fteqw/engine/common/net_ice.c

4830 lines
144 KiB
C
Raw Normal View History

#include "quakedef.h"
#include "netinc.h"
typedef struct
{
unsigned short msgtype;
unsigned short msglen;
unsigned int magiccookie;
unsigned int transactid[3];
} stunhdr_t;
//class
#define STUN_REQUEST 0x0000
#define STUN_REPLY 0x0100
#define STUN_ERROR 0x0110
#define STUN_INDICATION 0x0010
//request
#define STUN_BINDING 0x0001
#define STUN_ALLOCATE 0x0003 //TURN
#define STUN_REFRESH 0x0004 //TURN
#define STUN_SEND 0x0006 //TURN
#define STUN_DATA 0x0007 //TURN
#define STUN_CREATEPERM 0x0008 //TURN
#define STUN_CHANBIND 0x0009 //TURN
//misc stuff...
#define STUN_MAGIC_COOKIE 0x2112a442
//attributes
#define STUNATTR_MAPPED_ADDRESS 0x0001
#define STUNATTR_USERNAME 0x0006
#define STUNATTR_MESSAGEINTEGRITIY 0x0008
#define STUNATTR_ERROR_CODE 0x0009
//#define STUNATTR_CHANNELNUMBER 0x000c //TURN
#define STUNATTR_LIFETIME 0x000d //TURN
#define STUNATTR_XOR_PEER_ADDRESS 0x0012 //TURN
#define STUNATTR_DATA 0x0013 //TURN
#define STUNATTR_REALM 0x0014 //TURN
#define STUNATTR_NONCE 0x0015 //TURN
#define STUNATTR_XOR_RELAYED_ADDRESS 0x0016 //TURN
#define STUNATTR_REQUESTED_ADDRFAM 0x0017 //TURN
//#define STUNATTR_EVEN_PORT 0x0018 //TURN
#define STUNATTR_REQUESTED_TRANSPORT 0x0019 //TURN
#define STUNATTR_DONT_FRAGMENT 0x001a //TURN
#define STUNATTR_XOR_MAPPED_ADDRESS 0x0020
#define STUNATTR_ICE_PRIORITY 0x0024 //ICE
#define STUNATTR_ICE_USE_CANDIDATE 0x0025 //ICE
//0x8000 attributes are optional, and may be silently ignored without issue.
#define STUNATTR_ADDITIONAL_ADDRFAM 0x8000 //TURN -- listen for ipv6 in addition to ipv4
#define STUNATTR_SOFTWARE 0x8022 //TURN
#define STUNATTR_FINGERPRINT 0x8028
#define STUNATTR_ICE_CONTROLLED 0x8029 //ICE
#define STUNATTR_ICE_CONTROLLING 0x802A //ICE
typedef struct
{
unsigned short attrtype;
unsigned short attrlen;
} stunattr_t;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#if defined(SUPPORT_ICE) || defined(MASTERONLY)
#include "zlib.h"
#endif
#ifdef SUPPORT_ICE
static cvar_t net_ice_exchangeprivateips = CVARFD("net_ice_exchangeprivateips", "", CVAR_NOTFROMSERVER, "Boolean. When set to 0, hides private IP addresses from your peers. Only addresses determined from the other side of your router will be shared. Setting it to 0 may be desirable but it can cause connections to fail when your router does not support hairpinning, whereas 1 fixes that at the cost of exposing private IP addresses.");
static cvar_t net_ice_allowstun = CVARFD("net_ice_allowstun", "1", CVAR_NOTFROMSERVER, "Boolean. When set to 0, prevents the use of stun to determine our public address (does not prevent connecting to our peer's server-reflexive candidates).");
static cvar_t net_ice_allowturn = CVARFD("net_ice_allowturn", "1", CVAR_NOTFROMSERVER, "Boolean. When set to 0, prevents registration of turn connections (does not prevent connecting to our peer's relay candidates).");
static cvar_t net_ice_relayonly = CVARFD("net_ice_relayonly", "0", CVAR_NOTFROMSERVER, "Boolean. When set to 1, blocks reporting non-relay local candidates, does not attempt to connect to remote candidates other than via a relay.");
static cvar_t net_ice_usewebrtc = CVARFD("net_ice_usewebrtc", "", CVAR_NOTFROMSERVER, "Use webrtc's extra overheads rather than simple ICE. This makes packets larger and is slower to connect, but is compatible with the web port.");
static cvar_t net_ice_servers = CVARFD("net_ice_servers", "", CVAR_NOTFROMSERVER, "A space-separated list of ICE servers, eg stun:host.example:3478 or turn:host.example:3478?user=foo?auth=blah");
static cvar_t net_ice_debug = CVARFD("net_ice_debug", "0", CVAR_NOTFROMSERVER, "0: Hide messy details.\n1: Show new candidates.\n2: Show each connectivity test.");
#if 0 //enable this for testing only. will generate excessive error messages with non-hacked turn servers...
#define ASCOPE_TURN_REQUIRESCOPE ASCOPE_HOST //try sending it any address.
#define ASCOPE_LCANDIDATE ASCOPE_HOST //report ALL addresses
#else
#define ASCOPE_TURN_REQUIRESCOPE ASCOPE_LAN //don't report loopback/link-local addresses to turn relays.
#define ASCOPE_LCANDIDATE ASCOPE_LAN //don't report loopback/link-local addresses over the internet.
//FIXME: replace with ASCOPE_NET and report mDNS .local names instead
#endif
/*
Interactive Connectivity Establishment (rfc 5245)
find out your peer's potential ports.
spam your peer with stun packets.
see what sticks.
the 'controller' assigns some final candidate pair to ensure that both peers send+receive from a single connection.
if no candidates are available, try using stun to find public nat addresses.
in fte, a 'pair' is actually in terms of each local socket and remote address. hopefully that won't cause too much weirdness.
this does limit which interfaces we can send packets from, which may cause issues with local TURN relays(does not result in extra prflx candidates) and VPNs(may need to be set up as the default route), and prevents us from being able to report reladdr in candidate offers (again mostly only of use with TURN)
lan connections should resolve to a host interface anyway
stun test packets must contain all sorts of info. username+integrity+fingerprint for validation. priority+usecandidate+icecontrol(ing) to decree the priority of any new remote candidates, whether its finished, and just who decides whether its finished.
peers don't like it when those are missing.
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
host candidates - addresses that are directly known (but are probably unroutable private things)
server reflexive candidates - addresses that we found from some public stun server (useful for NATs that use a single public port for each unique private port)
peer reflexive candidates - addresses that our peer finds out about as we spam them
relayed candidates - some sort of socks5 or something proxy.
Note: Even after the ICE connection becomes active, you should continue to collect local candidates and transmit them to the peer out of band.
this allows the connection to pick a new route if some router dies (like a relay kicking us).
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
FIXME: the client currently disconnects from the broker. the server tracks players via ip rather than ICE.
tcp rtp framing should generally be done with a 16-bit network-endian length prefix followed by the data.
*/
struct icecandidate_s
{
struct icecandinfo_s info;
struct icecandidate_s *next;
netadr_t peer;
//peer needs telling or something.
qboolean dirty;
//these are bitmasks. one bit for each local socket.
unsigned int reachable; //looked like it was up a while ago...
unsigned int reached; //timestamp of when it was last reachable. pick a new one if too old.
unsigned int tried; //don't try again this round
};
struct icestate_s
{
struct icestate_s *next;
void *module;
netadr_t qadr; //address reported to the rest of the engine (packets from our peer get remapped to this)
netadr_t chosenpeer; //address we're sending our data to.
struct iceserver_s
{ //stun/turn servers
netadr_t addr;
qboolean isstun;
unsigned int stunretry; //once a second, extended to once a minite on reply
unsigned int stunrnd[3];
//turn stuff.
ftenet_generic_connection_t *con; //TURN needs unique client addresses otherwise it gets confused and starts reporting errors about client ids. make sure these connections are unique.
//FIXME: should be able to get away with only one if the addr field is unique... move to icestate_s
netadrtype_t family; //ipv4 or ipv6. can't send to other peers.
char *user, *auth;
enum {
TURN_UNINITED, //never tried to poke server, need to send a quicky allocate request.
TURN_HAVE_NONCE, //nonce should be valid, need to send a full allocate request.
TURN_ALLOCATED, //we have an allocation that we need to refresh every now and then.
TURN_TERMINATING, //we have an allocation that we need to refresh every now and then.
} state;
unsigned int expires; //allocation expiry time.
char *nonce, *realm; //gathered from the server
// netadr_t relay, srflx; //our relayed port, and our local address for the sake of it.
unsigned int peers;
struct
{
unsigned int expires; //once a second, extended to once a minite on reply
unsigned int retry; //gotta keep retrying
unsigned int stunrnd[3]; //so we know when to stop retrying.
struct icecandidate_s *rc; //if its not rc then we need to reauth now.
} peer[32];
} server[8];
unsigned int servers;
unsigned int timeout; //time when we consider the connection dead
unsigned int keepalive; //sent periodically...
unsigned int retries; //bumped after each round of connectivity checks. affects future intervals.
enum iceproto_e proto;
enum icemode_e mode;
qboolean initiator; //sends the initial sdpoffer.
qboolean controlled; //controller chooses final ports.
enum icestate_e state;
char *conname; //internal id.
char *friendlyname; //who you're talking to.
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
unsigned int originid; //should be randomish
unsigned int originversion;//bumped each time something in the sdp changes.
char originaddress[16];
struct icecandidate_s *lc;
char *lpwd;
char *lufrag;
struct icecandidate_s *rc;
char *rpwd;
char *rufrag;
unsigned int tiehigh;
unsigned int tielow;
int foundation;
qboolean blockcandidates; //don't send candidates yet.
#ifdef HAVE_DTLS
void *dtlsstate;
struct sctp_s *sctp; //ffs! extra processing needed.
const dtlsfuncs_t *dtlsfuncs;
qboolean dtlspassive; //true=server, false=client (separate from ice controller and whether we're hosting. yay...)
dtlscred_t cred; //credentials info for dtls (both peer and local info)
quint16_t mysctpport;
quint16_t peersctpport;
qboolean sctpoptional;
qboolean peersctpoptional; //still uses dtls at least.
#endif
ftenet_connections_t *connections; //used only for PRIVATE sockets.
struct icecodecslot_s
{
//FIXME: we should probably include decode state in here somehow so multiple connections don't clobber each other.
int id;
char *name;
} codecslot[34]; //96-127. don't really need to care about other ones.
};
typedef struct sctp_s
{
quint16_t myport, peerport;
qboolean peerhasfwdtsn;
double nextreinit;
void *cookie; //sent part of the handshake (might need resending).
size_t cookiesize;
struct
{
quint32_t verifycode;
qboolean writable;
quint32_t tsn; //Transmit Sequence Number
quint32_t ctsn; //acked tsn
quint32_t losttsn; //total gap size...
} o;
struct
{
quint32_t verifycode;
int ackneeded;
quint32_t ctsn;
quint32_t htsn; //so we don't have to walk so many packets to count gaps.
#define SCTP_RCVSIZE 2048 //cannot be bigger than 65536
qbyte received[SCTP_RCVSIZE/8];
struct
{
quint32_t firsttns; //so we only ack fragments that were complete.
quint32_t tsn; //if a continuation doesn't match, we drop it.
quint32_t ppid;
quint16_t sid;
quint16_t seq;
size_t size;
qboolean toobig;
qbyte buf[65536];
} r;
} i;
unsigned short qstreamid; //in network endian.
} sctp_t;
#ifdef HAVE_DTLS
static neterr_t SCTP_Transmit(sctp_t *sctp, struct icestate_s *peer, const void *data, size_t length);
#endif
static neterr_t ICE_Transmit(void *cbctx, const qbyte *data, size_t datasize);
static neterr_t TURN_Encapsulate(struct icestate_s *ice, netadr_t *to, const qbyte *data, size_t datasize);
static void TURN_AuthorisePeer(struct icestate_s *con, struct iceserver_s *srv, int peer);
static struct icestate_s *icelist;
static struct icecodecslot_s *ICE_GetCodecSlot(struct icestate_s *ice, int slot)
{
if (slot >= 96 && slot < 96+32)
return &ice->codecslot[slot-96];
else if (slot == 0)
return &ice->codecslot[32];
else if (slot == 8)
return &ice->codecslot[33];
return NULL;
}
#if !defined(SERVERONLY) && defined(VOICECHAT)
extern cvar_t snd_voip_send;
struct rtpheader_s
{
unsigned char v2_p1_x1_cc4;
unsigned char m1_pt7;
unsigned short seq;
unsigned int timestamp;
unsigned int ssrc;
unsigned int csrc[1]; //sized according to cc
};
void S_Voip_RTP_Parse(unsigned short sequence, const char *codec, const unsigned char *data, unsigned int datalen);
qboolean S_Voip_RTP_CodecOkay(const char *codec);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static qboolean NET_RTP_Parse(void)
{
struct rtpheader_s *rtpheader = (void*)net_message.data;
if (net_message.cursize >= sizeof(*rtpheader) && (rtpheader->v2_p1_x1_cc4 & 0xc0) == 0x80)
{
int hlen;
int padding = 0;
struct icestate_s *con;
//make sure this really came from an accepted rtp stream
//note that an rtp connection equal to the game connection will likely mess up when sequences start to get big
//(especially problematic in sane clients that start with a random sequence)
for (con = icelist; con; con = con->next)
{
if (con->state != ICE_INACTIVE && (con->proto == ICEP_VIDEO || con->proto == ICEP_VOICE) && NET_CompareAdr(&net_from, &con->chosenpeer))
{
struct icecodecslot_s *codec = ICE_GetCodecSlot(con, rtpheader->m1_pt7 & 0x7f);
if (codec) //untracked slot
{
char *codecname = codec->name;
if (!codecname) //inactive slot
continue;
if (rtpheader->v2_p1_x1_cc4 & 0x20)
padding = net_message.data[net_message.cursize-1];
hlen = sizeof(*rtpheader);
hlen += ((rtpheader->v2_p1_x1_cc4 & 0xf)-1) * sizeof(int);
if (con->proto == ICEP_VOICE)
S_Voip_RTP_Parse((unsigned short)BigShort(rtpheader->seq), codecname, hlen+(char*)(rtpheader), net_message.cursize - padding - hlen);
// if (con->proto == ICEP_VIDEO)
// S_Video_RTP_Parse((unsigned short)BigShort(rtpheader->seq), codecname, hlen+(char*)(rtpheader), net_message.cursize - padding - hlen);
return true;
}
}
}
}
return false;
}
qboolean NET_RTP_Active(void)
{
struct icestate_s *con;
for (con = icelist; con; con = con->next)
{
if (con->state == ICE_CONNECTED && con->proto == ICEP_VOICE)
return true;
}
return false;
}
qboolean NET_RTP_Transmit(unsigned int sequence, unsigned int timestamp, const char *codec, char *cdata, int clength)
{
sizebuf_t buf;
char pdata[512];
int i;
struct icestate_s *con;
qboolean built = false;
memset(&buf, 0, sizeof(buf));
buf.maxsize = sizeof(pdata);
buf.cursize = 0;
buf.allowoverflow = true;
buf.data = pdata;
for (con = icelist; con; con = con->next)
{
if (con->state == ICE_CONNECTED && con->proto == ICEP_VOICE)
{
for (i = 0; i < countof(con->codecslot); i++)
{
if (con->codecslot[i].name && !strcmp(con->codecslot[i].name, codec))
{
if (!built)
{
built = true;
MSG_WriteByte(&buf, (2u<<6) | (0u<<5) | (0u<<4) | (0<<0)); //v2_p1_x1_cc4
MSG_WriteByte(&buf, (0u<<7) | (con->codecslot[i].id<<0)); //m1_pt7
MSG_WriteShort(&buf, BigShort(sequence&0xffff)); //seq
MSG_WriteLong(&buf, BigLong(timestamp)); //timestamp
MSG_WriteLong(&buf, BigLong(0)); //ssrc
SZ_Write(&buf, cdata, clength);
if (buf.overflowed)
return built;
}
ICE_Transmit(con, buf.data, buf.cursize);
break;
}
}
}
}
return built;
}
#endif
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static struct icestate_s *QDECL ICE_Find(void *module, const char *conname)
{
struct icestate_s *con;
for (con = icelist; con; con = con->next)
{
if (con->module == module && !strcmp(con->conname, conname))
return con;
}
return NULL;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static ftenet_connections_t *ICE_PickConnection(struct icestate_s *con)
{
if (con->connections)
return con->connections;
switch(con->proto)
{
default:
break;
#ifndef SERVERONLY
case ICEP_VOICE:
case ICEP_QWCLIENT:
return cls.sockets;
#endif
#ifndef CLIENTONLY
case ICEP_QWSERVER:
return svs.sockets;
#endif
}
return NULL;
}
static qboolean TURN_AddXorAddressAttrib(sizebuf_t *buf, unsigned int attr, netadr_t *to)
{ //12 or 24 bytes.
int alen, atype, aofs, i;
if (to->type == NA_IP)
{
alen = 4;
atype = 1;
aofs = 0;
}
else if (to->type == NA_IPV6 &&
!*(int*)&to->address.ip6[0] &&
!*(int*)&to->address.ip6[4] &&
!*(short*)&to->address.ip6[8] &&
*(short*)&to->address.ip6[10] == (short)0xffff)
{ //just because we use an ipv6 address for ipv4 internally doesn't mean we should tell the peer that they're on ipv6...
alen = 4;
atype = 1;
aofs = sizeof(to->address.ip6) - sizeof(to->address.ip);
}
else if (to->type == NA_IPV6)
{
alen = 16;
atype = 2;
aofs = 0;
}
else
return false;
MSG_WriteShort(buf, BigShort(attr));
MSG_WriteShort(buf, BigShort(4+alen));
MSG_WriteShort(buf, BigShort(atype));
MSG_WriteShort(buf, to->port ^ *(short*)(buf->data+4));
for (i = 0; i < alen; i++)
MSG_WriteByte(buf, ((qbyte*)&to->address)[aofs+i] ^ (buf->data+4)[i]);
return true;
}
#define FTEENGINE
#include "../../plugins/emailnot/md5.c"
static qboolean TURN_AddAuth(sizebuf_t *buf, struct iceserver_s *srv)
{ //adds auth info to a stun packet
unsigned short len;
char integrity[20];
if (!srv->user || !srv->nonce || !srv->realm)
return false;
MSG_WriteShort(buf, BigShort(STUNATTR_USERNAME));
len = strlen(srv->user);
MSG_WriteShort(buf, BigShort(len));
SZ_Write (buf, srv->user, len);
if (len&3)
SZ_Write (buf, "\0\0\0\0", 4-(len&3));
MSG_WriteShort(buf, BigShort(STUNATTR_REALM));
len = strlen(srv->realm);
MSG_WriteShort(buf, BigShort(len));
SZ_Write (buf, srv->realm, len);
if (len&3)
SZ_Write (buf, "\0\0\0\0", 4-(len&3));
MSG_WriteShort(buf, BigShort(STUNATTR_NONCE));
len = strlen(srv->nonce);
MSG_WriteShort(buf, BigShort(len));
SZ_Write (buf, srv->nonce, len);
if (len&3)
SZ_Write (buf, "\0\0\0\0", 4-(len&3));
//message integrity is a bit annoying
buf->data[2] = ((buf->cursize+4+sizeof(integrity)-20)>>8)&0xff; //hashed header length is up to the end of the hmac attribute
buf->data[3] = ((buf->cursize+4+sizeof(integrity)-20)>>0)&0xff;
//but the hash is to the start of the attribute's header
{ //long-term credentials do stuff weird.
char *tmpkey = va("%s:%s:%s", srv->user, srv->realm, srv->auth);
MD5_ToBinary(tmpkey, strlen(tmpkey), integrity, 16);
}
CalcHMAC(&hash_sha1, integrity, sizeof(integrity), buf->data, buf->cursize, integrity,16);
MSG_WriteShort(buf, BigShort(STUNATTR_MESSAGEINTEGRITIY));
MSG_WriteShort(buf, BigShort(sizeof(integrity))); //sha1 key length
SZ_Write(buf, integrity, sizeof(integrity)); //integrity data
return true;
}
static const char *ICE_NetworkToName(struct icestate_s *ice, int network)
{
network -= 1;
if (network >= 0 && network < MAX_CONNECTIONS)
{ //return the cvar name
ftenet_connections_t *col = ICE_PickConnection(ice);
if (col && col->conn[network])
return col->conn[network]->name;
}
else if (network >= MAX_CONNECTIONS)
{
network -= MAX_CONNECTIONS;
if (network >= countof(ice->server))
{ //a peer-reflexive address from poking a TURN server...
network -= countof(ice->server);
if (network < ice->servers)
return "turn-reflexive";
}
else
return va("turn:%s", ice->server[network].realm);
}
return "<UNKNOWN>";
}
static neterr_t TURN_Encapsulate(struct icestate_s *ice, netadr_t *to, const qbyte *data, size_t datasize)
{
sizebuf_t buf;
unsigned int network = to->connum-1;
struct iceserver_s *srv;
if (to->type == NA_INVALID)
return NETERR_NOROUTE;
if (network >= MAX_CONNECTIONS)
{ //fancy turn-related gubbins
network -= MAX_CONNECTIONS;
if (network >= countof(ice->server))
{ //really high, its from the raw socket, unstunned.
network -= countof(ice->server);
if (network >= countof(ice->server))
{
Con_Printf("TURN_Encapsulate: %i\n", to->connum);
return NETERR_NOROUTE;
}
srv = &ice->server[network];
if (!srv->con || net_ice_relayonly.ival)
return NETERR_CLOGGED;
return srv->con->SendPacket(srv->con, datasize, data, &srv->addr);
}
srv = &ice->server[network];
memset(&buf, 0, sizeof(buf));
buf.maxsize = 20+//stun header
8+16+//(max)peeraddr
4+((datasize+3)&~3);//data
buf.cursize = 0;
buf.data = alloca(buf.maxsize);
MSG_WriteShort(&buf, BigShort(STUN_INDICATION|STUN_SEND));
MSG_WriteShort(&buf, 0); //fill in later
MSG_WriteLong(&buf, BigLong(STUN_MAGIC_COOKIE));
MSG_WriteLong(&buf, 0); //randomid
MSG_WriteLong(&buf, 0); //randomid
MSG_WriteLong(&buf, 0); //randomid
if (!TURN_AddXorAddressAttrib(&buf, STUNATTR_XOR_PEER_ADDRESS, to))
return NETERR_NOROUTE;
MSG_WriteShort(&buf, BigShort(STUNATTR_DATA));
MSG_WriteShort(&buf, BigShort(datasize));
SZ_Write(&buf, data, datasize);
if (datasize&3)
SZ_Write(&buf, "\0\0\0\0", 4-(datasize&3));
//fill in the length (for the final time, after filling in the integrity and fingerprint)
buf.data[2] = ((buf.cursize-20)>>8)&0xff;
buf.data[3] = ((buf.cursize-20)>>0)&0xff;
if (!srv->con)
return NETERR_CLOGGED;
return srv->con->SendPacket(srv->con, buf.cursize, buf.data, &srv->addr);
}
if (net_ice_relayonly.ival)
return NETERR_CLOGGED;
return NET_SendPacket(ICE_PickConnection(ice), datasize, data, to);
}
static neterr_t ICE_Transmit(void *cbctx, const qbyte *data, size_t datasize)
{
struct icestate_s *ice = cbctx;
if (ice->chosenpeer.type == NA_INVALID)
{
if (ice->state == ICE_FAILED)
return NETERR_DISCONNECTED;
else
return NETERR_CLOGGED;
}
return TURN_Encapsulate(ice, &ice->chosenpeer, data, datasize);
}
static struct icestate_s *QDECL ICE_Create(void *module, const char *conname, const char *peername, enum icemode_e mode, enum iceproto_e proto, qboolean initiator)
{
ftenet_connections_t *collection;
struct icestate_s *con;
static unsigned int icenum;
//only allow modes that we actually support.
if (mode != ICEM_RAW && mode != ICEM_ICE && mode != ICEM_WEBRTC)
return NULL;
#ifndef HAVE_DTLS
if (mode == ICEM_WEBRTC)
return NULL;
#endif
//only allow protocols that we actually support.
switch(proto)
{
default:
return NULL;
// case ICEP_VIDEO:
// collection = NULL;
// break;
#if !defined(SERVERONLY) && defined(VOICECHAT)
case ICEP_VOICE:
case ICEP_VIDEO:
collection = cls.sockets;
if (!collection)
{
NET_InitClient(false);
collection = cls.sockets;
}
break;
#endif
#ifndef SERVERONLY
case ICEP_QWCLIENT:
collection = cls.sockets;
if (!collection)
{
NET_InitClient(false);
collection = cls.sockets;
}
break;
#endif
#ifndef CLIENTONLY
case ICEP_QWSERVER:
collection = svs.sockets;
break;
#endif
}
if (!collection)
return NULL; //not initable or something
if (!conname)
{
int rnd[2];
Sys_RandomBytes((void*)rnd, sizeof(rnd));
conname = va("fte%08x%08x", rnd[0], rnd[1]);
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (ICE_Find(module, conname))
return NULL; //don't allow dupes.
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
con = Z_Malloc(sizeof(*con));
con->conname = Z_StrDup(conname);
con->friendlyname = peername?Z_StrDup(peername):Z_StrDupf("%i", icenum++);
con->proto = proto;
con->rpwd = Z_StrDup("");
con->rufrag = Z_StrDup("");
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
Sys_RandomBytes((void*)&con->originid, sizeof(con->originid));
con->originversion = 1;
Q_strncpyz(con->originaddress, "127.0.0.1", sizeof(con->originaddress));
con->initiator = initiator;
con->controlled = !initiator;
con->blockcandidates = true; //until offers/answers are sent.
#ifdef HAVE_DTLS
con->dtlspassive = (proto == ICEP_QWSERVER); //note: may change later.
if (mode == ICEM_WEBRTC)
{ //dtls+sctp is a mandatory part of our connection, sadly.
if (!con->dtlsfuncs)
{
if (con->dtlspassive)
con->dtlsfuncs = DTLS_InitServer();
else
con->dtlsfuncs = DTLS_InitClient(); //credentials are a bit different, though fingerprints make it somewhat irrelevant.
}
if (con->dtlsfuncs && con->dtlsfuncs->GenTempCertificate && !con->cred.local.certsize)
{
Con_DPrintf("Generating dtls certificate...\n");
if (!con->dtlsfuncs->GenTempCertificate(NULL, &con->cred.local))
{
con->dtlsfuncs = NULL;
mode = ICEM_ICE;
Con_DPrintf("Failed\n");
}
else
Con_DPrintf("Done\n");
}
else
{ //failure if we can't do the whole dtls thing.
con->dtlsfuncs = NULL;
Con_Printf(CON_WARNING"DTLS %s support unavailable, disabling encryption (and webrtc compat).\n", con->dtlspassive?"server":"client");
mode = ICEM_ICE; //fall back on unencrypted (this doesn't depend on the peer, so while shitty it hopefully shouldn't be exploitable with a downgrade-attack)
}
if (mode == ICEM_WEBRTC && !net_ice_usewebrtc.ival)
con->sctpoptional = true;
if (mode == ICEM_WEBRTC)
con->mysctpport = 27500;
}
con->qadr.port = htons(con->mysctpport);
#endif
con->qadr.type = NA_ICE;
con->qadr.prot = NP_DGRAM;
Q_strncpyz(con->qadr.address.icename, con->friendlyname, sizeof(con->qadr.address.icename));
con->mode = mode;
con->next = icelist;
icelist = con;
{
int rnd[1]; //'must have at least 24 bits randomness'
Sys_RandomBytes((void*)rnd, sizeof(rnd));
con->lufrag = Z_StrDupf("%08x", rnd[0]);
}
{
int rnd[4]; //'must have at least 128 bits randomness'
Sys_RandomBytes((void*)rnd, sizeof(rnd));
con->lpwd = Z_StrDupf("%08x%08x%08x%08x", rnd[0], rnd[1], rnd[2], rnd[3]);
}
Sys_RandomBytes((void*)&con->tiehigh, sizeof(con->tiehigh));
Sys_RandomBytes((void*)&con->tielow, sizeof(con->tielow));
if (collection)
{
int i, m;
netadr_t addr[64];
struct ftenet_generic_connection_s *gcon[sizeof(addr)/sizeof(addr[0])];
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
unsigned int flags[sizeof(addr)/sizeof(addr[0])];
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
const char *params[sizeof(addr)/sizeof(addr[0])];
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
m = NET_EnumerateAddresses(collection, gcon, flags, addr, params, sizeof(addr)/sizeof(addr[0]));
for (i = 0; i < m; i++)
{
if (addr[i].type == NA_IP || addr[i].type == NA_IPV6)
{
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
// if (flags[i] & ADDR_REFLEX)
// ICE_AddLCandidateInfo(con, &addr[i], i, ICE_SRFLX); //FIXME: needs reladdr relport info
// else
ICE_AddLCandidateInfo(con, &addr[i], i, ICE_HOST);
}
}
}
return con;
}
//if either remotecand is null, new packets will be sent to all.
static qboolean ICE_SendSpam(struct icestate_s *con)
{
struct icecandidate_s *rc;
int i;
int bestlocal = -1;
struct icecandidate_s *bestpeer = NULL;
ftenet_connections_t *collection = ICE_PickConnection(con);
if (!collection)
return false;
//only send one ping to each.
for(rc = con->rc; rc; rc = rc->next)
{
for (i = 0; i < MAX_CONNECTIONS; i++)
{
if (collection->conn[i] && collection->conn[i]->prot == NP_DGRAM && (collection->conn[i]->addrtype[0]==NA_IP||collection->conn[i]->addrtype[0]==NA_IPV6))
{
if (!(rc->tried & (1u<<i)))
{
if (!bestpeer || bestpeer->info.priority < rc->info.priority)
{
if (NET_ClassifyAddress(&rc->peer, NULL) < ASCOPE_TURN_REQUIRESCOPE)
{ //don't waste time asking the relay to poke its loopback. if only because it'll report lots of errors.
rc->tried |= (1u<<i);
continue;
}
bestpeer = rc;
bestlocal = i;
}
}
}
}
//send via appropriate turn servers
if (rc->info.type == ICE_RELAY || rc->info.type == ICE_SRFLX)
{
for (i = 0; i < con->servers; i++)
{
if (con->server[i].state!=TURN_ALLOCATED)
continue; //not ready yet...
if (!con->server[i].con)
continue; //can't...
if (!(rc->tried & (1u<<(MAX_CONNECTIONS+i))))
{
//fixme: no local priority. a multihomed machine will try the same ip from different ports.
if (!bestpeer || bestpeer->info.priority < rc->info.priority)
{
if (con->server[i].family && rc->peer.type != con->server[i].family)
continue; //if its ipv4-only then don't send it ipv6 packets or whatever.
bestpeer = rc;
bestlocal = MAX_CONNECTIONS+i;
}
}
}
}
}
if (bestpeer && bestlocal >= 0)
{
neterr_t err;
netadr_t to;
sizebuf_t buf;
char data[512];
char integ[20];
int crc;
qboolean usecandidate = false;
const char *candtype;
unsigned int priority;
memset(&buf, 0, sizeof(buf));
buf.maxsize = sizeof(data);
buf.cursize = 0;
buf.data = data;
bestpeer->tried |= (1u<<bestlocal);
to = bestpeer->peer;
to.connum = 1+bestlocal;
if (to.type == NA_INVALID)
return true; //erk?
safeswitch(bestpeer->info.type)
{
case ICE_HOST: candtype="host"; break;
case ICE_SRFLX: candtype="srflx"; break;
case ICE_PRFLX: candtype="prflx"; break;
case ICE_RELAY: candtype="relay"; break;
safedefault: candtype="?";
}
if (bestlocal >= MAX_CONNECTIONS)
{
struct iceserver_s *srv = &con->server[bestlocal-MAX_CONNECTIONS];
unsigned int i;
unsigned int now = Sys_Milliseconds();
for (i = 0; ; i++)
{
if (i == srv->peers)
{
if (i == countof(srv->peer))
return true;
srv->peer[i].expires = now;
srv->peer[i].rc = bestpeer;
Sys_RandomBytes((char*)srv->peer[i].stunrnd, sizeof(srv->peer[i].stunrnd));
srv->peer[i].retry = now;
srv->peers++;
}
if (srv->peer[i].rc == bestpeer)
{
break;
}
}
if ((int)(srv->peer[i].retry-now) <= 0)
{
srv->peer[i].retry = now + (con->state==ICE_CONNECTED?2000:50); //keep retrying till we get an ack. be less agressive once it no longer matters so much
TURN_AuthorisePeer(con, srv, i);
}
}
if (!con->controlled && NET_CompareAdr(&to, &con->chosenpeer) && to.connum == con->chosenpeer.connum)
usecandidate = true;
MSG_WriteShort(&buf, BigShort(STUN_BINDING));
MSG_WriteShort(&buf, 0); //fill in later
MSG_WriteLong(&buf, BigLong(STUN_MAGIC_COOKIE)); //magic
MSG_WriteLong(&buf, BigLong(0)); //randomid
MSG_WriteLong(&buf, BigLong(0)); //randomid
MSG_WriteLong(&buf, BigLong(0x80000000|bestlocal)); //randomid
if (usecandidate)
{
MSG_WriteShort(&buf, BigShort(STUNATTR_ICE_USE_CANDIDATE));//ICE-USE-CANDIDATE
MSG_WriteShort(&buf, BigShort(0)); //just a flag, so no payload to this attribute
}
//username
MSG_WriteShort(&buf, BigShort(STUNATTR_USERNAME)); //USERNAME
MSG_WriteShort(&buf, BigShort(strlen(con->rufrag) + 1 + strlen(con->lufrag)));
SZ_Write(&buf, con->rufrag, strlen(con->rufrag));
MSG_WriteChar(&buf, ':');
SZ_Write(&buf, con->lufrag, strlen(con->lufrag));
while(buf.cursize&3)
MSG_WriteChar(&buf, 0); //pad
//priority
priority =
//FIXME. should be set to:
// priority = (2^24)*(type preference) +
// (2^8)*(local preference) +
// (2^0)*(256 - component ID)
//type preference should be 126 and is a function of the candidate type (direct sending should be highest priority at 126)
//local preference should reflect multi-homed preferences. ipv4+ipv6 count as multihomed.
//component ID should be 1 (rtcp would be 2 if we supported it)
(1<<24)*(bestlocal>=MAX_CONNECTIONS?0:126) +
(1<<8)*((bestpeer->peer.type == NA_IP?32768:0)+bestlocal*256+(255/*-adrno*/)) +
(1<<0)*(256 - bestpeer->info.component);
MSG_WriteShort(&buf, BigShort(STUNATTR_ICE_PRIORITY));//ICE-PRIORITY
MSG_WriteShort(&buf, BigShort(4));
MSG_WriteLong(&buf, BigLong(priority));
//these two attributes carry a random 64bit tie-breaker.
//the controller is the one with the highest number.
if (con->controlled)
{
MSG_WriteShort(&buf, BigShort(STUNATTR_ICE_CONTROLLED));//ICE-CONTROLLED
MSG_WriteShort(&buf, BigShort(8));
MSG_WriteLong(&buf, BigLong(con->tiehigh));
MSG_WriteLong(&buf, BigLong(con->tielow));
}
else
{
MSG_WriteShort(&buf, BigShort(STUNATTR_ICE_CONTROLLING));//ICE-CONTROLLING
MSG_WriteShort(&buf, BigShort(8));
MSG_WriteLong(&buf, BigLong(con->tiehigh));
MSG_WriteLong(&buf, BigLong(con->tielow));
}
//message integrity is a bit annoying
data[2] = ((buf.cursize+4+sizeof(integ)-20)>>8)&0xff; //hashed header length is up to the end of the hmac attribute
data[3] = ((buf.cursize+4+sizeof(integ)-20)>>0)&0xff;
//but the hash is to the start of the attribute's header
openxr plugin: tweaked - inputs should be working properly now, and are visible to csqc. subject to further breaking changes, however. _pext_vrinputs: added cvar to enable vr inputs protocol extension allowing vr inputs to be networked to ssqc too. defaults to 0 for now, will be renamed when deemed final. updates menu: the prompt to enable sources is now more explicit instead of expecting the user to have a clue. updates menu: added a v3 sources format, which should be more maintainable. not final. updates menu: try to give reasons why sources might be failing (to help blame ISPs if they try fucking over TTH dns again). presets menu: no longer closes the instant a preset is chosen. some presets have a couple of modifiers listed. force the demo loop in the background to serve as a preview. prompts menus: now does word wrapping. ftemaster: support importing server lists from other master servers (requested by Eukara). server: try to detect when non-reply inbound packets are blocked by firewalls/nats/etc (using ftemaster to do so). qcvm: added pointcontentsmask builtin, allowing it to probe more than just world, with fte's full contentbit range instead of just q1 legacy. qcvm: memfill8 builtin now works on createbuffer() pointers. qcvm: add missing unsigned ops. Fixed double comparison ops. fixed bug with op_store_i64. added missing OP_LOADP_I64 qcc: added '#pragma framerate RATE' for overriding implicit nextthink durations. qcc: fixed '#pragma DONT_COMPILE_THIS_FILE' to not screw up comments. qcc: added __GITURL__ __GITHASH__ __GITDATE__ __GITDATETIME__ __GITDESC__ for any mods that might want to make use of that. qcc: fix up -Fhashonly a little setrenderer: support for vulkan gpu enumeration. rulesets: reworked to support custom rulesets (using hashes to catch haxxors, though still nothing prevents just changing the client to ignore rulesets) bspx: use our BIH code for the bspx BRUSHLIST lump instead of the older less efficient code. (static)iqm+obj: these model formats can now be used for the worldmodel (with a suitable .ent file). Also using BIH for much better collision performance. pmove: tried to optimise PM_NudgePosition, should boost fps in stress tests. wayland: fix a crash on startup. mousegrabs now works better. imagetool: uses sdl for previews. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5813 fc73d0e0-1445-4013-8a0c-d673dee63da5
2021-04-14 05:21:04 +00:00
CalcHMAC(&hash_sha1, integ, sizeof(integ), data, buf.cursize, con->rpwd, strlen(con->rpwd));
MSG_WriteShort(&buf, BigShort(STUNATTR_MESSAGEINTEGRITIY)); //MESSAGE-INTEGRITY
MSG_WriteShort(&buf, BigShort(20)); //sha1 key length
SZ_Write(&buf, integ, sizeof(integ)); //integrity data
data[2] = ((buf.cursize+8-20)>>8)&0xff; //dummy length
data[3] = ((buf.cursize+8-20)>>0)&0xff;
crc = crc32(0, data, buf.cursize)^0x5354554e;
MSG_WriteShort(&buf, BigShort(STUNATTR_FINGERPRINT)); //FINGERPRINT
MSG_WriteShort(&buf, BigShort(sizeof(crc)));
MSG_WriteLong(&buf, BigLong(crc));
//fill in the length (for the fourth time, after filling in the integrity and fingerprint)
data[2] = ((buf.cursize-20)>>8)&0xff;
data[3] = ((buf.cursize-20)>>0)&0xff;
err = TURN_Encapsulate(con, &to, buf.data, buf.cursize);
switch(err)
{
case NETERR_SENT:
if (net_ice_debug.ival >= 2)
Con_Printf(S_COLOR_GRAY"[%s]: checking %s -> %s:%i (%s)\n", con->friendlyname, ICE_NetworkToName(con, to.connum), bestpeer->info.addr, bestpeer->info.port, candtype);
break;
case NETERR_CLOGGED: //oh well... we retry anyway.
break;
default:
if (net_ice_debug.ival >= 2)
Con_Printf("ICE send error to %s:%i(%s)\n", bestpeer->info.addr, bestpeer->info.port, candtype);
break;
}
return true;
}
return false;
}
extern ftenet_generic_connection_t *FTENET_Datagram_EstablishConnection(ftenet_connections_t *col, const char *address, netadr_t adr);
#ifdef TCPCONNECT
struct turntcp_connection_s
{ //this sends packets only to the relay, and accepts them only from there too. all packets must be stun packets (for byte-count framing)
ftenet_generic_connection_t pub;
vfsfile_t *f;
struct
{
qbyte buf[20+65536];
unsigned int offset;
unsigned int avail;
} recv, send;
netadr_t adr;
};
static qboolean TURN_TCP_GetPacket(struct ftenet_generic_connection_s *con)
{
struct turntcp_connection_s *n = (struct turntcp_connection_s*)con;
qboolean tried = false;
int err;
if (n->send.avail)
{
err = VFS_WRITE(n->f, n->send.buf+n->send.offset, n->send.avail);
if (err >= 0)
{
n->send.avail -= err;
n->send.offset += err;
}
}
for (;;)
{
if (n->recv.avail >= 20) //must be a stun packet...
{
unsigned int psize = 20 + (n->recv.buf[n->recv.offset+2]<<8)+n->recv.buf[n->recv.offset+3];
if (psize <= n->recv.avail)
{
memcpy(net_message.data, n->recv.buf+n->recv.offset, psize);
n->recv.avail -= psize;
n->recv.offset += psize;
net_message.cursize = psize;
net_from = n->adr;
return true;
}
}
if (n->recv.offset && n->recv.offset+n->recv.avail == sizeof(n->recv.buf))
{
memmove(n->recv.buf, n->recv.buf+n->recv.offset, n->recv.avail);
n->recv.offset = 0;
}
else if (tried)
return false; //don't infinitely loop.
err = VFS_READ(n->f, n->recv.buf+n->recv.offset+n->recv.avail, sizeof(n->recv.buf) - (n->recv.offset+n->recv.avail));
if (!err) //no data...
return false;
else if (err < 0)
return false;
else
{
tried = true;
n->recv.avail += err;
}
}
}
static neterr_t TURN_TCP_SendPacket(struct ftenet_generic_connection_s *con, int length, const void *data, netadr_t *to)
{
int err;
struct turntcp_connection_s *n = (struct turntcp_connection_s*)con;
if (!NET_CompareAdr(to, &n->adr))
return NETERR_NOROUTE;
//validate the packet - make sure its a TURN one. only checking size here cos we're lazy and that's enough.
if (length < 20 || length != 20 + (((const qbyte*)data)[2]<<8)+((const qbyte*)data)[3])
return NETERR_NOROUTE;
if (!n->send.avail && length < sizeof(n->send.buf))
{ //avoid copying if we have to
err = VFS_WRITE(n->f, data, length);
if (err >= 0 && err < length)
{ //but sometimes its partial.
data = (const char*)data + err;
length -= err;
n->send.offset = 0;
memcpy(n->send.buf, data, length);
n->send.avail = length;
}
if (!err)
return NETERR_CLOGGED; //mostly so we don't spam while still doing tcp/tls handshakes.
}
else
{
if ( n->send.avail+length > sizeof(n->send.buf))
return NETERR_CLOGGED; //can't possibly fit.
if (n->send.offset+n->send.avail+length > sizeof(n->send.buf))
{ //move it down if we need.
memmove(n->send.buf, n->send.buf+n->send.offset, n->send.avail);
n->send.offset = 0;
}
memcpy(n->send.buf+n->send.offset, data, length);
n->send.avail += length;
err = VFS_WRITE(n->f, n->send.buf+n->send.offset, n->send.avail);
if (err >= 0)
{
n->send.offset += err;
n->send.avail -= err;
}
}
if (err >= 0)
{
return NETERR_SENT; //sent something at least.
}
else
{
//one of the VFS_ERROR_* codes
return NETERR_DISCONNECTED;
}
}
static void TURN_TCP_Close(struct ftenet_generic_connection_s *con)
{
struct turntcp_connection_s *n = (struct turntcp_connection_s*)con;
if (n->f)
VFS_CLOSE(n->f);
}
static ftenet_generic_connection_t *TURN_TCP_EstablishConnection(ftenet_connections_t *col, const char *address, netadr_t adr)
{
struct turntcp_connection_s *n = Z_Malloc(sizeof(*n));
n->pub.thesocket = TCP_OpenStream(&adr, address);
n->f = FS_WrapTCPSocket(n->pub.thesocket, true, address);
#ifdef HAVE_SSL
//convert to tls...
if (adr.prot == NP_TLS || adr.prot == NP_RTC_TLS)
n->f = FS_OpenSSL(address, n->f, false);
#endif
if (!n->f)
{
Z_Free(n);
return NULL;
}
n->pub.owner = col;
n->pub.SendPacket = TURN_TCP_SendPacket;
n->pub.GetPacket = TURN_TCP_GetPacket;
n->pub.Close = TURN_TCP_Close;
n->adr = adr;
return &n->pub;
}
#endif
static void ICE_ToStunServer(struct icestate_s *con, struct iceserver_s *srv)
{
sizebuf_t buf;
char data[512];
int crc;
ftenet_connections_t *collection = ICE_PickConnection(con);
neterr_t err;
if (!collection)
return;
memset(&buf, 0, sizeof(buf));
buf.maxsize = sizeof(data);
buf.cursize = 0;
buf.data = data;
if (srv->isstun)
{
if (!net_ice_allowstun.ival || net_ice_relayonly.ival)
return;
if (net_ice_debug.ival >= 2)
Con_Printf(S_COLOR_GRAY"[%s]: STUN: Checking public IP via %s\n", con->friendlyname, NET_AdrToString(data, sizeof(data), &srv->addr));
MSG_WriteShort(&buf, BigShort(STUN_BINDING));
}
else
{
if (!net_ice_allowturn.ival)
{
if (net_ice_relayonly.ival)
{
Con_Printf("%s: forcing %s on\n", net_ice_relayonly.name, net_ice_allowturn.name);
Cvar_Set(&net_ice_allowturn, "1");
}
return;
}
if (!srv->con)
{
if (srv->addr.type == NA_INVALID)
return; //nope...
if (srv->addr.prot != NP_DGRAM)
srv->con = TURN_TCP_EstablishConnection(collection, srv->realm, srv->addr);
else
{
netadr_t localadr;
memset(&localadr, 0, sizeof(localadr));
localadr.type = srv->addr.type;
localadr.prot = srv->addr.prot;
srv->con = FTENET_Datagram_EstablishConnection(collection, srv->realm, localadr);
}
if (!srv->con)
{
srv->addr.type = NA_INVALID; //fail it.
return;
}
srv->con->connum = 1+MAX_CONNECTIONS+countof(con->server)+(srv-con->server); //*sigh*
}
if (srv->state==TURN_TERMINATING)
{
if (net_ice_debug.ival >= 2)
Con_Printf(S_COLOR_GRAY"[%s]: TURN: Terminating %s\n", con->friendlyname, NET_AdrToString(data, sizeof(data), &srv->addr));
MSG_WriteShort(&buf, BigShort(STUN_REFRESH));
}
else if (srv->state==TURN_ALLOCATED)
{
if (net_ice_debug.ival >= 2)
Con_Printf(S_COLOR_GRAY"[%s]: TURN: Refreshing %s\n", con->friendlyname, NET_AdrToString(data, sizeof(data), &srv->addr));
MSG_WriteShort(&buf, BigShort(STUN_REFRESH));
}
else
{
if (net_ice_debug.ival >= 2)
Con_Printf(S_COLOR_GRAY "[%s]: TURN: Allocating %s\n", con->friendlyname, NET_AdrToString(data, sizeof(data), &srv->addr));
MSG_WriteShort(&buf, BigShort(STUN_ALLOCATE));
}
}
Sys_RandomBytes((char*)srv->stunrnd, sizeof(srv->stunrnd));
MSG_WriteShort(&buf, 0); //fill in later
MSG_WriteLong(&buf, BigLong(STUN_MAGIC_COOKIE));
MSG_WriteLong(&buf, srv->stunrnd[0]); //randomid
MSG_WriteLong(&buf, srv->stunrnd[1]); //randomid
MSG_WriteLong(&buf, srv->stunrnd[2]); //randomid
if (!srv->isstun)
{
if (srv->state<TURN_ALLOCATED)
{
MSG_WriteShort(&buf, BigShort(STUNATTR_REQUESTED_TRANSPORT));
MSG_WriteShort(&buf, BigShort(4));
MSG_WriteLong(&buf, 17/*udp*/);
switch (srv->family)
{
case NA_IP:
//MSG_WriteShort(&buf, BigShort(STUNATTR_REQUESTED_ADDRFAM));
//MSG_WriteShort(&buf, BigShort(4));
//MSG_WriteLong(&buf, 1/*ipv4*/);
break;
case NA_IPV6:
MSG_WriteShort(&buf, BigShort(STUNATTR_REQUESTED_ADDRFAM));
MSG_WriteShort(&buf, BigShort(4));
MSG_WriteLong(&buf, 2/*ipv6*/);
break;
case NA_INVALID: //ask for both ipv4+ipv6.
MSG_WriteShort(&buf, BigShort(STUNATTR_ADDITIONAL_ADDRFAM));
MSG_WriteShort(&buf, BigShort(4));
MSG_WriteLong(&buf, 2/*ipv6*/);
break;
default:
return; //nope... not valid.
}
}
// MSG_WriteShort(&buf, BigShort(STUNATTR_DONT_FRAGMENT));
// MSG_WriteShort(&buf, BigShort(0));
/* MSG_WriteShort(&buf, BigShort(STUNATTR_SOFTWARE));
crc = strlen(FULLENGINENAME);
MSG_WriteShort(&buf, BigShort(crc));
SZ_Write (&buf, FULLENGINENAME, crc);
if (crc&3)
SZ_Write (&buf, "\0\0\0\0", 4-(crc&3));
*/
if (srv->state==TURN_TERMINATING)
{
MSG_WriteShort(&buf, BigShort(STUNATTR_LIFETIME));
MSG_WriteShort(&buf, BigShort(4));
MSG_WriteLong(&buf, 0);
}
else
{
// MSG_WriteShort(&buf, BigShort(STUNATTR_LIFETIME));
// MSG_WriteShort(&buf, BigShort(4));
// MSG_WriteLong(&buf, BigLong(300));
if (srv->state != TURN_UNINITED)
{
if (!TURN_AddAuth(&buf, srv))
return;
}
}
}
else
{
data[2] = ((buf.cursize+8-20)>>8)&0xff; //dummy length
data[3] = ((buf.cursize+8-20)>>0)&0xff;
crc = crc32(0, data, buf.cursize)^0x5354554e;
MSG_WriteShort(&buf, BigShort(STUNATTR_FINGERPRINT));
MSG_WriteShort(&buf, BigShort(sizeof(crc)));
MSG_WriteLong(&buf, BigLong(crc));
}
//fill in the length (for the final time, after filling in the integrity and fingerprint)
data[2] = ((buf.cursize-20)>>8)&0xff;
data[3] = ((buf.cursize-20)>>0)&0xff;
if (srv->isstun)
err = NET_SendPacket(collection, buf.cursize, data, &srv->addr);
else if (srv->con)
err = srv->con->SendPacket(srv->con, buf.cursize, data, &srv->addr);
else
return;
if (err == NETERR_CLOGGED)
srv->stunretry = Sys_Milliseconds(); //just keep retrying until it actually goes through.
}
static void TURN_AuthorisePeer(struct icestate_s *con, struct iceserver_s *srv, int peer)
{
struct icecandidate_s *rc = srv->peer[peer].rc;
sizebuf_t buf;
netadr_t to2;
char data[512];
if (srv->state != TURN_ALLOCATED)
return;
memset(&buf, 0, sizeof(buf));
buf.maxsize = sizeof(data);
buf.cursize = 0;
buf.data = data;
MSG_WriteShort(&buf, BigShort(STUN_CREATEPERM));
MSG_WriteShort(&buf, 0); //fill in later
MSG_WriteLong(&buf, BigLong(STUN_MAGIC_COOKIE)); //magic
MSG_WriteLong(&buf, srv->peer[peer].stunrnd[0]); //randomid
MSG_WriteLong(&buf, srv->peer[peer].stunrnd[1]); //randomid
MSG_WriteLong(&buf, srv->peer[peer].stunrnd[2]); //randomid
if (!TURN_AddXorAddressAttrib(&buf, STUNATTR_XOR_PEER_ADDRESS, &rc->peer))
return;
if (*rc->info.reladdr && strcmp(rc->info.addr, rc->info.reladdr)) //allow the relay to bypass the peer's relay if its different (TURN doesn't care about port permissions).
if (NET_StringToAdr(rc->info.reladdr, rc->info.relport, &to2))
TURN_AddXorAddressAttrib(&buf, STUNATTR_XOR_PEER_ADDRESS, &to2);
if (!TURN_AddAuth(&buf, srv))
return;
buf.data[2] = ((buf.cursize-20)>>8)&0xff;
buf.data[3] = ((buf.cursize-20)>>0)&0xff;
srv->con->SendPacket(srv->con, buf.cursize, buf.data, &srv->addr);
if (net_ice_debug.ival >= 1)
{
const char *candtype;
safeswitch(rc->info.type)
{
case ICE_HOST: candtype="host"; break;
case ICE_SRFLX: candtype="srflx"; break;
case ICE_PRFLX: candtype="prflx"; break;
case ICE_RELAY: candtype="relay"; break;
safedefault: candtype="?";
}
Con_Printf(S_COLOR_GRAY"[%s]: (re)registering %s -> %s:%i (%s)\n", con->friendlyname, srv->realm, rc->info.addr, rc->info.port, candtype);
}
}
int ParsePartialIP(const char *s, netadr_t *a);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static void QDECL ICE_AddRCandidateInfo(struct icestate_s *con, struct icecandinfo_s *n)
{
struct icecandidate_s *o;
qboolean isnew;
netadr_t peer;
int peerbits;
//I don't give a damn about rtpc.
if (n->component != 1)
return;
if (n->transport != 0)
return; //only UDP is supported.
//don't use the regular string->addr, because browsers seem to shove internal gibberish names in there that waste time to resolve. hostnames don't really make sense here anyway.
peerbits = ParsePartialIP(n->addr, &peer);
peer.prot = NP_DGRAM;
peer.port = htons(n->port);
if (peer.type == NA_IP && peerbits == 32)
{
//ignore invalid addresses
if (!peer.address.ip[0] && !peer.address.ip[1] && !peer.address.ip[2] && !peer.address.ip[3])
return;
}
else if (peer.type == NA_IPV6 && peerbits == 128)
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
{
//ignore invalid addresses
int i;
for (i = 0; i < countof(peer.address.ip6); i++)
if (peer.address.ip6[i])
break;
if (i == countof(peer.address.ip6))
return; //all clear. in6_addr_any
}
else
return; //bad address type, or partial.
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (*n->candidateid)
{
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
for (o = con->rc; o; o = o->next)
{
//not sure that updating candidates is particuarly useful tbh, but hey.
if (!strcmp(o->info.candidateid, n->candidateid))
break;
}
}
else
{
for (o = con->rc; o; o = o->next)
{
//avoid dupes.
if (!strcmp(o->info.addr, n->addr) && o->info.port == n->port)
break;
}
}
if (!o)
{
o = Z_Malloc(sizeof(*o));
o->next = con->rc;
con->rc = o;
Q_strncpyz(o->info.candidateid, n->candidateid, sizeof(o->info.candidateid));
isnew = true;
}
else
{
isnew = false;
}
Q_strncpyz(o->info.addr, n->addr, sizeof(o->info.addr));
o->info.port = n->port;
o->info.type = n->type;
o->info.priority = n->priority;
o->info.network = n->network;
o->info.generation = n->generation;
o->info.foundation = n->foundation;
o->info.component = n->component;
o->info.transport = n->transport;
o->dirty = true;
o->peer = peer;
o->tried = 0;
o->reachable = 0;
if (net_ice_debug.ival >= 1)
Con_Printf(S_COLOR_GRAY"[%s]: %s remote candidate %s: [%s]:%i\n", con->friendlyname, isnew?"Added":"Updated", o->info.candidateid, o->info.addr, o->info.port);
if (n->type == ICE_RELAY && *n->reladdr && (strcmp(n->addr, n->reladdr) || n->port != n->relport))
{ //for relay candidates, add an srflx candidate too.
struct icecandinfo_s t = o->info;
t.type = ICE_SRFLX;
strcpy(t.addr, n->reladdr);
t.port = n->relport;
*t.reladdr = 0;
t.relport = 0;
t.priority |= 1<<24; //nudge its priority up slightly to favour more direct links when we can.
*t.candidateid = 0; //anonymous...
ICE_AddRCandidateInfo(con, &t);
}
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static qboolean QDECL ICE_Set(struct icestate_s *con, const char *prop, const char *value);
static void ICE_ParseSDPLine(struct icestate_s *con, const char *value)
{
if (!strncmp(value, "a=ice-pwd:", 10))
ICE_Set(con, "rpwd", value+10);
else if (!strncmp(value, "a=ice-ufrag:", 12))
ICE_Set(con, "rufrag", value+12);
#ifdef HAVE_DTLS
else if (!strncmp(value, "a=setup:", 8))
{ //this is their state, so we want the opposite.
if (!strncmp(value+8, "passive", 7))
con->dtlspassive = false;
else if (!strncmp(value+8, "active", 6))
con->dtlspassive = true;
}
else if (!strncmp(value, "a=fingerprint:", 14))
{
char name[64];
value = COM_ParseOut(value+14, name, sizeof(name));
if (!strcasecmp(name, "sha-1"))
con->cred.peer.hash = &hash_sha1;
else if (!strcasecmp(name, "sha-224"))
con->cred.peer.hash = &hash_sha224;
else if (!strcasecmp(name, "sha-256"))
con->cred.peer.hash = &hash_sha256;
else if (!strcasecmp(name, "sha-384"))
con->cred.peer.hash = &hash_sha384;
else if (!strcasecmp(name, "sha-512"))
con->cred.peer.hash = &hash_sha512;
else
con->cred.peer.hash = NULL; //hash not recognised
if (con->cred.peer.hash)
{
int b, o, v;
while (*value == ' ')
value++;
for (b = 0; b < con->cred.peer.hash->digestsize; )
{
v = *value;
if (v >= '0' && v <= '9')
o = (v-'0');
else if (v >= 'A' && v <= 'F')
o = (v-'A'+10);
else if (v >= 'a' && v <= 'f')
o = (v-'a'+10);
else
break;
o <<= 4;
v = *++value;
if (v >= '0' && v <= '9')
o |= (v-'0');
else if (v >= 'A' && v <= 'F')
o |= (v-'A'+10);
else if (v >= 'a' && v <= 'f')
o |= (v-'a'+10);
else
break;
con->cred.peer.digest[b++] = o;
v = *++value;
if (v != ':')
break;
value++;
}
if (b != con->cred.peer.hash->digestsize)
con->cred.peer.hash = NULL; //bad!
}
}
else if (!strncmp(value, "a=sctp-port:", 12))
con->peersctpport = atoi(value+12);
else if (!strncmp(value, "a=sctp-optional:", 16))
con->peersctpoptional = atoi(value+16);
#endif
else if (!strncmp(value, "a=rtpmap:", 9))
{
char name[64];
int codec;
char *sl;
value += 9;
codec = strtoul(value, (char**)&value, 0);
if (*value == ' ') value++;
COM_ParseOut(value, name, sizeof(name));
sl = strchr(name, '/');
if (sl)
*sl = '@';
ICE_Set(con, va("codec%i", codec), name);
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
else if (!strncmp(value, "a=candidate:", 12))
{
struct icecandinfo_s n;
memset(&n, 0, sizeof(n));
value += 12;
n.foundation = strtoul(value, (char**)&value, 0);
if(*value == ' ')value++;
n.component = strtoul(value, (char**)&value, 0);
if(*value == ' ')value++;
if (!strncasecmp(value, "UDP ", 4))
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
{
n.transport = 0;
value += 3;
}
else
return;
if(*value == ' ')value++;
n.priority = strtoul(value, (char**)&value, 0);
if(*value == ' ')value++;
value = COM_ParseOut(value, n.addr, sizeof(n.addr));
if (!value) return;
if(*value == ' ')value++;
n.port = strtoul(value, (char**)&value, 0);
if(*value == ' ')value++;
if (strncmp(value, "typ ", 4)) return;
value += 3;
if(*value == ' ')value++;
if (!strncmp(value, "host", 4))
n.type = ICE_HOST;
else if (!strncmp(value, "srflx", 4))
n.type = ICE_SRFLX;
else if (!strncmp(value, "prflx", 4))
n.type = ICE_PRFLX;
else if (!strncmp(value, "relay", 4))
n.type = ICE_RELAY;
else
return;
while (*value)
{
if(*value == ' ')value++;
if (!strncmp(value, "raddr ", 6))
{
value += 6;
value = COM_ParseOut(value, n.reladdr, sizeof(n.reladdr));
if (!value)
break;
}
else if (!strncmp(value, "rport ", 6))
{
value += 6;
n.relport = strtoul(value, (char**)&value, 0);
}
/*else if (!strncmp(value, "network-cost ", 13))
{
value += 13;
n.netcost = strtoul(value, (char**)&value, 0);
}*/
/*else if (!strncmp(value, "ufrag ", 6))
{
value += 6;
while (*value && *value != ' ')
value++;
}*/
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
else
{
//this is meant to be extensible.
while (*value && *value != ' ')
value++;
if(*value == ' ')value++;
while (*value && *value != ' ')
value++;
}
}
ICE_AddRCandidateInfo(con, &n);
}
}
void CL_Transfer(netadr_t *adr);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static qboolean QDECL ICE_Set(struct icestate_s *con, const char *prop, const char *value)
{
if (!strcmp(prop, "state"))
{
int oldstate = con->state;
if (!strcmp(value, STRINGIFY(ICE_CONNECTING)))
con->state = ICE_CONNECTING;
else if (!strcmp(value, STRINGIFY(ICE_INACTIVE)))
con->state = ICE_INACTIVE;
else if (!strcmp(value, STRINGIFY(ICE_FAILED)))
con->state = ICE_FAILED;
else if (!strcmp(value, STRINGIFY(ICE_CONNECTED)))
con->state = ICE_CONNECTED;
else
{
Con_Printf("ICE_Set invalid state %s\n", value);
con->state = ICE_INACTIVE;
}
con->timeout = Sys_Milliseconds() + 30;
con->retries = 0;
#ifndef SERVERONLY
if (con->state == ICE_CONNECTING && (con->proto == ICEP_QWCLIENT || con->proto == ICEP_VOICE))
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (!cls.sockets)
NET_InitClient(false);
#endif
if (con->state >= ICE_CONNECTING)
{
#ifndef SERVERONLY
if (con->proto == ICEP_QWCLIENT)
CL_Transfer(&con->qadr); //okay, the client should be using this ice connection now.
#endif
#ifdef HAVE_DTLS
if (con->mode == ICEM_WEBRTC)
{
if (!con->dtlsstate && con->dtlsfuncs)
{
extern cvar_t net_enable_dtls;
if (con->cred.peer.hash)
con->dtlsstate = con->dtlsfuncs->CreateContext(&con->cred, con, ICE_Transmit, con->dtlspassive);
else if (net_enable_dtls.ival)
{ //peer doesn't seem to support dtls.
con->state = ICE_FAILED;
Con_Printf(CON_WARNING"WARNING: peer does not support dtls. Set net_enable_dtls to 0 to make optional.\n");
}
else if (con->state == ICE_CONNECTING)
Con_Printf(CON_WARNING"WARNING: peer does not support dtls.\n");
}
if (!con->sctp && (!con->sctpoptional || !con->peersctpoptional) && con->mysctpport && con->peersctpport)
{
con->sctp = Z_Malloc(sizeof(*con->sctp));
con->sctp->myport = htons(con->mysctpport);
con->sctp->peerport = htons(con->peersctpport);
con->sctp->o.tsn = rand() ^ (rand()<<16);
Sys_RandomBytes((void*)&con->sctp->o.verifycode, sizeof(con->sctp->o.verifycode));
Sys_RandomBytes((void*)&con->sctp->i.verifycode, sizeof(con->sctp->i.verifycode));
}
}
#endif
}
if (oldstate != con->state && con->state == ICE_CONNECTED)
{
if (con->chosenpeer.type == NA_INVALID)
{
con->state = ICE_FAILED;
Con_Printf("ICE failed. peer not valid.\n");
}
#ifndef CLIENTONLY
else if (con->proto == ICEP_QWSERVER && con->mode != ICEM_WEBRTC)
{
net_from = con->qadr;
SVC_GetChallenge(false);
}
#endif
if (con->state == ICE_CONNECTED)
{
if (con->proto >= ICEP_VOICE || net_ice_debug.ival)
{
char msg[256];
Con_Printf(S_COLOR_GRAY "[%s]: %s connection established (peer %s, via %s).\n", con->friendlyname, con->proto == ICEP_VOICE?"voice":"data", NET_AdrToString(msg, sizeof(msg), &con->chosenpeer), ICE_NetworkToName(con, con->chosenpeer.connum));
}
}
}
#if !defined(SERVERONLY) && defined(VOICECHAT)
snd_voip_send.ival = (snd_voip_send.ival & ~4) | (NET_RTP_Active()?4:0);
#endif
}
else if (!strcmp(prop, "controlled"))
con->controlled = !!atoi(value);
else if (!strcmp(prop, "controller"))
con->controlled = !atoi(value);
else if (!strncmp(prop, "codec", 5))
{
struct icecodecslot_s *codec = ICE_GetCodecSlot(con, atoi(prop+5));
if (!codec)
return false;
codec->id = atoi(prop+5);
#if !defined(SERVERONLY) && defined(VOICECHAT)
if (!S_Voip_RTP_CodecOkay(value))
#endif
{
Z_Free(codec->name);
codec->name = NULL;
return false;
}
Z_Free(codec->name);
codec->name = Z_StrDup(value);
}
else if (!strcmp(prop, "rufrag"))
{
Z_Free(con->rufrag);
con->rufrag = Z_StrDup(value);
}
else if (!strcmp(prop, "rpwd"))
{
Z_Free(con->rpwd);
con->rpwd = Z_StrDup(value);
}
else if (!strcmp(prop, "server"))
{
netadr_t hostadr[1];
qboolean okay;
qboolean tcp = false;
qboolean tls;
qboolean stun;
char *s, *next;
const char *user=NULL, *auth=NULL;
const char *host;
netadrtype_t family = NA_INVALID;
if (!strncmp(value, "stun:", 5))
stun=true, tls=false, value+=5;
else if (!strncmp(value, "turn:", 5))
stun=false, tls=false, value+=5;
else if (!strncmp(value, "turns:", 6))
stun=false, tls=true, value+=6;
else
return false; //nope, uri not known.
host = Z_StrDup(value);
s = strchr(host, '?');
for (;s;s=next)
{
*s++ = 0;
next = strchr(s, '?');
if (next)
*next = 0;
if (!strncmp(s, "transport=", 10))
{
if (!strcmp(s+10, "udp"))
tcp=false;
else if (!strcmp(s+10, "tcp"))
tcp=true;
}
else if (!strncmp(s, "user=", 5))
user = s+5;
else if (!strncmp(s, "auth=", 5))
auth = s+5;
else if (!strncmp(s, "fam=", 4))
{
if (!strcmp(s+4, "ipv4") || !strcmp(s+4, "ip4") || !strcmp(s+4, "ip") || !strcmp(s+4, "4"))
family=NA_IP;
else if (!strcmp(s+4, "ipv6") || !strcmp(s+4, "ip6") || !strcmp(s+4, "6"))
family=NA_IPV6;
}
}
okay = !strchr(host, '/');
if (con->servers == countof(con->server))
okay = false;
else if (okay)
{
struct iceserver_s *srv = &con->server[con->servers];
//handily both stun and turn default to the same port numbers.
//FIXME: worker thread...
okay = NET_StringToAdr2(host, tls?5349:3478, hostadr, countof(hostadr), NULL);
if (okay)
{
if (tls)
hostadr->prot = tcp?NP_TLS:NP_DTLS;
else
hostadr->prot = tcp?NP_STREAM:NP_DGRAM;
con->servers++;
srv->isstun = stun;
srv->family = family;
srv->realm = Z_StrDup(host);
Sys_RandomBytes((char*)srv->stunrnd, sizeof(srv->stunrnd));
srv->stunretry = Sys_Milliseconds(); //'now'...
srv->addr = *hostadr;
srv->user = user?Z_StrDup(user):NULL;
srv->auth = auth?Z_StrDup(auth):NULL;
}
}
Z_Free(s);
return !!okay;
}
else if (!strcmp(prop, "sdp") || !strcmp(prop, "sdpoffer") || !strcmp(prop, "sdpanswer"))
{
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
char line[8192];
const char *eol;
for (; *value; value = eol)
{
eol = strchr(value, '\n');
if (!eol)
eol = value+strlen(value);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (eol-value < sizeof(line))
{
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
memcpy(line, value, eol-value);
line[eol-value] = 0;
if (eol>value && line[eol-value-1] == '\r')
line[eol-value-1] = 0;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
ICE_ParseSDPLine(con, line);
}
if (eol && *eol)
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
eol++;
}
}
else
return false;
return true;
}
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
static char *ICE_CandidateToSDP(struct icecandidate_s *can, char *value, size_t valuelen)
{
char *ctype = "?";
switch(can->info.type)
{
default:
case ICE_HOST: ctype = "host"; break;
case ICE_SRFLX: ctype = "srflx"; break;
case ICE_PRFLX: ctype = "prflx"; break;
case ICE_RELAY: ctype = "relay"; break;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
Q_snprintfz(value, valuelen, "a=candidate:%i %i %s %i %s %i typ %s",
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
can->info.foundation,
can->info.component,
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
can->info.transport==0?"UDP":"ERROR",
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
can->info.priority,
can->info.addr,
can->info.port,
ctype
);
Q_strncatz(value, va(" generation %i", can->info.generation), valuelen);
if (can->info.type != ICE_HOST)
{
if (net_ice_relayonly.ival)
{ //don't leak srflx info (technically this info is mandatory)
Q_strncatz(value, va(" raddr %s", can->info.addr), valuelen);
Q_strncatz(value, va(" rport %i", can->info.port), valuelen);
}
else
{
if (*can->info.reladdr)
Q_strncatz(value, va(" raddr %s", can->info.reladdr), valuelen);
Q_strncatz(value, va(" rport %i", can->info.relport), valuelen);
}
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
}
return value;
}
static qboolean ICE_LCandidateIsPrivate(struct icecandinfo_s *caninfo)
{ //return true for the local candidates that we're actually allowed to report. they'll stay flagged as 'dirty' otherwise.
if (!net_ice_exchangeprivateips.ival && caninfo->type == ICE_HOST)
return true;
if (net_ice_relayonly.ival && caninfo->type != ICE_RELAY)
return true;
return false;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static qboolean QDECL ICE_Get(struct icestate_s *con, const char *prop, char *value, size_t valuelen)
{
if (!strcmp(prop, "sid"))
Q_strncpyz(value, con->conname, valuelen);
else if (!strcmp(prop, "state"))
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
{
switch(con->state)
{
case ICE_INACTIVE:
Q_strncpyz(value, STRINGIFY(ICE_INACTIVE), valuelen);
break;
case ICE_FAILED:
Q_strncpyz(value, STRINGIFY(ICE_FAILED), valuelen);
break;
case ICE_CONNECTING:
Q_strncpyz(value, STRINGIFY(ICE_CONNECTING), valuelen);
break;
case ICE_CONNECTED:
Q_strncpyz(value, STRINGIFY(ICE_CONNECTED), valuelen);
break;
}
}
else if (!strcmp(prop, "lufrag"))
Q_strncpyz(value, con->lufrag, valuelen);
else if (!strcmp(prop, "lpwd"))
Q_strncpyz(value, con->lpwd, valuelen);
else if (!strncmp(prop, "codec", 5))
{
int codecid = atoi(prop+5);
struct icecodecslot_s *codec = ICE_GetCodecSlot(con, atoi(prop+5));
if (!codec || codec->id != codecid)
return false;
if (codec->name)
Q_strncpyz(value, codec->name, valuelen);
else
Q_strncpyz(value, "", valuelen);
}
else if (!strcmp(prop, "newlc"))
{
struct icecandidate_s *can;
Q_strncpyz(value, "0", valuelen);
for (can = con->lc; can; can = can->next)
{
if (can->dirty && !ICE_LCandidateIsPrivate(&can->info))
{
Q_strncpyz(value, "1", valuelen);
break;
}
}
}
else if (!strcmp(prop, "peersdp"))
{ //for debugging.
Q_strncpyz(value, "", valuelen);
if ((con->proto == ICEP_QWSERVER || con->proto == ICEP_QWCLIENT) && con->mode == ICEM_WEBRTC)
{
#ifdef HAVE_DTLS
if (con->cred.peer.hash)
{
int b;
Q_strncatz(value, "a=fingerprint:", valuelen);
if (con->cred.peer.hash == &hash_sha1)
Q_strncatz(value, "sha-1", valuelen);
else if (con->cred.peer.hash == &hash_sha256)
Q_strncatz(value, "sha-256", valuelen);
else if (con->cred.peer.hash == &hash_sha512)
Q_strncatz(value, "sha-512", valuelen);
else
Q_strncatz(value, "UNKNOWN", valuelen);
for (b = 0; b < con->cred.peer.hash->digestsize; b++)
Q_strncatz(value, va(b?":%02X":" %02X", con->cred.peer.digest[b]), valuelen);
Q_strncatz(value, "\n", valuelen);
}
#endif
}
Q_strncatz(value, va("a=ice-pwd:%s\n", con->rpwd), valuelen);
Q_strncatz(value, va("a=ice-ufrag:%s\n", con->rufrag), valuelen);
#ifdef HAVE_DTLS
if (con->peersctpport)
Q_strncatz(value, va("a=sctp-port:%i\n", con->peersctpport), valuelen); //stupid hardcoded thing.
if (con->peersctpoptional)
Q_strncatz(value, "a=sctp-optional:1\n", valuelen);
#endif
}
else if (!strcmp(prop, "sdp") || !strcmp(prop, "sdpoffer") || !strcmp(prop, "sdpanswer"))
{
struct icecandidate_s *can;
netadr_t sender;
char tmpstr[MAX_QPATH], *at;
int i;
{
netadr_t addr[1];
struct ftenet_generic_connection_s *gcon[countof(addr)];
int flags[countof(addr)];
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
const char *params[countof(addr)];
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (!NET_EnumerateAddresses(ICE_PickConnection(con), gcon, flags, addr, params, countof(addr)))
{
sender.type = NA_INVALID;
sender.port = 0;
}
else
sender = *addr;
}
Q_strncpyz(value, "v=0\n", valuelen); //version...
Q_strncatz(value, va("o=%s %u %u IN IP4 %s\n", "-", con->originid, con->originversion, con->originaddress), valuelen); //originator. usually just dummy info.
Q_strncatz(value, va("s=%s\n", con->conname), valuelen); //session name.
Q_strncatz(value, "t=0 0\n", valuelen); //start+end times...
Q_strncatz(value, va("a=ice-options:trickle\n"), valuelen);
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
if ((con->proto == ICEP_QWSERVER || con->proto == ICEP_QWCLIENT) && con->mode == ICEM_WEBRTC)
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
{
#ifdef HAVE_DTLS
if (con->cred.local.certsize)
{
qbyte fingerprint[DIGEST_MAXSIZE];
int b;
CalcHash(&hash_sha256, fingerprint, sizeof(fingerprint), con->cred.local.cert, con->cred.local.certsize);
Q_strncatz(value, "a=fingerprint:sha-256", valuelen);
for (b = 0; b < hash_sha256.digestsize; b++)
Q_strncatz(value, va(b?":%02X":" %02X", fingerprint[b]), valuelen);
Q_strncatz(value, "\n", valuelen);
}
Q_strncatz(value, "m=application 9 UDP/DTLS/SCTP webrtc-datachannel\n", valuelen);
if (con->sctpoptional)
Q_strncatz(value, "a=sctp-optional:1\n", valuelen);
#endif
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
}
// Q_strncatz(value, va("c=IN %s %s\n", sender.type==NA_IPV6?"IP6":"IP4", NET_BaseAdrToString(tmpstr, sizeof(tmpstr), &sender)), valuelen);
Q_strncatz(value, "c=IN IP4 0.0.0.0\n", valuelen);
Q_strncatz(value, va("a=ice-pwd:%s\n", con->lpwd), valuelen);
Q_strncatz(value, va("a=ice-ufrag:%s\n", con->lufrag), valuelen);
#ifdef HAVE_DTLS
if (con->dtlsfuncs)
{
if (!strcmp(prop, "sdpanswer"))
{ //answerer decides.
if (con->dtlspassive)
Q_strncatz(value, va("a=setup:passive\n"), valuelen);
else
Q_strncatz(value, va("a=setup:active\n"), valuelen);
}
else if (!strcmp(prop, "sdpoffer"))
Q_strncatz(value, va("a=setup:actpass\n"), valuelen); //don't care if we're active or passive
}
if (con->mysctpport)
Q_strncatz(value, va("a=sctp-port:%i\n", con->mysctpport), valuelen); //stupid hardcoded thing.
#endif
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
/*fixme: merge the codecs into a single media line*/
for (i = 0; i < countof(con->codecslot); i++)
{
int id = con->codecslot[i].id;
if (!con->codecslot[i].name)
continue;
Q_strncatz(value, va("m=audio %i RTP/AVP %i\n", sender.port, id), valuelen);
Q_strncatz(value, va("b=RS:0\n"), valuelen);
Q_strncatz(value, va("b=RR:0\n"), valuelen);
Q_strncpyz(tmpstr, con->codecslot[i].name, sizeof(tmpstr));
at = strchr(tmpstr, '@');
if (at)
{
*at = '/';
Q_strncatz(value, va("a=rtpmap:%i %s\n", id, tmpstr), valuelen);
}
else
Q_strncatz(value, va("a=rtpmap:%i %s/%i\n", id, tmpstr, 8000), valuelen);
for (can = con->lc; can; can = can->next)
{
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
char canline[256];
can->dirty = false; //doesn't matter now.
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
ICE_CandidateToSDP(can, canline, sizeof(canline));
Q_strncatz(value, canline, valuelen);
Q_strncatz(value, "\n", valuelen);
}
}
}
else
return false;
return true;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static void ICE_PrintSummary(struct icestate_s *con, qboolean islisten)
{
char msg[64];
Con_Printf(S_COLOR_GRAY" ^[[%s]\\ice\\%s^]: ", con->friendlyname, con->friendlyname);
switch(con->proto)
{
case ICEP_VOICE: Con_Printf(S_COLOR_GRAY"(voice) "); break;
case ICEP_VIDEO: Con_Printf(S_COLOR_GRAY"(video) "); break;
default: break;
}
switch(con->state)
{
case ICE_INACTIVE: Con_Printf(S_COLOR_RED "inactive"); break;
case ICE_FAILED: Con_Printf(S_COLOR_RED "failed"); break;
case ICE_CONNECTING: Con_Printf(S_COLOR_YELLOW "connecting"); break;
case ICE_CONNECTED: Con_Printf(S_COLOR_GRAY"%s via %s", NET_AdrToString(msg,sizeof(msg), &con->chosenpeer), ICE_NetworkToName(con, con->chosenpeer.connum)); break;
}
if (con->dtlsstate)
Con_Printf(S_COLOR_GREEN " (encrypted)");
else
Con_Printf(S_COLOR_RED " (plain-text)");
Con_Printf("\n");
}
static void ICE_Debug(struct icestate_s *con)
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
{
struct icecandidate_s *can;
char buf[65536];
ICE_Get(con, "state", buf, sizeof(buf));
Con_Printf("ICE [%s] (%s):\n", con->friendlyname, buf);
if (net_ice_debug.ival >= 2)
{ //rather uninteresting really...
if (con->initiator)
ICE_Get(con, "sdpoffer", buf, sizeof(buf));
else
ICE_Get(con, "sdpanswer", buf, sizeof(buf));
Con_Printf("sdp:\n"S_COLOR_YELLOW "%s\n", buf);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
//incomplete anyway
ICE_Get(con, "peersdp", buf, sizeof(buf));
Con_Printf("peer:\n"S_COLOR_YELLOW"%s\n", buf);
}
Con_Printf(" local:\n");
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
for (can = con->lc; can; can = can->next)
{
ICE_CandidateToSDP(can, buf, sizeof(buf));
if (con->chosenpeer.type!=NA_INVALID && con->chosenpeer.connum == can->info.network)
Con_Printf(S_COLOR_GREEN" %s\n", buf);
else if (can->dirty)
Con_Printf(S_COLOR_RED" %s\n", buf);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
else
Con_Printf(S_COLOR_YELLOW" %s\n", buf);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
Con_Printf(" remote:\n");
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
for (can = con->rc; can; can = can->next)
{
ICE_CandidateToSDP(can, buf, sizeof(buf));
if (can->reachable)
{
if (con->chosenpeer.type!=NA_INVALID && NET_CompareAdr(&can->peer,&con->chosenpeer))
Con_Printf(S_COLOR_GREEN" %s\n", buf);
else
Con_Printf(S_COLOR_YELLOW" %s\n", buf);
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
else
Con_Printf(S_COLOR_RED" %s\n", buf);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
}
static void ICE_Show_f(void)
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
{
const char *findname = Cmd_Argv(1);
struct icestate_s *ice;
for (ice = icelist; ice; ice = ice->next)
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
{
if (!*findname || !strcmp(findname, ice->friendlyname))
ICE_Debug(ice);
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
}
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static struct icecandinfo_s *QDECL ICE_GetLCandidateInfo(struct icestate_s *con)
{
struct icecandidate_s *can;
for (can = con->lc; can; can = can->next)
{
if (can->dirty)
{
if (ICE_LCandidateIsPrivate(&can->info))
continue;
can->dirty = false;
return &can->info;
}
}
return NULL;
}
static qboolean QDECL ICE_GetLCandidateSDP(struct icestate_s *con, char *out, size_t outsize)
{
struct icecandinfo_s *info = ICE_GetLCandidateInfo(con);
if (info)
{
struct icecandidate_s *can = (struct icecandidate_s*)info;
ICE_CandidateToSDP(can, out, outsize);
return true;
}
return false;
}
static unsigned int ICE_ComputePriority(netadr_t *adr, struct icecandinfo_s *info)
{
int tpref, lpref;
switch(info->type)
{
case ICE_HOST: tpref = 126; break; //ideal
case ICE_PRFLX: tpref = 110; break;
case ICE_SRFLX: tpref = 100; break;
default:
case ICE_RELAY: tpref = 0; break; //relays suck
}
lpref = 0;
if (info->transport == 0)
lpref += 0x8000; //favour udp the most (tcp sucks for stalls)
lpref += (255-info->network)<<16; //favour the first network/socket specified
lpref += (255-NET_ClassifyAddress(adr, NULL))<<8;
if (adr->type == NA_IP)
lpref += 0x0001; //favour ipv4 over ipv6 (less header/mtu overhead...). this is only slight,
return (tpref<<24) + (lpref<<lpref) + ((256 - info->component)<<0);
}
//adrno is 0 if the type is anything but host.
void QDECL ICE_AddLCandidateInfo(struct icestate_s *con, netadr_t *adr, int adrno, int type)
{
int rnd[2];
struct icecandidate_s *cand;
if (!con)
return;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
switch(adr->type)
{
case NA_IP:
case NA_IPV6:
if (NET_ClassifyAddress(adr, NULL) < ASCOPE_LCANDIDATE)
{ //don't waste time asking the relay to poke its loopback. if only because it'll report lots of errors.
return;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
break;
default: //bad protocols
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return; //no, just no.
}
switch(adr->prot)
{
case NP_DTLS:
case NP_DGRAM:
break;
default:
return; //don't report any tcp/etc connections...
}
for (cand = con->lc; cand; cand = cand->next)
{
if (NET_CompareAdr(adr, &cand->peer))
break;
}
if (cand)
return; //DUPE
cand = Z_Malloc(sizeof(*cand));
cand->next = con->lc;
con->lc = cand;
NET_BaseAdrToString(cand->info.addr, sizeof(cand->info.addr), adr);
cand->info.port = ntohs(adr->port);
cand->info.type = type;
cand->info.generation = 0;
cand->info.foundation = 1;
cand->info.component = 1;
cand->info.network = adr->connum;
cand->dirty = true;
Sys_RandomBytes((void*)rnd, sizeof(rnd));
Q_strncpyz(cand->info.candidateid, va("x%08x%08x", rnd[0], rnd[1]), sizeof(cand->info.candidateid));
cand->info.priority = ICE_ComputePriority(adr, &cand->info);
}
void QDECL ICE_AddLCandidateConn(ftenet_connections_t *col, netadr_t *addr, int type)
{
struct icestate_s *ice;
for (ice = icelist; ice; ice = ice->next)
{
if (ICE_PickConnection(ice) == col)
ICE_AddLCandidateInfo(ice, addr, 0, type);
}
}
static void ICE_Destroy(struct icestate_s *con)
{
struct icecandidate_s *c;
#ifdef HAVE_DTLS
if (con->sctp)
{
Z_Free(con->sctp->cookie);
Z_Free(con->sctp);
}
if (con->dtlsstate)
con->dtlsfuncs->DestroyContext(con->dtlsstate);
if (con->cred.local.cert)
Z_Free(con->cred.local.cert);
if (con->cred.local.key)
Z_Free(con->cred.local.key);
#endif
while(con->rc)
{
c = con->rc;
con->rc = c->next;
Z_Free(c);
}
while(con->lc)
{
c = con->lc;
con->lc = c->next;
Z_Free(c);
}
while (con->servers)
{
struct iceserver_s *s = &con->server[--con->servers];
if (s->con)
{ //make sure we tell the TURN server to release our allocation.
s->state = TURN_TERMINATING;
ICE_ToStunServer(con, s);
s->con->Close(s->con);
}
Z_Free(s->user);
Z_Free(s->auth);
Z_Free(s->realm);
Z_Free(s->nonce);
}
if (con->connections)
FTENET_CloseCollection(con->connections);
Z_Free(con->lufrag);
Z_Free(con->lpwd);
Z_Free(con->rufrag);
Z_Free(con->rpwd);
Z_Free(con->friendlyname);
Z_Free(con->conname);
//has already been unlinked
Z_Free(con);
}
//send pings to establish/keep the connection alive
void ICE_Tick(void)
{
struct icestate_s *con;
for (con = icelist; con; con = con->next)
{
switch(con->mode)
{
case ICEM_RAW:
//raw doesn't do handshakes or keepalives. it should just directly connect.
//raw just uses the first (assumed only) option
if (con->state == ICE_CONNECTING)
{
struct icecandidate_s *rc;
rc = con->rc;
if (!rc || !NET_StringToAdr(rc->info.addr, rc->info.port, &con->chosenpeer))
con->chosenpeer.type = NA_INVALID;
ICE_Set(con, "state", STRINGIFY(ICE_CONNECTED));
}
break;
case ICEM_WEBRTC:
case ICEM_ICE:
if (con->state == ICE_CONNECTING || con->state == ICE_CONNECTED)
{
unsigned int curtime = Sys_Milliseconds();
size_t i, j;
struct iceserver_s *srv;
for (i = 0; i < con->servers; i++)
{
srv = &con->server[i];
if ((signed int)(srv->stunretry-curtime) < 0)
{
srv->stunretry = curtime + 2*1000;
ICE_ToStunServer(con, srv);
}
for (j = 0; j < srv->peers; j++)
{
if ((signed int)(srv->peer[j].retry-curtime) < 0)
{
TURN_AuthorisePeer(con, srv, j);
srv->peer[j].retry = curtime + 2*1000;
}
}
if (srv->con)
{
while (srv->con->GetPacket(srv->con))
{
net_from.connum = srv->con->connum;
net_from_connection = srv->con;
srv->con->owner->ReadGamePacket();
}
}
}
if (con->keepalive < curtime)
{
if (!ICE_SendSpam(con))
{
struct icecandidate_s *rc;
struct icecandidate_s *best = NULL;
for (rc = con->rc; rc; rc = rc->next)
{ //FIXME:
if (rc->reachable && (!best || rc->info.priority > best->info.priority))
best = rc;
}
if (best)
{
netadr_t nb = best->peer;
for (i = 0; ; i++)
{
if (best->reachable&(1<<i))
{
best->tried &= ~(1<<i); //keep poking it...
nb.connum = i+1;
break;
}
}
if (memcmp(&con->chosenpeer, &nb, sizeof(nb)) && (con->chosenpeer.type==NA_INVALID || !con->controlled))
{ //it actually changed... let them know NOW.
best->tried &= ~(1<<(con->chosenpeer.connum-1)); //make sure we resend this one.
con->chosenpeer = nb;
ICE_SendSpam(con);
if (net_ice_debug.ival >= 1)
{
char msg[64];
Con_Printf(S_COLOR_GRAY"[%s]: New peer chosen %s, via %s.\n", con->friendlyname, NET_AdrToString(msg, sizeof(msg), &con->chosenpeer), ICE_NetworkToName(con, con->chosenpeer.connum));
}
}
}
/*if (con->state == ICE_CONNECTED && best)
{ //once established, back off somewhat
for (rc = con->rc; rc; rc = rc->next)
rc->tried &= ~rc->reachable;
}
else*/
{
for (rc = con->rc; rc; rc = rc->next)
rc->tried = 0;
}
con->retries++;
if (con->retries > 32)
con->retries = 32;
con->keepalive = curtime + 200*(con->retries); //RTO... ish.
}
else
con->keepalive = curtime + 50*(con->retries+1); //Ta... absmin of 5ms
}
}
if (con->state == ICE_CONNECTED)
{
#ifdef HAVE_DTLS
if (con->sctp)
SCTP_Transmit(con->sctp, con, NULL,0); //try to keep it ticking...
if (con->dtlsstate)
con->dtlsfuncs->Timeouts(con->dtlsstate);
#endif
//FIXME: We should be sending a stun binding indication every 15 secs with a fingerprint attribute
}
break;
}
}
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static void QDECL ICE_Close(struct icestate_s *con)
{
struct icestate_s **link;
ICE_Set(con, "state", STRINGIFY(ICE_INACTIVE));
for (link = &icelist; *link; )
{
if (con == *link)
{
*link = con->next;
ICE_Destroy(con);
return;
}
else
link = &(*link)->next;
}
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static void QDECL ICE_CloseModule(void *module)
{
struct icestate_s **link, *con;
for (link = &icelist; *link; )
{
con = *link;
if (con->module == module)
{
*link = con->next;
ICE_Destroy(con);
}
else
link = &(*link)->next;
}
}
icefuncs_t iceapi =
{
ICE_Create,
ICE_Set,
ICE_Get,
ICE_GetLCandidateInfo,
ICE_AddRCandidateInfo,
ICE_Close,
playdemo accepts https urls now. will start playing before the file has finished downloading, to avoid unnecessary delays. reworked network addresses to separate address family and connection type. this should make banning people more reliable, as well as simplifying a whole load of logic (no need to check for ipv4 AND ipv6). tcpconnect will keep trying to connect even if the connection wasn't instant, instead of giving up instantly. rewrote tcp connections quite a bit. sv_port_tcp now handles qtv+qizmo+http+ws+rtcbroker+tls equivalents. qtv_streamport is now a legacy cvar and now acts equivalently to sv_port_tcp (but still separate). rewrote screenshot and video capture code to use strides. this solves image-is-upside down issues with vulkan. ignore alt key in browser port. oh no! no more red text! oh no! no more alt-being-wrongly-down-and-being-unable-to-type-anything-without-forcing-alt-released! reworked audio decoder interface. now has clearly defined success/unavailable/end-of-file results. this should solve a whole load of issues with audio streaming. fixed various openal audio streaming issues too. openal also got some workarounds for emscripten's poor emulation. fixed ogg decoder to retain sync properly if seeked. updated menu_media a bit. now reads vorbis comments/id3v1 tags to get proper track names. also saves the playlist so you don't have to manually repopulate the list so it might actually be usable now (after how many years?) r_stains now defaults to 0, and is no longer enabled by presets. use decals if you want that sort of thing. added fs_noreexec cvar, so configs will not be reexeced on gamedir change. this also means defaults won't be reapplied, etc. added 'nvvk' renderer on windows, using nvidia's vulkan-inside-opengl gl extension. mostly just to see how much slower it is. fixed up the ftp server quite a lot. more complete, more compliant, and should do ipv6 properly to-boot. file transfers also threaded. fixed potential crash inside runclientphys. experimental sv_antilag=3 setting. totally untested. the aim is to avoid missing due to lagged knockbacks. may be expensive for the server. browser port's websockets support fixed. experimental support for webrtc ('works for me', requires a broker server). updated avplug(renamed to ffmpeg so people know what it is) to use ffmpeg 3.2.4 properly, with its new encoder api. should be much more robust... also added experimental audio decoder for game music etc (currently doesn't resample, so playback rates are screwed, disabled by cvar). git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5097 fc73d0e0-1445-4013-8a0c-d673dee63da5
2017-05-10 02:08:58 +00:00
ICE_CloseModule,
ICE_GetLCandidateSDP
};
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#endif
#if defined(SUPPORT_ICE) && defined(HAVE_DTLS)
//========================================
//WebRTC's interpretation of SCTP. its annoying, but hey its only 28 wasted bytes... along with the dtls overhead too. most of this is redundant.
//we only send unreliably.
//there's no point in this code without full webrtc code.
struct sctp_header_s
{
quint16_t srcport;
quint16_t dstport;
quint32_t verifycode;
quint32_t crc;
};
struct sctp_chunk_s
{
qbyte type;
#define SCTP_TYPE_DATA 0
#define SCTP_TYPE_INIT 1
#define SCTP_TYPE_INITACK 2
#define SCTP_TYPE_SACK 3
#define SCTP_TYPE_PING 4
#define SCTP_TYPE_PONG 5
#define SCTP_TYPE_ABORT 6
#define SCTP_TYPE_SHUTDOWN 7
#define SCTP_TYPE_SHUTDOWNACK 8
#define SCTP_TYPE_ERROR 9
#define SCTP_TYPE_COOKIEECHO 10
#define SCTP_TYPE_COOKIEACK 11
#define SCTP_TYPE_SHUTDOWNDONE 14
#define SCTP_TYPE_FORWARDTSN 192
qbyte flags;
quint16_t length;
//value...
};
struct sctp_chunk_data_s
{
struct sctp_chunk_s chunk;
quint32_t tsn;
quint16_t sid;
quint16_t seq;
quint32_t ppid;
#define SCTP_PPID_DCEP 50 //datachannel establishment protocol
#define SCTP_PPID_DATA 53 //our binary quake data.
};
struct sctp_chunk_init_s
{
struct sctp_chunk_s chunk;
quint32_t verifycode;
quint32_t arwc;
quint16_t numoutstreams;
quint16_t numinstreams;
quint32_t tsn;
};
struct sctp_chunk_sack_s
{
struct sctp_chunk_s chunk;
quint32_t tsn;
quint32_t arwc;
quint16_t gaps;
quint16_t dupes;
/*struct {
quint16_t first;
quint16_t last;
} gapblocks[]; //actually received rather than gaps, but same meaning.
quint32_t dupe_tsns[];*/
};
struct sctp_chunk_fwdtsn_s
{
struct sctp_chunk_s chunk;
quint32_t tsn;
/*struct
{
quint16_t sid;
quint16_t seq;
} streams[];*/
};
static neterr_t SCTP_PeerSendPacket(struct icestate_s *peer, int length, const void *data)
{ //sends to the dtls layer (which will send to the generic ice dispatcher that'll send to the dgram stuff... layers on layers.
if (peer)
{
if (peer->dtlsstate)
return peer->dtlsfuncs->Transmit(peer->dtlsstate, data, length);
else if (peer->chosenpeer.type != NA_INVALID)
return ICE_Transmit(peer, data, length);
else if (peer->state < ICE_CONNECTING)
return NETERR_DISCONNECTED;
else
return NETERR_CLOGGED;
}
else
return NETERR_NOROUTE;
}
static quint32_t SCTP_Checksum(const struct sctp_header_s *h, size_t size)
{
int k;
const qbyte *buf = (const qbyte*)h;
size_t ofs;
uint32_t crc = 0xFFFFFFFF;
for (ofs = 0; ofs < size; ofs++)
{
if (ofs >= 8 && ofs < 8+4)
; //the header's crc should be read as 0.
else
crc ^= buf[ofs];
for (k = 0; k < 8; k++) //CRC-32C polynomial 0x1EDC6F41 in reversed bit order.
crc = crc & 1 ? (crc >> 1) ^ 0x82f63b78 : crc >> 1;
}
return ~crc;
}
static neterr_t SCTP_Transmit(sctp_t *sctp, struct icestate_s *peer, const void *data, size_t length)
{
qbyte pkt[65536];
size_t pktlen = 0;
struct sctp_header_s *h = (void*)pkt;
struct sctp_chunk_data_s *d;
struct sctp_chunk_fwdtsn_s *fwd;
if (length > sizeof(pkt))
return NETERR_MTU;
h->dstport = sctp->peerport;
h->srcport = sctp->myport;
h->verifycode = sctp->o.verifycode;
pktlen += sizeof(*h);
//advance our ctsn if we're received the relevant packets
while(sctp->i.htsn)
{
quint32_t tsn = sctp->i.ctsn+1;
if (!(sctp->i.received[(tsn>>3)%sizeof(sctp->i.received)] & 1<<(tsn&7)))
break;
//advance our cumulative ack.
sctp->i.received[(tsn>>3)%sizeof(sctp->i.received)] &= ~(1<<(tsn&7));
sctp->i.ctsn = tsn;
sctp->i.htsn--;
}
if (!sctp->o.writable)
{
double time = Sys_DoubleTime();
if (time > sctp->nextreinit)
{
sctp->nextreinit = time + 0.5;
if (!sctp->cookie)
{
struct sctp_chunk_init_s *init = (struct sctp_chunk_init_s*)&pkt[pktlen];
struct {
quint16_t ptype;
quint16_t plen;
} *ftsn = (void*)(init+1);
h->verifycode = 0;
init->chunk.type = SCTP_TYPE_INIT;
init->chunk.flags = 0;
init->chunk.length = BigShort(sizeof(*init)+sizeof(*ftsn));
init->verifycode = sctp->i.verifycode;
init->arwc = BigLong(65535);
init->numoutstreams = BigShort(2);
init->numinstreams = BigShort(2);
init->tsn = BigLong(sctp->o.tsn);
ftsn->ptype = BigShort(49152);
ftsn->plen = BigShort(sizeof(*ftsn));
pktlen += sizeof(*init) + sizeof(*ftsn);
h->crc = SCTP_Checksum(h, pktlen);
return SCTP_PeerSendPacket(peer, pktlen, h);
}
else
{
struct sctp_chunk_s *cookie = (struct sctp_chunk_s*)&pkt[pktlen];
if (pktlen + sizeof(*cookie) + sctp->cookiesize > sizeof(pkt))
return NETERR_DISCONNECTED;
cookie->type = SCTP_TYPE_COOKIEECHO;
cookie->flags = 0;
cookie->length = BigShort(sizeof(*cookie)+sctp->cookiesize);
memcpy(cookie+1, sctp->cookie, sctp->cookiesize);
pktlen += sizeof(*cookie) + sctp->cookiesize;
h->crc = SCTP_Checksum(h, pktlen);
return SCTP_PeerSendPacket(peer, pktlen, h);
}
}
return NETERR_CLOGGED; //nope, not ready yet
}
if (sctp->peerhasfwdtsn && (int)(sctp->o.ctsn-sctp->o.tsn) < -5 && sctp->o.losttsn)
{
fwd = (struct sctp_chunk_fwdtsn_s*)&pkt[pktlen];
fwd->chunk.type = SCTP_TYPE_FORWARDTSN;
fwd->chunk.flags = 0;
fwd->chunk.length = BigShort(sizeof(*fwd));
fwd->tsn = BigLong(sctp->o.tsn-1);
//we only send unordered unreliables, so this stream stuff here is irrelevant.
// fwd->streams[0].sid = sctp->qstreamid;
// fwd->streams[0].seq = BigShort(0);
pktlen += sizeof(*fwd);
}
if (sctp->i.ackneeded >= 2)
{
struct sctp_chunk_sack_s *rsack;
struct sctp_chunk_sack_gap_s {
uint16_t first;
uint16_t last;
} *rgap;
quint32_t otsn;
rsack = (struct sctp_chunk_sack_s*)&pkt[pktlen];
rsack->chunk.type = SCTP_TYPE_SACK;
rsack->chunk.flags = 0;
rsack->chunk.length = BigShort(sizeof(*rsack));
rsack->tsn = BigLong(sctp->i.ctsn);
rsack->arwc = BigLong(65535);
rsack->gaps = 0;
rsack->dupes = BigShort(0);
pktlen += sizeof(*rsack);
rgap = (struct sctp_chunk_sack_gap_s*)&pkt[pktlen];
for (otsn = 0; otsn < sctp->i.htsn; otsn++)
{
quint32_t tsn = sctp->i.ctsn+otsn;
if (!(sctp->i.received[(tsn>>3)%sizeof(sctp->i.received)] & 1<<(tsn&7)))
continue; //missing, don't report it in the 'gaps'... yeah, backwards naming.
if (rsack->gaps && rgap[-1].last == otsn-1)
rgap[-1].last = otsn; //merge into the last one.
else
{
rgap->first = otsn; //these values are Offset from the Cumulative TSN, to save storage.
rgap->last = otsn;
rgap++;
rsack->gaps++;
pktlen += sizeof(*rgap);
if (pktlen >= 500)
break; //might need fragmentation... just stop here.
}
}
for (otsn = 0, rgap = (struct sctp_chunk_sack_gap_s*)&pkt[pktlen]; otsn < rsack->gaps; otsn++)
{
rgap[otsn].first = BigShort(rgap[otsn].first);
rgap[otsn].last = BigShort(rgap[otsn].last);
}
rsack->gaps = BigShort(rsack->gaps);
sctp->i.ackneeded = 0;
}
if (pktlen + sizeof(*d) + length >= 500 && length && pktlen != sizeof(*h))
{ //probably going to result in fragmentation issues. send separate packets.
h->crc = SCTP_Checksum(h, pktlen);
SCTP_PeerSendPacket(peer, pktlen, h);
//reset to the header
pktlen = sizeof(*h);
}
if (length)
{
d = (void*)&pkt[pktlen];
d->chunk.type = SCTP_TYPE_DATA;
d->chunk.flags = 3|4;
d->chunk.length = BigShort(sizeof(*d) + length);
d->tsn = BigLong(sctp->o.tsn++);
d->sid = sctp->qstreamid;
d->seq = BigShort(0); //not needed for unordered
d->ppid = BigLong(SCTP_PPID_DATA);
memcpy(d+1, data, length);
pktlen += sizeof(*d) + length;
//chrome insists on pointless padding at the end. its annoying.
while(pktlen&3)
pkt[pktlen++]=0;
}
if (pktlen == sizeof(*h))
return NETERR_SENT; //nothing to send...
h->crc = SCTP_Checksum(h, pktlen);
return SCTP_PeerSendPacket(peer, pktlen, h);
}
static void SCTP_DecodeDCEP(sctp_t *sctp, struct icestate_s *peer, qbyte *resp)
{ //send an ack...
size_t pktlen = 0;
struct sctp_header_s *h = (void*)resp;
struct sctp_chunk_data_s *d;
char *data = "\02";
size_t length = 1; //*sigh*...
struct
{
qbyte type;
qbyte chantype;
quint16_t priority;
quint32_t relparam;
quint16_t labellen;
quint16_t protocollen;
} *dcep = (void*)sctp->i.r.buf;
if (dcep->type == 3)
{
char *label = (qbyte*)(dcep+1);
char *prot = label + strlen(label)+1;
sctp->qstreamid = sctp->i.r.sid;
if (net_ice_debug.ival >= 1)
Con_Printf("New SCTP Channel: \"%s\" (%s)\n", label, prot);
h->dstport = sctp->peerport;
h->srcport = sctp->myport;
h->verifycode = sctp->o.verifycode;
pktlen += sizeof(*h);
pktlen = (pktlen+3)&~3; //pad.
d = (void*)&resp[pktlen];
d->chunk.type = SCTP_TYPE_DATA;
d->chunk.flags = 3;
d->chunk.length = BigShort(sizeof(*d) + length);
d->tsn = BigLong(sctp->o.tsn++);
d->sid = sctp->qstreamid;
d->seq = BigShort(0); //not needed for unordered
d->ppid = BigLong(SCTP_PPID_DCEP);
memcpy(d+1, data, length);
pktlen += sizeof(*d) + length;
h->crc = SCTP_Checksum(h, pktlen);
SCTP_PeerSendPacket(peer, pktlen, h);
}
}
struct sctp_errorcause_s
{
quint16_t cause;
quint16_t length;
};
static void SCTP_ErrorChunk(const char *errortype, struct sctp_errorcause_s *s, size_t totallen)
{
quint16_t cc, cl;
while(totallen > 0)
{
if (totallen < sizeof(*s))
return; //that's an error in its own right
cc = BigShort(s->cause);
cl = BigShort(s->length);
if (totallen < cl)
return; //err..
switch(cc)
{
case 1: Con_Printf("%s: Invalid Stream Identifier\n", errortype); break;
case 2: Con_Printf("%s: Missing Mandatory Parameter\n", errortype); break;
case 3: Con_Printf("%s: Stale Cookie Error\n", errortype); break;
case 4: Con_Printf("%s: Out of Resource\n", errortype); break;
case 5: Con_Printf("%s: Unresolvable Address\n", errortype); break;
case 6: Con_Printf("%s: Unrecognized Chunk Type\n", errortype); break;
case 7: Con_Printf("%s: Invalid Mandatory Parameter\n", errortype); break;
case 8: Con_Printf("%s: Unrecognized Parameters\n", errortype); break;
case 9: Con_Printf("%s: No User Data\n", errortype); break;
case 10: Con_Printf("%s: Cookie Received While Shutting Down\n", errortype); break;
case 11: Con_Printf("%s: Restart of an Association with New Addresses\n", errortype); break;
case 12: Con_Printf("%s: User Initiated Abort\n", errortype); break;
case 13: Con_Printf("%s: Protocol Violation [%s]\n", errortype, (char*)(s+1)); break;
default: Con_Printf("%s: Unknown Reason\n", errortype); break;
}
totallen -= cl;
totallen &= ~3;
s = (struct sctp_errorcause_s*)((qbyte*)s + ((cl+3)&~3));
}
}
static void SCTP_Decode(sctp_t *sctp, struct icestate_s *peer, ftenet_connections_t *col)
{
qbyte resp[4096];
qbyte *msg = net_message.data;
qbyte *msgend = net_message.data+net_message.cursize;
struct sctp_header_s *h = (struct sctp_header_s*)msg;
struct sctp_chunk_s *c = (struct sctp_chunk_s*)(h+1);
quint16_t clen;
if ((qbyte*)c+1 > msgend)
return; //runt
if (h->dstport != sctp->myport)
return; //not for us...
if (h->srcport != sctp->peerport)
return; //not from them... we could check for a INIT but its over dtls anyway so why give a damn.
if (h->verifycode != ((c->type == SCTP_TYPE_INIT)?0:sctp->i.verifycode))
return; //wrong cookie... (be prepared to parse dupe inits if our ack got lost...
if (h->crc != SCTP_Checksum(h, net_message.cursize))
return; //crc wrong. assume corruption.
if (net_message.cursize&3)
{
if (net_ice_debug.ival >= 2)
Con_Printf("SCTP: packet not padded\n");
return; //mimic chrome, despite it being pointless.
}
while ((qbyte*)(c+1) <= msgend)
{
clen = BigShort(c->length);
if ((qbyte*)c + clen > msgend)
break; //corrupt
safeswitch(c->type)
{
case SCTP_TYPE_DATA:
if (clen >= sizeof(struct sctp_chunk_data_s))
{
struct sctp_chunk_data_s *dc = (void*)c;
quint32_t tsn = BigLong(dc->tsn), u;
qint32_t adv = tsn - sctp->i.ctsn;
sctp->i.ackneeded++;
if (adv >= SCTP_RCVSIZE)
Con_DPrintf("SCTP: Future Packet\n");/*too far in the future. we can't track such things*/
else if (adv <= 0)
Con_DPrintf("SCTP: PreCumulative\n");/*already acked this*/
else if (sctp->i.received[(tsn>>3)%sizeof(sctp->i.received)] & 1<<(tsn&7))
Con_DPrintf("SCTP: Dupe\n");/*already processed it. FIXME: Make a list for the next SACK*/
else
{
qboolean err = false;
if (c->flags & 2)
{ //beginning...
sctp->i.r.firsttns = tsn;
sctp->i.r.tsn = tsn;
sctp->i.r.size = 0;
sctp->i.r.ppid = dc->ppid;
sctp->i.r.sid = dc->sid;
sctp->i.r.seq = dc->seq;
sctp->i.r.toobig = false;
}
else
{
if (sctp->i.r.tsn != tsn || sctp->i.r.ppid != dc->ppid)
err = true;
}
if (err)
; //don't corrupt anything in case we get a quick resend that fixes it.
else
{
size_t dlen = clen-sizeof(*dc);
if (adv > sctp->i.htsn) //weird maths in case it wraps.
sctp->i.htsn = adv;
sctp->i.r.tsn++;
if (sctp->i.r.size + clen-sizeof(*dc) > sizeof(sctp->i.r.buf))
{
Con_DPrintf("SCTP: Oversized\n");
sctp->i.r.toobig = true; //reassembled packet was too large, just corrupt it.
}
else
{
memcpy(sctp->i.r.buf+sctp->i.r.size, dc+1, dlen); //include the dc header
sctp->i.r.size += dlen;
}
if (c->flags & 1) //an ending. we have the complete packet now.
{
for (u = sctp->i.r.tsn - sctp->i.r.firsttns; u --> 0; )
{
tsn = sctp->i.r.firsttns + u;
sctp->i.received[(tsn>>3)%sizeof(sctp->i.received)] |= 1<<(tsn&7);
}
if (sctp->i.r.toobig)
;/*ignore it when it cannot be handled*/
else if (sctp->i.r.ppid == BigLong(SCTP_PPID_DATA))
{
memmove(net_message.data, sctp->i.r.buf, sctp->i.r.size);
net_message.cursize = sctp->i.r.size;
col->ReadGamePacket();
if (net_message.cursize != sctp->i.r.size)
{
net_message.cursize = 0;
return; //something weird happened...
}
}
else if (sctp->i.r.ppid == BigLong(SCTP_PPID_DCEP))
SCTP_DecodeDCEP(sctp, peer, resp);
}
}
//FIXME: we don't handle reordering properly at all.
// if (c->flags & 4)
// Con_Printf("\tUnordered\n");
// Con_Printf("\tStream Id %i\n", BigShort(dc->sid));
// Con_Printf("\tStream Seq %i\n", BigShort(dc->seq));
// Con_Printf("\tPPID %i\n", BigLong(dc->ppid));
}
}
break;
case SCTP_TYPE_INIT:
case SCTP_TYPE_INITACK:
if (clen >= sizeof(struct sctp_chunk_init_s))
{
qboolean isack = c->type==SCTP_TYPE_INITACK;
struct sctp_chunk_init_s *init = (void*)c;
struct {
quint16_t ptype;
quint16_t plen;
} *p = (void*)(init+1);
sctp->i.ctsn = BigLong(init->tsn)-1;
sctp->i.htsn = 0;
sctp->o.verifycode = init->verifycode;
(void)BigLong(init->arwc);
(void)BigShort(init->numoutstreams);
(void)BigShort(init->numinstreams);
while ((qbyte*)p+sizeof(*p) <= (qbyte*)c+clen)
{
unsigned short ptype = BigShort(p->ptype);
unsigned short plen = BigShort(p->plen);
switch(ptype)
{
case 7: //init cookie
if (sctp->cookie)
Z_Free(sctp->cookie);
sctp->cookiesize = plen - sizeof(*p);
sctp->cookie = Z_Malloc(sctp->cookiesize);
memcpy(sctp->cookie, p+1, sctp->cookiesize);
break;
case 32776: //ASCONF
break;
case 49152:
sctp->peerhasfwdtsn = true;
break;
default:
Con_DPrintf("SCTP: Found unknown init parameter %i||%#x\n", ptype, ptype);
break;
}
p = (void*)((qbyte*)p + ((plen+3)&~3));
}
if (isack)
{
sctp->nextreinit = 0;
if (sctp->cookie)
SCTP_Transmit(sctp, peer, NULL, 0); //make sure we send acks occasionally even if we have nothing else to say.
}
else
{
struct sctp_header_s *rh = (void*)resp;
struct sctp_chunk_init_s *rinit = (void*)(rh+1);
struct {
quint16_t ptype;
quint16_t plen;
struct {
qbyte data[16];
} cookie;
} *rinitcookie = (void*)(rinit+1);
struct {
quint16_t ptype;
quint16_t plen;
} *rftsn = (void*)(rinitcookie+1);
qbyte *end = sctp->peerhasfwdtsn?(void*)(rftsn+1):(void*)(rinitcookie+1);
rh->srcport = sctp->myport;
rh->dstport = sctp->peerport;
rh->verifycode = sctp->o.verifycode;
rh->crc = 0;
*rinit = *init;
rinit->chunk.type = SCTP_TYPE_INITACK;
rinit->chunk.flags = 0;
rinit->chunk.length = BigShort(end-(qbyte*)rinit);
rinit->verifycode = sctp->i.verifycode;
rinit->arwc = BigLong(65536);
rinit->numoutstreams = init->numoutstreams;
rinit->numinstreams = init->numinstreams;
rinit->tsn = BigLong(sctp->o.tsn);
rinitcookie->ptype = BigShort(7);
rinitcookie->plen = BigShort(sizeof(*rinitcookie));
memcpy(&rinitcookie->cookie, "deadbeefdeadbeef", sizeof(rinitcookie->cookie)); //frankly the contents of the cookie are irrelevant to anything. we've already verified the peer's ice pwd/ufrag stuff as well as their dtls certs etc.
rftsn->ptype = BigShort(49152);
rftsn->plen = BigShort(sizeof(*rftsn));
//complete. calc the proper crc and send it off.
rh->crc = SCTP_Checksum(rh, end-resp);
SCTP_PeerSendPacket(peer, end-resp, rh);
}
}
break;
case SCTP_TYPE_SACK:
if (clen >= sizeof(struct sctp_chunk_sack_s))
{
struct sctp_chunk_sack_s *sack = (void*)c;
quint32_t tsn = BigLong(sack->tsn);
sctp->o.ctsn = tsn;
sctp->o.losttsn = BigShort(sack->gaps); //if there's a gap then they're telling us they got a later one.
//Con_Printf(CON_ERROR"Sack %#x (%i in flight)\n"
// "\tgaps: %i, dupes %i\n",
// tsn, sctp->o.tsn-tsn,
// BigShort(sack->gaps), BigShort(sack->dupes));
}
break;
case SCTP_TYPE_PING:
if (clen >= sizeof(struct sctp_chunk_s))
{
struct sctp_chunk_s *ping = (void*)c;
struct sctp_header_s *pongh = Z_Malloc(sizeof(*pongh) + clen);
pongh->srcport = sctp->myport;
pongh->dstport = sctp->peerport;
pongh->verifycode = sctp->o.verifycode;
pongh->crc = 0;
memcpy(pongh+1, ping, clen);
((struct sctp_chunk_s*)(pongh+1))->type = SCTP_TYPE_PONG;
//complete. calc the proper crc and send it off.
pongh->crc = SCTP_Checksum(pongh, sizeof(*pongh) + clen);
SCTP_PeerSendPacket(peer, sizeof(*pongh) + clen, pongh);
Z_Free(pongh);
}
break;
// case SCTP_TYPE_PONG: //we don't send pings
case SCTP_TYPE_ABORT:
ICE_Set(peer, "state", STRINGIFY(ICE_FAILED));
SCTP_ErrorChunk("Abort", (struct sctp_errorcause_s*)(c+1), clen-sizeof(*c));
break;
case SCTP_TYPE_SHUTDOWN: //FIXME. we should send an ack...
ICE_Set(peer, "state", STRINGIFY(ICE_FAILED));
Con_DPrintf(CON_ERROR"SCTP: Shutdown\n");
break;
// case SCTP_TYPE_SHUTDOWNACK: //we don't send shutdowns, cos we're lame like that...
case SCTP_TYPE_ERROR:
//not fatal...
SCTP_ErrorChunk("Error", (struct sctp_errorcause_s*)(c+1), clen-sizeof(*c));
break;
case SCTP_TYPE_COOKIEECHO:
if (clen >= sizeof(struct sctp_chunk_s))
{
struct sctp_header_s *rh = (void*)resp;
struct sctp_chunk_s *rack = (void*)(rh+1);
qbyte *end = (void*)(rack+1);
rh->srcport = sctp->myport;
rh->dstport = sctp->peerport;
rh->verifycode = sctp->o.verifycode;
rh->crc = 0;
rack->type = SCTP_TYPE_COOKIEACK;
rack->flags = 0;
rack->length = BigShort(sizeof(*rack));
//complete. calc the proper crc and send it off.
rh->crc = SCTP_Checksum(rh, end-resp);
SCTP_PeerSendPacket(peer, end-resp, rh);
sctp->o.writable = true; //channel SHOULD now be open for data.
}
break;
case SCTP_TYPE_COOKIEACK:
sctp->o.writable = true; //we know the other end is now open.
break;
case SCTP_TYPE_FORWARDTSN:
if (clen >= sizeof(struct sctp_chunk_fwdtsn_s))
{
struct sctp_chunk_fwdtsn_s *fwd = (void*)c;
quint32_t tsn = BigLong(fwd->tsn), count;
count = tsn - sctp->i.ctsn;
if ((int)count < 0)
break; //overflow? don't go backwards.
if (count > 1024)
count = 1024; //don't advance too much in one go. we'd block and its probably an error anyway.
while(count --> 0)
{
tsn = ++sctp->i.ctsn;
sctp->i.received[(tsn>>3)%sizeof(sctp->i.received)] &= ~(1<<(tsn&7));
if (sctp->i.htsn)
sctp->i.htsn--;
sctp->i.ackneeded++; //flag for a sack if we actually completed something here.
}
}
break;
// case SCTP_TYPE_SHUTDOWNDONE:
safedefault:
//no idea what this chunk is, just ignore it...
Con_DPrintf("SCTP: Unsupported chunk %i\n", c->type);
break;
}
c = (struct sctp_chunk_s*)((qbyte*)c + ((clen+3)&~3)); //next chunk is 4-byte aligned.
}
if (sctp->i.ackneeded >= 5)
SCTP_Transmit(sctp, peer, NULL, 0); //make sure we send acks occasionally even if we have nothing else to say.
//we already made sense of it all.
net_message.cursize = 0;
}
//========================================
#endif
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#if defined(SUPPORT_ICE) || defined(MASTERONLY)
qboolean ICE_WasStun(ftenet_connections_t *col)
{
#ifdef SUPPORT_ICE
if (net_from.type == NA_ICE)
return false; //this stuff over an ICE connection doesn't make sense.
#endif
// if (net_from.prot != NP_DGRAM)
// return false;
#if defined(HAVE_CLIENT) && defined(VOICECHAT)
if (col == cls.sockets)
{
if (NET_RTP_Parse())
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return true;
}
#endif
if ((net_from.type == NA_IP || net_from.type == NA_IPV6) && net_message.cursize >= 20 && *net_message.data<2)
{
stunhdr_t *stun = (stunhdr_t*)net_message.data;
int stunlen = BigShort(stun->msglen);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#ifdef SUPPORT_ICE
if ((stun->msgtype == BigShort(STUN_BINDING|STUN_REPLY) || stun->msgtype == BigShort(STUN_BINDING|STUN_ERROR)) && net_message.cursize == stunlen + sizeof(*stun))
{
//binding reply (or error)
netadr_t adr = net_from;
char xor[16];
short portxor;
stunattr_t *attr = (stunattr_t*)(stun+1);
int alen;
unsigned short attrval;
int err = 0;
char errmsg[64];
*errmsg = 0;
adr.type = NA_INVALID;
while(stunlen)
{
stunlen -= sizeof(*attr);
alen = (unsigned short)BigShort(attr->attrlen);
if (alen > stunlen)
return false;
stunlen -= (alen+3)&~3;
attrval = BigShort(attr->attrtype);
switch(attrval)
{
case STUNATTR_USERNAME:
case STUNATTR_MESSAGEINTEGRITIY:
break;
default:
if (attrval & 0x8000)
break; //okay to ignore
return true;
case STUNATTR_MAPPED_ADDRESS:
if (adr.type != NA_INVALID)
break; //ignore it if we already got an address...
//fallthrough
case STUNATTR_XOR_MAPPED_ADDRESS:
if (attrval == STUNATTR_XOR_MAPPED_ADDRESS)
{
portxor = *(short*)&stun->magiccookie;
memcpy(xor, &stun->magiccookie, sizeof(xor));
}
else
{
portxor = 0;
memset(xor, 0, sizeof(xor));
}
if (alen == 8 && ((qbyte*)attr)[5] == 1) //ipv4
{
adr.type = NA_IP;
adr.port = (((short*)attr)[3]) ^ portxor;
*(int*)adr.address.ip = *(int*)(&((qbyte*)attr)[8]) ^ *(int*)xor;
}
else if (alen == 20 && ((qbyte*)attr)[5] == 2) //ipv6
{
adr.type = NA_IPV6;
adr.port = (((short*)attr)[3]) ^ portxor;
((int*)adr.address.ip6)[0] = ((int*)&((qbyte*)attr)[8])[0] ^ ((int*)xor)[0];
((int*)adr.address.ip6)[1] = ((int*)&((qbyte*)attr)[8])[1] ^ ((int*)xor)[1];
((int*)adr.address.ip6)[2] = ((int*)&((qbyte*)attr)[8])[2] ^ ((int*)xor)[2];
((int*)adr.address.ip6)[3] = ((int*)&((qbyte*)attr)[8])[3] ^ ((int*)xor)[3];
}
break;
case STUNATTR_ERROR_CODE:
{
unsigned short len = BigShort(attr->attrlen)-4;
if (len > sizeof(errmsg)-1)
len = sizeof(errmsg)-1;
memcpy(errmsg, &((qbyte*)attr)[8], len);
errmsg[len] = 0;
if (err==0)
err = (((qbyte*)attr)[6]*100) + (((qbyte*)attr)[7]%100);
}
break;
}
alen = (alen+3)&~3;
attr = (stunattr_t*)((char*)(attr+1) + alen);
}
if (err)
{
char sender[256];
if (net_ice_debug.ival >= 1)
Con_Printf("%s: Stun error code %u : %s\n", NET_AdrToString(sender, sizeof(sender), &net_from), err, errmsg);
}
else if (adr.type!=NA_INVALID && !err)
{
struct icestate_s *con;
for (con = icelist; con; con = con->next)
{
struct icecandidate_s *rc;
size_t i;
struct iceserver_s *s;
if (con->mode == ICEM_RAW)
continue;
for (i = 0; i < con->servers; i++)
{
s = &con->server[i];
if (NET_CompareAdr(&net_from, &s->addr))
{ //check to see if this is a new server-reflexive address, which happens when the peer is behind a nat.
for (rc = con->lc; rc; rc = rc->next)
{
if (NET_CompareAdr(&adr, &rc->peer))
break;
}
if (!rc)
{
//netadr_t reladdr;
//int relflags;
//const char *relpath;
int rnd[2];
struct icecandidate_s *src; //server Reflexive Candidate
char str[256];
src = Z_Malloc(sizeof(*src));
src->next = con->lc;
con->lc = src;
src->peer = adr;
NET_BaseAdrToString(src->info.addr, sizeof(src->info.addr), &adr);
src->info.port = ntohs(adr.port);
//if (net_from.connum >= 1 && net_from.connum < 1+MAX_CONNECTIONS && col->conn[net_from.connum-1])
// col->conn[net_from.connum-1]->GetLocalAddresses(col->conn[net_from.connum-1], &relflags, &reladdr, &relpath, 1);
//FIXME: we don't really know which one... NET_BaseAdrToString(src->info.reladdr, sizeof(src->info.reladdr), &reladdr);
//src->info.relport = ntohs(reladdr.port);
src->info.type = ICE_SRFLX;
src->info.component = 1;
src->info.network = net_from.connum;
src->dirty = true;
src->info.priority = ICE_ComputePriority(&src->peer, &src->info); //FIXME
Sys_RandomBytes((void*)rnd, sizeof(rnd));
Q_strncpyz(src->info.candidateid, va("x%08x%08x", rnd[0], rnd[1]), sizeof(src->info.candidateid));
if (net_ice_debug.ival >= 1)
Con_Printf(S_COLOR_GRAY"[%s]: Public address: %s\n", con->friendlyname, NET_AdrToString(str, sizeof(str), &adr));
}
s->stunretry = Sys_Milliseconds() + 60*1000;
return true;
}
}
//check to see if this is a new peer-reflexive address, which happens when the peer is behind a nat.
for (rc = con->rc; rc; rc = rc->next)
{
if (NET_CompareAdr(&net_from, &rc->peer))
{
if (!(rc->reachable & (1u<<(net_from.connum-1))))
{
char str[256];
if (net_ice_debug.ival >= 1)
Con_Printf(S_COLOR_GRAY"[%s]: We can reach %s\n", con->friendlyname, NET_AdrToString(str, sizeof(str), &net_from));
}
rc->reachable |= 1u<<(net_from.connum-1);
rc->reached = Sys_Milliseconds();
if (NET_CompareAdr(&net_from, &con->chosenpeer) && (stun->transactid[2] & BigLong(0x80000000)))
{
if (con->state == ICE_CONNECTING)
ICE_Set(con, "state", STRINGIFY(ICE_CONNECTED));
}
}
}
}
}
return true;
}
else if (stun->msgtype == BigShort(STUN_BINDING|STUN_INDICATION))// && net_message.cursize == stunlen + sizeof(*stun) && stun->magiccookie == BigLong(0x2112a442))
{
//binding indication. used as an rtp keepalive. should have a fingerprint
return true;
}
else if (stun->msgtype == BigShort(STUN_DATA|STUN_INDICATION)
&& net_message.cursize == stunlen + sizeof(*stun) && stun->magiccookie == BigLong(STUN_MAGIC_COOKIE))
{
//TURN relayed data
//these MUST come from a _known_ turn server.
netadr_t adr;
char xor[16];
short portxor;
void *data = NULL;
unsigned short datasize = 0;
unsigned short attrval;
stunattr_t *attr = (stunattr_t*)(stun+1);
int alen;
unsigned int network = net_from.connum-1; //also net_from_connection->connum
struct icestate_s *con;
if (network < MAX_CONNECTIONS)
return true; //don't handle this if its on the non-turn sockets.
network -= MAX_CONNECTIONS;
if (network < countof(con->server))
return true; //don't double-decapsulate...
network -= countof(con->server);
for (con = icelist; con; con = con->next)
{
if (network < con->servers && net_from_connection == con->server[network].con)
break;
}
if (!con)
return true; //don't know what it was. just ignore it.
if (network >= con->servers || !NET_CompareAdr(&net_from, &con->server[network].addr))
return true; //right socket, but not from the server that we expected...
adr.type = NA_INVALID;
while(stunlen>0)
{
stunlen -= sizeof(*attr);
alen = (unsigned short)BigShort(attr->attrlen);
if (alen > stunlen)
return false;
stunlen -= (alen+3)&~3;
attrval = BigShort(attr->attrtype);
switch(attrval)
{
default:
if (attrval & 0x8000)
break; //okay to ignore
return true;
case STUNATTR_DATA:
data = attr+1;
datasize = alen;
break;
case STUNATTR_XOR_PEER_ADDRESS:
//always xor
portxor = *(short*)&stun->magiccookie;
memcpy(xor, &stun->magiccookie, sizeof(xor));
adr.prot = NP_DGRAM;
adr.connum = net_from.connum;
adr.scopeid = net_from.scopeid;
if (alen == 8 && ((qbyte*)attr)[5] == 1) //ipv4
{
adr.type = NA_IP;
adr.port = (((short*)attr)[3]) ^ portxor;
*(int*)adr.address.ip = *(int*)(&((qbyte*)attr)[8]) ^ *(int*)xor;
}
else if (alen == 20 && ((qbyte*)attr)[5] == 2) //ipv6
{
adr.type = NA_IPV6;
adr.port = (((short*)attr)[3]) ^ portxor;
((int*)adr.address.ip6)[0] = ((int*)&((qbyte*)attr)[8])[0] ^ ((int*)xor)[0];
((int*)adr.address.ip6)[1] = ((int*)&((qbyte*)attr)[8])[1] ^ ((int*)xor)[1];
((int*)adr.address.ip6)[2] = ((int*)&((qbyte*)attr)[8])[2] ^ ((int*)xor)[2];
((int*)adr.address.ip6)[3] = ((int*)&((qbyte*)attr)[8])[3] ^ ((int*)xor)[3];
}
break;
}
alen = (alen+3)&~3;
attr = (stunattr_t*)((char*)(attr+1) + alen);
}
if (data)
{
memmove(net_message.data, data, net_message.cursize = datasize);
adr.connum = net_from.connum-countof(con->server); //came via the relay.
net_from = adr;
col->ReadGamePacket();
return true;
}
}
else if ((stun->msgtype == BigShort(STUN_CREATEPERM|STUN_REPLY) || stun->msgtype == BigShort(STUN_CREATEPERM|STUN_ERROR))
&& net_message.cursize == stunlen + sizeof(*stun) && stun->magiccookie == BigLong(STUN_MAGIC_COOKIE))
{
//TURN CreatePermissions reply (or error)
unsigned short attrval;
stunattr_t *attr = (stunattr_t*)(stun+1), *nonce=NULL, *realm=NULL;
int alen;
struct iceserver_s *s = NULL;
int i, j;
struct icestate_s *con;
char errmsg[128];
int err = 0;
*errmsg = 0;
//make sure it makes sense.
while(stunlen>0)
{
stunlen -= sizeof(*attr);
alen = (unsigned short)BigShort(attr->attrlen);
if (alen > stunlen)
return false;
stunlen -= (alen+3)&~3;
attrval = BigShort(attr->attrtype);
switch(attrval)
{
default:
if (attrval & 0x8000)
break; //okay to ignore
return true;
case STUNATTR_NONCE:
nonce = attr;
break;
case STUNATTR_REALM:
realm = attr;
break;
case STUNATTR_ERROR_CODE:
{
unsigned short len = BigShort(attr->attrlen)-4;
if (len > sizeof(errmsg)-1)
len = sizeof(errmsg)-1;
memcpy(errmsg, &((qbyte*)attr)[8], len);
errmsg[len] = 0;
if (err==0)
err = (((qbyte*)attr)[6]*100) + (((qbyte*)attr)[7]%100);
}
break;
case STUNATTR_MESSAGEINTEGRITIY:
break;
}
alen = (alen+3)&~3;
attr = (stunattr_t*)((char*)(attr+1) + alen);
}
//now figure out what it acked.
for (con = icelist; con; con = con->next)
{
for (i = 0; i < con->servers; i++)
{
s = &con->server[i];
if (NET_CompareAdr(&net_from, &s->addr))
for (j = 0; j < s->peers; j++)
{
if (s->peer[j].stunrnd[0] == stun->transactid[0] && s->peer[j].stunrnd[1] == stun->transactid[1] && s->peer[j].stunrnd[2] == stun->transactid[2])
{ //the lifetime of a permission is a fixed 5 mins (this is separately from the port allocation)
unsigned int now = Sys_Milliseconds();
if (err)
{
if (err == 438 && realm && nonce)
{
alen = BigShort(nonce->attrlen);
Z_Free(s->nonce);
s->nonce = Z_Malloc(alen+1);
memcpy(s->nonce, nonce+1, alen);
s->nonce[alen] = 0;
alen = BigShort(realm->attrlen);
Z_Free(s->realm);
s->realm = Z_Malloc(alen+1);
memcpy(s->realm, realm+1, alen);
s->realm[alen] = 0;
s->peer[j].retry = now; //retry fast.
}
}
else
{
now -= 25; //we don't know when it acked, so lets pretend we're a few MS ago.
s->peer[j].expires = now + 5*60*1000;
s->peer[j].retry = now + 4*60*1000; //start trying to refresh it a min early (which will do resends).
}
//next attempt will use a different id.
Sys_RandomBytes((char*)s->peer[i].stunrnd, sizeof(s->peer[i].stunrnd));
return true;
}
}
}
if (i < con->servers)
break;
}
return true;
}
else if ((stun->msgtype == BigShort(STUN_ALLOCATE|STUN_REPLY) || stun->msgtype == BigShort(STUN_ALLOCATE|STUN_ERROR)||
(stun->msgtype == BigShort(STUN_REFRESH|STUN_REPLY) || stun->msgtype == BigShort(STUN_REFRESH|STUN_ERROR)))
&& net_message.cursize == stunlen + sizeof(*stun) && stun->magiccookie == BigLong(STUN_MAGIC_COOKIE))
{
//TURN allocate reply (or error)
netadr_t adrs[2], ladr, *adr; //the last should be our own ip.
char xor[16];
short portxor;
unsigned short attrval;
stunattr_t *attr = (stunattr_t*)(stun+1);
int alen;
int err = 0;
char errmsg[64];
struct iceserver_s *s = NULL;
int i;
struct icestate_s *con;
qboolean noncechanged = false;
unsigned int lifetime = 0;
//gotta have come from our private socket.
unsigned int network = net_from.connum-1; //also net_from_connection->connum
if (network < MAX_CONNECTIONS)
return true; //don't handle this if its on the non-turn sockets.
network -= MAX_CONNECTIONS;
if (network < countof(con->server))
return true; //TURN-over-TURN is bad...
network -= countof(con->server);
for (con = icelist; con; con = con->next)
{
if (network < con->servers && net_from_connection == con->server[network].con)
{
s = &con->server[network];
if (s->stunrnd[0] == stun->transactid[0] && s->stunrnd[1] == stun->transactid[1] && s->stunrnd[2] == stun->transactid[2] && NET_CompareAdr(&net_from, &s->addr))
break;
Con_Printf("Stale transaction id (got %x, expected %x)\n", stun->transactid[0], s->stunrnd[0]);
}
}
if (!con)
return true; //don't know what it was. just ignore it.
network += 1 + MAX_CONNECTIONS; //fix it up to refer to the relay rather than the private socket.
adrs[0].type = NA_INVALID;
adrs[1].type = NA_INVALID;
ladr.type = NA_INVALID;
while(stunlen>0)
{
stunlen -= sizeof(*attr);
alen = (unsigned short)BigShort(attr->attrlen);
if (alen > stunlen)
return false;
stunlen -= (alen+3)&~3;
attrval = BigShort(attr->attrtype);
switch(attrval)
{
case STUNATTR_LIFETIME:
if (alen >= 4)
lifetime = BigLong(*(int*)(attr+1));
break;
// case STUNATTR_SOFTWARE:
case STUNATTR_MESSAGEINTEGRITIY:
// case STUNATTR_FINGERPRINT:
break;
default:
if (attrval & 0x8000)
break; //okay to ignore
err = -1; //got an attribute we 'must' handle...
return true;
case STUNATTR_NONCE:
Z_Free(s->nonce);
s->nonce = Z_Malloc(alen+1);
memcpy(s->nonce, attr+1, alen);
s->nonce[alen] = 0;
noncechanged = true;
break;
case STUNATTR_REALM:
Z_Free(s->realm);
s->realm = Z_Malloc(alen+1);
memcpy(s->realm, attr+1, alen);
s->realm[alen] = 0;
break;
case STUNATTR_XOR_RELAYED_ADDRESS:
case STUNATTR_XOR_MAPPED_ADDRESS:
if (BigShort(attr->attrtype) == STUNATTR_XOR_MAPPED_ADDRESS)
adr = &ladr;
else
adr = adrs[0].type?&adrs[1]:&adrs[0];
//always xor
portxor = *(short*)&stun->magiccookie;
memcpy(xor, &stun->magiccookie, sizeof(xor));
adr->prot = NP_DGRAM;
adr->connum = net_from.connum;
adr->scopeid = net_from.scopeid;
if (alen == 8 && ((qbyte*)attr)[5] == 1) //ipv4
{
adr->type = NA_IP;
adr->port = (((short*)attr)[3]) ^ portxor;
*(int*)adr->address.ip = *(int*)(&((qbyte*)attr)[8]) ^ *(int*)xor;
}
else if (alen == 20 && ((qbyte*)attr)[5] == 2) //ipv6
{
adr->type = NA_IPV6;
adr->port = (((short*)attr)[3]) ^ portxor;
((int*)adr->address.ip6)[0] = ((int*)&((qbyte*)attr)[8])[0] ^ ((int*)xor)[0];
((int*)adr->address.ip6)[1] = ((int*)&((qbyte*)attr)[8])[1] ^ ((int*)xor)[1];
((int*)adr->address.ip6)[2] = ((int*)&((qbyte*)attr)[8])[2] ^ ((int*)xor)[2];
((int*)adr->address.ip6)[3] = ((int*)&((qbyte*)attr)[8])[3] ^ ((int*)xor)[3];
}
break;
case STUNATTR_ERROR_CODE:
{
unsigned short len = BigShort(attr->attrlen)-4;
if (len > sizeof(errmsg)-1)
len = sizeof(errmsg)-1;
memcpy(errmsg, &((qbyte*)attr)[8], len);
errmsg[len] = 0;
if (err==0)
err = (((qbyte*)attr)[6]*100) + (((qbyte*)attr)[7]%100);
}
break;
}
alen = (alen+3)&~3;
attr = (stunattr_t*)((char*)(attr+1) + alen);
}
if (err)
{
char sender[256];
if (err == 438/*stale nonce*/)
{ //reset everything.
s->state = noncechanged?TURN_HAVE_NONCE:TURN_UNINITED;
s->stunretry = Sys_Milliseconds();
if (net_ice_debug.ival >= 1)
Con_Printf("[%s]: %s: TURN error code %u : %s\n", con->friendlyname, NET_AdrToString(sender, sizeof(sender), &net_from), err, errmsg);
}
else if (err == 403/*forbidden*/) //something bad...
{
s->state = TURN_UNINITED, s->stunretry = Sys_Milliseconds() + 60*1000;
if (net_ice_debug.ival >= 1)
Con_Printf("[%s]: %s: TURN error code %u : %s\n", con->friendlyname, NET_AdrToString(sender, sizeof(sender), &net_from), err, errmsg);
}
else if (err == 401 && s->state == TURN_UNINITED && s->nonce) //failure when sending auth... give up for a min
{ //this happens from initial auth. we need to reply with the real auth request now.
s->state = TURN_HAVE_NONCE, s->stunretry = Sys_Milliseconds();
}
else
{
s->stunretry = Sys_Milliseconds() + 60*1000;
// if (net_ice_debug.ival >= 1)
Con_Printf(CON_ERROR"[%s]: %s: TURN error code %u : %s\n", con->friendlyname, NET_AdrToString(sender, sizeof(sender), &net_from), err, errmsg);
}
}
else
{
struct icecandidate_s *lc;
for (i = 0; i < countof(adrs); i++)
{
if (adrs[i].type != NA_INVALID && stun->msgtype == BigShort(STUN_ALLOCATE|STUN_REPLY))
{
s->state = TURN_ALLOCATED;
if (!i)
s->family = adrs[i].type;
if (s->family != adrs[i].type)
s->family = NA_INVALID; //send it both types.
if (ladr.type != NA_INVALID)
adr = &ladr; //can give a proper reflexive address
else
adr = &adrs[i]; //no info... give something.
//check to see if this is a new server-reflexive address, which happens when the peer is behind a nat.
for (lc = con->lc; lc; lc = lc->next)
{
if (NET_CompareAdr(&adrs[i], &lc->peer))
break;
}
if (!lc)
{
int rnd[2];
struct icecandidate_s *src; //server Reflexive Candidate
char str[256];
src = Z_Malloc(sizeof(*src));
src->next = con->lc;
con->lc = src;
src->peer = adrs[i];
NET_BaseAdrToString(src->info.addr, sizeof(src->info.addr), &adrs[i]);
src->info.port = ntohs(adrs[i].port);
NET_BaseAdrToString(src->info.reladdr, sizeof(src->info.reladdr), adr);
src->info.relport = ntohs(adr->port);
src->info.type = ICE_RELAY;
src->info.component = 1;
src->info.network = network;
src->dirty = true;
src->info.priority = ICE_ComputePriority(&adrs[i], &src->info);
Sys_RandomBytes((void*)rnd, sizeof(rnd));
Q_strncpyz(src->info.candidateid, va("x%08x%08x", rnd[0], rnd[1]), sizeof(src->info.candidateid));
if (net_ice_debug.ival >= 1)
Con_Printf(S_COLOR_GRAY"[%s]: Relayed local candidate: %s\n", con->friendlyname, NET_AdrToString(str, sizeof(str), &adrs[i]));
}
}
}
if (lifetime < 60) //don't spam reauth requests too often...
lifetime = 60;
s->stunretry = Sys_Milliseconds() + (lifetime-50)*1000;
return true;
}
return true;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#endif
if (stun->msgtype == BigShort(STUN_BINDING|STUN_REQUEST) && net_message.cursize == stunlen + sizeof(*stun) && stun->magiccookie == BigLong(STUN_MAGIC_COOKIE))
{
char username[256];
char integrity[20];
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#ifdef SUPPORT_ICE
struct icestate_s *con;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
int role = 0;
unsigned int tiehigh = 0;
unsigned int tielow = 0;
qboolean usecandidate = false;
unsigned int priority = 0;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#endif
char *integritypos = NULL;
int error = 0;
sizebuf_t buf;
char data[512];
int alen = 0, atype = 0, aofs = 0;
int crc;
//binding request
stunattr_t *attr = (stunattr_t*)(stun+1);
*username = 0;
while(stunlen)
{
alen = (unsigned short)BigShort(attr->attrlen);
if (alen+sizeof(*attr) > stunlen)
return false;
switch((unsigned short)BigShort(attr->attrtype))
{ case 0xc057: /*'network cost'*/ break;
default:
//unknown attributes < 0x8000 are 'mandatory to parse', and such packets must be dropped in their entirety.
//other ones are okay.
if (!((unsigned short)BigShort(attr->attrtype) & 0x8000))
return false;
break;
case STUNATTR_USERNAME:
if (alen < sizeof(username))
{
memcpy(username, attr+1, alen);
username[alen] = 0;
// Con_Printf("Stun username = \"%s\"\n", username);
}
break;
case STUNATTR_MESSAGEINTEGRITIY:
memcpy(integrity, attr+1, sizeof(integrity));
integritypos = (char*)(attr+1);
break;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#ifdef SUPPORT_ICE
case STUNATTR_ICE_PRIORITY:
// Con_Printf("priority = \"%i\"\n", priority);
priority = BigLong(*(int*)(attr+1));
break;
case STUNATTR_ICE_USE_CANDIDATE:
usecandidate = true;
break;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#endif
case STUNATTR_FINGERPRINT:
// Con_Printf("fingerprint = \"%08x\"\n", BigLong(*(int*)(attr+1)));
break;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#ifdef SUPPORT_ICE
case STUNATTR_ICE_CONTROLLED:
case STUNATTR_ICE_CONTROLLING:
role = (unsigned short)BigShort(attr->attrtype);
tiehigh = BigLong(((int*)(attr+1))[0]);
tielow = BigLong(((int*)(attr+1))[1]);
break;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#endif
}
alen = (alen+3)&~3;
attr = (stunattr_t*)((char*)(attr+1) + alen);
stunlen -= alen+sizeof(*attr);
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#ifdef SUPPORT_ICE
if (*username || integritypos)
{
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
//we need to know which connection its from in order to validate the integrity
for (con = icelist; con; con = con->next)
{
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (!strcmp(va("%s:%s", con->lufrag, con->rufrag), username))
break;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (!con)
{
if (net_ice_debug.ival >= 2)
Con_Printf("Received STUN request from unknown user \"%s\"\n", username);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return true;
}
/*else if (con->chosenpeer.type != NA_INVALID)
{ //got one.
if (!NET_CompareAdr(&net_from, &con->chosenpeer))
return true; //FIXME: we're too stupid to handle switching. pretend to be dead.
}*/
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
else if (con->state == ICE_INACTIVE)
return true; //bad timing
else
{
struct icecandidate_s *rc;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (net_ice_debug.ival >= 2)
Con_Printf(S_COLOR_GRAY"[%s]: got binding request on %s from %s\n", con->friendlyname, ICE_NetworkToName(con, net_from.connum), NET_AdrToString(username,sizeof(username), &net_from));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (integritypos)
{
char key[20];
//the hmac is a bit weird. the header length includes the integrity attribute's length, but the checksum doesn't even consider the attribute header.
stun->msglen = BigShort(integritypos+sizeof(integrity) - (char*)stun - sizeof(*stun));
openxr plugin: tweaked - inputs should be working properly now, and are visible to csqc. subject to further breaking changes, however. _pext_vrinputs: added cvar to enable vr inputs protocol extension allowing vr inputs to be networked to ssqc too. defaults to 0 for now, will be renamed when deemed final. updates menu: the prompt to enable sources is now more explicit instead of expecting the user to have a clue. updates menu: added a v3 sources format, which should be more maintainable. not final. updates menu: try to give reasons why sources might be failing (to help blame ISPs if they try fucking over TTH dns again). presets menu: no longer closes the instant a preset is chosen. some presets have a couple of modifiers listed. force the demo loop in the background to serve as a preview. prompts menus: now does word wrapping. ftemaster: support importing server lists from other master servers (requested by Eukara). server: try to detect when non-reply inbound packets are blocked by firewalls/nats/etc (using ftemaster to do so). qcvm: added pointcontentsmask builtin, allowing it to probe more than just world, with fte's full contentbit range instead of just q1 legacy. qcvm: memfill8 builtin now works on createbuffer() pointers. qcvm: add missing unsigned ops. Fixed double comparison ops. fixed bug with op_store_i64. added missing OP_LOADP_I64 qcc: added '#pragma framerate RATE' for overriding implicit nextthink durations. qcc: fixed '#pragma DONT_COMPILE_THIS_FILE' to not screw up comments. qcc: added __GITURL__ __GITHASH__ __GITDATE__ __GITDATETIME__ __GITDESC__ for any mods that might want to make use of that. qcc: fix up -Fhashonly a little setrenderer: support for vulkan gpu enumeration. rulesets: reworked to support custom rulesets (using hashes to catch haxxors, though still nothing prevents just changing the client to ignore rulesets) bspx: use our BIH code for the bspx BRUSHLIST lump instead of the older less efficient code. (static)iqm+obj: these model formats can now be used for the worldmodel (with a suitable .ent file). Also using BIH for much better collision performance. pmove: tried to optimise PM_NudgePosition, should boost fps in stress tests. wayland: fix a crash on startup. mousegrabs now works better. imagetool: uses sdl for previews. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5813 fc73d0e0-1445-4013-8a0c-d673dee63da5
2021-04-14 05:21:04 +00:00
CalcHMAC(&hash_sha1, key, sizeof(key), (qbyte*)stun, integritypos-4 - (char*)stun, con->lpwd, strlen(con->lpwd));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (memcmp(key, integrity, sizeof(integrity)))
{
Con_DPrintf(CON_WARNING"Integrity is bad! needed %x got %x\n", *(int*)key, *(int*)integrity);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return true;
}
}
//check to see if this is a new peer-reflexive address, which happens when the peer is behind a nat.
for (rc = con->rc; rc; rc = rc->next)
{
if (NET_CompareAdr(&net_from, &rc->peer))
break;
}
if (!rc)
{
//netadr_t reladdr;
//int relflags;
//const char *relpath;
struct icecandidate_s *rc;
rc = Z_Malloc(sizeof(*rc));
rc->next = con->rc;
con->rc = rc;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
rc->peer = net_from;
NET_BaseAdrToString(rc->info.addr, sizeof(rc->info.addr), &net_from);
rc->info.port = ntohs(net_from.port);
//if (net_from.connum >= 1 && net_from.connum < 1+MAX_CONNECTIONS && col->conn[net_from.connum-1])
// col->conn[net_from.connum-1]->GetLocalAddresses(col->conn[net_from.connum-1], &relflags, &reladdr, &relpath, 1);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
//FIXME: we don't really know which one... NET_BaseAdrToString(rc->info.reladdr, sizeof(rc->info.reladdr), &reladdr);
//rc->info.relport = ntohs(reladdr.port);
rc->info.type = ICE_PRFLX;
rc->dirty = true;
rc->info.priority = priority;
}
//flip ice control role, if we're wrong.
if (role && role != (con->controlled?STUNATTR_ICE_CONTROLLING:STUNATTR_ICE_CONTROLLED))
{
if (tiehigh == con->tiehigh && tielow == con->tielow)
{
Con_Printf("ICE: Evil loopback hack enabled\n");
if (usecandidate)
{
if ((con->chosenpeer.connum != net_from.connum || !NET_CompareAdr(&con->chosenpeer, &net_from)) && net_ice_debug.ival >= 1)
{
char msg[64];
if (con->chosenpeer.connum-1 >= MAX_CONNECTIONS)
Con_Printf(S_COLOR_GRAY"[%s]: New peer imposed %s, via %s.\n", con->friendlyname, NET_AdrToString(msg, sizeof(msg), &net_from), ICE_NetworkToName(con, con->chosenpeer.connum));
else
Con_Printf(S_COLOR_GRAY"[%s]: New peer imposed %s.\n", con->friendlyname, NET_AdrToString(msg, sizeof(msg), &net_from));
}
con->chosenpeer = net_from;
if (con->state == ICE_CONNECTING)
ICE_Set(con, "state", STRINGIFY(ICE_CONNECTED));
}
}
else
{
con->controlled = (tiehigh > con->tiehigh) || (tiehigh == con->tiehigh && tielow > con->tielow);
if (net_ice_debug.ival >= 1)
Con_Printf(S_COLOR_GRAY"[%s]: role conflict detected. We should be %s\n", con->friendlyname, con->controlled?"controlled":"controlling");
error = 87;
}
}
else if (usecandidate && con->controlled)
{
//in the controlled role, we're connected once we're told the pair to use (by the usecandidate flag).
//note that this 'nominates' candidate pairs, from which the highest priority is chosen.
//so we just pick select the highest.
//this is problematic, however, as we don't actually know the real priority that the peer thinks we'll nominate it with.
if ((con->chosenpeer.connum != net_from.connum || !NET_CompareAdr(&con->chosenpeer, &net_from)) && net_ice_debug.ival >= 1)
{
char msg[64];
Con_Printf(S_COLOR_GRAY"[%s]: New peer imposed %s, via %s.\n", con->friendlyname, NET_AdrToString(msg, sizeof(msg), &net_from), ICE_NetworkToName(con, net_from.connum));
}
con->chosenpeer = net_from;
if (con->state == ICE_CONNECTING)
ICE_Set(con, "state", STRINGIFY(ICE_CONNECTED));
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
}//otherwise its just an ip check
else
con = NULL;
#else
(void)integritypos;
#endif
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
memset(&buf, 0, sizeof(buf));
buf.maxsize = sizeof(data);
buf.cursize = 0;
buf.data = data;
if (net_from.type == NA_IP)
{
alen = 4;
atype = 1;
aofs = 0;
}
else if (net_from.type == NA_IPV6 &&
!*(int*)&net_from.address.ip6[0] &&
!*(int*)&net_from.address.ip6[4] &&
!*(short*)&net_from.address.ip6[8] &&
*(short*)&net_from.address.ip6[10] == (short)0xffff)
{ //just because we use an ipv6 address for ipv4 internally doesn't mean we should tell the peer that they're on ipv6...
alen = 4;
atype = 1;
aofs = sizeof(net_from.address.ip6) - sizeof(net_from.address.ip);
}
else if (net_from.type == NA_IPV6)
{
alen = 16;
atype = 2;
aofs = 0;
}
else
{
alen = 0;
atype = 0;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
//Con_DPrintf("STUN from %s\n", NET_AdrToString(data, sizeof(data), &net_from));
MSG_WriteShort(&buf, BigShort(error?(STUN_BINDING|STUN_ERROR):(STUN_BINDING|STUN_REPLY)));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(0)); //fill in later
MSG_WriteLong(&buf, stun->magiccookie);
MSG_WriteLong(&buf, stun->transactid[0]);
MSG_WriteLong(&buf, stun->transactid[1]);
MSG_WriteLong(&buf, stun->transactid[2]);
if (error == 87)
{
char *txt = "Role Conflict";
MSG_WriteShort(&buf, BigShort(STUNATTR_ERROR_CODE));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(4 + strlen(txt)));
MSG_WriteShort(&buf, 0); //reserved
MSG_WriteByte(&buf, 0); //class
MSG_WriteByte(&buf, error); //code
SZ_Write(&buf, txt, strlen(txt)); //readable
while(buf.cursize&3) //padding
MSG_WriteChar(&buf, 0);
}
else if (1)
{ //xor mapped
netadr_t xored = net_from;
int i;
xored.port ^= *(short*)(data+4);
for (i = 0; i < alen; i++)
((qbyte*)&xored.address)[aofs+i] ^= ((qbyte*)data+4)[i];
MSG_WriteShort(&buf, BigShort(STUNATTR_XOR_MAPPED_ADDRESS));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(4+alen));
MSG_WriteShort(&buf, BigShort(atype));
MSG_WriteShort(&buf, xored.port);
SZ_Write(&buf, (char*)&xored.address + aofs, alen);
}
else
{ //non-xor mapped
MSG_WriteShort(&buf, BigShort(STUNATTR_MAPPED_ADDRESS));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(4+alen));
MSG_WriteShort(&buf, BigShort(atype));
MSG_WriteShort(&buf, net_from.port);
SZ_Write(&buf, (char*)&net_from.address + aofs, alen);
}
MSG_WriteShort(&buf, BigShort(STUNATTR_USERNAME)); //USERNAME
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(strlen(username)));
SZ_Write(&buf, username, strlen(username));
while(buf.cursize&3)
MSG_WriteChar(&buf, 0);
#ifdef SUPPORT_ICE
if (con)
{
//message integrity is a bit annoying
data[2] = ((buf.cursize+4+sizeof(integrity)-20)>>8)&0xff; //hashed header length is up to the end of the hmac attribute
data[3] = ((buf.cursize+4+sizeof(integrity)-20)>>0)&0xff;
//but the hash is to the start of the attribute's header
openxr plugin: tweaked - inputs should be working properly now, and are visible to csqc. subject to further breaking changes, however. _pext_vrinputs: added cvar to enable vr inputs protocol extension allowing vr inputs to be networked to ssqc too. defaults to 0 for now, will be renamed when deemed final. updates menu: the prompt to enable sources is now more explicit instead of expecting the user to have a clue. updates menu: added a v3 sources format, which should be more maintainable. not final. updates menu: try to give reasons why sources might be failing (to help blame ISPs if they try fucking over TTH dns again). presets menu: no longer closes the instant a preset is chosen. some presets have a couple of modifiers listed. force the demo loop in the background to serve as a preview. prompts menus: now does word wrapping. ftemaster: support importing server lists from other master servers (requested by Eukara). server: try to detect when non-reply inbound packets are blocked by firewalls/nats/etc (using ftemaster to do so). qcvm: added pointcontentsmask builtin, allowing it to probe more than just world, with fte's full contentbit range instead of just q1 legacy. qcvm: memfill8 builtin now works on createbuffer() pointers. qcvm: add missing unsigned ops. Fixed double comparison ops. fixed bug with op_store_i64. added missing OP_LOADP_I64 qcc: added '#pragma framerate RATE' for overriding implicit nextthink durations. qcc: fixed '#pragma DONT_COMPILE_THIS_FILE' to not screw up comments. qcc: added __GITURL__ __GITHASH__ __GITDATE__ __GITDATETIME__ __GITDESC__ for any mods that might want to make use of that. qcc: fix up -Fhashonly a little setrenderer: support for vulkan gpu enumeration. rulesets: reworked to support custom rulesets (using hashes to catch haxxors, though still nothing prevents just changing the client to ignore rulesets) bspx: use our BIH code for the bspx BRUSHLIST lump instead of the older less efficient code. (static)iqm+obj: these model formats can now be used for the worldmodel (with a suitable .ent file). Also using BIH for much better collision performance. pmove: tried to optimise PM_NudgePosition, should boost fps in stress tests. wayland: fix a crash on startup. mousegrabs now works better. imagetool: uses sdl for previews. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5813 fc73d0e0-1445-4013-8a0c-d673dee63da5
2021-04-14 05:21:04 +00:00
CalcHMAC(&hash_sha1, integrity, sizeof(integrity), data, buf.cursize, con->lpwd, strlen(con->lpwd));
MSG_WriteShort(&buf, BigShort(STUNATTR_MESSAGEINTEGRITIY));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(sizeof(integrity))); //sha1 key length
SZ_Write(&buf, integrity, sizeof(integrity)); //integrity data
}
#endif
data[2] = ((buf.cursize+8-20)>>8)&0xff; //dummy length
data[3] = ((buf.cursize+8-20)>>0)&0xff;
crc = crc32(0, data, buf.cursize)^0x5354554e;
MSG_WriteShort(&buf, BigShort(STUNATTR_FINGERPRINT)); //FINGERPRINT
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
MSG_WriteShort(&buf, BigShort(sizeof(crc)));
MSG_WriteLong(&buf, BigLong(crc));
data[2] = ((buf.cursize-20)>>8)&0xff;
data[3] = ((buf.cursize-20)>>0)&0xff;
#ifdef SUPPORT_ICE
TURN_Encapsulate(con, &net_from, data, buf.cursize);
#else
NET_SendPacket(col, buf.cursize, data, &net_from);
#endif
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return true;
}
}
#ifdef SUPPORT_ICE
{
struct icestate_s *con;
struct icecandidate_s *rc;
for (con = icelist; con; con = con->next)
{
for (rc = con->rc; rc; rc = rc->next)
{
if (NET_CompareAdr(&net_from, &rc->peer))
{
// if (rc->reachable)
{ //found it. fix up its source address to our ICE connection (so we don't have path-switching issues) and keep chugging along.
con->timeout = Sys_Milliseconds() + 32; //not dead yet...
#ifdef HAVE_DTLS
if (con->dtlsstate)
{
switch(con->dtlsfuncs->Received(con->dtlsstate, &net_message))
{
case NETERR_SENT:
break; //
case NETERR_NOROUTE:
return false; //not a dtls packet at all. don't de-ICE it when we're meant to be using ICE.
case NETERR_DISCONNECTED: //dtls failure. ICE failed.
iceapi.ICE_Set(con, "state", STRINGIFY(ICE_FAILED));
return true;
default: //some kind of failure decoding the dtls packet. drop it.
return true;
}
}
#endif
net_from = con->qadr;
#ifdef HAVE_DTLS
if (con->sctp)
SCTP_Decode(con->sctp, con, col);
#endif
if (net_message.cursize)
col->ReadGamePacket();
return true;
}
}
}
}
}
#endif
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return false;
}
#ifdef SUPPORT_ICE
qboolean ICE_IsEncrypted(netadr_t *to)
{
#ifdef HAVE_DTLS
struct icestate_s *con;
for (con = icelist; con; con = con->next)
{
if (NET_CompareAdr(to, &con->qadr))
{
if (con->dtlsstate)
return true;
}
}
#endif
return false;
}
neterr_t ICE_SendPacket(ftenet_connections_t *col, size_t length, const void *data, netadr_t *to)
{
struct icestate_s *con;
for (con = icelist; con; con = con->next)
{
if (NET_CompareAdr(to, &con->qadr))
{
#ifdef HAVE_DTLS
if (con->sctp)
return SCTP_Transmit(con->sctp, con, data, length);
if (con->dtlsstate)
return SCTP_PeerSendPacket(con, length, data);
#endif
if (con->chosenpeer.type != NA_INVALID)
return ICE_Transmit(con, data, length);
if (con->state < ICE_CONNECTING)
return NETERR_DISCONNECTED;
return NETERR_CLOGGED; //still pending
}
}
return NETERR_DISCONNECTED;
}
#endif
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#endif
#ifdef SUPPORT_ICE
//this is the clientside part of our custom accountless broker protocol
//basically just keeps the broker processing, but doesn't send/receive actual game packets.
//inbound messages can change ice connection states.
//clients only handle one connection. servers need to handle multiple
typedef struct {
ftenet_generic_connection_t generic;
//config state
char brokername[64]; //dns name
netadr_t brokeradr; //actual ip
char gamename[64]; //what we're trying to register as/for with the broker
//broker connection state
vfsfile_t *broker;
qboolean handshaking;
double nextping; //send heartbeats every now and then
double heartbeat;
double timeout; //detect if the broker goes dead, so we can reconnect reliably (instead of living for two hours without anyone able to connect).
qbyte in[8192];
size_t insize;
qbyte out[8192];
size_t outsize;
int error; //outgoing data is corrupt. kill it.
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
//client state...
struct icestate_s *ice;
int serverid;
//server state...
struct
{
struct icestate_s *ice;
} *clients;
size_t numclients;
} ftenet_ice_connection_t;
static void FTENET_ICE_Close(ftenet_generic_connection_t *gcon)
{
ftenet_ice_connection_t *b = (void*)gcon;
int cl;
if (b->broker)
VFS_CLOSE(b->broker);
for (cl = 0; cl < b->numclients; cl++)
if (b->clients[cl].ice)
iceapi.ICE_Close(b->clients[cl].ice);
Z_Free(b->clients);
if (b->ice)
iceapi.ICE_Close(b->ice);
Z_Free(b);
}
static void FTENET_ICE_Flush(ftenet_ice_connection_t *b)
{
int r;
if (!b->outsize || b->error || !b->broker)
return;
r = VFS_WRITE(b->broker, b->out, b->outsize);
if (r > 0)
{
b->outsize -= r;
memmove(b->out, b->out+r, b->outsize);
}
if (r < 0)
b->error = true;
}
static neterr_t FTENET_ICE_SendPacket(ftenet_generic_connection_t *gcon, int length, const void *data, netadr_t *to)
{
ftenet_ice_connection_t *b = (void*)gcon;
if (to->prot != NP_RTC_TCP && to->prot != NP_RTC_TLS)
return NETERR_NOROUTE;
if (!NET_CompareAdr(to, &b->brokeradr))
return NETERR_NOROUTE; //its using some other broker, don't bother trying to handle it here.
if (b->error)
return NETERR_DISCONNECTED;
return NETERR_CLOGGED; //we'll switch to a connect localcmd when the connection completes, so we don't really need to send any packets when they're using ICE. Just make sure the client doesn't give up.
}
static void FTENET_ICE_SplurgeRaw(ftenet_ice_connection_t *b, const qbyte *data, size_t len)
{
//0: dropclient (cl=-1 drops entire connection)
if (b->outsize+len > sizeof(b->out))
b->error = true;
else
{
memcpy(b->out+b->outsize, data, len);
b->outsize += len;
}
}
static void FTENET_ICE_SplurgeWS(ftenet_ice_connection_t *b, enum websocketpackettype_e pkttype, const qbyte *data1, size_t len1, const qbyte *data2, size_t len2)
{
size_t tlen = len1+len2;
qbyte header[8];
header[0] = 0x80|pkttype;
if (tlen >= 126)
{
header[1] = 126;
header[2] = tlen>>8; //bigendian
header[3] = tlen&0xff;
FTENET_ICE_SplurgeRaw(b, header, 4);
}
else
{ //small data
header[1] = tlen;
FTENET_ICE_SplurgeRaw(b, header, 2);
}
FTENET_ICE_SplurgeRaw(b, data1, len1);
FTENET_ICE_SplurgeRaw(b, data2, len2);
}
static void FTENET_ICE_SplurgeCmd(ftenet_ice_connection_t *b, int icemsg, int cl, const char *data)
{
qbyte msg[3] = {icemsg, cl&0xff, (cl>>8)&0xff}; //little endian...
FTENET_ICE_SplurgeWS(b, WS_PACKETTYPE_BINARYFRAME, msg, sizeof(msg), data, strlen(data));
}
static void FTENET_ICE_Heartbeat(ftenet_ice_connection_t *b)
{
b->heartbeat = realtime+30;
#ifdef HAVE_SERVER
if (b->generic.islisten)
{
extern cvar_t maxclients;
char info[2048];
int i;
client_t *cl;
int numclients = 0;
for (i=0 ; i<svs.allocated_client_slots ; i++)
{
cl = &svs.clients[i];
if ((cl->state == cs_connected || cl->state == cs_spawned || cl->name[0]) && !cl->spectator)
numclients++;
}
*info = 0;
Info_SetValueForKey(info, "protocol", SV_GetProtocolVersionString(), sizeof(info));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
Info_SetValueForKey(info, "maxclients", maxclients.string, sizeof(info));
Info_SetValueForKey(info, "clients", va("%i", numclients), sizeof(info));
Info_SetValueForKey(info, "hostname", hostname.string, sizeof(info));
Info_SetValueForKey(info, "modname", FS_GetGamedir(true), sizeof(info));
Info_SetValueForKey(info, "mapname", InfoBuf_ValueForKey(&svs.info, "map"), sizeof(info));
Info_SetValueForKey(info, "needpass", InfoBuf_ValueForKey(&svs.info, "needpass"), sizeof(info));
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
FTENET_ICE_SplurgeCmd(b, ICEMSG_SERVERINFO, -1, info);
}
#endif
}
static void FTENET_ICE_SendOffer(ftenet_ice_connection_t *b, int cl, struct icestate_s *ice, const char *type)
{
char buf[8192];
//okay, now send the sdp to our peer.
if (iceapi.ICE_Get(ice, type, buf, sizeof(buf)))
{
char json[8192+256];
if (ice->mode == ICEM_WEBRTC)
{
Q_strncpyz(json, va("{\"type\":\"%s\",\"sdp\":\"", type+3), sizeof(json));
COM_QuotedString(buf, json+strlen(json), sizeof(json)-strlen(json)-2, true);
Q_strncatz(json, "\"}", sizeof(json));
FTENET_ICE_SplurgeCmd(b, ICEMSG_OFFER, cl, json);
}
else
FTENET_ICE_SplurgeCmd(b, ICEMSG_OFFER, cl, buf);
ice->blockcandidates = false;
}
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static void FTENET_ICE_Establish(ftenet_ice_connection_t *b, int cl, struct icestate_s **ret)
{ //sends offer
struct icestate_s *ice;
qboolean usewebrtc;
char *s;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (*ret)
iceapi.ICE_Close(*ret);
#ifndef HAVE_DTLS
usewebrtc = false;
#else
if (!*net_ice_usewebrtc.string)
usewebrtc = true; //let the peer decide. this means we use dtls, but not sctp.
else
usewebrtc = net_ice_usewebrtc.ival;
#endif
ice = *ret = iceapi.ICE_Create(b, NULL, b->generic.islisten?NULL:va("/%s", b->gamename), usewebrtc?ICEM_WEBRTC:ICEM_ICE, b->generic.islisten?ICEP_QWSERVER:ICEP_QWCLIENT, !b->generic.islisten);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (!*ret)
return; //some kind of error?!?
iceapi.ICE_Set(ice, "server", va("stun:%s:%i", b->brokername, BigShort(b->brokeradr.port)));
s = net_ice_servers.string;
while((s=COM_Parse(s)))
iceapi.ICE_Set(ice, "server", com_token);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (!b->generic.islisten)
FTENET_ICE_SendOffer(b, cl, ice, "sdpoffer");
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
static void FTENET_ICE_Refresh(ftenet_ice_connection_t *b, int cl, struct icestate_s *ice)
{ //sends offer
char buf[8192];
if (ice->blockcandidates)
return; //don't send candidates before the offers...
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
while (ice && iceapi.ICE_GetLCandidateSDP(ice, buf, sizeof(buf)))
{
char json[8192+256];
if (ice->mode == ICEM_WEBRTC)
{
Q_strncpyz(json, "{\"candidate\":\"", sizeof(json));
COM_QuotedString(buf+2, json+strlen(json), sizeof(json)-strlen(json)-2, true);
Q_strncatz(json, "\",\"sdpMid\":\"0\",\"sdpMLineIndex\":0}", sizeof(json));
FTENET_ICE_SplurgeCmd(b, ICEMSG_CANDIDATE, cl, json);
}
else
FTENET_ICE_SplurgeCmd(b, ICEMSG_CANDIDATE, cl, buf);
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
static qboolean FTENET_ICE_GetPacket(ftenet_generic_connection_t *gcon)
{
ftenet_ice_connection_t *b = (void*)gcon;
int ctrl, len, cmd, cl, ofs;
char *data, n;
if (!b->broker)
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
{
const char *s;
if (b->timeout > realtime)
return false;
b->generic.thesocket = TCP_OpenStream(&b->brokeradr, b->brokername); //save this for select.
b->broker = FS_WrapTCPSocket(b->generic.thesocket, true, b->brokername);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
#ifdef HAVE_SSL
//convert to tls...
if (b->brokeradr.prot == NP_TLS || b->brokeradr.prot == NP_RTC_TLS)
b->broker = FS_OpenSSL(b->brokername, b->broker, false);
#endif
if (!b->broker)
{
b->timeout = realtime + 30;
Con_Printf("rtc broker connection to %s failed (retry: 30 secs)\n", b->brokername);
return false;
}
b->insize = b->outsize = 0;
COM_Parse(com_protocolname.string);
b->handshaking = true;
s = va("GET /%s/%s HTTP/1.1\r\n"
"Host: %s\r\n"
"Connection: Upgrade\r\n"
"Upgrade: websocket\r\n"
"Sec-WebSocket-Version: 13\r\n"
"Sec-WebSocket-Protocol: %s\r\n"
"\r\n", com_token, b->gamename, b->brokername, b->generic.islisten?"rtc_host":"rtc_client");
FTENET_ICE_SplurgeRaw(b, s, strlen(s));
b->heartbeat = realtime;
b->nextping = realtime + 100;
b->timeout = realtime + 270;
}
if (b->error)
{
handleerror:
b->generic.thesocket = INVALID_SOCKET;
if (b->broker)
VFS_CLOSE(b->broker);
b->broker = NULL;
for (cl = 0; cl < b->numclients; cl++)
{
if (b->clients[cl].ice)
iceapi.ICE_Close(b->clients[cl].ice);
b->clients[cl].ice = NULL;
}
if (b->ice)
iceapi.ICE_Close(b->ice);
b->ice = NULL;
if (b->error != 1 || !b->generic.islisten)
return false; //permanant error...
b->error = false;
b->insize = b->outsize = 0;
b->timeout = realtime + 30;
return false;
}
//keep checking for new candidate info.
if (b->ice)
FTENET_ICE_Refresh(b, b->serverid, b->ice);
for (cl = 0; cl < b->numclients; cl++)
if (b->clients[cl].ice)
FTENET_ICE_Refresh(b, cl, b->clients[cl].ice);
if (realtime >= b->heartbeat)
FTENET_ICE_Heartbeat(b);
len = VFS_READ(b->broker, b->in+b->insize, sizeof(b->in)-1-b->insize);
if (!len)
{
FTENET_ICE_Flush(b);
if (realtime > b->nextping)
{ //nothing happening... make sure the connection isn't dead...
FTENET_ICE_SplurgeWS(b, WS_PACKETTYPE_PING, NULL, 0, NULL, 0);
b->nextping = realtime + 100;
}
return false; //nothing new
}
if (len < 0)
{
if (!b->error)
Con_Printf("rtc broker connection to %s failed (retry: 30 secs)\n", b->brokername);
b->error = true;
goto handleerror;
}
b->insize += len;
b->in[b->insize] = 0;
ofs = 0;
b->nextping = realtime + 100;
b->timeout = max(b->timeout, realtime + 270);
if (b->handshaking)
{ //we're still waiting for an http 101 code. websocket data starts straight after.
char *end = strstr(b->in, "\r\n\r\n");
if (!end)
return false; //not available yet...
if (strncmp(b->in, "HTTP/1.1 101 ", 13))
{
b->error = ~0;
return false;
}
end+=4;
b->handshaking = false; //done...
ofs = (qbyte*)end-b->in;
}
while (b->insize >= ofs+2)
{
ctrl = b->in[ofs+0];
len = b->in[ofs+1];
ofs+=2;
if (len > 126)
{//unsupported
b->error = 1;
break;
}
else if (len == 126)
{
if (b->insize <= 4)
break;
len = (b->in[ofs+0]<<8)|(b->in[ofs+1]);
ofs+=2;
}
if (b->insize < ofs+len)
break;
n = b->in[ofs+len];
b->in[ofs+len] = 0;
switch(ctrl & 0xf)
{
case WS_PACKETTYPE_PING:
FTENET_ICE_SplurgeWS(b, WS_PACKETTYPE_PONG, NULL, 0, b->in+ofs, len);
FTENET_ICE_Flush(b);
break;
case WS_PACKETTYPE_CLOSE:
b->error = true;
break;
default:
break;
case WS_PACKETTYPE_BINARYFRAME:
cmd = b->in[ofs];
cl = (short)(b->in[ofs+1] | (b->in[ofs+2]<<8));
data = b->in+ofs+3;
switch(cmd)
{
case ICEMSG_PEERDROP: //connection closing...
if (cl == -1)
{
b->error = true;
// Con_Printf("Broker closed connection: %s\n", data);
}
else if (cl >= 0 && cl < b->numclients)
{
if (b->clients[cl].ice)
iceapi.ICE_Close(b->clients[cl].ice);
b->clients[cl].ice = NULL;
// Con_Printf("Broker closing connection: %s\n", data);
}
break;
case ICEMSG_NAMEINUSE:
Con_Printf("Unable to listen on /%s - name already taken\n", b->gamename);
b->error = true; //try again later.
break;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
case ICEMSG_GREETING: //reports the trailing url we're 'listening' on. anyone else using that url will connect to us.
data = strchr(data, '/');
if (data++)
Q_strncpyz(b->gamename, data, sizeof(b->gamename));
Con_Printf("Publicly listening on /%s\n", b->gamename);
break;
case ICEMSG_NEWPEER: //relay connection established with a new peer
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
//note that the server ought to wait for an offer from the client before replying with any ice state, but it doesn't really matter for our use-case.
if (b->generic.islisten)
{
// Con_DPrintf("Client connecting: %s\n", data);
if (cl < 1024 && cl >= b->numclients)
{ //looks like a new one... but don't waste memory
Z_ReallocElements((void**)&b->clients, &b->numclients, cl+1, sizeof(b->clients[0]));
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (cl >= 0 && cl < b->numclients)
FTENET_ICE_Establish(b, cl, &b->clients[cl].ice);
}
else
{
// Con_DPrintf("Server found: %s\n", data);
FTENET_ICE_Establish(b, cl, &b->ice);
b->serverid = cl;
}
break;
case ICEMSG_OFFER: //we received an offer from a client
if (!strncmp(data, "{\"type\":\"offer\",\"sdp\":\"", 23))
{
data += 22;
COM_ParseCString(data, com_token, sizeof(com_token), NULL);
data = com_token;
}
else if (!strncmp(data, "{\"type\":\"answer\",\"sdp\":\"", 24))
{
data += 23;
COM_ParseCString(data, com_token, sizeof(com_token), NULL);
data = com_token;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (b->generic.islisten)
{
if (cl >= 0 && cl < b->numclients && b->clients[cl].ice)
{
iceapi.ICE_Set(b->clients[cl].ice, "sdpoffer", data);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
iceapi.ICE_Set(b->clients[cl].ice, "state", STRINGIFY(ICE_CONNECTING));
FTENET_ICE_SendOffer(b, cl, b->clients[cl].ice, "sdpanswer");
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
else
{
if (b->ice)
{
iceapi.ICE_Set(b->ice, "sdpanswer", data);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
iceapi.ICE_Set(b->ice, "state", STRINGIFY(ICE_CONNECTING));
}
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
break;
case ICEMSG_CANDIDATE:
if (!strncmp(data, "{\"candidate\":\"", 14))
{
data += 13;
com_token[0]='a';
com_token[1]='=';
COM_ParseCString(data, com_token+2, sizeof(com_token)-2, NULL);
data = com_token;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
// Con_Printf("Candidate update: %s\n", data);
if (b->generic.islisten)
{
if (cl >= 0 && cl < b->numclients && b->clients[cl].ice)
iceapi.ICE_Set(b->clients[cl].ice, "sdp", data);
}
else
{
if (b->ice)
iceapi.ICE_Set(b->ice, "sdp", data);
}
break;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
break;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
ofs+=len;
b->in[ofs] = n;
}
if (ofs)
{ //and eat the newly parsed data...
b->insize -= ofs;
memmove(b->in, b->in+ofs, b->insize);
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
FTENET_ICE_Flush(b);
return false;
}
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
static void FTENET_ICE_PrintStatus(ftenet_generic_connection_t *gcon)
{
ftenet_ice_connection_t *b = (void*)gcon;
size_t c;
if (b->ice)
ICE_Debug(b->ice);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (b->numclients)
{
size_t activeice = 0;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
for (c = 0; c < b->numclients; c++)
if (b->clients[c].ice)
{
activeice++;
ICE_PrintSummary(b->clients[c].ice, b->generic.islisten);
}
Con_Printf("%u ICE connections\n", (unsigned)activeice);
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
}
}
static int FTENET_ICE_GetLocalAddresses(struct ftenet_generic_connection_s *gcon, unsigned int *adrflags, netadr_t *addresses, const char **adrparms, int maxaddresses)
{
ftenet_ice_connection_t *b = (void*)gcon;
if (maxaddresses < 1)
return 0;
*addresses = b->brokeradr;
*adrflags = 0;
*adrparms = b->gamename;
return 1;
}
static qboolean FTENET_ICE_ChangeLocalAddress(struct ftenet_generic_connection_s *gcon, const char *address, netadr_t *newadr)
{
ftenet_ice_connection_t *b = (void*)gcon;
netadr_t adr;
const char *path;
if (!NET_StringToAdr2(address, PORT_ICEBROKER, &adr, 1, &path))
return true; //err... something failed? don't break what works!
if (!NET_CompareAdr(&adr, &b->brokeradr))
return false; //the broker changed! zomg! just kill it all!
if (path && *path++=='/')
{
if (*path && strcmp(path, b->gamename))
return false; //it changed! and we care! break everything!
}
return true;
}
ftenet_generic_connection_t *FTENET_ICE_EstablishConnection(ftenet_connections_t *col, const char *address, netadr_t adr)
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
{
ftenet_ice_connection_t *newcon;
const char *path;
char *c;
if (!NET_StringToAdr2(address, PORT_ICEBROKER, &adr, 1, &path))
return NULL;
/* if (adr.prot == NP_ICES)
adr.prot = NP_TLS;
else if (adr.prot == NP_ICE)
adr.prot = NP_STREAM;
*/
newcon = Z_Malloc(sizeof(*newcon));
if (!strncmp(address, "ice://", 6)||!strncmp(address, "rtc://", 6))
address+=6;
else if (!strncmp(address, "ices://", 7)||!strncmp(address, "rtcs://", 7))
address+=7;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
if (address == path && *path=='/' && fs_manifest->rtcbroker)
{
if (!strncmp(fs_manifest->rtcbroker, "tls://", 6) || !strncmp(fs_manifest->rtcbroker, "tcp://", 6))
Q_strncpyz(newcon->brokername, fs_manifest->rtcbroker+6, sizeof(newcon->brokername)); //name is for prints only.
else
Q_strncpyz(newcon->brokername, fs_manifest->rtcbroker, sizeof(newcon->brokername)); //name is for prints only.
Q_strncpyz(newcon->gamename, path+1, sizeof(newcon->gamename)); //so we know what to tell the broker.
}
else
{
Q_strncpyz(newcon->brokername, address, sizeof(newcon->brokername)); //name is for prints only.
if (path && *path == '/' && path-address < sizeof(newcon->brokername))
{
newcon->brokername[path-address] = 0;
Q_strncpyz(newcon->gamename, path+1, sizeof(newcon->gamename)); //so we know what to tell the broker.
}
else
*newcon->gamename = 0;
}
c = strchr(newcon->brokername, ':');
if (c) *c = 0;
newcon->brokeradr = adr;
newcon->broker = NULL;
newcon->timeout = realtime;
newcon->heartbeat = realtime;
newcon->nextping = realtime;
newcon->generic.owner = col;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
newcon->generic.thesocket = INVALID_SOCKET;
newcon->generic.addrtype[0] = NA_INVALID;
newcon->generic.addrtype[1] = NA_INVALID;
newcon->generic.GetPacket = FTENET_ICE_GetPacket;
newcon->generic.SendPacket = FTENET_ICE_SendPacket;
newcon->generic.Close = FTENET_ICE_Close;
newcon->generic.PrintStatus = FTENET_ICE_PrintStatus;
newcon->generic.GetLocalAddresses = FTENET_ICE_GetLocalAddresses;
newcon->generic.ChangeLocalAddress = FTENET_ICE_ChangeLocalAddress;
newcon->generic.islisten = col->islisten;
fixed eztv md4 incompatibility. reimplemented qtvreverse command. fixed some stuffcmds being handled by the wrong splitscreen seats (was noticable in TF). rework smartjump to try to be more predictable... rework relighting to try to be more robust (and more self-contained). allow the csqc to actually use VF_PROJECTIONOFFSET. jump now moves upwards instead of trying to lock on to a nearby player when spectating. assume 32 fullbright pixels when running with a palette.lmp yet no colormap.lmp (happens with some total conversions). tweaked scoreboard for fainter backgrounds. rearranged autoid, to be smaller etc. hacked around dodgy conchars.lmp - don't treat 128*128 qpics as qpics to work around workarounds for buggy wad tools (with a warning). fixed missing fullbrights on h2holey models. avoided warning about mod_h2holey_bugged on dedicated servers. added net_ice_exchangeprivateips, for people worried about exposing lan IPs when using ICE. sv_public 2: implemented client support for our webrtc broker in order to use our own ICE implementation without needing to faff around with irc accounts or plugins etc. TODO: ensure at least one ephemerial udp port when using ice or come up with some better sv_port handling fixed multiple tls bugs (one could cause server problems). change net_enable_tls to disabled by default anyway (reenable for the server to be able to respond to https/wss/tls schemes again). don't colourmap when there appears to be a highres diffusemap on q1 models. imgtool now understands exporting from qpics in wads, as well as just mips. implemented speed-o-meter in ezhud. added removeinstant builtin to avoid the half-second rule. git-svn-id: https://svn.code.sf.net/p/fteqw/code/trunk@5614 fc73d0e0-1445-4013-8a0c-d673dee63da5
2020-02-11 18:06:10 +00:00
return &newcon->generic;
}
void ICE_Init(void)
{
Cvar_Register(&net_ice_exchangeprivateips, "networking");
Cvar_Register(&net_ice_allowstun, "networking");
Cvar_Register(&net_ice_allowturn, "networking");
Cvar_Register(&net_ice_relayonly, "networking");
Cvar_Register(&net_ice_usewebrtc, "networking");
Cvar_Register(&net_ice_servers, "networking");
Cvar_Register(&net_ice_debug, "networking");
Cmd_AddCommand("net_ice_show", ICE_Show_f);
}
#endif