PCx86 User with PCI Device Needed for Testing

gregory.menke at gsfc.nasa.gov gregory.menke at gsfc.nasa.gov
Fri Sep 17 17:57:36 UTC 2004

"Joel Sherrill <joel at OARcorp.com>" <joel.sherrill at OARcorp.com> writes:
 > Hi,
 > I need some feedback on the path in GNATS for PR608.
 > Greg.. any comments on it.

I retrieved the diff to re-examine the issue.  I don't have an
environment to reasonably test this, but I did build-test the pc386
bsp before submitting the patch.

I wrote it as a quick-fix so the elnk and dec2114x drivers would
compile on pc386, where they should also reasonably be expected to

BusCountPCI() tries to find the # of PCI busses on the system;
basically it tries to find the maximum bus number & uses that as a
basis.  If the pci scanning functions don't return something
reasonable (> 0 and < 255), it assumes 1 bus in an attempt to preserve
some semblance of sanity- if there is PCI related code in the bsp, you
should reasonably expect at least 1 pci bus to be present.  The
function relies on existing i386 pci functions for the bus probes, so
its not too gruesome a hack.

I think a minimal test would be for someone who uses the i386 bsp on
hardware where the pci bus is supported to print the result of
BusCountPCI() after bootup.  A good test would be to put a 3com or dec
tulip board in the system and try out the driver.


More information about the users mailing list