Re: rfc: a driver to access MSRs on i386

[Mindaugas Rasiukevicius <rmind%netbsd.org@localhost>]

Joerg Sonnenberger <joerg%britannica.bec.de@localhost> wrote:
> On Sun, Jan 24, 2010 at 01:13:10PM -0600, David Young wrote:
> > Can you do without an ioctl for invoking rdmsr/wrmsr if you trap the
> > instructions and simulate them in the kernel?
> 
> That is IMO a very bad idea, especially for wrmsr. This instructions
> require proper access control and doing that in the fault handler sounds
> like a very stupid idea.

Agree - it is potentially error-prone (thus insecure).  Our trap handler
code is already quite complicated, there is no need to complicate it more,
especially with access control bits.  Also, I think it is not really what
Unix kernel, by the concept, should do.

-- 
Mindaugas