Merge pull request #1661 from matcabral/master

PSM and PSM2 MTLs to detect drivers and link

Author: rhc54

config/ompi_check_psm.m4

@@ -16,6 +16,8 @@ dnl Copyright (c) 2015      Research Organization for Information Science
 dnl                         and Technology (RIST). All rights reserved.
 dnl Copyright (c) 2016      Los Alamos National Security, LLC. All rights
 dnl                         reserved.
+dnl Copyright (c) 2016      Intel Corporation. All rights reserved.
+dnl 
 dnl $COPYRIGHT$
 dnl
 dnl Additional copyrights may follow
@@ -68,6 +70,13 @@ AC_DEFUN([OMPI_CHECK_PSM],[
               [AC_MSG_WARN([PSM driver does not currently support progress threads.  Disabling BTL.])
                ompi_check_psm_happy="no"])
 
+    	AS_IF([test "$ompi_check_psm_happy" = "yes"],
+              [AC_CHECK_HEADERS(
+               glob.h,
+               [],
+               [AC_MSG_WARN([glob.h not found.  Can not build component.])
+               ompi_check_psm_happy="no"])])
+
 	OPAL_SUMMARY_ADD([[Transports]],[[Intel TrueScale (PSM)]],[$1],[$ompi_check_psm_happy])
     fi
 

config/ompi_check_psm2.m4

@@ -12,7 +12,7 @@
 #                         All rights reserved.
 # Copyright (c) 2006      QLogic Corp. All rights reserved.
 # Copyright (c) 2009-2016 Cisco Systems, Inc.  All rights reserved.
-# Copyright (c) 2014      Intel Corporation. All rights reserved.
+# Copyright (c) 2016      Intel Corporation. All rights reserved.
 # Copyright (c) 2015      Research Organization for Information Science
 #                         and Technology (RIST). All rights reserved.
 # Copyright (c) 2016      Los Alamos National Security, LLC. All rights
@@ -69,6 +69,13 @@ AC_DEFUN([OMPI_CHECK_PSM2],[
               [AC_MSG_WARN([PSM2 driver does not currently support progress threads.  Disabling MTL.])
                ompi_check_psm2_happy="no"])
 
+    	AS_IF([test "$ompi_check_psm2_happy" = "yes"],
+              [AC_CHECK_HEADERS(
+               glob.h,
+               [],
+               [AC_MSG_WARN([glob.h not found.  Can not build component.])
+               ompi_check_psm2_happy="no"])])
+
 	OPAL_SUMMARY_ADD([[Transports]],[[Intel Omnipath (PSM2)]],[$1],[$ompi_check_psm2_happy])
     fi
 

ompi/mca/mtl/psm/mtl_psm_component.c

@@ -37,6 +37,7 @@
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <unistd.h>
+#include <glob.h>
 
 static int param_priority;
 
@@ -185,12 +186,41 @@ ompi_mtl_psm_component_open(void)
     }
 
   /* Component available only if Truescale hardware is present */
-  if (0 == stat("/dev/ipath", &st)) {
-    return OMPI_SUCCESS;
+  if (0 != stat("/dev/ipath", &st)) {
+    return OPAL_ERR_NOT_AVAILABLE;
+  }
+
+  /* Component available only if at least one qib port is ACTIVE */
+  bool foundOnlineQibPort = false;
+  size_t i;
+  char portState[128];
+  FILE *devFile;
+  glob_t globbuf;
+  globbuf.gl_offs = 0;
+  if (glob("/sys/class/infiniband/qib*/ports/*/state",
+        GLOB_DOOFFS, NULL, &globbuf) != 0) {
+    return OPAL_ERR_NOT_AVAILABLE;
+  }
+
+  for (i=0;i < globbuf.gl_pathc; i++) {
+    devFile = fopen(globbuf.gl_pathv[i], "r");
+    fgets(portState, sizeof(portState), devFile);
+    fclose(devFile);
+
+    if (strstr(portState, "ACTIVE") != NULL) {
+      /* Found at least one ACTIVE port */
+      foundOnlineQibPort = true;
+      break;
+    }
   }
-  else {
+
+  globfree(&globbuf);
+
+  if (!foundOnlineQibPort) {
     return OPAL_ERR_NOT_AVAILABLE;
   }
+
+  return OMPI_SUCCESS;
 }
 
 static int

ompi/mca/mtl/psm2/mtl_psm2_component.c

@@ -37,6 +37,7 @@
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <unistd.h>
+#include <glob.h>
 
 static int param_priority;
 
@@ -101,15 +102,46 @@ ompi_mtl_psm2_component_register(void)
 static int
 ompi_mtl_psm2_component_open(void)
 {
-  struct stat st;
+  glob_t globbuf;
+  globbuf.gl_offs = 0;
 
   /* Component available only if Omni-Path hardware is present */
-  if (0 == stat("/dev/hfi1", &st)) {
-    return OMPI_SUCCESS;
+  if ((glob("/dev/hfi1_[0-9]", GLOB_DOOFFS, NULL, &globbuf) != 0) &&
+      (glob("/dev/hfi1_[0-9][0-9]", GLOB_APPEND, NULL, &globbuf) != 0)) {
+    return OPAL_ERR_NOT_AVAILABLE;
+  }
+
+  globfree(&globbuf);
+
+  /* Component available only if at least one hfi1 port is ACTIVE */
+  bool foundOnlineHfi1Port = false;
+  size_t i;
+  char portState[128];
+  FILE *devFile;
+  if (glob("/sys/class/infiniband/hfi1_*/ports/*/state",
+        GLOB_DOOFFS, NULL, &globbuf) != 0) {
+    return OPAL_ERR_NOT_AVAILABLE;
+  }
+
+  for (i=0;i < globbuf.gl_pathc; i++) {
+    devFile = fopen(globbuf.gl_pathv[i], "r");
+    fgets(portState, sizeof(portState), devFile);
+    fclose(devFile);
+
+    if (strstr(portState, "ACTIVE") != NULL) {
+      /* Found at least one ACTIVE port */
+      foundOnlineHfi1Port = true;
+      break;
+    }
   }
-  else {
+
+  globfree(&globbuf);
+
+  if (!foundOnlineHfi1Port) {
     return OPAL_ERR_NOT_AVAILABLE;
   }
+
+  return OMPI_SUCCESS;
 }
 
 static int