Re: [RFC PATCH 05/11] mm/mempolicy: modify set_mempolicy_home_node to take a task argument

On Wed 22-11-23 16:11:54, Gregory Price wrote:
[...]

+
+	/*
+	 * Behavior when task == current allows a task modifying itself
+	 * to bypass the check in get_task_mm and acquire the mm directly
+	 */
+	if (task == current) {
+		mm = task->mm;
+		mmget(mm);
+	} else
+		mm = get_task_mm(task);

Similar question as in the previous patch. When is a kthread legit to do
manipulate borrowed mm memory policy?

+
+	if (!mm)
+		return -ENODEV;

Is this an expected error? No mm means task dying and mm has been
already released. Should we simply return ESRCH wouldbe a better error
code IMO. This should never happen for the current task so it sould be
safe to add.

-- 
Michal Hocko
SUSE Labs