[coyotos-dev] syscall/sysret issues

[Jonathan S. Shapiro]

On Sun, 2005-11-06 at 21:50 +0000, David Hopwood wrote:
> Jonathan S. Shapiro wrote:
> > On many earlier hardware systems, sysenter/sysexit was broken. In some
> > cases, the behavior could vary depending on the microcode stepping in
> > effect.
> 
> That particular problem could be addressed by unconditionally loading a
> fixed microcode version at boot [*]. The "virtual dynamic library" approach
> may still be needed because of the differences between Intel and AMD,
> though.
> 
> 
> [*] Only Intel systems have loadable microcode, but it is only Intel systems
>     that have the broken sysenter/sysexit.

I'm not sure that all of the historically broken processors *had*
loadable microcode, but this may be good enough for processors currently
in actual use.

shap