KVM: fix XSAVE bit scanning (now properly)
authorAndre Przywara <andre.przywara@amd.com>
Fri, 10 Jun 2011 09:35:30 +0000 (11:35 +0200)
committerAvi Kivity <avi@redhat.com>
Tue, 12 Jul 2011 10:16:20 +0000 (13:16 +0300)
commit 123108f1c1aafd51d6a5c79cc04d7999dd88a930 tried to fix KVMs
XSAVE valid feature scanning, but it was wrong. It was not considering
the sparse nature of this bitfield, instead reading values from
uninitialized members of the entries array.
This patch now separates subleaf indicies from KVM's array indicies
and fills the entry before querying it's value.
This fixes AVX support in KVM guests.

Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
arch/x86/kvm/x86.c

index ff4623b1b10242da251e409f9cbb8813b8828326..84f46074ca7485485a33cd8062bd29d57d7abf6d 100644 (file)
@@ -2447,16 +2447,17 @@ static void do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
                break;
        }
        case 0xd: {
-               int i;
+               int idx, i;
 
                entry->flags |= KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
-               for (i = 1; *nent < maxnent && i < 64; ++i) {
-                       if (entry[i].eax == 0 || !supported_xcr0_bit(i))
+               for (idx = 1, i = 1; *nent < maxnent && idx < 64; ++idx) {
+                       do_cpuid_1_ent(&entry[i], function, idx);
+                       if (entry[i].eax == 0 || !supported_xcr0_bit(idx))
                                continue;
-                       do_cpuid_1_ent(&entry[i], function, i);
                        entry[i].flags |=
                               KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
                        ++*nent;
+                       ++i;
                }
                break;
        }