Re: [PATCH] tty: tty_buffer: add interface to queue work on specific wq and cpu

[Greg KH <gregkh@linuxfoundation.org>]

On Tue, Oct 21, 2025 at 12:21:10PM +0800, Xin Zhao wrote:
> On the embedded platform, certain critical data, such as IMU data, is
> transmitted through UART. We have customized our own 8250 driver, but
> we still need to reuse the kernel's TTY layer. The tty_flip_buffer_push
> interface in the TTY layer uses system_unbound_wq to handle the flipping
> of the TTY buffer. Although the unbound workqueue can create new threads
> on demand and wake up the kworker thread on an idle CPU, the priority of
> the kworker thread itself is not high. Even if the CPU running this work
> was idle just a moment ago, it may be preempted by real-time tasks or
> other high-priority tasks.
> In our system, the processing interval for each frame of IMU data
> transmitted via UART can experience significant jitter due to this issue.
> Instead of the expected 10 to 15 ms frame processing interval, we see
> spikes up to 30 to 35 ms. Moreover, in just one or two hours, there can
> be 2 to 3 occurrences of such high jitter, which is quite frequent. This
> jitter exceeds the software's tolerable limit of 20 ms.
> Adding the tty_flip_buffer_push_wq interface would allow our 8250 driver
> code to push work to a specified CPU. Since CPU tasks, especially real-time
> tasks, often have a certain distribution pattern in the system, we found
> that by pushing to a specified CPU with a WQ_HIGHPRI workqueue, we can
> stably eliminate the jitter and ensure that the frame processing interval
> remains between 10 and 15 ms. Furthermore, if we implement an RT workqueue
> based on this, using the tty_flip_buffer_push_wq interface to push to the
> RT workqueue(FIFO 1) we can further stabilize the processing interval,
> significantly reducing the occurrences of 14-15 ms intervals and increasing
> the frequency of 11-12 ms intervals.
> 
> Signed-off-by: Xin Zhao <jackzxcui1989@163.com>
> ---
>  drivers/tty/tty_buffer.c | 23 +++++++++++++++++++++++
>  include/linux/tty_flip.h |  3 +++
>  2 files changed, 26 insertions(+)
> 
> diff --git a/drivers/tty/tty_buffer.c b/drivers/tty/tty_buffer.c
> index 67271fc0b..112a650e2 100644
> --- a/drivers/tty/tty_buffer.c
> +++ b/drivers/tty/tty_buffer.c
> @@ -534,6 +534,29 @@ void tty_flip_buffer_push(struct tty_port *port)
>  }
>  EXPORT_SYMBOL(tty_flip_buffer_push);
>  
> +/**
> + * tty_flip_buffer_push		-	push terminal buffers
> + * @port: tty port to push
> + * @wq: workqueue on which to queue work
> + * @cpu: cpu on which to queue work
> + *
> + * Queue a push of the terminal flip buffers to the line discipline. Can be
> + * called from IRQ/atomic context.
> + *
> + * In the event of the queue being busy for flipping the work will be held off
> + * and retried later.
> + */
> +void tty_flip_buffer_push_wq(struct tty_port *port,
> +			     struct workqueue_struct *wq,
> +			     int cpu)
> +{
> +	struct tty_bufhead *buf = &port->buf;
> +
> +	tty_flip_buffer_commit(buf->tail);
> +	queue_work_on(cpu, wq, &buf->work);
> +}
> +EXPORT_SYMBOL(tty_flip_buffer_push_wq);

For obvious reasons, we can not accept new kernel apis without any
in-kernel user.  Please submit this as part of a patch series that adds
the new user.

thanks,

greg k-h