Todo List

Group Basic infrastructure

add some propaganda what's so great on this approach... I personally think it is broken and confusing (see TODOs below).

Global cgroup_change_cgroup_flags: Determine thread-safeness and fix of not safe.

Global cgroup_change_cgroup_path: should this function be really public?

Global cgroup_change_cgroup_uid_gid_flags: Determine thread-safeness and fix if not safe.

Global cgroup_create_cgroup: what is ignore_ownership good for?

Global cgroup_create_cgroup_from_parent

what is this good for? Why the list of controllers added by cgroup_add_controller() is not used, like in cgroup_create_cgroup()? I can't crate subgroup of root group in just one hierarchy with this function!

what is ignore_ownership good for?

Global cgroup_delete_cgroup: what is ignore_migration good for? rmdir() will fail if tasks were not moved.

Global CGROUP_FILE_TYPE_OTHER: really?

Global cgroup_free_controllers: should this function be public???

Global cgroup_get_cgroup

what is this function good for? Why is not considered only the list of controllers attached by cgroup_add_controller()? What owners will return cgroup_get_uid_gid() if the group is in multiple hierarchies, each with different owner of tasks file?

Global cgroup_get_value_string

rephrase, it's too vague... How big is the buffer actually?

Global cgroup_init_rules_cache: add parameter with the filename?

Global cgroup_new_cgroup: suggest one preferred way, either "/foo" or "foo".

Global cgroup_walk_tree_begin: why base_level is not hidden in **handle?

Group Configuration file: add this description?: These functions are mostly intended to be used by internal libcgroup tools, however they are fully supported and applications can benefit from them.

Group group_iterators: not all iterators follow this pattern, e.g. cgroup_walk_tree_begin() can result both in a state that cgroup_walk_tree_end() is not needed and will sigsegv and in a state that cgroup_walk_tree_end() is needed to free allocated memory. Complete review is needed!

Group Initialization: this is very bad... There should be at least way how to refresh the cache and/or an option to refresh it automatically (does kernel provide any indication, when a filesystem is mounted/unmounted?). Dtto the cleanup on exit.