Messages in this thread |  | | Date | Sun, 4 Dec 2022 18:36:05 +0100 | From | Andrew Lunn <> | Subject | Re: [PATCH net-next 4/4] driver/ncn26000: add PLCA support |
| |
> It also seems odd to have this "positive numbers mean update" thing > when every other ethtool API just programs whatever is passed. It seems > to be a different API theory to the rest of the ethtool established > principle that if one wishes to modify, one calls the _GET followed > by a _SET. This applies throughout this function.
Hi Russell
That was my idea. This is a netlink only API, which gives some flexibility which the old ioctl interface did not. With netlink attributes, we can pass only the attributes we want to change.
The question then is, what is the interface between the netlink code and the PHY driver. How do you express what attributes were in the request? I suggested -1 for not present.
We could keep with the old read/modify/write model, but then we are not making use of what netlink actually offers.
Andrew
|  |