open-mpi · ggouaillardet · Jul 16, 2019 · Jul 16, 2019
diff --git a/opal/mca/pmix/pmix4x/pmix/NEWS b/opal/mca/pmix/pmix4x/pmix/NEWS
@@ -76,6 +76,45 @@ Master (not on release branches yet)
  - Fix a bug when registering default event handlers
 
 
+3.1.0 -- 17 Jan 2019
+----------------------
+**** THIS RELEASE MARKS THE STARTING POINT FOR FULL COMPLIANCE
+**** WITH THE PMIX v3 STANDARD. ALL API BEHAVIORS AND ATTRIBUTE
+**** DEFINITIONS MEET THE v3 STANDARD SPECIFICATIONS.
+ - Add a new, faster dstore GDS component 'ds21'
+ - Performance optimizations for the dstore GDS components.
+ - Plug miscellaneous memory leaks
+ - Silence an unnecessary warning message when checking connection
+   to a non-supporting server
+ - Ensure lost-connection events get delivered to default event
+   handlers
+ - Correctly handle cache refresh for queries
+ - Protect against race conditions between host and internal library
+   when dealing with async requests
+ - Cleanup tool operations and add support for connections to
+   remote servers. Initial support for debugger direct/indirect
+   launch verified with PRRTE. Cleanup setting of tmpdir options.
+   Drop rendezvous files when acting as a launcher
+ - Automatically store the server URI for easy access by client
+ - Provide MCA parameter to control TCP connect retry/timeout
+ - Update event notification system to properly evict oldest events
+   when more space is needed
+ - Fix a number of error paths
+ - Update IOF cache code to properly drop oldest message. Provide
+   MCA parameter for setting cache size.
+ - Handle setsockopt(SO_RCVTIMEO) not being supported
+ - Ensure that epilogs get run even when connections unexpectedly
+   terminate. Properly split epilog strings to process multiple
+   paths
+ - Pass the tool's command line to the server so it can be returned
+   in queries
+ - Add support for C11 atomics
+ - Support collection and forwarding of fabric-specific envars
+ - Improve handling of hwloc configure option
+ - Fix PMIx_server_generate_regex to preserve node ordering
+ - Fix a bug when registering default event handlers
+
+
 3.0.2 -- 18 Sept 2018
 ----------------------
 - Ensure we cleanup any active sensors when a peer departs. Allow the

diff --git a/opal/mca/pmix/pmix4x/pmix/VERSION b/opal/mca/pmix/pmix4x/pmix/VERSION
@@ -30,7 +30,7 @@ greek=a1
 # command, or with the date (if "git describe" fails) in the form of
 # "date<date>".
 
-repo_rev=git99971222
+repo_rev=git03a8b5da
 
 # If tarball_version is not empty, it is used as the version string in
 # the tarball filename, regardless of all other versions listed in
@@ -44,7 +44,7 @@ tarball_version=
 
 # The date when this release was created
 
-date="Jun 27, 2019"
+date="Jul 16, 2019"
 
 # The shared library version of each of PMIx's public libraries.
 # These versions are maintained in accordance with the "Library

diff --git a/opal/mca/pmix/pmix4x/pmix/config/pmix.m4 b/opal/mca/pmix/pmix4x/pmix/config/pmix.m4
@@ -191,7 +191,7 @@ AC_DEFUN([PMIX_SETUP_CORE],[
                                [Link the output PMIx library to this extra lib (used in embedded mode)]))
     AC_MSG_CHECKING([for extra lib])
     AS_IF([test ! -z "$with_pmix_extra_lib"],
-          [AS_IF([test "$with_pmix_extra_lib" == "yes" || test "$with_pmix_extra_lib" == "no"],
+          [AS_IF([test "$with_pmix_extra_lib" = "yes" || test "$with_pmix_extra_lib" = "no"],
                  [AC_MSG_RESULT([ERROR])
                   AC_MSG_WARN([Invalid value for --with-extra-pmix-lib:])
                   AC_MSG_WARN([    $with_pmix_extra_lib])
@@ -209,7 +209,7 @@ AC_DEFUN([PMIX_SETUP_CORE],[
                                [Link any embedded components/tools that require it to the provided libtool lib (used in embedded mode)]))
     AC_MSG_CHECKING([for extra ltlib])
     AS_IF([test ! -z "$with_pmix_extra_ltlib"],
-          [AS_IF([test "$with_pmix_extra_ltlib" == "yes" || test "$with_pmix_extra_ltlib" == "no"],
+          [AS_IF([test "$with_pmix_extra_ltlib" = "yes" || test "$with_pmix_extra_ltlib" = "no"],
                  [AC_MSG_RESULT([ERROR])
                   AC_MSG_WARN([Invalid value for --with-pmix-extra-ltlib:])
                   AC_MSG_WARN([    $with_pmix_extra_ltlib])
@@ -664,7 +664,7 @@ AC_DEFUN([PMIX_SETUP_CORE],[
     # -lrt might be needed for clock_gettime
     PMIX_SEARCH_LIBS_CORE([clock_gettime], [rt])
 
-    AC_CHECK_FUNCS([asprintf snprintf vasprintf vsnprintf strsignal socketpair strncpy_s usleep statfs statvfs getpeereid getpeerucred strnlen posix_fallocate tcgetpgrp setpgid ptsname openpty])
+    AC_CHECK_FUNCS([asprintf snprintf vasprintf vsnprintf strsignal socketpair strncpy_s usleep statfs statvfs getpeereid getpeerucred strnlen posix_fallocate tcgetpgrp setpgid ptsname openpty setenv])
 
     # On some hosts, htonl is a define, so the AC_CHECK_FUNC will get
     # confused.  On others, it's in the standard library, but stubbed with
@@ -1245,7 +1245,7 @@ AC_MSG_CHECKING([if want to support dlopen of non-global namespaces])
 AC_ARG_ENABLE([nonglobal-dlopen],
               AC_HELP_STRING([--enable-nonglobal-dlopen],
                              [enable non-global dlopen (default: enabled)]))
-if test "$enable_nonglobal_dlopen" == "no"; then
+if test "$enable_nonglobal_dlopen" = "no"; then
     AC_MSG_RESULT([no])
     pmix_need_libpmix=0
 else

diff --git a/opal/mca/pmix/pmix4x/pmix/config/pmix_setup_hwloc.m4 b/opal/mca/pmix/pmix4x/pmix/config/pmix_setup_hwloc.m4
@@ -63,7 +63,7 @@ AC_DEFUN([_PMIX_HWLOC_EXTERNAL],[
     if test "$with_hwloc" != "no"; then
         AC_MSG_CHECKING([for hwloc in])
         if test ! -z "$with_hwloc" && test "$with_hwloc" != "yes"; then
-            pmix_hwloc_dir=$with_hwloc
+            pmix_hwloc_dir=$with_hwloc/include
             pmix_hwloc_standard_header_location=no
             pmix_hwloc_standard_lib_location=no
             AS_IF([test -z "$with_hwloc_libdir" || test "$with_hwloc_libdir" = "yes"],

diff --git a/opal/mca/pmix/pmix4x/pmix/configure.ac b/opal/mca/pmix/pmix4x/pmix/configure.ac
@@ -171,7 +171,7 @@ LT_PREREQ([2.2.6])
 
 pmix_enable_shared="$enable_shared"
 pmix_enable_static="$enable_static"
-AS_IF([test ! -z "$enable_static" && test "$enable_static" == "yes"],
+AS_IF([test ! -z "$enable_static" && test "$enable_static" = "yes"],
       [CFLAGS="$CFLAGS -fPIC"])
 
 AM_ENABLE_SHARED
@@ -201,7 +201,6 @@ AS_IF([test "$pmix_debug" = "1"],
 
 LT_INIT()
 LT_LANG([C])
-LT_LANG([C++])
 
 ############################################################################
 # Setup the core

diff --git a/opal/mca/pmix/pmix4x/pmix/contrib/whitespace-purge.sh b/opal/mca/pmix/pmix4x/pmix/contrib/whitespace-purge.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 #
-# Copyright (c) 2015      Intel, Inc. All rights reserved.
+# Copyright (c) 2015-2019 Intel, Inc.  All rights reserved.
 # Copyright (c) 2015      Los Alamos National Security, LLC. All rights
 #                         reserved
 # Copyright (c) 2015 Cisco Systems, Inc.
@@ -18,7 +18,7 @@ for file in $(git ls-files) ; do
     # skip sym links, pdfs, etc. If any other file types should be
     # skipped add the check here.
     type=$(file -b --mime-type -h $file)
-    if test ${type::4} == "text" ; then
+    if test ${type::4} = "text" ; then
         # Eliminate whitespace at the end of lines
         perl -pi -e 's/\s*$/\n/' $file
     fi

diff --git a/opal/mca/pmix/pmix4x/pmix/include/pmix_common.h.in b/opal/mca/pmix/pmix4x/pmix/include/pmix_common.h.in
@@ -268,14 +268,19 @@ typedef uint32_t pmix_rank_t;
 #define PMIX_PARENT_ID                      "pmix.parent"           // (pmix_proc_t*) identifier of the process that called PMIx_Spawn
                                                                     //                to launch this proc's application
 #define PMIX_EXIT_CODE                      "pmix.exit.code"        // (int) exit code returned when proc terminated
-#define PMIX_NETWORK_COORDINATE             "pmix.net.coord"        // (pmix_coord_t*) Network coordinate of the specified process
-#define PMIX_NETWORK_COORD_SYSTEM           "pmix.net.coordsys"     // (pmix_coord_system_t) Network coordinate system being employed to
-                                                                    //                describe the specified network plane
+#define PMIX_NETWORK_COORDINATE             "pmix.net.coord"        // (pmix_coord_t*) Network coordinate of the specified process in
+                                                                    //       the given view type (e.g., logical vs physical)
+#define PMIX_NETWORK_VIEW                   "pmix.net.view"         // (pmix_coord_view_t) Requested view type (e.g., logical vs physical)
+#define PMIX_NETWORK_DIMS                   "pmix.net.dims"         // (uint32_t) Number of dimensions in the specified network plane/view
 #define PMIX_NETWORK_PLANE                  "pmix.net.plane"        // (char*) string ID of a network plane
 #define PMIX_NETWORK_SWITCH                 "pmix.net.switch"       // (char*) string ID of a network switch
 #define PMIX_NETWORK_NIC                    "pmix.net.nic"          // (char*) string ID of a NIC
 #define PMIX_NETWORK_ENDPT                  "pmix.net.endpt"        // (assigned) network endpt for process - type assigned by
                                                                     //            fabric provider
+#define PMIX_NETWORK_SHAPE                  "pmix.net.shape"        // (pmix_data_array_t*) number of interfaces (uint32_t) on each dimension of the
+                                                                    //        specified network plane in the requested view
+#define PMIX_NETWORK_SHAPE_STRING           "pmix.net.shapestr"     // (char*) network shape expressed as a string (e.g., "10x12x2")
+
 
 /* size info */
 #define PMIX_UNIV_SIZE                      "pmix.univ.size"        // (uint32_t) #procs in this nspace
@@ -1079,19 +1084,6 @@ typedef enum {
 } pmix_group_operation_t;
 
 
-/* define a structure for a proc's network coordinate */
-typedef struct pmix_coord {
-    int x;
-    int y;
-    int z;
-} pmix_coord_t;
-
-/* define coordinate system values */
-typedef uint8_t pmix_coord_system_t;
-#define PMIX_COORD_CARTESIAN    0x00
-#define PMIX_COORD_TOROID       0x01
-#define PMIX_COORD_CYLINDRICAL  0x02
-
 
 /* define some "hooks" external libraries can use to
  * intercept memory allocation/release operations */
@@ -1143,6 +1135,60 @@ static inline void* pmix_calloc(size_t n, size_t m)
     (PMIX_CHECK_NSPACE((a)->nspace, (b)->nspace) && ((a)->rank == (b)->rank || (PMIX_RANK_WILDCARD == (a)->rank || PMIX_RANK_WILDCARD == (b)->rank)))
 
 
+/****    PMIX COORD    ****/
+/* define coordinate system views */
+typedef uint8_t pmix_coord_view_t;
+#define PMIX_COORD_VIEW_UNDEF       0x00
+#define PMIX_COORD_LOGICAL_VIEW     0x01
+#define PMIX_COORD_PHYSICAL_VIEW    0x02
+
+/* define a structure for a proc's network coordinate */
+typedef struct pmix_coord {
+    pmix_coord_view_t view;
+    int *coord;
+    size_t dims;
+} pmix_coord_t;
+
+#define PMIX_COORD_CREATE(m, d, n)                                      \
+    do {                                                                \
+        (m) = (pmix_coord_t*)pmix_calloc((n), sizeof(pmix_coord_t));    \
+        if (NULL != (m)) {                                              \
+            (m)->view = PMIX_COORD_VIEW_UNDEF;                          \
+            (m)->dims = (d);                                            \
+            (m)->coord = (int*)pmix_calloc((m)->dims, sizeof(int));     \
+        }                                                               \
+    } while(0)
+
+#define PMIX_COORD_CONSTRUCT(m, d)          \
+    do {                                    \
+        (m)->view = PMIX_COORD_VIEW_UNDEF;  \
+        (m)->coord = NULL;                  \
+        (m)->dims = 0;                      \
+    } while(0)
+
+#define PMIX_COORD_DESTRUCT(m)              \
+    do {                                    \
+        (m)->view = PMIX_COORD_VIEW_UNDEF;  \
+        if (NULL != (m)->coord) {           \
+            pmix_free((m)->coord);          \
+            (m)->coord = NULL;              \
+            (m)->dims = 0;                  \
+        }                                   \
+    } while(0)
+
+#define PMIX_COORD_FREE(m, n)                       \
+    do {                                            \
+        size_t _nc_;                                \
+        if (NULL != (m)) {                          \
+            for (_nc_ = 0; _nc_ < (n); _nc_++) {    \
+                PMIX_COORD_DESTRUCT(&(m)[_nc_]);    \
+            }                                       \
+            free((m));                              \
+            (m) = NULL;                             \
+        }                                           \
+    } while(0)
+
+
 /****    PMIX BYTE OBJECT    ****/
 typedef struct pmix_byte_object {
     char *bytes;

diff --git a/opal/mca/pmix/pmix4x/pmix/include/pmix_sched.h b/opal/mca/pmix/pmix4x/pmix/include/pmix_sched.h
@@ -64,16 +64,20 @@ extern "C" {
 typedef struct pmix_fabric_s {
     /* user-supplied name for this fabric */
     char *name;
-    /* revision - tracks how many times the
-     * fabric info has been updated. Used to detect
-     * that a change has occurred since the last
-     * time the data was accessed. Restricted to
-     * PMIx-internal use */
-    uint64_t revision;
-    /* PMIx server-defined object for internal use */
+    /* communication cost array - the number of vertices
+     * (nverts) equals the number of interfaces in the
+     * fabric. This equates to the number of columns & rows
+     * in the commcost array as the matrix is symmetric */
+    uint16_t **commcost;
+    uint32_t nverts;
+    /* object pointer for use by the PMIx server library */
     void *module;
 } pmix_fabric_t;
 
+/* convenience macros to support pmix_fabric_t */
+#define PMIX_FABRIC_CONSTRUCT(x) \
+    memset(x, 0, sizeof(pmix_fabric_t))
+
 /* Register for access to fabric-related information, including
  * communication cost matrix. This call must be made prior to
  * requesting information from a fabric.
@@ -83,9 +87,9 @@ typedef struct pmix_fabric_s {
  *          utilize this field
  *
  * directives - an optional array of values indicating desired
- *              behaviors and/or fabric to be accessed via
- *              the returned struct. If NULL, then the highest
- *              priority available fabric will return the struct
+ *              behaviors and/or fabric to be accessed. If NULL,
+ *              then the highest priority available fabric will
+ *              be used
  *
  * ndirs - number of elements in the directives array
  *
@@ -106,62 +110,9 @@ PMIX_EXPORT pmix_status_t PMIx_server_register_fabric(pmix_fabric_t *fabric,
  */
 PMIX_EXPORT pmix_status_t PMIx_server_deregister_fabric(pmix_fabric_t *fabric);
 
-/* Get the number of vertices in the provided fabric.
- * To avoid blocking the caller, this function will
- * always return immediately, returning a PMIX_ERR_RESOURCE_BUSY
- * status if the matrix is in the process of being updated.
- *
- * fabric - pointer to the pmix_fabric_t struct provided
- *          to the registration function
- *
- * nverts - pointer to the location where the number of
- *          vertices is to be returned
- *
- * Return values include:
- *
- * PMIX_SUCCESS - indicates return of a valid value
- * PMIX_ERR_RESOURCE_BUSY - matrix is being updated
- * PMIX_ERR_FABRIC_UPDATED - fabric info has been updated since
- *                           last call involving this pmix_fabric_t
- */
-PMIX_EXPORT pmix_status_t PMIx_server_get_num_vertices(pmix_fabric_t *fabric,
-                                                       uint32_t *nverts);
-
-/* Obtain communication cost for messages transmitted from indicated
- * source to destination across the provided fabric - i.e.,
- * the value of the (src,dest) entry of that fabric's communication
- * cost matrix. To avoid blocking the caller, this function will
- * always return immediately, returning a PMIX_ERR_RESOURCE_BUSY
- * status if the matrix is in the process of being updated.
- *
- * fabric - pointer to the pmix_fabric_t struct provided to
- *          the registration function
- *
- * src - the index of the originating vertex for the communication
- *
- * dest - the index of the destination vertex for the communication
- *
- * cost - pointer to the location where the cost is to be returned
- *
- * Return values include:
- *
- * PMIX_SUCCESS - indicates return of a valid value
- * PMIX_ERR_BAD_PARAM - src and/or dest is out of bounds
- * PMIX_ERR_RESOURCE_BUSY - matrix is being updated
- * PMIX_ERR_FABRIC_UPDATED - fabric info has been updated since
- *                           last call involving this pmix_fabric_t
- */
-PMIX_EXPORT pmix_status_t PMIx_server_get_comm_cost(pmix_fabric_t *fabric,
-                                                    uint32_t src, uint32_t dest,
-                                                    uint16_t *cost);
-
-/* Given a communication cost matrix index, return the corresponding
- * vertex info in the provided fabric and the name of the node upon
+/* Given a communication cost matrix index for a specified fabric,
+ * return the corresponding vertex info and the name of the node upon
  * which it resides.
- * If the PMIX_ERR_RESOURCE_BUSY or PMIX_ERR_FABRIC_UPDATED status is
- * returned, then the caller should update their cost information
- * before re-issuing this request to ensure accurate correlation
- * between cost and LID
  *
  * fabric - pointer to the pmix_fabric_t struct provided to
  *          the registration function
@@ -179,20 +130,14 @@ PMIX_EXPORT pmix_status_t PMIx_server_get_comm_cost(pmix_fabric_t *fabric,
  *
  * PMIX_SUCCESS - indicates return of a valid value
  * PMIX_ERR_BAD_PARAM - provided index is out of bounds
- * PMIX_ERR_RESOURCE_BUSY - matrix is being updated
- * PMIX_ERR_FABRIC_UPDATED - fabric info has been updated since
- *                           last call involving this pmix_fabric_t
  */
 PMIX_EXPORT pmix_status_t PMIx_server_get_vertex_info(pmix_fabric_t *fabric,
                                                       uint32_t i, pmix_value_t *vertex,
                                                       char **nodename);
 
-/* Given vertex info, return the corresponding communication cost matrix
- * index and the name of the node upon which it resides.
- * If the PMIX_ERR_RESOURCE_BUSY or PMIX_ERR_FABRIC_UPDATED status is
- * returned, then the caller should update their cost information
- * before re-issuing this request to ensure accurate correlation
- * between cost and LID
+/* Given vertex info and the name of the device upon which that
+ * vertex resides, return the corresponding communication cost matrix
+ * index
  *
  * fabric - pointer to the pmix_fabric_t struct provided to
  *          the registration function
@@ -201,10 +146,6 @@ PMIX_EXPORT pmix_status_t PMIx_server_get_vertex_info(pmix_fabric_t *fabric,
  *
  * i - pointer to the location where the index is to be returned
  *
- * nodename - pointer to the location where the string nodename
- *            is to be returned. The caller is responsible for
- *            releasing the string when done
- *
  * Return values include:
  *
  * PMIX_SUCCESS - indicates return of a valid value
@@ -214,8 +155,7 @@ PMIX_EXPORT pmix_status_t PMIx_server_get_vertex_info(pmix_fabric_t *fabric,
  *                           last call involving this pmix_fabric_t
   */
 PMIX_EXPORT pmix_status_t PMIx_server_get_index(pmix_fabric_t *fabric,
-                                                pmix_value_t *vertex, uint32_t *i,
-                                                char **nodename);
+                                                pmix_value_t *vertex, uint32_t *i);
 
 #if defined(c_plusplus) || defined(__cplusplus)
 }