> Bob Palowoda wrote: > > John Martin wrote: > >> Bob Palowoda wrote: > >>> Just curious if any one else is running into > this. I have a couple > >>> of ECS dual core Athlon motherboards (VIA based) > with 7600gs/gtx > >>> pci-e cards. I noticed booting build 102 I get a > hard hang when the > >>> kernel is loading. kdb indicate the hard hang is > right > >>> when it loads the gfx_private module. I take the > video card out and > >>> it boots fine with > >>> the vga onboard card. Now was there some changes > that happened to > >>> this build > >>> that would involve the pci-e bus nexus layer in > combination with the > >>> nvidia pci-e card? > >> Try disabling MSIs in /kernel/drv/nvidia.conf > (uncomment the last > > line in the file). > >> This is the default starting in b103 until known > MSI interrupt > >> delivery issues > >> are fixed. (However, none so far have been > reported on VIA chipsets). > >> > >> > >> > > Disabling the MSI interrupt was one of the first > things I tried. > > Unfortunate it did not work. > > I failed to mention that during a normal boot from > the dvd it will > > reset the machine when > > it loads that driver. I did noticed it only > affected ECS motherboards > > K8M890x which are > > common these days. The nforce chipset did not show > this failure > > mode. Will wait to see > > if anybody else hits this. > Did this motherboard work with an earlier build? If > so, does it work in > b102 > with the NVDAgraphics+NVDAgraphicsr packages from > b99? > > The code in the nvidia driver that uses the vgatext > module hasn't > changed in over a year, > but I suspect the vgatext module load is just > resolving a dependency for > the driver. >
Just a followup on this. I turned the ACPI off in the bios and it works now. I suspect it does have something to do with the acpi changes in build 102. I also tried disabling the acpi-user-options but that did not cure the problem. I don't know what I lose other than the powernow features which doesn't even work on this platform to begin with. So for now it is an acceptable workaround. ---Bob -- This message posted from opensolaris.org
