Andreas Ferber wrote:
>
> Hi,
>
> I was very pleased that 802.1q VLAN tagging support has been included
> into the kernel recently :-)
>
> Though this is only part of the work needed.
>
> Let me start from the beginning, for those not familiar with 802.1q
> tagging. The tag adds an additional header field (4 Bytes) to the
> ethernet header, so that the maximum frame size is 1518 Octets instead
> of 1514 (with a standard MTU of 1500).
>
> The problem is that most network cards by default will flag those
> frames as oversized and drop them (though /sending/ tagged frames is
> not a problem with any card I know of). There are several
> possibilities to resolve this problem:
>
> - reduce the MTU on all servers connected to the VLAN (not always
> possible)
> - patch the NIC driver so that it doesn't drop oversized frames
> - set the MTU on the physical interface to 1504, which is a) not
> possible with every driver, and b) untagged frames sent via the
> interface will be oversized now.
>
> Below my signature you will find a patch that adds the necessary bits
> to the 3c59x.c driver. It sets up the NIC to receive VLAN tagged
> frames and (if supported by the NIC) correctly checksum them.
>
> Newer cards (those with HW checksumming support) have a register to
> configure the maximum frame size. Older cards don't have this
> register, but they have a flag that lets the NIC receive FDDI sized
> frames (4500 Octets IIRC).
>
> It was discussed on the VLAN mailinglist if (at least in the case of
> hardware like the old 3Com NICs where you can basically only disable
> checking the frame size) it is feasible to only enable the large
> frame receipt if VLAN tags are really used on the NIC. We have come to
> the conclusion that uncoditionally enabling it should not pose any
> performance penalties with any sane hardware implementation, so we
> decided to not introduce a VLAN enable callback into NIC drivers.
>
> Just to be sure, the 3c59x patch only enables the new parts if 802.1q
> support has been enabled in the kernel configuration. Testing proved
> the patch quite stable (no problem reports so far).
>
> There is no maintainer mentioned for the 3c59x driver in the
> MAINTAINERS file, so I send this directly to the list. Please consider
> the patch for kernel inclusion.
>
> Andreas
> --
> Andreas Ferber - dev/consulting GmbH - Bielefeld, FRG
> ---------------------------------------------------------
> +49 521 1365800 - af@devcon.net - www.devcon.net
>
> --- linux.orig/drivers/net/3c59x.c Sun Sep 30 21:26:06 2001
> +++ linux/drivers/net/3c59x.c Wed Oct 24 21:52:10 2001
> @@ -308,6 +308,9 @@
> code size of a per-interface flag is not worthwhile. */
> static char mii_preamble_required;
>
> +/* The Ethernet Type used for 802.1q tagged frames */
> +#define VLAN_ETHER_TYPE 0x8100
This needs to be ETH_P_8021Q from if_ether.h.
Have you tested this? I should think you would need a dev->change_mtu
also.
-- Jeff Garzik | Only so many songs can be sung Building 1024 | with two lips, two lungs, and one tongue. MandrakeSoft | - nomeansno- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Nov 07 2001 - 21:00:35 EST