6 ALWAYS_EXCEPT="52 60 $RECOVERY_SMALL_EXCEPT"
8 export MULTIOP=${MULTIOP:-multiop}
9 PTLDEBUG=${PTLDEBUG:--1}
10 LUSTRE=${LUSTRE:-`dirname $0`/..}
11 . $LUSTRE/tests/test-framework.sh
13 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
16 require_dsh_mds || exit 0
18 # also long tests: 19, 21a, 21e, 21f, 23, 27
19 # 1 2.5 2.5 4 4 (min)"
20 [ "$SLOW" = "no" ] && EXCEPT_SLOW="17 26a 26b 50 51 57"
22 [ $(facet_fstype $SINGLEMDS) = "zfs" ] &&
23 # bug number for skipped test: LU-2194 LU-2547
24 ALWAYS_EXCEPT="$ALWAYS_EXCEPT 19b 24a 24b"
28 # Allow us to override the setup if we already have a mounted system by
29 # setting SETUP=" " and CLEANUP=" "
31 CLEANUP=${CLEANUP:-""}
33 check_and_setup_lustre
36 rm -rf $DIR/d[0-9]* $DIR/f.${TESTSUITE}*
39 local f1="$DIR/$tfile"
40 local f2="$DIR/$tfile.2"
42 drop_request "mcreate $f1" ||
43 error_noexit "create '$f1': drop req"
45 drop_reint_reply "mcreate $f2" ||
46 error_noexit "create '$f2': drop rep"
48 drop_request "tchmod 111 $f2" ||
49 error_noexit "chmod '$f2': drop req"
51 drop_reint_reply "tchmod 666 $f2" ||
52 error_noexit "chmod '$f2': drop rep"
54 drop_request "statone $f2" ||
55 error_noexit "stat '$f2': drop req"
57 drop_reply "statone $f2" ||
58 error_noexit "stat '$f2': drop rep"
60 run_test 1 "create, chmod, stat: drop req, drop rep"
64 do_facet_create_file client $t 10K ||
65 error_noexit "Create file $t"
67 drop_request "cat $t > /dev/null" ||
68 error_noexit "Open request for $t file"
70 drop_reply "cat $t > /dev/null" ||
71 error_noexit "Open replay for $t file"
73 run_test 4 "open: drop req, drop rep"
78 local RR="$T-renamed-again"
79 do_facet_create_file client $T 10K ||
80 error_noexit "Create file $T"
82 drop_request "mv $T $R" ||
83 error_noexit "Rename $T"
85 drop_reint_reply "mv $R $RR" ||
86 error_noexit "Failed rename replay on $R"
88 do_facet client "checkstat -v $RR" ||
89 error_noexit "checkstat error on $RR"
91 do_facet client "rm $RR" ||
92 error_noexit "Can't remove file $RR"
94 run_test 5 "rename: drop req, drop rep"
98 local LINK1=$DIR/$tfile.link1
99 local LINK2=$DIR/$tfile.link2
101 do_facet_create_file client $T 10K ||
102 error_noexit "Create file $T"
104 drop_request "mlink $T $LINK1" ||
105 error_noexit "mlink request for $T"
107 drop_reint_reply "mlink $T $LINK2" ||
108 error_noexit "mlink reply for $T"
110 drop_request "munlink $LINK1" ||
111 error_noexit "munlink request for $T"
113 drop_reint_reply "munlink $LINK2" ||
114 error_noexit "munlink reply for $T"
116 do_facet client "rm $T" ||
117 error_noexit "Can't remove file $T"
119 run_test 6 "link, unlink: drop req, drop rep"
123 drop_reint_reply "touch $DIR/$tfile" || return 1
125 run_test 8 "touch: drop rep (bug 1423)"
129 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
133 do_facet_random_file client $TMP/$tfile 1K ||
134 error_noexit "Create random file $TMP/$tfile"
135 # make this big, else test 9 doesn't wait for bulk -- bz 5595
136 do_facet_create_file client $TMP/$t1 4M ||
137 error_noexit "Create file $TMP/$t1"
138 do_facet client "cp $TMP/$t1 $DIR/$t1" ||
139 error_noexit "Can't copy to $DIR/$t1 file"
140 pause_bulk "cp $TMP/$tfile $DIR/$tfile" ||
141 error_noexit "Can't pause_bulk copy"
142 do_facet client "cp $TMP/$t1 $DIR/$t2" ||
143 error_noexit "Can't copy file"
144 do_facet client "sync"
145 do_facet client "rm $DIR/$tfile $DIR/$t2 $DIR/$t1" ||
146 error_noexit "Can't remove files"
147 do_facet client "rm $TMP/$t1 $TMP/$tfile"
149 run_test 9 "pause bulk on OST (bug 1420)"
153 local before=$(date +%s)
156 do_facet client "stat $DIR > /dev/null" ||
157 error "failed to stat $DIR: $?"
158 drop_bl_callback "chmod 0777 $DIR" ||
159 error "failed to chmod $DIR: $?"
161 # let the client reconnect
163 evict=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
164 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
165 [ ! -z "$evict" ] && [[ $evict -gt $before ]] ||
166 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
167 error "no eviction: $evict before:$before")
169 do_facet client checkstat -v -p 0777 $DIR ||
170 error "client checkstat failed: $?"
172 run_test 10a "finish request on server after client eviction (bug 1521)"
175 local before=$(date +%s)
178 do_facet client "stat $DIR > /dev/null" ||
179 error "failed to stat $DIR: $?"
180 drop_bl_callback_once "chmod 0777 $DIR" ||
181 error "failed to chmod $DIR: $?"
183 # let the client reconnect
185 evict=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
186 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
188 [ -z "$evict" ] || [[ $evict -le $before ]] ||
189 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
190 error "eviction happened: $evict before:$before")
192 do_facet client checkstat -v -p 0777 $DIR ||
193 error "client checkstat failed: $?"
195 run_test 10b "re-send BL AST"
198 # wake up a thread waiting for completion after eviction
200 do_facet client $MULTIOP $DIR/$tfile Ow ||
201 { error "multiop write failed: $?"; return 1; }
202 do_facet client $MULTIOP $DIR/$tfile or ||
203 { error "multiop read failed: $?"; return 2; }
207 do_facet client $MULTIOP $DIR/$tfile or ||
208 { error "multiop read failed: $?"; return 3; }
209 drop_bl_callback_once $MULTIOP $DIR/$tfile Ow ||
210 echo "evicted as expected"
212 do_facet client munlink $DIR/$tfile ||
213 { error "munlink failed: $?"; return 4; }
214 # allow recovery to complete
215 client_up || client_up || sleep $TIMEOUT
217 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
221 $LCTL mark $MULTIOP $DIR/$tfile OS_c
222 do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
223 clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
224 multiop_bg_pause $DIR/$tfile OS_c ||
225 { error "multiop failed: $?"; return 1; }
227 #define OBD_FAIL_MDS_CLOSE_NET 0x115
229 echo "waiting for multiop $PID"
230 wait $PID || { error "wait for multiop faile: $?"; return 2; }
231 do_facet client munlink $DIR/$tfile ||
232 { error "client munlink failed: $?"; return 3; }
233 # allow recovery to complete
234 client_up || client_up || sleep $TIMEOUT
236 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
238 # Bug 113, check that readdir lost recv timeout works.
240 mkdir -p $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
241 touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
242 # OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
243 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
244 ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
245 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
246 rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
248 run_test 13 "mdc_readpage restart test (bug 1138)"
250 # Bug 113, check that readdir lost send timeout works.
253 touch $DIR/$tdir/newentry
254 # OBD_FAIL_MDS_SENDPAGE|OBD_FAIL_ONCE
255 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000106"
256 ls $DIR/$tdir || return 1
257 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
259 run_test 14 "mdc_readpage resend test (bug 1138)"
262 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
263 touch $DIR/$tfile && return 1
266 run_test 15 "failed open (-ENOMEM)"
268 READ_AHEAD=`lctl get_param -n llite.*.max_read_ahead_mb | head -n 1`
270 lctl set_param -n llite.*.max_read_ahead_mb 0
274 lctl set_param -n llite.*.max_read_ahead_mb $READ_AHEAD
278 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
280 do_facet_random_file client $TMP/$tfile 100K ||
281 { error_noexit "Create random file $TMP/$T" ; return 0; }
282 do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
283 { error_noexit "Copy to $DIR/$tfile file" ; return 0; }
287 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
288 do_facet ost1 "lctl set_param fail_loc=0x80000504"
290 # OST bulk will time out here, client resends
291 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
292 do_facet ost1 lctl set_param fail_loc=0
293 # give recovery a chance to finish (shouldn't take long)
295 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
299 run_test 16 "timeout bulk put, don't evict client (2732)"
304 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
306 local SAMPLE_FILE=$TMP/$tfile
307 do_facet_random_file client $SAMPLE_FILE 20K ||
308 { error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
310 # With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
311 if at_is_enabled; then
312 at_max_saved=$(at_max_get ost1)
313 at_max_set $TIMEOUT ost1
316 # OBD_FAIL_PTLRPC_BULK_GET_NET 0x0503 | OBD_FAIL_ONCE
317 # OST bulk will time out here, client retries
318 do_facet ost1 lctl set_param fail_loc=0x80000503
319 # need to ensure we send an RPC
320 do_facet client cp $SAMPLE_FILE $DIR/$tfile
323 # with AT, client will wait adaptive_max*factor+net_latency before
324 # expiring the req, hopefully timeout*2 is enough
325 sleep $(($TIMEOUT*2))
327 do_facet ost1 lctl set_param fail_loc=0
328 do_facet client "df $DIR"
329 # expect cmp to succeed, client resent bulk
330 do_facet client "cmp $SAMPLE_FILE $DIR/$tfile" || return 3
331 do_facet client "rm $DIR/$tfile" || return 4
332 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
335 run_test 17 "timeout bulk get, don't evict client (2732)"
338 [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
340 do_facet_create_file client $TMP/$tfile 20K ||
341 { error_noexit "Create file $TMP/$tfile" ; return 0; }
343 do_facet client mkdir -p $DIR/$tdir
347 pgcache_empty || return 1
350 $LFS setstripe -i 1 -c 1 $f
351 stripe_index=$($LFS getstripe -i $f)
352 if [ $stripe_index -ne 1 ]; then
354 error "$f: stripe_index $stripe_index != 1" && return
357 do_facet client cp $TMP/$tfile $f
359 local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
360 $LCTL --device $osc2dev deactivate || return 3
361 # my understanding is that there should be nothing in the page
362 # cache after the client reconnects?
364 pgcache_empty || rc=2
365 $LCTL --device $osc2dev activate
369 run_test 18a "manual ost invalidate clears page cache immediately"
372 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
374 do_facet_create_file client $TMP/$tfile 20K ||
375 { error_noexit "Create file $TMP/$tfile" ; return 0; }
377 do_facet client mkdir -p $DIR/$tdir
381 pgcache_empty || return 1
383 $LFS setstripe -i 0 -c 1 $f
384 stripe_index=$($LFS getstripe -i $f)
385 if [ $stripe_index -ne 0 ]; then
387 error "$f: stripe_index $stripe_index != 0" && return
390 do_facet client cp $TMP/$tfile $f
393 # allow recovery to complete
394 sleep $((TIMEOUT + 2))
395 # my understanding is that there should be nothing in the page
396 # cache after the client reconnects?
398 pgcache_empty || rc=2
402 run_test 18b "eviction and reconnect clears page cache (2766)"
405 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
407 do_facet_create_file client $TMP/$tfile 20K ||
408 { error_noexit "Create file $TMP/$tfile" ; return 0; }
410 do_facet client mkdir -p $DIR/$tdir
414 pgcache_empty || return 1
416 $LFS setstripe -i 0 -c 1 $f
417 stripe_index=$($LFS getstripe -i $f)
418 if [ $stripe_index -ne 0 ]; then
420 error "$f: stripe_index $stripe_index != 0" && return
423 do_facet client cp $TMP/$tfile $f
427 # OBD_FAIL_OST_CONNECT_NET2
428 # lost reply to connect request
429 do_facet ost1 lctl set_param fail_loc=0x80000225
432 df $MOUNT > /dev/null 2>&1
434 # my understanding is that there should be nothing in the page
435 # cache after the client reconnects?
437 pgcache_empty || rc=2
441 run_test 18c "Dropped connect reply after eviction handing (14755)"
444 local BEFORE=`date +%s`
447 mount_client $DIR2 || error "failed to mount $DIR2"
449 # cancel cached locks from OST to avoid eviction from it
452 do_facet client "stat $DIR > /dev/null" ||
453 error "failed to stat $DIR: $?"
454 drop_ldlm_cancel "chmod 0777 $DIR2" ||
455 error "failed to chmod $DIR2"
459 # let the client reconnect
461 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
462 awk -F"[ [,]" '/EVICTED ]$/ \
463 { if (mx<$5) {mx=$5;} } END { print mx }')
465 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
466 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
467 error "no eviction: $EVICT before:$BEFORE")
469 run_test 19a "test expired_lock_main on mds (2867)"
472 local BEFORE=`date +%s`
475 mount_client $DIR2 || error "failed to mount $DIR2: $?"
477 # cancel cached locks from MDT to avoid eviction from it
480 do_facet client $MULTIOP $DIR/$tfile Ow ||
481 error "failed to run multiop: $?"
482 drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow ||
483 error "failed to ldlm_cancel: $?"
485 umount_client $DIR2 || error "failed to unmount $DIR2: $?"
486 do_facet client munlink $DIR/$tfile ||
487 error "failed to unlink $DIR/$tfile: $?"
489 # let the client reconnect
491 EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state |
492 awk -F"[ [,]" '/EVICTED ]$/ \
493 { if (mx < $5) {mx = $5;} } END { print mx }')
495 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
496 (do_facet client $LCTL get_param osc.$FSNAME-OST*.state;
497 error "no eviction: $EVICT before:$BEFORE")
499 run_test 19b "test expired_lock_main on ost (2867)"
502 local BEFORE=`date +%s`
505 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0
507 mkdir -p $DIR1/$tfile
510 #define OBD_FAIL_PTLRPC_CANCEL_RESEND 0x516
511 do_facet mds $LCTL set_param fail_loc=0x80000516
513 touch $DIR2/$tfile/file1 &
515 # let touch to get blocked on the server
519 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1
522 # let the client reconnect
524 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
525 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
527 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
529 run_test 19c "check reconnect and lock resend do not trigger expired_lock_main"
531 test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
532 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
535 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
536 multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
539 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
540 do_facet ost1 lctl set_param fail_loc=0x80000308
541 kill -USR1 $MULTI_PID
544 [ $rc -eq 0 ] && error "multiop didn't fail enqueue: rc $rc" || true
546 run_test 20a "ldlm_handle_enqueue error (should return error)"
548 test_20b() { # bug 2986 - ldlm_handle_enqueue error during open
549 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
552 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
554 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
555 do_facet ost1 lctl set_param fail_loc=0x80000308
556 dd if=/etc/hosts of=$DIR/$tdir/$tfile && \
557 error "didn't fail open enqueue" || true
559 run_test 20b "ldlm_handle_enqueue error (should return error)"
562 mkdir -p $DIR/$tdir-1
563 mkdir -p $DIR/$tdir-2
564 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
567 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
568 $MULTIOP $DIR/$tdir-2/f Oc &
571 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
573 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
574 kill -USR1 $close_pid
576 wait $close_pid || return 1
577 wait $open_pid || return 2
578 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
580 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
581 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
585 run_test 21a "drop close request while close and open are both in flight"
588 mkdir -p $DIR/$tdir-1
589 mkdir -p $DIR/$tdir-2
590 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
593 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
594 mcreate $DIR/$tdir-2/f &
597 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
599 kill -USR1 $close_pid
601 wait $close_pid || return 1
602 wait $open_pid || return 3
604 $CHECKSTAT -t file $DIR/$tdir-1/f || return 4
605 $CHECKSTAT -t file $DIR/$tdir-2/f || return 5
608 run_test 21b "drop open request while close and open are both in flight"
611 mkdir -p $DIR/$tdir-1
612 mkdir -p $DIR/$tdir-2
613 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
616 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
617 mcreate $DIR/$tdir-2/f &
620 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
622 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
623 kill -USR1 $close_pid
625 wait $close_pid || return 1
626 wait $open_pid || return 2
628 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
630 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
631 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
634 run_test 21c "drop both request while close and open are both in flight"
637 mkdir -p $DIR/$tdir-1
638 mkdir -p $DIR/$tdir-2
639 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
642 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
643 $MULTIOP $DIR/$tdir-2/f Oc &
645 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
647 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
650 wait $pid || return 1
651 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
653 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
654 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
658 run_test 21d "drop close reply while close and open are both in flight"
661 mkdir -p $DIR/$tdir-1
662 mkdir -p $DIR/$tdir-2
663 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
666 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
667 touch $DIR/$tdir-2/f &
669 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
673 wait $pid || return 1
676 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
677 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
680 run_test 21e "drop open reply while close and open are both in flight"
683 mkdir -p $DIR/$tdir-1
684 mkdir -p $DIR/$tdir-2
685 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
688 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
689 touch $DIR/$tdir-2/f &
691 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
693 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
696 wait $pid || return 1
697 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
699 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
700 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
703 run_test 21f "drop both reply while close and open are both in flight"
706 mkdir -p $DIR/$tdir-1
707 mkdir -p $DIR/$tdir-2
708 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
711 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
712 touch $DIR/$tdir-2/f &
714 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
716 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
719 wait $pid || return 1
720 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
722 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
723 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
726 run_test 21g "drop open reply and close request while close and open are both in flight"
729 mkdir -p $DIR/$tdir-1
730 mkdir -p $DIR/$tdir-2
731 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
734 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
735 touch $DIR/$tdir-2/f &
738 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
740 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
743 wait $pid || return 1
744 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
746 wait $touch_pid || return 2
748 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
749 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
752 run_test 21h "drop open request and close reply while close and open are both in flight"
754 # bug 3462 - multiple MDC requests
759 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
764 $MULTIOP $f1 msu || return 1
767 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
769 wait $close_pid || return 2
770 rm -rf $f2 || return 4
772 run_test 22 "drop close request and do mknod"
775 multiop_bg_pause $DIR/$tfile O_c || return 1
777 # give a chance for open
781 drop_request "kill -USR1 $pid"
784 wait $pid || return 1
787 run_test 23 "client hang when close a file after mds crash"
789 test_24a() { # bug 11710 details correct fsync() behavior
790 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
793 $LFS setstripe -i 0 -c 1 $DIR/$tdir
795 multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
798 kill -USR1 $MULTI_PID
801 lctl set_param fail_loc=0x0
804 error_ignore bz5494 "multiop didn't fail fsync: rc $rc" || true
806 run_test 24a "fsync error (should return error)"
808 wait_client_evicted () {
811 local varsvc=${facet}_svc
813 wait_update $(facet_active_host $facet) \
814 "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
819 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
823 lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
825 multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
826 error "mulitop Ow8192_yc failed"
829 multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
830 error "mulitop Ow8192_c failed"
835 kill -USR1 $MULTI_PID1
838 kill -USR1 $MULTI_PID2
841 lctl set_param fail_loc=0x0
843 [ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
844 error_ignore bz5494 "multiop didn't fail fsync: $rc1 or close: $rc2" ||
847 dmesg | grep "dirty page discard:" ||
848 error "no discarded dirty page found!"
850 run_test 24b "test dirty page discard due to client eviction"
852 test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
853 # this test can only run from a client on a separate node.
854 remote_ost || { skip "local OST" && return 0; }
855 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
856 remote_mds || { skip "local MDS" && return 0; }
858 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
859 skip "msg and ost1 are at the same node"
863 check_timeout || return 1
865 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
867 echo starting with $OST_NEXP OST exports
868 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
869 do_facet client lctl set_param fail_loc=0x505
870 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
871 # But if there's a race to start the evictor from various obds,
872 # the loser might have to wait for the next ping.
875 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
877 do_facet client lctl set_param fail_loc=0x0
878 [ $rc -eq 0 ] || error "client not evicted from OST"
880 run_test 26a "evict dead exports"
882 test_26b() { # bug 10140 - evict dead exports by pinger
883 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
885 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
886 skip "msg and ost1 are at the same node"
890 check_timeout || return 1
892 zconf_mount `hostname` $MOUNT2 ||
893 { error "Failed to mount $MOUNT2"; return 2; }
894 sleep 1 # wait connections being established
896 local MDS_NEXP=$(do_facet $SINGLEMDS lctl get_param -n mdt.${mds1_svc}.num_exports | cut -d' ' -f2)
897 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
899 echo starting with $OST_NEXP OST and $MDS_NEXP MDS exports
901 zconf_umount `hostname` $MOUNT2 -f
903 # PING_INTERVAL max(obd_timeout / 4, 1U)
904 # PING_EVICT_TIMEOUT (PING_INTERVAL * 6)
906 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
907 # But if there's a race to start the evictor from various obds,
908 # the loser might have to wait for the next ping.
909 # = 9 * PING_INTERVAL + PING_INTERVAL
910 # = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
911 # let's wait $((TIMEOUT * 3)) # bug 19887
913 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
914 error "Client was not evicted by ost" rc=1
915 wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
916 error "Client was not evicted by mds"
918 run_test 26b "evict dead exports"
922 writemany -q -a $DIR/$tdir/$tfile 0 5 &
925 local save_FAILURE_MODE=$FAILURE_MODE
927 facet_failover $SINGLEMDS
928 #define OBD_FAIL_OSC_SHUTDOWN 0x407
929 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
930 # need to wait for reconnect
931 echo waiting for fail_loc
932 wait_update_facet $SINGLEMDS "lctl get_param -n fail_loc" "-2147482617"
933 facet_failover $SINGLEMDS
935 kill -USR1 $CLIENT_PID
938 FAILURE_MODE=$save_FAILURE_MODE
940 run_test 27 "fail LOV while using OSC's"
942 test_28() { # bug 6086 - error adding new clients
943 do_facet client mcreate $DIR/$tfile || return 1
944 drop_bl_callback_once "chmod 0777 $DIR/$tfile" ||
945 echo "evicted as expected"
946 #define OBD_FAIL_MDS_CLIENT_ADD 0x12f
947 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012f"
948 # fail once (evicted), reconnect fail (fail_loc), ok
949 client_up || (sleep 10; client_up) || (sleep 10; client_up) || error "reconnect failed"
951 fail $SINGLEMDS # verify MDS last_rcvd can be loaded
953 run_test 28 "handle error adding new clients (bug 6086)"
955 test_29a() { # bug 22273 - error adding new clients
956 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
957 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000711"
958 # fail abort so client will be new again
959 fail_abort $SINGLEMDS
960 client_up || error "reconnect failed"
961 wait_osc_import_state $SINGLEMDS ost FULL
964 run_test 29a "error adding new clients doesn't cause LBUG (bug 22273)"
966 test_29b() { # bug 22273 - error adding new clients
967 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
968 do_facet ost1 "lctl set_param fail_loc=0x80000711"
969 # fail abort so client will be new again
971 client_up || error "reconnect failed"
974 run_test 29b "error adding new clients doesn't cause LBUG (bug 22273)"
978 # put a load of file creates/writes/deletes
979 writemany -q $DIR/$tdir/$tfile 0 5 &
981 echo writemany pid $CLIENT_PID
985 # wait for client to reconnect to MDS
990 # client process should see no problems even though MDS went down
992 kill -USR1 $CLIENT_PID
995 echo writemany returned $rc
996 #these may fail because of eviction due to slow AST response.
998 error_ignore bz13652 "writemany returned rc $rc" || true
1000 run_test 50 "failover MDS under load"
1003 #define OBD_FAIL_MDS_SYNC_CAPA_SL 0x1310
1004 do_facet ost1 lctl set_param fail_loc=0x00001310
1007 # put a load of file creates/writes/deletes
1008 writemany -q $DIR/$tdir/$tfile 0 5 &
1012 facet_failover $SINGLEMDS
1013 # failover at various points during recovery
1014 SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
1015 echo will failover at $SEQ
1018 #echo failover in $i sec
1019 log "test_$testnum: failover in $i sec"
1021 facet_failover $SINGLEMDS
1023 # client process should see no problems even though MDS went down
1024 # and recovery was interrupted
1026 kill -USR1 $CLIENT_PID
1029 echo writemany returned $rc
1031 error_ignore bz13652 "writemany returned rc $rc" || true
1033 run_test 51 "failover MDS during recovery"
1036 do_facet client "mkdir -p $DIR/$tdir"
1037 do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
1039 echo writemany pid $CLIENT_PID
1044 wait $CLIENT_PID || rc=$?
1045 # active client process should see an EIO for down OST
1046 [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }
1047 # but timing or failover setup may allow success
1048 [ $rc -eq 0 ] && { echo "writemany succeeded" && return 0; }
1049 echo "writemany returned $rc"
1054 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1059 [ $rc -ne 0 ] && { return $rc; }
1060 # wait for client to reconnect to OST
1064 [ $rc -ne 0 ] && { return $rc; }
1071 run_test 52 "failover OST under load"
1073 # test of open reconstruct
1076 drop_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||\
1079 run_test 53 "touch: drop rep"
1082 zconf_mount `hostname` $MOUNT2
1084 touch $DIR2/$tfile.1
1086 cat $DIR2/$tfile.missing # save transno = 0, rc != 0 into last_rcvd
1089 ERROR=`dmesg | egrep "(test 54|went back in time)" | tail -n1 | grep "went back in time"`
1090 [ x"$ERROR" == x ] || error "back in time occured"
1092 run_test 54 "back in time"
1094 # bug 11330 - liblustre application death during I/O locks up OST
1096 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1100 # Minimum pass speed is 2MBps
1102 # LU-2887/LU-3089 - set min pass speed to 500KBps
1103 [ "$(facet_fstype ost1)" = "zfs" ] && ddtimeout=256
1105 # first dd should be finished quickly
1106 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
1107 dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
1110 echo "step1: testing ......"
1111 while kill -0 $DDPID 2> /dev/null; do
1113 if [ $count -gt $ddtimeout ]; then
1114 error "dd should be finished!"
1118 echo "(dd_pid=$DDPID, time=$count)successful"
1120 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
1121 #define OBD_FAIL_OST_DROP_REQ 0x21d
1122 do_facet ost1 lctl set_param fail_loc=0x0000021d
1123 # second dd will be never finished
1124 dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
1127 echo "step2: testing ......"
1128 while [ $count -le $ddtimeout ]; do
1129 if ! kill -0 $DDPID 2> /dev/null; then
1131 error "dd shouldn't be finished! (time=$count)"
1136 echo "(dd_pid=$DDPID, time=$count)successful"
1138 #Recover fail_loc and dd will finish soon
1139 do_facet ost1 lctl set_param fail_loc=0
1141 echo "step3: testing ......"
1142 while kill -0 $DDPID 2> /dev/null; do
1144 if [ $count -gt $((ddtimeout + 440)) ]; then
1145 error "dd should be finished!"
1149 echo "(dd_pid=$DDPID, time=$count)successful"
1153 run_test 55 "ost_brw_read/write drops timed-out read/write request"
1155 test_56() { # b=11277
1156 #define OBD_FAIL_MDS_RESEND 0x136
1158 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000136"
1159 stat $DIR/$tfile || error "stat failed"
1160 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1163 run_test 56 "do not fail on getattr resend"
1166 # no oscs means no client or mdt
1167 while lctl get_param osc.*.* > /dev/null 2>&1; do
1168 : # loop until proc file is removed
1172 test_57() { # bug 10866
1176 #define OBD_FAIL_LPROC_REMOVE 0xB00
1177 lctl set_param fail_loc=0x80000B00
1178 zconf_umount `hostname` $DIR
1179 lctl set_param fail_loc=0x80000B00
1180 fail_abort $SINGLEMDS
1182 lctl set_param fail_loc=0
1184 do_facet client "df $DIR"
1186 run_test 57 "read procfs entries causes kernel crash"
1188 test_58() { # bug 11546
1189 #define OBD_FAIL_MDC_ENQUEUE_PAUSE 0x801
1192 lctl set_param fail_loc=0x80000801
1193 cp $DIR/$tfile /dev/null &
1196 lctl set_param fail_loc=0
1197 drop_bl_callback_once rm -f $DIR/$tfile
1199 # the first 'df' could tigger the eviction caused by
1200 # 'drop_bl_callback_once', and it's normal case.
1201 # but the next 'df' should return successfully.
1202 do_facet client "df $DIR" || do_facet client "df $DIR"
1204 run_test 58 "Eviction in the middle of open RPC reply processing"
1206 test_59() { # bug 10589
1207 zconf_mount `hostname` $MOUNT2 || error "Failed to mount $MOUNT2"
1208 echo $DIR2 | grep -q $MOUNT2 || error "DIR2 is not set properly: $DIR2"
1209 #define OBD_FAIL_LDLM_CANCEL_EVICT_RACE 0x311
1210 lctl set_param fail_loc=0x311
1211 writes=$(LANG=C dd if=/dev/zero of=$DIR2/$tfile count=1 2>&1)
1212 [ $? = 0 ] || error "dd write failed"
1213 writes=$(echo $writes | awk -F '+' '/out/ {print $1}')
1214 lctl set_param fail_loc=0
1216 zconf_umount `hostname` $MOUNT2 -f
1217 reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1)
1218 [ $? = 0 ] || error "dd read failed"
1219 reads=$(echo $reads | awk -F '+' '/in/ {print $1}')
1220 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes
1222 run_test 59 "Read cancel race on client eviction"
1225 # we assume that all md changes are in the MDT0 changelog
1226 if [ $MDSCOUNT -gt 1 ]; then
1227 error_ignore bz17935 $*
1234 MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid |
1235 awk '{ gsub(/_UUID/,""); print $1 }' | head -n1)
1240 # Register (and start) changelog
1241 USER=$(do_facet $SINGLEMDS lctl --device $MDT0 changelog_register -n)
1242 echo "Registered as $MDT0 changelog user $USER"
1244 # Generate a large number of changelog entries
1245 createmany -o $DIR/$tdir/$tfile $NUM_FILES
1249 # Unlink files in the background
1250 unlinkmany $DIR/$tdir/$tfile $NUM_FILES &
1254 # Failover the MDS while unlinks are happening
1255 facet_failover $SINGLEMDS
1257 # Wait for unlinkmany to finish
1260 # Check if all the create/unlink events were recorded
1262 $LFS changelog $MDT0 >> $DIR/$tdir/changelog
1263 local cl_count=$(grep UNLNK $DIR/$tdir/changelog | wc -l)
1264 echo "$cl_count unlinks in $MDT0 changelog"
1266 do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $USER
1267 USERS=$(( $(do_facet $SINGLEMDS lctl get_param -n \
1268 mdd.$MDT0.changelog_users | wc -l) - 2 ))
1269 if [ $USERS -eq 0 ]; then
1270 [ $cl_count -eq $NUM_FILES ] || \
1271 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1272 # Also make sure we can clear large changelogs
1273 cl_count=$($LFS changelog $FSNAME | wc -l)
1274 [ $cl_count -le 2 ] || \
1275 error "Changelog not empty: $cl_count entries"
1277 # If there are other users, there may be other unlinks in the log
1278 [ $cl_count -ge $NUM_FILES ] || \
1279 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1280 echo "$USERS other changelog users; can't verify clear"
1283 run_test 60 "Add Changelog entries during MDS failover"
1287 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000)
1288 mdtosc=${mdtosc/-MDT*/-MDT\*}
1289 local cflags="osc.$mdtosc.connect_flags"
1290 do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
1291 [ $? -ne 0 ] && skip "don't have skip orphan feature" && return
1293 mkdir -p $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
1294 # Set the default stripe of $DIR/$tdir to put the files to ost1
1295 $LFS setstripe -c 1 -i 0 $DIR/$tdir
1297 replay_barrier $SINGLEMDS
1298 createmany -o $DIR/$tdir/$tfile-%d 10
1299 local oid=`do_facet ost1 "lctl get_param -n obdfilter.${ost1_svc}.last_id"`
1301 fail_abort $SINGLEMDS
1303 touch $DIR/$tdir/$tfile
1304 local id=`$LFS getstripe $DIR/$tdir/$tfile | awk '$1 == 0 { print $2 }'`
1305 [ $id -le $oid ] && error "the orphan objid was reused, failed"
1310 run_test 61 "Verify to not reuse orphan objects - bug 17025"
1312 # test_62 as seen it b2_1 please do not reuse test_62
1315 # zconf_umount `hostname` $DIR
1316 # #define OBD_FAIL_PTLRPC_DELAY_IMP_FULL 0x516
1317 # lctl set_param fail_loc=0x516
1320 #run_test 62 "Verify connection flags race - bug LU-1716"
1322 check_cli_ir_state()
1324 local NODE=${1:-$HOSTNAME}
1326 st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
1327 awk '/imperative_recovery:/ { print \\\$2}'")
1328 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1329 error "Error state $st, must be ENABLED or DISABLED"
1333 check_target_ir_state()
1336 local name=${target}_svc
1337 local recovery_proc=obdfilter.${!name}.recovery_status
1340 st=$(do_facet $target "lctl get_param -n $recovery_proc |
1341 awk '/IR:/{ print \\\$2}'")
1342 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1343 error "Error state $st, must be ENABLED or DISABLED"
1349 do_facet mgs lctl set_param -n mgs.MGS.live.$FSNAME="state=$1"
1354 local state=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1355 awk '/state:/{ print \\\$2 }'")
1359 nidtbl_version_mgs()
1361 local ver=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1362 awk '/nidtbl_version:/{ print \\\$2 }'")
1366 # nidtbl_version_client <mds1|client> [node]
1367 nidtbl_version_client()
1370 local node=${2:-$HOSTNAME}
1372 if [ X$cli = Xclient ]; then
1375 local obdtype=${cli/%[0-9]*/}
1376 [ $obdtype != mds ] && error "wrong parameters $cli"
1378 node=$(facet_active_host $cli)
1383 local vers=$(do_node $node "lctl get_param -n mgc.*.ir_state" |
1384 awk "/$cli/{print \$6}" |sort -u)
1386 # in case there are multiple mounts on the client node
1388 [ ${#arr[@]} -ne 1 ] && error "versions on client node mismatch"
1392 nidtbl_versions_match()
1394 [ $(nidtbl_version_mgs) -eq $(nidtbl_version_client ${1:-client}) ]
1397 target_instance_match()
1403 obdtype=${srv/%[0-9]*/}
1414 error "invalid target type" $srv
1419 local target=${srv}_svc
1420 local si=$(do_facet $srv lctl get_param -n $obdname.${!target}.instance)
1421 local ci=$(lctl get_param -n $cliname.${!target}-${cliname}-*.import |
1422 awk '/instance/{ print $2 }' | head -n1)
1424 return $([ $si -eq $ci ])
1429 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1430 { skip "MGS without IR support"; return 0; }
1432 # MDT was just restarted in the previous test, make sure everything
1435 while [ $cnt -gt 0 ]; do
1436 nidtbl_versions_match && break
1442 set_ir_status disabled
1444 local prev_ver=$(nidtbl_version_client client)
1446 local saved_FAILURE_MODE=$FAILURE_MODE
1447 [ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
1451 [ $(nidtbl_version_client client) -eq $prev_ver ] ||
1452 error "version must not change due to IR disabled"
1453 target_instance_match ost1 || error "instance mismatch"
1457 FAILURE_MODE=$saved_FAILURE_MODE
1459 run_test 100 "IR: Make sure normal recovery still works w/o IR"
1463 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1464 { skip "MGS without IR support"; return 0; }
1468 local OST1_IMP=$(get_osc_import_name client ost1)
1470 # disable pinger recovery
1471 lctl set_param -n osc.$OST1_IMP.pinger_recov=0
1475 target_instance_match ost1 || error "instance mismatch"
1476 nidtbl_versions_match || error "version must match"
1478 lctl set_param -n osc.$OST1_IMP.pinger_recov=1
1480 run_test 101 "IR: Make sure IR works w/o normal recovery"
1484 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1485 { skip "MGS without IR support"; return 0; }
1487 local clients=${CLIENTS:-$HOSTNAME}
1494 # let's have a new nidtbl version
1497 # sleep for a while so that clients can see the failure of ost
1498 # it must be MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC.
1499 # int mgc_request.c:
1500 # define MGC_TIMEOUT_MIN_SECONDS 5
1501 # define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 *
1502 local count=30 # 20 seconds at most
1503 while [ $count -gt 0 ]; do
1504 nidtbl_versions_match && break
1509 nidtbl_versions_match || error "nidtbl mismatch"
1512 old_version=$(nidtbl_version_client client)
1514 zconf_umount_clients $clients $MOUNT || error "Cannot umount client"
1517 combined_mgs_mds && mgsdev=mds1
1518 remount_facet $mgsdev
1521 zconf_mount_clients $clients $MOUNT || error "Cannot mount client"
1524 new_version=$(nidtbl_version_client client)
1525 [ $new_version -lt $old_version ] &&
1526 error "nidtbl version wrong after mgs restarts"
1529 run_test 102 "IR: New client gets updated nidtbl after MGS restart"
1533 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1534 { skip "MGS without IR support"; return 0; }
1536 combined_mgs_mds && skip "mgs and mds on the same target" && return 0
1538 # workaround solution to generate config log on the mds
1544 # We need this test because mds is like a client in IR context.
1545 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS ||
1546 error "MDS should start w/o mgs"
1548 # start mgs and remount mds w/ ir
1549 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
1552 # remount client so that fsdb will be created on the MGS
1553 umount_client $MOUNT || error "umount failed"
1554 mount_client $MOUNT || error "mount failed"
1556 # sleep 30 seconds so the MDS has a chance to detect MGS restarting
1558 while [ $count -gt 0 ]; do
1559 [ $(nidtbl_version_client mds1) -ne 0 ] && break
1564 # after a while, mds should be able to reconnect to mgs and fetch
1565 # up-to-date nidtbl version
1566 nidtbl_versions_match mds1 || error "mds nidtbl mismatch"
1571 run_test 103 "IR: MDS can start w/o MGS and get updated nidtbl later"
1575 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1576 { skip "MGS without IR support"; return 0; }
1581 start ost1 $(ostdevname 1) "$OST_MOUNT_OPTS -onoir" ||
1582 error "OST1 cannot start"
1585 local ir_state=$(check_target_ir_state ost1)
1586 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1587 error "ir status on ost1 should be DISABLED"
1589 run_test 104 "IR: ost can disable IR voluntarily"
1593 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
1594 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1595 { skip "MGS without IR support"; return 0; }
1599 # get one of the clients from client list
1600 local rcli=$(echo $RCLIENTS |cut -d' ' -f 1)
1602 local mount_opts=${MOUNT_OPTS:+$MOUNT_OPTS,}noir
1603 zconf_umount $rcli $MOUNT || error "umount failed"
1604 zconf_mount $rcli $MOUNT $mount_opts || error "mount failed"
1606 # make sure lustre mount at $rcli disabling IR
1607 local ir_state=$(check_cli_ir_state $rcli)
1608 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1609 error "IR state must be DISABLED at $rcli"
1611 # Since the client just mounted, its last_rcvd entry is not on disk.
1612 # Send an RPC so exp_need_sync forces last_rcvd to commit this export
1613 # so the client can reconnect during OST recovery (LU-924, LU-1582)
1614 $SETSTRIPE -i 0 $DIR/$tfile
1615 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
1617 # make sure MGS's state is Partial
1618 [ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
1621 # make sure IR on ost1 is DISABLED
1622 local ir_state=$(check_target_ir_state ost1)
1623 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1624 error "IR status on ost1 should be DISABLED"
1626 # remount with the default MOUNT_OPTS
1627 zconf_umount $rcli $MOUNT || error "umount failed"
1628 zconf_mount $rcli $MOUNT || error "mount failed"
1630 # make sure MGS's state is full
1631 [ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
1634 # make sure IR on ost1 is ENABLED
1635 local ir_state=$(check_target_ir_state ost1)
1636 [ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
1637 error "IR status on ost1 should be ENABLED"
1641 run_test 105 "IR: NON IR clients support"
1649 test_106() { # LU-1789
1650 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.50) ]] ||
1651 { skip "Need MDS version at least 2.3.50"; return 0; }
1653 #define OBD_FAIL_MDC_LIGHTWEIGHT 0x805
1654 $LCTL set_param fail_loc=0x805
1657 trap cleanup_106 EXIT
1659 # enable lightweight flag on mdc connection
1662 local MDS_NEXP=$(do_facet $SINGLEMDS \
1663 lctl get_param -n mdt.${mds1_svc}.num_exports |
1665 $LCTL set_param fail_loc=0
1667 touch $DIR2/$tfile || error "failed to create empty file"
1668 replay_barrier $SINGLEMDS
1670 $LCTL set_param debug=console
1672 facet_failover $SINGLEMDS
1674 # lightweight connection must be evicted
1675 touch -c $DIR2/$tfile || true
1676 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1677 evicted=`awk '/This client was evicted by .*MDT0000/ {
1679 }' $TMP/lustre-log-$TESTNAME.log`
1680 [ -z "$evicted" ] && error "lightweight client not evicted by mds"
1682 # and all operations performed by lightweight client should be
1683 # synchronous, so the file created before mds restart should be there
1684 $CHECKSTAT -t file $DIR/$tfile || error "file not present"
1689 run_test 106 "lightweight connection support"
1696 # OBD_FAIL_MDS_REINT_NET_REP 0x119
1697 do_facet $SINGLEMDS lctl set_param fail_loc=0x119
1698 multiop $DIR/$tdir D_c &
1700 mkdir $DIR/$tdir/dir_106 &
1702 do_facet $SINGLEMDS lctl set_param fail_loc=0
1705 wait $CLIENT_PID || rc=$?
1706 checkstat -t dir $DIR/$tdir/dir_106 || return 1
1708 kill -USR1 $close_pid
1709 wait $close_pid || return 2
1713 run_test 107 "drop reint reply, then restart MDT"
1716 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1717 local remote_dir=$DIR/$tdir/remote_dir
1721 #prepare for 110 test, which need set striped dir on remote MDT.
1722 for num in $(seq $MDSCOUNT); do
1724 lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
1729 drop_request "$LFS mkdir -i $MDTIDX -c2 $remote_dir" ||
1730 error "lfs mkdir failed"
1731 local diridx=$($GETSTRIPE -M $remote_dir)
1732 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
1734 rm -rf $DIR/$tdir || error "rmdir failed"
1736 run_test 110a "create remote directory: drop client req"
1739 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1740 local remote_dir=$DIR/$tdir/remote_dir
1744 drop_reint_reply "$LFS mkdir -i $MDTIDX -c2 $remote_dir" ||
1745 error "lfs mkdir failed"
1747 diridx=$($GETSTRIPE -M $remote_dir)
1748 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
1750 rm -rf $DIR/$tdir || error "rmdir failed"
1752 run_test 110b "create remote directory: drop Master rep"
1755 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1756 local remote_dir=$DIR/$tdir/remote_dir
1760 drop_update_reply $MDTIDX "$LFS mkdir -i $MDTIDX -c2 $remote_dir" ||
1761 error "lfs mkdir failed"
1763 diridx=$($GETSTRIPE -M $remote_dir)
1764 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
1766 rm -rf $DIR/$tdir || error "rmdir failed"
1768 run_test 110c "create remote directory: drop update rep on slave MDT"
1771 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1772 local remote_dir=$DIR/$tdir/remote_dir
1776 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
1778 drop_request "rm -rf $remote_dir" || error "rm remote dir failed"
1780 rm -rf $DIR/$tdir || error "rmdir failed"
1784 run_test 110d "remove remote directory: drop client req"
1787 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1788 local remote_dir=$DIR/$tdir/remote_dir
1792 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
1793 drop_reint_reply "rm -rf $remote_dir" || error "rm remote dir failed"
1795 rm -rf $DIR/$tdir || error "rmdir failed"
1799 run_test 110e "remove remote directory: drop master rep"
1802 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1803 local remote_dir=$DIR/$tdir/remote_dir
1807 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
1808 drop_update_reply $MDTIDX "rm -rf $remote_dir" ||
1809 error "rm remote dir failed"
1811 rm -rf $DIR/$tdir || error "rmdir failed"
1813 run_test 110f "remove remote directory: drop slave rep"
1816 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1817 local remote_dir=$DIR/$tdir/remote_dir
1820 mkdir -p $remote_dir
1822 createmany -o $remote_dir/f 100
1824 #define OBD_FAIL_MIGRATE_NET_REP 0x1702
1825 do_facet mds$MDTIDX lctl set_param fail_loc=0x1702
1826 $LFS mv -M $MDTIDX $remote_dir || error "migrate failed"
1827 do_facet mds$MDTIDX lctl set_param fail_loc=0x0
1829 for file in $(find $remote_dir); do
1830 mdt_index=$($LFS getstripe -M $file)
1831 [ $mdt_index == $MDTIDX ] ||
1832 error "$file is not on MDT${MDTIDX}"
1835 rm -rf $DIR/$tdir || error "rmdir failed"
1837 run_test 110g "drop reply during migration"
1839 # LU-2844 mdt prepare fail should not cause umount oops
1842 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.62) ]] ||
1843 { skip "Need MDS version at least 2.3.62"; return 0; }
1845 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
1846 #define OBD_FAIL_MDS_CHANGELOG_INIT 0x151
1847 do_facet $SINGLEMDS lctl set_param fail_loc=0x151
1848 stop $SINGLEMDS || error "stop MDS failed"
1849 start $SINGLEMDS $mdsdev && error "start MDS should fail"
1850 do_facet $SINGLEMDS lctl set_param fail_loc=0
1851 start $SINGLEMDS $mdsdev || error "start MDS failed"
1853 run_test 111 "mdd setup fail should not cause umount oops"
1857 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1859 do_facet_random_file client $TMP/$tfile 100K ||
1860 error_noexit "Create random file $TMP/$tfile"
1862 pause_bulk "cp $TMP/$tfile $DIR/$tfile" $TIMEOUT ||
1863 error_noexit "Can't pause_bulk copy"
1866 # expect cmp to succeed, client resent bulk
1867 cmp $TMP/$tfile $DIR/$tfile ||
1868 error_noexit "Wrong data has been written"
1870 error_noexit "Can't remove file"
1873 run_test 112a "bulk resend while orignal request is in progress"
1875 # parameters: fail_loc CMD RC
1882 #define OBD_FAIL_LDLM_CP_CB_WAIT2 0x320
1883 #define OBD_FAIL_LDLM_CP_CB_WAIT3 0x321
1884 #define OBD_FAIL_LDLM_CP_CB_WAIT4 0x322
1885 #define OBD_FAIL_LDLM_CP_CB_WAIT5 0x323
1888 echo -n "** FLOCK REPLY vs. EVICTION race, lock $2"
1889 [ "$1" = "CLEANUP" ] &&
1890 fail=0x80000320 && echo ", $1 cp first"
1891 [ "$1" = "REPLY" ] &&
1892 fail=0x80000321 && echo ", $1 cp first"
1893 [ "$1" = "DEADLOCK CLEANUP" ] &&
1894 fail=0x80000322 && echo " DEADLOCK, CLEANUP cp first"
1895 [ "$1" = "DEADLOCK REPLY" ] &&
1896 fail=0x80000323 && echo " DEADLOCK, REPLY cp first"
1898 if [ x"$2" = x"get" ]; then
1899 #for TEST lock, take a conflict in advance
1900 # sleep longer than evictor to not confuse fail_loc: 2+2+4
1901 echo "** Taking conflict **"
1902 flocks_test 5 set read sleep 10 $DIR/$tfile &
1908 $LCTL set_param fail_loc=$fail
1910 flocks_test 5 $2 write $DIR/$tfile &
1914 echo "** Evicting and re-connecting client **"
1919 if [ x"$2" = x"get" ]; then
1926 # check if the return value is allowed
1927 [ $rc -eq $3 ] && rc=0
1929 $LCTL set_param fail_loc=0
1933 # a lock is taken, unlock vs. cleanup_resource() race for destroying
1934 # the ORIGINAL lock.
1939 flocks_test 5 set write sleep 4 $DIR/$tfile &
1943 # let unlock to sleep in CP CB
1944 $LCTL set_param fail_loc=$1
1947 # let cleanup to cleep in CP CB
1955 $LCTL set_param fail_loc=0
1960 flock_is_enabled || { skip "mount w/o flock enabled" && return; }
1963 test_120_reply "CLEANUP" set 5 || error "SET race failed"
1964 test_120_reply "CLEANUP" get 5 || error "GET race failed"
1965 test_120_reply "CLEANUP" unlock 5 || error "UNLOCK race failed"
1967 test_120_reply "REPLY" set 5 || error "SET race failed"
1968 test_120_reply "REPLY" get 5 || error "GET race failed"
1969 test_120_reply "REPLY" unlock 5 || error "UNLOCK race failed"
1972 test_120_reply "DEADLOCK CLEANUP" set 5 || error "DEADLOCK race failed"
1973 test_120_reply "DEADLOCK REPLY" set 35 || error "DEADLOCK race failed"
1975 test_120_destroy 0x320 || error "unlock-cleanup race failed"
1977 run_test 120 "flock race: completion vs. evict"
1980 local BEFORE=$(date +%s)
1983 # modify dir so that next revalidate would not obtain UPDATE lock
1986 # drop 1 reply with UPDATE lock,
1987 # resend should not create 2nd lock on server
1988 mcreate $DIR/$tfile || error "mcreate failed: $?"
1989 drop_ldlm_reply_once "stat $DIR/$tfile" || error "stat failed: $?"
1991 # 2 BL AST will be sent to client, both must find the same lock,
1992 # race them to not get EINVAL for 2nd BL AST
1993 #define OBD_FAIL_LDLM_PAUSE_CANCEL2 0x31f
1994 $LCTL set_param fail_loc=0x8000031f
1996 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0 > /dev/null
1997 chmod 0777 $DIR/$tfile || error "chmod failed: $?"
1998 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1 > /dev/null
2000 # let the client reconnect
2002 EVICT=$($LCTL get_param mdc.$FSNAME-MDT*.state |
2003 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
2005 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
2007 run_test 113 "ldlm enqueue dropped reply should not cause deadlocks"
2010 check_and_cleanup_lustre