cvs commit: src/sys/dev/re if_re.c

[Brian Somers]

I have a few of these bogus cards.

RELENG_5:
re0: <RealTek 8169S Single-chip Gigabit Ethernet> port 0xa000-0xa0ff mem 0xe3400000-0xe34000ff irq 16 at device 5.0 on pci0
miibus0: <MII bus> on re0
rgephy0: <RTL8169S/8110S media interface> on miibus0
rgephy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseTX, 1000baseTX-FDX, auto
re0: Ethernet address: 00:40:f4:b1:1c:85
re0: [GIANT-LOCKED]

-current:
re0: <RealTek 8169S Single-chip Gigabit Ethernet> port 0xd400-0xd4ff mem 0xf3800000-0xf38000ff irq 9 at device 11.0 on pci2
miibus0: <MII bus> on re0
rgephy0: <RTL8169S/8110S media interface> on miibus0
rgephy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseTX, 1000baseTX-FDX, auto
re0: Ethernet address: 00:40:f4:b1:10:da

They were sold as ``Airlink'' cards, but I haven't got the packaging any
more (just the driver floppy disks).

With HW checksumming enabled they not only got the odd checksum wrong
(we can all deal with that!), but on occasion they *truncated* packets
by 1-3 bytes and then checksummed them - in a repeatable way!!

I put one in a Windows box too, but I have no idea if HW checksumming is
enabled, let alone knowing if the card is working correctly.  It performs
a little better than a 100Mbit card....

On Sat, 15 Jan 2005 23:30:45 -0600 (CST), Mike Silbersack <silby at silby.com> wrote:
> 
> On Tue, 4 Jan 2005, John-Mark Gurney wrote:
> 
> > Olivier Houchard wrote this message on Wed, Jan 05, 2005 at 00:06 +0000:
> >>   Disable checksum offloading by default. It seems to produce corrupted packets
> >>   with some revisions of the chip (particularly when using multiple TX
> >>   descriptors).
> >
> > Ummm... do you have documentation on this?  i.e. PR numbers or message
> > id's, or even better, an errata from RealTek?
> >
> > Also, make sure you update the documentation...
> >
> > -- 
> >  John-Mark Gurney				Voice: +1 415 225 5579
> 
> Maybe, and I'm just throwing this out there... maybe when INVARIANTS is 
> compiled in, the kernel should re-check RX checksums, and print a warning 
> if the card goofed.  It'd be interesting to see if that showed previously 
> unknown bugs.
> 
> Mike "Silby" Silbersack


-- 
Brian Somers                                          <brian at Awfulhak.org>
Don't _EVER_ lose your sense of humour !               <brian at FreeBSD.org>