Ping.

-----Original Message-----
From: Qemu-devel [mailto:qemu-devel-bounces+zhi.a.wang=intel....@nongnu.org] On 
Behalf Of Zhi Wang
Sent: Monday, May 29, 2017 4:29 PM
To: qemu-devel@nongnu.org
Cc: Tian, Kevin <kevin.t...@intel.com>; daniel.vet...@ffwll.ch; 
joonas.lahti...@linux.intel.com; zhen...@linux.intel.com; Alex Williamson 
<alex.william...@redhat.com>; Lv, Zhiyuan <zhiyuan...@intel.com>; Dong, 
Chuanxiao <chuanxiao.d...@intel.com>; Zhang, Xiong Y <xiong.y.zh...@intel.com>
Subject: Re: [Qemu-devel] [RFC 0/6] Refine IGD stolen memory support in VFIO

Hi Alex:
     This the RFC patch of my ides in previous email. Could you shed some 
lights here? :)

Thanks,
Zhi.

On 05/30/17 01:30, Zhi Wang wrote:
> This patch series are aim to refine the IGD stolen memory support in QEMU.
>
> Some features of Intel HW relies on the stolen memory to work, but 
> some of registers used to indicate the base address of stolen memory 
> can only be programmed once after system is powered-up. This will 
> cause problems if the base of the stolen memory is changed during IGD 
> passthrough.
>
> The solution of this patch series is allocating a new VM dedicated 
> memory as "guest stolen memory" in QEMU for isolation and let the 
> "guest stolen memory" sits in the same PA as host.
>
> Also it introduces Intel platform definitions and clears some old 
> funtions and definitions.
>
> Zhi Wang (6):
>    vfio: Add Intel platform definitions
>    vfio: Setup IGD quirks earlier
>    vfio: Setup IGD stolen memory
>    vfio: Refine GGTT clearing
>    vfio: Remove extra quirks and old definitions
>    vfio: Setup IGD stolen memory under secondary mode
>
>   hw/vfio/Makefile.objs    |   2 +-
>   hw/vfio/intel-platform.c | 366 
> +++++++++++++++++++++++++++++++++++++++++++++++
>   hw/vfio/intel-platform.h |  38 +++++
>   hw/vfio/pci-quirks.c     | 322 ++++++++---------------------------------
>   4 files changed, 463 insertions(+), 265 deletions(-)
>   create mode 100644 hw/vfio/intel-platform.c
>   create mode 100644 hw/vfio/intel-platform.h
>


Reply via email to