diff --git a/doc/html/taskplugins.shtml b/doc/html/taskplugins.shtml
index 4599bb8beb1fa602fa3875f1f535ab5e22992197..bcc3d0f1eb424426d32d4e7664ae2c62c4a01370 100644
--- a/doc/html/taskplugins.shtml
+++ b/doc/html/taskplugins.shtml
@@ -84,6 +84,28 @@ ID of the node on which the resources are being acquired
On failure, the plugin should return SLURM_ERROR and set the errno to an
appropriate value to indicate the reason for failure.</p>
+<p class="commandline">int task_slurmd_suspend_job (uint32_t job_id);</p>
+<p style="margin-left:.2in"><b>Description</b>: Temporarily release resources
+previously reserved for a job.
+Executed by the <b>slurmd</b> daemon as user root.</p>
+<p style="margin-left:.2in"><b>Arguments</b>:
+<span class="commandline">job_id</span> (input)
+ID of the job which is being suspended.</p>
+<p style="margin-left:.2in"><b>Returns</b>: SLURM_SUCCESS if successful.
+On failure, the plugin should return SLURM_ERROR and set the errno to an
+appropriate value to indicate the reason for failure.</p>
+
+<p class="commandline">int task_slurmd_resume_job (uint32_t job_id);</p>
+<p style="margin-left:.2in"><b>Description</b>: Reclaim resources which
+were previously released using the task_slurmd_suspend_job function.
+Executed by the <b>slurmd</b> daemon as user root.</p>
+<p style="margin-left:.2in"><b>Arguments</b>:
+<span class="commandline">job_id</span> (input)
+ID of the job which is being resumed.</p>
+<p style="margin-left:.2in"><b>Returns</b>: SLURM_SUCCESS if successful.
+On failure, the plugin should return SLURM_ERROR and set the errno to an
+appropriate value to indicate the reason for failure.</p>
+
<p class="commandline">int task_slurmd_release_resources (uint32_t job_id);</p>
<p style="margin-left:.2in"><b>Description</b>: Release resources previously
reserved for a job. Executed by the <b>slurmd</b> daemon as user root.</p>
@@ -142,6 +164,6 @@ appropriate value to indicate the reason for failure.</p>
Future releases of SLURM may revise this API.</p>
<p class="footer"><a href="#top">top</a></p>
-<p style="text-align:center;">Last modified 27 March 2007</p>
+<p style="text-align:center;">Last modified 28 May 2008</p>
<!--#include virtual="footer.txt"-->
diff --git a/src/plugins/task/affinity/dist_tasks.c b/src/plugins/task/affinity/dist_tasks.c
index 03d86c0873cd0fa4cfd23ec75942f0808d926723..9a725dc1afcfde72ffe89d2a1dea3a7e9f0600d0 100644
--- a/src/plugins/task/affinity/dist_tasks.c
+++ b/src/plugins/task/affinity/dist_tasks.c
@@ -35,13 +35,12 @@
#include "src/plugins/task/affinity/dist_tasks.h"
static slurm_lllp_ctx_t *lllp_ctx = NULL; /* binding context */
-static struct node_gids *lllp_tasks = NULL; /* Keep track of the task count for
- * logical processors
- * socket/core/thread.
- */
-static uint32_t lllp_reserved_size = 0;/* size of lllp reserved array */
-static uint32_t *lllp_reserved = NULL; /* count of Reserved lllps (socket,
- * core, threads) */
+static struct node_gids *lllp_tasks = NULL; /* Keep track of the task count
+ * for logical processors
+ * socket/core/thread. */
+static uint32_t lllp_reserved_size = 0; /* lllp reserved array size */
+static uint32_t *lllp_reserved = NULL; /* count of Reserved lllps
+ * (socket, core, threads) */
static void _task_layout_display_masks(launch_tasks_request_msg_t *req,
diff --git a/src/plugins/task/affinity/task_affinity.c b/src/plugins/task/affinity/task_affinity.c
index 054c753956d24f57a4077c81f98baf6951b1310b..466622d180b4616f9dd98ba350a4844b77318d1a 100644
--- a/src/plugins/task/affinity/task_affinity.c
+++ b/src/plugins/task/affinity/task_affinity.c
@@ -2,11 +2,11 @@
* task_affinity.c - Library for task pre-launch and post_termination
* functions for task affinity support
*****************************************************************************
- * Copyright (C) 2005 Hewlett-Packard Development Company, L.P.
+ * Copyright (C) 2005-2008 Hewlett-Packard Development Company, L.P.
* Modified by Hewlett-Packard for task affinity support using task_none.c
- * Copyright (C) 2005 The Regents of the University of California and
+ * Copyright (C) 2005-2007 The Regents of the University of California
+ * Copyright (C) 2008 Lawrence Livermore National Security.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
- * task_none.c Written by Morris Jette <jette1@llnl.gov>.
* LLNL-CODE-402394.
*
* This file is part of SLURM, a resource management program.
@@ -84,7 +84,7 @@ const uint32_t plugin_version = 100;
* init() is called when the plugin is loaded, before any other functions
* are called. Put global initialization here.
*/
-int init ( void )
+extern int init (void)
{
lllp_ctx_alloc();
verbose("%s loaded", plugin_name);
@@ -95,7 +95,7 @@ int init ( void )
* fini() is called when the plugin is removed. Clear any allocated
* storage here.
*/
-int fini ( void )
+extern int fini (void)
{
lllp_ctx_destroy();
verbose("%s unloaded", plugin_name);
@@ -195,8 +195,9 @@ static void _update_bind_type(launch_tasks_request_msg_t *req)
/*
* task_slurmd_launch_request()
*/
-int task_slurmd_launch_request (uint32_t job_id,
- launch_tasks_request_msg_t *req, uint32_t node_id)
+extern int task_slurmd_launch_request (uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id)
{
int hw_sockets, hw_cores, hw_threads;
char buf_type[100];
@@ -230,18 +231,37 @@ int task_slurmd_launch_request (uint32_t job_id,
/*
* task_slurmd_reserve_resources()
*/
-int task_slurmd_reserve_resources ( uint32_t job_id,
- launch_tasks_request_msg_t *req, uint32_t node_id)
+extern int task_slurmd_reserve_resources (uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id)
{
debug("task_slurmd_reserve_resources: %u", job_id);
cr_reserve_lllp(job_id, req, node_id);
return SLURM_SUCCESS;
}
+/*
+ * task_slurmd_suspend_job()
+ */
+extern int task_slurmd_suspend_job (uint32_t job_id)
+{
+ debug("task_slurmd_suspend_job: %u", job_id);
+ return SLURM_SUCCESS;
+}
+
+/*
+ * task_slurmd_resume_job()
+ */
+extern int task_slurmd_resume_job (uint32_t job_id)
+{
+ debug("task_slurmd_resume_job: %u", job_id);
+ return SLURM_SUCCESS;
+}
+
/*
* task_slurmd_release_resources()
*/
-int task_slurmd_release_resources ( uint32_t job_id )
+extern int task_slurmd_release_resources (uint32_t job_id)
{
debug("task_slurmd_release_resources: %u", job_id);
cr_release_lllp(job_id);
@@ -253,7 +273,7 @@ int task_slurmd_release_resources ( uint32_t job_id )
* user to launch his jobs. Use this to create the CPUSET directory
* and set the owner appropriately.
*/
-int task_pre_setuid ( slurmd_job_t *job )
+extern int task_pre_setuid (slurmd_job_t *job)
{
char path[PATH_MAX];
@@ -274,7 +294,7 @@ int task_pre_setuid ( slurmd_job_t *job )
* It is followed by TaskProlog program (from slurm.conf) and
* --task-prolog (from srun command line).
*/
-int task_pre_launch ( slurmd_job_t *job )
+extern int task_pre_launch (slurmd_job_t *job)
{
char base[PATH_MAX], path[PATH_MAX];
@@ -358,7 +378,7 @@ int task_pre_launch ( slurmd_job_t *job )
* It is preceeded by --task-epilog (from srun command line)
* followed by TaskEpilog program (from slurm.conf).
*/
-int task_post_term ( slurmd_job_t *job )
+extern int task_post_term (slurmd_job_t *job)
{
debug("affinity task_post_term: %u.%u, task %d",
job->jobid, job->stepid, job->envtp->procid);
diff --git a/src/plugins/task/none/task_none.c b/src/plugins/task/none/task_none.c
index 177fce59ac02a225458e21e941654e775812652c..5e852eab6596b349b6eebdf1efb47719b34bb228 100644
--- a/src/plugins/task/none/task_none.c
+++ b/src/plugins/task/none/task_none.c
@@ -2,7 +2,8 @@
* task_none.c - Library for task pre-launch and post_termination functions
* with no actions
*****************************************************************************
- * Copyright (C) 2005 The Regents of the University of California.
+ * Copyright (C) 2005-2007 The Regents of the University of California.
+ * Copyright (C) 2008 Lawrence Livermore National Security.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* LLNL-CODE-402394.
@@ -82,7 +83,7 @@ const uint32_t plugin_version = 100;
* init() is called when the plugin is loaded, before any other functions
* are called. Put global initialization here.
*/
-int init ( void )
+extern int init (void)
{
verbose("%s loaded", plugin_name);
return SLURM_SUCCESS;
@@ -92,7 +93,7 @@ int init ( void )
* fini() is called when the plugin is removed. Clear any allocated
* storage here.
*/
-int fini ( void )
+extern int fini (void)
{
return SLURM_SUCCESS;
}
@@ -100,30 +101,49 @@ int fini ( void )
/*
* task_slurmd_launch_request()
*/
-int task_slurmd_launch_request ( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id)
+extern int task_slurmd_launch_request (uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id)
{
- debug("task_slurmd_launch_request: %u %u",
- job_id, node_id);
+ debug("task_slurmd_launch_request: %u %u", job_id, node_id);
return SLURM_SUCCESS;
}
/*
* task_slurmd_reserve_resources()
*/
-int task_slurmd_reserve_resources ( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id )
+extern int task_slurmd_reserve_resources (uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id)
{
- debug("task_slurmd_reserve_resources: %u %u",
- job_id, node_id);
+ debug("task_slurmd_reserve_resources: %u %u", job_id, node_id);
+ return SLURM_SUCCESS;
+}
+
+/*
+ * task_slurmd_suspend_job()
+ */
+extern int task_slurmd_suspend_job (uint32_t job_id)
+{
+ debug("task_slurmd_suspend_job: %u", job_id);
+ return SLURM_SUCCESS;
+}
+
+/*
+ * task_slurmd_resume_job()
+ */
+extern int task_slurmd_resume_job (uint32_t job_id)
+{
+ debug("task_slurmd_resume_job: %u", job_id);
return SLURM_SUCCESS;
}
/*
* task_slurmd_release_resources()
*/
-int task_slurmd_release_resources ( uint32_t job_id )
+extern int task_slurmd_release_resources (uint32_t job_id)
{
- debug("task_slurmd_release_resources: %u",
- job_id);
+ debug("task_slurmd_release_resources: %u", job_id);
return SLURM_SUCCESS;
}
@@ -132,7 +152,7 @@ int task_slurmd_release_resources ( uint32_t job_id )
* user to launch his jobs. Use this to create the CPUSET directory
* and set the owner appropriately.
*/
-int task_pre_setuid ( slurmd_job_t *job )
+extern int task_pre_setuid (slurmd_job_t *job)
{
return SLURM_SUCCESS;
}
@@ -142,7 +162,7 @@ int task_pre_setuid ( slurmd_job_t *job )
* It is followed by TaskProlog program (from slurm.conf) and
* --task-prolog (from srun command line).
*/
-int task_pre_launch ( slurmd_job_t *job )
+extern int task_pre_launch (slurmd_job_t *job)
{
debug("task_pre_launch: %u.%u, task %d",
job->jobid, job->stepid, job->envtp->procid);
@@ -154,7 +174,7 @@ int task_pre_launch ( slurmd_job_t *job )
* It is preceeded by --task-epilog (from srun command line)
* followed by TaskEpilog program (from slurm.conf).
*/
-int task_post_term ( slurmd_job_t *job )
+extern int task_post_term (slurmd_job_t *job)
{
debug("task_post_term: %u.%u, task %d",
job->jobid, job->stepid, job->envtp->procid);
diff --git a/src/slurmd/common/task_plugin.c b/src/slurmd/common/task_plugin.c
index 1a7505c017507dbfd76aa90209ffb1baba045054..1882a9b0555a1044a77eff929214fbcdb7330bc2 100644
--- a/src/slurmd/common/task_plugin.c
+++ b/src/slurmd/common/task_plugin.c
@@ -1,7 +1,8 @@
/*****************************************************************************\
* task_plugin.h - task launch plugin stub.
*****************************************************************************
- * Copyright (C) 2005 The Regents of the University of California.
+ * Copyright (C) 2005-2007 The Regents of the University of California.
+ * Copyright (C) 2008 Lawrence Livermore National Security.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* LLNL-CODE-402394.
@@ -46,13 +47,19 @@
#include "src/slurmd/slurmstepd/slurmstepd_job.h"
typedef struct slurmd_task_ops {
- int (*slurmd_launch_request) ( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id);
- int (*slurmd_reserve_resources) ( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id );
- int (*slurmd_release_resources) ( uint32_t job_id);
-
- int (*pre_setuid) ( slurmd_job_t *job );
- int (*pre_launch) ( slurmd_job_t *job );
- int (*post_term) ( slurmd_job_t *job );
+ int (*slurmd_launch_request) (uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id);
+ int (*slurmd_reserve_resources) (uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id);
+ int (*slurmd_suspend_job) (uint32_t job_id);
+ int (*slurmd_resume_job) (uint32_t job_id);
+ int (*slurmd_release_resources) (uint32_t job_id);
+
+ int (*pre_setuid) (slurmd_job_t *job);
+ int (*pre_launch) (slurmd_job_t *job);
+ int (*post_term) (slurmd_job_t *job);
} slurmd_task_ops_t;
@@ -68,7 +75,7 @@ static pthread_mutex_t g_task_context_lock = PTHREAD_MUTEX_INITIALIZER;
static slurmd_task_ops_t *
-_slurmd_task_get_ops( slurmd_task_context_t *c )
+_slurmd_task_get_ops(slurmd_task_context_t *c)
{
/*
* Must be synchronized with slurmd_task_ops_t above.
@@ -76,6 +83,8 @@ _slurmd_task_get_ops( slurmd_task_context_t *c )
static const char *syms[] = {
"task_slurmd_launch_request",
"task_slurmd_reserve_resources",
+ "task_slurmd_suspend_job",
+ "task_slurmd_resume_job",
"task_slurmd_release_resources",
"task_pre_setuid",
"task_pre_launch",
@@ -118,7 +127,7 @@ _slurmd_task_get_ops( slurmd_task_context_t *c )
static slurmd_task_context_t *
-_slurmd_task_context_create( const char *task_plugin_type )
+_slurmd_task_context_create(const char *task_plugin_type)
{
slurmd_task_context_t *c;
@@ -137,7 +146,7 @@ _slurmd_task_context_create( const char *task_plugin_type )
static int
-_slurmd_task_context_destroy( slurmd_task_context_t *c )
+_slurmd_task_context_destroy(slurmd_task_context_t *c)
{
/*
* Must check return code here because plugins might still
@@ -161,7 +170,7 @@ _slurmd_task_context_destroy( slurmd_task_context_t *c )
*
* RET - slurm error code
*/
-extern int slurmd_task_init( void )
+extern int slurmd_task_init(void)
{
int retval = SLURM_SUCCESS;
char *task_plugin_type = NULL;
@@ -198,7 +207,7 @@ extern int slurmd_task_init( void )
*
* RET - slurm error code
*/
-extern int slurmd_task_fini( void )
+extern int slurmd_task_fini(void)
{
int rc;
@@ -215,9 +224,11 @@ extern int slurmd_task_fini( void )
*
* RET - slurm error code
*/
-extern int slurmd_launch_request( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id)
+extern int slurmd_launch_request(uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id)
{
- if ( slurmd_task_init() )
+ if (slurmd_task_init())
return SLURM_ERROR;
return (*(g_task_context->ops.slurmd_launch_request))(job_id, req, node_id);
@@ -228,22 +239,50 @@ extern int slurmd_launch_request( uint32_t job_id, launch_tasks_request_msg_t *r
*
* RET - slurm error code
*/
-extern int slurmd_reserve_resources( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id )
+extern int slurmd_reserve_resources(uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id )
{
- if ( slurmd_task_init() )
+ if (slurmd_task_init())
return SLURM_ERROR;
return (*(g_task_context->ops.slurmd_reserve_resources))(job_id, req, node_id);
}
+/*
+ * Slurmd is suspending a job.
+ *
+ * RET - slurm error code
+ */
+extern int slurmd_suspend_job(uint32_t job_id)
+{
+ if (slurmd_task_init())
+ return SLURM_ERROR;
+
+ return (*(g_task_context->ops.slurmd_suspend_job))(job_id);
+}
+
+/*
+ * Slurmd is resuming a previously suspended job.
+ *
+ * RET - slurm error code
+ */
+extern int slurmd_resume_job(uint32_t job_id)
+{
+ if (slurmd_task_init())
+ return SLURM_ERROR;
+
+ return (*(g_task_context->ops.slurmd_resume_job))(job_id);
+}
+
/*
* Slurmd is releasing resources for the task.
*
* RET - slurm error code
*/
-extern int slurmd_release_resources( uint32_t job_id )
+extern int slurmd_release_resources(uint32_t job_id)
{
- if ( slurmd_task_init() )
+ if (slurmd_task_init())
return SLURM_ERROR;
return (*(g_task_context->ops.slurmd_release_resources))(job_id);
@@ -255,9 +294,9 @@ extern int slurmd_release_resources( uint32_t job_id )
*
* RET - slurm error code
*/
-extern int pre_setuid( slurmd_job_t *job )
+extern int pre_setuid(slurmd_job_t *job)
{
- if ( slurmd_task_init() )
+ if (slurmd_task_init())
return SLURM_ERROR;
return (*(g_task_context->ops.pre_setuid))(job);
@@ -268,9 +307,9 @@ extern int pre_setuid( slurmd_job_t *job )
*
* RET - slurm error code
*/
-extern int pre_launch( slurmd_job_t *job )
+extern int pre_launch(slurmd_job_t *job)
{
- if ( slurmd_task_init() )
+ if (slurmd_task_init())
return SLURM_ERROR;
return (*(g_task_context->ops.pre_launch))(job);
@@ -281,9 +320,9 @@ extern int pre_launch( slurmd_job_t *job )
*
* RET - slurm error code
*/
-extern int post_term( slurmd_job_t *job )
+extern int post_term(slurmd_job_t *job)
{
- if ( slurmd_task_init() )
+ if (slurmd_task_init())
return SLURM_ERROR;
return (*(g_task_context->ops.post_term))(job);
diff --git a/src/slurmd/common/task_plugin.h b/src/slurmd/common/task_plugin.h
index 4cca516ac64fc08cc2e0edbf0ad1689285ebd39e..121876569de2e577f7b4507ec9fd0ce4b82e6529 100644
--- a/src/slurmd/common/task_plugin.h
+++ b/src/slurmd/common/task_plugin.h
@@ -1,7 +1,8 @@
/*****************************************************************************\
* task_plugin.h - Define plugin functions for task pre_launch and post_term.
*****************************************************************************
- * Copyright (C) 2005 The Regents of the University of California.
+ * Copyright (C) 2005-2007 The Regents of the University of California.
+ * Copyright (C) 2008 Lawrence Livermore National Security.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* LLNL-CODE-402394.
@@ -65,21 +66,39 @@ extern int slurmd_task_fini(void);
*
* RET - slurm error code
*/
-extern int slurmd_launch_request( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id );
+extern int slurmd_launch_request(uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id );
/*
* Slurmd is reserving resources for the task.
*
* RET - slurm error code
*/
-extern int slurmd_reserve_resources( uint32_t job_id, launch_tasks_request_msg_t *req, uint32_t node_id );
+extern int slurmd_reserve_resources(uint32_t job_id,
+ launch_tasks_request_msg_t *req,
+ uint32_t node_id );
+
+/*
+ * Slurmd is suspending a job.
+ *
+ * RET - slurm error code
+ */
+extern int slurmd_suspend_job(uint32_t job_id);
+
+/*
+ * Slurmd is resuming a previously suspended job.
+ *
+ * RET - slurm error code
+ */
+extern int slurmd_resume_job(uint32_t job_id);
/*
* Slurmd is releasing resources for the task.
*
* RET - slurm error code
*/
-extern int slurmd_release_resources( uint32_t job_id );
+extern int slurmd_release_resources(uint32_t job_id);
/*
* Note that a task launch is about to occur.
diff --git a/src/slurmd/slurmd/req.c b/src/slurmd/slurmd/req.c
index c3c70a372dd1bcd6bdf105d2dceaf19dba2ca11e..6852e43f6db1ab7131c032a175a738aee85dda8d 100644
--- a/src/slurmd/slurmd/req.c
+++ b/src/slurmd/slurmd/req.c
@@ -2390,6 +2390,12 @@ _rpc_suspend_job(slurm_msg_t *msg)
sleep(1);
}
+ /* Release or reclaim resources bound to these tasks (task affinity) */
+ if (req->op == SUSPEND_JOB)
+ (void) slurmd_suspend_job(req->job_id);
+ else
+ (void) slurmd_resume_job(req->job_id);
+
/*
* Loop through all job steps and call stepd_suspend or stepd_resume
* as appropriate. Since the "suspend" action contains a 'sleep 1',