Re: [PATCH 2/7] mm: shrinker: Add a .to_text() method for shrinkers

[Roman Gushchin]

On Tue, Nov 28, 2023 at 02:23:36PM +0800, Qi Zheng wrote:
> 
> 
> On 2023/11/28 11:53, Kent Overstreet wrote:
> > On Tue, Nov 28, 2023 at 11:27:11AM +0800, Muchun Song wrote:
> > > 
> > > 
> > > > On Nov 25, 2023, at 08:30, Kent Overstreet <kent.overstreet@xxxxxxxxx> wrote:
> > > > 
> > > > On Fri, Nov 24, 2023 at 11:08:11AM +0800, Qi Zheng wrote:
> > > > > Hi Kent,
> > > > > 
> > > > > On 2023/11/24 05:24, Kent Overstreet wrote:
> > > > > > On Thu, Nov 23, 2023 at 11:32:59AM +0800, Qi Zheng wrote:
> > > > > > > > + void (*to_text)(struct seq_buf *, struct shrinker *);
> > > > > > > 
> > > > > > > The "to_text" looks a little strange, how about naming it
> > > > > > > "stat_objects"?
> > > > > > 
> > > > > > The convention I've been using heavily in bcachefs is
> > > > > > typename_to_text(), or type.to_text(), for debug reports. The
> > > > > 
> > > > > OK.
> > > > > 
> > > > > > consistency is nice.
> > > > > 
> > > > > However, this is inconsistent with the name style of other
> > > > > shrinker callbacks. Please use the "objects" suffix. As for
> > > > > bcachefs's own callback function, you can use typename_to_text()
> > > > > to ensure consistency.
> > > > 
> > > > That would be inconsistent with introducing a convention to the wider
> > > > kernel.
> > > > 
> > > 
> > > I don not think .to_text is a good name. I really do not know what it means
> > > when I first look at this name. I knew you want to report the objects of
> > > shrinks, so why not use .report_objects or stat_objects proposed by Qi.
> > > Although .to_text is only used by bcachefs now, shrinker is a general module
> > > which is not only serving the bcachefs itself. I think it should be better
> > > to use a more straightforward name.
> > 
> > No, .report_objects or .stat_objects would be wrong; this isn't
> > generating a report on the objects owned by the shrinker, it's just a
> > report on the statistics of the shrinker itself.
> 
> Now I think adding this method might not be a good idea. If we allow
> shrinkers to report thier own private information, OOM logs may become
> cluttered. Most people only care about some general information when
> troubleshooting OOM problem, but not the private information of a
> shrinker.

I agree with that.

It seems that the feature is mostly useful for kernel developers and it's easily
achievable by attaching a bpf program to the oom handler. If it requires a bit
of work on the bpf side, we can do that instead, but probably not. And this
solution can potentially provide way more information in a more flexible way.

So I'm not convinced it's a good idea to make the generic oom handling code
more complicated and fragile for everybody, as well as making oom reports differ
more between kernel versions and configurations.

Thanks!