{lib,mpi}[GCC/11.2.0,GCCcore/11.2.0] OpenMPI v4.1.1, PMIx v4.1.0

easybuild/easyconfigs/o/OpenMPI/OpenMPI-4.1.1-GCC-11.2.0.eb

@@ -0,0 +1,46 @@
+name = 'OpenMPI'
+version = '4.1.1'
+
+homepage = 'https://www.open-mpi.org/'
+description = """The Open MPI Project is an open source MPI-3 implementation."""
+
+toolchain = {'name': 'GCC', 'version': '11.2.0'}
+
+source_urls = ['https://www.open-mpi.org/software/ompi/v%(version_major_minor)s/downloads']
+sources = [SOURCELOWER_TAR_BZ2]
+patches = [
+    'OpenMPI-4.1.1_fix-bufferoverflow-in-common_ofi.patch',
+    'OpenMPI-4.0.6_remove-pmix-check-in-pmi-switch.patch',
+    'OpenMPI-4.1.1_opal-pmix-package-rank.patch',
+    'OpenMPI-4.1.1_pmix3x-protection.patch',
+]
+checksums = [
+    'e24f7a778bd11a71ad0c14587a7f5b00e68a71aa5623e2157bafee3d44c07cda',  # openmpi-4.1.1.tar.bz2
+    # OpenMPI-4.1.1_fix-bufferoverflow-in-common_ofi.patch
+    'a189d834506f3d7c31eda6aa184598a3631ea24a94bc551d5ed1f053772ca49e',
+    # OpenMPI-4.0.6_remove-pmix-check-in-pmi-switch.patch
+    '8acee6c9b2b4bf12873a39b85a58ca669de78e90d26186e52f221bb4853abc4d',
+    '04353672cf7be031e5306c94068d7012d99e6cd94b69d93230797ffcd7f31903',  # OpenMPI-4.1.1_opal-pmix-package-rank.patch
+    '384ef9f1fa803b0d71dae2ec0748d0f20295992437532afedf21478bda164ff8',  # OpenMPI-4.1.1_pmix3x-protection.patch
+]
+
+builddependencies = [
+    ('pkg-config', '0.29.2'),
+]
+
+dependencies = [
+    ('zlib', '1.2.11'),
+    ('hwloc', '2.5.0'),
+    ('libevent', '2.1.12'),
+    ('UCX', '1.11.0'),
+    ('libfabric', '1.13.0'),
+    ('PMIx', '4.1.0'),
+]
+
+# disable MPI1 compatibility for now, see what breaks...
+# configopts = '--enable-mpi1-compatibility '
+
+# to enable SLURM integration (site-specific)
+# configopts += '--with-slurm --with-pmi=/usr/include/slurm --with-pmi-libdir=/usr'
+
+moduleclass = 'mpi'

easybuild/easyconfigs/o/OpenMPI/OpenMPI-4.1.1_opal-pmix-package-rank.patch

@@ -0,0 +1,66 @@
+see https://github.com/open-mpi/ompi/pull/9212
+
+From 8b07775d1002cc9be887c41768aea9ac896c5ecd Mon Sep 17 00:00:00 2001
+From: Ralph Castain <[email protected]>
+Date: Tue, 10 Aug 2021 08:38:15 -0700
+Subject: [PATCH] Use an OPAL-prefixed abstraction for PMIX_PACKAGE_RANK
+
+If someone configures against PMIx v4.1 or above, the configure
+logic will correctly detect the presence of PMIX_PACKAGE_RANK.
+However, the internal code only includes the opal/mca/pmix headers
+and thus only the OPAL-prefixed PMIx abstractions are available.
+Add an OPAL_PMIX_PACKAGE_RANK abstraction and update the common/ofi
+code to use it.
+
+Signed-off-by: Ralph Castain <[email protected]>
+---
+ opal/mca/common/ofi/common_ofi.c | 6 ++----
+ opal/mca/pmix/pmix_types.h       | 2 ++
+ 2 files changed, 4 insertions(+), 4 deletions(-)
+
+diff --git a/opal/mca/common/ofi/common_ofi.c b/opal/mca/common/ofi/common_ofi.c
+index 12bed7eb6b8..769ecb75239 100644
+--- a/opal/mca/common/ofi/common_ofi.c
++++ b/opal/mca/common/ofi/common_ofi.c
+@@ -312,19 +312,17 @@ static uint32_t get_package_rank(int32_t num_local_peers, uint16_t my_local_rank
+     char *local_peers = NULL;
+     char *locality_string = NULL;
+     char *mylocality = NULL;
++    uint16_t *package_rank_ptr;
+
+     pname.jobid = OPAL_PROC_MY_NAME.jobid;
+     pname.vpid = OPAL_VPID_WILDCARD;
+
+-#if HAVE_DECL_PMIX_PACKAGE_RANK
+-    uint16_t *package_rank_ptr;
+     // Try to get the PACKAGE_RANK from PMIx
+-    OPAL_MODEX_RECV_VALUE_OPTIONAL(rc, PMIX_PACKAGE_RANK,
++    OPAL_MODEX_RECV_VALUE_OPTIONAL(rc, OPAL_PMIX_PACKAGE_RANK,
+                                    &pname, &package_rank_ptr, OPAL_UINT16);
+     if (OPAL_SUCCESS == rc) {
+         return (uint32_t)*package_rank_ptr;
+     }
+-#endif
+
+     // Get the local peers
+     OPAL_MODEX_RECV_VALUE(rc, OPAL_PMIX_LOCAL_PEERS,
+diff --git a/opal/mca/pmix/pmix_types.h b/opal/mca/pmix/pmix_types.h
+index d2643b2e549..816e386cbf1 100644
+--- a/opal/mca/pmix/pmix_types.h
++++ b/opal/mca/pmix/pmix_types.h
+@@ -2,6 +2,7 @@
+  * Copyright (c) 2014-2018 Intel, Inc. All rights reserved.
+  * Copyright (c) 2016      Research Organization for Information Science
+  *                         and Technology (RIST). All rights reserved.
++ * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
+  * $COPYRIGHT$
+  *
+  * Additional copyrights may follow
+@@ -130,6 +131,7 @@ BEGIN_C_DECLS
+ #define OPAL_PMIX_NPROC_OFFSET                  "pmix.offset"           // (uint32_t) starting global rank of this job
+ #define OPAL_PMIX_LOCAL_RANK                    "pmix.lrank"            // (uint16_t) rank on this node within this job
+ #define OPAL_PMIX_NODE_RANK                     "pmix.nrank"            // (uint16_t) rank on this node spanning all jobs
++#define OPAL_PMIX_PACKAGE_RANK                  "pmix.pkgrank"          // (uint16_t) rank within this job on the package where this proc resides
+ #define OPAL_PMIX_LOCALLDR                      "pmix.lldr"             // (uint64_t) opal_identifier of lowest rank on this node within this job
+ #define OPAL_PMIX_APPLDR                        "pmix.aldr"             // (uint32_t) lowest rank in this app within this job
+ #define OPAL_PMIX_PROC_PID                      "pmix.ppid"             // (pid_t) pid of specified proc

easybuild/easyconfigs/o/OpenMPI/OpenMPI-4.1.1_pmix3x-protection.patch

@@ -0,0 +1,81 @@
+see https://github.com/open-mpi/ompi/pull/9213
+
+From bd48a06e25c1058b7d3c3ba4414af388836c8219 Mon Sep 17 00:00:00 2001
+From: Ralph Castain <[email protected]>
+Date: Tue, 10 Aug 2021 13:01:11 -0700
+Subject: [PATCH] Protect the pmix3x component from a PMIx v4 attribute
+
+PMIx v4 introduced a new PMIX_TOPO2 attribute that takes a data type
+unknown to PMIx 3. Unfortunately, that attribute can be provided by
+the RM without our knowledge. We will update PMIx to try and detect
+it and prevent it from slipping down to this level. Meantime, add
+some simple protection here.
+
+Signed-off-by: Ralph Castain <[email protected]>
+---
+ opal/mca/pmix/pmix3x/pmix3x.c | 29 +++++++++++++++++------------
+ 1 file changed, 17 insertions(+), 12 deletions(-)
+
+diff --git a/opal/mca/pmix/pmix3x/pmix3x.c b/opal/mca/pmix/pmix3x/pmix3x.c
+index 90670f6adb6..3b2a5353cc6 100644
+--- a/opal/mca/pmix/pmix3x/pmix3x.c
++++ b/opal/mca/pmix/pmix3x/pmix3x.c
+@@ -8,6 +8,7 @@
+  * Copyright (c) 2016      Cisco Systems, Inc.  All rights reserved.
+  * Copyright (c) 2017      Los Alamos National Security, LLC. All rights
+  *                         reserved.
++ * Copyright (c) 2021      Nanook Consulting.  All rights reserved.
+  * $COPYRIGHT$
+  *
+  * Additional copyrights may follow
+@@ -966,6 +967,7 @@ int pmix3x_value_unload(opal_value_t *kv,
+     opal_list_t *lt;
+     opal_value_t *ival;
+     size_t n;
++    pmix_info_t *iptr;
+
+     switch(v->type) {
+     case PMIX_UNDEF:
+@@ -1178,27 +1180,30 @@ int pmix3x_value_unload(opal_value_t *kv,
+         kv->data.pinfo.state = pmix3x_convert_state(v->data.pinfo->state);
+         break;
+     case PMIX_DATA_ARRAY:
+-        if (NULL == v->data.darray || NULL == v->data.darray->array) {
++        if (NULL == v->data.darray || NULL == v->data.darray->array ||
++            PMIX_INFO != v->data.darray->type) {
+             kv->data.ptr = NULL;
+             break;
+         }
+         lt = OBJ_NEW(opal_list_t);
+         kv->type = OPAL_PTR;
+         kv->data.ptr = (void*)lt;
++        iptr = (pmix_info_t*)v->data.darray->array;
+         for (n=0; n < v->data.darray->size; n++) {
++            if (0 == strcmp("pmix.topo2", iptr[n].key)) {
++                /* we do not know (yet) how to convert the pmix.topo2 key from PMIx 4.0.0
++                 * but since we are not going to use it, simply ignore it and move on */
++                continue;
++            }
+             ival = OBJ_NEW(opal_value_t);
+             opal_list_append(lt, &ival->super);
+-            /* handle the various types */
+-            if (PMIX_INFO == v->data.darray->type) {
+-                pmix_info_t *iptr = (pmix_info_t*)v->data.darray->array;
+-                ival->key = strdup(iptr[n].key);
+-                rc = pmix3x_value_unload(ival, &iptr[n].value);
+-                if (OPAL_SUCCESS != rc) {
+-                    OPAL_LIST_RELEASE(lt);
+-                    kv->type = OPAL_UNDEF;
+-                    kv->data.ptr = NULL;
+-                    break;
+-                }
++            ival->key = strdup(iptr[n].key);
++            rc = pmix3x_value_unload(ival, &iptr[n].value);
++            if (OPAL_SUCCESS != rc) {
++                OPAL_LIST_RELEASE(lt);
++                kv->type = OPAL_UNDEF;
++                kv->data.ptr = NULL;
++                break;
+             }
+         }
+         break;

easybuild/easyconfigs/p/PMIx/PMIx-4.1.0-GCCcore-11.2.0.eb

@@ -0,0 +1,46 @@
+easyblock = 'ConfigureMake'
+
+name = 'PMIx'
+version = '4.1.0'
+
+homepage = 'https://pmix.org/'
+description = """Process Management for Exascale Environments
+PMI Exascale (PMIx) represents an attempt to
+provide an extended version of the PMI standard specifically designed
+to support clusters up to and including exascale sizes. The overall
+objective of the project is not to branch the existing pseudo-standard
+definitions - in fact, PMIx fully supports both of the existing PMI-1
+and PMI-2 APIs - but rather to (a) augment and extend those APIs to
+eliminate some current restrictions that impact scalability, and (b)
+provide a reference implementation of the PMI-server that demonstrates
+the desired level of scalability.
+"""
+
+toolchain = {'name': 'GCCcore', 'version': '11.2.0'}
+toolchainopts = {'pic': True}
+
+source_urls = ['https://github.com/openpmix/openpmix/releases/download/v%(version)s']
+sources = ['%(namelower)s-%(version)s.tar.bz2']
+checksums = ['145f05a6c621bfb3fc434776b615d7e6d53260cc9ba340a01f55b383e07c842e']
+
+builddependencies = [('binutils', '2.37')]
+
+dependencies = [
+    ('libevent', '2.1.12'),
+    ('zlib', '1.2.11'),
+    ('hwloc', '2.5.0'),
+]
+
+configopts = ' --with-libevent=$EBROOTLIBEVENT --with-zlib=$EBROOTZLIB'
+configopts += ' --with-hwloc=$EBROOTHWLOC'
+configopts += ' --enable-pmix-binaries'
+configopts += ' --disable-man-pages'
+
+buildopts = 'V=1'
+
+sanity_check_paths = {
+    'files': ['bin/pevent', 'bin/plookup', 'bin/pmix_info', 'bin/pps'],
+    'dirs': ['etc', 'include', 'lib', 'share']
+}
+
+moduleclass = 'lib'