t/zbd: Mandate blkzone capacity report for devices with zone capacity
[fio.git] / t / zbd / test-zbd-support
CommitLineData
191d1d1a
BVA
1#!/bin/bash
2#
3# Copyright (C) 2018 Western Digital Corporation or its affiliates.
4#
5# This file is released under the GPL.
6
7usage() {
6dcb098d 8 echo "Usage: $(basename "$0") [-d] [-e] [-l] [-r] [-v] [-t <test>] [-z] <SMR drive device node>"
191d1d1a
BVA
9}
10
11max() {
12 if [ "$1" -gt "$2" ]; then
13 echo "$1"
14 else
15 echo "$2"
16 fi
17}
18
19min() {
20 if [ "$1" -lt "$2" ]; then
21 echo "$1"
22 else
23 echo "$2"
24 fi
25}
26
6dcb098d
DF
27ioengine() {
28 if [ -n "$use_libzbc" ]; then
29 echo -n "--ioengine=libzbc"
30 else
31 echo -n "--ioengine=$1"
32 fi
33}
34
191d1d1a
BVA
35set_io_scheduler() {
36 local dev=$1 sched=$2
37
38 [ -e "/sys/block/$dev" ] || return $?
39 if [ -e "/sys/block/$dev/mq" ]; then
40 case "$sched" in
41 noop) sched=none;;
42 deadline) sched=mq-deadline;;
43 esac
44 else
45 case "$sched" in
46 none) sched=noop;;
47 mq-deadline) sched=deadline;;
48 esac
49 fi
50
51 echo "$sched" >"/sys/block/$dev/queue/scheduler"
52}
53
54check_read() {
55 local read
56
57 read=$(fio_read <"${logfile}.${test_number}")
58 echo "read: $read <> $1" >> "${logfile}.${test_number}"
59 [ "$read" = "$1" ]
60}
61
62check_written() {
63 local written
64
65 written=$(fio_written <"${logfile}.${test_number}")
66 echo "written: $written <> $1" >> "${logfile}.${test_number}"
67 [ "$written" = "$1" ]
68}
69
70# Compare the reset count from the log file with reset count $2 using operator
71# $1 (=, -ge, -gt, -le, -lt).
72check_reset_count() {
73 local reset_count
74
75 reset_count=$(fio_reset_count <"${logfile}.${test_number}")
76 echo "reset_count: test $reset_count $1 $2" >> "${logfile}.${test_number}"
77 eval "[ '$reset_count' '$1' '$2' ]"
78}
79
80# Whether or not $1 (/dev/...) is a SCSI device.
81is_scsi_device() {
82 local d f
83
84 d=$(basename "$dev")
85 for f in /sys/class/scsi_device/*/device/block/"$d"; do
86 [ -e "$f" ] && return 0
87 done
88 return 1
89}
90
91run_fio() {
97e41ee4 92 local fio opts
191d1d1a
BVA
93
94 fio=$(dirname "$0")/../../fio
95
fc3bf67a
DLM
96 opts=("--max-jobs=16" "--aux-path=/tmp" "--allow_file_create=0" \
97 "--significant_figures=10" "$@")
6dcb098d 98 opts+=(${var_opts[@]})
97e41ee4 99 { echo; echo "fio ${opts[*]}"; echo; } >>"${logfile}.${test_number}"
191d1d1a 100
97e41ee4 101 "${dynamic_analyzer[@]}" "$fio" "${opts[@]}"
191d1d1a
BVA
102}
103
104run_one_fio_job() {
105 local r
106
107 r=$(((RANDOM << 16) | RANDOM))
108 run_fio --name="$dev" --filename="$dev" "$@" --randseed="$r" \
109 --thread=1 --direct=1
110}
111
8f601b44
SK
112write_and_run_one_fio_job() {
113 local r
114 local write_offset="${1}"
115 local write_size="${2}"
116
117 shift 2
118 r=$(((RANDOM << 16) | RANDOM))
119 run_fio --filename="$dev" --randseed="$r" --name="write_job" --rw=write \
120 "$(ioengine "psync")" --bs="${logical_block_size}" \
121 --zonemode=zbd --zonesize="${zone_size}" --thread=1 --direct=1 \
122 --offset="${write_offset}" --size="${write_size}" \
123 --name="$dev" --wait_for="write_job" "$@" --thread=1 --direct=1
124}
125
191d1d1a
BVA
126# Run fio on the first four sequential zones of the disk.
127run_fio_on_seq() {
128 local opts=()
129
130 opts+=("--offset=$((first_sequential_zone_sector * 512))")
131 opts+=("--size=$((4 * zone_size))" "--zonemode=zbd")
132 if [ -z "$is_zbd" ]; then
133 opts+=("--zonesize=${zone_size}")
134 fi
135 run_one_fio_job "${opts[@]}" "$@"
136}
137
138# Check whether buffered writes are refused.
139test1() {
140 run_fio --name=job1 --filename="$dev" --rw=write --direct=0 --bs=4K \
6dcb098d 141 "$(ioengine "psync")" --size="${zone_size}" --thread=1 \
191d1d1a
BVA
142 --zonemode=zbd --zonesize="${zone_size}" 2>&1 |
143 tee -a "${logfile}.${test_number}" |
144 grep -q 'Using direct I/O is mandatory for writing to ZBD drives'
145 local fio_rc=${PIPESTATUS[0]} grep_rc=${PIPESTATUS[2]}
146 case "$fio_rc" in
147 0|1) ;;
148 *) return "$fio_rc"
149 esac
150 if [ -n "$is_zbd" ]; then
151 [ "$grep_rc" = 0 ]
152 else
153 [ "$grep_rc" != 0 ]
154 fi
155}
156
157# Block size exceeds zone size.
158test2() {
159 local bs off opts=() rc
160
161 off=$(((first_sequential_zone_sector + 2 * sectors_per_zone) * 512))
162 bs=$((2 * zone_size))
6dcb098d 163 opts+=("$(ioengine "psync")")
191d1d1a
BVA
164 opts+=("--name=job1" "--filename=$dev" "--rw=write" "--direct=1")
165 opts+=("--zonemode=zbd" "--offset=$off" "--bs=$bs" "--size=$bs")
166 if [ -z "$is_zbd" ]; then
167 opts+=("--zonesize=${zone_size}")
168 fi
f572c8f6
DLM
169 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
170 ! grep -q 'WRITE:' "${logfile}.${test_number}"
191d1d1a
BVA
171}
172
173# Run fio against an empty zone. This causes fio to report "No I/O performed".
174test3() {
175 local off opts=() rc
176
177 off=$((first_sequential_zone_sector * 512 + 128 * zone_size))
178 size=$((zone_size))
179 [ -n "$is_zbd" ] && reset_zone "$dev" $((off / 512))
180 opts+=("--name=$dev" "--filename=$dev" "--offset=$off" "--bs=4K")
181 opts+=("--size=$size" "--zonemode=zbd")
6dcb098d 182 opts+=("$(ioengine "psync")" "--rw=read" "--direct=1" "--thread=1")
191d1d1a
BVA
183 if [ -z "$is_zbd" ]; then
184 opts+=("--zonesize=${zone_size}")
185 fi
186 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
bbd6f1bb 187 ! grep -q 'READ:' "${logfile}.${test_number}"
191d1d1a
BVA
188}
189
190# Run fio with --read_beyond_wp=1 against an empty zone.
191test4() {
192 local off opts=()
193
194 off=$((first_sequential_zone_sector * 512 + 129 * zone_size))
195 size=$((zone_size))
196 [ -n "$is_zbd" ] && reset_zone "$dev" $((off / 512))
197 opts+=("--name=$dev" "--filename=$dev" "--offset=$off" "--bs=$size")
198 opts+=("--size=$size" "--thread=1" "--read_beyond_wp=1")
6dcb098d 199 opts+=("$(ioengine "psync")" "--rw=read" "--direct=1" "--disable_lat=1")
191d1d1a
BVA
200 opts+=("--zonemode=zbd" "--zonesize=${zone_size}")
201 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
202 check_read $size || return $?
203}
204
205# Sequential write to sequential zones.
206test5() {
207 local size
208
209 size=$((4 * zone_size))
6dcb098d 210 run_fio_on_seq "$(ioengine "psync")" --iodepth=1 --rw=write \
191d1d1a
BVA
211 --bs="$(max $((zone_size / 64)) "$logical_block_size")"\
212 --do_verify=1 --verify=md5 \
213 >>"${logfile}.${test_number}" 2>&1 || return $?
214 check_written $size || return $?
215 check_read $size || return $?
216}
217
0123751c 218# Sequential read from sequential zones.
191d1d1a
BVA
219test6() {
220 local size
221
222 size=$((4 * zone_size))
0123751c
SK
223 write_and_run_one_fio_job \
224 $((first_sequential_zone_sector * 512)) "${size}" \
225 --offset=$((first_sequential_zone_sector * 512)) \
226 --size="${size}" --zonemode=zbd --zonesize="${zone_size}" \
227 "$(ioengine "psync")" --iodepth=1 --rw=read \
228 --bs="$(max $((zone_size / 64)) "$logical_block_size")" \
229 >>"${logfile}.${test_number}" 2>&1 || return $?
191d1d1a
BVA
230 check_read $size || return $?
231}
232
233# Random write to sequential zones, libaio, queue depth 1.
234test7() {
235 local size=$((zone_size))
236
6dcb098d 237 run_fio_on_seq "$(ioengine "libaio")" --iodepth=1 --rw=randwrite \
191d1d1a
BVA
238 --bs="$(min 16384 "${zone_size}")" \
239 --do_verify=1 --verify=md5 --size="$size" \
240 >>"${logfile}.${test_number}" 2>&1 || return $?
241 check_written $size || return $?
242 check_read $size || return $?
243}
244
245# Random write to sequential zones, libaio, queue depth 64.
246test8() {
247 local size
248
249 size=$((4 * zone_size))
6dcb098d 250 run_fio_on_seq "$(ioengine "libaio")" --iodepth=64 --rw=randwrite \
191d1d1a
BVA
251 --bs="$(min 16384 "${zone_size}")" \
252 --do_verify=1 --verify=md5 \
253 >>"${logfile}.${test_number}" 2>&1 || return $?
254 check_written $size || return $?
255 check_read $size || return $?
256}
257
258# Random write to sequential zones, sg, queue depth 1.
259test9() {
260 local size
261
262 if ! is_scsi_device "$dev"; then
263 echo "$dev is not a SCSI device" >>"${logfile}.${test_number}"
264 return 0
265 fi
266
267 size=$((4 * zone_size))
6dcb098d
DF
268 run_fio_on_seq --ioengine=sg \
269 --iodepth=1 --rw=randwrite --bs=16K \
191d1d1a
BVA
270 --do_verify=1 --verify=md5 \
271 >>"${logfile}.${test_number}" 2>&1 || return $?
272 check_written $size || return $?
273 check_read $size || return $?
274}
275
276# Random write to sequential zones, sg, queue depth 64.
277test10() {
278 local size
279
280 if ! is_scsi_device "$dev"; then
281 echo "$dev is not a SCSI device" >>"${logfile}.${test_number}"
282 return 0
283 fi
284
285 size=$((4 * zone_size))
6dcb098d
DF
286 run_fio_on_seq --ioengine=sg \
287 --iodepth=64 --rw=randwrite --bs=16K \
191d1d1a
BVA
288 --do_verify=1 --verify=md5 \
289 >>"${logfile}.${test_number}" 2>&1 || return $?
290 check_written $size || return $?
291 check_read $size || return $?
292}
293
294# Random write to sequential zones, libaio, queue depth 64, random block size.
295test11() {
296 local size
297
298 size=$((4 * zone_size))
6dcb098d 299 run_fio_on_seq "$(ioengine "libaio")" --iodepth=64 --rw=randwrite \
191d1d1a
BVA
300 --bsrange=4K-64K --do_verify=1 --verify=md5 \
301 --debug=zbd >>"${logfile}.${test_number}" 2>&1 || return $?
302 check_written $size || return $?
303 check_read $size || return $?
304}
305
306# Random write to sequential zones, libaio, queue depth 64, max 1 open zone.
307test12() {
308 local size
309
310 size=$((8 * zone_size))
6dcb098d 311 run_fio_on_seq "$(ioengine "libaio")" --iodepth=64 --rw=randwrite --bs=16K \
191d1d1a
BVA
312 --max_open_zones=1 --size=$size --do_verify=1 --verify=md5 \
313 --debug=zbd >>"${logfile}.${test_number}" 2>&1 || return $?
314 check_written $size || return $?
315 check_read $size || return $?
316}
317
318# Random write to sequential zones, libaio, queue depth 64, max 4 open zones.
319test13() {
320 local size
321
322 size=$((8 * zone_size))
6dcb098d 323 run_fio_on_seq "$(ioengine "libaio")" --iodepth=64 --rw=randwrite --bs=16K \
191d1d1a
BVA
324 --max_open_zones=4 --size=$size --do_verify=1 --verify=md5 \
325 --debug=zbd \
326 >>"${logfile}.${test_number}" 2>&1 || return $?
327 check_written $size || return $?
328 check_read $size || return $?
329}
330
331# Random write to conventional zones.
332test14() {
333 local size
334
335 size=$((16 * 2**20)) # 20 MB
336 if [ $size -gt $((first_sequential_zone_sector * 512)) ]; then
337 echo "$dev does not have enough sequential zones" \
338 >>"${logfile}.${test_number}"
339 return 0
340 fi
6dcb098d 341 run_one_fio_job "$(ioengine "libaio")" --iodepth=64 --rw=randwrite --bs=16K \
191d1d1a
BVA
342 --zonemode=zbd --zonesize="${zone_size}" --do_verify=1 \
343 --verify=md5 --size=$size \
344 >>"${logfile}.${test_number}" 2>&1 || return $?
345 check_written $((size)) || return $?
346 check_read $((size)) || return $?
347}
348
349# Sequential read on a mix of empty and full zones.
350test15() {
351 local i off size
291aa0a7 352 local w_off w_size
191d1d1a
BVA
353
354 for ((i=0;i<4;i++)); do
355 [ -n "$is_zbd" ] &&
356 reset_zone "$dev" $((first_sequential_zone_sector +
357 i*sectors_per_zone))
358 done
291aa0a7
SK
359 w_off=$(((first_sequential_zone_sector + 2 * sectors_per_zone) * 512))
360 w_size=$((2 * zone_size))
191d1d1a
BVA
361 off=$((first_sequential_zone_sector * 512))
362 size=$((4 * zone_size))
291aa0a7
SK
363 write_and_run_one_fio_job "${w_off}" "${w_size}" \
364 "$(ioengine "psync")" --rw=read --bs=$((zone_size / 16)) \
191d1d1a
BVA
365 --zonemode=zbd --zonesize="${zone_size}" --offset=$off \
366 --size=$((size)) >>"${logfile}.${test_number}" 2>&1 ||
367 return $?
291aa0a7
SK
368 check_written $((w_size)) || return $?
369 check_read $((size / 2))
191d1d1a
BVA
370}
371
7285b445 372# Random read on a mix of empty and full zones.
191d1d1a
BVA
373test16() {
374 local off size
7285b445 375 local i w_off w_size
191d1d1a 376
7285b445
SK
377 for ((i=0;i<4;i++)); do
378 [ -n "$is_zbd" ] &&
379 reset_zone "$dev" $((first_sequential_zone_sector +
380 i*sectors_per_zone))
381 done
382 w_off=$(((first_sequential_zone_sector + 2 * sectors_per_zone) * 512))
383 w_size=$((2 * zone_size))
191d1d1a
BVA
384 off=$((first_sequential_zone_sector * 512))
385 size=$((4 * zone_size))
7285b445
SK
386 write_and_run_one_fio_job "${w_off}" "${w_size}" \
387 "$(ioengine "libaio")" --iodepth=64 --rw=randread --bs=16K \
191d1d1a
BVA
388 --zonemode=zbd --zonesize="${zone_size}" --offset=$off \
389 --size=$size >>"${logfile}.${test_number}" 2>&1 || return $?
7285b445 390 check_written $w_size || return $?
191d1d1a
BVA
391 check_read $size || return $?
392}
393
394# Random reads and writes in the last zone.
395test17() {
396 local io off read size written
397
398 off=$(((disk_size / zone_size - 1) * zone_size))
399 size=$((disk_size - off))
191d1d1a
BVA
400 if [ -n "$is_zbd" ]; then
401 reset_zone "$dev" $((off / 512)) || return $?
402 fi
6dcb098d 403 run_one_fio_job "$(ioengine "libaio")" --iodepth=8 --rw=randrw --bs=4K \
191d1d1a
BVA
404 --zonemode=zbd --zonesize="${zone_size}" \
405 --offset=$off --loops=2 --norandommap=1\
406 >>"${logfile}.${test_number}" 2>&1 || return $?
407 written=$(fio_written <"${logfile}.${test_number}")
408 read=$(fio_read <"${logfile}.${test_number}")
409 io=$((written + read))
410 echo "Total number of bytes read and written: $io <> $size" \
411 >>"${logfile}.${test_number}"
412 [ $io = $((size * 2)) ];
413}
414
415# Out-of-range zone reset threshold and frequency parameters.
416test18() {
417 run_fio_on_seq --zone_reset_threshold=-1 |&
418 tee -a "${logfile}.${test_number}" |
419 grep -q 'value out of range' || return $?
420}
421
422test19() {
423 run_fio_on_seq --zone_reset_threshold=2 |&
424 tee -a "${logfile}.${test_number}" |
425 grep -q 'value out of range' || return $?
426}
427
428test20() {
429 run_fio_on_seq --zone_reset_threshold=.4:.6 |&
430 tee -a "${logfile}.${test_number}" |
431 grep -q 'the list exceeding max length' || return $?
432}
433
434test21() {
435 run_fio_on_seq --zone_reset_frequency=-1 |&
436 tee -a "${logfile}.${test_number}" |
437 grep -q 'value out of range' || return $?
438}
439
440test22() {
441 run_fio_on_seq --zone_reset_frequency=2 |&
442 tee -a "${logfile}.${test_number}" |
443 grep -q 'value out of range' || return $?
444}
445
446test23() {
447 run_fio_on_seq --zone_reset_frequency=.4:.6 |&
448 tee -a "${logfile}.${test_number}" |
449 grep -q 'the list exceeding max length' || return $?
450}
451
452test24() {
453 local bs loops=9 size=$((zone_size))
454
455 bs=$(min $((256*1024)) "$zone_size")
6dcb098d
DF
456 run_fio_on_seq "$(ioengine "psync")" --rw=write --bs="$bs" \
457 --size=$size --loops=$loops \
191d1d1a
BVA
458 --zone_reset_frequency=.01 --zone_reset_threshold=.90 \
459 >> "${logfile}.${test_number}" 2>&1 || return $?
460 check_written $((size * loops)) || return $?
461 check_reset_count -eq 8 ||
462 check_reset_count -eq 9 ||
463 check_reset_count -eq 10 || return $?
464}
465
466# Multiple non-overlapping sequential write jobs for the same drive.
467test25() {
468 local i opts=()
469
470 for ((i=0;i<16;i++)); do
471 [ -n "$is_zbd" ] &&
472 reset_zone "$dev" $((first_sequential_zone_sector + i*sectors_per_zone))
473 done
474 for ((i=0;i<16;i++)); do
475 opts+=("--name=job$i" "--filename=$dev" "--thread=1" "--direct=1")
476 opts+=("--offset=$((first_sequential_zone_sector*512 + zone_size*i))")
6dcb098d 477 opts+=("--size=$zone_size" "$(ioengine "psync")" "--rw=write" "--bs=16K")
191d1d1a 478 opts+=("--zonemode=zbd" "--zonesize=${zone_size}" "--group_reporting=1")
6dcb098d 479 opts+=(${var_opts[@]})
191d1d1a
BVA
480 done
481 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
482}
483
484write_to_first_seq_zone() {
485 local loops=4 r
486
487 r=$(((RANDOM << 16) | RANDOM))
6dcb098d 488 run_fio --name="$dev" --filename="$dev" "$(ioengine "psync")" --rw="$1" \
191d1d1a
BVA
489 --thread=1 --do_verify=1 --verify=md5 --direct=1 --bs=4K \
490 --offset=$((first_sequential_zone_sector * 512)) \
491 "--size=$zone_size" --loops=$loops --randseed="$r" \
492 --zonemode=zbd --zonesize="${zone_size}" --group_reporting=1 \
493 --gtod_reduce=1 >> "${logfile}.${test_number}" 2>&1 || return $?
494 check_written $((loops * zone_size)) || return $?
495}
496
497# Overwrite the first sequential zone four times sequentially.
498test26() {
499 write_to_first_seq_zone write
500}
501
502# Overwrite the first sequential zone four times using random writes.
503test27() {
504 write_to_first_seq_zone randwrite
505}
506
507# Multiple overlapping random write jobs for the same drive.
508test28() {
509 local i jobs=16 off opts
510
511 off=$((first_sequential_zone_sector * 512 + 64 * zone_size))
512 [ -n "$is_zbd" ] && reset_zone "$dev" $((off / 512))
513 opts=("--debug=zbd")
514 for ((i=0;i<jobs;i++)); do
515 opts+=("--name=job$i" "--filename=$dev" "--offset=$off" "--bs=16K")
6dcb098d 516 opts+=("--size=$zone_size" "$(ioengine "psync")" "--rw=randwrite")
191d1d1a
BVA
517 opts+=("--thread=1" "--direct=1" "--zonemode=zbd")
518 opts+=("--zonesize=${zone_size}" "--group_reporting=1")
6dcb098d 519 opts+=(${var_opts[@]})
191d1d1a
BVA
520 done
521 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
522 check_written $((jobs * zone_size)) || return $?
523 check_reset_count -eq $jobs ||
524 check_reset_count -eq $((jobs - 1)) ||
525 return $?
526}
527
528# Multiple overlapping random write jobs for the same drive and with a limited
529# number of open zones.
530test29() {
531 local i jobs=16 off opts=()
532
533 off=$((first_sequential_zone_sector * 512 + 64 * zone_size))
534 size=$((16*zone_size))
535 [ -n "$is_zbd" ] && reset_zone "$dev" $((off / 512))
536 opts=("--debug=zbd")
537 for ((i=0;i<jobs;i++)); do
538 opts+=("--name=job$i" "--filename=$dev" "--offset=$off" "--bs=16K")
539 opts+=("--size=$size" "--io_size=$zone_size" "--thread=1")
6dcb098d 540 opts+=("$(ioengine "psync")" "--rw=randwrite" "--direct=1")
191d1d1a
BVA
541 opts+=("--max_open_zones=4" "--group_reporting=1")
542 opts+=("--zonemode=zbd" "--zonesize=${zone_size}")
6dcb098d 543 opts+=(${var_opts[@]})
191d1d1a
BVA
544 done
545 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
546 check_written $((jobs * zone_size)) || return $?
547}
548
549# Random reads and writes across the entire disk for 30s.
550test30() {
551 local off
552
553 off=$((first_sequential_zone_sector * 512))
6dcb098d 554 run_one_fio_job "$(ioengine "libaio")" --iodepth=8 --rw=randrw \
191d1d1a
BVA
555 --bs="$(max $((zone_size / 128)) "$logical_block_size")"\
556 --zonemode=zbd --zonesize="${zone_size}" --offset=$off\
557 --loops=2 --time_based --runtime=30s --norandommap=1\
558 >>"${logfile}.${test_number}" 2>&1
559}
560
561# Random reads across all sequential zones for 30s. This is not only a fio
562# test but also allows to verify the performance of a drive.
563test31() {
564 local bs inc nz off opts size
565
566 # Start with writing 128 KB to 128 sequential zones.
567 bs=128K
568 nz=128
569 # shellcheck disable=SC2017
570 inc=$(((disk_size - (first_sequential_zone_sector * 512)) / (nz * zone_size)
571 * zone_size))
572 opts=()
573 for ((off = first_sequential_zone_sector * 512; off < disk_size;
574 off += inc)); do
575 opts+=("--name=$dev" "--filename=$dev" "--offset=$off" "--io_size=$bs")
6dcb098d 576 opts+=("--bs=$bs" "--size=$zone_size" "$(ioengine "libaio")")
191d1d1a
BVA
577 opts+=("--rw=write" "--direct=1" "--thread=1" "--stats=0")
578 opts+=("--zonemode=zbd" "--zonesize=${zone_size}")
6dcb098d 579 opts+=(${var_opts[@]})
191d1d1a
BVA
580 done
581 "$(dirname "$0")/../../fio" "${opts[@]}" >> "${logfile}.${test_number}" 2>&1
582 # Next, run the test.
583 off=$((first_sequential_zone_sector * 512))
584 size=$((disk_size - off))
585 opts=("--name=$dev" "--filename=$dev" "--offset=$off" "--size=$size")
6dcb098d 586 opts+=("--bs=$bs" "$(ioengine "psync")" "--rw=randread" "--direct=1")
191d1d1a
BVA
587 opts+=("--thread=1" "--time_based" "--runtime=30" "--zonemode=zbd")
588 opts+=("--zonesize=${zone_size}")
589 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
590}
591
592# Random writes across all sequential zones. This is not only a fio test but
593# also allows to verify the performance of a drive.
594test32() {
595 local off opts=() size
596
597 off=$((first_sequential_zone_sector * 512))
598 size=$((disk_size - off))
599 opts+=("--name=$dev" "--filename=$dev" "--offset=$off" "--size=$size")
6dcb098d 600 opts+=("--bs=128K" "$(ioengine "psync")" "--rw=randwrite" "--direct=1")
191d1d1a
BVA
601 opts+=("--thread=1" "--time_based" "--runtime=30")
602 opts+=("--max_open_zones=$max_open_zones" "--zonemode=zbd")
603 opts+=("--zonesize=${zone_size}")
604 run_fio "${opts[@]}" >> "${logfile}.${test_number}" 2>&1 || return $?
605}
606
607# Write to sequential zones with a block size that is not a divisor of the
608# zone size.
609test33() {
610 local bs io_size size
611
612 size=$((2 * zone_size))
613 io_size=$((5 * zone_size))
614 bs=$((3 * zone_size / 4))
6dcb098d
DF
615 run_fio_on_seq "$(ioengine "psync")" --iodepth=1 --rw=write \
616 --size=$size --io_size=$io_size --bs=$bs \
191d1d1a
BVA
617 >> "${logfile}.${test_number}" 2>&1 || return $?
618 check_written $(((io_size + bs - 1) / bs * bs)) || return $?
619}
620
621# Write to sequential zones with a block size that is not a divisor of the
622# zone size and with data verification enabled.
623test34() {
624 local size
625
626 size=$((2 * zone_size))
6dcb098d 627 run_fio_on_seq "$(ioengine "psync")" --iodepth=1 --rw=write --size=$size \
191d1d1a
BVA
628 --do_verify=1 --verify=md5 --bs=$((3 * zone_size / 4)) \
629 >> "${logfile}.${test_number}" 2>&1 && return 1
630 grep -q 'not a divisor of' "${logfile}.${test_number}"
631}
632
633# Test 1/4 for the I/O boundary rounding code: $size < $zone_size.
634test35() {
635 local bs off io_size size
636
637 off=$(((first_sequential_zone_sector + 1) * 512))
638 size=$((zone_size - 2 * 512))
639 bs=$((zone_size / 4))
6dcb098d
DF
640 run_one_fio_job --offset=$off --size=$size "$(ioengine "psync")" \
641 --iodepth=1 --rw=write --do_verify=1 --verify=md5 \
642 --bs=$bs --zonemode=zbd --zonesize="${zone_size}" \
191d1d1a
BVA
643 >> "${logfile}.${test_number}" 2>&1 && return 1
644 grep -q 'io_size must be at least one zone' "${logfile}.${test_number}"
645}
646
647# Test 2/4 for the I/O boundary rounding code: $size < $zone_size.
648test36() {
649 local bs off io_size size
650
651 off=$(((first_sequential_zone_sector) * 512))
652 size=$((zone_size - 512))
653 bs=$((zone_size / 4))
6dcb098d
DF
654 run_one_fio_job --offset=$off --size=$size "$(ioengine "psync")" \
655 --iodepth=1 --rw=write --do_verify=1 --verify=md5 \
656 --bs=$bs --zonemode=zbd --zonesize="${zone_size}" \
191d1d1a
BVA
657 >> "${logfile}.${test_number}" 2>&1 && return 1
658 grep -q 'io_size must be at least one zone' "${logfile}.${test_number}"
659}
660
661# Test 3/4 for the I/O boundary rounding code: $size > $zone_size.
662test37() {
663 local bs off size
664
665 if [ "$first_sequential_zone_sector" = 0 ]; then
666 off=0
667 else
668 off=$(((first_sequential_zone_sector - 1) * 512))
669 fi
670 size=$((zone_size + 2 * 512))
671 bs=$((zone_size / 4))
6dcb098d
DF
672 run_one_fio_job --offset=$off --size=$size "$(ioengine "psync")" \
673 --iodepth=1 --rw=write --do_verify=1 --verify=md5 \
674 --bs=$bs --zonemode=zbd --zonesize="${zone_size}" \
191d1d1a
BVA
675 >> "${logfile}.${test_number}" 2>&1
676 check_written $((zone_size)) || return $?
677}
678
679# Test 4/4 for the I/O boundary rounding code: $offset > $disk_size - $zone_size
680test38() {
681 local bs off size
682
683 size=$((logical_block_size))
684 off=$((disk_size - logical_block_size))
685 bs=$((logical_block_size))
6dcb098d
DF
686 run_one_fio_job --offset=$off --size=$size "$(ioengine "psync")" \
687 --iodepth=1 --rw=write --do_verify=1 --verify=md5 \
688 --bs=$bs --zonemode=zbd --zonesize="${zone_size}" \
191d1d1a
BVA
689 >> "${logfile}.${test_number}" 2>&1 && return 1
690 grep -q 'io_size must be at least one zone' "${logfile}.${test_number}"
691}
692
693# Read one block from a block device.
694read_one_block() {
695 local bs
696
697 bs=$((logical_block_size))
6dcb098d 698 run_one_fio_job --rw=read "$(ioengine "psync")" --bs=$bs --size=$bs "$@" 2>&1 |
191d1d1a
BVA
699 tee -a "${logfile}.${test_number}"
700}
701
702# Check whether fio accepts --zonemode=none for zoned block devices.
703test39() {
704 [ -n "$is_zbd" ] || return 0
705 read_one_block --zonemode=none >/dev/null || return $?
706 check_read $((logical_block_size)) || return $?
707}
708
709# Check whether fio accepts --zonemode=strided for zoned block devices.
710test40() {
711 local bs
712
713 bs=$((logical_block_size))
714 [ -n "$is_zbd" ] || return 0
715 read_one_block --zonemode=strided |
716 grep -q 'fio: --zonesize must be specified when using --zonemode=strided' ||
717 return $?
718 read_one_block --zonemode=strided --zonesize=$bs >/dev/null || return $?
719 check_read $bs || return $?
720}
721
722# Check whether fio checks the zone size for zoned block devices.
723test41() {
724 [ -n "$is_zbd" ] || return 0
725 read_one_block --zonemode=zbd --zonesize=$((2 * zone_size)) |
726 grep -q 'job parameter zonesize.*does not match disk zone size'
727}
728
729# Check whether fio handles --zonesize=0 correctly for regular block devices.
730test42() {
731 [ -n "$is_zbd" ] && return 0
732 read_one_block --zonemode=zbd --zonesize=0 |
733 grep -q 'Specifying the zone size is mandatory for regular block devices with --zonemode=zbd'
734}
735
a4b7f12b 736# Check whether fio handles --zonesize=1 correctly for regular block devices.
191d1d1a 737test43() {
a4b7f12b 738 [ -n "$is_zbd" ] && return 0
191d1d1a
BVA
739 read_one_block --zonemode=zbd --zonesize=1 |
740 grep -q 'zone size must be at least 512 bytes for --zonemode=zbd'
741}
742
743# Check whether fio handles --zonemode=none --zonesize=1 correctly.
744test44() {
745 read_one_block --zonemode=none --zonesize=1 |
746 grep -q 'fio: --zonemode=none and --zonesize are not compatible'
747}
748
749test45() {
750 local bs i
751
752 [ -z "$is_zbd" ] && return 0
753 bs=$((logical_block_size))
6dcb098d 754 run_one_fio_job "$(ioengine "psync")" --iodepth=1 --rw=randwrite --bs=$bs\
191d1d1a
BVA
755 --offset=$((first_sequential_zone_sector * 512)) \
756 --size="$zone_size" --do_verify=1 --verify=md5 2>&1 |
757 tee -a "${logfile}.${test_number}" |
758 grep -q "fio: first I/O failed. If .* is a zoned block device, consider --zonemode=zbd"
759}
760
bc596cbc
DLM
761# Random write to sequential zones, libaio, 8 jobs, queue depth 64 per job
762test46() {
763 local size
764
765 size=$((4 * zone_size))
6dcb098d 766 run_fio_on_seq "$(ioengine "libaio")" --iodepth=64 --rw=randwrite --bs=4K \
bc596cbc
DLM
767 --group_reporting=1 --numjobs=8 \
768 >> "${logfile}.${test_number}" 2>&1 || return $?
769 check_written $((size * 8)) || return $?
770}
771
4d37720a
DLM
772# Check whether fio handles --zonemode=zbd --zoneskip=1 correctly.
773test47() {
774 local bs
775
4d37720a 776 bs=$((logical_block_size))
e0424151 777 run_fio_on_seq "$(ioengine "psync")" --rw=write --bs=$bs --zoneskip=1 \
4d37720a
DLM
778 >> "${logfile}.${test_number}" 2>&1 && return 1
779 grep -q 'zoneskip 1 is not a multiple of the device zone size' "${logfile}.${test_number}"
780}
781
3bd2078b
NA
782# Multiple overlapping random write jobs for the same drive and with a
783# limited number of open zones. This is similar to test29, but uses libaio
784# to stress test zone locking.
785test48() {
786 local i jobs=16 off opts=()
787
788 off=$((first_sequential_zone_sector * 512 + 64 * zone_size))
789 size=$((16*zone_size))
790 [ -n "$is_zbd" ] && reset_zone "$dev" $((off / 512))
791 opts=("--aux-path=/tmp" "--allow_file_create=0" "--significant_figures=10")
792 opts+=("--debug=zbd")
6dcb098d 793 opts+=("$(ioengine "libaio")" "--rw=randwrite" "--direct=1")
3bd2078b
NA
794 opts+=("--time_based" "--runtime=30")
795 opts+=("--zonemode=zbd" "--zonesize=${zone_size}")
796 opts+=("--max_open_zones=4")
797 for ((i=0;i<jobs;i++)); do
798 opts+=("--name=job$i" "--filename=$dev" "--offset=$off" "--bs=16K")
799 opts+=("--io_size=$zone_size" "--iodepth=256" "--thread=1")
800 opts+=("--group_reporting=1")
801 done
802
803 fio=$(dirname "$0")/../../fio
804
805 { echo; echo "fio ${opts[*]}"; echo; } >>"${logfile}.${test_number}"
806
807 timeout -v -s KILL 45s \
808 "${dynamic_analyzer[@]}" "$fio" "${opts[@]}" \
809 >> "${logfile}.${test_number}" 2>&1 || return $?
810}
811
191d1d1a
BVA
812tests=()
813dynamic_analyzer=()
814reset_all_zones=
6dcb098d
DF
815use_libzbc=
816zbd_debug=
191d1d1a
BVA
817
818while [ "${1#-}" != "$1" ]; do
819 case "$1" in
820 -d) dynamic_analyzer=(valgrind "--read-var-info=yes" "--tool=drd"
821 "--show-confl-seg=no");
822 shift;;
823 -e) dynamic_analyzer=(valgrind "--read-var-info=yes" "--tool=helgrind");
824 shift;;
6dcb098d 825 -l) use_libzbc=1; shift;;
191d1d1a
BVA
826 -r) reset_all_zones=1; shift;;
827 -t) tests+=("$2"); shift; shift;;
828 -v) dynamic_analyzer=(valgrind "--read-var-info=yes");
829 shift;;
6dcb098d 830 -z) zbd_debug=1; shift;;
191d1d1a
BVA
831 --) shift; break;;
832 esac
833done
834
835if [ $# != 1 ]; then
836 usage
837 exit 1
838fi
839
840# shellcheck source=functions
841source "$(dirname "$0")/functions" || exit $?
842
6dcb098d
DF
843var_opts=()
844if [ -n "$zbd_debug" ]; then
845 var_opts+=("--debug=zbd")
846fi
191d1d1a
BVA
847dev=$1
848realdev=$(readlink -f "$dev")
849basename=$(basename "$realdev")
6dcb098d
DF
850
851if [[ -b "$realdev" ]]; then
852 major=$((0x$(stat -L -c '%t' "$realdev"))) || exit $?
853 minor=$((0x$(stat -L -c '%T' "$realdev"))) || exit $?
854 disk_size=$(($(<"/sys/dev/block/$major:$minor/size")*512))
855
856 # When the target is a partition device, get basename of its
857 # holder device to access sysfs path of the holder device
858 if [[ -r "/sys/dev/block/$major:$minor/partition" ]]; then
859 realsysfs=$(readlink "/sys/dev/block/$major:$minor")
860 basename=$(basename "${realsysfs%/*}")
861 fi
862 logical_block_size=$(<"/sys/block/$basename/queue/logical_block_size")
863 case "$(<"/sys/class/block/$basename/queue/zoned")" in
864 host-managed|host-aware)
865 is_zbd=true
552e214c
SK
866 if ! check_blkzone "${dev}"; then
867 exit 1
868 fi
6dcb098d
DF
869 if ! result=($(first_sequential_zone "$dev")); then
870 echo "Failed to determine first sequential zone"
871 exit 1
872 fi
873 first_sequential_zone_sector=${result[0]}
874 sectors_per_zone=${result[1]}
875 zone_size=$((sectors_per_zone * 512))
876 if ! max_open_zones=$(max_open_zones "$dev"); then
877 echo "Failed to determine maximum number of open zones"
878 exit 1
879 fi
880 set_io_scheduler "$basename" deadline || exit $?
881 if [ -n "$reset_all_zones" ]; then
882 reset_zone "$dev" -1
883 fi
884 ;;
885 *)
886 first_sequential_zone_sector=$(((disk_size / 2) &
887 (logical_block_size - 1)))
888 zone_size=$(max 65536 "$logical_block_size")
889 sectors_per_zone=$((zone_size / 512))
890 max_open_zones=128
891 set_io_scheduler "$basename" none || exit $?
892 ;;
893 esac
894elif [[ -c "$realdev" ]]; then
895 # For an SG node, we must have libzbc option specified
896 if [[ ! -n "$use_libzbc" ]]; then
897 echo "Character device files can only be used with -l (libzbc) option"
898 exit 1
899 fi
900
901 if ! $(is_zbc "$dev"); then
902 echo "Device is not a ZBC disk"
903 exit 1
904 fi
191d1d1a 905 is_zbd=true
6dcb098d
DF
906
907 if ! disk_size=($(( $(zbc_disk_sectors "$dev") * 512))); then
908 echo "Failed to determine disk size"
909 exit 1
910 fi
911 if ! logical_block_size=($(zbc_logical_block_size "$dev")); then
912 echo "Failed to determine logical block size"
913 exit 1
914 fi
191d1d1a 915 if ! result=($(first_sequential_zone "$dev")); then
6dcb098d
DF
916 echo "Failed to determine first sequential zone"
917 exit 1
191d1d1a
BVA
918 fi
919 first_sequential_zone_sector=${result[0]}
920 sectors_per_zone=${result[1]}
921 zone_size=$((sectors_per_zone * 512))
922 if ! max_open_zones=$(max_open_zones "$dev"); then
6dcb098d
DF
923 echo "Failed to determine maximum number of open zones"
924 exit 1
191d1d1a 925 fi
191d1d1a 926 if [ -n "$reset_all_zones" ]; then
6dcb098d 927 reset_zone "$dev" -1
191d1d1a 928 fi
6dcb098d
DF
929fi
930
931echo -n "First sequential zone starts at sector $first_sequential_zone_sector;"
932echo " zone size: $((zone_size >> 20)) MB"
191d1d1a
BVA
933
934if [ "${#tests[@]}" = 0 ]; then
106a71cd
SK
935 readarray -t tests < <(declare -F | grep "test[0-9]*" | \
936 tr -c -d "[:digit:]\n" | sort -n)
191d1d1a
BVA
937fi
938
939logfile=$0.log
940
72176dc7
BVA
941passed=0
942failed=0
332fa178
DF
943if [ -t 1 ]; then
944 red="\e[1;31m"
945 green="\e[1;32m"
946 end="\e[m"
947else
948 red=""
949 green=""
950 end=""
951fi
191d1d1a 952rc=0
332fa178 953
c96b385b
DF
954intr=0
955trap 'intr=1' SIGINT
956
191d1d1a
BVA
957for test_number in "${tests[@]}"; do
958 rm -f "${logfile}.${test_number}"
332fa178 959 echo -n "Running test $(printf "%02d" $test_number) ... "
191d1d1a
BVA
960 if eval "test$test_number"; then
961 status="PASS"
332fa178 962 cc_status="${green}${status}${end}"
72176dc7 963 ((passed++))
191d1d1a
BVA
964 else
965 status="FAIL"
332fa178 966 cc_status="${red}${status}${end}"
72176dc7 967 ((failed++))
191d1d1a
BVA
968 rc=1
969 fi
332fa178 970 echo -e "$cc_status"
191d1d1a 971 echo "$status" >> "${logfile}.${test_number}"
c96b385b 972 [ $intr -ne 0 ] && exit 1
191d1d1a
BVA
973done
974
72176dc7
BVA
975echo "$passed tests passed"
976if [ $failed -gt 0 ]; then
977 echo " and $failed tests failed"
978fi
191d1d1a 979exit $rc