[PATCH] KVM: selftests: Randomize dirty_log_test's delay before reaping the bitmap

[PATCH] KVM: selftests: Randomize dirty_log_test's delay before reaping the bitmap

[Sean Christopherson]

In the dirty log test, randomize the delay before the initial call to get
the dirty log bitmap for a given iteration, so that the amount of memory
dirtied by the guest varies from iteration to iteration, and so that the
user can effectively control the duration (by increasing the interval).

Always waiting 1ms effectively hides a KVM RISC-V bug as the test reaps the
dirty bitmap before the guest has a chance to trigger the problematic flow
in KVM.

Reported-by: Wu Fei <wu.fei9@sanechips.com.cn>
Closes: https://lore.kernel.org/all/202605111130.64BBUXDN013040@mse-fl2.zte.com.cn
Cc: Wu Fei <atwufei@163.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 tools/testing/selftests/kvm/dirty_log_test.c | 26 +++++++++++++++-----
 1 file changed, 20 insertions(+), 6 deletions(-)

diff --git a/tools/testing/selftests/kvm/dirty_log_test.c b/tools/testing/selftests/kvm/dirty_log_test.c
index 12446a4b6e8d..74ca096bf976 100644
--- a/tools/testing/selftests/kvm/dirty_log_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_test.c
@@ -694,7 +694,17 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 	pthread_create(&vcpu_thread, NULL, vcpu_worker, vcpu);
 
 	for (iteration = 1; iteration <= p->iterations; iteration++) {
-		unsigned long i;
+		unsigned long i, reap_i;
+
+		/*
+		 * Select a random point in the time interval to reap the dirty
+		 * bitmap/ring while the guest is running, i.e. randomize how
+		 * long the guest gets to initially run and thus how many pages
+		 * it can dirty, before collecting the dirty bitmap/ring.  See
+		 * the loop below for details.
+		 */
+		reap_i = random() % p->interval;
+		printf("Reaping after a %lu ms delay\n", reap_i);
 
 		sync_global_to_guest(vm, iteration);
 
@@ -729,13 +739,17 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 			 * that's effectively blocked.  Collecting while the
 			 * guest is running also verifies KVM doesn't lose any
 			 * state.
-			 *
+			 */
+			if (i < reap_i)
+				continue;
+
+			/*
 			 * For bitmap modes, KVM overwrites the entire bitmap,
 			 * i.e. collecting the bitmaps is destructive.  Collect
-			 * the bitmap only on the first pass, otherwise this
-			 * test would lose track of dirty pages.
+			 * the bitmap while the guest is running only once,
+			 * otherwise this test would lose track of dirty pages.
 			 */
-			if (i && host_log_mode != LOG_MODE_DIRTY_RING)
+			if (i > reap_i && host_log_mode != LOG_MODE_DIRTY_RING)
 				continue;
 
 			/*
@@ -745,7 +759,7 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 			 * the ring on every pass would make it unlikely the
 			 * vCPU would ever fill the fing).
 			 */
-			if (i && !READ_ONCE(dirty_ring_vcpu_ring_full))
+			if (i > reap_i && !READ_ONCE(dirty_ring_vcpu_ring_full))
 				continue;
 
 			log_mode_collect_dirty_pages(vcpu, TEST_MEM_SLOT_INDEX,

base-commit: 66939c1603bd5579e63278f9dc72cba5b79da9b5
-- 
2.54.0.794.g4f17f83d09-goog

Re: [PATCH] KVM: selftests: Randomize dirty_log_test's delay before reaping the bitmap

[Sean Christopherson]

On Fri, 22 May 2026 10:02:30 -0700, Sean Christopherson wrote:
> In the dirty log test, randomize the delay before the initial call to get
> the dirty log bitmap for a given iteration, so that the amount of memory
> dirtied by the guest varies from iteration to iteration, and so that the
> user can effectively control the duration (by increasing the interval).
> 
> Always waiting 1ms effectively hides a KVM RISC-V bug as the test reaps the
> dirty bitmap before the guest has a chance to trigger the problematic flow
> in KVM.
> 
> [...]

Applied to kvm-x86 selftests, thanks!

[1/1] KVM: selftests: Randomize dirty_log_test's delay before reaping the bitmap
      https://github.com/kvm-x86/linux/commit/4e0fdd9b0d7d

--
https://github.com/kvm-x86/linux/tree/next