From cb68cce85aa466cb72a74eab36439f038d470542 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Sat, 26 Mar 2011 02:57:23 +0000
Subject: [PATCH] Patch #21: Prevents nesting of ALPS sessions within salloc.
Please see separate email correspondence with Ben - to me this
now looks much similar to the way the SGI pagg IDs are used on
Cray: * upper part = node_id (corresponding to alloc_node)
* lower part = 32 bit job ID (corresponding to alloc_sid)
---
src/slurmctld/job_mgr.c | 36 ++++++++++++++++++++++++++++++++++++
src/slurmctld/proc_req.c | 12 ++++++++++++
src/slurmctld/slurmctld.h | 7 +++++++
3 files changed, 55 insertions(+)
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index d1f22d1605b..ea05b8b825c 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -2166,6 +2166,42 @@ extern bool partition_in_use(char *part_name)
return false;
}
+/*
+ * allocated_session_in_use - check if an interactive session is already running
+ * IN new_alloc - allocation (alloc_node:alloc_sid) to test for
+ * Returns true if an interactive session of the same node:sid already is in use
+ * by a RUNNING, PENDING, or SUSPENDED job. Provides its own locking.
+ */
+extern bool allocated_session_in_use(job_desc_msg_t *new_alloc)
+{
+ ListIterator job_iter;
+ struct job_record *job_ptr;
+ /* Locks: Read job */
+ slurmctld_lock_t job_read_lock = {
+ NO_LOCK, READ_LOCK, NO_LOCK, NO_LOCK };
+
+ if ((new_alloc->script != NULL) || (new_alloc->alloc_node == NULL))
+ return false;
+
+ lock_slurmctld(job_read_lock);
+ job_iter = list_iterator_create(job_list);
+ if (job_iter == NULL)
+ fatal("list_iterator_create: malloc failure");
+
+ while ((job_ptr = (struct job_record *)list_next(job_iter))) {
+ if (job_ptr->batch_flag || IS_JOB_FINISHED(job_ptr))
+ continue;
+ if (job_ptr->alloc_node &&
+ (strcmp(job_ptr->alloc_node, new_alloc->alloc_node) == 0) &&
+ (job_ptr->alloc_sid == new_alloc->alloc_sid))
+ break;
+ }
+ list_iterator_destroy(job_iter);
+ unlock_slurmctld(job_read_lock);
+
+ return job_ptr != NULL;
+}
+
/*
* kill_running_job_by_node_name - Given a node name, deallocate RUNNING
* or COMPLETING jobs from the node or kill them
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index a98a3b0d52c..42d8a376481 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -769,6 +769,18 @@ static void _slurm_rpc_allocate_resources(slurm_msg_t * msg)
error("REQUEST_RESOURCE_ALLOCATE lacks alloc_node from uid=%d",
uid);
}
+#if HAVE_CRAY
+ /*
+ * We are using the alloc_sid as unique identifier to confirm the ALPS
+ * reservation. ALPS will refuse any attempt to create a second session
+ * with the same identifier, hence sessions may not be nested.
+ */
+ if (allocated_session_in_use(job_desc_msg)) {
+ error_code = ESLURM_RESERVATION_BUSY;
+ error("attempt to nest ALPS allocation on %s:%d by uid=%d",
+ job_desc_msg->alloc_node, job_desc_msg->alloc_sid, uid);
+ }
+#endif
slurm_get_peer_addr(msg->conn_fd, &resp_addr);
job_desc_msg->resp_host = xmalloc(16);
slurm_get_ip_str(&resp_addr, &port, job_desc_msg->resp_host, 16);
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index 9c012ef9eec..462f4ca0272 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -875,6 +875,13 @@ extern bool is_node_down (char *name);
*/
extern bool is_node_resp (char *name);
+/*
+ * allocated_session_in_use - check if an interactive session is already running
+ * IN new_alloc - allocation (alloc_node:alloc_sid) to test for
+ * Returns true if an interactive session of the same node:sid already exists.
+ */
+extern bool allocated_session_in_use(job_desc_msg_t *new_alloc);
+
/*
* job_alloc_info - get details about an existing job allocation
* IN uid - job issuing the code
--
GitLab