-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, On Jul 25, 2009, at 6:23 AM, Mihai Chelaru wrote:
On Thu, 23 Jul 2009 00:10:05 -0400, Michael wrote:With my changes to libpciaccess we should be able to use any PCI graphics device on any architecture, no matter if it's the console oreven if it has any driver attached at all. Effectively it adds supportfor what the server code calls 'PCI domains'.Finally, I checkout a fresh xsrc and got X compiled and started. At firstglance everything looks OK except DRI. Looks something related to PCI BusID mismatch between X and DRM: drmOpenByBusid: Searching for BusID pci:0001:01:00.0 drmOpenDevice: node name is /dev/dri/card0 drmOpenDevice: open result is 4, (OK) drmOpenByBusid: drmOpenMinor returns 4 drmOpenByBusid: drmGetBusid reports pci:0000:01:00.0
Ouch, I need to fix that. The problem is that right now userland has no sane way to check if a bus is a subordinate of another bus or if it's a root bus so it puts every bus in its own 'domain'. Not sure if the DRM code always assigns domain 0 or if it really counts host bridges.
have fun Michael -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (Darwin) iQEVAwUBSmtqx8pnzkX8Yg2nAQLNgAf/XGIP9D0NjkVH8QXjEDQMoCQiuQH3lbdc 3LPiN73S53nViVjr/TdsYuVmw3Vu4dIhBHQMZA1Gu8Bbv20bYeFuwgV+u6onWqNk UWRDIr/PnsUKaFn7kMcv2YE/pkpfUe7xtThXf1idW9iZOXvESPbzdxtiHa+7v9zD SJ83YGdI8EJuu2SXIScmG3mPnuBvlukCAk7LXmcn1iODrFdLsvT1K+P/yRUCbUeH btcLqGiEQmfTjLOifHR35sKcCxKXpVQnIxRIg4XqPFmrt7LP4e8wosW2Yyo79ZWZ JY0Z8Wt3j6LzfsW/Eol4m245/lhTzf2tA+dW5DIBRHSCXkkoFBnDkw== =KcdT -----END PGP SIGNATURE-----