netdev - Re: [PATCH net-next 09/11] sfc: Prefetch RX skb header area

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]

Date:	Thu, 20 Jun 2013 00:10:45 +0100
From:	Ben Hutchings <bhutchings@...arflare.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
CC:	David Miller <davem@...emloft.net>, <netdev@...r.kernel.org>,
	<linux-net-drivers@...arflare.com>
Subject: Re: [PATCH net-next 09/11] sfc: Prefetch RX skb header area

On Wed, 2013-06-19 at 15:56 -0700, Eric Dumazet wrote:
> On Wed, 2013-06-19 at 23:39 +0100, Ben Hutchings wrote:
> 
> > You can't use build_skb() if you put multiple RX buffers in a page.  See
> > <1365805319.2791.18.camel@...-desktop.uk.solarflarecom.com>.
> 
> Well, you use small skb (used to have 64 bytes headroom and you changed
> it to 128) and attach a frag on it, so anyway build_skb() brings
> nothing.
> 
> I suspect you have a win on workloads with flood of rx, but on moderate
> load sk_buff will be cold, so the cache adds latencies.
> 
> My feeling is that this kind of tradeoff should not be per driver.

Jon's testing showed a significant latency improvement.

In Linux 3.10 sfc has reduced UDP performance (both latency and
throughput) as we removed the support for skb pre-allocation in
preparation for the implementation for DMA scatter and recycling of
IOMMU-mapped buffers.  The RX changes in this series are intended to
recover that, and our benchmark results suggest that we succeeded.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html