xinglp wrote:
> 2013/1/10 Bruce Dubbs <bruce.du...@gmail.com>:
>> Bruce Dubbs wrote:
>>> xinglp wrote:
>>>> 2013/1/10 Bruce Dubbs <bruce.du...@gmail.com>:
>>>>> xinglp wrote:
>>>>>> If so, the script init-net-rules.sh in udev-lfs-197-1 also need some
>>>>>> change.
>>>>>>
>>>>>> DEVICES=$(eval echo /sys/class/net/{eth*,ath*,wlan*[0-9], ...
>>>>>> -->
>>>>>> DEVICES=$(eval echo /sys/class/net/{enp*,eth*,ath*,wlan*[0-9], ...
>>>>> What kind of network card do you have?  I've never heard of enp*
>>>> outputs of lspci
>>>> 00:19.0 Ethernet controller: Intel Corporation 82567LM Gigabit Network
>>>> Connection (rev 03)
>>>>          Subsystem: Hewlett-Packard Company Device 30e1
>>>>          Flags: bus master, fast devsel, latency 0, IRQ 42
>>>>          Memory at 90700000 (32-bit, non-prefetchable) [size=128K]
>>>>          Memory at 90727000 (32-bit, non-prefetchable) [size=4K]
>>>>          I/O ports at 20e0 [size=32]
>>>>          Capabilities: <access denied>
>>>>          Kernel driver in use: e1000e
>>>> It was named to eth0 when I use udev 196.
>>> See
>>> http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames
>>>
>>>
>>> We're going to need to figure this out.  The init-net-rules.sh may go
>>> away completely.  I'm just reading this now.

It should either go away completely, or we should disable the new setup.
:-)

For the record, the new default setup only works if (IIRC from the
thread on systemd-devel) the BIOS exports the slot number for the card.
Not all do (mine definitely does not -- the NIC sysfs directories do not
have a acpi_index or index attributes, and there are no entries in
/sys/bus/pci/slots at all).

From the wiki page, that eliminates 1) and 2).

3) is the same old by-path persistence, and because it's encoded into
the NIC name, it has the same length limit of 15 bytes.  So anything on
a USB bus just isn't going to work, both because it changes when you
change hub ports, and because too many hubs between the PCI device and
the NIC will make the name too long.

That leaves 4), by-mac, which only works for NICs that have unique MACs.

(FWIW I do think we should get rid of the NIC rename stuff, but I think
we should replace it with a shell script to look up an arbitrarily-long
identifier from the udev database instead, and translate that into an
eth* or wlan* or whatever name.  That ID can be a storage-style path, a
MAC, or any other unique identifier that the script knows how to
dereference by mapping to a udev property.  I threw something together a
few months back, but never submitted it to contrib/ or anything like
that.)

Attachment: signature.asc
Description: OpenPGP digital signature

-- 
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to