Hi,
Ague Mill wrote (20 May 2012 17:55:03 GMT) :
> I have noticed that there is some seconds in the shutdown sequence
> that might be wasted by properly shutting down services that we
> don't care much about.
I don't object to this work being done, but I'd like to make sure we
carefully balance what looks to me like a pretty tiny gain (a few
seconds in the non-emergency shutdown sequence) against the need to
prepare, test and maintain the hacks that would implement the removal
of a bunch of shutdown scripts. Arguably, the needed hack is trivial,
so we can probably take the risk: if this becomes a maintenance mess
in the future, nothing prevents us from removing the hacks.
> Here is the current init scripts that are currently called (same
> scripts and order for shutdown and reboot sequences) in Tails 0.11
> and an analysis of what might be removed:
This analysis looks correct to me.
> * memlockd
> Stop memlockd. Memwipe files will not kept in memory after that.
> Better let sendsigs kill it, or even, add its PID to
> /var/run/sendsigs.omit. **Remove**
Something has to be done in this area to fix what looks like a serious
bug, independently of the optimization of the shutdown sequence speed.
The sendsigs.omit way looks right: we really don't want these files
unmapped from memory until tails-kexec has had a chance to do its
work. Are you taking care of that one?
> Any advice on how to implement the removal of those extra scripts?
> A hook with a bunch of `rm` commands? It looks like `update-rc.d` really
> wants to have shutdown scripts in place...
`rm' looks good enough.
Cheers,
--
intrigeri
| GnuPG key @ https://gaffer.ptitcanardnoir.org/intrigeri/intrigeri.asc
| OTR fingerprint @ https://gaffer.ptitcanardnoir.org/intrigeri/otr.asc
