5 PTLDEBUG=${PTLDEBUG:--1}
6 LUSTRE=${LUSTRE:-$(dirname $0)/..}
7 . $LUSTRE/tests/test-framework.sh
11 ALWAYS_EXCEPT="$RECOVERY_SMALL_EXCEPT "
14 always_except LU-17141 133
19 require_dsh_mds || exit 0
21 # Allow us to override the setup if we already have a mounted system by
22 # setting SETUP=" " and CLEANUP=" "
24 CLEANUP=${CLEANUP:-""}
26 check_and_setup_lustre
29 rm -rf $DIR/d[0-9]* $DIR/f.${TESTSUITE}*
34 local f1="$DIR/$tfile"
35 local f2="$DIR/$tfile.2"
37 drop_request "mcreate $f1" ||
38 error_noexit "create '$f1': drop req"
40 drop_reint_reply "mcreate $f2" ||
41 error_noexit "create '$f2': drop rep"
43 drop_request "tchmod 111 $f2" ||
44 error_noexit "chmod '$f2': drop req"
46 drop_reint_reply "tchmod 666 $f2" ||
47 error_noexit "chmod '$f2': drop rep"
49 drop_request "statone $f2" ||
50 error_noexit "stat '$f2': drop req"
52 drop_reply "statone $f2" ||
53 error_noexit "stat '$f2': drop rep"
55 run_test 1 "create, chmod, stat: drop req, drop rep"
59 do_facet_create_file client $t 10K ||
60 error_noexit "Create file $t"
62 drop_request "cat $t > /dev/null" ||
63 error_noexit "Open request for $t file"
65 drop_reply "cat $t > /dev/null" ||
66 error_noexit "Open replay for $t file"
68 run_test 4 "open: drop req, drop rep"
73 local RR="$T-renamed-again"
74 do_facet_create_file client $T 10K ||
75 error_noexit "Create file $T"
77 drop_request "mv $T $R" ||
78 error_noexit "Rename $T"
80 drop_reint_reply "mv $R $RR" ||
81 error_noexit "Failed rename replay on $R"
83 do_facet client "checkstat -v $RR" ||
84 error_noexit "checkstat error on $RR"
86 do_facet client "rm $RR" ||
87 error_noexit "Can't remove file $RR"
89 run_test 5 "rename: drop req, drop rep"
93 local LINK1=$DIR/$tfile.link1
94 local LINK2=$DIR/$tfile.link2
96 do_facet_create_file client $T 10K || error_noexit "Create file $T"
97 drop_request "link $T $LINK1" || error_noexit "link request for $T"
98 drop_reint_reply "link $T $LINK2" || error_noexit "link reply for $T"
99 drop_request "unlink $LINK1" || error_noexit "unlink request for $T"
100 drop_reint_reply "unlink $LINK2" || error_noexit "unlink reply for $T"
101 do_facet client "rm $T" || error_noexit "Can't remove file $T"
103 run_test 6 "link, unlink: drop req, drop rep"
107 drop_reint_reply "touch $DIR/$tfile" || return 1
109 run_test 8 "touch: drop rep (bug 1423)"
113 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
117 do_facet_random_file client $TMP/$tfile 1K ||
118 error_noexit "Create random file $TMP/$tfile"
119 # make this big, else test 9 doesn't wait for bulk -- bz 5595
120 do_facet_create_file client $TMP/$t1 4M ||
121 error_noexit "Create file $TMP/$t1"
122 do_facet client "cp $TMP/$t1 $DIR/$t1" ||
123 error_noexit "Can't copy to $DIR/$t1 file"
124 pause_bulk "cp $TMP/$tfile $DIR/$tfile" ||
125 error_noexit "Can't pause_bulk copy"
126 do_facet client "cp $TMP/$t1 $DIR/$t2" ||
127 error_noexit "Can't copy file"
128 do_facet client "sync"
129 do_facet client "rm $DIR/$tfile $DIR/$t2 $DIR/$t1" ||
130 error_noexit "Can't remove files"
131 do_facet client "rm $TMP/$t1 $TMP/$tfile"
133 run_test 9 "pause bulk on OST (bug 1420)"
137 local before=$(date +%s)
140 do_facet client "stat $DIR > /dev/null" ||
141 error "failed to stat $DIR: $?"
142 drop_bl_callback "chmod 0777 $DIR" ||
143 error "failed to chmod $DIR: $?"
145 # let the client reconnect
147 evict=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
148 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
149 [ ! -z "$evict" ] && [[ $evict -gt $before ]] ||
150 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
151 error "no eviction: $evict before:$before")
153 do_facet client checkstat -v -p 0777 $DIR ||
154 error "client checkstat failed: $?"
156 run_test 10a "finish request on server after client eviction (bug 1521)"
159 local before=$(date +%s)
162 [[ "$MDS1_VERSION" -lt $(version_code 2.6.53) ]] &&
163 skip "Need MDS version at least 2.6.53"
164 do_facet client "stat $DIR > /dev/null" ||
165 error "failed to stat $DIR: $?"
166 drop_bl_callback_once "chmod 0777 $DIR" ||
167 error "failed to chmod $DIR: $?"
169 # let the client reconnect
171 evict=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
172 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
174 [ -z "$evict" ] || [[ $evict -le $before ]] ||
175 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
176 error "eviction happened: $evict before:$before")
178 do_facet client checkstat -v -p 0777 $DIR ||
179 error "client checkstat failed: $?"
181 run_test 10b "re-send BL AST"
184 local before=$(date +%s)
193 workdir="${DIR}/${tdir}"
194 mkdir -p ${workdir} || error "can't create workdir $?"
195 stat ${workdir} > /dev/null ||
196 error "failed to stat ${workdir}: $?"
197 mdtidx=$($LFS getdirstripe -i ${workdir})
198 mdtname=$($LFS mdts ${workdir} | grep -e "^$mdtidx:" |
199 awk '{sub("_UUID", "", $2); print $2;}')
201 mdccli=$($LCTL dl | grep "${mdtname}-mdc" | awk '{print $4;}')
202 conn_uuid=$($LCTL get_param -n mdc.${mdccli}.conn_uuid)
203 mdcpath="mdc.${mdccli}.import=connection=${conn_uuid}"
205 drop_bl_callback_once "chmod 0777 ${workdir}" &
210 # force client reconnect
211 $LCTL set_param "${mdcpath}"
213 # wait client reconnect
217 evict=$($LCTL get_param mdc.${mdccli}.state |
218 awk -F"[ [,]" '/EVICTED]$/ { if (t<$4) {t=$4;} } END { print t }')
220 [[ $evict -le $before ]] ||
221 ( $LCTL get_param mdc.$FSNAME-MDT*.state;
222 error "eviction happened: $EVICT before:$BEFORE" )
224 [ $rc -eq 0 ] || error "chmod must finished OK"
225 checkstat -v -p 0777 "${workdir}" ||
226 error "client checkstat failed: $?"
228 run_test 10c "re-send BL AST vs reconnect race (LU-5569)"
231 local before=$(date +%s)
234 [[ "$MDS1_VERSION" -lt $(version_code 2.6.90) ]] &&
235 skip "Need MDS version at least 2.6.90"
237 # sleep 1 is to make sure that BEFORE is not equal to EVICTED below
240 echo -n ", world" | dd of=$TMP/$tfile bs=1c seek=5
242 remount_client $MOUNT
246 $LFS setstripe -i 0 -c 1 $DIR1/$tfile
247 echo -n hello | dd of=$DIR1/$tfile bs=5
249 stat $DIR2/$tfile >& /dev/null
250 $LCTL set_param fail_err=71
251 drop_bl_callback "echo -n \\\", world\\\" >> $DIR2/$tfile"
256 cmp -l $DIR1/$tfile $DIR2/$tfile || error "file contents differ"
257 cmp -l $DIR1/$tfile $TMP/$tfile || error "wrong content found"
259 evict=$(do_facet client $LCTL get_param osc.$FSNAME-OST0000*.state | \
261 awk -F"[ [,]" '/EVICTED$/ { if (mx<$1) {mx=$1;} } END { print mx }')
263 [[ $evict -gt $before ]] ||
264 (do_facet client $LCTL get_param osc.$FSNAME-OST0000*.state;
265 error "no eviction: $evict before:$before")
267 $LCTL set_param fail_err=0
269 umount_client $MOUNT2
271 run_test 10d "test failed blocking ast"
275 [[ "$OST1_VERSION" -le $(version_code 2.8.58) ]] &&
276 skip "Need OST version at least 2.8.59"
277 [ $CLIENTCOUNT -lt 2 ] && skip "need two clients"
278 [ $(facet_host client) == $(facet_host ost1) ] &&
279 skip "need ost1 and client on different nodes"
280 local -a clients=(${CLIENTS//,/ })
281 local client1=${clients[0]}
282 local client2=${clients[1]}
284 $LFS setstripe -c 1 -i 0 $DIR/$tfile-1 $DIR/$tfile-2
285 $MULTIOP $DIR/$tfile-1 Ow1048576c
287 #define OBD_FAIL_LDLM_BL_CALLBACK_NET 0x305
288 $LCTL set_param fail_loc=0x80000305
290 #define OBD_FAIL_LDLM_ENQUEUE_OLD_EXPORT 0x30e
291 do_facet ost1 "$LCTL set_param fail_loc=0x1000030e"
292 # hit OBD_FAIL_LDLM_ENQUEUE_OLD_EXPORT twice:
293 # 1. to return ENOTCONN from ldlm_handle_enqueue0
294 # 2. to pause reconnect handling between resend and setting
295 # import to LUSTRE_IMP_FULL state
296 do_facet ost1 "$LCTL set_param fail_val=3"
298 # client1 fails ro respond to bl ast
299 do_node $client2 "$MULTIOP $DIR/$tfile-1 Ow1048576c" &
302 # ost1 returns error on enqueue, which causes client1 to reconnect
303 do_node $client1 "$MULTIOP $DIR/$tfile-2 Ow1048576c" ||
304 error "multiop failed"
307 do_facet ost1 "$LCTL set_param fail_loc=0"
308 do_facet ost1 "$LCTL set_param fail_val=0"
310 run_test 10e "re-send BL AST vs reconnect race 2"
313 # wake up a thread waiting for completion after eviction
315 do_facet client $MULTIOP $DIR/$tfile Ow ||
316 { error "multiop write failed: $?"; return 1; }
317 do_facet client $MULTIOP $DIR/$tfile or ||
318 { error "multiop read failed: $?"; return 2; }
322 do_facet client $MULTIOP $DIR/$tfile or ||
323 { error "multiop read failed: $?"; return 3; }
324 drop_bl_callback_once $MULTIOP $DIR/$tfile Ow ||
325 echo "evicted as expected"
327 do_facet client unlink $DIR/$tfile ||
328 { error "unlink failed: $?"; return 4; }
329 # allow recovery to complete
330 client_up || client_up || sleep $TIMEOUT
332 run_test 11 "wake up a thread waiting for completion after eviction (b=2460)"
336 $LCTL mark $MULTIOP $DIR/$tfile OS_c
337 do_facet $SINGLEMDS "lctl set_param fail_loc=0x115"
338 clear_failloc $SINGLEMDS $((TIMEOUT * 2)) &
339 multiop_bg_pause $DIR/$tfile OS_c ||
340 { error "multiop failed: $?"; return 1; }
342 #define OBD_FAIL_MDS_CLOSE_NET 0x115
344 echo "waiting for multiop $PID"
345 wait $PID || { error "wait for multiop faile: $?"; return 2; }
346 do_facet client unlink $DIR/$tfile ||
347 { error "client unlink failed: $?"; return 3; }
348 # allow recovery to complete
349 client_up || client_up || sleep $TIMEOUT
351 run_test 12 "recover from timed out resend in ptlrpcd (b=2494)"
353 # Bug 113, check that readdir lost recv timeout works.
355 mkdir_on_mdt0 $DIR/$tdir || { error "mkdir failed: $?"; return 1; }
356 touch $DIR/$tdir/newentry || { error "touch failed: $?"; return 2; }
357 # OBD_FAIL_MDS_READPAGE_NET|CFS_FAIL_ONCE
358 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000104"
359 ls $DIR/$tdir || { error "ls failed: $?"; return 3; }
360 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
361 rm -rf $DIR/$tdir || { error "remove test dir failed: $?"; return 4; }
363 run_test 13 "mdc_readpage restart test (bug 1138)"
365 # Bug 113, check that readdir lost send timeout works.
367 mkdir_on_mdt0 $DIR/$tdir
368 touch $DIR/$tdir/newentry
369 # OBD_FAIL_MDS_SENDPAGE|CFS_FAIL_ONCE
370 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000106"
371 ls $DIR/$tdir || return 1
372 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
374 run_test 14 "mdc_readpage resend test (bug 1138)"
377 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000128"
378 touch $DIR/$tfile && return 1
381 run_test 15 "failed open (-ENOMEM)"
383 READ_AHEAD=`lctl get_param -n llite.*.max_read_ahead_mb | head -n 1`
385 lctl set_param -n llite.*.max_read_ahead_mb 0
389 lctl set_param -n llite.*.max_read_ahead_mb $READ_AHEAD
393 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
395 do_facet_random_file client $TMP/$tfile 100K ||
396 { error_noexit "Create random file $TMP/$T" ; return 0; }
397 do_facet client "cp $TMP/$tfile $DIR/$tfile" ||
398 { error_noexit "Copy to $DIR/$tfile file" ; return 0; }
402 #define OBD_FAIL_PTLRPC_BULK_PUT_NET 0x504 | CFS_FAIL_ONCE
403 do_facet ost1 "lctl set_param fail_loc=0x80000504"
405 # OST bulk will time out here, client resends
406 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 1
407 do_facet ost1 lctl set_param fail_loc=0
408 # give recovery a chance to finish (shouldn't take long)
410 do_facet client "cmp $TMP/$tfile $DIR/$tfile" || return 2
414 run_test 16 "timeout bulk put, don't evict client (2732)"
419 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
421 local SAMPLE_FILE=$TMP/$tfile
422 do_facet_random_file client $SAMPLE_FILE 20K ||
423 { error_noexit "Create random file $SAMPLE_FILE" ; return 0; }
425 # With adaptive timeouts, bulk_get won't expire until
426 # adaptive_timeout_max
427 if at_is_enabled; then
428 at_max_saved=$(at_max_get ost1)
429 at_max_set $TIMEOUT ost1
432 # OBD_FAIL_PTLRPC_BULK_GET_NET 0x0503 | CFS_FAIL_ONCE
433 # OST bulk will time out here, client retries
434 do_facet ost1 lctl set_param fail_loc=0x80000503
435 # need to ensure we send an RPC
436 do_facet client cp $SAMPLE_FILE $DIR/$tfile
439 # with AT, client will wait adaptive_max*factor+net_latency before
440 # expiring the req, hopefully timeout*2 is enough
441 sleep $(($TIMEOUT*2))
443 do_facet ost1 lctl set_param fail_loc=0
444 do_facet client "df $DIR"
445 # expect cmp to succeed, client resent bulk
446 do_facet client "cmp $SAMPLE_FILE $DIR/$tfile" || return 3
447 do_facet client "rm $DIR/$tfile" || return 4
448 [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved ost1
451 run_test 17a "timeout bulk get, don't evict client (2732)"
454 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
456 # get one of the clients from client list
457 local rcli=$(echo $RCLIENTS | cut -d ' ' -f 1)
458 local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
459 local ldlm_enqueue_min=$(do_facet ost1 find /sys -name ldlm_enqueue_min)
460 [ -z "$ldlm_enqueue_min" ] &&
461 skip "missing /sys/.../ldlm_enqueue_min" && return 0
463 $LFS setstripe -i 0 -c 1 -S 1048576 $DIR/$tfile ||
464 error "setstripe failed"
465 $LFS setstripe -i 0 -c 1 -S 1048576 $DIR/${tfile}2 ||
466 error "setstripe 2 failed"
468 save_lustre_params ost1 "at_history" > $p
469 save_lustre_params ost1 "bulk_timeout" >> $p
470 local dev="${FSNAME}-OST0000"
471 save_lustre_params ost1 "obdfilter.$dev.brw_size" >> $p
472 local ldlm_enqueue_min_save=$(do_facet ost1 cat $ldlm_enqueue_min)
474 local new_at_history=15
476 do_facet ost1 "$LCTL set_param at_history=$new_at_history"
477 do_facet ost1 "$LCTL set_param bulk_timeout=30"
478 do_facet ost1 "echo 30 > /sys/module/ptlrpc/parameters/ldlm_enqueue_min"
480 # brw_size is required to be 4m so that bulk transfer timeout
481 # could be simulated with OBD_FAIL_PTLRPC_CLIENT_BULK_CB3
482 local brw_size=$($LCTL get_param -n \
483 osc.$FSNAME-OST0000-osc-[^M]*.import |
484 awk '/max_brw_size/{print $2}')
485 if [ $brw_size -ne 4194304 ]
487 save_lustre_params ost1 "obdfilter.$dev.brw_size" >> $p
488 do_facet ost1 "$LCTL set_param obdfilter.$dev.brw_size=4"
489 remount_client $MOUNT
492 # get service estimate expanded
493 #define OBD_FAIL_OST_BRW_PAUSE_PACK 0x224
494 do_facet ost1 "$LCTL set_param fail_loc=0x80000224 fail_val=35"
495 echo "delay rpc servicing by 35 seconds"
496 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=fdatasync
499 estimate=$($LCTL get_param -n osc.$dev-osc-*.timeouts |
500 awk '/portal 6/ {print $5}')
501 echo "service estimates increased to $estimate"
503 # let current worst service estimate to get closer to obliteration
504 sleep $((new_at_history / 3))
506 # start i/o and simulate bulk transfer loss
507 #define OBD_FAIL_PTLRPC_CLIENT_BULK_CB3 0x520
508 do_facet ost1 "$LCTL set_param fail_loc=0xa0000520 fail_val=1"
509 dd if=/dev/zero of=$DIR/$tfile bs=2M count=1 conv=fdatasync,notrunc &
512 # start lock conflict handling
513 sleep $((new_at_history / 3))
514 do_node $rcli "dd if=$DIR/$tfile of=/dev/null bs=1M count=1" &
517 # obliterate the worst service estimate
518 sleep $((new_at_history / 3 + 1))
519 dd if=/dev/zero of=$DIR/${tfile}2 bs=1M count=1
521 estimate=$($LCTL get_param -n osc.$dev-osc-*.timeouts |
522 awk '/portal 6/ {print $5}')
523 echo "service estimate dropped to $estimate"
526 [[ $? == 0 ]] || error "write failed"
528 [[ $? == 0 ]] || error "read failed"
530 restore_lustre_params <$p
531 if [ $brw_size -ne 4194304 ]
533 remount_client $MOUNT || error "remount_client failed"
535 do_facet ost1 "echo $ldlm_enqueue_min_save > $ldlm_enqueue_min"
537 run_test 17b "timeout bulk get, dont evict client (3582)"
540 [ -z ${ost2_svc} ] && skip_env "needs 2 osts" && return 0
542 do_facet_create_file client $TMP/$tfile 20K ||
543 { error_noexit "Create file $TMP/$tfile" ; return 0; }
545 do_facet client mkdir -p $DIR/$tdir
549 pgcache_empty || return 1
552 $LFS setstripe -i 1 -c 1 $f
553 stripe_index=$($LFS getstripe -i $f)
554 if [ $stripe_index -ne 1 ]; then
556 error "$f: stripe_index $stripe_index != 1" && return
559 do_facet client cp $TMP/$tfile $f
561 local osc2dev=`lctl get_param -n devices | grep ${ost2_svc}-osc- | egrep -v 'MDT' | awk '{print $1}'`
562 $LCTL --device $osc2dev deactivate || return 3
563 # my understanding is that there should be nothing in the page
564 # cache after the client reconnects?
566 pgcache_empty || rc=2
567 $LCTL --device $osc2dev activate
571 run_test 18a "manual ost invalidate clears page cache immediately"
574 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
576 do_facet_create_file client $TMP/$tfile 20K ||
577 { error_noexit "Create file $TMP/$tfile" ; return 0; }
579 do_facet client mkdir -p $DIR/$tdir
583 pgcache_empty || return 1
585 $LFS setstripe -i 0 -c 1 $f
586 stripe_index=$($LFS getstripe -i $f)
587 if [ $stripe_index -ne 0 ]; then
589 error "$f: stripe_index $stripe_index != 0" && return
592 do_facet client cp $TMP/$tfile $f
595 # allow recovery to complete
596 sleep $((TIMEOUT + 2))
597 # my understanding is that there should be nothing in the page
598 # cache after the client reconnects?
600 pgcache_empty || rc=2
604 run_test 18b "eviction and reconnect clears page cache (2766)"
607 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
609 do_facet_create_file client $TMP/$tfile 20K ||
610 { error_noexit "Create file $TMP/$tfile" ; return 0; }
612 do_facet client mkdir -p $DIR/$tdir
616 pgcache_empty || return 1
618 $LFS setstripe -i 0 -c 1 $f
619 stripe_index=$($LFS getstripe -i $f)
620 if [ $stripe_index -ne 0 ]; then
622 error "$f: stripe_index $stripe_index != 0" && return
625 do_facet client cp $TMP/$tfile $f
629 # OBD_FAIL_OST_CONNECT_NET2
630 # lost reply to connect request
631 do_facet ost1 lctl set_param fail_loc=0x80000225
634 $LFS df $MOUNT > /dev/null 2>&1
636 # my understanding is that there should be nothing in the page
637 # cache after the client reconnects?
639 pgcache_empty || rc=2
643 run_test 18c "Dropped connect reply after eviction handing (14755)"
646 local BEFORE=`date +%s`
649 mount_client $DIR2 || error "failed to mount $DIR2"
651 # cancel cached locks from OST to avoid eviction from it
654 do_facet client "stat $DIR > /dev/null" ||
655 error "failed to stat $DIR: $?"
656 drop_ldlm_cancel "chmod 0777 $DIR2" ||
657 error "failed to chmod $DIR2"
661 # let the client reconnect
663 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
664 awk -F"[ [,]" '/EVICTED ]$/ \
665 { if (mx<$5) {mx=$5;} } END { print mx }')
667 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
668 (do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state;
669 error "no eviction: $EVICT before:$BEFORE")
671 run_test 19a "test expired_lock_main on mds (2867)"
674 local BEFORE=`date +%s`
677 mount_client $DIR2 || error "failed to mount $DIR2: $?"
679 # cancel cached locks from MDT to avoid eviction from it
682 do_facet client $MULTIOP $DIR/$tfile Ow ||
683 error "failed to run multiop: $?"
684 drop_ldlm_cancel $MULTIOP $DIR2/$tfile Ow ||
685 error "failed to ldlm_cancel: $?"
687 umount_client $DIR2 || error "failed to unmount $DIR2: $?"
688 do_facet client unlink $DIR/$tfile ||
689 error "failed to unlink $DIR/$tfile: $?"
691 # let the client reconnect
693 EVICT=$(do_facet client $LCTL get_param osc.$FSNAME-OST*.state |
694 awk -F"[ [,]" '/EVICTED ]$/ \
695 { if (mx < $5) {mx = $5;} } END { print mx }')
697 [ ! -z "$EVICT" ] && [[ $EVICT -gt $BEFORE ]] ||
698 (do_facet client $LCTL get_param osc.$FSNAME-OST*.state;
699 error "no eviction: $EVICT before:$BEFORE")
701 run_test 19b "test expired_lock_main on ost (2867)"
704 local BEFORE=`date +%s`
707 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0
709 mkdir -p $DIR1/$tfile
712 #define OBD_FAIL_PTLRPC_CANCEL_RESEND 0x516
713 do_facet mds $LCTL set_param fail_loc=0x80000516
715 touch $DIR2/$tfile/file1 &
717 # let touch to get blocked on the server
721 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1
724 # let the client reconnect
726 EVICT=$(do_facet client $LCTL get_param mdc.$FSNAME-MDT*.state |
727 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
729 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
731 run_test 19c "check reconnect and lock resend do not trigger expired_lock_main"
733 test_20a() { # bug 2983 - ldlm_handle_enqueue cleanup
734 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
737 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
738 multiop_bg_pause $DIR/$tdir/${tfile} O_wc || return 1
741 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
742 do_facet ost1 lctl set_param fail_loc=0x80000308
743 kill -USR1 $MULTI_PID
746 [ $rc -eq 0 ] && error "multiop didn't fail enqueue: rc $rc" || true
748 run_test 20a "ldlm_handle_enqueue error (should return error)"
750 test_20b() { # bug 2986 - ldlm_handle_enqueue error during open
751 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
754 $LFS setstripe -i 0 -c 1 $DIR/$tdir/${tfile}
756 #define OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR 0x308
757 do_facet ost1 lctl set_param fail_loc=0x80000308
758 dd if=/etc/hosts of=$DIR/$tdir/$tfile && \
759 error "didn't fail open enqueue" || true
761 run_test 20b "ldlm_handle_enqueue error (should return error)"
764 mkdir_on_mdt0 $DIR/$tdir-1
765 mkdir_on_mdt0 $DIR/$tdir-2
766 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
769 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
770 $MULTIOP $DIR/$tdir-2/f Oc &
773 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
775 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
776 kill -USR1 $close_pid
778 wait $close_pid || return 1
779 wait $open_pid || return 2
780 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
782 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
783 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
787 run_test 21a "drop close request while close and open are both in flight"
790 mkdir_on_mdt0 $DIR/$tdir-1
791 mkdir_on_mdt0 $DIR/$tdir-2
792 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
795 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
796 mcreate $DIR/$tdir-2/f &
799 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
801 kill -USR1 $close_pid
803 wait $close_pid || return 1
804 wait $open_pid || return 3
806 $CHECKSTAT -t file $DIR/$tdir-1/f || return 4
807 $CHECKSTAT -t file $DIR/$tdir-2/f || return 5
810 run_test 21b "drop open request while close and open are both in flight"
813 mkdir_on_mdt0 $DIR/$tdir-1
814 mkdir_on_mdt0 $DIR/$tdir-2
815 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
818 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
819 mcreate $DIR/$tdir-2/f &
822 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
824 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
825 kill -USR1 $close_pid
827 wait $close_pid || return 1
828 wait $open_pid || return 2
830 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
832 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
833 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
836 run_test 21c "drop both request while close and open are both in flight"
839 mkdir_on_mdt0 $DIR/$tdir-1
840 mkdir_on_mdt0 $DIR/$tdir-2
841 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
844 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000129"
845 $MULTIOP $DIR/$tdir-2/f Oc &
847 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
849 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
852 wait $pid || return 1
853 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
855 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
856 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
860 run_test 21d "drop close reply while close and open are both in flight"
863 mkdir_on_mdt0 $DIR/$tdir-1
864 mkdir_on_mdt0 $DIR/$tdir-2
865 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
868 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
869 touch $DIR/$tdir-2/f &
871 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
875 wait $pid || return 1
878 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
879 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
882 run_test 21e "drop open reply while close and open are both in flight"
885 mkdir_on_mdt0 $DIR/$tdir-1
886 mkdir_on_mdt0 $DIR/$tdir-2
887 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
890 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
891 touch $DIR/$tdir-2/f &
893 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
895 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
898 wait $pid || return 1
899 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
901 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
902 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
905 run_test 21f "drop both reply while close and open are both in flight"
908 mkdir_on_mdt0 $DIR/$tdir-1
909 mkdir_on_mdt0 $DIR/$tdir-2
910 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
913 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000119"
914 touch $DIR/$tdir-2/f &
916 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
918 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
921 wait $pid || return 1
922 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
924 $CHECKSTAT -t file $DIR/$tdir-1/f || return 2
925 $CHECKSTAT -t file $DIR/$tdir-2/f || return 3
928 run_test 21g "drop open reply and close request while close and open are both in flight"
931 mkdir_on_mdt0 $DIR/$tdir-1
932 mkdir_on_mdt0 $DIR/$tdir-2
933 multiop_bg_pause $DIR/$tdir-1/f O_c || return 1
936 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000107"
937 touch $DIR/$tdir-2/f &
940 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
942 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000122"
945 wait $pid || return 1
946 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
948 wait $touch_pid || return 2
950 $CHECKSTAT -t file $DIR/$tdir-1/f || return 3
951 $CHECKSTAT -t file $DIR/$tdir-2/f || return 4
954 run_test 21h "drop open request and close reply while close and open are both in flight"
956 # bug 3462 - multiple MDC requests
961 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000115"
966 $MULTIOP $f1 msu || return 1
969 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
971 wait $close_pid || return 2
972 rm -rf $f2 || return 4
974 run_test 22 "drop close request and do mknod"
977 multiop_bg_pause $DIR/$tfile O_c || return 1
979 # give a chance for open
983 drop_request "kill -USR1 $pid"
986 wait $pid || return 1
989 run_test 23 "client hang when close a file after mds crash"
991 test_24a() { # bug 11710 details correct fsync() behavior
992 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
995 $LFS setstripe -i 0 -c 1 $DIR/$tdir
997 multiop_bg_pause $DIR/$tdir/$tfile Owy_wyc || return 1
1000 kill -USR1 $MULTI_PID
1003 lctl set_param fail_loc=0x0
1006 error_ignore bz5494 "multiop didn't fail fsync: rc $rc" || true
1008 run_test 24a "fsync error (should return error)"
1010 wait_client_evicted () {
1013 local varsvc=${facet}_svc
1015 wait_update $(facet_active_host $facet) \
1016 "lctl get_param -n *.${!varsvc}.num_exports | cut -d' ' -f2" \
1021 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1023 dmesg -c > /dev/null
1025 lfs setstripe $DIR/$tdir -S 0 -i 0 -c 1 ||
1026 error "$LFS setstripe failed"
1027 cancel_lru_locks osc
1028 multiop_bg_pause $DIR/$tdir/$tfile-1 Ow8192_yc ||
1029 error "mulitop Ow8192_yc failed"
1032 multiop_bg_pause $DIR/$tdir/$tfile-2 Ow8192_c ||
1033 error "mulitop Ow8192_c failed"
1038 kill -USR1 $MULTI_PID1
1041 kill -USR1 $MULTI_PID2
1044 lctl set_param fail_loc=0x0
1046 [ $rc1 -eq 0 -o $rc2 -eq 0 ] &&
1047 error_ignore bz5494 "multiop didn't fail fsync: $rc1 or close: $rc2" ||
1050 dmesg | grep "dirty page discard:" ||
1051 error "no discarded dirty page found!"
1053 run_test 24b "test dirty page discard due to client eviction"
1055 test_26a() { # was test_26 bug 5921 - evict dead exports by pinger
1056 # this test can only run from a client on a separate node.
1057 remote_ost || { skip "local OST" && return 0; }
1058 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1059 remote_mds || { skip "local MDS" && return 0; }
1061 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
1062 skip "msg and ost1 are at the same node"
1066 check_timeout || return 1
1068 # make sure all imports are connected and not IDLE
1069 do_facet client lfs df > /dev/null
1070 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
1071 do_facet client lctl set_param fail_loc=0x505
1072 local before=$(date +%s)
1075 # evictor takes PING_EVICT_TIMEOUT to evict.
1076 # But if there's a race to start the evictor from various obds,
1077 # the loser might have to wait for the next ping.
1078 sleep $((TIMEOUT * 2))
1079 do_facet client lctl set_param fail_loc=0x0
1080 do_facet client lfs df > /dev/null
1082 local oscs=$(lctl dl | awk '/-osc-/ {print $4}')
1083 check_clients_evicted "$before ${oscs[@]}"
1084 check_clients_full 10 "${oscs[@]}"
1086 run_test 26a "evict dead exports"
1088 test_26b() { # bug 10140 - evict dead exports by pinger
1089 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1091 if [ $(facet_host mgs) = $(facet_host ost1) ]; then
1092 skip "msg and ost1 are at the same node"
1096 check_timeout || return 1
1098 zconf_mount `hostname` $MOUNT2 ||
1099 { error "Failed to mount $MOUNT2"; return 2; }
1100 sleep 1 # wait connections being established
1102 local MDS_NEXP=$(do_facet $SINGLEMDS lctl get_param -n mdt.${mds1_svc}.num_exports | cut -d' ' -f2)
1103 local OST_NEXP=$(do_facet ost1 lctl get_param -n obdfilter.${ost1_svc}.num_exports | cut -d' ' -f2)
1105 echo starting with $OST_NEXP OST and $MDS_NEXP MDS exports
1107 zconf_umount `hostname` $MOUNT2 -f
1109 # PING_INTERVAL max(obd_timeout / 4, 1U)
1110 # PING_EVICT_TIMEOUT (PING_INTERVAL * 6)
1112 # evictor takes PING_EVICT_TIMEOUT to evict.
1113 # But if there's a race to start the evictor from various obds,
1114 # the loser might have to wait for the next ping.
1115 # = 6 * PING_INTERVAL + PING_INTERVAL
1116 # = 7 PING_INTERVAL = 7 obd_timeout / 4 = (1+3/4)obd_timeout
1117 # let's wait $((TIMEOUT * 2)) # bug 19887
1118 wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 2)) ||
1119 error "Client was not evicted by ost"
1120 wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 2)) ||
1121 error "Client was not evicted by mds"
1123 run_test 26b "evict dead exports"
1126 mkdir_on_mdt0 $DIR/$tdir
1127 writemany -q -a $DIR/$tdir/$tfile 0 5 &
1130 local save_FAILURE_MODE=$FAILURE_MODE
1132 facet_failover $SINGLEMDS
1133 #define OBD_FAIL_OSC_SHUTDOWN 0x407
1134 do_facet $SINGLEMDS lctl set_param fail_loc=0x80000407
1135 # need to wait for reconnect
1136 echo waiting for fail_loc
1137 wait_update_facet $SINGLEMDS "lctl get_param -n fail_loc" "-2147482617"
1138 facet_failover $SINGLEMDS
1139 #no crashes allowed!
1140 kill -USR1 $CLIENT_PID
1143 FAILURE_MODE=$save_FAILURE_MODE
1145 run_test 27 "fail LOV while using OSC's"
1147 test_28() { # bug 6086 - error adding new clients
1148 do_facet client mcreate $DIR/$tfile || return 1
1149 drop_bl_callback_once "chmod 0777 $DIR/$tfile" ||
1150 echo "evicted as expected"
1151 #define OBD_FAIL_MDS_CLIENT_ADD 0x12f
1152 do_facet $SINGLEMDS "lctl set_param fail_loc=0x8000012f"
1153 # fail once (evicted), reconnect fail (fail_loc), ok
1154 client_up || (sleep 10; client_up) || (sleep 10; client_up) || error "reconnect failed"
1156 fail $SINGLEMDS # verify MDS last_rcvd can be loaded
1158 run_test 28 "handle error adding new clients (bug 6086)"
1160 test_29a() { # bug 22273 - error adding new clients
1161 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
1162 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000711"
1163 # fail abort so client will be new again
1164 fail_abort $SINGLEMDS
1165 client_up || error "reconnect failed"
1166 wait_osc_import_state $SINGLEMDS ost FULL
1169 run_test 29a "error adding new clients doesn't cause LBUG (bug 22273)"
1171 test_29b() { # bug 22273 - error adding new clients
1172 #define OBD_FAIL_TGT_CLIENT_ADD 0x711
1173 do_facet ost1 "lctl set_param fail_loc=0x80000711"
1174 # fail abort so client will be new again
1176 client_up || error "reconnect failed"
1179 run_test 29b "error adding new clients doesn't cause LBUG (bug 22273)"
1183 # put a load of file creates/writes/deletes
1184 writemany -q $DIR/$tdir/$tfile 0 5 &
1186 echo writemany pid $CLIENT_PID
1190 # wait for client to reconnect to MDS
1195 # client process should see no problems even though MDS went down
1197 kill -USR1 $CLIENT_PID
1200 echo writemany returned $rc
1201 #these may fail because of eviction due to slow AST response.
1203 error_ignore bz13652 "writemany returned rc $rc" || true
1205 run_test 50 "failover MDS under load"
1208 #define OBD_FAIL_MDS_SYNC_CAPA_SL 0x1310
1209 do_facet ost1 lctl set_param fail_loc=0x00001310
1211 mkdir_on_mdt0 $DIR/$tdir
1212 # put a load of file creates/writes/deletes
1213 writemany -q $DIR/$tdir/$tfile 0 5 &
1217 facet_failover $SINGLEMDS
1218 # failover at various points during recovery
1219 SEQ="1 5 10 $(seq $TIMEOUT 5 $(($TIMEOUT+10)))"
1220 echo will failover at $SEQ
1222 #echo failover in $i sec
1223 log "$TESTNAME: failover in $i sec"
1225 facet_failover $SINGLEMDS
1227 # client process should see no problems even though MDS went down
1228 # and recovery was interrupted
1230 kill -USR1 $CLIENT_PID
1233 echo writemany returned $rc
1235 error_ignore bz13652 "writemany returned rc $rc" || true
1237 run_test 51 "failover MDS during recovery"
1240 do_facet client "mkdir -p $DIR/$tdir"
1241 do_facet client "writemany -q -a $DIR/$tdir/$tfile 300 5" &
1243 echo writemany pid $CLIENT_PID
1248 wait $CLIENT_PID || rc=$?
1249 # active client process should see an EIO for down OST
1250 [ $rc -eq 5 ] && { echo "writemany correctly failed $rc" && return 0; }
1251 # but timing or failover setup may allow success
1252 [ $rc -eq 0 ] && { echo "writemany succeeded" && return 0; }
1253 echo "writemany returned $rc"
1258 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1263 [ $rc -ne 0 ] && { return $rc; }
1264 # wait for client to reconnect to OST
1268 [ $rc -ne 0 ] && { return $rc; }
1275 run_test 52 "failover OST under load"
1277 # test of open reconstruct
1280 drop_mdt_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||\
1283 run_test 53a "touch: drop rep"
1288 drop_mdt_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||
1291 run_test 53b "touch: drop rep"
1296 drop_mdt_ldlm_reply "openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tfile" ||
1299 run_test 53c "touch: drop rep"
1302 zconf_mount $(hostname) $MOUNT2
1304 touch $DIR2/$tfile.1
1306 cat $DIR2/$tfile.missing # save transno = 0, rc != 0 into last_rcvd
1309 ERROR=$(dmesg | egrep "(test 54|went back in time)" | tail -n1 |
1310 grep "went back in time")
1311 [ x"$ERROR" == x ] || error "back in time occured"
1313 run_test 54 "back in time"
1315 # bug 11330 - liblustre application death during I/O locks up OST
1317 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
1321 # This test assumes relatively small max_dirty_mb setting
1322 # which we want to walk away from, so just for it we will
1323 # temporarily lower the value
1324 local max_dirty_mb=$(lctl get_param -n osc.*.max_dirty_mb | head -n 1)
1325 lctl set_param -n osc.*.max_dirty_mb=32
1326 stack_trap "lctl set_param osc.*.max_dirty_mb=$max_dirty_mb" EXIT
1328 # Minimum pass speed is 2MBps
1330 # LU-2887/LU-3089 - set min pass speed to 500KBps
1331 [ "$ost1_FSTYPE" = zfs ] && ddtimeout=256
1333 # first dd should be finished quickly
1334 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-1
1335 dd if=/dev/zero of=$DIR/$tdir/$tfile-1 bs=32M count=4 &
1338 echo "step1: testing ......"
1339 while kill -0 $DDPID 2> /dev/null; do
1341 if [ $count -gt $ddtimeout ]; then
1342 error "dd should be finished!"
1346 echo "(dd_pid=$DDPID, time=$count)successful"
1348 $LFS setstripe -c 1 -i 0 $DIR/$tdir/$tfile-2
1349 #define OBD_FAIL_OST_DROP_REQ 0x21d
1350 do_facet ost1 lctl set_param fail_loc=0x0000021d
1351 # second dd will be never finished
1352 dd if=/dev/zero of=$DIR/$tdir/$tfile-2 bs=32M count=4 &
1355 echo "step2: testing ......"
1356 while [ $count -le $ddtimeout ]; do
1357 if ! kill -0 $DDPID 2> /dev/null; then
1359 error "dd shouldn't be finished! (time=$count)"
1364 echo "(dd_pid=$DDPID, time=$count)successful"
1366 #Recover fail_loc and dd will finish soon
1367 do_facet ost1 lctl set_param fail_loc=0
1369 echo "step3: testing ......"
1370 while kill -0 $DDPID 2> /dev/null; do
1372 if [ $count -gt $((ddtimeout + 440)) ]; then
1373 error "dd should be finished!"
1377 echo "(dd_pid=$DDPID, time=$count)successful"
1381 run_test 55 "ost_brw_read/write drops timed-out read/write request"
1383 test_56() { # b=11277
1384 #define OBD_FAIL_MDS_RESEND 0x136
1386 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000136"
1387 stat $DIR/$tfile || error "stat failed"
1388 do_facet $SINGLEMDS "lctl set_param fail_loc=0"
1391 run_test 56 "do not fail on getattr resend"
1394 # no oscs means no client or mdt
1395 while lctl get_param osc.*.* > /dev/null 2>&1; do
1396 : # loop until proc file is removed
1400 test_57() { # bug 10866
1404 #define OBD_FAIL_LPROC_REMOVE 0xB00
1405 lctl set_param fail_loc=0x80000B00
1406 zconf_umount `hostname` $DIR
1407 lctl set_param fail_loc=0x80000B00
1408 fail_abort $SINGLEMDS
1410 lctl set_param fail_loc=0
1412 do_facet client "df $DIR"
1414 run_test 57 "read procfs entries causes kernel crash"
1416 test_58() { # bug 11546
1417 #define OBD_FAIL_MDC_ENQUEUE_PAUSE 0x801
1420 lctl set_param fail_loc=0x80000801
1421 cp $DIR/$tfile /dev/null &
1424 lctl set_param fail_loc=0
1425 drop_bl_callback_once rm -f $DIR/$tfile
1427 # the first 'df' could tigger the eviction caused by
1428 # 'drop_bl_callback_once', and it's normal case.
1429 # but the next 'df' should return successfully.
1430 do_facet client "df $DIR" || do_facet client "df $DIR"
1432 run_test 58 "Eviction in the middle of open RPC reply processing"
1434 test_59() { # bug 10589
1435 zconf_mount `hostname` $MOUNT2 || error "Failed to mount $MOUNT2"
1436 echo $DIR2 | grep -q $MOUNT2 || error "DIR2 is not set properly: $DIR2"
1437 #define OBD_FAIL_LDLM_CANCEL_EVICT_RACE 0x311
1438 lctl set_param fail_loc=0x311
1439 writes=$(LANG=C dd if=/dev/zero of=$DIR2/$tfile count=1 2>&1)
1440 [ $? = 0 ] || error "dd write failed"
1441 writes=$(echo $writes | awk -F '+' '/out/ {print $1}')
1442 lctl set_param fail_loc=0
1444 zconf_umount `hostname` $MOUNT2 -f
1445 reads=$(LANG=C dd if=$DIR/$tfile of=/dev/null 2>&1)
1446 [ $? = 0 ] || error "dd read failed"
1447 reads=$(echo $reads | awk -F '+' '/in/ {print $1}')
1448 [ "$reads" -eq "$writes" ] || error "read" $reads "blocks, must be" $writes
1450 run_test 59 "Read cancel race on client eviction"
1453 local num_files=${COUNT:-5000}
1454 test_mkdir $DIR/$tdir
1456 # Register (and start) changelog
1457 changelog_register || error "changelog_register failed"
1459 # Generate a large number of changelog entries
1460 createmany -o $DIR/$tdir/$tfile $num_files
1464 # Unlink files in the background
1465 unlinkmany $DIR/$tdir/$tfile $num_files &
1469 # Failover the MDS while unlinks are happening
1470 facet_failover $SINGLEMDS
1472 # Wait for unlinkmany to finish
1475 # Check if all the create/unlink events were recorded
1477 local cl_count=$(changelog_dump | grep -c UNLNK)
1478 echo "$cl_count unlinks in changelog"
1480 changelog_deregister || error "changelog_deregister failed"
1481 if ! changelog_users $SINGLEMDS | grep -q "^cl"; then
1482 [ $cl_count -eq $num_files ] ||
1483 error "Recorded $cl_count of $num_files unlinks"
1484 # Also make sure we can clear large changelogs
1485 cl_count=$(changelog_dump | wc -l)
1486 [ $cl_count -le 2 ] ||
1487 error "Changelog not empty: $cl_count entries"
1488 else # If other users, there may be other unlinks in the log
1489 [ $cl_count -ge $num_files ] ||
1490 error "Recorded $cl_count of $num_files unlinks"
1491 changelog_users $SINGLEMDS
1492 echo "other changelog users; can't verify clear"
1495 run_test 60 "Add Changelog entries during MDS failover"
1499 local mdtosc=$(get_mdtosc_proc_path $SINGLEMDS $FSNAME-OST0000)
1500 mdtosc=${mdtosc/-MDT*/-MDT\*}
1501 local cflags="osc.$mdtosc.connect_flags"
1502 do_facet $SINGLEMDS "lctl get_param -n $cflags" |grep -q skip_orphan
1503 [ $? -ne 0 ] && skip "don't have skip orphan feature" && return
1505 mkdir_on_mdt0 $DIR/$tdir || error "mkdir dir $DIR/$tdir failed"
1506 # Set the default stripe of $DIR/$tdir to put the files to ost1
1507 $LFS setstripe -c 1 -i 0 $DIR/$tdir
1509 replay_barrier $SINGLEMDS
1510 createmany -o $DIR/$tdir/$tfile-%d 10
1511 local oid=$(do_facet ost1 "lctl get_param -n \
1512 obdfilter.${ost1_svc}.last_id" | sed -e 's/.*://')
1514 fail_abort $SINGLEMDS
1516 touch $DIR/$tdir/$tfile
1517 local id=$($LFS getstripe $DIR/$tdir/$tfile |
1518 awk '$1 == 0 { print $2 }')
1519 [ $id -le $oid ] && error "the orphan objid was reused, failed"
1524 run_test 61 "Verify to not reuse orphan objects - bug 17025"
1526 # test_62 as seen it b2_1 please do not reuse test_62
1529 # zconf_umount `hostname` $DIR
1530 # #define OBD_FAIL_PTLRPC_DELAY_IMP_FULL 0x516
1531 # lctl set_param fail_loc=0x516
1534 #run_test 62 "Verify connection flags race - bug LU-1716"
1539 #grant lock1, export2
1540 $LFS setstripe -i -0 $DIR2/$tfile || error "setstripe failed"
1541 $MULTIOP $DIR2/$tfile Ow || error "multiop failed"
1543 #define OBD_FAIL_LDLM_BL_EVICT 0x31e
1544 do_facet ost $LCTL set_param fail_loc=0x31e
1545 #get waiting lock2, export1
1546 $MULTIOP $DIR/$tfile Ow &
1548 # let enqueue to get asleep
1552 $MULTIOP $DIR2/$tfile Ow &
1560 do_facet ost $LCTL set_param fail_loc=0
1567 run_test 65 "lock enqueue for destroyed export"
1571 [[ "$MDS1_VERSION" -ge $(version_code 2.7.51) ]] ||
1572 skip "Need MDS version at least 2.7.51"
1574 local list=$(comma_list $(osts_nodes))
1576 # modify dir so that next revalidate would not obtain UPDATE lock
1579 # drop 1 reply with UPDATE lock
1580 mcreate $DIR/$tfile || error "mcreate failed: $?"
1581 drop_mdt_ldlm_reply_once "stat $DIR/$tfile" &
1584 # make the re-sent lock to sleep
1585 #define OBD_FAIL_MDS_RESEND 0x136
1586 do_nodes $list $LCTL set_param fail_loc=0x80000136
1588 #initiate the re-connect & re-send
1589 local mdtname="MDT0000"
1590 local mdccli=$($LCTL dl | grep "${mdtname}-mdc" | awk '{print $4;}')
1591 local conn_uuid=$($LCTL get_param -n mdc.${mdccli}.conn_uuid)
1592 $LCTL set_param "mdc.${mdccli}.import=connection=${conn_uuid}"
1595 #initiate the client eviction while enqueue re-send is in progress
1601 run_test 66 "lock enqueue re-send vs client eviction"
1605 #define OBD_FAIL_PTLRPC_CONNECT_RACE 0x531
1606 $LCTL set_param fail_loc=0x80000531
1608 local mdtname="MDT0000"
1609 local mdccli=$($LCTL dl | grep "${mdtname}-mdc" | awk '{print $4;}')
1610 local conn_uuid=$($LCTL get_param -n mdc.${mdccli}.mds_conn_uuid)
1611 $LCTL set_param "mdc.${mdccli}.import=connection=${conn_uuid}" &
1620 run_test 67 "connect vs import invalidate race"
1622 check_cli_ir_state()
1624 local NODE=${1:-$HOSTNAME}
1626 st=$(do_node $NODE "lctl get_param mgc.*.ir_state |
1627 awk '/imperative_recovery:/ { print \\\$2}'")
1628 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1629 error "Error state $st, must be ENABLED or DISABLED"
1633 check_target_ir_state()
1636 local name=${target}_svc
1637 local recovery_proc=obdfilter.${!name}.recovery_status
1641 st=$(do_facet $target "$LCTL get_param -n $recovery_proc |
1642 awk '/status:/{ print \\\$2}'")
1643 [ x$st = xRECOVERING ] || break
1645 st=$(do_facet $target "lctl get_param -n $recovery_proc |
1646 awk '/IR:/{ print \\\$2}'")
1647 [ $st != ON -o $st != OFF -o $st != ENABLED -o $st != DISABLED ] ||
1648 error "Error state $st, must be ENABLED or DISABLED"
1654 do_facet mgs lctl set_param -n mgs.MGS.live.$FSNAME="state=$1"
1659 local state=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1660 awk '/state:/{ print \\\$2 }'")
1664 nidtbl_version_mgs()
1666 local ver=$(do_facet mgs "lctl get_param -n mgs.MGS.live.$FSNAME |
1667 awk '/nidtbl_version:/{ print \\\$2 }'")
1671 # nidtbl_version_client <mds1|client> [node]
1672 nidtbl_version_client()
1675 local node=${2:-$HOSTNAME}
1677 if [ X$cli = Xclient ]; then
1680 local obdtype=${cli/%[0-9]*/}
1681 [ $obdtype != mds ] && error "wrong parameters $cli"
1683 node=$(facet_active_host $cli)
1688 local vers=$(do_node $node "lctl get_param -n mgc.*.ir_state" |
1689 awk "/$cli/{print \$6}" |sort -u)
1691 # in case there are multiple mounts on the client node
1693 [ ${#arr[@]} -ne 1 ] && error "versions on client node mismatch"
1697 nidtbl_versions_match()
1699 [ $(nidtbl_version_mgs) -eq $(nidtbl_version_client ${1:-client}) ]
1702 target_instance_match()
1708 obdtype=${srv/%[0-9]*/}
1719 error "invalid target type" $srv
1724 local target=${srv}_svc
1725 local si=$(do_facet $srv lctl get_param -n $obdname.${!target}.instance)
1726 local ci=$(lctl get_param -n $cliname.${!target}-${cliname}-*.import |
1727 awk '/instance/{ print $2 }' | head -n1)
1729 return $([ $si -eq $ci ])
1734 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1735 { skip "MGS without IR support"; return 0; }
1737 # MDT was just restarted in the previous test, make sure everything
1740 while [ $cnt -gt 0 ]; do
1741 nidtbl_versions_match && break
1747 set_ir_status disabled
1749 local prev_ver=$(nidtbl_version_client client)
1751 local saved_FAILURE_MODE=$FAILURE_MODE
1752 [ $(facet_host mgs) = $(facet_host ost1) ] && FAILURE_MODE="SOFT"
1756 [ $(nidtbl_version_client client) -eq $prev_ver ] ||
1757 error "version must not change due to IR disabled"
1758 target_instance_match ost1 || error "instance mismatch"
1762 FAILURE_MODE=$saved_FAILURE_MODE
1764 run_test 100 "IR: Make sure normal recovery still works w/o IR"
1768 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1769 skip "MGS without IR support"
1773 local ost1_imp=$(get_osc_import_name client ost1)
1775 # disable pinger recovery
1776 lctl set_param -n osc.$ost1_imp.pinger_recov=0
1777 stack_trap "$LCTL set_param -n osc.$ost1_imp.pinger_recov=1" EXIT
1781 target_instance_match ost1 || error "instance mismatch"
1782 nidtbl_versions_match || error "version must match"
1784 run_test 101a "IR: Make sure IR works w/o normal recovery"
1788 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1789 skip "MGS without IR support"
1793 local ost1_imp=$(get_osc_import_name client ost1)
1795 #define OBD_FAIL_OST_PREPARE_DELAY 0x247
1796 do_facet ost1 "$LCTL set_param fail_loc=0x247"
1797 # disable pinger recovery
1798 $LCTL set_param -n osc.$ost1_imp.pinger_recov=0
1799 stack_trap "$LCTL set_param -n osc.$ost1_imp.pinger_recov=1" EXIT
1801 #OST may return EAGAIN if it is not configured yet
1804 run_test 101b "IR: Make sure IR works w/o normal recovery and proceed EAGAIN"
1808 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1809 { skip "MGS without IR support"; return 0; }
1811 local clients=${CLIENTS:-$HOSTNAME}
1818 # let's have a new nidtbl version
1821 # sleep for a while so that clients can see the failure of ost
1822 # it must be MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC.
1823 # int mgc_request.c:
1824 # define MGC_TIMEOUT_MIN_SECONDS 5
1825 # define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 *
1826 local count=30 # 20 seconds at most
1827 while [ $count -gt 0 ]; do
1828 nidtbl_versions_match && break
1833 nidtbl_versions_match || error "nidtbl mismatch"
1836 old_version=$(nidtbl_version_client client)
1838 zconf_umount_clients $clients $MOUNT || error "Cannot umount client"
1841 combined_mgs_mds && mgsdev=mds1
1842 remount_facet $mgsdev
1845 zconf_mount_clients $clients $MOUNT || error "Cannot mount client"
1848 new_version=$(nidtbl_version_client client)
1849 [ $new_version -lt $old_version ] &&
1850 error "nidtbl version wrong after mgs restarts"
1853 run_test 102 "IR: New client gets updated nidtbl after MGS restart"
1857 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1858 { skip "MGS without IR support"; return 0; }
1860 combined_mgs_mds && skip "needs separate mgs and mds" && return 0
1862 # workaround solution to generate config log on the mds
1868 # We need this test because mds is like a client in IR context.
1869 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS ||
1870 error "MDS should start w/o mgs"
1872 # start mgs and remount mds w/ ir
1873 start mgs $(mgsdevname) $MGS_MOUNT_OPTS
1876 # remount client so that fsdb will be created on the MGS
1877 umount_client $MOUNT || error "umount failed"
1878 mount_client $MOUNT || error "mount failed"
1880 # sleep 30 seconds so the MDS has a chance to detect MGS restarting
1882 while [ $count -gt 0 ]; do
1883 [ $(nidtbl_version_client mds1) -ne 0 ] && break
1888 # after a while, mds should be able to reconnect to mgs and fetch
1889 # up-to-date nidtbl version
1890 nidtbl_versions_match mds1 || error "mds nidtbl mismatch"
1895 run_test 103 "IR: MDS can start w/o MGS and get updated nidtbl later"
1899 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1900 { skip "MGS without IR support"; return 0; }
1905 start ost1 $(ostdevname 1) "$OST_MOUNT_OPTS -onoir" ||
1906 error "OST1 cannot start"
1909 local ir_state=$(check_target_ir_state ost1)
1911 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1912 error "ir status on ost1 should be DISABLED"
1914 run_test 104 "IR: ost can disable IR voluntarily"
1918 [ -z "$RCLIENTS" ] && skip "Needs multiple clients" && return 0
1919 do_facet mgs $LCTL list_param mgs.*.ir_timeout ||
1920 { skip "MGS without IR support"; return 0; }
1924 # get one of the clients from client list
1925 local rcli=$(echo $RCLIENTS |cut -d' ' -f 1)
1927 local mount_opts=${MOUNT_OPTS:+$MOUNT_OPTS,}noir
1928 zconf_umount $rcli $MOUNT || error "umount failed"
1929 zconf_mount $rcli $MOUNT $mount_opts || error "mount failed"
1931 # make sure lustre mount at $rcli disabling IR
1932 local ir_state=$(check_cli_ir_state $rcli)
1933 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1934 error "IR state must be DISABLED at $rcli"
1936 # Since the client just mounted, its last_rcvd entry is not on disk.
1937 # Send an RPC so exp_need_sync forces last_rcvd to commit this export
1938 # so the client can reconnect during OST recovery (LU-924, LU-1582)
1939 $LFS setstripe -i 0 $DIR/$tfile
1940 dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
1942 # make sure MGS's state is Partial
1943 [ $(get_ir_status) = "partial" ] || error "MGS IR state must be partial"
1946 # make sure IR on ost1 is DISABLED
1947 local ir_state=$(check_target_ir_state ost1)
1948 [ $ir_state = "DISABLED" -o $ir_state = "OFF" ] ||
1949 error "IR status on ost1 should be DISABLED"
1951 # remount with the default MOUNT_OPTS
1952 zconf_umount $rcli $MOUNT || error "umount failed"
1953 zconf_mount $rcli $MOUNT || error "mount failed"
1955 # make sure MGS's state is full
1956 [ $(get_ir_status) = "full" ] || error "MGS IR status must be full"
1959 # make sure IR on ost1 is ENABLED
1960 local ir_state=$(check_target_ir_state ost1)
1961 [ $ir_state = "ENABLED" -o $ir_state = "ON" ] ||
1962 error "IR status on ost1 should be ENABLED"
1966 run_test 105 "IR: NON IR clients support"
1974 test_106() { # LU-1789
1975 [[ "$MDS1_VERSION" -ge $(version_code 2.3.50) ]] ||
1976 skip "Need MDS version at least 2.3.50"
1978 #define OBD_FAIL_MDC_LIGHTWEIGHT 0x805
1979 $LCTL set_param fail_loc=0x805
1982 trap cleanup_106 EXIT
1984 # enable lightweight flag on mdc connection
1987 local MDS_NEXP=$(do_facet $SINGLEMDS \
1988 lctl get_param -n mdt.${mds1_svc}.num_exports |
1990 $LCTL set_param fail_loc=0
1992 touch $DIR2/$tfile || error "failed to create empty file"
1993 replay_barrier $SINGLEMDS
1995 $LCTL set_param debug=ha
1997 facet_failover $SINGLEMDS
1999 # lightweight goes through LUSTRE_IMP_RECOVER during failover
2000 touch -c $DIR2/$tfile || true
2001 $LCTL dk $TMP/lustre-log-$TESTNAME.log
2002 recovered=$(awk '/MDT0000-mdc-[0-9a-f]*. lwp recover/ { print }' \
2003 $TMP/lustre-log-$TESTNAME.log)
2004 [ -z "$recovered" ] && error "lightweight client was not recovered"
2006 # and all operations performed by lightweight client should be
2007 # synchronous, so the file created before mds restart should be there
2008 $CHECKSTAT -t file $DIR/$tfile || error "file not present"
2013 run_test 106 "lightweight connection support"
2019 mkdir_on_mdt0 $DIR/$tdir
2020 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2021 do_facet $SINGLEMDS lctl set_param fail_loc=0x119
2022 multiop $DIR/$tdir D_c &
2024 mkdir $DIR/$tdir/dir_106 &
2026 do_facet $SINGLEMDS lctl set_param fail_loc=0
2029 wait $CLIENT_PID || rc=$?
2030 checkstat -t dir $DIR/$tdir/dir_106 || return 1
2032 kill -USR1 $close_pid
2033 wait $close_pid || return 2
2037 run_test 107 "drop reint reply, then restart MDT"
2041 $LFS setstripe -c 1 -i 0 $DIR/$tdir
2043 dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=256 &
2051 client_up || error "reconnect failed"
2052 rm -f $DIR/$tdir/$tfile
2054 run_test 108 "client eviction don't crash"
2057 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2058 local remote_dir=$DIR/$tdir/remote_dir
2062 #prepare for 110 test, which need set striped dir on remote MDT.
2063 for num in $(seq $MDSCOUNT); do
2065 lctl set_param -n mdt.$FSNAME*.enable_remote_dir=1 \
2069 mkdir_on_mdt0 $DIR/$tdir
2070 drop_request "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
2071 error "lfs mkdir failed"
2072 local diridx=$($LFS getstripe -m $remote_dir)
2073 [ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
2075 rm -rf $DIR/$tdir || error "rmdir failed"
2077 run_test 110a "create remote directory: drop client req"
2080 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2081 local remote_dir=$DIR/$tdir/remote_dir
2084 mkdir_on_mdt0 $DIR/$tdir
2085 drop_reint_reply "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
2086 error "lfs mkdir failed"
2088 diridx=$($LFS getstripe -m $remote_dir)
2089 [ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
2091 rm -rf $DIR/$tdir || error "rmdir failed"
2093 run_test 110b "create remote directory: drop Master rep"
2096 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2097 local remote_dir=$DIR/$tdir/remote_dir
2100 mkdir_on_mdt0 $DIR/$tdir
2101 drop_update_reply $mdtidx "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
2102 error "lfs mkdir failed"
2104 diridx=$($LFS getstripe -m $remote_dir)
2105 [ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
2107 rm -rf $DIR/$tdir || error "rmdir failed"
2109 run_test 110c "create remote directory: drop update rep on slave MDT"
2112 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2113 local remote_dir=$DIR/$tdir/remote_dir
2116 mkdir_on_mdt0 $DIR/$tdir
2117 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
2119 drop_request "rm -rf $remote_dir" || error "rm remote dir failed"
2121 rm -rf $DIR/$tdir || error "rmdir failed"
2125 run_test 110d "remove remote directory: drop client req"
2128 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2129 local remote_dir=$DIR/$tdir/remote_dir
2132 mkdir_on_mdt0 $DIR/$tdir
2133 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
2134 drop_reint_reply "rm -rf $remote_dir" || error "rm remote dir failed"
2136 rm -rf $DIR/$tdir || error "rmdir failed"
2140 run_test 110e "remove remote directory: drop master rep"
2143 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2144 local remote_dir=$DIR/$tdir/remote_dir
2147 mkdir_on_mdt0 $DIR/$tdir
2148 $LFS mkdir -i $MDTIDX -c2 $remote_dir || error "lfs mkdir failed"
2149 drop_update_reply $MDTIDX "rm -rf $remote_dir" ||
2150 error "rm remote dir failed"
2152 rm -rf $DIR/$tdir || error "rmdir failed"
2154 run_test 110f "remove remote directory: drop slave rep"
2157 [[ "$MDS1_VERSION" -ge $(version_code 2.11.0) ]] ||
2158 skip "Need MDS version at least 2.11.0"
2160 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2162 mkdir_on_mdt0 $DIR/$tdir
2163 touch $DIR/$tdir/$tfile
2165 # OBD_FAIL_MDS_REINT_NET_REP 0x119
2166 do_facet mds1 $LCTL set_param fail_loc=0x119
2167 $LFS migrate -m 1 $DIR/$tdir &
2170 do_facet mds1 $LCTL set_param fail_loc=0
2174 mdt_index=$($LFS getstripe -m $DIR/$tdir)
2175 [ $mdt_index == 1 ] || error "$tdir is not on MDT1"
2176 mdt_index=$($LFS getstripe -m $DIR/$tdir/$tfile)
2177 [ $mdt_index == 1 ] || error "$tfile is not on MDT1"
2179 rm -rf $DIR/$tdir || error "rmdir failed"
2181 run_test 110g "drop reply during migration"
2184 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2185 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
2186 skip "Need MDS version at least 2.7.56"
2188 local src_dir=$DIR/$tdir/source_dir
2189 local tgt_dir=$DIR/$tdir/target_dir
2193 $LFS mkdir -i $MDTIDX $tgt_dir
2195 dd if=/etc/hosts of=$src_dir/src_file
2196 touch $tgt_dir/tgt_file
2197 drop_update_reply $MDTIDX \
2198 "mrename $src_dir/src_file $tgt_dir/tgt_file" ||
2199 error "mrename failed"
2201 $CHECKSTAT -t file $src_dir/src_file &&
2202 error "src_file present after rename"
2204 diff /etc/hosts $tgt_dir/tgt_file ||
2205 error "file changed after rename"
2208 run_test 110h "drop update reply during cross-MDT file rename"
2211 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2212 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
2213 skip "Need MDS version at least 2.7.56"
2215 local src_dir=$DIR/$tdir/source_dir
2216 local tgt_dir=$DIR/$tdir/target_dir
2220 $LFS mkdir -i $MDTIDX $tgt_dir
2222 mkdir $src_dir/src_dir
2223 touch $src_dir/src_dir/a
2224 mkdir $tgt_dir/tgt_dir
2225 drop_update_reply $MDTIDX \
2226 "mrename $src_dir/src_dir $tgt_dir/tgt_dir" ||
2227 error "mrename failed"
2229 $CHECKSTAT -t dir $src_dir/src_dir &&
2230 error "src_dir present after rename"
2232 $CHECKSTAT -t dir $tgt_dir/tgt_dir ||
2233 error "tgt_dir not present after rename"
2235 $CHECKSTAT -t file $tgt_dir/tgt_dir/a ||
2236 error "a not present after rename"
2238 run_test 110i "drop update reply during cross-MDT dir rename"
2241 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
2242 [[ "$MDS1_VERSION" -ge $(version_code 2.7.56) ]] ||
2243 skip "Need MDS version at least 2.7.56"
2245 local remote_dir=$DIR/$tdir/remote_dir
2246 local local_dir=$DIR/$tdir/local_dir
2249 mkdir_on_mdt0 $DIR/$tdir
2250 mkdir $DIR/$tdir/local_dir
2251 $LFS mkdir -i $MDTIDX $remote_dir
2253 touch $local_dir/local_file
2254 drop_update_reply $MDTIDX \
2255 "ln $local_dir/local_file $remote_dir/remote_file" ||
2258 $CHECKSTAT -t file $remote_dir/remote_file ||
2259 error "remote not present after ln"
2261 run_test 110j "drop update reply during cross-MDT ln"
2264 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTS"
2265 [[ "$MDS1_VERSION" -ge $(version_code 2.12.55) ]] ||
2266 skip "Need MDS version at least 2.12.55"
2269 stop mds2 || error "stop mds2 failed"
2271 #define OBD_FAIL_FLD_QUERY_REQ 0x1103
2272 do_facet mds2 lctl set_param fail_loc=0x80001103
2273 local OPTS="$MDS_MOUNT_OPTS -o abort_recovery"
2274 start mds2 $(mdsdevname 2) $OPTS ||
2275 error "start MDS with abort_recovery should succeed"
2276 do_facet mds2 lctl set_param fail_loc=0
2279 stop mds2 || error "cleanup: stop mds2 failed"
2280 start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS ||
2281 error "cleanup: start mds2 failed"
2282 zconf_mount $(hostname) $MOUNT || error "cleanup: mount failed"
2283 client_up || error "post-failover df failed"
2286 run_test 110k "FID_QUERY failed during recovery"
2289 (( $(lustre_version_code $SINGLEMDS) >= $(version_code 2.14.52) )) ||
2290 skip "Need MDS version at least 2.14.52"
2291 (( $MDSCOUNT >= 2 )) || skip "needs at least 2 MDTs"
2292 local remote_dir=$DIR/$tdir/remote_dir
2297 mkdir_on_mdt0 $DIR/$tdir
2299 #define OBD_FAIL_PTLRPC_RESEND_RACE 0x0525
2300 do_facet mds1 $LCTL set_param fail_loc=0x80000525
2301 $LFS mkdir -i 1 -c2 $remote_dir &
2304 # initiate the re-connect & re-send
2305 mdccli=$(do_facet mds2 $LCTL dl |
2306 awk '/MDT0000-osp-MDT0001/ {print $4;}')
2307 uuid=$(do_facet mds2 $LCTL get_param -n osp.$mdccli.mdt_conn_uuid)
2308 echo "conn_uuid=$uuid"
2309 do_facet mds2 $LCTL set_param "osp.$mdccli.import=connection=$uuid"
2312 (( $? == 0 )) || error "mkdir failed"
2314 diridx=$($LFS getstripe -m $remote_dir)
2315 (( $diridx == 1 )) || error "$diridx != 1"
2316 rm -rf $DIR/$tdir || error "rmdir failed"
2318 run_test 110m "update resent vs original RPC race"
2320 # LU-2844 mdt prepare fail should not cause umount oops
2323 [[ "$MDS1_VERSION" -ge $(version_code 2.3.62) ]] ||
2324 skip "Need MDS version at least 2.3.62"
2326 #define OBD_FAIL_MDS_CHANGELOG_INIT 0x151
2327 do_facet $SINGLEMDS lctl set_param fail_loc=0x151
2328 stop $SINGLEMDS || error "stop MDS failed"
2329 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS &&
2330 error "start MDS should fail"
2331 do_facet $SINGLEMDS lctl set_param fail_loc=0
2332 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
2333 error "start MDS failed"
2335 run_test 111 "mdd setup fail should not cause umount oops"
2339 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
2341 do_facet_random_file client $TMP/$tfile 100K ||
2342 error_noexit "Create random file $TMP/$tfile"
2344 pause_bulk "cp $TMP/$tfile $DIR/$tfile" $TIMEOUT ||
2345 error_noexit "Can't pause_bulk copy"
2348 # expect cmp to succeed, client resent bulk
2349 cmp $TMP/$tfile $DIR/$tfile ||
2350 error_noexit "Wrong data has been written"
2352 error_noexit "Can't remove file"
2355 run_test 112a "bulk resend while orignal request is in progress"
2362 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1
2363 cancel_lru_locks osc
2366 $LCTL set_param fail_loc=$fail1 fail_val=3
2367 dd of=/dev/null if=$DIR/$tfile bs=4096 count=1 &
2371 set_nodes_failloc "$(osts_nodes)" $fail2
2373 wait $pid || error "dd failed"
2381 local fail_val2=${4:-0}
2387 $LCTL set_param fail_loc=$fail1 fail_val=4
2388 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=dsync &
2393 set_nodes_failloc "$(osts_nodes)" $fail2 $fail_val2
2397 [ $error -eq 0 ] && [ $rc -ne 0 ] && error "dd error ($rc)"
2398 [ $error -ne 0 ] && [ $rc -eq 0 ] && error "dd success"
2403 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2404 skip "need at least 2.8.50 on OST"
2406 #define OBD_FAIL_PTLRPC_LONG_REQ_UNLINK 0x51b
2407 #define OBD_FAIL_PTLRPC_DROP_BULK 0x51a
2408 test_115_read 0x8000051b 0x8000051a
2410 run_test 115a "read: late REQ MDunlink and no bulk"
2413 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2414 skip "need at least 2.8.50 on OST"
2416 #define OBD_FAIL_PTLRPC_LONG_REQ_UNLINK 0x51b
2417 #define OBD_FAIL_OST_ENOSPC 0x215
2419 # pass $OSTCOUNT for the fail_loc to be caught
2420 # appropriately by the IO thread
2421 test_115_write 0x8000051b 0x80000215 1 $OSTCOUNT
2423 run_test 115b "write: late REQ MDunlink and no bulk"
2426 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2427 skip "need at least 2.8.50 on OST"
2429 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
2430 #define OBD_FAIL_PTLRPC_DROP_BULK 0x51a
2431 test_115_read 0x8000050f 0x8000051a
2433 run_test 115c "read: late Reply MDunlink and no bulk"
2436 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2437 skip "need at least 2.8.50 on OST"
2439 #define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f
2440 #define OBD_FAIL_OST_ENOSPC 0x215
2441 test_115_write 0x8000050f 0x80000215 0
2443 run_test 115d "write: late Reply MDunlink and no bulk"
2446 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2447 skip "need at least 2.8.50 on OST"
2449 #define OBD_FAIL_PTLRPC_LONG_BULK_UNLINK 0x510
2450 #define OBD_FAIL_OST_ALL_REPLY_NET 0x211
2451 test_115_read 0x80000510 0x80000211
2453 run_test 115e "read: late Bulk MDunlink and no reply"
2456 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2457 skip "need at least 2.8.50 on OST"
2459 #define OBD_FAIL_PTLRPC_LONG_REQ_UNLINK 0x51b
2460 #define OBD_FAIL_OST_ALL_REPLY_NET 0x211
2461 test_115_read 0x8000051b 0x80000211
2463 run_test 115f "read: late REQ MDunlink and no reply"
2466 [ "$OST1_VERSION" -lt $(version_code 2.8.50) ] &&
2467 skip "need at least 2.8.50 on OST"
2469 #define OBD_FAIL_PTLRPC_LONG_BOTH_UNLINK 0x51c
2470 test_115_read 0x8000051c 0
2472 run_test 115g "read: late REQ MDunlink and Reply MDunlink"
2474 # parameters: fail_loc CMD RC
2481 #define OBD_FAIL_LDLM_CP_CB_WAIT2 0x320
2482 #define OBD_FAIL_LDLM_CP_CB_WAIT3 0x321
2483 #define OBD_FAIL_LDLM_CP_CB_WAIT4 0x322
2484 #define OBD_FAIL_LDLM_CP_CB_WAIT5 0x323
2487 echo -n "** FLOCK REPLY vs. EVICTION race, lock $2"
2488 [ "$1" = "CLEANUP" ] &&
2489 fail=0x80000320 && echo ", $1 cp first"
2490 [ "$1" = "REPLY" ] &&
2491 fail=0x80000321 && echo ", $1 cp first"
2492 [ "$1" = "DEADLOCK CLEANUP" ] &&
2493 fail=0x80000322 && echo " DEADLOCK, CLEANUP cp first"
2494 [ "$1" = "DEADLOCK REPLY" ] &&
2495 fail=0x80000323 && echo " DEADLOCK, REPLY cp first"
2497 if [ x"$2" = x"get" ]; then
2498 #for TEST lock, take a conflict in advance
2499 # sleep longer than evictor to not confuse fail_loc: 2+2+4
2500 echo "** Taking conflict **"
2501 flocks_test 5 set read sleep 10 $DIR/$tfile &
2507 $LCTL set_param fail_loc=$fail
2509 flocks_test 5 $2 write $DIR/$tfile &
2513 echo "** Evicting and re-connecting client **"
2518 if [ x"$2" = x"get" ]; then
2525 # check if the return value is allowed
2526 [ $rc -eq $3 ] && rc=0
2528 $LCTL set_param fail_loc=0
2532 # a lock is taken, unlock vs. cleanup_resource() race for destroying
2533 # the ORIGINAL lock.
2538 flocks_test 5 set write sleep 4 $DIR/$tfile &
2542 # let unlock to sleep in CP CB
2543 $LCTL set_param fail_loc=$1
2546 # let cleanup to cleep in CP CB
2554 $LCTL set_param fail_loc=0
2559 flock_is_enabled || { skip "mount w/o flock enabled" && return; }
2562 test_120_reply "CLEANUP" set 5 || error "SET race failed"
2563 test_120_reply "CLEANUP" get 5 || error "GET race failed"
2564 test_120_reply "CLEANUP" unlock 5 || error "UNLOCK race failed"
2566 test_120_reply "REPLY" set 5 || error "SET race failed"
2567 test_120_reply "REPLY" get 5 || error "GET race failed"
2568 test_120_reply "REPLY" unlock 5 || error "UNLOCK race failed"
2571 test_120_reply "DEADLOCK CLEANUP" set 5 || error "DEADLOCK race failed"
2572 test_120_reply "DEADLOCK REPLY" set 35 || error "DEADLOCK race failed"
2574 test_120_destroy 0x320 || error "unlock-cleanup race failed"
2576 run_test 120 "flock race: completion vs. evict"
2579 local BEFORE=$(date +%s)
2582 # modify dir so that next revalidate would not obtain UPDATE lock
2585 # drop 1 reply with UPDATE lock,
2586 # resend should not create 2nd lock on server
2587 mcreate $DIR/$tfile || error "mcreate failed: $?"
2588 drop_mdt_ldlm_reply_once "stat $DIR/$tfile" || error "stat failed: $?"
2590 # 2 BL AST will be sent to client, both must find the same lock,
2591 # race them to not get EINVAL for 2nd BL AST
2592 #define OBD_FAIL_LDLM_PAUSE_CANCEL2 0x31f
2593 $LCTL set_param fail_loc=0x8000031f
2595 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=0 > /dev/null
2596 chmod 0777 $DIR/$tfile || error "chmod failed: $?"
2597 $LCTL set_param ldlm.namespaces.*.early_lock_cancel=1 > /dev/null
2599 # let the client reconnect
2601 EVICT=$($LCTL get_param mdc.$FSNAME-MDT*.state |
2602 awk -F"[ [,]" '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
2604 [ -z "$EVICT" ] || [[ $EVICT -le $BEFORE ]] || error "eviction happened"
2606 run_test 113 "ldlm enqueue dropped reply should not cause deadlocks"
2610 test_mkdir -p -c1 $DIR/$tdir
2612 # Prevent interference from layout intent RPCs due to
2613 # asynchronous writeback. These will be tested in 130c below.
2614 do_nodes ${CLIENTS:-$HOSTNAME} sync
2616 # get only LOOKUP lock on $tdir
2617 cancel_lru_locks mdc
2618 ls $DIR/$tdir/$tfile 2>/dev/null
2620 # get getattr by fid on $tdir
2622 # we need to race with unlink, unlink must complete before we will
2623 # take a DLM lock, otherwise unlink will wait until getattr will
2624 # complete; but later than getattr starts so that getattr found
2626 #define OBD_FAIL_MDS_INTENT_DELAY 0x160
2627 set_nodes_failloc "$(mdts_nodes)" 0x80000160
2634 # drop the reply so that resend happens on an unlinked file.
2635 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2636 set_nodes_failloc "$(mdts_nodes)" 0x80000157
2640 remote_mds_nodsh && skip "remote MDS with nodsh"
2641 [[ "$MDS1_VERSION" -ge $(version_code 2.7.2) ]] ||
2642 skip "Need server version newer than 2.7.1"
2646 wait $T130_PID || [ $? -eq 0 ] && error "stat should fail"
2649 run_test 130a "enqueue resend on not existing file"
2652 remote_mds_nodsh && skip "remote MDS with nodsh"
2653 [[ "$MDS1_VERSION" -ge $(version_code 2.7.2) ]] ||
2654 skip "Need server version newer than 2.7.1"
2657 # let the reply to be dropped
2660 #define OBD_FAIL_SRV_ENOENT 0x217
2661 set_nodes_failloc "$(mdts_nodes)" 0x80000217
2663 wait $T130_PID || [ $? -eq 0 ] && error "stat should fail"
2666 run_test 130b "enqueue resend on a stale inode"
2669 remote_mds_nodsh && skip "remote MDS with nodsh" && return
2671 do_nodes ${CLIENTS:-$HOSTNAME} sync
2672 echo XXX > $DIR/$tfile
2674 cancel_lru_locks mdc
2676 # Trigger writeback on $tfile.
2678 # we need to race with unlink, unlink must complete before we will
2679 # take a DLM lock, otherwise unlink will wait until intent will
2680 # complete; but later than intent starts so that intent found
2682 #define OBD_FAIL_MDS_INTENT_DELAY 0x160
2683 set_nodes_failloc "$(mdts_nodes)" 0x80000160
2690 # drop the reply so that resend happens on an unlinked file.
2691 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2692 set_nodes_failloc "$(mdts_nodes)" 0x80000157
2694 # let the reply to be dropped
2697 #define OBD_FAIL_SRV_ENOENT 0x217
2698 set_nodes_failloc "$(mdts_nodes)" 0x80000217
2704 run_test 130c "layout intent resend on a stale inode"
2707 local before=$(date +%s)
2710 mount_client $MOUNT2 || error "mount filed"
2713 # get a lock on client so that export would reach the stale list
2714 $LFS setstripe -i 0 $DIR/$tfile || error "setstripe failed"
2715 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync ||
2718 #define OBD_FAIL_OST_PAUSE_PUNCH 0x236
2719 do_facet ost1 $LCTL set_param fail_val=120 fail_loc=0x80000236
2721 $TRUNCATE $DIR/$tfile 100 &
2724 dd if=/dev/zero of=$DIR2/$tfile bs=4096 count=1 conv=notrunc ||
2728 umount_client $MOUNT2
2730 evict=$(do_facet client $LCTL get_param \
2731 osc.$FSNAME-OST0000-osc-*/state |
2732 awk -F"[ [,]" '/EVICTED ]$/ { if (t<$5) {t=$5;} } END { print t }')
2734 [ -z "$evict" ] || [[ $evict -le $before ]] ||
2735 (do_facet client $LCTL get_param \
2736 osc.$FSNAME-OST0000-osc-*/state;
2737 error "eviction happened: $evict before:$before")
2739 run_test 132 "long punch"
2742 remote_ost_nodsh && skip "remote OST with nodsh" && return 0
2745 # get a lock on client so that export would reach the stale list
2746 $LFS setstripe -i 0 $DIR/$tfile || error "setstripe failed"
2747 dd if=/dev/zero of=$DIR/$tfile count=1 || error "dd failed"
2749 # another IO under the same lock
2750 #define OBD_FAIL_OSC_DELAY_IO 0x414
2751 $LCTL set_param fail_loc=0x80000414
2752 $LCTL set_param fail_val=4 fail_loc=0x80000414
2753 dd if=/dev/zero of=$DIR/$tfile count=1 conv=notrunc oflag=dsync &
2757 #define OBD_FAIL_LDLM_BL_EVICT 0x31e
2758 set_nodes_failloc "$(osts_nodes)" 0x8000031e
2762 wait $pid && error "dd succeeded"
2765 run_test 131 "IO vs evict results to IO under staled lock"
2768 local list=$(comma_list $(mdts_nodes))
2770 local t=$((TIMEOUT * 2))
2773 flock $DIR/$tfile -c "echo bl lock;sleep $t;echo bl flock unlocked" &
2775 multiop_bg_pause $DIR/$tfile O_jc || return 1
2778 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2779 do_nodes $list $LCTL set_param fail_loc=0x80000157
2781 echo "waiting for multiop $PID"
2782 wait $PID || return 2
2788 run_test 133 "don't fail on flock resend"
2791 [ $CLIENTCOUNT -lt 2 ] &&
2792 { skip "Need 2+ clients, have $CLIENTCOUNT" && return; }
2794 mkdir -p $MOUNT/$tdir/1 $MOUNT/$tdir/2 || error "mkdir failed"
2795 touch $MOUNT/$tdir/1/$tfile $MOUNT/$tdir/2/$tfile ||
2796 error "touch failed"
2797 zconf_umount_clients $CLIENTS $MOUNT
2798 zconf_mount_clients $CLIENTS $MOUNT
2800 #define OBD_FAIL_TGT_REPLY_DATA_RACE 0x722
2801 # assume commit interval is 5
2802 do_facet mds1 "$LCTL set_param fail_loc=0x722 fail_val=5"
2804 local -a clients=(${CLIENTS//,/ })
2805 local client1=${clients[0]}
2806 local client2=${clients[1]}
2808 do_node $client1 rm $MOUNT/$tdir/1/$tfile &
2810 do_node $client2 mv $MOUNT/$tdir/2/$tfile $MOUNT/$tdir/2/${tfile}_2 &
2813 wait $rmpid || error "rm failed"
2814 wait $mvpid || error "mv failed"
2817 run_test 134 "race between failover and search for reply data free slot"
2820 [ "$MDS1_VERSION" -lt $(version_code 2.12.51) ] &&
2821 skip "Need MDS version at least 2.12.51"
2824 $LFS setstripe -E 1M -L mdt $DIR/$tdir
2825 # to have parent dir write lock before open/resend
2826 touch $DIR/$tdir/$tfile
2827 #define OBD_FAIL_MDS_LDLM_REPLY_NET 0x157
2828 do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param fail_loc=0x80000157
2829 openfile -f O_RDWR:O_CREAT -m 0755 $DIR/$tdir/$tfile ||
2830 error "Failed to open DOM file"
2832 run_test 135 "DOM: open/create resend to return size"
2835 remote_mds_nodsh && skip "remote MDS with nodsh"
2836 [[ "$MDS1_VERSION" -ge $(version_code 2.12.52) ]] ||
2837 skip "Need MDS version at least 2.12.52"
2839 local mdts=$(comma_list $(mdts_nodes))
2840 local MDT0=$(facet_svc $SINGLEMDS)
2842 local clog=$(do_facet mds1 $LCTL --device $MDT0 changelog_register -n)
2843 [ -n "$clog" ] || error "changelog_register failed"
2844 cl_mask=$(do_facet mds1 $LCTL get_param \
2845 mdd.$MDT0.changelog_mask -n)
2846 changelog_chmask "ALL"
2848 # generate some changelog records to accumulate
2849 test_mkdir -i 0 -c 0 $DIR/$tdir || error "mkdir $tdir failed"
2850 createmany -m $DIR/$tdir/$tfile 10000 ||
2851 error "create $DIR/$tdir/$tfile failed"
2853 local size1=$(do_facet $SINGLEMDS \
2854 $LCTL get_param -n mdd.$MDT0.changelog_size)
2855 echo "Changelog size $size1"
2857 #define OBD_FAIL_LLOG_PURGE_DELAY 0x1318
2858 do_nodes $mdts $LCTL set_param fail_loc=0x1318 fail_val=30
2860 # launch changelog_deregister in background on MDS
2861 do_facet mds1 "nohup $LCTL --device $MDT0 changelog_deregister $clog \
2862 > foo.out 2> foo.err < /dev/null &"
2863 # give time to reach fail_loc
2866 # fail_loc will make MDS sleep in the middle of changelog_deregister
2867 # take this opportunity to abruptly kill MDS
2868 FAILURE_MODE_save=$FAILURE_MODE
2871 FAILURE_MODE=$FAILURE_MODE_save
2873 do_nodes $mdts $LCTL set_param fail_loc=0x0 fail_val=0
2875 local size2=$(do_facet $SINGLEMDS \
2876 $LCTL get_param -n mdd.$MDT0.changelog_size)
2877 echo "Changelog size $size2"
2878 local clog2=$(do_facet $SINGLEMDS "$LCTL get_param -n \
2879 mdd.$MDT0.changelog_users | grep $clog")
2880 echo "After crash, changelog user $clog2"
2882 [ -n "$clog2" -o $size2 -lt $size1 ] ||
2883 error "changelog record count unchanged"
2885 do_facet mds1 $LCTL set_param mdd.$MDT0.changelog_mask=\'$cl_mask\' -n
2887 run_test 136 "changelog_deregister leaving pending records"
2891 mkdir_on_mdt0 $DIR/$tdir
2894 dd if=/dev/zero of=$DIR/$tdir/d1/$tfile bs=4096 count=1
2895 dd if=/dev/zero of=$DIR/$tdir/d2/$tfile bs=4096 count=1
2896 cancel_lru_locks osc
2898 #define OBD_FAIL_PTLRPC_RESEND_RACE 0x525
2899 do_facet $SINGLEMDS "lctl set_param fail_loc=0x80000525"
2901 # RPC1: any reply is to be delayed to disable last_xid logic
2902 ln $DIR/$tdir/d1/$tfile $DIR/$tdir/d1/f2 &
2905 # RPC2: setattr1 reply is delayed & resent
2906 # original reply comes to client; the resend get asleep
2907 chmod 666 $DIR/$tdir/d2/$tfile
2909 # RPC3: setattr2 on the same file; run ahead of RPC2 resend
2910 chmod 777 $DIR/$tdir/d2/$tfile
2912 # RPC2 resend wakes up
2914 [ $(stat -c "%a" $DIR/$tdir/d2/$tfile) == 777 ] ||
2915 error "resend got applied"
2917 run_test 137 "late resend must be skipped if already applied"
2920 remote_mds_nodsh && skip "remote MDS with nodsh"
2921 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2922 [[ "$MDS1_VERSION" -ge $(version_code 2.12.59) ]] ||
2923 skip "Need server version newer than 2.12.59"
2925 zconf_umount_clients $CLIENTS $MOUNT
2927 #define OBD_FAIL_TGT_RECOVERY_CONNECT 0x724
2928 #delay a first step of recovey when MDS waiting clients
2929 #and failing to get osp logs
2930 do_facet $SINGLEMDS $LCTL set_param fail_loc=0x724 fail_val=5
2932 facet_failover $SINGLEMDS
2934 #waiting failover and recovery timer
2935 #the valuse is based on target_recovery_overseer() wait_event timeout
2937 stop $SINGLEMDS || error "stop MDS failed"
2938 do_facet $SINGLEMDS $LCTL set_param fail_loc=0
2939 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
2940 error "start MDS failed"
2941 zconf_mount_clients $CLIENTS $MOUNT
2943 run_test 138 "Umount MDT during recovery"
2946 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
2947 [ $MDS1_VERSION -lt $(version_code 2.13.50) ] &&
2948 skip "Need MDS version at least 2.13.50"
2950 stop $SINGLEMDS || error "stop $SINGLEMDS failed"
2952 #define OBD_FAIL_OSP_INVALID_LOGID 0x2106
2953 do_facet $SINGLEMDS $LCTL set_param fail_val=0x68 fail_loc=0x80002106
2954 start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
2955 error "Fail to start $SINGLEMDS"
2957 run_test 139 "corrupted catid won't cause crash"
2960 [ $MDS1_VERSION -lt $(version_code 2.12.58) ] &&
2961 skip "Need MDS version at least 2.13.50"
2963 slr=$(do_facet mds1 \
2964 $LCTL get_param -n mdt.$FSNAME-MDT0000.local_recovery)
2965 stack_trap "do_facet mds1 $LCTL set_param \
2966 mdt.*.local_recovery=$slr" EXIT
2968 # disable recovery for local clients
2969 # so local clients should be marked with no_recovery flag
2970 do_facet mds1 $LCTL set_param mdt.*.local_recovery=0
2974 cnt=$(do_facet mds1 $LCTL get_param "mdt.*MDT0000.exports.*.export" |
2975 grep export_flags.*no_recovery | wc -l)
2976 echo "$cnt clients with recovery disabled"
2978 [ $cnt -eq 0 ] && error "no clients with recovery disabled"
2980 # enable recovery for local clients
2981 # so no local clients should be marked with no_recovery flag
2982 do_facet mds1 $LCTL set_param mdt.*.local_recovery=1
2985 cnt=$(do_facet mds1 $LCTL get_param "mdt.*MDT0000.exports.*.export" |
2986 grep export_flags.*no_recovery | wc -l)
2987 echo "$cnt clients with recovery disabled"
2989 [ $cnt -eq 0 ] || error "$cnt clients with recovery disabled"
2991 run_test 140a "local mount is flagged properly"
2994 [ $MDS1_VERSION -lt $(version_code 2.12.58) ] &&
2995 skip "Need MDS version at least 2.13.50"
2997 slr=$(do_facet mds1 \
2998 $LCTL get_param -n mdt.$FSNAME-MDT0000.local_recovery)
2999 stack_trap "do_facet mds1 $LCTL set_param \
3000 mdt.*.local_recovery=$slr" EXIT
3002 # disable recovery for local clients
3003 do_facet mds1 $LCTL set_param mdt.*.local_recovery=0
3009 # Lustre: tfs-MDT0000: Recovery over after 0:03, of 2 clients 2 rec...
3010 local recovery=$(do_facet mds1 dmesg |
3011 awk '/Recovery over after/ { print $6 }' | tail -1 |
3012 awk -F: '{ print $1 * 60 + $2 }')
3013 (( recovery < TIMEOUT * 2 + 5 )) ||
3014 error "recovery took too long $recovery > $((TIMEOUT * 2 + 5))"
3016 run_test 140b "local mount is excluded from recovery"
3022 [ "$PARALLEL" == "yes" ] && skip "skip parallel run"
3023 combined_mgs_mds || skip "needs combined MGS/MDT"
3024 ( local_mode || from_build_tree ) &&
3025 skip "cannot run in local mode or from build tree"
3027 # some get_param have a bug to handle dot in param name
3028 do_rpc_nodes $(facet_active_host $SINGLEMDS) cancel_lru_locks MGC
3029 oldc=$(do_facet $SINGLEMDS $LCTL get_param -n \
3030 'ldlm.namespaces.MGC*.lock_count')
3032 do_rpc_nodes $(facet_active_host $SINGLEMDS) cancel_lru_locks MGC
3033 newc=$(do_facet $SINGLEMDS $LCTL get_param -n \
3034 'ldlm.namespaces.MGC*.lock_count')
3036 [ $oldc -eq $newc ] || error "mgc lost locks ($oldc != $newc)"
3039 run_test 141 "do not lose locks on MGS restart"
3042 [ $MDS1_VERSION -lt $(version_code 2.11.56) ] &&
3043 skip "Need MDS version at least 2.11.56"
3045 #define OBD_FAIL_MDS_ORPHAN_DELETE 0x165
3046 do_facet mds1 $LCTL set_param fail_loc=0x165
3047 $MULTIOP $DIR/$tfile Ouc || error "multiop failed"
3050 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
3052 wait_update_facet mds1 "pgrep orph_.*-MDD | wc -l" "0" ||
3053 error "MDD orphan cleanup thread not quit"
3055 run_test 142 "orphan name stub can be cleaned up in startup"
3058 (( $MDS1_VERSION >= $(version_code 2.13.0) )) ||
3059 skip "Need MDS version at least 2.13.0"
3060 [ "$PARALLEL" == "yes" ] && skip "skip parallel run"
3062 local mntpt=$(facet_mntpt $SINGLEMDS)
3064 mount_fstype $SINGLEMDS || error "mount as fstype $SINGLEMDS failed"
3065 do_facet $SINGLEMDS touch $mntpt/PENDING/$tfile
3066 unmount_fstype $SINGLEMDS
3067 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS || error "mds1 start fail"
3069 wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
3070 wait_update_facet mds1 "pgrep orph_.*-MDD | wc -l" "0" ||
3071 error "MDD orphan cleanup thread not quit"
3073 run_test 143 "orphan cleanup thread shouldn't be blocked even delete failed"
3076 [[ $($LCTL get_param mdc.*.import) =~ connect_flags.*overstriping ]] ||
3077 skip "server does not support overstriping"
3078 (( MDS1_VERSION >= $(version_code 2.15.50.49) )) ||
3079 skip "Need MDS version at least 2.15.50.49"
3088 large_xattr_enabled || skip_env "ea_inode feature disabled"
3089 test_mkdir -i 0 -c 1 -p $DIR/$tdir
3090 stack_trap "rm -rf $DIR/$tdir" EXIT
3092 mds_timeout=$(do_facet mds1 $LCTL get_param -n timeout)
3093 do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param timeout=300
3094 stack_trap "do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param timeout=$mds_timeout" EXIT
3096 $LFS setstripe -i 0 -C $setcount $DIR/$tdir || error "setstripe failed"
3098 for (( i = 0; i < 50; i++)); do
3099 touch $DIR/$tdir/$tfile_$i &
3106 for pid in $pids; do
3107 kill -9 $pid >/dev/null 2>&1
3113 # here we measure MDT stop + MDT start time. For error case MDT stop takes
3114 # about obd_timeout-60 (240) seconds. Without error - less than 30s.
3115 # MDT start takes different time depends on a configuration, let's check
3117 diff=$((after - before))
3118 (( $diff < 240 )) || error "MDT failover took $diff seconds"
3120 # failover mds1 back to the primary server
3123 run_test 144a "MDT failover should stop precreation threads"
3126 [[ $($LCTL get_param mdc.*.import) =~ connect_flags.*overstriping ]] ||
3127 skip "server does not support overstriping"
3128 (( MDS1_VERSION >= $(version_code 2.15.51.50) )) ||
3129 skip "Need MDS version at least 2.15.51.50"
3136 large_xattr_enabled || skip_env "ea_inode feature disabled"
3137 test_mkdir -i 0 -c 1 -p $DIR/$tdir
3138 stack_trap "rm -rf $DIR/$tdir" EXIT
3140 mds_timeout=$(do_facet mds1 $LCTL get_param -n timeout)
3141 do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param timeout=300
3142 stack_trap "do_nodes $(comma_list $(mdts_nodes)) $LCTL set_param timeout=$mds_timeout" EXIT
3144 $LFS setstripe -i 0 -C $setcount $DIR/$tdir || error "setstripe failed"
3146 for (( i = 0; i < 50; i++)); do
3147 touch $DIR/$tdir/$tfile_$i &
3153 # 60s is not enought with ZFS which is still significatnly
3154 # slower on some operations like record writing
3156 [[ "$mds1_FSTYPE" != "zfs" ]] || stime=120
3159 for pid in $pids; do
3160 ps -p $pid > /dev/null
3161 (( $? == 0 )) && rc=1
3162 kill -9 $pid >/dev/null 2>&1
3165 (( $rc == 0 )) || { fail mds1; error "Create thread still running"; }
3167 run_test 144b "orphan cleanup shouldn't be blocked for no objects+failover situation"
3170 [ "$PARALLEL" == "yes" ] && skip "skip parallel run"
3171 remote_mds_nodsh && skip "remote MDS with nodsh"
3172 remote_ost_nodsh && skip "remote OST with nodsh"
3173 (( OST1_VERSION >= $(version_code 2.15.59.53) )) ||
3174 skip "need OSS >= v2_15_59.53 for reconnect fix"
3177 #increase a precreation window
3178 mkdir_on_mdt0 $DIR/$tdir
3179 $LFS setstripe -c 1 -i 0 $DIR/$tdir
3180 createmany -o $DIR/$tdir/$tfile 9000
3183 #define OBD_FAIL_OST_DELORPHAN_DELAY 0x254
3184 #delay delorphan request to reconnection 5seconds
3185 do_facet ost1 $LCTL set_param fail_loc=0x0000254 fail_val=5
3187 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS || error "mds1 start fail"
3189 wait_recovery_complete mds1 || error "MDS recovery not done"
3193 do_facet mds1 $LCTL --device $FSNAME-OST0000-osc-MDT0000 recover
3195 do_facet ost1 $LCTL set_param fail_loc=0 fail_val=0
3197 #first and second orphan request delayed for 5seconds
3200 local testid=${TESTNAME//_/ }
3202 do_facet ost1 "dmesg | tac | sed '/$testid/,$ d'" |
3203 grep "trust the OST"
3205 if (( rc == 0 )); then
3207 error "LAST_ID synchronization failed"
3212 run_test 144c "reconnection during orphan cleanup shouldn't lose LAST_ID synchronization"
3215 [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
3216 [ $(facet_active_host mds2) = $(facet_active_host mds3) ] &&
3217 skip "needs mds2 and mds3 on separate nodes"
3223 #define OBD_FAIL_PTLRPC_DELAY_RECOV 0x507
3224 echo block mds_connect from mds2
3225 do_facet mds2 "$LCTL set_param fail_loc=0x507"
3227 #define OBD_FAIL_OUT_UPDATE_DROP 0x1707
3228 echo block recovery updates from mds3
3229 do_facet mds3 "$LCTL set_param fail_loc=0x1707"
3231 local hard_timeout=\
3232 $(do_facet mds1 $LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_time_hard)
3236 local get_soft_timeout_cmd=\
3237 "$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_time_soft 2>/dev/null"
3239 echo wait until mds1 recovery_time_soft is $hard_timeout
3240 wait_update $(facet_host mds1) "$get_soft_timeout_cmd" \
3241 "$hard_timeout" $hard_timeout
3243 echo unblock mds_connect from mds2
3244 do_facet mds2 "$LCTL set_param fail_loc=0"
3246 echo upblock recovery updates from mds3
3247 do_facet mds3 "$LCTL set_param fail_loc=0"
3250 [ -f $DIR/$tfile ] || error "$DIR/$tfile does not exist"
3252 run_test 145 "connect mdtlovs and process update logs after recovery expire"
3255 (( $MDS1_VERSION >= $(version_code 2.15.54.6) )) ||
3256 skip "Need MDS >= v2_15_54-6-g3c69d46e17 for eviction_count"
3258 local prev_count=$(do_facet $SINGLEMDS \
3259 $LCTL get_param -n "mdt.${mds1_svc}.eviction_count")
3265 local next_count=$(do_facet $SINGLEMDS \
3266 $LCTL get_param -n "mdt.${mds1_svc}.eviction_count")
3268 [ "$prev_count" -lt "$next_count" ] ||
3269 error "wrong eviction count ($prev_count >= $next_count)"
3271 run_test 146 "test eviction is counted properly"
3274 local obd_timeout=200
3275 local old=$($LCTL get_param -n timeout)
3277 local connection_count
3279 $LFS setstripe -i 0 -c 1 $f
3280 stripe_index=$($LFS getstripe -i $f)
3281 if [ $stripe_index -ne 0 ]; then
3283 error "$f: stripe_index $stripe_index != 0" && return
3286 $LCTL set_param timeout=$obd_timeout
3287 stack_trap "$LCTL set_param timeout=$old && client_reconnect" EXIT
3289 # OBD_FAIL_OST_CONNECT_NET2
3290 # lost reply to connect request
3291 do_facet ost1 lctl set_param fail_loc=0x00000225 timeout=$obd_timeout
3292 stack_trap "do_facet ost1 $LCTL set_param fail_loc=0 timeout=$old" EXIT
3297 $LFS df $MOUNT > /dev/null 2>&1 &
3298 sleep $((obd_timeout * 3 / 4))
3300 $LCTL get_param osc.$FSNAME-OST0000-osc-*.state
3301 connection_count=$($LCTL get_param osc.$FSNAME-OST0000-osc-*.state |
3302 tac | sed "/FULL/,$ d" | grep CONNECTING | wc -l)
3304 echo $connection_count
3305 (($connection_count >= 6)) || error "Client reconnected too slow"
3307 run_test 147 "Check client reconnect"
3310 local wce_param="obdfilter.$FSNAME-OST0000.writethrough_cache_enable"
3311 local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
3312 local amc=$(at_max_get client)
3313 local amo=$(at_max_get ost1)
3318 timeout=$(request_timeout client)
3320 [ "$(facet_fstype ost1)" = "ldiskfs" ] && {
3321 # save old r/o cache settings
3322 save_lustre_params ost1 $wce_param > $p
3325 do_facet ost1 "$LCTL set_param -n $wce_param=0"
3328 $LFS setstripe -i 0 -c 1 $DIR/$tfile
3329 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=direct
3330 cp $DIR/$tfile $TMP/$tfile
3331 #define OBD_FAIL_OST_BRW_PAUSE_BULK2 0x227
3332 do_facet ost1 $LCTL set_param fail_loc=0x80000227
3333 do_facet ost1 $LCTL set_param fail_val=$((timeout+2))
3334 dd if=/dev/urandom of=$DIR/$tfile bs=4096 count=1 conv=notrunc,fdatasync
3335 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=notrunc,fdatasync
3337 cancel_lru_locks osc
3338 cmp -b $DIR/$tfile $TMP/$tfile || error "wrong data"
3340 rm -f $DIR/$tfile $TMP/$tfile
3342 at_max_set $amc client
3343 at_max_set $amo ost1
3345 [ "$(facet_fstype ost1)" = "ldiskfs" ] && {
3346 # restore initial r/o cache settings
3347 restore_lustre_params < $p
3352 run_test 148 "data corruption through resend"
3355 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
3357 test_mkdir -i 0 -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed"
3359 # make an orphan striped dir
3360 $MULTIOP $DIR/$tdir D_c &
3363 rmdir $DIR/$tdir || error "can't rmdir"
3365 # stop a slave MDT where one ons stripe is located
3368 # stopping should not cause orphan as another MDT can
3372 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS || error "mds1 start fail"
3373 start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS || error "mds1 start fail"
3381 run_test 149 "skip orphan removal at umount"
3384 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
3388 lazystatfs=$($LCTL get_param -n llite.$FSNAME-*.lazystatfs | head -1)
3389 max=$($LCTL get_param -n llite.$FSNAME-*.statahead_max | head -1)
3391 $LCTL set_param llite.$FSNAME-*.lazystatfs=1
3392 $LCTL set_param llite.$FSNAME-*.statahead_max=0
3393 stack_trap "$LCTL set_param llite.$FSNAME-*.lazystatfs=$lazystatfs" EXIT
3394 stack_trap "$LCTL set_param llite.$FSNAME-*.statahead_max=$max" EXIT
3395 # stop a slave MDT where one ons stripe is located
3398 stack_trap "start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS && \
3399 wait_recovery_complete mds1 && clients_up && true" EXIT
3401 df $MOUNT || error "statfs failed"
3404 run_test 150 "statfs when MDT0 offline with lazystatfs option"
3407 [[ $($LCTL get_param mdc.*.import) =~ connect_flags.*overstriping ]] ||
3408 skip "server does not support overstriping"
3419 large_xattr_enabled || skip_env "ea_inode feature disabled"
3420 version=$(do_facet mds1 \
3421 uname -r | sed -e "s/\([0-9]*\.[0-9]*\.[0-9]*\).*/\1/")
3422 version=$(version_code ${version//\./ })
3423 if (( $version < $(version_code 4.6.0) )); then
3424 skip "MDS Linux kernel does not support killable semaphore"
3427 test_mkdir -i 0 -c 1 -p $DIR/$tdir
3428 test_mkdir -i 0 -c 1 -p $DIR/$tdir/rr
3429 test_mkdir -i 0 -c 1 -p $DIR/$tdir/qos
3430 stack_trap "rm -rf $DIR/$tdir" EXIT
3432 $LFS setstripe -C $setcount $DIR/$tdir/rr/ || error "setstripe failed"
3435 #define OBD_FAIL_MDS_LOD_CREATE_PAUSE 0x173
3436 #Simulate OST failover and sleep RR allocation under lq_rw_sem
3437 do_facet mds1 $LCTL set_param fail_loc=0x80000173 fail_val=20
3439 for (( i = 0; i < 2; i++)); do
3440 touch $DIR/$tdir/rr/$tfile_$i &
3441 pids_rr="$pids_rr $!"
3445 saved=$(do_facet mds1 $LCTL get_param -n lov.*0000*.qos_threshold_rr)
3446 do_facet mds1 $LCTL set_param lov.*.qos_threshold_rr=0
3447 stack_trap "do_facet mds1 $LCTL set_param lov.*.qos_threshold_rr=$saved" EXIT
3449 #create files with QoS algo, killable semaphore sleeps for 2seconds
3450 for (( i = 0; i < 3; i++)); do
3451 touch $DIR/$tdir/qos/$tfile_$i &
3452 pids_qos="$pids_qos $!"
3455 for pid in $pids_qos; do
3460 diff=$((after - before))
3461 echo "QoS allocation took $diff seconds"
3462 for pid in $pids_rr; do
3467 error "QoS allocation slower than RR, killable semaphore doesn't work"
3469 run_test 152 "QoS object allocation could be awakened in case of OST failover"
3472 #define OBD_FAIL_MDS_CONNECT_VS_EVICT 0x174
3473 do_facet mds1 "$LCTL set_param fail_loc=0x174"
3474 # first drop ping reply from MDS and then
3475 # evict on the subsequent reconnect
3476 # (see target_handle_connect)
3477 sleep $((TIMEOUT + 3))
3479 do_facet mds1 "$LCTL set_param fail_loc=0"
3480 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS ||
3481 error "Fail to start $SINGLEMDS"
3482 wait_recovery_complete mds1 || error "MDS recovery not done"
3484 run_test 153 "evict vs reconnect race"
3487 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
3491 local mnt=$(facet_mntpt mds2)
3494 local ula=($(do_facet mds2 od -t u8 $mnt/update_log | \
3495 awk '{ if($2 > 0) printf "0x%x:0x%x:0x0 ", $3, $2 }'))
3496 [ ${#ula[@]} -lt 2 ] && error "update_log file corruption"
3498 local catfile=$mnt/update_log_dir/\[${ula[0]}\]
3500 echo replace file $catfile
3501 do_facet mds2 dd if=/dev/urandom of=$catfile count=100 ||
3502 error "Write file $catfile error"
3505 start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS || error "mds2 start fail"
3508 start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS || error "mds1 start fail"
3510 echo waiting mds1 recovery....
3511 wait_recovery_complete mds1 20
3513 do_facet mds1 $LCTL get_param mdt.$FSNAME-MDT0000.recovery_status | \
3514 grep -w COMPLETE || error "Recovery was blocked"
3516 run_test 154a "corruption update llog can be skipped"
3519 [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs"
3523 stack_trap "do_facet mds1 $LCTL set_param fail_loc=0"
3524 #define OBD_FAIL_TGT_RECOVERY_CONNECT 0x724
3525 do_facet mds1 $LCTL set_param fail_loc=0x0724 fail_val=5
3526 local OPTS="$MDS_MOUNT_OPTS -o abort_recov"
3527 start mds1 $(mdsdevname 1) $OPTS ||
3528 error "start MDS with abort_recovery failed"
3530 echo waiting mds1 recovery....
3531 wait_recovery_complete mds1 30
3532 do_facet mds1 $LCTL get_param mdt.$FSNAME-MDT0000.recovery_status | \
3533 grep -w COMPLETE || error "Recovery was blocked"
3535 mount_client $MOUNT2
3536 umount_client $MOUNT2
3537 remount_client $MOUNT
3539 run_test 154b "restore update llog after failed recovery"
3546 lsoutput1=$(ls -l $DIR)
3548 zconf_umount $HOSTNAME $MOUNT || error "umount failed"
3549 # make sure that last_rcvd update is committed
3551 zconf_mount $HOSTNAME $MOUNT || error "mount failed"
3553 replay_barrier_nosync mds1
3557 lsoutput2=$(ls -l $DIR) || error "ls failed"
3558 [[ $lsoutput1 == $lsoutput2 ]] || error "$lsoutput1 != $lsoutput2"
3560 run_test 155 "failover after client remount"
3564 # on failover recovery time hard will be 9 * 5
3565 local saved_timeout=$(do_facet ost1 $LCTL get_param -n timeout)
3567 do_facet mgs $LCTL set_param -P timeout=5 ||
3568 error "failed to set obd_timeout"
3569 stack_trap "do_facet mgs $LCTL set_param -P timeout=$saved_timeout" \
3572 $LFS setstripe -c 1 -i 0 $DIR/$tfile || error "setstripe failed"
3574 # this is to sync last_rcvd, so that the client will have to
3575 # send replay on recovery
3581 $MULTIOP $DIR/$tfile oO_RDWR:O_SYNC:w1048576c || error "multiop failed"
3583 # delay write replay for 45 sec (OBD_RECOVERY_TIME_HARD) to
3584 # get the client evicted as not sending replays
3586 #define OBD_FAIL_PTLRPC_REPLAY_PAUSE 0x536
3587 $LCTL set_param fail_loc=0x80000536 fail_val=45
3591 # check that ost1 evicted the client in recovery
3593 clients=($(do_facet ost1 \
3594 $LCTL get_param -n obdfilter.$FSNAME-OST0000.recovery_status |
3595 awk '/completed_clients/ { print $2 }' | tr '/' '\n'))
3596 [[ $((${clients[0]} + 1)) == ${clients[1]} ]] ||
3597 error "client not evicted by ost1"
3599 local testid=$(echo $TESTNAME | tr '_' ' ')
3600 do_facet ost1 dmesg | tac | sed "/$testid/,$ d" |
3601 grep "ofd_obd_disconnect: tot_granted" &&
3602 error "grant miscount" || true
3604 run_test 156 "tot_granted miscount after client eviction"
3608 $LFS setstripe -i 0 -c 1 $DIR/$tfile || error "setstripe failed"
3609 dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1
3610 cancel_lru_locks osc
3612 #define OBD_FAIL_LLITE_FAULT_PAUSE 0x1432
3613 $LCTL set_param fail_loc=0x80001432 fail_val=3
3614 $MULTIOP $DIR/$tfile soO_RDWR:MRUc &
3619 [[ $? == 135 ]] || error "multiop failed with not SIGBUS"
3621 run_test 157 "eviction during mmaped i/o"
3623 complete_test $SECONDS
3624 check_and_cleanup_lustre