Re: [PATCH] mm: workingset: ignore slab memory size when calculating shadows pressure

[Johannes Weiner <hannes@cmpxchg.org>]

On Wed, Sep 09, 2020 at 09:55:20AM -0700, Roman Gushchin wrote:
> On Wed, Sep 09, 2020 at 10:55:34AM -0400, Johannes Weiner wrote:
> > On Thu, Sep 03, 2020 at 04:00:55PM -0700, Roman Gushchin wrote:
> > > In the memcg case count_shadow_nodes() sums the number of pages in lru
> > > lists and the amount of slab memory (reclaimable and non-reclaimable)
> > > as a baseline for the allowed number of shadow entries.
> > > 
> > > It seems to be a good analogy for the !memcg case, where
> > > node_present_pages() is used. However, it's not quite true, as there
> > > two problems:
> > > 
> > > 1) Due to slab reparenting introduced by commit fb2f2b0adb98 ("mm:
> > > memcg/slab: reparent memcg kmem_caches on cgroup removal") local
> > > per-lruvec slab counters might be inaccurate on non-leaf levels.
> > > It's the only place where local slab counters are used.
> > 
> > Hm, that sounds like a bug tbh. We're reparenting the kmem caches and
> > the individual objects on the list_lru when a cgroup is removed -
> > shouldn't we also reparent the corresponding memory counters?
> 
> It's definitely an option too, the question is if the added code complexity
> is really worth it. I'd say no had we talk only about slab counters,
> but when we'll eventually start reparenting pagecache, we'll need to reparent
> other counters as well, so we'll need it anyway. So, ok, let's drop
> this patch for now.
> 
> > 
> > > 2) Shadow nodes by themselves are backed by slabs. So there is a loop
> > > dependency: the more shadow entries are there, the less pressure the
> > > kernel applies to reclaim them.
> > 
> > This effect is negligible in practice.
> > 
> > The permitted shadow nodes are a tiny percentage of memory consumed by
> > the cgroup. If shadow nodes make up a significant part of the cgroup's
> > footprint, or are the only thing left, they will be pushed out fast.
> > 
> > The formula is max_nodes = total_pages >> 3, and one page can hold 28
> > nodes. So if the cgroup holds nothing but 262,144 pages (1G) of shadow
> > nodes, the shrinker target is 32,768 nodes, which is 32,768 pages
> > (128M) in the worst packing case and 1,170 pages (4M) at best.
> > 
> > However, if you don't take slab into account here, it can evict shadow
> > entries with undue aggression when they are needed the most. If, say,
> > the inode or dentry cache explode temporarily and displace the page
> > cache, it would be a big problem to drop the cache's non-resident info
> > at the same time! This is when it's at its most important.
> 
> Just curious, have you seen this in the real life?

I have seen it with anon pages back when we targeted the page cache
instead of the total memory footprint. Especially in the context of
psi it missed thrashing situations, see:

commit 95f9ab2d596e8cbb388315e78c82b9a131bf2928
Author: Johannes Weiner <jweiner@fb.com>
Date:   Fri Oct 26 15:05:59 2018 -0700

    mm: workingset: don't drop refault information prematurely

I can't remember if I saw slabs doing the same, but it's equally
plausible.