Re: Debian Crash
On Mon, 16 Nov 1998, Rob Collins wrote:
> On the same topic, I've recently had a crash in debian
[...]
> 1. boot with xdm. Log in.
> 2. <Ctrl><Alt><F[1...6]> to switch to a different terminal.
> 3. Log in, do my stuff, log out.
> 4. <Ctrl><Alt><F[7,8]> to either return to X or look at the console log...
> it will never get to X (or the console log)... the video freezes blank and
> that's that.
[...]
> I doubt anyone could tell, just by description, what that problem is
There can be a problem with switching into and out of X.
Unfortunately, almost all PC video hardware is really, really stupid.
There's no way to read the card and figure out what video mode it's in.
You can't just reset it, either. If you don't know what mode it's in, you
can't safely change it. If you try to change the mode and you're wrong
about what mode it's currently in, you'll probably lock up the card. And
only powering down will get you out of it.
What this means is you can get into race conditions when you switch video
modes. Switching from X to VC and back again can trigger it. The VC code
starts running before X has finished letting go of the card, or vice
versa, and blammo, your display is locked until you power down.
These races cannot be eliminated, I'm told, given the current design of
Linux and X. There are developments in progress to change this, notably
the kernel framebuffer idea, where the kernel controls all access to the
video card and presents a virtual interface to other programs. (This would
actually help security since X wouldn't have to run as root... but might
hurt performance.)
What I'd suggest is making sure you're running absolutely the latest X
server for your video card. Some of these races have been at least
minimized (though as I said, not eliminated)in later version of the X
servers.
Sincerely,
Ray Ingles (248) 377-7735 ray.ingles@fanucrobotics.com
Anagrams of "FANUC Robotics North America":
bun of characteristic maroon omaha reconstruction fabric
a cubic transformation chore cherub fractionation macros
cube of anachronistic mortar
Reply to: