3Com 3c905C Tornado (newish 2.4.x) improper frame processing
A frame with a destination address of FF:00:00:00:00:00 shouldn't be processed, unless the card has been told to listen to that multicast address.
Driver/Card information:
-------------------
3c59x: Donald Becker and others. www.scyld.com/network/vortex.html 00:10.0: 3Com PCI 3c905C Tornado at 0xe000. Vers LK1.1.1600:10.0 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 78)
-------------------
I tried:
# ifconfig eth0 -mulicast
And STILL the frame was processed.
The linux/drivers/net/3c59x.c has this comment:
/* Pre-Cyclone chips have no documented multicast filter, so the only
multicast setting is to receive all multicast frames. At least the chip has a very clean way to set the mode, unlike many others. */
However, this card is Tornado, the latest rev, so this comment shouldn't apply.
The families go: Vortex -> Boomerang -> Cyclone -> Tornado
Also, I tried using the latest Donald Becker driver with the same results:
(dmesg output from latest Donald Becker driver)
3c59x.c:v0.99X 6/21/2002 Donald Becker, [email protected]
http://www.scyld.com/network/vortex.html
eth0: 3Com 3c905C Tornado at 0xe000, 00:01:03:de:57:37, IRQ 7
8K buffer 5:3 Rx:Tx split, autoselect/Autonegotiate interface. MII transceiver found at address 24, status 782d. Using bus-master transmits and whole-frame receives.
Quoted 11 lines · Dax Kelson
A frame with a destination address of FF:00:00:00:00:00 shouldn't be processed, unless the card has been told to listen to that multicast address.
Driver/Card information:
-------------------
3c59x: Donald Becker and others. www.scyld.com/network/vortex.html 00:10.0: 3Com PCI 3c905C Tornado at0xe000. Vers LK1.1.1600:10.0 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 78)
-------------------
There's a full implementation of hardware hashed filtering in 3com's GPL'ed 3c90x driver. But nobody got around to pilfering it.
-