Mon, Jun 01, 2020 at 11:44:36PM CEST, k...@kernel.org wrote: >On Mon, 1 Jun 2020 08:39:18 +0200 Jiri Pirko wrote: >> > If the permanent (NVRAM) parameter is true, all loaded new drivers >> > will indicate support for this feature and set the runtime value to >> > true by default. The runtime value would not be true if any loaded >> > driver is too old or has set the runtime value to false. >> >> This is a bit odd. It is a configuration, not an indication. When you >> want to indicate what you support something, I think it should be done >> in a different place. I think that "devlink dev info" is the place to >> put it, I think that we need "capabilities" there. >
First of all, I think we cleared that up, params are not used like that in this patchset. >Could you explain the need for "capabilities" under dev info? > >I don't like catch-all mechanisms in principle. Better if capabilities >are expressed by the API dedicated to configuration of a given feature. I see. That makes sense. I was thinking about that, some capabilities are per port. I think that a simple attribute would do. > >In this particular example the ability to do live reset is clearly >expressed by the presence of the parameter (as implemented by this set). >
