Jonathan Stone <jonathan@DSG.Stanford.EDU> writes:
> Hm. We could solve that generally by creating new bdevsw (and cdevsw?)
> entrypoints
>    	const  struct bdevsw  * (d_device)
> 
> and let the callee fish out the dv_xname.  Would you buy that, from an
> abstraction viewpoint?

you mean struct device *?  if so: no, that doesn't really seem
like the right solution.  (many disk-ish devices don't even really
have struct devices associated with them, e.g. ccd, vnd, etc., though
i'm not sure what the labelling considerations there are.)


> Or should we just change the readdisklabel/writedisklabel() signature
> and have the caller pass dv_xname?

You could probably use some of the same machinery that the RB_ASKNAME
code uses, or something...


cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.