This sloppy diff enables astfb on my Talos. It works around the
firmware's wrong sc_fbaddr by mapping PCI BAR 0x10 and using it as a
framebuffer. Mark, you might make a better diff?
"pcidump -v" didn't show my ast vga device until I made the missing
/dev/pci5 node. Then it showed some BARs. I tried BAR 0x10 because
its size seemed big enough for a framebuffer. I got the idea from
some code in /usr/xenocara/driver/xf86-video-ast that seemed to use
a PCI region as a framebuffer.
When I started xenodm, the colors were wrong. I added some code to
xf86-video-wsfb that fixed some colors, but caused emacs to crash
in a weird way. --George
Index: arch/powerpc64/dev/astfb.c
===================================================================
RCS file: /cvs/src/sys/arch/powerpc64/dev/astfb.c,v
retrieving revision 1.3
diff -u -p -r1.3 astfb.c
--- arch/powerpc64/dev/astfb.c 30 Oct 2020 13:36:45 -0000 1.3
+++ arch/powerpc64/dev/astfb.c 23 Aug 2021 02:49:48 -0000
@@ -118,8 +118,17 @@ astfb_attach(struct device *parent, stru
sc->sc_iot = pa->pa_memt;
if (bus_space_map(sc->sc_iot, sc->sc_fbaddr, sc->sc_fbsize,
BUS_SPACE_MAP_LINEAR, &sc->sc_ioh)) {
- printf(": can't map framebuffer\n");
- return;
+ bus_size_t size;
+ pcireg_t maptype;
+
+ maptype = pci_mapreg_type(pa->pa_pc, pa->pa_tag, ASTFB_PCI_FB);
+ if (pci_mapreg_map(pa, ASTFB_PCI_FB, maptype,
+ BUS_SPACE_MAP_LINEAR, &sc->sc_iot, &sc->sc_ioh,
+ &sc->sc_fbaddr, &size, 0) != 0) {
+ printf(": can't map framebuffer\n");
+ return;
+ }
+ printf(": BAR size 0x%zx", (size_t)size);
}
printf("\n");
