On Mon, Nov 08, 2021 at 01:19:57PM -0800, Mina Almasry wrote:
> On remote ooms (OOMs due to remote charging), the oom-killer will attempt
> to find a task to kill in the memcg under oom, if the oom-killer
> is unable to find one, the oom-killer should simply return ENOMEM to the
> allocating process.
>
> If we're in pagefault path and we're unable to return ENOMEM to the
> allocating process, we instead kill the allocating process.
>
> Signed-off-by: Mina Almasry <almasrymina@xxxxxxxxxx>
>
> Cc: Michal Hocko <mhocko@xxxxxxxx>
> Cc: Theodore Ts'o <tytso@xxxxxxx>
> Cc: Greg Thelen <gthelen@xxxxxxxxxx>
> Cc: Shakeel Butt <shakeelb@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Hugh Dickins <hughd@xxxxxxxxxx>
> Cc: Roman Gushchin <songmuchun@xxxxxxxxxxxxx>
> Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
> Cc: Hugh Dickins <hughd@xxxxxxxxxx>
> Cc: Tejun Heo <tj@xxxxxxxxxx>
> Cc: Vladimir Davydov <vdavydov.dev@xxxxxxxxx>
> Cc: Muchun Song <songmuchun@xxxxxxxxxxxxx>
> Cc: riel@xxxxxxxxxxx
> Cc: linux-mm@xxxxxxxxx
> Cc: linux-fsdevel@xxxxxxxxxxxxxxx
> Cc: cgroups@xxxxxxxxxxxxxxx
>
> ---
> mm/memcontrol.c | 21 +++++++++++++++++++++
> mm/oom_kill.c | 21 +++++++++++++++++++++
> 2 files changed, 42 insertions(+)
>
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 2e4c20d09f959..fc9c6280266b6 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2664,6 +2664,27 @@ int mem_cgroup_get_name_from_sb(struct super_block *sb, char *buf, size_t len)
> return ret < 0 ? ret : 0;
> }
>
> +/*
> + * Returns true if current's mm is a descendant of the memcg_under_oom (or
> + * equal to it). False otherwise. This is used by the oom-killer to detect
> + * ooms due to remote charging.
> + */
> +bool is_remote_oom(struct mem_cgroup *memcg_under_oom)
> +{
> + struct mem_cgroup *current_memcg;
> + bool is_remote_oom;
> +
> + if (!memcg_under_oom)
> + return false;
> +
> + current_memcg = get_mem_cgroup_from_mm(current->mm);
> + is_remote_oom =
> + !mem_cgroup_is_descendant(current_memcg, memcg_under_oom);
> + css_put(¤t_memcg->css);
> +
> + return is_remote_oom;
You'll be probably better with mem_cgroup_from_task(current) within an rcu read
section?
> +}
> +
> /*
> * Set or clear (if @memcg is NULL) charge association from file system to
> * memcg. If @memcg != NULL, then a css reference must be held by the caller to
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> index 0a7e16b16b8c3..556329dee273f 100644
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -1106,6 +1106,27 @@ bool out_of_memory(struct oom_control *oc)
> }
>
> select_bad_process(oc);
> +
> + /*
> + * For remote ooms in userfaults, we have no choice but to kill the
> + * allocating process.
> + */
> + if (!oc->chosen && is_remote_oom(oc->memcg) && current->in_user_fault &&
> + !oom_unkillable_task(current)) {
> + get_task_struct(current);
> + oc->chosen = current;
> + oom_kill_process(
> + oc, "Out of memory (Killing remote allocating task)");
> + return true;
> + }
> +
> + /*
> + * For remote ooms in non-userfaults, simply return ENOMEM to the
> + * caller.
> + */
> + if (!oc->chosen && is_remote_oom(oc->memcg))
> + return false;
> +
> /* Found nothing?!?! */
> if (!oc->chosen) {
I'd move both if's here:
if (is_remote_oom(oc->memcg)) {
if (current->in_user_fault && !oom_unkillable_task(current)) {
...
}
return false;
}
> dump_header(oc, NULL);
> --
> 2.34.0.rc0.344.g81b53c2807-goog
