new kernel ... / meaning of stable-tracking

Sean McNeil sean at mcneil.com
Sun Sep 28 00:31:16 CEST 2008


Werner Almesberger wrote:
> Michael 'Mickey' Lauer wrote:
>   
>> I would not protest on removing the system runlevel 6 trigger,
>>     
>
> Good, that's already half the messiness :-)
>
>   
>> however I
>> would certainly veto on removing the 8 seconds emergency shutdown, since
>> that's for cases where userland is entirely unresponsive (no daemon to
>> catch that state expected).
>>     
>
> Hmm, how about solving the problem that the daemon is unresponsive
> instead ?
>
> If you separate it from the rest of the framework, the code will be
> simple, small, and won't break easily because of some unrelated
> changes. If you mlockall, you're immune to out of memory conditions
> or a file system failure. If you raise the scheduling priority high
> enough, you'll win even against fork bombs.

Like a watchdog process?




More information about the openmoko-kernel mailing list