linux-kernel - Re: [PATCH] mm/vmscan: restore zone_reclaim

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: <3ba94f19-3b18-9d52-a070-f652620c88e6@intel.com>
Date:   Mon, 29 Jun 2020 07:27:41 -0700
From:   Dave Hansen <dave.hansen@...el.com>
To:     Baoquan He <bhe@...hat.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>
Cc:     linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        ben.widawsky@...el.com, alex.shi@...ux.alibaba.com,
        dwagner@...e.de, tobin@...nel.org, cl@...ux.com,
        akpm@...ux-foundation.org, stable@...nel.org
Subject: Re: [PATCH] mm/vmscan: restore zone_reclaim_mode ABI

On 6/28/20 11:52 PM, Baoquan He wrote:
> On 06/25/20 at 05:34pm, Dave Hansen wrote:
>>
>> From: Dave Hansen <dave.hansen@...ux.intel.com>
>>
>> I went to go add a new RECLAIM_* mode for the zone_reclaim_mode
>> sysctl.  Like a good kernel developer, I also went to go update the
>> documentation.  I noticed that the bits in the documentation didn't
>> match the bits in the #defines.
>>
>> The VM evidently stopped caring about RECLAIM_ZONE at some point (or
>> never cared) and the #define itself was later removed as a cleanup.
> 
>>>From git history, it seems to never care about the RECLAIM_ZONE bit.
> 
> I think this patch is justified. I have one question about adding back
> the RECLAIM_ZONE bit. Since we introduced RECLAIM_ZONE in the first
> place, but never use it, removing it truly may fail some existing
> script, does it mean we will never have chance to fix/clean up this kind
> of mess?

Our #1 rule is "don't break userspace".  We only break userspace when we
have no other choice.

This case a bit fuzzier because we don't know if anyone is depending on
the ignored bit to do anything.  But, it *was* documented.  To me, that
means it might have been used, even though it would have been a
"placebo" bit.

> Do we have possibility to remove it in mainline tree, let distos or
> stable kernel maintainer take care of the back porting? Like this, any
> stable kernel after 5.8, or any distrols which chooses post v5.8 kenrel
> as base won't have this confusion. I am not objecting this patch, just
> be curious if we have a way to fix/clean up for this type of issue.

The only way I can plausibly think of "cleaning up" the RECLAIM_ZONE bit
would be to raise our confidence that it is truly unused.  That takes
time, and probably a warning if we see it being set.  If we don't run
into anybody setting it or depending on it being set in a few years, we
can remove it.

Backporting a _warning_ into the -stable trees might be an interesting
way to find users of older kernels mucking with it.