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-next>] [day] [month] [year] [list]
Message-Id: <1329393696-4802-1-git-send-email-daniel.vetter@ffwll.ch>
Date:	Thu, 16 Feb 2012 13:01:35 +0100
From:	Daniel Vetter <daniel.vetter@...ll.ch>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	Intel Graphics Development <intel-gfx@...ts.freedesktop.org>,
	DRI Development <dri-devel@...ts.freedesktop.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Linux MM <linux-mm@...ck.org>,
	Daniel Vetter <daniel.vetter@...ll.ch>
Subject: [PATCH] extend prefault helpers to fault in more than PAGE_SIZE

Hi all,

drm/i915 has write/read paths to upload/download data to/from gpu buffer
objects. For a bunch of reasons we have special fastpaths with decent setup
costs, so when we fall back to the slow-path we don't fully recover to the
fastest fast-path when grabbing our locks again. This is also in parts due to
that we have multiple fallbacks to slower paths, so control-flow in our code
would get really ugly.

As part of a larger rewrite and re-tuning of these functions we've noticed that
the prefault helpers in pagemap.h only prefault up to PAGE_SIZE because that's
all the other users need. The follow-up patch extends this to abritary sizes so
that we can fully exploit our special fastpaths in more cases (we typically see
reads and writes of a few pages up to a few mb).

I'd like to get this in for 3.4. There's no functional dependency between this
patch and the drm/i915 rework (only performance effects), so this can go in
through -mm without causing merge issues.

If this or something similar isn't acceptable, plan B is to hand-roll these 2
functions in drm/i915. But I don't like nih these things in driver code much.

Comments highly welcome.

Yours, Daniel

For reference, the drm/i915 read/write rework is avaialable at:

http://cgit.freedesktop.org/~danvet/drm/log/?h=pwrite-pread

Unfortunately cgit.fd.o is currently on hiatus.

Daniel Vetter (1):
  mm: extend prefault helpers to fault in more than PAGE_SIZE

 include/linux/pagemap.h |   28 ++++++++++++++++++----------
 1 files changed, 18 insertions(+), 10 deletions(-)

-- 
1.7.7.5

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ