- OpenBSC - lists.osmocom.org

OpenBSC status report / Handover
by Harald Welte 21 Dec '09

21 Dec '09

Hi! Just a "quick" status report about the progress I've made with the code over the last three days, and what is now in 'master': 1) handover is not only working on a signalling level, but also fully working with actual voice traffic This turned out to be a major headache, since we don't [yet?] have a proper RTP media gateway and so far decided to simply relay the RTP frames rather than doing proper processing. In case of handover though, the RTP source changes (different SSRC / timestamp / sequence) - plus we also get some dropped frames and have to ensure that we don't hide that fact from the receiver. In order to make it work, I added a new "MNCC RTP Proxy" mode, which is automatically selected if you a) enable the RTP Proxy mode with "-P" on the command line, and b) enable handover using the 'handover 1' command on the VTY The Handover decision (handover_decision.c) is currently still based on a single MEASUREMENT REPORT, i.e. way too volatile. This was intended for testing, where we need as many handovers as possible. I've meanwhile written a lot of additional code to implement proper measurement averaging and neighbor cell selection. This is not committed yet, but will be committed by tomorrow. 2) Support for MNCC (lcr) with RTP based BTS is in 'master' There's no need for any old branches or patches for this anymore 3) Support for EFR frames with RTP based BTS in MNCC 4) No half-rate support at 26C3. The integration with lcr is likely too much work, and I think we should try not to implement even more features that have not yet received lots of testing 5) fixing tons of compiler warnings all over the code 6) fix segfault in RRLP in case of unsuccessful paging (sorry sylvain, I fixed this independently before seeing your patch, but I actually prefer your solution, will look at it tomorrow) 7) correct System Information Type 5 and 6 on ip.access The two BTS types implement the RSL SACCH FILLing slightly different, resulting in broken neighbor cell lists on ip.access systems. Any testing of current master that you can do is very much appreciated, let's hope we can catch more bugs before we go live at 26C3. Regards, Harald -- - Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/ ============================================================================ "Privacy in residential applications is a desirable marketing option." (ETSI EN 300 175-7 Ch. A6)

1 0

The need of RTP payload type IE in CRCX and MDCX
by Sylvain Munaut 20 Dec '09

20 Dec '09

Hi, I have a nanobts unit 139 with a software that somehow only accepts GSM FR (and not EFR), unless I also send the RTP payload type IE ( RSL_IE_IPAC_RTP_PAYLOAD ) in the CRCX and MDCX messages. (and only the "RTP payload type" IE, the "RTP payload type 2" has no effect I can see).

2 2

[PATCH] ipaccess: Send RTP Payload IE for CRCX & MDCX
by Sylvain Munaut 20 Dec '09

20 Dec '09

From: Sylvain Munaut <tnt(a)246tNt.com> For GSM V1 FR, the payload type is fixed to 3 in the RFC. But for the other codecs, the payload type is dynamically assigned between 96 and 127. Here, we use a static mapping internal to OpenBSC. This patch is needed to make a rather old 139 unit (with sw version 120a002_v149b42d0) work with something else than FR codec. I also tested this patch on a newer 139 (with sw version 120a352_v267b22d0) to make sure it didn't add a regression. More testing with newer EDGE units should be done by whoever has some of theses. Signed-off-by: Sylvain Munaut <tnt(a)246tNt.com> --- openbsc/include/openbsc/gsm_data.h | 1 + openbsc/include/openbsc/rtp_proxy.h | 6 ++++ openbsc/src/abis_rsl.c | 53 ++++++++++++++++++++++++++++++++-- openbsc/src/rtp_proxy.c | 3 -- 4 files changed, 56 insertions(+), 7 deletions(-) diff --git a/openbsc/include/openbsc/gsm_data.h b/openbsc/include/openbsc/gsm_data.h index bc1dbe6..16d6d94 100644 --- a/openbsc/include/openbsc/gsm_data.h +++ b/openbsc/include/openbsc/gsm_data.h @@ -204,6 +204,7 @@ struct gsm_lchan { u_int16_t bound_port; u_int16_t connect_port; u_int16_t conn_id; + u_int8_t rtp_payload; u_int8_t rtp_payload2; u_int8_t speech_mode; struct rtp_socket *rtp_socket; diff --git a/openbsc/include/openbsc/rtp_proxy.h b/openbsc/include/openbsc/rtp_proxy.h index a3e4596..07cf5a2 100644 --- a/openbsc/include/openbsc/rtp_proxy.h +++ b/openbsc/include/openbsc/rtp_proxy.h @@ -28,6 +28,12 @@ #include <openbsc/linuxlist.h> #include <openbsc/select.h> +#define RTP_PT_GSM_FULL 3 +#define RTP_PT_GSM_HALF 96 +#define RTP_PT_GSM_EFR 97 +#define RTP_PT_AMR_FULL 98 +#define RTP_PT_AMR_HALF 99 + enum rtp_rx_action { RTP_NONE, RTP_PROXY, diff --git a/openbsc/src/abis_rsl.c b/openbsc/src/abis_rsl.c index 2d55806..f4c42b8 100644 --- a/openbsc/src/abis_rsl.c +++ b/openbsc/src/abis_rsl.c @@ -39,6 +39,7 @@ #include <openbsc/paging.h> #include <openbsc/signal.h> #include <openbsc/meas_rep.h> +#include <openbsc/rtp_proxy.h> #define RSL_ALLOC_SIZE 1024 #define RSL_ALLOC_HEADROOM 128 @@ -1496,6 +1497,44 @@ static u_int8_t ipa_smod_s_for_lchan(struct gsm_lchan *lchan) return 0; } +static u_int8_t ipa_rtp_pt_for_lchan(struct gsm_lchan *lchan) +{ + switch (lchan->tch_mode) { + case GSM48_CMODE_SPEECH_V1: + switch (lchan->type) { + case GSM_LCHAN_TCH_F: + return RTP_PT_GSM_FULL; + case GSM_LCHAN_TCH_H: + return RTP_PT_GSM_HALF; + default: + break; + } + case GSM48_CMODE_SPEECH_EFR: + switch (lchan->type) { + case GSM_LCHAN_TCH_F: + return RTP_PT_GSM_EFR; + /* there's no half-rate EFR */ + default: + break; + } + case GSM48_CMODE_SPEECH_AMR: + switch (lchan->type) { + case GSM_LCHAN_TCH_F: + return RTP_PT_AMR_FULL; + case GSM_LCHAN_TCH_H: + return RTP_PT_AMR_HALF; + default: + break; + } + default: + break; + } + LOGP(DRSL, LOGL_ERROR, "Cannot determine ip.access rtp payload type for " + "tch_mode == 0x%02x\n & lchan_type == %d", + lchan->tch_mode, lchan->type); + return 0; +} + /* ip.access specific RSL extensions */ static void ipac_parse_rtp(struct gsm_lchan *lchan, struct tlv_parsed *tv) { @@ -1562,11 +1601,14 @@ int rsl_ipacc_crcx(struct gsm_lchan *lchan) /* 0x1- == receive-only, 0x-1 == EFR codec */ lchan->abis_ip.speech_mode = 0x10 | ipa_smod_s_for_lchan(lchan); + lchan->abis_ip.rtp_payload = ipa_rtp_pt_for_lchan(lchan); msgb_tv_put(msg, RSL_IE_IPAC_SPEECH_MODE, lchan->abis_ip.speech_mode); + msgb_tv_put(msg, RSL_IE_IPAC_RTP_PAYLOAD, lchan->abis_ip.rtp_payload); DEBUGP(DRSL, "channel=%s chan_nr=0x%02x IPAC_BIND " - "speech_mode=0x%02x\n", gsm_ts_name(lchan->ts), - dh->chan_nr, lchan->abis_ip.speech_mode); + "RTP_PAYLOAD=%d speech_mode=0x%02x\n", + gsm_ts_name(lchan->ts), dh->chan_nr, + lchan->abis_ip.rtp_payload, lchan->abis_ip.speech_mode); msg->trx = lchan->ts->trx; @@ -1593,12 +1635,14 @@ int rsl_ipacc_mdcx(struct gsm_lchan *lchan, u_int32_t ip, u_int16_t port, /* 0x0- == both directions, 0x-1 == EFR codec */ lchan->abis_ip.speech_mode = 0x00 | ipa_smod_s_for_lchan(lchan); + lchan->abis_ip.rtp_payload = ipa_rtp_pt_for_lchan(lchan); ia.s_addr = htonl(ip); DEBUGP(DRSL, "channel=%s chan_nr=0x%02x IPAC_MDCX " - "IP=%s PORT=%d RTP_PAYLOAD2=%d CONN_ID=%d speech_mode=0x%02x\n", + "IP=%s PORT=%d RTP_PAYLOAD=%d RTP_PAYLOAD2=%d CONN_ID=%d speech_mode=0x%02x\n", gsm_ts_name(lchan->ts), dh->chan_nr, inet_ntoa(ia), port, - rtp_payload2, lchan->abis_ip.conn_id, lchan->abis_ip.speech_mode); + lchan->abis_ip.rtp_payload, rtp_payload2, + lchan->abis_ip.conn_id, lchan->abis_ip.speech_mode); msgb_tv16_put(msg, RSL_IE_IPAC_CONN_ID, lchan->abis_ip.conn_id); msgb_v_put(msg, RSL_IE_IPAC_REMOTE_IP); @@ -1606,6 +1650,7 @@ int rsl_ipacc_mdcx(struct gsm_lchan *lchan, u_int32_t ip, u_int16_t port, *att_ip = ia.s_addr; msgb_tv16_put(msg, RSL_IE_IPAC_REMOTE_PORT, port); msgb_tv_put(msg, RSL_IE_IPAC_SPEECH_MODE, lchan->abis_ip.speech_mode); + msgb_tv_put(msg, RSL_IE_IPAC_RTP_PAYLOAD, lchan->abis_ip.rtp_payload); if (rtp_payload2) msgb_tv_put(msg, RSL_IE_IPAC_RTP_PAYLOAD2, rtp_payload2); diff --git a/openbsc/src/rtp_proxy.c b/openbsc/src/rtp_proxy.c index 5513497..6896755 100644 --- a/openbsc/src/rtp_proxy.c +++ b/openbsc/src/rtp_proxy.c @@ -81,9 +81,6 @@ struct rtp_x_hdr { #define RTP_VERSION 2 -#define RTP_PT_GSM_FULL 3 -#define RTP_PT_GSM_EFR 97 - /* decode an rtp frame and create a new buffer with payload */ static int rtp_decode(struct msgb *msg, u_int32_t callref, struct msgb **data) { -- 1.6.5.1

1 0

fix of lcr_rtp and systeminfo
by Andreas.Eversberg 20 Dec '09

20 Dec '09

hi, here are two patches (attached): the lcr_rtp.fix will make the current lcr_rtp branch compile. also it works with LCR. the systeminfo.fix contains a pointer fix and a question why there is a value cleared again. (don't apply it, just apply the line with the pointer fix) regards, andreas

1 0

[PATCH 0/7] Preliminary fixes
by Sylvain Munaut 19 Dec '09

19 Dec '09

Here's a small series with some minor fixes I found when looking at multiple SMS in a single RR connection. For the MO case, it actually now works as described in the spec in the no-error case. But what can happen is that after the second CM SERVICE REQUEST, we may never receive the last CP-ACK of the previous transaction (and thus leak a transaction). The spec says than when receiving a CP-DATA after a CM SERVICE REQUEST we should consider we also implicitely received the last CP-ACK of the previous transaction ... that's not implemented yet. For the MT case, it currently makes several RR connection which is quite bad ... to be fixed. Sylvain

3 9

payable freelance consulting on openbsc ?
by Viptel ApS - Kasper H. Hansen 18 Dec '09

18 Dec '09

Is there anyone well knowing about the signaling part in openbsc project there might be interested in the solving of a few specific tasks against payment? the results will be posted in the openbsc project also. Please contact me. Med venlig hilsen / Best regards Kasper Hessellund Hansen Denmark Khh at viptel.dk

1 0

payable freelance consulting on openbsc ?
by Viptel ApS - Kasper H. Hansen 18 Dec '09

18 Dec '09

Is there anyone well knowing about the signaling part in openbsc project there might be interested in the solving of a few specific tasks against payment? the results will be posted in the openbsc project also. Please contact me. Med venlig hilsen / Best regards Kasper Hessellund Hansen Denmark

1 0

LCR / RTP interface
by Harald Welte 17 Dec '09

17 Dec '09

Hi Andreas + List, I have now studied the new "RTP capable" lcr integration in detail. My sincere apologies once again for the long delay that this has been taking me. I really like the interface (like its predecessor). The major restriction at the moment seems to be its limitation to support only GSM Version 1 Full Rate. After doing some testing with the current 'lcr_rtp' branch in git, as well as the attempt to integrate handover support with the lcr/mncc interface, I am considering to add support for other codec types, too. GSM EFR (only exists in full-rate) is interesting for everyone, as almost all phones support it, and it renders significantly higher quality. GSM V1 halfrate is important for BS-11 users who want increased capacity, since this BTS does not support AMR-halfrate. GSM V1 full-rate is important for ip.access, since it is the only half-rate codec it supports, and thus the only way to increase capacity for this BTS. GSM AMR full-rate is probably of the least interest, but still nice to have. For the 26C3, I would love to run in a EFR-fullrate and AMR-halfrate configuration, depending on the phone's capabilities. Regards, Harald -- - Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/ ============================================================================ "Privacy in residential applications is a desirable marketing option." (ETSI EN 300 175-7 Ch. A6)

2 2

First successful handover
by Harald Welte 17 Dec '09

17 Dec '09

Hi! I've just committed the current state of the handover code to git. What is working so far: * receiving + parsing the measurement reports from both MS and BTS * performing the actual handover process, i.e. allocating / activating the new logical channel on the new BTS, waiting for the ACK, sending handover command to the MS, waiting for HO complete, HO fail or timeout, ... * a minimalistic handover decision making. without any averaging: If we have a cell that's >= 3dB better than the current one, try a handover. What's missing: * switching the actual traffic channel contents (i.e. re-configuring the RTP streams in the ip.access case, or reconfiguring the TRAU mux in E1 case) * better handover algorithm implementation including averaging I'm on a business trip tomorrow, but the code is already seeming to work quite fine, and it looks like: <0400> abis_rsl.c:980 MEASUREMENT RESULT NR=13 RXL-FULL-ul=-47dBm RXL-SUB-ul=-47dBm RXQ-FULL-ul=0 RXQ-SUB-ul=0 BS_POWER=0 L1_MS_PWR= 10dBm L1_FPC=0 L1_TA=0 RXL-FULL-dl=-61dBm RXL-SUB-dl=-56dBm RXQ-FULL-dl=6 RXQ-SUB-dl=4 NUM_NEIGH=1 <0400> abis_rsl.c:1010 ARFCN=868 BSIC=63 => -70 dBm <80000> handover_decision.c:61 process meas res: No better cell <0400> abis_rsl.c:980 MEASUREMENT RESULT NR=14 RXL-FULL-ul=-47dBm RXL-SUB-ul=-47dBm RXQ-FULL-ul=0 RXQ-SUB-ul=0 BS_POWER=0 L1_MS_PWR= 10dBm L1_FPC=0 L1_TA=0 RXL-FULL-dl=-87dBm RXL-SUB-dl=-83dBm RXQ-FULL-dl=6 RXQ-SUB-dl=4 NUM_NEIGH=1 <0400> abis_rsl.c:1010 ARFCN=868 BSIC=63 => -69 dBm <80000> handover_decision.c:61 process meas res: Cell on ARFCN 868 is better, starting handover <80000> handover_logic.c:92 (old_lchan on BTS 0, new BTS 1): <0010> abis_rsl.c:1107 channel=(bts=1,trx=0,ts=2) chan_nr=0x0a CHANNEL ACTIVATE ACK <80000> handover_logic.c:151 handover activate ack, send HO Command <0002> gsm_04_08_utils.c:144 (bts 0 trx 0 ts 2 pd 06) Sending 0x2b to MS. <0010> abis_rsl.c:1107 channel=(bts=1,trx=0,ts=2) chan_nr=0x0a <0010> abis_rsl.c:1082 HANDOVER DETECT access delay = 0 <0010> abis_rsl.c:1107 channel=(bts=1,trx=0,ts=2) chan_nr=0x0a <0010> abis_rsl.c:1082 HANDOVER DETECT access delay = 0 <0001> abis_rsl.c:1396 channel=(bts=1,trx=0,ts=2) chan_nr=0x0a sapi=0 ESTABLISH INDICATION <0001> abis_rsl.c:1396 channel=(bts=1,trx=0,ts=2) chan_nr=0x0a sapi=0 DATA INDICATION <0008> gsm_04_08.c:1595 HANDOVER COMPLETE cause = Normal event <0002> handover_logic.c:203 lchan (bts=0,trx=0,ts=2,ch=0) decreases usage to: 0 Regards, Harald -- - Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/ ============================================================================ "Privacy in residential applications is a desirable marketing option." (ETSI EN 300 175-7 Ch. A6)

1 0

Update on GSM network at 26C3
by Harald Welte 16 Dec '09

16 Dec '09

Dear All, as you will know, we have applied for an experimental GSM license for the 26C3 (http://events.ccc.de/congress/2009/) Last week, the license was granted. However, only two ARFCN's have been granted (I applied for 8). As it turns out, this was a communication problem. Today they have acknowledged a third ARFCN. So we can have at least one single-TRX BTS on each floor. On wednesday they will determine whether we can have more than three, or if we'll have to live with three ARFCN only. If we get 6, it will be one dual-TRX BTS for each floor. If we get even more, the surplus ones can be used for random experiments. TODO items before 26C3 (for the next couple of days): * finish handover implementation (critical) * do some more testing with LCR integration (Andreas did it, but I also want to have more hands-on experience with it) * introduce log levels and/or per-subscriber tracing functionality to "see the signal among all the noise" while debugging problems in an otherwise busy network * ensure SMS implementation is more robust than at HAR * log all measurement reports to database for later analysis of handover performance. We could also get them from pcap files, so not strictly neccessary. Optional: * finish AMR-halfrate implementation, as this will increase our capacity. This includes dynamically selecting the codec that is used for each call. I don't think we'll manage finishing this, especially with the LCR integration. Standalone might be possible - but that is useless for 26C3 -- - Harald Welte <laforge(a)gnumonks.org> http://laforge.gnumonks.org/ ============================================================================ "Privacy in residential applications is a desirable marketing option." (ETSI EN 300 175-7 Ch. A6)

4 9