linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Vinayak Menon <vinmenon@codeaurora.org>
To: Minchan Kim <minchan@kernel.org>
Cc: linux-mm@kvack.org
Subject: Re: [PATCH] mm: fix the race between swapin_readahead and SWP_SYNCHRONOUS_IO path
Date: Tue, 17 Sep 2019 11:08:49 +0530	[thread overview]
Message-ID: <4788d556-1b53-8d3e-121c-de2c286bac43@codeaurora.org> (raw)
In-Reply-To: <20190916200555.GA254094@google.com>


On 9/17/2019 1:35 AM, Minchan Kim wrote:
> Hi Vinayak,
>
> On Fri, Sep 13, 2019 at 02:35:41PM +0530, Vinayak Menon wrote:
>> On 9/12/2019 10:44 PM, Minchan Kim wrote:
>>> Hi Vinayak,
>>>
>>> On Wed, Sep 11, 2019 at 03:37:23PM +0530, Vinayak Menon wrote:
>>>
>>> < snip >
>>>
>>>>>> Can swapcache check be done like below, before taking the SWP_SYNCHRONOUS_IO path, as an alternative ?
>>>>> With your approach, what prevent below scenario?
>>>>>
>>>>> A                                                       B
>>>>>
>>>>>                                             do_swap_page
>>>>>                                             SWP_SYNCHRONOUS_IO && __swap_count == 1
>>>> As shrink_page_list is picking the page from LRU and B is trying to read from swap simultaneously, I assume someone had read
>>>>
>>>> the page from swap prior to B, when its swap_count was say 2 (for it to be reclaimed by shrink_page_list now)
>>> It could happen after B saw __swap_count == 1. Think about forking new process.
>>> In that case, swap_count is 2 and the forked process will access the page(it
>>> ends up freeing zram slot but the page would be swap cache. However, B process
>>> doesn't know it).
>>
>> Okay, so when B has read __swap_count == 1, it means that it has taken down_read on mmap_sem in fault path
>>
>> already. This means fork will not be able to proceed which needs to have down_write on parent's mmap_sem ?
>>
> You are exactly right. However, I still believe better option to solve
> the issue is to check swap_count and delte only if swap_count == 1
> in swap_slot_free_notify because it's zram specific issue and more safe
> without depending other lock scheme.


Sure. Let me know if you want me to post a patch for that.




  reply	other threads:[~2019-09-17  5:38 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-30 12:43 Vinayak Menon
2019-09-02 13:21 ` Michal Hocko
2019-09-03  6:13   ` Vinayak Menon
2019-09-03 11:41     ` Michal Hocko
2019-09-03 12:17       ` Vinayak Menon
2019-09-09  4:05         ` Vinayak Menon
2019-09-09 11:23           ` Michal Hocko
2019-09-09 23:26 ` Minchan Kim
2019-09-10  8:22   ` Vinayak Menon
2019-09-10 17:51     ` Minchan Kim
2019-09-11 10:07       ` Vinayak Menon
2019-09-12 17:14         ` Minchan Kim
2019-09-13  9:05           ` Vinayak Menon
2019-09-16 20:05             ` Minchan Kim
2019-09-17  5:38               ` Vinayak Menon [this message]
2019-09-18  1:12                 ` Minchan Kim

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4788d556-1b53-8d3e-121c-de2c286bac43@codeaurora.org \
    --to=vinmenon@codeaurora.org \
    --cc=linux-mm@kvack.org \
    --cc=minchan@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox