On Fri, Nov 19, 2021 at 09:43:12AM +0100, Jan Kiszka wrote:

Digging a bit deeper: One issue of the original patch was definitely
that it tried to sanitized mctrl inside serial8250_do_set_mctrl, rather
than serial8250_set_mctrl. That caused alternative set_mctrl handler to
become out of sync /wrt mctrl. Just look at omap8250_set_mctrl, how it
will work on the unsanitized value.

All omap8250_set_mctrl does is enable or disable autoRTS if autoRTS is
used.  Are you using autoRTS on this port?  That's one of several pieces
of information that I'm missing here to understand what's going on.

Thanks,

Lukas