linux-kernel - Re: [PATCH] mm/memory-failure: fix crash in split_huge_page_to_list from soft_offline

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <aaff56ad-7bbe-8aa9-cd18-2ac9f4ea23f4@huawei.com>
Date: Mon, 22 Jan 2024 20:57:06 +0800
From: Miaohe Lin <linmiaohe@...wei.com>
To: Matthew Wilcox <willy@...radead.org>
CC: <naoya.horiguchi@....com>, <akpm@...ux-foundation.org>,
	<linux-mm@...ck.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] mm/memory-failure: fix crash in split_huge_page_to_list
 from soft_offline_page

On 2024/1/21 10:00, Matthew Wilcox wrote:
> On Sat, Jan 20, 2024 at 02:57:29PM +0800, Miaohe Lin wrote:
>> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
>> index 636280d04008..20058f7ac3e9 100644
>> --- a/mm/memory-failure.c
>> +++ b/mm/memory-failure.c
>> @@ -1377,8 +1377,13 @@ void ClearPageHWPoisonTakenOff(struct page *page)
>>   */
>>  static inline bool HWPoisonHandlable(struct page *page, unsigned long flags)
>>  {
>> -	/* Soft offline could migrate non-LRU movable pages */
>> -	if ((flags & MF_SOFT_OFFLINE) && __PageMovable(page))
>> +	/*
>> +	 * Soft offline could migrate non-LRU movable pages.
>> +	 * Note that page->mapping is overloaded with slab->slab_list or slabs
>> +	 * fields which might make slab pages appear like non-LRU movable pages.
>> +	 * So __PageMovable() has to be done after PageSlab() is checked.
>> +	 */
>> +	if ((flags & MF_SOFT_OFFLINE) && !PageSlab(page) && __PageMovable(page))
>>  		return true;
>>  
>>  	return PageLRU(page) || is_free_buddy_page(page);
> 
> I think would make more sense as
> 
> +	if (PageSlab(page))
> +		return false;

Do you mean add PageSlab check above "if ((flags & MF_SOFT_OFFLINE) && __PageMovable(page))" block
so we don't need to add more comment?

> 
> ... and then leave the rest alone (including not touching the comment)> .

I have a concern that __PageMovable() seems unreliable now if we access page from random context.
This might introduce some potential problems. For example, offline_pages() might be stumped with
such pages without any progress until signal occurs IIUC:

  offline_pages
    ..
    do {
      scan_movable_pages
        if (__PageMovable(page)) -- It might be slab page here. ret will also be set to 0.
          goto found;
      do_migrate_range -- Failed to isolate slab page and retry.
    } while (!ret) -- retry since ret is 0.

There might be many similar scenes, but I haven't taken them more closely. Maybe these are
just dumb problems...

Thanks.