linux-kernel - Re: [PATCH 4/4] mm/slab: move slab merge from slab

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 for Android: free password hash cracker in your pocket

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

Date:   Tue, 21 Nov 2023 11:03:04 +0100
From:   Vlastimil Babka <vbabka@...e.cz>
To:     Hyeonggon Yoo <42.hyeyoo@...il.com>, sxwjean@...com
Cc:     cl@...ux.com, penberg@...nel.org, rientjes@...gle.com,
        iamjoonsoo.kim@....com, roman.gushchin@...ux.dev, corbet@....net,
        linux-mm@...ck.org, linux-doc@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/4] mm/slab: move slab merge from slab_common.c to slub.c

On 11/21/23 09:54, Hyeonggon Yoo wrote:
> On Mon, Nov 20, 2023 at 6:13 PM <sxwjean@...com> wrote:
>>
>> From: Xiongwei Song <xiongwei.song@...driver.com>
>>
>> Since slab allocator has been removed. There is no users about slab
>> merge except slub. This commit is almost to revert
>> commit 423c929cbbec ("mm/slab_common: commonize slab merge logic").
>>
>> Also change all prefix of slab merge related functions, variables and
>> definitions from "slab/SLAB" to"slub/SLUB".
> 
> Could you please elaborate a little bit?
> I am not sure if I understand what the last two patches of this series
> are useful for.
> 
> - Why rename variable/function/macro names?
> - Why move merge related functions from slab_common.c to slub.c?

In my series I have moved functions that were part of allocation/free hot
paths as there should be performance benefits if they are all in the same
compilation unit.

>   (I mean merging slab_common.c and slub.c into single file might make sense
>    but why move only some parts of one into the other?)

OTOH slub.c becomes quite big, so I think it would make sense to not merge
mm/slab_common.c fully. The non-hot code that's handling e.g. the caches
creation and management, such as what this patch is moving, could certainly
stay away from mm/slub.c. We could just pick a more descriptive name for
slab_common.c.

I'd even investigate if more parts of slub.c could be split out (to a new
file/files) without compromising the hot paths, i.e. sysfs, debugging etc.