[PATCH v11 00/10] Support blob memory and venus on qemu

[PATCH v11 00/10] Support blob memory and venus on qemu

[Dmitry Osipenko]

Hello,

This series enables Vulkan Venus context support on virtio-gpu.

All virglrender and almost all Linux kernel prerequisite changes
needed by Venus are already in upstream. For kernel there is a pending
KVM patchset that fixes mapping of compound pages needed for DRM drivers
using TTM [1], othewrwise hostmem blob mapping will fail with a KVM error
from Qemu.

[1] https://lore.kernel.org/kvm/20240229025759.1187910-1-stevensd@google.com/

You'll need to use recent Mesa version containing patch that removes
dependency on cross-device feature from Venus that isn't supported by
Qemu [2].

[2] https://gitlab.freedesktop.org/mesa/mesa/-/commit/087e9a96d13155e26987befae78b6ccbb7ae242b

Example Qemu cmdline that enables Venus:

  qemu-system-x86_64 -device virtio-vga-gl,hostmem=4G,blob=true,venus=true \
      -machine q35,accel=kvm,memory-backend=mem1 \
      -object memory-backend-memfd,id=mem1,size=8G -m 8G


Changes from V10 to V11

- Replaced cmd_resume bool in struct ctrl_command with
  "cmd->finished + !VIRTIO_GPU_FLAG_FENCE" checking as was requested
  by Akihiko Odaki.

- Reworked virgl_cmd_resource_unmap/unref_blob() to avoid re-adding
  the 'async_unmap_in_progress' flag that was dropped in v9:

    1. virgl_cmd_resource_[un]map_blob() now doesn't check itself whether
       resource was previously mapped and lets virglrenderer to do the
       checking.

    2. error returned by virgl_renderer_resource_unmap() is now handled
       and reported properly, previously the error wasn't checked. The
       virgl_renderer_resource_unmap() fails if resource wasn't mapped.

    3. virgl_cmd_resource_unref_blob() now doesn't allow to unref resource
       that is mapped, it's a error condition if guest didn't unmap resource
       before doing the unref. Previously unref was implicitly unmapping
       resource.

Changes from V9 to V10

- Dropped 'async_unmap_in_progress' variable and switched to use
  aio_bh_new() isntead of oneshot variant in the "blob commands" patch.

- Further improved error messages by printing error code when actual error
  occurrs and using ERR_UNSPEC instead of ERR_ENOMEM when we don't really
  know if it was ENOMEM for sure.

- Added vdc->unrealize for the virtio GL device and freed virgl data.

- Dropped UUID and doc/migration patches. UUID feature isn't needed
  anymore, instead we changed Mesa Venus driver to not require UUID.

- Renamed virtio-gpu-gl "vulkan" property name back to "venus".

Changes from V8 to V9

- Added resuming of cmdq processing when hostmem MR is freed,
  as was suggested by Akihiko Odaki.

- Added more error messages, suggested by Akihiko Odaki

- Dropped superfluous 'res->async_unmap_completed', suggested
  by Akihiko Odaki.

- Kept using cmd->suspended flag. Akihiko Odaki suggested to make
  virtio_gpu_virgl_process_cmd() return false if cmd processing is
  suspended, but it's not easy to implement due to ubiquitous
  VIRTIO_GPU_FILL_CMD() macros that returns void, requiring to change
  all the virtio-gpu processing code.

- Added back virtio_gpu_virgl_resource as was requested by Akihiko Odaki,
  though I'm not convinced it's really needed.

- Switched to use GArray, renamed capset2_max_ver/size vars and moved
  "vulkan" property definition to the virtio-gpu-gl device in the Venus
  patch, like was suggested by Akihiko Odaki.

- Moved UUID to virtio_gpu_virgl_resource and dropped UUID save/restore
  since it will require bumping VM version and virgl device isn't miratable
  anyways.

- Fixed exposing UUID feature with Rutabaga

- Dropped linux-headers update patch because headers were already updated
  in Qemu/staging.

- Added patch that updates virtio migration doc with a note about virtio-gpu
  migration specifics, suggested by Akihiko Odaki.

- Addressed coding style issue noticed by Akihiko Odaki

Changes from V7 to V8

- Supported suspension of virtio-gpu commands processing and made
  unmapping of hostmem region asynchronous by blocking/suspending
  cmd processing until region is unmapped. Suggested by Akihiko Odaki.

- Fixed arm64 building of x86 targets using updated linux-headers.
  Corrected the update script. Thanks to Rob Clark for reporting
  the issue.

- Added new patch that makes registration of virgl capsets dynamic.
  Requested by Antonio Caggiano and Pierre-Eric Pelloux-Prayer.

- Venus capset now isn't advertised if Vulkan is disabled with vulkan=false

Changes from V6 to V7

- Used scripts/update-linux-headers.sh to update Qemu headers based
  on Linux v6.8-rc3 that adds Venus capset definition to virtio-gpu
  protocol, was requested by Peter Maydel

- Added r-bs that were given to v6 patches. Corrected missing s-o-bs

- Dropped context_init Qemu's virtio-gpu device configuration flag,
  was suggested by Marc-André Lureau

- Added missing error condition checks spotted by Marc-André Lureau
  and Akihiko Odaki, and few more

- Returned back res->mr referencing to memory_region_init_ram_ptr() like
  was suggested by Akihiko Odaki. Incorporated fix suggested by Pierre-Eric
  to specify the MR name

- Dropped the virgl_gpu_resource wrapper, cleaned up and simplified
  patch that adds blob-cmd support

- Fixed improper blob resource removal from resource list on resource_unref
  that was spotted by Akihiko Odaki

- Change order of the blob patches, was suggested by Akihiko Odaki.
  The cmd_set_scanout_blob support is enabled first

- Factored out patch that adds resource management support to virtio-gpu-gl,
  was requested by Marc-André Lureau

- Simplified and improved the UUID support patch, dropped the hash table
  as we don't need it for now. Moved QemuUUID to virtio_gpu_simple_resource.
  This all was suggested by Akihiko Odaki and Marc-André Lureau

- Dropped console_has_gl() check, suggested by Akihiko Odaki

- Reworked Meson cheking of libvirglrender features, made new features
  available based on virglrender pkgconfig version instead of checking
  symbols in header. This should fix build error using older virglrender
  version, reported by Alex Bennée

- Made enabling of Venus context configrable via new virtio-gpu device
  "vulkan=true" flag, suggested by Marc-André Lureau. The flag is disabled
  by default because it requires blob and hostmem options to be enabled
  and configured

Changes from V5 to V6

- Move macros configurations under virgl.found() and rename
  HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS.

- Handle the case while context_init is disabled.

- Enable context_init by default.

- Move virtio_gpu_virgl_resource_unmap() into
  virgl_cmd_resource_unmap_blob().

- Introduce new struct virgl_gpu_resource to store virgl specific members.

- Remove erro handling of g_new0, because glib will abort() on OOM.

- Set resource uuid as option.

- Implement optional subsection of vmstate_virtio_gpu_resource_uuid_state
  for virtio live migration.

- Use g_int_hash/g_int_equal instead of the default

- Add scanout_blob function for virtio-gpu-virgl

- Resolve the memory leak on virtio-gpu-virgl

- Remove the unstable API flags check because virglrenderer is already 1.0

- Squash the render server flag support into "Initialize Venus"

Changes from V4 (virtio gpu V4) to V5

- Inverted patch 5 and 6 because we should configure
  HAVE_VIRGL_CONTEXT_INIT firstly.

- Validate owner of memory region to avoid slowing down DMA.

- Use memory_region_init_ram_ptr() instead of
  memory_region_init_ram_device_ptr().

- Adjust sequence to allocate gpu resource before virglrender resource
  creation

- Add virtio migration handling for uuid.

- Send kernel patch to define VIRTIO_GPU_CAPSET_VENUS.
  https://lore.kernel.org/lkml/20230915105918.3763061-1-ray.huang@amd.com/

- Add meson check to make sure unstable APIs defined from 0.9.0.

Changes from V1 to V2 (virtio gpu V4)

- Remove unused #include "hw/virtio/virtio-iommu.h"

- Add a local function, called virgl_resource_destroy(), that is used
  to release a vgpu resource on error paths and in resource_unref.

- Remove virtio_gpu_virgl_resource_unmap from
  virtio_gpu_cleanup_mapping(),
  since this function won't be called on blob resources and also because
  blob resources are unmapped via virgl_cmd_resource_unmap_blob().

- In virgl_cmd_resource_create_blob(), do proper cleanup in error paths
  and move QTAILQ_INSERT_HEAD(&g->reslist, res, next) after the resource
  has been fully initialized.

- Memory region has a different life-cycle from virtio gpu resources
  i.e. cannot be released synchronously along with the vgpu resource.
  So, here the field "region" was changed to a pointer and is allocated
  dynamically when the blob is mapped.
  Also, since the pointer can be used to indicate whether the blob
  is mapped, the explicite field "mapped" was removed.

- In virgl_cmd_resource_map_blob(), add check on the value of
  res->region, to prevent beeing called twice on the same resource.

- Add a patch to enable automatic deallocation of memory regions to resolve
  use-after-free memory corruption with a reference.

Antonio Caggiano (2):
  virtio-gpu: Handle resource blob commands
  virtio-gpu: Support Venus context

Dmitry Osipenko (4):
  virtio-gpu: Unrealize GL device
  virtio-gpu: Use pkgconfig version to decide which virgl features are
    available
  virtio-gpu: Don't require udmabuf when blobs and virgl are enabled
  virtio-gpu: Support suspension of commands processing

Huang Rui (2):
  virtio-gpu: Support context-init feature with virglrenderer
  virtio-gpu: Add virgl resource management

Pierre-Eric Pelloux-Prayer (1):
  virtio-gpu: Register capsets dynamically

Robert Beckett (1):
  virtio-gpu: Support blob scanout using dmabuf fd

 hw/display/virtio-gpu-gl.c     |  23 +-
 hw/display/virtio-gpu-virgl.c  | 541 +++++++++++++++++++++++++++++++--
 hw/display/virtio-gpu.c        |  35 ++-
 include/hw/virtio/virtio-gpu.h |  17 +-
 meson.build                    |  10 +-
 5 files changed, 592 insertions(+), 34 deletions(-)

-- 
2.44.0

[PATCH v11 01/10] virtio-gpu: Unrealize GL device

[Dmitry Osipenko]

Even though GL GPU doesn't support hotplugging today, free virgl
resources when GL device is unrealized. For consistency.

Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-gl.c     | 11 +++++++++++
 hw/display/virtio-gpu-virgl.c  |  9 +++++++++
 include/hw/virtio/virtio-gpu.h |  1 +
 3 files changed, 21 insertions(+)

diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
index e06be60dfbfc..0c0a8d136954 100644
--- a/hw/display/virtio-gpu-gl.c
+++ b/hw/display/virtio-gpu-gl.c
@@ -136,6 +136,16 @@ static Property virtio_gpu_gl_properties[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
+{
+    VirtIOGPU *g = VIRTIO_GPU(qdev);
+    VirtIOGPUGL *gl = VIRTIO_GPU_GL(qdev);
+
+    if (gl->renderer_inited) {
+        virtio_gpu_virgl_deinit(g);
+    }
+}
+
 static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
@@ -149,6 +159,7 @@ static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
     vgc->update_cursor_data = virtio_gpu_gl_update_cursor_data;
 
     vdc->realize = virtio_gpu_gl_device_realize;
+    vdc->unrealize = virtio_gpu_gl_device_unrealize;
     vdc->reset = virtio_gpu_gl_reset;
     device_class_set_props(dc, virtio_gpu_gl_properties);
 }
diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index 9f34d0e6619c..b0500eccf8e0 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -665,3 +665,12 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
 
     return capset2_max_ver ? 2 : 1;
 }
+
+void virtio_gpu_virgl_deinit(VirtIOGPU *g)
+{
+    if (g->fence_poll) {
+        timer_free(g->fence_poll);
+    }
+
+    virgl_renderer_cleanup(NULL);
+}
diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
index ed44cdad6b34..b657187159d9 100644
--- a/include/hw/virtio/virtio-gpu.h
+++ b/include/hw/virtio/virtio-gpu.h
@@ -336,6 +336,7 @@ void virtio_gpu_virgl_fence_poll(VirtIOGPU *g);
 void virtio_gpu_virgl_reset_scanout(VirtIOGPU *g);
 void virtio_gpu_virgl_reset(VirtIOGPU *g);
 int virtio_gpu_virgl_init(VirtIOGPU *g);
+void virtio_gpu_virgl_deinit(VirtIOGPU *g);
 int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g);
 
 #endif
-- 
2.44.0

[PATCH v11 02/10] virtio-gpu: Use pkgconfig version to decide which virgl features are available

[Dmitry Osipenko]

New virglrerenderer features were stabilized with release of v1.0.0.
Presence of symbols in virglrenderer.h doesn't guarantee ABI compatibility
with pre-release development versions of libvirglerender. Use virglrenderer
version to decide reliably which virgl features are available.

Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 meson.build | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/meson.build b/meson.build
index 83ae4347c7f9..ca2798dbac37 100644
--- a/meson.build
+++ b/meson.build
@@ -2286,11 +2286,8 @@ config_host_data.set('CONFIG_PNG', png.found())
 config_host_data.set('CONFIG_VNC', vnc.found())
 config_host_data.set('CONFIG_VNC_JPEG', jpeg.found())
 config_host_data.set('CONFIG_VNC_SASL', sasl.found())
-if virgl.found()
-  config_host_data.set('HAVE_VIRGL_D3D_INFO_EXT',
-                       cc.has_member('struct virgl_renderer_resource_info_ext', 'd3d_tex2d',
-                                     prefix: '#include <virglrenderer.h>',
-                                     dependencies: virgl))
+if virgl.version().version_compare('>=1.0.0')
+  config_host_data.set('HAVE_VIRGL_D3D_INFO_EXT', 1)
 endif
 config_host_data.set('CONFIG_VIRTFS', have_virtfs)
 config_host_data.set('CONFIG_VTE', vte.found())
-- 
2.44.0

[PATCH v11 03/10] virtio-gpu: Support context-init feature with virglrenderer

[Dmitry Osipenko]

From: Huang Rui <ray.huang@amd.com>

Patch "virtio-gpu: CONTEXT_INIT feature" has added the context_init
feature flags. Expose this feature and support creating virglrenderer
context with flags using context_id if libvirglrenderer is new enough.

Originally-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Antonio Caggiano <quic_acaggian@quicinc.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-gl.c    |  4 ++++
 hw/display/virtio-gpu-virgl.c | 20 ++++++++++++++++++--
 meson.build                   |  1 +
 3 files changed, 23 insertions(+), 2 deletions(-)

diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
index 0c0a8d136954..95806999189e 100644
--- a/hw/display/virtio-gpu-gl.c
+++ b/hw/display/virtio-gpu-gl.c
@@ -127,6 +127,10 @@ static void virtio_gpu_gl_device_realize(DeviceState *qdev, Error **errp)
     VIRTIO_GPU_BASE(g)->virtio_config.num_capsets =
         virtio_gpu_virgl_get_num_capsets(g);
 
+#ifdef HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS
+    g->parent_obj.conf.flags |= 1 << VIRTIO_GPU_FLAG_CONTEXT_INIT_ENABLED;
+#endif
+
     virtio_gpu_device_realize(qdev, errp);
 }
 
diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index b0500eccf8e0..8306961ad502 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -106,8 +106,24 @@ static void virgl_cmd_context_create(VirtIOGPU *g,
     trace_virtio_gpu_cmd_ctx_create(cc.hdr.ctx_id,
                                     cc.debug_name);
 
-    virgl_renderer_context_create(cc.hdr.ctx_id, cc.nlen,
-                                  cc.debug_name);
+    if (cc.context_init) {
+        if (!virtio_gpu_context_init_enabled(g->parent_obj.conf)) {
+            qemu_log_mask(LOG_GUEST_ERROR, "%s: context_init disabled",
+                          __func__);
+            cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
+            return;
+        }
+
+#ifdef HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS
+        virgl_renderer_context_create_with_flags(cc.hdr.ctx_id,
+                                                 cc.context_init,
+                                                 cc.nlen,
+                                                 cc.debug_name);
+        return;
+#endif
+    }
+
+    virgl_renderer_context_create(cc.hdr.ctx_id, cc.nlen, cc.debug_name);
 }
 
 static void virgl_cmd_context_destroy(VirtIOGPU *g,
diff --git a/meson.build b/meson.build
index ca2798dbac37..5448eb17f39b 100644
--- a/meson.build
+++ b/meson.build
@@ -2288,6 +2288,7 @@ config_host_data.set('CONFIG_VNC_JPEG', jpeg.found())
 config_host_data.set('CONFIG_VNC_SASL', sasl.found())
 if virgl.version().version_compare('>=1.0.0')
   config_host_data.set('HAVE_VIRGL_D3D_INFO_EXT', 1)
+  config_host_data.set('HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS', 1)
 endif
 config_host_data.set('CONFIG_VIRTFS', have_virtfs)
 config_host_data.set('CONFIG_VTE', vte.found())
-- 
2.44.0

[PATCH v11 04/10] virtio-gpu: Don't require udmabuf when blobs and virgl are enabled

[Dmitry Osipenko]

The udmabuf usage is mandatory when virgl is disabled and blobs feature
enabled in the Qemu machine configuration. If virgl and blobs are enabled,
then udmabuf requirement is optional. Since udmabuf isn't widely supported
by a popular Linux distros today, let's relax the udmabuf requirement for
blobs=on,virgl=on. Now, a full-featured virtio-gpu acceleration is
available to Qemu users without a need to have udmabuf available in the
system.

Reviewed-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Antonio Caggiano <quic_acaggian@quicinc.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index ae831b6b3e3e..dac272ecadb1 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -1472,6 +1472,7 @@ void virtio_gpu_device_realize(DeviceState *qdev, Error **errp)
 
     if (virtio_gpu_blob_enabled(g->parent_obj.conf)) {
         if (!virtio_gpu_rutabaga_enabled(g->parent_obj.conf) &&
+            !virtio_gpu_virgl_enabled(g->parent_obj.conf) &&
             !virtio_gpu_have_udmabuf()) {
             error_setg(errp, "need rutabaga or udmabuf for blob resources");
             return;
-- 
2.44.0

[PATCH v11 05/10] virtio-gpu: Add virgl resource management

[Dmitry Osipenko]

From: Huang Rui <ray.huang@amd.com>

In a preparation to adding host blobs support to virtio-gpu, add virgl
resource management that allows to retrieve resource based on its ID
and virgl resource wrapper on top of simple resource that will be contain
fields specific to virgl.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Antonio Caggiano <quic_acaggian@quicinc.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-virgl.c | 76 +++++++++++++++++++++++++++++++++++
 1 file changed, 76 insertions(+)

diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index 8306961ad502..c6c63ca1c373 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -22,6 +22,23 @@
 
 #include <virglrenderer.h>
 
+struct virtio_gpu_virgl_resource {
+    struct virtio_gpu_simple_resource base;
+};
+
+static struct virtio_gpu_virgl_resource *
+virtio_gpu_virgl_find_resource(VirtIOGPU *g, uint32_t resource_id)
+{
+    struct virtio_gpu_simple_resource *res;
+
+    res = virtio_gpu_find_resource(g, resource_id);
+    if (!res) {
+        return NULL;
+    }
+
+    return container_of(res, struct virtio_gpu_virgl_resource, base);
+}
+
 #if VIRGL_RENDERER_CALLBACKS_VERSION >= 4
 static void *
 virgl_get_egl_display(G_GNUC_UNUSED void *cookie)
@@ -35,11 +52,34 @@ static void virgl_cmd_create_resource_2d(VirtIOGPU *g,
 {
     struct virtio_gpu_resource_create_2d c2d;
     struct virgl_renderer_resource_create_args args;
+    struct virtio_gpu_virgl_resource *res;
 
     VIRTIO_GPU_FILL_CMD(c2d);
     trace_virtio_gpu_cmd_res_create_2d(c2d.resource_id, c2d.format,
                                        c2d.width, c2d.height);
 
+    if (c2d.resource_id == 0) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource id 0 is not allowed\n",
+                      __func__);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    res = virtio_gpu_virgl_find_resource(g, c2d.resource_id);
+    if (res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource already exists %d\n",
+                      __func__, c2d.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    res = g_new0(struct virtio_gpu_virgl_resource, 1);
+    res->base.width = c2d.width;
+    res->base.height = c2d.height;
+    res->base.format = c2d.format;
+    res->base.resource_id = c2d.resource_id;
+    QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
+
     args.handle = c2d.resource_id;
     args.target = 2;
     args.format = c2d.format;
@@ -59,11 +99,34 @@ static void virgl_cmd_create_resource_3d(VirtIOGPU *g,
 {
     struct virtio_gpu_resource_create_3d c3d;
     struct virgl_renderer_resource_create_args args;
+    struct virtio_gpu_virgl_resource *res;
 
     VIRTIO_GPU_FILL_CMD(c3d);
     trace_virtio_gpu_cmd_res_create_3d(c3d.resource_id, c3d.format,
                                        c3d.width, c3d.height, c3d.depth);
 
+    if (c3d.resource_id == 0) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource id 0 is not allowed\n",
+                      __func__);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    res = virtio_gpu_virgl_find_resource(g, c3d.resource_id);
+    if (res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource already exists %d\n",
+                      __func__, c3d.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    res = g_new0(struct virtio_gpu_virgl_resource, 1);
+    res->base.width = c3d.width;
+    res->base.height = c3d.height;
+    res->base.format = c3d.format;
+    res->base.resource_id = c3d.resource_id;
+    QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
+
     args.handle = c3d.resource_id;
     args.target = c3d.target;
     args.format = c3d.format;
@@ -82,12 +145,21 @@ static void virgl_cmd_resource_unref(VirtIOGPU *g,
                                      struct virtio_gpu_ctrl_command *cmd)
 {
     struct virtio_gpu_resource_unref unref;
+    struct virtio_gpu_virgl_resource *res;
     struct iovec *res_iovs = NULL;
     int num_iovs = 0;
 
     VIRTIO_GPU_FILL_CMD(unref);
     trace_virtio_gpu_cmd_res_unref(unref.resource_id);
 
+    res = virtio_gpu_virgl_find_resource(g, unref.resource_id);
+    if (!res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not exist %d\n",
+                      __func__, unref.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
     virgl_renderer_resource_detach_iov(unref.resource_id,
                                        &res_iovs,
                                        &num_iovs);
@@ -95,6 +167,10 @@ static void virgl_cmd_resource_unref(VirtIOGPU *g,
         virtio_gpu_cleanup_mapping_iov(g, res_iovs, num_iovs);
     }
     virgl_renderer_resource_unref(unref.resource_id);
+
+    QTAILQ_REMOVE(&g->reslist, &res->base, next);
+
+    g_free(res);
 }
 
 static void virgl_cmd_context_create(VirtIOGPU *g,
-- 
2.44.0

[PATCH v11 06/10] virtio-gpu: Support blob scanout using dmabuf fd

[Dmitry Osipenko]

From: Robert Beckett <bob.beckett@collabora.com>

Support displaying blob resources by handling SET_SCANOUT_BLOB
command.

Signed-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Signed-off-by: Robert Beckett <bob.beckett@collabora.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Antonio Caggiano <quic_acaggian@quicinc.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-virgl.c  | 109 +++++++++++++++++++++++++++++++++
 hw/display/virtio-gpu.c        |  12 ++--
 include/hw/virtio/virtio-gpu.h |   7 +++
 meson.build                    |   1 +
 4 files changed, 123 insertions(+), 6 deletions(-)

diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index c6c63ca1c373..524f15220b7f 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -17,6 +17,8 @@
 #include "trace.h"
 #include "hw/virtio/virtio.h"
 #include "hw/virtio/virtio-gpu.h"
+#include "hw/virtio/virtio-gpu-bswap.h"
+#include "hw/virtio/virtio-gpu-pixman.h"
 
 #include "ui/egl-helpers.h"
 
@@ -78,6 +80,7 @@ static void virgl_cmd_create_resource_2d(VirtIOGPU *g,
     res->base.height = c2d.height;
     res->base.format = c2d.format;
     res->base.resource_id = c2d.resource_id;
+    res->base.dmabuf_fd = -1;
     QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
 
     args.handle = c2d.resource_id;
@@ -125,6 +128,7 @@ static void virgl_cmd_create_resource_3d(VirtIOGPU *g,
     res->base.height = c3d.height;
     res->base.format = c3d.format;
     res->base.resource_id = c3d.resource_id;
+    res->base.dmabuf_fd = -1;
     QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
 
     args.handle = c3d.resource_id;
@@ -509,6 +513,106 @@ static void virgl_cmd_get_capset(VirtIOGPU *g,
     g_free(resp);
 }
 
+#ifdef HAVE_VIRGL_RESOURCE_BLOB
+static void virgl_cmd_set_scanout_blob(VirtIOGPU *g,
+                                       struct virtio_gpu_ctrl_command *cmd)
+{
+    struct virtio_gpu_framebuffer fb = { 0 };
+    struct virgl_renderer_resource_info info;
+    struct virtio_gpu_virgl_resource *res;
+    struct virtio_gpu_set_scanout_blob ss;
+    uint64_t fbend;
+
+    VIRTIO_GPU_FILL_CMD(ss);
+    virtio_gpu_scanout_blob_bswap(&ss);
+    trace_virtio_gpu_cmd_set_scanout_blob(ss.scanout_id, ss.resource_id,
+                                          ss.r.width, ss.r.height, ss.r.x,
+                                          ss.r.y);
+
+    if (ss.scanout_id >= g->parent_obj.conf.max_outputs) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: illegal scanout id specified %d",
+                      __func__, ss.scanout_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_SCANOUT_ID;
+        return;
+    }
+
+    if (ss.resource_id == 0) {
+        virtio_gpu_disable_scanout(g, ss.scanout_id);
+        return;
+    }
+
+    if (ss.width < 16 ||
+        ss.height < 16 ||
+        ss.r.x + ss.r.width > ss.width ||
+        ss.r.y + ss.r.height > ss.height) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: illegal scanout %d bounds for"
+                      " resource %d, rect (%d,%d)+%d,%d, fb %d %d\n",
+                      __func__, ss.scanout_id, ss.resource_id,
+                      ss.r.x, ss.r.y, ss.r.width, ss.r.height,
+                      ss.width, ss.height);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_PARAMETER;
+        return;
+    }
+
+    res = virtio_gpu_virgl_find_resource(g, ss.resource_id);
+    if (!res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not exist %d\n",
+                      __func__, ss.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+    if (virgl_renderer_resource_get_info(ss.resource_id, &info)) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not have info %d\n",
+                      __func__, ss.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+    if (res->base.dmabuf_fd < 0) {
+        res->base.dmabuf_fd = info.fd;
+    }
+    if (res->base.dmabuf_fd < 0) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource not backed by dmabuf %d\n",
+                      __func__, ss.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    fb.format = virtio_gpu_get_pixman_format(ss.format);
+    if (!fb.format) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: pixel format not supported %d\n",
+                      __func__, ss.format);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_PARAMETER;
+        return;
+    }
+
+    fb.bytes_pp = DIV_ROUND_UP(PIXMAN_FORMAT_BPP(fb.format), 8);
+    fb.width = ss.width;
+    fb.height = ss.height;
+    fb.stride = ss.strides[0];
+    fb.offset = ss.offsets[0] + ss.r.x * fb.bytes_pp + ss.r.y * fb.stride;
+
+    fbend = fb.offset;
+    fbend += fb.stride * (ss.r.height - 1);
+    fbend += fb.bytes_pp * ss.r.width;
+    if (fbend > res->base.blob_size) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: fb end out of range\n",
+                      __func__);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_PARAMETER;
+        return;
+    }
+
+    g->parent_obj.enable = 1;
+    if (virtio_gpu_update_dmabuf(g, ss.scanout_id, &res->base, &fb, &ss.r)) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: failed to update dmabuf\n",
+                      __func__);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_PARAMETER;
+        return;
+    }
+
+    virtio_gpu_update_scanout(g, ss.scanout_id, &res->base, &fb, &ss.r);
+}
+#endif /* HAVE_VIRGL_RESOURCE_BLOB */
+
 void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
                                       struct virtio_gpu_ctrl_command *cmd)
 {
@@ -575,6 +679,11 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
     case VIRTIO_GPU_CMD_GET_EDID:
         virtio_gpu_get_edid(g, cmd);
         break;
+#ifdef HAVE_VIRGL_RESOURCE_BLOB
+    case VIRTIO_GPU_CMD_SET_SCANOUT_BLOB:
+        virgl_cmd_set_scanout_blob(g, cmd);
+        break;
+#endif /* HAVE_VIRGL_RESOURCE_BLOB */
     default:
         cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
         break;
diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index dac272ecadb1..1e57a53d346c 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -380,7 +380,7 @@ static void virtio_gpu_resource_create_blob(VirtIOGPU *g,
     QTAILQ_INSERT_HEAD(&g->reslist, res, next);
 }
 
-static void virtio_gpu_disable_scanout(VirtIOGPU *g, int scanout_id)
+void virtio_gpu_disable_scanout(VirtIOGPU *g, int scanout_id)
 {
     struct virtio_gpu_scanout *scanout = &g->parent_obj.scanout[scanout_id];
     struct virtio_gpu_simple_resource *res;
@@ -597,11 +597,11 @@ static void virtio_unref_resource(pixman_image_t *image, void *data)
     pixman_image_unref(data);
 }
 
-static void virtio_gpu_update_scanout(VirtIOGPU *g,
-                                      uint32_t scanout_id,
-                                      struct virtio_gpu_simple_resource *res,
-                                      struct virtio_gpu_framebuffer *fb,
-                                      struct virtio_gpu_rect *r)
+void virtio_gpu_update_scanout(VirtIOGPU *g,
+                               uint32_t scanout_id,
+                               struct virtio_gpu_simple_resource *res,
+                               struct virtio_gpu_framebuffer *fb,
+                               struct virtio_gpu_rect *r)
 {
     struct virtio_gpu_simple_resource *ores;
     struct virtio_gpu_scanout *scanout;
diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
index b657187159d9..ba36497c477f 100644
--- a/include/hw/virtio/virtio-gpu.h
+++ b/include/hw/virtio/virtio-gpu.h
@@ -329,6 +329,13 @@ int virtio_gpu_update_dmabuf(VirtIOGPU *g,
                              struct virtio_gpu_framebuffer *fb,
                              struct virtio_gpu_rect *r);
 
+void virtio_gpu_update_scanout(VirtIOGPU *g,
+                               uint32_t scanout_id,
+                               struct virtio_gpu_simple_resource *res,
+                               struct virtio_gpu_framebuffer *fb,
+                               struct virtio_gpu_rect *r);
+void virtio_gpu_disable_scanout(VirtIOGPU *g, int scanout_id);
+
 /* virtio-gpu-3d.c */
 void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
                                   struct virtio_gpu_ctrl_command *cmd);
diff --git a/meson.build b/meson.build
index 5448eb17f39b..0433ea200dd7 100644
--- a/meson.build
+++ b/meson.build
@@ -2289,6 +2289,7 @@ config_host_data.set('CONFIG_VNC_SASL', sasl.found())
 if virgl.version().version_compare('>=1.0.0')
   config_host_data.set('HAVE_VIRGL_D3D_INFO_EXT', 1)
   config_host_data.set('HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS', 1)
+  config_host_data.set('HAVE_VIRGL_RESOURCE_BLOB', 1)
 endif
 config_host_data.set('CONFIG_VIRTFS', have_virtfs)
 config_host_data.set('CONFIG_VTE', vte.found())
-- 
2.44.0

[PATCH v11 07/10] virtio-gpu: Support suspension of commands processing

[Dmitry Osipenko]

Check whether command processing has been finished; otherwise, stop
processing commands and retry the command again next time. This allows
us to support asynchronous execution of non-fenced commands needed for
unmapping host blobs safely.

Suggested-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index 1e57a53d346c..6c8c7213bafa 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -1054,6 +1054,11 @@ void virtio_gpu_process_cmdq(VirtIOGPU *g)
         /* process command */
         vgc->process_cmd(g, cmd);
 
+        /* command suspended */
+        if (!cmd->finished && !(cmd->cmd_hdr.flags & VIRTIO_GPU_FLAG_FENCE)) {
+            break;
+        }
+
         QTAILQ_REMOVE(&g->cmdq, cmd, next);
         if (virtio_gpu_stats_enabled(g->parent_obj.conf)) {
             g->stats.requests++;
-- 
2.44.0

[PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Dmitry Osipenko]

From: Antonio Caggiano <antonio.caggiano@collabora.com>

Support BLOB resources creation, mapping and unmapping by calling the
new stable virglrenderer 0.10 interface. Only enabled when available and
via the blob config. E.g. -device virtio-vga-gl,blob=true

Signed-off-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Signed-off-by: Xenia Ragiadakou <xenia.ragiadakou@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-virgl.c  | 274 ++++++++++++++++++++++++++++++++-
 hw/display/virtio-gpu.c        |   4 +-
 include/hw/virtio/virtio-gpu.h |   2 +
 3 files changed, 277 insertions(+), 3 deletions(-)

diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index 524f15220b7f..3f2e406be3a4 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -26,6 +26,7 @@
 
 struct virtio_gpu_virgl_resource {
     struct virtio_gpu_simple_resource base;
+    MemoryRegion *mr;
 };
 
 static struct virtio_gpu_virgl_resource *
@@ -49,6 +50,117 @@ virgl_get_egl_display(G_GNUC_UNUSED void *cookie)
 }
 #endif
 
+#ifdef HAVE_VIRGL_RESOURCE_BLOB
+struct virtio_gpu_virgl_hostmem_region {
+    MemoryRegion mr;
+    struct VirtIOGPU *g;
+    struct virtio_gpu_virgl_resource *res;
+};
+
+static void virtio_gpu_virgl_resume_cmdq_bh(void *opaque)
+{
+    VirtIOGPU *g = opaque;
+
+    virtio_gpu_process_cmdq(g);
+}
+
+static void virtio_gpu_virgl_hostmem_region_free(void *obj)
+{
+    MemoryRegion *mr = MEMORY_REGION(obj);
+    struct virtio_gpu_virgl_hostmem_region *vmr;
+    VirtIOGPUBase *b;
+
+    vmr = container_of(mr, struct virtio_gpu_virgl_hostmem_region, mr);
+    vmr->res->mr = NULL;
+
+    b = VIRTIO_GPU_BASE(vmr->g);
+    b->renderer_blocked--;
+
+    /*
+     * memory_region_unref() is executed from RCU thread context, while
+     * virglrenderer works only on the main-loop thread that's holding GL
+     * context.
+     */
+    qemu_bh_schedule(vmr->g->cmdq_resume_bh);
+    g_free(vmr);
+}
+
+static int
+virtio_gpu_virgl_map_resource_blob(VirtIOGPU *g,
+                                   struct virtio_gpu_virgl_resource *res,
+                                   uint64_t offset)
+{
+    struct virtio_gpu_virgl_hostmem_region *vmr;
+    VirtIOGPUBase *b = VIRTIO_GPU_BASE(g);
+    MemoryRegion *mr;
+    uint64_t size;
+    void *data;
+    int ret;
+
+    if (!virtio_gpu_hostmem_enabled(b->conf)) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: hostmem disabled\n", __func__);
+        return -EOPNOTSUPP;
+    }
+
+    ret = virgl_renderer_resource_map(res->base.resource_id, &data, &size);
+    if (ret) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: failed to map virgl resource: %s\n",
+                      __func__, strerror(-ret));
+        return ret;
+    }
+
+    vmr = g_new0(struct virtio_gpu_virgl_hostmem_region, 1);
+    vmr->res = res;
+    vmr->g = g;
+
+    mr = &vmr->mr;
+    memory_region_init_ram_ptr(mr, OBJECT(mr), "blob", size, data);
+    memory_region_add_subregion(&b->hostmem, offset, mr);
+    memory_region_set_enabled(mr, true);
+
+    /*
+     * MR could outlive the resource if MR's reference is held outside of
+     * virtio-gpu. In order to prevent unmapping resource while MR is alive,
+     * and thus, making the data pointer invalid, we will block virtio-gpu
+     * command processing until MR is fully unreferenced and freed.
+     */
+    OBJECT(mr)->free = virtio_gpu_virgl_hostmem_region_free;
+
+    res->mr = mr;
+
+    return 0;
+}
+
+static int
+virtio_gpu_virgl_async_unmap_resource_blob(VirtIOGPU *g,
+                                           struct virtio_gpu_virgl_resource *res)
+{
+    VirtIOGPUBase *b = VIRTIO_GPU_BASE(g);
+    MemoryRegion *mr = res->mr;
+    int ret;
+
+    if (mr) {
+        /* render will be unblocked once MR is freed */
+        b->renderer_blocked++;
+
+        /* memory region owns self res->mr object and frees it by itself */
+        memory_region_set_enabled(mr, false);
+        memory_region_del_subregion(&b->hostmem, mr);
+        object_unparent(OBJECT(mr));
+    } else {
+        ret = virgl_renderer_resource_unmap(res->base.resource_id);
+        if (ret) {
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          "%s: failed to unmap virgl resource: %s\n",
+                          __func__, strerror(-ret));
+            return ret;
+        }
+    }
+
+    return 0;
+}
+#endif /* HAVE_VIRGL_RESOURCE_BLOB */
+
 static void virgl_cmd_create_resource_2d(VirtIOGPU *g,
                                          struct virtio_gpu_ctrl_command *cmd)
 {
@@ -146,7 +258,8 @@ static void virgl_cmd_create_resource_3d(VirtIOGPU *g,
 }
 
 static void virgl_cmd_resource_unref(VirtIOGPU *g,
-                                     struct virtio_gpu_ctrl_command *cmd)
+                                     struct virtio_gpu_ctrl_command *cmd,
+                                     bool *cmd_suspended)
 {
     struct virtio_gpu_resource_unref unref;
     struct virtio_gpu_virgl_resource *res;
@@ -164,6 +277,12 @@ static void virgl_cmd_resource_unref(VirtIOGPU *g,
         return;
     }
 
+    if (res->mr) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource is mapped\n", __func__);
+        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
+        return;
+    }
+
     virgl_renderer_resource_detach_iov(unref.resource_id,
                                        &res_iovs,
                                        &num_iovs);
@@ -514,6 +633,134 @@ static void virgl_cmd_get_capset(VirtIOGPU *g,
 }
 
 #ifdef HAVE_VIRGL_RESOURCE_BLOB
+static void virgl_cmd_resource_create_blob(VirtIOGPU *g,
+                                           struct virtio_gpu_ctrl_command *cmd)
+{
+    struct virgl_renderer_resource_create_blob_args virgl_args = { 0 };
+    struct virtio_gpu_resource_create_blob cblob;
+    struct virtio_gpu_virgl_resource *res;
+    int ret;
+
+    if (!virtio_gpu_blob_enabled(g->parent_obj.conf)) {
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_PARAMETER;
+        return;
+    }
+
+    VIRTIO_GPU_FILL_CMD(cblob);
+    virtio_gpu_create_blob_bswap(&cblob);
+    trace_virtio_gpu_cmd_res_create_blob(cblob.resource_id, cblob.size);
+
+    if (cblob.resource_id == 0) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource id 0 is not allowed\n",
+                      __func__);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    res = virtio_gpu_virgl_find_resource(g, cblob.resource_id);
+    if (res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource already exists %d\n",
+                      __func__, cblob.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    res = g_new0(struct virtio_gpu_virgl_resource, 1);
+    res->base.resource_id = cblob.resource_id;
+    res->base.blob_size = cblob.size;
+    res->base.dmabuf_fd = -1;
+
+    if (cblob.blob_mem != VIRTIO_GPU_BLOB_MEM_HOST3D) {
+        ret = virtio_gpu_create_mapping_iov(g, cblob.nr_entries, sizeof(cblob),
+                                            cmd, &res->base.addrs,
+                                            &res->base.iov, &res->base.iov_cnt);
+        if (!ret) {
+            g_free(res);
+            cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
+            return;
+        }
+    }
+
+    QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
+
+    virgl_args.res_handle = cblob.resource_id;
+    virgl_args.ctx_id = cblob.hdr.ctx_id;
+    virgl_args.blob_mem = cblob.blob_mem;
+    virgl_args.blob_id = cblob.blob_id;
+    virgl_args.blob_flags = cblob.blob_flags;
+    virgl_args.size = cblob.size;
+    virgl_args.iovecs = res->base.iov;
+    virgl_args.num_iovs = res->base.iov_cnt;
+
+    ret = virgl_renderer_resource_create_blob(&virgl_args);
+    if (ret) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: virgl blob create error: %s\n",
+                      __func__, strerror(-ret));
+        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
+    }
+}
+
+static void virgl_cmd_resource_map_blob(VirtIOGPU *g,
+                                        struct virtio_gpu_ctrl_command *cmd)
+{
+    struct virtio_gpu_resource_map_blob mblob;
+    struct virtio_gpu_virgl_resource *res;
+    struct virtio_gpu_resp_map_info resp;
+    int ret;
+
+    VIRTIO_GPU_FILL_CMD(mblob);
+    virtio_gpu_map_blob_bswap(&mblob);
+
+    res = virtio_gpu_virgl_find_resource(g, mblob.resource_id);
+    if (!res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not exist %d\n",
+                      __func__, mblob.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    ret = virtio_gpu_virgl_map_resource_blob(g, res, mblob.offset);
+    if (ret) {
+        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
+        return;
+    }
+
+    memset(&resp, 0, sizeof(resp));
+    resp.hdr.type = VIRTIO_GPU_RESP_OK_MAP_INFO;
+    virgl_renderer_resource_get_map_info(mblob.resource_id, &resp.map_info);
+    virtio_gpu_ctrl_response(g, cmd, &resp.hdr, sizeof(resp));
+}
+
+static void virgl_cmd_resource_unmap_blob(VirtIOGPU *g,
+                                          struct virtio_gpu_ctrl_command *cmd,
+                                          bool *cmd_suspended)
+{
+    struct virtio_gpu_resource_unmap_blob ublob;
+    struct virtio_gpu_virgl_resource *res;
+    int ret;
+
+    VIRTIO_GPU_FILL_CMD(ublob);
+    virtio_gpu_unmap_blob_bswap(&ublob);
+
+    res = virtio_gpu_virgl_find_resource(g, ublob.resource_id);
+    if (!res) {
+        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not exist %d\n",
+                      __func__, ublob.resource_id);
+        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
+        return;
+    }
+
+    ret = virtio_gpu_virgl_async_unmap_resource_blob(g, res);
+    if (ret) {
+        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
+        return;
+    }
+
+    if (res->mr) {
+        *cmd_suspended = true;
+    }
+}
+
 static void virgl_cmd_set_scanout_blob(VirtIOGPU *g,
                                        struct virtio_gpu_ctrl_command *cmd)
 {
@@ -616,6 +863,8 @@ static void virgl_cmd_set_scanout_blob(VirtIOGPU *g,
 void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
                                       struct virtio_gpu_ctrl_command *cmd)
 {
+    bool cmd_suspended = false;
+
     VIRTIO_GPU_FILL_CMD(cmd->cmd_hdr);
 
     virgl_renderer_force_ctx_0();
@@ -657,7 +906,7 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
         virgl_cmd_resource_flush(g, cmd);
         break;
     case VIRTIO_GPU_CMD_RESOURCE_UNREF:
-        virgl_cmd_resource_unref(g, cmd);
+        virgl_cmd_resource_unref(g, cmd, &cmd_suspended);
         break;
     case VIRTIO_GPU_CMD_CTX_ATTACH_RESOURCE:
         /* TODO add security */
@@ -680,6 +929,15 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
         virtio_gpu_get_edid(g, cmd);
         break;
 #ifdef HAVE_VIRGL_RESOURCE_BLOB
+    case VIRTIO_GPU_CMD_RESOURCE_CREATE_BLOB:
+        virgl_cmd_resource_create_blob(g, cmd);
+        break;
+    case VIRTIO_GPU_CMD_RESOURCE_MAP_BLOB:
+        virgl_cmd_resource_map_blob(g, cmd);
+        break;
+    case VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB:
+        virgl_cmd_resource_unmap_blob(g, cmd, &cmd_suspended);
+        break;
     case VIRTIO_GPU_CMD_SET_SCANOUT_BLOB:
         virgl_cmd_set_scanout_blob(g, cmd);
         break;
@@ -689,6 +947,9 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
         break;
     }
 
+    if (cmd_suspended) {
+        return;
+    }
     if (cmd->finished) {
         return;
     }
@@ -854,6 +1115,11 @@ int virtio_gpu_virgl_init(VirtIOGPU *g)
                                       virtio_gpu_print_stats, g);
         timer_mod(g->print_stats, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + 1000);
     }
+
+    g->cmdq_resume_bh = aio_bh_new(qemu_get_aio_context(),
+                                   virtio_gpu_virgl_resume_cmdq_bh,
+                                   g);
+
     return 0;
 }
 
@@ -869,6 +1135,10 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
 
 void virtio_gpu_virgl_deinit(VirtIOGPU *g)
 {
+    if (g->cmdq_resume_bh) {
+        qemu_bh_delete(g->cmdq_resume_bh);
+    }
+
     if (g->fence_poll) {
         timer_free(g->fence_poll);
     }
diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index 6c8c7213bafa..052ab493a00b 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -1483,10 +1483,12 @@ void virtio_gpu_device_realize(DeviceState *qdev, Error **errp)
             return;
         }
 
+#ifndef HAVE_VIRGL_RESOURCE_BLOB
         if (virtio_gpu_virgl_enabled(g->parent_obj.conf)) {
-            error_setg(errp, "blobs and virgl are not compatible (yet)");
+            error_setg(errp, "old virglrenderer, blob resources unsupported");
             return;
         }
+#endif
     }
 
     if (!virtio_gpu_base_device_realize(qdev,
diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
index ba36497c477f..c0b0b0eac08b 100644
--- a/include/hw/virtio/virtio-gpu.h
+++ b/include/hw/virtio/virtio-gpu.h
@@ -210,6 +210,8 @@ struct VirtIOGPU {
         QTAILQ_HEAD(, VGPUDMABuf) bufs;
         VGPUDMABuf *primary[VIRTIO_GPU_MAX_SCANOUTS];
     } dmabuf;
+
+    QEMUBH *cmdq_resume_bh;
 };
 
 struct VirtIOGPUClass {
-- 
2.44.0

[PATCH v11 09/10] virtio-gpu: Register capsets dynamically

[Dmitry Osipenko]

From: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

virtio_gpu_virgl_get_num_capsets will return "num_capsets", but we can't
assume that capset_index 1 is always VIRGL2 once we'll support more capsets,
like Venus and DRM capsets. Register capsets dynamically to avoid that problem.

Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-gl.c     |  6 ++++--
 hw/display/virtio-gpu-virgl.c  | 33 +++++++++++++++++++++------------
 include/hw/virtio/virtio-gpu.h |  4 +++-
 3 files changed, 28 insertions(+), 15 deletions(-)

diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
index 95806999189e..431fc2881a00 100644
--- a/hw/display/virtio-gpu-gl.c
+++ b/hw/display/virtio-gpu-gl.c
@@ -124,8 +124,8 @@ static void virtio_gpu_gl_device_realize(DeviceState *qdev, Error **errp)
     }
 
     g->parent_obj.conf.flags |= (1 << VIRTIO_GPU_FLAG_VIRGL_ENABLED);
-    VIRTIO_GPU_BASE(g)->virtio_config.num_capsets =
-        virtio_gpu_virgl_get_num_capsets(g);
+    g->capset_ids = virtio_gpu_virgl_get_capsets(g);
+    VIRTIO_GPU_BASE(g)->virtio_config.num_capsets = g->capset_ids->len;
 
 #ifdef HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS
     g->parent_obj.conf.flags |= 1 << VIRTIO_GPU_FLAG_CONTEXT_INIT_ENABLED;
@@ -148,6 +148,8 @@ static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
     if (gl->renderer_inited) {
         virtio_gpu_virgl_deinit(g);
     }
+
+    g_array_unref(g->capset_ids);
 }
 
 static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index 3f2e406be3a4..135974431492 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -590,19 +590,13 @@ static void virgl_cmd_get_capset_info(VirtIOGPU *g,
     VIRTIO_GPU_FILL_CMD(info);
 
     memset(&resp, 0, sizeof(resp));
-    if (info.capset_index == 0) {
-        resp.capset_id = VIRTIO_GPU_CAPSET_VIRGL;
-        virgl_renderer_get_cap_set(resp.capset_id,
-                                   &resp.capset_max_version,
-                                   &resp.capset_max_size);
-    } else if (info.capset_index == 1) {
-        resp.capset_id = VIRTIO_GPU_CAPSET_VIRGL2;
+
+    if (info.capset_index < g->capset_ids->len) {
+        resp.capset_id = g_array_index(g->capset_ids, uint32_t,
+                                       info.capset_index);
         virgl_renderer_get_cap_set(resp.capset_id,
                                    &resp.capset_max_version,
                                    &resp.capset_max_size);
-    } else {
-        resp.capset_max_version = 0;
-        resp.capset_max_size = 0;
     }
     resp.hdr.type = VIRTIO_GPU_RESP_OK_CAPSET_INFO;
     virtio_gpu_ctrl_response(g, cmd, &resp.hdr, sizeof(resp));
@@ -1123,14 +1117,29 @@ int virtio_gpu_virgl_init(VirtIOGPU *g)
     return 0;
 }
 
-int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
+static void virtio_gpu_virgl_add_capset(GArray *capset_ids, uint32_t capset_id)
+{
+    g_array_append_val(capset_ids, capset_id);
+}
+
+GArray *virtio_gpu_virgl_get_capsets(VirtIOGPU *g)
 {
     uint32_t capset2_max_ver, capset2_max_size;
+    GArray *capset_ids;
+
+    capset_ids = g_array_new(false, false, sizeof(uint32_t));
+
+    /* VIRGL is always supported. */
+    virtio_gpu_virgl_add_capset(capset_ids, VIRTIO_GPU_CAPSET_VIRGL);
+
     virgl_renderer_get_cap_set(VIRTIO_GPU_CAPSET_VIRGL2,
                               &capset2_max_ver,
                               &capset2_max_size);
+    if (capset2_max_ver) {
+        virtio_gpu_virgl_add_capset(capset_ids, VIRTIO_GPU_CAPSET_VIRGL2);
+    }
 
-    return capset2_max_ver ? 2 : 1;
+    return capset_ids;
 }
 
 void virtio_gpu_virgl_deinit(VirtIOGPU *g)
diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
index c0b0b0eac08b..7bbc6ef268eb 100644
--- a/include/hw/virtio/virtio-gpu.h
+++ b/include/hw/virtio/virtio-gpu.h
@@ -212,6 +212,8 @@ struct VirtIOGPU {
     } dmabuf;
 
     QEMUBH *cmdq_resume_bh;
+
+    GArray *capset_ids;
 };
 
 struct VirtIOGPUClass {
@@ -346,6 +348,6 @@ void virtio_gpu_virgl_reset_scanout(VirtIOGPU *g);
 void virtio_gpu_virgl_reset(VirtIOGPU *g);
 int virtio_gpu_virgl_init(VirtIOGPU *g);
 void virtio_gpu_virgl_deinit(VirtIOGPU *g);
-int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g);
+GArray *virtio_gpu_virgl_get_capsets(VirtIOGPU *g);
 
 #endif
-- 
2.44.0

[PATCH v11 10/10] virtio-gpu: Support Venus context

[Dmitry Osipenko]

From: Antonio Caggiano <antonio.caggiano@collabora.com>

Request Venus when initializing VirGL and if venus=true flag is set for
virtio-gpu-gl device.

Signed-off-by: Antonio Caggiano <antonio.caggiano@collabora.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
---
 hw/display/virtio-gpu-gl.c     |  2 ++
 hw/display/virtio-gpu-virgl.c  | 22 ++++++++++++++++++----
 hw/display/virtio-gpu.c        | 13 +++++++++++++
 include/hw/virtio/virtio-gpu.h |  3 +++
 meson.build                    |  1 +
 5 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
index 431fc2881a00..1deb3e64b781 100644
--- a/hw/display/virtio-gpu-gl.c
+++ b/hw/display/virtio-gpu-gl.c
@@ -137,6 +137,8 @@ static void virtio_gpu_gl_device_realize(DeviceState *qdev, Error **errp)
 static Property virtio_gpu_gl_properties[] = {
     DEFINE_PROP_BIT("stats", VirtIOGPU, parent_obj.conf.flags,
                     VIRTIO_GPU_FLAG_STATS_ENABLED, false),
+    DEFINE_PROP_BIT("venus", VirtIOGPU, parent_obj.conf.flags,
+                    VIRTIO_GPU_FLAG_VENUS_ENABLED, false),
     DEFINE_PROP_END_OF_LIST(),
 };
 
diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
index 135974431492..f1cad005dd8f 100644
--- a/hw/display/virtio-gpu-virgl.c
+++ b/hw/display/virtio-gpu-virgl.c
@@ -1094,6 +1094,11 @@ int virtio_gpu_virgl_init(VirtIOGPU *g)
         flags |= VIRGL_RENDERER_D3D11_SHARE_TEXTURE;
     }
 #endif
+#ifdef VIRGL_RENDERER_VENUS
+    if (virtio_gpu_venus_enabled(g->parent_obj.conf)) {
+        flags |= VIRGL_RENDERER_VENUS | VIRGL_RENDERER_RENDER_SERVER;
+    }
+#endif
 
     ret = virgl_renderer_init(g, flags, &virtio_gpu_3d_cbs);
     if (ret != 0) {
@@ -1124,7 +1129,7 @@ static void virtio_gpu_virgl_add_capset(GArray *capset_ids, uint32_t capset_id)
 
 GArray *virtio_gpu_virgl_get_capsets(VirtIOGPU *g)
 {
-    uint32_t capset2_max_ver, capset2_max_size;
+    uint32_t capset_max_ver, capset_max_size;
     GArray *capset_ids;
 
     capset_ids = g_array_new(false, false, sizeof(uint32_t));
@@ -1133,12 +1138,21 @@ GArray *virtio_gpu_virgl_get_capsets(VirtIOGPU *g)
     virtio_gpu_virgl_add_capset(capset_ids, VIRTIO_GPU_CAPSET_VIRGL);
 
     virgl_renderer_get_cap_set(VIRTIO_GPU_CAPSET_VIRGL2,
-                              &capset2_max_ver,
-                              &capset2_max_size);
-    if (capset2_max_ver) {
+                               &capset_max_ver,
+                               &capset_max_size);
+    if (capset_max_ver) {
         virtio_gpu_virgl_add_capset(capset_ids, VIRTIO_GPU_CAPSET_VIRGL2);
     }
 
+    if (virtio_gpu_venus_enabled(g->parent_obj.conf)) {
+        virgl_renderer_get_cap_set(VIRTIO_GPU_CAPSET_VENUS,
+                                   &capset_max_ver,
+                                   &capset_max_size);
+        if (capset_max_size) {
+            virtio_gpu_virgl_add_capset(capset_ids, VIRTIO_GPU_CAPSET_VENUS);
+        }
+    }
+
     return capset_ids;
 }
 
diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index 052ab493a00b..0518bb858e88 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -1491,6 +1491,19 @@ void virtio_gpu_device_realize(DeviceState *qdev, Error **errp)
 #endif
     }
 
+    if (virtio_gpu_venus_enabled(g->parent_obj.conf)) {
+#ifdef HAVE_VIRGL_VENUS
+        if (!virtio_gpu_blob_enabled(g->parent_obj.conf) ||
+            !virtio_gpu_hostmem_enabled(g->parent_obj.conf)) {
+            error_setg(errp, "venus requires enabled blob and hostmem options");
+            return;
+        }
+#else
+        error_setg(errp, "old virglrenderer, venus unsupported");
+        return;
+#endif
+    }
+
     if (!virtio_gpu_base_device_realize(qdev,
                                         virtio_gpu_handle_ctrl_cb,
                                         virtio_gpu_handle_cursor_cb,
diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
index 7bbc6ef268eb..9940bbcbaacc 100644
--- a/include/hw/virtio/virtio-gpu.h
+++ b/include/hw/virtio/virtio-gpu.h
@@ -99,6 +99,7 @@ enum virtio_gpu_base_conf_flags {
     VIRTIO_GPU_FLAG_BLOB_ENABLED,
     VIRTIO_GPU_FLAG_CONTEXT_INIT_ENABLED,
     VIRTIO_GPU_FLAG_RUTABAGA_ENABLED,
+    VIRTIO_GPU_FLAG_VENUS_ENABLED,
 };
 
 #define virtio_gpu_virgl_enabled(_cfg) \
@@ -117,6 +118,8 @@ enum virtio_gpu_base_conf_flags {
     (_cfg.flags & (1 << VIRTIO_GPU_FLAG_RUTABAGA_ENABLED))
 #define virtio_gpu_hostmem_enabled(_cfg) \
     (_cfg.hostmem > 0)
+#define virtio_gpu_venus_enabled(_cfg) \
+    (_cfg.flags & (1 << VIRTIO_GPU_FLAG_VENUS_ENABLED))
 
 struct virtio_gpu_base_conf {
     uint32_t max_outputs;
diff --git a/meson.build b/meson.build
index 0433ea200dd7..20241931730f 100644
--- a/meson.build
+++ b/meson.build
@@ -2290,6 +2290,7 @@ if virgl.version().version_compare('>=1.0.0')
   config_host_data.set('HAVE_VIRGL_D3D_INFO_EXT', 1)
   config_host_data.set('HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS', 1)
   config_host_data.set('HAVE_VIRGL_RESOURCE_BLOB', 1)
+  config_host_data.set('HAVE_VIRGL_VENUS', 1)
 endif
 config_host_data.set('CONFIG_VIRTFS', have_virtfs)
 config_host_data.set('CONFIG_VTE', vte.found())
-- 
2.44.0

Re: [PATCH v11 01/10] virtio-gpu: Unrealize GL device

[Akihiko Odaki]

On 2024/05/12 3:22, Dmitry Osipenko wrote:
> Even though GL GPU doesn't support hotplugging today, free virgl
> resources when GL device is unrealized. For consistency.
> 
> Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
> ---
>   hw/display/virtio-gpu-gl.c     | 11 +++++++++++
>   hw/display/virtio-gpu-virgl.c  |  9 +++++++++
>   include/hw/virtio/virtio-gpu.h |  1 +
>   3 files changed, 21 insertions(+)
> 
> diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
> index e06be60dfbfc..0c0a8d136954 100644
> --- a/hw/display/virtio-gpu-gl.c
> +++ b/hw/display/virtio-gpu-gl.c
> @@ -136,6 +136,16 @@ static Property virtio_gpu_gl_properties[] = {
>       DEFINE_PROP_END_OF_LIST(),
>   };
>   
> +static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
> +{
> +    VirtIOGPU *g = VIRTIO_GPU(qdev);
> +    VirtIOGPUGL *gl = VIRTIO_GPU_GL(qdev);
> +
> +    if (gl->renderer_inited) {
> +        virtio_gpu_virgl_deinit(g);
> +    }
> +}
> +
>   static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
>   {
>       DeviceClass *dc = DEVICE_CLASS(klass);
> @@ -149,6 +159,7 @@ static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
>       vgc->update_cursor_data = virtio_gpu_gl_update_cursor_data;
>   
>       vdc->realize = virtio_gpu_gl_device_realize;
> +    vdc->unrealize = virtio_gpu_gl_device_unrealize;
>       vdc->reset = virtio_gpu_gl_reset;
>       device_class_set_props(dc, virtio_gpu_gl_properties);
>   }
> diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
> index 9f34d0e6619c..b0500eccf8e0 100644
> --- a/hw/display/virtio-gpu-virgl.c
> +++ b/hw/display/virtio-gpu-virgl.c
> @@ -665,3 +665,12 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
>   
>       return capset2_max_ver ? 2 : 1;
>   }
> +
> +void virtio_gpu_virgl_deinit(VirtIOGPU *g)
> +{
> +    if (g->fence_poll) {

Isn't g->fence_poll always non-NULL when this function is called?

> +        timer_free(g->fence_poll);
> +    }
> +
> +    virgl_renderer_cleanup(NULL);
> +}
> diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
> index ed44cdad6b34..b657187159d9 100644
> --- a/include/hw/virtio/virtio-gpu.h
> +++ b/include/hw/virtio/virtio-gpu.h
> @@ -336,6 +336,7 @@ void virtio_gpu_virgl_fence_poll(VirtIOGPU *g);
>   void virtio_gpu_virgl_reset_scanout(VirtIOGPU *g);
>   void virtio_gpu_virgl_reset(VirtIOGPU *g);
>   int virtio_gpu_virgl_init(VirtIOGPU *g);
> +void virtio_gpu_virgl_deinit(VirtIOGPU *g);
>   int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g);
>   
>   #endif

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Akihiko Odaki]

On 2024/05/12 3:22, Dmitry Osipenko wrote:
> From: Antonio Caggiano <antonio.caggiano@collabora.com>
> 
> Support BLOB resources creation, mapping and unmapping by calling the
> new stable virglrenderer 0.10 interface. Only enabled when available and
> via the blob config. E.g. -device virtio-vga-gl,blob=true
> 
> Signed-off-by: Antonio Caggiano <antonio.caggiano@collabora.com>
> Signed-off-by: Xenia Ragiadakou <xenia.ragiadakou@amd.com>
> Signed-off-by: Huang Rui <ray.huang@amd.com>
> Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
> ---
>   hw/display/virtio-gpu-virgl.c  | 274 ++++++++++++++++++++++++++++++++-
>   hw/display/virtio-gpu.c        |   4 +-
>   include/hw/virtio/virtio-gpu.h |   2 +
>   3 files changed, 277 insertions(+), 3 deletions(-)
> 
> diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
> index 524f15220b7f..3f2e406be3a4 100644
> --- a/hw/display/virtio-gpu-virgl.c
> +++ b/hw/display/virtio-gpu-virgl.c
> @@ -26,6 +26,7 @@
>   
>   struct virtio_gpu_virgl_resource {
>       struct virtio_gpu_simple_resource base;
> +    MemoryRegion *mr;
>   };
>   
>   static struct virtio_gpu_virgl_resource *
> @@ -49,6 +50,117 @@ virgl_get_egl_display(G_GNUC_UNUSED void *cookie)
>   }
>   #endif
>   
> +#ifdef HAVE_VIRGL_RESOURCE_BLOB
> +struct virtio_gpu_virgl_hostmem_region {
> +    MemoryRegion mr;
> +    struct VirtIOGPU *g;
> +    struct virtio_gpu_virgl_resource *res;
> +};
> +
> +static void virtio_gpu_virgl_resume_cmdq_bh(void *opaque)
> +{
> +    VirtIOGPU *g = opaque;
> +
> +    virtio_gpu_process_cmdq(g);
> +}
> +
> +static void virtio_gpu_virgl_hostmem_region_free(void *obj)
> +{
> +    MemoryRegion *mr = MEMORY_REGION(obj);
> +    struct virtio_gpu_virgl_hostmem_region *vmr;
> +    VirtIOGPUBase *b;
> +
> +    vmr = container_of(mr, struct virtio_gpu_virgl_hostmem_region, mr);
> +    vmr->res->mr = NULL;
> +
> +    b = VIRTIO_GPU_BASE(vmr->g);
> +    b->renderer_blocked--;
> +
> +    /*
> +     * memory_region_unref() is executed from RCU thread context, while
> +     * virglrenderer works only on the main-loop thread that's holding GL
> +     * context.
> +     */
> +    qemu_bh_schedule(vmr->g->cmdq_resume_bh);
> +    g_free(vmr);
> +}
> +
> +static int
> +virtio_gpu_virgl_map_resource_blob(VirtIOGPU *g,
> +                                   struct virtio_gpu_virgl_resource *res,
> +                                   uint64_t offset)
> +{
> +    struct virtio_gpu_virgl_hostmem_region *vmr;
> +    VirtIOGPUBase *b = VIRTIO_GPU_BASE(g);
> +    MemoryRegion *mr;
> +    uint64_t size;
> +    void *data;
> +    int ret;
> +
> +    if (!virtio_gpu_hostmem_enabled(b->conf)) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: hostmem disabled\n", __func__);
> +        return -EOPNOTSUPP;
> +    }
> +
> +    ret = virgl_renderer_resource_map(res->base.resource_id, &data, &size);
> +    if (ret) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: failed to map virgl resource: %s\n",
> +                      __func__, strerror(-ret));
> +        return ret;
> +    }
> +
> +    vmr = g_new0(struct virtio_gpu_virgl_hostmem_region, 1);
> +    vmr->res = res;
> +    vmr->g = g;
> +
> +    mr = &vmr->mr;
> +    memory_region_init_ram_ptr(mr, OBJECT(mr), "blob", size, data);
> +    memory_region_add_subregion(&b->hostmem, offset, mr);
> +    memory_region_set_enabled(mr, true);
> +
> +    /*
> +     * MR could outlive the resource if MR's reference is held outside of
> +     * virtio-gpu. In order to prevent unmapping resource while MR is alive,
> +     * and thus, making the data pointer invalid, we will block virtio-gpu
> +     * command processing until MR is fully unreferenced and freed.
> +     */
> +    OBJECT(mr)->free = virtio_gpu_virgl_hostmem_region_free;
> +
> +    res->mr = mr;
> +
> +    return 0;
> +}
> +
> +static int
> +virtio_gpu_virgl_async_unmap_resource_blob(VirtIOGPU *g,
> +                                           struct virtio_gpu_virgl_resource *res)
> +{
> +    VirtIOGPUBase *b = VIRTIO_GPU_BASE(g);
> +    MemoryRegion *mr = res->mr;
> +    int ret;
> +
> +    if (mr) {
> +        /* render will be unblocked once MR is freed */
> +        b->renderer_blocked++;
> +
> +        /* memory region owns self res->mr object and frees it by itself */
> +        memory_region_set_enabled(mr, false);
> +        memory_region_del_subregion(&b->hostmem, mr);
> +        object_unparent(OBJECT(mr));
> +    } else {
> +        ret = virgl_renderer_resource_unmap(res->base.resource_id);
> +        if (ret) {
> +            qemu_log_mask(LOG_GUEST_ERROR,
> +                          "%s: failed to unmap virgl resource: %s\n",
> +                          __func__, strerror(-ret));
> +            return ret;
> +        }
> +    }
> +
> +    return 0;
> +}
> +#endif /* HAVE_VIRGL_RESOURCE_BLOB */
> +
>   static void virgl_cmd_create_resource_2d(VirtIOGPU *g,
>                                            struct virtio_gpu_ctrl_command *cmd)
>   {
> @@ -146,7 +258,8 @@ static void virgl_cmd_create_resource_3d(VirtIOGPU *g,
>   }
>   
>   static void virgl_cmd_resource_unref(VirtIOGPU *g,
> -                                     struct virtio_gpu_ctrl_command *cmd)
> +                                     struct virtio_gpu_ctrl_command *cmd,
> +                                     bool *cmd_suspended)

This parameter is not used as it returns an error if the resource is 
still mapped.

It may be better to actually implement unmapping instead of returning an 
error for consistency with the iov operation. Apparently crosvm also 
unmaps blobs with VIRTIO_GPU_CMD_RESOURCE_UNREF.

>   {
>       struct virtio_gpu_resource_unref unref;
>       struct virtio_gpu_virgl_resource *res;
> @@ -164,6 +277,12 @@ static void virgl_cmd_resource_unref(VirtIOGPU *g,
>           return;
>       }
>   
> +    if (res->mr) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource is mapped\n", __func__);
> +        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
> +        return;
> +    }
> +
>       virgl_renderer_resource_detach_iov(unref.resource_id,
>                                          &res_iovs,
>                                          &num_iovs);
> @@ -514,6 +633,134 @@ static void virgl_cmd_get_capset(VirtIOGPU *g,
>   }
>   
>   #ifdef HAVE_VIRGL_RESOURCE_BLOB
> +static void virgl_cmd_resource_create_blob(VirtIOGPU *g,
> +                                           struct virtio_gpu_ctrl_command *cmd)
> +{
> +    struct virgl_renderer_resource_create_blob_args virgl_args = { 0 };
> +    struct virtio_gpu_resource_create_blob cblob;
> +    struct virtio_gpu_virgl_resource *res;
> +    int ret;
> +
> +    if (!virtio_gpu_blob_enabled(g->parent_obj.conf)) {
> +        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_PARAMETER;
> +        return;
> +    }
> +
> +    VIRTIO_GPU_FILL_CMD(cblob);
> +    virtio_gpu_create_blob_bswap(&cblob);
> +    trace_virtio_gpu_cmd_res_create_blob(cblob.resource_id, cblob.size);
> +
> +    if (cblob.resource_id == 0) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource id 0 is not allowed\n",
> +                      __func__);
> +        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
> +        return;
> +    }
> +
> +    res = virtio_gpu_virgl_find_resource(g, cblob.resource_id);
> +    if (res) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource already exists %d\n",
> +                      __func__, cblob.resource_id);
> +        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
> +        return;
> +    }
> +
> +    res = g_new0(struct virtio_gpu_virgl_resource, 1);
> +    res->base.resource_id = cblob.resource_id;
> +    res->base.blob_size = cblob.size;
> +    res->base.dmabuf_fd = -1;
> +
> +    if (cblob.blob_mem != VIRTIO_GPU_BLOB_MEM_HOST3D) {
> +        ret = virtio_gpu_create_mapping_iov(g, cblob.nr_entries, sizeof(cblob),
> +                                            cmd, &res->base.addrs,
> +                                            &res->base.iov, &res->base.iov_cnt);
> +        if (!ret) {
> +            g_free(res);
> +            cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
> +            return;
> +        }
> +    }
> +
> +    QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
> +
> +    virgl_args.res_handle = cblob.resource_id;
> +    virgl_args.ctx_id = cblob.hdr.ctx_id;
> +    virgl_args.blob_mem = cblob.blob_mem;
> +    virgl_args.blob_id = cblob.blob_id;
> +    virgl_args.blob_flags = cblob.blob_flags;
> +    virgl_args.size = cblob.size;
> +    virgl_args.iovecs = res->base.iov;
> +    virgl_args.num_iovs = res->base.iov_cnt;
> +
> +    ret = virgl_renderer_resource_create_blob(&virgl_args);
> +    if (ret) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: virgl blob create error: %s\n",
> +                      __func__, strerror(-ret));
> +        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;

reslist keeps the stale res even if an error happens.

> +    }
> +}
> +
> +static void virgl_cmd_resource_map_blob(VirtIOGPU *g,
> +                                        struct virtio_gpu_ctrl_command *cmd)
> +{
> +    struct virtio_gpu_resource_map_blob mblob;
> +    struct virtio_gpu_virgl_resource *res;
> +    struct virtio_gpu_resp_map_info resp;
> +    int ret;
> +
> +    VIRTIO_GPU_FILL_CMD(mblob);
> +    virtio_gpu_map_blob_bswap(&mblob);
> +
> +    res = virtio_gpu_virgl_find_resource(g, mblob.resource_id);
> +    if (!res) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not exist %d\n",
> +                      __func__, mblob.resource_id);
> +        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
> +        return;
> +    }
> +
> +    ret = virtio_gpu_virgl_map_resource_blob(g, res, mblob.offset);
> +    if (ret) {
> +        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
> +        return;
> +    }
> +
> +    memset(&resp, 0, sizeof(resp));
> +    resp.hdr.type = VIRTIO_GPU_RESP_OK_MAP_INFO;
> +    virgl_renderer_resource_get_map_info(mblob.resource_id, &resp.map_info);
> +    virtio_gpu_ctrl_response(g, cmd, &resp.hdr, sizeof(resp));
> +}
> +
> +static void virgl_cmd_resource_unmap_blob(VirtIOGPU *g,
> +                                          struct virtio_gpu_ctrl_command *cmd,
> +                                          bool *cmd_suspended)
> +{
> +    struct virtio_gpu_resource_unmap_blob ublob;
> +    struct virtio_gpu_virgl_resource *res;
> +    int ret;
> +
> +    VIRTIO_GPU_FILL_CMD(ublob);
> +    virtio_gpu_unmap_blob_bswap(&ublob);
> +
> +    res = virtio_gpu_virgl_find_resource(g, ublob.resource_id);
> +    if (!res) {
> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: resource does not exist %d\n",
> +                      __func__, ublob.resource_id);
> +        cmd->error = VIRTIO_GPU_RESP_ERR_INVALID_RESOURCE_ID;
> +        return;
> +    }
> +
> +    ret = virtio_gpu_virgl_async_unmap_resource_blob(g, res);
> +    if (ret) {
> +        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
> +        return;
> +    }
> +
> +    if (res->mr) {
> +        *cmd_suspended = true;
> +    }
> +}
> +
>   static void virgl_cmd_set_scanout_blob(VirtIOGPU *g,
>                                          struct virtio_gpu_ctrl_command *cmd)
>   {
> @@ -616,6 +863,8 @@ static void virgl_cmd_set_scanout_blob(VirtIOGPU *g,
>   void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
>                                         struct virtio_gpu_ctrl_command *cmd)
>   {
> +    bool cmd_suspended = false;
> +
>       VIRTIO_GPU_FILL_CMD(cmd->cmd_hdr);
>   
>       virgl_renderer_force_ctx_0();
> @@ -657,7 +906,7 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
>           virgl_cmd_resource_flush(g, cmd);
>           break;
>       case VIRTIO_GPU_CMD_RESOURCE_UNREF:
> -        virgl_cmd_resource_unref(g, cmd);
> +        virgl_cmd_resource_unref(g, cmd, &cmd_suspended);
>           break;
>       case VIRTIO_GPU_CMD_CTX_ATTACH_RESOURCE:
>           /* TODO add security */
> @@ -680,6 +929,15 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
>           virtio_gpu_get_edid(g, cmd);
>           break;
>   #ifdef HAVE_VIRGL_RESOURCE_BLOB
> +    case VIRTIO_GPU_CMD_RESOURCE_CREATE_BLOB:
> +        virgl_cmd_resource_create_blob(g, cmd);
> +        break;
> +    case VIRTIO_GPU_CMD_RESOURCE_MAP_BLOB:
> +        virgl_cmd_resource_map_blob(g, cmd);
> +        break;
> +    case VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB:
> +        virgl_cmd_resource_unmap_blob(g, cmd, &cmd_suspended);
> +        break;
>       case VIRTIO_GPU_CMD_SET_SCANOUT_BLOB:
>           virgl_cmd_set_scanout_blob(g, cmd);
>           break;
> @@ -689,6 +947,9 @@ void virtio_gpu_virgl_process_cmd(VirtIOGPU *g,
>           break;
>       }
>   
> +    if (cmd_suspended) {
> +        return;
> +    }
>       if (cmd->finished) {
>           return;
>       }

Nitpick: I would write cmd_suspended || cmd->finished to save a few lines.

> @@ -854,6 +1115,11 @@ int virtio_gpu_virgl_init(VirtIOGPU *g)
>                                         virtio_gpu_print_stats, g);
>           timer_mod(g->print_stats, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + 1000);
>       }
> +
> +    g->cmdq_resume_bh = aio_bh_new(qemu_get_aio_context(),
> +                                   virtio_gpu_virgl_resume_cmdq_bh,
> +                                   g);
> +
>       return 0;
>   }
>   
> @@ -869,6 +1135,10 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
>   
>   void virtio_gpu_virgl_deinit(VirtIOGPU *g)
>   {
> +    if (g->cmdq_resume_bh) {
> +        qemu_bh_delete(g->cmdq_resume_bh);
> +    }
> +
>       if (g->fence_poll) {
>           timer_free(g->fence_poll);
>       }
> diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
> index 6c8c7213bafa..052ab493a00b 100644
> --- a/hw/display/virtio-gpu.c
> +++ b/hw/display/virtio-gpu.c
> @@ -1483,10 +1483,12 @@ void virtio_gpu_device_realize(DeviceState *qdev, Error **errp)
>               return;
>           }
>   
> +#ifndef HAVE_VIRGL_RESOURCE_BLOB
>           if (virtio_gpu_virgl_enabled(g->parent_obj.conf)) {
> -            error_setg(errp, "blobs and virgl are not compatible (yet)");
> +            error_setg(errp, "old virglrenderer, blob resources unsupported");
>               return;
>           }
> +#endif
>       }
>   
>       if (!virtio_gpu_base_device_realize(qdev,
> diff --git a/include/hw/virtio/virtio-gpu.h b/include/hw/virtio/virtio-gpu.h
> index ba36497c477f..c0b0b0eac08b 100644
> --- a/include/hw/virtio/virtio-gpu.h
> +++ b/include/hw/virtio/virtio-gpu.h
> @@ -210,6 +210,8 @@ struct VirtIOGPU {
>           QTAILQ_HEAD(, VGPUDMABuf) bufs;
>           VGPUDMABuf *primary[VIRTIO_GPU_MAX_SCANOUTS];
>       } dmabuf;
> +
> +    QEMUBH *cmdq_resume_bh;

Move this to VirtIOGPUGL.

>   };
>   
>   struct VirtIOGPUClass {

Re: [PATCH v11 09/10] virtio-gpu: Register capsets dynamically

[Akihiko Odaki]

On 2024/05/12 3:22, Dmitry Osipenko wrote:
> From: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
> 
> virtio_gpu_virgl_get_num_capsets will return "num_capsets", but we can't
> assume that capset_index 1 is always VIRGL2 once we'll support more capsets,
> like Venus and DRM capsets. Register capsets dynamically to avoid that problem.
> 
> Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
> Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
> ---
>   hw/display/virtio-gpu-gl.c     |  6 ++++--
>   hw/display/virtio-gpu-virgl.c  | 33 +++++++++++++++++++++------------
>   include/hw/virtio/virtio-gpu.h |  4 +++-
>   3 files changed, 28 insertions(+), 15 deletions(-)
> 
> diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
> index 95806999189e..431fc2881a00 100644
> --- a/hw/display/virtio-gpu-gl.c
> +++ b/hw/display/virtio-gpu-gl.c
> @@ -124,8 +124,8 @@ static void virtio_gpu_gl_device_realize(DeviceState *qdev, Error **errp)
>       }
>   
>       g->parent_obj.conf.flags |= (1 << VIRTIO_GPU_FLAG_VIRGL_ENABLED);
> -    VIRTIO_GPU_BASE(g)->virtio_config.num_capsets =
> -        virtio_gpu_virgl_get_num_capsets(g);
> +    g->capset_ids = virtio_gpu_virgl_get_capsets(g);
> +    VIRTIO_GPU_BASE(g)->virtio_config.num_capsets = g->capset_ids->len;
>   
>   #ifdef HAVE_VIRGL_CONTEXT_CREATE_WITH_FLAGS
>       g->parent_obj.conf.flags |= 1 << VIRTIO_GPU_FLAG_CONTEXT_INIT_ENABLED;
> @@ -148,6 +148,8 @@ static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
>       if (gl->renderer_inited) {
>           virtio_gpu_virgl_deinit(g);
>       }
> +
> +    g_array_unref(g->capset_ids);
>   }
>   
>   static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
> diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
> index 3f2e406be3a4..135974431492 100644
> --- a/hw/display/virtio-gpu-virgl.c
> +++ b/hw/display/virtio-gpu-virgl.c
> @@ -590,19 +590,13 @@ static void virgl_cmd_get_capset_info(VirtIOGPU *g,
>       VIRTIO_GPU_FILL_CMD(info);
>   
>       memset(&resp, 0, sizeof(resp));
> -    if (info.capset_index == 0) {
> -        resp.capset_id = VIRTIO_GPU_CAPSET_VIRGL;
> -        virgl_renderer_get_cap_set(resp.capset_id,
> -                                   &resp.capset_max_version,
> -                                   &resp.capset_max_size);
> -    } else if (info.capset_index == 1) {
> -        resp.capset_id = VIRTIO_GPU_CAPSET_VIRGL2;
> +
> +    if (info.capset_index < g->capset_ids->len) {
> +        resp.capset_id = g_array_index(g->capset_ids, uint32_t,
> +                                       info.capset_index);
>           virgl_renderer_get_cap_set(resp.capset_id,
>                                      &resp.capset_max_version,
>                                      &resp.capset_max_size);
> -    } else {
> -        resp.capset_max_version = 0;
> -        resp.capset_max_size = 0;
>       }
>       resp.hdr.type = VIRTIO_GPU_RESP_OK_CAPSET_INFO;
>       virtio_gpu_ctrl_response(g, cmd, &resp.hdr, sizeof(resp));
> @@ -1123,14 +1117,29 @@ int virtio_gpu_virgl_init(VirtIOGPU *g)
>       return 0;
>   }
>   
> -int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
> +static void virtio_gpu_virgl_add_capset(GArray *capset_ids, uint32_t capset_id)
> +{
> +    g_array_append_val(capset_ids, capset_id);
> +}

Is it worthwhile to have a function for this?

Re: [PATCH v11 01/10] virtio-gpu: Unrealize GL device

[Dmitry Osipenko]

On 5/13/24 11:44, Akihiko Odaki wrote:
> On 2024/05/12 3:22, Dmitry Osipenko wrote:
>> Even though GL GPU doesn't support hotplugging today, free virgl
>> resources when GL device is unrealized. For consistency.
>>
>> Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
>> ---
>>   hw/display/virtio-gpu-gl.c     | 11 +++++++++++
>>   hw/display/virtio-gpu-virgl.c  |  9 +++++++++
>>   include/hw/virtio/virtio-gpu.h |  1 +
>>   3 files changed, 21 insertions(+)
>>
>> diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
>> index e06be60dfbfc..0c0a8d136954 100644
>> --- a/hw/display/virtio-gpu-gl.c
>> +++ b/hw/display/virtio-gpu-gl.c
>> @@ -136,6 +136,16 @@ static Property virtio_gpu_gl_properties[] = {
>>       DEFINE_PROP_END_OF_LIST(),
>>   };
>>   +static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
>> +{
>> +    VirtIOGPU *g = VIRTIO_GPU(qdev);
>> +    VirtIOGPUGL *gl = VIRTIO_GPU_GL(qdev);
>> +
>> +    if (gl->renderer_inited) {
>> +        virtio_gpu_virgl_deinit(g);
>> +    }
>> +}
>> +
>>   static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
>>   {
>>       DeviceClass *dc = DEVICE_CLASS(klass);
>> @@ -149,6 +159,7 @@ static void virtio_gpu_gl_class_init(ObjectClass
>> *klass, void *data)
>>       vgc->update_cursor_data = virtio_gpu_gl_update_cursor_data;
>>         vdc->realize = virtio_gpu_gl_device_realize;
>> +    vdc->unrealize = virtio_gpu_gl_device_unrealize;
>>       vdc->reset = virtio_gpu_gl_reset;
>>       device_class_set_props(dc, virtio_gpu_gl_properties);
>>   }
>> diff --git a/hw/display/virtio-gpu-virgl.c
>> b/hw/display/virtio-gpu-virgl.c
>> index 9f34d0e6619c..b0500eccf8e0 100644
>> --- a/hw/display/virtio-gpu-virgl.c
>> +++ b/hw/display/virtio-gpu-virgl.c
>> @@ -665,3 +665,12 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
>>         return capset2_max_ver ? 2 : 1;
>>   }
>> +
>> +void virtio_gpu_virgl_deinit(VirtIOGPU *g)
>> +{
>> +    if (g->fence_poll) {
> 
> Isn't g->fence_poll always non-NULL when this function is called?

virtio_gpu_virgl_init() is invoked when first cmd is executed, please
see virtio_gpu_gl_handle_ctrl() that invokes it. Hence g->fence_poll can
be NULL.

-- 
Best regards,
Dmitry

Re: [PATCH v11 01/10] virtio-gpu: Unrealize GL device

[Akihiko Odaki]

On 2024/05/16 1:18, Dmitry Osipenko wrote:
> On 5/13/24 11:44, Akihiko Odaki wrote:
>> On 2024/05/12 3:22, Dmitry Osipenko wrote:
>>> Even though GL GPU doesn't support hotplugging today, free virgl
>>> resources when GL device is unrealized. For consistency.
>>>
>>> Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
>>> ---
>>>    hw/display/virtio-gpu-gl.c     | 11 +++++++++++
>>>    hw/display/virtio-gpu-virgl.c  |  9 +++++++++
>>>    include/hw/virtio/virtio-gpu.h |  1 +
>>>    3 files changed, 21 insertions(+)
>>>
>>> diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
>>> index e06be60dfbfc..0c0a8d136954 100644
>>> --- a/hw/display/virtio-gpu-gl.c
>>> +++ b/hw/display/virtio-gpu-gl.c
>>> @@ -136,6 +136,16 @@ static Property virtio_gpu_gl_properties[] = {
>>>        DEFINE_PROP_END_OF_LIST(),
>>>    };
>>>    +static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
>>> +{
>>> +    VirtIOGPU *g = VIRTIO_GPU(qdev);
>>> +    VirtIOGPUGL *gl = VIRTIO_GPU_GL(qdev);
>>> +
>>> +    if (gl->renderer_inited) {
>>> +        virtio_gpu_virgl_deinit(g);
>>> +    }
>>> +}
>>> +
>>>    static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
>>>    {
>>>        DeviceClass *dc = DEVICE_CLASS(klass);
>>> @@ -149,6 +159,7 @@ static void virtio_gpu_gl_class_init(ObjectClass
>>> *klass, void *data)
>>>        vgc->update_cursor_data = virtio_gpu_gl_update_cursor_data;
>>>          vdc->realize = virtio_gpu_gl_device_realize;
>>> +    vdc->unrealize = virtio_gpu_gl_device_unrealize;
>>>        vdc->reset = virtio_gpu_gl_reset;
>>>        device_class_set_props(dc, virtio_gpu_gl_properties);
>>>    }
>>> diff --git a/hw/display/virtio-gpu-virgl.c
>>> b/hw/display/virtio-gpu-virgl.c
>>> index 9f34d0e6619c..b0500eccf8e0 100644
>>> --- a/hw/display/virtio-gpu-virgl.c
>>> +++ b/hw/display/virtio-gpu-virgl.c
>>> @@ -665,3 +665,12 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
>>>          return capset2_max_ver ? 2 : 1;
>>>    }
>>> +
>>> +void virtio_gpu_virgl_deinit(VirtIOGPU *g)
>>> +{
>>> +    if (g->fence_poll) {
>>
>> Isn't g->fence_poll always non-NULL when this function is called?
> 
> virtio_gpu_virgl_init() is invoked when first cmd is executed, please
> see virtio_gpu_gl_handle_ctrl() that invokes it. Hence g->fence_poll can
> be NULL.
> 

But it already checks renderer_inited, doesn't it? And I think it's 
better to utilize one single flag to represent that virgl is enabled 
instead of checking several variables (fence_poll and cmdq_resume_bh in 
the future).

Re: [PATCH v11 09/10] virtio-gpu: Register capsets dynamically

[Dmitry Osipenko]

On 5/13/24 12:20, Akihiko Odaki wrote:
...
>>   -int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
>> +static void virtio_gpu_virgl_add_capset(GArray *capset_ids, uint32_t
>> capset_id)
>> +{
>> +    g_array_append_val(capset_ids, capset_id);
>> +}
> 
> Is it worthwhile to have a function for this?

It's necessary to have it because g_array_append_val() is actually a macro that takes &capset_id ptr internally. 

I.e. g_array_append_val(capset_ids, VIRTIO_GPU_CAPSET_VIRGL2) will fail to compile with:

/usr/include/glib-2.0/glib/garray.h:66:59: error: lvalue required as unary '&' operand
   66 | #define g_array_append_val(a,v)   g_array_append_vals (a, &(v), 1)

-- 
Best regards,
Dmitry

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Dmitry Osipenko]

On 5/13/24 12:18, Akihiko Odaki wrote:
>>     static void virgl_cmd_resource_unref(VirtIOGPU *g,
>> -                                     struct virtio_gpu_ctrl_command
>> *cmd)
>> +                                     struct virtio_gpu_ctrl_command
>> *cmd,
>> +                                     bool *cmd_suspended)
> 
> This parameter is not used as it returns an error if the resource is
> still mapped.

Missed to remove it by accident

> It may be better to actually implement unmapping instead of returning an
> error for consistency with the iov operation. Apparently crosvm also
> unmaps blobs with VIRTIO_GPU_CMD_RESOURCE_UNREF.

Then I'll add back `async_unmap_in_progress` because resource can be
both mapped/unmapped on unref, and we'll need flag to know whether async
unmapping has been finished to do the final unmapping of the resource.

...
>> +    QTAILQ_INSERT_HEAD(&g->reslist, &res->base, next);
>> +
>> +    virgl_args.res_handle = cblob.resource_id;
>> +    virgl_args.ctx_id = cblob.hdr.ctx_id;
>> +    virgl_args.blob_mem = cblob.blob_mem;
>> +    virgl_args.blob_id = cblob.blob_id;
>> +    virgl_args.blob_flags = cblob.blob_flags;
>> +    virgl_args.size = cblob.size;
>> +    virgl_args.iovecs = res->base.iov;
>> +    virgl_args.num_iovs = res->base.iov_cnt;
>> +
>> +    ret = virgl_renderer_resource_create_blob(&virgl_args);
>> +    if (ret) {
>> +        qemu_log_mask(LOG_GUEST_ERROR, "%s: virgl blob create error:
>> %s\n",
>> +                      __func__, strerror(-ret));
>> +        cmd->error = VIRTIO_GPU_RESP_ERR_UNSPEC;
> 
> reslist keeps the stale res even if an error happens.

Good catch

-- 
Best regards,
Dmitry

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Akihiko Odaki]

On 2024/05/16 1:39, Dmitry Osipenko wrote:
> On 5/13/24 12:18, Akihiko Odaki wrote:
>>>      static void virgl_cmd_resource_unref(VirtIOGPU *g,
>>> -                                     struct virtio_gpu_ctrl_command
>>> *cmd)
>>> +                                     struct virtio_gpu_ctrl_command
>>> *cmd,
>>> +                                     bool *cmd_suspended)
>>
>> This parameter is not used as it returns an error if the resource is
>> still mapped.
> 
> Missed to remove it by accident
> 
>> It may be better to actually implement unmapping instead of returning an
>> error for consistency with the iov operation. Apparently crosvm also
>> unmaps blobs with VIRTIO_GPU_CMD_RESOURCE_UNREF.
> 
> Then I'll add back `async_unmap_in_progress` because resource can be
> both mapped/unmapped on unref, and we'll need flag to know whether async
> unmapping has been finished to do the final unmapping of the resource.

Such a situation should be already handled since unmapping in progress 
blocks all commands (not just VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB but 
literally all, including VIRTIO_GPU_CMD_RESOURCE_UNREF).

Re: [PATCH v11 01/10] virtio-gpu: Unrealize GL device

[Dmitry Osipenko]

On 5/15/24 19:22, Akihiko Odaki wrote:
> On 2024/05/16 1:18, Dmitry Osipenko wrote:
>> On 5/13/24 11:44, Akihiko Odaki wrote:
>>> On 2024/05/12 3:22, Dmitry Osipenko wrote:
>>>> Even though GL GPU doesn't support hotplugging today, free virgl
>>>> resources when GL device is unrealized. For consistency.
>>>>
>>>> Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
>>>> ---
>>>>    hw/display/virtio-gpu-gl.c     | 11 +++++++++++
>>>>    hw/display/virtio-gpu-virgl.c  |  9 +++++++++
>>>>    include/hw/virtio/virtio-gpu.h |  1 +
>>>>    3 files changed, 21 insertions(+)
>>>>
>>>> diff --git a/hw/display/virtio-gpu-gl.c b/hw/display/virtio-gpu-gl.c
>>>> index e06be60dfbfc..0c0a8d136954 100644
>>>> --- a/hw/display/virtio-gpu-gl.c
>>>> +++ b/hw/display/virtio-gpu-gl.c
>>>> @@ -136,6 +136,16 @@ static Property virtio_gpu_gl_properties[] = {
>>>>        DEFINE_PROP_END_OF_LIST(),
>>>>    };
>>>>    +static void virtio_gpu_gl_device_unrealize(DeviceState *qdev)
>>>> +{
>>>> +    VirtIOGPU *g = VIRTIO_GPU(qdev);
>>>> +    VirtIOGPUGL *gl = VIRTIO_GPU_GL(qdev);
>>>> +
>>>> +    if (gl->renderer_inited) {
>>>> +        virtio_gpu_virgl_deinit(g);
>>>> +    }
>>>> +}
>>>> +
>>>>    static void virtio_gpu_gl_class_init(ObjectClass *klass, void *data)
>>>>    {
>>>>        DeviceClass *dc = DEVICE_CLASS(klass);
>>>> @@ -149,6 +159,7 @@ static void virtio_gpu_gl_class_init(ObjectClass
>>>> *klass, void *data)
>>>>        vgc->update_cursor_data = virtio_gpu_gl_update_cursor_data;
>>>>          vdc->realize = virtio_gpu_gl_device_realize;
>>>> +    vdc->unrealize = virtio_gpu_gl_device_unrealize;
>>>>        vdc->reset = virtio_gpu_gl_reset;
>>>>        device_class_set_props(dc, virtio_gpu_gl_properties);
>>>>    }
>>>> diff --git a/hw/display/virtio-gpu-virgl.c
>>>> b/hw/display/virtio-gpu-virgl.c
>>>> index 9f34d0e6619c..b0500eccf8e0 100644
>>>> --- a/hw/display/virtio-gpu-virgl.c
>>>> +++ b/hw/display/virtio-gpu-virgl.c
>>>> @@ -665,3 +665,12 @@ int virtio_gpu_virgl_get_num_capsets(VirtIOGPU *g)
>>>>          return capset2_max_ver ? 2 : 1;
>>>>    }
>>>> +
>>>> +void virtio_gpu_virgl_deinit(VirtIOGPU *g)
>>>> +{
>>>> +    if (g->fence_poll) {
>>>
>>> Isn't g->fence_poll always non-NULL when this function is called?
>>
>> virtio_gpu_virgl_init() is invoked when first cmd is executed, please
>> see virtio_gpu_gl_handle_ctrl() that invokes it. Hence g->fence_poll can
>> be NULL.
>>
> 
> But it already checks renderer_inited, doesn't it? And I think it's
> better to utilize one single flag to represent that virgl is enabled
> instead of checking several variables (fence_poll and cmdq_resume_bh in
> the future).

The virtio_gpu_virgl_init() will have to be changed to do that because
virgl_renderer_init() can fail before fence_poll/cmdq_resume_bh are inited.

Though, the error returned by virtio_gpu_virgl_init() isn't checked by
virtio_gpu_gl_handle_ctrl(), which leads to a further Qemu crash due to
fence_poll=NULL. I'll try to improve it all in v12, thanks.

-- 
Best regards,
Dmitry

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Dmitry Osipenko]

On 5/15/24 19:42, Akihiko Odaki wrote:
>>> It may be better to actually implement unmapping instead of returning an
>>> error for consistency with the iov operation. Apparently crosvm also
>>> unmaps blobs with VIRTIO_GPU_CMD_RESOURCE_UNREF.
>>
>> Then I'll add back `async_unmap_in_progress` because resource can be
>> both mapped/unmapped on unref, and we'll need flag to know whether async
>> unmapping has been finished to do the final unmapping of the resource.
> 
> Such a situation should be already handled since unmapping in progress
> blocks all commands (not just VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB but
> literally all, including VIRTIO_GPU_CMD_RESOURCE_UNREF).

The async unmapping consists of 3 parts:

1. begin async unmapping with memory_region_del_subregion() and suspend
2. wait for res->mr to be freed and resume
3. finish the unmapping with final virgl_renderer_resource_unmap()

Parts 1 and 3 are handled by  virtio_gpu_virgl_async_unmap_resource_blob()


The VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB is different because we know that
blob is mapped in the first place. Hence we can safely perform the part
3, assuming that parts 1/2 has been completed.

In case of VIRTIO_GPU_CMD_RESOURCE_UNREF, blob can be unmapped in the
first place and we can't do the part 3 because it will error out for
unmapped resource since parts 1/2 were not performed.

-- 
Best regards,
Dmitry

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Akihiko Odaki]

On 2024/05/16 2:01, Dmitry Osipenko wrote:
> On 5/15/24 19:42, Akihiko Odaki wrote:
>>>> It may be better to actually implement unmapping instead of returning an
>>>> error for consistency with the iov operation. Apparently crosvm also
>>>> unmaps blobs with VIRTIO_GPU_CMD_RESOURCE_UNREF.
>>>
>>> Then I'll add back `async_unmap_in_progress` because resource can be
>>> both mapped/unmapped on unref, and we'll need flag to know whether async
>>> unmapping has been finished to do the final unmapping of the resource.
>>
>> Such a situation should be already handled since unmapping in progress
>> blocks all commands (not just VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB but
>> literally all, including VIRTIO_GPU_CMD_RESOURCE_UNREF).
> 
> The async unmapping consists of 3 parts:
> 
> 1. begin async unmapping with memory_region_del_subregion() and suspend
> 2. wait for res->mr to be freed and resume
> 3. finish the unmapping with final virgl_renderer_resource_unmap()
> 
> Parts 1 and 3 are handled by  virtio_gpu_virgl_async_unmap_resource_blob()
> 
> 
> The VIRTIO_GPU_CMD_RESOURCE_UNMAP_BLOB is different because we know that
> blob is mapped in the first place. Hence we can safely perform the part
> 3, assuming that parts 1/2 has been completed.
> 
> In case of VIRTIO_GPU_CMD_RESOURCE_UNREF, blob can be unmapped in the
> first place and we can't do the part 3 because it will error out for
> unmapped resource since parts 1/2 were not performed.
> 

VIRTIO_GPU_CMD_RESOURCE_UNREF should also call 
virtio_gpu_virgl_async_unmap_resource_blob(). I guess that's the 
original intention of having a function for this instead of inlining the 
content of this function to virgl_cmd_resource_unmap_blob().

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Dmitry Osipenko]

On 5/15/24 20:04, Akihiko Odaki wrote:
>>
> 
> VIRTIO_GPU_CMD_RESOURCE_UNREF should also call
> virtio_gpu_virgl_async_unmap_resource_blob(). I guess that's the
> original intention of having a function for this instead of inlining the
> content of this function to virgl_cmd_resource_unmap_blob().

Correct, previous patchset versions unmapped resource on unref.

In v11 I dropped unmapping from unref to avoid adding additional
`async_unmap_in_progress` flag because normally map/unmap will be
balanced by guest anyways.

The virtio-gpu spec doesn't tell that resource have to be implicitly
unmapped on unref. In a case of Linux guest, it actually will be a bug
to unref a mapped resource because guest will continue to map and use
the destroyed resource.

-- 
Best regards,
Dmitry

Re: [PATCH v11 08/10] virtio-gpu: Handle resource blob commands

[Akihiko Odaki]

On 2024/05/16 2:15, Dmitry Osipenko wrote:
> On 5/15/24 20:04, Akihiko Odaki wrote:
>>>
>>
>> VIRTIO_GPU_CMD_RESOURCE_UNREF should also call
>> virtio_gpu_virgl_async_unmap_resource_blob(). I guess that's the
>> original intention of having a function for this instead of inlining the
>> content of this function to virgl_cmd_resource_unmap_blob().
> 
> Correct, previous patchset versions unmapped resource on unref.
> 
> In v11 I dropped unmapping from unref to avoid adding additional
> `async_unmap_in_progress` flag because normally map/unmap will be
> balanced by guest anyways.
> 
> The virtio-gpu spec doesn't tell that resource have to be implicitly
> unmapped on unref. In a case of Linux guest, it actually will be a bug
> to unref a mapped resource because guest will continue to map and use
> the destroyed resource.
> 

Additional `async_unmap_in_progress` flag should not be necessary as 
explained earlier.

It is a valid design not to issue UNMAP_BLOB before UNREF if the 
automatically performs the unmapping operation. A guest needs to ensure 
the blob is not mapped in a guest userspace virtual address space, but 
it does not require issuing UNMAP_BLOB, which is to unmap the blob from 
the guest physical address space.

In case of Linux, virtio_gpu_vram_free() calls virtio_gpu_cmd_unmap() to 
issue UNMAP_BLOB before UNREF, which is actually not necessary. Linux 
still needs to ensure that the blob is not mapped in a guest userspace 
virtual address space, but that is done before virtio_gpu_vram_free() 
gets called, and virtio_gpu_cmd_unmap() has nothing to do with that.

It is still a good practice for a guest to issue UNMAP_BLOB in such a 
case because the spec does not say the VMM will automatically unmap the 
blob for UNREF, and that's what Linux does. From the VMM perspective, 
it's better to perform the unmapping operation for UNREF because the 
spec does not say the guest always issue UNMAP_BLOB before UNREF.