I have a bunch of servers, mostly Fedora Core 5, some Debian, some Ubuntu, and got tired of having as many screens as servers, so I got a Belkin 4-port KVM.

I found that in some cases if I booted a server when the KVM was not pointed at it, the server would boot (could ssh in OK), but when the KVM pointed at it, the screen would not work (just black, and obviously under those circumstances, I cannot be sure whether the keyboard or mouse works!!!)

I Googled about and it looks like using the FrameBuffer device is the way to fix this. I also found that if I appended the string "vga-0x318" or similar to the kernel command in Grub, the boot process was run through the Frame Buffer, but as soon as X-org took over the plot, I was again plunged into darkness.

What I need is the simplest, most dumbed-down howto to get framebuffer working at normal runtime, preferably on Fed Core 5, which seems to exhibit the problem most. I'm not a Unix Newbie, but I'm not *at all* knowledgeable about the video subsys, so treat me like a true dummy!

Thanks,
Jimbo

what makes you think it's the frame buffer? i'd look a lot more towards the kvm itself. what happens when you boot up a machine with the kvm pointing at it? do you see the bios status screens? does it continue to boot correctly? no real reason to look towards the framebuffer in my own experience. what kind of kvm is it? a good old fachion ps/2 kvm is fairly bomb proff but for the more recent usb only ones it gets a lot more confusing for the kvm to present the correct dummy devices to the machine when the kvm is not live.

Thanks for the quick response!

The reason for suspecting Framebuffer may be the answer is that the only place I found a precise description of the problem in Google (after *hours* of search) was on an HP site, but it described the problem precisely, and the remedy was to deploy fb...as I mentioned, I'm a total dummy on X.org, so I could be straw-clutching.

I've tried a couple of Belkins, they are the common-or-garden 4 port variety, no frills sort (no sound, no USB), one my own, where the ports are ranged round the sides of the box (2 on one side, 2 on the other, screen, key,mouse on the ends) and the other (borrowed, with all ports down 1 side)

The KVM works fine on Windoze, and better on some versions of Linux, I'm wondering what happens when the boot process lets go and passes control to the X.org when x.org can't contact the screen...its like the screen and server don't get to sync to each other???

ahh ok i think i see what you're getting at. essentially the logic is to use the most bog standard X driver as possible, so as to make it the most compatible. if you open /etc/X11/xorg.conf and find the Device section, change the Drive to "vesa" then restart X that will use the vesa 2.0 video hooks and nothing more. in fc5 in gnome you can also run system-config-display and choose vesa from the hardware list.

Already tried that....vesa works OK with a head directly connected, but barfs when KVM'ed. Incidentally, the server has an Intel mobo, with an E1000 on-board display subsys...don't know if that makes a diff?