Function wait_for_writes 

pub fn wait_for_writes(
    device: &Device,
    queue: &Queue,
    plane: BorrowedFd<'_>,
) -> Result<Option<WriteFence>, ImportError>

Make the next submission on queue wait on the GPU for the writes still pending on the single-plane dma-buf plane (the producer’s implicit fence), so a sample of the buffer in it or any later submission reads what the producer wrote. No thread waits: the fence is exported as a sync file and imported as a semaphore the submission waits for.

Ok(None) when nothing is pending, or the kernel cannot export the fence (the caller then relies on having watched the buffer become readable).

Errors

ImportError::NotVulkan, or ImportError::Vulkan when the device lacks VK_KHR_external_semaphore_fd or the semaphore cannot be made.