diff --git a/NEWS b/NEWS
index 78ca5dc4d6107f76bca7e546a7d55da77cd2ab62..dd70abc76fe130d7d4da2ef0da1c04013d99a5d3 100644
--- a/NEWS
+++ b/NEWS
@@ -58,6 +58,8 @@ documents those changes that are of interest to users and admins.
-- Permit node suspend/resume logic to be enabled through "scontrol reconfig"
given appropriate changes to slurm configuration file.
-- Check for return codes on functions with warn_unused_result set
+ -- Fix memory leak in getting step information
+ -- Better logging for when job's request bad output locations.
* Changes in SLURM 2.0.3
========================
diff --git a/contribs/perlapi/libslurm-perl/alloc.c b/contribs/perlapi/libslurm-perl/alloc.c
index e6194392ab128f82fd6e2f9265b1a0fcc716b958..417e5da2673ed67ebf6ad934bca480a7f6da95e0 100644
--- a/contribs/perlapi/libslurm-perl/alloc.c
+++ b/contribs/perlapi/libslurm-perl/alloc.c
@@ -140,7 +140,7 @@ hv_to_job_desc_msg(HV* hv, job_desc_msg_t* job_desc_msg)
avp = (AV*)SvRV(*svp);
for(i = 0; i < SYSTEM_DIMENSIONS; i ++) {
if(! (svp = av_fetch(avp, i, FALSE))) {
- Perl_warn(aTHX_ "geometry of dimension %s missing in job descriptor", i);
+ Perl_warn(aTHX_ "geometry of dimension %d missing in job descriptor", i);
free_job_desc_msg_memory(job_desc_msg);
return -1;
}
diff --git a/src/common/slurm_protocol_defs.c b/src/common/slurm_protocol_defs.c
index 905057b0d9921cc43e5854c11e5881f80489b5d8..5c8871643c68522aa85d0ce04a3bd511d669c8d4 100644
--- a/src/common/slurm_protocol_defs.c
+++ b/src/common/slurm_protocol_defs.c
@@ -1501,7 +1501,10 @@ static void _slurm_free_job_step_info_members (job_step_info_t * msg)
{
if (msg != NULL) {
xfree(msg->partition);
+ xfree(msg->resv_ports);
xfree(msg->nodes);
+ xfree(msg->name);
+ xfree(msg->network);
xfree(msg->ckpt_dir);
}
}
diff --git a/src/common/slurm_protocol_pack.c b/src/common/slurm_protocol_pack.c
index 8dcea5962b8a699c9ffec5c17f64b362ea2df168..851eaa429cbb7640046e47680a2bb17fa9bda555 100644
--- a/src/common/slurm_protocol_pack.c
+++ b/src/common/slurm_protocol_pack.c
@@ -2694,6 +2694,7 @@ _unpack_job_step_info_members(job_step_info_t * step, Buf buffer)
unpack_error:
xfree(step->partition);
+ xfree(step->resv_ports);
xfree(step->nodes);
xfree(step->name);
xfree(step->network);
diff --git a/src/slurmd/slurmstepd/io.c b/src/slurmd/slurmstepd/io.c
index 680faec492ed51624536dad7123befbac0040278..a238ab3596a589dc7ad267be7b2dbb4c6f95bedd 100644
--- a/src/slurmd/slurmstepd/io.c
+++ b/src/slurmd/slurmstepd/io.c
@@ -945,7 +945,7 @@ _init_task_stdio_fds(slurmd_task_info_t *task, slurmd_job_t *job)
/* open file on task's stdin */
debug5(" stdin file name = %s", task->ifname);
if ((task->stdin_fd = open(task->ifname, O_RDONLY)) == -1) {
- error("Could not open stdin file: %m");
+ error("Could not open stdin file %s: %m", task->ifname);
return SLURM_ERROR;
}
fd_set_close_on_exec(task->stdin_fd);
@@ -999,6 +999,8 @@ _init_task_stdio_fds(slurmd_task_info_t *task, slurmd_job_t *job)
debug5(" stdout file name = %s", task->ofname);
task->stdout_fd = open(task->ofname, file_flags, 0666);
if (task->stdout_fd == -1) {
+ error("Could not open stdout file %s: %m",
+ task->ofname);
return SLURM_ERROR;
}
fd_set_close_on_exec(task->stdout_fd);
@@ -1055,6 +1057,8 @@ _init_task_stdio_fds(slurmd_task_info_t *task, slurmd_job_t *job)
debug5(" stderr file name = %s", task->efname);
task->stderr_fd = open(task->efname, file_flags, 0666);
if (task->stderr_fd == -1) {
+ error("Could not open stderr file %s: %m",
+ task->efname);
return SLURM_ERROR;
}
fd_set_close_on_exec(task->stderr_fd);
@@ -1278,7 +1282,7 @@ io_close_all(slurmd_job_t *job)
* and log facility may still try to write to stderr.
*/
if ((devnull = open("/dev/null", O_RDWR)) < 0) {
- error("Unable to open /dev/null: %m");
+ error("Could not open /dev/null: %m");
} else {
if (dup2(devnull, STDERR_FILENO) < 0)
error("Unable to dup /dev/null onto stderr\n");
diff --git a/src/slurmd/slurmstepd/mgr.c b/src/slurmd/slurmstepd/mgr.c
index b29735a4edf57f7cb95a3f2529db387ed346b92d..7684dc2693070ad372646d3b7b48f10567bdb670 100644
--- a/src/slurmd/slurmstepd/mgr.c
+++ b/src/slurmd/slurmstepd/mgr.c
@@ -174,7 +174,7 @@ static int _drop_privileges(slurmd_job_t *job, bool do_setuid,
struct priv_state *state);
static int _reclaim_privileges(struct priv_state *state);
static void _send_launch_resp(slurmd_job_t *job, int rc);
-static void _slurmd_job_log_init(slurmd_job_t *job);
+static int _slurmd_job_log_init(slurmd_job_t *job);
static void _wait_for_io(slurmd_job_t *job);
static int _send_exit_msg(slurmd_job_t *job, uint32_t *tid, int n,
int status);
@@ -414,8 +414,12 @@ _setup_normal_io(slurmd_job_t *job)
file_flags, job, job->labelio,
job->task[ii]->id,
same ? job->task[ii]->id : -2);
- if (rc != SLURM_SUCCESS)
+ if (rc != SLURM_SUCCESS) {
+ error("Could not open output "
+ "file %s: %m",
+ job->task[ii]->ofname);
return ESLURMD_IO_ERROR;
+ }
}
srun_stdout_tasks = -2;
if (same)
@@ -426,9 +430,12 @@ _setup_normal_io(slurmd_job_t *job)
job->task[0]->ofname,
file_flags, job, job->labelio,
-1, same ? -1 : -2);
- if (rc != SLURM_SUCCESS)
+ if (rc != SLURM_SUCCESS) {
+ error("Could not open output "
+ "file %s: %m",
+ job->task[0]->ofname);
return ESLURMD_IO_ERROR;
-
+ }
srun_stdout_tasks = -2;
if (same)
srun_stderr_tasks = -2;
@@ -443,8 +450,14 @@ _setup_normal_io(slurmd_job_t *job)
file_flags, job,
job->labelio,
-2, job->task[ii]->id);
- if (rc != SLURM_SUCCESS)
+ if (rc != SLURM_SUCCESS) {
+ error("Could not "
+ "open error "
+ "file %s: %m",
+ job->task[ii]->
+ efname);
return ESLURMD_IO_ERROR;
+ }
}
srun_stderr_tasks = -2;
} else if (errpattern == SLURMD_ALL_SAME) {
@@ -453,14 +466,16 @@ _setup_normal_io(slurmd_job_t *job)
job->task[0]->efname,
file_flags, job, job->labelio,
-2, -1);
- if (rc != SLURM_SUCCESS)
+ if (rc != SLURM_SUCCESS) {
+ error("Could not open error "
+ "file %s: %m",
+ job->task[0]->efname);
return ESLURMD_IO_ERROR;
-
+ }
srun_stderr_tasks = -2;
}
}
}
-
rc = io_initial_client_connect(srun, job, srun_stdout_tasks,
srun_stderr_tasks);
if (rc < 0)
@@ -476,7 +491,7 @@ _setup_normal_io(slurmd_job_t *job)
if (io_thread_start(job) < 0)
return ESLURMD_IO_ERROR;
}
-
+
debug2("Leaving _setup_normal_io");
return SLURM_SUCCESS;
}
@@ -811,6 +826,13 @@ job_manager(slurmd_job_t *job)
goto fail1;
}
+#ifndef NDEBUG
+# ifdef PR_SET_DUMPABLE
+ if (prctl(PR_SET_DUMPABLE, 1) < 0)
+ debug ("Unable to set dumpable to 1");
+# endif /* PR_SET_DUMPABLE */
+#endif /* !NDEBUG */
+
set_umask(job); /* set umask for stdout/err files */
if (job->user_managed_io)
rc = _setup_user_managed_io(job);
@@ -819,15 +841,9 @@ job_manager(slurmd_job_t *job)
/*
* Initialize log facility to copy errors back to srun
*/
- _slurmd_job_log_init(job);
+ if(!rc)
+ rc = _slurmd_job_log_init(job);
-#ifndef NDEBUG
-# ifdef PR_SET_DUMPABLE
- if (prctl(PR_SET_DUMPABLE, 1) < 0)
- debug ("Unable to set dumpable to 1");
-# endif /* PR_SET_DUMPABLE */
-#endif /* !NDEBUG */
-
if (rc) {
error("IO setup failed: %m");
rc = SLURM_SUCCESS; /* drains node otherwise */
@@ -1734,7 +1750,7 @@ _reclaim_privileges(struct priv_state *ps)
}
-static void
+static int
_slurmd_job_log_init(slurmd_job_t *job)
{
char argv0[64];
@@ -1761,15 +1777,16 @@ _slurmd_job_log_init(slurmd_job_t *job)
log_alter(conf->log_opts, 0, NULL);
log_set_argv0(argv0);
-
+
/* Connect slurmd stderr to job's stderr */
if (!job->user_managed_io && job->task != NULL) {
if (dup2(job->task[0]->stderr_fd, STDERR_FILENO) < 0) {
error("job_log_init: dup2(stderr): %m");
- return;
+ return ESLURMD_IO_ERROR;
}
}
verbose("debug level = %d", conf->log_opts.stderr_level);
+ return SLURM_SUCCESS;
}
diff --git a/testsuite/expect/test1.29 b/testsuite/expect/test1.29
index 828d2b79456a7cc88b46a7f4f6f37f4821197f3e..76d8a9114d3731efd8627e0941c5c87b06b1f877 100755
--- a/testsuite/expect/test1.29
+++ b/testsuite/expect/test1.29
@@ -220,8 +220,8 @@ expect {
if {$matches != 5} {
send_user "\nFAILURE: User limits not propagated got $matches matches\n"
- send_user" Check $file_err for errors\n"
- send_user" A long running slurmd could cause a file size limit error\n"
+ send_user "Check $file_err for errors\n"
+ send_user "A long running slurmd could cause a file size limit error\n"
set exit_code 1
}