From owner-cvs-all Sun Aug 4 10:38:16 2002 Delivered-To: cvs-all@freebsd.org Received: from mx1.FreeBSD.org (mx1.FreeBSD.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 84E5C37B400; Sun, 4 Aug 2002 10:38:13 -0700 (PDT) Received: from iguana.icir.org (iguana.icir.org [192.150.187.36]) by mx1.FreeBSD.org (Postfix) with ESMTP id 3C08643E4A; Sun, 4 Aug 2002 10:38:13 -0700 (PDT) (envelope-from rizzo@iguana.icir.org) Received: (from rizzo@localhost) by iguana.icir.org (8.11.6/8.11.3) id g74Hc6211937; Sun, 4 Aug 2002 10:38:06 -0700 (PDT) (envelope-from rizzo) Date: Sun, 4 Aug 2002 10:38:06 -0700 From: Luigi Rizzo To: Archie Cobbs Cc: cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: cvs commit: src/sys/sys mbuf.h src/sys/kern uipc_mbuf.c Message-ID: <20020804103806.A11861@iguana.icir.org> References: <200208040536.g745aMb8037891@freefall.freebsd.org> <200208041715.g74HF6781436@arch20m.dellroad.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5.1i In-Reply-To: <200208041715.g74HF6781436@arch20m.dellroad.org>; from archie@dellroad.org on Sun, Aug 04, 2002 at 10:15:06AM -0700 Sender: owner-cvs-all@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Sun, Aug 04, 2002 at 10:15:06AM -0700, Archie Cobbs wrote: ... > > Now that we have this interface both in -current and -stable, each > > network device driver can replace the call to MGETHDR/MCLGET with > > calls to m_getcl(), saving (on -stable) about 400 bytes and the > > associated cache pollution. > > Cool! Are you planning to do a massive search & replace on all > existing instances of MGETHDR()/MCLGET() ... etc. ? I think that more or less, though the idea scares me :) Also because having to touch all drivers, I would also like to: + remove the code which strips off the MAC ethernet from packets before calling ether_input, given that in so many places (netgraph, bridging etc.) we have to put it back in place; + possibly, consider leaving some extra space at the beginning of the cluster before passing it to the driver, so if one has to implement a bridge/router which does encapsulation of any kind (including vlan), he does not need to prepend a new buffer for the additional header fields when there is so much empty space at the end. So my plan is to start tonight with the drivers i use most (fxp, dc, sis, rl, maybe xl) and then continue with the others as necessity arises. In other words, feel free to help if you have a way to test things, i am not putting any locks in this work :) cheers luigi To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message