OSDev.org

is it possible to do this? I realize that you can setup a double fault stack but that doesn't help. the idea is i would like to have a way of detecting kernel stack overruns by using gaurd pages, but it seems impossible to do right.

proxy

proxy wrote: is it possible to do this? I realize that you can setup a double fault stack but that doesn't help. the idea is i would like to have a way of detecting kernel stack overruns by using gaurd pages, but it seems impossible to do right.

proxy

Why doesn't setting up a double fault handler help? If you set up a separate task with its own stack that is always resident, and use a task gate, it should work AFAIK... Ugly, but doable.

but i can't resume from that, another long term goal i have in mind is to grow the kernel stack as needed (up to a limit).

proxy

Well, you can't do this using a trap gate or interrupt gate, because when it gets a page fault on the stack, the processor will try to push the return address and error code onto the stack, and because the stack is not valid...*boom*! You'll get a double fault, and the processor will once again try to push stuff onto the stack, so it'll triple fault.

As far as I know, the only way to do this on x86 is with a task gate. You need to set up a TSS with a different stack from your kernel's main one. Then make your page fault handler a task gate to that. Whenever there's a page fault, you'll end up on the other stack, where you can handle it.

By the way, this problem doesn't seem to exist on most other processors - most of them put the return address in a register rather than on the stack, so there's no need to have a valid stack at all. x86-64 has a flag which can be set to tell the processor to switch stacks on some particular interrupt, so there's no need there for the TSS.

ahh, i'll have to look at that, i like it as a solution, thanks.

proxy