Re: [PATCH] mm: drop hotplug lock from lru_add_drain_all

[Michal Hocko]

On Tue 14-11-17 15:13:27, Thomas Gleixner wrote:
> On Tue, 14 Nov 2017, Michal Hocko wrote:
> > From: Michal Hocko <mhocko@xxxxxxxx>
> > 
> > Pulling cpu hotplug locks inside the mm core function like
> > lru_add_drain_all just asks for problems and the recent lockdep splat
> > [1] just proves this. While the usage in that particular case might
> > be wrong we should prevent from locking as lru_add_drain_all is used
> > at many places. It seems that this is not all that hard to achieve
> > actually.
> > 
> > We have done the same thing for drain_all_pages which is analogous by
> > a459eeb7b852 ("mm, page_alloc: do not depend on cpu hotplug locks inside
> > the allocator"). All we have to care about is to handle
> >       - the work item might be executed on a different cpu in worker from
> >         unbound pool so it doesn't run on pinned on the cpu
> > 
> >       - we have to make sure that we do not race with page_alloc_cpu_dead
> >         calling lru_add_drain_cpu
> > 
> > the first part is already handled because the worker calls lru_add_drain
> > which disables preemption when calling lru_add_drain_cpu on the local
> > cpu it is draining. The later is true because page_alloc_cpu_dead
> > is called on the controlling CPU after the hotplugged CPU vanished
> > completely.
> > 
> > [1] http://lkml.kernel.org/r/089e0825eec8955c1f055c83d476@xxxxxxxxxx
> > 
> > Signed-off-by: Michal Hocko <mhocko@xxxxxxxx>
> > ---
> > Hi,
> > this has been posted as 2 patch series [1] previously. It turned out
> > that the first patch was simply broken and the second one could be
> > simplified because the irq disabling is just pointless. There were
> > no other objections so I am resending this patch which should remove
> > quite a large space of potential lockups as lru_add_drain_all is used
> > at many places so removing the hoptlug locking is a good thing in
> > general.
> > 
> > Can we have this merged or there are still some objections?
> 
> No objections. The explanation makes sense, but it might be worth to have a
> comment at lru_add_drain_all() which explains the protection rules.

Do you mean wrt. cpu hotplug? Something like

diff --git a/mm/swap.c b/mm/swap.c
index 8bfdcab9f83e..fe6d645e8536 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -688,6 +688,11 @@ static void lru_add_drain_per_cpu(struct work_struct *dummy)
 
 static DEFINE_PER_CPU(struct work_struct, lru_add_drain_work);
 
+/*
+ * Doesn't need any cpu hotplug locking because we do rely on per-cpu
+ * kworkers being shut down before our page_alloc_cpu_dead callback is
+ * executed on the offlined cpu
+ */
 void lru_add_drain_all(void)
 {
 	static DEFINE_MUTEX(lock);
-- 
Michal Hocko
SUSE Labs