latest udev (0.114-2) problem

Wayne Topa

unread,

Aug 23, 2007, 12:30:15 PM8/23/07

to

I have an Athlon k7 box, running testing, as my gateway. Due to a new
system install, I had not upgraded in a few weeks. On the 21st I
decided that it was time and did the upgrade overnight. This morning
I rebooted the box to upgrade my sid partition and when I rebooted
testing, I had lost my my localnet connections.

Dmesg shows that the new udev package had decided to change eth0 to
eth1, for some reason. To use the localnet, I changed eth0 to eth1 in
/etc/network/interfaces, restarted networking and killed the firewall.
My localnet was back. I then rebooted and found that udev had now
changed the interface to eth3. This will be a PITA to users that don't
run their boxes 24/7.

I have not seen this posted yet so it may be something wrong here.
Has anyone else see this udev behavior?

Wayne

--
In English, every word can be verbed. Would that it were so in our
programming languages.
_______________________________________________________

--
To UNSUBSCRIBE, email to debian-us...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listm...@lists.debian.org

Mathias Brodala

unread,

Aug 23, 2007, 2:00:17 PM8/23/07

to

Hi Wayne.

Wayne Topa, 23.08.2007 18:21:

> I have an Athlon k7 box, running testing, as my gateway. Due to a new
> system install, I had not upgraded in a few weeks. On the 21st I
> decided that it was time and did the upgrade overnight. This morning
> I rebooted the box to upgrade my sid partition and when I rebooted
> testing, I had lost my my localnet connections.
>
> Dmesg shows that the new udev package had decided to change eth0 to
> eth1, for some reason. To use the localnet, I changed eth0 to eth1 in
> /etc/network/interfaces, restarted networking and killed the firewall.
> My localnet was back. I then rebooted and found that udev had now
> changed the interface to eth3. This will be a PITA to users that don't
> run their boxes 24/7.

This definitely doesn’t sound like udev were working correctly. Is there a
/etc/udev/rules.d/z25_persistent-net.rules file? This one should contain the
mapping of the ethernet devices based on their MAC address. Manual reordering
should be done here, the file usually never ever gets touched after initial
generation.

Regards, Mathias

--
debian/rules

signature.asc

Wayne Topa

unread,

Aug 23, 2007, 3:00:18 PM8/23/07

to

Mathias Brodala(in...@noctus.net) is reported to have said:
> Hi Wayne.
>
> Wayne Topa, 23.08.2007 18:21:
> > I have an Athlon k7 box, running testing, as my gateway. Due to a new
> > system install, I had not upgraded in a few weeks. On the 21st I
> > decided that it was time and did the upgrade overnight. This morning
> > I rebooted the box to upgrade my sid partition and when I rebooted
> > testing, I had lost my my localnet connections.
> >
> > Dmesg shows that the new udev package had decided to change eth0 to
> > eth1, for some reason. To use the localnet, I changed eth0 to eth1 in
> > /etc/network/interfaces, restarted networking and killed the firewall.
> > My localnet was back. I then rebooted and found that udev had now
> > changed the interface to eth3. This will be a PITA to users that don't
> > run their boxes 24/7.

Correction: udev stopped at eth2 NOT eth3. (I mistyped yet again) On a reboot
it came up again as eth2. So it is not as bad as I first thought.

>
> This definitely doesn't sound like udev were working correctly. Is there a
> /etc/udev/rules.d/z25_persistent-net.rules file? This one should contain the
> mapping of the ethernet devices based on their MAC address. Manual reordering
> should be done here, the file usually never ever gets touched after initial
> generation.

Yes there is. I've had problems with udev on systems that I was
trying out various interface cards (usb, pcmcia) on, so thats the first
file I checked. It is ignoring the, original, eth0 stansa that it was
not having a problem with and adding new eth interfaces.

Here it is, currently.
---------
# PCI device 10ec:8139 (8139too)
SUBSYSTEM=="net", DRIVER=="?*", SYSFS{address}=="00:48:54:d1:35:1c", NAME="eth0"

# USB device 2001:3c00 (rt2500usb)
#SUBSYSTEM=="net", DRIVERS=="?*", ATTRS{address}=="00:11:95:e6:19:12", ATTRS{type}=="1", NAME="wlan0"

# PCI device 0x10ec:0x8139 (8139too)
#SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:48:54:d1:35:1c", NAME="eth1"

# PCI device 0x10ec:0x8139 (8139too)
SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:48:54:d1:35:1c", NAME="eth2"
------

Prior to this udev version if I removed all the non existant
interfaces from the /etc/udev/rules.d/z25_persistent-net.rules file,
it would work as expected. Something changed in this version as that
no longer works.

I found a fix (?) with a google search but I can't try it until I
finish downloading a new kernel. It was on an Ubuntu Forum so just in
case I'll list it here. Not Tested Yet so take it with a grain of
salt.

----- Possible Fix ------------------
Create/Edit a file /etc/udev/rules.d/10-local.rules with the following content Code:

KERNEL="eth*", SYSFS{address}=="00:10:5a:33:44:55" NAME="eth0"
KERNEL="eth*", SYSFS{address}=="00:0c:44:55:66:77" NAME="eth1"

Just change the mac addresses to your own values from ifconfig.

----------------------------

Thanks for the reply!

Wayne

--
As far as we know, our computer has never had an undetected error.

Wackojacko

unread,

Aug 23, 2007, 3:30:15 PM8/23/07

to

Wayne

It looks as though udev is now detecting the card differently as the PCI
device now has a '0x' before 8139 so effectively the mapping for eth0 is
for a non existent card. I would try deleting all the eth* entries and
restart.

HTH

Wackojacko

Wayne Topa

unread,

Aug 23, 2007, 4:40:08 PM8/23/07

to

Wackojacko(wackoj...@ntlworld.com) is reported to have said:
> Wayne Topa wrote:
>> Mathias Brodala(in...@noctus.net) is reported to have said:
>>> Hi Wayne.
>>>
>>> Wayne Topa, 23.08.2007 18:21:
>>>> I have an Athlon k7 box, running testing, as my gateway. Due to a new
>>>> system install, I had not upgraded in a few weeks. On the 21st I
>>>> decided that it was time and did the upgrade overnight. This morning
>>>> I rebooted the box to upgrade my sid partition and when I rebooted
>>>> testing, I had lost my my localnet connections.

<--<snip>--> s

>> Here it is, currently.
>> ---------
>> # PCI device 10ec:8139 (8139too)
>> SUBSYSTEM=="net", DRIVER=="?*", SYSFS{address}=="00:48:54:d1:35:1c",
>> NAME="eth0"
>> # USB device 2001:3c00 (rt2500usb)
>> #SUBSYSTEM=="net", DRIVERS=="?*", ATTRS{address}=="00:11:95:e6:19:12",
>> ATTRS{type}=="1", NAME="wlan0"
>> # PCI device 0x10ec:0x8139 (8139too)
>> #SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:48:54:d1:35:1c",
>> NAME="eth1"
>> # PCI device 0x10ec:0x8139 (8139too)
>> SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:48:54:d1:35:1c",
>> NAME="eth2"
>> ------

<--<snip>-->

>
> It looks as though udev is now detecting the card differently as the PCI
> device now has a '0x' before 8139 so effectively the mapping for eth0 is
> for a non existent card. I would try deleting all the eth* entries and
> restart.

The 0x was added by the new version, the non 0x was inserted by the
previous version.

I have just rebooted with the 2.6.21-2 kernel, on the k7, and it
is still showing this odd stuff in dmesg.

dmesg |grep eth
eth0: RealTek RTL8139 at 0xdc00, 00:48:54:d1:35:1c, IRQ 11
eth0: Identified 8139 chip type 'RTL-8139C'
udev: renamed network interface eth0 to eth2
eth2: link up, 10Mbps, half-duplex, lpa 0x0000
eth2: no IPv6 routers present

I will try your suggestion next.

Thanks

wt

--
Daddy, what does FORMATTING DRIVE C mean?
_______________________________________________________

Wayne Topa

unread,

Aug 23, 2007, 4:40:12 PM8/23/07

to

Wayne Topa(linu...@intergate.com) is reported to have said:
> Mathias Brodala(in...@noctus.net) is reported to have said:
> > Hi Wayne.
> >
> > Wayne Topa, 23.08.2007 18:21:
> > > I have an Athlon k7 box, running testing, as my gateway. Due to a new
> > > system install, I had not upgraded in a few weeks. On the 21st I
> > > decided that it was time and did the upgrade overnight. This morning
> > > I rebooted the box to upgrade my sid partition and when I rebooted
> > > testing, I had lost my my localnet connections.

--<snip>-->

>
> I found a fix (?) with a google search but I can't try it until I
> finish downloading a new kernel. It was on an Ubuntu Forum so just in
> case I'll list it here. Not Tested Yet so take it with a grain of
> salt.
>
> ----- Possible Fix ------------------
> Create/Edit a file /etc/udev/rules.d/10-local.rules with the following content Code:
>
> KERNEL="eth*", SYSFS{address}=="00:10:5a:33:44:55" NAME="eth0"
> KERNEL="eth*", SYSFS{address}=="00:0c:44:55:66:77" NAME="eth1"
>
> Just change the mac addresses to your own values from ifconfig.
>
> ----------------------------

The fix did not work on this problem. I don't know enough about udev
but I did think it looked like it would.

The problem was not fixed by installing the 2.6.21-2 k7 kernel.

The suggestion made by Wackojacko did work.

I was able to boot with and without the above fix after I removed all
rules in the /etc/udev/rules.d/z25_persistent-net.rules file.

Commenting out the rules did not work, removing them did. I have to
test if it still works with the old 2.6.18 kernel, but if your
interfaces get changed with the new udev on the 2.6.21-2 k7 kernel,
this should fix it.

Thanks Wackojacko!

WT

--
"Imagine if every Thursday your shoes exploded if you tied them the
usual way. This happens to us all the time with computers, and nobody
thinks of complaining."
-- Jeff Raskin, interviewed in Doctor Dobb's Journal

Wayne Topa

unread,

Aug 23, 2007, 5:30:13 PM8/23/07

to

Wackojacko(wackoj...@ntlworld.com) is reported to have said:
> Wayne Topa wrote:
>> Mathias Brodala(in...@noctus.net) is reported to have said:
>>> Hi Wayne.
>>>

<--<snip>-->

>> ---------
>> # PCI device 10ec:8139 (8139too)
>> SUBSYSTEM=="net", DRIVER=="?*", SYSFS{address}=="00:48:54:d1:35:1c",
>> NAME="eth0"
>> # USB device 2001:3c00 (rt2500usb)
>> #SUBSYSTEM=="net", DRIVERS=="?*", ATTRS{address}=="00:11:95:e6:19:12",
>> ATTRS{type}=="1", NAME="wlan0"
>> # PCI device 0x10ec:0x8139 (8139too)
>> #SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:48:54:d1:35:1c",
>> NAME="eth1"
>> # PCI device 0x10ec:0x8139 (8139too)
>> SUBSYSTEM=="net", DRIVERS=="?*", ATTR{address}=="00:48:54:d1:35:1c",
>> NAME="eth2"

<--<snip>-->

>
> It looks as though udev is now detecting the card differently as the PCI
> device now has a '0x' before 8139 so effectively the mapping for eth0 is
> for a non existent card. I would try deleting all the eth* entries and
> restart.

I forget to mention that I had commented all but the first eth out and
had added the 0x in the commented line. ie
# PCI device 10ec:8139 (8139too) -> # PCI device 0x10ec:8139 (8139too)
but it didn't work, but the other lines were just commented out, not
removed as you suggested.

Thanks again.

Wayne

--
Daddy, what does FORMATTING DRIVE C mean?
_______________________________________________________

Wayne Topa

unread,

Aug 23, 2007, 6:30:16 PM8/23/07

to

Confirmed. removing all of the entries in the
/etc/udev/rules.d/z25_persistent-net.rules file works in the 2.6.18
kernels as well.

The question is, why would commented rules get read by udev and why
would it think a rule, made by a previous version of udev, wasn't
correct? Well I guess thats what _testing_ is for. It sure didn't
get caught in Sid.

Thanks to all that replied!

Wayne

--
I am a computer, dumber than any human and smarter than any
administrator.