Does anybody have good ideas about how to fix this case?
Everything in the udevinfo output is identical except:
- ATTR{link_mode}=="1"
+ ATTR{link_mode}=="0"
- ATTR{carrier}=="1"
- ATTR{dormant}=="0"
- ATTR{flags}=="0x1003"
+ ATTR{flags}=="0x1002"
(The first one is the "real" interface.)
----- Forwarded message from Celejar <celejar@xxxxxxxxx> -----
From: Celejar <celejar@xxxxxxxxx>
To: Debian Bug Tracking System <submit@xxxxxxxxxxxxxxx>
Subject: Bug#499868: udev: multiple mac80211 interfaces created on the same
physical device are incorrectly renamed
Package: udev
Version: 0.125-6
Severity: normal
mac80211 drivers allow multiple interfaces to be created on the same physical
device, so the underlying wmaster0 device can have eth0, eth1 and eth2. When
the driver is first loaded, a udev rule like this is written:
# PCI device 0x14e4:0x4318 (b43)
SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", ATTR{address}=="xx:xx:xx:xx:xx:xx", ATTR{type}=="1", KERNEL=="wlan*", NAME="eth0"
(originally 'NAME="wlan0"', I changed it to eth0)
When I subsequently add an interface with
iw dev wmaster0 interface add eth1 type station
a udev rule like this is added:
# PCI device 0x14e4:0x4318 (b43)
SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", ATTR{address}=="xx:xx:xx:xx:xx:xx", ATTR{type}=="1", KERNEL=="eth**", NAME="eth1"
Then, if I add another interface with
iw dev wmaster0 interface add eth2 type station
it gets caught by the existing rule and I get the infamous 'eth2_rename'
instead of 'eth2'.
I have seen this behavior with both b43 and ath5k. Perhaps creating multiple
ethX interfaces is a corner case, but the documentation should at least warn
about it.
----- End forwarded message -----
--
ciao,
Marco
Attachment:
signature.asc
Description: Digital signature
