b=22656 Prevent failover nids from registering with MGS first.

Make the check in mgs_handle_target_reg()

o=Joseph Herring
i=nathan.rutman
i=andreas.dilger

diff --git a/lustre/mgs/mgs_handler.c b/lustre/mgs/mgs_handler.c
index 274b6fb..35d8da9 100644 (file)
--- a/lustre/mgs/mgs_handler.c
+++ b/lustre/mgs/mgs_handler.c
@@ -59,7 +59,7 @@
 #include <lustre_fsfilt.h>
 #include <lustre_disk.h>
 #include "mgs_internal.h"
-
+#include <lustre_param.h>
 
 /* Establish a connection to the MGS.*/
 static int mgs_connect(const struct lu_env *env,
@@ -401,6 +401,30 @@ static int mgs_check_target(struct obd_device *obd, struct mgs_target_info *mti)
         RETURN(rc);
 }
 
+/* Ensure this is not a failover node that is connecting first*/
+static int mgs_check_failover_reg(struct mgs_target_info *mti)
+{
+        lnet_nid_t nid;
+        char *ptr;
+        int i;
+
+        ptr = mti->mti_params;
+        while (class_find_param(ptr, PARAM_FAILNODE, &ptr) == 0) {
+                while (class_parse_nid(ptr, &nid, &ptr) == 0) {
+                        for (i = 0; i < mti->mti_nid_count; i++) {
+                                if (nid == mti->mti_nids[i]) {
+                                        LCONSOLE_WARN("Denying initial registra"
+                                                      "tion attempt from nid %s"
+                                                      ", specified as failover"
+                                                      "\n",libcfs_nid2str(nid));
+                                        return -EADDRNOTAVAIL;
+                                }
+                        }
+                }
+        }
+        return 0;
+}
+
 /* Called whenever a target starts up.  Flags indicate first connect, etc. */
 static int mgs_handle_target_reg(struct ptlrpc_request *req)
 {
@@ -427,6 +451,9 @@ static int mgs_handle_target_reg(struct ptlrpc_request *req)
                 if (rc <= 0)
                         /* Nothing wrong, or fatal error */
                         GOTO(out_nolock, rc);
+        } else {
+                if ((rc = mgs_check_failover_reg(mti)))
+                        GOTO(out_nolock, rc);
         }
 
         OBD_FAIL_TIMEOUT(OBD_FAIL_MGS_PAUSE_TARGET_REG, 10);

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 05b73c7..721cee6 100644 (file)
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -2363,10 +2363,6 @@ test_53b() {
 }
 run_test 53b "check MDT thread count params"
 
-if ! combined_mgs_mds ; then
-       stop mgs
-fi
-
 run_llverfs()
 {
         local dir=$1
@@ -2447,6 +2443,20 @@ test_56() {
 }
 run_test 56 "check big indexes"
 
+test_57() { # bug 22656
+       local NID=$($LCTL list_nids | head -1)
+       writeconf
+       do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed"
+       if ! combined_mgs_mds ; then
+               start_mgs
+       fi
+       start_mds
+       start_ost && error "OST registration from failnode should fail"
+       stop_mds
+       reformat
+}
+run_test 57 "initial registration from failnode should fail (should return errs)"
+
 count_osts() {
         do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l
 }
@@ -2481,6 +2491,10 @@ test_59() {
 }
 run_test 59 "writeconf mount option"
 
+
+if ! combined_mgs_mds ; then
+       stop mgs
+fi
 cleanup_gss
 equals_msg `basename $0`: test complete
 [ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true