Re: [tboot-devel] VT-d PMRs

[Cihula, J. <jos...@in...>]

> From: Jonathan M. McCune [mailto:jon...@cm...]
> Sent: Thursday, October 09, 2008 10:27 AM
> 
> Hi Joe,
> 
> Cihula, Joseph wrote:
> > It is really:
> > 	Current values of VT-d PMR registers do not match requested
> > values in SinitMleData
> >
> > which means that some code has already programmed the PMRs but not
to
> > the same values that the MLE is requesting.  Because the PMRs cannot
> be
> > changed reliably once they are set/enabled, this is not an allowed
> > condition.
> 
> 
> I am successfully invoking SENTER / SEXIT and then returning control
to
> the legacy OS once per boot cycle.  When I try to execute SENTER a
> second time, the system reboots and LT.ERRORCODE is populated with the
> above error.  However, I cannot figure out what is wrong with my PMR
> values.
> 
> Is there a way to read the current values, so that I can see how they
> are set following the first SENTER? Looking through the MLE manual and
> the Sw Dev Manual Vol 2b has left me without much insight.

VT-d registers are described in the VT-d spec at:
http://download.intel.com/technology/computing/vptech/Intel(r)_VT_for_Di
rect_IO.pdf

> My MLE resides at physical address 0x00c00000 and consumes less than
> 0x200000 (2MB) bytes.
> 
> TBOOT:   vtd_pmr_lo_base: 0xc00000
> TBOOT:   vtd_pmr_lo_size: 0x200000
> TBOOT:   vtd_pmr_hi_base: 0x0
> TBOOT:   vtd_pmr_hi_size: 0x0
> 
> I'm fairly confident that the os_sinit_data_t.vtd_pmr_* values are
> being
> set identically prior to both invocations of SENTER (when I print them
> out, the above is what I see).

If you could use the VT-d spec to read and output the PMRs before
launch, that should determine if they are somehow getting changed
between invocations.

> 
> Thanks,
> -Jon