LU-16266 lfsck: fix the PROJID of OST object

The PROJID of the OST object should be fixed along with UID/GID
when LFSCK finds the UID/GID/PROJID of the OST object is different
from the value of the MDT object.

Signed-off-by: Hongchao Zhang <hongchao@whamcloud.com>
Change-Id: Ifbdc89797ccfa4d80515ec1a67062016e7dc30fb
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/58519
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Li Dongyang <dongyangli@ddn.com>

diff --git a/lustre/lfsck/lfsck_layout.c b/lustre/lfsck/lfsck_layout.c
index 5a682e3..ddbdf46 100644 (file)
--- a/lustre/lfsck/lfsck_layout.c
+++ b/lustre/lfsck/lfsck_layout.c
@@ -4092,7 +4092,8 @@ static int lfsck_layout_repair_owner(const struct lu_env *env,
 
        tla->la_uid = pla->la_uid;
        tla->la_gid = pla->la_gid;
-       tla->la_valid = LA_UID | LA_GID;
+       tla->la_projid = pla->la_projid;
+       tla->la_valid = LA_UID | LA_GID | LA_PROJID;
        handle = lfsck_trans_create(env, dev, com->lc_lfsck);
        if (IS_ERR(handle))
                GOTO(log, rc = PTR_ERR(handle));
@@ -4539,7 +4540,8 @@ check_owner:
         * It can be handled later inside the lfsck_layout_repair_owner().
         */
        if (unlikely(cla->la_uid != pla->la_uid ||
-                    cla->la_gid != pla->la_gid)) {
+                    cla->la_gid != pla->la_gid ||
+                    cla->la_projid != pla->la_projid)) {
                type = LLIT_INCONSISTENT_OWNER;
                goto repair;
        }

diff --git a/lustre/tests/sanity-lfsck.sh b/lustre/tests/sanity-lfsck.sh
index 94fefb9..8051ea9 100755 (executable)
--- a/lustre/tests/sanity-lfsck.sh
+++ b/lustre/tests/sanity-lfsck.sh
@@ -6371,6 +6371,115 @@ test_44() {
 }
 run_test 44 "umount while lfsck is stopping"
 
+test_45() {
+       (( $OST1_VERSION >= $(version_code 2.16.53) )) ||
+               skip "Need OST version at least 2.16.53"
+
+       local prjid=1000
+       local testfile="$DIR/$tdri/$tfile-0"
+       local cnt=30
+       local usage
+
+       $LFS setstripe $testfile -i 0 -c 1 || error "setstripe $testfile failed"
+       chown $RUNAS_ID.$RUNAS_GID $testfile || error "chown $testfile failed"
+       is_project_quota_supported && change_project -p $prjid $testfile
+       $RUNAS $DD of=$testfile count=$cnt || error "write $testfile failed"
+
+       cancel_lru_locks osc
+       sync; sync_all_data || true
+       sleep 5
+
+       echo "check the quota usage after initial write"
+       usage=$(getquota -u $RUNAS_ID global curspace)
+       ((usage > cnt * 1024 * 9 / 10)) ||
+               error "quota USR usage $usage for $RUNAS_ID is wrong"
+
+       usage=$(getquota -g $RUNAS_GID global curspace)
+       ((usage > cnt * 1024 * 9 / 10)) ||
+               error "quota GRP usage $usage for $RUNAS_GID is wrong"
+
+       is_project_quota_supported && {
+               usage=$(getquota -p $prjid global curspace)
+               ((usage > cnt * 1024 * 9 / 10)) ||
+                       error "quota PRJ usage $usage for $prjid is wrong"
+       }
+
+       local fids=($($LFS getstripe $testfile | grep 0x))
+       local fid="${fids[3]}:${fids[2]}:0"
+       local objpath=$(ost_fid2_objpath ost1 $fid)
+
+       stop ost1 || error "failed to stop ost1"
+
+       echo "clear the UID/GID/PROJID of the test file"
+       mount_fstype ost1 || return 1
+       do_facet ost1  chown root:root $(facet_mntpt ost1)/$objpath
+       do_facet ost1 ls -l $(facet_mntpt ost1)/$objpath
+       is_project_quota_supported && {
+               do_facet ost1 chattr -p 0 $(facet_mntpt ost1)/$objpath
+               do_facet ost1 lsattr -p $(facet_mntpt ost1)/$objpath
+       }
+       unmount_fstype ost1
+
+       start ost1 $(ostdevname 1) $OST_MOUNT_OPTS ||
+               error "failed to start ost1"
+
+       echo "check the quota usage after UID/GID/PROJID is cleared"
+       usage=$(getquota -u $RUNAS_ID global curspace)
+       ((usage < cnt * 1024 * 1 / 10)) ||
+               error "quota USR usage $usage for $RUNAS_ID is wrong"
+
+       usage=$(getquota -g $RUNAS_GID global curspace)
+       ((usage < cnt * 1024 * 1 / 10)) ||
+               error "quota GRP usage $usage for $RUNAS_GID is wrong"
+
+       is_project_quota_supported && {
+               usage=$(getquota -p $prjid global curspace)
+               ((usage < cnt * 1024 * 1 / 10)) ||
+                       error "quota PRJ usage $usage for $prjid is wrong"
+       }
+
+       echo "the quota usage should be transferred to root"
+       usage=$(getquota -u root global curspace)
+       ((usage > cnt * 1024 * 9 / 10)) ||
+               error "quota USR usage $usage for root is wrong"
+
+       usage=$(getquota -g root global curspace)
+       ((usage > cnt * 1024 * 9 / 10)) ||
+               error "quota GRP usage $usage for root is wrong"
+
+       is_project_quota_supported && {
+               usage=$(getquota -p 0 global curspace)
+               ((usage > cnt * 1024 * 9 / 10)) ||
+                       error "quota PRJ usage $usage for 0 is wrong"
+       }
+
+       echo "fix the UID/GID/PROJID by LFSCK"
+       $START_LAYOUT -r -A || error "failed to start LFSCK"
+
+       wait_update_facet ost1 \
+               "$LCTL get_param -n obdfilter.$ost1_svc.lfsck_layout |
+                awk '/^status/ { print \\\$2 }'" "completed" 32 || {
+               $SHOW_LAYOUT_ON_OST
+               error "unexpected status of LFSCK on OST1"
+       }
+
+       echo "the quota usage should be fixed"
+       usage=$(getquota -u $RUNAS_ID global curspace)
+       ((usage > cnt * 1024 * 9 / 10)) ||
+               error "quota USR usage $usage for $RUNAS_ID is wrong"
+
+       usage=$(getquota -g $RUNAS_GID global curspace)
+       ((usage > cnt * 1024 * 9 / 10)) ||
+               error "quota GRP usage $usage for $RUNAS_GID is wrong"
+
+       is_project_quota_supported && {
+               usage=$(getquota -p $prjid global curspace)
+               ((usage > cnt * 1024 * 9 / 10)) ||
+                       error "quota PRJ usage $usage for $prjid is wrong"
+       }
+}
+run_test 45 "LFSCK should fix UID/GID/PROJID of OST object"
+
 # restore MDS/OST size
 MDSSIZE=${SAVED_MDSSIZE}
 OSTSIZE=${SAVED_OSTSIZE}