btl/ugni: protect against re-entry and races in connections

This commit fixes two issues that can occur during a connection:

 - Re-entry to connection progress from modex lookup. Added an
   additional endpoint state that will keep the code from re-entering
   the common endpoint create.

 - Fixed a race between a process posting a directed datagram through
   a send and a connection being progressed through opal_progress().
   The progress code was not obtaining the endpoint lock before
   attempting to update the endpoint. To limit the amount of code
   changed for 2.0.1 this commit makes the endpoint lock recursive. In
   a future update this may be changed.

Signed-off-by: Nathan Hjelm <[email protected]>

Author: hjelmn

opal/mca/btl/ugni/btl_ugni_component.c

@@ -489,7 +489,14 @@ mca_btl_ugni_progress_datagram (mca_btl_ugni_module_t *ugni_module)
                  data, (void *) ep, remote_id));
 
     /* NTH: TODO -- error handling */
+    opal_mutex_lock (&ep->lock);
+    if (handle != ugni_module->wildcard_ep) {
+        /* directed post complete */
+        ep->dg_posted = false;
+    }
+
     (void) mca_btl_ugni_ep_connect_progress (ep);
+    opal_mutex_unlock (&ep->lock);
 
     if (MCA_BTL_UGNI_EP_STATE_CONNECTED == ep->state) {
         /*  process messages waiting in the endpoint's smsg mailbox */

opal/mca/btl/ugni/btl_ugni_endpoint.c

@@ -202,11 +202,14 @@ int mca_btl_ugni_ep_connect_progress (mca_btl_base_endpoint_t *ep) {
 
     if (GNI_SMSG_TYPE_INVALID == ep->remote_attr.smsg_attr.msg_type) {
         /* use datagram to exchange connection information with the remote peer */
-        rc = mca_btl_ugni_directed_ep_post (ep);
-        if (OPAL_SUCCESS == rc) {
-            rc = OPAL_ERR_RESOURCE_BUSY;
+        if (!ep->dg_posted) {
+            rc = mca_btl_ugni_directed_ep_post (ep);
+            if (OPAL_SUCCESS == rc) {
+                ep->dg_posted = true;
+                rc = OPAL_ERR_RESOURCE_BUSY;
+            }
+            return rc;
         }
-        return rc;
     }
 
     return mca_btl_ugni_ep_connect_finish (ep);

opal/mca/btl/ugni/btl_ugni_endpoint.h

@@ -1,6 +1,6 @@
 /* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
 /*
- * Copyright (c) 2011-2014 Los Alamos National Security, LLC. All rights
+ * Copyright (c) 2011-2016 Los Alamos National Security, LLC. All rights
  *                         reserved.
  * Copyright (c) 2011      UT-Battelle, LLC. All rights reserved.
  * $COPYRIGHT$
@@ -17,6 +17,7 @@
 
 enum mca_btl_ugni_endpoint_state_t {
     MCA_BTL_UGNI_EP_STATE_INIT = 0,
+    MCA_BTL_UGNI_EP_STATE_START,
     MCA_BTL_UGNI_EP_STATE_RDMA,
     MCA_BTL_UGNI_EP_STATE_CONNECTING,
     MCA_BTL_UGNI_EP_STATE_CONNECTED
@@ -30,7 +31,10 @@ typedef struct mca_btl_base_endpoint_t {
 
     opal_proc_t *peer_proc;
 
-    opal_mutex_t lock;
+    /** may need to lock recursively as the modex lookup could call opal_progress
+     * and hence our progress function. if this changes modify this mutex to not
+     * be recursive. also need to update the constructor function. */
+    opal_recursive_mutex_t lock;
     mca_btl_ugni_endpoint_state_t state;
 
     opal_common_ugni_endpoint_t *common;
@@ -48,6 +52,8 @@ typedef struct mca_btl_base_endpoint_t {
 
     opal_list_t frag_wait_list;
     bool wait_listed;
+    /** protect against race on connection */
+    bool dg_posted;
 
     int32_t smsg_progressing;
 
@@ -74,7 +80,6 @@ static inline int mca_btl_ugni_init_ep (mca_btl_ugni_module_t *ugni_module,
 
     endpoint->btl = btl;
     endpoint->peer_proc = peer_proc;
-    endpoint->common = NULL;
     endpoint->index = opal_pointer_array_add (&ugni_module->endpoints, endpoint);
 
     *ep = endpoint;
@@ -116,6 +121,7 @@ static inline int mca_btl_ugni_check_endpoint_state (mca_btl_ugni_endpoint_t *ep
     switch (ep->state) {
     case MCA_BTL_UGNI_EP_STATE_INIT:
     case MCA_BTL_UGNI_EP_STATE_RDMA:
+    case MCA_BTL_UGNI_EP_STATE_START:
         rc = mca_btl_ugni_ep_connect_progress (ep);
         if (OPAL_SUCCESS != rc) {
             break;
@@ -139,7 +145,15 @@ static inline int mca_btl_ugni_ep_connect_rdma (mca_btl_base_endpoint_t *ep) {
         return OPAL_SUCCESS;
     }
 
-    /* get the modex info for this endpoint and setup a ugni endpoint */
+    /* protect against re-entry from opal_progress */
+    if (OPAL_UNLIKELY(MCA_BTL_UGNI_EP_STATE_START == ep->state)) {
+        return OPAL_ERR_RESOURCE_BUSY;
+    }
+
+    ep->state = MCA_BTL_UGNI_EP_STATE_START;
+
+    /* get the modex info for this endpoint and setup a ugni endpoint. this call may lead
+     * to re-entry through opal_progress(). */
     rc = opal_common_ugni_endpoint_for_proc (ep->btl->device, ep->peer_proc, &ep->common);
     if (OPAL_SUCCESS != rc) {
         assert (0);