On 5/20/25 01:20, Yiwei Zhang wrote: >> Is it possible that virglrenderer will ever write a context fence in >> virgl_renderer_create_fence()? Do we really need this check? > I assume you were referring to whether a context fence can be written > in virgl_write_fence(). Yes, the fenceq contains both the global > fences and context specific fences. Without the check, > virgl_write_fence() will signal any context fence with a fence id > smaller than the latest signaled virgl global fence.
Indeed, I was comparing with the async-cb variant and forgot that async one uses separate fence list. -- Best regards, Dmitry