* [PATCH 2/5] Xen/MCE: vMCE injection
@ 2012-09-19 8:03 Liu, Jinsong
2012-09-20 15:23 ` Christoph Egger
0 siblings, 1 reply; 7+ messages in thread
From: Liu, Jinsong @ 2012-09-19 8:03 UTC (permalink / raw)
To: Jan Beulich, xen-devel@lists.xensource.com
Cc: Christoph Egger, keir@xen.org, Ian.Campbell@citrix.com
[-- Attachment #1: Type: text/plain, Size: 2868 bytes --]
Xen/MCE: vMCE injection
In our test for win8 guest mce, we find a bug that no matter what SRAO/SRAR
error xen inject to win8 guest, it always reboot.
The root cause is, current Xen vMCE logic inject vMCE# only to vcpu0, this is
not correct for Intel MCE (Under Intel arch, h/w generate MCE# to all CPUs).
This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
--- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
@@ -340,48 +340,27 @@
int inject_vmce(struct domain *d)
{
- int cpu = smp_processor_id();
+ struct vcpu *v;
- /* PV guest and HVM guest have different vMCE# injection methods. */
- if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
+ /* inject vMCE to all vcpus */
+ for_each_vcpu(d, v)
{
- if ( d->is_hvm )
+ if ( !test_and_set_bool(v->mce_pending) &&
+ ((d->is_hvm) ||
+ guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
- d->domain_id);
- vcpu_kick(d->vcpu[0]);
+ mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ vcpu_kick(v);
}
else
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
- d->domain_id);
- if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
- {
- cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
- d->vcpu[0]->cpu_affinity);
- mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
- cpu, d->vcpu[0]->processor);
- vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
- vcpu_kick(d->vcpu[0]);
- }
- else
- {
- mce_printk(MCE_VERBOSE,
- "MCE: Kill PV guest with No MCE handler\n");
- domain_crash(d);
- }
+ mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ return -1;
}
}
- else
- {
- /* new vMCE comes while first one has not been injected yet,
- * in this case, inject fail. [We can't lose this vMCE for
- * the mce node's consistency].
- */
- mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
- " to this DOM%d!\n", d->domain_id);
- return -1;
- }
+
return 0;
}
[-- Attachment #2: 2_vmce_injection.patch --]
[-- Type: application/octet-stream, Size: 2793 bytes --]
Xen/MCE: vMCE injection
In our test for win8 guest mce, we find a bug that no matter what SRAO/SRAR
error xen inject to win8 guest, it always reboot.
The root cause is, current Xen vMCE logic inject vMCE# only to vcpu0, this is
not correct for Intel MCE (Under Intel arch, h/w generate MCE# to all CPUs).
This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
--- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
@@ -340,48 +340,27 @@
int inject_vmce(struct domain *d)
{
- int cpu = smp_processor_id();
+ struct vcpu *v;
- /* PV guest and HVM guest have different vMCE# injection methods. */
- if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
+ /* inject vMCE to all vcpus */
+ for_each_vcpu(d, v)
{
- if ( d->is_hvm )
+ if ( !test_and_set_bool(v->mce_pending) &&
+ ((d->is_hvm) ||
+ guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
- d->domain_id);
- vcpu_kick(d->vcpu[0]);
+ mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ vcpu_kick(v);
}
else
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
- d->domain_id);
- if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
- {
- cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
- d->vcpu[0]->cpu_affinity);
- mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
- cpu, d->vcpu[0]->processor);
- vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
- vcpu_kick(d->vcpu[0]);
- }
- else
- {
- mce_printk(MCE_VERBOSE,
- "MCE: Kill PV guest with No MCE handler\n");
- domain_crash(d);
- }
+ mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ return -1;
}
}
- else
- {
- /* new vMCE comes while first one has not been injected yet,
- * in this case, inject fail. [We can't lose this vMCE for
- * the mce node's consistency].
- */
- mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
- " to this DOM%d!\n", d->domain_id);
- return -1;
- }
+
return 0;
}
[-- Attachment #3: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH 2/5] Xen/MCE: vMCE injection
2012-09-19 8:03 [PATCH 2/5] Xen/MCE: vMCE injection Liu, Jinsong
@ 2012-09-20 15:23 ` Christoph Egger
2012-09-20 19:15 ` Liu, Jinsong
0 siblings, 1 reply; 7+ messages in thread
From: Christoph Egger @ 2012-09-20 15:23 UTC (permalink / raw)
To: Liu, Jinsong
Cc: xen-devel@lists.xensource.com, keir@xen.org,
Ian.Campbell@citrix.com, Jan Beulich
On 09/19/12 10:03, Liu, Jinsong wrote:
> Xen/MCE: vMCE injection
>
> In our test for win8 guest mce, we find a bug that no matter what SRAO/SRAR
> error xen inject to win8 guest, it always reboot.
>
> The root cause is, current Xen vMCE logic inject vMCE# only to vcpu0, this is
> not correct for Intel MCE (Under Intel arch, h/w generate MCE# to all CPUs).
>
> This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
This breaks the AMD way. The AMD way is to only inject it to vcpu0.
I suggest to add a flag argument to inject_vmce() that says whether
to inject to all vcpus or just vcpu0.
Then set/clear that flag from the caller side depending on whether you
run on Intel or AMD.
Christoph
>
> Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
>
> diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
> --- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
> +++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
> @@ -340,48 +340,27 @@
>
> int inject_vmce(struct domain *d)
> {
> - int cpu = smp_processor_id();
> + struct vcpu *v;
>
> - /* PV guest and HVM guest have different vMCE# injection methods. */
> - if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
> + /* inject vMCE to all vcpus */
> + for_each_vcpu(d, v)
> {
> - if ( d->is_hvm )
> + if ( !test_and_set_bool(v->mce_pending) &&
> + ((d->is_hvm) ||
> + guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
> {
> - mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
> - d->domain_id);
> - vcpu_kick(d->vcpu[0]);
> + mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
> + d->domain_id, v->vcpu_id);
> + vcpu_kick(v);
> }
> else
> {
> - mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
> - d->domain_id);
> - if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
> - {
> - cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
> - d->vcpu[0]->cpu_affinity);
> - mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
> - cpu, d->vcpu[0]->processor);
> - vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
> - vcpu_kick(d->vcpu[0]);
> - }
> - else
> - {
> - mce_printk(MCE_VERBOSE,
> - "MCE: Kill PV guest with No MCE handler\n");
> - domain_crash(d);
> - }
> + mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
> + d->domain_id, v->vcpu_id);
> + return -1;
> }
> }
> - else
> - {
> - /* new vMCE comes while first one has not been injected yet,
> - * in this case, inject fail. [We can't lose this vMCE for
> - * the mce node's consistency].
> - */
> - mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
> - " to this DOM%d!\n", d->domain_id);
> - return -1;
> - }
> +
> return 0;
> }
>
--
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Einsteinring 24, 85689 Dornach b. Muenchen
Geschaeftsfuehrer: Alberto Bozzo
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH 2/5] Xen/MCE: vMCE injection
2012-09-20 15:23 ` Christoph Egger
@ 2012-09-20 19:15 ` Liu, Jinsong
2012-09-21 7:38 ` Christoph Egger
0 siblings, 1 reply; 7+ messages in thread
From: Liu, Jinsong @ 2012-09-20 19:15 UTC (permalink / raw)
To: Christoph Egger
Cc: xen-devel@lists.xensource.com, keir@xen.org,
Ian.Campbell@citrix.com, Jan Beulich
Christoph Egger wrote:
> On 09/19/12 10:03, Liu, Jinsong wrote:
>
>> Xen/MCE: vMCE injection
>>
>> In our test for win8 guest mce, we find a bug that no matter what
>> SRAO/SRAR error xen inject to win8 guest, it always reboot.
>>
>> The root cause is, current Xen vMCE logic inject vMCE# only to
>> vcpu0, this is not correct for Intel MCE (Under Intel arch, h/w
>> generate MCE# to all CPUs).
>>
>> This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
>
>
> This breaks the AMD way. The AMD way is to only inject it to vcpu0.
> I suggest to add a flag argument to inject_vmce() that says whether
> to inject to all vcpus or just vcpu0.
> Then set/clear that flag from the caller side depending on whether you
> run on Intel or AMD.
>
> Christoph
>
No, it didn't breaks AMD since it only called by intel_memerr_dhandler().
Thanks,
Jinsong
>
>>
>> Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
>>
>> diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
>> --- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
>> +++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
>> @@ -340,48 +340,27 @@
>>
>> int inject_vmce(struct domain *d)
>> {
>> - int cpu = smp_processor_id();
>> + struct vcpu *v;
>>
>> - /* PV guest and HVM guest have different vMCE# injection
>> methods. */
>> - if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
>> + /* inject vMCE to all vcpus */
>> + for_each_vcpu(d, v)
>> {
>> - if ( d->is_hvm )
>> + if ( !test_and_set_bool(v->mce_pending) && +
>> ((d->is_hvm) || + guest_has_trap_callback(d,
>> v->vcpu_id, TRAP_machine_check)) ) {
>> - mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM
>> %d\n",
>> - d->domain_id);
>> - vcpu_kick(d->vcpu[0]);
>> + mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d
>> vcpu%d\n", + d->domain_id, v->vcpu_id);
>> + vcpu_kick(v);
>> }
>> else
>> {
>> - mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV
>> DOM%d\n",
>> - d->domain_id);
>> - if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
>> - {
>> - cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
>> - d->vcpu[0]->cpu_affinity);
>> - mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity,
>> old %d\n",
>> - cpu, d->vcpu[0]->processor);
>> - vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
>> - vcpu_kick(d->vcpu[0]);
>> - }
>> - else
>> - {
>> - mce_printk(MCE_VERBOSE,
>> - "MCE: Kill PV guest with No MCE
>> handler\n");
>> - domain_crash(d);
>> - }
>> + mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d
>> vcpu%d\n", + d->domain_id, v->vcpu_id);
>> + return -1;
>> }
>> }
>> - else
>> - {
>> - /* new vMCE comes while first one has not been injected yet,
>> - * in this case, inject fail. [We can't lose this vMCE for
>> - * the mce node's consistency].
>> - */
>> - mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be
>> injected "
>> - " to this DOM%d!\n", d->domain_id);
>> - return -1;
>> - }
>> +
>> return 0;
>> }
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH 2/5] Xen/MCE: vMCE injection
2012-09-20 19:15 ` Liu, Jinsong
@ 2012-09-21 7:38 ` Christoph Egger
0 siblings, 0 replies; 7+ messages in thread
From: Christoph Egger @ 2012-09-21 7:38 UTC (permalink / raw)
To: Liu, Jinsong
Cc: xen-devel@lists.xensource.com, keir@xen.org,
Ian.Campbell@citrix.com, Jan Beulich
On 09/20/12 21:15, Liu, Jinsong wrote:
> Christoph Egger wrote:
>> On 09/19/12 10:03, Liu, Jinsong wrote:
>>
>>> Xen/MCE: vMCE injection
>>>
>>> In our test for win8 guest mce, we find a bug that no matter what
>>> SRAO/SRAR error xen inject to win8 guest, it always reboot.
>>>
>>> The root cause is, current Xen vMCE logic inject vMCE# only to
>>> vcpu0, this is not correct for Intel MCE (Under Intel arch, h/w
>>> generate MCE# to all CPUs).
>>>
>>> This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
>>
>>
>> This breaks the AMD way. The AMD way is to only inject it to vcpu0.
>> I suggest to add a flag argument to inject_vmce() that says whether
>> to inject to all vcpus or just vcpu0.
>> Then set/clear that flag from the caller side depending on whether you
>> run on Intel or AMD.
>>
>> Christoph
>>
>
> No, it didn't breaks AMD since it only called by intel_memerr_dhandler().
But it will with the mce patches I still have in my queue.
Christoph
> Thanks,
> Jinsong
>
>>
>>>
>>> Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
>>>
>>> diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
>>> --- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
>>> +++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
>>> @@ -340,48 +340,27 @@
>>>
>>> int inject_vmce(struct domain *d)
>>> {
>>> - int cpu = smp_processor_id();
>>> + struct vcpu *v;
>>>
>>> - /* PV guest and HVM guest have different vMCE# injection
>>> methods. */
>>> - if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
>>> + /* inject vMCE to all vcpus */
>>> + for_each_vcpu(d, v)
>>> {
>>> - if ( d->is_hvm )
>>> + if ( !test_and_set_bool(v->mce_pending) && +
>>> ((d->is_hvm) || + guest_has_trap_callback(d,
>>> v->vcpu_id, TRAP_machine_check)) ) {
>>> - mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM
>>> %d\n",
>>> - d->domain_id);
>>> - vcpu_kick(d->vcpu[0]);
>>> + mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d
>>> vcpu%d\n", + d->domain_id, v->vcpu_id);
>>> + vcpu_kick(v);
>>> }
>>> else
>>> {
>>> - mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV
>>> DOM%d\n",
>>> - d->domain_id);
>>> - if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
>>> - {
>>> - cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
>>> - d->vcpu[0]->cpu_affinity);
>>> - mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity,
>>> old %d\n",
>>> - cpu, d->vcpu[0]->processor);
>>> - vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
>>> - vcpu_kick(d->vcpu[0]);
>>> - }
>>> - else
>>> - {
>>> - mce_printk(MCE_VERBOSE,
>>> - "MCE: Kill PV guest with No MCE
>>> handler\n");
>>> - domain_crash(d);
>>> - }
>>> + mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d
>>> vcpu%d\n", + d->domain_id, v->vcpu_id);
>>> + return -1;
>>> }
>>> }
>>> - else
>>> - {
>>> - /* new vMCE comes while first one has not been injected yet,
>>> - * in this case, inject fail. [We can't lose this vMCE for
>>> - * the mce node's consistency].
>>> - */
>>> - mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be
>>> injected "
>>> - " to this DOM%d!\n", d->domain_id);
>>> - return -1;
>>> - }
>>> +
>>> return 0;
>>> }
>
>
--
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Einsteinring 24, 85689 Dornach b. Muenchen
Geschaeftsfuehrer: Alberto Bozzo
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 2/5] Xen/MCE: vMCE injection
@ 2012-09-26 3:16 Liu, Jinsong
2012-09-26 10:10 ` Jan Beulich
0 siblings, 1 reply; 7+ messages in thread
From: Liu, Jinsong @ 2012-09-26 3:16 UTC (permalink / raw)
To: Jan Beulich, Christoph Egger, xen-devel@lists.xensource.com
Cc: keir@xen.org, Ian.Campbell@citrix.com
[-- Attachment #1: Type: text/plain, Size: 4425 bytes --]
Xen/MCE: vMCE injection
for Intel MCE, broadcast vMCE to all vcpus;
for AMD MCE, only inject vMCE to 1 vcpu, say, vcpu0
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
Suggested_by: Christoph Egger <Christoph.Egger@amd.com>
Suggested_by: Jan Beulich <jbeulich@suse.com>
diff -r 570d98e2f1cf xen/arch/x86/cpu/mcheck/mce.h
--- a/xen/arch/x86/cpu/mcheck/mce.h Wed Sep 19 23:22:57 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce.h Wed Sep 26 18:59:03 2012 +0800
@@ -168,7 +168,7 @@
int fill_vmsr_data(struct mcinfo_bank *mc_bank, struct domain *d,
uint64_t gstatus);
-int inject_vmce(struct domain *d);
+int inject_vmce(struct domain *d, int vcpuid);
static inline int mce_vendor_bank_msr(const struct vcpu *v, uint32_t msr)
{
diff -r 570d98e2f1cf xen/arch/x86/cpu/mcheck/mce_intel.c
--- a/xen/arch/x86/cpu/mcheck/mce_intel.c Wed Sep 19 23:22:57 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce_intel.c Wed Sep 26 18:59:03 2012 +0800
@@ -359,7 +359,7 @@
}
/* We will inject vMCE to DOMU*/
- if ( inject_vmce(d) < 0 )
+ if ( inject_vmce(d, -1) < 0 )
{
mce_printk(MCE_QUIET, "inject vMCE to DOM%d"
" failed\n", d->domain_id);
diff -r 570d98e2f1cf xen/arch/x86/cpu/mcheck/vmce.c
--- a/xen/arch/x86/cpu/mcheck/vmce.c Wed Sep 19 23:22:57 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/vmce.c Wed Sep 26 18:59:03 2012 +0800
@@ -338,51 +338,44 @@
HVM_REGISTER_SAVE_RESTORE(VMCE_VCPU, vmce_save_vcpu_ctxt,
vmce_load_vcpu_ctxt, 1, HVMSR_PER_VCPU);
-int inject_vmce(struct domain *d)
+/*
+ * for Intel MCE, broadcast vMCE to all vcpus
+ * for AMD MCE, only inject vMCE to 1 vcpu, say, vcpu0
+ * @ d, domain to which would inject vmce
+ * @ vcpuid,
+ * < 0, broadcast vMCE to all vcpus
+ * >= 0, vcpu who would be injected vMCE
+ * return 0 for success injection, -1 for fail injection
+ */
+int inject_vmce(struct domain *d, int vcpuid)
{
- int cpu = smp_processor_id();
+ struct vcpu *v;
+ int ret = -1;
- /* PV guest and HVM guest have different vMCE# injection methods. */
- if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
+ for_each_vcpu(d, v)
{
- if ( d->is_hvm )
+ if ( (vcpuid < 0) || (vcpuid == v->vcpu_id) )
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
- d->domain_id);
- vcpu_kick(d->vcpu[0]);
- }
- else
- {
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
- d->domain_id);
- if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
+ if ( !test_and_set_bool(v->mce_pending) &&
+ ((d->is_hvm) ||
+ guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
{
- cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
- d->vcpu[0]->cpu_affinity);
- mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
- cpu, d->vcpu[0]->processor);
- vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
- vcpu_kick(d->vcpu[0]);
+ mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ vcpu_kick(v);
+ ret = 0;
}
else
{
- mce_printk(MCE_VERBOSE,
- "MCE: Kill PV guest with No MCE handler\n");
- domain_crash(d);
+ mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ ret = -1;
+ break;
}
}
}
- else
- {
- /* new vMCE comes while first one has not been injected yet,
- * in this case, inject fail. [We can't lose this vMCE for
- * the mce node's consistency].
- */
- mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
- " to this DOM%d!\n", d->domain_id);
- return -1;
- }
- return 0;
+
+ return ret;
}
int fill_vmsr_data(struct mcinfo_bank *mc_bank, struct domain *d,
[-- Attachment #2: 2_vmce_injection.patch --]
[-- Type: application/octet-stream, Size: 4312 bytes --]
Xen/MCE: vMCE injection
for Intel MCE, broadcast vMCE to all vcpus;
for AMD MCE, only inject vMCE to 1 vcpu, say, vcpu0
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
Suggested_by: Christoph Egger <Christoph.Egger@amd.com>
Suggested_by: Jan Beulich <jbeulich@suse.com>
diff -r 570d98e2f1cf xen/arch/x86/cpu/mcheck/mce.h
--- a/xen/arch/x86/cpu/mcheck/mce.h Wed Sep 19 23:22:57 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce.h Wed Sep 26 18:59:03 2012 +0800
@@ -168,7 +168,7 @@
int fill_vmsr_data(struct mcinfo_bank *mc_bank, struct domain *d,
uint64_t gstatus);
-int inject_vmce(struct domain *d);
+int inject_vmce(struct domain *d, int vcpuid);
static inline int mce_vendor_bank_msr(const struct vcpu *v, uint32_t msr)
{
diff -r 570d98e2f1cf xen/arch/x86/cpu/mcheck/mce_intel.c
--- a/xen/arch/x86/cpu/mcheck/mce_intel.c Wed Sep 19 23:22:57 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/mce_intel.c Wed Sep 26 18:59:03 2012 +0800
@@ -359,7 +359,7 @@
}
/* We will inject vMCE to DOMU*/
- if ( inject_vmce(d) < 0 )
+ if ( inject_vmce(d, -1) < 0 )
{
mce_printk(MCE_QUIET, "inject vMCE to DOM%d"
" failed\n", d->domain_id);
diff -r 570d98e2f1cf xen/arch/x86/cpu/mcheck/vmce.c
--- a/xen/arch/x86/cpu/mcheck/vmce.c Wed Sep 19 23:22:57 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/vmce.c Wed Sep 26 18:59:03 2012 +0800
@@ -338,51 +338,44 @@
HVM_REGISTER_SAVE_RESTORE(VMCE_VCPU, vmce_save_vcpu_ctxt,
vmce_load_vcpu_ctxt, 1, HVMSR_PER_VCPU);
-int inject_vmce(struct domain *d)
+/*
+ * for Intel MCE, broadcast vMCE to all vcpus
+ * for AMD MCE, only inject vMCE to 1 vcpu, say, vcpu0
+ * @ d, domain to which would inject vmce
+ * @ vcpuid,
+ * < 0, broadcast vMCE to all vcpus
+ * >= 0, vcpu who would be injected vMCE
+ * return 0 for success injection, -1 for fail injection
+ */
+int inject_vmce(struct domain *d, int vcpuid)
{
- int cpu = smp_processor_id();
+ struct vcpu *v;
+ int ret = -1;
- /* PV guest and HVM guest have different vMCE# injection methods. */
- if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
+ for_each_vcpu(d, v)
{
- if ( d->is_hvm )
+ if ( (vcpuid < 0) || (vcpuid == v->vcpu_id) )
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
- d->domain_id);
- vcpu_kick(d->vcpu[0]);
- }
- else
- {
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
- d->domain_id);
- if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
+ if ( !test_and_set_bool(v->mce_pending) &&
+ ((d->is_hvm) ||
+ guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
{
- cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
- d->vcpu[0]->cpu_affinity);
- mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
- cpu, d->vcpu[0]->processor);
- vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
- vcpu_kick(d->vcpu[0]);
+ mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ vcpu_kick(v);
+ ret = 0;
}
else
{
- mce_printk(MCE_VERBOSE,
- "MCE: Kill PV guest with No MCE handler\n");
- domain_crash(d);
+ mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ ret = -1;
+ break;
}
}
}
- else
- {
- /* new vMCE comes while first one has not been injected yet,
- * in this case, inject fail. [We can't lose this vMCE for
- * the mce node's consistency].
- */
- mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
- " to this DOM%d!\n", d->domain_id);
- return -1;
- }
- return 0;
+
+ return ret;
}
int fill_vmsr_data(struct mcinfo_bank *mc_bank, struct domain *d,
[-- Attachment #3: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH 2/5] Xen/MCE: vMCE injection
2012-09-26 3:16 Liu, Jinsong
@ 2012-09-26 10:10 ` Jan Beulich
0 siblings, 0 replies; 7+ messages in thread
From: Jan Beulich @ 2012-09-26 10:10 UTC (permalink / raw)
To: Christoph Egger, Jinsong Liu
Cc: keir@xen.org, Ian.Campbell@citrix.com, xen-devel
>>> On 26.09.12 at 05:16, "Liu, Jinsong" <jinsong.liu@intel.com> wrote:
> Xen/MCE: vMCE injection
>
> for Intel MCE, broadcast vMCE to all vcpus;
> for AMD MCE, only inject vMCE to 1 vcpu, say, vcpu0
Please double check what got committed.
Jan
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH 2/5] Xen/MCE: vMCE injection
@ 2012-09-18 13:14 Liu, Jinsong
0 siblings, 0 replies; 7+ messages in thread
From: Liu, Jinsong @ 2012-09-18 13:14 UTC (permalink / raw)
To: Jan Beulich, xen-devel@lists.xensource.com
Cc: keir@xen.org, Ian.Campbell@citrix.com
[-- Attachment #1: Type: text/plain, Size: 2868 bytes --]
Xen/MCE: vMCE injection
In our test for win8 guest mce, we find a bug that no matter what SRAO/SRAR
error xen inject to win8 guest, it always reboot.
The root cause is, current Xen vMCE logic inject vMCE# only to vcpu0, this is
not correct for Intel MCE (Under Intel arch, h/w generate MCE# to all CPUs).
This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
--- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
@@ -340,48 +340,27 @@
int inject_vmce(struct domain *d)
{
- int cpu = smp_processor_id();
+ struct vcpu *v;
- /* PV guest and HVM guest have different vMCE# injection methods. */
- if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
+ /* inject vMCE to all vcpus */
+ for_each_vcpu(d, v)
{
- if ( d->is_hvm )
+ if ( !test_and_set_bool(v->mce_pending) &&
+ ((d->is_hvm) ||
+ guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
- d->domain_id);
- vcpu_kick(d->vcpu[0]);
+ mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ vcpu_kick(v);
}
else
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
- d->domain_id);
- if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
- {
- cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
- d->vcpu[0]->cpu_affinity);
- mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
- cpu, d->vcpu[0]->processor);
- vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
- vcpu_kick(d->vcpu[0]);
- }
- else
- {
- mce_printk(MCE_VERBOSE,
- "MCE: Kill PV guest with No MCE handler\n");
- domain_crash(d);
- }
+ mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ return -1;
}
}
- else
- {
- /* new vMCE comes while first one has not been injected yet,
- * in this case, inject fail. [We can't lose this vMCE for
- * the mce node's consistency].
- */
- mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
- " to this DOM%d!\n", d->domain_id);
- return -1;
- }
+
return 0;
}
[-- Attachment #2: 2_vmce_injection.patch --]
[-- Type: application/octet-stream, Size: 2793 bytes --]
Xen/MCE: vMCE injection
In our test for win8 guest mce, we find a bug that no matter what SRAO/SRAR
error xen inject to win8 guest, it always reboot.
The root cause is, current Xen vMCE logic inject vMCE# only to vcpu0, this is
not correct for Intel MCE (Under Intel arch, h/w generate MCE# to all CPUs).
This patch fix vMCE injection bug, injecting vMCE# to all vcpus.
Signed-off-by: Liu, Jinsong <jinsong.liu@intel.com>
diff -r 133664c6bfb4 xen/arch/x86/cpu/mcheck/vmce.c
--- a/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 22:39:11 2012 +0800
+++ b/xen/arch/x86/cpu/mcheck/vmce.c Tue Sep 18 23:46:38 2012 +0800
@@ -340,48 +340,27 @@
int inject_vmce(struct domain *d)
{
- int cpu = smp_processor_id();
+ struct vcpu *v;
- /* PV guest and HVM guest have different vMCE# injection methods. */
- if ( !test_and_set_bool(d->vcpu[0]->mce_pending) )
+ /* inject vMCE to all vcpus */
+ for_each_vcpu(d, v)
{
- if ( d->is_hvm )
+ if ( !test_and_set_bool(v->mce_pending) &&
+ ((d->is_hvm) ||
+ guest_has_trap_callback(d, v->vcpu_id, TRAP_machine_check)) )
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to HVM DOM %d\n",
- d->domain_id);
- vcpu_kick(d->vcpu[0]);
+ mce_printk(MCE_VERBOSE, "MCE: inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ vcpu_kick(v);
}
else
{
- mce_printk(MCE_VERBOSE, "MCE: inject vMCE to PV DOM%d\n",
- d->domain_id);
- if ( guest_has_trap_callback(d, 0, TRAP_machine_check) )
- {
- cpumask_copy(d->vcpu[0]->cpu_affinity_tmp,
- d->vcpu[0]->cpu_affinity);
- mce_printk(MCE_VERBOSE, "MCE: CPU%d set affinity, old %d\n",
- cpu, d->vcpu[0]->processor);
- vcpu_set_affinity(d->vcpu[0], cpumask_of(cpu));
- vcpu_kick(d->vcpu[0]);
- }
- else
- {
- mce_printk(MCE_VERBOSE,
- "MCE: Kill PV guest with No MCE handler\n");
- domain_crash(d);
- }
+ mce_printk(MCE_QUIET, "Fail to inject vMCE to dom%d vcpu%d\n",
+ d->domain_id, v->vcpu_id);
+ return -1;
}
}
- else
- {
- /* new vMCE comes while first one has not been injected yet,
- * in this case, inject fail. [We can't lose this vMCE for
- * the mce node's consistency].
- */
- mce_printk(MCE_QUIET, "There's a pending vMCE waiting to be injected "
- " to this DOM%d!\n", d->domain_id);
- return -1;
- }
+
return 0;
}
[-- Attachment #3: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2012-09-26 10:10 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-09-19 8:03 [PATCH 2/5] Xen/MCE: vMCE injection Liu, Jinsong
2012-09-20 15:23 ` Christoph Egger
2012-09-20 19:15 ` Liu, Jinsong
2012-09-21 7:38 ` Christoph Egger
-- strict thread matches above, loose matches on Subject: below --
2012-09-26 3:16 Liu, Jinsong
2012-09-26 10:10 ` Jan Beulich
2012-09-18 13:14 Liu, Jinsong
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.