performance-metrics: Add QCOW2 read micro benchmark

Add micro_bench_qcow_read which times read_vectored calls through
QcowSync on a prepopulated QCOW2 image. This exercises the hot
read path including L2 lookup, pread64 for allocated clusters and
iovec scatter.

Two TEST_LIST entries: micro_block_qcow_read_128_us and
micro_block_qcow_read_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; 62] = [
+const TEST_LIST: [PerformanceTest; 64] = [
     PerformanceTest {
         name: "boot_time_ms",
         func_ptr: performance_boot_time,
@@ -1253,6 +1253,30 @@ const TEST_LIST: [PerformanceTest; 62] = [
         },
         unit_adjuster: adjuster::s_to_us,
     },
+    PerformanceTest {
+        name: "micro_block_qcow_read_128_us",
+        func_ptr: micro_bench_block::micro_bench_qcow_read,
+        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_read_256_us",
+        func_ptr: micro_bench_block::micro_bench_qcow_read,
+        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

@@ -11,10 +11,13 @@ use std::os::unix::io::AsRawFd;
 use std::time::Instant;
 
 use block::async_io::AsyncIo;
+use block::disk_file::AsyncDiskFile;
 use block::raw_async_aio::RawFileAsyncAio;
 
 use crate::PerformanceTestControl;
-use crate::util::{self, BLOCK_SIZE};
+use crate::util::{
+    self, BLOCK_SIZE, QCOW_CLUSTER_SIZE, drain_completions, read_iovec, submit_reads,
+};
 
 /// Submit num_ops AIO writes, wait for them all to land, then time
 /// how long it takes to drain every completion via next_completed_request().
@@ -51,3 +54,28 @@ pub fn micro_bench_aio_drain(control: &PerformanceTestControl) -> f64 {
     }
     start.elapsed().as_secs_f64()
 }
+
+/// Read num_ops clusters from a prepopulated qcow2 image through the
+/// QcowSync async_io path and time the total read_vectored wall clock.
+///
+/// This exercises the hot read path: L2 lookup via map_clusters_for_read,
+/// pread64 for allocated data, and iovec scatter.
+///
+/// Returns the total read wall clock time in seconds.
+pub fn micro_bench_qcow_read(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 mut buf = vec![0u8; QCOW_CLUSTER_SIZE as usize];
+    let iovec = read_iovec(&mut buf);
+
+    let start = Instant::now();
+    submit_reads(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
+}