[lxc-devel] [PATCH 1/1] cgroupfs: support older kernels without cgroup.clone_children

[Robert Vogelgesang]

Hi Serge,

your patch uses the same concept as mine, but it won't work, as noted
below.

If you can wait until Monday, I could send you a cleaned-up and tested
version of my patch; I have to do other things today and tomorrow.


On Wed, Jan 22, 2014 at 10:49:00AM -0600, Serge Hallyn wrote:
> When creating a cgroup, detect whether cgroup.clone_children
> exists.  If not, then manually copy the parent's cpuset.cpus
> and cpuset.mems values.
> 
> Signed-off-by: Serge Hallyn <serge.hallyn at ubuntu.com>
> ---
>  src/lxc/cgroup.c | 72 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 72 insertions(+)
> 
> diff --git a/src/lxc/cgroup.c b/src/lxc/cgroup.c
> index 4482b32..49f1ba1 100644
> --- a/src/lxc/cgroup.c
> +++ b/src/lxc/cgroup.c
> @@ -74,6 +74,7 @@ static int do_setup_cgroup_limits(struct lxc_handler *h, struct lxc_list *cgroup
>  static int cgroup_recursive_task_count(const char *cgroup_path);
>  static int count_lines(const char *fn);
>  static int handle_cgroup_settings(struct cgroup_mount_point *mp, char *cgroup_path);
> +static void setup_cpuset_if_needed(char **subsystems, char *path);
>  
>  static struct cgroup_ops cgfs_ops;
>  struct cgroup_ops *active_cg_ops = &cgfs_ops;
> @@ -723,6 +724,71 @@ static char *cgroup_rename_nsgroup(const char *mountpath, const char *oldname, p
>  	return newname;
>  }
>  
> +static long get_value(const char *dir, const char *file)
> +{
> +	FILE *f;
> +	char path[MAXPATHLEN];
> +	int ret, retv;
> +
> +	retv = snprintf(path, MAXPATHLEN, "%s/%s", dir, file);
> +	if (retv < 0 || retv >= MAXPATHLEN)
> +		return 0;
> +	f = fopen(path, "r");
> +	ret = fscanf(f, "%d", &retv);

This is not sufficient, because cpuset.cpus and cpuset.mems do not contain
plain decimals, but lists and ranges of decimals.  You have to use %s here.

I used lxc_read_from_file() to read the values, a buffer of size 128,
checked if the buffer was large enough, and errored out if it was not.


> +	fclose(f);
> +	if (ret != 1)
> +		return 0;
> +	return retv;
> +}
> +
> +static void set_value(const char *dir, const char *file, long v)
> +{
> +	FILE *f;
> +	char path[MAXPATHLEN];
> +	int retv;
> +
> +	retv = snprintf(path, MAXPATHLEN, "%s/%s", dir, file);
> +	if (retv < 0 || retv >= MAXPATHLEN)
> +		return;
> +	f = fopen(path, "w");
> +	fprintf(f, "%ld\n", v);
> +	fclose(f);
> +}

This could be replaced with lxc_write_to_file().

> +
> +static bool file_exists(const char *dir, const char *file)
> +{
> +	char path[MAXPATHLEN];
> +	struct stat sb;
> +	int ret;
> +
> +	ret = snprintf(path, MAXPATHLEN, "%s/%s", dir, file);
> +	if (ret < 0 || ret >= MAXPATHLEN)
> +		return true;
> +	ret = stat(path, &sb);
> +	return ret == 0;
> +}
> +
> +static void setup_cpuset_if_needed(char **subsystems, char *path)
> +{
> +	char *parentpath, *p;
> +	long v;
> +	
> +	if (!lxc_string_in_array("cpuset", (const char **) subsystems))
> +		return;
> +	if (file_exists(path, "cgroup.clone_children"))
> +		return;
> +	parentpath = strdup(path);
> +	if (!parentpath)
> +		return;
> +	if ((p = rindex(parentpath, '/')))
> +		*p = '\0';
> +	v = get_value(parentpath, "cpuset.mems");
> +	set_value(path, "cpuset.mems", v);
> +	v = get_value(parentpath, "cpuset.cpus");
> +	set_value(path, "cpuset.cpus", v);
> +	free(parentpath);
> +}
> +
>  /* create a new cgroup */
>  struct cgroup_process_info *lxc_cgroupfs_create(const char *name, const char *path_pattern, struct cgroup_meta_data *meta_data, const char *sub_pattern)
>  {
> @@ -898,6 +964,8 @@ struct cgroup_process_info *lxc_cgroupfs_create(const char *name, const char *pa
>  				if (r < 0)
>  					goto cleanup_from_error;
>  				info_ptr->created_paths[info_ptr->created_paths_count++] = current_entire_path;
> +				setup_cpuset_if_needed(info_ptr->hierarchy->subsystems,
> +						current_entire_path);

As Qiang Huang already wrote to the list, you need the full path here.


>  			} else {
>  				/* if we didn't create the cgroup, then we have to make sure that
>  				 * further cgroups will be created properly

What about this "else" code path, shouldn't setup_cpuset_if_needed()
be called here, too?

Note: This comment is followed by a call to handle_cgroup_settings()
that has the first argument wrong, as I already wrote to the list
yesterday in the "Containers do not start with lxc-1.0.0.beta2 on
RHEL-6.5" thread.

	Robert


> @@ -2039,8 +2107,12 @@ static int handle_cgroup_settings(struct cgroup_mount_point *mp,
>  	 */
>  	if (lxc_string_in_array("cpuset", (const char **)mp->hierarchy->subsystems)) {
>  		char *cc_path = cgroup_to_absolute_path(mp, cgroup_path, "/cgroup.clone_children");
> +		struct stat sb;
> +
>  		if (!cc_path)
>  			return -1;
> +		if (stat(cc_path, &sb) != 0 && errno == ENOENT)
> +			return 0;
>  		r = lxc_read_from_file(cc_path, buf, 1);
>  		if (r == 1 && buf[0] == '1') {
>  			free(cc_path);
> -- 
> 1.8.5.3