On Sunday 04 May 2008, Shachar Shemesh wrote:
> Ira Abramov wrote:
> > Howdie folks,
> >
> > My client needs to develop tools that communicate directly with QLA2xxx
> > cards. at the moment, working on regular kernels he gets the machine
> > stuck at times. I read somewhere that with Xen, there's a way to let the
> > VMs talk to the hardware directly and that will make getting unstuck a
> > bit easier.
>
> Yes (not that I know the details), but that requires a machine that
> supports IOMMU. Not many do.
It does work without an IOMMU, it just doesn't protect you from a buggy driver 
causing DMA access to the wrong address. For other kinds of driver bugs, this 
might still be better than no virtualization.

>
> >  Xen however is a bit of a chore. should/could I use KVM
> > instead? will it support the hardware access I need?
>
> AFAIR, yes.
KVM doesn't AFAIK support direct hardware access by guest VMs. This is related 
to the fact that KVM and Xen have different architectures.

Xen boots a small hypervisor which then brings up several domains including 
the privileged domain 0. The hypervisor is responsible for PCI enumeration 
and mapping devices to domains, so another domain can be given (exclusive) 
access to a device by passing the appropriate bootloader parameters to the 
hypervisor.

KVM is a Linux kernel module that sets up an ordinary Linux machine as a VM 
host, much like VMWare Workstation. The host always claims all devices at 
boot, and so guests can't be given direct access to them. (I suppose the 
architecture doesn't make this inherently impossible, but it's not 
supported.)

-- 
Dan Armak

=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to