Re: [PATCH v1 02/19] cxl: Add skeletal features driver

[Dan Williams <dan.j.williams@intel.com>]

Dave Jiang wrote:
> 
> 
> On 1/22/25 8:59 PM, Dan Williams wrote:
> > Dave Jiang wrote:
> >> Add the basic bits of a features driver to handle all CXL feature related
> >> services. The driver is expected to handle all CXL mailbox feature command
> >> related operations.
> >>
> >> Suggested-by: Dan Williams <dan.j.williams@intel.com>
> >> Signed-off-by: Dave Jiang <dave.jiang@intel.com>
[..]
> > 
> > Does this need its own ida? I expect it could just use the same id as
> > the memdev, and that saves some potential confusion of which
> > memory-devices support features and which features interface correlates
> > with which device.
> > 
> > ...or how do you imagine that working?
> 
> One of the issue was that the feature commands query happens before
> memdev is created. So that ida doesn't exist yet. In order to share,
> we'll need to do a split initialization of the feature device. But
> really I was trying to separate the association of features from the
> memdev as they are independent of each other. i.e a type2 may not have
> memdev but have mailbox and features.

We could do what platform_device does and let the id be optionally
auto-selected or specified by the caller?

However, I think that's a can we can kick down the road until a Type-2
user with a mailbox, but without a memdev shows up. No signs of that on
the horizon unless someone wants to chime in to the contrary.

The Features commands of most interest to date are around memory RAS, so
it is likely those will have a memdev. The Type-2 series from Alejandro
registers a memdev for example.