[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

xviddetect for sparc (was: Re: autodectection of modules at boot time)

Adam Di Carlo wrote:
> In message <[🔎] 19991022182441.E5993@tausq.org> you wrote:
> >xvidetect was installed into the archives yesterday, so you should be able
> >to get it from your favorite mirror. Please note that the copy in the
> >archive is slightly different from the one that's in my homedir on master.
> >(it uses /usr/share/doc instead of /usr/doc)
> Excellent.
> >It'd be great if we
> >can get this in. One concern though is that it really only works well for
> >i386 at the moment.
> Don't you mean it only works for PCI video cards?  I tested it on
> sparc (ultra5, mach64 video chip) and it worked great.  Remember that
> PPC, new Sparcs, etc are all PCI.
> [...]
> >The good thing perhaps is that the other archs have a
> >much smaller number of X servers, and the typical {alpha,sparc,m68k,ppc}
> >user is probably more knowledgable about what to pick anyway.
> Yes -- to be cute, on Sparc/SBUS machines, you should be able to detect
> the proper video card from proc.  Not sure on the details -- ask the
> right lists.  Not sure about m68k.
> Anyhow, if it works, great, if not, oh well we tried.
> A solution that works for 70% of the time is a success, in my book.
> Right now people have no where at all to turn...

I added preliminary support for sparc SBUS to xviddetect (SBUS is a plug-n-play
system bus in case you was not aware of).  The patch is attached to this mail.
I wrote it by scanning /proc/openpromfs for display devices entries.  It works
at least on sparc2 (bwtwo framebuffer) & SparcClassic (cgthree), the only
hardware I can run it on, but I need other testers to be sure it also works on
CG2, CG4, CG6, CG8, CG14, TCX, Creator & Creator3D framebuffers.
I supposed that the names of display cards reported by OpenPROM are matching
their commercial names but I could be wrong, especially for Creator cards.

Comm. name       OpenPROM name
  BW2        ->    bwtwo
  CG2        ->    cgtwo
  CG3        ->    cgthree
  CG4        ->    cgfour
  CG6        ->    cgsix
  CG8        ->    cgeight
  CG14       ->    cgfourteen
  TCX        ->    tcx
  Creator    ->    creator
  Creator3D  ->    creator3d

I uploaded a patched release of xviddetect to my ftp site:


Any help in testing would be really appreciated.
Too know OpenPROM name of one device, look in /proc/openprom/iommu/sbus/<dev>/name
file and report unmatched names back to me.

Thanks in advance.

PS: what about multi-headed systems?  I can plug more than one card in my
    sparc, but xviddetect only reports the first one.  Anyway, I don't know if
    xfree is working in a multi-headed mode.

 Eric Delaunay                 | "La guerre justifie l'existence des militaires.
 delaunay@lix.polytechnique.fr | En les supprimant." Henri Jeanson (1900-1970)
diff -unr xviddetect-0.2.orig/debian/rules xviddetect-0.2/debian/rules
--- xviddetect-0.2.orig/debian/rules	Tue Aug 31 06:09:02 1999
+++ xviddetect-0.2/debian/rules	Sat Oct 23 19:13:34 1999
@@ -22,6 +22,7 @@
 	install -m 0755 xviddetect $(DESTDIR)/usr/bin/xviddetect
 	install -m 0644 video.ids $(DESTDIR)/usr/share/misc/video.ids
+	install -m 0644 pci.ids $(DESTDIR)/usr/share/misc/pci.ids
 	install -m 0644 README $(DESTDIR)/usr/doc/$(PACKAGE)/README
 	install -m 0644 debian/changelog $(DESTDIR)/usr/doc/$(PACKAGE)/changelog
 	install -m 0644 debian/copyright $(DESTDIR)/usr/doc/$(PACKAGE)/copyright
diff -uNr xviddetect-0.2.orig/video.ids xviddetect-0.2/video.ids
--- xviddetect-0.2.orig/video.ids	Sat Aug 21 22:36:27 1999
+++ xviddetect-0.2/video.ids	Sat Oct 23 18:26:16 1999
@@ -117,3 +117,14 @@
 s3v|all|53338c01|S3 Inc.|ViRGE/MX
 s3v|all|53338c02|S3 Inc.|ViRGE/MX+
 s3v|all|53338c03|S3 Inc.|ViRGE/MX+MV
+xsun-mono|sparc|bwtwo|Sun|BW2 Monochrome Sun framebuffer
+xsun|sparc|cgtwo|Sun|CG2 8-bit Sun framebuffer
+xsun|sparc|cgthree|Sun|CG3 8-bit Sun framebuffer
+xsun|sparc|cgfour|Sun|CG4 8-bit Sun framebuffer
+xsun|sparc|cgsix|Sun|CG6 8-bit Sun framebuffer
+xsun|sparc|tcx|Sun|TCX Sun framebuffer in 8-bit color mode
+xsun24|sparc|tcx|Sun|TCX Sun framebuffer in 24-bit color mode
+xsun24|sparc|cgeight|Sun|CG8 24-bit Sun framebuffer
+xsun24|sparc|cgfourteen|Sun|CG14 24-bit Sun framebuffer
+xsun24|sparc|creator|Sun|Creator 24-bit Sun framebuffer
+xsun24|sparc|creator3d|Sun|Creator3D 24-bit Sun framebuffer
diff -uNr xviddetect-0.2.orig/xviddetect xviddetect-0.2/xviddetect
--- xviddetect-0.2.orig/xviddetect	Tue Aug 31 06:11:25 1999
+++ xviddetect-0.2/xviddetect	Sat Oct 23 19:13:44 1999
@@ -15,6 +15,7 @@
 sub GetArch {
   my $arch = `uname -m`;
+  chop $arch;
   $arch = 'i386' if ($arch =~ /^i\d86/);
   return $arch;
@@ -65,6 +66,9 @@
+  # no report of error on sparc (could be an sbus machine)
+  return if ($arch eq "sparc");
   print "Cannot open $pcidevice. You probably have an old kernel. Support\n";
   print "for the old Linux PCI interface should be coming soon\n\n";
   exit 0;
@@ -94,6 +98,59 @@
   close F;
+# Read the SPARC SBus configuration to fill the device list with detected
+# display devices.
+sub ReadSBUS {
+  my $dir;
+  my $dev;
+  # first check for openpromfs
+  -f "/proc/openprom/.node" || die "This system is lacking openpromfs support.";
+  # then parse the tree to looking at display type entries
+  # Old sparc have their devices referenced under sbus@x,y and newer ones have
+  # one more level of indirection (iommu@x,y/sbus@x,y), so look in both trees.
+  opendir(D, "/proc/openprom");
+  my @bus = readdir(D);
+  closedir D;
+  my @io = grep { s/^iommu/\/proc\/openprom\/iommu/ } @bus;
+  my @bus = grep { s/^sbus/\/proc\/openprom\/sbus/ } @bus;
+  foreach $dir (@io) {
+    opendir(D, $dir);
+    push @bus, grep { s:^sbus:$dir/sbus: } readdir(D);
+    closedir D;
+  }
+  foreach $dir (@bus) {
+    # In the sbus dir, there is one subdir per device.
+    # Each subdir name is build from the name of the device, sbus slot number
+    # and sbus first address in slot (dev@x,y).
+    # Each device subdir contains may contains 2 files named device_type &
+    # name. The former is telling the type of device (eg. 'display' for a
+    # graphic card) and the latter gives his name ('cgthree', ...).
+    opendir(D, $dir);
+    # to know whether dir entry is a device subdir, we cannot use -d $_.  It
+    # simply doesn't work with openpromfs :(  One alternative is to looking for
+    # an '@' in the name.
+    my @devlist = grep { s:(.*@):$dir/\1: } readdir(D);
+    closedir D;
+    foreach $dev ( @devlist ) {
+      if (open(G, "<$dev/device_type")) {
+	my $id = <G>;
+	chop $id;
+        close G;
+	# a display device is referenced as such in the device_type file
+	if ($id eq "'display'") {
+	  open(G, "<$dev/name") || die "Bad OpenPROM contents.";
+	  # get the name of the device (remove surrounding quotes)
+	  $id = <G>;
+	  chop $id;
+	  $id =~ s/'//g;
+	  # register this device in the database
+	  push @devices, $id;
+	}
+      }
+    }
+  }
 sub Help {
   print <<EOF;
 video.pl [--arch <architecture>] [--help]
@@ -118,6 +175,7 @@
+&ReadSBUS if $arch eq "sparc";
 if (!$quiet) {
   print "Your architecture seems to be $arch.\n";

Reply to: