Back To Eth Shuffling …

Home » CentOS » Back To Eth Shuffling …
CentOS 8 Comments

So I’m back to this problem. A quick run down of what the original problem was:

I have a machine that I’m configuring to use kickstart to setup. It has two builtin ethernet ports (labeled ports 1 and 2) and I’m adding a third one on its PCIe bus. Originally I was using an r8169 clone a default kickstart always put it as eth0 with the builtin ones as eth1 and eth2 respectively. After some fiddling with blacklisting the driver during the kickstart process, I was able to get the machine to boot up with the r8169 as eth3
(!!!) while the other two were correctly identified as eth0 and eth1 (this is what I want except the PCIe card should’ve been eth2 but it never came up that way)

However I ran into other issues with the r8169 driver and ultimately replaced it for an IntelPRO, which happens to use the same e1000e module as the builtin ones. Great. Except …

Now when I kickstart the machine, I get this:
builtin port1 -> eth0
PCIe card -> eth1
builtin port2 -> eth2

What I want is:
builtin port1 -> eth0
builtin port2 -> eth1
PCIe card -> eth2

I should also point out here that when PXE kicks in on this machine, the machine correctly enumerates the ethernet devices as port1, port2 and then the PCIe card. So somewhere in the kickstart process things change.

So I’m back to trying to figure out how to achieve this.

8 thoughts on - Back To Eth Shuffling …

  • Have you tried having kickstart set up a more appropriate
    /etc/udev/rules/70-persistent-net.rules?
    This is normally written by /lib/udev/write_net_rules. You should be able to modify the automatically generated one to match what you need.

    K

  • When I was working on this last time (with the r8169 driver), someone on this list provided the following script which is what “fixed” the issue at the time by creating a new 70-persistent-net.rules file with the devices enumerated in order. However, this no longer works now.

    echo “[KICKSTART] Binding eth interfaces to the expected MAC address in UDEV”
    echo “## Created by Kickstart to keep network interfaces in an expected order” > \
    /etc/udev/rules.d/70-persistent-net.rules echo “” >> \
    /etc/udev/rules.d/70-persistent-net.rules

    cd /sys/class/net/
    for NETDEV in $(ls | /bin/grep eth | sort)
    do
    ## Create a UDEV rule for each eth interface
    echo “## ${NETDEV} interface” >> \
    /etc/udev/rules.d/70-persistent-net.rules

    ## We throw this one in here as it can contain some useful information
    echo “## $(dmesg | /bin/grep ${NETDEV} | /bin/grep -i -v -e “console”
    -e “Command line” | head -1)” >> \
    /etc/udev/rules.d/70-persistent-net.rules

    echo -n “SUBSYSTEM==\”net\”, ACTION==\”add\”, DRIVERS==\”?*\”, ” >> \
    /etc/udev/rules.d/70-persistent-net.rules
    echo -n “ATTR{address}==\”$(cat ${NETDEV}/address)\”, ” >> \
    /etc/udev/rules.d/70-persistent-net.rules
    echo -n “ATTR{dev_id}==\”0x0\”, ATTR{type}==\”1\”, KERNEL==\”eth?*\”, ”
    /etc/udev/rules.d/70-persistent-net.rules
    echo -e “NAME=\”${NETDEV}\”\n” >> \
    /etc/udev/rules.d/70-persistent-net.rules

    ## Make a log of the devices present during install
    echo -e “${NETDEV} $(cat ${NETDEV}/address)\n” >> /root/ksnet-devices

    ## Also remove the HWADDR line from all of the net config files
    /bin/grep -v -e NAME -e HWADDR -e NM_CONTROLLED \
    /etc/sysconfig/network-scripts/ifcfg-${NETDEV} | sed ‘s/\”//g’ \
    > /etc/sysconfig/network-scripts/ifcfg-${NETDEV}-tmp
    echo “NM_CONTROLLED=no” >>
    /etc/sysconfig/network-scripts/ifcfg-${NETDEV}-tmp
    #/usr/bin/perl -p -i -e ‘s/dhcp/none/’
    /etc/sysconfig/network-scripts/ifcfg-${NETDEV}-tmp
    /bin/mv -f /etc/sysconfig/network-scripts/ifcfg-${NETDEV}-tmp \
    /etc/sysconfig/network-scripts/ifcfg-${NETDEV}
    done

    After the machine boots and I look in /root/ksnet-devices, I see the MAC
    addresses for the devices as:
    Port1 -> eth0
    PCIe Card-> eth1
    Port2 -> eth2

    And yet, during the machine’s POST (which can verify by the PXE boot up of each device), it correctly enumerates the ethernet devices as:
    Port1
    Port2
    PCIe card

    So where and why is the order changed when the kernel kicks in and start booting the machine? And how can I stop it, or change its behavior? I’m not seeing any ‘renaming’ going on in dmesg like I sometimes find. So something is causing this and I can’t figure it out.

  • ​Yeah, my understanding is that the kernel does not really trust all the information that the BIOS tells it and probes for the network devices itself.​

    I’m thinking, since you are using kickstart and PXE boot, you’re not going to know the HWADDR for your NICs before hand, but you probably do have a predictable pci bus layout.

    So grab pciutils package and run:

    lspci | grep Ethernet

    I get something like

    00:0a.0 Ethernet controller: NVIDIA Corporation MCP67 Ethernet (rev a2)
    01:06.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05)
    01:07.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05)

    The first number is the bus id which is going to be consistent across machines with the same mainboard and nic layout.

    So I can make a persistent-net.rules file like:

    ACTION==”add”, SUBSYSTEM==”net”, BUS==”pci”, ID==”0000:00:0a.0″, NAME=”eth0″
    ACTION==”add”, SUBSYSTEM==”net”, BUS==”pci”, ID==”0000:01:06.0″, NAME=”eth1″
    ACTION==”add”, SUBSYSTEM==”net”, BUS==”pci”, ID==”0000:01:07.0″, NAME=”eth2″

    Hope this helps.

    K

  • Actually, I know what the MAC is for the builtin Port1 and 2. Those are listed in the BIOS. But ultimately I don’t want to rely on them as I want the same kickstart file to work for other machines, so hardcoding those in the kickstart file wouldn’t quite work, unless I start writing multiple kickstart files, one per machine.

    Anyway, lspci reports this:
    00:19.0 Ethernet controller: Intel Corporation 82566DM-2 Gigabit Network Connection (rev 02)
    03:00.0 Ethernet controller: Intel Corporation 82572EI Gigabit Ethernet Controller (Copper) (rev 06)
    08:00.0 Ethernet controller: Intel Corporation 82573V Gigabit Ethernet Controller (Copper) (rev 03)

    Now when I look at the enumeration from dmesg, I get this:
    e1000e 0000:00:19.0: eth0: (PCI Express:2.5GT/s:Width x1)
    00:1e:68:58:00:4c <-- this is Port1 e1000e 0000:03:00.0: eth1: (PCI Express:2.5GT/s:Width x1) 00:15:17:d3:43:62 <-- PCIe Card e1000e 0000:08:00.0: eth2: (PCI Express:2.5GT/s:Width x1) 00:1e:68:58:00:4d <-- this is Port2 So this is how it gets enumerated for some reason. But during the machine's POST, it always brings up the PXE boot order as Port1, Port2, and finally the PCIe card, which is the correct order that I want it in. It also only has one single expansion slot so it's not like I can try a different one to see if it makes a difference, unfortunately.

  • So a 70-persistent-net.rules like

    # onboard port 1 -> eth0
    ACTION==”add”, SUBSYSTEM==”net”, BUS==”pci”, ID==”0000:00:19.0″, NAME=”eth0″

    # PCIe card -> eth2
    ACTION==”add”, SUBSYSTEM==”net”, BUS==”pci”, ID==”0000:03:00.0″, NAME=”eth2″

    # onboard port 2 -> eth1
    ACTION==”add”, SUBSYSTEM==”net”, BUS==”pci”, ID==”0000:08:00.0″, NAME=”eth1″

    will do what you want.

    Note: I’ve just changed the ID and NAME values to match what you have and what you want.

    Kahlil (Kal) Hodgson GPG: C9A02289
    Head of Technology (m) +61 (0) 4 2573 0382
    DealMax Pty Ltd GitHub: @tartansandal

    Suite 1416
    401 Docklands Drive Docklands VIC 3008 Australia

    “All parts should go together without forcing. You must remember that the parts you are reassembling were disassembled by you. Therefore, if you can’t get them together again, there must be a reason. By all means, do not use a hammer.” — IBM maintenance manual, 1925

    CentOS mailing list CentOS@CentOS.org http://lists.CentOS.org/mailman/listinfo/CentOS

  • Right, I understand that part. However I believe I’m now in the realm of making this specific to this machine as I have no guarantee that another identical machine will pop up with those same bus IDs. Maybe for the internal ports, but I don’t know if the same will happen for the PCIe bus. Would that be correct?

  • another identical machine will have the same bus ids. that’s why this works.

    Kahlil (Kal) Hodgson GPG: C9A02289
    Head of Technology (m) +61 (0) 4 2573 0382
    DealMax Pty Ltd GitHub: @tartansandal

    Suite 1416
    401 Docklands Drive Docklands VIC 3008 Australia

    “All parts should go together without forcing. You must remember that the parts you are reassembling were disassembled by you. Therefore, if you can’t get them together again, there must be a reason. By all means, do not use a hammer.” — IBM maintenance manual, 1925

    CentOS mailing list CentOS@CentOS.org http://lists.CentOS.org/mailman/listinfo/CentOS