No IOMMU/DMAR with DELL 3020
Slawa Olhovchenkov
slw at zxy.spb.ru
Thu Apr 9 10:50:30 UTC 2015
On Thu, Apr 09, 2015 at 01:03:35PM +0300, Konstantin Belousov wrote:
> On Thu, Apr 09, 2015 at 11:52:33AM +0300, Slawa Olhovchenkov wrote:
> > On Thu, Apr 09, 2015 at 11:35:12AM +0300, Konstantin Belousov wrote:
> >
> > > On Thu, Apr 09, 2015 at 10:28:10AM +0200, Gustau P?rez wrote:
> > > > Yup, sorry for the error. I checked the micro in the ark and it has vt-d:
> > > >
> > > > http://goo.gl/CZZRHz
> > > It only indicates that the CPU/northbridge has the hardware, but BIOS must
> > > do a work to configure it and to inform the OS about the configuration.
> > > Your BIOS did not.
> > >
> > > >
> > > > in the bios, there's only one option to enable virtualization
> > > > support, which is ticked.
> > > >
> > > > The complete log is here:
> > > >
> > > > http://dpaste.com/28FDMJQ
> > > >
> > > Dmesg would not give you any useful information there. A DMAR table
> > > is either present, or is it not. In the later case, OS cannot use the
> > > hardware, and if no option in BIOS is present, your only choice is to
> > > complain to the machine/BIOS vendor.
> >
> > May be some OS utilites can do same work?
> > This is theoretically capable?
>
> No, OS must know the peculiarities of the particular chipset. But also,
Someone may be know this and wrote support in utility.
May be chipset datashit available.
I am don't talk about 'universal, out of box support in OS'.
I am talk about theoretically utility, that perform some operations
after OS load.
Also, I am interesting by OS-control interleaving memory (in
multi-socket configuration).
This is totaly imposible or just very complex?
> it must know the intimate details of the BIOS operation, since several
> facilities perform background DMA transfers not under the OS control.
> Examples are USB legacy emulation, BMC working with the main memory,
> or UMA GPU in BIOS-configured mode. Normally, BIOS communicates the
> requirements of such facilities to OS using RMRR records in the DMAR
> table.
More information about the freebsd-current
mailing list