>On Tue, 9 Apr 2002, Corey Minyard wrote:
>
>>Why is that too fine grained? You would just set the values from 1000
>>to 255000 instead of 1 to 255, and round up.
>>
>>I have a board that sets the time value in wierd times (like 225ms,
>>450ms, 900ms, 1800ms, 3600ms, etc.). I wouldn't be against the
>>WDIOS_TIMEINMILLI option, but milliseconds should be good enough for anyone.
>>
>
>Yet Another Brainfart. I've been having a lot of them recently.
>
>I don't feel comfortable changing the API that much in a stable kernel
>series. Also, some other boards that have very small timeout windows
>emulate a larger userspace timeout since it's quite possible that a
>process won't get scheduled every 250ms. I guess the only reason I can see
>for such a small timeout window is if one needs 99.9999% uptime and the 29
>extra seconds that the watchdog waits before kicking off is important.
>
The actual reason for a small timeout is if the system is in a
high-throughput highly available application, to get the board out of
commission as soon as possible and avoid a big delay in message handling
and/or lose as little traffic as possible. Or if the system can fail
and start misbehaving, to kill it as soon as possible to minimize the
damage. With the preemptable kernel and real-time processes, it's not
unreasonable to schedule something every 250ms.
-Corey
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/