PCI resources above 4GB

Steven Newbury steve at snewbury.org.uk
Thu Apr 12 09:40:41 PDT 2012

On Thu, 12 Apr 2012, 17:07:33 BST, Yinghai Lu <yinghai at kernel.org> wrote:

> On Thu, Apr 12, 2012 at 4:22 AM, Steven Newbury <steve at snewbury.org.uk>
> wrote:
> > Thanks, that fixed it! :) I had a similar patch I've been working on
> > but I had my fix in the wrong place!
> > 
> > In the working case, initially the BIOS has set GMA to within the low
> > system DRAM 0xC0000000 obviously invalid.  This conflict is detected
> > and it's relallocated to 0x12000000.
> > 
> > I've attempted to modify probe.c to disable 64-bit BARs not allocated
> > above 4G so they get reallocated above when possible later.  It seemed
> > to work, but again broke GMA despite the BAR originally containing an
> > invalid address as mentioned above, it seems for some reason something
> > is different when the conflict is detected and rellocated, compared to
> > disabling it early then allocating a valid value..?
> > 
> > It would be useful to preserve as much low PCI memory address space as
> > possible for hotplug devices (like my Radeon), but the other problem
> > is small regions get allocated at the bottom, resulting in the
> > inability to find large aligned regions later on.  I see code to
> > default to top-down allocation was reverted, I guess I'm going to have
> > to dig into the archive to find out why...
> for hotplug case, You can work around like:
> after hotplug add,
> 1. use lspci and /proc/iomem to find out offending device and bridge.
> 2. use /sys/.../unbind etc to stop driver for those devices.
> 3. use setpci to clear related BAR
> 4. use /sys/devices/pci000..../remove to remove those devices
> 5. echo 1 > /sys/bus/pci/rescan
> then it should work...
Good idea, I can easily hook that up into the dock event. Is it possible to disable the auto bus scan on hotplug, then trigger it manually as above? But it still leaves me needing to have at least the Intel GMA cleanly reallocated high from boot.

More information about the dri-devel mailing list