First steps towards a new GTP hub. The aim is to mux GTP connections, so that
multiple SGSN <--> GGSN links can pass through a single point. Background:
allow having more than one SGSN, possibly in various remote locations.
The recent addition of OAP to GSUP is related to the same background idea.
(This is a collapsed patch of various changes that do not make sense to review
in chronological order anymore, since a lot of it has thorougly transmorphed
after it was first committed.)
Sponsored-by: On-Waves ehf
Extend the ul/dl counting to count the usual messages on the
Gb interface. Add counters for the attach, routing area update,
pdp context activation and deactivation procedures. Update the
test result with the new counters.
Trigger an OAP registration upon IPA connect. Feed incoming OAP messages to
oap_handle() and send replies returned by it.
Add oap_config to sgsn_config (todo: vty).
Sponsored-by: On-Waves ehf
[hfreyther: Fix coding style]
Add new kitchen sink openbsc/utils.h and libcommon/utils.c to make three so far
static functions public (so I can use them in the upcoming OAP code).
A place to put them could have been the gprs_utils.h, but all general functions
in there have a gprs_ prefix, and todo markings to move them away. All other
libcommon headers are too specific, so I opened up this kitchen sink header.
Replace the implementation of encode_big_endian() with a call to
osmo_store64be_ext(). See comments.
Apply the change in Makefiles and C files.
This change has some implications for the test case. It manipulated
bss_ptmsi_state and sgsn_tlli_state variables to make the output of
rand_r() and thus the TLLI/TMSI used predictable.
This possibility is gone when using RAND_bytes() so instead it is
overridden by a function that returns a deterministic sequence of values
(0x00dead00, 0x00dead01, ...). The test cases are adapted to expect
these values instead of the pseudo random values before.
The gbproxy_test stdout file changes as well, but only where the
TLLI/TMSI is displayed (in the hex dumps as well as the TLLI cache
entries). All other output is the same.
This API is a bit unfortunate as the caller will also
access the endpoint directly. E.g. like this:
output = bsc_mgcp_rewrite(...,
mgcp_net_src_addr(endp),
endp->net_end.local_port, -1,
In terms of "terminology" the "net" was meant to be bad
internet and the "bts" is the local and trusted network
segment. With this terminology the "bts" would be the
call-agent/MGW and "net" where the BSCs will send data
to but that is not the case and terminology actuallys
refers to:
* net: The addresses exposed to the entity that
made the MGCP call
* bts: The system where we get our data for the
local audio flow.
Fix the method but leave the API as it is. Use the net_end
in the net_src method and the bts_end in the bts_src method.
There appears to be a leak of CIDs:
<000b> mgcp_osmux.c:544 All Osmux circuits are in use!
There are paths that a CID had been requested and never released
of the NAT. Remember the allocated CID inside the endpoint so it
can always be released. It is using a new variable as the behavior
for the NAT and MGCP MGW is different.
The allocated_cid must be signed so that we can assign outside
of the 0-255 range of it.
Fixes: OW#1493
Some systems only want to use Osmux. In case only Osmux
should be used fail if it has not be offered/acked.
Client:
Verified On, Off and Only with X-Osmux: 3 and without this field.
<000b> mgcp_protocol.c:823 Osmux only and no osmux offered on 0x14
<000b> mgcp_protocol.c:884 Resource error on 0x14
NAT:
Not tested and implemented
Fixes: OW#1492
* Test that one can get an id
* That they are assigned predicatble right now
* That returning them will make the number of used ones go down
* That allocating more will fail
The signature of mr_config and the BSC implementation didn't
match and the compiler was warning about it:
osmo_bsc_api.c:530:2: warning: initialization from incompatible pointer type
.mr_config = bsc_mr_config,
^
osmo_bsc_api.c:530:2: warning: (near initialization for ‘bsc_handler.mr_config’)
Change the mr_config again and provide an implementation
that will set the ms and bts data structure. It would be
better to put the size outside of the IE but I am not going
to change it right now. It would also be nice to either move
the AMR setting into the "nitb" structure or have the msc
data be used _after_ the bts settings. This needs to be
cleaned up in the next step.
Manually verified by placing a MO call and checking that
both the channel mode modify and the mode modify request
contain the multi rate config with the rate mr config
(length two bytes, version 1, icmi==1, no start mode being
set).
This way a lot of if/else can just be killed by the caller deciding
which of the two instances to use.
I have copied both branches to new files, replace bts for ms in one
of them and ran diff on it. There is no difference.
In case of the RTP bridge mode we need to select the codec
ourselves. Rely on the same (incomplete) codec selection that
can be done using the mncc-int configuration node. This might
gain bearer capabilities support.
In case of a SDCCH a TCH/F will be attempted to be assigned.
This is an open issue for both modes and there should be a
preference for full or half-rate channels somewhere.
Implement sending MDCX on the newly allocated channel and send
the data to the same destination as the currently connected one.
This way the receiver can implement RTP RFC Appendix A.1 and
deal with the new source.
For the LCR rtp-bridge audio should directly flow to the
remote system. In contrast to the original patch audio
will now flow directly from the BTS to the remote system.
This assumes that BTS and the remote system are in the
same network segment and can directly communicate.
There are various limitations in the first iteration of
the implementation:
We could (and in the future) should delay the assignment
but currently we are forced to pick the channel and move
it to the audio state. In case we are located on a SDCCH
we always need to change but if we are on a TCH we could
send the ipa.CRCX and change the audio state a lot later.
The net effect is that the audio codec selection needs to
be done in the NITB code and not in the system connected
to it.
This only works with ip based systems. For E1 systems one
could still use the RTP socket or even try to move this
out of the process.
There is no code for handover handling and it relies on
the remote system dealing with the SSRC change of the
system.
This adds the protocol definition for the RTP bridge extension
of Andreas Eversberg and bumps the protocol version.
I added the missing mncc mappings from value to string.
[ 5cf8fb10ea3addcae74d37f4dbf1c1be664df53e protocol extension
5dac90de38990b188f499c602bf18a4f232070e8 payload extension]
When using multiple interfaces on a system one can now configure
which will be served for the BTS ports and which will be served
for the network. The direct usage of source_addr is now only to
initialize the MGCP receiving port itself.
Make it possible to bind the call-agent to a specific IP address
and the network and bts end to different ip addresses. Begin by
clarifying which source ip address we want to have.
Use the existing ulaw encode/decode to support PCMU as well.
The MERA VoIP switch has some severe issues with the GSM codec
and it appears easier to enable transcoding for it.
The mera switch doesn't appear to cope with codec change
between a SIP 180 trying and the 200 ok connection result.
Inserting the codec is touching too many places. Ideally we
should have the transcoding function as pointer in the struct
as well but the arguments differ.. so it is not a direct way
forward.
Iridium is a satellite network which operates a GPRS-like that allows you to
get speeds up to 128kbit/s. However, it takes from 5 to 6 secs to get the
bandwidth allocated, so the conversation is garbled during the time.
This patch uses the new dummy padding support in libosmo-netif that is
controlled through the osmux osmux_xfrm_input_open_circuit().
This includes a new VTY option for osmux.
We have a lot of legacy that I am afraid to break. We have
everything in place to make a good codec selection (e.g. if
we can avoid transcoding, pick the one with best quality or
the lowest speed). Right now I have a specific case where
from all options I want to pick GSM. Guard the codec compat
check behind the disallow transcoding option to make sure
to not break legacy application.
The SDP file handling will get more complicated in terms of
codec selection so let's remove it from the protocol handling
before we start blowing it up in size.
The parsing code assumed that there will be a single payload
type and this assumption is clearly wrong. Forward all of the
payload types. The code is still only extracting the first
type from the list. The variable name has been renamed to
reflect this.
In case foreign simcards are used we can not do authentication
and ciphering. In case a TMSI is re-used too early and we do
page using TMSI we can't know which of the two MS is responding
to us. We could change the "secure channel" routine to ask for
the IMSI and only then stop the paging.
As we don't have ciphering there is not much use in using the
TMSI. Add a mode "no assign-tmsi" that will not assign the TMSI
during LU. Now CM Service Request and Paging Response will
work using the IMSI. There can't be a clash with that.
[ciaby fixed the vty write to use the right name]
in_address is not 'accidently' included by FreeBSD when we include
the osmocom/core/select.h header file. We need to include a bit
more.
In file included from mgcp_protocol.c:38:
../../include/openbsc/mgcp_internal.h:134:21: error: field has incomplete type 'struct sockaddr_in'
struct sockaddr_in forward;
Struct osmo_msc_data contains int core_ncc, which is actually the
MNC part of the PLMN, not to be confused with the Network Colour
Code.
The following patch renames this field for clarity and consistency
with the standards.
We don't need to consume all the entropy of the kernel but can
use libcrypto (OpenSSL) to generate random data. It is not clear
if we need to call RAND_load_file but I think we can assume that
our Unices have a /dev/urandom.
This takes less CPU time, provides good enough entropy (in theory)
and leaves some in the kernel entropy pool.
We are using the token to find the right bsc_config and
then we can use the last_rand of the bsc_connection to
calculate the expected result and try to compare it with
a time constant(???) memcmp.
Check if the NAT has sent 16 bytes of RAND and if a key
has been configured in the system and then generate a
result using milenage. The milenage res will be sent and
noth the four byte GSM SRES derivation.
Generate 16 byte of random data to be used for A3A8 by
the BSC in the response. We can't know which BSC it is
at this point and I don't want to send another message
once the token has been received so always send the data
with an undefined code. The old BSCs don't parse the
message and will happily ignore the RAND.
/dev/urandom can give short reads on Linux so loop
around it until the bytes have been read from the kernel.
Instead of doing open/read/close all the time, open the
FD in the beginning and keep it open. To scare me even
more I have seen /dev/urandom actually providing a short
read and then blocking but it seems to be the best way
to get the random byes we need for authentication.
So one should/could run the cheap random generator on
the system (e.g. haveged) or deal with the NAT process
to block.
Unfortunately the basic structure of the response is broken.
There is a two byte length followed by data. The concept of
a 'tag' happens to be the first byte of the data.
This means we want to write strlen of the token, then we
want to write the NUL and then we need to account for the
tag in front.
Introduce a flag if the new or old format should be used.
This will allow to have new BSCs talk to old NATs without
an additional change. In the long run we can clean that up.
The code to do that doesn't belong to the control interface, so
abstract it out to a separate function gsm_bts_set_system_infos().
[hfreyther: Fix the coding style...]
For real networks we need to check if the requested APN string
is allowed and then resolve the GGSN address through DNS. There
are countries with two or three digit MNCs and one could either
try to keep a list of countries that have two/three digits or
just try both of them. I have opted for the later for the ease
of the implementation.
C-Ares doesn't allow to cancel a request so we will need to
have the MMCTX and the Lookup have different lifetimes. We simply
set ->mmctx to NULL in case the MMCTX dies more early.
The selected and verified apn_str will be copied into the out
parameter. In case no static APN/GGSN config is present and the
dynamic mode is enabled a request will be made.
c-ares is an asynchronous DNS resolver and we need it to
resolve the GGSN address. This is integrating the library
into our infrastructure. We will create and maintain a list
of registered FDs (c-ares is currently only using one of
them) and (re-)schedule the timer after events occurred.
Include the hlr-Number of the subscriber in the CDR. This is useful
for debugging and understanding which equipment was used during the
test. In contrast to the MSISDN the '+' is emitted as the number
must be in international format already.
Copy the hlr-Number into the sgsn_data and use it during
the purgeMS. There is no unit test that looks at the data
we send so I manually verified this by looking at the output.
Below is the output of the test that purges the subscriber.
<000f> gprs_subscriber.c:170 SUBSCR(123456789012345) Sending GSUP, will send: 0c 01 08 21 43 65 87 09 21 43 f5 09 07 91 83 61 26 31 23 f3
The charging_id is provided by the GGSN. Copy it into the CDR
part of the data structure so it will remain present until after
the pdp context has been deleted.
This is consuming the new signals and allows to install several
different CDR/observing/event/audit modules in the future. For
getting the bytes in/out the code would have had to undo what the
rate counter is doing and at the same time adding a "total" to
the ratecounter didn't look like a good idea, the same went for
making it a plain counter.
Begin writing the values one by one and open/closing a new FILE
for every log messages. This is not efficient but easily deals
with external truncation/rotation of the file (no fstat for and
checking the links and size). As usual we will wait and see if
this is an issue.
Add some new members to our PDP context structure to see what it
is about.
sgsn_create_pdp_ctx should use the subscribed QoS. When selecting
the PDP context we inject the QoS to be used into the TLV structure
and use it during the request. Assume a "qos-Subscribed" structure
only with three bytes and prepend the Allocation/Retention policy
to the request.
The MSISDN should be present for "security" reasons in the first
activation of a PDP context. Take the encoded MSISDN, store it for
future use and then put it into the PDP activation request.
The MM Context contains a field for a decoded MSISDN already. As
we need to forward the data to the GGSN I want to avoid having to
store TON and NPI in another place. Simply store the data in the
encoded form.
It is a bit arbitary to decide which one is the global
and which one is the local one. We might change it around.
I don't think we want to introduce it based on BTS.
For the BSC we will have the gsm48_hdr and don't need to
find data within SCCP. For legacy reasons we need to
initialize con_type, imsi, reject causes early on and
need to do the same in the filter method.
This means we need to require a talloc context and
simply operate on the list. I had considered creating
a structure to hold the list head but I didn't find
any other members so omitted it for now.
Move the filter methods to the filter module. This is
still only usable for the NAT and the _dt/_cr filter
routines need to move back to the bsc_nat in the long
run.
ENDPOINT_NUMBER takes the difference of two pointers. On 64bit
builds the difference is a long and the compiler then complains
about the usage of abs. We will never have thousands of endpoints
so silence the warning by casting the ENDPOINT_NUMBER to int.
mgcp_vty.c:1381:34: warning: absolute value function 'abs' given an argument of type 'long' but has parameter of
type 'int' which may cause truncation of value [-Wabsolute-value]
rtp_port = rtp_calculate_port(ENDPOINT_NUMBER(endp),
^
../../include/openbsc/mgcp_internal.h:206:31: note: expanded from macro 'ENDPOINT_NUMBER'
#define ENDPOINT_NUMBER(endp) abs(endp - endp->tcfg->endpoints)
^
mgcp_vty.c:1381:34: note: use function 'labs' instead
The idea of "subscriber_get_channel" was that different
requests would be coordinated. At the same time we have
seen that the "queue" can get stuck at both 31C3 and the
rhizomatica installations.
Voice calls and SMS do not need coordination. We should
be able to send SMS on a voice channel and switch the MS
from a SDCCH to a TCH in case we establish a voice call.
The SMS code itself needs to coordinate to obey the limit
of one SMS per direction but this should be enforced in
the sms layer and not on the subscriber.
Modify the code to have a simple paging coordination. The
subscriber code will schedule the paging and register who
would like to know about success/failure.
This allowed to greatly simplify the paging response
handling for the transaction code (and in fact we could
move the transaction list into the subscriber structure
now). The code gained to support to cancel the notification
of a request (but not the paging itself yet).
TODO: Cancel paging request in case no one cares about it
anymore.
We might have compiled transcoding into the MGW but
we don't want to enable it for a given user. Add a new
switch that should allow that.
I had manually tested the allow-transcoding/no allow
VTY interface for the primary interface and a new trunk
using show running-config.
We had issues with odd behavior on the nanoBTS which lead
to the introduction of the "broken" state. On busy multi
BTS cells (e.g. rhizomatica) with wifi backhaul the timeout
we set to wait for a RF Channe Release ACK is sometimes too
little and channels are marked broken that look to be okay
(besides the still to be determined delay).
In case of a sysmoBTS we now know that we can change the
state of a broken channel back to normal in case we do
receive the right response.
Manually verified using the Smalltalk BTS code
PackageLoader fileInPackage: 'FakeBTS'
bts := FakeBTS.BTS new.
bts btsId: '1903/0/0'.
bts connect: 'localhost'.
bts waitForBTSReady.
test := FakeBTS.OpenBSCTest new.
test bts: bts.
test requireAnyChannel
... wait for NITB output
<0004> abis_rsl.c:223 (bts=0,trx=0,ts=0,ss=0) Timeout during deactivation! Marked as broken.
... process pending messages
stdin next
<0004> abis_rsl.c:735 (bts=0,trx=0,ts=0,ss=0) CHAN REL ACK for broken channel. Releasing it.
So the channel went from broken to unallocated.
We need to use different LAC/CI towards the core network.
It is a bit problematic as LAC/CI is a per BTS attribute
so this feature only works if a BSC manages everything in
the same LAC.
Related: SYS#1398
This reverts commit f81cacc681.
Since the PURGE MS retry mechanism had been removed, this feature
is not used anymore. It just makes the code more complex.
Conflicts:
openbsc/include/openbsc/gprs_sgsn.h
openbsc/src/gprs/gprs_subscriber.c
openbsc/tests/sgsn/sgsn_test.c
Currently the APN IE in the Activate PDP Contex Request and the PDP
data that is stored with the subscriber is ignored completely.
This commit adds the sgsn_mm_ctx_find_ggsn_ctx that checks the APN IE
against the subscriber's PDP data entries if both are present. If
there is no match, the request is rejected.
If an APN IE has not been included but PDP data entries are present,
the function checks all of these entries against the static 'apn'
configuration to find a suitable entry.
If an APN has not been determined so far and any APN is allowed, the
configuration is checked with an empty APN string, to allow for
default configurations based on the IMSI prefix only.
If nothing of this succeeded but the request wasn't rejected either,
and there is no 'apn' configuration at all or if any APN is allowed
but a default configuration ist not present, the GGSN with id 0 is
used (if present).
Otherwise the request is rejected ('missing APN').
Ticket: OW#1334
Sponsored-by: On-Waves ehf
This commit adds the exported functions apn_ctx_find_alloc,
apn_ctx_free, apn_ctx_by_name, and apn_ctx_match to manage and
retrieve APN to GGSN mappings.
The following VTY commands are added to 'config-sgsn':
- apn APN ggsn <0-255>
- apn APN imsi-prefix PREFIX ggsn <0-255>
which maps an APN gateway string to an SGSN id. The SGSN must be
configured in advance. When matching an APN string, entries with a
leading '*' are used for suffix matching, otherwise an exact match is
done. When a prefix is given, it is matched against the IMSI. If
several entries match, a longer matching IMSI prefix has precedence.
If there are several matching entries with the same PREFIX, the entry
with longest matching APN is returned.
Ticket: OW#1334
Sponsored-by: On-Waves ehf
Currently the PDP info that is transmitted via GSUP is just parsed
and then discarded.
This commit adds a new data structure sgsn_subscriber_pdp_data and
maintains a list of those in sgsn_subscriber_data. The PDP data is
copied from an incoming GSUP UpdateLocationResult message. If that
message contains the PDPInfoComplete flag, the list is cleared before
new entries are added. The 'show subscriber cache' output now also
shows the PDP data entries.
Note that the InsertSubscriberData message is still not supported.
[hfreyther: Added talloc_free in gprs_subscr_pdp_data_clear]
Sponsored-by: On-Waves ehf
Keep track if the power level has been "fixed" by the BSC,
otherwise keep track of the currently ordered one. The ms_power
is the initial value set by the BSC and continues to be used.
Create a control command to read and modify the gprs mode. Use
the get_string_value to indicate if the value was found or not.
This is useful for the ctrl interface where I didn't want to
replicate "none", "gprs" and "egprs". Share code to verify that
a BTS supports the mode.
Related: SYS#591
Currently the MM context cleanup code is distributed over several
functions. sgsn_mm_ctx_free not only frees data structure but also
eventually stops the timer and does the subscriber clean-up.
mm_ctx_cleanup_free (gprs_gmm.c) cleans up the PDP contexts and
unassign the TLLI.
This commit moves the cleanup code from both functions into a new
unifying function sgsn_mm_ctx_cleanup_free that cares about the
clean-up of all related sub-systems.
Sponsored-by: On-Waves ehf
Currently, sgsn_update_subscriber_data can be called with mmctx ==
NULL and will find and associate the right context (if present) based
on the subscriber's IMSI. This will not happen in regular use
any more, since sgsn_update_subscriber_data will only be called when
subscribers are used (auth mode 'remote') and in this case
gprs_subscr_get_or_create_by_mmctx will already be called by
sgsn_auth_request. Therefore, MM context and subscriber are always
associated except for some test cases and experimental VTY usage.
The current implementation of sgsn_update_subscriber_data also causes
additional complexity for the deletion on MM contexts to avoid a
ipossible double-free MM contexts.
This commit removes the MM context <-> subscriber association code
from sgsn_update_subscriber_data. That function must always be called
with mmctx != NULL, now. To avoid problems with VTY and test usage,
the calling subscriber function now only call
sgsn_update_subscriber_data when mmctx != NULL, since the purpose of
that function is to update that state of an existing MM context after
subscriber data has been changed.
Sponsored-by: On-Waves ehf
The old name is somewhat misleading. The function is rather preparing
the subscriber for a subsequent subscr_free, that is possibly invoked
by a subscr_put. It detaches the subscriber from the MM context and
optionally invokes a PURGE_MS procedure. Therefore the _cleanup
suffix is chosen (see mm_ctx_cleanup_free).
Sponsored-by: On-Waves ehf
Currently gprs_subscr_delete implicitely calls subscr_put, which
makes the code more complex than necessary (additional subscr_get) in
a few places. It also makes it more difficult to see, whether get/put
are balanced within a function. In addition, the functions are not
named consistently (gprs_subscr_delete vs.
gprs_subscr_put_and_cancel).
This commit changes the semantics of gprs_subscr_delete and
indirectly of gprs_subscr_put_and_cancel to not call subscr_put on
their argument, but to leave that for the caller to do it
explicitely.
It renames gprs_subscr_put_and_cancel to gprs_subscr_cancel to
reflect that change in the name, too.
Sponsored-by: On-Waves ehf
The subscriber cache would help in case:
* GPRS DETACH, GPRS ATTACH. In that case we might still
have some cached authentication tuples we avoid another
sendAuthenticationInfo request.
* After a detach the cache expiry would make sure to
eventually send a purgeMS to the HLR (which might be
ignored).
At the same time to make the cache work we will need to
make sure to start and stop timers. In case we don't
start we might accumulate subscribers. I am afraid that
the above two benefits do not outweight the complexity
of this implementation.
Currently old LLMEs and MM contexts that haven't been explicitly
detached or cancelled are not removed until another request with the
same IMSI is made. These stale entries may accumulate over time and
severely compromise the operation of the SGSN.
This patch implements age based LLME expiry, when the maximum age has
been reached, the corresponding MM context is cancelled. If such an MM
context doesn't exist, the LLME is unassigned directly.
The implementation works as follows.
- llme->age_timestamp is reset on each received PTP LLC message
- sgsn_llme_check_cb is invoked periodically (each 30s)
- sgsn_llme_check_cb sets the age_timestamp to the current time if
it has been reset
- sgsn_llme_check_cb computes the age and expires the LLME if
it exceeds gprs_max_time_to_idle()
Ticket: OW#1364
Sponsored-by: On-Waves ehf
[hfreyther: Fix typo in comment LMME -> LLME]
Currently, all GPRS timer values are hard-coded. To make these values
configurable in seconds and to show them, conversion functions from
and to seconds are needed.
This patch adds gprs_tmr_to_secs and gprs_secs_to_tmr_floor. Due to
the limited number of bits used to encode GPRS timer values, only a
few durations can be represented. gprs_secs_to_tmr_floor therefore
always returns the timer value that represents either the exact
number (if an exact representation exists) or the next lower number
for that an exact representation exists.
Sponsored-by: On-Waves ehf
Currently an error_cause of 0 is being used to indicate normal
operation. Albeit this is not a defined GMM cause, the value is not
explicitly reserved.
This commit adds the macro SGSN_ERROR_CAUSE_NONE and uses it for
initialisation (instead of relying on talloc_zero) and comparisons.
The value is set to -1 to be on the safe side. The VTY code is
updated to set the error_cause when using the
'update-subscriber imsi IMSI update-location-result CAUSE' command.
Sponsored-by: On-Waves ehf
Currently, an incoming GSUP request message isn't answered at all if
it is not handled due to an error or missing implementation.
This patch adds GSUP error replies for these requests (and only for
requests). It also adds tests for these cases.
Note that several of these tests check for
GMM_CAUSE_MSGT_NOTEXIST_NOTIMPL, which will have to be changed, when
the features are implemented.
Sponsored-by: On-Waves ehf
GSM 09.02, 19.4.1.4 mandates that no other MAP procedures shall be
started until the PURGE_MS procedure has been completed.
This patch implements this by adding corresponding state and checks
to gprs_subscr_purge, gprs_subscr_location_update, and
gprs_subscr_update_auth_info. If an Update Location or a Send Auth
Info Req procedure is not started because of blocking, the retry
mechanism is aborted to shorten the blocking time. The outstanding
Purge MS procedure itself is not aborted.
Sponsored-by: On-Waves ehf
Currently, when the PURGE_MS_REQ to the HLR gets lost (e.g. by a
connection or peer failure), the expired subscriber entry will not get
deleted.
This commit adds a retry mechanism then restarts the procedure after
a timeout (currently 10s). The maximum number of retries is limited
(currently to 3 PURGE_MS messages). If none of these procedures is
completed (either with success or error), the subscriber data is
deleted.
Sponsored-by: On-Waves ehf
When a subscriber entry is going to be deleted by SGSN and when the
subscriber info has been obtained from a remote peer via GSUP, the
peer should be informed before the entry is really deleted. For this
purpose, MAP defines the PURGE MS procedure (see GSM 09.02, 19.1.4).
This patch adds support for the PURGE_MS_REQ/_ERR/_RES messages and
invokes the procedure when the subscriber entry is going to be
removed. This only applies if GSUP is being used, the Update
Location procedure has been completed successfully, and the
subscriber has not been cancelled. The removal of the entry is
delayed until a PURGE_MS_RES or PURGE_MS_ERR message is received.
Note that GSM 09.02, 19.1.4.4 implies that the subscriber data is not
to be removed when the procedure fails which is not the way the
feature has been implemented.
Note that handling 'P-TMSI freezing' is not implemented.
Ticket: OW#1338
Sponsored-by: On-Waves ehf
This commit implements the encoding and decoding of the messages
- Purge MS Request
- Purge MS Error
- Purge MS Result
and adds corresponding tests.
Sponsored-by: On-Waves ehf
Set the expiry delay after the subscriber has been deleted (e.g. by
freeing the MM context). If cancelled, the subscriber will be deleted
immediately and no timeout will be set. If the expiry time is set to
SGSN_TIMEOUT_NEVER, no timer will be started and the subscriber entry
will be kept until it is cancelled.
The following VTY command is added to the sgsn node:
- subscriber-expiry-time <0-999999> set expiry time in seconds
- no subscriber-expiry-time set to SGSN_TIMEOUT_NEVER
The default is an expiry time of 0 seconds, which means that the
subscriber entries are wiped out immediately after an MM context is
destroyed.
Note that unused MM contexts are not expired yet. Therefore the
subscriber will only be expired after a successful MM detach.
Sponsored-by: On-Waves ehf
This patch adds a new logging macro, that logs to DGPRS and provides
a uniform prefix containing the IMSI without using the mm reference.
This is an improvement over using LOGMMCTXP, since the new macro also
provides an IMSI if no MM context is attached.
Sponsored-by: On-Waves ehf
This patch extends gsm0408_gprs_access_denied and
gsm0408_gprs_access_cancelled to accept GMM cause codes. These are
then passed to the MS, unless gsm0408_gprs_access_cancelled is called
with cause 0 (no error -> updateProcedure).
Since gsm0408_gprs_access_denied uses GMM_CAUSE_GPRS_NOTALLOWED if
the cause is not set, and the subscriber's error_cause is never set
(and thus always 0), the SGSN's behaviour does not change with this
patch.
Sponsored-by: On-Waves ehf
Conflicts:
openbsc/include/openbsc/gprs_sgsn.h
[hfreyther: Conflict due the removal of the unused
authenticate flag]
Currently the mapping between GSM 04.08 (GPRS) protocol specific
numbers and their textual description was put into gprs_gmm.c and not
exported.
This commit moves the mappings to a new file gsm_04_08_gprs.c,
renames some of them, and exports them via gsm_04_08_gprs.h.
The following identifiers are renamed to match the corresponding type
names:
- gmm_cause_names -> gsm48_gmm_cause_names
- gsm_cause_names -> gsm48_gsm_cause_names
Sponsored-by: On-Waves ehf
Currently, the reconnect mechanism relies on gsup_client_updown_cb
which in turn gets called based on the OS' view of connection state.
This patch adds a timer based PING mechanism that regularly sends
PING messages and forces a reconnect if a PONG message won't be
received until the next PING message is scheduled. The current ping
interval is 20s.
Sponsored-by: On-Waves ehf
Conflicts:
openbsc/src/gprs/gprs_gsup_client.c
[hfreyther: Conflicts due the potential memleak fix by me. Removed
another TODO from the code as we stop the ping/pong timer]
Currently the GSUP connection to a server is not restarted if the
connection cannot be established or is terminated during operation.
This commit adds a timer based connection mechanism, basically
consisting of a timer callback that calls gsup_client_connect. The
timer is eventually triggered (up == 0) or cleared (up != 0) by
gsup_client_updown_cb. It adds calls to osmo_timer_del() to
gsup_client_connect and gprs_gsup_client_destroy. The latter is now
called instead of talloc_free in gprs_gsup_client_create on error to
be on the safe side.
Sponsored-by: On-Waves ehf
This flag is used to determine, whether the Update Location procedure
shall be invoked. This is currently only set, when the 'remote'
authorization policy is set. When the flag is set, sgsn_auth_update
will not never be called directly by sgsn_auth_request, if an Attach
Request procedure is pending, even if the remote connection fails for
some reason.
Sponsored-by: On-Waves ehf
Currently the flag 'authenticate' is managed per subscriber.
This patch replaces that flag by a global cfg.require_authentication
flag that enables/disables the use of the Auth & Ciph procedure for
every subscriber. The flag is set by the VTY, if and only if the
authorization policy is 'remote'.
The VTY command
- update-subscriber imsi IMSI insert authenticate <0-1>
is removed.
Sponsored-by: On-Waves ehf
This commit adds GSUP client configuration (via VTY), connection set
up, and real message sending.
The following configuration commands are added:
- gsup remote-ip A.B.C.D set server IP address
- gsup remote-port PORT set server TCP port
Ticket: OW#1338
Sponsored-by: On-Waves ehf
This commit adds the client code to get subscriber information from a
remote server. It provides an IPA over TCP connection to transmit and
receive GSUP messages.
Sponsored-by: On-Waves ehf
This patch extends gprs_subscr_query_auth_info and
gprs_subscr_location_update to create GSUP messages with the help of
a static gprs_subscr_tx_gsup_message function. A corresponding
gprs_subscr_rx_gsup_message is added which takes a messages, gets the
subscr, and updates it accordingly.
Sponsored-by: On-Waves ehf
[hfreyther: Added a msgb_free gprs_subscr_tx_gsup_message]
This patch adds functions to encode and decode GSUP messages. This
does not include the layer 1 framing (IPA). The messages so far
supported are: send_auth_info_*, update_location_*,
location_cancellation_*.
Sponsored-by: On-Waves ehf
Move the "logging filter imsi IMSI" into the BTS/NITB code to
allow to set the gsm_subscriber and only compare it. This way
we simply compare the subscriber address and don't have to care
if the subscriber data is still valid.
This extends osmo_nitb to offer a UDP feed of real-time measurement
reports, which can be used by (a variety of) external tools for
visualization or other processing.
We also add a small ncurses based tool (meas_vis) which shows a
baragraph display of the last few mobile stations that were active,
indicating their uplink/downlink receive level and quality.
<WARNING>
This sends non-portable structures like gsm_meas_rep over UDP
and assumes the receiver has identical alignment and endianness! Before
this feature is merged, it either needs to be converted to a unix domain
socket (but they don't do multicast, which would be nice) or the wire
format needs to change into something portable with defined alignment
and encoding
</WARNING>
The idea of ts_alloc()/ts_free() dates back to the very early days of
OpenBSC, where we didn't yet have a fixed PCHAN type assigned for every
lchan in a BTS. However, ever since, PCHAN types (channel combinations)
are configured by OML in a certain way, and we only allocate LCHANs
inside PCHANs of a matching type. There should be no PCHANs with
type GSM_PCHAN_NONE, unless those that you don't want to use for
administraive reasons or the like.
show net with an CCCH+SDCCH/4+CBCH channel active caused bts_chan_load to read
from invalid memory. Fix this by making sure the pchan array is large enough.
==30346==ERROR: AddressSanitizer: stack-buffer-overflow on address 0x7fff9bdc5dc8 at pc 0x5aeece bp 0x7fff9bdc5350 sp 0x7fff9bdc5348
READ of size 4 at 0x7fff9bdc5dc8 thread T0
#0 0x5aeecd in bts_chan_load /home/alphaone/scm/osmo/openbsc/openbsc/src/libbsc/chan_alloc.c:490
#1 0x5af706 in network_chan_load /home/alphaone/scm/osmo/openbsc/openbsc/src/libbsc/chan_alloc.c:511
#2 0x4b7410 in net_dump_vty /home/alphaone/scm/osmo/openbsc/openbsc/src/libbsc/bsc_vty.c:208
#3 0x4b5f23 in show_net /home/alphaone/scm/osmo/openbsc/openbsc/src/libbsc/bsc_vty.c:227
#4 0x7fdabaa425bd in cmd_execute_command_real /home/alphaone/scm/osmo/libosmocore/src/vty/command.c:2042
#5 0x7fdabaa3f124 in cmd_execute_command /home/alphaone/scm/osmo/libosmocore/src/vty/command.c:2077
#6 0x7fdabaa850e9 in vty_command /home/alphaone/scm/osmo/libosmocore/src/vty/vty.c:402
#7 0x7fdabaa75962 in vty_execute /home/alphaone/scm/osmo/libosmocore/src/vty/vty.c:666
#8 0x7fdabaa6d947 in vty_read /home/alphaone/scm/osmo/libosmocore/src/vty/vty.c:1408
#9 0x7fdabaa9165f in client_data /home/alphaone/scm/osmo/libosmocore/src/vty/telnet_interface.c:119
#10 0x7fdaba7860b6 in osmo_select_main /home/alphaone/scm/osmo/libosmocore/src/select.c:160
#11 0x43c656 in main /home/alphaone/scm/osmo/openbsc/openbsc/src/osmo-nitb/bsc_hack.c:355
#12 0x7fdab92604bc (/lib64/libc.so.6+0x224bc)
#13 0x43b6cc (/home/alphaone/local/osmo-asan/bin/osmo-nitb+0x43b6cc)
Address 0x7fff9bdc5dc8 is located in stack of thread T0 at offset 232 in frame
#0 0x4b5faf in net_dump_vty /home/alphaone/scm/osmo/openbsc/openbsc/src/libbsc/bsc_vty.c:182
This frame has 3 object(s):
[32, 40) ''
[96, 104) ''
[160, 224) 'pl'
HINT: this may be a false positive if your program uses some custom stack unwind mechanism or swapcontext
(longjmp and C++ exceptions *are* supported)
SUMMARY: AddressSanitizer: stack-buffer-overflow /home/alphaone/scm/osmo/openbsc/openbsc/src/libbsc/chan_alloc.c:490 bts_chan_load
We can now configure the pyisical channel types for CBCH either in the
CCCH+SDCCH4 or in the SDCCH8 chanel combination.
Depending on whether a CBCH exists on the BTS, we also generate the SI4
with matching CBCH channel description to notify the phones of the
existance of the CBCH.
There is now a VTY command how a SMS-CB message can be sent to a given
BTS.
We do not yet have any logic at all for actual scheduling of multiple
CBCH RSL messages towards one or multiple BTSs yet, though.
This implements the MAP way of subscriber validation when the MS
tries to perform an Attach Request:
1. perform authentication (optionally invoke the sendAuthInfo
procedure), starts the Auth & Ciph procedure
2. perform update location
3. insert subscriber data
4. finish the update location
5. Attach Accept / Attach Reject
The authentication triplets are used and eventually updated if all of
them have been used.
This is currently accessible via the VTY interface by the following
commands:
- update-subscriber imsi IMSI update-auth-info
- update-subscriber imsi IMSI update-location-result (ok|ERR-CAUSE)
Sponsored-by: On-Waves ehf
This commit add data structures, functions, initialization, and VTY
commands for per subscriber authentication triplets.
The following VTY command is added:
- update-subscriber imsi IMSI \
insert auth-triplet <1-5> sres SRES rand RAND kc KC
Note that the triplets are not really used by the SGSN yet.
Sponsored-by: On-Waves ehf
Currently the parser functions for single information elements are
defined within gprs_gb_parse.c and not exported explicitely. In
addition they are named like libosmocore's TLV parser functions and
do not have a proper name prefix. Since it is planned to use them for
other protocols, they need to be globally accessible.
This patch moves them to gprs_utils.c and renames them.
The new names are:
lv_shift -> gprs_shift_lv
v_fixed_shift -> gprs_shift_v_fixed
lv_shift -> gprs_shift_lv
v_fixed_shift -> gprs_shift_v_fixed
In the long term, these functions should be moved to libosmocore (and
renamed again).
Sponsored-by: On-Waves ehf
In case a BTS is dropped, iterate over the list of BTS and check
if a dependency is now missing and then drop the BTS. This check
could lead to check of 256*256 checks (e.g. all BTS on each other
in the chain and the master is being dropped). The performance
aspect of it doesn't matter for our usecase. We expect to have
pairs of BTS right now.
E.g. for the sysmoBTS2050 we have the requirement that the first
board connects before the second due clocking. The easiest point
to enforce this is the BSC. Add a simple bitmask based system to
allow to express dependencies for IP based systems.
This commit mainly extends sgsn_auth.c to use and support the
auth_state SGSN_AUTH_AUTHENTICATE. It will be activated when IMSI and
IMEI are available, authentication is required
(subscr->sgsn_data->authenticate is set), but the MM context is not
marked as authenticated. If the state has been set to
SGSN_AUTH_AUTHENTICATE and sgsn_auth_update() is called, the GMM
layer will be informed by invoking gsm0408_gprs_authenticate().
Sponsored-by: On-Waves ehf
Currently the Authentication and Ciphering procedure is not yet
invoked by the GMM layer.
This patch starts this procedure from within gsm48_gmm_authorize when
the mm->auth_state has been set to SGSN_AUTH_AUTHENTICATE and a call
to gsm0408_gprs_authenticate has been issued directly or indirectly
by the call to sgsn_auth_request.
Sponsored-by: On-Waves ehf
There will be an increasing number of SGSN related fields per
subscriber. Instead of extending gsm_subscriber accordingly, a single
struct sgsn_subscriber_data object is assigned to it. The talloc
context used to allocated that object is the subscr object itself.
Therefore it will be freed automatically along with the subscr
object.
Sponsored-by: On-Waves ehf
This function is called to delete an established MM context
silently without invoking a detach procedure.
It is called when a subscriber is cancelled by the HLR. This
generally happens, when an MS has moved to another routing area and
has to use another SGSN.
Sponsored-by: On-Waves ehf
This commit adds a new authorization policy 'remote' and uses
the subscriber cache for authorization when this policy is being used.
Note that there is no remote backend implemented yet. After the
IMSI/IMEI have been acquired, a request would be sent to the remote
peer. The attach/auth-ciph procedure continues when authorization
info has been received from the peer. This means, that
gprs_subscr_update() must be called then to tell the GMM layer
that it can proceed. A later commit will add VTY commands to do this
manually.
Sponsored-by: On-Waves ehf
This patch adds GPRS specific functions for gsm_subscriber objects
(allocation, retrieval, deletion) and subscriber data
requests/updates. The sgsn_update_subscriber_data callback is used to
notify the sgsn about updates and is extended by a parameter that
passes a reference to a gsm_subscriber.
Sponsored-by: On-Waves ehf
Currently the keep_subscr flag in gsm_subscriber_group refers to a
whole group of subscribers which makes it difficult to really delete
single entries if the flag is set.
This patch adds a keep_in_ram field to gsm_subscriber which allows for
keeping subscriber objects in RAM while deleting others.
Note that really deleting an entry requires that both flags
(subscr_group->keep_subscr and subscr->keep_in_ram) are set to 0. So
only the latter should be used if a specification requires the
deletion of a subscriber entry.
Sponsored-by: On-Waves ehf
The trans_alloc function still uses the subscr object to access the
network object.
This patch adds an explicit net parameter to this function and
removes the access to subscr to obtain it.
Sponsored-by: On-Waves ehf
Currently the SRES value in the Auth & Ciph Response is ignored.
This patch checks the SRES value in response against the value stored
in mm->auth_triplet.sres. If they don't match, an Auth & Ciph Reject
message is sent to the MS. If they match, the mm->is_authenticated
flag is set.
Note that the procedure will not be started yet.
Sponsored-by: On-Waves ehf
Currently mmctx_timer_start is called from within
gsm48_tx_gmm_auth_ciph_req which differs from the way e.g. the
identification procedure is implemented. It also makes it more
difficult to restart the procedure after timeout, which is not
implemented yet. In addition, the timer is not properly stopped when
an AUTH & CIPH response is received.
This patch removes this timer start from gsm48_tx_gmm_auth_ciph_req,
adds the retransmission of Auth & Ciph requests to the timer callback
function, and properly stops the timer in
gsm48_rx_gmm_auth_ciph_resp.
Sponsored-by: On-Waves ehf
Currently every subcriber object directly refers to the gsm_network
which contains a flag shared by every related subscriber
(keep_subscr). This adds a dependency on gsm_network even if only the
function defined in gsm_subscriber_base.c are used.
This patch adds a new struct gsm_subscriber_group which contains the
keep_subscr flag and a back reference to the network object. The
latter is not dereferenced in gsm_subscriber_base.c, so it can safely
be set to NULL when only that part of the gsm_subscriber API is being
used. It also changes that API to use gsm_subscriber_group instead of
gsm_network parameters.
Since there are some places where a pointer to the gsm_network is
needed but where only a gsm_subscriber is available, a 'net' back
pointer is added to the group struct, too. Nevertheless subscr group
and network could be separated completely, but this is not the topic
of this commit.
Sponsored-by: On-Waves ehf
Currently the net pointer is obtained from trans->subscr->net. On the
other hand, the list gsm_trans object is managed by the net object.
This patch adds the back pointer to the structure and replaces all
trans->subscr->net by trans->net expressions. In trans_alloc() the
trans->net pointer is obtained from the subscr object.
Sponsored-by: On-Waves ehf
Increase the bcch_change_mark and generate a new copy of the
system information. Make the method public, add a small test
case. Manually verified using the FakeBTS. I don't know if
the MS will re-read these SIs.
Related: SYS#739
To implement subscriber based authorization a data structure is
needed that keeps the subscriber data. The MSC already uses a similar
struct named gsm_subscriber whose implementation is split into a
generic part (allocation, retrieval, reference counting, list
maintenance) and MSC related parts. For GPRS, only the generic part
will be used and specific fields may be added when needed.
This patch adds a field mm to struct gsm_subscriber that will be used
by the SGSN to store a reference to the current MM context (or NULL
if there is none). This also adds a field subscr to struct
sgsn_mm_ctx that reversely points to a gsm_subscriber (or NULL if
there is none). Either both fields are NULL or both fields are
non-NULL. Note that subscr is being reference counted.
Sponsored-by: On-Waves ehf
Currently the function in sgsn_auth.c either have an sgsn_config or
an sgsn_instance parameter. Since then global sgsn variable is
already being used in that file and since other parts of the SGSN
related code also rely on a global sgsn singleton, these parameters
pretend to provide a flexibility that is not really supported.
Therefore this patch removes these parameters except for the ACL
related functions, which do not call code that uses the sgsn
variable.
Sponsored-by: On-Waves ehf
Currently the authorization is done in sgsn_auth_request for ACL
based authorization. This doesn't match the way remote authorization
would work, so that there is a second call to sgsn_auth_state already
present in sgsn_auth_update.
This patch removes the autorization check completely from
sgsn_auth_request which in turn calls sgsn_auth_update directly now.
Sponsored-by: On-Waves ehf
Currently the VTY 'auth-policy' command results in setting or clearing
the acl_enabled flag. This also enables the matching of the MCC/MNC
prefix of the IMSI.
This patch adds an additional policy 'acl-only' which disables the
MCC/MNC matching and relies on the ACL only.
Sponsored-by: On-Waves ehf
Currently the authorization of an IMSI is done by checking ACLs
synchronously which is not feasible when the subscriber data has to
be retrieved from an external source.
This patch changes this by using a callback when the information is
available. This is also done when only ACL are checked, in this case
the callback is invoked from within sgsn_auth_request(). The callback
function sgsn_update_subscriber_data calls sgsn_auth_update which
in turn calls either gsm0408_gprs_access_granted or
gsm0408_gprs_access_denied. gsm48_gmm_authorize is extended by a call
to sgsn_auth_request when IMSI and IMEI are available but the
auth_state is unknown.
The change has been successfully tested with single phones (E71 and
IPhone 5c).
Sponsored-by: On-Waves ehf
Currently the IMSI is only checked immediately when an Attach Request
is received that contains an IMSI IE. If it contains a P-TMSI
instead, access is always granted.
This commit moves the IMSI check to gsm48_gmm_authorize where it is
applied when IMSI and IMEI have been acquired. This fixes the
authorization when the Attach Accept doesn't contain an IMSI.
Sponsored-by: On-Waves ehf
This patch replaces gsm0408_gprs_force_reattach(msg, mmctx) by two
functions
- gsm0408_gprs_force_reattach(mmctx)
- gsm0408_gprs_force_reattach_oldmsg(msg)
The old function basically consists of the code of the two new
functions, where the code path selected depends on mmctx == NULL,
which is harder to maintain, less obvious to use, and not consistent
with many other SGSN functions.
Sponsored-by: On-Waves ehf
Currently the order of the 'if' clauses in gsm48_gmm_authorize
doesn't match the order in which the conditional parts are entered.
This makes it difficult to maintain. In addition the t3350_mode is
not stored in every path, so that this information is lost when the
identification procedure is started. Since the default value
coincidentally is GMM_T3350_MODE_ATT, this doesn't hurt for Attach
Requests which are the only messages that initially trigger the
authentication yet.
This patch changes the order of the 'if' clause to match the
processing order, it removes the t3350_mode parameter entirely and
introduces a mm->pending_req field. The latter must be set when the
request that causes the authorization before calling
gsm48_gmm_authorize. The gprs_t3350_mode enum is extended by
GMM_T3350_MODE_NONE (value 0, which is the default) to make it
possible to detect related initialisation errors or race conditions.
Sponsored-by: On-Waves ehf
Currently, when a Detach Request is received with an unknown TLLI,
it is answered by another Detach Request (!), even when a power_off
Type is used.
This patch uses gsm48_rx_gmm_det_req to handle the message instead.
So this function is changed to cope with a NULL mmctx. In that case
it doesn't unassign the llme, so this must be done manually
afterwards.
Sponsored-by: On-Waves ehf
Currently the ACL code is located in sgsn_vty.c.
This commit moves this to a new file sgsn_auth.c as a first step to
make authorization more flexible in order to implement remote
acquisition on subsciber data.
Sponsored-by: On-Waves ehf
Currently the P-TMSI IE in PAGING_PS is not patched.
This commit adds code to patch BSSGP P-TMSI IE in
gbproxy_patch_bssgp independently from the P-TMSI patching at the LLC
layer. It also extends gbproxy_update_link_state_dl to use the IMSI
to find the link_info if the TLLI is not present in the message.
Note that the spec (GSM 08.18, 7.2) requires to use of the P-TMSI
instead of the IMSI to select the MS if that IE is available.
Nevertheless as long as the IMSI is always present in downlink BSSGP
messages and as long as the optional P-TMSI refers to the same MS
(which is the case currently), this is not an issue.
Sponsored-by: On-Waves ehf
The osmo-sgsn sends Status messages (or nothing in case of non
GMM/GSM) when the TLLI is unknown. This prevents the MS from
reconnecting.
This patch adds the initiation of an MT detach procedure to force a
re-attach to set up a valid LLE context if an LLE or an MM context
cannot be found. Since this can also be triggered by non-GMM SAPI
messages, a GPRS application callback sgsn_force_reattach_oldmsg is
added which in turn calls the GMM layer to generate the GSM 04.08
specific messages.
Note that the MS can be left in REGISTERED state after initially
wanting to detach itself, since it will receive a Detach Req
(re-attach) when sending a DEACT PDP CTX REQ after the SGSN or
gbproxy (P-TMSI patching enabled) has been restarted. This same
behaviour has been observed with another SGSN.
Sponsored-by: On-Waves ehf
Currently the PDP contexts are hard freed (via sgsn_pdp_ctx_free)
at some places in gprs_gmm.c on the reception of a Detach Req and on
re-use of an IMSI that is already associated with an MM context. This
can lead to segfaults when there is a pending request or a data
indication at libgtp.
This patch add a new function sgsn_pdp_ctx_terminate that de-associates
the PTP context from the MM context, deactivates SNDCP, sets pdp->mm
to NULL and then calls sgsn_delete_pdp_ctx. sgsn_libgtp is updated to
check for pdp->mm being non-NULL before dereferencing it. The
sgsn_pdp_ctx_terminate function will be called for each PDP context of
an MM context before this context is going to be deleted via
sgsn_mm_ctx_free. To ensure, that the ctx->llme (which is accessed
during the deactivation of SNDCP) remains valid, the call to
gprs_llgmm_assign is moved after the call to sgsn_mm_ctx_free. The
handling of re-used IMSIs is changed to mimic the processing of a
Detach Req.
Addresses:
<0002> gprs_gmm.c:654 MM(/f6b31ab0) Deleting old MM Context for same
IMSI p_tmsi_old=0xc6f19134
<000f> gprs_sgsn.c:259 PDP freeing PDP context that still has a
libgtp handle attached to it, this shouldn't happen!
[...]
SEGFAULT
Ticket: OW#1311
Sponsored-by: On-Waves ehf
Currently, the log level is always LOGL_DEBUG. In case of errors it
would be helpful to use a higher log level.
This patch adds a log_level parameter to gprs_gb_log_parse_context to
let the caller decide about the level.
Ticket: OW#1307
Sponsored-by: On-Waves ehf
This function tries to get an accurate name for the message even if
the parsing has been aborted due to message errors.
The patch also moves the settings of the BSSGP related fields in
parse_ctx from behind to the front of bssgp_tlv_parse, to get more
information in the case of failure. This is now consistent with the
handling of the llc and g48_hdr fields.
Id addition, gprs_gb_log_parse_context now uses the new function to
derive a more accurate message name.
Ticket: OW#1307
Sponsored-by: On-Waves ehf
Currently, ptmsi_enc and new_ptmsi_enc point to the beginning of the
mobile identity. Since all P-TMSI in 04.08 (MM) are encoded this way (1
byte header + 4 byte P-TMSI value). This is different to the P-TMSI
encoding in 08.18 (BSSGP), where the P-TMSI is encoded into 4 byte
without MI header.
This patch changes the code to use pointers to the P-TMSI value,
which is encoded in the same way in both specifications.
Sponsored-by: On-Waves ehf
Currently one regexp ('patching') is used for all matching.
This patch adds a second category 'routing' which is exclusively used
for SGSN selection. It also adds a corresponding VTY command:
- match-imsi patching RE : MS related patching (currently APN)
- match-imsi routing RE : Select secondary SGSN on match only
- no match-imsi : Clear all filter expressions
Ticket: OW#1258
Sponsored-by: On-Waves ehf
The current implementation makes it difficult to add further match
expressions.
This patch adds a new struct gbproxy_match that contains the fields
needed for each match expression. The matches (config) and the
results (link_info) are stored in arrays. All related functions are
updated to use them. The old fields in the config structure are
removed.
Sponsored-by: On-Waves ehf
This patch adds und uses the function gbproxy_gsm48_to_peer() which
takes a GSM 04.08 message, encapsulates it in BSSGP and LLC, and
sends it to the BSS peer. This function increments vu_gen_tx_bss
which is now used instead of imsi_acq_retries to set the N(U) of the
outgoing message.
Since imsi_acq_retries isn't currently incremented before a Detach
Accept is generated, this patch also fixes the N(U) of such messages.
Sponsored-by: On-Waves ehf
Currently then link_info is not cleaned up completely, when
gbproxy_unregister_link_info is called.
This patch adds a function gbproxy_reset_link that must be defined
externally. This is done in gb_proxy.c, where it resets the IMSI
acquisition.
Sponsored-by: On-Waves ehf
This field in struct gbproxy_patch_state has involved and holds a
list of all tracked logical links now. Thus the name is modified
accordingly.
Sponsored-by: On-Waves ehf
Currently in many places where 'tlli' (Temporary Logical Link
Identifier) within identifiers is used, the logical link itself is
meant instead. For instance, the tlli_info contain information about
an LLC logical link including up to four individual TLLI.
To avoid confusion between these concepts, this patch replaces all
'tlli_info' by 'link_info' and a few 'tlli' by 'link'.
Sponsored-by: On-Waves ehf
This patch replaces 'tlli' by 'tlli_info' within the following
function identifiers:
- gbproxy_delete_tlli
- gbproxy_delete_tllis
- gbproxy_remove_stale_tllis
- gbproxy_touch_tlli
- gbproxy_unregister_tlli
- gbproxy_remove_matching_tllis
- gbproxy_find_tlli -> gbproxy_tlli_info_by_tlli
- gbproxy_find_tlli_by_* -> gbproxy_tlli_info_by_*
These functions refer to the whole logical link info rather than to a
certain TLLI. So they are renamed to be named consistently with
gbproxy_attach_tlli_info and others.
Sponsored-by: On-Waves ehf
This patch renames gbproxy_check_tlli() to
gbproxy_imsi_matches() and struct tlli_info's
enable_patching to imsi_matches.
It's meant to be more obvious and consistent this way.
Sponsored-by: On-Waves ehf
The Annex A code has a probation period but we don't have it. When
starting with seq_no==0 do not assume that the sequence numbers
have wrapped. Do it by moving the entire checking code into the
else.
mgcp_patch_and_count has grown due supporting linearizing timestamps,
ssrc and other things for equipment like the ip.access nanoBTS. Fight
back and move the Annex A code into a dedicated method.
The result is updated as we now count after all the patching and for
the Annex A code no change in SSRC can be detected.
Currently the SGSN side message's TLLI are searched without checking
the originating SGSN. This leads to collisions if both SGSN use the
same P-TMSI for different MS.
With this patch, the SGSN NSEI is stored within the tlli_info and is
used in comparisons to separate the namespaces.
Note that this type of collision cannot happen with BSS numbers,
since the tlli_info are already separated and stored per (BSS) peer.
Sponsored-by: On-Waves ehf
Currently a tlli_info entry is deleted when the TLLI gets invalidated
by a Detach message.
This patch introduces the possibility to keep tlli_info entries in
the list. Those entries then have cleared TLLI fields, are marked as
de-registered, and can only be retrieved by a message containing an
IMSI or a P-TMSI.
The following VTY configuration commands are added to the gbproxy
node:
- tlli-list keep-mode never : Don't keep the entries (default)
- tlli-list keep-mode re-attach : Only keep them, when a Detach
message with re-attach required has been received
- tlli-list keep-mode identified : Only keep entries which are
associated with an IMSI
- tlli-list keep-mode always : Keep all entries
Note that at least one of max-length or max-age should be set when
this feature is used to limit the number of entries.
Sponsored-by: On-Waves ehf
This function is a remainder of the initial implemenation that was
not meant for TLLI patching and can be used for the BSS side only.
The SGSN side is already using a composition of more flexible
single purpose functions.
This patch changes the implementation to use a similar approach. The
function is moved to gbproxy_test.c and renamed to register_tlli to
keep the tests intact.
Sponsored-by: On-Waves ehf
Currently the RAI in the LLC part of the message is not updated if
the message has been taken from the list of stored messages. The
reason is, that old_raid_matches is update in
gbprox_process_bssgp_ul() but not in gbproxy_flush_stored_messages().
This patch moves the check to gprs_gb_parse_bssgp() which is called
at both places and where other fields like parse_ctx->tlli are set,
too.
In addition, old_raid_matches is replaced by old_raid_is_foreign
since this is clearer in the case when there is no old RAI at all.
Several RAI patch counter assertions are also added to
test_gbproxy_ra_patching().
Sponsored-by: On-Waves ehf
Since at all places where mi_data/mi_data_len is used it will always
contain an IMSI. Thus the names of the identifiers have been updated
accordingly for clarity.
Sponsored-by: On-Waves ehf
This allows us to know what number of messages and bytes has been
received per active osmux endpoint.
Note that an Osmux message is composed of several chunks. Each chunk
contains an osmux header plus several voice data frames.
P: PS=385, OS=11188, PR=195, OR=5655, PL=0, JI=49
X-Osmo-CP: EC TIS=0, TOS=0, TIR=0, TOR=0
X-Osmux-ST: CR=51, BR=3129
The new 'X-Osmux-ST:' notifies the received chunks and bytes.
This patch removes the patch_mode feature including the related VTY
command patch-mode. Where sensible, the other configuration flags are
queried instead.
In addition, this initial checks in gbprox_process_bssgp_dl() and
gbprox_process_bssgp_ul() have been updated.
The patch mode feature has not been used and was increasingly
difficult to maintain.
Sponsored-by: On-Waves ehf
This patch add explicit tests for
- gbproxy_peer_by_bvci
- gbproxy_peer_by_nsei
- gbproxy_cleanup_peers
- gbproxy_peer_by_rai
- gbproxy_peer_by_lai
- gbproxy_peer_by_lac
and for messages with an unknown TLLI sent by the SGSN.
Sponsored-by: On-Waves ehf
Currently the stored messages are only removed, when IMSI acquisition
has succeeded. In addition, receiving two ATTACH_REQ messages in
sequence (e.g. due to loss of a Identity Req/Resp message) will not
restart the IMSI acquisition procedure.
This patch adds gbproxy_tlli_info_discard_messages() to clean up the
message list and calls it from gbproxy_delete_tlli() fixing a
potential memory leak. It is also called when an Attach Request
message has been received. In that case the imsi_acq_pending flag is
cleared, too. This would (re-)trigger the IMSI acquisition procedure
at each of these messages. If an Ident Response has been lost,
resending the Ident Request with the same N(U) will not work.
Therefore the N(U) gets incremented on each Ident Request generated
by the gbproxy. The first N(U) used is 256 which shouldn't collide
with the V(UT) used by the SGSN given that P-TMSI patching is enabled
(since a new random TLLI is used initially on every new (no
tlli_info) connection and V(U) starts with zero then).
Ticket: OW#1261
Sponsored-by: On-Waves ehf
Currently the numeric TLLI or tlli_info's enable_patching flag is
used to decide, whether a APN shall be patched or the secondary SGSN
shall be used. Using the numeric TLLI imposes a problem, when
TLLI/P-TMSI patching is used, since gbproxy_check_tlli uses the BSS
side TLLI namespace when trying to get the tlli_info.
This patch modifies the gbproxy_check_tlli() function to accept a
tlli_info pointer instead of a numeric TLLI. The tlli_info is already
available when the function is called. Since this a similar approach
has been used by accessing the enable_patching flag directly, this
commit unifies checking by always using this function instead of the
flag outside of gb_proxy_tlli.c.
This fixes the APN patching that doesn't work currently when P-TMSI
patching is enabled.
Sponsored-by: On-Waves ehf
This patch refactors SGSN NSEI handling to support a secondary SGSN.
It adds the following VTY commands:
- secondary-sgsn nsei <0-65534>
- no secondary-sgsn
Sending messages to the secondary SGSN is not yet implemented, but
received messages from such a SGSN would be forwarded to the BSS
peers.
Sponsored-by: On-Waves ehf
We want to reduce the background traffic and might set the ping
interval to be in the range of minutes. But this means that if
the TCP connection is frozen several "SCCP CR CM Service Requests"
will be stuck in the send queue without ever being answered. I
could have used the logic of not receiving the "SCCP CC" to close
the connection but instead I am introducing an overload to schedule
the ping as part of the normal SCCP connection establishment.
The VTY write case has been manually verified, I have also looked
at a single trace to see that the SCCP CR and the IPA PING is
transfered in the same ethernet frame.
To modify or route messages based on the IMSI the latter must be known
when the action shall take place.
This patch modifies the gbproxy to optionally retain and enqueue
messages from the MS while initiating an identification procedure.
Further message processing of the LLC PTP link towards the SGSN will
be done, when the identity of the MS has been acquired.
Note that the N(U) of the LLC GMM SAPI are not adjusted, so it is
possible that adjacent messages of a single LLC link arriving either
at the BSS or the SGSN have the same N(U) and might get discarded,
leading to retransmissions and additional delay.
Note also that retransmissions and packet loss are not yet handled
explicitely. If for instance the generated IDENT REQ gets lost, the
gbproxy will not act on its own. In this case, the MS will time out
and eventually resend the Attach Request on which the gbproxy will
act exactly like before (thus having two Attach Req messages in its
queue, which will both be sent after the Ident Resp arrives).
This has been tested successfully with an E71, needing one
retransmission by the SGSN due to an N(U) collision.
Ticket: OW#1261
Sponsored-by: On-Waves ehf
Currently when patching is enabled and an error happens when
receiving a message from the SGSN, the patched message is sent back
with the PDU_IN_ERROR IE.
This patch modifies gbprox_rx_sig_from_sgsn() to copy the message
before it is patched, so that the original message can be used with
the STATUS message. gbprox_rx_ptp_from_sgsn() does all checks before
the message is patched, so copying is not necessary.
Since gbprox_rx_sig_from_sgsn() is not called for BSSGP UNITDATA
messages and the msgb is already been copied in the gbprox_relay2peer
function, the relative performance impact is expected to be low.
Note that the PDU IE of STATUS messages received from an MS and
forwarded to the SGSN will not be patched. STATUS messages from the
SGSN are only logged and not forwarded to the MS.
Sponsored-by: On-Waves ehf
The compiler also uses this attribute for code elimination. If the
nonnull attribute has been given erroneously for an parameter, that
is later been checked against NULL, this check is removed silently
by the gcc if optimization is enabled. This can lead to hard-to-find
segmentation violation faults.
To be on the safe side, this patch removes all uses of the nonnull
attribute in openbsc.
Compiler:
- gcc 4.8.2 (Ubuntu 4.8.2-19ubuntu1): no warning, segfault
- clang 3.4 (3.4-1ubuntu3): no warning, no segfault, asm ok
Example:
/* foo.c */
int f(int* p) __attribute((nonnull));
int f(int *p) {
if (!p)
return 0;
return *p;
}
/* main.c */
int f(int* p) __attribute((nonnull));
int g () {
return f(arg);
}
int main() {
return g(NULL);
}
When these files are compiled into an executable, no warnungs are
issued but it will fail with a segfault when -O2 is used (unless LTO
is active).
Compiler output (gcc -O2):
int f(int *p) {
0: 8b 44 24 04 mov 0x4(%esp),%eax
4: 8b 00 mov (%eax),%eax
6: c3 ret
}
Sponsored-by: On-Waves ehf
In case of some RTP proxy from time to time we are offered both
G729 and G711 but only one of them will work. I intend to adjust
the codec at runtime in case we receive the wrong codec.
We might be offered multiple codecs by the remote and need to
switch between them once we receive data. Do this by moving it
to a struct so we can separate between proposed and current
codec. In SDP we can have multiple codecs but a global ptime.
The current code doesn't separate that clearly instead we write
it to the main codec.
Use the rtp_hdr structure. The basic alignment issue remains
and I need to merge/cherry-pick Jacob's getters for the ts,
sequence number and other attributes.
The talloc_free on the nat lead to the freeing of the bsc_config
which lead to freeing of the rate_ctr_group. The rate_ctr_group
remained in a global list and the next creation of a bsc_config
would access dead memory. Fix it.
The free routine is only meant to be used by the test, for the
real nat we would need to make sure that all connections and
other state that refers to the cfg is removed/closed first.
Fix various memleaks in the test while we are at it. There are
still some to fix.
==7195== Invalid write of size 4
==7195== at 0x4043171: rate_ctr_group_alloc (linuxlist.h:65)
==7195== by 0x804D893: bsc_config_alloc (bsc_nat_utils.c:174)
==7195== by 0x804B5D2: main (bsc_nat_test.c:954)
==7195== Address 0x4311cbc is 52 bytes inside a block of size 208 free'd
==7195== at 0x4029D28: free (in /usr/lib/valgrind/vgpreload_memcheck-x86-linux.so)
==7195== by 0x4048D98: _talloc_free (talloc.c:609)
==7195== by 0x4052806: talloc_free (talloc.c:578)
==7195== by 0x804B58A: main (bsc_nat_test.c:940)
This patch includes several osmux fixes that are interdependent:
1) This adds Osmux circuit ID, this is allocated from the bsc-nat. This
announces the circuit ID in the CRCX MGCP message. This aims to resolve
the lack of uniqueness due to the use of endp->ci, which is local to
the bsc. This ID is notified via X-Osmux: NUM where NUM is the osmux
circuit ID.
2) The dummy load routines are now used to setup osmux both in bsc and
bsc-nat to resolve source port NAT issues as suggested by Holger. The
source port that is used from the bsc is not known until the first
voice message is sent to the bsc-nat, therefore enabling osmux from
the MGCP plane breaks when a different source port is used.
3) Add refcnt to struct osmux_handle, several endpoints can be using the
same input RTP osmux handle to perform the batching. Remove it from the
osmux handle list once nobody is using it anymore to clean it up.
4) Add a simple Osmux state-machine with three states. The initial state
is disabled, then if the bsc-nat requests Osmux, both sides enters
activating. The final enabled state is reached once the bsc-nat sees
the dummy load message that tells what source port is used by the bsc.
5) The osmux input handle (which transforms RTP messages to one Osmux batch)
is now permanently attached to the endpoint when Osmux is set up from the
dummy load path, so we skip a lookup for each message. This simplifies
osmux_xfrm_to_osmux().
After this patch, the workflow to setup Osmux is the following:
bsc bsc-nat
| |
|<------ CRCX ----------|
| X-Osmux: 3 | (where 3 is the Osmux circuit ID
| | that the bsc-nat has allocated)
|------- resp --------->|
| X-Osmux: 3 | (the bsc confirm that it can
| | use Osmux).
. .
| |
setup osmux |----- dummy load ----->| setup osmux
| Osmux CID: 3 |
In two steps:
1st) Allocate the Osmux Circuit ID (CID): The bsc-nat allocates an unique
Osmux CID that is notified to the bsc through the 'X-Osmux:' extension.
The bsc-nat annotates this circuit ID in the endpoint object. The bsc
replies back with the 'X-Osmux:' to confirm that it agrees to use Osmux.
If the bsc doesn't want to use Osmux, it doesn't include the extension
so the bsc-nat knows that it has to use to RTP.
2nd) The dummy load is used to convey the Osmux CID. This needs to happen
at this stage since the bsc-nat needs to know what source port the bsc
uses to get this working since the bsc may use a different source
port due to NAT. Unfortunately, this can't be done from the MGCP signal
plane since the real source port is not known that the bsc uses is not
known.
This patch also reverts the MDCX handling until it is clear that we need
this special handling for this case.