Re: [PATCH v5 6/9] gfxstream + rutabaga: add initial support for gfxstream

[Akihiko Odaki <akihiko.odaki@gmail.com>]

On 2023/08/16 0:50, Gurchetan Singh wrote:
> 
> 
> On Tue, Aug 15, 2023 at 8:07 AM Akihiko Odaki <akihiko.odaki@gmail.com 
> <mailto:akihiko.odaki@gmail.com>> wrote:
> 
>     On 2023/08/15 9:35, Gurchetan Singh wrote:
>      > This adds initial support for gfxstream and cross-domain.  Both
>      > features rely on virtio-gpu blob resources and context types, which
>      > are also implemented in this patch.
>      >
>      > gfxstream has a long and illustrious history in Android graphics
>      > paravirtualization.  It has been powering graphics in the Android
>      > Studio Emulator for more than a decade, which is the main developer
>      > platform.
>      >
>      > Originally conceived by Jesse Hall, it was first known as "EmuGL"
>     [a].
>      > The key design characteristic was a 1:1 threading model and
>      > auto-generation, which fit nicely with the OpenGLES spec.  It also
>      > allowed easy layering with ANGLE on the host, which provides the GLES
>      > implementations on Windows or MacOS enviroments.
>      >
>      > gfxstream has traditionally been maintained by a single engineer, and
>      > between 2015 to 2021, the goldfish throne passed to Frank Yang.
>      > Historians often remark this glorious reign ("pax gfxstreama" is the
>      > academic term) was comparable to that of Augustus and both Queen
>      > Elizabeths.  Just to name a few accomplishments in a resplendent
>      > panoply: higher versions of GLES, address space graphics, snapshot
>      > support and CTS compliant Vulkan [b].
>      >
>      > One major drawback was the use of out-of-tree goldfish drivers.
>      > Android engineers didn't know much about DRM/KMS and especially
>     TTM so
>      > a simple guest to host pipe was conceived.
>      >
>      > Luckily, virtio-gpu 3D started to emerge in 2016 due to the work of
>      > the Mesa/virglrenderer communities.  In 2018, the initial virtio-gpu
>      > port of gfxstream was done by Cuttlefish enthusiast Alistair Delva.
>      > It was a symbol compatible replacement of virglrenderer [c] and named
>      > "AVDVirglrenderer".  This implementation forms the basis of the
>      > current gfxstream host implementation still in use today.
>      >
>      > cross-domain support follows a similar arc.  Originally conceived by
>      > Wayland aficionado David Reveman and crosvm enjoyer Zach Reizner in
>      > 2018, it initially relied on the downstream "virtio-wl" device.
>      >
>      > In 2020 and 2021, virtio-gpu was extended to include blob resources
>      > and multiple timelines by yours truly, features
>     gfxstream/cross-domain
>      > both require to function correctly.
>      >
>      > Right now, we stand at the precipice of a truly fantastic
>     possibility:
>      > the Android Emulator powered by upstream QEMU and upstream Linux
>      > kernel.  gfxstream will then be packaged properfully, and app
>      > developers can even fix gfxstream bugs on their own if they encounter
>      > them.
>      >
>      > It's been quite the ride, my friends.  Where will gfxstream head
>     next,
>      > nobody really knows.  I wouldn't be surprised if it's around for
>      > another decade, maintained by a new generation of Android graphics
>      > enthusiasts.
>      >
>      > Technical details:
>      >    - Very simple initial display integration: just used Pixman
>      >    - Largely, 1:1 mapping of virtio-gpu hypercalls to rutabaga
>     function
>      >      calls
>      >
>      > Next steps for Android VMs:
>      >    - The next step would be improving display integration and UI
>     interfaces
>      >      with the goal of the QEMU upstream graphics being in an emulator
>      >      release [d].
>      >
>      > Next steps for Linux VMs for display virtualization:
>      >    - For widespread distribution, someone needs to package
>     Sommelier or the
>      >      wayland-proxy-virtwl [e] ideally into Debian main. In
>     addition, newer
>      >      versions of the Linux kernel come with DRM_VIRTIO_GPU_KMS
>     option,
>      >      which allows disabling KMS hypercalls.  If anyone cares
>     enough, it'll
>      >      probably be possible to build a custom VM variant that uses
>     this display
>      >      virtualization strategy.
>      >
>      > [a]
>     https://android-review.googlesource.com/c/platform/development/+/34470 <https://android-review.googlesource.com/c/platform/development/+/34470>
>      > [b]
>     https://android-review.googlesource.com/q/topic:%22vulkan-hostconnection-start%22 <https://android-review.googlesource.com/q/topic:%22vulkan-hostconnection-start%22>
>      > [c]
>     https://android-review.googlesource.com/c/device/generic/goldfish-opengl/+/761927 <https://android-review.googlesource.com/c/device/generic/goldfish-opengl/+/761927>
>      > [d] https://developer.android.com/studio/releases/emulator
>     <https://developer.android.com/studio/releases/emulator>
>      > [e] https://github.com/talex5/wayland-proxy-virtwl
>     <https://github.com/talex5/wayland-proxy-virtwl>
>      >
>      > Signed-off-by: Gurchetan Singh <gurchetansingh@chromium.org
>     <mailto:gurchetansingh@chromium.org>>
>      > Tested-by: Alyssa Ross <hi@alyssa.is <mailto:hi@alyssa.is>>
>      > ---
>      > v1: Incorported various suggestions by Akihiko Odaki and Bernard
>     Berschow
>      >      - Removed GET_VIRTIO_GPU_GL / GET_RUTABAGA macros
>      >      - Used error_report(..)
>      >      - Used g_autofree to fix leaks on error paths
>      >      - Removed unnecessary casts
>      >      - added virtio-gpu-pci-rutabaga.c + virtio-vga-rutabaga.c files
>      >
>      > v2: Incorported various suggestions by Akihiko Odaki, Marc-André
>     Lureau and
>      >      Bernard Berschow:
>      >      - Parenthesis in CHECK macro
>      >      - CHECK_RESULT(result, ..) --> CHECK(!result, ..)
>      >      - delay until g->parent_obj.enable = 1
>      >      - Additional cast fixes
>      >      - initialize directly in virtio_gpu_rutabaga_realize(..)
>      >      - add debug callback to hook into QEMU error's APIs
>      >
>      > v3: Incorporated feedback from Akihiko Odaki and Alyssa Ross:
>      >      - Autodetect Wayland socket when not explicitly specified
>      >      - Fix map_blob error paths
>      >      - Add comment why we need both `res` and `resource` in
>     create blob
>      >      - Cast and whitespace fixes
>      >      - Big endian check comes before virtio_gpu_rutabaga_init().
>      >      - VirtIOVGARUTABAGA --> VirtIOVGARutabaga
>      >
>      > v4: Incorporated feedback from Akihiko Odaki and Alyssa Ross:
>      >      - Double checked all casts
>      >      - Remove unnecessary parenthesis
>      >      - Removed `resource` in create_blob
>      >      - Added comment about failure case
>      >      - Pass user-provided socket as-is
>      >      - Use stack variable rather than heap allocation
>      >      - Future-proofed map info API to give access flags as well
>      >
>      > v5: Incorporated feedback from Akihiko Odaki:
>      >      - Check (ss.scanout_id < VIRTIO_GPU_MAX_SCANOUTS)
>      >      - Simplify num_capsets check
>      >      - Call cleanup mapping on error paths
>      >      - uint64_t --> void* for rutabaga_map(..)
>      >      - Removed unnecessary parenthesis
>      >      - Removed unnecessary cast
>      >      - #define UNIX_PATH_MAX sizeof((struct sockaddr_un) {}.sun_path)
>      >      - Reuse result variable
>      >      - Report error when default socket doesn't fit in Wayland
>     socket path
>      >
>      >   hw/display/virtio-gpu-pci-rutabaga.c |   48 ++
>      >   hw/display/virtio-gpu-rutabaga.c     | 1114
>     ++++++++++++++++++++++++++
>      >   hw/display/virtio-vga-rutabaga.c     |   51 ++
>      >   3 files changed, 1213 insertions(+)
>      >   create mode 100644 hw/display/virtio-gpu-pci-rutabaga.c
>      >   create mode 100644 hw/display/virtio-gpu-rutabaga.c
>      >   create mode 100644 hw/display/virtio-vga-rutabaga.c
>      >
>      > diff --git a/hw/display/virtio-gpu-pci-rutabaga.c
>     b/hw/display/virtio-gpu-pci-rutabaga.c
>      > new file mode 100644
>      > index 0000000000..c71173d8ca
>      > --- /dev/null
>      > +++ b/hw/display/virtio-gpu-pci-rutabaga.c
>      > @@ -0,0 +1,48 @@
>      > +// SPDX-License-Identifier: GPL-2.0
>      > +
>      > +#include "qemu/osdep.h"
>      > +#include "qapi/error.h"
>      > +#include "qemu/module.h"
>      > +#include "hw/pci/pci.h"
>      > +#include "hw/qdev-properties.h"
>      > +#include "hw/virtio/virtio.h"
>      > +#include "hw/virtio/virtio-bus.h"
>      > +#include "hw/virtio/virtio-gpu-pci.h"
>      > +#include "qom/object.h"
>      > +
>      > +#define TYPE_VIRTIO_GPU_RUTABAGA_PCI "virtio-gpu-rutabaga-pci"
>      > +typedef struct VirtIOGPURutabagaPCI VirtIOGPURutabagaPCI;
>      > +DECLARE_INSTANCE_CHECKER(VirtIOGPURutabagaPCI,
>     VIRTIO_GPU_RUTABAGA_PCI,
>      > +                         TYPE_VIRTIO_GPU_RUTABAGA_PCI)
>      > +
>      > +struct VirtIOGPURutabagaPCI {
>      > +    VirtIOGPUPCIBase parent_obj;
>      > +    VirtIOGPURutabaga vdev;
>      > +};
>      > +
>      > +static void virtio_gpu_rutabaga_initfn(Object *obj)
>      > +{
>      > +    VirtIOGPURutabagaPCI *dev = VIRTIO_GPU_RUTABAGA_PCI(obj);
>      > +
>      > +    virtio_instance_init_common(obj, &dev->vdev, sizeof(dev->vdev),
>      > +                                TYPE_VIRTIO_GPU_RUTABAGA);
>      > +    VIRTIO_GPU_PCI_BASE(obj)->vgpu = VIRTIO_GPU_BASE(&dev->vdev);
>      > +}
>      > +
>      > +static const VirtioPCIDeviceTypeInfo
>     virtio_gpu_rutabaga_pci_info = {
>      > +    .generic_name = TYPE_VIRTIO_GPU_RUTABAGA_PCI,
>      > +    .parent = TYPE_VIRTIO_GPU_PCI_BASE,
>      > +    .instance_size = sizeof(VirtIOGPURutabagaPCI),
>      > +    .instance_init = virtio_gpu_rutabaga_initfn,
>      > +};
>      > +module_obj(TYPE_VIRTIO_GPU_RUTABAGA_PCI);
>      > +module_kconfig(VIRTIO_PCI);
>      > +
>      > +static void virtio_gpu_rutabaga_pci_register_types(void)
>      > +{
>      > +    virtio_pci_types_register(&virtio_gpu_rutabaga_pci_info);
>      > +}
>      > +
>      > +type_init(virtio_gpu_rutabaga_pci_register_types)
>      > +
>      > +module_dep("hw-display-virtio-gpu-pci");
>      > diff --git a/hw/display/virtio-gpu-rutabaga.c
>     b/hw/display/virtio-gpu-rutabaga.c
>      > new file mode 100644
>      > index 0000000000..3f9f106abc
>      > --- /dev/null
>      > +++ b/hw/display/virtio-gpu-rutabaga.c
>      > @@ -0,0 +1,1114 @@
>      > +// SPDX-License-Identifier: GPL-2.0
>      > +
>      > +#include "qemu/osdep.h"
>      > +#include "qapi/error.h"
>      > +#include "qemu/error-report.h"
>      > +#include "qemu/iov.h"
>      > +#include "trace.h"
>      > +#include "hw/virtio/virtio.h"
>      > +#include "hw/virtio/virtio-gpu.h"
>      > +#include "hw/virtio/virtio-gpu-pixman.h"
>      > +#include "hw/virtio/virtio-iommu.h"
>      > +
>      > +#include <glib/gmem.h>
>      > +#include <rutabaga_gfx/rutabaga_gfx_ffi.h>
>      > +
>      > +#define CHECK(condition, cmd)                                   
>                   \
>      > +    do {                                                       
>                    \
>      > +        if (!(condition)) {                                     
>                   \
>      > +            error_report("CHECK failed in %s() %s:" "%d",
>     __func__,           \
>      > +                         __FILE__, __LINE__);                   
>                   \
>      > +            (cmd)->error = VIRTIO_GPU_RESP_ERR_UNSPEC;         
>                    \
>      > +            return;                                             
>                   \
>      > +       }                                                       
>                    \
>      > +    } while (0)
>      > +
>      > +/*
>      > + * This is the size of the char array in struct sock_addr_un. No
>     Wayland socket
>      > + * can be created with a path longer than this, including the
>     null terminator.
>      > + */
>      > +#ifndef UNIX_PATH_MAX
> 
>     You shouldn't need this ifndef.
> 
>      > +#define UNIX_PATH_MAX sizeof((struct sockaddr_un) {} .sun_path)
>      > +#endif
>      > +
>      > +struct rutabaga_aio_data {
>      > +    struct VirtIOGPURutabaga *vr;
>      > +    struct rutabaga_fence fence;
>      > +};
>      > +
>      > +static void
>      > +virtio_gpu_rutabaga_update_cursor(VirtIOGPU *g, struct
>     virtio_gpu_scanout *s,
>      > +                                  uint32_t resource_id)
>      > +{
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct rutabaga_transfer transfer = { 0 };
>      > +    struct iovec transfer_iovec;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    res = virtio_gpu_find_resource(g, resource_id);
>      > +    if (!res) {
>      > +        return;
>      > +    }
>      > +
>      > +    if (res->width != s->current_cursor->width ||
>      > +        res->height != s->current_cursor->height) {
>      > +        return;
>      > +    }
>      > +
>      > +    transfer.x = 0;
>      > +    transfer.y = 0;
>      > +    transfer.z = 0;
>      > +    transfer.w = res->width;
>      > +    transfer.h = res->height;
>      > +    transfer.d = 1;
>      > +
>      > +    transfer_iovec.iov_base = s->current_cursor->data;
>      > +    transfer_iovec.iov_len = res->width * res->height * 4;
>      > +
>      > +    rutabaga_resource_transfer_read(vr->rutabaga, 0,
>      > +                                    resource_id, &transfer,
>      > +                                    &transfer_iovec);
>      > +}
>      > +
>      > +static void
>      > +virtio_gpu_rutabaga_gl_flushed(VirtIOGPUBase *b)
>      > +{
>      > +    VirtIOGPU *g = VIRTIO_GPU(b);
>      > +    virtio_gpu_process_cmdq(g);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_create_resource_2d(VirtIOGPU *g,
>      > +                                struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct rutabaga_create_3d rc_3d = { 0 };
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_resource_create_2d c2d;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(c2d);
>      > +    trace_virtio_gpu_cmd_res_create_2d(c2d.resource_id, c2d.format,
>      > +                                       c2d.width, c2d.height);
>      > +
>      > +    rc_3d.target = 2;
>      > +    rc_3d.format = c2d.format;
>      > +    rc_3d.bind = (1 << 1);
>      > +    rc_3d.width = c2d.width;
>      > +    rc_3d.height = c2d.height;
>      > +    rc_3d.depth = 1;
>      > +    rc_3d.array_size = 1;
>      > +    rc_3d.last_level = 0;
>      > +    rc_3d.nr_samples = 0;
>      > +    rc_3d.flags = VIRTIO_GPU_RESOURCE_FLAG_Y_0_TOP;
>      > +
>      > +    result = rutabaga_resource_create_3d(vr->rutabaga,
>     c2d.resource_id, &rc_3d);
>      > +    CHECK(!result, cmd);
>      > +
>      > +    res = g_new0(struct virtio_gpu_simple_resource, 1);
>      > +    res->width = c2d.width;
>      > +    res->height = c2d.height;
>      > +    res->format = c2d.format;
>      > +    res->resource_id = c2d.resource_id;
>      > +
>      > +    QTAILQ_INSERT_HEAD(&g->reslist, res, next);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_create_resource_3d(VirtIOGPU *g,
>      > +                                struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct rutabaga_create_3d rc_3d = { 0 };
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_resource_create_3d c3d;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(c3d);
>      > +
>      > +    trace_virtio_gpu_cmd_res_create_3d(c3d.resource_id, c3d.format,
>      > +                                       c3d.width, c3d.height,
>     c3d.depth);
>      > +
>      > +    rc_3d.target = c3d.target;
>      > +    rc_3d.format = c3d.format;
>      > +    rc_3d.bind = c3d.bind;
>      > +    rc_3d.width = c3d.width;
>      > +    rc_3d.height = c3d.height;
>      > +    rc_3d.depth = c3d.depth;
>      > +    rc_3d.array_size = c3d.array_size;
>      > +    rc_3d.last_level = c3d.last_level;
>      > +    rc_3d.nr_samples = c3d.nr_samples;
>      > +    rc_3d.flags = c3d.flags;
>      > +
>      > +    result = rutabaga_resource_create_3d(vr->rutabaga,
>     c3d.resource_id, &rc_3d);
>      > +    CHECK(!result, cmd);
>      > +
>      > +    res = g_new0(struct virtio_gpu_simple_resource, 1);
>      > +    res->width = c3d.width;
>      > +    res->height = c3d.height;
>      > +    res->format = c3d.format;
>      > +    res->resource_id = c3d.resource_id;
>      > +
>      > +    QTAILQ_INSERT_HEAD(&g->reslist, res, next);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_resource_unref(VirtIOGPU *g,
>      > +                            struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_resource_unref unref;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(unref);
>      > +
>      > +    trace_virtio_gpu_cmd_res_unref(unref.resource_id);
>      > +
>      > +    res = virtio_gpu_find_resource(g, unref.resource_id);
>      > +    CHECK(res, cmd);
>      > +
>      > +    result = rutabaga_resource_unref(vr->rutabaga,
>     unref.resource_id);
>      > +    CHECK(!result, cmd);
>      > +
>      > +    if (res->image) {
>      > +        pixman_image_unref(res->image);
>      > +    }
>      > +
>      > +    QTAILQ_REMOVE(&g->reslist, res, next);
>      > +    g_free(res);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_context_create(VirtIOGPU *g,
>      > +                            struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_ctx_create cc;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(cc);
>      > +    trace_virtio_gpu_cmd_ctx_create(cc.hdr.ctx_id,
>      > +                                    cc.debug_name);
>      > +
>      > +    result = rutabaga_context_create(vr->rutabaga, cc.hdr.ctx_id,
>      > +                                     cc.context_init,
>     cc.debug_name, cc.nlen);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_context_destroy(VirtIOGPU *g,
>      > +                             struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_ctx_destroy cd;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(cd);
>      > +    trace_virtio_gpu_cmd_ctx_destroy(cd.hdr.ctx_id);
>      > +
>      > +    result = rutabaga_context_destroy(vr->rutabaga, cd.hdr.ctx_id);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_resource_flush(VirtIOGPU *g, struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result, i;
>      > +    struct virtio_gpu_scanout *scanout = NULL;
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct rutabaga_transfer transfer = { 0 };
>      > +    struct iovec transfer_iovec;
>      > +    struct virtio_gpu_resource_flush rf;
>      > +    bool found = false;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +    if (vr->headless) {
>      > +        return;
>      > +    }
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(rf);
>      > +    trace_virtio_gpu_cmd_res_flush(rf.resource_id,
>      > +                                   rf.r.width, rf.r.height,
>     rf.r.x, rf.r.y);
>      > +
>      > +    res = virtio_gpu_find_resource(g, rf.resource_id);
>      > +    CHECK(res, cmd);
>      > +
>      > +    for (i = 0; i < g->parent_obj.conf.max_outputs; i++) {
>      > +        scanout = &g->parent_obj.scanout[i];
>      > +        if (i == res->scanout_bitmask) {
>      > +            found = true;
>      > +            break;
>      > +        }
>      > +    }
>      > +
>      > +    if (!found) {
>      > +        return;
>      > +    }
>      > +
>      > +    transfer.x = 0;
>      > +    transfer.y = 0;
>      > +    transfer.z = 0;
>      > +    transfer.w = res->width;
>      > +    transfer.h = res->height;
>      > +    transfer.d = 1;
>      > +
>      > +    transfer_iovec.iov_base = pixman_image_get_data(res->image);
>      > +    transfer_iovec.iov_len = res->width * res->height * 4;
>      > +
>      > +    result = rutabaga_resource_transfer_read(vr->rutabaga, 0,
>      > +                                             rf.resource_id,
>     &transfer,
>      > +                                             &transfer_iovec);
>      > +    CHECK(!result, cmd);
>      > +    dpy_gfx_update_full(scanout->con);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_set_scanout(VirtIOGPU *g, struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_scanout *scanout = NULL;
>      > +    struct virtio_gpu_set_scanout ss;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +    if (vr->headless) {
>      > +        return;
>      > +    }
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(ss);
>      > +    trace_virtio_gpu_cmd_set_scanout(ss.scanout_id, ss.resource_id,
>      > +                                     ss.r.width, ss.r.height,
>     ss.r.x, ss.r.y);
>      > +
>      > +    CHECK(ss.scanout_id < VIRTIO_GPU_MAX_SCANOUTS, cmd);
>      > +    scanout = &g->parent_obj.scanout[ss.scanout_id];
>      > +
>      > +    if (ss.resource_id == 0) {
> 
>     You should disable scanout as virtio-gpu-virgl does.
> 
>      > +        return;
>      > +    }
>      > +
>      > +    res = virtio_gpu_find_resource(g, ss.resource_id);
>      > +    CHECK(res, cmd);
>      > +
>      > +    if (!res->image) {
>      > +        pixman_format_code_t pformat;
>      > +        pformat = virtio_gpu_get_pixman_format(res->format);
>      > +        CHECK(pformat, cmd);
>      > +
>      > +        res->image = pixman_image_create_bits(pformat,
>      > +                                              res->width,
>      > +                                              res->height,
>      > +                                              NULL, 0);
>      > +        CHECK(res->image, cmd);
>      > +        pixman_image_ref(res->image);
>      > +    }
>      > +
>      > +    g->parent_obj.enable = 1;
>      > +
>      > +    /* realloc the surface ptr */
>      > +    scanout->ds = qemu_create_displaysurface_pixman(res->image);
>      > +    dpy_gfx_replace_surface(scanout->con, NULL);
>      > +    dpy_gfx_replace_surface(scanout->con, scanout->ds);
>      > +    res->scanout_bitmask = ss.scanout_id;
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_submit_3d(VirtIOGPU *g,
>      > +                       struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_cmd_submit cs;
>      > +    struct rutabaga_command rutabaga_cmd = { 0 };
>      > +    g_autofree uint8_t *buf = NULL;
>      > +    size_t s;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(cs);
>      > +    trace_virtio_gpu_cmd_ctx_submit(cs.hdr.ctx_id, cs.size);
>      > +
>      > +    buf = g_new0(uint8_t, cs.size);
>      > +    s = iov_to_buf(cmd->elem.out_sg, cmd->elem.out_num,
>      > +                   sizeof(cs), buf, cs.size);
>      > +    CHECK(s == cs.size, cmd);
>      > +
>      > +    rutabaga_cmd.ctx_id = cs.hdr.ctx_id;
>      > +    rutabaga_cmd.cmd = buf;
>      > +    rutabaga_cmd.cmd_size = cs.size;
>      > +
>      > +    result = rutabaga_submit_command(vr->rutabaga, &rutabaga_cmd);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_transfer_to_host_2d(VirtIOGPU *g,
>      > +                                 struct virtio_gpu_ctrl_command
>     *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct rutabaga_transfer transfer = { 0 };
>      > +    struct virtio_gpu_transfer_to_host_2d t2d;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(t2d);
>      > +    trace_virtio_gpu_cmd_res_xfer_toh_2d(t2d.resource_id);
>      > +
>      > +    transfer.x = t2d.r.x;
>      > +    transfer.y = t2d.r.y;
>      > +    transfer.z = 0;
>      > +    transfer.w = t2d.r.width;
>      > +    transfer.h = t2d.r.height;
>      > +    transfer.d = 1;
>      > +
>      > +    result = rutabaga_resource_transfer_write(vr->rutabaga, 0,
>     t2d.resource_id,
>      > +                                              &transfer);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_transfer_to_host_3d(VirtIOGPU *g,
>      > +                                 struct virtio_gpu_ctrl_command
>     *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct rutabaga_transfer transfer = { 0 };
>      > +    struct virtio_gpu_transfer_host_3d t3d;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(t3d);
>      > +    trace_virtio_gpu_cmd_res_xfer_toh_3d(t3d.resource_id);
>      > +
>      > +    transfer.x = t3d.box.x;
>      > +    transfer.y = t3d.box.y;
>      > +    transfer.z = t3d.box.z;
>      > +    transfer.w = t3d.box.w;
>      > +    transfer.h = t3d.box.h;
>      > +    transfer.d = t3d.box.d;
>      > +    transfer.level = t3d.level;
>      > +    transfer.stride = t3d.stride;
>      > +    transfer.layer_stride = t3d.layer_stride;
>      > +    transfer.offset = t3d.offset;
>      > +
>      > +    result = rutabaga_resource_transfer_write(vr->rutabaga,
>     t3d.hdr.ctx_id,
>      > +                                              t3d.resource_id,
>     &transfer);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_transfer_from_host_3d(VirtIOGPU *g,
>      > +                                   struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct rutabaga_transfer transfer = { 0 };
>      > +    struct virtio_gpu_transfer_host_3d t3d;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(t3d);
>      > +    trace_virtio_gpu_cmd_res_xfer_fromh_3d(t3d.resource_id);
>      > +
>      > +    transfer.x = t3d.box.x;
>      > +    transfer.y = t3d.box.y;
>      > +    transfer.z = t3d.box.z;
>      > +    transfer.w = t3d.box.w;
>      > +    transfer.h = t3d.box.h;
>      > +    transfer.d = t3d.box.d;
>      > +    transfer.level = t3d.level;
>      > +    transfer.stride = t3d.stride;
>      > +    transfer.layer_stride = t3d.layer_stride;
>      > +    transfer.offset = t3d.offset;
>      > +
>      > +    result = rutabaga_resource_transfer_read(vr->rutabaga,
>     t3d.hdr.ctx_id,
>      > +                                             t3d.resource_id,
>     &transfer, NULL);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_attach_backing(VirtIOGPU *g, struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    struct rutabaga_iovecs vecs = { 0 };
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_resource_attach_backing att_rb;
>      > +    int ret;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(att_rb);
>      > +    trace_virtio_gpu_cmd_res_back_attach(att_rb.resource_id);
>      > +
>      > +    res = virtio_gpu_find_resource(g, att_rb.resource_id);
>      > +    CHECK(res, cmd);
>      > +    CHECK(!res->iov, cmd);
>      > +
>      > +    ret = virtio_gpu_create_mapping_iov(g, att_rb.nr_entries,
>     sizeof(att_rb),
>      > +                                        cmd, NULL, &res->iov,
>     &res->iov_cnt);
>      > +    CHECK(!ret, cmd);
>      > +
>      > +    vecs.iovecs = res->iov;
>      > +    vecs.num_iovecs = res->iov_cnt;
>      > +
>      > +    ret = rutabaga_resource_attach_backing(vr->rutabaga,
>     att_rb.resource_id,
>      > +                                           &vecs);
>      > +    if (ret != 0) {
>      > +        virtio_gpu_cleanup_mapping(g, res);
>      > +    }
>      > +
>      > +    CHECK(!ret, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_detach_backing(VirtIOGPU *g, struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_resource_detach_backing detach_rb;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(detach_rb);
>      > +    trace_virtio_gpu_cmd_res_back_detach(detach_rb.resource_id);
>      > +
>      > +    res = virtio_gpu_find_resource(g, detach_rb.resource_id);
>      > +    CHECK(res, cmd);
>      > +
>      > +    rutabaga_resource_detach_backing(vr->rutabaga,
>      > +                                     detach_rb.resource_id);
>      > +
>      > +    virtio_gpu_cleanup_mapping(g, res);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_ctx_attach_resource(VirtIOGPU *g,
>      > +                                 struct virtio_gpu_ctrl_command
>     *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_ctx_resource att_res;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(att_res);
>      > +    trace_virtio_gpu_cmd_ctx_res_attach(att_res.hdr.ctx_id,
>      > +                                        att_res.resource_id);
>      > +
>      > +    result = rutabaga_context_attach_resource(vr->rutabaga,
>     att_res.hdr.ctx_id,
>      > +                                              att_res.resource_id);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_ctx_detach_resource(VirtIOGPU *g,
>      > +                                 struct virtio_gpu_ctrl_command
>     *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_ctx_resource det_res;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(det_res);
>      > +    trace_virtio_gpu_cmd_ctx_res_detach(det_res.hdr.ctx_id,
>      > +                                        det_res.resource_id);
>      > +
>      > +    result = rutabaga_context_detach_resource(vr->rutabaga,
>     det_res.hdr.ctx_id,
>      > +                                              det_res.resource_id);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_get_capset_info(VirtIOGPU *g, struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_get_capset_info info;
>      > +    struct virtio_gpu_resp_capset_info resp;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(info);
>      > +
>      > +    result = rutabaga_get_capset_info(vr->rutabaga,
>     info.capset_index,
>      > +                                      &resp.capset_id,
>     &resp.capset_max_version,
>      > +                                      &resp.capset_max_size);
>      > +    CHECK(!result, cmd);
>      > +
>      > +    resp.hdr.type = VIRTIO_GPU_RESP_OK_CAPSET_INFO;
>      > +    virtio_gpu_ctrl_response(g, cmd, &resp.hdr, sizeof(resp));
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_get_capset(VirtIOGPU *g, struct
>     virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    struct virtio_gpu_get_capset gc;
>      > +    struct virtio_gpu_resp_capset *resp;
>      > +    uint32_t capset_size;
>      > +    uint32_t current_id;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(gc);
>      > +    for (uint32_t i = 0; i < vr->num_capsets; i++) {
>      > +        CHECK(i < vr->num_capsets, cmd);
> 
>     This CHECK will not work since you have i < vr->num_capsets as the loop
>     condition.
> 
>      > +        result = rutabaga_get_capset_info(vr->rutabaga, i,
>      > +                                          &current_id, &capset_size,
>      > +                                          &capset_size);
>      > +        CHECK(!result, cmd);
>      > +
>      > +        if (current_id == gc.capset_id) {
>      > +            break;
>      > +        }
>      > +    }
>      > +
>      > +    resp = g_malloc0(sizeof(*resp) + capset_size);
>      > +    resp->hdr.type = VIRTIO_GPU_RESP_OK_CAPSET;
>      > +    rutabaga_get_capset(vr->rutabaga, gc.capset_id,
>     gc.capset_version,
>      > +                        resp->capset_data, capset_size);
>      > +
>      > +    virtio_gpu_ctrl_response(g, cmd, &resp->hdr, sizeof(*resp) +
>     capset_size);
>      > +    g_free(resp);
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_resource_create_blob(VirtIOGPU *g,
>      > +                                  struct virtio_gpu_ctrl_command
>     *cmd)
>      > +{
>      > +    int result;
>      > +    struct rutabaga_iovecs vecs = { 0 };
>      > +    g_autofree struct virtio_gpu_simple_resource *res = NULL;
>      > +    struct virtio_gpu_resource_create_blob cblob;
>      > +    struct rutabaga_create_blob rc_blob = { 0 };
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(cblob);
>      > +    trace_virtio_gpu_cmd_res_create_blob(cblob.resource_id,
>     cblob.size);
>      > +
>      > +    CHECK(cblob.resource_id != 0, cmd);
>      > +
>      > +    res = g_new0(struct virtio_gpu_simple_resource, 1);
>      > +
>      > +    res->resource_id = cblob.resource_id;
>      > +    res->blob_size = cblob.size;
>      > +
>      > +    if (cblob.blob_mem != VIRTIO_GPU_BLOB_MEM_HOST3D) {
>      > +        result = virtio_gpu_create_mapping_iov(g, cblob.nr_entries,
>      > +                                               sizeof(cblob),
>     cmd, &res->addrs,
>      > +                                               &res->iov,
>     &res->iov_cnt);
>      > +        CHECK(!result, cmd);
>      > +    }
>      > +
>      > +    rc_blob.blob_id = cblob.blob_id;
>      > +    rc_blob.blob_mem = cblob.blob_mem;
>      > +    rc_blob.blob_flags = cblob.blob_flags;
>      > +    rc_blob.size = cblob.size;
>      > +
>      > +    vecs.iovecs = res->iov;
>      > +    vecs.num_iovecs = res->iov_cnt;
>      > +
>      > +    result = rutabaga_resource_create_blob(vr->rutabaga,
>     cblob.hdr.ctx_id,
>      > +                                           cblob.resource_id,
>     &rc_blob, &vecs,
>      > +                                           NULL);
>      > +
>      > +    if (result && cblob.blob_mem != VIRTIO_GPU_BLOB_MEM_HOST3D) {
>      > +        virtio_gpu_cleanup_mapping(g, res);
>      > +    }
>      > +
>      > +    CHECK(!result, cmd);
>      > +
>      > +    QTAILQ_INSERT_HEAD(&g->reslist, res, next);
>      > +    res = NULL;
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_resource_map_blob(VirtIOGPU *g,
>      > +                               struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    int32_t result;
>      > +    uint32_t map_info = 0;
>      > +    uint32_t slot = 0;
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct rutabaga_mapping mapping = { 0 };
>      > +    struct virtio_gpu_resource_map_blob mblob;
>      > +    struct virtio_gpu_resp_map_info resp = { 0 };
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(mblob);
>      > +
>      > +    CHECK(mblob.resource_id != 0, cmd);
>      > +
>      > +    res = virtio_gpu_find_resource(g, mblob.resource_id);
>      > +    CHECK(res, cmd);
>      > +
>      > +    result = rutabaga_resource_map_info(vr->rutabaga,
>     mblob.resource_id,
>      > +                                        &map_info);
>      > +    CHECK(!result, cmd);
>      > +
>      > +    /*
>      > +     * RUTABAGA_MAP_ACCESS_* flags are not part of the
>     virtio-gpu spec, but do
>      > +     * exist to potentially allow the hypervisor to restrict
>     write access to
>      > +     * memory. QEMU does not need to use this functionality at
>     the moment.
>      > +     */
>      > +    resp.map_info = map_info & RUTABAGA_MAP_CACHE_MASK;
>      > +
>      > +    result = rutabaga_resource_map(vr->rutabaga,
>     mblob.resource_id, &mapping);
>      > +    CHECK(!result, cmd);
>      > +
>      > +    for (slot = 0; slot < MAX_SLOTS; slot++) {
>      > +        if (vr->memory_regions[slot].used) {
>      > +            continue;
>      > +        }
>      > +
>      > +        MemoryRegion *mr = &(vr->memory_regions[slot].mr);
>      > +        memory_region_init_ram_ptr(mr, NULL, "blob", mapping.size,
>      > +                                   mapping.ptr);
>      > +        memory_region_add_subregion(&g->parent_obj.hostmem,
>      > +                                    mblob.offset, mr);
>      > +        vr->memory_regions[slot].resource_id = mblob.resource_id;
>      > +        vr->memory_regions[slot].used = 1;
>      > +        break;
>      > +    }
>      > +
>      > +    if (slot >= MAX_SLOTS) {
>      > +        result = rutabaga_resource_unmap(vr->rutabaga,
>     mblob.resource_id);
>      > +        CHECK(!result, cmd);
>      > +    }
>      > +
>      > +    CHECK(slot < MAX_SLOTS, cmd);
>      > +
>      > +    resp.hdr.type = VIRTIO_GPU_RESP_OK_MAP_INFO;
>      > +    virtio_gpu_ctrl_response(g, cmd, &resp.hdr, sizeof(resp));
>      > +}
>      > +
>      > +static void
>      > +rutabaga_cmd_resource_unmap_blob(VirtIOGPU *g,
>      > +                                 struct virtio_gpu_ctrl_command
>     *cmd)
>      > +{
>      > +    int32_t result;
>      > +    uint32_t slot = 0;
>      > +    struct virtio_gpu_simple_resource *res;
>      > +    struct virtio_gpu_resource_unmap_blob ublob;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(ublob);
>      > +
>      > +    CHECK(ublob.resource_id != 0, cmd);
>      > +
>      > +    res = virtio_gpu_find_resource(g, ublob.resource_id);
>      > +    CHECK(res, cmd);
>      > +
>      > +    for (slot = 0; slot < MAX_SLOTS; slot++) {
>      > +        if (vr->memory_regions[slot].resource_id !=
>     ublob.resource_id) {
>      > +            continue;
>      > +        }
>      > +
>      > +        MemoryRegion *mr = &(vr->memory_regions[slot].mr);
>      > +        memory_region_del_subregion(&g->parent_obj.hostmem, mr);
>      > +
>      > +        vr->memory_regions[slot].resource_id = 0;
>      > +        vr->memory_regions[slot].used = 0;
>      > +        break;
>      > +    }
>      > +
>      > +    CHECK(slot < MAX_SLOTS, cmd);
>      > +    result = rutabaga_resource_unmap(vr->rutabaga,
>     res->resource_id);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +virtio_gpu_rutabaga_process_cmd(VirtIOGPU *g,
>      > +                                struct virtio_gpu_ctrl_command *cmd)
>      > +{
>      > +    struct rutabaga_fence fence = { 0 };
>      > +    int32_t result;
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    VIRTIO_GPU_FILL_CMD(cmd->cmd_hdr);
>      > +
>      > +    switch (cmd->cmd_hdr.type) {
>      > +    case VIRTIO_GPU_CMD_CTX_CREATE:
>      > +        rutabaga_cmd_context_create(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_CTX_DESTROY:
>      > +        rutabaga_cmd_context_destroy(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_CREATE_2D:
>      > +        rutabaga_cmd_create_resource_2d(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_CREATE_3D:
>      > +        rutabaga_cmd_create_resource_3d(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_SUBMIT_3D:
>      > +        rutabaga_cmd_submit_3d(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_TRANSFER_TO_HOST_2D:
>      > +        rutabaga_cmd_transfer_to_host_2d(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_TRANSFER_TO_HOST_3D:
>      > +        rutabaga_cmd_transfer_to_host_3d(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_TRANSFER_FROM_HOST_3D:
>      > +        rutabaga_cmd_transfer_from_host_3d(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_ATTACH_BACKING:
>      > +        rutabaga_cmd_attach_backing(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_DETACH_BACKING:
>      > +        rutabaga_cmd_detach_backing(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_SET_SCANOUT:
>      > +        rutabaga_cmd_set_scanout(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_FLUSH:
>      > +        rutabaga_cmd_resource_flush(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_UNREF:
>      > +        rutabaga_cmd_resource_unref(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_CTX_ATTACH_RESOURCE:
>      > +        rutabaga_cmd_ctx_attach_resource(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_CTX_DETACH_RESOURCE:
>      > +        rutabaga_cmd_ctx_detach_resource(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_GET_CAPSET_INFO:
>      > +        rutabaga_cmd_get_capset_info(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_GET_CAPSET:
>      > +        rutabaga_cmd_get_capset(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_GET_DISPLAY_INFO:
>      > +        virtio_gpu_get_display_info(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_GET_EDID:
>      > +        virtio_gpu_get_edid(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_CREATE_BLOB:
>      > +        rutabaga_cmd_resource_create_blob(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_MAP_BLOB:
>      > +        rutabaga_cmd_resource_map_blob(g, cmd);
>      > +        break;
>      > +    case VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB:
>      > +        rutabaga_cmd_resource_unmap_blob(g, cmd);
>      > +        break;
>      > +    default:
>      > +        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
>      > +        break;
>      > +    }
>      > +
>      > +    if (cmd->finished) {
>      > +        return;
>      > +    }
>      > +    if (cmd->error) {
>      > +        error_report("%s: ctrl 0x%x, error 0x%x", __func__,
>      > +                     cmd->cmd_hdr.type, cmd->error);
>      > +        virtio_gpu_ctrl_response_nodata(g, cmd, cmd->error);
>      > +        return;
>      > +    }
>      > +    if (!(cmd->cmd_hdr.flags & VIRTIO_GPU_FLAG_FENCE)) {
>      > +        virtio_gpu_ctrl_response_nodata(g, cmd,
>     VIRTIO_GPU_RESP_OK_NODATA);
>      > +        return;
>      > +    }
>      > +
>      > +    fence.flags = cmd->cmd_hdr.flags;
>      > +    fence.ctx_id = cmd->cmd_hdr.ctx_id;
>      > +    fence.fence_id = cmd->cmd_hdr.fence_id;
>      > +    fence.ring_idx = cmd->cmd_hdr.ring_idx;
>      > +
>      > +    trace_virtio_gpu_fence_ctrl(cmd->cmd_hdr.fence_id,
>     cmd->cmd_hdr.type);
>      > +
>      > +    result = rutabaga_create_fence(vr->rutabaga, &fence);
>      > +    CHECK(!result, cmd);
>      > +}
>      > +
>      > +static void
>      > +virtio_gpu_rutabaga_aio_cb(void *opaque)
>      > +{
>      > +    struct rutabaga_aio_data *data = opaque;
>      > +    VirtIOGPU *g = VIRTIO_GPU(data->vr);
>      > +    struct rutabaga_fence fence_data = data->fence;
>      > +    struct virtio_gpu_ctrl_command *cmd, *tmp;
>      > +
>      > +    uint32_t signaled_ctx_specific = fence_data.flags &
>      > +                                     RUTABAGA_FLAG_INFO_RING_IDX;
>      > +
>      > +    QTAILQ_FOREACH_SAFE(cmd, &g->fenceq, next, tmp) {
>      > +        /*
>      > +         * Due to context specific timelines.
>      > +         */
>      > +        uint32_t target_ctx_specific = cmd->cmd_hdr.flags &
>      > +                                       RUTABAGA_FLAG_INFO_RING_IDX;
>      > +
>      > +        if (signaled_ctx_specific != target_ctx_specific) {
>      > +            continue;
>      > +        }
>      > +
>      > +        if (signaled_ctx_specific &&
>      > +           (cmd->cmd_hdr.ring_idx != fence_data.ring_idx)) {
>      > +            continue;
>      > +        }
>      > +
>      > +        if (cmd->cmd_hdr.fence_id > fence_data.fence_id) {
>      > +            continue;
>      > +        }
>      > +
>      > +        trace_virtio_gpu_fence_resp(cmd->cmd_hdr.fence_id);
>      > +        virtio_gpu_ctrl_response_nodata(g, cmd,
>     VIRTIO_GPU_RESP_OK_NODATA);
>      > +        QTAILQ_REMOVE(&g->fenceq, cmd, next);
>      > +        g_free(cmd);
>      > +    }
>      > +
>      > +    g_free(data);
>      > +}
>      > +
>      > +static void
>      > +virtio_gpu_rutabaga_fence_cb(uint64_t user_data,
>      > +                             const struct rutabaga_fence *fence) {
>      > +    struct rutabaga_aio_data *data;
>      > +    VirtIOGPU *g = (VirtIOGPU *)user_data;
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +
>      > +    /*
>      > +     * gfxstream and both cross-domain (and even newer versions
>     virglrenderer:
>      > +     * see VIRGL_RENDERER_ASYNC_FENCE_CB) like to signal fence
>     completion on
>      > +     * threads ("callback threads") that are different from the
>     thread that
>      > +     * processes the command queue ("main thread").
>      > +     *
>      > +     * crosvm and other virtio-gpu 1.1 implementations enable
>     callback threads
>      > +     * via locking.  However, on QEMU a deadlock is observed if
>      > +     * virtio_gpu_ctrl_response_nodata(..) [used in the fence
>     callback] is used
>      > +     * from a thread that is not the main thread.
>      > +     *
>      > +     * The reason is QEMU's internal locking is designed to work
>     with QEMU
>      > +     * threads (see rcu_register_thread()) and not generic
>     C/C++/Rust threads.
>      > +     * For now, we can workaround this by scheduling the return
>     of the
>      > +     * fence descriptors on the main thread.
>      > +     */
>      > +
>      > +    data = g_new0(struct rutabaga_aio_data, 1);
>      > +    data->vr = vr;
>      > +    data->fence = *fence;
>      > +    aio_bh_schedule_oneshot_full(qemu_get_aio_context(),
>      > +                                 virtio_gpu_rutabaga_aio_cb,
>      > +                                 data, "aio");
>      > +}
>      > +
>      > +static void
>      > +virtio_gpu_rutabaga_debug_cb(uint64_t user_data,
>      > +                             const struct rutabaga_debug *debug) {
>      > +
>      > +    if (debug->debug_type == RUTABAGA_DEBUG_ERROR) {
>      > +        error_report("%s", debug->message);
>      > +    } else if (debug->debug_type == RUTABAGA_DEBUG_WARN) {
>      > +        warn_report("%s", debug->message);
>      > +    } else if (debug->debug_type == RUTABAGA_DEBUG_INFO) {
>      > +        info_report("%s", debug->message);
>      > +    }
>      > +}
>      > +
>      > +static int virtio_gpu_rutabaga_init(VirtIOGPU *g, Error **errp)
>      > +{
>      > +    int result;
>      > +    uint64_t capset_mask;
>      > +    struct rutabaga_builder builder = { 0 };
>      > +    char wayland_socket_path[UNIX_PATH_MAX];
>      > +    struct rutabaga_channel channel = { 0 };
>      > +    struct rutabaga_channels channels = { 0 };
>      > +
>      > +    VirtIOGPURutabaga *vr = VIRTIO_GPU_RUTABAGA(g);
>      > +    vr->rutabaga = NULL;
>      > +
>      > +    if (!vr->capset_names) {
>      > +        error_setg(errp, "a capset name from the virtio-gpu spec
>     is required");
>      > +        return -EINVAL;
>      > +    }
>      > +
>      > +    builder.wsi = RUTABAGA_WSI_SURFACELESS;
>      > +    /*
>      > +     * Currently, if WSI is specified, the only valid strings
>     are "surfaceless"
>      > +     * or "headless".  Surfaceless doesn't create a native
>     window surface, but
>      > +     * does copy from the render target to the Pixman buffer if
>     a virtio-gpu
>      > +     * 2D hypercall is issued.  Surfacless is the default.
>      > +     *
>      > +     * Headless is like surfaceless, but doesn't copy to the
>     Pixman buffer. The
>      > +     * use case is automated testing environments where there is
>     no need to view
>      > +     * results.
>      > +     *
>      > +     * In the future, more performant virtio-gpu 2D UI
>     integration may be added.
>      > +     */
>      > +    if (vr->wsi) {
>      > +        if (g_str_equal(vr->wsi, "surfaceless")) {
>      > +            vr->headless = false;
>      > +        } else if (g_str_equal(vr->wsi, "headless")) {
>      > +            vr->headless = true;
>      > +        } else {
>      > +            error_setg(errp, "invalid wsi option selected");
>      > +            return -EINVAL;
>      > +        }
>      > +    }
>      > +
>      > +    result = rutabaga_calculate_capset_mask(vr->capset_names,
>     &capset_mask);
>      > +    if (result) {
>      > +        error_setg(errp, "invalid capset names: %s",
>     vr->capset_names);
>      > +        return result;
>      > +    }
>      > +
>      > +    builder.fence_cb = virtio_gpu_rutabaga_fence_cb;
>      > +    builder.debug_cb = virtio_gpu_rutabaga_debug_cb;
>      > +    builder.capset_mask = capset_mask;
>      > +    builder.user_data = (uint64_t)g;
>      > +
>      > +    /*
>      > +     * If the user doesn't specify the wayland socket path, we
>     try to infer
>      > +     * the socket via a process similar to the one used by
>     libwayland.
>      > +     * libwayland does the following:
>      > +     *
>      > +     * 1) If $WAYLAND_DISPLAY is set, attempt to connect to
>      > +     *    $XDG_RUNTIME_DIR/$WAYLAND_DISPLAY
>      > +     * 2) Otherwise, attempt to connect to
>     $XDG_RUNTIME_DIR/wayland-0
>      > +     * 3) Otherwise, don't pass a wayland socket to rutabaga. If
>     a guest
>      > +     *    wayland proxy is launched, it will fail to work.
>      > +     */
>      > +    channel.channel_type = RUTABAGA_CHANNEL_TYPE_WAYLAND;
>      > +    if (!vr->wayland_socket_path) {
>      > +        const char *runtime_dir = getenv("XDG_RUNTIME_DIR");
>      > +        const char *display = getenv("WAYLAND_DISPLAY");
>      > +        if (!display) {
>      > +            display = "wayland-0";
>      > +        }
>      > +
>      > +        if (runtime_dir) {
>      > +            result = snprintf(wayland_socket_path, UNIX_PATH_MAX,
>      > +                              "%s/%s", runtime_dir, display);
>      > +            if (result > 0 && result < UNIX_PATH_MAX) {
>      > +                channel.channel_name = wayland_socket_path;
>      > +            }
> 
>     As I said in the last email, I think it's better to make a hard
>     error if
>     the user requested RUTABAGA_CAPSET_CROSS_DOMAIN but the default socket
>     path does not fit in wayland_socket_path.
> 
> 
> I mentioned this in the cover-letter to v5, but messed up the commit 
> message of patch 6.  But we do want to be possible to use cross_domain 
> without wayland for following reason:
> 
> "It is possible to use cross-domain without enabling Wayland, for guest
> swapchain allocation [a] which often needs special handling.  We might
> We want to use this capability on MacOS/Windows in the future without
> Wayland, hence we shouldn't have hard errors if the default wayland socket
> isn't found.
> 
> [a] 
> https://android.googlesource.com/platform/external/minigbm/+/refs/heads/main/virtgpu_cross_domain.c <https://android.googlesource.com/platform/external/minigbm/+/refs/heads/main/virtgpu_cross_domain.c>"
> 
> I'll incorporate your other suggestions in v6 though.

What happens on X.org system then? Such a system may have 
XDG_RUNTIME_DIR but should not have WAYLAND_DISPLAY, resulting in 
$XDG_RUNTIME_DIR/wayland-0 passed to Rutabaga.