On Sun, Oct 16, 2016 at 12:56 AM, Guilhem Moulin <guil...@guilhem.org> wrote:
> On Sat, 15 Oct 2016 at 08:13:27 +0700, Theppitak Karoonboonyanan wrote:
>> Reassigning the bug to udev to hear its maintainer's opinion.
> I think this is a initscripts bug, not a udev bug.  Both the
> ‘cryptdisks-early’ and ‘cryptdisks’ LSB init scripts have had a
> “Should-Stop: udev” header.  Shouldn't this be enough to have udevd
> still running during cryptdisks shutdown?  If that's the case then the
> culprit is sendsigs as it violates shutdown dependencies.

Given its function to "killall5", how can sendsigs discriminate
between processes to be killed and those to be preserved?
How can it know which daemons were started by a named service
(udev in this case) so it can skip sending SIGTERM to them?

I think what sendsigs has provided for this purpose is the
/run/sendsigs.omit.d/ directory, as documented in

And to add a process to the omission list, a most appropriate
place to do so is udev's init script itself, where the daemon was
started. That's why I reassign the bug to it.

Theppitak Karoonboonyanan

Pkg-systemd-maintainers mailing list

Reply via email to