[LU-13277] Potential deadlock in lnet_peer_discovery

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.14.0
• Severity: 3
• Assignee: Chris Horn
• Reporter: Chris Horn

Description

Potential deadlock here when LNet is shutting down:

static int lnet_peer_discovery(void *arg)
...
        for (;;) {
...
                lnet_net_lock(LNET_LOCK_EX);
                if (the_lnet.ln_dc_state == LNET_DC_STATE_STOPPING)
                        break;
...
        }

        CDEBUG(D_NET, "stopping\n");
        /*
         * Clean up before telling lnet_peer_discovery_stop() that
         * we're done. Use wake_up() below to somewhat reduce the
         * size of the thundering herd if there are multiple threads
         * waiting on discovery of a single peer.
         */

        /* Queue cleanup 1: stop all pending pings and pushes. */
        lnet_net_lock(LNET_LOCK_EX); <<< Deadlock
...