On 8/30/2013 3:09 AM, Jan Beulich wrote:

Sorry for didn't get a chance to follow up with this sooner.  Ok, I see 
> the code you mentioned.
> However, if I understand this correctly, it is also mapping the bridge 
> to the domU
> along with the end-device.  This is not the same as what you mentioned 
> that the bridge should
> be mapped to only Dom0.

I don't think I ever said this. Whether we're talking about DomU or
Dom0 here is irrelevant. I'm just pointing out that bridges do get
mappings established for them whenever a device behind them gets
mapped.

Jan

Ok, I think I understand it a bit more now. I'll modify the code to skip only the host bridge when they are not specified in the IVRS as it is not needed to be mapped to a domain. I'll send out the code shortly. However, I also found out that the current AMD IOMMU driver maps other type of bridges (i.e. DEV_TYPE_PCIe_BRIDGE, DEV_TYPE_PCIe2PCI_BRIDGE, and DEV_TYPE_LEGACY_PCI_BRIDGE) _only_ to Dom0, and they won't get reassigned when the end point devices downstream are assigned to other domain. I don't think this is correct. I'll work on this and and send out a separate patch for this. Thank you, Suravee