Re: [PATCH RFC v3 07/10] tty: serial: Add Cortina UART driver

[Fil Dunsky <filipp.dunsky@gmail.com>]

On 12. 10. 26, 9:13, Jiri Slaby wrote:
> On 12. 10. 26, 9:10, sashiko-bot@kernel.org wrote:
>>> +static void cortina_uart_set_termios(struct uart_port *port,
>>> +				     struct ktermios *termios,
>>> +				     const struct ktermios *old)
>>> +{
>> [ ... ]
>>> +	uart_port_lock_irqsave(port, &flags);
>
> Seeing this, why not to use guard()?

No good reason. v4 uses guard()/scoped_guard() for the port lock in
startup, shutdown and set_termios, the sysrq-aware
guard(uart_port_lock_check_sysrq_irqsave) in the interrupt handler, and
guard(mutex) for the port table in probe and remove. console_write()
keeps the explicit lock/trylock pair for the oops case.

v4 also addresses the RX points sashiko raised in this mail: the driver
now sets read_status_mask/ignore_status_mask from INPCK/IGNPAR, reports
parity and framing errors as TTY_PARITY/TTY_FRAME, and inserts the
overrun marker after the characters drained from the FIFO. The hardware
latches these errors only in the interrupt status register (RXDAT has
just the data and a valid bit), so the error is charged to the first
character read in that interrupt.

Thanks,
Fil