RE: [PATCH v2 tty-next 3/3] 8250: microchip: pci1xxxx: Add power management functions to quad-uart driver.

[Kumaravel.Thiagarajan]

> -----Original Message-----
> From: Ilpo Järvinen <ilpo.jarvinen@xxxxxxxxxxxxxxx>
> Sent: Monday, October 3, 2022 3:21 PM
> To: Kumaravel Thiagarajan - I21417 <Kumaravel.Thiagarajan@xxxxxxxxxxxxx>
> Subject: Re: [PATCH v2 tty-next 3/3] 8250: microchip: pci1xxxx: Add power
> management functions to quad-uart driver.
>  
> On Sat, 1 Oct 2022, Kumaravel Thiagarajan wrote:
> 
> > pci1xxxx's quad-uart function has the capability to wake up the host
> > from suspend state. Enable wakeup before entering into suspend and
> > disable wakeup on resume.
> >
> > Signed-off-by: Kumaravel Thiagarajan
> > <kumaravel.thiagarajan@xxxxxxxxxxxxx>
> > ---
> > Changes in v2:
> > - Use DEFINE_SIMPLE_DEV_PM_OPS instead of SIMPLE_DEV_PM_OPS.
> > - Use pm_sleep_ptr instead of CONFIG_PM_SLEEP.
> > - Change the return data type of pci1xxxx_port_suspend to bool from int.
> > ---
> >  drivers/tty/serial/8250/8250_pci1xxxx.c | 112
> > ++++++++++++++++++++++++
> >  1 file changed, 112 insertions(+)
> >
> > diff --git a/drivers/tty/serial/8250/8250_pci1xxxx.c
> > b/drivers/tty/serial/8250/8250_pci1xxxx.c
> > index 999e5a284266..0a0459f66177 100644
> > --- a/drivers/tty/serial/8250/8250_pci1xxxx.c
> > +++ b/drivers/tty/serial/8250/8250_pci1xxxx.c
> > @@ -352,6 +352,112 @@ static void pci1xxxx_irq_assign(struct
> pci1xxxx_8250 *priv,
> >       }
> >  }
> >
> > +static bool pci1xxxx_port_suspend(int line) {
> > +     struct uart_8250_port *up = serial8250_get_port(line);
> > +     struct uart_port *port = &up->port;
> > +     unsigned long flags;
> > +     u8 wakeup_mask;
> > +     bool ret = false;
> > +
> > +     if (port->suspended == 0 && port->dev) {
> > +             wakeup_mask = readb(up->port.membase +
> > + UART_WAKE_MASK_REG);
> > +
> > +             spin_lock_irqsave(&port->lock, flags);
> > +             port->mctrl &= ~TIOCM_OUT2;
> > +             port->ops->set_mctrl(port, port->mctrl);
> > +             spin_unlock_irqrestore(&port->lock, flags);
> > +
> > +             if ((wakeup_mask & UART_WAKE_SRCS) != UART_WAKE_SRCS)
> > +                     ret = true;
> > +     }
> > +
> > +     writeb(UART_WAKE_SRCS, port->membase + UART_WAKE_REG);
> > +
> > +     return ret;
> > +}
> > +
> > +static void pci1xxxx_port_resume(int line) {
> > +     struct uart_8250_port *up = serial8250_get_port(line);
> > +     struct uart_port *port = &up->port;
> > +     unsigned long flags;
> > +
> > +     writeb(UART_WAKE_SRCS, port->membase + UART_WAKE_REG);
> > +
> > +     if (port->suspended == 0) {
> 
> Is this check the right way around?
Yes. I think port->suspended is not set for wake-up capable ports and the code in this if block gets executed for those ports.
I will check this again.
> 
> > +             spin_lock_irqsave(&port->lock, flags);
> > +             port->mctrl |= TIOCM_OUT2;
> > +             port->ops->set_mctrl(port, port->mctrl);
> > +             spin_unlock_irqrestore(&port->lock, flags);
> > +     }
> > +}
> > +
> > +static int pci1xxxx_suspend(struct device *dev) {
> > +     struct pci1xxxx_8250 *priv = dev_get_drvdata(dev);
> > +     struct pci_dev *pcidev = to_pci_dev(dev);
> > +     unsigned int data;
> > +     void __iomem *p;
> > +     bool wakeup = false;
> > +     int i;
> > +
> > +     for (i = 0; i < priv->nr; i++) {
> > +             if (priv->line[i] >= 0) {
> > +                     serial8250_suspend_port(priv->line[i]);
> > +                     wakeup |= pci1xxxx_port_suspend(priv->line[i]);
> 
> So first serial8250_suspend_port() calls into uart_suspend_port() that sets
> port->suspended to 1, then pci1xxxx_port_suspend() checks if it's 0.
> Is this intentional?
Yes. I think port->suspended does not seem to be set for wake-up capable ports and only 
for those ports, inside pci1xxxx_port_suspend, TIOCM_OUT2 is cleared.
But I must check for the race condition as Andy had pointed out.
Please let me know if there are any questions.

Thank You.

Regards,
Kumar