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;
From:   Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
To:     Oscar Salvador <osalvador@suse.de>
CC:     Michal Hocko <mhocko@kernel.org>,
        "mike.kravetz@oracle.com" <mike.kravetz@oracle.com>,
        "linux-mm@kvack.org" <linux-mm@kvack.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH v2 10/16] mm,hwpoison: Rework soft offline for free
 pages
Thread-Topic: [RFC PATCH v2 10/16] mm,hwpoison: Rework soft offline for free
 pages
Thread-Index: AQHVhPYxuKrdanK4Pk2pyFsBlimO86dfuBWAgATFroCAAC2WAIABDXAAgAALI4CAAAKLgIAADVAAgAAKCgCAAAdKgIAAAl2AgAEHHQA=
Date:   Wed, 23 Oct 2019 02:15:08 +0000
Message-ID: <20191023021508.GA29387@hori.linux.bs1.fc.nec.co.jp>
References: <20191018120615.GM5017@dhcp22.suse.cz>
 <20191021125842.GA11330@linux> <20191021154158.GV9379@dhcp22.suse.cz>
 <20191022074615.GA19060@linux> <20191022082611.GD9379@dhcp22.suse.cz>
 <20191022083505.GA19708@linux> <20191022092256.GH9379@dhcp22.suse.cz>
 <20191022095852.GB20429@linux> <20191022102457.GJ9379@dhcp22.suse.cz>
 <20191022103319.GA21736@linux>
In-Reply-To: <20191022103319.GA21736@linux>
Accept-Language: en-US, ja-JP
Content-Language: ja-JP
Content-Type: text/plain; charset="iso-2022-jp"
Content-ID: <E800C125A5CD5A43AD8AFB60C05422A0@gisp.nec.co.jp>
Content-Transfer-Encoding: 8BIT
MIME-Version: 1.0
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Oct 22, 2019 at 12:33:25PM +0200, Oscar Salvador wrote:
> On Tue, Oct 22, 2019 at 12:24:57PM +0200, Michal Hocko wrote:
> > Yes, that makes a perfect sense. What I am saying that the migration
> > (aka trying to recover) is the main and only difference. The soft
> > offline should poison page tables when not able to migrate as well
> > IIUC.
> 
> Yeah, I see your point.
> I do not really why soft-offline strived so much to left the page
> untouched unless it was able to content the problem.
> 
> Note that if we start now to poison pages even if we could not 
> content them (in soft-offline mode), that is a big and visible user
> change.

It's declared that soft offline never disrupts userspace by design,
so if poisoning page tables in migration failure, we could break this
and send SIGBUSs.  Then end users would complain that their processes
are killed by corrected (so non-urgent) errors.

Thanks,
Naoya Horiguchi