Re: [PATCH] fix book E watchdog to take WDIOC_SETTIMEOUT arg in seconds

[Josh Boyer <jwboyer@linux.vnet.ibm.com>]

On Thu, Aug 27, 2009 at 01:14:58PM +0200, Wim Van Sebroeck wrote:
>Hi Chris,
>
>> The WDIOC_SETTIMEOUT argument is supposed to be a "seconds" value.
>> However, the book E wdt currently treats it as a "period" which is
>> interpreted in a board-specific way.
>> 
>> This patch allows the user to pass in a "seconds" value and the driver
>> will set the smallest timeout that is at least as large as specified
>> by the user.  It's been tested on e500 hardware and works as
>> expected.
>> 
>> The patch only modifies the CONFIG_FSL_BOOKE case, the CONFIG_4xx case
>> is left unmodified as I don't have any hardware to test it on.
>> 
>> Signed-off-by: Chris Friesen <cfriesen@nortel.com>
>
>Added with some small changes to keep checkpatch happy. (removed trailing spaces + changed sizeof(struct watchdog_info) to sizeof(ident) and changed some spaces in tabs).
>
>Now we only need someone that can look at the CONFIG_4xx cases still :-)

It seems the FSL watchdog is much more flexible than the one found in 4xx
cores.  On 4xx, you basically have 4 static choices that represent specific
times determined by the clock frequency.  I'm concerned that the lack of
granularity here will result in less than desirable behavior.

For example, with a 400MHz clock you would get choices of roughly:

5.2 ms
83.9 ms
1.34 s
21.47 s

Personally, I consider the first two options basically unusable.  Considering
the second two, if a user were to say "Set the timeout for 2 seconds" they
would then get a timeout of 21 seconds with the framework Chris' patch has
set up.  That doesn't really seem to be ideal to me.

josh