update docs

docs/changes.txt

@@ -253,3 +253,9 @@ v0.9.4.2
 
 * builder_linux: simple scripts to build static linux bins for any supported architecture
 * zapret-auto:  incompatible change.  cond_lua "code" parameter => "cond_code". to avoid collision with luaexec
+
+v0.9.4.3
+
+* nfqws2: fix broken wifi ssid update
+* github: revert to upx 4.2.4 for all archs except riscv64
+* zapret-lib: apply_fooling throws error if tcp_ts,tcp_seq,tcp_ack,ip_ttl,ip6_ttl,ip_autottl,ip6_autottl are empty or invalid

lua/zapret-lib.lua

@@ -875,7 +875,11 @@ function apply_fooling(desync, dis, fooling_options)
 				if type(desync.track.lua_state.autottl_cache)~="table" then desync.track.lua_state.autottl_cache={} end
 				if type(desync.track.lua_state.autottl_cache[desync.func_instance])~="table" then desync.track.lua_state.autottl_cache[desync.func_instance]={} end
 				if not desync.track.lua_state.autottl_cache[desync.func_instance].autottl_found then
-					desync.track.lua_state.autottl_cache[desync.func_instance].autottl = autottl(desync.track.incoming_ttl,parse_autottl(arg_autottl))
+					attl = parse_autottl(arg_autottl)
+					if not attl then
+						error("apply_fooling: invalid autottl value '"..arg_autottl.."'")
+					end
+					desync.track.lua_state.autottl_cache[desync.func_instance].autottl = autottl(desync.track.incoming_ttl,attl)
 					if desync.track.lua_state.autottl_cache[desync.func_instance].autottl then
 						desync.track.lua_state.autottl_cache[desync.func_instance].autottl_found = true
 							DLOG("apply_fooling: discovered autottl "..desync.track.lua_state.autottl_cache[desync.func_instance].autottl)
@@ -890,8 +894,11 @@ function apply_fooling(desync, dis, fooling_options)
 				DLOG("apply_fooling: cannot apply autottl because incoming ttl unknown")
 			end
 		end
-		if not ttl and tonumber(arg_ttl) then
+		if not ttl and arg_ttl then
 			ttl = tonumber(arg_ttl)
+			if not ttl or ttl<0 or ttl>255 then
+				error("apply_fooling: ip_ttl and ip6_ttl require valid value")
+			end
 		end
 		--io.stderr:write("TTL "..tostring(ttl).."\n")
 		return ttl
@@ -908,11 +915,19 @@ function apply_fooling(desync, dis, fooling_options)
 	-- use current packet if dissect not given
 	if not dis then dis = desync.dis end
 	if dis.tcp then
-		if tonumber(fooling_options.tcp_seq) then
-			dis.tcp.th_seq = u32add(dis.tcp.th_seq, fooling_options.tcp_seq)
+		if fooling_options.tcp_seq then
+			if tonumber(fooling_options.tcp_seq) then
+				dis.tcp.th_seq = u32add(dis.tcp.th_seq, fooling_options.tcp_seq)
+			else
+				error("apply_fooling: tcp_seq requires increment parameter. there's no default value.")
+			end
 		end
-		if tonumber(fooling_options.tcp_ack) then
-			dis.tcp.th_ack = u32add(dis.tcp.th_ack, fooling_options.tcp_ack)
+		if fooling_options.tcp_ack then
+			if tonumber(fooling_options.tcp_ack) then
+				dis.tcp.th_ack = u32add(dis.tcp.th_ack, fooling_options.tcp_ack)
+			else
+				error("apply_fooling: tcp_ack requires increment parameter. there's no default value.")
+			end
 		end
 		if fooling_options.tcp_flags_unset then
 			dis.tcp.th_flags = bitand(dis.tcp.th_flags, bitnot(parse_tcp_flags(fooling_options.tcp_flags_unset)))
@@ -927,12 +942,16 @@ function apply_fooling(desync, dis, fooling_options)
 				end
 			end
 		end
-		if tonumber(fooling_options.tcp_ts) then
-			local idx = find_tcp_option(dis.tcp.options,TCP_KIND_TS)
-			if idx and (dis.tcp.options[idx].data and #dis.tcp.options[idx].data or 0)==8 then
-				dis.tcp.options[idx].data = bu32(u32add(u32(dis.tcp.options[idx].data),fooling_options.tcp_ts))..string.sub(dis.tcp.options[idx].data,5)
+		if fooling_options.tcp_ts then
+			if tonumber(fooling_options.tcp_ts) then
+				local idx = find_tcp_option(dis.tcp.options,TCP_KIND_TS)
+				if idx and (dis.tcp.options[idx].data and #dis.tcp.options[idx].data or 0)==8 then
+					dis.tcp.options[idx].data = bu32(u32add(u32(dis.tcp.options[idx].data),fooling_options.tcp_ts))..string.sub(dis.tcp.options[idx].data,5)
+				else
+					DLOG("apply_fooling: timestamp tcp option not present or invalid")
+				end
 			else
-				DLOG("apply_fooling: timestamp tcp option not present or invalid")
+				error("apply_fooling: tcp_ts requires increment parameter. there's no default value.")
 			end
 		end
 		if fooling_options.tcp_md5 then

nfq2/conntrack.c

@@ -148,7 +148,7 @@ static void ConntrackApplyPos(t_ctrack *t, bool bReverse, const struct dissect *
 		if (scale != SCALE_NONE) direct->scale = scale;
 		direct->mss = tcp_find_mss(dis->tcp);
 	}
-	else if (direct->scale != SCALE_NONE)
+	else
 		// apply scale only outside of the SYN stage
 		direct->winsize_calc <<= direct->scale;
 
@@ -367,8 +367,8 @@ void ConntrackPoolDump(const t_conntrack *p)
 				t->track.pos.client.seq0, t->track.pos.client.seq_last - t->track.pos.client.seq0, t->track.pos.client.pos - t->track.pos.client.seq0,
 				t->track.pos.server.seq0, t->track.pos.server.seq_last - t->track.pos.server.seq0, t->track.pos.server.pos - t->track.pos.server.seq0,
 				t->track.pos.client.mss, t->track.pos.server.mss,
-				t->track.pos.client.winsize, t->track.pos.client.scale == SCALE_NONE ? -1 : t->track.pos.client.scale,
-				t->track.pos.server.winsize, t->track.pos.server.scale == SCALE_NONE ? -1 : t->track.pos.server.scale);
+				t->track.pos.client.winsize, t->track.pos.client.scale,
+				t->track.pos.server.winsize, t->track.pos.server.scale);
 		else
 			printf("rseq=%u client.pos=%u rack=%u server.pos=%u",
 				t->track.pos.client.seq_last, t->track.pos.client.pos,

nfq2/conntrack_base.h

@@ -29,7 +29,7 @@ typedef struct
 	uint16_t winsize;	// last seen window size
 	uint16_t mss;
 	uint32_t winsize_calc;	// calculated window size
-	uint8_t scale;		// last seen window scale factor. SCALE_NONE if none
+	uint8_t scale;		// last seen window scale factor
 	bool rseq_over_2G;
 } t_ctrack_position;
 

nfq2/darkmagic.c

@@ -2244,6 +2244,7 @@ static bool scan_info(struct mnl_socket* nl, uint16_t wlan_family_id, struct wla
 	// wlan_info does not return ssid since kernel 5.19
 	// it's used to enumerate all wifi interfaces then call scan_info on each
 	if (!wlan_info(nl, wlan_family_id, &wc_all, false)) return false;
+	w->count=0;
 	for(int i=0;i<wc_all.count;i++)
 		if (!netlink_genl_simple_transact(nl, wlan_family_id, NLM_F_REQUEST | NLM_F_ACK | NLM_F_DUMP, NL80211_CMD_GET_SCAN, 0, scan_prepare, (void*)&wc_all.wlan[i].ifindex, scan_info_cb, w))
 			return false;

nfq2/desync.c

@@ -1648,9 +1648,12 @@ static const uint8_t *dns_extract_name(const uint8_t *a, const uint8_t *b, const
 {
 	size_t nl, off;
 	const uint8_t *p;
-	bool bptr = (*a & 0xC0)==0xC0;
+	bool bptr;
 	uint8_t x,y;
 
+	if (!name_size) return NULL;
+
+	bptr = (*a & 0xC0)==0xC0;
 	if (bptr)
 	{
 		if (a+1>=e) return NULL;
@@ -1665,66 +1668,115 @@ static const uint8_t *dns_extract_name(const uint8_t *a, const uint8_t *b, const
 	if (p>=e) return NULL;
 	for (nl=0; *p ;)
 	{
+		if (nl)
+		{
+			if (nl>=name_size) return NULL;
+			name[nl++] = '.';
+		}
 		// do not support mixed ptr+real
 		if ((*p & 0xC0) || (p+*p+1)>=e || (*p+1)>=(name_size-nl)) return NULL;
-		if (nl)	name[nl++] = '.';
 		for(y=*p++,x=0 ; x<y ; x++,p++) name[nl+x] = tolower(*p);
 		nl += y;
 	}
+	if (nl>=name_size) return NULL;
 	name[nl] = 0;
 	return bptr ? a+2 : p+1;
 }
+static bool dns_skip_name(const uint8_t **a, size_t *len)
+{
+	// 11 higher bits indicate pointer
+	// lazy skip name. mixed compressed/uncompressed names are supported
+	for(;;)
+	{
+		if (*len<2) return false;
+		if ((**a & 0xC0)==0xC0)
+		{
+			// pointer is the end
+			(*a)+=2; (*len)-=2;
+			break;
+		}
+		if (!**a)
+		{
+			// zero length is the end
+			(*a)++; (*len)--;
+			break;
+		}
+		if (*len<(**a+1)) return false;
+		*len-=**a+1;
+		*a+=**a+1;
+	}
+	return true;
+}
+
 static bool feed_dns_response(const uint8_t *a, size_t len)
 {
 	if (!params.cache_hostname) return true;
 
 	// check of minimum header length and response flag
-	uint16_t k, off, dlen, qcount = a[4]<<8 | a[5], acount = a[6]<<8 | a[7];
+	uint16_t k, typ, off, dlen, qcount = a[4]<<8 | a[5], acount = a[6]<<8 | a[7];
 	char s_ip[INET6_ADDRSTRLEN];
 	const uint8_t *b = a, *p;
 	const uint8_t *e = b + len;
 	size_t nl;
 	char name[256] = "";
 
-	if (len<12 || !(a[2]&0x80)) return false;
-	a+=12; len-=12;
-	for(k=0;k<qcount;k++)
+	if (!qcount || len<12 || !(a[2]&0x80)) return false;
+	if (!acount)
 	{
+		DLOG("skipping DNS response without answer\n");
+		return false;
+	}
+	a+=12; len-=12;
+	for(k=0,*name = 0 ; k<qcount ; k++)
+	{
+		if (*name) return false; // we do not support multiple queries with names
 		// remember original query name
 		if (!(p = dns_extract_name(a, b, e, name, sizeof(name)))) return false;
 		len -= p-a;
+		if ((len<4) || p[2] || p[3]!=1)	return false;
+		typ = pntoh16(p);
 		// must be A or AAAA query. others are not interesting
-		if ((len<4) || p[0] || p[1]!=1 && p[1]!=28 || p[2] || p[3]!=1) return false;
+		if (typ!=1 && typ!=28)
+		{
+			DLOG("skipping DNS query type %u for '%s'\n", typ, name);
+			return false;
+		}
+		else
+		{
+			DLOG("DNS query type %u for '%s'\n", typ, name);
+		}
 		// skip type, class
 		a=p+4; len-=4;
 	}
 	if (!*name) return false;
 	for(k=0;k<acount;k++)
 	{
-		// 11 higher bits indicate pointer
-		if (len<12 || (*a & 0xC0)!=0xC0) return false;
-
-		dlen = a[10]<<8 | a[11];
-		if (len<(dlen+12)) return false;
-		if (a[4]==0 && a[5]==1 && a[2]==0) // IN class and higher byte of type = 0
+		if (!dns_skip_name(&a,&len)) return false;
+		if (len<10) return false;
+		dlen = a[8]<<8 | a[9];
+		if (len<(dlen+10)) return false;
+		if (a[2]==0 && a[3]==1) // IN class
 		{
-			switch(a[3])
+			typ = pntoh16(a);
+			switch(typ)
 			{
 				case 1: // A
 					if (dlen!=4) break;
-					if (params.debug && inet_ntop(AF_INET, a+12, s_ip, sizeof(s_ip)))
-						DLOG("DNS response : %s\n", s_ip);
-					ipcache_put_hostname((struct in_addr *)(a+12), NULL, name, false);
+					if (params.debug && inet_ntop(AF_INET, a+10, s_ip, sizeof(s_ip)))
+						DLOG("DNS response type %u : %s\n", typ, s_ip);
+					ipcache_put_hostname((struct in_addr *)(a+10), NULL, name, false);
 					break;
 				case 28: // AAAA
 					if (dlen!=16) break;
-					if (params.debug && inet_ntop(AF_INET6, a+12, s_ip, sizeof(s_ip)))
-						DLOG("DNS response : %s\n", s_ip);
-					ipcache_put_hostname(NULL, (struct in6_addr *)(a+12), name, false);
+					if (params.debug && inet_ntop(AF_INET6, a+10, s_ip, sizeof(s_ip)))
+						DLOG("DNS response type %u : %s\n", typ, s_ip);
+					ipcache_put_hostname(NULL, (struct in6_addr *)(a+10), name, false);
 					break;
+				default:
+					DLOG("skipping DNS response type %u\n", typ);
 			}
 		}
-		len -= 12+dlen; a += 12+dlen;
+		len -= 10+dlen; a += 10+dlen;
 	}
 	return true;
 }

nfq2/hostlist.c

@@ -8,8 +8,10 @@ static bool addpool(hostlist_pool **hostlist, char **s, const char *end, int *ct
 {
 	char *p=*s;
 
+	for (; p<end && (*p==' ' || *p=='\t') ; p++);
+	*s = p;
 	// comment line ?
-	if ( *p != '#' && *p != ';' && *p != '/' && *p != '\r' && *p != '\n')
+	if (p<end && *p != '#' && *p != ';' && *p != '/' && *p != '\r' && *p != '\n')
 	{
 		// advance until eol lowering all chars
 		uint32_t flags = 0;
@@ -64,18 +66,21 @@ bool AppendHostList(hostlist_pool **hostlist, const char *filename)
 		{
 			DLOG_CONDUP("zlib compression detected. uncompressed size : %zu\n", zsize);
 
-			p = zbuf;
-			e = zbuf + zsize;
-			while(p<e)
+			if (zbuf)
 			{
-				if (!addpool(hostlist,&p,e,&ct))
+				p = zbuf;
+				e = zbuf + zsize;
+				while(p<e)
 				{
-					DLOG_ERR("Not enough memory to store host list : %s\n", filename);
-					free(zbuf);
-					return false;
+					if (!addpool(hostlist,&p,e,&ct))
+					{
+						DLOG_ERR("Not enough memory to store host list : %s\n", filename);
+						free(zbuf);
+						return false;
+					}
 				}
+				free(zbuf);
 			}
-			free(zbuf);
 		}
 		else
 		{

nfq2/ipset.c

@@ -11,36 +11,42 @@ static bool addpool(ipset *ips, char **s, const char *end, int *ct)
 	struct cidr4 c4;
 	struct cidr6 c6;
 
-	for (p=*s; p<end && *p && *p!=' ' && *p!='\t' && *p!='\r' && *p != '\n'; p++);
-
-	// comment line
-	if (!(**s == '#' || **s == ';' || **s == '/' || **s == '\r' || **s == '\n' ))
+	for (p=*s; p<end && (*p==' ' || *p=='\t') ; p++);
+	if (p<end)
 	{
-		l = p-*s;
-		if (l>=sizeof(cidr)) l=sizeof(cidr)-1;
-		memcpy(cidr,*s,l);
-		cidr[l]=0;
+		// comment line
+		if (!(*p == '#' || *p == ';' || *p == '/' || *p == '\r' || *p == '\n' ))
+		{
+			*s=p;
+			// advance to the token's end
+			for (; p<end && *p && *p!=' ' && *p!='\t' && *p!='\r' && *p != '\n'; p++);
 
-		if (parse_cidr4(cidr,&c4))
-		{
-			if (!ipset4AddCidr(&ips->ips4, &c4))
+			l = p-*s;
+			if (l>=sizeof(cidr)) l=sizeof(cidr)-1;
+			memcpy(cidr,*s,l);
+			cidr[l]=0;
+
+			if (parse_cidr4(cidr,&c4))
 			{
-				ipsetDestroy(ips);
-				return false;
+				if (!ipset4AddCidr(&ips->ips4, &c4))
+				{
+					ipsetDestroy(ips);
+					return false;
+				}
+				if (ct) (*ct)++;
 			}
-			if (ct) (*ct)++;
-		}
-		else if (parse_cidr6(cidr,&c6))
-		{
-			if (!ipset6AddCidr(&ips->ips6, &c6))
+			else if (parse_cidr6(cidr,&c6))
 			{
-				ipsetDestroy(ips);
-				return false;
+				if (!ipset6AddCidr(&ips->ips6, &c6))
+				{
+					ipsetDestroy(ips);
+					return false;
+				}
+				if (ct) (*ct)++;
 			}
-			if (ct) (*ct)++;
+			else
+				DLOG_ERR("bad ip or subnet : %s\n",cidr);
 		}
-		else
-			DLOG_ERR("bad ip or subnet : %s\n",cidr);
 	}
 
 	// skip remaining non-eol chars
@@ -81,18 +87,21 @@ static bool AppendIpset(ipset *ips, const char *filename)
 		{
 			DLOG_CONDUP("zlib compression detected. uncompressed size : %zu\n", zsize);
 
-			p = zbuf;
-			e = zbuf + zsize;
-			while(p<e)
+			if (zbuf)
 			{
-				if (!addpool(ips,&p,e,&ct))
+				p = zbuf;
+				e = zbuf + zsize;
+				while(p<e)
 				{
-					DLOG_ERR("Not enough memory to store ipset : %s\n", filename);
-					free(zbuf);
-					return false;
+					if (!addpool(ips,&p,e,&ct))
+					{
+						DLOG_ERR("Not enough memory to store ipset : %s\n", filename);
+						free(zbuf);
+						return false;
+					}
 				}
+				free(zbuf);
 			}
-			free(zbuf);
 		}
 		else
 		{

nfq2/nfqws.c

@@ -43,6 +43,7 @@
 #endif
 
 #ifdef __linux__
+#include <sys/ioctl.h>
 #include <libnetfilter_queue/libnetfilter_queue.h>
 #define NF_DROP 0
 #define NF_ACCEPT 1
@@ -240,6 +241,13 @@ static int write_pidfile(FILE **Fpid)
 
 
 #ifdef __linux__
+
+struct nfq_cb_data
+{
+	uint8_t *mod;
+	int sock;
+};
+
 // cookie must point to mod buffer with size RECONSTRUCT_MAX_SIZE
 static int nfq_cb(struct nfq_q_handle *qh, struct nfgenmsg *nfmsg, struct nfq_data *nfa, void *cookie)
 {
@@ -247,11 +255,10 @@ static int nfq_cb(struct nfq_q_handle *qh, struct nfgenmsg *nfmsg, struct nfq_da
 	size_t len;
 	struct nfqnl_msg_packet_hdr *ph;
 	uint8_t *data;
-	uint32_t ifidx_out, ifidx_in;
-	char ifout[IFNAMSIZ], ifin[IFNAMSIZ];
 	size_t modlen;
-	uint8_t *mod = (uint8_t*)cookie;
+	struct nfq_cb_data *cbdata = (struct nfq_cb_data*)cookie;
 	uint32_t mark;
+	struct ifreq ifr_in, ifr_out;
 
 	ph = nfq_get_msg_packet_hdr(nfa);
 	id = ph ? ntohl(ph->packet_id) : 0;
@@ -259,15 +266,21 @@ static int nfq_cb(struct nfq_q_handle *qh, struct nfgenmsg *nfmsg, struct nfq_da
 	mark = nfq_get_nfmark(nfa);
 	ilen = nfq_get_payload(nfa, &data);
 
-	ifidx_out = nfq_get_outdev(nfa);
-	*ifout = 0;
-	if (ifidx_out) if_indextoname(ifidx_out, ifout);
+	// if_indextoname creates socket, calls ioctl, closes socket
+	// code below prevents socket() and close() syscalls on every packet
+	// this saves CPU 5-10 times
 
-	ifidx_in = nfq_get_indev(nfa);
-	*ifin = 0;
-	if (ifidx_in) if_indextoname(ifidx_in, ifin);
+	*ifr_out.ifr_name = 0;
+	ifr_out.ifr_ifindex = nfq_get_outdev(nfa);
+	if (ifr_out.ifr_ifindex && ioctl(cbdata->sock, SIOCGIFNAME, &ifr_out)<0)
+		DLOG_PERROR("ioctl(SIOCGIFNAME)");
 
-	DLOG("\npacket: id=%d len=%d mark=%08X ifin=%s(%u) ifout=%s(%u)\n", id, ilen, mark, ifin, ifidx_in, ifout, ifidx_out);
+	*ifr_in.ifr_name = 0;
+	ifr_in.ifr_ifindex = nfq_get_indev(nfa);
+	if (ifr_in.ifr_ifindex && ioctl(cbdata->sock, SIOCGIFNAME, &ifr_in)<0)
+		DLOG_PERROR("ioctl(SIOCGIFNAME)");
+
+	DLOG("\npacket: id=%d len=%d mark=%08X ifin=%s(%u) ifout=%s(%u)\n", id, ilen, mark, ifr_in.ifr_name, ifr_in.ifr_ifindex, ifr_out.ifr_name, ifr_out.ifr_ifindex);
 
 	if (ilen >= 0)
 	{
@@ -277,12 +290,12 @@ static int nfq_cb(struct nfq_q_handle *qh, struct nfgenmsg *nfmsg, struct nfq_da
 		// to support increased sizes use separate mod buffer
 		// this is not a problem because only LUA code can trigger VERDICT_MODIFY (and postnat workaround too, once a connection if first packet is dropped)
 		// in case of VERIDCT_MODIFY packet is always reconstructed from dissect, so no difference where to save the data => no performance loss
-		uint8_t verdict = processPacketData(&mark, ifin, ifout, data, len, mod, &modlen);
+		uint8_t verdict = processPacketData(&mark, ifr_in.ifr_name, ifr_out.ifr_name, data, len, cbdata->mod, &modlen);
 		switch (verdict & VERDICT_MASK)
 		{
 		case VERDICT_MODIFY:
 			DLOG("packet: id=%d pass modified. len %zu => %zu\n", id, len, modlen);
-			return nfq_set_verdict2(qh, id, NF_ACCEPT, mark, (uint32_t)modlen, mod);
+			return nfq_set_verdict2(qh, id, NF_ACCEPT, mark, (uint32_t)modlen, cbdata->mod);
 		case VERDICT_DROP:
 			DLOG("packet: id=%d drop\n", id);
 			return nfq_set_verdict2(qh, id, NF_DROP, mark, 0, NULL);
@@ -306,7 +319,7 @@ static void nfq_deinit(struct nfq_handle **h, struct nfq_q_handle **qh)
 		*h = NULL;
 	}
 }
-static bool nfq_init(struct nfq_handle **h, struct nfq_q_handle **qh, uint8_t *mod_buffer)
+static bool nfq_init(struct nfq_handle **h, struct nfq_q_handle **qh, struct nfq_cb_data *cbdata)
 {
 	nfq_deinit(h, qh);
 
@@ -343,7 +356,7 @@ static bool nfq_init(struct nfq_handle **h, struct nfq_q_handle **qh, uint8_t *m
 	}
 
 	DLOG_CONDUP("binding this socket to queue '%u'\n", params.qnum);
-	*qh = nfq_create_queue(*h, params.qnum, &nfq_cb, mod_buffer);
+	*qh = nfq_create_queue(*h, params.qnum, &nfq_cb, cbdata);
 	if (!*qh) {
 		DLOG_PERROR("nfq_create_queue()");
 		goto exiterr;
@@ -365,6 +378,8 @@ static bool nfq_init(struct nfq_handle **h, struct nfq_q_handle **qh, uint8_t *m
 		// dot not fail. not supported in old linuxes <3.6 
 	}
 
+	nfnl_rcvbufsiz(nfq_nfnlh(*h), Q_RCVBUF);
+
 	int yes = 1, fd = nfq_fd(*h);
 
 #if defined SOL_NETLINK && defined NETLINK_NO_ENOBUFS
@@ -388,7 +403,7 @@ static void notify_ready(void)
 }
 
 // extra space for netlink headers
-#define NFQ_MAX_RECV_SIZE (RECONSTRUCT_MAX_SIZE+512)
+#define NFQ_MAX_RECV_SIZE (RECONSTRUCT_MAX_SIZE+4096)
 static int nfq_main(void)
 {
 	struct nfq_handle *h = NULL;
@@ -397,6 +412,7 @@ static int nfq_main(void)
 	ssize_t rd;
 	FILE *Fpid = NULL;
 	uint8_t *buf=NULL, *mod=NULL;
+	struct nfq_cb_data cbdata = { .sock = -1 };
 
 	if (*params.pidfile && !(Fpid = fopen(params.pidfile, "w")))
 	{
@@ -438,13 +454,19 @@ static int nfq_main(void)
 		goto exok;
 	}
 
-	if (!(buf = malloc(NFQ_MAX_RECV_SIZE)) || !(mod = malloc(RECONSTRUCT_MAX_SIZE)))
+	if (!(buf = malloc(NFQ_MAX_RECV_SIZE)) || !(cbdata.mod = malloc(RECONSTRUCT_MAX_SIZE)))
 	{
 		DLOG_ERR("out of memory\n");
 		goto err;
 	}
 
-	if (!nfq_init(&h, &qh, mod))
+	if ((cbdata.sock = socket(AF_INET, SOCK_DGRAM, 0)) < 0)
+	{
+		DLOG_PERROR("socket");
+		goto err;
+	}
+
+	if (!nfq_init(&h, &qh, &cbdata))
 		goto err;
 
 #ifdef HAS_FILTER_SSID
@@ -483,7 +505,7 @@ static int nfq_main(void)
 					DLOG_ERR("cannot get wlan info\n");
 #endif
 			int r = nfq_handle_packet(h, (char *)buf, (int)rd);
-			if (r) DLOG_ERR("nfq_handle_packet error %d\n", r);
+			if (r<0) DLOG_ERR("nfq_handle_packet result %d, errno %d : %s\n", r, errno, strerror(errno));
 			if (bQuit) goto quit;
 		}
 		if (errno==EINTR)
@@ -498,7 +520,8 @@ static int nfq_main(void)
 exok:
 	res=0;
 ex:
-	free(mod);
+	free(cbdata.mod);
+	if (cbdata.sock>=0) close(cbdata.sock);
 	free(buf);
 	nfq_deinit(&h, &qh);
 	lua_shutdown();

nfq2/params.h

@@ -23,7 +23,8 @@
 
 #define RAW_SNDBUF	(64*1024)	// in bytes
 
-#define Q_MAXLEN	1024		// in packets
+#define Q_MAXLEN	4096		// in packets
+#define Q_RCVBUF	(1024*1024)	// in bytes
 
 #define HOSTLIST_AUTO_FAIL_THRESHOLD_DEFAULT	3
 #define	HOSTLIST_AUTO_FAIL_TIME_DEFAULT 	60

nfq2/protocol.c

@@ -1445,7 +1445,7 @@ bool IsStunMessage(const uint8_t *data, size_t len)
 		(data[0]&0xC0)==0 && // 2 most significant bits must be zeroes
 		(data[3]&3)==0 && // length must be a multiple of 4
 		pntoh32(data+4)==0x2112A442 && // magic cookie
-		pntoh16(data+2)==(len-20);
+		pntoh16(data+2)<=(len-20);
 }
 #if defined(__GNUC__) && !defined(__llvm__)
 __attribute__((optimize ("no-strict-aliasing")))
@@ -1460,7 +1460,7 @@ bool IsMTProto(const uint8_t *data, size_t len)
 		return !memcmp(decrypt+56,"\xEF\xEF\xEF\xEF",4);
 */
 		// this way requires only one AES instead of 4
-		uint8_t decrypt[16] __attribute__((aligned)), iv[16];
+		uint8_t decrypt[16] __attribute__((aligned(16))), iv[16] __attribute__((aligned(16)));
 		aes_context ctx;
 
 		memcpy(iv, data+40, 16);