2007/477: Drivers registering devids

[Jerry Gilliam]

>> >The function has been in the DDI for more than ten years ... has it
>> >been unusable that whole time due to the static list?
>> >
>> 
>> This is a piece of work required to complete the devid cache discovery
>> mechanism for devid providers not bundled in the ON consolidation.
>
>OK.  The surprising thing to me was that we added this to the DDI long
>ago -- making it theoretically available for others to use -- but if
>they ever tried using it, it wouldn't have worked as expected, at
>least according to this project.
>
>I had though that meant I was likely missing something (such as a
>subsequent change to device enumeration that broke the DDI).  But if
>that's not the case, then "ok."
>

A more complete autopsy here is that pre-devfs, solaris relied on
bottom-up full attach, for all intents and purposes a reconfiguration
boot, to implement open-by-devid.

Devfs itself did not fix this, so open-by-devid still required
full attach.  The devid cache eliminated that so that with a
cache hit, an open-by-devid goes directly to the device.

With a cache miss though, discovery is needed.  The embedded list
of drivers providing devids was a short-term solution to the problem
that has left unbundled drivers out in the cold.  In retrospect,
yes I should have done this as part of the original devid cache
work.


-jg