LU-14399 hsm: process hsm_actions in coordinator

Wait for mdd setup in a separate thread to don't block mount.
The patch adds conf-sanity_131 to verify the fix.

Change-Id: Ifd0e8969d7ed4f8944ab61ab0e0ebe2655bad003
Fixes: a558006b ("LU-13920 hsm: process hsm_actions only after mdd setup")
HPE-bug-id: LUS-9750
Signed-off-by: Sergey Cheremencev <sergey.cheremencev@hpe.com>
Reviewed-on: https://review.whamcloud.com/41445
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Alexander Boyko <alexander.boyko@hpe.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Artem Blagodarenko <artem.blagodarenko@hpe.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/mdt/mdt_coordinator.c b/lustre/mdt/mdt_coordinator.c
index 87296cb..767af80 100644 (file)
--- a/lustre/mdt/mdt_coordinator.c
+++ b/lustre/mdt/mdt_coordinator.c
@@ -549,7 +549,31 @@ static int set_cdt_state(struct coordinator *cdt, enum cdt_states new_state)
        return rc;
 }
 
+static int mdt_hsm_pending_restore(struct mdt_thread_info *mti);
 
+static void cdt_start_pending_restore(struct mdt_device *mdt,
+                                     struct coordinator *cdt)
+{
+       struct mdt_thread_info *cdt_mti;
+       unsigned int i = 0;
+       int rc;
+
+       /* wait until MDD initialize hsm actions llog */
+       while (!test_bit(MDT_FL_CFGLOG, &mdt->mdt_state) && i < obd_timeout) {
+               schedule_timeout_interruptible(cfs_time_seconds(1));
+               i++;
+       }
+       if (!test_bit(MDT_FL_CFGLOG, &mdt->mdt_state))
+               CWARN("%s: trying to init HSM before MDD\n", mdt_obd_name(mdt));
+
+       /* set up list of started restore requests */
+       cdt_mti = lu_context_key_get(&cdt->cdt_env.le_ctx, &mdt_thread_key);
+       rc = mdt_hsm_pending_restore(cdt_mti);
+       if (rc)
+               CERROR("%s: cannot take the layout locks needed for registered restore: %d\n",
+                      mdt_obd_name(mdt), rc);
+
+}
 
 /**
  * coordinator thread
@@ -579,6 +603,7 @@ static int mdt_coordinator(void *data)
 
        /* Inform mdt_hsm_cdt_start(). */
        wake_up(&cdt->cdt_waitq);
+       cdt_start_pending_restore(mdt, cdt);
 
        while (1) {
                int i;
@@ -1103,7 +1128,6 @@ static int mdt_hsm_cdt_start(struct mdt_device *mdt)
 {
        struct coordinator *cdt = &mdt->mdt_coordinator;
        struct mdt_thread_info *cdt_mti;
-       unsigned int i = 0;
        int rc;
        void *ptr;
        struct task_struct *task;
@@ -1135,28 +1159,13 @@ static int mdt_hsm_cdt_start(struct mdt_device *mdt)
        cdt->cdt_group_request_mask = (1UL << HSMA_RESTORE);
        cdt->cdt_other_request_mask = (1UL << HSMA_RESTORE);
 
-       /* wait until MDD initialize hsm actions llog */
-       while (!test_bit(MDT_FL_CFGLOG, &mdt->mdt_state) && i < obd_timeout) {
-               schedule_timeout_interruptible(cfs_time_seconds(1));
-               i++;
-       }
-       if (!test_bit(MDT_FL_CFGLOG, &mdt->mdt_state))
-               CWARN("%s: trying to init HSM before MDD\n", mdt_obd_name(mdt));
-
        /* to avoid deadlock when start is made through sysfs
         * sysfs entries are created by the coordinator thread
         */
-       /* set up list of started restore requests */
-       cdt_mti = lu_context_key_get(&cdt->cdt_env.le_ctx, &mdt_thread_key);
-       rc = mdt_hsm_pending_restore(cdt_mti);
-       if (rc)
-               CERROR("%s: cannot take the layout locks needed"
-                      " for registered restore: %d\n",
-                      mdt_obd_name(mdt), rc);
-
        if (mdt->mdt_bottom->dd_rdonly)
                RETURN(0);
 
+       cdt_mti = lu_context_key_get(&cdt->cdt_env.le_ctx, &mdt_thread_key);
        task = kthread_run(mdt_coordinator, cdt_mti, "hsm_cdtr");
        if (IS_ERR(task)) {
                rc = PTR_ERR(task);

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index e3a56c7..7a9a105 100644 (file)
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -9397,6 +9397,28 @@ test_131() {
 }
 run_test 131 "MDT backup restore with project ID"
 
+test_132() {
+       local err_cnt
+       local err_cnt2
+
+       reformat
+       combined_mgs_mds || start_mgs || error "unable to start MGS"
+       start_mdt 1 || error "unable to start mdt1"
+
+       err_cnt=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
+       stop_mdt 1 || error "stop mdt1 failed"
+
+       [ "$mds1_FSTYPE" == zfs ] && import_zpool mds1
+       do_facet mds1 $TUNEFS --param mdt.hsm_control=enabled $(mdsdevname 1) ||
+               error "tunefs failed"
+       start_mdt 1 || error "cannot start mdt1"
+
+       err_cnt2=$(do_facet mds1 dmesg | grep -c "cannot take the layout locks")
+       [ $err_cnt -eq $err_cnt2 ] || error "Can not take the layout lock"
+       stop_mdt 1 || error "stop mdt1 failed"
+}
+run_test 132 "hsm_actions processed after failover"
+
 if ! combined_mgs_mds ; then
        stop mgs
 fi