LU-17184 mgc: remove damaged local configs

If local config llog is damaged it can't be removed and
prevents target from mounting. This happens because
mgc_llog_local_copy() uses llog_erase() to remove llogs
which can't do the job if llog header is damaged.

Patch changes are:
- llog_erase() to don't initialize header but just destroy
  llog file
- mgc_llog_local_copy() to don't exit on backup to temp
  file but continue with remote llog copying anyway
- conf-sanity test_151 is added to check that target can
  mount with damaged local config

Signed-off-by: Mikhail Pershin <mpershin@whamcloud.com>
Change-Id: I637749c38fd5ed03bdac5ca1cd60196f724ab0d1
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/52697
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Alexander Boyko <alexander.boyko@hpe.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/mgc/mgc_request_server.c b/lustre/mgc/mgc_request_server.c
index 15b5fa6..6eadd9d 100644 (file)
--- a/lustre/mgc/mgc_request_server.c
+++ b/lustre/mgc/mgc_request_server.c
@@ -528,32 +528,38 @@ static int mgc_llog_local_copy(const struct lu_env *env,
 
        ENTRY;
        /*
-        * - copy it to backup using llog_backup()
+        * NB: mgc_process_server_cfg_log() always needs valid local copy
+        * and works only on it, so that defines the process:
+        * - copy current local copy to temp_log using llog_backup()
         * - copy remote llog to logname using llog_backup()
-        * - if failed then move bakup to logname again
+        * - if failed then restore logname from backup
         */
+
        OBD_ALLOC(temp_log, strlen(logname) + 2);
        if (!temp_log)
                RETURN(-ENOMEM);
        sprintf(temp_log, "%sT", logname);
 
-       /* make a copy of local llog at first */
+       /* copy current local llog to temp_log */
        rc = llog_backup(env, obd, lctxt, lctxt, logname, temp_log);
        if (rc < 0 && rc != -ENOENT)
-               GOTO(out, rc);
-       /* copy remote llog to the local copy */
+               CWARN("%s: failed to backup local config %s: rc = %d\n",
+                     obd->obd_name, logname, rc);
+
+       /* build new local llog */
        rc = llog_backup(env, obd, rctxt, lctxt, logname, logname);
        if (rc == -ENOENT) {
-               /* no remote llog, delete local one too */
+               CWARN("%s: no remote llog for %s, check MGS config\n",
+                     obd->obd_name, logname);
                llog_erase(env, lctxt, NULL, logname);
        } else if (rc < 0) {
                /* error during backup, get local one back from the copy */
-               llog_backup(env, obd, lctxt, lctxt, temp_log, logname);
-out:
-               CERROR("%s: failed to copy remote log %s: rc = %d\n",
+               CWARN("%s: failed to copy new config %s: rc = %d\n",
                       obd->obd_name, logname, rc);
+               llog_backup(env, obd, lctxt, lctxt, temp_log, logname);
        }
        llog_erase(env, lctxt, NULL, temp_log);
+
        OBD_FREE(temp_log, strlen(logname) + 2);
        return rc;
 }

diff --git a/lustre/obdclass/llog.c b/lustre/obdclass/llog.c
index f8197e5..61e836e 100644 (file)
--- a/lustre/obdclass/llog.c
+++ b/lustre/obdclass/llog.c
@@ -1288,8 +1288,8 @@ EXPORT_SYMBOL(llog_open_create);
 int llog_erase(const struct lu_env *env, struct llog_ctxt *ctxt,
               struct llog_logid *logid, char *name)
 {
-       struct llog_handle      *handle;
-       int                      rc = 0, rc2;
+       struct llog_handle *handle;
+       int rc;
 
        ENTRY;
 
@@ -1301,13 +1301,9 @@ int llog_erase(const struct lu_env *env, struct llog_ctxt *ctxt,
        if (rc < 0)
                RETURN(rc);
 
-       rc = llog_init_handle(env, handle, LLOG_F_IS_PLAIN, NULL);
-       if (rc == 0)
-               rc = llog_destroy(env, handle);
+       rc = llog_destroy(env, handle);
+       llog_close(env, handle);
 
-       rc2 = llog_close(env, handle);
-       if (rc == 0)
-               rc = rc2;
        RETURN(rc);
 }
 EXPORT_SYMBOL(llog_erase);

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 0c228d6..b90be4b 100644 (file)
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -10780,6 +10780,34 @@ test_150() {
 }
 run_test 150 "test setting max_cached_mb to a %"
 
+test_151() {
+       (( MDS1_VERSION >= $(version_code 2.15.58) )) ||
+               skip "need MDS version at least 2.15.58"
+       [[ "$ost1_FSTYPE" == ldiskfs ]] || skip "ldiskfs only test"
+
+       cleanup
+       if ! combined_mgs_mds ; then
+               stop mgs
+       fi
+
+       echo "Damage ost1 local config log"
+       do_facet ost1 "$DEBUGFS -w -R 'punch CONFIGS/$FSNAME-OST0000 0 1' \
+                     $(ostdevname 1) || return \$?" ||
+               error "do_facet ost1 failed with $?"
+
+       # expect OST to fail mount with no MGS and bad local config
+       start_ost && error "OST start should fail"
+
+       if ! combined_mgs_mds ; then
+               start_mgs
+       fi
+       start_mds || error "MDS start failed"
+       # now it should start with MGS config
+       start_ost || error "OST start failed"
+       reformat_and_config
+}
+run_test 151 "damaged local config doesn't prevent mounting"
+
 #
 # (This was sanity/802a)
 #