Netgate Discussion Forum
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Search
    • Register
    • Login

    Em(4) NIC link cycling when member of bridge

    Scheduled Pinned Locked Moved 2.3-RC Snapshot Feedback and Issues - ARCHIVED
    18 Posts 5 Posters 3.3k Views
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • U
      UltramaticOrange
      last edited by

      If I'm reading this thread correctly, it sounds like issues with the network interface on the pfSense box going up then immediately back down after a network cable has been hot plugged. I'm seeing this issue as well.

      Asus N3150I-C

      WAN:
      Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 15)

      Bridged & assigned as LAN:
      4 port Gb HP PCIe(x4) NIC using Intel Corporation 82571EB Gigabit Ethernet Controller (Copper) (rev 06)

      rebooting the system keeps the interfaces up and DHCP successfully leases IPs to clients.

      1 Reply Last reply Reply Quote 0
      • C
        cmb
        last edited by

        I split this out into its own thread as the other thread doesn't sound related, it just link cycled when the interface statistics widget was enabled on the dashboard.

        It sounds like this affects your LAN? Which is a lagg? Which lagg type?

        Are you running 2.3? That's where the original thread was, so where this went.

        1 Reply Last reply Reply Quote 0
        • U
          UltramaticOrange
          last edited by

          Correct, 2.3 Beta.

          I'm not configuring lagg (but the 4 port HP (intel) NIC has the option, iirc), I have the 4 port HP NIC (intel based) bridged and br0 is set to to LAN. Switching ports, plugging in something new, or even just unplugging then replugging causes the port on the NIC to endlessly cycle. Although I haven't found the correct logs to confirm.

          Software engineer here, so it's entirely possible that my lack of networking (and or the lagg functionality) is what is screwing me over here.

          1 Reply Last reply Reply Quote 0
          • C
            cmb
            last edited by

            Sounds like an old bug back in a new way. Years back there was an issue with em NICs in a bridge where if they lost link, they'd continually link cycle because in the process of adding the NIC to the bridge, it'd set things on the NIC that made it cycle link. Rinse and repeat forever.

            Old bug, new issue is at least somewhat different.
            https://redmine.pfsense.org/issues/846

            If you edit /etc/rc.linkup and comment out the line:

            interface_configure($iface, true, true);
            

            Right under the comment from that ticket. "// Do not try to readd to bridge otherwise em(4) has problems"

            does it no longer link cycle? That's not a proper fix but would at least narrow it down. If you already got it in a situation where it's cycling link, reboot after making that change. Then unplug and replug one of the NICs and see what happens.

            1 Reply Last reply Reply Quote 0
            • U
              UltramaticOrange
              last edited by

              I updated to 2.3.b.20160211.0929, verified the issue still existed, but it's been endlessly rebooting since. I'll sit on my hands until the next build, reinstall and retry.

              1 Reply Last reply Reply Quote 0
              • U
                UltramaticOrange
                last edited by

                2.3-BETA (amd64)
                built on Thu Feb 11 18:43:21 CST 2016
                FreeBSD 10.3-BETA1

                commented out the specified line (line 100).

                No change in behavior. I found a spare hdmi cable, so I wired up the second input to my monitor and noticed it's complaining about ipv6. Most relevent looking lines were as below:

                @console:

                bridge0: IPv6 address on em2 have been removed before adding it as a member to
                prevent IPv6 address scope violation.

                Same note for em0, em1, and em3. Not sure if there's missing text due to a lack of a wordwrap on-screen, or if the message is just oddly worded.

                1 Reply Last reply Reply Quote 0
                • H
                  Harvy66
                  last edited by

                  if the NIC is bridged, maybe the switch is shutting down a loop? I know my HP switch, by default, shuts down loops to save the network.

                  1 Reply Last reply Reply Quote 0
                  • C
                    cmb
                    last edited by

                    Confirmed there is definitely an issue there with em NICs in a bridge. Lose and re-gain link on the NIC and you'll end up with non-stop cycling.

                    In my test case, after doing that to one NIC, all the NICs in the bridge are cycling link. That the case for you as well?

                    I'll look into it, though it might be early next week, several other things I have to get finished today.

                    1 Reply Last reply Reply Quote 0
                    • U
                      UltramaticOrange
                      last edited by

                      Added a second machine to my setup and confirmed that all NICs on the bridge cycle when I swap ports on one.

                      It may be worth noting at this point that the reason why I'm on the 2.3 beta is because on 2.2.x, the realtek NIC disappears when the HP/Intel NIC is added to the machine. Seems like any previous fixes for the issue in this thread might be related to the issue with the missing realtek NIC.

                      1 Reply Last reply Reply Quote 0
                      • M
                        maverick_slo
                        last edited by

                        Hmmm I have similar problem with em nics but they are not in the bridge. When I unplug one (wan dhcp) web interface stops to respond and never come back.
                        If I unplug pppoe, link never comes back in console I see something about watchdog resseting…
                        I never had any issues with nics until now.
                        Can this be related?

                        1 Reply Last reply Reply Quote 0
                        • C
                          cmb
                          last edited by

                          @maverick_slo:

                          Can this be related?

                          No, this strictly affects bridges, please start your own thread describing what you're doing and seeing if you're still having issues after the change back in the other thread.

                          1 Reply Last reply Reply Quote 0
                          • M
                            maverick_slo
                            last edited by

                            Hi cmb all fine here with latest snaps.

                            1 Reply Last reply Reply Quote 0
                            • C
                              cmb
                              last edited by

                              This should be fixed now.

                              1 Reply Last reply Reply Quote 0
                              • U
                                UltramaticOrange
                                last edited by

                                Updated to 2.3.b.20160217.1500 and tested with one DHCP client. Looks good!  :D Thanks

                                1 Reply Last reply Reply Quote 0
                                • J
                                  jwt Netgate
                                  last edited by

                                  patch was reverted.

                                  1 Reply Last reply Reply Quote 0
                                  • U
                                    UltramaticOrange
                                    last edited by

                                    @jwt:

                                    patch was reverted.

                                    I'm taking this to mean that I should hold off on updating until the patch/fix is reintroduced.

                                    1 Reply Last reply Reply Quote 0
                                    • C
                                      cmb
                                      last edited by

                                      No, it was the change that introduced the problem that was reverted, and the problem that change fixed was fixed in a different way that doesn't trigger this problem. You're fine.

                                      1 Reply Last reply Reply Quote 0
                                      • U
                                        UltramaticOrange
                                        last edited by

                                        Cool. Thanks.

                                        1 Reply Last reply Reply Quote 0
                                        • First post
                                          Last post
                                        Copyright 2025 Rubicon Communications LLC (Netgate). All rights reserved.