From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Subject: Re: [PATCH net v2 2/6] bnxt_en: do not allow wildcard matches for L2
 flows
Date: Wed, 11 Apr 2018 16:31:52 -0400
Message-ID: <20180411203152.GD33938@C02RW35GFVH8.dhcp.broadcom.net>
References: <1523461818-15774-1-git-send-email-michael.chan@broadcom.com>
 <1523461818-15774-3-git-send-email-michael.chan@broadcom.com>
 <20180411114303.6f927c45@cakuba.netronome.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: Michael Chan <michael.chan@broadcom.com>, davem@davemloft.net,
        netdev@vger.kernel.org
To: Jakub Kicinski <kubakici@wp.pl>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mail-qt0-f181.google.com ([209.85.216.181]:37780 "EHLO
        mail-qt0-f181.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1756345AbeDKUb4 (ORCPT
        <rfc822;netdev@vger.kernel.org>); Wed, 11 Apr 2018 16:31:56 -0400
Received: by mail-qt0-f181.google.com with SMTP id w12so3538276qti.4
        for <netdev@vger.kernel.org>; Wed, 11 Apr 2018 13:31:56 -0700 (PDT)
Content-Disposition: inline
In-Reply-To: <20180411114303.6f927c45@cakuba.netronome.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On Wed, Apr 11, 2018 at 11:43:14AM -0700, Jakub Kicinski wrote:
> On Wed, 11 Apr 2018 11:50:14 -0400, Michael Chan wrote:
> > @@ -764,6 +788,41 @@ static bool bnxt_tc_can_offload(struct bnxt *bp, struct bnxt_tc_flow *flow)
> >  		return false;
> >  	}
> >  
> > +	/* Currently source/dest MAC cannot be partial wildcard  */
> > +	if (bits_set(&flow->l2_key.smac, sizeof(flow->l2_key.smac)) &&
> > +	    !is_exactmatch(flow->l2_mask.smac, sizeof(flow->l2_mask.smac))) {
> > +		netdev_info(bp->dev, "Wildcard match unsupported for Source MAC\n");
> 
> This wouldn't be something to do in net, but how do you feel about
> using extack for messages like this?
> 

I agree 'net' would not have been the place for a change like that, but
I do think that would be a good idea.  It looks like we could easily
change the ndo_setup_tc to something like this:

        int                     (*ndo_setup_tc)(struct net_device *dev,
                                                enum tc_setup_type type,
                                                void *type_data,
						struct netlink_ext_ack *extack);

It also looks like most of the callers of ndo_setup_tc have infra in
place to pass extack easily when the call is sourced from a netlink
message.   The others can just pass in NULL or define a local
netlink_ext_ack variable for short-term use.