linux-kernel - Re: [PATCH] Slab infrastructure for array operations

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: <20141219113113.477fd18f@redhat.com>
Date:	Fri, 19 Dec 2014 11:31:13 +0100
From:	Jesper Dangaard Brouer <brouer@...hat.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	Christoph Lameter <cl@...ux.com>, akpm@...uxfoundation.org,
	Steven Rostedt <rostedt@...dmis.org>,
	LKML <linux-kernel@...r.kernel.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	Linux Memory Management List <linux-mm@...ck.org>,
	Pekka Enberg <penberg@...nel.org>,
	Joonsoo Kim <js1304@...il.com>, brouer@...hat.com
Subject: Re: [PATCH] Slab infrastructure for array operations

On Thu, 18 Dec 2014 14:06:29 -0800 Andrew Morton <akpm@...ux-foundation.org> wrote:

> On Thu, 18 Dec 2014 10:33:23 -0600 (CST) Christoph Lameter <cl@...ux.com> wrote:
> 
> > This patch adds the basic infrastructure for alloc / free operations
> > on pointer arrays.
> 
> Please provide the justification/reason for making this change.

I agree that this needs more justification.

I (think) the reason behind this is a first step towards "bulk" alloc
and free.  And the reason behind that is to save/amortize the cost of
the locking/CAS operations.

> > Allocators must define _HAVE_SLAB_ALLOCATOR_OPERATIONS in their
> > header files in order to implement their own fast version for
> > these array operations.

I would like to see an implementation of a fast-version.  Else it is
difficult to evaluate if the API is the right one.  E.g. if it would be
beneficial for the MM system, we could likely restrict the API to only
work with power-of-two, from the beginning.

> Why?  What's driving this?

The network stack have a pattern of allocating 64 SKBs while pulling
out packets of the NICs RX-ring.  Packets are placing into the TX-ring,
and later at TX-completing time, we free up-to 256 SKBs (depending on
driver).

Another use-case, which need smaller bulk's, could be tree-structures
that need to expand, allocating two elems in one-shot should cut the
alloc overhead in half.

I'm implemented a prove-of-concept[1] lockless bulk alloc and free
scheme, that demonstrate this can benefit the network stack.  Now,
Christoph and I are trying to integrate some of the ideas into the slub
allocator.

[1] http://thread.gmane.org/gmane.linux.network/342347/focus=126138 
-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Sr. Network Kernel Developer at Red Hat
  Author of http://www.iptv-analyzer.org
  LinkedIn: http://www.linkedin.com/in/brouer
--
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/