[Qemu-devel] [PATCH] test for lm capable kernel.

[Qemu-devel] [PATCH] test for lm capable kernel.

[Glauber Costa]

Don't save x86_64-specific msrs if our kernel does not support them.
Code is already in qemu-kvm.git, but the test function is slightly different.
It follows what the test for msr_start does.

Signed-off-by: Glauber Costa <glommer@redhat.com>
---
 target-i386/kvm.c |   37 +++++++++++++++++++++++++++----------
 1 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/target-i386/kvm.c b/target-i386/kvm.c
index 7010999..0e69b57 100644
--- a/target-i386/kvm.c
+++ b/target-i386/kvm.c
@@ -15,6 +15,7 @@
 #include <sys/types.h>
 #include <sys/ioctl.h>
 #include <sys/mman.h>
+#include <sys/utsname.h>
 
 #include <linux/kvm.h>
 
@@ -266,6 +267,20 @@ static int kvm_has_msr_star(CPUState *env)
     return 0;
 }
 
+static int lm_capable_kernel(void)
+{
+    static int lm_capable = -1;
+    struct utsname utsname;
+
+    if (lm_capable != -1) {
+        return lm_capable;
+    }
+
+    uname(&utsname);
+    lm_capable = (strcmp(utsname.machine, "x86_64") == 0);
+    return lm_capable;
+}
+
 int kvm_arch_init(KVMState *s, int smp_cpus)
 {
     int ret;
@@ -478,11 +493,12 @@ static int kvm_put_msrs(CPUState *env)
 	kvm_msr_entry_set(&msrs[n++], MSR_STAR, env->star);
     kvm_msr_entry_set(&msrs[n++], MSR_IA32_TSC, env->tsc);
 #ifdef TARGET_X86_64
-    /* FIXME if lm capable */
-    kvm_msr_entry_set(&msrs[n++], MSR_CSTAR, env->cstar);
-    kvm_msr_entry_set(&msrs[n++], MSR_KERNELGSBASE, env->kernelgsbase);
-    kvm_msr_entry_set(&msrs[n++], MSR_FMASK, env->fmask);
-    kvm_msr_entry_set(&msrs[n++], MSR_LSTAR, env->lstar);
+    if (lm_capable_kernel()) {
+        kvm_msr_entry_set(&msrs[n++], MSR_CSTAR, env->cstar);
+        kvm_msr_entry_set(&msrs[n++], MSR_KERNELGSBASE, env->kernelgsbase);
+        kvm_msr_entry_set(&msrs[n++], MSR_FMASK, env->fmask);
+        kvm_msr_entry_set(&msrs[n++], MSR_LSTAR, env->lstar);
+    }
 #endif
     msr_data.info.nmsrs = n;
 
@@ -611,11 +627,12 @@ static int kvm_get_msrs(CPUState *env)
 	msrs[n++].index = MSR_STAR;
     msrs[n++].index = MSR_IA32_TSC;
 #ifdef TARGET_X86_64
-    /* FIXME lm_capable_kernel */
-    msrs[n++].index = MSR_CSTAR;
-    msrs[n++].index = MSR_KERNELGSBASE;
-    msrs[n++].index = MSR_FMASK;
-    msrs[n++].index = MSR_LSTAR;
+    if (lm_capable_kernel()) {
+        msrs[n++].index = MSR_CSTAR;
+        msrs[n++].index = MSR_KERNELGSBASE;
+        msrs[n++].index = MSR_FMASK;
+        msrs[n++].index = MSR_LSTAR;
+    }
 #endif
     msr_data.info.nmsrs = n;
     ret = kvm_vcpu_ioctl(env, KVM_GET_MSRS, &msr_data);
-- 
1.6.2.5

Re: [Qemu-devel] [PATCH] test for lm capable kernel.

[Anthony Liguori]

Glauber Costa wrote:
> Don't save x86_64-specific msrs if our kernel does not support them.
> Code is already in qemu-kvm.git, but the test function is slightly different.
> It follows what the test for msr_start does.
>   

Repeat after me: "I will not use --target-list when submitting qemu 
patches.  I will not use --target-list when submitting qemu patches."

Fedora enables ccache by default.  After one build, all future builds 
are extremely fast in most cases.  It's really not that bad :-)

> Signed-off-by: Glauber Costa <glommer@redhat.com>
> ---
>  target-i386/kvm.c |   37 +++++++++++++++++++++++++++----------
>  1 files changed, 27 insertions(+), 10 deletions(-)
>
> diff --git a/target-i386/kvm.c b/target-i386/kvm.c
> index 7010999..0e69b57 100644
> --- a/target-i386/kvm.c
> +++ b/target-i386/kvm.c
> @@ -15,6 +15,7 @@
>  #include <sys/types.h>
>  #include <sys/ioctl.h>
>  #include <sys/mman.h>
> +#include <sys/utsname.h>
>  
>  #include <linux/kvm.h>
>  
> @@ -266,6 +267,20 @@ static int kvm_has_msr_star(CPUState *env)
>      return 0;
>  }
>  
> +static int lm_capable_kernel(void)
> +{
> +    static int lm_capable = -1;
> +    struct utsname utsname;
> +
> +    if (lm_capable != -1) {
> +        return lm_capable;
> +    }
> +
> +    uname(&utsname);
> +    lm_capable = (strcmp(utsname.machine, "x86_64") == 0);
> +    return lm_capable;
> +}
> +
>   

This function is unused for qemu-system-i386.

Regards,

Anthony Liguori

Re: [Qemu-devel] [PATCH] test for lm capable kernel.

[Avi Kivity]

On 10/14/2009 07:09 PM, Glauber Costa wrote:
> Don't save x86_64-specific msrs if our kernel does not support them.
> Code is already in qemu-kvm.git, but the test function is slightly different.
> It follows what the test for msr_start does.
>    

Can't we use KVM_GET_MSR_LIST to see which msrs are supported instead?

-- 
error compiling committee.c: too many arguments to function

Re: [Qemu-devel] [PATCH] test for lm capable kernel.

[Glauber Costa]

On Thu, Oct 22, 2009 at 06:51:46PM +0200, Avi Kivity wrote:
> On 10/14/2009 07:09 PM, Glauber Costa wrote:
>> Don't save x86_64-specific msrs if our kernel does not support them.
>> Code is already in qemu-kvm.git, but the test function is slightly different.
>> It follows what the test for msr_start does.
>>    
>
> Can't we use KVM_GET_MSR_LIST to see which msrs are supported instead?
that would be good, indeed.

Right now we're doing it for star, but we could keep the list and always
query it.