As a result we move from:
INSERT=O(1) SEARCH=O(n) REMOVE=O(1)
to:
INSERT=O(log(N)) SEARCH=O(log(N)) REMOVE=O(log(N))
So we get rid of O(n) complexity every time we need to find a conn
object based on conn_id. When a big number of SCCP conns is handled,
this saves a lot of CPU time every time a conn needs to be looked up,
for instance each time a message is received.
For instance, given 1500 SCCP conns, searching is ~10 steps while it
took 1500 steps beforehand.
Morever, since when creating a new conn_id the code looks if it exists
prior to it, that means in practice inserting used to took O(n), while
now it takes O(log(N)).
Change-Id: I28ac67038207e2fad89ea291629cec5b2f912461
A DUPU message in SUA and M3UA indicates the unavailability of
a (MTP-level) user, i.e. the entire SCCP, ISUP, ... is not available.
If we receive a DUPU (destination user part unavailable) message in ASP
role, then we must
* distribute it to any other ASPs for which we operate in SG mode
* pass it as MTP-STATUS.ind to SCCP, which can then generates
N-PCSTATE.ind to the SCCP User
Change-Id: I1559ed0f761a8495b222df48c6bd43798e220471
Unlike M3UA, in SUA a DUNA/DAVA message can contain not just the point
code that became available / unavailable, but it can also include a SSN.
In that case, it is just the SSN that became available/unavailable, and
not the entire point code. Hence, a N-STATE.ind and not a N-PCSTATE.ind
must be delivered to the SCCP user.
Change-Id: Ie9a45b905bc17e7b695e15fe12ba4bbadcd032bf
SCMG (SCCP Management) is a special sub-system that normally resides
at SSN=1. In Osmocom we so far ignored its existence. However,
in terms of interop with other implementation, we should implement
at least some basic features.
The only procedure implemented in this initial commit is the response
to an incoming SST (Subsystem Test) message. If we don't respond to
this message, a remote SCCP entity could assume the SSN is dead on
our side, rendering communication impossible.
Change-Id: I04b162476f7652ef0540b5ea7299e9447efd1d09
* add N-PCSTATE.ind and N-STATE.ind definitions to SCCP user SAP
* add minimal SCMG (SCCP Management) and LBCS (Local Broadcast)
* generate MTP-PAUSE.ind/MTP-RESUME.ind based on received xUA DUNA/DAVA
* generate N-PCSTATE.ind towards the local SCCP users
Change-Id: Idb799f7d7ab329ad12f07b7cbe6336da0891ae92
Related: OS#2623, OS#3403, OS#4701
Add osmo_sccp_user_sap_down_nofree(), which is identical to
osmo_sccp_user_sap_down(), but doesn't imply a msgb_free().
To implement that, sccp_sclc_user_sap_down_nofree() with the same msgb
semantics is required.
Rationale:
Avoiding msgb leaks is easiest if the caller retains ownership of the msgb.
Take this hypothetical chain where leaks are obviously avoided:
void send()
{
msg = msgb_alloc();
dispatch(msg);
msgb_free(msg);
}
void dispatch(msg)
{
osmo_fsm_inst_dispatch(fi, msg);
}
void fi_on_event(fi, data)
{
if (socket_is_ok)
socket_write((struct msgb*)data);
}
void socket_write(msgb)
{
if (!ok1)
return;
if (ok2) {
if (!ok3)
return;
write(sock, msg->data);
}
}
However, if the caller passes ownership down to the msgb consumer, things
become nightmarishly complex:
void send()
{
msg = msgb_alloc();
rc = dispatch(msg);
/* dispatching event failed? */
if (rc)
msgb_free(msg);
}
int dispatch(msg)
{
if (osmo_fsm_inst_dispatch(fi, msg))
return -1;
if (something_else())
return -1; // <-- double free!
}
void fi_on_event(fi, data)
{
if (socket_is_ok) {
socket_write((struct msgb*)data);
else
/* socket didn't consume? */
msgb_free(data);
}
int socket_write(msgb)
{
if (!ok1)
return -1; // <-- leak!
if (ok2) {
if (!ok3)
goto out;
write(sock, msg->data);
}
out:
msgb_free(msg);
return -2;
}
If any link in this call chain fails to be aware of the importance to return a
failed RC or to free a msgb if the chain is broken, or to not return a failed
RC if the msgb is consumed, we have a hidden msgb leak or double free.
This is the case with osmo_sccp_user_sap_down(). In new osmo-msc, passing data
through various FSM instances, there is high potential for leak/double-free
bugs. A very large brain is required to track down every msgb path.
osmo_sccp_user_sap_down_nofree() makes this problem trivial to solve even for
humans.
Change-Id: Ic818efa78b90f727e1a94c18b60d9a306644f340
Remove all comments, that claim conn_id is the local reference. This
is only a hack, and should not be something to rely on. A properly
separated local reference will be introduced shortly.
Related: OS#3871
Change-Id: I900124037da76caaaf5ecee323eb82e1c6d2e368
The previous hardcoded SCCP timers may cause SCCP connection releases, if the
peer is configured with far lower timers than libosmo-sccp. Testing with a
specific SCCPlite MSC, I experienced an iar of just over three minutes, meaning
that calls would be cut off by the MSC, since the osmo-bsc failed to send an
Inactivity Timer message until seven minutes have passed.
With this patch, SCCP timers are configurable by the user.
Define constant global default timers, and variable user-configurable timers
with each osmo_sccp_instance.
Add VTY UI to configure the timers. Users must call osmo_sccp_vty_init() to get
the sccp-timer config nodes under the 'cs7' node. Show the new UI in
ss7_asp_test.vty.
Note that even though this function is not new at all, until recently, all of
our SCCP users (osmo-bsc, osmo-msc, osmo-sgsn, osmo-hnbgw) failed to call
osmo_sccp_vty_init(), and thus also missed out on the various 'show' commands
defined in sccp_vty.c. In other words, to benefit from the timer
configurability, the patches to call osmo_sccp_vty_init() must first be merged
to the corresponding master branches.
If a 'sccp-timer' config command occurs, the cs7 instance must allocate an SCCP
instance in order to store the timer config. Do that by calling the recently
added osmo_ss7_ensure_sccp() function.
Hence remove the limitation that the SCCP instance must not be populated from
the "simple" setup function. If we want to configure SCCP timers beforehand,
there must be an SCCP instance for that, and there is no hard reason to require
a NULL SCCP instance, besides the desire to prevent this function from being
invoked twice.
Change-Id: I28a7362aa838e648ecc9b26ee53dbcade81a9d65
Define SCCP_STR in sccp_internal.h, because I know that I want to also use it
in osmo_ss7_vty.c.
Fix "Signaling" to "Signalling" upon copying the old string.
Change-Id: Ic93e4771147070a9222c73f80b5f7c29ae7eec35
A previous patch added ss7_instance primary_pc validity checks by means of
OSMO_SS7_PC_INVALID. To be consistent, also adjust sccp_user accordingly.
(see I7f0f0c89b7335d9da24161bfac8234be214ca00c)
Remove the osmo_sccp_user's pc_valid field, replaced by pc=OSMO_SS7_PC_INVALID.
Adjust all code paths.
Simplify some log printing, using the fact that osmo_ss7_pointcode_print() now
outputs "(no PC)" for unset point codes.
Change-Id: I8684c9b559712072c772012890bbf7efa7c8eb35
This is an implementation of SCCP as specified in ITO-T Q.71x,
particularly the SCRC (routing), SCLC (Connectionless) and SCOC
(Connection Oriented) portions. the elaborate state machines of
SCOC are implemented using osmo_fsm, with one state machine for each
connection.
Interfaces to the top (user application) are the SCCP-USER-SAP and on
the bottom (network) side the MTP-USER-SAP as provided by osmo_ss7.
Contrary to a straight-forward implementation, the code internally
always uses a SUA representation of all messages (in struct xua_msg).
This enables us to have one common implementation of all related state
machines and use them for both SUA and SCCP. If used with real SCCP
wire format, all messages are translated from SCCP to SUA on ingress and
translated from SUA to SCCP on egress. As SUA is a super-set of SCCP,
this can be done "lossless".
Change-Id: I916e895d9a4914b05483fe12ab5251f206d10dee
Harald Welte
Pau Espin
Neels Hofmeyr
Oliver Smith
Neels Hofmeyr