> > > The SB1500 will definitely not function right > now > > > (pci-bridge not supported by sparcPci.c). > > > > > > > Does the SB2500 have the same issue? > > Yes. So do the U25 and U45 and also T1000/T2000. > A Blade 1000/2000, however, *is* known to work with > the current Xorg's sparcPci.c (SPARC32PLUS build Xorg > - only). *All* pre-Blade1000/2000 Ultra workstations > will also work(Schizo, Psycho, Hummingbird and so > on). > > > #A) ... > #B) ... > So the question is, how to make /dev/fb's knowledge and functions usable to > Xorg. > It should be doable. [SNIP]
The difficulty lies in trying to find the Least Common Denominator among the various device specific /dev/fb kernel drivers (which isn't too easy without access to the sources other than cgsix). Unfortunately there doesn't seem to be too much of such a thing anymore on most modern frame buffer's /dev/fb drivers, as Sun apparently didn't continue to keep them standards conform to the historically well-organized "frame buffer interface as defined in fbio(7I)". The last drivers providing the fbio interface were cgsix http://cvs.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/sun/io/cgsix.c , maybe partially also m64. But that's it. Without the driver sources it is also impossible to learn more about all the device specific offsets, consequently making it impossible to replace the global aperture driver mmap()'ings with the less global/more secure xsvc, which can be built on SPARC, but doesn't properly mmap(). Not to mention going the most local/device-specific/secure way of directly mmap()'ing through /dev/fb, making any additional aperture or xsvc driver superfluous. One of my experiments (as I still don't have any Xsun sources [except for the DDK and sunInit.c]) : I found out, that Xsun doesn't have any device-specific knowledge at all. Nor does it perform any general bus scans (as I can see from comparing Xsun's sunInit.c versus in Xorg's init c file). I took Solaris 2.5.1's Xsun packages and tested them with /dev/fb drivers, ddx modules and actual hardware that came out 6 yours after that old Xsun server. But that vintage Xsun server from 1997 is still able to use drivers and hardware which didn't show up before 2003!!! This tells a quite lot about how the server's and the driver's sources must be (vaguely) organized. That's the first step in reverse engineering whatever. And the only option left when stuff is closed source, sigh ... But back to Xorg and the non-functional pci-bridges. Meanwhile SPARC-Linux (Debian Etch and Gentoo) folks get Xorg working on Sun Blade 1500 (with XVR-100 Radeon). Thanks to big names like David S. Miller. This is, where the future lies: --->> APPROACH #3) http://www.x.org/wiki/PciReworkProposal Solaris Prototyping: edward.shu at sun.com Driver Status: ati/ati Bugzilla fufutos ati/atimisc Bugzilla fufutos I wonder if Edward Shu is reading this list and, if so, why he doesn't seem to show up. This message posted from opensolaris.org
