From: Jeff D. <jd...@ka...> - 1999-12-21 20:27:39
|
Basically, the kernel doesn't boot on them. I didn't test this explicitly because the filesystem that I do most of my testing on is a mkrootfs-made filesystem from a while ago. I've added a bunch of things by hand (like swapping, virtual console, Apache) since then, and something in there is making it a lot less likely that the kernel hits the bug. As far as I've diagnosed it, a process ends up with a page mapped that its page tables don't know about. It's a page left over from something else, and it contains garbage, as far as the process is concerned. Jeff |