[systemd-devel] [PATCH 1/4] Adding halt binary to shutdown the system

[Lennart Poettering]

On Fri, 01.10.10 18:16, Michael Biebl (mbiebl at gmail.com) wrote:

> 
> 2010/10/1 Fabiano Fidêncio <fidencio at profusion.mobi>:
> > About fuse, Which is the problem in try to umount using umount2?
> 
> I'm not an expert regarding fuse, but say I have a partition mounted
> using ntfs-3g.
> If I kill the ntfs-3g process, the mount will go away.
> During your "kill" stage, the order of processes being killed is
> random, I guess.
> So there still might be processes accessing that ntfs partition.
> 
> It would definitely be nicer, if you kill all running processes
> (besides the ntfs-3g process), and then unmount the NTFS partition.
> The nfs case is similar.
> killall5 (at least in Debian) has an -o flag [1], and e.g. portmap or
> ntfs-3g use that mechanism to not be killed by the killall script.
> 
> As I already wrote for the LVM/mdadm/cryptsetup case, imo we need a
> mechanism how those tools can hook into the shutdown process.
> Maybe having a single binary doing all steps in on go does not offer
> the necessary flexibility.

As mentioned, Fabianos code is intended as last resort. The proper order
in which to shut down stuff should be ensured with with the usual
brefore/after dependencies.

Lennart

-- 
Lennart Poettering - Red Hat, Inc.