Re: [PATCH 01/28] serial: earlycon: allow MEM32 I/O for DT earlycon

[Peter Hurley <peter@hurleysoftware.com>]

On 11/30/2015 05:52 PM, Rob Herring wrote:
> On Mon, Nov 30, 2015 at 10:21 AM, Paul Burton <paul.burton@imgtec.com> wrote:
>> Read the reg-io-width property when earlycon is setup via device tree,
>> and set the I/O type to UPIO_MEM32 when 4 is read. This behaviour
>> matches that of the of_serial driver, and is needed for DT configured
>> earlycon on the MIPS Boston board.
>>
>> Note that this is only possible when CONFIG_LIBFDT is enabled, but
>> enabling it everywhere seems like overkill. Thus systems that need this
>> functionality should select CONFIG_LIBFDT for themselves.
> 
> libfdt is enabled if you are booting from DT, so checking this
> property should not add anything.
> 
>>
>> Signed-off-by: Paul Burton <paul.burton@imgtec.com>
>> ---
>>
>>  drivers/of/fdt.c              |  2 +-
>>  drivers/tty/serial/Makefile   |  1 +
>>  drivers/tty/serial/earlycon.c | 15 ++++++++++++++-
>>  include/linux/serial_core.h   |  2 +-
>>  4 files changed, 17 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/of/fdt.c b/drivers/of/fdt.c
>> index d243029..71c7f0d 100644
>> --- a/drivers/of/fdt.c
>> +++ b/drivers/of/fdt.c
>> @@ -833,7 +833,7 @@ static int __init early_init_dt_scan_chosen_serial(void)
>>                 if (addr == OF_BAD_ADDR)
>>                         return -ENXIO;
>>
>> -               of_setup_earlycon(addr, match->data);
>> +               of_setup_earlycon(fdt, offset, addr, match->data);
>>                 return 0;
>>         }
>>         return -ENODEV;
>> diff --git a/drivers/tty/serial/Makefile b/drivers/tty/serial/Makefile
>> index 5ab4111..1d290d6 100644
>> --- a/drivers/tty/serial/Makefile
>> +++ b/drivers/tty/serial/Makefile
>> @@ -7,6 +7,7 @@ obj-$(CONFIG_SERIAL_21285) += 21285.o
>>
>>  obj-$(CONFIG_SERIAL_EARLYCON) += earlycon.o
>>  obj-$(CONFIG_SERIAL_EARLYCON_ARM_SEMIHOST) += earlycon-arm-semihost.o
>> +CFLAGS_earlycon.o += -I$(srctree)/scripts/dtc/libfdt
> 
> This is no longer necessary.
> 
>>
>>  # These Sparc drivers have to appear before others such as 8250
>>  # which share ttySx minor node space.  Otherwise console device
>> diff --git a/drivers/tty/serial/earlycon.c b/drivers/tty/serial/earlycon.c
>> index f096360..2b936a7 100644
>> --- a/drivers/tty/serial/earlycon.c
>> +++ b/drivers/tty/serial/earlycon.c
>> @@ -17,6 +17,7 @@
>>  #include <linux/kernel.h>
>>  #include <linux/init.h>
>>  #include <linux/io.h>
>> +#include <linux/libfdt.h>
>>  #include <linux/serial_core.h>
>>  #include <linux/sizes.h>
>>  #include <linux/mod_devicetable.h>
>> @@ -196,17 +197,29 @@ static int __init param_setup_earlycon(char *buf)
>>  }
>>  early_param("earlycon", param_setup_earlycon);
>>
>> -int __init of_setup_earlycon(unsigned long addr,
>> +int __init of_setup_earlycon(const void *fdt, int offset, unsigned long addr,
> 
> I would add iotype as a parameter instead, and then...
> 
>>                              int (*setup)(struct earlycon_device *, const char *))
>>  {
>>         int err;
>>         struct uart_port *port = &early_console_dev.port;
>> +       const __be32 *prop;
>>
>>         port->iotype = UPIO_MEM;
>>         port->mapbase = addr;
>>         port->uartclk = BASE_BAUD * 16;
>>         port->membase = earlycon_map(addr, SZ_4K);
>>
>> +       if (config_enabled(CONFIG_LIBFDT)) {
>> +               prop = fdt_getprop(fdt, offset, "reg-io-width", NULL);
>> +               if (prop) {
>> +                       switch (be32_to_cpup(prop)) {
>> +                       case 4:
>> +                               port->iotype = UPIO_MEM32;
>> +                               break;
>> +                       }
>> +               }
> 
> ...move this parsing into fdt.c where we parse the address.

FWIW, all of of_setup_earlycon() should only be #ifdef CONFIG_OF_EARLY_FLATTREE

Regards,
Peter Hurley