From e3bd00b4348795e5d988156fa0f5666eef57f09a Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Fri, 12 Dec 2003 01:36:02 +0000
Subject: [PATCH] Add support for BNR communications. This initial version
supports all functions, but only for a single task. Multi-task support will
need to wait for modifications to srun and its communications protocol.
---
Makefile.am | 1 +
slurm/bnr.h | 23 ++++
src/api/Makefile.am | 3 +-
src/api/bnr.c | 269 ++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 295 insertions(+), 1 deletion(-)
create mode 100644 slurm/bnr.h
create mode 100644 src/api/bnr.c
diff --git a/Makefile.am b/Makefile.am
index f5ced457d25..034797d74d1 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -20,6 +20,7 @@ EXTRA_DIST = \
META
pkginclude_HEADERS = \
+ slurm/bnr.h \
slurm/slurm.h \
slurm/slurm_errno.h
diff --git a/slurm/bnr.h b/slurm/bnr.h
new file mode 100644
index 00000000000..25b4b727125
--- /dev/null
+++ b/slurm/bnr.h
@@ -0,0 +1,23 @@
+/*
+ * BNR interface definitions, based upon
+ * Interfacing Prallel Jobs to Process Managers
+ * Brian Toonen, et. al.
+ *
+ * http://www-unix.globus.org/mail_archive/mpich-g/2001/Archive/ps00000.ps
+ */
+
+typedef int BNR_gid;
+#define BNR_MAXATTRLEN 64
+#define BNR_MAXVALLEN 3*1024
+
+#define BNR_SUCCESS 0
+#define BNR_ERROR 1
+
+int BNR_Init(BNR_gid *mygid);
+int BNR_Put(BNR_gid gid, char *attr, char *val);
+int BNR_Fence(BNR_gid gid);
+int BNR_Get(BNR_gid gid, char *attr, char *val);
+int BNR_Finalize();
+int BNR_Rank(BNR_gid group, int *myrank);
+int BNR_Nprocs(BNR_gid group, int *nprocs);
+
diff --git a/src/api/Makefile.am b/src/api/Makefile.am
index 9364a6a8e5e..15dc990061c 100644
--- a/src/api/Makefile.am
+++ b/src/api/Makefile.am
@@ -20,7 +20,7 @@ INCLUDES = -I$(top_srcdir)
#
# If any interfaces have been removed, set age = 0
#
-current = 1
+current = 2
rev = 0
age = 0
@@ -28,6 +28,7 @@ lib_LTLIBRARIES = libslurm.la
libslurm_la_SOURCES = \
allocate.c \
+ bnr.c \
cancel.c \
complete.c \
config_info.c \
diff --git a/src/api/bnr.c b/src/api/bnr.c
new file mode 100644
index 00000000000..936d4f1b087
--- /dev/null
+++ b/src/api/bnr.c
@@ -0,0 +1,269 @@
+/*****************************************************************************\
+ * bnr.c - SLURM implemenation of BNR interface, based upon
+ * Interfacing Prallel Jobs to Process Managers
+ * Brian Toonen, et. al.
+ *
+ * http://www-unix.globus.org/mail_archive/mpich-g/2001/Archive/ps00000.ps
+ *****************************************************************************
+ * Copyright (C) 2003 The Regents of the University of California.
+ * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
+ * Written by Moe Jette <jette1@llnl.gov>, et. al
+ * UCRL-CODE-2002-040.
+ *
+ * This file is part of SLURM, a resource management program.
+ * For details, see <http://www.llnl.gov/linux/slurm/>.
+ *
+ * SLURM is free software; you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * SLURM is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with SLURM; if not, write to the Free Software Foundation, Inc.,
+ * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
+\*****************************************************************************/
+
+#if HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#ifdef WITH_PTHREADS
+# include <pthread.h>
+#endif
+
+#include <limits.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <slurm/bnr.h>
+
+#include "src/common/macros.h"
+#include "src/common/xmalloc.h"
+#include "src/common/xstring.h"
+
+#define BNR_MAX_GROUPS 256
+
+struct bnr_group_info {
+ int active;
+ int my_rank;
+ int nprocs;
+ int keys;
+ char ** attr;
+ char ** val;
+};
+struct bnr_group_info bnr_groups[BNR_MAX_GROUPS];
+
+static pthread_mutex_t bnr_lock = PTHREAD_MUTEX_INITIALIZER;
+
+int BNR_Init(BNR_gid *mygid)
+{
+ int group = -1, i;
+ char *my_rank_char, *nprocs_char;
+
+ slurm_mutex_lock(&bnr_lock);
+
+ /* get a group id */
+ for (i=0; i<BNR_MAX_GROUPS; i++) {
+ if (bnr_groups[i].active)
+ continue;
+ group = i;
+ break;
+ }
+ if (group < 0) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "Exhausted supply of BNR groups\n");
+ return BNR_ERROR;
+ }
+
+ /* get the rank */
+ bnr_groups[group].my_rank = -1;
+ my_rank_char = getenv("SLURM_PROCID");
+ if (my_rank_char) {
+ long my_rank_long;
+ char *end_ptr;
+ my_rank_long = strtol(my_rank_char, &end_ptr, 10);
+ if ( (my_rank_long != LONG_MAX) &&
+ (my_rank_long != LONG_MIN) &&
+ (end_ptr[0] == '\0') )
+ bnr_groups[group].my_rank = my_rank_long;
+ }
+ if (bnr_groups[group].my_rank < 0) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "SLURM_PROCID environment variable not set\n");
+ return BNR_ERROR;
+ }
+
+ /* get the number of tasks */
+ bnr_groups[group].nprocs = -1;
+ nprocs_char = getenv("SLURM_NPROCS");
+ if (nprocs_char) {
+ long nprocs_long;
+ char *end_ptr;
+ nprocs_long = strtol(nprocs_char, &end_ptr, 10);
+ if ( (nprocs_long != LONG_MAX) &&
+ (nprocs_long != LONG_MIN) &&
+ (end_ptr[0] == '\0') )
+ bnr_groups[group].nprocs = nprocs_long;
+ }
+ if (bnr_groups[group].nprocs < 0) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "SLURM_NPROCS environment variable not set\n");
+ return BNR_ERROR;
+ }
+
+ bnr_groups[group].active = 1;
+ *mygid = group;
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+}
+
+int BNR_Put(BNR_gid gid, char *attr, char *val)
+{
+ int i;
+
+ slurm_mutex_lock(&bnr_lock);
+
+ if ( (gid < 0) || (gid >= BNR_MAX_GROUPS) ||
+ (bnr_groups[gid].active == 0) ) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "BNR_Put: Invalid group id %d\n", gid);
+ return BNR_ERROR;
+ }
+
+ if ( (strlen(attr) > BNR_MAXATTRLEN) ||
+ (strlen(val) > BNR_MAXVALLEN) ) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "BNR_Put: argument too large\n");
+ return BNR_ERROR;
+ }
+
+ /* look for duplicate attr */
+ for (i=0; i<bnr_groups[gid].keys; i++) {
+ if (strcmp(bnr_groups[gid].attr[i], attr))
+ continue;
+ xfree(bnr_groups[gid].val[i]);
+ bnr_groups[gid].val[i] = xstrdup(val);
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+ }
+
+ /* add new record */
+ xrealloc(bnr_groups[gid].attr,
+ (sizeof(char *) * (bnr_groups[gid].keys+1)));
+ xrealloc(bnr_groups[gid].val,
+ (sizeof(char *) * (bnr_groups[gid].keys+1)));
+ bnr_groups[gid].attr[bnr_groups[gid].keys] = xstrdup(attr);
+ bnr_groups[gid].val [bnr_groups[gid].keys] = xstrdup(val);
+ bnr_groups[gid].keys++;
+
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+}
+
+int BNR_Fence(BNR_gid gid)
+{
+ slurm_mutex_lock(&bnr_lock);
+
+ if ( (gid < 0) || (gid >= BNR_MAX_GROUPS) ||
+ (bnr_groups[gid].active == 0) ) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "BNR_Fence: Invalid group id %d\n", gid);
+ return BNR_ERROR;
+ }
+
+ if (bnr_groups[gid].nprocs == 1) { /* no other tasks to sync */
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+ }
+
+ /* upload keypairs to central DB
+ * wait for all tasks to register
+ * download aggregate key pair DB */
+ fprintf(stderr, "BNR_Fence code needed here\n");
+
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+}
+
+int BNR_Get(BNR_gid gid, char *attr, char *val)
+{
+ int i;
+
+ slurm_mutex_lock(&bnr_lock);
+
+ if ( (gid < 0) || (gid >= BNR_MAX_GROUPS) ||
+ (bnr_groups[gid].active == 0) ) {
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "BNR_Get: Invalid group id %d\n", gid);
+ return BNR_ERROR;
+ }
+
+ for (i=0; i<bnr_groups[gid].keys; i++) {
+ if (strcmp(attr, bnr_groups[gid].attr[i]))
+ continue;
+ strcpy(val, bnr_groups[gid].val[i]);
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+ }
+
+ slurm_mutex_unlock(&bnr_lock);
+ fprintf(stderr, "BNR_GET: No such attr %s\n", attr);
+ return BNR_ERROR;
+}
+
+int BNR_Finalize()
+{
+ int i, j;
+
+ slurm_mutex_lock(&bnr_lock);
+ for (i=0; i<BNR_MAX_GROUPS; i++) {
+ if (bnr_groups[i].active == 0)
+ continue;
+ for (j=0; j<bnr_groups[i].keys; j++) {
+ xfree(bnr_groups[i].attr[i]);
+ xfree(bnr_groups[i].val[i]);
+ }
+ xfree(bnr_groups[i].attr);
+ xfree(bnr_groups[i].val);
+ bnr_groups[i].active = 0;
+ }
+
+ slurm_mutex_unlock(&bnr_lock);
+ return BNR_SUCCESS;
+}
+
+int BNR_Rank(BNR_gid group, int *myrank)
+{
+ int rc = -1;
+
+ slurm_mutex_lock(&bnr_lock);
+ if ( (group >= 0) && (group < BNR_MAX_GROUPS) &&
+ (bnr_groups[group].active) )
+ rc = bnr_groups[group].my_rank;
+ else
+ fprintf(stderr, "BNR_Rank: Invalid group id %d\n", group);
+
+ slurm_mutex_unlock(&bnr_lock);
+ return rc;
+}
+
+int BNR_Nprocs(BNR_gid group, int *nprocs)
+{
+ int rc = -1;
+
+ slurm_mutex_lock(&bnr_lock);
+
+ if ( (group >= 0) && (group < BNR_MAX_GROUPS) &&
+ (bnr_groups[group].active) )
+ rc = bnr_groups[group].nprocs;
+ else
+ fprintf(stderr, "BNR_Nprocsk: Invalid group id %d\n", group);
+
+ slurm_mutex_unlock(&bnr_lock);
+ return -1;
+}
--
GitLab