Messages in this thread | | | Date | Fri, 25 Jun 2010 01:45:41 -0400 | From | Ben Blum <> | Subject | [RFC] [PATCH v3 0/2] cgroups: implement moving a threadgroup's threads atomically with cgroup.procs |
| |
This patch series is a revision of http://lkml.org/lkml/2010/5/29/126 .
These patches use an rwlock in signal_struct which access is dependent on Oleg's recent changes to signal_struct's lifetime rules.
It is okay to write the tid of any task in the threadgroup. This is implemented by taking task->group_leader right after find_task_by_vpid while still rcu_read-side. This makes it necessary to check if thread_group_leader(leader) every time we want to iterate over ->thread_group; each of these checks can fail with -EAGAIN. Unfortunately this also means I needed to put these checks in can_attach for each subsystem that needs to check each thread in the group.
I handle EAGAIN in the file's write handler, since hey, it's a super expensive operation, might as well make it unbounded-time to boot - this is optional and would work just as well with -EAGAIN sent to userspace.
-- bblum
--- Documentation/cgroups/cgroups.txt | 13 - include/linux/cgroup.h | 15 - include/linux/init_task.h | 9 include/linux/sched.h | 10 kernel/cgroup.c | 449 +++++++++++++++++++++++++++++++++----- kernel/cgroup_freezer.c | 4 kernel/cpuset.c | 4 kernel/fork.c | 10 kernel/ns_cgroup.c | 4 kernel/sched.c | 4 10 files changed, 462 insertions(+), 60 deletions(-)
| |