Hi Michal,
Merry Christmas and thanks a lot for the prompt reply.
After upgraded the kernel to 5.10 and we did see the expected logs:
[root@shlei-dev-machine ~]# ethtool -s ens32 master-slave preferred-master
netlink error: master/slave configuration not supported by device (offset 36)
netlink error: Operation not supported
Best Regards,
Rudy
On 2020/12/27, 5:46 PM, "Michal Kubecek" <mkube...@suse.cz> wrote:
On Sun, Dec 27, 2020 at 12:34:09PM +0800, Bruce LIU wrote:
> Hi Michal Kubecek and Network dev team,
>
> Good day! Hope you are doing well.
> This is Bruce from China, and please allow me to cc Rudy from Cisco
Systems
> in China team.
>
> We are facing a weird behavior about "master-slave configuration" function
> in ethtool.
> Please correct me if I am wrong....
>
> As you know, start from ethtool 5.8, "master/slave configuration support"
> added.
> https://lwn.net/Articles/828044/
>
> ========================================================================
> Appeal:
> Confirm and discuss workaround
>
> ========================================================================
> Issue description:
> As we test in lab, no "master-slave" option supported.
>
> ========================================================================
> Issue reproduce:
> root@raspberrypi:~# ethtool -s eth0 master-slave master-preferred
> ethtool: bad command line argument(s)
> For more information run ethtool -h
>
> ========================================================================
> Environment:
> debian-live-10.7.0-amd64-standard.iso
> Kernel 5.4.79
This is the problem. Kernel support for this feature was added in
5.8-rc1 so that your kernel does not have it and there is no chance it
could possibly work. Newer ethtool has support for this feature but
kernel must support it as well for it to actually work.
But I agree that the error message is misleading. We handle subcommands
supported only in netlink with proper error message when ioctl fallback
is used but we don't do the same for new parameters of existing
subcommands which are generally supported by ioctl code. That's why the
command line parser used by ioctl code does not recognize the new
parameter and handles it as a syntax error.
We'll need to handle new parameters in ioctl parser so that it produces
more meaningful error for parameters only supported via netlink. Long
term, the proper solution would probably be using one parser for both
netlink and ioctl but that was something I wanted to avoid for now to
reduce the risk of introducing subtle changes in behaviour of existing
code.
Michal
