linux-kernel - Re: [PATCH 1/3] Introduce & use new SG page iterators

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: <20160509095134.GD8414@nuc-i3427.alporthouse.com>
Date:	Mon, 9 May 2016 10:51:34 +0100
From:	Chris Wilson <chris@...is-wilson.co.uk>
To:	Dave Gordon <david.s.gordon@...el.com>
Cc:	intel-gfx@...ts.freedesktop.org, Imre Deak <imre.deak@...el.com>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/3] Introduce & use new SG page iterators

On Mon, May 09, 2016 at 10:33:31AM +0100, Dave Gordon wrote:
> The existing for_each_sg_page() iterator is somewhat inconvenient; in
> particular, the 'nents' parameters is not expressed in any useful way,
> and so there is no way to get a precise (maximum) number of iterations
> (and therefore pages) without knowing that the SGL has been constructed
> in a specific way.
> 
> So here we introduce for_each_sgt_page(), which simplifies the
> parameters down to just two: the iterator and an sg_table to iterate
> over. This is ideal where we simply need to process *all* the pages,
> regardless of how many there might be (e.g. put_pages()).

It seems to be a lot of churn for something that doesn't address my core
complaint with the iterator: it is too slow and *is* one of the
rate-limiting steps in some heavy workloads (games where we have to
rewrite the page tables frequently as their textures do not fit within
the GTT).
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre