Re: [RFC PATCH 1/4] cgroups: Allow a cgroup subsys to reject a fork

[Frederic Weisbecker]

On Tue, Jun 21, 2011 at 10:39:04AM -0700, Paul Menage wrote:
> On Sun, Jun 19, 2011 at 4:51 PM, Frederic Weisbecker <fweisbec@xxxxxxxxx> wrote:
> > Make the cgroup subsystem fork callback return a value
> > so that subsystems are able to accept or reject a fork
> > completion with a custom error value.
> 
> This is unnecessary complexity in the cgroup subsystem (and seems to
> miss cleanup for subsystems that have previously had their fork()
> method return success).

No it seems only the freezer subsystem had this fork callback implemented.
But indeed it adds complexity because if a subsystem can cancel a fork,
then previous subsystems that called ->fork() would need to have a
kind cancel_fork() callback to call.

I haven't looked very deep but the freezer doesn't seem to need any
rollback. Future subsystems using the fork() callback may need it
though.

> If you want a subsystem to be able to reject a fork, I think it's
> better to have that subsystem be called explicitly from do_fork(), and
> keep that logic out of cgroups.

Ok I can do that.
--
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/