[Openchrome-devel] PCI bus corruption?

[Peter Stuge]

I saw the start of this on the b43 list.

Yuval Hager wrote:
> When the screen is waken up after blanking (e.g. by calling 'xset
> dpms force standby' and then moving the mouse a bit) the output
> from lspci on a specific card becomes all 1's.

I asked for some more output and noticed this:

00:01.0 PCI bridge: VIA Technologies, Inc. VT8237 PCI Bridge...
	Memory behind bridge: fbf00000-fcffffff

00:02.0 PCI bridge: VIA Technologies, Inc. P4M900 PCI to PCI...
	Memory behind bridge: fd000000-fdffffff

01:00.0 VGA compatible controller: VIA Technologies, Inc. Chrome9...
	Memory at fc000000 (32-bit, non-prefetchable) [size=16M]

02:00.0 Network controller: Broadcom Corporation BCM4312 802.11a/b/g...
	Memory at fdffc000 (64-bit, non-prefetchable) [size=16K]

So, the bridge to b43 lies immediately after the Chrome MMIO space,
and the problem can be provoked when asking openchrome to do things
which likely generate a bit of scattered IO (panel off/on).

My question is if there are a few strategic places in openchrome
where MMIO accesses could be logged and/or validated to actually be
within the device's MMIO range, to find a starting point for further
debugging of this problem?

I grabbed openchrome svn just a while ago but I haven't looked at the
code before. grep -ri pci gives a lot of hits.


//Peter