[RFC 2/2] Add MCE simulation support to qemu/kvm

[Huang Ying <ying.huang@xxxxxxxxx>]

MCE features are detected, initialized and injected via the
corresponding KVM ioctl.

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

---
 kvm-all.c            |   24 ++++++++++++++++++
 kvm.h                |    4 +++
 target-i386/helper.c |    8 +++++-
 target-i386/kvm.c    |   67 ++++++++++++++++++++++++++++++++++++++++++++++++++-
 4 files changed, 101 insertions(+), 2 deletions(-)

--- a/kvm-all.c
+++ b/kvm-all.c
@@ -765,6 +765,30 @@ int kvm_has_sync_mmu(void)
     return 0;
 }
 
+int kvm_has_mce(void)
+{
+#ifdef KVM_CAP_MCE
+    KVMState *s = kvm_state;
+    int r;
+
+    r = kvm_ioctl(s, KVM_CHECK_EXTENSION, KVM_CAP_MCE);
+    if (r > 0)
+        return r;
+#endif
+    return 0;
+}
+
+int kvm_get_mce_cap_supported(uint64_t *mce_cap)
+{
+#ifdef KVM_CAP_MCE
+    KVMState *s = kvm_state;
+
+    return kvm_ioctl(s, KVM_X86_GET_MCE_CAP_SUPPORTED, mce_cap);
+#else
+    return -ENOSYS;
+#endif
+}
+
 #ifdef KVM_CAP_SET_GUEST_DEBUG
 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *env,
                                                  target_ulong pc)
--- a/kvm.h
+++ b/kvm.h
@@ -47,6 +47,10 @@ int kvm_log_start(target_phys_addr_t phy
 int kvm_log_stop(target_phys_addr_t phys_addr, ram_addr_t size);
 
 int kvm_has_sync_mmu(void);
+int kvm_has_mce(void);
+int kvm_get_mce_cap_supported(uint64_t *mce_cap);
+void kvm_inject_x86_mce(CPUState *cenv, int bank, uint64_t status,
+                        uint64_t mcg_status, uint64_t addr, uint64_t misc);
 
 int kvm_coalesce_mmio_region(target_phys_addr_t start, ram_addr_t size);
 int kvm_uncoalesce_mmio_region(target_phys_addr_t start, ram_addr_t size);
--- a/target-i386/helper.c
+++ b/target-i386/helper.c
@@ -1440,6 +1440,11 @@ 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;
 
@@ -1757,7 +1762,8 @@ CPUX86State *cpu_x86_init(const char *cp
         cpu_x86_close(env);
         return NULL;
     }
-    mce_init(env);
+    if (!kvm_enabled())
+        mce_init(env);
     cpu_reset(env);
 #ifdef CONFIG_KQEMU
     kqemu_init(env);
--- a/target-i386/kvm.c
+++ b/target-i386/kvm.c
@@ -34,6 +34,42 @@
     do { } while (0)
 #endif
 
+static void kvm_arch_setup_mce(CPUState *env)
+{
+    int banks;
+    int ret;
+    uint64_t mcg_cap;
+
+#ifdef KVM_CAP_MCE
+    if (((env->cpuid_version >> 8) & 0xf) < 6)
+        return;
+
+    if ((env->cpuid_features & (CPUID_MCE|CPUID_MCA)) != (CPUID_MCE|CPUID_MCA))
+        return;
+
+    banks = kvm_has_mce();
+    if (banks <= 0)
+        return;
+
+    ret = kvm_get_mce_cap_supported(&mcg_cap);
+    if (ret) {
+        fprintf(stderr, "kvm_get_mce_cap_supported FAILED\n");
+        return;
+    }
+
+    if (banks > MCE_BANKS_DEF)
+        banks = MCE_BANKS_DEF;
+    mcg_cap &= MCE_CAP_DEF;
+    mcg_cap |= banks;
+
+    if (kvm_vcpu_ioctl(env, KVM_X86_SETUP_MCE, &mcg_cap)) {
+        fprintf(stderr, "kvm: setup mce FAILED\n");
+        return;
+    }
+    env->mcg_cap = mcg_cap;
+#endif
+}
+
 int kvm_arch_init_vcpu(CPUState *env)
 {
     struct {
@@ -42,6 +78,7 @@ int kvm_arch_init_vcpu(CPUState *env)
     } __attribute__((packed)) cpuid_data;
     uint32_t limit, i, j, cpuid_i;
     uint32_t unused;
+    int ret;
 
     cpuid_i = 0;
 
@@ -107,7 +144,13 @@ int kvm_arch_init_vcpu(CPUState *env)
 
     cpuid_data.cpuid.nent = cpuid_i;
 
-    return kvm_vcpu_ioctl(env, KVM_SET_CPUID2, &cpuid_data);
+    ret = kvm_vcpu_ioctl(env, KVM_SET_CPUID2, &cpuid_data);
+    if (ret < 0)
+        return ret;
+
+    kvm_arch_setup_mce(env);
+
+    return 0;
 }
 
 static int kvm_has_msr_star(CPUState *env)
@@ -665,6 +708,28 @@ int kvm_arch_handle_exit(CPUState *env, 
     return ret;
 }
 
+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,
+    };
+    int ret;
+
+    if (kvm_has_mce() <= 0)
+        return;
+
+    ret = kvm_vcpu_ioctl(cenv, KVM_X86_SET_MCE, &mce);
+    if (ret < 0)
+        fprintf(stderr, "kvm: inject mce FAILED\n");
+#endif
+}
+
 #ifdef KVM_CAP_SET_GUEST_DEBUG
 int kvm_arch_insert_sw_breakpoint(CPUState *env, struct kvm_sw_breakpoint *bp)
 {

Attachment: signature.asc
Description: This is a digitally signed message part