From 3fe018ae6d127506a7aa25c442bad00c9a2a2c52 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Sat, 5 Jun 2004 01:05:39 +0000
Subject: [PATCH] Added srun option --debugger-test to confirm that slurm's
debugger infrastructure is operational.
---
NEWS | 2 ++
src/srun/msg.c | 30 +++++++++++++++++++++++++++---
src/srun/opt.c | 13 ++++++++++++-
src/srun/opt.h | 1 +
4 files changed, 42 insertions(+), 4 deletions(-)
diff --git a/NEWS b/NEWS
index 83d91d43de6..32bede9be7b 100644
--- a/NEWS
+++ b/NEWS
@@ -8,6 +8,8 @@ documents those changes that are of interest to users and admins.
-- Fix sinfo -R, --list-reasons to list all relevant node states.
-- Fix slurmctld to honor srun's node configuration specifications
with FastSchedule==0 configuration.
+ -- Added srun option --debugger-test to confirm that slurm's debugger
+ infrastructure is operational.
-- Removed debugging hacks for srun.wrapper.c. Temporarily use
RPM's debugedit utility if available for similar effect.
diff --git a/src/srun/msg.c b/src/srun/msg.c
index 04e120d4b56..cbb9d7d4c94 100644
--- a/src/srun/msg.c
+++ b/src/srun/msg.c
@@ -74,6 +74,7 @@ static slurm_fd slurmctld_fd = (slurm_fd) NULL;
*/
static void _accept_msg_connection(job_t *job, int fdnum);
static void _confirm_launch_complete(job_t *job);
+static void _dump_proctable(job_t *job);
static void _exit_handler(job_t *job, slurm_msg_t *exit_msg);
static void _handle_msg(job_t *job, slurm_msg_t *msg);
static inline bool _job_msg_done(job_t *job);
@@ -131,11 +132,34 @@ _build_proctable(job_t *job, char *host, int nodeid, int ntasks, uint32_t *pid)
if (tasks_recorded == opt.nprocs) {
MPIR_debug_state = MPIR_DEBUG_SPAWNED;
- MPIR_Breakpoint();
+ MPIR_Breakpoint();
+ if (opt.debugger_test)
+ _dump_proctable(job);
}
}
-
+static void _dump_proctable(job_t *job)
+{
+ int node_inx, task_inx, taskid, max_task;
+ MPIR_PROCDESC *tv;
+
+ if (opt.overcommit)
+ max_task = opt.nprocs;
+
+ for (node_inx=0; node_inx<job->nhosts; node_inx++) {
+ if (!opt.overcommit)
+ max_task = job->cpus[node_inx];
+ for (task_inx=0; ; task_inx++) {
+ taskid = job->tids[node_inx][task_inx];
+ if ((task_inx > 0) && (taskid == 0))
+ break;
+ tv = &MPIR_proctable[taskid];
+ info("task:%d, host:%s, pid:%d",
+ taskid, tv->host_name, tv->pid);
+ }
+ }
+}
+
void debugger_launch_failure(void)
{
if (opt.parallel_debug) {
@@ -148,7 +172,7 @@ void MPIR_Breakpoint(void)
{
debug("In MPIR_Breakpoint");
/* This just notifies parallel
- * debugger that some event of interest occured */
+ * debugger that some event of interest occurred */
}
/*
diff --git a/src/srun/opt.c b/src/srun/opt.c
index d3fb2a83d55..6e3f2b80231 100644
--- a/src/srun/opt.c
+++ b/src/srun/opt.c
@@ -103,7 +103,7 @@
#define LONG_OPT_MPI 0x10c
#define LONG_OPT_CORE 0x10e
#define LONG_OPT_NOSHELL 0x10f
-
+#define LONG_OPT_DEBUG_TS 0x110
/*---- forward declarations of static functions ----*/
typedef struct env_vars env_vars_t;
@@ -651,6 +651,7 @@ static void _opt_args(int argc, char **argv)
{"max-exit-timeout", required_argument, 0, LONG_OPT_XTO},
{"uid", required_argument, 0, LONG_OPT_UID},
{"gid", required_argument, 0, LONG_OPT_GID},
+ {"debugger-test", no_argument, 0, LONG_OPT_DEBUG_TS},
{"help", no_argument, 0, LONG_OPT_HELP},
{"usage", no_argument, 0, LONG_OPT_USAGE},
{NULL, 0, 0, 0}
@@ -882,6 +883,16 @@ static void _opt_args(int argc, char **argv)
if (opt.egid == (gid_t) -1)
fatal ("--gid=\"%s\" invalid", optarg);
break;
+ case LONG_OPT_DEBUG_TS:
+ opt.debugger_test = true;
+ /* make other parameters look like debugger
+ * is really attached */
+ opt.parallel_debug = true;
+ MPIR_being_debugged = 1;
+ opt.max_launch_time = 120;
+ opt.max_threads = 1;
+ opt.msg_timeout = 15;
+ break;
case LONG_OPT_HELP:
_help();
exit(0);
diff --git a/src/srun/opt.h b/src/srun/opt.h
index 54b0f3b8d19..281429f9246 100644
--- a/src/srun/opt.h
+++ b/src/srun/opt.h
@@ -135,6 +135,7 @@ typedef struct srun_options {
bool quit_on_intr; /* --quit-on-interrupt, -q */
int quiet;
bool parallel_debug; /* srun controlled by debugger */
+ bool debugger_test; /* --debugger-test */
/* constraint options */
int mincpus; /* --mincpus=n */
--
GitLab