From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH] skbuff: Add new tc classify variable Date: Wed, 08 Feb 2012 10:21:56 +0100 Message-ID: <1328692916.2284.5.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> References: <1328639948-25232-1-git-send-email-siwu@hrz.tu-chemnitz.de> <20120207.135841.1753473725075272902.davem@davemloft.net> <20120207201633.GB26533@pandem0nium> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , netdev@vger.kernel.org, Pablo Neira Ayuso , Patrick McHardy , Jamal Hadi Salim , Johannes Berg , "John W. Linville" , Marek Lindner , Sven Eckelmann , netfilter-devel@vger.kernel.org, netfilter@vger.kernel.org, coreteam@netfilter.org, linux-wireless@vger.kernel.org To: Simon Wunderlich Return-path: In-Reply-To: <20120207201633.GB26533@pandem0nium> Sender: netfilter-devel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Le mardi 07 f=C3=A9vrier 2012 =C3=A0 21:16 +0100, Simon Wunderlich a =C3= =A9crit : > Hello David, >=20 > On Tue, Feb 07, 2012 at 01:58:41PM -0500, David Miller wrote: > > From: Simon Wunderlich > > Date: Tue, 7 Feb 2012 19:39:08 +0100 > >=20 > > > The linux traffic control mechanism has different ways to select = the > > > correct class of a qdisc. A common way to do this is to use tc fi= lters > > > that are directly attached to a qdisc. Another approach is to use= the > > > iptables classify module. The latter one can reduce the amount of= work > > > necessary to process a packet when iptables is already involved i= n the > > > packet classification. > >=20 > > Do not bloat up sk_buff any more. Add this, and the other existing > > tc_* members to the qdisc SKB control block instead. > >=20 >=20 > Thanks for your feedback! >=20 > I guess you mean skb->cb, but this is also used within mac80211 for v= arious things > (quoting include/net/mac80211.h): >=20 > * struct ieee80211_tx_info - skb transmit information > * > * This structure is placed in skb->cb for three uses: > * (1) mac80211 TX control - mac80211 tells the driver what to do > * (2) driver internal use (if applicable) > * (3) TX status information - driver tells mac80211 what happened >=20 > We could give it a try, but we most probably run into conflicts again= =2E >=20 > I've messed up the CCs in my initial mail (and just resent it) - sorr= y about that. > Maybe the mac80211 guys have a suggestion as well :) >=20 I really dont understand what the conflict might be. As long as skb is in Qdisc layer, Qdisc owns skb->cb[] (a part of it actually, see current discussions about IB on netdev) As soon as packet is given to device (mac80211), cb[] can be reused. Or are you saying tc_class also might be needed by mac80211 in the future ? -- To unsubscribe from this list: send the line "unsubscribe netfilter-dev= el" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html