Date: Thu, 20 Jan 2005 17:39:48 +0300 (MSK) From: Maxim Konovalov <maxim@macomnet.ru> To: Robert Watson <rwatson@freebsd.org> Cc: Charlie Schluting <charlie@schluting.com> Subject: Re: vlans changed? Message-ID: <20050120173819.Q62580@mp2.macomnet.net> In-Reply-To: <Pine.NEB.3.96L.1050120102946.84093B-100000@fledge.watson.org> References: <Pine.NEB.3.96L.1050120102946.84093B-100000@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 20 Jan 2005, 10:33-0000, Robert Watson wrote: > > On Wed, 19 Jan 2005, Charlie Schluting wrote: > > > Now, in 5.3, the only thing I can get working is to configure the em0 > > int with the IP, and set the trunk to have the native vlan corresponding > > to that IP. Weird. > > > > Also, is there a way to stop em(4) from stripping dot1q tags in > > hardware? I'd like to see them with tcpdump. What kind of a performance > > hit does this involve? > > Try "ifconfig em0 -vlanhwtag" and see if that helps. If not, take a look Yep, it works for Alex (CC'ed), thanks for the tip. > in if_em.c:em_setup_interface(), and you'll see two lines like this: > > #if __FreeBSD_version >= 500000 > ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU; > ifp->if_capenable |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU; > #endif > > Delete the contents "|FCAP_VLAN_HWTAGGING |" from each line, and that > should disable support for hardware vlan tagging and stripping in the > driver. There are several bugs relating to the handling of hardware vlan > tagging and promiscuous mode in both if_re and if_em. I had hoped to have > a chance to resolve them over the past couple of months but have not as > yet been able to do so. I measured a small performance hit last time I > tried disabling the hardware tagging, perhaps a couple of percent, but > mileage may vary -- for in-bound packets, there's a small amount > additional work, but for outgoing packets you may see an extra memory > allocation for each encapsulated packet (it depends a bit on what you > send). If this appears to work properly for you, we should probably > commit the change so that what's in the tree works properly, even if it's > slightly slower. IMO this is a good idea. -- Maxim Konovalov
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050120173819.Q62580>