Received-SPF: pass (google.com: domain of linux-kernel+bounces-139652-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Subject: Re: [PATCH] mm/memory-failure: fix deadlock when
 hugetlb_optimize_vmemmap is enabled
To: Oscar Salvador <osalvador@suse.de>
CC: <akpm@linux-foundation.org>, <naoya.horiguchi@nec.com>,
	<linux-mm@kvack.org>, <linux-kernel@vger.kernel.org>
References: <20240407085456.2798193-1-linmiaohe@huawei.com>
 <ZhVMThr9TNeP6SWj@localhost.localdomain>
 <ZhVoatdJZ1RWu2r3@localhost.localdomain>
 <13aa38af-46a1-3894-32bd-c3eb6ef67359@huawei.com>
 <ZhZTZtzyMpMMowoD@localhost.localdomain>
From: Miaohe Lin <linmiaohe@huawei.com>
Message-ID: <247259aa-9c78-d1ae-c829-aa72adc75922@huawei.com>
Date: Thu, 11 Apr 2024 10:26:44 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Thunderbird/78.6.0
Precedence: bulk
MIME-Version: 1.0
In-Reply-To: <ZhZTZtzyMpMMowoD@localhost.localdomain>
Content-Type: text/plain; charset="utf-8"
Content-Language: en-US
Content-Transfer-Encoding: 7bit

On 2024/4/10 16:52, Oscar Salvador wrote:
> On Wed, Apr 10, 2024 at 03:52:14PM +0800, Miaohe Lin wrote:
>> AFAICS, iff check_pages_enabled static key is enabled and in hard offline mode,
>> check_new_pages() will prevent those pages from ending up in a PCP queue again
>> when refilling PCP list. Because PageHWPoison pages will be taken as 'bad' pages
>> and skipped when refill PCP list.
> 
> Yes, but check_pages_enabled static key is only enabled when
> either CONFIG_DEBUG_PAGEALLOC or CONFIG_DEBUG_VM are set, which means
> that under most of the systems that protection will not take place.
> 
> Which takes me to a problem we had in the past where we were handing
> over hwpoisoned pages from PCP lists happily.
> Now, with for soft-offline mode, we worked hard to stop doing that
> because soft-offline is a non-disruptive operation and no one should get 
> killed.
> hard-offline is another story, but still I think that extending the
> comment to include the following would be a good idea:
> 
> "Disabling pcp before dissolving the page was a deterministic approach
>  because we made sure that those pages cannot end up in any PCP list.
>  Draining PCP lists expels those pages to the buddy system, but nothing
>  guarantees that those pages do not get back to a PCP queue if we need
>  to refill those."

This really helps. Will add it in v2.
Thanks Oscar.

> 
>  Just to remind ourselves of the dangers of a non-deterministic
>  approach.
> 
> 
> Thanks
> 
>