Merge pull request #2351 from larrystevenwise/v2.x

openib btl: remove BTL_OPENIB_FAILOVER_ENABLED code

Author: hppritcha

opal/mca/btl/openib/Makefile.am

@@ -66,13 +66,6 @@ sources = \
     connect/btl_openib_connect_empty.h \
     connect/connect.h
 
-# If we have failover support, build that file
-if MCA_btl_openib_enable_failover
-sources += \
-    btl_openib_failover.c \
-    btl_openib_failover.h
-endif
-
 # If we have rdmacm support, build that CPC
 if MCA_btl_openib_have_rdmacm
 sources += \

opal/mca/btl/openib/btl_openib.c

@@ -1846,23 +1846,13 @@ int mca_btl_openib_sendi( struct mca_btl_base_module_t* btl,
         assert(max_data == payload_size);
     }
 
-#if BTL_OPENIB_FAILOVER_ENABLED
-    send_signaled = 1;
-#else
     send_signaled = qp_need_signal(ep, qp, payload_size + header_size, do_rdma);
-#endif
     ib_rc = post_send(ep, to_send_frag(item), do_rdma, send_signaled);
 
     if (!ib_rc) {
         if (0 == send_signaled) {
             MCA_BTL_IB_FRAG_RETURN(frag);
         }
-#if BTL_OPENIB_FAILOVER_ENABLED
-        else {
-            /* Return up in case needed for failover */
-            *descriptor = (struct mca_btl_base_descriptor_t *) frag;
-        }
-#endif
         OPAL_THREAD_UNLOCK(&ep->endpoint_lock);
 
         return OPAL_SUCCESS;

opal/mca/btl/openib/btl_openib.h

@@ -243,9 +243,6 @@ struct mca_btl_openib_component_t {
     opal_event_base_t *async_evbase; /**< Async event base */
     bool use_async_event_thread;     /**< Use the async event handler */
     mca_btl_openib_srq_manager_t srq_manager;     /**< Hash table for all BTL SRQs */
-#if BTL_OPENIB_FAILOVER_ENABLED
-    bool port_error_failover;        /**< Report port errors to speed up failover */
-#endif
     /* declare as an int instead of btl_openib_device_type_t since there is no
        guarantee about the size of an enum. this value will be registered as an
        integer with the MCA variable system */
@@ -312,9 +309,6 @@ struct mca_btl_openib_component_t {
     int memory_registration_verbose_level;
     int memory_registration_verbose;
     int ignore_locality;
-#if BTL_OPENIB_FAILOVER_ENABLED
-    int verbose_failover;
-#endif
 #if OPAL_CUDA_SUPPORT
     bool cuda_async_send;
     bool cuda_async_recv;

opal/mca/btl/openib/btl_openib_component.c

@@ -84,9 +84,6 @@
 #include "btl_openib_ini.h"
 #include "btl_openib_mca.h"
 #include "btl_openib_xrc.h"
-#if BTL_OPENIB_FAILOVER_ENABLED
-#include "btl_openib_failover.h"
-#endif
 #include "btl_openib_async.h"
 #include "connect/base.h"
 #include "btl_openib_ip.h"
@@ -502,12 +499,6 @@ static void btl_openib_control(mca_btl_base_module_t* btl,
             mca_btl_openib_endpoint_connected(ep);
         }
         break;
-#if BTL_OPENIB_FAILOVER_ENABLED
-    case MCA_BTL_OPENIB_CONTROL_EP_BROKEN:
-    case MCA_BTL_OPENIB_CONTROL_EP_EAGER_RDMA_ERROR:
-        btl_openib_handle_failover_control_messages(ctl_hdr, ep);
-        break;
-#endif
     default:
         BTL_ERROR(("Unknown message type received by BTL"));
        break;
@@ -3470,20 +3461,8 @@ static void handle_wc(mca_btl_openib_device_t* device, const uint32_t cq,
                 opal_list_item_t *i;
                 while((i = opal_list_remove_first(&to_send_frag(des)->coalesced_frags))) {
                     btl_ownership = (to_base_frag(i)->base.des_flags & MCA_BTL_DES_FLAGS_BTL_OWNERSHIP);
-#if BTL_OPENIB_FAILOVER_ENABLED
-                    /* The check for the callback flag is only needed when running
-                     * with the failover case because there is a chance that a fragment
-                     * generated from a sendi call (which does not set the flag) gets
-                     * coalesced.  In normal operation, this cannot happen as the sendi
-                     * call will never queue up a fragment which could potentially become
-                     * a coalesced fragment.  It will revert to a regular send. */
-                    if (to_base_frag(i)->base.des_flags & MCA_BTL_DES_SEND_ALWAYS_CALLBACK) {
-#endif
                         to_base_frag(i)->base.des_cbfunc(&openib_btl->super, endpoint,
                                 &to_base_frag(i)->base, OPAL_SUCCESS);
-#if BTL_OPENIB_FAILOVER_ENABLED
-                    }
-#endif
                     if( btl_ownership ) {
                         mca_btl_openib_free(&openib_btl->super, &to_base_frag(i)->base);
                     }
@@ -3608,14 +3587,9 @@ static void handle_wc(mca_btl_openib_device_t* device, const uint32_t cq,
         }
     }
 
-#if BTL_OPENIB_FAILOVER_ENABLED
-    mca_btl_openib_handle_endpoint_error(openib_btl, des, qp,
-                                         remote_proc, endpoint);
-#else
     if(openib_btl)
         openib_btl->error_cb(&openib_btl->super, MCA_BTL_ERROR_FLAGS_FATAL,
                              (struct opal_proc_t*)remote_proc, NULL);
-#endif
 }
 
 static int poll_device(mca_btl_openib_device_t* device, int count)
@@ -3826,9 +3800,6 @@ static int btl_openib_component_progress(void)
         if(openib_btl->device->got_port_event) {
             /* These are non-fatal so just ignore it. */
             openib_btl->device->got_port_event = false;
-#if BTL_OPENIB_FAILOVER_ENABLED
-            mca_btl_openib_handle_btl_error(openib_btl);
-#endif
         }
     }
     return count;

opal/mca/btl/openib/btl_openib_endpoint.h

@@ -581,13 +581,6 @@ static inline int post_send(mca_btl_openib_endpoint_t *ep,
             BTL_OPENIB_FOOTER_HTON(*ftr);
 
         sr_desc->wr.rdma.rkey = ep->eager_rdma_remote.rkey;
-#if BTL_OPENIB_FAILOVER_ENABLED
-        /* frag->ftr is unused on the sending fragment, so use it
-         * to indicate it is an eager fragment.  A non-zero value
-         * indicates it is eager, and the value indicates the
-         * location in the eager RDMA array that it lives. */
-        frag->ftr = (mca_btl_openib_footer_t*)(long)(1 + head);
-#endif
         sr_desc->wr.rdma.remote_addr =
             ep->eager_rdma_remote.base.lval +
             head * openib_btl->eager_rdma_frag_size +