Date: Tue, 10 Feb 2026 10:20:08 -0800 From: Mark Millard <marklmi@yahoo.com> To: Andrew Turner <andrew@FreeBSD.org>, dev-commits-src-all@FreeBSD.org, dev-commits-src-main@FreeBSD.org Subject: Re: git: 1ea0721e1a56 - main - vtgpu: Support virtio gpu on Parallels Desktop Message-ID: <5ac3abe1-0a95-414b-9c5e-9f2ba03854eb@yahoo.com> In-Reply-To: <698b65c4.2198b.4fc2ec98@gitrepo.freebsd.org>
index | next in thread | previous in thread | raw e-mail
On 2/10/26 09:07, Andrew Turner wrote: > The branch main has been updated by andrew: > > URL: https://cgit.FreeBSD.org/src/commit/?id=1ea0721e1a566fdb552b0a919c22667844a894d9 > > commit 1ea0721e1a566fdb552b0a919c22667844a894d9 > Author: Andrew Turner <andrew@FreeBSD.org> > AuthorDate: 2026-02-10 17:06:31 +0000 > Commit: Andrew Turner <andrew@FreeBSD.org> > CommitDate: 2026-02-10 17:06:31 +0000 > > vtgpu: Support virtio gpu on Parallels Desktop > > The Parallels Desktop Virtio GPU implementation doesn't handle > enqueuing the VIRTIO_GPU_CMD_RESOURCE_ATTACH_BACKING request and > memory list together. > > Work around this by splitting them before sending them to be enqueued. > > Sponsored by: Arm Ltd > Differential Revision: https://reviews.freebsd.org/D55147 > --- > sys/dev/virtio/gpu/virtio_gpu.c | 51 ++++++++++++++++++++++++++++++----------- > 1 file changed, 38 insertions(+), 13 deletions(-) > > diff --git a/sys/dev/virtio/gpu/virtio_gpu.c b/sys/dev/virtio/gpu/virtio_gpu.c > index 668eb170304a..b95bae9fe286 100644 > --- a/sys/dev/virtio/gpu/virtio_gpu.c > +++ b/sys/dev/virtio/gpu/virtio_gpu.c > @@ -445,21 +445,33 @@ vtgpu_alloc_virtqueue(struct vtgpu_softc *sc) > } > > static int > -vtgpu_req_resp(struct vtgpu_softc *sc, void *req, size_t reqlen, > - void *resp, size_t resplen) > +vtgpu_req_resp2(struct vtgpu_softc *sc, void *req1, size_t req1len, > + void *req2, size_t req2len, void *resp, size_t resplen) > { > struct sglist sg; > - struct sglist_seg segs[2]; > - int error; > + struct sglist_seg segs[3]; > + int error, rcount; > > - sglist_init(&sg, 2, segs); > + sglist_init(&sg, 3, segs); > > - error = sglist_append(&sg, req, reqlen); > + rcount = 1; > + error = sglist_append(&sg, req1, req1len); > if (error != 0) { > device_printf(sc->vtgpu_dev, > - "Unable to append the request to the sglist: %d\n", error); > + "Unable to append the request to the sglist: %d\n", > + error); > return (error); > } > + if (req2 != NULL) { > + error = sglist_append(&sg, req2, req2len); > + if (error != 0) { > + device_printf(sc->vtgpu_dev, > + "Unable to append the request to the sglist: %d\n", > + error); > + return (error); > + } > + rcount++; > + } > error = sglist_append(&sg, resp, resplen); > if (error != 0) { > device_printf(sc->vtgpu_dev, > @@ -467,7 +479,7 @@ vtgpu_req_resp(struct vtgpu_softc *sc, void *req, size_t reqlen, > error); > return (error); > } > - error = virtqueue_enqueue(sc->vtgpu_ctrl_vq, resp, &sg, 1, 1); > + error = virtqueue_enqueue(sc->vtgpu_ctrl_vq, resp, &sg, rcount, 1); > if (error != 0) { > device_printf(sc->vtgpu_dev, "Enqueue failed: %d\n", error); > return (error); > @@ -479,6 +491,13 @@ vtgpu_req_resp(struct vtgpu_softc *sc, void *req, size_t reqlen, > return (0); > } > > +static int > +vtgpu_req_resp(struct vtgpu_softc *sc, void *req, size_t reqlen, > + void *resp, size_t resplen) > +{ > + return (vtgpu_req_resp2(sc, req, reqlen, NULL, 0, resp, resplen)); > +} > + > static int > vtgpu_get_display_info(struct vtgpu_softc *sc) > { > @@ -559,9 +578,15 @@ static int > vtgpu_attach_backing(struct vtgpu_softc *sc) > { > struct { > + /* > + * Split the backing and mem request arguments as some > + * hypervisors, e.g. Parallels Desktop, don't work when > + * they are enqueued together. > + */ > struct { > struct virtio_gpu_resource_attach_backing backing; > - struct virtio_gpu_mem_entry mem[1]; > + char pad; > + struct virtio_gpu_mem_entry mem; > } req; > char pad; > struct virtio_gpu_ctrl_hdr resp; > @@ -577,11 +602,11 @@ vtgpu_attach_backing(struct vtgpu_softc *sc) > s.req.backing.resource_id = htole32(VTGPU_RESOURCE_ID); > s.req.backing.nr_entries = htole32(1); > > - s.req.mem[0].addr = htole64(sc->vtgpu_fb_info.fb_pbase); > - s.req.mem[0].length = htole32(sc->vtgpu_fb_info.fb_size); > + s.req.mem.addr = htole64(sc->vtgpu_fb_info.fb_pbase); > + s.req.mem.length = htole32(sc->vtgpu_fb_info.fb_size); > > - error = vtgpu_req_resp(sc, &s.req, sizeof(s.req), &s.resp, > - sizeof(s.resp)); > + error = vtgpu_req_resp2(sc, &s.req.backing, sizeof(s.req.backing), > + &s.req.mem, sizeof(s.req.mem), &s.resp, sizeof(s.resp)); > if (error != 0) > return (error); > > > Up to now, for the aarch64 macOS context, I've been using the below /boot/device.hints to only get the console that results from the change. since virtio_pci1 was not working. That has been sufficient for my context: Mostly I ssh in and do not use the console for much. (The long "virtio_pci1:" comment line below will wrap twice.) # cat /boot/device.hints # This is for virtio_gpu --for avoiding its use under Parallels: # dmesg -a | grep -i "virtio.*gpu" # virtio_pci1: <VirtIO PCI (modern) GPU adapter> mem 0x10000000-0x17ffffff,0x18008000-0x18008fff,0x18000000-0x18003fff at device 10.0 on pci0 hint.virtio_pci.1.disabled="1" Should the change committed mean that things would likely work if I then comment out the hint.virtio_pci.1.disabled assignment (once I manage to upgrade)? -- === Mark Millard marklmi at yahoo.comhome | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5ac3abe1-0a95-414b-9c5e-9f2ba03854eb>