behavier of halt/reboot

[walter harms]

Denys Vlasenko wrote:
> On Friday 01 August 2008 11:30, walter harms wrote:
>> hi list,
>> we use sysvinit and found some strange behaver of busybox on our system.
>>
>> reboot -> causes reboot but init seems not to be informed and services does not get shutdown
>> halt   -> no reaction
>>
>> The sysvinit commands work as expected.
>>
>> I guess this happens because of
>>
>>   if (ENABLE_INIT && !(flags & 4)) {
>>
>> the author expected the busybox init to be enabled.
>> That is wrong in this case.
> 
> Fully agree, will remove "ENABLE_INIT && " there.
> 
>> even with ENABLE_INIT i still have strange problems.
> 
> 
> Please:
> 
> 1. Describe the problems.
> 
> 2. Post strace logs of busybox's halt/reboot and sysvinit's halt/reboot:
> 
> strace -o halt.log -tt halt
> 
> (My personal view continues to be that halts/reboots by signalling init
> is an idiotic design. killall5 -TERM; sleep N; killall5 -KILL
> is a logical Unix way to do it. But then, early Unix gave us plenty of
> badly designed things (sendmail, bind,...).
> 
> http://busybox.net/~vda/init_vs_runsv.html
> 
> But I do want to make busybox halt/reboot to work "compatibly".
> Who knows, maybe *I'm* stupid and don't see the higher reason why
> to use them).


Hello Denys,
i am not aware of a working strace for the cris architecture. Unfortunately i can only
continue early in 4 days since i will be away.
I am fully aware of the alternatives for sysvinit and we are on the way to remove it
but i will take some more time to get familiar with the other systems.
(NTL i like the runlevel stuff for debuging :) ).

i hope this helps to improve busybox.
re,
 wh