The changelog says:
src: axgbe: fix link issues for gigabit external SFP PHYs and 100/1000 fiber modules
src: axgbe: apply RRC to miibus attached PHYs and add support for variable bitrate 25G SFP+ DACs
src: axgbe: properly release resource in error case
My dmesg with this kernel shows one of my links constantly flapping:
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
ax1: Link is UP - 10Gbps/Full - flow control off
ax1: link state changed to UP
ax1: Link is DOWN
ax1: link state changed to DOWN
Booting into the previous kernel fixes it.
My hardware is the official DEC840
Which module (vendor) are you using?
Any netmap involved? VLANs? LAGGs? etc. that could be relevant.
Cheers,
Franco
One more here: https://www.reddit.com/r/opnsense/comments/13rgxcs/comment/jlm3p5w/?utm_source=reddit&utm_medium=web2x&context=3
QuoteUpdated this afternoon, seems to have caused a link flap with a 10Gtek ASF-10G-T SFP. Using a StarTech SFP1000TXST for the time being.
EDIT: I'm using an OPNsense (Deciso) DEC740.
No netmap, no VLANs on that interface. That's my A&TT upstream interface and I know I had to turn off hardware vlan filtering because I'm bypassing their CPE hardware and there may be some VLAN 0 voodoo going on? I may have to capture some packets and see what they look like.
I will get you the exact model of the SFP+ module tomorrow
To limit the exposure of the previous patches perhaps this can help? https://github.com/opnsense/src/commit/fb81510bd
To install...
# opnsense-update -zkr 23.1.8_1
(keep in mind this will delete the old kernel, if you want to retain an older kernel try to revert to that first)
# opnsense-update -kr 23.1.6
(reboot)
# opnsense-update -zkr 23.1.8_1
Cheers,
Franco
The SFP+ module is SFP-10G-T from FS.com that is Juniper coded
https://www.fs.com/products/73107.html?attribute=27&id=425
Quote from: franco on May 26, 2023, 09:24:29 AM
To limit the exposure of the previous patches perhaps this can help? https://github.com/opnsense/src/commit/fb81510bd
To install...
# opnsense-update -zkr 23.1.8_1
This kernel is working. The interface is not flapping now.
Quote from: feld on May 26, 2023, 05:16:17 AM
there may be some VLAN 0 voodoo going on?
I have confirmed there is vlan0, if anyone is curious
13:41:09.525982 a0:f3:e4:63:0b:7b > f4:90:ea:00:62:2e, ethertype 802.1Q (0x8100), length 70: vlan 0, p 0, ethertype IPv4, 78.192.134.61.13103 > 75.13.68.65.13000: Flags [R.], seq 1, ack 59, win 509, options [nop,nop,TS val 3436954555 ecr 1177160148], length 0
13:41:09.553622 a0:f3:e4:63:0b:7b > f4:90:ea:00:62:2e, ethertype 802.1Q (0x8100), length 60: vlan 0, p 0, ethertype IPv4, 94.102.61.38.50380 > 75.13.68.70.5004: Flags [S], seq 794064754, win 65535, length 0
13:41:09.605744 a0:f3:e4:63:0b:7b > f4:90:ea:00:62:2e, ethertype 802.1Q (0x8100), length 180: vlan 0, p 0, ethertype IPv4, 205.251.197.161.53 > 172.13.126.189.32944: 2930 NXDomain*-$ 0/1/1 (134)
Thanks we may check the mentioned modules as time permits but if the fix works we will replace the kernel accordingly now.
VLAN 0 doesn't appear to be an issue at first glance, but thanks for mentioning that.
Cheers,
Franco