Skip to contentSkip to conversation

KVM: selftests: Fixes for access tracking perf test

PATCH In review Sean Christopherson 2 participants 4 messages Nov 29, 2022 – Dec 1, 2022

Subsystems (2)
lore
Share & tools
lore

Sean Christopherson

author reviewer
#0 lore raw

cover KVM: selftests: Fixes for access tracking perf test

Fix an inverted check in the access tracking perf test, and restore the assert that there aren't too many dangling idle pages when running the test on x86-64 bare metal.

Sean Christopherson (2):
KVM: selftests: Fix inverted "warning" in access tracking perf test
KVM: selftests: Restore assert for non-nested VMs in access tracking
test

 .../selftests/kvm/access_tracking_perf_test.c | 22 ++++++++++++-------
 .../selftests/kvm/include/x86_64/processor.h  |  1 +
 2 files changed, 15 insertions(+), 8 deletions(-)

base-commit: 3e04435fe60590a1c79ec94d60e9897c3ff7d73b

Signature · 1 line
2.38.1.584.g0f3c55d4c2-goog
author reviewer
#1 lore raw

PATCH 1/2 KVM: selftests: Fix inverted "warning" in access tracking perf test

Warn if the number of idle pages is greater than or equal to 10% of the total number of pages, not if the percentage of idle pages is less than 10%. The original code asserted that less than 10% of pages were still idle, but the check got inverted when the assert was converted to a warning.

Opportunistically clean up the warning; selftests are 64-bit only, there is no need to use "%PRIu64" instead of "%lu".

Fixes 6336a810db5c ("KVM: selftests: replace assertion with warning in access_tracking_perf_test")
Cc Emanuele Giuseppe Esposito <[email protected]>
---
 tools/testing/selftests/kvm/access_tracking_perf_test.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
tools/testing/selftests/kvm/access_tracking_perf_test.c +3 -4
index 02d3587cab0a..d45ef319a68f 100644
@@ -185,10 +185,9 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,
 	 * happens, much more pages are cached there and guest won't see the
 	 * "idle" bit cleared.
 	 */
-	if (still_idle < pages / 10)
-		printf("WARNING: vCPU%d: Too many pages still idle (%" PRIu64
-		       "out of %" PRIu64 "), this will affect performance results"
-		       ".\n",
+	if (still_idle >= pages / 10)
+		printf("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), "
+		       "this will affect performance results.\n",
 		       vcpu_idx, still_idle, pages);
 
 	close(page_idle_fd);
Signature · 1 line
2.38.1.584.g0f3c55d4c2-goog
author reviewer
#2 lore raw

PATCH 2/2 KVM: selftests: Restore assert for non-nested VMs in access tracking test

In reply to Sean Christopherson

Restore the assert (on x86-64) that <10% of pages are still idle when NOT running as a nested VM in the access tracking test. The original assert was converted to a "warning" to avoid false failures when running the test in a VM, but the non-nested case does not suffer from the same "infinite TLB size" issue.

Using the HYPERVISOR flag isn't infallible as VMMs aren't strictly required to enumerate the "feature" in CPUID, but practically speaking anyone that is running KVM selftests in VMs is going to be using a VMM and hypervisor that sets the HYPERVISOR flag.

Cc David Matlack <[email protected]>
Cc Emanuele Giuseppe Esposito <[email protected]>
---
 .../selftests/kvm/access_tracking_perf_test.c   | 17 ++++++++++++-----
 .../selftests/kvm/include/x86_64/processor.h    |  1 +
 2 files changed, 13 insertions(+), 5 deletions(-)
tools/testing/selftests/kvm/access_tracking_perf_test.c +12 -5
index d45ef319a68f..9f9503e40ca5 100644
@@ -46,6 +46,7 @@
 #include "test_util.h"
 #include "memstress.h"
 #include "guest_modes.h"
+#include "processor.h"
 
 /* Global variable used to synchronize all of the vCPU threads. */
 static int iteration;
@@ -180,15 +181,21 @@ static void mark_vcpu_memory_idle(struct kvm_vm *vm,
 	 * access tracking but low enough as to not make the test too brittle
 	 * over time and across architectures.
 	 *
-	 * Note that when run in nested virtualization, this check will trigger
-	 * much more frequently because TLB size is unlimited and since no flush
-	 * happens, much more pages are cached there and guest won't see the
-	 * "idle" bit cleared.
+	 * When running the guest as a nested VM, "warn" instead of asserting
+	 * as the TLB size is effectively unlimited and the KVM doesn't
+	 * explicitly flush the TLB when aging SPTEs.  As a result, more pages
+	 * are cached and the guest won't see the "idle" bit cleared.
 	 */
-	if (still_idle >= pages / 10)
+	if (still_idle >= pages / 10) {
+#ifdef __x86_64__
+		TEST_ASSERT(this_cpu_has(X86_FEATURE_HYPERVISOR),
+			    "vCPU%d: Too many pages still idle (%lu out of %lu)",
+			    vcpu_idx, still_idle, pages);
+#endif
 		printf("WARNING: vCPU%d: Too many pages still idle (%lu out of %lu), "
 		       "this will affect performance results.\n",
 		       vcpu_idx, still_idle, pages);
+	}
 
 	close(page_idle_fd);
 	close(pagemap_fd);
tools/testing/selftests/kvm/include/x86_64/processor.h +1 -0
index 5d310abe6c3f..22852bd32d7b 100644
@@ -94,6 +94,7 @@ struct kvm_x86_cpu_feature {
 #define	X86_FEATURE_XSAVE		KVM_X86_CPU_FEATURE(0x1, 0, ECX, 26)
 #define	X86_FEATURE_OSXSAVE		KVM_X86_CPU_FEATURE(0x1, 0, ECX, 27)
 #define	X86_FEATURE_RDRAND		KVM_X86_CPU_FEATURE(0x1, 0, ECX, 30)
+#define	X86_FEATURE_HYPERVISOR		KVM_X86_CPU_FEATURE(0x1, 0, ECX, 31)
 #define X86_FEATURE_PAE			KVM_X86_CPU_FEATURE(0x1, 0, EDX, 6)
 #define	X86_FEATURE_MCE			KVM_X86_CPU_FEATURE(0x1, 0, EDX, 7)
 #define	X86_FEATURE_APIC		KVM_X86_CPU_FEATURE(0x1, 0, EDX, 9)
Signature · 1 line
2.38.1.584.g0f3c55d4c2-goog
Reviewed-by on the v1 series
#3 lore raw
In reply to Sean Christopherson

Am 29/11/2022 um 18:52 schrieb Sean Christopherson:

Quoted 9 lines

Fix an inverted check in the access tracking perf test, and restore the assert that there aren't too many dangling idle pages when running the test on x86-64 bare metal.

Sean Christopherson (2):
KVM: selftests: Fix inverted "warning" in access tracking perf test
KVM: selftests: Restore assert for non-nested VMs in access tracking
test

Quoted 7 lines
 .../selftests/kvm/access_tracking_perf_test.c | 22 ++++++++++++-------
 .../selftests/kvm/include/x86_64/processor.h  |  1 +
 2 files changed, 15 insertions(+), 8 deletions(-)

base-commit: 3e04435fe60590a1c79ec94d60e9897c3ff7d73b

Makes sense, apologies for inverting the check.