On 10/4/23 12:01, Tony Lindgren wrote:
* Johan Hovold <johan@xxxxxxxxxx> [231004 09:14]:
On Wed, Oct 04, 2023 at 12:03:20PM +0300, Tony Lindgren wrote:
The serial port device and serdev device are siblings of the physical
serial port controller device as seen in the hierarcy printed out by
Maximilian.
Yeah, and that's precisely the broken part. Keeping the serdev
controller active is supposed to keep the serial controller active. Your
serial core rework appears to have broken just that.
Hmm OK good point, tx can currently have an extra delay if a serdev
device is active, and the serial port controller device is not active.
So we can check for active port->dev instead of &port_dev->dev though
to know when when start_tx() is safe to do as below.
I can confirm that this works as well.
Thanks,
Max
8< -----------------
diff --git a/drivers/tty/serial/serial_core.c b/drivers/tty/serial/serial_core.c
index 6207f0051f23d..defecc5b04422 100644
--- a/drivers/tty/serial/serial_core.c
+++ b/drivers/tty/serial/serial_core.c
@@ -156,7 +156,7 @@ static void __uart_start(struct uart_state *state)
* enabled, serial_port_runtime_resume() calls start_tx() again
* after enabling the device.
*/
- if (pm_runtime_active(&port_dev->dev))
+ if (!pm_runtime_enabled(port->dev) || pm_runtime_active(port->dev))
port->ops->start_tx(port);
pm_runtime_mark_last_busy(&port_dev->dev);
pm_runtime_put_autosuspend(&port_dev->dev);
