hme(4) broken on non-sparc64 systems in -current

[Andrew R. Reiter]

On Mon, 20 Mar 2006, Scott Long wrote:

:Joerg Wunsch wrote:
:> As Pyun YongHyeon wrote:
:> 
:> 
:> > How about backing out rev. 1.46(if_hme.c)?
:> 
:> 
:> > For the same patch sent to bard at OpenBSD I got a positive report
:> > so it's strange to me though.(brad at OpenBSD reported Rx checksum
:> > offload breakage on little endian systems.)
:> 
:> 
:> Yes, backing that out helps.  I'm not sure what this change was trying
:> to fix.  I've noticed before that tools like ethereal reported the
:> checksum as invalid but the traffic itself was unaffected.  Anyway, as
:> it was now, the traffic was blocked, so perhaps there's more than one
:> spot where this needs to be fixed?
:> 
:> I'll look a bit further into it tonight.  Thanks!
:> 
:
:When tx checksum offloading is enabled, you'll always get checksum errors when
:capturing the local end of the transmission.  That expected,
:since the checksum never gets computed until long after the bpf tap has
:captured it.
:
:Scott
:

Something to note (from the application side) is that many applications 
that use libnids don't offer the option to deal with cksum offloading.  
Things like dsniff will want you to turn offloading off since there is no 
simple way (unless you modify the source, which then weakens the 
reliabliity of the stream assembly) to deal with this.

Just an extra thought to thing of.

Cheers,
Andrew


--
arr at watson.org