Add more output when calls to the MX library fails.

Move the connection status from theproc into the endpoint.

This commit was SVN r12924.

ompi/mca/btl/mx/btl_mx.c

@@ -31,36 +31,6 @@
 #include "ompi/mca/mpool/base/base.h" 
 #include "ompi/mca/mpool/mpool.h" 
 
-
-mca_btl_mx_module_t mca_btl_mx_module = {
-    {
-        &mca_btl_mx_component.super,
-        0, /* max size of first fragment */
-        0, /* min send fragment size */
-        0, /* max send fragment size */
-        0, /* min rdma fragment size */
-        0, /* max rdma fragment size */
-        0, /* exclusivity */
-        0, /* latency */
-        0, /* bandwidth */
-        MCA_BTL_FLAGS_SEND_INPLACE | MCA_BTL_FLAGS_PUT, /* flags */
-        mca_btl_mx_add_procs,
-        mca_btl_mx_del_procs,
-        mca_btl_mx_register, 
-        mca_btl_mx_finalize,
-        mca_btl_mx_alloc, 
-        mca_btl_mx_free, 
-        mca_btl_mx_prepare_src,
-        mca_btl_mx_prepare_dst,
-        mca_btl_mx_send,
-        NULL, /* put */
-        NULL, /* get */
-        mca_btl_base_dump,
-        NULL, /* mpool */
-        NULL /* register error */
-    }
-};
-
 /**
  *
  */
@@ -378,8 +348,8 @@ int mca_btl_mx_send( struct mca_btl_base_module_t* btl,
     mx_return_t mx_return;
     uint64_t total_length;
 
-    if( MCA_BTL_MX_CONNECTED != ((mca_btl_mx_endpoint_t*)endpoint)->endpoint_proc->status ) {
-        if( MCA_BTL_MX_NOT_REACHEABLE == ((mca_btl_mx_endpoint_t*)endpoint)->endpoint_proc->status )
+    if( MCA_BTL_MX_CONNECTED != ((mca_btl_mx_endpoint_t*)endpoint)->status ) {
+        if( MCA_BTL_MX_NOT_REACHEABLE == ((mca_btl_mx_endpoint_t*)endpoint)->status )
             return OMPI_ERROR;
         if( OMPI_SUCCESS != mca_btl_mx_proc_connect( (mca_btl_mx_endpoint_t*)endpoint ) )
             return OMPI_ERROR;
@@ -433,3 +403,32 @@ int mca_btl_mx_finalize( struct mca_btl_base_module_t* btl )
     return OMPI_SUCCESS;
 }
 
+mca_btl_mx_module_t mca_btl_mx_module = {
+    {
+        &mca_btl_mx_component.super,
+        0, /* max size of first fragment */
+        0, /* min send fragment size */
+        0, /* max send fragment size */
+        0, /* min rdma fragment size */
+        0, /* max rdma fragment size */
+        0, /* exclusivity */
+        0, /* latency */
+        0, /* bandwidth */
+        MCA_BTL_FLAGS_SEND_INPLACE | MCA_BTL_FLAGS_PUT, /* flags */
+        mca_btl_mx_add_procs,
+        mca_btl_mx_del_procs,
+        mca_btl_mx_register, 
+        mca_btl_mx_finalize,
+        mca_btl_mx_alloc, 
+        mca_btl_mx_free, 
+        mca_btl_mx_prepare_src,
+        mca_btl_mx_prepare_dst,
+        mca_btl_mx_send,
+        NULL, /* put */
+        NULL, /* get */
+        mca_btl_base_dump,
+        NULL, /* mpool */
+        NULL /* register error */
+    }
+};
+

ompi/mca/btl/mx/btl_mx_component.c

@@ -336,6 +336,8 @@ mca_btl_base_module_t** mca_btl_mx_component_init(int *num_btl_modules,
         return NULL;
     if( (status = mx_get_info( NULL, MX_NIC_IDS, NULL, 0,
                                nic_addrs, size)) != MX_SUCCESS) {
+        opal_output(0, "MX BTL error (mx_get_info failed) size = %ld [%s] #cards %d\n",
+                    size, mx_strerror(status), mca_btl_mx_component.mx_num_btls );
         free(nic_addrs);
         return NULL;
     }

ompi/mca/btl/mx/btl_mx_endpoint.c

@@ -35,18 +35,17 @@
  * Initialize state of the endpoint instance.
  *
  */
-
 static void mca_btl_mx_endpoint_construct(mca_btl_base_endpoint_t* endpoint)
 {
-    endpoint->endpoint_btl = NULL;
+    endpoint->endpoint_btl  = NULL;
     endpoint->endpoint_proc = NULL;
+    endpoint->status        = MCA_BTL_MX_NOT_CONNECTED;
 }
 
 /*
  * Destroy a endpoint
  *
  */
-
 static void mca_btl_mx_endpoint_destruct(mca_btl_base_endpoint_t* endpoint)
 {
 }

ompi/mca/btl/mx/btl_mx_endpoint.h

@@ -32,6 +32,10 @@
 extern "C" {
 #endif
 
+#define MCA_BTL_MX_NOT_CONNECTED   0x0000
+#define MCA_BTL_MX_NOT_REACHEABLE  0x0001
+#define MCA_BTL_MX_CONNECTED       0x0002
+
 /**
  * Structure used to publish MX information to peers
  */
@@ -62,6 +66,8 @@ struct mca_btl_base_endpoint_t {
 
     mx_endpoint_addr_t          mx_peer_addr;
     /** the remote MX endpoint address */
+
+    int status;  /**< status of the endpoint */
 };
 
 typedef struct mca_btl_base_endpoint_t mca_btl_base_endpoint_t;

ompi/mca/btl/mx/btl_mx_proc.c

@@ -159,7 +159,6 @@ int mca_btl_mx_proc_insert( mca_btl_mx_proc_t* module_proc,
         return OMPI_ERROR;
     }
 
-    module_proc->status = MCA_BTL_MX_NOT_CONNECTED;
     module_proc->mx_peers = mx_peers;
 
     if( NULL == module_proc->proc_endpoints ) {
@@ -169,7 +168,6 @@ int mca_btl_mx_proc_insert( mca_btl_mx_proc_t* module_proc,
             return OMPI_ERR_OUT_OF_RESOURCE;
         }
     }
-    /* insert into endpoint array */
     module_endpoint->endpoint_proc = module_proc;
 
     return OMPI_SUCCESS;
@@ -206,13 +204,12 @@ int mca_btl_mx_proc_connect( mca_btl_mx_endpoint_t* module_endpoint )
         module_endpoint->mx_peer.nic_id      = module_proc->mx_peers[i].nic_id;
         module_endpoint->mx_peer.endpoint_id = module_proc->mx_peers[i].endpoint_id;
         module_endpoint->mx_peer_addr        = mx_remote_addr;
+        module_endpoint->status              = MCA_BTL_MX_CONNECTED;
         module_proc->proc_addr_index         = i;
-        module_proc->status                  = MCA_BTL_MX_CONNECTED;
         break;
     }
 
     if( i == module_proc->mx_peers_count ) {  /* no available connection */
-        module_proc->status = MCA_BTL_MX_NOT_REACHEABLE;
         return OMPI_ERROR;
     }
 

ompi/mca/btl/mx/btl_mx_proc.h

@@ -29,10 +29,6 @@
 extern "C" {
 #endif
 
-#define MCA_BTL_MX_NOT_CONNECTED   0x0000
-#define MCA_BTL_MX_NOT_REACHEABLE  0x0001
-#define MCA_BTL_MX_CONNECTED       0x0002
-
     /**
      * Represents the state of a remote process and the set of addresses
      * that it exports. Also cache an instance of mca_btl_base_endpoint_t for
@@ -46,8 +42,6 @@ extern "C" {
         ompi_proc_t *proc_ompi;                  
         /**< pointer to corresponding ompi_proc_t */
 
-        int status;  /**< status of the connection */
-
         mca_btl_mx_addr_t  *mx_peers;  /**< peers addresses */
         int mx_peers_count;