Re: [PATCH 6/10] watchdog: bcm63xx_wdt: Obtain watchdog clock HZ from "periph" clk

[Florian Fainelli <f.fainelli@gmail.com>]

On 23/11/15 07:02, Jonas Gorski wrote:
> Hi,
> 
> On Sun, Nov 22, 2015 at 3:07 PM, Simon Arlott <simon@fire.lp0.eu> wrote:
>> Instead of using a fixed clock HZ in the driver, obtain it from the
>> "periph" clk that the watchdog timer uses.
>>
>> Signed-off-by: Simon Arlott <simon@fire.lp0.eu>
>> ---
>>  drivers/watchdog/bcm63xx_wdt.c | 36 +++++++++++++++++++++++++++++++-----
>>  1 file changed, 31 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/watchdog/bcm63xx_wdt.c b/drivers/watchdog/bcm63xx_wdt.c
>> index 1d2a501..eb5e551 100644
>> --- a/drivers/watchdog/bcm63xx_wdt.c
>> +++ b/drivers/watchdog/bcm63xx_wdt.c
>> @@ -13,6 +13,7 @@
>>
>>  #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
>>
>> +#include <linux/clk.h>
>>  #include <linux/errno.h>
>>  #include <linux/io.h>
>>  #include <linux/kernel.h>
>> @@ -32,11 +33,13 @@
>>
>>  #define PFX KBUILD_MODNAME
>>
>> -#define WDT_HZ                 50000000                /* Fclk */
>> +#define WDT_CLK_NAME           "periph"
> 
> @Florian:
> Is this correct? The comment for the watchdog in 6358_map_part.h and
> earlier claims that the clock is 40 MHz there, but the code uses 50MHz
> - is this a bug in the comments or is it a bug taken over from the
> original broadcom code? I'm sure that the periph clock being 50 MHz
> even on the older chips is correct, else we'd have noticed that in
> serial output (where it's also used).

There are references to a Fbus2 clock in documentation, but I could not
find any actual documentation for its actual clock frequency, I would be
surprised if this chip would have diverged from the previous and future
ones and used a 40Mhz clock. 6345 started with a peripheral clock
running at 50Mhz, and that is true for all chips since then AFAICT.

I agree we would have noticed this with the UART or SPI controllers if
that was not true, so probably a code glitch here...
-- 
Florian