Subject: Re: pci probe
To: None <itojun@itojun.org>
From: None <cgd@broadcom.com>
List: port-i386
Date: 08/15/2003 14:42:41
At Sat, 16 Aug 2003 06:38:31 +0900 (JST), itojun@itojun.org wrote:
> 	my (and imp's) intent was just to filter hdrtype we don't support,

"How does the generic code know what we support?"

Somebody might want to produce a drop-in driver source file, or
loadable module, for a new device which happens to use something
different, for an already-shipped release...  (assuming people ever
define type 3.)



> 	and avoid touching registers that are not defined in yet-to-be-known
> 	hdrtypes.

If generic code is touching registers outside of the first 16 bytes of
each function's configuration space in a manner that doesn't include
having the header type checked first, then it is broken.

However, from previously-quoted standard text, the first 16 bytes for
*all* header types -- including those that we don't yet know about --
are defined to be the same.



cgd