On Mon, 2013-04-08 at 15:21 +1000, Michael Ellerman wrote: > Looking at x86, ia64 and ours, there's a fair bit of difference. > > x86/ia64 both walk up the parents checking PCI_BRIDGE_CTL_VGA, which > powerpc doesn't (though maybe it should?). Unclear for several reasons. That bit indicates that the bridge forwards the hard coded legacy VGA IO ports and memory aperture. These are not necessary to get modern video cards going. Only if legacy modes are used. On x86, that tends to be the case (is it always even with EFI nowadays ?). On other architectures that is not necessarily the case. The firmware can (and will) initialize the card using MMIO entirely and even if possible disabling the legacy stuff, which means that turning those bits on in the bridge is also unnecessary. In fact, on such setups, the isn't really a concept of a "primary" video card to begin with. On the other hand, that also means that a video card initialized like that is pretty much out of the grasp of the vga arbiter which has no effect on it either. Also be careful that while it may be relevant on x86, the VGA fwd bit is not on the PCIe root complex on IBM machines. Finally, P8 has no IO space at all... > x86/ia64 set IORESOURCE_ROM_SHADOW, which powerpc doesn't. > > ia64 doesn't call vga_set_default_device(), x86 and powerpc do. > > So we'll merge this and maybe someone can tease out the common bits, but > personally I don't see that there's an obvious chunk of generic logic. Cheers, Ben. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
