Age | Commit message (Collapse) | Author | Files | Lines |
|
This patch modifies openBSC code to use msg->dst which stores the
pointer to the signalling link structure instead of the pointer to
the transceiver structure.
This patch prepares the introduction of libosmo-abis.
|
|
|
|
|
|
It is possible that MNCC sends a MNCC_LCHAN_MODIFY and
wants a channel mode that is not possible on the current
lchan, in that case a new channel is assigned. We now crash
as the osmo-nitb is not having an assignment complete handler,
add a NULL check.
|
|
I'm not sure if that is an abnormal condition or not, but it seems
that lchan state and type have to be none for the lchan to be
considered idle.
|
|
|
|
The Nokia metrosite BTS seem to keep the channels open indefinitely.
If osmo-nitb is restarted while one of the channel was still active
and tries to activate that channel again the bts would return a
CHANNEL ACTIVATE NACK with "Radio channel already activated". This
accumulated over the restarts so soon enough no more channels were
available.
This patch sends a release request to the bts so the channel
becomes available again.
|
|
This makes sure that the lchans can be used again. (state and type
are set to NONE)
|
|
The function lchan_alloc only considers lchans to be available if both
the type and state are NONE. So change show lchan to list all lchans
that are not considered available.
|
|
This will cause the remote end to read 0 bytes, which is interpreted as
if we cleanly closed the socket, making the remote end close their side
of the socket, which would lead to us closing our side of the socket,
so we should never send such a packet.
|
|
The timer callback will simply reset the lchan state to NONE in order
to prevent channels getting stuck in 'activation requested' or
'deactivation requested' states.
|
|
|
|
|
|
|
|
|
|
So far, there was no way to set the line->name field at all.
|
|
|
|
During the GSM deployment in the CCC Camp, Daniel Willmann
noticed that the LCR and the MNCC were closing the local
connection over unix sockets communication quite so often.
After some debugging, Peter Stuge noticed that openBSC was
closing the connection since write was returning 0.
Then, I suggested that it could be a malformed message with
zero length. By skipping empty messages, Peter confirmed that
the connection between the LCR and the MNCC was not closing
anymore. However, there was no voice in the calls that went
over MNCC.
After some more debugging I found that we were not building
GSM_TCHF_FRAME over MNCC appropriately in the TRAU multiplexer
code, since we forgot to msgb_put() the message.
|
|
|
|
Scenario: BTS are configured and working, then the BSC stops working
for some reason (crash or administrative stop).
If the BSC comes back to life, LAPD among other things does not know
about the previous existing TEIs. Instead of ignoring these frames,
we notify the driver that we are seeing frames with unknown TEIs, so
it can try to recover, e.g. by resending the SABM message.
|
|
(thanks to Gus Bourg)
|
|
|
|
|
|
|
|
This is one step in the direction of supporting multiple Nokia BTS
|
|
|
|
Running the entire bts_nokia_site.c through the 'Lindent' script
to match indent/coding style with remainder of project.
There are still lots of other cleanups pending, but this one is
a purely cosmetic one.
|
|
This includes the MetroSite, but also other Nokia BTS models.
|
|
this way you can actually see them...
|
|
|
|
For a system-level daemon, no protocol parser error should ever call
assert, which would take down the entire process.
|
|
DAHDI creates one device node for every E1 timeslot, starting from '1',
and keeps incrementing that number even for additional ports/cards.
Thus, we have to use the e1inp_line number multiplied by 31 as a base.
|
|
The byte ordering is a bit odd: The least significant byte is ahead of
the most significant byte, different from everything else in GSM that
seems to be big-endian.
Thanks to Seungju Kim <admin@manateeshome.com> for repoerting the bug.
|
|
|
|
This is particularly important in case of the Nokia BTS, as they seem
to drop the RF/signalling channel if they don't get proper TRAU
frames.
|
|
|
|
This fixes a bug introduced more than one year ago in commit
e38bd6caa34005816a9336f021fd17d328d5c901:
The RSL_IE_CHAN_IDENT is a TLV, but the GSM48_IE_CHANDESC_2 contained in
it, is a mere TV type IE with fixed length.
The problem specifically has caused problems on Nokia MetroSite BTS,
which apparently read the TSC out of this Layer3 IE.
|
|
|
|
The MS Radio Access Capability IE can be _very_ long in some recent
high-end mobile phones, way beyond the old 14-byte limit. We increase
our array to 52 bytes, and make sure not to overflow that buffer.
|
|
It was possible to set the LAC=0 on a subscriber that just has
done a LU because it did not respond to a paging request.
E.g. when a phone is rebooting, a SMS being delivered, the phone
is doing the LU, sub_ready_for_sm will try to send a SMS (but the
phone is not ready yet and it will timeout), then the paging code will
send us an expiration note and we might set the LAC=0 for this
subscriber.
Ideally we would be able to stop the paging request once the subscriber
is authenticated and then hand this to the SMS layer, right now the
best thing to do is to detect that we will run into this problem and
not send the SMS, not try to set the LAC=0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|