netdev - Re: [PATCH] sky2: skb recycling

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:	Tue, 21 Oct 2008 08:09:50 -0700
From:	Stephen Hemminger <shemminger@...tta.com>
To:	Eric Dumazet <dada1@...mosbay.com>
Cc:	Jeff Garzik <jgarzik@...ox.com>, netdev@...r.kernel.org
Subject: Re: [PATCH] sky2: skb recycling

On Tue, 21 Oct 2008 07:54:18 +0200
Eric Dumazet <dada1@...mosbay.com> wrote:

> Stephen Hemminger a écrit :
> > Add support for recycling tx buffers into receive buffers.
> > This is experimental at this point.
> > 
> 
> I really like this skb recycling
> 
> For best performance, driver should perform TX completion before RX completion, so that
> freshly added skb in recycle queue have a chance being reused right after. Apparently sky2
> tx handling is finegrained (events posted by NIC)

The NIC always reports TX completion after RX, so recycling doesn't work that well.

> But still, this mechanism also use more skbs per device, especially on dormant ones.
> 
> If your RX ring has 256 skb, then adding the recycle queue can adds 256 more skbs... One MB or event more...
> 
> Maybe we should flush the recycle queue, every 5 seconds or so, especially if device is dormant
> (no RX on it, litle heartbeat TX)

I think dropping the recycle queue down to the typical max number of receive frames per interrupt
which is between 4 and 16 would be enough.
--
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