Re: [PATCH v2] mm: optimize the redundant loop of mm_update_next_owner()

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 21 Jun 2024 10:50:10 +0200 Michal Hocko <mhocko@suse.com> wrote:

> On Thu 20-06-24 19:30:19, Oleg Nesterov wrote:
> > Can't review, I forgot everything about mm_update_next_owner().
> > So I am sorry for the noise I am going to add, feel free to ignore.
> > Just in case, I see nothing wrong in this patch.
> > 
> > I think this deserves a comment to explain that this is optimization
> > for the case we race with the pending mmput(). mm_update_next_owner()
> > checks mm_users at the start.
> > 
> > And. Can we drop tasklist and use rcu_read_lock() before for_each_process?
> > Yes, this will probably need more changes even if possible...
> > 
> > 
> > Or even better. Can't we finally kill mm_update_next_owner() and turn the
> > ugly mm->owner into mm->mem_cgroup ?
> 
> Yes, dropping the mm->owner should be a way to go. Replacing that by
> mem_cgroup sounds like an improvemnt. I have a vague recollection that
> this has some traps on the way. E.g. tasks sharing the mm but living in
> different cgroups. Things have changes since the last time I've checked
> and for example memcg charge migration on task move will be deprecated
> soon so chances are that there are less roadblocks on the way.

I think this was alexjlzheng's first kernel contribution and as such we
might not be hearing from him(?) again.  And that's OK, thanks for the
bug report - it helps Linux.

Meanwhile we have a stalled patch in mm-unstable.  If someone could add
this issue to their todo list, that would be great.