@@ -783,6 +783,13 @@ int xc_cpuid_apply_policy(xc_interface *
}
}
+ /*
+ * Do not try to shrink the policy if restoring, as that could cause
+ * guest visible changes in the maximum leaf fields.
+ */
+ if ( !restore )
+ x86_cpu_policy_shrink_max_leaves(&p->policy);
+
nr_leaves = ARRAY_SIZE(p->leaves);
rc = x86_cpuid_copy_to_buffer(&p->policy, p->leaves, &nr_leaves);
if ( rc )
@@ -8,10 +8,13 @@
#include <err.h>
#include <xen-tools/common-macros.h>
+#include <xen/asm/x86-defns.h>
#include <xen/asm/x86-vendors.h>
#include <xen/lib/x86/cpu-policy.h>
#include <xen/domctl.h>
+#define XSTATE_FP_SSE (X86_XCR0_X87 | X86_XCR0_SSE)
+
static unsigned int nr_failures;
#define fail(fmt, ...) \
({ \
@@ -576,6 +579,103 @@ static void test_cpuid_out_of_range_clea
}
}
+static void test_cpuid_maximum_leaf_shrinking(void)
+{
+ static const struct test {
+ const char *name;
+ struct cpu_policy p;
+ } tests[] = {
+ {
+ .name = "basic",
+ .p = {
+ /* Very basic information only. */
+ .basic.max_leaf = 1,
+ .basic.raw_fms = 0xc2,
+ },
+ },
+ {
+ .name = "cache",
+ .p = {
+ /* Cache subleaves present. */
+ .basic.max_leaf = 4,
+ .cache.subleaf[0].type = 1,
+ },
+ },
+ {
+ .name = "feat#0",
+ .p = {
+ /* Subleaf 0 only with some valid bit. */
+ .basic.max_leaf = 7,
+ .feat.max_subleaf = 0,
+ .feat.fsgsbase = 1,
+ },
+ },
+ {
+ .name = "feat#1",
+ .p = {
+ /* Subleaf 1 only with some valid bit. */
+ .basic.max_leaf = 7,
+ .feat.max_subleaf = 1,
+ .feat.avx_vnni = 1,
+ },
+ },
+ {
+ .name = "topo",
+ .p = {
+ /* Topology subleaves present. */
+ .basic.max_leaf = 0xb,
+ .topo.subleaf[0].type = 1,
+ },
+ },
+ {
+ .name = "xstate",
+ .p = {
+ /* First subleaf always valid (and then non-zero). */
+ .basic.max_leaf = 0xd,
+ .xstate.xcr0_low = XSTATE_FP_SSE,
+ },
+ },
+ {
+ .name = "extd",
+ .p = {
+ /* Commonly available information only. */
+ .extd.max_leaf = 0x80000008,
+ .extd.maxphysaddr = 0x28,
+ .extd.maxlinaddr = 0x30,
+ },
+ },
+ };
+
+ printf("Testing CPUID maximum leaf shrinking:\n");
+
+ for ( size_t i = 0; i < ARRAY_SIZE(tests); ++i )
+ {
+ const struct test *t = &tests[i];
+ struct cpu_policy *p = memdup(&t->p);
+
+ p->basic.max_leaf = ARRAY_SIZE(p->basic.raw) - 1;
+ p->feat.max_subleaf = ARRAY_SIZE(p->feat.raw) - 1;
+ p->extd.max_leaf = 0x80000000 | (ARRAY_SIZE(p->extd.raw) - 1);
+
+ x86_cpu_policy_shrink_max_leaves(p);
+
+ /* Check the the resulting max (sub)leaf values against expecations. */
+ if ( p->basic.max_leaf != t->p.basic.max_leaf )
+ fail(" Test %s basic fail - expected %#x, got %#x\n",
+ t->name, t->p.basic.max_leaf, p->basic.max_leaf);
+
+ if ( p->extd.max_leaf != t->p.extd.max_leaf )
+ fail(" Test %s extd fail - expected %#x, got %#x\n",
+ t->name, t->p.extd.max_leaf, p->extd.max_leaf);
+
+ if ( p->feat.max_subleaf != t->p.feat.max_subleaf )
+ fail(" Test %s feat fail - expected %#x, got %#x\n",
+ t->name, t->p.feat.max_subleaf, p->feat.max_subleaf);
+
+ free(p);
+ }
+}
+
static void test_is_compatible_success(void)
{
static struct test {
@@ -671,6 +771,7 @@ int main(int argc, char **argv)
test_cpuid_serialise_success();
test_cpuid_deserialise_failure();
test_cpuid_out_of_range_clearing();
+ test_cpuid_maximum_leaf_shrinking();
test_msr_serialise_success();
test_msr_deserialise_failure();
@@ -383,6 +383,8 @@ static void __init calculate_host_policy
if ( vpmu_mode == XENPMU_MODE_OFF )
p->basic.raw[0xa] = EMPTY_LEAF;
+ x86_cpu_policy_shrink_max_leaves(p);
+
/* 0x000000ce MSR_INTEL_PLATFORM_INFO */
/* probe_cpuid_faulting() sanity checks presence of MISC_FEATURES_ENABLES */
p->platform_info.cpuid_faulting = cpu_has_cpuid_faulting;
@@ -576,12 +578,6 @@ static void __init calculate_pv_max_poli
*p = host_cpu_policy;
- /*
- * Some VMs may have a larger-than-necessary feat max_subleaf. Allow them
- * to migrate in.
- */
- p->feat.max_subleaf = ARRAY_SIZE(p->feat.raw) - 1;
-
x86_cpu_policy_to_featureset(p, fs);
for ( i = 0; i < ARRAY_SIZE(fs); ++i )
@@ -613,6 +609,14 @@ static void __init calculate_pv_max_poli
recalculate_xstate(p);
p->extd.raw[0xa] = EMPTY_LEAF; /* No SVM for PV guests. */
+
+ x86_cpu_policy_shrink_max_leaves(p);
+
+ /*
+ * Some VMs may have a larger-than-necessary feat max_subleaf. Allow them
+ * to migrate in.
+ */
+ p->feat.max_subleaf = ARRAY_SIZE(p->feat.raw) - 1;
}
static void __init calculate_pv_def_policy(void)
@@ -663,12 +667,6 @@ static void __init calculate_hvm_max_pol
*p = host_cpu_policy;
- /*
- * Some VMs may have a larger-than-necessary feat max_subleaf. Allow them
- * to migrate in.
- */
- p->feat.max_subleaf = ARRAY_SIZE(p->feat.raw) - 1;
-
x86_cpu_policy_to_featureset(p, fs);
mask = hvm_hap_supported() ?
@@ -774,6 +772,14 @@ static void __init calculate_hvm_max_pol
x86_cpu_featureset_to_policy(fs, p);
recalculate_xstate(p);
+ x86_cpu_policy_shrink_max_leaves(p);
+
+ /*
+ * Some VMs may have a larger-than-necessary feat max_subleaf. Allow them
+ * to migrate in.
+ */
+ p->feat.max_subleaf = ARRAY_SIZE(p->feat.raw) - 1;
+
/* It's always possible to emulate CPUID faulting for HVM guests */
p->platform_info.cpuid_faulting = true;
}
@@ -1060,13 +1060,15 @@ void cpuid_hypervisor_leaves(const struc
uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
uint32_t idx = leaf - base;
unsigned int limit = is_viridian_domain(d) ? p->hv2_limit : p->hv_limit;
+ unsigned int dflt = is_pv_domain(d) ? XEN_CPUID_MAX_PV_NUM_LEAVES
+ : XEN_CPUID_MAX_HVM_NUM_LEAVES;
if ( limit == 0 )
/* Default number of leaves */
- limit = XEN_CPUID_MAX_NUM_LEAVES;
+ limit = dflt;
else
/* Clamp toolstack value between 2 and MAX_NUM_LEAVES. */
- limit = min(max(limit, 2u), XEN_CPUID_MAX_NUM_LEAVES + 0u);
+ limit = min(max(limit, 2u), dflt);
if ( idx > limit )
return;
@@ -117,6 +117,10 @@
/* Max. address width in bits taking memory hotplug into account. */
#define XEN_CPUID_MACHINE_ADDRESS_WIDTH_MASK (0xffu << 0)
-#define XEN_CPUID_MAX_NUM_LEAVES 5
+#define XEN_CPUID_MAX_PV_NUM_LEAVES 5
+#define XEN_CPUID_MAX_HVM_NUM_LEAVES 4
+#define XEN_CPUID_MAX_NUM_LEAVES \
+ (XEN_CPUID_MAX_PV_NUM_LEAVES > XEN_CPUID_MAX_HVM_NUM_LEAVES ? \
+ XEN_CPUID_MAX_PV_NUM_LEAVES : XEN_CPUID_MAX_HVM_NUM_LEAVES)
#endif /* __XEN_PUBLIC_ARCH_X86_CPUID_H__ */
@@ -438,6 +438,13 @@ void x86_cpu_policy_fill_native(struct c
*/
void x86_cpu_policy_clear_out_of_range_leaves(struct cpu_policy *p);
+/**
+ * Shrink max leaf/subleaf values such that the last respective valid entry
+ * isn't all blank. While permitted by the spec, such extraneous leaves may
+ * provide undue "hints" to guests.
+ */
+void x86_cpu_policy_shrink_max_leaves(struct cpu_policy *p);
+
#ifdef __XEN__
#include <public/xen.h>
typedef XEN_GUEST_HANDLE_64(xen_cpuid_leaf_t) cpuid_leaf_buffer_t;
@@ -291,6 +291,45 @@ void x86_cpu_policy_clear_out_of_range_l
ARRAY_SIZE(p->extd.raw) - 1);
}
+void x86_cpu_policy_shrink_max_leaves(struct cpu_policy *p)
+{
+ unsigned int i;
+
+ p->basic.raw[0x4] = p->cache.raw[0];
+
+ for ( i = p->feat.max_subleaf; i; --i )
+ if ( p->feat.raw[i].a | p->feat.raw[i].b |
+ p->feat.raw[i].c | p->feat.raw[i].d )
+ break;
+ p->feat.max_subleaf = i;
+ p->basic.raw[0x7] = p->feat.raw[i];
+
+ p->basic.raw[0xb] = p->topo.raw[0];
+
+ /*
+ * Due to the way xstate gets handled in the hypervisor (see
+ * recalculate_xstate()) there is (for now at least) no need to fiddle
+ * with the xstate subleaves (IOW we assume they're already in consistent
+ * shape, for coming from either hardware or recalculate_xstate()).
+ */
+ p->basic.raw[0xd] = p->xstate.raw[0];
+
+ for ( i = p->basic.max_leaf; i; --i )
+ if ( p->basic.raw[i].a | p->basic.raw[i].b |
+ p->basic.raw[i].c | p->basic.raw[i].d )
+ break;
+ p->basic.max_leaf = i;
+
+ for ( i = p->extd.max_leaf & 0xffff; i; --i )
+ if ( p->extd.raw[i].a | p->extd.raw[i].b |
+ p->extd.raw[i].c | p->extd.raw[i].d )
+ break;
+ if ( i | p->extd.raw[0].b | p->extd.raw[0].c | p->extd.raw[0].d )
+ p->extd.max_leaf = 0x80000000U | i;
+ else
+ p->extd.max_leaf = 0;
+}
+
const uint32_t *x86_cpu_policy_lookup_deep_deps(uint32_t feature)
{
static const uint32_t deep_features[] = INIT_DEEP_FEATURES;
Zapping leaf data for out of range leaves is just one half of it: To avoid guests (bogusly or worse) inferring information from mere leaf presence, also shrink maximum indicators such that the respective trailing entry is not all blank (unless of course it's the initial subleaf of a leaf that's not the final one). This is also in preparation of bumping the maximum basic leaf we support, to ensure guests not getting exposed related features won't observe a change in behavior. Signed-off-by: Jan Beulich <jbeulich@suse.com> --- TBD: The use in xc_cpuid_apply_policy() supposedly needs to further respect user (or migration source) max (sub)leaf values (i.e. the function needs calling unconditionally and then higher inputs need putting back); question is where those original values would be recorded / come from. (Supposedly there was a patch from Roger under similar or even the same title, which should have eliminated the need for the patch here, but upon searching I can't find such a patch.) --- v5: Re-base. v4: Re-instate with default policy shrinking dropped, and with 81da2b544cbb folded in. Re-base. v3: Record the actual non-empty subleaf in p->basic.raw[0x7], rather than subleaf 0. Re-base over Viridian leaf 40000005 addition. v2: New.