Christoph ,here is the updated patch, as stated in previous mail, if there are
different Mci_CTL in different CPU, vMCE will be disabled
Any comments?
--jyh
Not GP fault when guest write non 0s or 1s to MCA CTL MSRs.
a) For Mci_CTL MSR, Guest can write any value to it. When read back, it will be
ANDed with the physical value. Some bit in physical value can be 0, either
because read-only in hardware (like masked by AMD's Mci_CTL_MASK), or because
Xen didn't enable it.
If guest write some bit as 0, while that bit is 1 in host, we will not
inject MCE corresponding that bank to guest, as we can't distinguish if the MCE
is caused by the guest-cleared bit.
b) For MCG_CTL MSR, guest can write any value to it. When read back, it will be
ANDed with the physical value.
If guest does not write all 1s. In mca_ctl_conflict(), we simply not inject
any vMCE to guest if some bit is set in physical MSR while is cleared in guest
's vMCG_CTL MSR.
Signed-off-by: Jiang, Yunhong <yunhong.jiang@xxxxxxxxx>
diff -r c814db5ffd8a xen/arch/x86/cpu/mcheck/mce.c
--- a/xen/arch/x86/cpu/mcheck/mce.c Thu Jan 28 18:57:01 2010 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce.c Thu Jan 28 23:54:35 2010 +0800
@@ -30,6 +30,13 @@ unsigned int nr_mce_banks;
unsigned int nr_mce_banks;
static uint64_t g_mcg_cap;
+/* We will try to inject vMCE to corresponding guest */
+int vmce_enable = 1;
+
+/* Real value in physical CTL MSR */
+static uint64_t h_mcg_ctl = 0UL;
+static uint64_t *h_mci_ctrl;
+int firstbank;
static void intpose_init(void);
static void mcinfo_clear(struct mc_info *);
@@ -642,6 +649,34 @@ void mcheck_init(struct cpuinfo_x86 *c)
break;
}
+ if ( !h_mci_ctrl )
+ {
+ h_mci_ctrl = xmalloc_array(uint64_t, nr_mce_banks);
+ if (!h_mci_ctrl)
+ {
+ dprintk(XENLOG_INFO, "Failed to alloc h_mci_ctrl\n");
+ return;
+ }
+ /* Don't care banks before firstbank */
+ memset(h_mci_ctrl, 0xff, sizeof(h_mci_ctrl));
+ for (i = firstbank; i < nr_mce_banks; i++)
+ rdmsrl(MSR_IA32_MC0_CTL + 4*i, h_mci_ctrl[i]);
+ } else {
+ uint64_t ctrl;
+
+ for ( i = firstbank; i < nr_mce_banks; i++ )
+ {
+ rdmsrl(MSR_IA32_MC0_CTL + 4*i, ctrl);
+ if ( ctrl != h_mci_ctrl[i] )
+ {
+ dprintk(XENLOG_WARNING,
+ "Different MCi ctl in pCPUs, no vMCE to guest\n");
+ vmce_enable = 0;
+ }
+ }
+ }
+ if (g_mcg_cap & MCG_CTL_P)
+ rdmsrl(MSR_IA32_MCG_CTL, h_mcg_ctl);
set_poll_bankmask(c);
if (!inited)
printk(XENLOG_INFO "CPU%i: No machine check initialization\n",
@@ -708,7 +743,8 @@ int mce_rdmsr(uint32_t msr, uint64_t *va
*val);
break;
case MSR_IA32_MCG_CTL:
- *val = d->arch.vmca_msrs.mcg_ctl;
+ /* Always 0 if no CTL support */
+ *val = d->arch.vmca_msrs.mcg_ctl & h_mcg_ctl;
mce_printk(MCE_VERBOSE, "MCE: rdmsr MCG_CTL 0x%"PRIx64"\n",
*val);
break;
@@ -723,7 +759,8 @@ int mce_rdmsr(uint32_t msr, uint64_t *va
switch (msr & (MSR_IA32_MC0_CTL | 3))
{
case MSR_IA32_MC0_CTL:
- *val = d->arch.vmca_msrs.mci_ctl[bank];
+ *val = d->arch.vmca_msrs.mci_ctl[bank] &
+ (h_mci_ctrl ? h_mci_ctrl[bank] : ~0UL);
mce_printk(MCE_VERBOSE, "MCE: rdmsr MC%u_CTL 0x%"PRIx64"\n",
bank, *val);
break;
@@ -805,13 +842,6 @@ int mce_wrmsr(u32 msr, u64 val)
switch ( msr )
{
case MSR_IA32_MCG_CTL:
- if ( val && (val + 1) )
- {
- mce_printk(MCE_QUIET, "MCE: val \"%"PRIx64"\" written "
- "to MCG_CTL should be all 0s or 1s\n", val);
- ret = -1;
- break;
- }
d->arch.vmca_msrs.mcg_ctl = val;
break;
case MSR_IA32_MCG_STATUS:
@@ -855,14 +885,6 @@ int mce_wrmsr(u32 msr, u64 val)
switch ( msr & (MSR_IA32_MC0_CTL | 3) )
{
case MSR_IA32_MC0_CTL:
- if ( val && (val + 1) )
- {
- mce_printk(MCE_QUIET, "MCE: val written to MC%u_CTL "
- "should be all 0s or 1s (is %"PRIx64")\n",
- bank, val);
- ret = -1;
- break;
- }
d->arch.vmca_msrs.mci_ctl[bank] = val;
break;
case MSR_IA32_MC0_STATUS:
@@ -1162,6 +1184,26 @@ void intpose_inval(unsigned int cpu_nr,
(r) <= MSR_IA32_MC0_MISC + (nr_mce_banks - 1) * 4 && \
((r) - MSR_IA32_MC0_CTL) % 4 != 0) /* excludes MCi_CTL */
+int mce_inject_vmce(struct mcinfo_bank *bank, struct domain *d)
+{
+ int bank_nr;
+
+ if ( !vmce_enable )
+ return 0;
+
+ if ( !bank || !d || !h_mci_ctrl )
+ return 0;
+
+ /* Will MCE happen in host if If host mcg_ctl is 0? */
+ if ( ~d->arch.vmca_msrs.mcg_ctl & h_mcg_ctl )
+ return 0;
+
+ bank_nr = bank->mc_bank;
+ if (~d->arch.vmca_msrs.mci_ctl[bank_nr] & h_mci_ctrl[bank_nr] )
+ return 0;
+ return 1;
+}
+
static int x86_mc_msrinject_verify(struct xen_mc_msrinject *mci)
{
struct cpuinfo_x86 *c;
diff -r c814db5ffd8a xen/arch/x86/cpu/mcheck/mce.h
--- a/xen/arch/x86/cpu/mcheck/mce.h Thu Jan 28 18:57:01 2010 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce.h Thu Jan 28 23:55:01 2010 +0800
@@ -39,6 +39,8 @@ void amd_nonfatal_mcheck_init(struct cpu
void amd_nonfatal_mcheck_init(struct cpuinfo_x86 *c);
u64 mce_cap_init(void);
+extern int firstbank;
+int mce_inject_vmce(struct mcinfo_bank *bank, struct domain *d);
int intel_mce_rdmsr(uint32_t msr, uint64_t *val);
int intel_mce_wrmsr(uint32_t msr, uint64_t val);
diff -r c814db5ffd8a xen/arch/x86/cpu/mcheck/mce_intel.c
--- a/xen/arch/x86/cpu/mcheck/mce_intel.c Thu Jan 28 18:57:01 2010 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce_intel.c Thu Jan 28 23:55:18 2010 +0800
@@ -20,7 +20,6 @@ int ser_support = 0;
int ser_support = 0;
static int nr_intel_ext_msrs = 0;
-static int firstbank;
/* Below are for MCE handling */
struct mce_softirq_barrier {
@@ -361,7 +360,15 @@ static void intel_UCR_handler(struct mci
* the mfn in question) */
BUG_ON( result->owner == DOMID_COW );
if ( result->owner != DOMID_XEN ) {
+
d = get_domain_by_id(result->owner);
+ if ( !mce_inject_vmce(bank, d) )
+ {
+ /* Guest has different MCE ctl with hypervisor */
+ put_domain(d);
+ return;
+ }
+
gfn =
mfn_to_gmfn(d, ((bank->mc_addr) >> PAGE_SHIFT));
bank->mc_addr =
>If this really happen, that means the physical CPU has different MCE setting.
>In that
>situation, I think we should not try to inject any vMCE to guest, because the
>vCPU
>can combine to any physical CPU.
>But, still I suspect if under any situation will that happen.
>
>--jyh
vmsr.patch
Description: vmsr.patch
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home