Date: Fri, 15 Sep 2000 16:47:13 -0700 From: Nick Sayer <nsayer@sftw.com> To: freebsd-emulation@freebsd.org Cc: archie@freebsd.org Subject: Re: vmnet, bridging and netgraph Message-ID: <39C2B501.58EC338E@sftw.com> References: <39C29F98.F3A0FF53@sftw.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Nick Sayer wrote:
>
> Holy crap! It works!
Not quite. I was a bit overenthusiastic. There are a couple issues:
1. You pretty much have to do the ngctl steps in the vmware.sh file at
startup, because they have to run as root. Adding an suid binary to do
this stuff at run time would be a little overkill, IMHO.
2. The kernel spits out lots of these:
arp: 209.157.37.4 is on xl0 but got reply from 00:bd:5a:e6:4d:91 on
vmnet1
This is because ARP checks the source of the packet and sees the wrong
interface. When ng_bridge is sending packets to an upper hook, it should
make the packet look like it was received on the interface belonging to
the upper hook in question. Either that or the arp code has not yet been
sufficiently disconnected from the interfaces to let the upper and lower
hooks work the way ng_bridge wants them to. Archie? :-)
3. The promisc flag on vmnet1 does not seem to persist. Given how it
works, this may not matter.
4. I forgot the need to sent setpromisc 1 and setautosrc 0 messages when
setting this whole thing up.
5. I don't yet have a correct list of modules to load in vmware.sh to
make this work. I have tried loading just netgraph.ko, but this fails
the _first_ time I do all of the hookups, but not the second. Archie?
:-)
6. There is a long standing bug in tap(?) that for some bizarre reason
keeps my win2k guest from being able to do DHCP. The DHCP server sends
its reply to 255.255.255.255, and somehow tap never gets it. I'm not
sure why not. Real win2k machines have no problem DHCPing from this
server.
7. Those of you using ed interfaces may notice that there is a couple
#ifdef BRIDGE sections in the driver. Not sure what for.
>
> For those of you who have an up-to-date install, you can hose out all of
> the bridge stuff in your vmware.sh file and just change the exec line at
> the bottom of /usr/local/bin/vmware into this:
>
> ETHER= your ethernet interface (e.g. xl0)
>
> ngctl mkpeer vmnet1: bridge lower link0
> ngctl name vmnet1:lower vmnet_bridge
> ngctl connect vmnet_bridge: ${ETHER}: link1 lower
> ngctl connect vmnet_bridge: ${ETHER}: link1 upper
>
> /usr/local/lib/vmware/bin/vmware "$@"
>
> ngctl shutdown vmnet_bridge:
>
> No more having to change the kernel, no more worries about having the
> bridge turned on when vmware isn't running. It's all gravy.
>
> Note that this will probably NOT work if your primary Ethernet interface
> is a wi0 card. That's because those cards can't transmit Ethernet frames
> with source addresses different than their own. Sorry. Blame Lucent.
> They don't want you to avoid buying an access point to do
> wireless-to-LAN bridges.
>
> THREE CHEERS FOR ARCHIE!!!! HUZZAH! HUZZAH! HUZZAH!
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-emulation" in the body of the message
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-emulation" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?39C2B501.58EC338E>