Re: How to make an Intel EtherExpress known to xe(4)?
Greg 'groggy' Lehey <[email protected]>
| Newsgroups | gmane.os.freebsd.devel.mobile |
|---|---|
| Organization | The FreeBSD Project |
| Message-ID | <[email protected]> |
On Monday, 27 January 2003 at 0:01:31 +0100, Joerg Wunsch wrote: > As Kevin Oberman wrote: > >> This is typical of receiving 802.1q tagged frames, but I have been >> getting this error on my current system with a Xircom RE-100 card >> and I am reasonably sure that i am not getting any tagged frames >> where it's located. > > Me too. The remote end is running a vx0 which i think couldn't > even emit 802.1q tags at all. > > It's a bug in the driver, or an oddity of the xe hardware, depending > on the point of view. Unlike many other devices, it seems these chips > do indeed deliver the FCS along with the packet, and set the length > appropriately. > > Browsing through the source reveals that our network code could > already handle this situation, only the driver needs to set the > M_HASFCS flag. The attached patch would do this. If nobody objects, > i'll commit that. > > The only remaining oddity is then that tcpdump displays the incoming > packet (and length) including the FCS. > > I verified that the xe(4) driver in 4.x (where the card worked > flawlessly so far) also experienced this behaviour, i. e. the upper > network layers obviously silently ignored the trailing FCS. Only now > in 5.x, the kernel complains about oversized packets and discards > them. OK, I've tried this one, and it now works on my machine. What would be the implication of using this code with a card which doesn't return the FCS? Greg -- See complete headers for address and phone numbers To Unsubscribe: send mail to [email protected] with "unsubscribe freebsd-mobile" in the body of the message