On 12/16/16 14:40 +0000, Andrew Cooper wrote:
On 16/12/16 13:43, Haozhong Zhang wrote:
diff --git a/tests/vvmx/cpuid.c b/tests/vvmx/cpuid.c
new file mode 100644
index 0000000..9a4cdae
--- /dev/null
+++ b/tests/vvmx/cpuid.c
@@ -0,0 +1,24 @@
+#include <xtf.h>
+
+bool test_cpuid_vmx_feat(void)
+{
+ uint32_t ecx = cpuid_ecx(1);
+
+ if ( !(ecx & X86_FEATURE_VMX) )
+ {
+ xtf_failure("Fail: cpuid.1:ecx[5] is not set.\n");
+ return false;
+ }
Some cpuid information is cached at boot.
It turns out that there is already a suitable cpu_has_vmx define.
Ah yes, it's what I want.
+#include <xtf.h>
+
+const char test_title[] = "Test vvmx";
+
+extern bool test_cpuid_vmx_feat(void);
+
+void test_main(void)
+{
+ if ( !vendor_is(X86_VENDOR_INTEL) )
There is a slightly shorter vendor_is_intel which you can use.
ditto
+ {
+ xtf_skip("Skip: non-Intel processors\n");
"processor"
will change
+ return;
Where it makes the code easier to read, I tend to use return
xtf_skip("Skip: non-Intel processor\n"), which in this case allows the
braces to be dropped. However, I am not overly fussed if you prefer
this style.
Your suggestion is more clear and I'll change.
+ }
+
+ if ( !test_cpuid_vmx_feat() )
+ goto fail;
Are you intending to do converse tests? We have had security issues in
the past where some of the nested-virt code in Xen was reachable from a
guest even through the feature was intended to be fully disabled.
I should follow the pseudo code in Intel SDM more closely, as the
pseudo code has a branch saying if vmxon is executed w/o VMX support,
there will be some errors. I'll add test cases for such circumstance.
Haozhong
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel
