Re: [PATCH 07/12] cfq-iosched: implement hierarchy-ready cfq_groupcharge scaling

From: Vivek Goyal
Date: Mon Dec 17 2012 - 16:49:07 EST

On Mon, Dec 17, 2012 at 01:33:14PM -0800, Tejun Heo wrote:
> Hello,
> On Mon, Dec 17, 2012 at 04:27:36PM -0500, Vivek Goyal wrote:
> > What I do care about is atleast being able to read and understand the
> > code easily. Right now, it is hard to understand. I am still struggling
> > to wrap my head around it.
> Hmm... I thought it was really simple. Maybe I'm just too familiar
> with it. You just walk up the tree multiplying the fraction you have
> at each level. It really doesn't get much simpler than that.
> > For example, while adding a group to service tree we calculate
> > cfqg->vfaction as follows.
> >
> > vfr = vfr * pos->leaf_weight / pos->level_weight;
> >
> > and then
> >
> > vfr = vfr * pos->weight / parent->level_weight;
> >
> > cfqg->vfraction = max_t(unsigned, vfr, 1)
> >
> > If cfqg->vfraction is about cfqg then why should we take into account
> > leaf_weight and level_weight. We should be just worried about pos->weight
> > and parent->level_weight and that should determine vfaction of cfqg.
> Eh? Then how would it compete with the children cfqgs? You can
> consider it as the hidden leaf node competing with the children cfqgs,
> right? So, you take the leaf weight and divide it by the total active
> weight of all the children (including the hidden leaf node). It isn't
> different from the rest of the calculation. It just looks different
> because leaf_weight is stored elsewhere and we look down there and
> then walke up. The calculation being done is exactly the same one.

Again it is coming from multiplexing cfqg and cfqg->task_group. So
effectively we are calculating the vfraction of task_group inside cfqg.
Once you say cfqg->vfraction, I immediately think of total share of
group (including task_group and all children cgroups).

May be cfqg->task_group_vfraction (or cfqg->tg_fraction) is a better name.

Also descrition says.

/* vfraction the fraction of vdisktime that a cfqg is entitled to */

May be we can clarify it that it is share of task_group of cfqg.

I guess i am too familiar with how cpu has done it. I think I am getting
confused with properties which belong to cfqg and properties which
belong to cfqg->task_group.

I guess if we prefix fields belong to task_group with somete suitable
string, it might become little easier to understand.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at