Re: Slight interface change on the watchdog fido

From: Poul-Henning Kamp <phk_at_phk.freebsd.dk>
Date: Sun, 10 Dec 2006 15:35:40 +0000
In message <20061210160457.W42195_at_localhost>, Nick Hibma writes:
>>>   cognet_at_freebsd.org	i80321_wdog.c (*)
>>> (*) The i80321_wdog.c cannot be disarmed. Is this correct?
>>
>> If true, then this is a poster-child for the WD_PASSIVE need, the idea
>> being that if userland says "I'll not pat the dog anymore" and the hardware
>> cannot be disabled, the kernel shoul do it.
>
>~he implementation of the WD_PASSIVE part is on my list.
>
>I don't quite agree with you on the kernel taking over though. When 
>testing watchdogs you should be able to see that you could not disarm 
>it, as you would otherwise get mysterious hard reboots. I'd rather have 
>watchdogd refuse to exit if it cannot disarm the watchdog. I'll put that 
>on my list too.

Watchdog[d](8) may not be the only program that calls the ioctl, in
many embedded apps the central application will do so itself.

It seems to me a much more intuitive behaviour if the kernel takes
over the job of patting the offending piece of hardware.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk_at_FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.
Received on Sun Dec 10 2006 - 14:35:53 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:03 UTC