4 Commits (60ecade84bc2c97ee2a9ef0d39d41f3cdeab715f)

Author	SHA1	Message	Date
Bryan Biedenkapp	f7023e2965	update network stack documentation;	2 months ago
Bryan Biedenkapp	9a31910526	fix formatting to make my OCD feel better about itself;	3 months ago
Bryan Biedenkapp	18ac540cd2	update techincal docs;	3 months ago
Bryan Biedenkapp	274b805517	Merge R05A04 (r05a04_dev) into master (#114) ... * rebuild r05a04_dev from nasty Git merge bullshit; implement handling of SNDCP on the FNE instead of dvmhost; add quick sanity Catch2 testcases; BUGFIX: NXDN SACCH was incorrectly handling the RAN and structure causing the structure value to become overwritten; correct badly set example IP range in FNE config; add AI generated documentation for the network statck, FNE REST and DVMHost REST; update version number for next dev version; * remove old documents; * BUGFIX: fix some unbounded array checking; * enhance InfluxDB implementation: fix bad handling on Windows; fix bad use of free() for address info freeaddrinfo(); fix partial data writes for bigger payloads; enhance getting responses back from InfluxDB (this makes it so that InfluxDB cannot fail silently easily); * EXPERIMENTAL: add an adaptive jitter buffer implementation to the FNE to better deal with peers on bad connections that may send packets out-of-sequence; * remove unnecessary garbage in AI generated doc; use references for autos when looping peers; normalize jitter buffer log messages; * cleanup doc; * properly move hastily erected debug YAML config to the peer ACL file instead; * update jitter buffer doc; * BUGFIX: correct naive implementation that did not consider the length of the packet being sent, which could result in a nullptr after casting the ip struct; BUGFIX: send actual bytes read from VTUN vs the default MTU size; * BUGFIX: correct similar ip struct casting problem; * implement a very preliminary DMR to VTUN IP dispatcher based on the P25 implementation (while I am sure this is probably going to be quite iffy, I am quite sure the VTUN TAP interface method we are using *is* the right direction; even though the P25 implementation is iffy, with iteration over time I suspect it is the correct approach, as such I have implemented a DMR equiviliant based on the P25 implementation); * use 3/4 rate instead of 1 rate; add some dev notes future me; * implement preliminary positive response to KMM HELLO (this still will fail with No-Service due to missing implementation for per-SU UKEK implementation from the crypto container); * add some extra logging for VTUN PDU; * better implement V(R) V(S) checking; * add prereq logging of last load time for various lookup tables; * add fne-stats; add reload peers and crypto; * update FNE REST API doc; * WIN32: fix ctime_r and add Win32 compat for ctime_s so we can compile on Win32 again; * Win32: FNE's P25PacketData clock() in its entirety should do nothing on Windows; * shut the compiler up about order of constructor initializers; * EXPERIMENTAL: add option for 1 second delayed network TDU when a subscriber dekeys; * slight TUI color palette change; * implement support to configure adaptive jitter buffer from peered; * initialize FinalCut palette earlier; * document that key IDs are hex not dec; * BUGFIX: lets not memset past the possible end of the key array....; * track active call counts (this might be slightly error prone, so we will at best call this an approximate number); * attempt to validate source IP/port during a RPTL, RPTK or RPTC event when RPTL, RPTK or RPTC is sent while not in the correct login state; * BUGFIX: dvmpatch did not properly release the network stream ID at the termination of a call, resulting in a hang up; * BUGFIX: for R05A04s 1s delayed TDU, make sure we have valid IDs before allowing a delayed TDU; * BUGFIX: add more stream ID lockout reset handling; * BUGFIX: add more stream ID lockout reset handling; * add some experiemental support to dvmpatch to support clear to enc one-way patching; * BUGFIX: only consider jitter buffer for RTP frames not carrying the end of call sequence (65535); * BUGFIX: always allow the network stack to process and pass RTP frames carrying the end of call sequence (65535); * BUGFIX: dvmbridge, ignore UDP frame timeouts when using RTP or USRP UDP streams; * add dropped call support to dvmpatch; refactor/rewrite dvmpatch encryption support; * hide debug messages; * BUGFIX: validate destination matches before handling TDU; * report ssrc for late packets; * prevent issue with 0 callDstId on dvmpatch; reject TDUs with no destination (these are invalid and can fuck off); * add separate packet dumping option to the network stack, to allow for less-verbose debugging messaging (debugigng where packet dump/trace is not required); * BUGFIX: fix condition in P25 call handler where if the TSBK was not decoded it could result in a crash; * add network debug trace around stream ID reset calls; * better handle framequeue debugging when not performing packet dumping; * implement proper packet reordering for P25 PDU data; BUGFIX: add boundary checking for P25 PDU disassembly; * BUGFIX: fix startup crash of dvmbridge when using analog audio mode; add transmit voice frame status for analog audio; correct core network issue determining length of analog audio frames; * bump copyright dates; * BUGFIX: for unauthorized, unconnected failures the NAK response was not sending the appropriate RTP sequence; * enhance REST API logging; add dev REST API commands to force send NAK packets; * add PDU block reordering support to dvmhost; * minor bugfix, correct incorrect log module for host sourced analog audio; * very experimental fix for #111, along with additional enhancements to repeat parrot traffic to all VCs of a trunk site to ensure parrot traffic repeats; * update bridge config and expressly document the need to properly configure timing parameters; * make the program status variable m_running a static s_running for dvmpatch; ensure PeerNetwork for dvmpatch uses m_packetDump instead of m_debug for packet dumping; * (we're still hunting the cause of bridge ceasing UDP socket operations after a period of time, it *seems* to occur on Win32 only though); make the program status variable m_running a static s_running for dvmbridge; ensure PeerNetwork for dvmbridge uses m_packetDump instead of m_debug for packet dumping; fix shutdown issue with local audio causing a crash for dvmbridge; refactor how RTP and uLaw frames work; remove and deprecate the no include length configuration parameter; refactor and centralize UDP audio writing functionality; add verbose log messages for if the s_running flag becomes non-true during operation; * fix concurrency lockup for DMR granting; * remove main thread sleep control this causes WinSock on Win32 to lose its mind when polled too fast; * move sample level dumping option to CLI; * always reset the call streams at the end of a call; * increase packet processing latency warning to 250ms (1/4 second); * begin adding support for P25P2 network transiting; * add some verbose debug trace for checkNetTrafficCollision(); * begin defining P25 Phase 2 primitives; * P2 abbreviated/explicit partition MCOs are essentially just TSBK opcodes; * add MFID specific partition type; * refactor MAC MCO defines; * correct comment; * make slot numerical; * whoops meant == not =; * implement RTP audio pacing by timestamp sent by source; * for maintainability split modes into separate source CPP files; BUGFIX: fix buffer overflow when copying PCM data; allow uLaw encoded RTP frames to also carry source and dest metadata; * preamble tones are not supported for UDP audio; * BUGFIX: dvmpatch do not send LDU2s until network is in non-idle state; * convert direct MMDVM TDU to a timed TDU; correct some handling of LDU1 srcId and dstId; * remove dvmmon, R05A02 will be the last version to support dvmmon, R05A04+ will not support it in favor of using sysview from the FNE for monitoring; * add more P25 Phase 2 primitives for S-OEMI sync and Hamming (8,4,4) for P2 DUID handling; * more P25 Phase 2 constants; * extreme preliminary work to handle P25 Phase 2 MAC messages (nothing uses this yet this is for future proofing); * continued work on preliminary instrumentation for Phase 2 MAC handling; * fix missing parens; * split the decodeVCH_MACPDU into two functions, one for IEMI the other for OEMI; * bugfix and correct encode/decode for RS 52,30,23, 46,26,21, 45,26,20, 44,16,29; implement catch2 cases for P25 Phase 2 RS codes; correct bad EC check for RS 24,12,13, 24,16,9 and 36,20,17 that could cause false positive failures in edge case conditions; * implement and add more EDAC verification test; * add DMR EMB and QR 16,7,6 tests; correct issue where QR 16,7,6 decode() was correcting and returning the wrong number of bits; correct issue where DMR EMB would not actually use the corrected QR 16,7,6 codeword; * finally what I was trying to get to -- add testcases for the new P2 MAC PDU logic; correct implementation problems with the P2 MAC PDU handling for OEMI and IEMI (the implementation still requires handling scrambling); * add missing Hamming FEC test cases; * finish out the base crypto test suite and include the DES algo; * (this first part is subject to Git revert if it becomes problematic, but because this is a dev branch I am gonna roll with it) revert FrameQueue timestamp list/map changes back to pre-R04J32, the original implementation was far cleaner and faster with O(1) speed for timestamp lookup vs the O(n) lookup, additionally, the newer R04J32+ implementation introduces a pointer that can be dereferenced incorrectly causing a edge case crash; fix the implementation for handling the mutex for timestamp map locking, the implementation used a incorrect instance mutex which could in high-traffic conditions cause a race condition and ultimately a crash; * for condition where a downstream peer is trying to connect to us, only validate the IP instead of IP/port; * EXPERIMENTAL (and untested): allow source ID overriding for parrot playback, this feature will rewrite the source ID to a static ID configured in the FNE configuration, for P25, NXDN and Analog this rewrite is striaghtfoward, for DMR this will likely work oddly due to the way DMR has source ID data embedded in the transited data frames; * validate the RPC data length includes the message; * BUGFIX: handle edge case where a opcode $27 is not an ISP_EMERG_ALARM_REQ but rather its a network sourced OSP_DENY_RSP; BUGFIX: handle incorrect decoding of OSP_DENY_RSP by SysView; BUGFIX: for DVRS operation properly set MFG_MOT for DENY and QUEUE outbounds; * do not use [] access for grantTimers and unitRegTimers, use find() lookups these are safer; * issue a notify CC channel release on a ChanRelease TDULC; * add explicit TDU TG release option, this option allows a CC to process incoming TDUs to determine if a channel grant should be released; add more explicit TDU process logging to the FNE, ensure TDUs being sent outside a call are being logged properly; during a TG unpermit on a non-authoritative VC, transmit a burst of TDUs on the outbound VC RF interface; add extra verbosity to the active TG logging; * add option displayModemDebugMessages to optionally disable or enable debug log messages coming *from* the modem; properly check if we're operating in DFSI or not and if we are do not attempt to set the FIFO buffer lengths or clear buffers related to DMR or NXDN; correct deletion of array types in the DFSICallData structure; correct bad length of the VHDR1 during DFSICallData initialization, the length defaulted to TIA mode which would cause a crash when trying to delete and deallocate VHDR1 when used in V.24 mode; * correct TSBK, TDULC and PDU transmission using V.24 DFSI, timing must be applied like IMBE; lengthen the high-level CC generation timer by 10ms, this has no impact on regular air and hotspot modems, but better times CC frames for V.24 DFSI modems; refactor naming to be clearer for the V.24 jitter buffer transmit types; because V.24 uses OSP_TSBK_RAW and LC_TDULC_RAW correct several off by 2 shift bugs; remove unnecessary V.24 start of stream and end of stream calls; * allow TDUs to be sent regardless of whether a V.24 call was started or not; * do work on V.24 PDU support (inbound should be working pretty much perfect now, outbound -- is still broken); * V.24 PDU work; * allow the user to adjust the frame timeout length when using frame timing at the bridge; ensure audio frames aren't greater then x2 the size of an expected audio frame; * after consideration entirely refactor how UDP frame timing is generated internally when we're using raw PCM; fix a variety of bugs related to locking and handling of the deque for UDP audio frames; better insert silence at the start and end of calls; * disable this debug trace, its too noisy; * do not consider active call or call counts for parrot; * add call collisions to the stats counters; add REST APIs to reset and zero call counters; * log non-call DMR terminators like P25; * correct endpoints for counter reset APIs; * reset active call count on a maintainence loop; * increase V.24 outbound rate for TSDUs; * properly set the V.24 Tx buffer size to the FIFO length like done for air and hotspot modems; pass whether or not the frame being written to the modem is an immediate frame; modify V.24 modem to properly maintain two independant buffers, one for background/normal priority frames, and one for immediate priority frames (this helps with CC mode delays, there are more buffering issues, likely at the modem to mitgate but thats a future project); * update gitignore to include some python stuff; add very very preliminary Python tool that helps generate dvmhost configurations; * add option to disable deny responses in conventional for radios that are braindead and interpret an OSP_DENY as a ISP_EMERG because of opcode reuse (thanks TIA-102); * delete array instead of delete; * add protections to prevent the network mutex from locking too long; add proper network watchdog to prevent unterminated calls from hanging bridge; * ICC and GRANT shouldnt NAK, as this will cause a straight disconnect/reconnect, ignore them instead; * fix issue where in-call control might start looping requests; * lets make sure RID0 is never used, treat as WUID_FNE instead; * only send ICC upstream when the ICC request originates on the local FNE; * Revise cross-compilation instructions in README Updated cross-compilation instructions for ARM 32-bit and ARM 64-bit, including examples and required packages. * relabel some source files in the FNE, FNENetwork -> TrafficNetwork, DiagNetwork -> MetadataNetwork; remove support for disabling the alternate port, this feature is mandatory now; * identify a software SDR; * add support to dvmcfggen for logging configuration; add support to dvmcfggen for user supplied answers files to automate skipping certain wizard questions; * typo; * HIGHLY_UNECESSARY_DISCLAIMER_FOR_THE_MENTAL; * the typo of that define was driving my OCD nuts; * include new statement in README.md; * enforce a users usage understanding by adding a startup configuration parameter that must be set; * these flags are root level; * make not be stupid error more clear; * copy legalese from config.yml to README.md; * policy updates * Update usage_guidelines.md * create issue & pr templates * remove dvmmon from bug & pr template * revert how timestamping is generated; * BUGFIX: fix incorrectly set length for RTP packets, we were incorrectly adding 4U bytes which does not apply to RTP frames; * enrage the entire community by cleaning up the bridge configuration to be more sane -- a lot of these parameters had nothing to do with FNE network configuration; * code cleanup (no functional change); * fix missing marker bit on RTP seq 0 to make jsb stop complaining; fix missing CTS controller teardown; * properly set the FIRST seq (when seq is reset to 0) as a marker; because JSB is *REALLY* hell bent on this, implement continuous RTP seq to make him happy; * wtf; * cleanup typo and bad spacing due to bad VSCode configuration; correct memory leak when handling buffer drop due to new stream packet wraparound; * correct issue where RTP sequences were not stepped appropriately; * ignore m_audioDetect flag for UDP audio outbound to the FNE; attempt to correct becoming stuck on incoming RTP timestamps; * whoops remove debug trace line; * remove incorrect shift; * ensure network watchdog resets all states; ignore TDU/TDULC when local audio or UDP traffic is running; * reset callStartTime for TGs who receive Non-Call TDU/Terminators; * remove duration log from Non-Call TDU; * disable the expected TS code for RTP frames; * add mutex locking when making a copy of the RadioId table to prevent thread-based race conditions; --------- Co-authored-by: Dev_Ranger <30966416+DevRanger@users.noreply.github.com>	5 months ago