dirty = true;
}
+ if (kvm_notify_vmexit_enabled(kvm))
+ vmcs_write32(NOTIFY_WINDOW, kvm->arch.notify_window);
+
vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0);
vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0);
vmcs_write32(CR3_TARGET_COUNT, 0); /* 22.2.1 */
@@ -5691,6 +5700,32 @@ static int handle_bus_lock_vmexit(struct kvm_vcpu *vcpu)
return 1;
}
+static int handle_notify(struct kvm_vcpu *vcpu)
+{
+ unsigned long exit_qual = vmx_get_exit_qual(vcpu);
+
+ ++vcpu->stat.notify_window_exits;
+ pr_warn_ratelimited("Notify window exits at address: 0x%lx\n",
+ kvm_rip_read(vcpu));
+
+ if (!(exit_qual & NOTIFY_VM_CONTEXT_INVALID)) {
+ /*
+ * Notify VM exit happened while executing iret from NMI,
+ * "blocked by NMI" bit has to be set before next VM entry.
+ */
+ if (enable_vnmi &&
+ (exit_qual & INTR_INFO_UNBLOCK_NMI))
+ vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO,
+ GUEST_INTR_STATE_NMI);
+
+ return 1;
+ }
+
+ vcpu->run->exit_reason = KVM_EXIT_NOTIFY;
+ vcpu->run->notify.data |= KVM_NOTIFY_CONTEXT_INVALID;
+ return 0;
+}
+
/*
* The exit handlers return 1 if the exit was handled fully and guest execution
* may resume. Otherwise they set the kvm_run parameter to indicate what needs
@@ -5748,6 +5783,7 @@ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = {
[EXIT_REASON_PREEMPTION_TIMER] = handle_preemption_timer,
[EXIT_REASON_ENCLS] = handle_encls,
[EXIT_REASON_BUS_LOCK] = handle_bus_lock_vmexit,
+ [EXIT_REASON_NOTIFY] = handle_notify,
};
static const int kvm_vmx_max_exit_handlers =
@@ -6112,7 +6148,8 @@ static int __vmx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
exit_reason.basic != EXIT_REASON_EPT_VIOLATION &&
exit_reason.basic != EXIT_REASON_PML_FULL &&
exit_reason.basic != EXIT_REASON_APIC_ACCESS &&
- exit_reason.basic != EXIT_REASON_TASK_SWITCH)) {
+ exit_reason.basic != EXIT_REASON_TASK_SWITCH &&
+ exit_reason.basic != EXIT_REASON_NOTIFY)) {
int ndata = 3;
vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
@@ -7987,6 +8024,7 @@ static __init int hardware_setup(void)
}
kvm_has_bus_lock_exit = cpu_has_vmx_bus_lock_detection();
+ kvm_has_notify_vmexit = cpu_has_notify_vmexit();
set_bit(0, vmx_vpid_bitmap); /* 0 is reserved for host */
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index fee402a700df..9fd693db6d9d 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -163,6 +163,8 @@ u64 __read_mostly kvm_default_tsc_scaling_ratio;
EXPORT_SYMBOL_GPL(kvm_default_tsc_scaling_ratio);
bool __read_mostly kvm_has_bus_lock_exit;
EXPORT_SYMBOL_GPL(kvm_has_bus_lock_exit);
+bool __read_mostly kvm_has_notify_vmexit;
+EXPORT_SYMBOL_GPL(kvm_has_notify_vmexit);
/* tsc tolerance in parts per million - default to 1/2 of the NTP threshold */
static u32 __read_mostly tsc_tolerance_ppm = 250;
@@ -291,7 +293,8 @@ const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
STATS_DESC_COUNTER(VCPU, nested_run),
STATS_DESC_COUNTER(VCPU, directed_yield_attempted),
STATS_DESC_COUNTER(VCPU, directed_yield_successful),
- STATS_DESC_ICOUNTER(VCPU, guest_mode)
+ STATS_DESC_ICOUNTER(VCPU, guest_mode),
+ STATS_DESC_COUNTER(VCPU, notify_window_exits),
};
const struct kvm_stats_header kvm_vcpu_stats_header = {
@@ -4359,10 +4362,13 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
if (r < sizeof(struct kvm_xsave))
r = sizeof(struct kvm_xsave);
break;
+ }
case KVM_CAP_PMU_CAPABILITY:
r = enable_pmu ? KVM_CAP_PMU_VALID_MASK : 0;
break;
- }
+ case KVM_CAP_X86_NOTIFY_VMEXIT:
+ r = kvm_has_notify_vmexit;
+ break;
default:
break;
}
@@ -6055,6 +6061,13 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
}
mutex_unlock(&kvm->lock);
break;
+ case KVM_CAP_X86_NOTIFY_VMEXIT:
+ r = -EINVAL;
+ if (!kvm_has_notify_vmexit)
+ break;
+ kvm->arch.notify_window = cap->args[0];
+ r = 0;
+ break;
default:
r = -EINVAL;
break;
@@ -11649,6 +11662,8 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
kvm->arch.guest_can_read_msr_platform_info = true;
kvm->arch.enable_pmu = enable_pmu;
+ kvm->arch.notify_window = -1;
+
#if IS_ENABLED(CONFIG_HYPERV)
spin_lock_init(&kvm->arch.hv_root_tdp_lock);
kvm->arch.hv_root_tdp = INVALID_PAGE;
diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
index aa86abad914d..cf115233ce18 100644
--- a/arch/x86/kvm/x86.h
+++ b/arch/x86/kvm/x86.h
@@ -349,6 +349,11 @@ static inline bool kvm_cstate_in_guest(struct kvm *kvm)
return kvm->arch.cstate_in_guest;
}
+static inline bool kvm_notify_vmexit_enabled(struct kvm *kvm)
+{
+ return kvm->arch.notify_window >= 0;
+}
+
enum kvm_intr_type {
/* Values are arbitrary, but must be non-zero. */
KVM_HANDLING_IRQ = 1,
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index d2f1efc3aa35..8f58196569a0 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -270,6 +270,7 @@ struct kvm_xen_exit {
#define KVM_EXIT_X86_BUS_LOCK 33
#define KVM_EXIT_XEN 34
#define KVM_EXIT_RISCV_SBI 35
+#define KVM_EXIT_NOTIFY 36
/* For KVM_EXIT_INTERNAL_ERROR */
/* Emulate instruction failed. */
@@ -487,6 +488,11 @@ struct kvm_run {
unsigned long args[6];
unsigned long ret[2];
} riscv_sbi;
+ /* KVM_EXIT_NOTIFY */
+ struct {
+#define KVM_NOTIFY_CONTEXT_INVALID (1 << 0)
+ __u32 data;
+ } notify;
/* Fix the size of the union. */
char padding[256];
};
@@ -1143,6 +1149,7 @@ struct kvm_ppc_resize_hpt {
#define KVM_CAP_PPC_AIL_MODE_3 210
#define KVM_CAP_S390_MEM_OP_EXTENSION 211
#define KVM_CAP_PMU_CAPABILITY 212
+#define KVM_CAP_X86_NOTIFY_VMEXIT 213
#ifdef KVM_CAP_IRQ_ROUTING
--
2.17.1
Add brief documentation for KVM_CAP_X86_NOTIFY_VMEXIT, as well as the
new field in kvm_run struct for the exit reason KVM_EXIT_NOTIFY.
Signed-off-by: Chenyi Qiang <[email protected]>
---
Documentation/virt/kvm/api.rst | 39 ++++++++++++++++++++++++++++++++++
1 file changed, 39 insertions(+)
diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
index 9682b0a438bd..d60b03b5a63e 100644
--- a/Documentation/virt/kvm/api.rst
+++ b/Documentation/virt/kvm/api.rst
@@ -6199,6 +6199,26 @@ array field represents return values. The userspace should update the return
values of SBI call before resuming the VCPU. For more details on RISC-V SBI
spec refer, https://github.com/riscv/riscv-sbi-doc.
+::
+
+ /* KVM_EXIT_NOTIFY */
+ struct {
+ #define KVM_NOTIFY_CONTEXT_INVALID (1 << 0)
+ __u32 data;
+ } notify;
+
+Used on x86 systems. When the VM capability KVM_CAP_X86_NOTIFY_VMEXIT is
+enabled and the parameter is non-negative, a VM exit generated if no event
+window occurs in VM non-root mode for a specified amount of time. In some
+special case, e.g. VM context invalid, it should exit to userspace with the
+exit reason KVM_EXIT_NOTIFY for further handling. The "data" field contains
+the more detailed info.
+
+Valid values for 'data' are:
+
+ - KVM_NOTIFY_CONTEXT_INVALID -- the VM context is corrupted and not valid
+ in VMCS. It would run into unknown result if resume the target VM.
+
::
/* Fix the size of the union. */
@@ -7085,6 +7105,25 @@ resource that is controlled with the H_SET_MODE hypercall.
This capability allows a guest kernel to use a better-performance mode for
handling interrupts and system calls.
+7.31 KVM_CAP_X86_NOTIFY_VMEXIT
+------------------------------
+
+:Architectures: x86
+:Target: VM
+:Parameters: args[0] is the value of notify window
+:Returns: 0 on success, -EINVAL if hardware doesn't support notify VM exit.
+
+This capability allows userspace to configure the notify VM exit on/off
+in per-VM scope during VM creation. Notify VM exit is disabled by default.
+When userspace provides a non-negative value in args[0], VMM would enable
+this feature to trigger VM exit if no event window occurs in VM non-root
+mode for a specified of time (notify window). The notify window is determined
+by args[0].
+
+This capability is aimed to mitigate the threat that malicious VMs can
+cause CPU stuck (due to event windows don't open up) and make the CPU
+unavailable to host or other VMs.
+
8. Other capabilities.
======================
--
2.17.1
On Fri, Mar 18, 2022, Chenyi Qiang wrote:
> diff --git a/arch/x86/include/asm/vmxfeatures.h b/arch/x86/include/asm/vmxfeatures.h
> index d9a74681a77d..15f0f2ab4f95 100644
> --- a/arch/x86/include/asm/vmxfeatures.h
> +++ b/arch/x86/include/asm/vmxfeatures.h
> @@ -84,5 +84,6 @@
> #define VMX_FEATURE_USR_WAIT_PAUSE ( 2*32+ 26) /* Enable TPAUSE, UMONITOR, UMWAIT in guest */
> #define VMX_FEATURE_ENCLV_EXITING ( 2*32+ 28) /* "" VM-Exit on ENCLV (leaf dependent) */
> #define VMX_FEATURE_BUS_LOCK_DETECTION ( 2*32+ 30) /* "" VM-Exit when bus lock caused */
> +#define VMX_FEATURE_NOTIFY_VM_EXITING ( 2*32+ 31) /* VM-Exit when no event windows after notify window */
>
> #endif /* _ASM_X86_VMXFEATURES_H */
> diff --git a/arch/x86/include/uapi/asm/vmx.h b/arch/x86/include/uapi/asm/vmx.h
> index 946d761adbd3..ef4c80f6553e 100644
> --- a/arch/x86/include/uapi/asm/vmx.h
> +++ b/arch/x86/include/uapi/asm/vmx.h
> @@ -91,6 +91,7 @@
> #define EXIT_REASON_UMWAIT 67
> #define EXIT_REASON_TPAUSE 68
> #define EXIT_REASON_BUS_LOCK 74
> +#define EXIT_REASON_NOTIFY 75
>
> #define VMX_EXIT_REASONS \
> { EXIT_REASON_EXCEPTION_NMI, "EXCEPTION_NMI" }, \
> @@ -153,7 +154,8 @@
> { EXIT_REASON_XRSTORS, "XRSTORS" }, \
> { EXIT_REASON_UMWAIT, "UMWAIT" }, \
> { EXIT_REASON_TPAUSE, "TPAUSE" }, \
> - { EXIT_REASON_BUS_LOCK, "BUS_LOCK" }
> + { EXIT_REASON_BUS_LOCK, "BUS_LOCK" }, \
> + { EXIT_REASON_NOTIFY, "NOTIFY"}
Uber nit, needs a space before the closing curly brace.
> #define VMX_EXIT_REASON_FLAGS \
> { VMX_EXIT_REASONS_FAILED_VMENTRY, "FAILED_VMENTRY" }
> diff --git a/arch/x86/kvm/vmx/capabilities.h b/arch/x86/kvm/vmx/capabilities.h
> index 3f430e218375..0102a6e8a194 100644
> --- a/arch/x86/kvm/vmx/capabilities.h
> +++ b/arch/x86/kvm/vmx/capabilities.h
> @@ -417,4 +417,10 @@ static inline u64 vmx_supported_debugctl(void)
> return debugctl;
> }
>
> +static inline bool cpu_has_notify_vmexit(void)
> +{
> + return vmcs_config.cpu_based_2nd_exec_ctrl &
> + SECONDARY_EXEC_NOTIFY_VM_EXITING;
> +}
> +
> #endif /* __KVM_X86_VMX_CAPS_H */
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index f18744f7ff82..1bcf086d2ed4 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -2134,6 +2134,8 @@ static u64 nested_vmx_calc_efer(struct vcpu_vmx *vmx, struct vmcs12 *vmcs12)
>
> static void prepare_vmcs02_constant_state(struct vcpu_vmx *vmx)
> {
> + struct kvm *kvm = vmx->vcpu.kvm;
> +
> /*
> * If vmcs02 hasn't been initialized, set the constant vmcs02 state
> * according to L0's settings (vmcs12 is irrelevant here). Host
> @@ -2176,6 +2178,9 @@ static void prepare_vmcs02_constant_state(struct vcpu_vmx *vmx)
> if (cpu_has_vmx_encls_vmexit())
> vmcs_write64(ENCLS_EXITING_BITMAP, INVALID_GPA);
>
> + if (kvm_notify_vmexit_enabled(kvm))
> + vmcs_write32(NOTIFY_WINDOW, kvm->arch.notify_window);
> +
> /*
> * Set the MSR load/store lists to match L0's settings. Only the
> * addresses are constant (for vmcs02), the counts can change based
> @@ -4218,8 +4223,15 @@ static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
> /*
> * Transfer the event that L0 or L1 may wanted to inject into
> * L2 to IDT_VECTORING_INFO_FIELD.
> + *
> + * Skip this if the exit is due to a NOTIFY_VM_CONTEXT_INVALID
> + * exit; in that case, L0 will synthesize a nested TRIPLE_FAULT
> + * vmexit to kill L2. No IDT vectoring info is recorded for
> + * triple faults, and __vmx_handle_exit does not expect it.
> */
> - vmcs12_save_pending_event(vcpu, vmcs12);
> + if (!(to_vmx(vcpu)->exit_reason.basic == EXIT_REASON_NOTIFY) &&
> + kvm_test_request(KVM_REQ_TRIPLE_FAULT, vcpu))
> + vmcs12_save_pending_event(vcpu, vmcs12);
Unless I'm misreading the parantheses, this is saving pending events if and only
if there's a pending triple fault. LOL, looks like you copy+pasted verbatim from
Paolo. https://lkml.kernel.org/r/[email protected]
Checking KVM_REQ_TRIPLE_FAULT is also flawed, KVM should never actually reach
this point with the triple fault pending (see my response to patch 1).
Explicitly checking for EXIT_REASON_TRIPLE_FAULT is more correct, but that's still
flawed as the VMCS needs to be explicitly written with zeros (well, bit 31 needs
to be cleared), i.e. this check needs to be moved inside vmcs12_save_pending_event().
Ha, and I think skipping this path on VMX_EXIT_REASONS_FAILED_VMENTRY is also
technically wrong. That's still a VM-Exit, I'm pretty sure failed VM-Entry only
skips the MSR load/store list processing (because the load lists on VM-Entry are
processed after the final consistency checks).
Lastly, this flaw is orthogonal to NOTIFY support, e.g. I'm guessing the MCE
shenanigans plus KVM_SET_VCPU_EVENTS can force this scenario. I.e. this needs to
be fixed in a separate patch.
All in all, I think doing the below across two patches would be do the trick.
I'll throw this into a small series along with the fix for the theoretical WARN
bugs in the forced exit paths.
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index f18744f7ff82..7e24bf7800fa 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -3695,12 +3695,16 @@ vmcs12_guest_cr4(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
}
static void vmcs12_save_pending_event(struct kvm_vcpu *vcpu,
- struct vmcs12 *vmcs12)
+ struct vmcs12 *vmcs12,
+ u32 vm_exit_reason)
{
u32 idt_vectoring;
unsigned int nr;
- if (vcpu->arch.exception.injected) {
+ if ((vm_exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY) ||
+ vm_exit_reason == EXIT_REASON_TRIPLE_FAULT) {
+ vmcs12->idt_vectoring_info_field = 0;
+ } else if (vcpu->arch.exception.injected) {
nr = vcpu->arch.exception.nr;
idt_vectoring = nr | VECTORING_INFO_VALID_MASK;
@@ -4215,12 +4219,6 @@ static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
* instead of reading the real value. */
vmcs12->vm_entry_intr_info_field &= ~INTR_INFO_VALID_MASK;
- /*
- * Transfer the event that L0 or L1 may wanted to inject into
- * L2 to IDT_VECTORING_INFO_FIELD.
- */
- vmcs12_save_pending_event(vcpu, vmcs12);
-
/*
* According to spec, there's no need to store the guest's
* MSRs if the exit is due to a VM-entry failure that occurs
@@ -4234,6 +4232,12 @@ static void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
VMX_ABORT_SAVE_GUEST_MSR_FAIL);
}
+ /*
+ * Transfer the event that L0 or L1 may wanted to inject into L2 to
+ * IDT_VECTORING_INFO_FIELD.
+ */
+ vmcs12_save_pending_event(vcpu, vmcs12, vm_exit_reason);
+
/*
* Drop what we picked up for L2 via vmx_complete_interrupts. It is
* preserved above and would only end up incorrectly in L1.
> /*
> * According to spec, there's no need to store the guest's
> @@ -6085,6 +6097,9 @@ static bool nested_vmx_l1_wants_exit(struct kvm_vcpu *vcpu,
> SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE);
> case EXIT_REASON_ENCLS:
> return nested_vmx_exit_handled_encls(vcpu, vmcs12);
> + case EXIT_REASON_NOTIFY:
> + /* Notify VM exit is not exposed to L1 */
> + return false;
> default:
> return true;
> }
...
> @@ -5691,6 +5700,32 @@ static int handle_bus_lock_vmexit(struct kvm_vcpu *vcpu)
> return 1;
> }
>
> +static int handle_notify(struct kvm_vcpu *vcpu)
> +{
> + unsigned long exit_qual = vmx_get_exit_qual(vcpu);
> +
> + ++vcpu->stat.notify_window_exits;
> + pr_warn_ratelimited("Notify window exits at address: 0x%lx\n",
> + kvm_rip_read(vcpu));
Gah, I'm pretty sure you added this printk at my suggestion[*]. That was a bad
suggestion, the host won't have any idea what that RIP means, and getting the info
to the right people/entity would be difficult, especially since it's ratelimited.
Unless someone objects, please drop the printk (I still like the stat though!)
Sorry for the churn :-(
[*] https://lkml.kernel.org/r/[email protected]
> +
> + if (!(exit_qual & NOTIFY_VM_CONTEXT_INVALID)) {
> + /*
> + * Notify VM exit happened while executing iret from NMI,
> + * "blocked by NMI" bit has to be set before next VM entry.
> + */
> + if (enable_vnmi &&
> + (exit_qual & INTR_INFO_UNBLOCK_NMI))
> + vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO,
> + GUEST_INTR_STATE_NMI);
> +
> + return 1;
> + }
> +
> + vcpu->run->exit_reason = KVM_EXIT_NOTIFY;
> + vcpu->run->notify.data |= KVM_NOTIFY_CONTEXT_INVALID;
> + return 0;
> +}
> +
> /*
> * The exit handlers return 1 if the exit was handled fully and guest execution
> * may resume. Otherwise they set the kvm_run parameter to indicate what needs
> @@ -5748,6 +5783,7 @@ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = {
> [EXIT_REASON_PREEMPTION_TIMER] = handle_preemption_timer,
> [EXIT_REASON_ENCLS] = handle_encls,
> [EXIT_REASON_BUS_LOCK] = handle_bus_lock_vmexit,
> + [EXIT_REASON_NOTIFY] = handle_notify,
> };
>
> static const int kvm_vmx_max_exit_handlers =
> @@ -6112,7 +6148,8 @@ static int __vmx_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
> exit_reason.basic != EXIT_REASON_EPT_VIOLATION &&
> exit_reason.basic != EXIT_REASON_PML_FULL &&
> exit_reason.basic != EXIT_REASON_APIC_ACCESS &&
> - exit_reason.basic != EXIT_REASON_TASK_SWITCH)) {
> + exit_reason.basic != EXIT_REASON_TASK_SWITCH &&
> + exit_reason.basic != EXIT_REASON_NOTIFY)) {
> int ndata = 3;
>
> vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
> @@ -7987,6 +8024,7 @@ static __init int hardware_setup(void)
> }
>
> kvm_has_bus_lock_exit = cpu_has_vmx_bus_lock_detection();
> + kvm_has_notify_vmexit = cpu_has_notify_vmexit();
Why is this info being cached? cpu_has_notify_vmexit() is not expensive, nor is
it called in a hot path. Oooh, it's so x86 can see it.
I'd prefer to add a flag to struct kvm_x86_ops, similar to cpu_dirty_log_size.
That avoids having all these global variables and their exports.
> set_bit(0, vmx_vpid_bitmap); /* 0 is reserved for host */
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index fee402a700df..9fd693db6d9d 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -163,6 +163,8 @@ u64 __read_mostly kvm_default_tsc_scaling_ratio;
> EXPORT_SYMBOL_GPL(kvm_default_tsc_scaling_ratio);
> bool __read_mostly kvm_has_bus_lock_exit;
> EXPORT_SYMBOL_GPL(kvm_has_bus_lock_exit);
> +bool __read_mostly kvm_has_notify_vmexit;
> +EXPORT_SYMBOL_GPL(kvm_has_notify_vmexit);
>
> /* tsc tolerance in parts per million - default to 1/2 of the NTP threshold */
> static u32 __read_mostly tsc_tolerance_ppm = 250;
> @@ -291,7 +293,8 @@ const struct _kvm_stats_desc kvm_vcpu_stats_desc[] = {
> STATS_DESC_COUNTER(VCPU, nested_run),
> STATS_DESC_COUNTER(VCPU, directed_yield_attempted),
> STATS_DESC_COUNTER(VCPU, directed_yield_successful),
> - STATS_DESC_ICOUNTER(VCPU, guest_mode)
> + STATS_DESC_ICOUNTER(VCPU, guest_mode),
> + STATS_DESC_COUNTER(VCPU, notify_window_exits),
> };
>
> const struct kvm_stats_header kvm_vcpu_stats_header = {
> @@ -4359,10 +4362,13 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
> if (r < sizeof(struct kvm_xsave))
> r = sizeof(struct kvm_xsave);
> break;
> + }
> case KVM_CAP_PMU_CAPABILITY:
> r = enable_pmu ? KVM_CAP_PMU_VALID_MASK : 0;
> break;
> - }
> + case KVM_CAP_X86_NOTIFY_VMEXIT:
> + r = kvm_has_notify_vmexit;
> + break;
> default:
> break;
> }
> @@ -6055,6 +6061,13 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
> }
> mutex_unlock(&kvm->lock);
> break;
> + case KVM_CAP_X86_NOTIFY_VMEXIT:
> + r = -EINVAL;
> + if (!kvm_has_notify_vmexit)
> + break;
> + kvm->arch.notify_window = cap->args[0];
Using -1 for "off" is both kludgy and flawed, e.g. it prevents usersepace from
configuring the full 32 bits of the field. Not that I expect userspace to do that,
but it's an unnecessary shortcoming.
Rather than -1, what about using bits 63:32 for the window value, and bits 31:0
for flags? Then we can have an explicit enable flag (no need to set notifi_window
to -1 during VM creation), and we can also reserve flags for future use, e.g. if
we want to support exiting to userspace on every NOTIFY exit. Actually, I don't
see any reason not to add that functionality straightaway. KVM already has the
new userpace exit reason, so it'd be like one or two lines of code here, plus the
same in the exit handler.
> + r = 0;
> + break;
> default:
> r = -EINVAL;
> break;
> @@ -11649,6 +11662,8 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
> kvm->arch.guest_can_read_msr_platform_info = true;
> kvm->arch.enable_pmu = enable_pmu;
>
> + kvm->arch.notify_window = -1;
> +
> #if IS_ENABLED(CONFIG_HYPERV)
> spin_lock_init(&kvm->arch.hv_root_tdp_lock);
> kvm->arch.hv_root_tdp = INVALID_PAGE;
> diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
> index aa86abad914d..cf115233ce18 100644
> --- a/arch/x86/kvm/x86.h
> +++ b/arch/x86/kvm/x86.h
> @@ -349,6 +349,11 @@ static inline bool kvm_cstate_in_guest(struct kvm *kvm)
> return kvm->arch.cstate_in_guest;
> }
>
> +static inline bool kvm_notify_vmexit_enabled(struct kvm *kvm)
> +{
> + return kvm->arch.notify_window >= 0;
> +}
> +
> enum kvm_intr_type {
> /* Values are arbitrary, but must be non-zero. */
> KVM_HANDLING_IRQ = 1,
> diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
> index d2f1efc3aa35..8f58196569a0 100644
> --- a/include/uapi/linux/kvm.h
> +++ b/include/uapi/linux/kvm.h
> @@ -270,6 +270,7 @@ struct kvm_xen_exit {
> #define KVM_EXIT_X86_BUS_LOCK 33
> #define KVM_EXIT_XEN 34
> #define KVM_EXIT_RISCV_SBI 35
> +#define KVM_EXIT_NOTIFY 36
>
> /* For KVM_EXIT_INTERNAL_ERROR */
> /* Emulate instruction failed. */
> @@ -487,6 +488,11 @@ struct kvm_run {
> unsigned long args[6];
> unsigned long ret[2];
> } riscv_sbi;
> + /* KVM_EXIT_NOTIFY */
> + struct {
> +#define KVM_NOTIFY_CONTEXT_INVALID (1 << 0)
> + __u32 data;
Probably better to use "flags" instead of "data".
> + } notify;
> /* Fix the size of the union. */
> char padding[256];
> };
> @@ -1143,6 +1149,7 @@ struct kvm_ppc_resize_hpt {
> #define KVM_CAP_PPC_AIL_MODE_3 210
> #define KVM_CAP_S390_MEM_OP_EXTENSION 211
> #define KVM_CAP_PMU_CAPABILITY 212
> +#define KVM_CAP_X86_NOTIFY_VMEXIT 213
>
> #ifdef KVM_CAP_IRQ_ROUTING
>
> --
> 2.17.1
>
On Wed, Apr 06, 2022, Sean Christopherson wrote:
> On Fri, Mar 18, 2022, Chenyi Qiang wrote:
> Ha, and I think skipping this path on VMX_EXIT_REASONS_FAILED_VMENTRY is also
> technically wrong. That's still a VM-Exit, I'm pretty sure failed VM-Entry only
> skips the MSR load/store list processing (because the load lists on VM-Entry are
> processed after the final consistency checks).
Nope, I'm wrong, the SDM explicit states that only EXIT_QUALIFICATION is modified
on VM-Entry failure VM-Exits. Which means we actually have the opposite bug as
KVM is clearing a bunch of fields.