Alan Coopersmith wrote:
> X.Org is already moving from the current pci scanning to libpciaccess,
> for which Edward Shu in Sun's Beijing video driver team contributed the
> Solaris port - I don't know if the interfaces it uses solves the SPARC
> bus scanning problem or not though.
In the file libpciaccess/solx_devfs.c from the current X.org git tree
there is the following curious code:
/*
* Identify problematic southbridges. These have device id 0x5249 and
* vendor id 0x10b9. Check for revision ID 0 and class code 060400 as well.
* Values are little endian, so they are reversed for SPARC.
*
* Check for these southbridges on all architectures, as the issue is a
* southbridge issue, independent of processor.
*
* If one of these is found during probing, skip probing other devs/funcs on
* the rest of the bus, since the southbridge and all devs underneath will
* otherwise disappear.
*/
#if (NATIVE_ENDIAN == PCITOOL_ACC_ATTR_ENDN_BIG)
#define U45_SB_DEVID_VID 0xb9104952
#define U45_SB_CLASS_RID 0x00000406
#else
#define U45_SB_DEVID_VID 0x524910b9
#define U45_SB_CLASS_RID 0x06040000
#endif
... and in probe_dev()
/*
* Special case for the type of Southbridge found on
* Ultra-45 and other sun4u fire workstations.
*/
if ((config_hdr.dwords[0] == U45_SB_DEVID_VID) &&
(config_hdr.dwords[2] == U45_SB_CLASS_RID)) {
rval = ECANCELED;
break;
}
What is special about the sun4u bridge?
Pat
---
