Re: [RFC PATCH 6/6] drm/i915/gvt: support QEMU getting the dmabuf

[Alex Williamson]

On Fri, 12 May 2017 06:56:03 +0000
"Chen, Xiaoguang" <xiaoguang.chen@xxxxxxxxx> wrote:

> Hi Gerd,
> 
> >-----Original Message-----
> >From: intel-gvt-dev [mailto:intel-gvt-dev-bounces@xxxxxxxxxxxxxxxxxxxxx] On
> >Behalf Of Gerd Hoffmann
> >Sent: Thursday, May 11, 2017 9:28 PM
> >To: Chen, Xiaoguang <xiaoguang.chen@xxxxxxxxx>
> >Cc: Tian, Kevin <kevin.tian@xxxxxxxxx>; intel-gfx@xxxxxxxxxxxxxxxxxxxxx; linux-
> >kernel@xxxxxxxxxxxxxxx; zhenyuw@xxxxxxxxxxxxxxx; Alex Williamson
> ><alex.williamson@xxxxxxxxxx>; Lv, Zhiyuan <zhiyuan.lv@xxxxxxxxx>; intel-gvt-
> >dev@xxxxxxxxxxxxxxxxxxxxx; Wang, Zhi A <zhi.a.wang@xxxxxxxxx>
> >Subject: Re: [RFC PATCH 6/6] drm/i915/gvt: support QEMU getting the dmabuf
> >
> >  Hi,
> >  
> >> While read the framebuffer region we have to tell the vendor driver which  
> >framebuffer we want to read? There are two framebuffers now in KVMGT that is
> >primary and cursor.  
> >> There are two methods to implement this:
> >> 1) write the plane id first and then read the framebuffer.
> >> 2) create 2 vfio regions one for primary and one for cursor.  
> >
> >(3) Place information for both planes into one vfio region.
> >    Which allows to fetch both with a single read() syscall.  
> That works too. Then using the ioctl to get the dmabuf fd if needed. And plane id can be ioctl's parameter.
> 
> How about method 2 primary plane and cursor plane are different and should generate different dmabuf for each of them.
> 
> >
> >The question is how you'll get the file descriptor then.  If the ioctl returns the
> >dma-buf fd only you have a racy interface:  Things can change between read(vfio-
> >region) and ioctl(need-dmabuf-fd).  
> You are right. So when creating the dmabuf we may have to decode the framebuffer and create the dmabuf using the latest framebuffer information and we must return the framebuffer information together with the dmabuf fd.
> 
> In the current implementation I saved the framebuffer information while user querying the framebuffer and generate the dmabuf using the saved information no error found yet but in theory there are sync problems.
> 
> >
> >ioctl(need-dma-buf) could return both dmabuf fd and plane info to fix the race,
> >but then it is easier to go with ioctl only interface (simliar to the orginal one from
> >dec last year) I think.  
> Yes. ioctl works for it.
> But based on the mail last week. If I understand correctly Alex hope to query the framebuffer information by reading the vfio device region and then get the dmabuf fd using ioctl.

No, I was explaining that I had questioned whether we could use a vfio
region in place of a separate dmabuf fd.  We can't.  I have no
particular desire to use a vfio region just for querying framebuffer
info.  I prefer the dmabuf manager fd idea that Gerd suggested.  Thanks,

Alex