Subject: Re: Floating point in the kernel
To: None <firstname.lastname@example.org>
From: Andrei Petrov <email@example.com>
Date: 09/18/1998 15:23:49
> Date: Fri, 18 Sep 1998 14:27:09 -0700 (PDT)
> From: "Eduardo E. Horvath" <firstname.lastname@example.org>
> cc: email@example.com
> On Fri, 18 Sep 1998, Andrei Petrov wrote:
> > I would expect that kernel saves/restores user's FPU state so it should be
> > safe just use FP in kernel mode assuming that you are the only part of kernel
> > who does that. Don't know what to do with interrupts from FP in kernel mode.
> First of all it is not safe to do so. Unless you know how floating pont
> registers are managed you can cause all sorts of problems. If the
> registers are disabled then using them will cause a trap. This may cause
> the dirty contents to be stored away properly, or it may cause a panic.
> Let's say you managed to get the registers saved off and enabled properly.
> If you switch back to user mode with the registers enabled then you have a
> silent user code data corruption. If you disable the fp regs you need to
> make sure that when user code attempts to access them again that the dirty
> registers you created are not mistakenly written to some process' fp
> register save area. And then what happens if your code is pre-empted and
> called back again? Or someone else comes up with your brilliant idea and
> wants to use fp code later?
> Eduardo Horvath firstname.lastname@example.org
> "I need to find a pithy new quote." -- me
I assume if you use fp in kernel you should be responsible for all what you mentioned before,
also I don't think kernel is the good place for math. One thing I wouldn't do is fp save/restore
during process context switch only; I think that safer to mark processes which use fpu and make
fp save/restore only for them, shouldn't be much overhead; then you would be more protected from lkm who wants math.
May be idea is not that brilliant but it could be fun
BTW, who is sparc64 going