Re: panic! panic! panic! sigh.. panic!

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Mon, 22 Sep 2008 09:51:50 -0400
"Martin S. Weber" <Ephaeton%gmx.net@localhost> wrote:

> On Mon, Sep 22, 2008 at 09:37:27AM +0100, Matthias Scheler wrote:
> > On Sun, Sep 21, 2008 at 01:59:40PM -0400, Martin S. Weber wrote:
> > > I thought SSE(2) would not hurt as I thought it also enables the
> > > use of SIMD Integer operations along with the MMX registers ...
> > 
> > Yes, and these require code to handle them.
> > 
> > The kernel doesn't save any of these registers on entry into the
> > kernel (via an interrupt or system call) and doesn't restore them
> > on exit. It also doesn't save and restore them on internal
> > scheduling. If you compile the kernel with MMX or SSE<n> enabled
> > the kernel will sooner or later globber some state and you get a
> > panic.
> 
> thank you for the explanation. How are we (the phaeton'ish users)
> supposed to know that? I didn't look but ... is this documented
> somewhere?
> 
It would be a good idea, I think, to put a warning
into /usr/src/BUILDING about compiler options.


                --Steve Bellovin, http://www.cs.columbia.edu/~smb