linux-kernel - Re: [RFC PATCH 0/2] remove SWAP_MAP

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] [day] [month] [year] [list]

Message-ID: <CAKEwX=NB-vf4zTTJ2KaRFGJmcfeDQpLLuiX=Rh6X+49ib8S=wA@mail.gmail.com>
Date: Wed, 25 Sep 2024 07:37:12 -0700
From: Nhat Pham <nphamcs@...il.com>
To: Baolin Wang <baolin.wang@...ux.alibaba.com>
Cc: Yosry Ahmed <yosryahmed@...gle.com>, akpm@...ux-foundation.org, hannes@...xchg.org, 
	hughd@...gle.com, shakeel.butt@...ux.dev, ryan.roberts@....com, 
	ying.huang@...el.com, chrisl@...nel.org, david@...hat.com, kasong@...cent.com, 
	willy@...radead.org, viro@...iv.linux.org.uk, baohua@...nel.org, 
	chengming.zhou@...ux.dev, linux-mm@...ck.org, kernel-team@...a.com, 
	linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 0/2] remove SWAP_MAP_SHMEM

On Tue, Sep 24, 2024 at 6:53 PM Baolin Wang
<baolin.wang@...ux.alibaba.com> wrote:
>
>
> One benefit I can mention is that removing 'SWAP_MAP_SHMEM' can help to
> batch free shmem swap entries in __swap_entries_free(), similar to the
> commit bea67dcc5eea ("mm: attempt to batch free swap entries for
> zap_pte_range()") did, which can improve the performance of shmem mTHP
> munmap() function based on my testing.

Yeah, the problem with having an extraneous state is you have to
constantly check for it in code, and/or keep it in mind when you
develop things. I've been constantly having to check for this state
when I develop code around this area, and it gets old fast.

If we can use it to optimize something, I can understand keeping it.
But it just seems like dead code to me :)

My preference is to do this as simply as possible - add another case
(usage == 1, nr > 1, and we need to add swap continuations) in the
check in __swap_duplicate()'s first loop, and just WARN right there.

That case CANNOT happen UNLESS we introduce a bug, or have a new use
case. When we actually have a use case, we can always introduce
handling/fallback logic for that case.

Barry, Yosry, Baolin, Ying, how do you feel about this?