Messages in this thread | | | Date | Thu, 18 Jun 2009 11:31:21 -0700 | From | Chris Wright <> | Subject | Re: [PATCH 1/2] IOMMU Identity Mapping Support: iommu_identity_mapping definition |
| |
* Muli Ben-Yehuda (muli@il.ibm.com) wrote: > On Thu, Jun 18, 2009 at 11:14:51AM -0700, Yu, Fenghua wrote: > > > > > >On Thu, Jun 18, 2009 at 11:05:14AM -0700, Fenghua Yu wrote: > > > > > >> IOMMU Identity Mapping Support: iommu_identity_mapping definition > > >> > > >> Identity mapping for IOMMU defines a single domain to 1:1 map all > > >> pci devices to all usable memory. > > > > > >Why use VT-d at all in this case? Do you have a use-case in mind? > > > > Some users want to use VT-d in KVM but are concerned of DMA > > remapping performance. They can use identity mapping and still have > > KVM on VT-d. They can also use pass through patch (sent out before) > > if hardware supports pass through. > > Sorry, I must be missing something. For the normal device assignment > case, we want the IOMMU page tables to have gpa->hpa mappings rather > than the 1-1 identity mapping. How do you envision the 1-1 mapping > being used in the device assignment case?
The 1-1 mapping is for all the host devices _not_ assigned to guests. To eliminate the i/o overhead imposed on all guests not using an assigned device or from i/o from host.
It's just the same as VT-d PassThrough mode for hardware that doesn't support it.
thanks, -chris
| |