Re: [PATCH v4 2/2] serial: amba-pl011: keep RS485 transmission running on reconfiguration

[Nicolai Buchwitz <nb@tipi-net.de>]

On 21.10.2026 14:12, sashiko-bot@kernel.org wrote:

> [...]

> [Severity: Medium]
> During a TIOCSRS485 ioctl, uart_set_rs485_config() invokes
> port->rs485_config() before assigning the new configuration to 
> port->rs485:
> 
> drivers/tty/serial/serial_core.c:uart_set_rs485_config() {
> 	...
> 	scoped_guard(uart_port_lock_irqsave, port) {
> 		ret = port->rs485_config(port, &tty->termios, &rs485);
> 		if (!ret) {
> 			port->rs485 = rs485;
> 	...
> }
> 
> In drivers/tty/serial/amba-pl011.c, pl011_rs485_config() calls
> pl011_rs485_tx_stop() -> pl011_rs485_tx_stop_now(), which reads
> port->rs485.flags rather than the new rs485->flags:
> 
> drivers/tty/serial/amba-pl011.c:pl011_rs485_tx_stop_now() {
> 	...
> 	if (port->rs485.flags & SER_RS485_RTS_AFTER_SEND)
> 		cr &= ~UART011_CR_RTS;
> 	else
> 		cr |= UART011_CR_RTS;
> 	...
> }
> 
> Doesn't this cause three regressions when configuring RS485 via 
> TIOCSRS485?
> 
> First, when RS485 is already enabled and idle (uap->rs485_tx_state == 
> OFF)
> and userspace changes SER_RS485_RTS_AFTER_SEND, 
> pl011_rs485_tx_stop_now()
> programs UART011_CR_RTS using the old port->rs485.flags instead of the 
> new
> rs485->flags.
> 
> Second, because the else-if condition in pl011_rs485_config() checks
> port->rs485.flags & SER_RS485_ENABLED instead of
> rs485->flags & SER_RS485_ENABLED, newly enabling RS485 on an open, idle 
> port
> (where pl011_tx_active(uap) is false and pl011_tx_empty(port) is true) 
> skips
> both pl011_rs485_tx_resume() and pl011_rs485_tx_stop(), leaving
> UART011_CR_TXE enabled and UART011_CR_RTS at its previous non-RS485 
> level.
> 
> Third, when RS485 is reconfigured while a transmission or break is 
> active
> (uap->rs485_tx_state != OFF), all branches in pl011_rs485_config() are
> skipped, so changes to SER_RS485_RTS_ON_SEND or SER_RS485_RX_DURING_TX 
> in
> rs485->flags are not applied to REG_CR for the ongoing transmission.
> 
> Should pl011_rs485_config() check rs485->flags & SER_RS485_ENABLED and 
> apply
> the new rs485->flags when setting the idle or active REG_CR state?

All three predate this series, the branch read port->rs485 the same way 
before.
The cover letter lists them as left for the RS485 rework, which fixes 
them by
applying the new settings in pl011_rs485_config().

IMHO that's too big for these stable fixes.

Thanks,
Nicolai