[PATCH v13 2/5] uart: pl011: Introduce register accessor

[Peter Hurley]

On 10/22/2015 07:36 PM, Timur Tabi wrote:
> On Fri, Sep 18, 2015 at 5:51 AM, Andre Przywara <andre.przywara at arm.com> wrote:
>>
>>>  static void pl011_putc(struct uart_port *port, int c)
>>>  {
>>> -       while (readl(port->membase + REG_FR) & UART01x_FR_TXFF)
>>> +       struct uart_amba_port *uap =
>>> +           container_of(port, struct uart_amba_port, port);
>>> +
>>> +       while (pl011_readw(uap, REG_FR) & UART01x_FR_TXFF)
>>>                 ;
>>> -       writeb(c, port->membase + REG_DR);
>>> -       while (readl(port->membase + REG_FR) & UART01x_FR_BUSY)
>>> +       pl011_writeb(uap, c, REG_DR);
>>> +       while (pl011_readw(uap, REG_FR) & UART01x_FR_BUSY)
>>>                 ;
>>>  }
>>
>> Just for the records, as this has been discussed before: pl011_putc() is
>> called by the earlycon code, before the uart_port is actually
>> initialized. So we cannot rely on the accessors, but have to use the
>> old-fashioned director accessors for this.
>>
>> Which means you cannot use that approach to get earlycon support for the
>> ZTE UART, if I get this correctly. It shouldn't be to hard to introduce
>> another earlycon type specificly for that one, copying pl011_early_write
>> and pl011_early_console_setup and changing pl011_putc into
>> zte_uart_putc. But of course this belongs into the final patch (or a
>> separate one), not in this. So I guess you just leave that function
>> unchanged in this patch.
> 
> How about something like this?  It adds the "sbsa32" option to the
> earlycon command-line parameter.
> 
> static void pl011_putc(struct uart_port *port, int c)
> {
>     while (readl(port->membase + UART01x_FR) & UART01x_FR_TXFF)
>         cpu_relax();
>     writeb(c, port->membase + UART01x_DR);
>     while (readl(port->membase + UART01x_FR) & UART01x_FR_BUSY)
>         cpu_relax();
> }
> 
> static void pl011_early_write(struct console *con, const char *s, unsigned n)
> {
>     struct earlycon_device *dev = con->data;
> 
>     uart_console_write(&dev->port, s, n, pl011_putc);
> }
> 
> static void pl011_putc_sbsa32(struct uart_port *port, int c)
> {
>     while (readl(port->membase + UART01x_FR) & UART01x_FR_TXFF)
>         cpu_relax();
>     writel(c, port->membase + UART01x_DR);
>     while (readl(port->membase + UART01x_FR) & UART01x_FR_BUSY)
>         cpu_relax();
> }
> 
> static void pl011_early_write_sbsa32(struct console *con, const char
> *s, unsigned n)
> {
>     struct earlycon_device *dev = con->data;
> 
>     uart_console_write(&dev->port, s, n, pl011_putc_sbsa32);
> }
> 
> static int __init pl011_early_console_setup(struct earlycon_device *device,
>                         const char *opt)
> {
>     if (!device->port.membase)
>         return -ENODEV;
> 
>     if (strcmp(device->options, "sbsa32"))
>         device->con->write = pl011_early_write_sbsa32;
>     else
>         device->con->write = pl011_early_write;
> 
>     return 0;
> }


For earlycon support, I'd prefer to see different OF_EARLYCON_DECLARE()
(and EARLYCON_DECLARE()) declarations with alternate setup() functions
for zte and sbsa32. For example:

static int __init zte_early_console_setup(struct earlycon_device *device,
	const char *opt)
{
	if (!device->port.membase)
		return -ENODEV;
	device->con->write = zte_early_write;
	return 0;
}

static int __init sbsa32_early_console_setup(struct earlycon_device *device,
	const char *opt)
{
	if (!device->port.membase)
		return -ENODEV;
	device->con->write = sbsa32_early_write;
	return 0;
}

EARLYCON_DECLARE(zte, zte_early_console_setup);
OF_EARLYCON_DECLARE(zte, ".......", zte_early_console_setup);
EARLYCON_DECLARE(sbsa32, sbsa32_early_console_setup);
OF_EARLYCON_DECLARE(sbsa32, "arm,sbsa-uart", sbsa32_early_console_setup);

The above assumes that the sbsa32 maintains the equivalence of
earlycon and console; ie., that 

/ {
	chosen {
		stdout-path = &uart0;
	};

	soc {
		uart0: serial at xxxxxx {
			compatible = "arm,sbsa-uart";
		};
	};
};


will start a sbsa32 earlycon and later replace that with
sbsa console.


Regards,
Peter Hurley