Re: [PATCH 1/1] mm, oom_adj: don't loop through tasks in __set_oom_adj when not necessary

From: Michal Hocko
Date: Thu Aug 20 2020 - 08:41:39 EST


On Thu 20-08-20 13:42:56, Michal Hocko wrote:
> On Thu 20-08-20 13:30:23, Christian Brauner wrote:
[...]
> > trying to rely on set_bit() and test_bit() in copy_mm() being atomic and
> > then calling it where Oleg said after the point of no return.
>
> No objections.

Would something like the following work for you?

diff --git a/kernel/fork.c b/kernel/fork.c
index 9177a76bf840..25b83f0912a6 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1403,15 +1403,6 @@ static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
if (clone_flags & CLONE_VM) {
mmget(oldmm);
mm = oldmm;
- if (!(clone_flags & CLONE_SIGHAND)) {
- /* We need to synchronize with __set_oom_adj */
- mutex_lock(&oom_adj_lock);
- set_bit(MMF_PROC_SHARED, &mm->flags);
- /* Update the values in case they were changed after copy_signal */
- tsk->signal->oom_score_adj = current->signal->oom_score_adj;
- tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min;
- mutex_unlock(&oom_adj_lock);
- }
goto good_mm;
}

@@ -1818,6 +1809,19 @@ static __always_inline void delayed_free_task(struct task_struct *tsk)
free_task(tsk);
}

+static void copy_oom_score_adj(u64 clone_flags, struct task_struct *tsk)
+{
+ if ((clone_flags & (CLONE_VM | CLONE_THREAD | CLONE_VFORK)) == CLONE_VM) {
+ /* We need to synchronize with __set_oom_adj */
+ mutex_lock(&oom_adj_lock);
+ set_bit(MMF_PROC_SHARED, &mm->flags);
+ /* Update the values in case they were changed after copy_signal */
+ tsk->signal->oom_score_adj = current->signal->oom_score_adj;
+ tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min;
+ mutex_unlock(&oom_adj_lock);
+ }
+}
+
/*
* This creates a new process as a copy of the old one,
* but does not actually start it yet.
@@ -2290,6 +2294,8 @@ static __latent_entropy struct task_struct *copy_process(
trace_task_newtask(p, clone_flags);
uprobe_copy_process(p, clone_flags);

+ copy_oom_score_adj(clone_flags, p);
+
return p;

bad_fork_cancel_cgroup:
--
Michal Hocko
SUSE Labs