Add an option "cgroup_nodelete" to not delete cgroups after job completion

o Add an option cgroup_nodelete to not remove cgroups created by fio after
  the job completion. This can help a user in inspecting various cgroup
  files after fio job completion.

Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

5 files changed, 28 insertions, 4 deletions

diff --git a/HOWTO b/HOWTO
index 6c7f05b9..922db2f0 100644
--- a/HOWTO
+++ b/HOWTO
@@ -1036,6 +1036,12 @@ cgroup_weight=int	Set the weight of the cgroup to this value. See
 		the documentation that comes with the kernel, allowed values
 		are in the range of 100..1000.
 
+cgroup_nodelete=bool Normally fio will delete the cgroups it has created after
+		the job completion. To override this behavior and to leave
+		cgroups around after the job completion, set cgroup_nodelete=1.
+		This can be useful if one wants to inspect various cgroup
+		files after job completion. Default: false
+
 uid=int		Instead of running as the invoking user, set the user ID to
 		this value before the thread/process does any work.
 
diff --git a/cgroup.c b/cgroup.c
index 14bbc57b..0e3a71eb 100644
--- a/cgroup.c
+++ b/cgroup.c
@@ -14,6 +14,7 @@ static struct fio_mutex *lock;
 struct cgroup_member {
 	struct flist_head list;
 	char *root;
+	unsigned int cgroup_nodelete;
 };
 
 static char *find_cgroup_mnt(struct thread_data *td)
@@ -43,14 +44,16 @@ static char *find_cgroup_mnt(struct thread_data *td)
 	return mntpoint;
 }
 
-static void add_cgroup(const char *name, struct flist_head *clist)
+static void add_cgroup(struct thread_data *td, const char *name,
+			struct flist_head *clist)
 {
 	struct cgroup_member *cm;
 
 	cm = smalloc(sizeof(*cm));
 	INIT_FLIST_HEAD(&cm->list);
 	cm->root = smalloc_strdup(name);
-
+	if (td->o.cgroup_nodelete)
+		cm->cgroup_nodelete = 1;
 	fio_mutex_down(lock);
 	flist_add_tail(&cm->list, clist);
 	fio_mutex_up(lock);
@@ -65,7 +68,8 @@ void cgroup_kill(struct flist_head *clist)
 
 	flist_for_each_safe(n, tmp, clist) {
 		cm = flist_entry(n, struct cgroup_member, list);
-		rmdir(cm->root);
+		if (!cm->cgroup_nodelete)
+			rmdir(cm->root);
 		flist_del(&cm->list);
 		sfree(cm->root);
 		sfree(cm);
@@ -144,7 +148,7 @@ int cgroup_setup(struct thread_data *td, struct flist_head *clist, char **mnt)
 			goto err;
 		}
 	} else
-		add_cgroup(root, clist);
+		add_cgroup(td, root, clist);
 
 	if (td->o.cgroup_weight) {
 		if (write_int_to_file(td, root, "blkio.weight",
diff --git a/fio.1 b/fio.1
index 5d0988b2..12146c87 100644
--- a/fio.1
+++ b/fio.1
@@ -762,6 +762,12 @@ your system doesn't have it mounted, you can do so with:
 Set the weight of the cgroup to this value. See the documentation that comes
 with the kernel, allowed values are in the range of 100..1000.
 .TP
+.BI cgroup_nodelete \fR=\fPbool
+Normally fio will delete the cgroups it has created after the job completion.
+To override this behavior and to leave cgroups around after the job completion,
+set cgroup_nodelete=1. This can be useful if one wants to inspect various
+cgroup files after job completion. Default: false
+.TP
 .BI uid \fR=\fPint
 Instead of running as the invoking user, set the user ID to this value before
 the thread/process does any work.
diff --git a/fio.h b/fio.h
index 5038e4de..5dc0352a 100644
--- a/fio.h
+++ b/fio.h
@@ -281,6 +281,7 @@ struct thread_options {
 	 */
 	char *cgroup;
 	unsigned int cgroup_weight;
+	unsigned int cgroup_nodelete;
 
 	unsigned int uid;
 	unsigned int gid;
diff --git a/options.c b/options.c
index 974df334..994f2a19 100644
--- a/options.c
+++ b/options.c
@@ -1818,6 +1818,13 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 		.maxval	= 1000,
 	},
 	{
+		.name	= "cgroup_nodelete",
+		.type	= FIO_OPT_BOOL,
+		.off1	= td_var_offset(cgroup_nodelete),
+		.help	= "Do not delete cgroups after job completion",
+		.def	= "0",
+	},
+	{
 		.name	= "uid",
 		.type	= FIO_OPT_INT,
 		.off1	= td_var_offset(uid),