Re: [PATCH v4] mm: oom: introduce cpuset oom

[Gang Li]

Hi,

On 2023/8/17 16:40, Gang Li wrote:
> On 2023/4/11 22:36, Michal Hocko wrote:
> > I believe it still wouldn't hurt to be more specific here.
> > CONSTRAINT_CPUSET is rather obscure. Looking at this just makes my head
> > spin.
> >          /* Check this allocation failure is caused by cpuset's wall 
> > function */
> >          for_each_zone_zonelist_nodemask(zone, z, oc->zonelist,
> >                          highest_zoneidx, oc->nodemask)
> >                  if (!cpuset_zone_allowed(zone, oc->gfp_mask))
> >                          cpuset_limited = true;
> > > Does this even work properly and why? prepare_alloc_pages sets
> > oc->nodemask to current->mems_allowed but the above gives us
> > cpuset_limited only if there is at least one zone/node that is not
> > oc->nodemask compatible. So it seems like this wouldn't ever get set
> > unless oc->nodemask got reset somewhere. This is a maze indeed.Is there
>
> In __alloc_pages:
> ```
> /*
>   * Restore the original nodemask if it was potentially replaced with
>   * &cpuset_current_mems_allowed to optimize the fast-path attempt.
>   */
> ac.nodemask = nodemask;
> page = __alloc_pages_slowpath(alloc_gfp, order, &ac);
>
> ```
>
> __alloc_pages set ac.nodemask back to mempolicy before call
> __alloc_pages_slowpath.
>
>
> > any reason why we cannot rely on __GFP_HARWALL here? Or should we
>
> In prepare_alloc_pages:
> ```
> if (cpusets_enabled()) {
>      *alloc_gfp |= __GFP_HARDWALL;
>      ...
> }
> ```
>
> Since __GFP_HARDWALL is set as long as cpuset is enabled, I think we can
> use it to determine if we are under the constraint of CPUSET.

We have two nodemasks: one from the parameters of __alloc_pages and
another from cpuset. If the node allowed by the parameters of
__alloc_pages is not allowed by cpuset, it means that this page
allocation is constrained by cpuset, and thus CONSTRAINT_CPUSET can be
returned.

I guess this piece of code is reasonable and we can keep the
code as it is.

Thanks,
Gang Li.