diff --git a/NEWS b/NEWS
index 545e9a65d895f44eeab47d7ae8231fc4cbd541d5..8578390dea5dcfa4a83dae6ea086ac85dbcca3bc 100644
--- a/NEWS
+++ b/NEWS
@@ -1,10 +1,12 @@
This file describes changes in recent versions of SLURM. It primarily
documents those changes that are of interest to users and admins.
-* Changes in SLURM 1.1.0-pre6
+* Changes in SLURM 1.1.0-pre7
=============================
-- Accounting works for aix systems, use jobacct/aix
-- Support large (over 2GB) files on 32-bit linux systems
+ -- changed all writes to safe_write in srun
+ -- added $float to globals.example in the testsuite
* Changes in SLURM 1.1.0-pre6
=============================
diff --git a/src/sacct/options.c b/src/sacct/options.c
index d95830b69c3363d8971c38ab2f4ea4a866c838a0..83e315bb130b26d05fa4b1237862797ce515ec41 100644
--- a/src/sacct/options.c
+++ b/src/sacct/options.c
@@ -472,12 +472,12 @@ int get_data(void)
show_full = 1;
list_iterator_destroy(itr);
goto foundjob;
- } else if (rec_type == JOB_STEP
+ } else if (rec_type != JOB_STEP
|| !strcmp(f[F_JOBSTEP],
selected_step->step)) {
list_iterator_destroy(itr);
goto foundjob;
- }
+ }
}
list_iterator_destroy(itr);
continue; /* no match */
@@ -914,7 +914,6 @@ void parse_command_line(int argc, char **argv)
selected_step = xmalloc(sizeof(selected_step_t));
list_append(selected_steps, selected_step);
- selected_step->job = xstrdup(start);
dot = strstr(start, ".");
if (dot == NULL) {
debug2("No jobstep requested");
@@ -923,6 +922,7 @@ void parse_command_line(int argc, char **argv)
*dot++ = 0;
selected_step->step = xstrdup(dot);
}
+ selected_step->job = xstrdup(start);
start = end + 1;
}
if (params.opt_verbose) {
@@ -1715,7 +1715,7 @@ void do_list(void)
print_fields(JOB, job);
}
- if (do_jobsteps && job->track_steps) {
+ if (do_jobsteps && (job->track_steps || !job->show_full)) {
itr_step = list_iterator_create(job->steps);
while((step = list_next(itr_step))) {
if (step->status == JOB_RUNNING
diff --git a/src/srun/io.c b/src/srun/io.c
index 3e5b153144fcac263d4eaf66993f615a0a0ee994..5bf4e663c5f030b9a5188fda5c6aaa98698c6ef3 100644
--- a/src/srun/io.c
+++ b/src/srun/io.c
@@ -380,7 +380,8 @@ _server_write(eio_obj_t *obj, List objs)
debug3("_server_write: nothing in the queue");
return SLURM_SUCCESS;
}
- debug3(" dequeue successful, s->out_msg->length = %d", s->out_msg->length);
+ debug3(" dequeue successful, s->out_msg->length = %d",
+ s->out_msg->length);
s->out_remaining = s->out_msg->length;
}
@@ -420,7 +421,6 @@ again:
s->out_msg = NULL;
return SLURM_SUCCESS;
-
}
/**********************************************************************
@@ -451,7 +451,10 @@ static void _write_label(int fd, int taskid)
snprintf(buf, 16, "%0*d: ", fmt_width, taskid);
/* FIXME - Need to handle return code */
- write(fd, buf, fmt_width+2);
+ safe_write(fd, buf, fmt_width+2);
+ return;
+rwfail:
+ error("_write_label: write from io process failed");
}
static void _write_newline(int fd)
@@ -468,6 +471,7 @@ again:
}
/* FIXME handle error */
}
+ return;
}
/*
diff --git a/src/srun/launch.c b/src/srun/launch.c
index a7970702c2d975f49fd830cf280aab59666d4ad6..1e809b2500cf8c862b4121fe0eef16d3957e2e6d 100644
--- a/src/srun/launch.c
+++ b/src/srun/launch.c
@@ -431,11 +431,11 @@ _update_failed_node(srun_job_t *j, int id)
j->host_state[id] = SRUN_HOST_UNREACHABLE;
if(message_thread) {
- write(j->forked_msg->par_msg->msg_pipe[1],
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
&pipe_enum,sizeof(int));
- write(j->forked_msg->par_msg->msg_pipe[1],
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
&id,sizeof(int));
- write(j->forked_msg->par_msg->msg_pipe[1],
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
&j->host_state[id],sizeof(int));
}
}
@@ -445,18 +445,23 @@ _update_failed_node(srun_job_t *j, int id)
j->task_state[j->step_layout->tids[id][i]] = SRUN_TASK_FAILED;
if(message_thread) {
- write(j->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum,sizeof(int));
- write(j->forked_msg->par_msg->msg_pipe[1],
- &j->step_layout->tids[id][i],sizeof(int));
- write(j->forked_msg->par_msg->msg_pipe[1],
- &j->task_state[j->step_layout->tids[id][i]],
- sizeof(int));
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
+ &j->step_layout->tids[id][i], sizeof(int));
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
+ &j->task_state[j->step_layout->tids[id][i]],
+ sizeof(int));
}
}
pthread_mutex_unlock(&j->task_mutex);
/* update_failed_tasks(j, id); */
+ return;
+rwfail:
+ pthread_mutex_unlock(&j->task_mutex);
+ error("_update_failed_node: "
+ "write from srun message-handler process failed");
}
static void
@@ -468,15 +473,20 @@ _update_contacted_node(srun_job_t *j, int id)
if (j->host_state[id] == SRUN_HOST_INIT) {
j->host_state[id] = SRUN_HOST_CONTACTED;
if(message_thread) {
- write(j->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum,sizeof(int));
- write(j->forked_msg->par_msg->msg_pipe[1],
- &id,sizeof(int));
- write(j->forked_msg->par_msg->msg_pipe[1],
- &j->host_state[id],sizeof(int));
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
+ &id, sizeof(int));
+ safe_write(j->forked_msg->par_msg->msg_pipe[1],
+ &j->host_state[id], sizeof(int));
}
}
pthread_mutex_unlock(&j->task_mutex);
+ return;
+rwfail:
+ pthread_mutex_unlock(&j->task_mutex);
+ error("_update_contacted_node: "
+ "write from srun message-handler process failed");
}
diff --git a/src/srun/msg.c b/src/srun/msg.c
index c79c6b9da5bc94b67bec839917f576c0721ba8e6..e422addbc060e0c901b72962ba2ff5f42d729a29 100644
--- a/src/srun/msg.c
+++ b/src/srun/msg.c
@@ -140,8 +140,7 @@ static void _update_mpir_proctable(int fd, srun_job_t *job,
return;
rwfail:
- error("write to srun main process failed");
- return;
+ error("_update_mpir_proctable: write to srun main process failed");
}
static void _handle_update_mpir_proctable(int fd, srun_job_t *job)
@@ -205,11 +204,12 @@ static void _handle_update_mpir_proctable(int fd, srun_job_t *job)
return;
rwfail:
- error("read from srun message-handler process failed");
- return;
+ error("_handle_update_mpir_proctable: "
+ "read from srun message-handler process failed");
}
-static void _update_step_layout(int fd, slurm_step_layout_t *layout, int nodeid)
+static void _update_step_layout(int fd, slurm_step_layout_t *layout,
+ int nodeid)
{
int msg_type = PIPE_UPDATE_STEP_LAYOUT;
int dummy = 0xdeadbeef;
@@ -234,8 +234,7 @@ static void _update_step_layout(int fd, slurm_step_layout_t *layout, int nodeid)
return;
rwfail:
- error("write to srun main process failed");
- return;
+ error("_update_step_layout: write to srun main process failed");
}
static void _handle_update_step_layout(int fd, slurm_step_layout_t *layout)
@@ -270,8 +269,8 @@ static void _handle_update_step_layout(int fd, slurm_step_layout_t *layout)
return;
rwfail:
- error("read from srun message-handler process failed");
- return;
+ error("_handle_update_step_layout: "
+ "read from srun message-handler process failed");
}
static void _dump_proctable(srun_job_t *job)
@@ -302,15 +301,20 @@ void debugger_launch_failure(srun_job_t *job)
if (opt.parallel_debug) {
if(message_thread && job) {
i = MPIR_DEBUG_ABORTING;
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &i,sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &i, sizeof(int));
} else if(!job) {
error("Hey I don't have a job to write to on the "
"failure of the debugger launch.");
}
}
+ return;
+rwfail:
+ error("debugger_launch_failure: "
+ "write from srun message-handler process failed");
+
}
/*
@@ -328,7 +332,7 @@ static void _timeout_handler(time_t timeout)
if (timeout != last_timeout) {
last_timeout = timeout;
verbose("job time limit to be reached at %s",
- ctime(&timeout));
+ ctime(&timeout));
}
}
@@ -374,12 +378,12 @@ _process_launch_resp(srun_job_t *job, launch_tasks_response_msg_t *msg)
pthread_mutex_unlock(&job->task_mutex);
if(message_thread) {
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum, sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &msg->srun_node_id, sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->host_state[msg->srun_node_id], sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &msg->srun_node_id, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->host_state[msg->srun_node_id], sizeof(int));
}
_update_mpir_proctable(job->forked_msg->par_msg->msg_pipe[1], job,
@@ -387,6 +391,10 @@ _process_launch_resp(srun_job_t *job, launch_tasks_response_msg_t *msg)
msg->local_pids, remote_argv[0]);
_print_pid_list( msg->node_name, msg->count_of_pids,
msg->local_pids, remote_argv[0] );
+ return;
+rwfail:
+ error("_process_launch_resp: "
+ "write from srun message-handler process failed");
}
@@ -396,21 +404,27 @@ update_tasks_state(srun_job_t *job, uint32_t nodeid)
int i;
pipe_enum_t pipe_enum = PIPE_TASK_STATE;
debug2("updating %d running tasks for node %d",
- job->step_layout->tasks[nodeid], nodeid);
+ job->step_layout->tasks[nodeid], nodeid);
slurm_mutex_lock(&job->task_mutex);
for (i = 0; i < job->step_layout->tasks[nodeid]; i++) {
uint32_t tid = job->step_layout->tids[nodeid][i];
if(message_thread) {
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &tid,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->task_state[tid],sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum,sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &tid,sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->task_state[tid],sizeof(int));
}
}
slurm_mutex_unlock(&job->task_mutex);
+ return;
+rwfail:
+ slurm_mutex_unlock(&job->task_mutex);
+ error("update_tasks_state: "
+ "write from srun message-handler process failed");
+
}
static void
@@ -419,22 +433,28 @@ update_running_tasks(srun_job_t *job, uint32_t nodeid)
int i;
pipe_enum_t pipe_enum = PIPE_TASK_STATE;
debug2("updating %d running tasks for node %d",
- job->step_layout->tasks[nodeid], nodeid);
+ job->step_layout->tasks[nodeid], nodeid);
slurm_mutex_lock(&job->task_mutex);
for (i = 0; i < job->step_layout->tasks[nodeid]; i++) {
uint32_t tid = job->step_layout->tids[nodeid][i];
job->task_state[tid] = SRUN_TASK_RUNNING;
if(message_thread) {
- write(job->forked_msg->
- par_msg->msg_pipe[1],&pipe_enum,sizeof(int));
- write(job->forked_msg->
- par_msg->msg_pipe[1],&tid,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->task_state[tid],sizeof(int));
+ safe_write(job->forked_msg->
+ par_msg->msg_pipe[1],
+ &pipe_enum,sizeof(int));
+ safe_write(job->forked_msg->
+ par_msg->msg_pipe[1],&tid, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->task_state[tid], sizeof(int));
}
}
slurm_mutex_unlock(&job->task_mutex);
+ return;
+rwfail:
+ slurm_mutex_unlock(&job->task_mutex);
+ error("update_running_tasks: "
+ "write from srun message-handler process failed");
}
static void
@@ -449,12 +469,12 @@ update_failed_tasks(srun_job_t *job, uint32_t nodeid)
job->task_state[tid] = SRUN_TASK_FAILED;
if(message_thread) {
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum,sizeof(int));
- write(job->forked_msg->
- par_msg->msg_pipe[1],&tid,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->task_state[tid],sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &tid, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->task_state[tid], sizeof(int));
}
tasks_exited++;
}
@@ -464,6 +484,11 @@ update_failed_tasks(srun_job_t *job, uint32_t nodeid)
debug2("all tasks exited");
update_job_state(job, SRUN_JOB_TERMINATED);
}
+rwfail:
+ slurm_mutex_unlock(&job->task_mutex);
+ error("update_failed_tasks: "
+ "write from srun message-handler process failed");
+
}
static void
@@ -479,28 +504,29 @@ _launch_handler(srun_job_t *job, slurm_msg_t *resp)
if (msg->return_code != 0) {
error("%s: launch failed: %s",
- msg->node_name, slurm_strerror(msg->return_code));
+ msg->node_name, slurm_strerror(msg->return_code));
slurm_mutex_lock(&job->task_mutex);
job->host_state[msg->srun_node_id] = SRUN_HOST_REPLIED;
slurm_mutex_unlock(&job->task_mutex);
if(message_thread) {
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &msg->srun_node_id,sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->host_state[msg->srun_node_id],sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &msg->srun_node_id, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->host_state[msg->srun_node_id],
+ sizeof(int));
}
update_failed_tasks(job, msg->srun_node_id);
/*
- if (!opt.no_kill) {
- job->rc = 124;
- update_job_state(job, SRUN_JOB_WAITING_ON_IO);
- } else
- update_failed_tasks(job, msg->srun_node_id);
+ if (!opt.no_kill) {
+ job->rc = 124;
+ update_job_state(job, SRUN_JOB_WAITING_ON_IO);
+ } else
+ update_failed_tasks(job, msg->srun_node_id);
*/
debugger_launch_failure(job);
return;
@@ -508,6 +534,11 @@ _launch_handler(srun_job_t *job, slurm_msg_t *resp)
_process_launch_resp(job, msg);
update_running_tasks(job, msg->srun_node_id);
}
+ return;
+rwfail:
+ error("_launch_handler: "
+ "write from srun message-handler process failed");
+
}
/* _confirm_launch_complete
@@ -555,12 +586,12 @@ _reattach_handler(srun_job_t *job, slurm_msg_t *msg)
if(message_thread) {
pipe_enum_t pipe_enum = PIPE_HOST_STATE;
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum, sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &resp->srun_node_id, sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->host_state[resp->srun_node_id], sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &resp->srun_node_id, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->host_state[resp->srun_node_id], sizeof(int));
}
if (resp->return_code != 0) {
@@ -589,8 +620,9 @@ _reattach_handler(srun_job_t *job, slurm_msg_t *msg)
for (i = 0; i < resp->ntasks; i++) {
job->step_layout->tids[resp->srun_node_id][i] = resp->gtids[i];
- job->step_layout->hostids[resp->gtids[i]] = resp->srun_node_id;
- info ("setting task%d on hostid %d\n", resp->gtids[i], resp->srun_node_id);
+ job->step_layout->hostids[resp->gtids[i]] = resp->srun_node_id;
+ info ("setting task%d on hostid %d\n",
+ resp->gtids[i], resp->srun_node_id);
}
_update_step_layout(job->forked_msg->par_msg->msg_pipe[1],
job->step_layout, resp->srun_node_id);
@@ -612,7 +644,10 @@ _reattach_handler(srun_job_t *job, slurm_msg_t *msg)
remote_argv[0]);
update_running_tasks(job, resp->srun_node_id);
-
+ return;
+rwfail:
+ error("_reattach_handler: "
+ "write from srun message-handler process failed");
}
@@ -679,13 +714,17 @@ _update_task_exitcode(srun_job_t *job, int taskid)
pipe_enum_t pipe_enum = PIPE_TASK_EXITCODE;
if(message_thread) {
- write(job->forked_msg->par_msg->msg_pipe[1],
- &pipe_enum, sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &taskid, sizeof(int));
- write(job->forked_msg->par_msg->msg_pipe[1],
- &job->tstatus[taskid], sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &taskid, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->tstatus[taskid], sizeof(int));
}
+ return;
+rwfail:
+ error("_update_task_exitcode: "
+ "write from srun message-handler process failed");
}
static void
@@ -729,7 +768,7 @@ _exit_handler(srun_job_t *job, slurm_msg_t *exit_msg)
tasks_exited++;
if ((tasks_exited == opt.nprocs)
- || (slurm_mpi_single_task_per_node ()
+ || (slurm_mpi_single_task_per_node ()
&& (tasks_exited == job->nhosts))) {
debug2("All tasks exited");
eio_signal_shutdown(job->eio);
@@ -782,56 +821,56 @@ _handle_msg(srun_job_t *job, slurm_msg_t *msg)
switch (msg->msg_type)
{
- case RESPONSE_LAUNCH_TASKS:
- _launch_handler(job, msg);
- slurm_free_launch_tasks_response_msg(msg->data);
- break;
- case MESSAGE_TASK_EXIT:
- _exit_handler(job, msg);
- slurm_free_task_exit_msg(msg->data);
- break;
- case RESPONSE_REATTACH_TASKS:
- debug2("recvd reattach response");
- _reattach_handler(job, msg);
- slurm_free_reattach_tasks_response_msg(msg->data);
- break;
- case SRUN_PING:
- debug3("slurmctld ping received");
- slurm_send_rc_msg(msg, SLURM_SUCCESS);
- slurm_free_srun_ping_msg(msg->data);
- break;
- case SRUN_TIMEOUT:
- to = msg->data;
- _timeout_handler(to->timeout);
- slurm_send_rc_msg(msg, SLURM_SUCCESS);
- slurm_free_srun_timeout_msg(msg->data);
- break;
- case SRUN_NODE_FAIL:
- nf = msg->data;
- _node_fail_handler(nf->nodelist, job);
- slurm_send_rc_msg(msg, SLURM_SUCCESS);
- slurm_free_srun_node_fail_msg(msg->data);
- break;
- case RESPONSE_RESOURCE_ALLOCATION:
- debug3("resource allocation response received");
- slurm_send_rc_msg(msg, SLURM_SUCCESS);
- slurm_free_resource_allocation_response_msg(msg->data);
- break;
- case PMI_KVS_PUT_REQ:
- debug3("PMI_KVS_PUT_REQ received");
- rc = pmi_kvs_put((struct kvs_comm_set *) msg->data);
- slurm_send_rc_msg(msg, rc);
- break;
- case PMI_KVS_GET_REQ:
- debug3("PMI_KVS_GET_REQ received");
- rc = pmi_kvs_get((kvs_get_msg_t *) msg->data);
- slurm_send_rc_msg(msg, rc);
- slurm_free_get_kvs_msg((kvs_get_msg_t *) msg->data);
- break;
- default:
- error("received spurious message type: %d\n",
- msg->msg_type);
- break;
+ case RESPONSE_LAUNCH_TASKS:
+ _launch_handler(job, msg);
+ slurm_free_launch_tasks_response_msg(msg->data);
+ break;
+ case MESSAGE_TASK_EXIT:
+ _exit_handler(job, msg);
+ slurm_free_task_exit_msg(msg->data);
+ break;
+ case RESPONSE_REATTACH_TASKS:
+ debug2("recvd reattach response");
+ _reattach_handler(job, msg);
+ slurm_free_reattach_tasks_response_msg(msg->data);
+ break;
+ case SRUN_PING:
+ debug3("slurmctld ping received");
+ slurm_send_rc_msg(msg, SLURM_SUCCESS);
+ slurm_free_srun_ping_msg(msg->data);
+ break;
+ case SRUN_TIMEOUT:
+ to = msg->data;
+ _timeout_handler(to->timeout);
+ slurm_send_rc_msg(msg, SLURM_SUCCESS);
+ slurm_free_srun_timeout_msg(msg->data);
+ break;
+ case SRUN_NODE_FAIL:
+ nf = msg->data;
+ _node_fail_handler(nf->nodelist, job);
+ slurm_send_rc_msg(msg, SLURM_SUCCESS);
+ slurm_free_srun_node_fail_msg(msg->data);
+ break;
+ case RESPONSE_RESOURCE_ALLOCATION:
+ debug3("resource allocation response received");
+ slurm_send_rc_msg(msg, SLURM_SUCCESS);
+ slurm_free_resource_allocation_response_msg(msg->data);
+ break;
+ case PMI_KVS_PUT_REQ:
+ debug3("PMI_KVS_PUT_REQ received");
+ rc = pmi_kvs_put((struct kvs_comm_set *) msg->data);
+ slurm_send_rc_msg(msg, rc);
+ break;
+ case PMI_KVS_GET_REQ:
+ debug3("PMI_KVS_GET_REQ received");
+ rc = pmi_kvs_get((kvs_get_msg_t *) msg->data);
+ slurm_send_rc_msg(msg, rc);
+ slurm_free_get_kvs_msg((kvs_get_msg_t *) msg->data);
+ break;
+ default:
+ error("received spurious message type: %d\n",
+ msg->msg_type);
+ break;
}
return;
}
@@ -875,7 +914,7 @@ _accept_msg_connection(srun_job_t *job, int fdnum)
* to 5 seconds for no_alloc option only */
if (opt.no_alloc)
timeout = 5;
- again:
+again:
ret_list = slurm_receive_msg(fd, msg, timeout);
if(!ret_list || errno != SLURM_SUCCESS) {
if (errno == EINTR) {
@@ -935,7 +974,7 @@ _do_poll(srun_job_t *job, struct pollfd *fds, int timeout)
case EINVAL:
case EFAULT: fatal("poll: %m");
default: error("poll: %m. Continuing...");
- continue;
+ continue;
}
}
@@ -960,8 +999,8 @@ _get_next_timeout(srun_job_t *job)
timeout = job->ltimeout - time(NULL);
else
timeout = job->ltimeout < job->etimeout ?
- job->ltimeout - time(NULL) :
- job->etimeout - time(NULL);
+ job->ltimeout - time(NULL) :
+ job->etimeout - time(NULL);
return timeout;
}
@@ -1064,7 +1103,8 @@ par_thr(void *arg)
//slurm_uid = (uid_t) slurm_get_slurm_user_id();
close(msg_par->msg_pipe[0]); // close read end of pipe
close(par_msg->msg_pipe[1]); // close write end of pipe
- while(read(par_msg->msg_pipe[0], &c, sizeof(int)) == sizeof(int)) {
+ while(read(par_msg->msg_pipe[0], &c, sizeof(int))
+ == sizeof(int)) {
// getting info from msg thread
if(type == PIPE_NONE) {
debug2("got type %d\n",c);
@@ -1216,7 +1256,7 @@ msg_thr_create(srun_job_t *job)
slurm_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
while ((errno = pthread_create(&job->jtid, &attr, &msg_thr,
- (void *)job))) {
+ (void *)job))) {
if (++retries > MAX_RETRIES)
fatal("Can't create pthread");
sleep(1);
@@ -1231,8 +1271,7 @@ msg_thr_create(srun_job_t *job)
* close.
*/
while(read(job->forked_msg->msg_par->msg_pipe[0],
- &c, sizeof(int))
- > 0)
+ &c, sizeof(int)) > 0)
; /* do nothing */
close(job->forked_msg->msg_par->msg_pipe[0]);
@@ -1251,7 +1290,7 @@ msg_thr_create(srun_job_t *job)
slurm_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
while ((errno = pthread_create(&job->jtid, &attr, &par_thr,
- (void *)job))) {
+ (void *)job))) {
if (++retries > MAX_RETRIES)
fatal("Can't create pthread");
sleep(1); /* sleep and try again */
@@ -1307,13 +1346,13 @@ extern slurm_fd slurmctld_msg_init(void)
fatal("slurm_get_stream_addr error %m");
fd_set_nonblocking(slurmctld_fd);
/* hostname is not set, so slurm_get_addr fails
- slurm_get_addr(&slurm_address, &port, hostname, sizeof(hostname)); */
+ slurm_get_addr(&slurm_address, &port, hostname, sizeof(hostname)); */
port = ntohs(slurm_address.sin_port);
slurmctld_comm_addr.hostname = xstrdup(opt.ctrl_comm_ifhn);
slurmctld_comm_addr.port = port;
debug2("slurmctld messages to host=%s,port=%u",
- slurmctld_comm_addr.hostname,
- slurmctld_comm_addr.port);
+ slurmctld_comm_addr.hostname,
+ slurmctld_comm_addr.port);
return slurmctld_fd;
}
diff --git a/src/srun/srun_job.c b/src/srun/srun_job.c
index 3ae218f8c09d0960eaec66deb0cbf55c190c2fbf..e90f270675f5b2fc41ffceef0a817aabb6869024 100644
--- a/src/srun/srun_job.c
+++ b/src/srun/srun_job.c
@@ -304,15 +304,21 @@ update_job_state(srun_job_t *job, srun_job_state_t state)
if (job->state < state) {
job->state = state;
if(message_thread) {
- write(job->forked_msg->
- par_msg->msg_pipe[1],&pipe_enum,sizeof(int));
- write(job->forked_msg->
- par_msg->msg_pipe[1],&job->state,sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &pipe_enum, sizeof(int));
+ safe_write(job->forked_msg->par_msg->msg_pipe[1],
+ &job->state, sizeof(int));
}
pthread_cond_signal(&job->state_cond);
}
pthread_mutex_unlock(&job->state_mutex);
+ return;
+rwfail:
+ pthread_mutex_unlock(&job->state_mutex);
+ error("update_job_state: "
+ "write from srun message-handler process failed");
+
}
srun_job_state_t
diff --git a/testsuite/expect/globals.example b/testsuite/expect/globals.example
index 2535af3d4313550296a56b144a99be6fc7583c8a..651cacea652ad511e1d6afe44b1676db50bd5e28 100755
--- a/testsuite/expect/globals.example
+++ b/testsuite/expect/globals.example
@@ -116,6 +116,7 @@ set alpha_numeric_under "\[a-zA-Z0-9_\]+"
set alpha_under "\[A-Z_\]+"
set end_of_line "\[\r\n\]"
set number "\[0-9\]+"
+set float "\[0-9\]+\\.?\[0-9\]+"
#
# Cache SlurmUser to check for SuperUser requests
diff --git a/testsuite/expect/test12.2 b/testsuite/expect/test12.2
index 358d4a725d745901177960ea44b0f9bfaac7e53c..f949dcb81bbc4732107cd90d16c2cb3d96c05bf2 100755
--- a/testsuite/expect/test12.2
+++ b/testsuite/expect/test12.2
@@ -39,7 +39,7 @@ set matches 0
# job paramters
set mem_size 100000
-set sleep_time 25
+set sleep_time 5
set ret_code 42
print_header $test_id
@@ -51,7 +51,11 @@ set supported 0
log_user 0
spawn $scontrol show config
expect {
- -re "jobacct/log" {
+ -re "jobacct/linux" {
+ set supported 1
+ exp_continue
+ }
+ -re "jobacct/aix" {
set supported 1
exp_continue
}
@@ -103,7 +107,7 @@ exec $bin_sleep 5
# Report basic sacct info
#
-spawn $sacct --noheader --jobstep=$job_id.0 --fields jobstep,jobname,status,error
+spawn $sacct --noheader --job=$job_id.0 --fields jobid,jobname,status,exitcode
expect {
-re "$job_id\.0" {
incr matches
@@ -138,17 +142,13 @@ if {$matches != 4} {
# Report the sacct accouting info
#
set elapsed_time 0
-spawn $sacct --noheader --jobstep=$job_id.0 --fields elapsed
+spawn $sacct --noheader --job=$job_id.0 --fields elapsed
expect {
- -re "($number):($number)\.($number)" {
+ -re "($number):($number)" {
set mins $expect_out(1,string)
set secs $expect_out(2,string)
set elapsed_time [expr $mins * 60 + $secs]
}
- -re "($number)\.($number)" {
- set elapsed_time $expect_out(1,string)
- exp_continue
- }
timeout {
send_user "\nFAILURE: sacct not responding\n"
set exit_code 1
@@ -167,11 +167,15 @@ if {$error_time > 5} {
send_user "\nSUCCESS: sacct elapsed time error of $error_time\n"
}
-set mem_used 0
-spawn $sacct --noheader --jobstep=$job_id.0 --fields vsize
+set mem_used -1
+set mem_task -1
+set ave_used -1
+spawn $sacct --noheader --job=$job_id.0 --fields vsize
expect {
- -re "($number)" {
+ -re "($float).*/($number) - ($float)" {
set mem_used $expect_out(1,string)
+ set mem_task $expect_out(2,string)
+ set ave_used $expect_out(3,string)
exp_continue
}
timeout {
@@ -183,6 +187,19 @@ expect {
}
}
+if { $mem_used == -1 } {
+ send_user "\nFAILURE: sacct memory not found\n"
+ set exit_code 1
+} elseif { $mem_task != 0 } {
+ send_user "\nFAILURE: sacct memory task not found\n"
+ set exit_code 1
+} elseif { $ave_used != $mem_used } {
+ send_user "\nFAILURE: sacct memory task not equal to ave memory\n"
+ set exit_code 1
+}
+
+set mem_used [expr $mem_used * 1024]
+
set diff_mem [expr $mem_used - $mem_size]
set error_mem [expr abs($diff_mem)]
if {$error_mem > 4000} {
diff --git a/testsuite/expect/test14.7 b/testsuite/expect/test14.7
index 46983d4df2ea0a9d092ad82b98b159e55ec0061d..ca30f15c90eec4e8203f9aec86367e90f0ca73c2 100755
--- a/testsuite/expect/test14.7
+++ b/testsuite/expect/test14.7
@@ -118,7 +118,12 @@ if {[wait_for_file $file_err] == 0} {
incr matches
exp_continue
}
- -re "No such file" | "does not exist" {
+ -re "No such file" {
+ send_user "These errors are expected, no worries\n"
+ incr matches
+ exp_continue
+ }
+ -re "does not exist" {
send_user "These errors are expected, no worries\n"
incr matches
exp_continue