From mboxrd@z Thu Jan 1 00:00:00 1970 From: jamal Subject: Re: patch: Action repeat Date: Sat, 30 Apr 2005 15:51:49 -0400 Message-ID: <1114890709.8929.147.camel@localhost.localdomain> References: <1114879817.8929.117.camel@localhost.localdomain> <4273BB30.1050402@trash.net> <4273BBAA.6060405@trash.net> <1114882045.8929.123.camel@localhost.localdomain> <4273CAB7.6080403@trash.net> Reply-To: hadi@cyberus.ca Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: netdev , "David S. Miller" Return-path: To: Patrick McHardy In-Reply-To: <4273CAB7.6080403@trash.net> Sender: netdev-bounce@oss.sgi.com Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Sat, 2005-30-04 at 20:13 +0200, Patrick McHardy wrote: > Since it only has such a short lifetime (action function sets it, > tcf_action_exec() clears it after changing classification result), > it would be less wasteful to just pass the classification result > to the actions. This would also avoid that skbs with tc_classid > already set can reach tcf_action_exec() (for example through mirred). > > What do you think? You mean not passing it back via skbs but through something else? What do you have in mind? It does sound distasteful for either changing the ->act() parametrization just so we can have a classid passed back or provide a spot for it in struct tc_action since only some actions will need to change it. I see the issue with classid leaking - perhaps specific actions could reset it when they steal packets? We should also reset it if the packet is stolen. cheers, jamal [In the long run really the classid setting should be per action setting; i.e when a user sets "flowid X:Y" they are explictly saying " action metaset class X:Y";]