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 do_facet client mcreate $DIR/$tfile ||
154 { error "mcreate failed: $?"; return 1; }
155 drop_bl_callback "chmod 0777 $DIR/$tfile" || echo "evicted as expected"
156 # wait for the mds to evict the client
157 #echo "sleep $(($TIMEOUT*2))"
158 #sleep $(($TIMEOUT*2))
159 do_facet client touch $DIR/$tfile || echo "touch failed, evicted"
160 do_facet client checkstat -v -p 0777 $DIR/$tfile ||
161 { error "client checkstat failed: $?"; return 3; }
162 do_facet client "munlink $DIR/$tfile"
163 # allow recovery to complete
164 client_up || client_up || sleep $TIMEOUT
166 run_test 10 "finish request on server after client eviction (bug 1521)"
169 # wake up a thread waiting for completion after eviction
171 do_facet client $MULTIOP $DIR/$tfile Ow ||
172 { error "multiop write failed: $?"; return 1; }
173 do_facet client $MULTIOP $DIR/$tfile or ||
174 { error "multiop read failed: $?"; return 2; }
178 do_facet client $MULTIOP $DIR/$tfile or ||
179 { error "multiop read failed: $?"; return 3; }
180 drop_bl_callback $MULTIOP $DIR/$tfile Ow || echo "evicted as expected"
182 do_facet client munlink $DIR/$tfile ||
183 { error "munlink failed: $?"; return 4; }
184 # allow recovery to complete
185 client_up || client_up || sleep $TIMEOUT
187 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
191 $LCTL mark $MULTIOP $DIR/$tfile OS_c
192 do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
193 clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
194 multiop_bg_pause $DIR/$tfile OS_c ||
195 { error "multiop failed: $?"; return 1; }
197 #define OBD_FAIL_MDS_CLOSE_NET 0x115
199 echo "waiting for multiop $PID"
200 wait $PID || { error "wait for multiop faile: $?"; return 2; }
201 do_facet client munlink $DIR/$tfile ||
202 { error "client munlink failed: $?"; return 3; }
203 # allow recovery to complete
204 client_up || client_up || sleep $TIMEOUT
206 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
208 # Bug 113, check that readdir lost recv timeout works.
210 mkdir -p $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
211 touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
212 # OBD_FAIL_MDS_READPAGE_NET|OBD_FAIL_ONCE
213 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
214 ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
215 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
216 rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
218 run_test 13 "mdc_readpage restart test (bug 1138)"
220 # Bug 113, check that readdir lost send timeout works.
223 touch $DIR/$tdir/newentry
224 # OBD_FAIL_MDS_SENDPAGE|OBD_FAIL_ONCE
225 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000106"
226 ls $DIR/$tdir || return 1
227 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
229 run_test 14 "mdc_readpage resend test (bug 1138)"
232 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
233 touch $DIR/$tfile && return 1
236 run_test 15 "failed open (-ENOMEM)"
238 READ_AHEAD=`lctl get_param -n llite.*.max_read_ahead_mb | head -n 1`
240 lctl set_param -n llite.*.max_read_ahead_mb 0
244 lctl set_param -n llite.*.max_read_ahead_mb $READ_AHEAD
248 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
250 do_facet_random_file client $TMP/$tfile 100K ||
251 { error_noexit "Create random file $TMP/$T" ; return 0; }
252 do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
253 { error_noexit "Copy to $DIR/$tfile file" ; return 0; }
257 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | OBD_FAIL_ONCE
258 do_facet ost1 "lctl set_param fail_loc=0x80000504"
260 # OST bulk will time out here, client resends
261 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
262 do_facet ost1 lctl set_param fail_loc=0
263 # give recovery a chance to finish (shouldn't take long)
265 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
269 run_test 16 "timeout bulk put, don't evict client (2732)"
274 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
276 local SAMPLE_FILE=$TMP/$tfile
277 do_facet_random_file client $SAMPLE_FILE 20K ||
278 { error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
280 # With adaptive timeouts, bulk_get won't expire until adaptive_timeout_max
281 if at_is_enabled; then
282 at_max_saved=$(at_max_get ost1)
283 at_max_set $TIMEOUT ost1
286 # OBD_FAIL_PTLRPC_BULK_GET_NET 0x0503 | OBD_FAIL_ONCE
287 # OST bulk will time out here, client retries
288 do_facet ost1 lctl set_param fail_loc=0x80000503
289 # need to ensure we send an RPC
290 do_facet client cp $SAMPLE_FILE $DIR/$tfile
293 # with AT, client will wait adaptive_max*factor+net_latency before
294 # expiring the req, hopefully timeout*2 is enough
295 sleep $(($TIMEOUT*2))
297 do_facet ost1 lctl set_param fail_loc=0
298 do_facet client "df $DIR"
299 # expect cmp to succeed, client resent bulk
300 do_facet client "cmp $SAMPLE_FILE $DIR/$tfile" || return 3
301 do_facet client "rm $DIR/$tfile" || return 4
302 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
305 run_test 17 "timeout bulk get, don't evict client (2732)"
308 [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
310 do_facet_create_file client $TMP/$tfile 20K ||
311 { error_noexit "Create file $TMP/$tfile" ; return 0; }
313 do_facet client mkdir -p $DIR/$tdir
317 pgcache_empty || return 1
320 $LFS setstripe -i 1 -c 1 $f
321 stripe_index=$($LFS getstripe -i $f)
322 if [ $stripe_index -ne 1 ]; then
324 error "$f: stripe_index $stripe_index != 1" && return
327 do_facet client cp $TMP/$tfile $f
329 local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
330 $LCTL --device $osc2dev deactivate || return 3
331 # my understanding is that there should be nothing in the page
332 # cache after the client reconnects?
334 pgcache_empty || rc=2
335 $LCTL --device $osc2dev activate
339 run_test 18a "manual ost invalidate clears page cache immediately"
342 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
344 do_facet_create_file client $TMP/$tfile 20K ||
345 { error_noexit "Create file $TMP/$tfile" ; return 0; }
347 do_facet client mkdir -p $DIR/$tdir
351 pgcache_empty || return 1
353 $LFS setstripe -i 0 -c 1 $f
354 stripe_index=$($LFS getstripe -i $f)
355 if [ $stripe_index -ne 0 ]; then
357 error "$f: stripe_index $stripe_index != 0" && return
360 do_facet client cp $TMP/$tfile $f
363 # allow recovery to complete
364 sleep $((TIMEOUT + 2))
365 # my understanding is that there should be nothing in the page
366 # cache after the client reconnects?
368 pgcache_empty || rc=2
372 run_test 18b "eviction and reconnect clears page cache (2766)"
375 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
377 do_facet_create_file client $TMP/$tfile 20K ||
378 { error_noexit "Create file $TMP/$tfile" ; return 0; }
380 do_facet client mkdir -p $DIR/$tdir
384 pgcache_empty || return 1
386 $LFS setstripe -i 0 -c 1 $f
387 stripe_index=$($LFS getstripe -i $f)
388 if [ $stripe_index -ne 0 ]; then
390 error "$f: stripe_index $stripe_index != 0" && return
393 do_facet client cp $TMP/$tfile $f
397 # OBD_FAIL_OST_CONNECT_NET2
398 # lost reply to connect request
399 do_facet ost1 lctl set_param fail_loc=0x80000225
402 df $MOUNT > /dev/null 2>&1
404 # my understanding is that there should be nothing in the page
405 # cache after the client reconnects?
407 pgcache_empty || rc=2
411 run_test 18c "Dropped connect reply after eviction handing (14755)"
414 local BEFORE=`date +%s`
417 mount_client $DIR2 || error "failed to mount $DIR2"
419 # cancel cached locks from OST to avoid eviction from it
422 do_facet client "stat $DIR > /dev/null" ||
423 error "failed to stat $DIR: $?"
424 drop_ldlm_cancel "chmod 0777 $DIR2" ||
425 error "failed to chmod $DIR2"
429 # let the client reconnect
431 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
432 awk -F"[ [,]" '/EVICTED ]$/ \
433 { if (mx<$5) {mx=$5;} } END { print mx }')
435 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
436 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
437 error "no eviction: $EVICT before:$BEFORE")
439 run_test 19a "test expired_lock_main on mds (2867)"
442 local BEFORE=`date +%s`
445 mount_client $DIR2 || error "failed to mount $DIR2: $?"
447 # cancel cached locks from MDT to avoid eviction from it
450 do_facet client $MULTIOP $DIR/$tfile Ow ||
451 error "failed to run multiop: $?"
452 drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow ||
453 error "failed to ldlm_cancel: $?"
455 umount_client $DIR2 || error "failed to unmount $DIR2: $?"
456 do_facet client munlink $DIR/$tfile ||
457 error "failed to unlink $DIR/$tfile: $?"
459 # let the client reconnect
461 EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state |
462 awk -F"[ [,]" '/EVICTED ]$/ \
463 { if (mx < $5) {mx = $5;} } END { print mx }')
465 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
466 (do_facet client $LCTL get_param osc.$FSNAME-OST*.state;
467 error "no eviction: $EVICT before:$BEFORE")
469 run_test 19b "test expired_lock_main on ost (2867)"
472 local BEFORE=`date +%s`
475 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0
477 mkdir -p $DIR1/$tfile
480 #define OBD_FAIL_PTLRPC_CANCEL_RESEND 0x516
481 do_facet mds $LCTL set_param fail_loc=0x80000516
483 touch $DIR2/$tfile/file1 &
485 # let touch to get blocked on the server
489 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1
492 # let the client reconnect
494 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
495 awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
497 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
499 run_test 19c "check reconnect and lock resend do not trigger expired_lock_main"
501 test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
502 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
505 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
506 multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
509 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
510 do_facet ost1 lctl set_param fail_loc=0x80000308
511 kill -USR1 $MULTI_PID
514 [ $rc -eq 0 ] && error "multiop didn't fail enqueue: rc $rc" || true
516 run_test 20a "ldlm_handle_enqueue error (should return error)"
518 test_20b() { # bug 2986 - ldlm_handle_enqueue error during open
519 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
522 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
524 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
525 do_facet ost1 lctl set_param fail_loc=0x80000308
526 dd if=/etc/hosts of=$DIR/$tdir/$tfile && \
527 error "didn't fail open enqueue" || true
529 run_test 20b "ldlm_handle_enqueue error (should return error)"
532 mkdir -p $DIR/$tdir-1
533 mkdir -p $DIR/$tdir-2
534 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
537 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
538 $MULTIOP $DIR/$tdir-2/f Oc &
541 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
543 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
544 kill -USR1 $close_pid
546 wait $close_pid || return 1
547 wait $open_pid || return 2
548 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
550 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
551 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
555 run_test 21a "drop close request while close and open are both in flight"
558 mkdir -p $DIR/$tdir-1
559 mkdir -p $DIR/$tdir-2
560 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
563 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
564 mcreate $DIR/$tdir-2/f &
567 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
569 kill -USR1 $close_pid
571 wait $close_pid || return 1
572 wait $open_pid || return 3
574 $CHECKSTAT -t file $DIR/$tdir-1/f || return 4
575 $CHECKSTAT -t file $DIR/$tdir-2/f || return 5
578 run_test 21b "drop open request while close and open are both in flight"
581 mkdir -p $DIR/$tdir-1
582 mkdir -p $DIR/$tdir-2
583 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
586 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
587 mcreate $DIR/$tdir-2/f &
590 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
592 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
593 kill -USR1 $close_pid
595 wait $close_pid || return 1
596 wait $open_pid || return 2
598 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
600 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
601 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
604 run_test 21c "drop both request while close and open are both in flight"
607 mkdir -p $DIR/$tdir-1
608 mkdir -p $DIR/$tdir-2
609 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
612 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
613 $MULTIOP $DIR/$tdir-2/f Oc &
615 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
617 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
620 wait $pid || return 1
621 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
623 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
624 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
628 run_test 21d "drop close reply while close and open are both in flight"
631 mkdir -p $DIR/$tdir-1
632 mkdir -p $DIR/$tdir-2
633 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
636 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
637 touch $DIR/$tdir-2/f &
639 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
643 wait $pid || return 1
646 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
647 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
650 run_test 21e "drop open reply while close and open are both in flight"
653 mkdir -p $DIR/$tdir-1
654 mkdir -p $DIR/$tdir-2
655 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
658 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
659 touch $DIR/$tdir-2/f &
661 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
663 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
666 wait $pid || return 1
667 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
669 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
670 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
673 run_test 21f "drop both reply while close and open are both in flight"
676 mkdir -p $DIR/$tdir-1
677 mkdir -p $DIR/$tdir-2
678 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
681 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
682 touch $DIR/$tdir-2/f &
684 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
686 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
689 wait $pid || return 1
690 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
692 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
693 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
696 run_test 21g "drop open reply and close request while close and open are both in flight"
699 mkdir -p $DIR/$tdir-1
700 mkdir -p $DIR/$tdir-2
701 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
704 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
705 touch $DIR/$tdir-2/f &
708 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
710 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
713 wait $pid || return 1
714 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
716 wait $touch_pid || return 2
718 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
719 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
722 run_test 21h "drop open request and close reply while close and open are both in flight"
724 # bug 3462 - multiple MDC requests
729 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
734 $MULTIOP $f1 msu || return 1
737 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
739 wait $close_pid || return 2
740 rm -rf $f2 || return 4
742 run_test 22 "drop close request and do mknod"
745 multiop_bg_pause $DIR/$tfile O_c || return 1
747 # give a chance for open
751 drop_request "kill -USR1 $pid"
754 wait $pid || return 1
757 run_test 23 "client hang when close a file after mds crash"
759 test_24a() { # bug 11710 details correct fsync() behavior
760 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
763 $LFS setstripe -i 0 -c 1 $DIR/$tdir
765 multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
768 kill -USR1 $MULTI_PID
771 lctl set_param fail_loc=0x0
774 error_ignore bz5494 "multiop didn't fail fsync: rc $rc" || true
776 run_test 24a "fsync error (should return error)"
778 wait_client_evicted () {
781 local varsvc=${facet}_svc
783 wait_update $(facet_active_host $facet) \
784 "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
789 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
793 lfs setstripe $DIR/$tdir -s 0 -i 0 -c 1
795 multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
796 error "mulitop Ow8192_yc failed"
799 multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
800 error "mulitop Ow8192_c failed"
805 kill -USR1 $MULTI_PID1
808 kill -USR1 $MULTI_PID2
811 lctl set_param fail_loc=0x0
813 [ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
814 error_ignore bz5494 "multiop didn't fail fsync: $rc1 or close: $rc2" ||
817 dmesg | grep "dirty page discard:" ||
818 error "no discarded dirty page found!"
820 run_test 24b "test dirty page discard due to client eviction"
822 test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
823 # this test can only run from a client on a separate node.
824 remote_ost || { skip "local OST" && return 0; }
825 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
826 remote_mds || { skip "local MDS" && return 0; }
828 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
829 skip "msg and ost1 are at the same node"
833 check_timeout || return 1
835 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
837 echo starting with $OST_NEXP OST exports
838 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
839 do_facet client lctl set_param fail_loc=0x505
840 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
841 # But if there's a race to start the evictor from various obds,
842 # the loser might have to wait for the next ping.
845 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
847 do_facet client lctl set_param fail_loc=0x0
848 [ $rc -eq 0 ] || error "client not evicted from OST"
850 run_test 26a "evict dead exports"
852 test_26b() { # bug 10140 - evict dead exports by pinger
853 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
855 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
856 skip "msg and ost1 are at the same node"
860 check_timeout || return 1
862 zconf_mount `hostname` $MOUNT2 ||
863 { error "Failed to mount $MOUNT2"; return 2; }
864 sleep 1 # wait connections being established
866 local MDS_NEXP=$(do_facet $SINGLEMDS lctl get_param -n mdt.${mds1_svc}.num_exports | cut -d' ' -f2)
867 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
869 echo starting with $OST_NEXP OST and $MDS_NEXP MDS exports
871 zconf_umount `hostname` $MOUNT2 -f
873 # PING_INTERVAL max(obd_timeout / 4, 1U)
874 # PING_EVICT_TIMEOUT (PING_INTERVAL * 6)
876 # evictor takes PING_EVICT_TIMEOUT + 3 * PING_INTERVAL to evict.
877 # But if there's a race to start the evictor from various obds,
878 # the loser might have to wait for the next ping.
879 # = 9 * PING_INTERVAL + PING_INTERVAL
880 # = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
881 # let's wait $((TIMEOUT * 3)) # bug 19887
883 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
884 error "Client was not evicted by ost" rc=1
885 wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
886 error "Client was not evicted by mds"
888 run_test 26b "evict dead exports"
892 writemany -q -a $DIR/$tdir/$tfile 0 5 &
895 local save_FAILURE_MODE=$FAILURE_MODE
897 facet_failover $SINGLEMDS
898 #define OBD_FAIL_OSC_SHUTDOWN 0x407
899 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
900 # need to wait for reconnect
901 echo waiting for fail_loc
902 wait_update_facet $SINGLEMDS "lctl get_param -n fail_loc" "-2147482617"
903 facet_failover $SINGLEMDS
905 kill -USR1 $CLIENT_PID
908 FAILURE_MODE=$save_FAILURE_MODE
910 run_test 27 "fail LOV while using OSC's"
912 test_28() { # bug 6086 - error adding new clients
913 do_facet client mcreate $DIR/$tfile || return 1
914 drop_bl_callback "chmod 0777 $DIR/$tfile" ||echo "evicted as expected"
915 #define OBD_FAIL_MDS_CLIENT_ADD 0x12f
916 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012f"
917 # fail once (evicted), reconnect fail (fail_loc), ok
918 client_up || (sleep 10; client_up) || (sleep 10; client_up) || error "reconnect failed"
920 fail $SINGLEMDS # verify MDS last_rcvd can be loaded
922 run_test 28 "handle error adding new clients (bug 6086)"
924 test_29a() { # bug 22273 - error adding new clients
925 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
926 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000711"
927 # fail abort so client will be new again
928 fail_abort $SINGLEMDS
929 client_up || error "reconnect failed"
930 wait_osc_import_state $SINGLEMDS ost FULL
933 run_test 29a "error adding new clients doesn't cause LBUG (bug 22273)"
935 test_29b() { # bug 22273 - error adding new clients
936 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
937 do_facet ost1 "lctl set_param fail_loc=0x80000711"
938 # fail abort so client will be new again
940 client_up || error "reconnect failed"
943 run_test 29b "error adding new clients doesn't cause LBUG (bug 22273)"
947 # put a load of file creates/writes/deletes
948 writemany -q $DIR/$tdir/$tfile 0 5 &
950 echo writemany pid $CLIENT_PID
954 # wait for client to reconnect to MDS
959 # client process should see no problems even though MDS went down
961 kill -USR1 $CLIENT_PID
964 echo writemany returned $rc
965 #these may fail because of eviction due to slow AST response.
967 error_ignore bz13652 "writemany returned rc $rc" || true
969 run_test 50 "failover MDS under load"
972 #define OBD_FAIL_MDS_SYNC_CAPA_SL 0x1310
973 do_facet ost1 lctl set_param fail_loc=0x00001310
976 # put a load of file creates/writes/deletes
977 writemany -q $DIR/$tdir/$tfile 0 5 &
981 facet_failover $SINGLEMDS
982 # failover at various points during recovery
983 SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
984 echo will failover at $SEQ
987 #echo failover in $i sec
988 log "test_$testnum: failover in $i sec"
990 facet_failover $SINGLEMDS
992 # client process should see no problems even though MDS went down
993 # and recovery was interrupted
995 kill -USR1 $CLIENT_PID
998 echo writemany returned $rc
1000 error_ignore bz13652 "writemany returned rc $rc" || true
1002 run_test 51 "failover MDS during recovery"
1005 do_facet client "mkdir -p $DIR/$tdir"
1006 do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
1008 echo writemany pid $CLIENT_PID
1013 wait $CLIENT_PID || rc=$?
1014 # active client process should see an EIO for down OST
1015 [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }
1016 # but timing or failover setup may allow success
1017 [ $rc -eq 0 ] && { echo "writemany succeeded" && return 0; }
1018 echo "writemany returned $rc"
1023 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1028 [ $rc -ne 0 ] && { return $rc; }
1029 # wait for client to reconnect to OST
1033 [ $rc -ne 0 ] && { return $rc; }
1040 run_test 52 "failover OST under load"
1042 # test of open reconstruct
1045 drop_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||\
1048 run_test 53 "touch: drop rep"
1051 zconf_mount `hostname` $MOUNT2
1053 touch $DIR2/$tfile.1
1055 cat $DIR2/$tfile.missing # save transno = 0, rc != 0 into last_rcvd
1058 ERROR=`dmesg | egrep "(test 54|went back in time)" | tail -n1 | grep "went back in time"`
1059 [ x"$ERROR" == x ] || error "back in time occured"
1061 run_test 54 "back in time"
1063 # bug 11330 - liblustre application death during I/O locks up OST
1065 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1069 # Minimum pass speed is 2MBps
1071 # LU-2887/LU-3089 - set min pass speed to 500KBps
1072 [ "$(facet_fstype ost1)" = "zfs" ] && ddtimeout=256
1074 # first dd should be finished quickly
1075 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
1076 dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
1079 echo "step1: testing ......"
1080 while kill -0 $DDPID 2> /dev/null; do
1082 if [ $count -gt $ddtimeout ]; then
1083 error "dd should be finished!"
1087 echo "(dd_pid=$DDPID, time=$count)successful"
1089 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
1090 #define OBD_FAIL_OST_DROP_REQ 0x21d
1091 do_facet ost1 lctl set_param fail_loc=0x0000021d
1092 # second dd will be never finished
1093 dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
1096 echo "step2: testing ......"
1097 while [ $count -le $ddtimeout ]; do
1098 if ! kill -0 $DDPID 2> /dev/null; then
1100 error "dd shouldn't be finished! (time=$count)"
1105 echo "(dd_pid=$DDPID, time=$count)successful"
1107 #Recover fail_loc and dd will finish soon
1108 do_facet ost1 lctl set_param fail_loc=0
1110 echo "step3: testing ......"
1111 while kill -0 $DDPID 2> /dev/null; do
1113 if [ $count -gt $((ddtimeout + 440)) ]; then
1114 error "dd should be finished!"
1118 echo "(dd_pid=$DDPID, time=$count)successful"
1122 run_test 55 "ost_brw_read/write drops timed-out read/write request"
1124 test_56() { # b=11277
1125 #define OBD_FAIL_MDS_RESEND 0x136
1127 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000136"
1128 stat $DIR/$tfile || error "stat failed"
1129 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1132 run_test 56 "do not fail on getattr resend"
1135 # no oscs means no client or mdt
1136 while lctl get_param osc.*.* > /dev/null 2>&1; do
1137 : # loop until proc file is removed
1141 test_57() { # bug 10866
1145 #define OBD_FAIL_LPROC_REMOVE 0xB00
1146 lctl set_param fail_loc=0x80000B00
1147 zconf_umount `hostname` $DIR
1148 lctl set_param fail_loc=0x80000B00
1149 fail_abort $SINGLEMDS
1151 lctl set_param fail_loc=0
1153 do_facet client "df $DIR"
1155 run_test 57 "read procfs entries causes kernel crash"
1157 test_58() { # bug 11546
1158 #define OBD_FAIL_MDC_ENQUEUE_PAUSE 0x801
1161 lctl set_param fail_loc=0x80000801
1162 cp $DIR/$tfile /dev/null &
1165 lctl set_param fail_loc=0
1166 drop_bl_callback rm -f $DIR/$tfile
1168 # the first 'df' could tigger the eviction caused by
1169 # 'drop_bl_callback', and it's normal case.
1170 # but the next 'df' should return successfully.
1171 do_facet client "df $DIR" || do_facet client "df $DIR"
1173 run_test 58 "Eviction in the middle of open RPC reply processing"
1175 test_59() { # bug 10589
1176 zconf_mount `hostname` $MOUNT2 || error "Failed to mount $MOUNT2"
1177 echo $DIR2 | grep -q $MOUNT2 || error "DIR2 is not set properly: $DIR2"
1178 #define OBD_FAIL_LDLM_CANCEL_EVICT_RACE 0x311
1179 lctl set_param fail_loc=0x311
1180 writes=$(LANG=C dd if=/dev/zero of=$DIR2/$tfile count=1 2>&1)
1181 [ $? = 0 ] || error "dd write failed"
1182 writes=$(echo $writes | awk -F '+' '/out/ {print $1}')
1183 lctl set_param fail_loc=0
1185 zconf_umount `hostname` $MOUNT2 -f
1186 reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1)
1187 [ $? = 0 ] || error "dd read failed"
1188 reads=$(echo $reads | awk -F '+' '/in/ {print $1}')
1189 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes
1191 run_test 59 "Read cancel race on client eviction"
1194 # we assume that all md changes are in the MDT0 changelog
1195 if [ $MDSCOUNT -gt 1 ]; then
1196 error_ignore bz17935 $*
1203 MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid |
1204 awk '{ gsub(/_UUID/,""); print $1 }' | head -n1)
1209 # Register (and start) changelog
1210 USER=$(do_facet $SINGLEMDS lctl --device $MDT0 changelog_register -n)
1211 echo "Registered as $MDT0 changelog user $USER"
1213 # Generate a large number of changelog entries
1214 createmany -o $DIR/$tdir/$tfile $NUM_FILES
1218 # Unlink files in the background
1219 unlinkmany $DIR/$tdir/$tfile $NUM_FILES &
1223 # Failover the MDS while unlinks are happening
1224 facet_failover $SINGLEMDS
1226 # Wait for unlinkmany to finish
1229 # Check if all the create/unlink events were recorded
1231 $LFS changelog $MDT0 >> $DIR/$tdir/changelog
1232 local cl_count=$(grep UNLNK $DIR/$tdir/changelog | wc -l)
1233 echo "$cl_count unlinks in $MDT0 changelog"
1235 do_facet $SINGLEMDS lctl --device $MDT0 changelog_deregister $USER
1236 USERS=$(( $(do_facet $SINGLEMDS lctl get_param -n \
1237 mdd.$MDT0.changelog_users | wc -l) - 2 ))
1238 if [ $USERS -eq 0 ]; then
1239 [ $cl_count -eq $NUM_FILES ] || \
1240 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1241 # Also make sure we can clear large changelogs
1242 cl_count=$($LFS changelog $FSNAME | wc -l)
1243 [ $cl_count -le 2 ] || \
1244 error "Changelog not empty: $cl_count entries"
1246 # If there are other users, there may be other unlinks in the log
1247 [ $cl_count -ge $NUM_FILES ] || \
1248 err17935 "Recorded ${cl_count} unlinks out of $NUM_FILES"
1249 echo "$USERS other changelog users; can't verify clear"
1252 run_test 60 "Add Changelog entries during MDS failover"
1256 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000)
1257 mdtosc=${mdtosc/-MDT*/-MDT\*}
1258 local cflags="osc.$mdtosc.connect_flags"
1259 do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
1260 [ $? -ne 0 ] && skip "don't have skip orphan feature" && return
1262 mkdir -p $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
1263 # Set the default stripe of $DIR/$tdir to put the files to ost1
1264 $LFS setstripe -c 1 -i 0 $DIR/$tdir
1266 replay_barrier $SINGLEMDS
1267 createmany -o $DIR/$tdir/$tfile-%d 10
1268 local oid=`do_facet ost1 "lctl get_param -n obdfilter.${ost1_svc}.last_id"`
1270 fail_abort $SINGLEMDS
1272 touch $DIR/$tdir/$tfile
1273 local id=`$LFS getstripe $DIR/$tdir/$tfile | awk '$1 == 0 { print $2 }'`
1274 [ $id -le $oid ] && error "the orphan objid was reused, failed"
1279 run_test 61 "Verify to not reuse orphan objects - bug 17025"
1281 # test_62 as seen it b2_1 please do not reuse test_62
1284 # zconf_umount `hostname` $DIR
1285 # #define OBD_FAIL_PTLRPC_DELAY_IMP_FULL 0x516
1286 # lctl set_param fail_loc=0x516
1289 #run_test 62 "Verify connection flags race - bug LU-1716"
1291 check_cli_ir_state()
1293 local NODE=${1:-$HOSTNAME}
1295 st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
1296 awk '/imperative_recovery:/ { print \\\$2}'")
1297 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1298 error "Error state $st, must be ENABLED or DISABLED"
1302 check_target_ir_state()
1305 local name=${target}_svc
1306 local recovery_proc=obdfilter.${!name}.recovery_status
1309 st=$(do_facet $target "lctl get_param -n $recovery_proc |
1310 awk '/IR:/{ print \\\$2}'")
1311 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1312 error "Error state $st, must be ENABLED or DISABLED"
1318 do_facet mgs lctl set_param -n mgs.MGS.live.$FSNAME="state=$1"
1323 local state=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1324 awk '/state:/{ print \\\$2 }'")
1328 nidtbl_version_mgs()
1330 local ver=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1331 awk '/nidtbl_version:/{ print \\\$2 }'")
1335 # nidtbl_version_client <mds1|client> [node]
1336 nidtbl_version_client()
1339 local node=${2:-$HOSTNAME}
1341 if [ X$cli = Xclient ]; then
1344 local obdtype=${cli/%[0-9]*/}
1345 [ $obdtype != mds ] && error "wrong parameters $cli"
1347 node=$(facet_active_host $cli)
1352 local vers=$(do_node $node "lctl get_param -n mgc.*.ir_state" |
1353 awk "/$cli/{print \$6}" |sort -u)
1355 # in case there are multiple mounts on the client node
1357 [ ${#arr[@]} -ne 1 ] && error "versions on client node mismatch"
1361 nidtbl_versions_match()
1363 [ $(nidtbl_version_mgs) -eq $(nidtbl_version_client ${1:-client}) ]
1366 target_instance_match()
1372 obdtype=${srv/%[0-9]*/}
1383 error "invalid target type" $srv
1388 local target=${srv}_svc
1389 local si=$(do_facet $srv lctl get_param -n $obdname.${!target}.instance)
1390 local ci=$(lctl get_param -n $cliname.${!target}-${cliname}-*.import |
1391 awk '/instance/{ print $2 }' | head -n1)
1393 return $([ $si -eq $ci ])
1398 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1399 { skip "MGS without IR support"; return 0; }
1401 # MDT was just restarted in the previous test, make sure everything
1404 while [ $cnt -gt 0 ]; do
1405 nidtbl_versions_match && break
1411 set_ir_status disabled
1413 local prev_ver=$(nidtbl_version_client client)
1415 local saved_FAILURE_MODE=$FAILURE_MODE
1416 [ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
1420 [ $(nidtbl_version_client client) -eq $prev_ver ] ||
1421 error "version must not change due to IR disabled"
1422 target_instance_match ost1 || error "instance mismatch"
1426 FAILURE_MODE=$saved_FAILURE_MODE
1428 run_test 100 "IR: Make sure normal recovery still works w/o IR"
1432 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1433 { skip "MGS without IR support"; return 0; }
1437 local OST1_IMP=$(get_osc_import_name client ost1)
1439 # disable pinger recovery
1440 lctl set_param -n osc.$OST1_IMP.pinger_recov=0
1444 target_instance_match ost1 || error "instance mismatch"
1445 nidtbl_versions_match || error "version must match"
1447 lctl set_param -n osc.$OST1_IMP.pinger_recov=1
1449 run_test 101 "IR: Make sure IR works w/o normal recovery"
1453 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1454 { skip "MGS without IR support"; return 0; }
1456 local clients=${CLIENTS:-$HOSTNAME}
1463 # let's have a new nidtbl version
1466 # sleep for a while so that clients can see the failure of ost
1467 # it must be MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC.
1468 # int mgc_request.c:
1469 # define MGC_TIMEOUT_MIN_SECONDS 5
1470 # define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 *
1471 local count=30 # 20 seconds at most
1472 while [ $count -gt 0 ]; do
1473 nidtbl_versions_match && break
1478 nidtbl_versions_match || error "nidtbl mismatch"
1481 old_version=$(nidtbl_version_client client)
1483 zconf_umount_clients $clients $MOUNT || error "Cannot umount client"
1486 combined_mgs_mds && mgsdev=mds1
1487 remount_facet $mgsdev
1490 zconf_mount_clients $clients $MOUNT || error "Cannot mount client"
1493 new_version=$(nidtbl_version_client client)
1494 [ $new_version -lt $old_version ] &&
1495 error "nidtbl version wrong after mgs restarts"
1498 run_test 102 "IR: New client gets updated nidtbl after MGS restart"
1502 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1503 { skip "MGS without IR support"; return 0; }
1505 combined_mgs_mds && skip "mgs and mds on the same target" && return 0
1507 # workaround solution to generate config log on the mds
1513 # We need this test because mds is like a client in IR context.
1514 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS ||
1515 error "MDS should start w/o mgs"
1517 # start mgs and remount mds w/ ir
1518 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
1521 # remount client so that fsdb will be created on the MGS
1522 umount_client $MOUNT || error "umount failed"
1523 mount_client $MOUNT || error "mount failed"
1525 # sleep 30 seconds so the MDS has a chance to detect MGS restarting
1527 while [ $count -gt 0 ]; do
1528 [ $(nidtbl_version_client mds1) -ne 0 ] && break
1533 # after a while, mds should be able to reconnect to mgs and fetch
1534 # up-to-date nidtbl version
1535 nidtbl_versions_match mds1 || error "mds nidtbl mismatch"
1540 run_test 103 "IR: MDS can start w/o MGS and get updated nidtbl later"
1544 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1545 { skip "MGS without IR support"; return 0; }
1550 start ost1 $(ostdevname 1) "$OST_MOUNT_OPTS -onoir" ||
1551 error "OST1 cannot start"
1554 local ir_state=$(check_target_ir_state ost1)
1555 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1556 error "ir status on ost1 should be DISABLED"
1558 run_test 104 "IR: ost can disable IR voluntarily"
1562 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
1563 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1564 { skip "MGS without IR support"; return 0; }
1568 # get one of the clients from client list
1569 local rcli=$(echo $RCLIENTS |cut -d' ' -f 1)
1571 local old_MOUNTOPT=$MOUNTOPT
1572 MOUNTOPT=${MOUNTOPT},noir
1573 zconf_umount $rcli $MOUNT || error "umount failed"
1574 zconf_mount $rcli $MOUNT || error "mount failed"
1576 # make sure lustre mount at $rcli disabling IR
1577 local ir_state=$(check_cli_ir_state $rcli)
1578 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1579 error "IR state must be DISABLED at $rcli"
1581 # Since the client just mounted, its last_rcvd entry is not on disk.
1582 # Send an RPC so exp_need_sync forces last_rcvd to commit this export
1583 # so the client can reconnect during OST recovery (LU-924, LU-1582)
1584 $SETSTRIPE -i 0 $DIR/$tfile
1585 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
1587 # make sure MGS's state is Partial
1588 [ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
1591 # make sure IR on ost1 is DISABLED
1592 local ir_state=$(check_target_ir_state ost1)
1593 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1594 error "IR status on ost1 should be DISABLED"
1597 MOUNTOPT=$old_MOUNTOPT
1598 zconf_umount $rcli $MOUNT || error "umount failed"
1599 zconf_mount $rcli $MOUNT || error "mount failed"
1601 # make sure MGS's state is full
1602 [ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
1605 # make sure IR on ost1 is ENABLED
1606 local ir_state=$(check_target_ir_state ost1)
1607 [ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
1608 error "IR status on ost1 should be ENABLED"
1612 run_test 105 "IR: NON IR clients support"
1620 test_106() { # LU-1789
1621 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.50) ]] ||
1622 { skip "Need MDS version at least 2.3.50"; return 0; }
1624 #define OBD_FAIL_MDC_LIGHTWEIGHT 0x805
1625 $LCTL set_param fail_loc=0x805
1628 trap cleanup_106 EXIT
1630 # enable lightweight flag on mdc connection
1633 local MDS_NEXP=$(do_facet $SINGLEMDS \
1634 lctl get_param -n mdt.${mds1_svc}.num_exports |
1636 $LCTL set_param fail_loc=0
1638 touch $DIR2/$tfile || error "failed to create empty file"
1639 replay_barrier $SINGLEMDS
1641 $LCTL set_param debug=console
1643 facet_failover $SINGLEMDS
1645 # lightweight connection must be evicted
1646 touch -c $DIR2/$tfile || true
1647 $LCTL dk $TMP/lustre-log-$TESTNAME.log
1648 evicted=`awk '/This client was evicted by .*MDT0000/ {
1650 }' $TMP/lustre-log-$TESTNAME.log`
1651 [ -z "$evicted" ] && error "lightweight client not evicted by mds"
1653 # and all operations performed by lightweight client should be
1654 # synchronous, so the file created before mds restart should be there
1655 $CHECKSTAT -t file $DIR/$tfile || error "file not present"
1660 run_test 106 "lightweight connection support"
1667 # OBD_FAIL_MDS_REINT_NET_REP 0x119
1668 do_facet $SINGLEMDS lctl set_param fail_loc=0x119
1669 multiop $DIR/$tdir D_c &
1671 mkdir $DIR/$tdir/dir_106 &
1673 do_facet $SINGLEMDS lctl set_param fail_loc=0
1676 wait $CLIENT_PID || rc=$?
1677 checkstat -t dir $DIR/$tdir/dir_106 || return 1
1679 kill -USR1 $close_pid
1680 wait $close_pid || return 2
1684 run_test 107 "drop reint reply, then restart MDT"
1687 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1688 local remote_dir=$DIR/$tdir/remote_dir
1692 #prepare for 110 test, which need set striped dir on remote MDT.
1693 for num in $(seq $MDSCOUNT); do
1695 lctl set_param -n mdt.${FSNAME}*.enable_remote_dir=1 \
1700 drop_request "$LFS mkdir -i $MDTIDX -c2 $remote_dir" ||
1701 error "lfs mkdir failed"
1702 local diridx=$($GETSTRIPE -M $remote_dir)
1703 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
1705 rm -rf $DIR/$tdir || error "rmdir failed"
1707 run_test 110a "create remote directory: drop client req"
1710 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1711 local remote_dir=$DIR/$tdir/remote_dir
1715 drop_reint_reply "$LFS mkdir -i $MDTIDX -c2 $remote_dir" ||
1716 error "lfs mkdir failed"
1718 diridx=$($GETSTRIPE -M $remote_dir)
1719 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
1721 rm -rf $DIR/$tdir || error "rmdir failed"
1723 run_test 110b "create remote directory: drop Master rep"
1726 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1727 local remote_dir=$DIR/$tdir/remote_dir
1731 drop_update_reply $MDTIDX "$LFS mkdir -i $MDTIDX -c2 $remote_dir" ||
1732 error "lfs mkdir failed"
1734 diridx=$($GETSTRIPE -M $remote_dir)
1735 [ $diridx -eq $MDTIDX ] || error "$diridx != $MDTIDX"
1737 rm -rf $DIR/$tdir || error "rmdir failed"
1739 run_test 110c "create remote directory: drop update rep on slave MDT"
1742 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1743 local remote_dir=$DIR/$tdir/remote_dir
1747 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
1749 drop_request "rm -rf $remote_dir" || error "rm remote dir failed"
1751 rm -rf $DIR/$tdir || error "rmdir failed"
1755 run_test 110d "remove remote directory: drop client req"
1758 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1759 local remote_dir=$DIR/$tdir/remote_dir
1763 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
1764 drop_reint_reply "rm -rf $remote_dir" || error "rm remote dir failed"
1766 rm -rf $DIR/$tdir || error "rmdir failed"
1770 run_test 110e "remove remote directory: drop master rep"
1773 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1774 local remote_dir=$DIR/$tdir/remote_dir
1778 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
1779 drop_update_reply $MDTIDX "rm -rf $remote_dir" ||
1780 error "rm remote dir failed"
1782 rm -rf $DIR/$tdir || error "rmdir failed"
1784 run_test 110f "remove remote directory: drop slave rep"
1787 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
1788 local remote_dir=$DIR/$tdir/remote_dir
1791 mkdir -p $remote_dir
1793 createmany -o $remote_dir/f 100
1795 #define OBD_FAIL_MIGRATE_NET_REP 0x1702
1796 do_facet mds$MDTIDX lctl set_param fail_loc=0x1702
1797 $LFS mv -M $MDTIDX $remote_dir || error "migrate failed"
1798 do_facet mds$MDTIDX lctl set_param fail_loc=0x0
1800 for file in $(find $remote_dir); do
1801 mdt_index=$($LFS getstripe -M $file)
1802 [ $mdt_index == $MDTIDX ] ||
1803 error "$file is not on MDT${MDTIDX}"
1806 rm -rf $DIR/$tdir || error "rmdir failed"
1808 run_test 110g "drop reply during migration"
1810 # LU-2844 mdt prepare fail should not cause umount oops
1813 [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.3.62) ]] ||
1814 { skip "Need MDS version at least 2.3.62"; return 0; }
1816 local mdsdev=$(mdsdevname ${SINGLEMDS//mds/})
1817 #define OBD_FAIL_MDS_CHANGELOG_INIT 0x151
1818 do_facet $SINGLEMDS lctl set_param fail_loc=0x151
1819 stop $SINGLEMDS || error "stop MDS failed"
1820 start $SINGLEMDS $mdsdev && error "start MDS should fail"
1821 do_facet $SINGLEMDS lctl set_param fail_loc=0
1822 start $SINGLEMDS $mdsdev || error "start MDS failed"
1824 run_test 111 "mdd setup fail should not cause umount oops"
1828 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1830 do_facet_random_file client $TMP/$tfile 100K ||
1831 error_noexit "Create random file $TMP/$tfile"
1833 pause_bulk "cp $TMP/$tfile $DIR/$tfile" $TIMEOUT ||
1834 error_noexit "Can't pause_bulk copy"
1837 # expect cmp to succeed, client resent bulk
1838 cmp $TMP/$tfile $DIR/$tfile ||
1839 error_noexit "Wrong data has been written"
1841 error_noexit "Can't remove file"
1844 run_test 112a "bulk resend while orignal request is in progress"
1846 # parameters: fail_loc CMD RC
1853 #define OBD_FAIL_LDLM_CP_CB_WAIT2 0x320
1854 #define OBD_FAIL_LDLM_CP_CB_WAIT3 0x321
1855 #define OBD_FAIL_LDLM_CP_CB_WAIT4 0x322
1856 #define OBD_FAIL_LDLM_CP_CB_WAIT5 0x323
1859 echo -n "** FLOCK REPLY vs. EVICTION race, lock $2"
1860 [ "$1" = "CLEANUP" ] &&
1861 fail=0x80000320 && echo ", $1 cp first"
1862 [ "$1" = "REPLY" ] &&
1863 fail=0x80000321 && echo ", $1 cp first"
1864 [ "$1" = "DEADLOCK CLEANUP" ] &&
1865 fail=0x80000322 && echo " DEADLOCK, CLEANUP cp first"
1866 [ "$1" = "DEADLOCK REPLY" ] &&
1867 fail=0x80000323 && echo " DEADLOCK, REPLY cp first"
1869 if [ x"$2" = x"get" ]; then
1870 #for TEST lock, take a conflict in advance
1871 # sleep longer than evictor to not confuse fail_loc: 2+2+4
1872 echo "** Taking conflict **"
1873 flocks_test 5 set read sleep 10 $DIR/$tfile &
1879 $LCTL set_param fail_loc=$fail
1881 flocks_test 5 $2 write $DIR/$tfile &
1885 echo "** Evicting and re-connecting client **"
1890 if [ x"$2" = x"get" ]; then
1897 # check if the return value is allowed
1898 [ $rc -eq $3 ] && rc=0
1900 $LCTL set_param fail_loc=0
1904 # a lock is taken, unlock vs. cleanup_resource() race for destroying
1905 # the ORIGINAL lock.
1910 flocks_test 5 set write sleep 4 $DIR/$tfile &
1914 # let unlock to sleep in CP CB
1915 $LCTL set_param fail_loc=$1
1918 # let cleanup to cleep in CP CB
1926 $LCTL set_param fail_loc=0
1931 flock_is_enabled || { skip "mount w/o flock enabled" && return; }
1934 test_120_reply "CLEANUP" set 5 || error "SET race failed"
1935 test_120_reply "CLEANUP" get 5 || error "GET race failed"
1936 test_120_reply "CLEANUP" unlock 5 || error "UNLOCK race failed"
1938 test_120_reply "REPLY" set 5 || error "SET race failed"
1939 test_120_reply "REPLY" get 5 || error "GET race failed"
1940 test_120_reply "REPLY" unlock 5 || error "UNLOCK race failed"
1943 test_120_reply "DEADLOCK CLEANUP" set 5 || error "DEADLOCK race failed"
1944 test_120_reply "DEADLOCK REPLY" set 35 || error "DEADLOCK race failed"
1946 test_120_destroy 0x320 || error "unlock-cleanup race failed"
1948 run_test 120 "flock race: completion vs. evict"
1951 local BEFORE=$(date +%s)
1954 # modify dir so that next revalidate would not obtain UPDATE lock
1957 # drop 1 reply with UPDATE lock,
1958 # resend should not create 2nd lock on server
1959 mcreate $DIR/$tfile || error "mcreate failed: $?"
1960 drop_ldlm_reply_once "stat $DIR/$tfile" || error "stat failed: $?"
1962 # 2 BL AST will be sent to client, both must find the same lock,
1963 # race them to not get EINVAL for 2nd BL AST
1964 #define OBD_FAIL_LDLM_PAUSE_CANCEL2 0x31f
1965 $LCTL set_param fail_loc=0x8000031f
1967 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0 > /dev/null
1968 chmod 0777 $DIR/$tfile || error "chmod failed: $?"
1969 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1 > /dev/null
1971 # let the client reconnect
1973 EVICT=$($LCTL get_param mdc.$FSNAME-MDT*.state |
1974 awk -F"[ [,]" '/EVICTED]$/ { if (mx<$4) {mx=$4;} } END { print mx }')
1976 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
1978 run_test 113 "ldlm enqueue dropped reply should not cause deadlocks"
1981 check_and_cleanup_lustre