Re: [RFC] [PATCH v3 1/2] cgroups: read-write lock CLONE_THREADforking per threadgroup
From: KAMEZAWA Hiroyuki
Date: Mon Jun 28 2010 - 03:15:27 EST
On Fri, 25 Jun 2010 01:46:54 -0400
Ben Blum <bblum@xxxxxxxxxxxxxx> wrote:
> Adds functionality to read/write lock CLONE_THREAD fork()ing per-threadgroup
> From: Ben Blum <bblum@xxxxxxxxxxxxxx>
> This patch adds an rwsem that lives in a threadgroup's signal_struct that's
> taken for reading in the fork path, under CONFIG_CGROUPS. If another part of
> the kernel later wants to use such a locking mechanism, the CONFIG_CGROUPS
> ifdefs should be changed to a higher-up flag that CGROUPS and the other system
> would both depend on.
> This is a pre-patch for cgroups-procs-write.patch.
Hmm, at adding a new lock, please describe its semantics.
Following is my understanding, right ?
This rwsem is read-locked from cgroup_fork() to cgroup_post_fork().
Most of works for fork() are between cgroup_fork() and cgroup_post_for().
This means if sig->threadgroup_fork_lock is held, no new do_work() can
make progress in this process groups. This rwsem is held only when
CLONE_THREAD is in clone_flags. IOW, this rwsem is taken only at creating
What we can do with this:
By locking sig->threadgroup_fork_lock, a code can visit _all_ threads
in a process group witout any races. So, if you want to implement an
atomic operation against a process, taking this lock is an idea.
To implement an atomic process move in cgroup, we need this lock.
Why this implemantation:
Considering cgroup, threads in a cgroup can be under several different
cgroup. So, we can't implement lock in cgroup-internal, we use signal
By the way, IMHO, hiding lock in cgroup_fork() and cgroup_post_fork() doesn't
seem good idea. How about a code like this ?
We may have chances to move these lock to better position if cgroup is
an only user.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/