This is with 20010124 rawhide and a Dell Optiplex GX110 with onboard i810 with an add-in nvidia pci card. Boot with bootnet and don't get framebuffer console. Select NFS install. Following happens as anaconda crashes out (typed by hand, so ignore my typos :) Provdes X server is XFree86 Card:RIVA TNT2 We can use framebuffer fb_check returned with value Card: RIVA TNT2 We can use framebuffer 0 Using Xserver /usr/X11R6/bin/XF86_FBDev Can't open /dev/fb0 Waiting for X server to start... X server started successfully. Gdk-ERROR **: Fatal IO error (Connection reset by peer) on X server :1. install exited abnormally and unmounts fs and allows me to reboot from there. Should at least fall back gracefully to text mode if XF86_FBDev fails to start
This defect is considered MUST-FIX for Florence Release-Candidate #1
This is strange because it is supposed to fall back to text mode. Actually, anaconda detects if you are in framebuffer mode and if not, then it starts a native X server such as SVGA. If that fails, only then does it fall back to text mode. I have a Dell OptiPlex GX110 test machine here that works correctly, but I am using the onboard (i810) video card. Can you try removing the extra video card and see what happens then? We've had problems in the past detecting the secondary video card on i810 systems in the past.
Removal of the add-in card does have the frame buffer server start properly, so it is related to the second video card. Unfortunately saying to remove teh second video card is probably not the best thing to tell to "normal users" :)
I agree totally. What surprises me is that your i810 card can start framebuffer at all. Mine can't. Are you sure it didn't fall back to the native X server instead? Go to VC2 and type 'cat /dev/fb0' If a bunch of text scrolls across the screen you're in framebuffer mode. If nothing happens, it must have started the native X server. On my test machine, I've added a second video card, but I can't get a signal from it. Did you set anything in your bios to send the video signal to the second card?
Sorry, openafs + 2.4 is screwing with my head and I just can't read today... it just starts the SVGA server. As to enabling a secondary video card, check in the BIOS -> Integrated Devices -> Primary Video Controller -> Auto. I think that should be all you need, though this is how it was shipped by Dell so I never changed it.
Strange. I've tried two different video cards and I've set the bios just like you described and no luck. I can't get any signal at all through the secondary card. The only options are Auto and Onboard, so I assumed Auto is what it needed to be, but that doesn't work. It's hard for me to fix the problem if I don't have a working test case. What pci slot are you putting the video card in? I know this probably shouldn't make a difference, but I'm out of other ideas.
Slot on the far left as you look at the back of the box. If you can't get the add-on card to work, I'll take a deeper look into the problem this weekend / beginning of net week since I do have an easily reproducible test case
Well, I looked around on the Dell site and found about 1000 people on their message boards with the same problem that I'm having. Turns out there is a bios update to fix this problem. You must have a newer bios than me. Anyway, the bios downloads are all in .exe format, even if you pick Red Hat Linux as your operating system. I don't have any windows machines around, so I can't fix this problem until I can get to my machine at home with vmWare. Arrg. I'll do that over the weekend.
Ok, I tried updating the bios of my machine and it doesn't fix the problem. I'm inclined to think that problems are related to the bios. Please try with beta 3 (Fisher).
Still happens with beta3 but I've got a much better handle on what's actually going on now (and why it didn't happen in 7.0) Stage 2 loads and checks the framebuffer. Obviously, no framebuffer so this fails. Note that fb_check() shows that the card in the machine is the TNT2 because fb_check() always takes the last card returned from the kudzu probe. Then, when the Xconfiguration is created, it instead uses the first card and so sets up the test XF86Config for the i810. X tries to start and fails due to the error in the XF86Config, and it fails after the gui tries to start, so then starting the gui fails and we die. The quick fix to at least stop anaconda from falling over and to get it to fallback to text mode is to add a short sleep after starting the X server (around line 308 of anaconda/anaconda). As to why it didn't happen in 7.0, in 7.0, XF86_SVGA was used as the X server instead of XFree86 and so it didn't care what the card was since it was the same driver either way. The real solution is fixing up the handling of multiple video cards, but that's probably not going to happen for this release :) Also, for some cross-referencing, this is related to bug 24468 (and the ones it's related to)
I think there are some inherent problems with having two or more video cards in your machine. We don't currently support this. Xconfigurator and kudzu need more intelligent probing routines in order to handle this. Deferring to a future release.
Heh, luckily it seems that the changes for disabled pci devices appears to have fixed the problems from the little bit of testing I did earlier today but just hadn't gotten around to dropping in bugzilla yet. The issue of multiple video cards does need looking at for a future release though since it's getting to be an extremely common setup.
*** This bug has been marked as a duplicate of 16123 ***