Re: [PATCH 08/24] proc: pick out a function to iterate task children

[Oleg Nesterov <oleg@redhat.com>]

On 07/17, Andrew Vagin wrote:
>
> On Tue, Jul 14, 2015 at 08:02:35PM +0200, Oleg Nesterov wrote:
> > On 07/06, Andrey Vagin wrote:
> > >
> > > -static struct pid *
> > > -get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
> > > +static struct task_struct *
> > > +task_next_child(struct task_struct *parent, struct task_struct *prev, unsigned int pos)
> > >  {
> >
> > I won't really argue, just a question...
> >
> > So this patch changes it to accept/return task_struct rather pid. Why?
> > it is better to get/put "struct pid" only, not the whole task_struct.
> >
> > If another caller want task_struct, the necessary conversion is simple.
>
> Another caller wants task_struct.
>
> Currently this function receives pid and converts it into task_struct, then
> gets the next child and returns its pid.

Exactly because we try to avoid get_task_struct() if possible.

> So I try to avoid extra
> conversion in task_diag code.

Which is simple. And perhaps even iter->task can actually be iter->pid.

But as I said I won't really argue. And just in case, I personally think
this series makes sense, although I can't review the netlink interface.

Oleg.