On Thu, 4 Aug 2022 15:36:48 +0200 Michal Hocko <mhocko@xxxxxxxx> wrote:
> On Thu 04-08-22 21:03:42, Feng Tang wrote:
> > Muchun Song found that after MPOL_PREFERRED_MANY policy was introduced
> > in commit b27abaccf8e8 ("mm/mempolicy: add MPOL_PREFERRED_MANY for multiple preferred nodes")
> > [1], the policy_nodemask_current()'s semantics for this new policy
> > has been changed, which returns 'preferred' nodes instead of 'allowed'
> > nodes, and could hurt the usage of its caller in hugetlb:
> > allowed_mems_nr().
>
> The acutal user visible effect description is missing here. AFAIU it
> would be this.
>
> With the changed semantic of policy_nodemask_current a taks with
> MPOL_PREFERRED_MANY policy could fail to get its reservation even though
> it can fall back to other nodes (either defined by cpusets or all online
> nodes) for that reservation failing mmap calles unnecessarily early.
>
> The fix is to not consider MPOL_PREFERRED_MANY for reservations at all
> because they, unlike MPOL_MBIND, do not pose any actual hard constrain.
And is this Fixes: b27abaccf8e8 ("mm/mempolicy: add MPOL_PREFERRED_MANY
for multiple preferred nodes")?
