virgl: native fence fd support

Following the support for fences on the virtio driver add support
for native fence on virgl. This was somewhat based on the freedeno one.

Signed-off-by: Gustavo Padovan <[email protected]>
Signed-off-by: Robert Foss <[email protected]>
Reviewed-by: Emil Velikov <[email protected]>

7 files changed, 166 insertions, 16 deletions

diff --git a/src/gallium/drivers/virgl/virgl_context.c b/src/gallium/drivers/virgl/virgl_context.c
index 96932c473d8..9be7775abd3 100644
--- a/src/gallium/drivers/virgl/virgl_context.c
+++ b/src/gallium/drivers/virgl/virgl_context.c
@@ -21,6 +21,7 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+#include <libsync.h>
 #include "pipe/p_shader_tokens.h"
 
 #include "pipe/p_context.h"
@@ -709,13 +710,20 @@ static void virgl_draw_vbo(struct pipe_context *ctx,
 
 }
 
-static void virgl_flush_eq(struct virgl_context *ctx, void *closure)
+static void virgl_flush_eq(struct virgl_context *ctx, void *closure,
+			   struct pipe_fence_handle **fence)
 {
    struct virgl_screen *rs = virgl_screen(ctx->base.screen);
+   int out_fence_fd = -1;
 
    /* send the buffer to the remote side for decoding */
    ctx->num_transfers = ctx->num_draws = 0;
-   rs->vws->submit_cmd(rs->vws, ctx->cbuf);
+
+   rs->vws->submit_cmd(rs->vws, ctx->cbuf, ctx->cbuf->in_fence_fd,
+                       ctx->cbuf->needs_out_fence_fd ? &out_fence_fd : NULL);
+
+   if (fence)
+      *fence = rs->vws->cs_create_fence(rs->vws, out_fence_fd);
 
    virgl_encoder_set_sub_ctx(ctx, ctx->hw_sub_ctx_id);
 
@@ -728,11 +736,10 @@ static void virgl_flush_from_st(struct pipe_context *ctx,
                                enum pipe_flush_flags flags)
 {
    struct virgl_context *vctx = virgl_context(ctx);
-   struct virgl_screen *rs = virgl_screen(ctx->screen);
    struct virgl_buffer *buf, *tmp;
 
-   if (fence)
-      *fence = rs->vws->cs_create_fence(rs->vws);
+   if (flags & PIPE_FLUSH_FENCE_FD)
+       vctx->cbuf->needs_out_fence_fd = true;
 
    LIST_FOR_EACH_ENTRY_SAFE(buf, tmp, &vctx->to_flush_bufs, flush_list) {
       struct pipe_resource *res = &buf->base.u.b;
@@ -742,7 +749,13 @@ static void virgl_flush_from_st(struct pipe_context *ctx,
       pipe_resource_reference(&res, NULL);
 
    }
-   virgl_flush_eq(vctx, vctx);
+   virgl_flush_eq(vctx, vctx, fence);
+
+   if (vctx->cbuf->in_fence_fd != -1) {
+      close(vctx->cbuf->in_fence_fd);
+      vctx->cbuf->in_fence_fd = -1;
+   }
+   vctx->cbuf->needs_out_fence_fd = false;
 }
 
 static struct pipe_sampler_view *virgl_create_sampler_view(struct pipe_context *ctx,
@@ -1016,6 +1029,23 @@ static void virgl_set_shader_buffers(struct pipe_context *ctx,
    virgl_encode_set_shader_buffers(vctx, shader, start_slot, count, buffers);
 }
 
+static void virgl_create_fence_fd(struct pipe_context *ctx,
+			        struct pipe_fence_handle **fence, int fd)
+{
+   struct virgl_screen *rs = virgl_screen(ctx->screen);
+
+   *fence = rs->vws->cs_create_fence(rs->vws, fd);
+}
+
+static void virgl_fence_server_sync(struct pipe_context *ctx,
+			            struct pipe_fence_handle *fence)
+{
+   struct virgl_context *vctx = virgl_context(ctx);
+   struct virgl_screen *rs = virgl_screen(ctx->screen);
+
+   rs->vws->fence_server_sync(rs->vws, vctx->cbuf, fence);
+}
+
 static void virgl_set_shader_images(struct pipe_context *ctx,
                                     enum pipe_shader_type shader,
                                     unsigned start_slot, unsigned count,
@@ -1108,7 +1138,7 @@ virgl_context_destroy( struct pipe_context *ctx )
    vctx->framebuffer.zsbuf = NULL;
    vctx->framebuffer.nr_cbufs = 0;
    virgl_encoder_destroy_sub_ctx(vctx, vctx->hw_sub_ctx_id);
-   virgl_flush_eq(vctx, vctx);
+   virgl_flush_eq(vctx, vctx, NULL);
 
    rs->vws->cmd_buf_destroy(vctx->cbuf);
    if (vctx->uploader)
@@ -1244,6 +1274,8 @@ struct pipe_context *virgl_context_create(struct pipe_screen *pscreen,
    vctx->base.resource_copy_region = virgl_resource_copy_region;
    vctx->base.flush_resource = virgl_flush_resource;
    vctx->base.blit =  virgl_blit;
+   vctx->base.create_fence_fd = virgl_create_fence_fd;
+   vctx->base.fence_server_sync = virgl_fence_server_sync;
 
    vctx->base.set_shader_buffers = virgl_set_shader_buffers;
    vctx->base.set_hw_atomic_buffers = virgl_set_hw_atomic_buffers;
diff --git a/src/gallium/drivers/virgl/virgl_screen.c b/src/gallium/drivers/virgl/virgl_screen.c
index e71883b06f1..8addbea9e4c 100644
--- a/src/gallium/drivers/virgl/virgl_screen.c
+++ b/src/gallium/drivers/virgl/virgl_screen.c
@@ -340,7 +340,7 @@ virgl_get_param(struct pipe_screen *screen, enum pipe_cap param)
    case PIPE_CAP_VIDEO_MEMORY:
       return 0;
    case PIPE_CAP_NATIVE_FENCE_FD:
-      return 0;
+      return !!vscreen->vws->supports_fences;
    default:
       return u_pipe_screen_get_param_defaults(screen, param);
    }
@@ -721,6 +721,15 @@ static boolean virgl_fence_finish(struct pipe_screen *screen,
    return vws->fence_wait(vws, fence, timeout);
 }
 
+static int virgl_fence_get_fd(struct pipe_screen *screen,
+            struct pipe_fence_handle *fence)
+{
+   struct virgl_screen *vscreen = virgl_screen(screen);
+   struct virgl_winsys *vws = vscreen->vws;
+
+   return vws->fence_get_fd(vws, fence);
+}
+
 static uint64_t
 virgl_get_timestamp(struct pipe_screen *_screen)
 {
@@ -765,6 +774,7 @@ virgl_create_screen(struct virgl_winsys *vws)
    screen->base.fence_reference = virgl_fence_reference;
    //screen->base.fence_signalled = virgl_fence_signalled;
    screen->base.fence_finish = virgl_fence_finish;
+   screen->base.fence_get_fd = virgl_fence_get_fd;
 
    virgl_init_screen_resource_functions(&screen->base);
 
diff --git a/src/gallium/drivers/virgl/virgl_winsys.h b/src/gallium/drivers/virgl/virgl_winsys.h
index 0e6cb7953f6..d13212c5e84 100644
--- a/src/gallium/drivers/virgl/virgl_winsys.h
+++ b/src/gallium/drivers/virgl/virgl_winsys.h
@@ -40,10 +40,13 @@ struct virgl_drm_caps {
 struct virgl_cmd_buf {
    unsigned cdw;
    uint32_t *buf;
+   int in_fence_fd;
+   bool needs_out_fence_fd;
 };
 
 struct virgl_winsys {
    unsigned pci_id;
+   int supports_fences; /* In/Out fences are supported */
 
    void (*destroy)(struct virgl_winsys *vws);
 
@@ -83,7 +86,8 @@ struct virgl_winsys {
    void (*cmd_buf_destroy)(struct virgl_cmd_buf *buf);
 
    void (*emit_res)(struct virgl_winsys *vws, struct virgl_cmd_buf *buf, struct virgl_hw_res *res, boolean write_buffer);
-   int (*submit_cmd)(struct virgl_winsys *vws, struct virgl_cmd_buf *buf);
+   int (*submit_cmd)(struct virgl_winsys *vws, struct virgl_cmd_buf *buf,
+                     int32_t in_fence_fd, int32_t *out_fence_fd);
 
    boolean (*res_is_referenced)(struct virgl_winsys *vws,
                                 struct virgl_cmd_buf *buf,
@@ -92,7 +96,7 @@ struct virgl_winsys {
    int (*get_caps)(struct virgl_winsys *vws, struct virgl_drm_caps *caps);
 
    /* fence */
-   struct pipe_fence_handle *(*cs_create_fence)(struct virgl_winsys *vws);
+   struct pipe_fence_handle *(*cs_create_fence)(struct virgl_winsys *vws, int fd);
    bool (*fence_wait)(struct virgl_winsys *vws,
                       struct pipe_fence_handle *fence,
                       uint64_t timeout);
@@ -107,6 +111,12 @@ struct virgl_winsys {
                              unsigned level, unsigned layer,
                              void *winsys_drawable_handle,
                              struct pipe_box *sub_box);
+   void (*fence_server_sync)(struct virgl_winsys *vws,
+                             struct virgl_cmd_buf *cbuf,
+                             struct pipe_fence_handle *fence);
+
+   int (*fence_get_fd)(struct virgl_winsys *vws,
+                       struct pipe_fence_handle *fence);
 };
 
 /* this defaults all newer caps,
diff --git a/src/gallium/winsys/virgl/drm/virgl_drm_winsys.c b/src/gallium/winsys/virgl/drm/virgl_drm_winsys.c
index 98e0e99f661..5b604044598 100644
--- a/src/gallium/winsys/virgl/drm/virgl_drm_winsys.c
+++ b/src/gallium/winsys/virgl/drm/virgl_drm_winsys.c
@@ -38,11 +38,17 @@
 #include "virgl/virgl_public.h"
 
 #include <xf86drm.h>
+#include <libsync.h>
 #include "virtgpu_drm.h"
 
 #include "virgl_drm_winsys.h"
 #include "virgl_drm_public.h"
 
+
+#define VIRGL_DRM_VERSION(major, minor) ((major) << 16 | (minor))
+#define VIRGL_DRM_VERSION_FENCE_FD      VIRGL_DRM_VERSION(1, 0)
+
+
 static inline boolean can_cache_resource(struct virgl_hw_res *res)
 {
    return res->cacheable == TRUE;
@@ -70,6 +76,9 @@ static void virgl_hw_res_destroy(struct virgl_drm_winsys *qdws,
       if (res->ptr)
          os_munmap(res->ptr, res->size);
 
+      if (res->fence_fd != -1)
+         close(res->fence_fd);
+
       memset(&args, 0, sizeof(args));
       args.handle = res->bo_handle;
       drmIoctl(qdws->fd, DRM_IOCTL_GEM_CLOSE, &args);
@@ -222,6 +231,7 @@ virgl_drm_winsys_resource_create(struct virgl_winsys *qws,
    res->stride = stride;
    pipe_reference_init(&res->reference, 1);
    res->num_cs_references = 0;
+   res->fence_fd = -1;
    return res;
 }
 
@@ -457,6 +467,7 @@ virgl_drm_winsys_resource_create_handle(struct virgl_winsys *qws,
    res->stride = info_arg.stride;
    pipe_reference_init(&res->reference, 1);
    res->num_cs_references = 0;
+   res->fence_fd = -1;
 
    util_hash_table_set(qdws->bo_handles, (void *)(uintptr_t)handle, res);
 
@@ -577,6 +588,7 @@ static struct virgl_cmd_buf *virgl_drm_cmd_buf_create(struct virgl_winsys *qws)
    }
 
    cbuf->base.buf = cbuf->buf;
+   cbuf->base.in_fence_fd = -1;
    return &cbuf->base;
 }
 
@@ -687,7 +699,8 @@ static boolean virgl_drm_res_is_ref(struct virgl_winsys *qws,
 }
 
 static int virgl_drm_winsys_submit_cmd(struct virgl_winsys *qws,
-                                       struct virgl_cmd_buf *_cbuf)
+                                       struct virgl_cmd_buf *_cbuf,
+                                       int in_fence_fd, int *out_fence_fd)
 {
    struct virgl_drm_winsys *qdws = virgl_drm_winsys(qws);
    struct virgl_drm_cmd_buf *cbuf = virgl_drm_cmd_buf(_cbuf);
@@ -702,12 +715,24 @@ static int virgl_drm_winsys_submit_cmd(struct virgl_winsys *qws,
    eb.size = cbuf->base.cdw * 4;
    eb.num_bo_handles = cbuf->cres;
    eb.bo_handles = (unsigned long)(void *)cbuf->res_hlist;
+   eb.fence_fd = -1;
+
+   if (in_fence_fd != -1) {
+       eb.flags |= VIRTGPU_EXECBUF_FENCE_FD_IN;
+       eb.fence_fd = in_fence_fd;
+   }
+
+   if (out_fence_fd != NULL)
+       eb.flags |= VIRTGPU_EXECBUF_FENCE_FD_OUT;
 
    ret = drmIoctl(qdws->fd, DRM_IOCTL_VIRTGPU_EXECBUFFER, &eb);
    if (ret == -1)
       fprintf(stderr,"got error from kernel - expect bad rendering %d\n", errno);
    cbuf->base.cdw = 0;
 
+   if (out_fence_fd != NULL)
+      *out_fence_fd = eb.fence_fd;
+
    virgl_drm_release_all_res(qdws, cbuf);
 
    memset(cbuf->is_handle_added, 0, sizeof(cbuf->is_handle_added));
@@ -759,7 +784,7 @@ static int handle_compare(void *key1, void *key2)
 }
 
 static struct pipe_fence_handle *
-virgl_cs_create_fence(struct virgl_winsys *vws)
+virgl_cs_create_fence(struct virgl_winsys *vws, int fd)
 {
    struct virgl_hw_res *res;
 
@@ -769,6 +794,7 @@ virgl_cs_create_fence(struct virgl_winsys *vws)
                                                 VIRGL_BIND_CUSTOM,
                                                 8, 1, 1, 0, 0, 0, 8);
 
+   res->fence_fd = fd;
    return (struct pipe_fence_handle *)res;
 }
 
@@ -793,6 +819,12 @@ static bool virgl_fence_wait(struct virgl_winsys *vws,
       return TRUE;
    }
    virgl_drm_resource_wait(vws, res);
+
+   if (res->fence_fd != -1) {
+      int ret = sync_wait(res->fence_fd, timeout / 1000000);
+      return ret == 0;
+   }
+
    return TRUE;
 }
 
@@ -805,11 +837,51 @@ static void virgl_fence_reference(struct virgl_winsys *vws,
                                 virgl_hw_res(src));
 }
 
+static void virgl_fence_server_sync(struct virgl_winsys *vws,
+		                    struct virgl_cmd_buf *cbuf,
+                                    struct pipe_fence_handle *fence)
+{
+   struct virgl_hw_res *hw_res = virgl_hw_res(fence);
+
+   /* if not an external fence, then nothing more to do without preemption: */
+   if (hw_res->fence_fd == -1)
+      return;
+
+   sync_accumulate("virgl", &cbuf->in_fence_fd, hw_res->fence_fd);
+}
+
+static int virgl_fence_get_fd(struct virgl_winsys *vws,
+                               struct pipe_fence_handle *fence)
+{
+        struct virgl_hw_res *hw_res = virgl_hw_res(fence);
+
+        return dup(hw_res->fence_fd);
+}
+
+static int virgl_drm_get_version(int fd)
+{
+	int ret;
+	drmVersionPtr version;
+
+	version = drmGetVersion(fd);
+
+	if (!version)
+		ret = -EFAULT;
+	else if (version->version_major != 0)
+		ret = -EINVAL;
+	else
+		ret = version->version_minor;
+
+	drmFreeVersion(version);
+
+	return ret;
+}
 
 static struct virgl_winsys *
 virgl_drm_winsys_create(int drmFD)
 {
    struct virgl_drm_winsys *qdws;
+   int drm_version;
    int ret;
    int gl = 0;
    struct drm_virtgpu_getparam getparam = {0};
@@ -820,6 +892,10 @@ virgl_drm_winsys_create(int drmFD)
    if (ret < 0 || !gl)
       return NULL;
 
+   drm_version = virgl_drm_get_version(qdws->fd);
+   if (drm_version < 0)
+      return NULL;
+
    qdws = CALLOC_STRUCT(virgl_drm_winsys);
    if (!qdws)
       return NULL;
@@ -851,9 +927,13 @@ virgl_drm_winsys_create(int drmFD)
    qdws->base.cs_create_fence = virgl_cs_create_fence;
    qdws->base.fence_wait = virgl_fence_wait;
    qdws->base.fence_reference = virgl_fence_reference;
+   qdws->base.fence_server_sync = virgl_fence_server_sync;
+   qdws->base.fence_get_fd = virgl_fence_get_fd;
+   qdws->base.supports_fences =  drm_version >= VIRGL_DRM_VERSION_FENCE_FD;
 
    qdws->base.get_caps = virgl_drm_get_caps;
 
+
    uint32_t value = 0;
    getparam.param = VIRTGPU_PARAM_CAPSET_QUERY_FIX;
    getparam.value = (uint64_t)(uintptr_t)&value;
diff --git a/src/gallium/winsys/virgl/drm/virgl_drm_winsys.h b/src/gallium/winsys/virgl/drm/virgl_drm_winsys.h
index b28e7127ca0..4316a74977d 100644
--- a/src/gallium/winsys/virgl/drm/virgl_drm_winsys.h
+++ b/src/gallium/winsys/virgl/drm/virgl_drm_winsys.h
@@ -50,12 +50,14 @@ struct virgl_hw_res {
    int64_t start, end;
    boolean flinked;
    uint32_t flink;
+   int fence_fd;
 };
 
 struct virgl_drm_winsys
 {
    struct virgl_winsys base;
    int fd;
+   int drm_version;
    struct list_head delayed;
    int num_delayed;
    unsigned usecs;
diff --git a/src/gallium/winsys/virgl/drm/virtgpu_drm.h b/src/gallium/winsys/virgl/drm/virtgpu_drm.h
index 8596febe9fd..4be22aa077c 100644
--- a/src/gallium/winsys/virgl/drm/virtgpu_drm.h
+++ b/src/gallium/winsys/virgl/drm/virtgpu_drm.h
@@ -44,6 +44,16 @@
 #define DRM_VIRTGPU_WAIT     0x08
 #define DRM_VIRTGPU_GET_CAPS  0x09
 
+/*
+ * virtgpu execbuffer flags
+ */
+#define VIRTGPU_EXECBUF_FENCE_FD_IN	0x01
+#define VIRTGPU_EXECBUF_FENCE_FD_OUT	0x02
+#define VIRTGPU_EXECBUF_FLAGS	(\
+		VIRTGPU_EXECBUF_FENCE_FD_IN |\
+		VIRTGPU_EXECBUF_FENCE_FD_OUT |\
+		0)
+
 struct drm_virtgpu_map {
 	uint64_t offset; /* use for mmap system call */
 	uint32_t handle;
@@ -56,7 +66,7 @@ struct drm_virtgpu_execbuffer {
 	uint64_t command; /* void* */
 	uint64_t bo_handles;
 	uint32_t num_bo_handles;
-	uint32_t pad;
+	int32_t fence_fd;
 };
 
 #define VIRTGPU_PARAM_3D_FEATURES 1 /* do we have 3D features in the hw */
@@ -130,7 +140,7 @@ struct drm_virtgpu_get_caps {
 	DRM_IOWR(DRM_COMMAND_BASE + DRM_VIRTGPU_MAP, struct drm_virtgpu_map)
 
 #define DRM_IOCTL_VIRTGPU_EXECBUFFER \
-	DRM_IOW(DRM_COMMAND_BASE + DRM_VIRTGPU_EXECBUFFER,\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VIRTGPU_EXECBUFFER,\
 		struct drm_virtgpu_execbuffer)
 
 #define DRM_IOCTL_VIRTGPU_GETPARAM \
diff --git a/src/gallium/winsys/virgl/vtest/virgl_vtest_winsys.c b/src/gallium/winsys/virgl/vtest/virgl_vtest_winsys.c
index f44d4d74ff1..c1c88151ad3 100644
--- a/src/gallium/winsys/virgl/vtest/virgl_vtest_winsys.c
+++ b/src/gallium/winsys/virgl/vtest/virgl_vtest_winsys.c
@@ -501,7 +501,8 @@ static void virgl_vtest_add_res(struct virgl_vtest_winsys *vtws,
 }
 
 static int virgl_vtest_winsys_submit_cmd(struct virgl_winsys *vws,
-                                         struct virgl_cmd_buf *_cbuf)
+                                         struct virgl_cmd_buf *_cbuf,
+                                         int in_fence_fd, int *out_fence_fd)
 {
    struct virgl_vtest_winsys *vtws = virgl_vtest_winsys(vws);
    struct virgl_vtest_cmd_buf *cbuf = virgl_vtest_cmd_buf(_cbuf);
@@ -510,6 +511,9 @@ static int virgl_vtest_winsys_submit_cmd(struct virgl_winsys *vws,
    if (cbuf->base.cdw == 0)
       return 0;
 
+   assert(in_fence_fd == -1);
+   assert(out_fence_fd == NULL);
+
    ret = virgl_vtest_submit_cmd(vtws, cbuf);
 
    virgl_vtest_release_all_res(vtws, cbuf);
@@ -552,7 +556,7 @@ static int virgl_vtest_get_caps(struct virgl_winsys *vws,
 }
 
 static struct pipe_fence_handle *
-virgl_cs_create_fence(struct virgl_winsys *vws)
+virgl_cs_create_fence(struct virgl_winsys *vws, int fd)
 {
    struct virgl_hw_res *res;
 
@@ -694,8 +698,10 @@ virgl_vtest_winsys_wrap(struct sw_winsys *sws)
    vtws->base.cs_create_fence = virgl_cs_create_fence;
    vtws->base.fence_wait = virgl_fence_wait;
    vtws->base.fence_reference = virgl_fence_reference;
+   vtws->base.supports_fences =  0;
 
    vtws->base.flush_frontbuffer = virgl_vtest_flush_frontbuffer;
 
+
    return &vtws->base;
 }