From 452ae1a863ccb817b430cf1f9739cfe6fc146aea Mon Sep 17 00:00:00 2001
From: Danny Auble <da@llnl.gov>
Date: Wed, 19 Jan 2011 23:17:11 +0000
Subject: [PATCH] -- BLUEGENE - This fixes a race condition with dynamic mode
to make it so we copy the booted and job list of blocks before trying to
create.
---
NEWS | 2 +-
.../select/bluegene/plugin/bg_job_place.c | 29 ++++++++++++++-----
.../select/bluegene/plugin/dynamic_block.c | 7 +++--
3 files changed, 27 insertions(+), 11 deletions(-)
diff --git a/NEWS b/NEWS
index a0932a9baa2..6ceaa08bd0e 100644
--- a/NEWS
+++ b/NEWS
@@ -42,7 +42,7 @@ documents those changes that are of interest to users and admins.
-- BLUEGENE - more robust checking for states when freeing blocks.
-- Added correct files to the slurm.spec file for correct perl api rpm
creation.
-
+
* Changes in SLURM 2.2.0
========================
-- Change format of Duration field in "scontrol show reservation" output from
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index 71ebf6f8f4f..126ab37aa29 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -729,6 +729,7 @@ static int _dynamically_request(List block_list, int *blocks_added,
List list_of_lists = NULL;
List temp_list = NULL;
List new_blocks = NULL;
+ List job_list = NULL, booted_list = NULL;
ListIterator itr = NULL;
int rc = SLURM_ERROR;
int create_try = 0;
@@ -746,20 +747,28 @@ static int _dynamically_request(List block_list, int *blocks_added,
if (SELECT_IS_PREEMPT_SET(query_mode)
&& SELECT_IS_CHECK_FULL_SET(query_mode)) {
list_append(list_of_lists, block_list);
- } else if (user_req_nodes)
- list_append(list_of_lists, bg_lists->job_running);
- else {
+ } else if (user_req_nodes) {
+ slurm_mutex_lock(&block_state_mutex);
+ job_list = copy_bg_list(bg_lists->job_running);
+ list_append(list_of_lists, job_list);
+ slurm_mutex_unlock(&block_state_mutex);
+ } else {
+ slurm_mutex_lock(&block_state_mutex);
list_append(list_of_lists, block_list);
if (list_count(block_list) != list_count(bg_lists->booted)) {
- list_append(list_of_lists, bg_lists->booted);
+ booted_list = copy_bg_list(bg_lists->booted);
+ list_append(list_of_lists, booted_list);
if (list_count(bg_lists->booted)
- != list_count(bg_lists->job_running))
- list_append(list_of_lists,
- bg_lists->job_running);
+ != list_count(bg_lists->job_running)) {
+ job_list = copy_bg_list(bg_lists->job_running);
+ list_append(list_of_lists, job_list);
+ }
} else if (list_count(block_list)
!= list_count(bg_lists->job_running)) {
- list_append(list_of_lists, bg_lists->job_running);
+ job_list = copy_bg_list(bg_lists->job_running);
+ list_append(list_of_lists, job_list);
}
+ slurm_mutex_unlock(&block_state_mutex);
}
itr = list_iterator_create(list_of_lists);
while ((temp_list = (List)list_next(itr))) {
@@ -838,6 +847,10 @@ static int _dynamically_request(List block_list, int *blocks_added,
if (list_of_lists)
list_destroy(list_of_lists);
+ if (job_list)
+ list_destroy(job_list);
+ if (booted_list)
+ list_destroy(booted_list);
return rc;
}
diff --git a/src/plugins/select/bluegene/plugin/dynamic_block.c b/src/plugins/select/bluegene/plugin/dynamic_block.c
index 8c17f5601ae..de672cb3004 100644
--- a/src/plugins/select/bluegene/plugin/dynamic_block.c
+++ b/src/plugins/select/bluegene/plugin/dynamic_block.c
@@ -75,11 +75,15 @@ extern List create_dynamic_block(List block_list,
}
memset(&blockreq, 0, sizeof(blockreq_t));
- slurm_mutex_lock(&block_state_mutex);
if (my_block_list) {
reset_ba_system(track_down_nodes);
itr = list_iterator_create(my_block_list);
while ((bg_record = list_next(itr))) {
+ if (bg_record->magic != BLOCK_MAGIC) {
+ error("got a block with bad magic");
+ list_remove(itr);
+ continue;
+ }
if (bg_record->free_cnt) {
if (bg_conf->slurm_debug_flags
& DEBUG_FLAG_BG_PICK) {
@@ -402,7 +406,6 @@ finished:
if (results)
list_destroy(results);
errno = rc;
- slurm_mutex_unlock(&block_state_mutex);
return new_blocks;
}
--
GitLab