xref: /qemu/tests/qemu-iotests/185 (revision 7653b1ea)
1#!/usr/bin/env bash
2# group: rw
3#
4# Test exiting qemu while jobs are still running
5#
6# Copyright (C) 2017 Red Hat, Inc.
7#
8# This program is free software; you can redistribute it and/or modify
9# it under the terms of the GNU General Public License as published by
10# the Free Software Foundation; either version 2 of the License, or
11# (at your option) any later version.
12#
13# This program is distributed in the hope that it will be useful,
14# but WITHOUT ANY WARRANTY; without even the implied warranty of
15# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
16# GNU General Public License for more details.
17#
18# You should have received a copy of the GNU General Public License
19# along with this program.  If not, see <http://www.gnu.org/licenses/>.
20#
21
22# creator
23owner=kwolf@redhat.com
24
25seq=`basename $0`
26echo "QA output created by $seq"
27
28status=1 # failure is the default!
29
30_cleanup()
31{
32    _rm_test_img "${TEST_IMG}.mid"
33    _rm_test_img "${TEST_IMG}.copy"
34    _cleanup_test_img
35    _cleanup_qemu
36
37    if [ -f "$TEST_DIR/qsd.pid" ]; then
38        kill -SIGKILL "$(cat "$TEST_DIR/qsd.pid")"
39        rm -f "$TEST_DIR/qsd.pid"
40    fi
41    rm -f "$SOCK_DIR/qsd.sock"
42}
43trap "_cleanup; exit \$status" 0 1 2 3 15
44
45# get standard environment, filters and checks
46. ./common.rc
47. ./common.filter
48. ./common.qemu
49
50_supported_fmt qcow2
51_supported_proto file
52_supported_os Linux
53
54size=$((64 * 1048576))
55TEST_IMG="${TEST_IMG}.base" _make_test_img $size
56
57echo
58echo === Starting VM ===
59echo
60
61qemu_comm_method="qmp"
62
63_launch_qemu \
64    -drive file="${TEST_IMG}.base",cache=$CACHEMODE,aio=$AIOMODE,driver=$IMGFMT,id=disk
65h=$QEMU_HANDLE
66_send_qemu_cmd $h "{ 'execute': 'qmp_capabilities' }" 'return'
67
68echo
69echo === Creating backing chain ===
70echo
71
72_send_qemu_cmd $h \
73    "{ 'execute': 'blockdev-snapshot-sync',
74       'arguments': { 'device': 'disk',
75                      'snapshot-file': '$TEST_IMG.mid',
76                      'format': '$IMGFMT',
77                      'mode': 'absolute-paths' } }" \
78    "return"
79
80_send_qemu_cmd $h \
81    "{ 'execute': 'human-monitor-command',
82       'arguments': { 'command-line':
83                      'qemu-io disk \"write 0 4M\"' } }" \
84    "return"
85
86_send_qemu_cmd $h \
87    "{ 'execute': 'blockdev-snapshot-sync',
88       'arguments': { 'device': 'disk',
89                      'snapshot-file': '$TEST_IMG',
90                      'format': '$IMGFMT',
91                      'mode': 'absolute-paths' } }" \
92    "return"
93
94echo
95echo === Start commit job and exit qemu ===
96echo
97
98# Note that the reference output intentionally includes the 'offset' field in
99# BLOCK_JOB_* events for all of the following block jobs. They are predictable
100# and any change in the offsets would hint at a bug in the job throttling code.
101#
102# In order to achieve these predictable offsets, all of the following tests
103# use speed=65536. Each job will perform exactly one iteration before it has
104# to sleep at least for a second, which is plenty of time for the 'quit' QMP
105# command to be received (after receiving the command, the rest runs
106# synchronously, so jobs can arbitrarily continue or complete).
107#
108# The buffer size for commit and streaming is 512k (waiting for 8 seconds after
109# the first request), for active commit and mirror it's large enough to cover
110# the full 4M, and for backup it's the qcow2 cluster size, which we know is
111# 64k. As all of these are at least as large as the speed, we are sure that the
112# offset advances exactly once before qemu exits.
113
114_send_qemu_cmd $h \
115    "{ 'execute': 'block-commit',
116       'arguments': { 'device': 'disk',
117                      'base':'$TEST_IMG.base',
118                      'top': '$TEST_IMG.mid',
119                      'speed': 65536 } }" \
120    "return"
121
122# If we don't sleep here 'quit' command races with disk I/O
123sleep 0.5
124
125# Ignore the JOB_STATUS_CHANGE events while shutting down the VM. Depending on
126# the timing, jobs may or may not transition through a paused state.
127_send_qemu_cmd $h "{ 'execute': 'quit' }" "return"
128wait=1 _cleanup_qemu | grep -v 'JOB_STATUS_CHANGE'
129
130echo
131echo === Start active commit job and exit qemu ===
132echo
133
134_launch_qemu \
135    -drive file="${TEST_IMG}",cache=$CACHEMODE,aio=$AIOMODE,driver=$IMGFMT,id=disk
136h=$QEMU_HANDLE
137_send_qemu_cmd $h "{ 'execute': 'qmp_capabilities' }" 'return'
138
139_send_qemu_cmd $h \
140    "{ 'execute': 'block-commit',
141       'arguments': { 'device': 'disk',
142                      'base':'$TEST_IMG.base',
143                      'speed': 65536 } }" \
144    "return"
145
146# If we don't sleep here 'quit' command races with disk I/O
147sleep 0.5
148
149_send_qemu_cmd $h "{ 'execute': 'quit' }" "return"
150wait=1 _cleanup_qemu | grep -v 'JOB_STATUS_CHANGE'
151
152echo
153echo === Start mirror job and exit qemu ===
154echo
155
156_launch_qemu \
157    -drive file="${TEST_IMG}",cache=$CACHEMODE,aio=$AIOMODE,driver=$IMGFMT,id=disk
158h=$QEMU_HANDLE
159_send_qemu_cmd $h "{ 'execute': 'qmp_capabilities' }" 'return'
160
161_send_qemu_cmd $h \
162    "{ 'execute': 'drive-mirror',
163       'arguments': { 'device': 'disk',
164                      'target': '$TEST_IMG.copy',
165                      'format': '$IMGFMT',
166                      'sync': 'full',
167                      'speed': 65536 } }" \
168    "return"
169
170# If we don't sleep here 'quit' command may be handled before
171# the first mirror iteration is done
172sleep 0.5
173
174_send_qemu_cmd $h "{ 'execute': 'quit' }" "return"
175wait=1 _cleanup_qemu | grep -v 'JOB_STATUS_CHANGE'
176
177echo
178echo === Start backup job and exit qemu ===
179echo
180
181_launch_qemu \
182    -drive file="${TEST_IMG}",cache=$CACHEMODE,aio=$AIOMODE,driver=$IMGFMT,id=disk
183h=$QEMU_HANDLE
184_send_qemu_cmd $h "{ 'execute': 'qmp_capabilities' }" 'return'
185
186_send_qemu_cmd $h \
187    "{ 'execute': 'drive-backup',
188       'arguments': { 'device': 'disk',
189                      'target': '$TEST_IMG.copy',
190                      'format': '$IMGFMT',
191                      'sync': 'full',
192                      'speed': 65536,
193                      'x-perf': {'max-chunk': 65536} } }" \
194    "return"
195
196# If we don't sleep here 'quit' command races with disk I/O
197sleep 0.5
198
199_send_qemu_cmd $h "{ 'execute': 'quit' }" "return"
200wait=1 _cleanup_qemu | grep -v 'JOB_STATUS_CHANGE'
201
202echo
203echo === Start streaming job and exit qemu ===
204echo
205
206_launch_qemu \
207    -drive file="${TEST_IMG}",cache=$CACHEMODE,aio=$AIOMODE,driver=$IMGFMT,id=disk
208h=$QEMU_HANDLE
209_send_qemu_cmd $h "{ 'execute': 'qmp_capabilities' }" 'return'
210
211_send_qemu_cmd $h \
212    "{ 'execute': 'block-stream',
213       'arguments': { 'device': 'disk',
214                      'speed': 65536 } }" \
215    "return"
216
217# If we don't sleep here 'quit' command races with disk I/O
218sleep 0.5
219
220_send_qemu_cmd $h "{ 'execute': 'quit' }" "return"
221wait=1 _cleanup_qemu | grep -v 'JOB_STATUS_CHANGE'
222
223_check_test_img
224
225echo
226echo === Start mirror to throttled QSD and exit qemu ===
227echo
228
229# Mirror to a throttled QSD instance (so that qemu cannot drain the
230# throttling), wait for READY, then write some data to the device,
231# and then quit qemu.
232# (qemu should force-cancel the job and not wait for the data to be
233# written to the target.)
234
235_make_test_img $size
236
237# Will be used by this and the next case
238set_up_throttled_qsd() {
239    $QSD \
240        --object throttle-group,id=thrgr,limits.bps-total=1048576 \
241        --blockdev null-co,node-name=null,size=$size \
242        --blockdev throttle,node-name=throttled,throttle-group=thrgr,file=null \
243        --nbd-server addr.type=unix,addr.path="$SOCK_DIR/qsd.sock" \
244        --export nbd,id=exp,node-name=throttled,name=target,writable=true \
245        --pidfile "$TEST_DIR/qsd.pid" \
246        --daemonize
247}
248
249set_up_throttled_qsd
250
251# Need a virtio-blk device so that qemu-io writes will not block the monitor
252_launch_qemu \
253    --blockdev file,node-name=source-proto,filename="$TEST_IMG" \
254    --blockdev qcow2,node-name=source-fmt,file=source-proto \
255    --device virtio-blk,id=vblk,drive=source-fmt \
256    --blockdev "{\"driver\": \"nbd\",
257                 \"node-name\": \"target\",
258                 \"server\": {
259                     \"type\": \"unix\",
260                     \"path\": \"$SOCK_DIR/qsd.sock\"
261                 },
262                 \"export\": \"target\"}"
263
264h=$QEMU_HANDLE
265_send_qemu_cmd $h '{"execute": "qmp_capabilities"}' 'return'
266
267# Use sync=top, so the first pass will not copy the whole image
268_send_qemu_cmd $h \
269    '{"execute": "blockdev-mirror",
270      "arguments": {
271          "job-id": "mirror",
272          "device": "source-fmt",
273          "target": "target",
274          "sync": "top"
275      }}' \
276    'return' \
277    | grep -v JOB_STATUS_CHANGE # Ignore these events during creation
278
279# This too will be used by this and the next case
280# $1: QEMU handle
281# $2: Image size
282wait_for_job_and_quit() {
283    h=$1
284    size=$2
285
286    # List of expected events
287    capture_events='BLOCK_JOB_READY JOB_STATUS_CHANGE'
288    _wait_event $h 'BLOCK_JOB_READY'
289    QEMU_EVENTS= # Ignore all JOB_STATUS_CHANGE events that came before READY
290
291    # Write something to the device for post-READY mirroring.  Write it in
292    # blocks matching the cluster size, each spaced one block apart, so
293    # that the mirror job will have to spawn one request per cluster.
294    # Because the number of concurrent requests is limited (to 16), this
295    # limits the number of bytes concurrently in flight, which speeds up
296    # cancelling the job (in-flight requests still are waited for).
297    # To limit the number of bytes in flight, we could alternatively pass
298    # something for blockdev-mirror's @buf-size parameter, but
299    # block-commit does not have such a parameter, so we need to figure
300    # something out that works for both.
301
302    cluster_size=65536
303    step=$((cluster_size * 2))
304
305    echo '--- Writing data to the virtio-blk device ---'
306
307    for ofs in $(seq 0 $step $((size - step))); do
308        qemu_io_cmd="qemu-io -d vblk/virtio-backend "
309        qemu_io_cmd+="\\\"aio_write $ofs $cluster_size\\\""
310
311        # Do not include these requests in the reference output
312        # (it's just too much)
313        silent=yes _send_qemu_cmd $h \
314            "{\"execute\": \"human-monitor-command\",
315              \"arguments\": {
316                  \"command-line\": \"$qemu_io_cmd\"
317              }}" \
318            'return'
319    done
320
321    # Wait until the job's length is updated to reflect the write requests
322
323    # We have written to half of the device, so this is the expected job length
324    final_len=$((size / 2))
325    timeout=100 # unit: 0.1 seconds
326    while true; do
327        len=$(
328            _send_qemu_cmd $h \
329                '{"execute": "query-block-jobs"}' \
330                'return.*"len": [0-9]\+' \
331                | grep 'return.*"len": [0-9]\+' \
332                | sed -e 's/.*"len": \([0-9]\+\).*/\1/'
333        )
334        if [ "$len" -eq "$final_len" ]; then
335            break
336        fi
337        timeout=$((timeout - 1))
338        if [ "$timeout" -eq 0 ]; then
339            echo "ERROR: Timeout waiting for job to reach len=$final_len"
340            break
341        fi
342        sleep 0.1
343    done
344
345    sleep 1
346
347    # List of expected events
348    capture_events='BLOCK_JOB_CANCELLED JOB_STATUS_CHANGE SHUTDOWN'
349
350    _send_qemu_cmd $h \
351        '{"execute": "quit"}' \
352        'return'
353
354    _wait_event $h 'SHUTDOWN'
355    _wait_event $h 'JOB_STATUS_CHANGE' # standby
356    _wait_event $h 'JOB_STATUS_CHANGE' # ready
357    _wait_event $h 'JOB_STATUS_CHANGE' # standby
358    _wait_event $h 'JOB_STATUS_CHANGE' # ready
359    _wait_event $h 'JOB_STATUS_CHANGE' # aborting
360    # Filter the offset (depends on when exactly `quit` was issued)
361    _wait_event $h 'BLOCK_JOB_CANCELLED' \
362        | sed -e 's/"offset": [0-9]\+/"offset": (filtered)/'
363    _wait_event $h 'JOB_STATUS_CHANGE' # concluded
364    _wait_event $h 'JOB_STATUS_CHANGE' # null
365
366    wait=yes _cleanup_qemu
367
368    kill -SIGTERM "$(cat "$TEST_DIR/qsd.pid")"
369}
370
371wait_for_job_and_quit $h $size
372
373echo
374echo === Start active commit to throttled QSD and exit qemu ===
375echo
376
377# Same as the above, but instead of mirroring, do an active commit
378
379_make_test_img $size
380
381set_up_throttled_qsd
382
383_launch_qemu \
384    --blockdev "{\"driver\": \"nbd\",
385                 \"node-name\": \"target\",
386                 \"server\": {
387                     \"type\": \"unix\",
388                     \"path\": \"$SOCK_DIR/qsd.sock\"
389                 },
390                 \"export\": \"target\"}" \
391    --blockdev file,node-name=source-proto,filename="$TEST_IMG" \
392    --blockdev qcow2,node-name=source-fmt,file=source-proto,backing=target \
393    --device virtio-blk,id=vblk,drive=source-fmt
394
395h=$QEMU_HANDLE
396_send_qemu_cmd $h '{"execute": "qmp_capabilities"}' 'return'
397
398_send_qemu_cmd $h \
399    '{"execute": "block-commit",
400      "arguments": {
401          "job-id": "commit",
402          "device": "source-fmt"
403      }}' \
404    'return' \
405    | grep -v JOB_STATUS_CHANGE # Ignore these events during creation
406
407wait_for_job_and_quit $h $size
408
409# success, all done
410echo "*** done"
411rm -f $seq.full
412status=0
413