Age | Commit message (Collapse) | Author | Files | Lines |
|
Change-Id: I4ea799c5fa61f81c404e6ef1b9ac86a8faa1fb49
|
|
Original libvlr code is by Harald Welte <laforge@gnumonks.org>,
polished and tweaked by Neels Hofmeyr <nhofmeyr@sysmocom.de>.
This is a long series of development collapsed in one patch. The original
history may still be available as branch neels/vlr_orig.
TODO: This commit may be split in several smaller changes before merging to
master.
Related: OS#1592
Change-Id: I702ba504ce2de93507312c28eca8d11f09f4ee8b
|
|
In a future commit, gsm_subscriber will be replaced by vlr_subscr, and it will
not make sense to use vlr_subscr in libbsc. Thus we need a dedicated BSC
subscriber: struct bsc_subscr.
Add rf_policy arg to bsc_grace_paging_request() because the bsc_subscr will no
longer have a backpointer to gsm_network (used to be via subscr->group).
Create a separate logging filter for the new BSC subscriber. The implementation
of adjusting the filter context is added in libbsc to not introduce
bsc_subscr_get/_put() dependencies to libcommon.
During Paging Response, fetch a bsc_subscr from the mobile identity, like we do
for the gsm_subscriber. It looks like a duplication now, but will make sense
for the VLR as well as for future MSC split patches.
Naming: it was requested to not name the new struct bsc_sub, because 'sub' is
too ambiguous. At the same time it would be fine to have 'bsc_sub_' as function
prefix. Instead of struct bsc_subscriber and bsc_sub_ prefix, I decided to
match both up as struct bsc_subscr and bsc_subscr_ function prefix. It's fast
to type, relatively short, unambiguous, and the naming is consistent.
Add bsc_subscr unit test.
Related: OS#1592, OS#1594
Change-Id: Ia61cc00e8bb186b976939a4fc8f7cf9ce6aa3d8e
|
|
In libosmocore, my patch was merged to master a bit too soon. To accomodate the
request for naming that matches the general "LOG" prefix instead of "LOGGING",
a fixup was committed to libosmocore. Adjust for that.
Original patch: change-id I5c343630020f4b108099696fd96c2111614c8067
The fixup: change-id I424fe3f12ea620338902b2bb8230544bde3f1a93
Change-Id: Ib2ec5e4884aa90f48051ee2f832af557aa525991
|
|
The LCHAN and BTS filter contexts are actually never used, so drop them until
someone adds them properly.
For now use only LOGGING_{FILTER,CTX}_VLR_SUBSCR. Some of these will change to
_BSC_SUBSCR once struct bsc_subscriber is introduced, and later on, struct
gsm_subscriber will be replaced by vlr_subscriber so that the names will match.
Depends: libosmocore change-id I5c343630020f4b108099696fd96c2111614c8067
Change-Id: Ifa82f6a461ad4c0eeddb8a38fb3833460432d16b
|
|
Use GSM_MACBLOCK_LEN consistently throughout abis_rsl.c
Change-Id: I96aec02748a0be0100dee2117f124ff32d5ee3f5
|
|
Log more data related to channel allocation:
- channel type
- number of paging attempts
- timers fired
Change-Id: Ib417a9c942c17b902dd80ff555cd9da5f91bff48
|
|
GSM48_CMODE_DATA_6k0 was not properly terminated and thus resulted in a
bug.
Change-Id: I4000f06d0b49c4afb0446beddd150521c4ba3cf0
Fixes: Coverity CID 148207
|
|
This is useful particularly in case where we deactivate PDCHs
which don't have a SACCH associated. The existin code would
always attempt to deactivate a SACCH even in those cases, leading
to the BTS responsding with related error messages.
Change-Id: Iaf46782329b38ba8f3d438e6c75c2d467b852734
|
|
Ericsson has introduced a propritary format to issue the S13 BCCH
information. Normally the system info type field for SI13 would
be encoded as 0x28. Ericsson encodes that field as 0x02 and ads
a bcch mapping parameter, (IEI=F2) This patch sets the BCCH mapping
to 0x00 (=BCCH Normal) statically (0xF200)
The new constands are added to libosmocore, see commit:
f0f9c8c29daaefbf9cff19177ade4a13ffb2e36c
Change-Id: Ie0900f9b810744172b3090ce1b0ef7b7a1132946
|
|
In case of the sysmoBTS and receiving a channel activation ack on a channel
that was marked as broken, release it again.
Use a normal release without SACCH deactivation and release the rqd_ta data.
Also add a local variable 'ts' to shorten some lines.
The typical situation where this would occur is with high latency between BTS
and BSC (or NITB). If a channel activation ack does not arrive in time, a
channel is marked broken, and never recovers after that. This patch will
release the channel again, which will remove the BROKEN_UNUSABLE state and
makes lchan available again. Reported by Rhizomatica.
However, in case of packet loss, i.e. when the channel activation ack never
arrives at the BSC, this patch does not provide a resolution of the
BROKEN_UNUSABLE state.
On dynamic timeslots: clearing the dyn ts state could possibly happen in
lchan_free() instead of in rsl_rx_chan_act_ack(). That's to be done in a
separate patch, if at all.
Tweaked-By: nhofmeyr
Change-Id: I63dc0deaf15ba7c21e20b1e0c7b85f0437e183ed
|
|
count_codecs() is called on every chan act ack, also for channels other than
TCH/F and TCH/H. So this logging happens a lot during normal operation but adds
no real information.
Also, RSL would be the wrong logging category for this -- RSL is about the RSL
communications, not whether our internal code tries to count lchan codecs for
the wrong channel types.
Change-Id: Ibdac3bbe48745fe6a1c31d6f87369c9066c0374a
|
|
Use channel type name instead of number and log it with DEBUG facility
otherwise it produces lots of irrelevant messages for SDCCH*
Change-Id: I11b04e0cb02bf6ed01f6076cb31a56d8921d735e
|
|
We count the codec when the channel was successful setted up
Using sign_link->trx->bts instead of msg->trx to get the bts.
Add OSMO_ASSERT for bts within count_codecs()
Change-Id: Ib49c7c337980a7d6f189d7a0551ca2e4c3822f45
|
|
Found this by coincidence, no actual failure case was observed.
lchan_lookup() does have a return NULL code path, so we should not blindly use
its returned pointer.
Change-Id: I34ce126d36420b8194c88c0faa865294334a6658
|
|
This reverts commit 38e9ea3f7f385c6660c5958970af5c71adc1682b.
Introduced a reproducable segfault, because msg->trx is not actually set/used
in the openbsc code paths.
Program received signal SIGSEGV, Segmentation fault.
count_codecs (lchan=0x1, bts=<optimized out>) at ../../../src/libbsc/abis_rsl.c:104
104 rate_ctr_inc(&bts->network->bsc_ctrs->ctr[BSC_CTR_CODEC_V1_FR]);
(gdb) bt
#0 count_codecs (lchan=0x1, bts=<optimized out>) at ../../../src/libbsc/abis_rsl.c:104
#1 0x0000000000425661 in abis_rsl_rx_dchan (msg=<optimized out>) at ../../../src/libbsc/abis_rsl.c:1516
#2 abis_rsl_rcvmsg (msg=0x8143f0) at ../../../src/libbsc/abis_rsl.c:2611
#3 0x00007ffff71420d0 in handle_ts1_read (bfd=<optimized out>) at ../../src/input/ipaccess.c:271
#4 ipaccess_fd_cb (bfd=0x815af8, what=1) at ../../src/input/ipaccess.c:386
#5 0x00007ffff7779b62 in osmo_fd_disp_fds (_eset=0x7fffffffe590, _wset=0x7fffffffe510, _rset=0x7fffffffe490) at ../../src/select.c:149
#6 osmo_select_main (polling=polling@entry=0) at ../../src/select.c:189
#7 0x0000000000406fac in main (argc=<optimized out>, argv=0x7fffffffe738) at ../../../src/osmo-nitb/bsc_hack.c:385
|
|
we count the codec when the channel was successful setted up
Change-Id: Ifc8a406a11dce16b9e7f3310841e470545550a2c
|
|
Add log_name to lchan_lookup() and pass such from the various RSL rx events
that call it to validate the RSL chan_nr.
Change-Id: Id81e7b8b9c27831923f050a78dfc7d650e687033
|
|
The message 'RF Channel Release due error 0' keeps catching my eye because
it says 'error' even though the error code is zero, i.e. no error.
This shall end now.
Change-Id: Ie0b9d62e8ce85a096c963931e0ae5527b8dc490a
|
|
Tweaked-By: Neels Hofmeyr <nhofmeyr@sysmocom.de>
Change-Id: I7361033cd1eb919ec3c2ea2652f40ab8c75b2f99
|
|
Change-Id: I57c3b7d27d857c96e3fa3dacf7b766bc43100fc3
|
|
In rsl_lchan_mark_broken(), call rsl_lchan_set_state() so the state transition
gets logged in the debug log.
Remove logging for the broken channel at the callers, instead log the error
actually in rsl_lchan_mark_broken() itself, with the reason message passed by
the caller anyway. (Removes code dup and ensures it's always logged.)
Change-Id: I54ae9bbd3f193bae7b1bda1fef3e33e62b353bf5
|
|
Change-Id: Ifbf31cde24b2d1022b7a472966c17959c96e6dda
|
|
Change-Id: I7ddcb41edce1cd7b22fe91e33bdcaedb21856222
|
|
Tested by hacking a REL ACK delay of a couple of seconds into osmo-bts' rsl.c
for the first TCH_H lchan:
[[[
diff --git a/include/osmo-bts/rsl.h b/include/osmo-bts/rsl.h
index 093e9cb..b35c3bb 100644
--- a/include/osmo-bts/rsl.h
+++ b/include/osmo-bts/rsl.h
@@ -22,6 +22,7 @@ int rsl_tx_est_ind(struct gsm_lchan *lchan, uint8_t link_id, uint8_t *data, int
int rsl_tx_chan_act_acknack(struct gsm_lchan *lchan, uint8_t cause);
int rsl_tx_conn_fail(struct gsm_lchan *lchan, uint8_t cause);
int rsl_tx_rf_rel_ack(struct gsm_lchan *lchan);
+int rsl_tx_rf_rel_ack_later(struct gsm_lchan *lchan);
int rsl_tx_hando_det(struct gsm_lchan *lchan, uint8_t *ho_delay);
/* call-back for LAPDm code, called when it wants to send msgs UP */
diff --git a/src/common/l1sap.c b/src/common/l1sap.c
index 3802e25..1f92b0d 100644
--- a/src/common/l1sap.c
+++ b/src/common/l1sap.c
@@ -491,7 +491,16 @@ static int l1sap_info_rel_cnf(struct gsm_bts_trx *trx,
lchan = get_lchan_by_chan_nr(trx, info_act_cnf->chan_nr);
- rsl_tx_rf_rel_ack(lchan);
+ static int yyy = 0;
+
+ DEBUGP(DRSL, "%s YYYYYYYYYYYYYYYYYYYYY %d %s\n",
+ gsm_lchan_name(lchan), yyy, gsm_lchant_name(lchan->type));
+
+ if (lchan->type == GSM_LCHAN_TCH_H && !yyy) {
+ yyy ++;
+ rsl_tx_rf_rel_ack_later(lchan);
+ } else
+ rsl_tx_rf_rel_ack(lchan);
/* During PDCH DEACT, this marks the deactivation of the PDTCH as
* requested by the PCU. Next up, we disconnect the TS completely and
diff --git a/src/common/rsl.c b/src/common/rsl.c
index 3c97af9..7926f21 100644
--- a/src/common/rsl.c
+++ b/src/common/rsl.c
@@ -534,6 +534,22 @@ int rsl_tx_rf_rel_ack(struct gsm_lchan *lchan)
return abis_bts_rsl_sendmsg(msg);
}
+struct osmo_timer_list yyy_timer;
+
+static void yyy_timer_cb(void *data)
+{
+ rsl_tx_rf_rel_ack(data);
+}
+
+int rsl_tx_rf_rel_ack_later(struct gsm_lchan *lchan)
+{
+ yyy_timer.cb = yyy_timer_cb;
+ yyy_timer.data = lchan;
+ osmo_timer_schedule(&yyy_timer, 10, 0);
+ return 0;
+}
+
+
/* 8.4.2 sending CHANnel ACTIVation ACKnowledge */
static int rsl_tx_chan_act_ack(struct gsm_lchan *lchan)
{
]]]
Change-Id: I87e07e1d54882f8f3d667fa300c6e3679f5c920d
Fixes: OS#1798
|
|
Actually schedule an activation timer for the activation part of a dyn TS
switchover. It needs to be restarted because the channel release procedure in
the first part of a switchover actually removes the activation timer.
Change-Id: Ibf50d13ba10298464a8b07e34716763161438990
|
|
The dyn_ts_switchover_*() functions made the impression that they act on a
specific lchan of a timeslot. The assumption that we would remember to use e.g.
lchan[1] across a PDCH deactivation is brain damaged to begin with; and
factually we always use lchan[0] anyway (the only case for using lchan[1] would
be when switching to TCH/H, but the channel allocator will always return
lchan[0] for that).
Instead of the brain damaged lchan args, use a ts arg across all
dyn_ts_switchover_*() functions, with one exception: The
dyn_ts_switchover_complete() actually receives an RSL activation ack message on
a specific lchan and needs to evaluate its lchan type. This will always be
lchan[0] as it is now, but we should stick with the lchan the message was sent
for.
For PDCH, a check to use lchan[0] already existed, when composing the ACT
message in rsl_chan_activate_lchan_as_pdch(). Replace with an assertion.
Adjust all callers to pass ts instead of lchan.
In dyn_ts_switchover_start(), there was a dead code check that jumps to
switchover_complete() in case the pchan already matches. This never hits,
because we only call dyn_ts_switchover_start() when pchans mismatch. So avoid
guessing at passing lchan[0] to dyn_ts_switchover_complete() by not calling it
at all but logging an error instead.
In rsl_chan_activate_lchan(), we remember some values before going into
switchover from PDCH. Explicitly store them in lchan[0], because after a PDCH
release we have always and will activate no other than lchan[0].
In dyn_ts_switchover_continue(), move the check for any existing lchan->rqd_ref
further above, and more correctly check all lchans that were so far valid on
the TS, instead of just one.
This partly prepares for a subsequent commit to fix the act_timer use for dyn
TS: with the old lchan arg, we might schedule an activation timer on lchan[1]
but receive an ack on lchan[0] (for PDCH), leading to an act_timer expiry.
Change-Id: I3f5d48a9bdaa49a42a1908d4a03744638c59796a
|
|
Tested by hacking a CHAN ACT ACK delay of a couple of seconds into osmo-bts'
rsl.c for the first TCH_H lchan:
[[[
diff --git a/src/common/rsl.c b/src/common/rsl.c
index 3c97af9..4bfd27a 100644
--- a/src/common/rsl.c
+++ b/src/common/rsl.c
@@ -559,6 +559,22 @@ static int rsl_tx_chan_act_ack(struct gsm_lchan *lchan)
return abis_bts_rsl_sendmsg(msg);
}
+struct osmo_timer_list xxx_timer;
+
+static void xxx_timer_cb(void *data)
+{
+ rsl_tx_chan_act_ack(data);
+}
+
+static int rsl_tx_chan_act_ack_later(struct gsm_lchan *lchan)
+{
+ xxx_timer.cb = xxx_timer_cb;
+ xxx_timer.data = lchan;
+ osmo_timer_schedule(&xxx_timer, 10, 0);
+ return 0;
+}
+
+
/* 8.4.7 sending HANDOver DETection */
int rsl_tx_hando_det(struct gsm_lchan *lchan, uint8_t *ho_delay)
{
@@ -614,6 +630,18 @@ int rsl_tx_chan_act_acknack(struct gsm_lchan *lchan, uint8_t cause)
if (cause)
return rsl_tx_chan_act_nack(lchan, cause);
+
+ static int xxx = 0;
+
+ DEBUGP(DRSL, "%s XXXXXXXXXXXXXXXXXXXXX %d %s\n",
+ gsm_lchan_name(lchan), xxx, gsm_lchant_name(lchan->type));
+
+ if (lchan->type == GSM_LCHAN_TCH_H) {
+ if (!xxx) {
+ xxx ++;
+ return rsl_tx_chan_act_ack_later(lchan);
+ }
+ }
return rsl_tx_chan_act_ack(lchan);
}
]]]
Change-Id: Ie82dec9c9fefc476fdf5b5afdad2246b9d6fe304
|
|
Prepares for an upcoming commit using the same check in error_timeout_cb().
Change-Id: I8abfa964631040f798212cc3e360f67f9e09b7c5
|
|
rate counters support the export to statsd and can have a delta value.
Change-Id: Ie749cebd53a0bb618d0e23d375885712078bf8dd
|
|
Init both TCH/F_PDCH and TCH/F_TCH/H_PDCH via dyn_ts_init(), which
refactors dyn_pdch_init().
Make dyn_ts_switchover_start from abis_rsl.c public in abis_rsl.h, so we can
start the initial switchover to PDCH from dyn_ts_init(); in abis_rsl.h include
gsm_utils.h for enum gsm_phys_chan_config.
Change-Id: I5c0b257ba8ff0e9c9a2268681a84b0681a778368
|
|
In struct gsm_lchan, add dyn.rqd_ref and dyn.rqd_ta. These save the Channel
Requested details across the PDCH deactivation dance.
abis_rsl.c: add static functions:
* dyn_ts_switchover*() for the various stages of switchover between pchans.
* pchan_for_lchant() to derive the desired pchan from the lchan type that was
set during lchan_alloc().
* rsl_chan_activate_lchan_as_pdch() to compose the simpler RSL CHAN ACT message
without introducing numerous special cases to the normal RSL CHAN ACT code.
In rsl_chan_activate_lchan(), detect and initiate required pchan switchovers if
requested pchan on a dyn TS differs.
In rsl_rx_rf_chan_rel_ack(), initiate or continue pchan switchovers after a
channel was released.
In rsl_rx_chan_act_ack(), notice that a switchover is complete.
In chan_alloc.c, add ts_subslots(): abis_rsl.c will need to know the number of
subslots per pchan, to verify that all lchans are free before dyn TS
switchover. The subslots_per_pchan[] array is static to lchan_alloc.c, and
since we need a non-trivial check for dyn TS anyway, add public ts_subslots()
to lchan_alloc.c, which also checks the current dyn pchan type.
Change-Id: I5c6bce13092a10204113d84678c587c65e35e4fd
|
|
Change-Id: Ica5ef2197b3e97d5e895f3e3221295d5d0ef8908
|
|
For upcoming dynamic TS, the pchan choice for RSL De-/Activation is not
trivial. So in order to pass the desired pchan to generate the RSL chan_nr,
introduce gsm_lchan_as_pchan2chan_nr().
To avoid code dup, this requires decoupling the gsm_ts2chan_nr() pchan from the
actual ts struct, so refactor gsm_ts2chan_nr() to gsm_pchan2chan_nr() with
explicit pchan, ts_nr and lchan_nr arguments.
Change-Id: I1a40e8452fe8120d350a27973e56be0b8c8c517f
|
|
This will also be used by the new dynamic TS type, so make the name more
general.
Change-Id: I2451b10519dff3e5cdf503b430574c0984d19000
|
|
Dyn TS will add a new type of chan activation, which does not need a Channel
Mode IE. Incidentally, the dyn PDCH also doesn't need this IE if it opts for
sending a PDCH ACT instead. So it makes sense to compose the Channel Mode IE
only after the dynamic decisions are done.
Change-Id: I66d88ad6a4ae7bee1e552960fd4e92aff953125c
|
|
Change-Id: I0f403b13ff9897770c0b855bf57a9440717b46e8
|
|
This function outputs a debug log without line ending, which should be
completed by a subsequent DEBUGPC(), so complete the started log line where
missing in three of the switch cases.
The three cases do print another log message, but since these don't start on a
new line when RLL is in debug level, the log output for these is hard(er) to
read without this patch.
Change-Id: I355647e77e1b2d8e75ae1a167fe87a507a38d82d
|
|
Change-Id: Ifa416eab76e6c26dc83e979d815ae778d0d7133b
|
|
lchan_lookup in abis_rsl.c and rsl_lchan_lookup() from osmo-bts rsl.c are the
same code, except for the log context, which is only set in abis_rsl.c.
Factor out the common code to rsl_lchan_lookup() in gsm_data_shared.c.
Openbsc and osmo-bts each define their own DRSL log constant, so add an int *rc
return code argument and keep the logging part in abis_rsl.c's thin lchan_lookup()
wrapper. Incidentally, this also removes code dup for logging.
To avoid duplicate symbols, the rsl_lchan_lookup() implementation needs to be
removed from osmo-bts, so older osmo-bts git revisions will not build with
this.
Change-Id: Ie89bc5bb9110a0e539d37991dedac6f913211b48
|
|
In preparation for an upcoming change.
Change-Id: I9ce71fd7dde42ad7d20f806ac70c150d11450efa
|
|
Change-Id: Idc8afc4e52e189f474077899eef896381ce238f7
|
|
In preparation for an upcoming change.
Change-Id: I11bd59492fa8d5b9392d9f2b511c8fa9585afe6c
|
|
A new type of dynamic channel will be introduced soon, so prepare some comments
to name the dynamic TS kind more specifically.
Change-Id: I51fa8c2ebba507299e55a5cb7e67e48a6c8471f7
|
|
Skip PDCH activation if the GPRS mode is 'none' at:
* TCH/F_PDCH init after OML Enable (dyn_pdch_init())
* after TCH/F_PDCH is released, in TCH/F mode
* in the T3111 error timer callback after a TCH/F_PDCH was released in error
state
Assert the GPRS mode in rsl_ipacc_pdch_activate() to make sure all callers
check the GPRS mode.
Closes: OS#1765
Change-Id: I970e5f9dbcb1c625209e914a4c7696294ed34e62
|
|
Rather use gsm_lchan_name().
Change-Id: I0334484eaa6a2c0f25925042c9c3c1a4e3e78ca4
|
|
Change-Id: If3cc40022e8283daa991fffe4b6baa734303d8a5
|
|
Change-Id: Ic70aca65b3796c90ba1a88ea67ac7a2ad9190b69
|
|
In rsl_rx_rf_chan_rel_ack(), only activate PDCH when in NONE state.
For the case of REL_ERR state, do the PDCH activation in the error timeout
callback after T3111 is done.
Change-Id: I4c55479b252a12039bb3d8c30a9cbf0199ca410e
|
|
Do the PDCH DE/ACT before we set the lchan->state to De-/Activation Requested.
It makes more sense semantically to change PDCH mode while the lchan is still
in NONE status. Thus slightly move some invocations:
PDCH ACT: Free the lchan before PDCH activation. Hence remove the lchan_free()
call from the rsl_rx_pdch_act_ack() code path; it used to do the PDCH
activation first and call lchan_free() in the callback.
PDCH DEACT: Set the (TCH) Activation Requested state only within
rsl_chan_activate_lchan(), after the PDCH deact is complete.
Channel allocator: don't pick channels that have a PDCH PENDING flag set, to
avoid using channels that are still in PDCH switchover (despite their state
being NONE).
The lchan_may_change_pdch() sanity checks are becoming a lot simpler.
Change-Id: I4206dd4808e21c3e59393ea7f5ab4f438afff066
|