v4.1.x: Update PMIx to v3.2.3

opal/mca/pmix/pmix3x/pmix/NEWS

@@ -1,5 +1,6 @@
 Copyright (c) 2015-2020 Intel, Inc.  All rights reserved.
 Copyright (c) 2017-2020 IBM Corporation.  All rights reserved.
+Copyright (c) 2021      Nanook Consulting.  All rights reserved.
 $COPYRIGHT$
 
 Additional copyrights may follow
@@ -20,11 +21,16 @@ other, a single NEWS-worthy item might apply to different series. For
 example, a bug might be fixed in the master, and then moved to
 multiple release branches.
 
-Master (not on release branches yet)
-------------------------------------
+
+3.2.3 -- 12 Feb 2021
+----------------------
+ - PR #1967: Fix thread conflict in event notification
+ - PR #2001: Finalize: set hostname in global to NULL to avoid segfault
+             in apps that call init/finalize multiple times
+ - PR #2045: Have developer builds use -O3 by default
 
 
-3.2.2 -- TBD
+3.2.2 -- 7 Dec 2020
 ----------------------
  - PR #1930: Remove man page setup as there are no manpages in v3.2
  - PR #1933: Remove stale config command
@@ -66,12 +72,13 @@ Master (not on release branches yet)
  - PR #1885: Fix immediate flag behavior at the server
 
 
-3.1.6 -- 20 Aug 2020
+3.1.6 -- 04 Jan 2021
 ----------------------
  - PR #1669: Silence unnecessary error log message
  - PR #1776: Fixed crash of dstore locks destructor
  - PR #1825: Add option to bypass local-only fence optimization
  - PR #1830: Increase timeout in test case
+ - PR #1990: Fix server/get for gds=hash
 
 
 3.1.5 -- 14 Feb 2020

opal/mca/pmix/pmix3x/pmix/VERSION

@@ -16,7 +16,7 @@
 
 major=3
 minor=2
-release=2
+release=3
 
 # greek is used for alpha or beta release tags.  If it is non-empty,
 # it will be appended to the version number.  It does not have to be
@@ -31,7 +31,7 @@ greek=
 # command, or with the date (if "git describe" fails) in the form of
 # "date<date>".
 
-repo_rev=git1703e6fb
+repo_rev=gitad8d1f3
 
 # If tarball_version is not empty, it is used as the version string in
 # the tarball filename, regardless of all other versions listed in
@@ -45,7 +45,7 @@ tarball_version=
 
 # The date when this release was created
 
-date="Dec 01, 2020"
+date="Feb 11, 2021"
 
 # The shared library version of each of PMIx's public libraries.
 # These versions are maintained in accordance with the "Library
@@ -76,7 +76,7 @@ date="Dec 01, 2020"
 # Version numbers are described in the Libtool current:revision:age
 # format.
 
-libpmix_so_version=4:32:2
+libpmix_so_version=4:33:2
 libpmi_so_version=1:1:0
 libpmi2_so_version=1:0:0
 

opal/mca/pmix/pmix3x/pmix/config/pmix.m4

@@ -23,6 +23,7 @@ dnl                         and Technology (RIST).  All rights reserved.
 dnl Copyright (c) 2016      Mellanox Technologies, Inc.
 dnl                         All rights reserved.
 dnl
+dnl Copyright (c) 2021      Nanook Consulting.  All rights reserved.
 dnl $COPYRIGHT$
 dnl
 dnl Additional copyrights may follow
@@ -1018,15 +1019,11 @@ else
     AC_MSG_RESULT([no])
     WANT_DEBUG=0
 fi
-#################### Early development override ####################
-if test "$WANT_DEBUG" = "0" && test -z "$enable_debug" && test "$PMIX_DEVEL" = "1"; then
-    WANT_DEBUG=1
-    echo "--> developer override: enable debugging code by default"
-fi
-#################### Early development override ####################
+
 if test "$WANT_DEBUG" = "0"; then
     CFLAGS="-DNDEBUG $CFLAGS"
 fi
+
 AC_DEFINE_UNQUOTED(PMIX_ENABLE_DEBUG, $WANT_DEBUG,
                    [Whether we want developer-level debugging code or not])
 

opal/mca/pmix/pmix3x/pmix/contrib/pmix.spec

@@ -12,9 +12,10 @@
 # Copyright (c) 2006-2016 Cisco Systems, Inc.  All rights reserved.
 # Copyright (c) 2013      Mellanox Technologies, Inc.
 #                         All rights reserved.
-# Copyright (c) 2015-2020 Intel, Inc.  All rights reserved.
+# Copyright (c) 2015-2018 Intel, Inc.  All rights reserved.
 # Copyright (c) 2015      Research Organization for Information Science
 #                         and Technology (RIST). All rights reserved.
+# Copyright (c) 2021      Nanook Consulting.  All rights reserved.
 # $COPYRIGHT$
 #
 # Additional copyrights may follow
@@ -192,7 +193,7 @@
 
 Summary: An extended/exascale implementation of PMI
 Name: %{?_name:%{_name}}%{!?_name:pmix}
-Version: 3.2.2
+Version: 3.2.3
 Release: 1%{?dist}
 License: BSD
 Group: Development/Libraries

opal/mca/pmix/pmix3x/pmix/src/event/pmix_event.h

@@ -12,6 +12,7 @@
  *                         All rights reserved.
  * Copyright (c) 2015-2020 Intel, Inc.  All rights reserved.
  * Copyright (c) 2020      IBM Corporation.  All rights reserved.
+ * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -140,7 +141,11 @@ typedef struct pmix_event_chain_t {
     size_t nallocated;
     pmix_info_t *results;
     size_t nresults;
+    pmix_info_t *interim;
+    size_t ninterim;
     pmix_event_hdlr_t *evhdlr;
+    pmix_op_cbfunc_t opcbfunc;
+    void *cbdata;
     pmix_op_cbfunc_t final_cbfunc;
     void *final_cbdata;
 } pmix_event_chain_t;

opal/mca/pmix/pmix3x/pmix/src/event/pmix_event_notification.c

@@ -5,6 +5,7 @@
  *                         and Technology (RIST).  All rights reserved.
  * Copyright (c) 2017      IBM Corporation. All rights reserved.
  *
+ * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -32,6 +33,10 @@ static pmix_status_t notify_server_of_event(pmix_status_t status,
                                             pmix_data_range_t range,
                                             const pmix_info_t info[], size_t ninfo,
                                             pmix_op_cbfunc_t cbfunc, void *cbdata);
+static void progress_local_event_hdlr(pmix_status_t status,
+                                      pmix_info_t *results, size_t nresults,
+                                      pmix_op_cbfunc_t cbfunc, void *thiscbdata,
+                                      void *notification_cbdata);
 
 /* if we are a client, we call this function to notify the server of
  * an event. If we are a server, our host RM will call this function
@@ -313,19 +318,13 @@ static pmix_status_t notify_server_of_event(pmix_status_t status,
 }
 
 
-static void progress_local_event_hdlr(pmix_status_t status,
-                                      pmix_info_t *results, size_t nresults,
-                                      pmix_op_cbfunc_t cbfunc, void *thiscbdata,
-                                      void *notification_cbdata)
+static void cycle_events(int sd, short args, void *cbdata)
 {
-    /* this may be in the host's thread, so we need to threadshift it
-     * before accessing our internal data */
-
-    pmix_event_chain_t *chain = (pmix_event_chain_t*)notification_cbdata;
+    pmix_event_chain_t *chain = (pmix_event_chain_t*)cbdata;
     size_t n, nsave, cnt;
-    pmix_info_t *newinfo;
     pmix_list_item_t *item;
     pmix_event_hdlr_t *nxt;
+    pmix_info_t *newinfo;
 
     pmix_output_verbose(2, pmix_client_globals.event_output,
                         "%s progressing local event",
@@ -345,7 +344,7 @@ static void progress_local_event_hdlr(pmix_status_t status,
      * the array to make space */
 
     /* add in any new results plus space for the returned status */
-    nsave += nresults + 1;
+    nsave += chain->ninterim + 1;
     /* create the new space */
     PMIX_INFO_CREATE(newinfo, nsave);
     /* transfer over the prior data */
@@ -364,11 +363,11 @@ static void progress_local_event_hdlr(pmix_status_t status,
         pmix_strncpy(newinfo[cnt].key, "UNKNOWN", PMIX_MAX_KEYLEN);
     }
     newinfo[cnt].value.type = PMIX_STATUS;
-    newinfo[cnt].value.data.status = status;
+    newinfo[cnt].value.data.status = chain->status;
     ++cnt;
     /* transfer across the new results */
-    for (n=0; n < nresults; n++) {
-        PMIX_INFO_XFER(&newinfo[cnt], &results[n]);
+    for (n=0; n < chain->ninterim; n++) {
+        PMIX_INFO_XFER(&newinfo[cnt], &chain->interim[n]);
         ++cnt;
     }
     /* release the prior results */
@@ -382,14 +381,27 @@ static void progress_local_event_hdlr(pmix_status_t status,
     chain->ninfo = chain->nallocated - 2;
     PMIX_INFO_DESTRUCT(&chain->info[chain->nallocated-2]);
     PMIX_INFO_DESTRUCT(&chain->info[chain->nallocated-1]);
-
+    // call their interim cbfunc
+    if (NULL != chain->opcbfunc) {
+        chain->opcbfunc(PMIX_SUCCESS, chain->cbdata);
+    }
+
     /* if the caller indicates that the chain is completed,
      * or we completed the "last" event */
-    if (PMIX_EVENT_ACTION_COMPLETE == status || chain->endchain) {
-        goto complete;
+    if (PMIX_EVENT_ACTION_COMPLETE == chain->status || chain->endchain) {
+        if (PMIX_EVENT_ACTION_COMPLETE == chain->status) {
+            chain->status = PMIX_SUCCESS;
+        }
+        /* we still have to call their final callback */
+        if (NULL != chain->final_cbfunc) {
+            chain->final_cbfunc(chain->status, chain->final_cbdata);
+        }
+        /* maintain acctng */
+        PMIX_RELEASE(chain);
+        return;
     }
-    item = NULL;
 
+    item = NULL;
     /* see if we need to continue, starting with the single code events */
     if (1 == chain->evhdlr->ncodes) {
         /* the last handler was for a single code - see if there are
@@ -591,18 +603,31 @@ static void progress_local_event_hdlr(pmix_status_t status,
         }
     }
 
-  complete:
-    /* we still have to call their final callback */
+    /* if we get here, there was nothing more to do, but
+     * we still have to call their final callback */
     if (NULL != chain->final_cbfunc) {
-        chain->final_cbfunc(PMIX_SUCCESS, chain->final_cbdata);
+        chain->final_cbfunc(chain->status, chain->final_cbdata);
         return;
     }
     /* maintain acctng */
     PMIX_RELEASE(chain);
-    /* let the caller know that we are done with their callback */
-    if (NULL != cbfunc) {
-        cbfunc(PMIX_SUCCESS, thiscbdata);
-    }
+}
+
+static void progress_local_event_hdlr(pmix_status_t status,
+                                      pmix_info_t *results, size_t nresults,
+                                      pmix_op_cbfunc_t cbfunc, void *thiscbdata,
+                                      void *notification_cbdata)
+{
+    /* this may be in the host's thread, so we need to threadshift it
+     * before accessing our internal data */
+
+    pmix_event_chain_t *chain = (pmix_event_chain_t*)notification_cbdata;
+
+    chain->interim = results;
+    chain->ninterim = nresults;
+    chain->opcbfunc = cbfunc;
+    chain->cbdata = thiscbdata;
+    PMIX_THREADSHIFT(chain, cycle_events);
 }
 
 /* given notification of an event, cycle thru our list of
@@ -1277,6 +1302,9 @@ void pmix_event_timeout_cb(int fd, short flags, void *arg)
     /* remove it from the list */
     pmix_list_remove_item(&pmix_globals.cached_events, &ch->super);
 
+    /* protect the chain */
+    PMIX_RETAIN(ch);
+
     /* process this event thru the regular channels */
     if (PMIX_PEER_IS_SERVER(pmix_globals.mypeer) &&
         !PMIX_PEER_IS_LAUNCHER(pmix_globals.mypeer)) {
@@ -1440,7 +1468,11 @@ static void chcon(pmix_event_chain_t *p)
     p->nallocated = 0;
     p->results = NULL;
     p->nresults = 0;
+    p->interim = NULL;
+    p->ninterim = 0;
     p->evhdlr = NULL;
+    p->opcbfunc = NULL;
+    p->cbdata = NULL;
     p->final_cbfunc = NULL;
     p->final_cbdata = NULL;
 }

opal/mca/pmix/pmix3x/pmix/src/event/pmix_event_registration.c

@@ -3,6 +3,7 @@
  * Copyright (c) 2014-2020 Intel, Inc.  All rights reserved.
  * Copyright (c) 2017-2018 Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -413,7 +414,7 @@ static void check_cached_events(pmix_rshift_caddy_t *cd)
         /* we always leave space for event hdlr name and a callback object */
         chain->nallocated = ncd->ninfo + 2;
         PMIX_INFO_CREATE(chain->info, chain->nallocated);
-        if (0 < cd->ninfo) {
+        if (0 < ncd->ninfo) {
             chain->ninfo = ncd->ninfo;
             /* need to copy the info */
             for (n=0; n < ncd->ninfo; n++) {
@@ -871,7 +872,6 @@ PMIX_EXPORT void PMIx_Register_event_handler(pmix_status_t codes[], size_t ncode
                                              void *cbdata)
 {
     pmix_rshift_caddy_t *cd;
-    size_t n;
 
     PMIX_ACQUIRE_THREAD(&pmix_global_lock);
 
@@ -900,9 +900,7 @@ PMIX_EXPORT void PMIx_Register_event_handler(pmix_status_t codes[], size_t ncode
             }
             return;
         }
-        for (n=0; n < ncodes; n++) {
-            cd->codes[n] = codes[n];
-        }
+        memcpy(cd->codes, codes, ncodes * sizeof(pmix_status_t));
     }
     cd->ncodes = ncodes;
     cd->info = info;

opal/mca/pmix/pmix3x/pmix/src/runtime/pmix_finalize.c

@@ -15,6 +15,8 @@
  * Copyright (c) 2013-2020 Intel, Inc.  All rights reserved.
  * Copyright (c) 2016-2019 Research Organization for Information Science
  *                         and Technology (RIST).  All rights reserved.
+ * Copyright (c) 2021      Triad National Security, LLC. All rights reserved.
+ * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -134,6 +136,7 @@ void pmix_rte_finalize(void)
     PMIX_LIST_DESTRUCT(&pmix_globals.stdin_targets);
     if (NULL != pmix_globals.hostname) {
         free(pmix_globals.hostname);
+        pmix_globals.hostname = NULL;
     }
     PMIX_LIST_DESTRUCT(&pmix_globals.nspaces);