[systemd-devel] [PATCH 2/5] udev: builtin - use RUN rather than IMPORT where it makes sense

[Kay Sievers]

On Mon, Mar 18, 2013 at 4:04 PM, Tom Gundersen <teg at jklm.no> wrote:
> On Mon, Mar 18, 2013 at 3:31 PM, Kay Sievers <kay at vrfy.org> wrote:
>> On Mon, Mar 18, 2013 at 3:24 PM, Tom Gundersen <teg at jklm.no> wrote:
>>> On Mon, Mar 18, 2013 at 3:01 PM, Kay Sievers <kay at vrfy.org> wrote:
>>
>>>> IMPORT is used to be called immediately and not be queued up to the
>>>> end of the rules handling, so that later rules could assume that this
>>>> rule has finished. PROGRAM would do the same as IMPORT and run
>>>> immediately, RUN always only happens at the very end, regardless where
>>>> it shows up in the sequence.
>>>>
>>>> Not sure if that matter in any interesting way, but it is surely a
>>>> different behaviour, and we should check that.
>>>
>>> Good point. I suppose it might be possible that later rules depend on
>>> modules/firmware being loaded, so RUN is probably the wrong choice.
>>> Any reason the ppdev module was loaded using RUN rather than IMPORT?
>>
>> Hmm, there is probably nothing that replies on that stuff. We should
>> use one of the versions everywhere. But which one? :)
>
> Well, was there a reason why the modules were loaded in
> 80-drivers.rules and not 99-drivers.rules?

We tried not to claim to be the last with the default set, 80 was
considered high enough that time to be *late* in the sequence. Now
systemd added a 99, so that rules does not apply anymore, I guess. :)

> If not, I guess we could
> rename it an use RUN, otherwise I suppose we'd have to stick with
> IMPORT... Anyone else have an opinion?

I would not expect any real problems; the modprobe causes new events
which matter, but should not change anything really for the device
that caused the modprobe.

Turn it around, if you think it's better, we will find out what happens. :)

Kay