Re: [PATCH] mm/mempolicy: fix lock contention on mems_allowed

[Abel Wu]

On 8/9/22 8:11 PM, Michal Hocko Wrote:
> On Tue 09-08-22 18:49:27, Abel Wu wrote:
> > The mems_allowed field can be modified by other tasks, so it
> > isn't safe to access it with alloc_lock unlocked even in the
> > current process context.
>
> It would be useful to describe the racing scenario and the effect it
> would have. 78b132e9bae9 hasn't really explained thinking behind and why
> it was considered safe to drop the lock. I assume it was based on the
> fact that the operation happens on the current task but this is hard to
> tell.

Sorry for my poor description. Say there are two tasks: A from cpusetA
is performing set_mempolicy(2), and B is changing cpusetA's cpuset.mems.

    A (set_mempolicy)		B (echo xx > cpuset.mems)

    pol = mpol_new();
				update_tasks_nodemask(cpusetA) {
				  foreach t in cpusetA {
				    cpuset_change_task_nodemask(t) {
				      task_lock(t); // t could be A
    mpol_set_nodemask(pol) {
      new = f(A->mems_allowed);
				      update t->mems_allowed;
      pol.create(pol, new);
    }
				      task_unlock(t);
    task_lock(A);
    A->mempolicy = pol;
    task_unlock(A);
				    }
				  }
				}

In this case A's pol->nodes is computed by old mems_allowed, and could
be inconsistent with A's new mems_allowed.

While it is different when replacing vmas' policy: the pol->nodes is
gone wild only when current_cpuset_is_being_rebound():

    A (mbind)			B (echo xx > cpuset.mems)

				cpuset_being_rebound = cpusetA;
				update_tasks_nodemask(cpusetA) {
				  foreach t in cpusetA {
				    cpuset_change_task_nodemask(t) {
				      task_lock(t); // t could be A
    pol = mpol_new();
    mmap_write_lock(A->mm);
    mpol_set_nodemask(pol) {
      mask = f(A->mems_allowed);
				      update t->mems_allowed;
      pol.create(pol, mask);
    }
				      task_unlock(t);
				    }
    foreach v in A->mm {
      if (current_cpuset_is_being_rebound())
        pol.rebind(pol, cpuset.mems);
      v->vma_policy = pol;
    }
    mmap_write_unlock(A->mm);
				    mmap_write_lock(t->mm);
				    mpol_rebind_mm(t->mm);
				    mmap_write_unlock(t->mm);
				  }
				}
				cpuset_being_rebound = NULL;

In this case, the cpuset.mems, which has already done updating, is
finally used for calculating pol->nodes, rather than A->mems_allowed.
So it is OK to call mpol_set_nodemask() with alloc_lock unlocked when
doing mbind(2).

Best Regards,
Abel