Re: PCI resource problems caused by improper address rounding

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

> That won't work, because PCI_BASE_ADDRESS_MEM_TYPE_64 controls how
> many bits need to be written back to the BAR.  If we changed that
> to PCI_BASE_ADDRESS_MEM_TYPE_32, we wouldn't clear the high 32-bits
> of the BAR.
> 
> > ... and that would be an X server issue!).
> 
> Of course, fixing the X server to *handle* 64-bit BARs is the correct
> solution.  I've no idea how involved that is, but I have a sneeking
> suspicion that it uses that damned CARD32 datatype for everything.

A lot more than X needs to be fixed to handle 64-bit BARs btw. There's a
whole load of places in drivers/pci/* where we just puke if we see a
value >4G being assigned.

Now, there is some hope that the new X with libpciaccess can cope with
that, and even if it is broken, it would be much easier to fix, as X in
that case is no longer trying to bypass the kernel, but instead uses
proper kernel interfaces to map device resources.

That used to be Xorg pci-rework branch, though it might have been merged
in the  trunk by now.
 
Ben.