iotests/185: Add post-READY quit tests

185 tests quitting qemu while a block job is active.  It does not
specifically test quitting qemu while a mirror or active commit job is
in its READY phase.

Add two test cases for this, where we respectively mirror or commit to
an external QSD instance, which provides a throttled block device.  qemu
is supposed to cancel the job so that it can quit as soon as possible
instead of waiting for the job to complete (which it did before 6.2).

Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220303164814.284974-5-hreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

1 files changed, 189 insertions, 1 deletions

diff --git a/tests/qemu-iotests/185 b/tests/qemu-iotests/185
index f2ec5c5..8b1143d 100755
--- a/tests/qemu-iotests/185
+++ b/tests/qemu-iotests/185
@@ -33,6 +33,12 @@ _cleanup()
     _rm_test_img "${TEST_IMG}.copy"
     _cleanup_test_img
     _cleanup_qemu
+
+    if [ -f "$TEST_DIR/qsd.pid" ]; then
+        kill -SIGKILL "$(cat "$TEST_DIR/qsd.pid")"
+        rm -f "$TEST_DIR/qsd.pid"
+    fi
+    rm -f "$SOCK_DIR/qsd.sock"
 }
 trap "_cleanup; exit \$status" 0 1 2 3 15
 
@@ -45,7 +51,7 @@ _supported_fmt qcow2
 _supported_proto file
 _supported_os Linux
 
-size=64M
+size=$((64 * 1048576))
 TEST_IMG="${TEST_IMG}.base" _make_test_img $size
 
 echo
@@ -216,6 +222,188 @@ wait=1 _cleanup_qemu | grep -v 'JOB_STATUS_CHANGE'
 
 _check_test_img
 
+echo
+echo === Start mirror to throttled QSD and exit qemu ===
+echo
+
+# Mirror to a throttled QSD instance (so that qemu cannot drain the
+# throttling), wait for READY, then write some data to the device,
+# and then quit qemu.
+# (qemu should force-cancel the job and not wait for the data to be
+# written to the target.)
+
+_make_test_img $size
+
+# Will be used by this and the next case
+set_up_throttled_qsd() {
+    $QSD \
+        --object throttle-group,id=thrgr,limits.bps-total=1048576 \
+        --blockdev null-co,node-name=null,size=$size \
+        --blockdev throttle,node-name=throttled,throttle-group=thrgr,file=null \
+        --nbd-server addr.type=unix,addr.path="$SOCK_DIR/qsd.sock" \
+        --export nbd,id=exp,node-name=throttled,name=target,writable=true \
+        --pidfile "$TEST_DIR/qsd.pid" \
+        --daemonize
+}
+
+set_up_throttled_qsd
+
+# Need a virtio-blk device so that qemu-io writes will not block the monitor
+_launch_qemu \
+    --blockdev file,node-name=source-proto,filename="$TEST_IMG" \
+    --blockdev qcow2,node-name=source-fmt,file=source-proto \
+    --device virtio-blk,id=vblk,drive=source-fmt \
+    --blockdev "{\"driver\": \"nbd\",
+                 \"node-name\": \"target\",
+                 \"server\": {
+                     \"type\": \"unix\",
+                     \"path\": \"$SOCK_DIR/qsd.sock\"
+                 },
+                 \"export\": \"target\"}"
+
+h=$QEMU_HANDLE
+_send_qemu_cmd $h '{"execute": "qmp_capabilities"}' 'return'
+
+# Use sync=top, so the first pass will not copy the whole image
+_send_qemu_cmd $h \
+    '{"execute": "blockdev-mirror",
+      "arguments": {
+          "job-id": "mirror",
+          "device": "source-fmt",
+          "target": "target",
+          "sync": "top"
+      }}' \
+    'return' \
+    | grep -v JOB_STATUS_CHANGE # Ignore these events during creation
+
+# This too will be used by this and the next case
+# $1: QEMU handle
+# $2: Image size
+wait_for_job_and_quit() {
+    h=$1
+    size=$2
+
+    # List of expected events
+    capture_events='BLOCK_JOB_READY JOB_STATUS_CHANGE'
+    _wait_event $h 'BLOCK_JOB_READY'
+    QEMU_EVENTS= # Ignore all JOB_STATUS_CHANGE events that came before READY
+
+    # Write something to the device for post-READY mirroring.  Write it in
+    # blocks matching the cluster size, each spaced one block apart, so
+    # that the mirror job will have to spawn one request per cluster.
+    # Because the number of concurrent requests is limited (to 16), this
+    # limits the number of bytes concurrently in flight, which speeds up
+    # cancelling the job (in-flight requests still are waited for).
+    # To limit the number of bytes in flight, we could alternatively pass
+    # something for blockdev-mirror's @buf-size parameter, but
+    # block-commit does not have such a parameter, so we need to figure
+    # something out that works for both.
+
+    cluster_size=65536
+    step=$((cluster_size * 2))
+
+    echo '--- Writing data to the virtio-blk device ---'
+
+    for ofs in $(seq 0 $step $((size - step))); do
+        qemu_io_cmd="qemu-io -d vblk/virtio-backend "
+        qemu_io_cmd+="\\\"aio_write $ofs $cluster_size\\\""
+
+        # Do not include these requests in the reference output
+        # (it's just too much)
+        silent=yes _send_qemu_cmd $h \
+            "{\"execute\": \"human-monitor-command\",
+              \"arguments\": {
+                  \"command-line\": \"$qemu_io_cmd\"
+              }}" \
+            'return'
+    done
+
+    # Wait until the job's length is updated to reflect the write requests
+
+    # We have written to half of the device, so this is the expected job length
+    final_len=$((size / 2))
+    timeout=100 # unit: 0.1 seconds
+    while true; do
+        len=$(
+            _send_qemu_cmd $h \
+                '{"execute": "query-block-jobs"}' \
+                'return.*"len": [0-9]\+' \
+                | grep 'return.*"len": [0-9]\+' \
+                | sed -e 's/.*"len": \([0-9]\+\).*/\1/'
+        )
+        if [ "$len" -eq "$final_len" ]; then
+            break
+        fi
+        timeout=$((timeout - 1))
+        if [ "$timeout" -eq 0 ]; then
+            echo "ERROR: Timeout waiting for job to reach len=$final_len"
+            break
+        fi
+        sleep 0.1
+    done
+
+    sleep 1
+
+    _send_qemu_cmd $h \
+        '{"execute": "quit"}' \
+        'return'
+
+    # List of expected events
+    capture_events='BLOCK_JOB_CANCELLED JOB_STATUS_CHANGE SHUTDOWN'
+    _wait_event $h 'SHUTDOWN'
+    QEMU_EVENTS= # Ignore all JOB_STATUS_CHANGE events that came before SHUTDOWN
+    _wait_event $h 'JOB_STATUS_CHANGE' # standby
+    _wait_event $h 'JOB_STATUS_CHANGE' # ready
+    _wait_event $h 'JOB_STATUS_CHANGE' # aborting
+    # Filter the offset (depends on when exactly `quit` was issued)
+    _wait_event $h 'BLOCK_JOB_CANCELLED' \
+        | sed -e 's/"offset": [0-9]\+/"offset": (filtered)/'
+    _wait_event $h 'JOB_STATUS_CHANGE' # concluded
+    _wait_event $h 'JOB_STATUS_CHANGE' # null
+
+    wait=yes _cleanup_qemu
+
+    kill -SIGTERM "$(cat "$TEST_DIR/qsd.pid")"
+}
+
+wait_for_job_and_quit $h $size
+
+echo
+echo === Start active commit to throttled QSD and exit qemu ===
+echo
+
+# Same as the above, but instead of mirroring, do an active commit
+
+_make_test_img $size
+
+set_up_throttled_qsd
+
+_launch_qemu \
+    --blockdev "{\"driver\": \"nbd\",
+                 \"node-name\": \"target\",
+                 \"server\": {
+                     \"type\": \"unix\",
+                     \"path\": \"$SOCK_DIR/qsd.sock\"
+                 },
+                 \"export\": \"target\"}" \
+    --blockdev file,node-name=source-proto,filename="$TEST_IMG" \
+    --blockdev qcow2,node-name=source-fmt,file=source-proto,backing=target \
+    --device virtio-blk,id=vblk,drive=source-fmt
+
+h=$QEMU_HANDLE
+_send_qemu_cmd $h '{"execute": "qmp_capabilities"}' 'return'
+
+_send_qemu_cmd $h \
+    '{"execute": "block-commit",
+      "arguments": {
+          "job-id": "commit",
+          "device": "source-fmt"
+      }}' \
+    'return' \
+    | grep -v JOB_STATUS_CHANGE # Ignore these events during creation
+
+wait_for_job_and_quit $h $size
+
 # success, all done
 echo "*** done"
 rm -f $seq.full