On Wed, Aug 22, 2018 at 04:12:13PM +0200, Michal Hocko wrote:
> On Tue 21-08-18 14:35:57, Roman Gushchin wrote:
> > If CONFIG_VMAP_STACK is set, kernel stacks are allocated
> > using __vmalloc_node_range() with __GFP_ACCOUNT. So kernel
> > stack pages are charged against corresponding memory cgroups
> > on allocation and uncharged on releasing them.
> >
> > The problem is that we do cache kernel stacks in small
> > per-cpu caches and do reuse them for new tasks, which can
> > belong to different memory cgroups.
> >
> > Each stack page still holds a reference to the original cgroup,
> > so the cgroup can't be released until the vmap area is released.
> >
> > To make this happen we need more than two subsequent exits
> > without forks in between on the current cpu, which makes it
> > very unlikely to happen. As a result, I saw a significant number
> > of dying cgroups (in theory, up to 2 * number_of_cpu +
> > number_of_tasks), which can't be released even by significant
> > memory pressure.
> >
> > As a cgroup structure can take a significant amount of memory
> > (first of all, per-cpu data like memcg statistics), it leads
> > to a noticeable waste of memory.
> >
> > Signed-off-by: Roman Gushchin <guro@xxxxxx>
> > Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
> > Cc: Michal Hocko <mhocko@xxxxxxxxxx>
> > Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> > Cc: Konstantin Khlebnikov <koct9i@xxxxxxxxx>
> > Cc: Tejun Heo <tj@xxxxxxxxxx>
> > Cc: Shakeel Butt <shakeelb@xxxxxxxxxx>
>
> Looks good to me. Two nits below.
>
> I am not sure stable tree backport is really needed but it would be nice
> to put
> Fixes: ac496bf48d97 ("fork: Optimize task creation by caching two thread stacks per CPU if CONFIG_VMAP_STACK=y")
>
> Acked-by: Michal Hocko <mhocko@xxxxxxxx>
Will add, thanks!
>
> > @@ -248,9 +253,20 @@ static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
> > static inline void free_thread_stack(struct task_struct *tsk)
> > {
> > #ifdef CONFIG_VMAP_STACK
> > - if (task_stack_vm_area(tsk)) {
> > + struct vm_struct *vm = task_stack_vm_area(tsk);
> > +
> > + if (vm) {
> > int i;
> >
> > + for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
> > + mod_memcg_page_state(vm->pages[i],
> > + MEMCG_KERNEL_STACK_KB,
> > + -(int)(PAGE_SIZE / 1024));
> > +
> > + memcg_kmem_uncharge(vm->pages[i],
> > + compound_order(vm->pages[i]));
>
> when do we have order > 0 here?
I guess, it's not possible, but hard-coded 1 looked a bit crappy.
Do you think it's better?
> Also I was wondering how come this
> doesn't blow up on partially charged stacks but both
> mod_memcg_page_state and memcg_kmem_uncharge check for page->mem_cgroup
> so this is safe. Maybe a comment would save people from scratching their
> heads.
Ok, will add.
Thank you!
