Summary: | Vlan Interface on ix1 doesn't pass any traffic on 12.0-ALPHA2 r338240 | ||
---|---|---|---|
Product: | Base System | Reporter: | Xavier Beaudouin <kiwi> |
Component: | kern | Assignee: | freebsd-net (Nobody) <net> |
Status: | Closed FIXED | ||
Severity: | Affects Only Me | CC: | bugzilla.freebsd, linimon, np, piotr.pietruszewski |
Priority: | --- | ||
Version: | CURRENT | ||
Hardware: | amd64 | ||
OS: | Any |
Description
Xavier Beaudouin
2018-08-24 06:12:33 UTC
Can you please check if reducing MTU to 1468 (on the if_vlan(4) child) has any influence? If so, please cross-link https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=230510 -harry vlan3402: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1468 options=600703<RXCSUM,TXCSUM,TSO4,TSO6,LRO,RXCSUM_IPV6,TXCSUM_IPV6> ether 0c:c4:7a:f4:34:a1 inet 10.194.101.101 netmask 0xffffff00 broadcast 10.194.101.255 groups: vlan vlan: 3402 vlanpcp: 0 parent interface: ix1 media: Ethernet autoselect (10Gbase-Twinax <full-duplex,rxpause,txpause>) status: active nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL> Reducting the MTU doesn't seems to have any influence... Note that r338333 have same issue :( > FreeBSD 12.0-ALPHA2 #0 r338182: Wed Aug 22 09:59:11 CEST 2018
VLAN tx was broken in r337943 and then fixed in r338305, so this particular
revision is definitely affected. That bug was tracked via PR 230794.
But the submitter mentioned that r338333 doesn't work either so this may
not be a dup.
Hello, does this problem persist? If so, have you had a chance to check it on 12.0-ALPHA4? What NIC are you using? Would you please show your configuration of parent interface (ix1)? Is the machine connected to switch or do you use back-to-back configuration? If the former, would you please check if the problem occurs also with a back-to-back connection? Peter Hi, # uname -a FreeBSD rsv-filer 12.0-ALPHA5 FreeBSD 12.0-ALPHA5 #0 r338557: Mon Sep 10 12:25:04 CEST 2018 root@rsv-filer:/usr/obj/usr/src/amd64.amd64/sys/GENERIC-NODEBUG amd64 Problem "seems" to have changed from pass no traffic to do not detect link on parent interface (I have this behavior on lagg on 11.2) : ix1: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500 options=e53fbb<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU,VLAN_HWCSUM,TSO4,TSO6,LRO,WOL_UCAST,WOL_MCAST,WOL_MAGIC,VLAN_HWFILTER,VLAN_HWTSO,RXCSUM_IPV6,TXCSUM_IPV6> ether 0c:c4:7a:f4:34:a1 media: Ethernet autoselect status: no carrier nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL> vlan3402: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500 options=600703<RXCSUM,TXCSUM,TSO4,TSO6,LRO,RXCSUM_IPV6,TXCSUM_IPV6> ether 0c:c4:7a:f4:34:a1 inet 10.194.101.101 netmask 0xffffff00 broadcast 10.194.101.255 groups: vlan vlan: 3402 vlanpcp: 0 parent interface: ix1 media: Ethernet autoselect status: no carrier nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL> Configuration is done using rc.conf : ifconfig_ix1="up" vlans_ix1="vlan3402" create_args_vlan3402="vlan 3402" ifconfig_vlan3402="10.194.101.101/24" This machine is connected to a Cisco 6800 thru a Twinax cable, so the media should be : media: Ethernet autoselect (10Gbase-Twinax <full-duplex,rxpause,txpause>) On the switch the configuration is pretty simple : interface TenGigabitEthernet3/5 description Server: rsv-filer switchport switchport trunk native vlan 3402 switchport mode trunk switchport trunk allowed vlan 3401-3403 switchport nonegotiate mtu 9216 logging event link-status flowcontrol send on spanning-tree portfast edge trunk spanning-tree bpduguard enable ! On the switch side the port is not shut because any security stuff... Regards Hum... Addition: # kldstat Id Refs Address Size Name 1 23 0xffffffff80200000 2450510 kernel 2 1 0xffffffff82652000 22b0 cc_htcp.ko 3 1 0xffffffff82655000 3a7ca8 zfs.ko 4 2 0xffffffff829fd000 a4f0 opensolaris.ko 5 1 0xffffffff82f12000 7d08 ioat.ko 6 1 0xffffffff82f1a000 1800 uhid.ko 7 1 0xffffffff82f1c000 23a8 ums.ko 8 1 0xffffffff82f1f000 acf mac_ntpd.ko Seems that if_vlan.ko is not loaded. Added it into loader.conf (and rebooted), link is up but issue is *still* there unfortunatly. Hello, we have trouble reproducing this bug in our environment. Could you please provide more details? 1. What NIC is causing the problem? 2. Have you tried connecting it back-to-back? If so, were there any differences in behaviour you previously described? 3. Is there any additional configuration applied on this interface? Hello 1. Both nic make me problems. ix0 and ix1 (see the first dmesg). I use usualy only ix1 to avoid time consuming issues to get back access to this servers (it is in a remote location, even with the IPMi, it can be a pain in the ass have access) 2. back-to-back is not possible to test, the servers is in a remote location :( and is "more or less" in production (shared storage dor esxi) 3. Only /etc/rc.conf configuration is done. The motherboard is a X11DPU from supermicro (with latest BIOS used, which wad a pain to update as well). Notice I get 4 servers like this (on 2 differents datacenter), 2 of them are currently working like a charm on ESX 6.7 (with vlan / mtu ....) the other two are used as storage on Freebsd. These 2 freebsd servers was never being stable on FreeBSD 11.2 (nvme issues with zfs cache / log stuff) and network on freebsd 11.2 was a bit tricky (lagg that doesn't up ix1 for example). Either this is an issue on FreeBSD with Supermicro X11DPU (maybe the hardware has a bug?) or something else ? Anyway on ESXi 6.7 (and 6.5) it working without any pain. Rergards, Xavier Is this a regression from an earlier install, or is this a new install? Well... Since I didn't try for more than 2 years I close this bug... Seems to be fixed... somewhere. |