Cellular Network Infrastructure » cni-legacy » OpenBSC

attaching pcap file with example.

don't be surprised if wireshark doesn't fully decode the messages. The non-standard RSL extensions require a patch (see my commit 89c71c635f0e85a66de3a176b0a0df2970d8058b in wireshark.git on osmocom.org, branch laforge/om2000-201610). However, a full decode of the RSL CHAN ACT message is not required to solve this bug.

With my default wireshark I basically see only "malformed packet", so I guess I'd need the extensions.

I'm not familiar with SACCH, but if you say that SACCH only exists for CS channels, then the
obvious answer is that we can't deact SACCH on a PDCH channel.

The conclusion would be that rsl_rx_conn_fail() calling rsl_rf_chan_release_err() in abis_rsl.c
must not pass SACCH_DEACTIVATE unconditionally. This patch could be all that's needed:

diff --git a/openbsc/src/libbsc/abis_rsl.c b/openbsc/src/libbsc/abis_rsl.c
index b9132a5..eb33b3a 100644
--- a/openbsc/src/libbsc/abis_rsl.c
+++ b/openbsc/src/libbsc/abis_rsl.c
@@ -859,7 +859,9 @@ static int rsl_rf_chan_release_err(struct gsm_lchan *lchan)
 {
        if (lchan->state != LCHAN_S_ACTIVE)
                return 0;
-       return rsl_rf_chan_release(lchan, 1, SACCH_DEACTIVATE);
+       return rsl_rf_chan_release(lchan, 1,
+                                  ts_is_pdch(lchan->ts) ?
+                                  SACCH_NONE : SACCH_DEACTIVATE);
 }

(haven't tested nor compiled it yet)

BTW...

From 08.58 8.4.4:

CONNECTION FAILURE INDICATION
This message is sent from BTS to BSC to indicate that an active connection has been broken for some reason.

That sounds like a channel was already fully established and then fails.
From a failure to switch on a PDCH lchan, I would expect an 'Channel Activation Nack' instead?
Anyway, if the RBS2000 replies with a conn failure, we have to be able to deal with it...