From ba96978eb3d168d2cc4bec2e2a4e11040a3b1ed6 Mon Sep 17 00:00:00 2001
From: Danny Auble <da@llnl.gov>
Date: Fri, 29 Feb 2008 17:31:55 +0000
Subject: [PATCH] svn merge -r13414:13423
https://eris.llnl.gov/svn/slurm/branches/slurm-1.2
---
NEWS | 2 ++
src/plugins/mpi/mvapich/mvapich.c | 13 +++++++++++--
2 files changed, 13 insertions(+), 2 deletions(-)
diff --git a/NEWS b/NEWS
index 177f780d36d..a1e8585776e 100644
--- a/NEWS
+++ b/NEWS
@@ -182,6 +182,8 @@ documents those changes that are of interest to users and admins.
-- Bug fix for setting exit code in accounting for batch script.
-- Add salloc option, --no-shell (for LSF).
-- Added new options for sacct output
+ -- mvapich: Ensure MPIRUN_ID is unique for all job steps within a job.
+ (Fixes crashes when running multiple job steps within a job on one node)
* Changes in SLURM 1.2.24
=========================
diff --git a/src/plugins/mpi/mvapich/mvapich.c b/src/plugins/mpi/mvapich/mvapich.c
index 38f1f5d8eca..b2e5b868484 100644
--- a/src/plugins/mpi/mvapich/mvapich.c
+++ b/src/plugins/mpi/mvapich/mvapich.c
@@ -1366,6 +1366,15 @@ static void mvapich_state_destroy(mvapich_state_t *st)
xfree(st);
}
+/*
+ * Create a unique MPIRUN_ID for jobid/stepid pairs.
+ * Combine the least significant bits of the jobid and stepid
+ */
+int mpirun_id_create(const mpi_plugin_client_info_t *job)
+{
+ return (int) ((job->jobid << 16) | (job->stepid & 0xffff));
+}
+
extern mvapich_state_t *mvapich_thr_create(const mpi_plugin_client_info_t *job,
char ***env)
{
@@ -1405,12 +1414,12 @@ extern mvapich_state_t *mvapich_thr_create(const mpi_plugin_client_info_t *job,
*/
env_array_overwrite_fmt(env, "MPIRUN_PORT", "%hu", port);
env_array_overwrite_fmt(env, "MPIRUN_NPROCS", "%d", st->nprocs);
- env_array_overwrite_fmt(env, "MPIRUN_ID", "%d", st->job->jobid);
+ env_array_overwrite_fmt(env, "MPIRUN_ID", "%d", mpirun_id_create(job));
if (st->connect_once) {
env_array_overwrite_fmt(env, "MPIRUN_CONNECT_ONCE", "1");
}
- verbose ("mvapich-0.9.[45] master listening on port %d", port);
+ verbose ("mvapich-0.9.[45] master listening on port %hu", port);
return st;
}
--
GitLab