diff --git a/libavfilter/vf_hwupload.c b/libavfilter/vf_hwupload.c
index 6aafac4d4d..3142aabb17 100644
--- a/libavfilter/vf_hwupload.c
+++ b/libavfilter/vf_hwupload.c
@@ -222,6 +222,15 @@ static av_cold void hwupload_uninit(AVFilterContext *avctx)
 {
     HWUploadContext *ctx = avctx->priv;
 
+    /**
+     * Ensure that all outstanding asynchronous uploads are properly completed.
+     * This is to avoid leaking the context, because nobody else will clean up
+     * after the hwfc while there are still pending asynchronous transfers
+     * referencing it.
+     */
+    if (ctx->hwframes_ref)
+        av_hwframe_transfer_wait_all(ctx->hwframes_ref, 0);
+
     av_buffer_unref(&ctx->hwframes_ref);
     av_buffer_unref(&ctx->hwdevice_ref);
 }
diff --git a/libavutil/hwcontext.c b/libavutil/hwcontext.c
index f06d49c45c..05c77ac7be 100644
--- a/libavutil/hwcontext.c
+++ b/libavutil/hwcontext.c
@@ -492,6 +492,15 @@ int av_hwframe_transfer_data(AVFrame *dst, const AVFrame *src, int flags)
     return 0;
 }
 
+int av_hwframe_transfer_wait_all(AVBufferRef *hwframe_ref, int flags)
+{
+    FFHWFramesContext *ctx = (FFHWFramesContext*) hwframe_ref->data;
+    if (!ctx->hw_type->frames_sync)
+        return 0;
+
+    return ctx->hw_type->frames_sync(&ctx->p);
+}
+
 int av_hwframe_get_buffer(AVBufferRef *hwframe_ref, AVFrame *frame, int flags)
 {
     FFHWFramesContext *ctxi = (FFHWFramesContext*)hwframe_ref->data;
diff --git a/libavutil/hwcontext.h b/libavutil/hwcontext.h
index 96042ba197..7a0d826e61 100644
--- a/libavutil/hwcontext.h
+++ b/libavutil/hwcontext.h
@@ -401,6 +401,16 @@ int av_hwframe_get_buffer(AVBufferRef *hwframe_ctx, AVFrame *frame, int flags);
  */
 int av_hwframe_transfer_data(AVFrame *dst, const AVFrame *src, int flags);
 
+/**
+ * Explicitly wait for all preceding (possibly asynchronous) transfers to be
+ * completed. No-op for synchronous hardware device types.
+ *
+ * @param hwframe_ctx a reference to an AVHWFramesContext
+ * @param flags currently unused, should be set to zero
+ * @return 0 on success, a negative AVERROR error code on failure.
+ */
+int av_hwframe_transfer_wait_all(AVBufferRef *hwframe_ctx, int flags);
+
 enum AVHWFrameTransferDirection {
     /**
      * Transfer the data from the queried hw frame.
diff --git a/libavutil/hwcontext_internal.h b/libavutil/hwcontext_internal.h
index db23579c9e..e1b8c12424 100644
--- a/libavutil/hwcontext_internal.h
+++ b/libavutil/hwcontext_internal.h
@@ -69,6 +69,7 @@ typedef struct HWContextType {
 
     int              (*frames_init)(AVHWFramesContext *ctx);
     void             (*frames_uninit)(AVHWFramesContext *ctx);
+    int              (*frames_sync)(AVHWFramesContext *ctx);
 
     int              (*frames_get_buffer)(AVHWFramesContext *ctx, AVFrame *frame);
     int              (*transfer_get_formats)(AVHWFramesContext *ctx,
diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index fec22a4f46..e1b3a6f346 100644
--- a/libavutil/hwcontext_vulkan.c
+++ b/libavutil/hwcontext_vulkan.c
@@ -2716,6 +2716,18 @@ static void vulkan_frames_uninit(AVHWFramesContext *hwfc)
     av_buffer_pool_uninit(&fp->tmp);
 }
 
+static int vulkan_frames_sync(AVHWFramesContext *hwfc)
+{
+    VulkanDevicePriv *p = hwfc->device_ctx->hwctx;
+    VulkanFramesPriv *fp = hwfc->hwctx;
+
+    ff_vk_exec_pool_wait(&p->vkctx, &fp->compute_exec);
+    ff_vk_exec_pool_wait(&p->vkctx, &fp->upload_exec);
+    ff_vk_exec_pool_wait(&p->vkctx, &fp->download_exec);
+
+    return 0;
+}
+
 static int vulkan_frames_init(AVHWFramesContext *hwfc)
 {
     int err;
@@ -4282,6 +4294,8 @@ static int vulkan_transfer_frame(AVHWFramesContext *hwfc,
         ff_vk_exec_wait(&p->vkctx, exec);
         if (!host_mapped)
             err = copy_buffer_data(hwfc, bufs[0], swf, region, planes, 0);
+    } else if (upload) {
+        ff_vk_exec_wait(&p->vkctx, exec);
     }
 
 end:
@@ -4478,6 +4492,7 @@ const HWContextType ff_hwcontext_type_vulkan = {
     .frames_init            = vulkan_frames_init,
     .frames_get_buffer      = vulkan_get_buffer,
     .frames_uninit          = vulkan_frames_uninit,
+    .frames_sync            = vulkan_frames_sync,
 
     .transfer_get_formats   = vulkan_transfer_get_formats,
     .transfer_data_to       = vulkan_transfer_data_to,
diff --git a/libavutil/vulkan.c b/libavutil/vulkan.c
index 0571ae64e1..4586d0edee 100644
--- a/libavutil/vulkan.c
+++ b/libavutil/vulkan.c
@@ -230,6 +230,16 @@ AVVulkanDeviceQueueFamily *ff_vk_qf_find(FFVulkanContext *s,
     return NULL;
 }
 
+void ff_vk_exec_pool_wait(FFVulkanContext *s, FFVkExecPool *pool)
+{
+    FFVulkanFunctions *vk = &s->vkfn;
+    for (int i = 0; i < pool->pool_size; i++) {
+        FFVkExecContext *e = &pool->contexts[i];
+        if (e->had_submission)
+            ff_vk_exec_wait(s, e);
+    }
+}
+
 void ff_vk_exec_pool_free(FFVulkanContext *s, FFVkExecPool *pool)
 {
     FFVulkanFunctions *vk = &s->vkfn;
diff --git a/libavutil/vulkan.h b/libavutil/vulkan.h
index 8690c13b3d..c7a194ad61 100644
--- a/libavutil/vulkan.h
+++ b/libavutil/vulkan.h
@@ -408,6 +408,11 @@ int ff_vk_exec_pool_init(FFVulkanContext *s, AVVulkanDeviceQueueFamily *qf,
                          const void *query_create_pnext);
 void ff_vk_exec_pool_free(FFVulkanContext *s, FFVkExecPool *pool);
 
+/**
+ * Wait until all prior submissions have finished.
+ */
+void ff_vk_exec_pool_wait(FFVulkanContext *s, FFVkExecPool *pool);
+
 /**
  * Retrieve an execution pool. Threadsafe.
  */