From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756047AbbLDKNj (ORCPT ); Fri, 4 Dec 2015 05:13:39 -0500 Received: from mx1.redhat.com ([209.132.183.28]:39764 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755177AbbLDKNf convert rfc822-to-8bit (ORCPT ); Fri, 4 Dec 2015 05:13:35 -0500 Message-ID: <1449224012.18669.55.camel@redhat.com> Subject: Re: [Intel-gfx] [Announcement] 2015-Q3 release of XenGT - a Mediated Graphics Passthrough Solution from Intel From: Gerd Hoffmann To: "Tian, Kevin" Cc: Alex Williamson , "Song, Jike" , "xen-devel@lists.xen.org" , "igvt-g@ml01.01.org" , "intel-gfx@lists.freedesktop.org" , "linux-kernel@vger.kernel.org" , "White, Michael L" , "Dong, Eddie" , "Li, Susie" , "Cowperthwaite, David J" , "Reddy, Raghuveer" , "Zhu, Libo" , "Zhou, Chao" , "Wang, Hongbo" , "Lv, Zhiyuan" , qemu-devel , Paolo Bonzini Date: Fri, 04 Dec 2015 11:13:32 +0100 In-Reply-To: References: <53D215D3.50608@intel.com> <547FCAAD.2060406@intel.com> <54AF967B.3060503@intel.com> <5527CEC4.9080700@intel.com> <559B3E38.1080707@intel.com> <562F4311.9@intel.com> <1447870341.4697.92.camel@redhat.com> <1447922452.25140.39.camel@redhat.com> <1448007960.6904.22.camel@redhat.com> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, > btw some questions here: > > for non-gl and gl rendering in Qemu, are they based on dma-buf already? > once we can export guest framebuffer in dma-buf, is there additional work > required or just straightforward to integrate with SPICE? Right now we are busy integrating dma-buf support into spice, which will be used for the gl rendering path, for virtio-gpu. For intel-vgpu the wireup inside qemu will be slightly different: We'll get a dma-buf handle from the igd driver, whereas virtio-gpu renders into a texture, then exports that texture as dma-buf. But in both cases we'll go pass the dma-buf with the guest framebuffer (and meta-data such as fourcc and size) to spice-server, which in turn will pass on the dma-buf to spice-client for (local) display. So we have a common code path in spice for both virtio-gpu and intel-vgpu, based on dma-bufs. spice-server even doesn't need to know what kind of graphics device the guest has, it'll go just process the dma-bufs. longer-term we also plan to support video-encoding for a remote display. Again based on dma-bufs, by sending them to the gpu video encoder. The non-gl rendering path needs to be figured out. With virtio-gpu we'll go simply turn off 3d support, so the guest will fallback to do software rendering, we'll get a classic DisplaySurface and the vnc server can work with that. That isn't going to fly with intel-vgpu though, so we need something else. Import dma-buf, then glReadPixels into a DisplaySurface would work. But as mentioned before I'd prefer a code path which doesn't require opengl support in qemu, and one option for that would be the special vfio region. I've written up a quick draft meanwhile: diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h index 751b69f..91b928d 100644 --- a/include/uapi/linux/vfio.h +++ b/include/uapi/linux/vfio.h @@ -596,6 +596,28 @@ struct vfio_iommu_spapr_tce_remove { }; #define VFIO_IOMMU_SPAPR_TCE_REMOVE _IO(VFIO_TYPE, VFIO_BASE + 20) +/* -------- Additional API for vGPU -------- */ + +/* + * framebuffer meta data + * subregion located at the end of the framebuffer region + */ +struct vfio_framebuffer { + __u32 argsz; + + /* out */ + __u32 format; /* drm fourcc */ + __u32 offset; /* relative to region start */ + __u32 width; /* in pixels */ + __u32 height; /* in pixels */ + __u32 stride; /* in bytes */ + + /* in+out */ +#define VFIO_FB_STATE_REQUEST_UPDATE 1 /* userspace requests update */ +#define VFIO_FB_STATE_UPDATE_COMPLETE 2 /* kernel signals completion */ + __u32 state; /* VFIO_FB_STATE_ */ +}; + /* ***************************************************************** */ #endif /* _UAPIVFIO_H */ cheers, Gerd