Hello,

On Thu, Mar 16, 2017 at 05:31:59PM +0100, Oleg Nesterov wrote:

Or we can add another "unsigned no_cgroups:1" bit into task_struct,
not sure.

To synchronize around initialization, a PF flag would be easier as we
can use wait_on_bit().

Anyway, I do not understand the PF_NO_SETAFFINITY check in
__cgroup_procs_write(). task_can_attach() checks it too, so cgroups
can't change the affinity. Imo something explicit like no_cgroups
makes more sense.

task_can_attach() predates the __cgroup_procs_write() and currently
doesn't do anything.  We can split the flag or rename it so that it's
more generic.  The reasons for disallowing cgroup migration have a lot
of crosssection with affinity, so it's not a complete misnomer.
Either way is fine by me.

Thanks.

-- 
tejun