DMARC-Filter: OpenDMARC Filter v1.3.2 mx1.redhat.com 4B4794E4C4
Message-ID: <1498459157.20591.6.camel@redhat.com>
Subject: Re: [Intel-gfx] [PATCH v9 5/7] vfio: Define vfio based dma-buf
 operations
From: Gerd Hoffmann <kraxel@redhat.com>
To: "Zhang, Tina" <tina.zhang@intel.com>,
        Alex Williamson <alex.williamson@redhat.com>
Cc: "intel-gfx@lists.freedesktop.org" <intel-gfx@lists.freedesktop.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        Kirti Wankhede <kwankhede@nvidia.com>,
        "Chen, Xiaoguang" <xiaoguang.chen@intel.com>,
        "intel-gvt-dev@lists.freedesktop.org" 
        <intel-gvt-dev@lists.freedesktop.org>,
        "Lv, Zhiyuan" <zhiyuan.lv@intel.com>,
        "Wang, Zhi A" <zhi.a.wang@intel.com>,
        "Wang, Zhenyu Z" <zhenyu.z.wang@intel.com>
Date: Mon, 26 Jun 2017 08:39:17 +0200
In-Reply-To: <237F54289DF84E4997F34151298ABEBC7C5731B3@SHSMSX101.ccr.corp.intel.com>
References: <1497513611-2814-1-git-send-email-xiaoguang.chen@intel.com>
         <1497513611-2814-6-git-send-email-xiaoguang.chen@intel.com>
         <f19da207-a7f5-d3d3-4577-b071e7d75cf7@nvidia.com>
         <1497542438.29252.1.camel@redhat.com> <20170615143833.7526351b@w520.home>
         <24c4880b-24f5-ea07-834c-c77d3e895c78@nvidia.com>
         <1497854312.4207.4.camel@redhat.com> <20170619085530.1f5e46dc@w520.home>
         <237F54289DF84E4997F34151298ABEBC7C56EBE0@SHSMSX101.ccr.corp.intel.com>
         <1497956256.16795.7.camel@redhat.com>
         <237F54289DF84E4997F34151298ABEBC7C5731B3@SHSMSX101.ccr.corp.intel.com>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1398
Lines: 37

  Hi,

> > With the generation we can also do something different:  Pass in
> > plane_type and
> > generation, and have VFIO_DEVICE_GET_DMABUF_FD return an error in
> > case
> > the generation doesn't match.  In that case it doesn't make much
> > sense any
> > more to have a separate plane_info struct, which was added so we
> > don't have
> > to duplicate things in query-plane and get- dmabuf ioctl structs.
> 
> Comparing with the current patch, this would make user space a little
> bit harder to
> get the dmabuf by calling VFIO_DEVICE_GET_DMABUF ioctl. Is it
> efficient for
> user mode usage?

user space has to call QUERY-PLANE first, then looks if it has a dma-
buf for that, if not call GET-DMABUF.

Problem is the guest could have changed the plane between the QUERY-
PLANE and GET-DMABUF ioctls.

Current patches (v8 series) just returns plane-info on GET-DMABUF too,
so userspace can at least detect something changed.

It would be easier for userspace if GET-DMABUF throws an error in case
the plane changed since the last QUERY-PLANE ioctl.  The generation id
would be one way to handle it, but possibly it is easier if the kernel
driver just keeps track internally.  So GET-DMABUF would be defined to
return a dmabuf for the plane returned by the previous QUERY-PLANE
ioctl (on the same file handle), or return an error in case the plane
has changed meanwhile.

cheers,
  Gerd