[PATCH 0/2] expose hypervisor information on cpuid

[PATCH 0/2] expose hypervisor information on cpuid

[Glauber Costa]

This discussion happened some days ago in kvm list.
We believe it's useful to advertise hypervisor information
somewhere. /proc/cpuinfo appears as one good candidate.

This patch shows this information for vmware, kvm, and the
string "none", otherwise. (someone would have to do it for xen,
and possibly others, if this patch is to be accepted).

[PATCH 1/2] show hypervisor information on cpuinfo

[Glauber Costa]

From: Glauber Costa <gcosta@redhat.com>

It is useful to easily grab information about whether or not
we're running on top of a hypervisor. And in case affirmative,
which one.

This patch shows it in a newly added "hypervisor" field to cpuinfo.
This seems to me like the best suited place for this.

We currently differentiate between vmware and none.

Signed-off-by: Glauber Costa <gcosta@redhat.com>
---
 arch/x86/include/asm/hypervisor.h |    1 +
 arch/x86/kernel/cpu/hypervisor.c  |    8 ++++++++
 arch/x86/kernel/cpu/proc.c        |    8 ++++++--
 3 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/arch/x86/include/asm/hypervisor.h b/arch/x86/include/asm/hypervisor.h
index 369f5c5..a0c0b0f 100644
--- a/arch/x86/include/asm/hypervisor.h
+++ b/arch/x86/include/asm/hypervisor.h
@@ -22,5 +22,6 @@
 
 extern unsigned long get_hypervisor_tsc_freq(void);
 extern void init_hypervisor(struct cpuinfo_x86 *c);
+extern char *hypervisor_str(struct cpuinfo_x86 *c);
 
 #endif
diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c
index fb5b86a..8c3fca7 100644
--- a/arch/x86/kernel/cpu/hypervisor.c
+++ b/arch/x86/kernel/cpu/hypervisor.c
@@ -51,6 +51,14 @@ hypervisor_set_feature_bits(struct cpuinfo_x86 *c)
 	}
 }
 
+char * __cpuinit hypervisor_str(struct cpuinfo_x86 *c)
+{
+	if (c->x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
+		return "VMWare";
+	else
+		return "none";
+}
+
 void __cpuinit init_hypervisor(struct cpuinfo_x86 *c)
 {
 	detect_hypervisor_vendor(c);
diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
index 01b1244..3700607 100644
--- a/arch/x86/kernel/cpu/proc.c
+++ b/arch/x86/kernel/cpu/proc.c
@@ -4,6 +4,8 @@
 #include <linux/seq_file.h>
 #include <linux/cpufreq.h>
 
+#include <asm/hypervisor.h>
+
 /*
  *	Get CPU information for use by the procfs.
  */
@@ -90,12 +92,14 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 		   "vendor_id\t: %s\n"
 		   "cpu family\t: %d\n"
 		   "model\t\t: %u\n"
-		   "model name\t: %s\n",
+		   "model name\t: %s\n"
+		   "hypervisor\t: %s\n",
 		   cpu,
 		   c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown",
 		   c->x86,
 		   c->x86_model,
-		   c->x86_model_id[0] ? c->x86_model_id : "unknown");
+		   c->x86_model_id[0] ? c->x86_model_id : "unknown",
+		   hypervisor_str(c));
 
 	if (c->x86_mask || c->cpuid_level >= 0)
 		seq_printf(m, "stepping\t: %d\n", c->x86_mask);
-- 
1.5.6.5

[PATCH 2/2] tell cpuinfo if we're running on top of KVM

[Glauber Costa]

From: Glauber Costa <gcosta@redhat.com>

KVM has a specific cpuid signature, for a long time now. It's currently
used in the kernel to advertise the possible availability of paravirt
functions, but it's safe to assume that any reasonably recent kvm
hypervisor will sign cpuid this way, regardless of any pv capability.

Use this information to fill in the hypervisor field in cpuinfo.

Signed-off-by: Glauber Costa <gcosta@redhat.com>
---
 arch/x86/include/asm/processor.h |    1 +
 arch/x86/kernel/cpu/hypervisor.c |    5 +++++
 2 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 091cd88..919c08d 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -126,6 +126,7 @@ struct cpuinfo_x86 {
 
 #define X86_HYPER_VENDOR_NONE  0
 #define X86_HYPER_VENDOR_VMWARE 1
+#define X86_HYPER_VENDOR_KVM	2
 
 /*
  * capabilities of CPUs
diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c
index 8c3fca7..251817a 100644
--- a/arch/x86/kernel/cpu/hypervisor.c
+++ b/arch/x86/kernel/cpu/hypervisor.c
@@ -23,6 +23,7 @@
 
 #include <asm/processor.h>
 #include <asm/vmware.h>
+#include <asm/kvm_para.h>
 #include <asm/hypervisor.h>
 
 static inline void __cpuinit
@@ -30,6 +31,8 @@ detect_hypervisor_vendor(struct cpuinfo_x86 *c)
 {
 	if (vmware_platform()) {
 		c->x86_hyper_vendor = X86_HYPER_VENDOR_VMWARE;
+	} else if (kvm_para_available()) {
+		c->x86_hyper_vendor = X86_HYPER_VENDOR_KVM;
 	} else {
 		c->x86_hyper_vendor = X86_HYPER_VENDOR_NONE;
 	}
@@ -55,6 +58,8 @@ char * __cpuinit hypervisor_str(struct cpuinfo_x86 *c)
 {
 	if (c->x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
 		return "VMWare";
+	else if (c->x86_hyper_vendor == X86_HYPER_VENDOR_KVM)
+		return "KVM";
 	else
 		return "none";
 }
-- 
1.5.6.5

Re: [PATCH 1/2] show hypervisor information on cpuinfo

[Alok Kataria]

I am all for it.

Just one nit, below.

>  
> +char * __cpuinit hypervisor_str(struct cpuinfo_x86 *c)
> +{
> +	if (c->x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
> +		return "VMWare";

Better to have "VMware" string instead.

Thanks,
Alok

Re: [PATCH 1/2] show hypervisor information on cpuinfo

[Ingo Molnar]

* Glauber Costa <glommer@redhat.com> wrote:

> +char * __cpuinit hypervisor_str(struct cpuinfo_x86 *c)
> +{
> +	if (c->x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
> +		return "VMWare";
> +	else
> +		return "none";
> +}

i'd suggest these variants instead:

 virtualization:	native kernel
 virtualization:	KVM guest
 virtualization:	VMWare guest
 virtualization:	Linux guest

	Ingo

Re: [PATCH 1/2] show hypervisor information on cpuinfo

[Chris Wright]

* Ingo Molnar (mingo@elte.hu) wrote:
>  virtualization:	Linux guest

That's confusing.  Meaning lguest?  These are all Linux guests ;-)

Re: [PATCH 1/2] show hypervisor information on cpuinfo

[Anthony Liguori]

Ingo Molnar wrote:
> * Glauber Costa <glommer@redhat.com> wrote:
>
>   
>> +char * __cpuinit hypervisor_str(struct cpuinfo_x86 *c)
>> +{
>> +	if (c->x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
>> +		return "VMWare";
>> +	else
>> +		return "none";
>> +}
>>     
>
> i'd suggest these variants instead:
>
>  virtualization:	native kernel
>   

I don't think "native kernel" really ports well to other architectures 
(like s390 or PPC) that always have hypervisors present in some form.  I 
think "none" makes more sense in the case of bare metal x86.

Regards,

Anthony Liguori

>  virtualization:	KVM guest
>  virtualization:	VMWare guest
>  virtualization:	Linux guest
>
> 	Ingo
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>   

Re: [PATCH 0/2] expose hypervisor information on cpuid

[Chris Wright]

* Glauber Costa (glommer@redhat.com) wrote:
> This discussion happened some days ago in kvm list.
> We believe it's useful to advertise hypervisor information
> somewhere. /proc/cpuinfo appears as one good candidate.

What's wrong with /sys/hypervisor?  That's what it's there for.

thanks,
-chris

Re: [PATCH 0/2] expose hypervisor information on cpuid

[Glauber Costa]

On Thu, Feb 05, 2009 at 10:52:07AM -0800, Chris Wright wrote:
> * Glauber Costa (glommer@redhat.com) wrote:
> > This discussion happened some days ago in kvm list.
> > We believe it's useful to advertise hypervisor information
> > somewhere. /proc/cpuinfo appears as one good candidate.
> 
> What's wrong with /sys/hypervisor?  That's what it's there for.
> 
I believe cpuinfo provides a more standard interface for that.
But I'm not opposed to your suggestion.