linux-kernel - Re: [PATCH v3 3/5] mm/memory-failure: improve memory failure action

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: <6928d21f-90c5-47ea-81e8-1838ad9ce618@oracle.com>
Date: Thu, 23 May 2024 10:38:10 -0700
From: Jane Chu <jane.chu@...cle.com>
To: Oscar Salvador <osalvador@...e.de>
Cc: linmiaohe@...wei.com, nao.horiguchi@...il.com, akpm@...ux-foundation.org,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 3/5] mm/memory-failure: improve memory failure
 action_result messages

On 5/22/2024 1:37 PM, Oscar Salvador wrote:

> On Tue, May 21, 2024 at 05:54:27PM -0600, Jane Chu wrote:
>> Added two explicit MF_MSG messages describing failure in get_hwpoison_page.
>> Attemped to document the definition of various action names, and made a few
>> adjustment to the action_result() calls.
>>
>> Signed-off-by: Jane Chu <jane.chu@...cle.com>
> This looks much better, thanks:
>
> Reviewed-by: Oscar Salvador <osalvador@...e.de>
>
> By the way, I was checking the block in memory_failure() that handles
> refcount=0 pages, concretely the piece of code that handles buddy pages.
>
> In there, if we fail to take the page off the buddy lists, we return
> MF_FAILED, but I really think we should be returning MF_IGNORED.

I guess you mean this code -
         if (has_extra_refcount(ps, p, false))
                 ret = MF_FAILED;
?

It appears in below code paths-
     hwpoison_user_mappings
       identify_page_state
         me_huge_page || me_swapcache_dirty || me_swapcache_clean
for LRU pages.

And for non-LRU
     if (!folio_test_lru(folio) && !folio_test_writeback(folio))
             goto identify_page_state;

My hunch is that the most common calling path would be: 
hwpoison_user_mappings has unmapped the page, then identify_page_state 
is called, but for some reason failed to take the page off the LRU.  The 
m-f() handler has isolated the page to avoid further MCE, so I think in 
general return MF_FAILED is okay.

That said, the line is not always clear, for example in the non-LRU 
case, where the m-f() handler may have done only a little, I guess I 
just need to let the case rest.

thanks,

-jane

>
> Thoughts?
>   
>