[Bcc to re@ because it would be good to have it fixed before 5.0]
Hi,
Pavlin Radoslavov recently discovered a problem that affects several
network drivers and tends to show up when running multicast routing.
The problem is that the multicast routing code calls iff_allmulti()
on the interfaces, which should enable them to receive all multicast
packets. The device independent code sets the IFF_ALLMULTI flag in
the ifp, and then calls the device-specific ioctl handler for
SIOCSIFFLAGS.
Individual drivers handle this call in the most various ways. Some
unconditionally reinitialize the hardware; some try to be smart and
only check which IFF_* flags have changed and perform appropriate
actions. And here is the problem -- several drivers forget to check
for a change in IFF_ALLMULTI, thus causing the change to be ignored
until some future action on the interface causes it (or its multicast
filter) to be reinitialized.
A list of broken which are broken and good is below. The 'broken'
list is worrysome as it includes a lot of new/high performance/wireless
cards.
Broken: dc mn sf sk ste ti tl xl an bge em gem gx ie lge sr
aue cue kue wi xe
Correct: de rl sis vr wb ar(?) cnw cs ed ep ex fe fxp
hme lnc my nge ray sbni sn tx txp vx wl
I believe the best way to fix this bug (in the *_ioctl handler,
for the SIOCSIFFLAGS case) is to follow the approach used by the
cs, ed, fe and vx drivers -- but if others have better suggestions
please let me know.
So, provided re@ approves, expect to see some commits to the drivers
in the 'broken' list related to this problem.
cheers
luigi
----------------------------------+-----------------------------------------
Luigi RIZZO, [EMAIL PROTECTED] . ICSI (on leave from Univ. di Pisa)
http://www.iet.unipi.it/~luigi/ . 1947 Center St, Berkeley CA 94704
Phone: (510) 666 2988
----------------------------------+-----------------------------------------
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-net" in the body of the message
