performance-metrics: Add QCOW2 write micro benchmark

Add micro_bench_qcow_write which times write_vectored calls through
QcowSync on an empty QCOW2 image. Each write allocates a new cluster
exercising map_cluster_for_write with L2 entry allocation and refcount
updates followed by pwrite_all.

Two TEST_LIST entries: micro_block_qcow_write_128_us and
micro_block_qcow_write_256_us with 128 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; 64] = [
+const TEST_LIST: [PerformanceTest; 66] = [
     PerformanceTest {
         name: "boot_time_ms",
         func_ptr: performance_boot_time,
@@ -1277,6 +1277,30 @@ const TEST_LIST: [PerformanceTest; 64] = [
         },
         unit_adjuster: adjuster::s_to_us,
     },
+    PerformanceTest {
+        name: "micro_block_qcow_write_128_us",
+        func_ptr: micro_bench_block::micro_bench_qcow_write,
+        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_write_256_us",
+        func_ptr: micro_bench_block::micro_bench_qcow_write,
+        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

@@ -17,6 +17,7 @@ use block::raw_async_aio::RawFileAsyncAio;
 use crate::PerformanceTestControl;
 use crate::util::{
     self, BLOCK_SIZE, QCOW_CLUSTER_SIZE, drain_completions, read_iovec, submit_reads,
+    submit_writes, write_iovec,
 };
 
 /// Submit num_ops AIO writes, wait for them all to land, then time
@@ -79,3 +80,29 @@ pub fn micro_bench_qcow_read(control: &PerformanceTestControl) -> f64 {
 
     elapsed
 }
+
+/// Write num_ops clusters into an empty qcow2 image through the
+/// QcowSync async_io path and time the total write_vectored wall clock.
+///
+/// This exercises the write allocation path: map_cluster_for_write
+/// allocates a new cluster and bumps refcounts, then pwrite_all writes
+/// the data.
+///
+/// Returns the total write wall clock time in seconds.
+pub fn micro_bench_qcow_write(control: &PerformanceTestControl) -> f64 {
+    let num_ops = control.num_ops.expect("num_ops required") as usize;
+    let (_tmp, disk) = util::empty_qcow_tempfile(num_ops);
+    let mut async_io = disk.new_async_io(1).expect("new_async_io failed");
+
+    let buf = vec![0xA5u8; QCOW_CLUSTER_SIZE as usize];
+    let iovec = write_iovec(&buf);
+
+    let start = Instant::now();
+    submit_writes(async_io.as_mut(), num_ops, QCOW_CLUSTER_SIZE, &[iovec]);
+    let elapsed = start.elapsed().as_secs_f64();
+
+    // Drain completions so Drop is clean.
+    drain_completions(async_io.as_mut(), num_ops);
+
+    elapsed
+}