Do you still see rx_discards?
I see them continuously, every / every other sec, on my mox with sfp module.
Didn’t checked since a long time.
Last reboot almost two days ago and here is the stats for eth2
RX packets:40684311 errors:0 dropped:0 overruns:0 frame:0
TX packets:7015215 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:532
RX bytes:52955186414 (49.3 GiB) TX bytes:1505944012 (1.4 GiB)
rx_discard: 14099
So still rx discard but not a lot. Didn’t seem to cause any issue.
Thx,
Could you be so kind to paste the output of ring buffer params for the ethX where you have the sfp.
ethtool -g ethX
RX 128 on my MOX looks quite low imho.
Would be Interesting to know if it is the same on omnias
Ring parameters for eth2:
Pre-set maximums:
RX: 128
RX Mini: 0
RX Jumbo: 0
TX: 532
Current hardware settings:
RX: 128
RX Mini: 0
RX Jumbo: 0
TX: 532
It seem it can’t be changed.
ethtool -G eth2 rx 532 for exemple do nothing
Has anyone gotten the Wiitek SFP+ 10g copper transceiver to work? I have a spare that I’d like to use, and there don’t seem to be any US sellers for the official 2.5G SFP module.
fwiw, the module gets recognized in ethtool -m eth2
, but in dmesg I get the dreaded eth2: validation with support 00,00000000,00002440 failed: -22
error and it refuses to recognize a link (even just 1gbit).
I agree that alternatives should be found for folks outside the EU! Perhaps someone can reveal the basic make + model that the RTROM01-RTSF-10G is re-branded from and Turris can be asked to provide firmware so we can source elsewhere? Shouldn’t be an issue for open source advocates!
As a follow-up: are there any 2.5g bidirectional fiber transceivers that work at the full 2.5g rate? I’m looking to move stuff around and thus a DAC cable won’t cut it anymore.
As an alternative to DAC I can report that a TP-Link TL-SM410U (2.5gbe SFP+) that I purchased from China via AliExpress for a little over $20 total delivered is working going from Omnia → TL-SM410U → Cat5e → native 2.5gb switch port (QNAP QSW-2104-2S)! Current OS v6.0 reports 2500 gbps connection in web interface & system logs show it with no drama. Green LED on switch indicates 2.5gb speed but I haven’t run iPerf yet. Seems to run cool.
It’s not fiber but at least high grade twisted pair copper can run much farther than Twinax!
TL-SM410U just arrived, however I can’t make it work with 1000 speed (which I need to use for some time):
ethtool eth2
Settings for eth2:
Supported ports: [ FIBRE ]
Supported link modes: 2500baseX/Full
Supported pause frame use: Symmetric
Supports auto-negotiation: Yes
Supported FEC modes: Not reported
Advertised link modes: 2500baseX/Full
Advertised pause frame use: Symmetric
Advertised auto-negotiation: Yes
Advertised FEC modes: Not reported
Speed: 2500Mb/s
Duplex: Full
Port: FIBRE
PHYAD: 0
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: d
Wake-on: d
Link detected: no
I have had issues with some sfp modules with Auto-negotiation: on
Try to hard set link speed on both ends
can’t set it. got invalid argument on each ethool -s option - autoneg off speed 1000 etc
weirdly enough it is being seen as fibre port:
Supported ports: [ FIBRE ]
I’m wondering if that can be changed somehow. Or it is a driver problem.
root@turris:~# dmesg | grep -E "sfp|mvneta"
[ 1.504147] mvneta_bm f10c8000.bm: Buffer Manager for network controller enabled
[ 1.513148] mvneta f1070000.ethernet eth0: Using hardware mac address d8:58:d7:00:3c:b9
[ 1.522453] mvneta f1030000.ethernet eth1: Using hardware mac address d8:58:d7:00:3c:b7
[ 1.531707] mvneta f1034000.ethernet eth2: Using hardware mac address d8:58:d7:00:3c:b8
[ 11.007817] sfp sfp: Host maximum power 3.0W
[ 11.337666] sfp sfp: module TP-LINK TL-SM410U rev 2.0 sn XXXXXXXXXXXXX dc 220621
[ 11.347024] mvneta f1034000.ethernet eth2: switched to inband/2500base-x link mode
[ 25.336119] mvneta f1030000.ethernet eth1: configuring for fixed/rgmii link mode
[ 25.336731] mvneta f1030000.ethernet eth1: Link is Up - 1Gbps/Full - flow control off
[ 26.379498] mvneta f1034000.ethernet eth2: configuring for inband/2500base-x link mode
I’m not expert on this stuff but my impression is that there is great variation amoung SFP based switches & routers and some marketing materials imply the products only support Fibre! Perhaps TP-Link marking these as such helps with compatability? I wouldn’t think it should matter ultimately though.
The negotiation failure can come from either end. What’s at the other end of the connection? If it’s a managed switch/router or HBA you might troubleshoot from there. Of course it’s possible that the TL-SM410U is crap but for the money it does at least do 2.5 Gigs! Can you pick up a cheap 1-Gig SFP and physically swap as needed?
Has anybody successfully tested a DAC with 2,5GBit speed?
Yes. Use ethtool -s eth2 speed 2500 autoneg off
to make it use 2.5g, and do the equivalent on the other side.
Makes sense for the Turris site of the cable, so for connecting two TO or the upcoming TO Enterprise it should work. Which DAC-manufacturer did you test? And to which endpoint did you connect your Turris device?
I don’t remember which DAC I used. It was just a generic 10g DAC. Maybe 10gtek? I connected it to a MikroTik CRS305, but according to release notes it should work on any 98DX or 98PX1024-based switch. It also must be manually configured (no auto-negotiation).
Do you think that will work also for a SFP+ NIC as endpoint?
Only if the NIC supports 2.5g. Unfortunately, I’m not aware of any that are confirmed to work at 2.5, at least not without a bunch of firmware and driver hacks.