[PATCH] tty/serial: change of_serial to use new of_property_read_u32() api

Grant Likely grant.likely at secretlab.ca
Fri Jul 1 07:01:45 EST 2011


On Thu, Jun 30, 2011 at 2:57 PM, Rob Herring <robherring2 at gmail.com> wrote:
> Grant,
>
> On 06/30/2011 02:00 PM, Grant Likely wrote:
>> Simplifies the code a bit and drops a few lines.
>>
>> Signed-off-by: Grant Likely <grant.likely at secretlab.ca>
>> ---
>>
>> I've only actually build tested this, but this shows some of the cleanup
>> achieved using the of_property_read_u32() API.  If this gets merged in the
>> v3.1 merge window then it will need to go via the devicetree/next branch.
>>
>> g.
>
> You can't give yourself bonus points. ;)
>
>>
>>  drivers/tty/serial/of_serial.c |   37 +++++++++++++++----------------------
>>  1 files changed, 15 insertions(+), 22 deletions(-)
>>
>> diff --git a/drivers/tty/serial/of_serial.c b/drivers/tty/serial/of_serial.c
>> index 36038ed..dbfbfda 100644
>> --- a/drivers/tty/serial/of_serial.c
>> +++ b/drivers/tty/serial/of_serial.c
>> @@ -32,17 +32,17 @@ static int __devinit of_platform_serial_setup(struct platform_device *ofdev,
>>  {
>>       struct resource resource;
>>       struct device_node *np = ofdev->dev.of_node;
>> -     const __be32 *clk, *spd;
>> -     const __be32 *prop;
>> -     int ret, prop_size;
>> +     u32 clk, spd, prop;
>> +     int ret;
>>
>>       memset(port, 0, sizeof *port);
>> -     spd = of_get_property(np, "current-speed", NULL);
>> -     clk = of_get_property(np, "clock-frequency", NULL);
>> -     if (!clk) {
>> +     if (of_property_read_u32(np, "clock-frequency", &clk)) {
>
> s/clk/port->uartclk/
>
> And below, then remove clk.
>
>>               dev_warn(&ofdev->dev, "no clock-frequency property set\n");
>>               return -ENODEV;
>>       }
>> +     /* If current-speed was set, then try not to change it. */
>> +     if (of_property_read_u32(np, "current-speed", &spd) == 0)
>> +             port->custom_divisor = clk / (16 * spd);
>>
>>       ret = of_address_to_resource(np, 0, &resource);
>>       if (ret) {
>> @@ -54,20 +54,17 @@ static int __devinit of_platform_serial_setup(struct platform_device *ofdev,
>>       port->mapbase = resource.start;
>>
>>       /* Check for shifted address mapping */
>> -     prop = of_get_property(np, "reg-offset", &prop_size);
>> -     if (prop && (prop_size == sizeof(u32)))
>> -             port->mapbase += be32_to_cpup(prop);
>> +     if (of_property_read_u32(np, "reg-offset", &prop) == 0)
>> +             port->mapbase += prop;
>>
>>       /* Check for registers offset within the devices address range */
>> -     prop = of_get_property(np, "reg-shift", &prop_size);
>> -     if (prop && (prop_size == sizeof(u32)))
>> -             port->regshift = be32_to_cpup(prop);
>> +     if (of_property_read_u32(np, "reg-shift", &prop) == 0)
>> +             port->regshift = prop;
>
> Can be further simplified:
>
> of_property_read_u32(np, "reg-shift", &port->regshift);

Hey Rob, thanks for the review.

I thought about that, but the new api specifically needs to be passed
a u32 pointer, not an unsigned char pointer which is what is in the
port structure.  That goes for the clk value too.  Even for unsigned
int values, i need the u32 bounce variable because it would break on
64 bit platforms.

g.


More information about the devicetree-discuss mailing list