Re: kern/49290: DRMKMS fails to work with R300

[mlelstv%serpens.de@localhost (Michael van Elst)]

The following reply was made to PR kern/49290; it has been noted by GNATS.

From: mlelstv%serpens.de@localhost (Michael van Elst)
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: kern/49290: DRMKMS fails to work with R300
Date: Thu, 8 Jan 2015 08:57:25 +0000 (UTC)

 martin%duskware.de@localhost (Martin Husemann) writes:
 
 >The following reply was made to PR kern/49290; it has been noted by GNATS.
 
 >From: Martin Husemann <martin%duskware.de@localhost>
 >To: gnats-bugs%NetBSD.org@localhost
 >Cc: 
 >Subject: Re: kern/49290: DRMKMS fails to work with R300
 >Date: Thu, 8 Jan 2015 08:58:49 +0100
 
 > The real fix IMHO is to make pcdisplay_match() return 0 in this configuration -
 > which it tries to do, there must be a bug somewhere.
 
 
 The vga and also pcdisplay driver seem to use the card IO space
 as a lock to prevent multiple drivers from attaching to the same
 hardware.
 
 When vga attaches at PCI, then the probe function for both ISA devices
 fail because they cannot map the IO space again.
 
 But when the DRM radeon driver attaches (at PCI), it provides a better
 match than the VGA driver, but at least at that time, it doesn't map
 the IO space because all the initialization is deferred to after mountroot
 to load the card firmware (or maybe it's not mapped at all?)
 
 This allows either the vga driver or the pcdisplay driver to attach
 at ISA.
 
 The most simple solution is to make the DRM radeon driver map the IO
 space when it attaches (and maybe to release it before the real
 initialization happens later).
 
 In an ideal world the autoconf framework would provide a mechanism
 for such locking.
 
 -- 
 -- 
                                 Michael van Elst
 Internet: mlelstv%serpens.de@localhost
                                 "A potential Snark may lurk in every tree."