2019-03-21 21:25:04

by Heiner Kallweit

[permalink] [raw]
Subject: New ethtool sub-feature and ethtool-netlink

Hi Michal / John,

I have an ethtool extension in the pipe that adds one more PHY tunable
sub-feature (for Fast Link Down support). Are smaller extensions still
ok for ethtool, or IOW: When do you think is ethtool-netlink going to
be stable enough so that other developers can add features?

Heiner


2019-03-22 07:39:01

by Michal Kubecek

[permalink] [raw]
Subject: Re: New ethtool sub-feature and ethtool-netlink

On Thu, Mar 21, 2019 at 10:23:36PM +0100, Heiner Kallweit wrote:
> I have an ethtool extension in the pipe that adds one more PHY tunable
> sub-feature (for Fast Link Down support). Are smaller extensions still
> ok for ethtool,

Adding a new PHY tunable is IMHO minor change which doesn't need to wait
for the netlink interface. It's a bit more complicated due to the way
tunables (and PHY tunables) are implemented in ioctl API (which requires
both sides to have a list of tunables and their interpretation) but that
is one of the things I would like to change. :-)

After all, I don't have support for (PHY) tunables in the netlink API
yet (not even in the part which hasn't been submitted yet).

> or IOW: When do you think is ethtool-netlink going to be stable enough
> so that other developers can add features?

I would say that once it reaches mainline and userspace starts to use
it, it will have to be stable (due to "Linus' first commandment"). When
that is going to be, that depends on how the review is going, it's hard
to guess.

Michal

2019-03-23 14:33:41

by Heiner Kallweit

[permalink] [raw]
Subject: Re: New ethtool sub-feature and ethtool-netlink

On 22.03.2019 08:38, Michal Kubecek wrote:
> On Thu, Mar 21, 2019 at 10:23:36PM +0100, Heiner Kallweit wrote:
>> I have an ethtool extension in the pipe that adds one more PHY tunable
>> sub-feature (for Fast Link Down support). Are smaller extensions still
>> ok for ethtool,
>
> Adding a new PHY tunable is IMHO minor change which doesn't need to wait
> for the netlink interface. It's a bit more complicated due to the way
> tunables (and PHY tunables) are implemented in ioctl API (which requires
> both sides to have a list of tunables and their interpretation) but that
> is one of the things I would like to change. :-)
>
> After all, I don't have support for (PHY) tunables in the netlink API
> yet (not even in the part which hasn't been submitted yet).
>
Thanks a lot, good to know. I think I can submit the new PHY tunable
next week.

>> or IOW: When do you think is ethtool-netlink going to be stable enough
>> so that other developers can add features?
>
> I would say that once it reaches mainline and userspace starts to use
> it, it will have to be stable (due to "Linus' first commandment"). When
> that is going to be, that depends on how the review is going, it's hard
> to guess.
>
> Michal
>
Heiner