linux-kernel - Re: x86: should clear_user() have alternatives?

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: <YgJnhB+bAxoNsiSB@zn.tnic>
Date:   Tue, 8 Feb 2022 13:52:32 +0100
From:   Borislav Petkov <bp@...en8.de>
To:     Hugh Dickins <hughd@...gle.com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        linux-kernel@...r.kernel.org, x86@...nel.org
Subject: Re: x86: should clear_user() have alternatives?

Hi Hugh,

On Mon, Feb 07, 2022 at 09:45:36PM -0800, Hugh Dickins wrote:
> I realize that dd'ing from /dev/zero to /dev/null, and sparse files on
> tmpfs, are not prime candidates for optimization; and I've no idea how
> much clear_user() normally gets used for long clears.

Right, we usually don't take such "optimizations" because the folks who
send them always come up with either microbenchmarks or only test on a
single machine.

> If I were capable of compiler asm, with alternatives, and knew at what
> length ERMS becomes advantageous when clearing, I would be sending you
> a proper patch.  As it is, I'm hoping to tempt someone else to do the
> work!  Or reject it as too niche to bother with.

Yap, looking at arch/x86/lib/clear_page_64.S - that's straight-forward
asm without special-cases noodles like __copy_user_nocache, for example,
so I wouldn't be opposed to someone

 - remodelling it so that you can have clear_user* variants there too,
 with the length supplied so that you can call a common function with
 arbitrary length and clear_page* can call it too. And then call them in
 a clear_user() version just like the clear_page() one which selects the
 proper target based on CPU feature flags.

 - testing this on bunch of modern machines with, say, a kernel build or
 some sensible benchmark so that we at least have some coverage

If the numbers are worth it - and judging by your quick testing above
they should be - then I don't mind taking that at all.

If only someone would have the time and diligence to do it properly...

:-)

Thx.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette