On 07/16/2017 02:10 PM, Matthijs Kooijman wrote:
I just tried Connman and oFono on a Debian stretch system for the first
time, hoping to get a more robust solution for keeping a cellular and/or
ethernet connection up. It seems the duo is quite capable of doing what
I need, but there were a few bumps in setting this up, where I think
things could be improved.
Some of these issues might be more connman issues, but they're mostly on
the cross-over area between the two, so I'll post them here for now.
In somewhat chronological order:
- I could not find any documentation on how to get a cellular
connection up. Plugging in a dongle made the "cellular" technology
appear in connman, but no services, with no indication of what needed
to happen next.
No services means either an incomplete modem driver, missing TUN module
or most likely that the APN hasn't been setup properly (e.g. automatic
provisioning has failed).
- When provisioning in oFono fails, a single gprs context is
without an APN. Setting the APN manually made the profile show up in
connman and (eventually) work.
oFono creates a single sentinel context when automatic provisioning
fails. This is an indication to the provisioning UI that the user
should be helped/asked/guided in providing a valid context configuration
(e.g. APN, etc)
It looks like Debian doesn't ship such a UI.
It might be better to cange connman to create a service for a
without an APN, and show an error when trying to connect it, instead
of completely hiding the context.
See above. Ideally the desktop environment should detect that the
provisioning failed and offer the user to configure the APNs with some
sort of helpful Wizard.
- There does not seem to be any official commandline interface to
to figure out what is happening. There are "test" python scripts in
the source repo, but these are not shipped in the Debian package
(reported here: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=868377
Correct. A C-based command line UI (similar to e.g. connmanctl) is
something that is on our TODO list but we haven't gotten around to it yet.
- When mobile-broadband-provider-info provision finds multiple
for the operator (that is, multiple apns all with the same usage
type), provisioning fails, disregarding *all* apns found. I believe
this even happens when there is just one type duplicate (e.g. if
multiple mms profiles are available and just one internet profile,
the internet one is not used). Not entirely sure, though.
MBPI database could be improved to make the chances of automatic
provisioning succeeding higher. However, a provisioning UI is still needed.
- When multiple profiles match, it would not make sense for oFono
just pick an arbitrary one to use. Ideally, it would just create
a context for each, and let the user select the right context. For
this to actually work, the user must be able to distinguish them, and
see all the relevant details for profile/context to allow choosing
the right one.
This was solved by using a provisioning UI. We don't believe that
pushing selection of the APN onto the user (even a power user) is the
I tried simply passing "allow_duplicates=TRUE" to ..., which does
indeed create all contexts, but seems to show only the first two as
aservices in connman (not sure why just these two), and there is no
way to really distinguish them (the service name is just the
currently connected network name).
ConnMan only manages internet contexts.
- When actually connecting from connman (or activating a context
the ofono test/activate-context script), the reported errors are very
unclear. I only got a "Not implemented" error, which is very
non-descriptive and, looking at the code, is returned in quite some
different places. There's also nothing in the log to indicate what
The activation error in my case was due to a missing tun kernel
module, I submitted a separate patch for that.
Missing tun support in the kernel is reported via ofono_error inside
drivers/atmodem/gprs-context.c. So you most definitely should be seeing
something in the log about it.