[systemd-devel] /usr on separate file system

[Kay Sievers]

On Fri, Feb 25, 2011 at 20:12, Bill Nottingham <notting at redhat.com> wrote:
> Kay Sievers (kay.sievers at vrfy.org) said:
>> >> > Well, it hasn't been working correctly in ages. It's really not new
>> >> > policy we came up with here. It's just a warning to the user that setups
>> >> > like this will break. End of story.
>> >>
>> >> Is this flagged for the Fedora 15 release notes?
>> >
>> > Speaking as a non-systemd maintainer, IMO:
>> >
>> > - Issues that arise with separate /usr should likely be fixed
>>
>> D-Bus has its config there, we rely on D-Bus, so this would be the
>> first thing to fix. Unless this is done, the warning is really useful.
>> It's more like a "taint" flag that tells "you are on your own here"
>> than anything else.
>
> This only matters if systemd relies on dbus, or a dbus-activated service,
> to bring up local-fs.target. If it doesn't, it's not an issue.

That's right today, but the use of D-Bus is growing. And I'm not even
sure that we currently handle the D-Bus daemon startup with an empty
/usr properly. And there should really no rule like: "You can't use a
service requiring D-Bus in basic.target" -- that would just be the
wrong message.

The most prominent /usr failure in udev land is unconfigured audio and
3G network cards. And the usual response is: "I don't need that.". :)

While all that might be true, an empty /usr on bootup is just
something that nobody wants to fix, and nobody wants to debug. I'm
seeing that for many years now, and I finally gave up on it. We should
just make it clear that this isn't properly supported, unless someone
is willing to fix all the issues, which I think will never happen.

Kay