From mboxrd@z Thu Jan  1 00:00:00 1970
From: Avi Kivity <avi@qumranet.com>
Subject: Re: [Qemu-devel] [PATCH 5/10] Add network announce function
Date: Wed, 10 Sep 2008 17:00:12 +0300
Message-ID: <48C7D2EC.3080001@qumranet.com>
References: <1220989802-13706-1-git-send-email-aliguori@us.ibm.com>	<1220989802-13706-6-git-send-email-aliguori@us.ibm.com> <48C7CB28.2090707@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: Chris Wright <chrisw@redhat.com>,
	Uri Lublin <uri.lublin@qumranet.com>,
	Anthony Liguori <aliguori@us.ibm.com>, kvm@vger.kernel.org
To: qemu-devel@nongnu.org
Return-path: <kvm-owner@vger.kernel.org>
Received: from il.qumranet.com ([212.179.150.194]:52104 "EHLO il.qumranet.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1751425AbYIJOAQ (ORCPT <rfc822;kvm@vger.kernel.org>);
	Wed, 10 Sep 2008 10:00:16 -0400
In-Reply-To: <48C7CB28.2090707@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Chris Lalancette wrote:
> Anthony Liguori wrote:
>   
>> +static int announce_self_create(uint8_t *buf, 
>> +				uint8_t *mac_addr)
>> +{
>> +    uint32_t magic = EXPERIMENTAL_MAGIC;
>> +    uint16_t proto = htons(ETH_P_EXPERIMENTAL);
>> +
>> +    /* FIXME: should we send a different packet (arp/rarp/ping)? */
>> +
>> +    memset(buf, 0xff, 6);         /* h_dst */
>> +    memcpy(buf + 6, mac_addr, 6); /* h_src */
>> +    memcpy(buf + 12, &proto, 2);  /* h_proto */
>> +    memcpy(buf + 14, &magic, 4);  /* magic */
>> +
>> +    return 18; /* len */
>> +}
>> +
>> +void qemu_announce_self(void)
>> +{
>> +    int i, j, len;
>> +    VLANState *vlan;
>> +    VLANClientState *vc;
>> +    uint8_t buf[256];
>> +
>> +    for (i = 0; i < nb_nics; i++) {
>> +        len = announce_self_create(buf, nd_table[i].macaddr);
>> +        vlan = nd_table[i].vlan;
>> +        for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
>> +            if (vc->fd_read == tap_receive)  /* send only if tap */
>> +                for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++)
>> +                    vc->fd_read(vc->opaque, buf, len);
>> +        }
>> +    }
>> +}
>> +
>>     
>
> This one is yucky, as the FIXME points out.  First, I'm guessing the point of
> this is to do an ARP poison on the switch?  If so, we probably want to do some
> kind of ARP packet, although I'm not an expert on this.

ARP is to announce an IP address moving to a different MAC address.  
Here the MAC address stays the same, but we want to announce to the 
switch that the MAC address is moving to a different port.

There may not even be an IP stack running in the guest, and we don't 
know the IP address(es) the guest is using, so I don't see how ARP fits 
here.

Since switches learn the MAC address to port association, all we need is 
to send _some_ packet.

>   Also, why are we trying
>  SELF_ANNOUNCE_ROUNDS times?  Is there some issue where some of these get
> dropped, or is it just being safe about it

Some switches drop the first packet; and if those switches are chained, 
we need to send multiple packets.

-- 
error compiling committee.c: too many arguments to function