[#LU-12293] Memory leak after router checker packet processing

If net_monitor_thr is stopped with a condition that router checker packet is waiting for retry,
resources for the packet is not released.

As a workaround, we correct to wait for completion of router checker shutdown(TIMEOUT is 10sec x 2). After that, purge retry packet.
Could you discuss how to fix this bug.

diff --git a/lnet/lnet/lib-move.c b/lnet/lnet/lib-move.c
index 5e990d9..3b16d89 100644
--- a/lnet/lnet/lib-move.c
+++ b/lnet/lnet/lib-move.c
@@ -3682,6 +3682,14 @@ void lnet_monitor_thr_stop(void)
        /* tell the monitor thread that we're shutting down */
        wake_up(&the_lnet.ln_mt_waitq);
 
+       /* wait tx completion for router checker */
+       if (atomic_read(&the_lnet.ln_routers_nsends)) {
+               set_current_state(TASK_UNINTERRUPTIBLE);
+               schedule_timeout(cfs_time_seconds(lnet_get_lnd_timeout() * 2));
+       }
+       /* purge resend messages */
+       lnet_clean_resendqs();
+
        /* block until monitor thread signals that it's done */
        down(&the_lnet.ln_mt_signal);
        LASSERT(the_lnet.ln_mt_state == LNET_MT_STATE_SHUTDOWN);
@@ -3691,7 +3699,6 @@ void lnet_monitor_thr_stop(void)
        lnet_rsp_tracker_clean();
        lnet_clean_local_ni_recoveryq();
        lnet_clean_peer_ni_recoveryq();
-       lnet_clean_resendqs();
        rc = LNetEQFree(the_lnet.ln_mt_eqh);
        LASSERT(rc == 0);
        return;

[LU-12293] Memory leak after router checker packet processing Created: 13/May/19 Updated: 30/Aug/19
Status:	Open
Project:	Lustre
Component/s:	None
Affects Version/s:	Lustre 2.12.1
Fix Version/s:	None

[LU-12293] Memory leak after router checker packet processing Created: 13/May/19 Updated: 30/Aug/19