Re: Requesting for freeze exception for RMRR

[Wei Liu <wei.liu2@citrix.com>]

On Mon, Jul 13, 2015 at 02:31:58PM +0800, Chen, Tiejun wrote:
> Hi Wei,
> 
> Here I'm trying to request the freeze exception for RMRR.
> 
> 1. clarify the state of patch series / feature.
> 
> Reviewed	Acked	RMRR series v7
> Y		Y	[v7][PATCH 01/16] xen: introduce XENMEM_reserved_device_memory_map
> Y		Y	[v7][PATCH 02/16] xen/vtd: create RMRR mapping
> Y		N	[v7][PATCH 03/16] xen/passthrough: extend hypercall to support rdm
> reservation policy
> Y		Y	[v7][PATCH 04/16] xen: enable XENMEM_memory_map in hvm
> Y		N	[v7][PATCH 05/16] hvmloader: get guest memory map into memory_map[]
> Y		N	[v7][PATCH 06/16] hvmloader/pci: skip reserved ranges
> Y		N	[v7][PATCH 07/16] hvmloader/e820: construct guest e820 table
> Y		Y	[v7][PATCH 08/16] tools/libxc: Expose new hypercall
> xc_reserved_device_memory_map
> Y		Y	[v7][PATCH 09/16] tools: extend xc_assign_device() to support rdm
> reservation policy
> Y		Y	[v7][PATCH 10/16] tools: introduce some new parameters to set rdm
> policy
> Y		Y	[v7][PATCH 11/16] tools/libxl: detect and avoid conflicts with RDM
> Y		Y	[v7][PATCH 12/16] tools: introduce a new parameter to set a predefined
> rdm boundary
> Y		Y	[v7][PATCH 13/16] libxl: construct e820 map with RDM information for
> HVM guest
> Y		Y	[v7][PATCH 14/16] xen/vtd: enable USB device assignment
> Y		Y	[v7][PATCH 15/16] xen/vtd: prevent from assign the device with shared
> rmrr
> Y		Y	[v7][PATCH 16/16] tools: parse to enable new rdm policy parameters	

Please work with maintainers to get those hvmloader patches acked or
reviewed.

> 
> Note Jackson and Campbell also raised some comments to improve current
> codes.
> 
> 2. explain why it needs to be in this release (benefits).
> 
> RMRR mechanism was broken for a long time and this makes VM always face
> security issues. In addition, those associated devices can't be passed
> through to VM and even result in VM crashes.
> 
> 3. explain why it doesn't break things (risks).
> 
> Our policy makes sure that system will work in the original way by default
> as without the RMRR patches. And especially, this series just impacts those
> platforms which have RMRR.
> 

Your patches touch crucial path in hvmloader to build memory map. There
is risk that it may break hvmloader even if it is turned off.

I would need at least some positive test results from you to confirm if
this feature is turned off everything works as before.

Wei.

> 4. CC relevant maintainers and release manager.
> Done
> 
> Cheers,
> Tiejun