diff --git a/NEWS b/NEWS
index 41051dc639daeb28f8a9ff62121e46960d34eab0..c3af6a16dcc5e32ccb929dd188831e48d4c85a18 100644
--- a/NEWS
+++ b/NEWS
@@ -238,6 +238,8 @@ documents those changes that are of interest to users and administrators.
-- power/cray - Disable power cap get and set operations on DOWN nodes.
-- Jobs preempted with PreemptMode=REQUEUE were incorrectly recorded as
REQUEUED in the accounting.
+ -- PMIX - Use volatile specifier to avoid flag caching.
+ -- PMIX - Make it possible to use %n or %h in a spool dir.
* Changes in Slurm 16.05.9
==========================
diff --git a/src/common/slurm_protocol_api.c b/src/common/slurm_protocol_api.c
index d91d9d92cbaba68b25b3e65eb10c04d4e89619b8..5abda4eb0d4a2efc06b46b06f1291ce8f4a1e860 100644
--- a/src/common/slurm_protocol_api.c
+++ b/src/common/slurm_protocol_api.c
@@ -2770,7 +2770,7 @@ char *slurm_get_job_container_plugin(void)
/* slurm_get_slurmd_spooldir
* RET slurmd_spooldir name, must be xfreed by caller */
-char *slurm_get_slurmd_spooldir(void)
+char *slurm_get_slurmd_spooldir(char *node_name)
{
char *slurmd_spooldir = NULL;
slurm_ctl_conf_t *conf;
@@ -2778,7 +2778,11 @@ char *slurm_get_slurmd_spooldir(void)
if (slurmdbd_conf) {
} else {
conf = slurm_conf_lock();
- slurmd_spooldir = xstrdup(conf->slurmd_spooldir);
+ if (!node_name)
+ slurmd_spooldir = xstrdup(conf->slurmd_spooldir);
+ else
+ slurmd_spooldir = slurm_conf_expand_slurmd_path(
+ conf->slurmd_spooldir, node_name);
slurm_conf_unlock();
}
return slurmd_spooldir;
diff --git a/src/common/slurm_protocol_api.h b/src/common/slurm_protocol_api.h
index fa69a20293a775ad29d1da0ef7661b02c27c8d0c..2f4c8c1cf03df1171099547e40eed0bab7f89a11 100644
--- a/src/common/slurm_protocol_api.h
+++ b/src/common/slurm_protocol_api.h
@@ -893,7 +893,7 @@ char *slurm_get_job_container_plugin(void);
/* slurm_get_slurmd_spooldir
* RET slurmd_spooldir name, must be xfreed by caller */
-char *slurm_get_slurmd_spooldir(void);
+char *slurm_get_slurmd_spooldir(char *node_name);
/* slurm_get_layouts
* RET comma seperated list of layouts in a string, must be xfreed by caller
diff --git a/src/plugins/mpi/pmi2/setup.c b/src/plugins/mpi/pmi2/setup.c
index 242769189be006ff2deffb728ee7326d604ad70a..7298bbd99d82a74b16bed9dbd87ce71e12efa3f3 100644
--- a/src/plugins/mpi/pmi2/setup.c
+++ b/src/plugins/mpi/pmi2/setup.c
@@ -272,7 +272,10 @@ _setup_stepd_sockets(const stepd_step_rec_t *job, char ***env)
}
sa.sun_family = PF_UNIX;
- spool = slurm_get_slurmd_spooldir();
+ /* FIXME: We need to handle %n and %h in the spool dir, we have
+ * job->node_name, but the other place we call get_slurmd_spooldir
+ * we don't so we aren't using it here */
+ spool = slurm_get_slurmd_spooldir(NULL);
snprintf(sa.sun_path, sizeof(sa.sun_path), PMI2_SOCK_ADDR_FMT,
spool, job->jobid, job->stepid);
unlink(sa.sun_path); /* remove possible old socket */
@@ -628,7 +631,9 @@ _setup_srun_tree_info(const mpi_plugin_client_info_t *job)
} else
tree_info.srun_addr = NULL;
- spool = slurm_get_slurmd_spooldir();
+ /* FIXME: We need to handle %n and %h in the spool dir, but don't have
+ * the node name here */
+ spool = slurm_get_slurmd_spooldir(NULL);
snprintf(tree_sock_addr, 128, PMI2_SOCK_ADDR_FMT,
spool, job->jobid, job->stepid);
xfree(spool);
diff --git a/src/plugins/mpi/pmix/pmixp_agent.c b/src/plugins/mpi/pmix/pmixp_agent.c
index 333e7ac8394bf5ea5300184499690d4176201e2e..09f82c678f8f706bb77d25a518c36b59a66d56a3 100644
--- a/src/plugins/mpi/pmix/pmixp_agent.c
+++ b/src/plugins/mpi/pmix/pmixp_agent.c
@@ -2,7 +2,7 @@
** pmix_agent.c - PMIx agent thread
*****************************************************************************
* Copyright (C) 2014-2015 Artem Polyakov. All rights reserved.
- * Copyright (C) 2015 Mellanox Technologies. All rights reserved.
+ * Copyright (C) 2015-2017 Mellanox Technologies. All rights reserved.
* Written by Artem Y. Polyakov <artpol84@gmail.com, artemp@mellanox.com>.
*
* This file is part of SLURM, a resource management program.
@@ -51,8 +51,8 @@
#define MAX_RETRIES 5
-static int _agent_is_running = 0;
-static int _timer_is_running = 0;
+static volatile int _agent_is_running = 0;
+static volatile int _timer_is_running = 0;
static eio_handle_t *_io_handle = NULL;
static int _agent_spawned = 0, _timer_spawned = 0;
diff --git a/src/plugins/mpi/pmix/pmixp_info.c b/src/plugins/mpi/pmix/pmixp_info.c
index edc0e47204fccbb2600d2051e22f5ea346dd4c1a..ecd17d97fd49c6def5bf443988243c0f053ea1d9 100644
--- a/src/plugins/mpi/pmix/pmixp_info.c
+++ b/src/plugins/mpi/pmix/pmixp_info.c
@@ -286,26 +286,27 @@ err_exit:
static int _env_set(char ***env)
{
char *p = NULL;
- char *spool = slurm_get_slurmd_spooldir();
+
+ xassert(_pmixp_job_info.hostname);
+
+ _pmixp_job_info.lib_tmpdir = slurm_get_slurmd_spooldir(
+ _pmixp_job_info.hostname);
/* ----------- Temp directories settings ------------- */
- _pmixp_job_info.lib_tmpdir = xstrdup_printf("%s/pmix.%d.%d/", spool,
- pmixp_info_jobid(), pmixp_info_stepid());
- xfree(spool);
+ xstrfmtcat(_pmixp_job_info.lib_tmpdir, "/pmix.%d.%d/",
+ pmixp_info_jobid(), pmixp_info_stepid());
/* save client temp directory if requested
* TODO: We want to get TmpFS value as well if exists.
* Need to sync with SLURM developers.
*/
p = getenvp(*env, PMIXP_TMPDIR_CLI);
- if (NULL != p) {
+
+ if (p)
_pmixp_job_info.cli_tmpdir_base = xstrdup(p);
- } else {
- p = slurm_get_tmp_fs();
- if (NULL != p) {
- _pmixp_job_info.cli_tmpdir_base = p;
- }
- }
+ else
+ _pmixp_job_info.cli_tmpdir_base = slurm_get_tmp_fs();
+
_pmixp_job_info.cli_tmpdir =
xstrdup_printf("%s/spmix_appdir_%d.%d",
_pmixp_job_info.cli_tmpdir_base,
diff --git a/src/plugins/mpi/pmix/pmixp_info.h b/src/plugins/mpi/pmix/pmixp_info.h
index e8e159567e88b99fb23f32219c5ceef2c043ba9c..69ea4d0a8af7b2b5649034cd1790ff63ce9490fc 100644
--- a/src/plugins/mpi/pmix/pmixp_info.h
+++ b/src/plugins/mpi/pmix/pmixp_info.h
@@ -272,12 +272,11 @@ static inline char *pmixp_info_job_host(int nodeid)
/* namespaces list operations */
static inline char *pmixp_info_nspace_usock(const char *nspace)
{
- char *spool, *usock = NULL;
+ char *spool;
debug("mpi/pmix: setup sockets");
- spool = slurm_get_slurmd_spooldir();
- xstrfmtcat(usock, "%s/stepd.%s", spool, nspace);
- xfree(spool);
- return usock;
+ spool = slurm_get_slurmd_spooldir(_pmixp_job_info.hostname);
+ xstrfmtcat(spool, "/stepd.%s", nspace);
+ return spool;
}
#endif /* PMIXP_INFO_H */