OpenVZ Forum


Home » Mailing lists » Devel » [RFC][patch 0/11][CFQ-cgroup]Yet another I/O bandwidth controlling subsystem for CGroups based on CF
RE: [RFC][patch 8/11][CFQ-cgroup] Control cfq_data per cgroup [message #29060 is a reply to message #29041] Fri, 04 April 2008 06:20 Go to previous messageGo to previous message
Satoshi UCHIDA is currently offline  Satoshi UCHIDA
Messages: 32
Registered: April 2008
Member
\> 
> On Tue, Apr 1, 2008 at 2:38 AM, Satoshi UCHIDA <s-uchida@ap.jp.nec.com>
> wrote:
> >  +
> >  +static void *cfq_cgroup_init_cfq_data(struct cfq_cgroup *cfqc, struct
> cfq_data *cfqd)
> >  +{
> >  +       struct cgroup *child;
> >  +
> >  +       /* setting cfq_data for cfq_cgroup */
> >  +       if (!cfqc) {
> >  +               cfqc =
> cgroup_to_cfq_cgroup(get_root_subsys(&cfq_cgroup_subsys));
> 
> Rather than adding get_root_subsys(), can't you just keep a reference
> locally to your root subsystem state?
> 


If the cfqc has not specific address, namely cfqc is null, 
the cfq_cgroup_init_cfq_data function is called when new device is plugged.
In this time, it needs to relate new cfq_data data with top cfq_cgroup data.
Probably, a running program will be "insmod" in its time.
However, its program is not in root group.

On the supposition that only current interface is usedm,
If using current process, the top cgroup can be calculated by

   task_cgroup(current, cfq_subsys_id)->top_cgroup .

Therefore cfq_cgroup data of top cgroup is calculated by

   cgroup_to_cfq_cgroup(task_cgroup(current, cfq_subsys_id)->top_cgroup) .


However,  It would be bad to use "current" variable in order to
calculate cfq_cgroup data of top cgroup.
Because relationship between "calculating cfq_cgroup data of top cgroup" and
"running(current) task" is weak.


So that you say, root subsystem state maybe keep a reference locally.
For example, create a variable for root subsystem state and
store the pointer when making subsystem state first.

However, I think that it is smart to calculate root group of subsystems
when needing its information.
Does the current code have any problem?


Satoshi UCHIDA.

_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: [PATCH 0/13 net-2.6.26] UDP/ICMP/TCP for a namespace v2
Next Topic: [RFC][PATCH 1/4] Provide a new procfs interface to set next id
Goto Forum:
  


Current Time: Sat Sep 06 22:12:53 GMT 2025

Total time taken to generate the page: 0.10371 seconds