linux-kernel - Re: [PATCH v5 00/11] Introduce huge zero page

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: <20121115072955.GA9387@otc-wbsnb-06>
Date:	Thu, 15 Nov 2012 09:29:55 +0200
From:	"Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	Andrea Arcangeli <aarcange@...hat.com>, linux-mm@...ck.org,
	Andi Kleen <ak@...ux.intel.com>,
	"H. Peter Anvin" <hpa@...ux.intel.com>,
	linux-kernel@...r.kernel.org,
	"Kirill A. Shutemov" <kirill@...temov.name>
Subject: Re: [PATCH v5 00/11] Introduce huge zero page

On Wed, Nov 14, 2012 at 01:33:42PM -0800, Andrew Morton wrote:
> On Wed,  7 Nov 2012 17:00:52 +0200
> "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com> wrote:
> 
> > Andrew, here's updated huge zero page patchset.
> 
> There is still a distinct lack of reviewed-by's and acked-by's on this
> patchset.
> 
> On 13 Sep, Andrea did indicate that he "reviewed the whole patchset and
> it looks fine to me".  But that information failed to make it into the
> changelogs, which is bad.

As I said before, I had to drop Andrea's reviewed-by on rebase to
v3.7-rc1. I had to solve few not-that-trivial conflicts and I was not sure
if the reviewed-by is still applicable.

> I grabbed the patchset.  I might hold it over until 3.9 depending on
> additional review/test feedback and upon whether Andrea can be
> persuaded to take another look at it all.
> 
> I'm still a bit concerned over the possibility that some workloads will
> cause a high-frequency free/alloc/memset cycle on that huge zero page. 
> We'll see how it goes...
> 
> For this reason and for general ease-of-testing: can and should we add
> a knob which will enable users to disable the feature at runtime?  That
> way if it causes problems or if we suspect it's causing problems, we
> can easily verify the theory and offer users a temporary fix.
> 
> Such a knob could be a boot-time option, but a post-boot /proc thing
> would be much nicer.

Okay, I'll add sysfs knob.

BTW, we already have build time knob: just revert last two patches in the
series. It will bring lazy allocation instead of refcounting.

-- 
 Kirill A. Shutemov

Download attachment "signature.asc" of type "application/pgp-signature" (837 bytes)