From 5c77bbfe330b1420791b12bba6ef78bea5108b13 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Wed, 21 Mar 2007 23:46:00 +0000
Subject: [PATCH] More work toward cpuset support.
---
src/plugins/task/affinity/affinity.c | 267 +---------------------
src/plugins/task/affinity/affinity.h | 16 +-
src/plugins/task/affinity/cpuset.c | 267 +++++++++++++++++++++-
src/plugins/task/affinity/cpuset.h | 41 ----
src/plugins/task/affinity/task_affinity.c | 64 +++---
5 files changed, 304 insertions(+), 351 deletions(-)
delete mode 100644 src/plugins/task/affinity/cpuset.h
diff --git a/src/plugins/task/affinity/affinity.c b/src/plugins/task/affinity/affinity.c
index 03103649da3..ef6593c1fdb 100644
--- a/src/plugins/task/affinity/affinity.c
+++ b/src/plugins/task/affinity/affinity.c
@@ -32,6 +32,7 @@
* with SLURM; if not, write to the Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
\*****************************************************************************/
+
#include "affinity.h"
#ifdef HAVE_PLPA
@@ -48,11 +49,13 @@ void slurm_chkaffinity(cpu_set_t *mask, slurmd_job_t *job, int statval)
int task_lid = job->envtp->localid;
pid_t mypid = job->envtp->task_pid;
- if (!(job->cpu_bind_type & CPU_BIND_VERBOSE)) return;
+ if (!(job->cpu_bind_type & CPU_BIND_VERBOSE))
+ return;
action[0] = '\0';
status[0] = '\0';
- if (statval) { strcpy(status, " FAILED"); }
+ if (statval)
+ strcpy(status, " FAILED");
if (job->cpu_bind_type & CPU_BIND_NONE) {
strcpy(action, "");
@@ -216,263 +219,3 @@ int slurm_getaffinity(pid_t pid, size_t size, cpu_set_t *mask)
debug3("sched_getaffinity(%d) = 0x%s", pid, cpuset_to_str(mask, mstr));
return (rval);
}
-
-#ifdef HAVE_CPUSETS_EXP
-int get_cpuset_mask(cs_cpumask_t *mask, slurmd_job_t *job)
-{
- int nbr_task_cpus;
- int nummasks, i, j;
- char *curstr, *selstr;
- char mstr[1 + CPU_SETSIZE / 4];
- int local_id = job->envtp->localid;
-
- debug2("get_cpuset_mask bind_type = %d, bind_list = %s\n",
- job->cpu_bind_type, job->cpu_bind);
-
- if ((!job->cpu_bind_type) || (job->cpu_bind_type & CPU_BIND_NONE)) {
- return false;
- }
-
- nbr_task_cpus = job->cpus / job->ntasks;
-
- if (job->cpu_bind_type & CPU_BIND_RANK) {
- for (i=0;i<nbr_task_cpus;i++){
- cs_cpumask_add(mask, local_id * nbr_task_cpus + i);
- }
- return true;
- }
-
- nummasks = 0;
- selstr = NULL;
-
- if (job->cpu_bind_type & CPU_BIND_MAPCPU) {
- unsigned int mycpu = 0;
- /* find first entry for this task */
- curstr = job->cpu_bind;
- while (*curstr) {
- if (nummasks == local_id*nbr_task_cpus) {
- selstr = curstr;
- break;
- }
- if (*curstr == ',')
- nummasks++;
- curstr++;
- }
- /* check if we found a cpu entry */
- if (!selstr){
- error("not enough entries in map_cpu:<list>");
- return false;
- }
- /* add cpus to mask from list for "nbr_task_cpus" */
- for (i=0;i<nbr_task_cpus;i++){
- if (!*selstr){
- error("not enough entries in map_cpu:<list>");
- return false;
- }
- /* extract the selected mask from the list */
- j = 0;
- curstr = mstr;
- while ((*selstr && *selstr != ',') && (j++ < (CPU_SETSIZE/4)))
- *curstr++ = *selstr++;
- *curstr = '\0';
- if (*selstr == ',')
- selstr++;
- if (strncmp(mstr, "0x", 2) == 0) {
- mycpu = strtoul (&(mstr[2]), NULL, 16);
- } else {
- mycpu = strtoul (mstr, NULL, 10);
- }
- cs_cpumask_add(mask, mycpu);
- }
- return true;
- }
-
- if (job->cpu_bind_type & CPU_BIND_MASKCPU) {
- /* find mask entry for this task */
- curstr = job->cpu_bind;
- while (*curstr) {
- if (nummasks == local_id) {
- selstr = curstr;
- break;
- }
- if (*curstr == ',')
- nummasks++;
- curstr++;
- }
- /* check if we found a cpu entry */
- if (!selstr){
- error("not enough entries in mask_cpu:<list>");
- return false;
- }
- /* extract the selected mask from the list */
- j = 0;
- curstr = mstr;
- while ((*selstr && *selstr != ',') && (j++ < (CPU_SETSIZE/4)))
- *curstr++ = *selstr++;
- *curstr = '\0';
- /* convert mask string into cpu_set_t mask */
- if (str_to_cpuset( (cpu_set_t *) mask, mstr) < 0) {
- error("str_to_cpuset %s", mstr);
- return false;
- }
- return true;
- }
-
- return false;
-}
-
-int get_memset_mask(cs_memmask_t *mem_mask, cs_cpumask_t *cpu_mask, slurmd_job_t *job)
-{
- cs_cpumask_t cur_mask, tst_mask;
- int nbr_nodes, nummasks, i, j;
- char *curstr, *selstr;
- char mstr[1 + CPU_SETSIZE / 4];
- int local_id = job->envtp->localid;
-
- debug2("get_memset_mask bind_type = %d, bind_list = %s\n",
- job->mem_bind_type, job->mem_bind);
-
- /* If "not specified" or "None" or "Rank", do not set a new memory mask in the CPUset */
- if ((!job->mem_bind_type) || (job->mem_bind_type & (MEM_BIND_NONE | MEM_BIND_RANK)))
- return false;
-
- /* For now, make LOCAL and MAP_CPU the same */
- if (job->mem_bind_type & (MEM_BIND_LOCAL | MEM_BIND_MAPCPU)) {
- nbr_nodes = cs_nr_nodes();
- for (i=0; i<nbr_nodes; i++) {
- cs_get_node_cpus(i, &cur_mask);
- cs_cpumask_and(&tst_mask, &cur_mask, cpu_mask);
- if (!cs_cpumask_empty(&tst_mask)) {
- cs_memmask_add(mem_mask, i);
- debug2("added node = %d to mem mask %08x \n",i,*mem_mask);
- }
- }
-
- return true;
- }
-
- /* allow user to set specific memory masks */
- if (job->mem_bind_type & MEM_BIND_MASKCPU) {
- /* find mask entry for this task */
- nummasks = 0;
- selstr = NULL;
-
- curstr = job->mem_bind;
- while (*curstr) {
- if (nummasks == local_id) {
- selstr = curstr;
- break;
- }
- if (*curstr == ',')
- nummasks++;
- curstr++;
- }
- /* check if we found a mem entry */
- if (!selstr){
- error("not enough entries in mask_mem:<list>");
- return false;
- }
- /* extract the selected mask from the list */
- j = 0;
- curstr = mstr;
- while ((*selstr && *selstr != ',') && (j++ < (CPU_SETSIZE/4)))
- *curstr++ = *selstr++;
- *curstr = '\0';
- /* convert mask string into cpu_set_t mask */
- if (str_to_cpuset( (cpu_set_t *) mem_mask, mstr) < 0) {
- error("str_to_cpuset %s", mstr);
- return false;
- }
- return true;
- }
- return false;
-}
-
-int make_task_cpuset(slurmd_job_t *job, cs_cpumask_t *cpu_mask, cs_memmask_t *mem_mask){
-
- char path[PATH_MAX];
- char *current_cs = NULL;
- int retval = 0;
- int success = 0;
-
- info("cpuset - cs_init called");
- cs_init();
-
- current_cs = cs_get_current();
- if (!current_cs)
- return -ENOMEM;
-
- int l = snprintf(path, PATH_MAX, "%sslurm%u_%d", current_cs, job->jobid,
- job->envtp->localid);
- if (l > PATH_MAX) {
- retval = -ENAMETOOLONG;
- goto out;
- }
-
- debug("cpuset path = %s",path);
-
- retval = cs_create(path);
- if (retval < 0)
- goto out;
- retval = cs_set_autoclean(path, CS_AUTOCLEAN);
- if (retval < 0)
- goto out;
-
- cs_lock_libcpuset();
- retval = cs_set_cpus(path, *cpu_mask);
- if (retval < 0)
- debug3("cpuset - error on cs_set_cpus = %d %s",retval,cs_strerror(retval));
- success = 1;
-
- if (cs_supports_mem()) {
-
- /* Check for mem_bind options */
- if (get_memset_mask(mem_mask, cpu_mask, job)) {
- debug("cpuset - mem_mask = %d (decimal) and %08x (hex)",
- *mem_mask, *mem_mask);
- retval = cs_set_mems(path, *mem_mask);
- if (retval < 0) {
- debug3("cpuset - error on cs_set_mems = %d %s",
- retval,cs_strerror(retval));
- success = 0;
- goto out_created;
- }
- } else {
- /* Copy parent of new cpuset (i.e current) mems mask */
- retval = cs_get_mems(current_cs, mem_mask);
- if (retval < 0) {
- debug3("cpuset - error on cs_get_mems = %d %s",
- retval,cs_strerror(retval));
- success = 0;
- goto out_created;
- }
- retval = cs_set_mems(path, *mem_mask);
- if (retval < 0) {
- debug3("cpuset - error on cs_set_mems = %d %s",
- retval,cs_strerror(retval));
- success = 0;
- goto out_created;
- }
- }
- }
-
- retval = cs_add_task(path, job->envtp->task_pid);
- if (retval < 0) {
- debug3("cpuset - error on cs_add_task = %d %s",retval,cs_strerror(retval));
- }
-
- out_created:
- if (!success)
- cs_destroy(path);
- cs_unlock_libcpuset();
-
- out:
- free(current_cs);
-
- current_cs = cs_get_current();
- debug("cpuset - exit make_task_cpuset retval = %d cpuset = %s",retval,current_cs);
- free(current_cs);
-
- return retval;
-}
-#endif
diff --git a/src/plugins/task/affinity/affinity.h b/src/plugins/task/affinity/affinity.h
index cad2f1cca5c..e46fc8fb27f 100644
--- a/src/plugins/task/affinity/affinity.h
+++ b/src/plugins/task/affinity/affinity.h
@@ -52,12 +52,14 @@
#include <sys/stat.h>
#include <sys/param.h>
#include <sys/poll.h>
-#include <unistd.h>
+#include <sys/stat.h>
+#include <fcntl.h>
#include <pwd.h>
#include <grp.h>
#include <stdio.h>
#include <string.h>
#include <sys/utsname.h>
+#include <unistd.h>
#define _GNU_SOURCE
#define __USE_GNU
@@ -67,10 +69,6 @@
# include <stdlib.h>
#endif
-#ifdef HAVE_CPUSETS_EXP
-# include <cpuset.h>
-#endif
-
#include <slurm/slurm_errno.h>
#include "src/common/slurm_xlator.h"
#include "src/slurmd/slurmd/slurmd.h"
@@ -95,11 +93,9 @@ void slurm_chkaffinity(cpu_set_t *mask, slurmd_job_t *job, int statval);
int get_cpuset(cpu_set_t *mask, slurmd_job_t *job);
int slurm_setaffinity(pid_t pid, size_t size, const cpu_set_t *mask);
int slurm_getaffinity(pid_t pid, size_t size, cpu_set_t *mask);
-#ifdef HAVE_CPUSETS_EXP
-int get_cpuset_mask(cs_cpumask_t *mask, slurmd_job_t *job);
-int get_memset_mask(cs_memmask_t *mem_mask, cs_cpumask_t *cpu_mask, slurmd_job_t *job);
-int make_task_cpuset(slurmd_job_t *job, cs_cpumask_t *cpu_mask, cs_memmask_t *mem_mask);
-#endif
+int slurm_set_cpuset(char *path, pid_t pid, size_t size,
+ const cpu_set_t *mask);
+int slurm_get_cpuset(char *path, pid_t pid, size_t size, cpu_set_t *mask);
/*** from numa.c ***/
#ifdef HAVE_NUMA
diff --git a/src/plugins/task/affinity/cpuset.c b/src/plugins/task/affinity/cpuset.c
index f98ebe77f89..97decac63fd 100644
--- a/src/plugins/task/affinity/cpuset.c
+++ b/src/plugins/task/affinity/cpuset.c
@@ -1,9 +1,10 @@
/*****************************************************************************\
* cpuset.c - Library for interacting with /dev/cpuset file system
*****************************************************************************
+ * Copyright (C) 2007 Bull
* Copyright (C) 2007 The Regents of the University of California.
- * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
- * Written by Morris Jette <jette1@llnl.gov>
+ * Written by Don Albert <Don.Albert@Bull.com> and
+ * Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
*
* This file is part of SLURM, a resource management program.
@@ -35,7 +36,263 @@
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
\*****************************************************************************/
-#if HAVE_CONFIG_H
-# include "config.h"
-#endif
+#include "affinity.h"
+
+int slurm_set_cpuset(char *path, pid_t pid, size_t size,
+ const cpu_set_t *mask)
+{
+ int fd, rc;
+ char file_path[PATH_MAX];
+ char mstr[1 + CPU_SETSIZE / 4];
+
+ if (mkdir(path, 0700) && (errno != EEXIST)) {
+ error("mkdir(%s): %m", path);
+ return -1;
+ }
+
+ snprintf(file_path, sizeof(file_path), "%s/notify_on_release", path);
+ fd = open(file_path, O_CREAT | O_WRONLY);
+ if (fd < 0) {
+ error("open(%s): %m", file_path);
+ return -1;
+ }
+ rc = write(fd, "1", 2); /* not sure if this is right */
+ close(fd);
+
+ snprintf(file_path, sizeof(file_path), "%s/cpus", path);
+ cpuset_to_str(mask, mstr);
+ fd = open(file_path, O_CREAT | O_WRONLY);
+ if (fd < 0) {
+ error("open(%s): %m", file_path);
+ return -1;
+ }
+ rc = write(fd, mstr, strlen(mstr));
+ close(fd);
+ if (rc < 1) {
+ error("write(%s): %m", file_path);
+ return -1;
+ }
+
+ snprintf(file_path, sizeof(file_path), "%s/tasks", path);
+ snprintf(mstr, sizeof(mstr), "%d", pid);
+ fd = open(file_path, O_CREAT);
+ if (fd < 0) {
+ error("open(%s): %m", file_path);
+ return -1;
+ }
+ rc = write(fd, mstr, strlen(mstr));
+ close(fd);
+ if (rc < 1) {
+ error("write(%s): %m", file_path);
+ return -1;
+ }
+
+ return 0;
+}
+
+int slurm_get_cpuset(char *path, pid_t pid, size_t size, cpu_set_t *mask)
+{
+ int fd, rc;
+ char file_path[PATH_MAX];
+ char mstr[1 + CPU_SETSIZE / 4];
+
+ snprintf(file_path, sizeof(file_path), "%s/cpus", path);
+ fd = open(file_path, O_RDONLY);
+ if (fd < 0) {
+ error("open(%s): %m", file_path);
+ return -1;
+ }
+ rc = read(fd, mstr, sizeof(mstr));
+ close(fd);
+ if (rc < 1) {
+ error("read(%s): %m", file_path);
+ return -1;
+ }
+ str_to_cpuset(mask, mstr);
+
+ snprintf(file_path, sizeof(file_path), "%s/tasks", path);
+ fd = open(file_path, O_RDONLY);
+ if (fd < 0) {
+ error("open(%s): %m", file_path);
+ return -1;
+ }
+ rc = read(fd, mstr, sizeof(mstr));
+ close(fd);
+ if (rc < 1) {
+ error("read(%s): %m", file_path);
+ return -1;
+ }
+ /* verify that pid is in mstr */
+
+ return 0;
+}
+
+#if 0
+int get_memset_mask(slurm_memmask_t *mem_mask, slurm_cpumask_t *cpu_mask,
+ slurmd_job_t *job)
+{
+ slurm_cpumask_t cur_mask, tst_mask;
+ int nbr_nodes, nummasks, i, j;
+ char *curstr, *selstr;
+ char mstr[1 + CPU_SETSIZE / 4];
+ int local_id = job->envtp->localid;
+
+ debug2("get_memset_mask bind_type = %d, bind_list = %s\n",
+ job->mem_bind_type, job->mem_bind);
+
+ /* If "not specified" or "None" or "Rank",
+ * do not set a new memory mask in the CPUset */
+ if ((!job->mem_bind_type)
+ || (job->mem_bind_type & (MEM_BIND_NONE | MEM_BIND_RANK)))
+ return false;
+
+ /* For now, make LOCAL and MAP_CPU the same */
+ if (job->mem_bind_type & (MEM_BIND_LOCAL | MEM_BIND_MAP)) {
+ nbr_nodes = cs_nr_nodes();
+ for (i=0; i<nbr_nodes; i++) {
+ cs_get_node_cpus(i, &cur_mask);
+ cs_cpumask_and(&tst_mask, &cur_mask, cpu_mask);
+ if (!cs_cpumask_empty(&tst_mask)) {
+ cs_memmask_add(mem_mask, i);
+ debug2("added node = %d to mem mask %08x \n",
+ i,*mem_mask);
+ }
+ }
+
+ return true;
+ }
+
+ /* allow user to set specific memory masks */
+ if (job->mem_bind_type & MEM_BIND_MASK) {
+ /* find mask entry for this task */
+ nummasks = 0;
+ selstr = NULL;
+
+ curstr = job->mem_bind;
+ while (*curstr) {
+ if (nummasks == local_id) {
+ selstr = curstr;
+ break;
+ }
+ if (*curstr == ',')
+ nummasks++;
+ curstr++;
+ }
+ /* check if we found a mem entry */
+ if (!selstr){
+ error("not enough entries in mask_mem:<list>");
+ return false;
+ }
+ /* extract the selected mask from the list */
+ j = 0;
+ curstr = mstr;
+ while ((*selstr && *selstr != ',') &&
+ (j++ < (CPU_SETSIZE/4))) {
+ *curstr++ = *selstr++;
+ }
+ *curstr = '\0';
+ /* convert mask string into cpu_set_t mask */
+ if (str_to_cpuset( (cpu_set_t *) mem_mask, mstr) < 0) {
+ error("str_to_cpuset %s", mstr);
+ return false;
+ }
+ return true;
+ }
+ return false;
+}
+int make_task_cpuset(slurmd_job_t *job, slurm_cpumask_t *cpu_mask,
+ slurm_memmask_t *mem_mask)
+{
+
+ char path[PATH_MAX];
+ char *current_cs = NULL;
+ int retval = 0;
+ int success = 0;
+
+ info("cpuset - cs_init called");
+ cs_init();
+
+ current_cs = cs_get_current();
+ if (!current_cs)
+ return -ENOMEM;
+
+ int l = snprintf(path, PATH_MAX, "%sslurm%u_%d", current_cs, job->jobid,
+ job->envtp->localid);
+ if (l > PATH_MAX) {
+ retval = -ENAMETOOLONG;
+ goto out;
+ }
+
+ debug("cpuset path = %s",path);
+
+ retval = cs_create(path);
+ if (retval < 0)
+ goto out;
+ retval = cs_set_autoclean(path, CS_AUTOCLEAN);
+ if (retval < 0)
+ goto out;
+
+ cs_lock_libcpuset();
+ retval = cs_set_cpus(path, *cpu_mask);
+ if (retval < 0) {
+ debug3("cpuset - error on cs_set_cpus = %d %s",
+ retval, cs_strerror(retval));
+ }
+ success = 1;
+
+ if (cs_supports_mem()) {
+
+ /* Check for mem_bind options */
+ if (get_memset_mask(mem_mask, cpu_mask, job)) {
+ debug("cpuset - mem_mask = %d (decimal) and %08x (hex)",
+ *mem_mask, *mem_mask);
+ retval = cs_set_mems(path, *mem_mask);
+ if (retval < 0) {
+ debug3("cpuset - error on cs_set_mems = %d %s",
+ retval,cs_strerror(retval));
+ success = 0;
+ goto out_created;
+ }
+ } else {
+ /* Copy parent of new cpuset (i.e current) mems mask */
+ retval = cs_get_mems(current_cs, mem_mask);
+ if (retval < 0) {
+ debug3("cpuset - error on cs_get_mems = %d %s",
+ retval,cs_strerror(retval));
+ success = 0;
+ goto out_created;
+ }
+ retval = cs_set_mems(path, *mem_mask);
+ if (retval < 0) {
+ debug3("cpuset - error on cs_set_mems = %d %s",
+ retval, cs_strerror(retval));
+ success = 0;
+ goto out_created;
+ }
+ }
+ }
+
+ retval = cs_add_task(path, job->envtp->task_pid);
+ if (retval < 0) {
+ debug3("cpuset - error on cs_add_task = %d %s",
+ retval, cs_strerror(retval));
+ }
+
+ out_created:
+ if (!success)
+ cs_destroy(path);
+ cs_unlock_libcpuset();
+
+ out:
+ free(current_cs);
+
+ current_cs = cs_get_current();
+ debug("cpuset - exit make_task_cpuset retval = %d cpuset = %s",
+ retval, current_cs);
+ free(current_cs);
+
+ return retval;
+}
+
+#endif
diff --git a/src/plugins/task/affinity/cpuset.h b/src/plugins/task/affinity/cpuset.h
deleted file mode 100644
index 72826e50dfe..00000000000
--- a/src/plugins/task/affinity/cpuset.h
+++ /dev/null
@@ -1,41 +0,0 @@
-/*****************************************************************************\
- * cpuset.h - Functions for interacting with /dev/cpuset file system
- *****************************************************************************
- * Copyright (C) 2007 The Regents of the University of California.
- * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
- * Written by Morris Jette <jette1@llnl.gov>
- * UCRL-CODE-226842.
- *
- * 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.
- *
- * In addition, as a special exception, the copyright holders give permission
- * to link the code of portions of this program with the OpenSSL library under
- * certain conditions as described in each individual source file, and
- * distribute linked combinations including the two. You must obey the GNU
- * General Public License in all respects for all of the code used other than
- * OpenSSL. If you modify file(s) with this exception, you may extend this
- * exception to your version of the file(s), but you are not obligated to do
- * so. If you do not wish to do so, delete this exception statement from your
- * version. If you delete this exception statement from all source files in
- * the program, then also delete it here.
- *
- * 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.,
- * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-\*****************************************************************************/
-
-#ifndef _HAVE_CPUSET_H
-#define _HAVE_CPUSET_H
-
-#endif
diff --git a/src/plugins/task/affinity/task_affinity.c b/src/plugins/task/affinity/task_affinity.c
index f0bcfda663b..0a27e81ef33 100644
--- a/src/plugins/task/affinity/task_affinity.c
+++ b/src/plugins/task/affinity/task_affinity.c
@@ -48,6 +48,8 @@
#include "affinity.h"
#include "dist_tasks.h"
+#define CPUSET_DIR "/dev/cpuset"
+
/*
* These variables are required by the generic plugin interface. If they
* are not found in the plugin, the plugin loader will ignore it.
@@ -156,10 +158,11 @@ int task_pre_launch ( slurmd_job_t *job )
debug("affinity task_pre_launch: %u.%u, task %d",
job->jobid, job->stepid, job->envtp->procid);
-#ifdef HAVE_CPUSETS_EXP
- if (!conf->use_cpusets) {
- info("Using sched_affinity for tasks");
-#endif
+ if (conf->use_cpusets)
+ info("Using cpuset affinity for tasks");
+ else
+ info("Using sched_affinity for tasks");
+
/*** CPU binding support ***/
if (job->cpu_bind_type) {
cpu_set_t new_mask, cur_mask;
@@ -168,15 +171,34 @@ int task_pre_launch ( slurmd_job_t *job )
int setval = 0;
slurm_getaffinity(mypid, sizeof(cur_mask), &cur_mask);
- if (get_cpuset(&new_mask, job)) {
- if (!(job->cpu_bind_type & CPU_BIND_NONE)) {
+ if (get_cpuset(&new_mask, job)
+ && (!(job->cpu_bind_type & CPU_BIND_NONE))) {
+ if (conf->use_cpusets) {
+ char path[PATH_MAX];
+ if (snprintf(path, PATH_MAX, "%sslurm%u_%d",
+ CPUSET_DIR, job->jobid,
+ job->envtp->localid) >
+ PATH_MAX) {
+ error("cpuset path too long");
+ return SLURM_ERROR;
+ }
+ setval = slurm_set_cpuset(path, mypid,
+ sizeof(new_mask),
+ &new_mask);
+ slurm_get_cpuset(path, mypid,
+ sizeof(cur_mask),
+ &cur_mask);
+ } else {
setval = slurm_setaffinity(mypid,
- sizeof(new_mask), &new_mask);
+ sizeof(new_mask),
+ &new_mask);
slurm_getaffinity(mypid,
- sizeof(cur_mask), &cur_mask);
+ sizeof(cur_mask),
+ &cur_mask);
}
}
- slurm_chkaffinity(setval ? &new_mask : &cur_mask, job, setval);
+ slurm_chkaffinity(setval ? &new_mask : &cur_mask,
+ job, setval);
}
#ifdef HAVE_NUMA
@@ -192,30 +214,6 @@ int task_pre_launch ( slurmd_job_t *job )
}
slurm_chk_memset(&cur_mask, job);
}
-#endif
-#ifdef HAVE_CPUSETS_EXP
- }
- else {
- cs_cpumask_t cpu_mask;
- cs_memmask_t mem_mask;
-
- info("Using cpuset affinity for tasks");
- debug("from job structure - nprocs=%u, ntasks=%u, cpus=%u",
- job->nprocs, job->ntasks, job->cpus);
- debug("from job->envtp - procid=%d, localid=%d, pid=%d",
- job->envtp->procid, job->envtp->localid, job->envtp->task_pid);
-
- cs_cpumask_clear(&cpu_mask);
- cs_memmask_clear(&mem_mask);
- if (get_cpuset_mask(&cpu_mask, job)) {
- debug2("cpu_mask = %d (decimal) and %08x (hex)", cpu_mask, cpu_mask);
- make_task_cpuset(job, &cpu_mask, &mem_mask);
- }
- slurm_chkaffinity(&cpu_mask, job, 0);
-#ifdef HAVE_NUMA
- slurm_chk_memset(&mem_mask, job);
-#endif
- }
#endif
return SLURM_SUCCESS;
}
--
GitLab