Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200311123848.118638-1-shile.zhang@linux.alibaba.com>
 <20200401154217.GQ22681@dhcp22.suse.cz> <dfc0014a-9b85-5eeb-70ea-d622ccf5d988@redhat.com>
 <20200401160048.GU22681@dhcp22.suse.cz> <20200401160929.jwekhr24tb44odea@ca-dmjordan1.us.oracle.com>
 <20200401161243.GW22681@dhcp22.suse.cz> <20200401161810.xvqikca2x46yqrlx@ca-dmjordan1.us.oracle.com>
 <20200401162655.GX22681@dhcp22.suse.cz>
In-Reply-To: <20200401162655.GX22681@dhcp22.suse.cz>
From:   Pavel Tatashin <pasha.tatashin@soleen.com>
Date:   Wed, 1 Apr 2020 12:41:13 -0400
Message-ID: <CA+CK2bCGpG6kBjkGd-QP06kNtwezj8mW13Jdvbxs6ExzRaJSpQ@mail.gmail.com>
Subject: Re: [PATCH v3] mm: fix tick timer stall during deferred page init
To:     Michal Hocko <mhocko@kernel.org>
Cc:     Daniel Jordan <daniel.m.jordan@oracle.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        David Hildenbrand <david@redhat.com>,
        Shile Zhang <shile.zhang@linux.alibaba.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Kirill Tkhai <ktkhai@virtuozzo.com>,
        linux-mm <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Apr 1, 2020 at 12:26 PM Michal Hocko <mhocko@kernel.org> wrote:
>
> On Wed 01-04-20 12:18:10, Daniel Jordan wrote:
> > On Wed, Apr 01, 2020 at 06:12:43PM +0200, Michal Hocko wrote:
> > > On Wed 01-04-20 12:09:29, Daniel Jordan wrote:
> > > > On Wed, Apr 01, 2020 at 06:00:48PM +0200, Michal Hocko wrote:
> > > > > On Wed 01-04-20 17:50:22, David Hildenbrand wrote:
> > > > > > On 01.04.20 17:42, Michal Hocko wrote:
> > > > > > > This needs a double checking but I strongly believe that the lock can be
> > > > > > > simply dropped in this path.
> > > >
> > > > This is what my fix does, it limits the time the resize lock is held.
> > >
> > > Just remove it from the deferred intialization and add a comment that we
> > > deliberately not taking the lock here because abc
> >
> > I think it has to be a little more involved because of the window where
> > interrupts might allocate during deferred init, as Vlastimil pointed out a few
> > years ago when the change was made.
>
> I do not remember any details but do we have any actual real allocation
> failure or was this mostly a theoretical concern. Vlastimil? For your
> context we are talking about 3a2d7fa8a3d5 ("mm: disable interrupts while
> initializing deferred pages")

I do not remember seeing any real failures, this was a theoretical
window. So, we could potentially simply remove these locks until we
see a real boot failure in some interrupt thread. The allocation has
to be rather large as well.

Pasha