[v6][PATCH 0/7] xen: reserve RMRR to avoid conflicting MMIO/RAM

[Tiejun Chen <tiejun.chen@intel.com>]

v6:

* Jan cleanup to regenerate one patch to replace the two original patches
  then rebase the others.
* Refine all patches short/long logs.
* Replace those info with 'reserved device memory maps' and s/RMRR/RDM. 
* Fix some code styles.
* Fix one interation error when we grab e820 info
  This is not valid when j == nr - 1 (last iteration). 
* Cleanup some codes.
* One comment to introduce a new field, nr_reserved_device_memory_map
  libxc/hvm_info_table, is still pending to wait the tools maintainer's
  further comment.

v5:

* Add patch #2 to introduce a global count, acpi_rmrr_unit_entries. 
* Refine hypercall return value to make sure the caller can distinguish
  clearly between "xen filled in N entries" and "xen said you need N
  entries for all information".
* Refine some structures
* Then Rebase 

v4:

* Drop the original patch #1. Instead, we use acpi_rmrr_units to get
  rmrr info directly.
* Refine the hypercall definition to make sure we can use it safely.
* Introduce introduce nr_reserved_device_memory_map in hvm_info_table,
  then we can avoid issue unnecessary hypercall, even we can know
  current RMRR entries to issue hypercall one time.
* Cleanup and rebase

v3:

* Use XENMEM_reserved_device_memory_map to replace XENMEM_RMRR_memory_map
* Then rebase all patches

v2:

* Don't use e820map to define RMRR maps directly to avoid any confusion.
* In patch #3 we introduce construct_rmrr_e820_maps() to check if we can
  insert RMRR maps and then we will sort all e820 entries.
* Clean patch #4
* In patch #5 we reuse check_mmio_hole() to check if current mmio range is
  fine to RMRR maps. If not, we just issue error to notify the user since
  mostly mmio should be configured again.

While we work for supporting RMRR mapping for Windows GFX driver in case
shared table,

http://osdir.com/ml/general/2014-07/msg55347.html
http://osdir.com/ml/general/2014-07/msg55348.html

we realize we should reserve RMRR range to avoid any potential MMIO/RAM
overlap with our discussion so here these preliminary patches are intended
to cover this.

----------------------------------------------------------------
Jan Beulich (1):
      introduce XENMEM_reserved_device_memory_map

Tiejun Chen (6):
      tools/libxc: introduce hypercall for xc_reserved_device_memory_map
      tools/libxc: check if mmio BAR is out of reserved device memory maps
      libxc/hvm_info_table: introduce a new field nr_reserved_device_memory_map
      hvmloader: introduce hypercall for xc_reserved_device_memory_map
      hvmloader: check to reserved device memory maps in e820
      xen/vtd: make USB RMRR mapping safe

 tools/firmware/hvmloader/e820.c         | 100 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/firmware/hvmloader/util.c         |  22 ++++++++++++++++++++++
 tools/firmware/hvmloader/util.h         |   3 +++
 tools/libxc/xc_domain.c                 |  29 +++++++++++++++++++++++++++++
 tools/libxc/xc_hvm_build_x86.c          |  82 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++--
 tools/libxc/xenctrl.h                   |   4 ++++
 xen/common/compat/memory.c              |  52 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 xen/common/memory.c                     |  49 +++++++++++++++++++++++++++++++++++++++++++++++++
 xen/drivers/passthrough/iommu.c         |  10 ++++++++++
 xen/drivers/passthrough/vtd/dmar.c      |  17 +++++++++++++++++
 xen/drivers/passthrough/vtd/extern.h    |   1 +
 xen/drivers/passthrough/vtd/iommu.c     |   9 +--------
 xen/include/public/hvm/hvm_info_table.h |   3 +++
 xen/include/public/memory.h             |  24 +++++++++++++++++++++++-
 xen/include/xen/iommu.h                 |   4 ++++
 xen/include/xlat.lst                    |   3 ++-
 16 files changed, 400 insertions(+), 12 deletions(-)

Thanks
Tiejun