Re: [PATCH mm-unstable v1 1/4] mm: don't hold css->refcnt during traversal

[Yosry Ahmed <yosryahmed@google.com>]

On Thu, Jul 25, 2024 at 1:43 PM Johannes Weiner <hannes@cmpxchg.org> wrote:
>
> On Wed, Jul 24, 2024 at 07:02:11PM +0000, Kinsey Ho wrote:
> > To obtain the pointer to the saved memcg position, mem_cgroup_iter()
> > currently holds css->refcnt during memcg traversal only to put
> > css->refcnt at the end of the routine. This isn't necessary as an
> > rcu_read_lock is already held throughout the function.
> >
> > Remove css->refcnt usage during traversal by leveraging RCU.
>
> Eh, I don't know about this.
>
> RCU ensures that the css memory isn't freed.
>
> The tryget ensures that the css is still alive and valid.
>
> In this case, it just so happens that the sibling linkage is also rcu
> protected. But accessing random css members when the refcount is 0 is
> kind of sketchy. On the other hand, the refcount is guaranteed to be
> valid, and rcu + tryget is a common pattern.

To be fair, the documentation of css_next_descendant_pre() mentions
that the requirements are:
- Either cgroup_mutex or RCU lock is held.
- Both @pos and @root are accessible.
- @pos is a descendant of @root.

This reads to me like it is intentional that RCU protection is enough
for @pos and @root, and that the sibling linkage is RCU protected by
design. Perhaps we could clarify this further (whether at
css_next_descendant_pre(), or above the definition of the linkage
members).

>
> What does this buy us? The tryget is cheap.

mem_cgroup_iter() is not an easy function to follow, so I personally
appreciate the simplicity gains tbh.