[PATCH -v4 2/2] QEMU-KVM: MCE: Add MCE simulation support to qemu/kvm

[Huang Ying <ying.huang@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 7813 bytes --]

KVM ioctls are used to initialize MCE simulation and inject MCE. The
real MCE simulation is implemented in Linux kernel.


ChangeLog:

v3:

- Re-based on qemu/tcg MCE support patch

v2:

- Use new kernel MCE capability exportion interface.


Signed-off-by: Huang Ying <ying.huang@intel.com>

---
 kvm/kernel/arch/x86/include/asm/kvm.h |    1 
 kvm/kernel/include/linux/kvm.h        |   21 +++++++++++++++++
 kvm/libkvm/libkvm-x86.c               |   39 +++++++++++++++++++++++++++++++++
 kvm/libkvm/libkvm.h                   |    4 +++
 qemu-kvm-x86.c                        |   23 +++++++++++++++++++
 qemu-kvm.c                            |   40 ++++++++++++++++++++++++++++++++++
 qemu-kvm.h                            |    3 ++
 target-i386/helper.c                  |    5 ++++
 target-i386/machine.c                 |    4 +--
 9 files changed, 138 insertions(+), 2 deletions(-)

--- a/kvm/libkvm/libkvm-x86.c
+++ b/kvm/libkvm/libkvm-x86.c
@@ -378,6 +378,45 @@ int kvm_set_msrs(kvm_context_t kvm, int 
     return r;
 }
 
+int kvm_get_mce_cap_supported(kvm_context_t kvm, uint64_t *mce_cap,
+                              int *max_banks)
+{
+#ifdef KVM_CAP_MCE
+    int r;
+
+    r = ioctl(kvm->fd, KVM_CHECK_EXTENSION, KVM_CAP_MCE);
+    if (r > 0) {
+        *max_banks = r;
+        return ioctl(kvm->fd, KVM_X86_GET_MCE_CAP_SUPPORTED, mce_cap);
+    }
+#endif
+    return -ENOSYS;
+}
+
+int kvm_setup_mce(kvm_context_t kvm, int vcpu, uint64_t *mcg_cap)
+{
+#ifdef KVM_CAP_MCE
+    int r;
+
+    r = ioctl(kvm->fd, KVM_CHECK_EXTENSION, KVM_CAP_MCE);
+    if (r > 0)
+        return ioctl(kvm->vcpu_fd[vcpu], KVM_X86_SETUP_MCE, mcg_cap);
+#endif
+    return -ENOSYS;
+}
+
+int kvm_set_mce(kvm_context_t kvm, int vcpu, struct kvm_x86_mce *m)
+{
+#ifdef KVM_CAP_MCE
+    int r;
+
+    r = ioctl(kvm->fd, KVM_CHECK_EXTENSION, KVM_CAP_MCE);
+    if (r > 0)
+        return ioctl(kvm->vcpu_fd[vcpu], KVM_X86_SET_MCE, m);
+#endif
+    return -ENOSYS;
+}
+
 static void print_seg(FILE *file, const char *name, struct kvm_segment *seg)
 {
     	fprintf(stderr,
--- a/kvm/libkvm/libkvm.h
+++ b/kvm/libkvm/libkvm.h
@@ -27,6 +27,10 @@ typedef struct kvm_context *kvm_context_
 struct kvm_msr_list *kvm_get_msr_list(kvm_context_t);
 int kvm_get_msrs(kvm_context_t, int vcpu, struct kvm_msr_entry *msrs, int n);
 int kvm_set_msrs(kvm_context_t, int vcpu, struct kvm_msr_entry *msrs, int n);
+int kvm_get_mce_cap_supported(kvm_context_t, uint64_t *mce_cap, int *max_banks);
+int kvm_setup_mce(kvm_context_t, int vcpu, uint64_t *mcg_cap);
+struct kvm_x86_mce;
+int kvm_set_mce(kvm_context_t, int vcpu, struct kvm_x86_mce *mce);
 #endif
 
 /*!
--- a/qemu-kvm-x86.c
+++ b/qemu-kvm-x86.c
@@ -566,6 +566,29 @@ int kvm_arch_qemu_init_env(CPUState *cen
 	do_cpuid_ent(&cpuid_ent[cpuid_nent++], i, 0, &copy);
 
     kvm_setup_cpuid2(kvm_context, cenv->cpu_index, cpuid_nent, cpuid_ent);
+
+#ifdef KVM_CAP_MCE
+    if (((cenv->cpuid_version >> 8)&0xF) >= 6
+        && (cenv->cpuid_features&(CPUID_MCE|CPUID_MCA)) == (CPUID_MCE|CPUID_MCA)
+        && kvm_check_extension(kvm_context, KVM_CAP_MCE) > 0) {
+        uint64_t mcg_cap;
+        int banks;
+
+        if (kvm_get_mce_cap_supported(kvm_context, &mcg_cap, &banks))
+            perror("kvm_get_mce_cap_supported FAILED");
+        else {
+            if (banks > MCE_BANKS_DEF)
+                banks = MCE_BANKS_DEF;
+            mcg_cap &= MCE_CAP_DEF;
+            mcg_cap |= banks;
+            if (kvm_setup_mce(kvm_context, cenv->cpu_index, &mcg_cap))
+                perror("kvm_setup_mce FAILED");
+            else
+                cenv->mcg_cap = mcg_cap;
+        }
+    }
+#endif
+
     return 0;
 }
 
--- a/qemu-kvm.h
+++ b/qemu-kvm.h
@@ -236,4 +236,7 @@ uint32_t kvm_arch_get_supported_cpuid(CP
                                       int reg);
 
 
+void kvm_inject_x86_mce(CPUState *cenv, int bank, uint64_t status,
+                        uint64_t mcg_status, uint64_t addr, uint64_t misc);
+
 #endif
--- a/target-i386/helper.c
+++ b/target-i386/helper.c
@@ -1460,6 +1460,11 @@ void cpu_inject_x86_mce(CPUState *cenv, 
     unsigned bank_num = mcg_cap & 0xff;
     uint64_t *banks = cenv->mce_banks;
 
+    if (kvm_enabled()) {
+        kvm_inject_x86_mce(cenv, bank, status, mcg_status, addr, misc);
+        return;
+    }
+
     if (bank >= bank_num || !(status & MCI_STATUS_VAL))
         return;
 
--- a/kvm/kernel/include/linux/kvm.h
+++ b/kvm/kernel/include/linux/kvm.h
@@ -455,6 +455,9 @@ struct kvm_trace_rec {
 #define KVM_CAP_ASSIGN_DEV_IRQ 29
 /* Another bug in KVM_SET_USER_MEMORY_REGION fixed: */
 #define KVM_CAP_JOIN_MEMORY_REGIONS_WORKS 30
+#ifdef __KVM_HAVE_MCE
+#define KVM_CAP_MCE 31
+#endif
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
@@ -494,6 +497,20 @@ struct kvm_irq_routing {
 
 #endif
 
+#ifdef KVM_CAP_MCE
+/* x86 MCE */
+struct kvm_x86_mce {
+	__u64 status;
+	__u64 addr;
+	__u64 misc;
+	__u64 mcg_status;
+	__u8 bank;
+	__u8 pad1;
+	__u16 pad2;
+	__u32 pad3;
+};
+#endif
+
 /*
  * ioctls for VM fds
  */
@@ -581,6 +598,10 @@ struct kvm_irq_routing {
 #define KVM_NMI                   _IO(KVMIO,  0x9a)
 /* Available with KVM_CAP_SET_GUEST_DEBUG */
 #define KVM_SET_GUEST_DEBUG       _IOW(KVMIO,  0x9b, struct kvm_guest_debug)
+/* MCE for x86 */
+#define KVM_X86_SETUP_MCE         _IOW(KVMIO,  0x9c, __u64)
+#define KVM_X86_GET_MCE_CAP_SUPPORTED _IOR(KVMIO,  0x9d, __u64)
+#define KVM_X86_SET_MCE           _IOW(KVMIO,  0x9e, struct kvm_x86_mce)
 
 /*
  * Deprecated interfaces
--- a/kvm/kernel/arch/x86/include/asm/kvm.h
+++ b/kvm/kernel/arch/x86/include/asm/kvm.h
@@ -57,6 +57,7 @@
 #define __KVM_HAVE_USER_NMI
 #define __KVM_HAVE_GUEST_DEBUG
 #define __KVM_HAVE_MSIX
+#define __KVM_HAVE_MCE
 
 /* Architectural interrupt line count. */
 #define KVM_NR_INTERRUPTS 256
--- a/qemu-kvm.c
+++ b/qemu-kvm.c
@@ -1431,3 +1431,43 @@ uint32_t kvm_arch_get_supported_cpuid(CP
 {
     return kvm_get_supported_cpuid(kvm_context, function, reg);
 }
+
+#ifdef TARGET_I386
+#ifdef KVM_CAP_MCE
+struct kvm_x86_mce_data
+{
+    CPUState *env;
+    struct kvm_x86_mce *mce;
+};
+
+static void kvm_do_inject_x86_mce(void *_data)
+{
+    struct kvm_x86_mce_data *data = _data;
+    int r;
+
+    r = kvm_set_mce(kvm_context, data->env->cpu_index, data->mce);
+    if (r < 0)
+        perror("kvm_set_mce FAILED");
+}
+#endif
+
+void kvm_inject_x86_mce(CPUState *cenv, int bank, uint64_t status,
+                        uint64_t mcg_status, uint64_t addr, uint64_t misc)
+{
+#ifdef KVM_CAP_MCE
+    struct kvm_x86_mce mce = {
+        .bank = bank,
+        .status = status,
+        .mcg_status = mcg_status,
+        .addr = addr,
+        .misc = misc,
+    };
+    struct kvm_x86_mce_data data = {
+            .env = cenv,
+            .mce = &mce,
+    };
+
+    on_vcpu(cenv, kvm_do_inject_x86_mce, &data);
+#endif
+}
+#endif
--- a/target-i386/machine.c
+++ b/target-i386/machine.c
@@ -162,7 +162,7 @@ void cpu_save(QEMUFile *f, void *opaque)
 
     /* MCE */
     qemu_put_be64s(f, &env->mcg_cap);
-    if (env->mcg_cap) {
+    if (env->mcg_cap && !kvm_enabled()) {
         qemu_put_be64s(f, &env->mcg_status);
         qemu_put_be64s(f, &env->mcg_ctl);
         for (i = 0; i < env->mcg_cap & 0xff; i++) {
@@ -378,7 +378,7 @@ int cpu_load(QEMUFile *f, void *opaque, 
 
     if (version_id >= 10) {
         qemu_get_be64s(f, &env->mcg_cap);
-        if (env->mcg_cap) {
+        if (env->mcg_cap && !kvm_enabled()) {
             qemu_get_be64s(f, &env->mcg_status);
             qemu_get_be64s(f, &env->mcg_ctl);
             for (i = 0; i < env->mcg_cap & 0xff; i++) {


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 197 bytes --]