performance-metrics: Add qcow2 punch hole micro benchmark

Add micro_bench_qcow_punch_hole which times punch_hole calls through
QcowSync on a prepopulated qcow2 image. Each call deallocates one
cluster exercising deallocate_bytes with refcount decrement and
fallocate punch_hole on the host file.

Two TEST_LIST entries: micro_block_qcow_punch_hole_64_us and
micro_block_qcow_punch_hole_256_us with 64 and 256 cluster workloads.

Signed-off-by: Anatol Belski <anbelski@linux.microsoft.com>

performance-metrics/src/main.rs

@@ -378,7 +378,7 @@ mod adjuster {
     }
 }
 
-const TEST_LIST: [PerformanceTest; 66] = [
+const TEST_LIST: [PerformanceTest; 68] = [
     PerformanceTest {
         name: "boot_time_ms",
         func_ptr: performance_boot_time,
@@ -1301,6 +1301,30 @@ const TEST_LIST: [PerformanceTest; 66] = [
         },
         unit_adjuster: adjuster::s_to_us,
     },
+    PerformanceTest {
+        name: "micro_block_qcow_punch_hole_128_us",
+        func_ptr: micro_bench_block::micro_bench_qcow_punch_hole,
+        control: PerformanceTestControl {
+            test_timeout: 10,
+            test_iterations: 20,
+            warmup_iterations: 5,
+            num_ops: Some(128),
+            ..PerformanceTestControl::default()
+        },
+        unit_adjuster: adjuster::s_to_us,
+    },
+    PerformanceTest {
+        name: "micro_block_qcow_punch_hole_256_us",
+        func_ptr: micro_bench_block::micro_bench_qcow_punch_hole,
+        control: PerformanceTestControl {
+            test_timeout: 10,
+            test_iterations: 20,
+            warmup_iterations: 5,
+            num_ops: Some(256),
+            ..PerformanceTestControl::default()
+        },
+        unit_adjuster: adjuster::s_to_us,
+    },
 ];
 
 fn run_test_with_timeout(

performance-metrics/src/micro_bench_block.rs

@@ -106,3 +106,29 @@ pub fn micro_bench_qcow_write(control: &PerformanceTestControl) -> f64 {
 
     elapsed
 }
+
+/// Punch holes for num_ops clusters in a prepopulated qcow2 image through
+/// the QcowSync async_io path and time the total punch_hole wall clock.
+///
+/// This exercises the discard path: deallocate_bytes decrements refcounts,
+/// frees clusters and issues fallocate punch_hole on the host file.
+///
+/// Returns the total punch_hole wall clock time in seconds.
+pub fn micro_bench_qcow_punch_hole(control: &PerformanceTestControl) -> f64 {
+    let num_ops = control.num_ops.expect("num_ops required") as usize;
+    let (_tmp, disk) = util::qcow_tempfile(num_ops);
+    let mut async_io = disk.new_async_io(1).expect("new_async_io failed");
+
+    let start = Instant::now();
+    for i in 0..num_ops {
+        async_io
+            .punch_hole(i as u64 * QCOW_CLUSTER_SIZE, QCOW_CLUSTER_SIZE, i as u64)
+            .expect("punch_hole failed");
+    }
+    let elapsed = start.elapsed().as_secs_f64();
+
+    // Drain completions so Drop is clean.
+    drain_completions(async_io.as_mut(), num_ops);
+
+    elapsed
+}