[LU-18697] Bug in b2_15 backport "b341288179 LU-14668 lnet: Lock primary NID logic"

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.15.7
• Severity: 3
• Assignee: Chris Horn
• Reporter: Chris Horn

Description

The backport causes reference loss on an lnet_peer_ni if its parent peer has the LNET_PEER_LOCK_PRIMARY bit set:

diff --git a/lnet/lnet/peer.c b/lnet/lnet/peer.c
index 7a438ea086..01eb9aa8aa 100644
--- a/lnet/lnet/peer.c
+++ b/lnet/lnet/peer.c
@@ -546,6 +546,15 @@ lnet_peer_del_nid(struct lnet_peer *lp, lnet_nid_t nid4, unsigned int flags)
        }

        lpni = lnet_peer_ni_find_locked(&nid);
+       /* If we're asked to lock down the primary NID we shouldn't be
+        * deleting it
+        */
+       if (lp->lp_state & LNET_PEER_LOCK_PRIMARY &&
+           nid_same(&primary_nid, &nid)) {
+               rc = -EPERM;
+               goto out; <<< Ref taken by lnet_peer_ni_find_locked() is not dropped
+       }
+
        if (!lpni) {
                rc = -ENOENT;
                goto out;