[SOLVED] Getting StarTech 10GBe card working with CentOS 7

Vanyel · 03-08-2019, 02:24 PM

I'm trying to get a Startech ST10000SPEX nic working under CentOS 7. It works fine under CentOS 6 with the supplied tn40xx-0.3.6.8 driver, but the Readme for that driver says it isn't for the kernel version in CentOS 7 (I'm at 7.6). And indeed it won't MAKE under CentOS 7.

I found a newer tn40xx-0.3.6.17 driver directly on Tehuti's site which does say it's for my kernel but though the driver installs fine, the card doesn't work. No link lights on the card. Does not acquire a DHCP address. If I plug the same network cable into other NICs on the machine, they have no problem.

After many days, Tech support is stumped. I was at the latest 3.10.0-957.5.1.el7.x86_64 kernel. They claim it worked there under an older 3.10.0-693.2.2.el7.x86_64 kernel, but I downgraded my kernel and re-installed the driver with no luck.

the card is recognized as a piece of hardware -

Code:

[root@<machine> ~]# dmesg | grep tn40
[   23.441672] tn40xx: loading out-of-tree module taints kernel.
[   23.441988] tn40xx: module verification failed: signature and/or required key missing - tainting kernel
[   23.442644] tn40xx: Tehuti Network Driver, 0.3.6.17
[   23.442912] tn40xx: Supported phys :    QT2025 TLK10232 AQR105 MUSTANG
[   23.443566] tn40xx: srom 0x0 HWver 16 build 0 lane# 4 max_pl 0x1 mrrs 0x2
[   23.444006] tn40xx 0000:22:00.0: irq 32 for MSI/MSI-X
[   23.968751] tn40xx: PHY detected on port 28 ID=1405896 - Native 10Gbps CX4
[   23.969037] tn40xx: PHY type by svid 7 found 1
[   24.069818] tn40xx: fw 0xe
[   24.070081] tn40xx: eth0, Port A
[   24.070383] tn40xx: 1 1fc9:4025:1fc9:3015
[   24.070656] tn40xx: detected 1 cards, 1 loaded

I've got the UUID and MAC address of the card and made an ifcfg file for it:

Code:

[root@<machine> working]# ip addr show
<snip>
2: enp35s0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast state DOWN group default qlen 1000
    link/ether 00:13:3b:10:21:af brd ff:ff:ff:ff:ff:ff
<snip>


[root@<machine> working]# nmcli con
NAME     UUID                                  TYPE      DEVICE
enp6s0   a510dbae-f568-4de5-aa4a-34d5417a58e8  ethernet  enp6s0
enp35s0  10333b96-7a74-ef57-bcfa-9fdf475e1ec6  ethernet  --


UUID="10333b96-7a74-ef57-bcfa-9fdf475e1ec6"
HWADDR=00:13:3b:10:21:af
TYPE=Ethernet
NAME=enp35s0
DEVICE=enp35s0
NM_CONTROLLED="yes"
BOOTPROTO=dhcp
PEERDNS=yes
PEERROUTES=yes
DEFROUTE=no
IPV4_FAILURE_FATAL=yes
#NETBOOT=yes
IPV6INIT=no
ONBOOT=yes

But the card won't come up on boot, or if I manually bounce it -

Code:

[root@<machine> working]# ifdown enp35s0
[root@<machine> working]# ifup enp35s0
Error: Connection activation failed: No suitable device found for this connection.

I've tried 2 different machines under CentOS 7. Same problem.

I have 2 cards, each works fine when swapped into a CentOS 6 machine. Just not under 7.

It's been days and I'm out of tricks. Anybody else have this card and gotten it to work? how?

berndbausch · 03-08-2019, 05:43 PM

I know neither this card nor its driver and how they are supposed to communicate with each other, but: "No suitable device found" points to a problem way up in the software stack, not the driver or the hardware. As if the card were not named enp35s0 (yes, I did see your ip output). Perhaps the problem is at udev level? Or it's related to Network Manager - it often has its own ideas about network configuration, which don't necessarily coincide with yours. Try NM_CONTROLLED=no.

A few other random suggestions:

Did you try to talk to the card with ethtool?
Does the failed ifup command leave any traces in the kernel buffer?
If the driver has a debug option, use it (but tech support should have directed you towards it already, I guess).
Look around under /sys/class/net/enp35s0 and see if you find anything suspect, compared to other NICs. There are tools like lshw that explore hardware as well.

Vanyel · 03-13-2019, 10:28 AM

Thanks for your suggestions! Here's the solution:

The driver is supplied by a separate company from StarTech. The problem was, there's an 88X3140-FW-R02-06-03.hdr firmware file that they bundle with the older driver on CD, but isn't present in the .zip file when you download the newest driver from the maker's site. I had to switch to emailing the driver vendor directly to find out. Without that file, the driver compiles without error. It just won't work!

I found the file on GitHub. https://github.com/acooks/tn40xx-driver

Works fine now!