Containers: css_put() dilemma

Balbir Singh balbir at linux.vnet.ibm.com
Tue Jul 17 03:28:33 PDT 2007


Paul (??) Menage wrote:
> On 7/17/07, Balbir Singh <balbir at linux.vnet.ibm.com> wrote:
>> >
>> > >                mutex_lock(&container_mutex);
>> > >                set_bit(CONT_RELEASABLE, &cont->flags);
>> > >-               if (atomic_dec_and_test(&css->refcnt)) {
>> > >-                       check_for_release(cont);
>> > >-               }
>> > >+               check_for_release(cont);
>> > >                mutex_unlock(&container_mutex);
>> > >
> 
> I think that this isn't safe as it stands, without a synchronize_rcu()
> in container_diput() prior to the kfree(). Also, it will break if
> anyone tries to use a release agent on a hierarchy that has your
> memory controller bound to it.
> 


Isn't the code functionally the same as before? We still do atomic_test_and_dec()
as before. We still set_bit() CONT_RELEASABLE, we take the container_mutex
and check_for_release(). I am not sure I understand what changed?

Could you please elaborate as to why using a release agent is broken
when the memory controller is attached to it? I am not quite sure why we
need the synchronize_rcu() either in container_diput().



> Paul
> _______________________________________________
> Containers mailing list
> Containers at lists.linux-foundation.org
> https://lists.linux-foundation.org/mailman/listinfo/containers


-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL


More information about the Containers mailing list