On Tue, Sep 14, 2021 at 09:13:48AM +0200, Michal Hocko [off-list ref] wrote:

"
This object can consume up to 2 pages, syscall is sleeping one,
size and duration can be controlled by user, and this allocation
can be repeated by many thread at the same time.
"

It sounds like a problem, except it is not because? A worst case
scenario evaluation would be beneficial for example

AFAICS, the offending allocation is in place only during the duration of
the syscall. So it's basically O(#tasks).
Considering at least 2 pages for task_struct + 2 pages for kernel stack,
back of the envelope calculation gives me the footprint amplification is
<1.5.
The factor would IMO be interesting if it was >> 2 (from the PoV of
excessive (ab)use, fine-grained accounting seems to be currently
unfeasible due to performance impact).

The commit message can be more explicit about this but to the patch
Reviewed-by: Michal Koutný <mkoutny-IBi9RG/b67k@public.gmane.org>