Add ipfw support for setting/matching DiffServ codepoints (DSCP).

Setting DSCP support is done via O_SETDSCP which works for both
IPv4 and IPv6 packets. Fast checksum recalculation (RFC 1624) is done for IPv4.
Dscp can be specified by name (AFXY, CSX, BE, EF), by value
(0..63) or via tablearg.

Matching DSCP is done via another opcode (O_DSCP) which accepts several
classes at once (af11,af22,be). Classes are stored in bitmask (2 u32 words).

Many people made their variants of this patch, the ones I'm aware of are
(in alphabetic order):

Dmitrii Tejblum
Marcelo Araujo
Roman Bogorodskiy (novel)
Sergey Matveichuk (sem)
Sergey Ryabin

PR:		kern/102471, kern/121122
MFC after:	2 weeks

7 files changed, 283 insertions, 4 deletions

diff --git a/sbin/ipfw/ipfw.8 b/sbin/ipfw/ipfw.8
index db0dfc0..a68d490 100644
--- a/sbin/ipfw/ipfw.8
+++ b/sbin/ipfw/ipfw.8
@@ -960,6 +960,61 @@ It is possible to use the
 keyword with setfib.
 If the tablearg value is not within the compiled range of fibs,
 the packet's fib is set to 0.
+.It Cm setdscp Ar DSCP | number | tablearg
+Set specified DiffServ codepoint for an IPv4/IPv6 packet.
+Processing continues at the next rule.
+Supported values are:
+.Pp
+.Cm CS0
+.Pq Dv 000000 ,
+.Cm CS1
+.Pq Dv 001000 ,
+.Cm CS2
+.Pq Dv 010000 ,
+.Cm CS3
+.Pq Dv 011000 ,
+.Cm CS4
+.Pq Dv 100000 ,
+.Cm CS5
+.Pq Dv 101000 ,
+.Cm CS6
+.Pq Dv 110000 ,
+.Cm CS7
+.Pq Dv 111000 ,
+.Cm AF11
+.Pq Dv 001010 ,
+.Cm AF12
+.Pq Dv 001100 ,
+.Cm AF13
+.Pq Dv 001110 ,
+.Cm AF21
+.Pq Dv 010010 ,
+.Cm AF22
+.Pq Dv 010100 ,
+.Cm AF23
+.Pq Dv 010110 ,
+.Cm AF31
+.Pq Dv 011010 ,
+.Cm AF32
+.Pq Dv 011100 ,
+.Cm AF33
+.Pq Dv 011110 ,
+.Cm AF41
+.Pq Dv 100010 ,
+.Cm AF42
+.Pq Dv 100100 ,
+.Cm AF43
+.Pq Dv 100110 ,
+.Cm EF
+.Pq Dv 101110 ,
+.Cm BE
+.Pq Dv 000000 .
+Additionally, DSCP value can be specified by number (0..64).
+It is also possible to use the
+.Cm tablearg
+keyword with setdscp.
+If the tablearg value is not within the 0..64 range, lower 6 bits of supplied
+value are used.
 .It Cm reass
 Queue and reassemble IP fragments.
 If the packet is not fragmented, counters are updated and
@@ -1454,6 +1509,17 @@ The supported IP types of service are:
 The absence of a particular type may be denoted
 with a
 .Ql \&! .
+.It Cm dscp spec Ns Op , Ns Ar spec
+Matches IPv4/IPv6 packets whose
+.Cm DS
+field value is contained in
+.Ar spec
+mask.
+Multiple values can be specified via 
+the comma separated list.
+Value can be one of keywords used in
+.Cm setdscp
+action or exact number.
 .It Cm ipttl Ar ttl-list
 Matches IPv4 packets whose time to live is included in
 .Ar ttl-list ,
@@ -2976,6 +3042,23 @@ configured on
 but coming in on
 .Li fxp1
 would be dropped.
+.Pp
+The
+.Cm setdscp
+option could be used to (re)mark user traffic,
+by adding the following to the appropriate place in ruleset:
+.Pp
+.Dl "ipfw add setdscp be ip from any to any dscp af11,af21"
+.Pp
+This rule drops all incoming packets that appear to be coming from another
+directly connected system but on the wrong interface.
+For example, a packet with a source address of
+.Li 192.168.0.0/24 ,
+configured on
+.Li fxp0 ,
+but coming in on
+.Li fxp1
+would be dropped.
 .Ss DYNAMIC RULES
 In order to protect a site from flood attacks involving fake
 TCP packets, it is safer to use dynamic rules:
diff --git a/sbin/ipfw/ipfw2.c b/sbin/ipfw/ipfw2.c
index ae5d2aa..39934ab 100644
--- a/sbin/ipfw/ipfw2.c
+++ b/sbin/ipfw/ipfw2.c
@@ -167,6 +167,32 @@ static struct _s_x f_iptos[] = {
 	{ NULL,	0 }
 };
 
+static struct _s_x f_ipdscp[] = {
+	{ "af11", IPTOS_DSCP_AF11 >> 2 },	/* 001010 */
+	{ "af12", IPTOS_DSCP_AF12 >> 2 },	/* 001100 */
+	{ "af13", IPTOS_DSCP_AF13 >> 2 },	/* 001110 */
+	{ "af21", IPTOS_DSCP_AF21 >> 2 },	/* 010010 */
+	{ "af22", IPTOS_DSCP_AF22 >> 2 },	/* 010100 */
+	{ "af23", IPTOS_DSCP_AF23 >> 2 },	/* 010110 */
+	{ "af31", IPTOS_DSCP_AF31 >> 2 },	/* 011010 */
+	{ "af32", IPTOS_DSCP_AF32 >> 2 },	/* 011100 */
+	{ "af33", IPTOS_DSCP_AF33 >> 2 },	/* 011110 */
+	{ "af41", IPTOS_DSCP_AF41 >> 2 },	/* 100010 */
+	{ "af42", IPTOS_DSCP_AF42 >> 2 },	/* 100100 */
+	{ "af43", IPTOS_DSCP_AF43 >> 2 },	/* 100110 */
+	{ "be", IPTOS_DSCP_CS0 >> 2 }, 	/* 000000 */
+	{ "ef", IPTOS_DSCP_EF >> 2 },	/* 101110 */
+	{ "cs0", IPTOS_DSCP_CS0 >> 2 },	/* 000000 */
+	{ "cs1", IPTOS_DSCP_CS1 >> 2 },	/* 001000 */
+	{ "cs2", IPTOS_DSCP_CS2 >> 2 },	/* 010000 */
+	{ "cs3", IPTOS_DSCP_CS3 >> 2 },	/* 011000 */
+	{ "cs4", IPTOS_DSCP_CS4 >> 2 },	/* 100000 */
+	{ "cs5", IPTOS_DSCP_CS5 >> 2 },	/* 101000 */
+	{ "cs6", IPTOS_DSCP_CS6 >> 2 },	/* 110000 */
+	{ "cs7", IPTOS_DSCP_CS7 >> 2 },	/* 100000 */
+	{ NULL, 0 }
+};
+
 static struct _s_x limit_masks[] = {
 	{"all",		DYN_SRC_ADDR|DYN_SRC_PORT|DYN_DST_ADDR|DYN_DST_PORT},
 	{"src-addr",	DYN_SRC_ADDR},
@@ -237,6 +263,7 @@ static struct _s_x rule_actions[] = {
 	{ "nat",		TOK_NAT },
 	{ "reass",		TOK_REASS },
 	{ "setfib",		TOK_SETFIB },
+	{ "setdscp",		TOK_SETDSCP },
 	{ "call",		TOK_CALL },
 	{ "return",		TOK_RETURN },
 	{ NULL, 0 }	/* terminator */
@@ -714,6 +741,51 @@ fill_newports(ipfw_insn_u16 *cmd, char *av, int proto, int cblen)
 	return (i);
 }
 
+/*
+ * Fill the body of the command with the list of DiffServ codepoints.
+ */
+static void
+fill_dscp(ipfw_insn *cmd, char *av, int cblen)
+{
+	uint32_t *low, *high;
+	char *s = av, *a;
+	int code;
+
+	cmd->opcode = O_DSCP;
+	cmd->len |= F_INSN_SIZE(ipfw_insn_u32) + 1;
+
+	CHECK_CMDLEN;
+
+	low = (uint32_t *)(cmd + 1);
+	high = low + 1;
+
+	*low = 0;
+	*high = 0;
+
+	while (s != NULL) {
+		a = strchr(s, ',');
+
+		if (a != NULL)
+			*a++ = '\0';
+
+		if (isalpha(*s)) {
+			if ((code = match_token(f_ipdscp, s)) == -1)
+				errx(EX_DATAERR, "Unknown DSCP code");
+		} else {
+			code = strtoul(s, NULL, 10);
+			if (code < 0 || code > 63)
+				errx(EX_DATAERR, "Invalid DSCP value");
+		}
+
+		if (code > 32)
+			*high |= 1 << (code - 32);
+		else
+			*low |= 1 << code;
+
+		s = a;
+	}
+}
+
 static struct _s_x icmpcodes[] = {
       { "net",			ICMP_UNREACH_NET },
       { "host",			ICMP_UNREACH_HOST },
@@ -972,6 +1044,32 @@ print_icmptypes(ipfw_insn_u32 *cmd)
 	}
 }
 
+static void
+print_dscp(ipfw_insn_u32 *cmd)
+{
+	int i, c;
+	uint32_t *v;
+	char sep= ' ';
+	const char *code;
+
+	printf(" dscp");
+	i = 0;
+	c = 0;
+	v = cmd->d;
+	while (i < 64) {
+		if (*v & (1 << i)) {
+			if ((code = match_value(f_ipdscp, i)) != NULL)
+				printf("%c%s", sep, code);
+			else
+				printf("%c%d", sep, i);
+			sep = ',';
+		}
+
+		if ((++i % 32) == 0)
+			v++;
+	}
+}
+
 /*
  * show_ipfw() prints the body of an ipfw rule.
  * Because the standard rule has at least proto src_ip dst_ip, we use
@@ -1205,6 +1303,17 @@ show_ipfw(struct ip_fw *rule, int pcwidth, int bcwidth)
 			PRINT_UINT_ARG("setfib ", cmd->arg1);
  			break;
 
+		case O_SETDSCP:
+		    {
+			const char *code;
+
+			if ((code = match_value(f_ipdscp, cmd->arg1)) != NULL)
+				printf("setdscp %s", code);
+			else
+				PRINT_UINT_ARG("setdscp ", cmd->arg1);
+		    }
+ 			break;
+
 		case O_REASS:
 			printf("reass");
 			break;
@@ -1500,6 +1609,10 @@ show_ipfw(struct ip_fw *rule, int pcwidth, int bcwidth)
 				printf(" ipprecedence %u", (cmd->arg1) >> 5 );
 				break;
 
+			case O_DSCP:
+				print_dscp((ipfw_insn_u32 *)cmd);
+	 			break;
+
 			case O_IPLEN:
 				if (F_LEN(cmd) == 1)
 				    printf(" iplen %u", cmd->arg1 );
@@ -3036,6 +3149,24 @@ chkarg:
 		break;
 	    }
 
+	case TOK_SETDSCP:
+	    {
+		int code;
+
+		action->opcode = O_SETDSCP;
+		NEED1("missing DSCP code");
+		if (_substrcmp(*av, "tablearg") == 0) {
+			action->arg1 = IP_FW_TABLEARG;
+		} else if (isalpha(*av[0])) {
+			if ((code = match_token(f_ipdscp, *av)) == -1)
+				errx(EX_DATAERR, "Unknown DSCP code");
+			action->arg1 = code;
+		} else
+		        action->arg1 = strtoul(*av, NULL, 10);
+		av++;
+		break;
+	    }
+
 	case TOK_REASS:
 		action->opcode = O_REASS;
 		break;
@@ -3448,6 +3579,12 @@ read_options:
 			av++;
 			break;
 
+		case TOK_DSCP:
+			NEED1("missing DSCP code");
+			fill_dscp(cmd, *av, cblen);
+			av++;
+			break;
+
 		case TOK_IPOPTS:
 			NEED1("missing argument for ipoptions");
 			fill_flags(cmd, O_IPOPT, f_ipopts, *av);
diff --git a/sbin/ipfw/ipfw2.h b/sbin/ipfw/ipfw2.h
index 81e0ba8..d592930 100644
--- a/sbin/ipfw/ipfw2.h
+++ b/sbin/ipfw/ipfw2.h
@@ -203,6 +203,7 @@ enum tokens {
 	TOK_SETFIB,
 	TOK_LOOKUP,
 	TOK_SOCKARG,
+	TOK_SETDSCP,
 };
 /*
  * the following macro returns an error message if we run out of
diff --git a/sys/netinet/ip_fw.h b/sys/netinet/ip_fw.h
index 9693523..14b08f5 100644
--- a/sys/netinet/ip_fw.h
+++ b/sys/netinet/ip_fw.h
@@ -218,6 +218,9 @@ enum ipfw_opcodes {		/* arguments (4 byte each)	*/
 
 	O_FORWARD_IP6,		/* fwd sockaddr_in6             */
 
+	O_DSCP,			/* 2 u32 = DSCP mask */
+	O_SETDSCP,		/* arg1=DSCP value */
+
 	O_LAST_OPCODE		/* not an opcode!		*/
 };
 
diff --git a/sys/netpfil/ipfw/ip_fw2.c b/sys/netpfil/ipfw/ip_fw2.c
index 1c096d9..5e0675b 100644
--- a/sys/netpfil/ipfw/ip_fw2.c
+++ b/sys/netpfil/ipfw/ip_fw2.c
@@ -1624,6 +1624,32 @@ do {								\
 				    flags_match(cmd, ip->ip_tos));
 				break;
 
+			case O_DSCP:
+			    {
+				uint32_t *p;
+				uint16_t x;
+
+				p = ((ipfw_insn_u32 *)cmd)->d;
+
+				if (is_ipv4)
+					x = ip->ip_tos >> 2;
+				else if (is_ipv6) {
+					uint8_t *v;
+					v = &((struct ip6_hdr *)ip)->ip6_vfc;
+					x = (*v & 0x0F) << 2;
+					v++;
+					x |= *v >> 6;
+				} else
+					break;
+
+				/* DSCP bitmask is stored as low_u32 high_u32 */
+				if (x > 32)
+					match = *(p + 1) & (1 << (x - 32));
+				else
+					match = *p & (1 << x);
+			    }
+				break;
+
 			case O_TCPDATALEN:
 				if (proto == IPPROTO_TCP && offset == 0) {
 				    struct tcphdr *tcp;
@@ -2353,6 +2379,32 @@ do {								\
 				break;
 		        }
 
+			case O_SETDSCP: {
+				uint16_t code;
+
+				code = IP_FW_ARG_TABLEARG(cmd->arg1) & 0x3F;
+				l = 0;		/* exit inner loop */
+				if (is_ipv4) {
+					uint16_t a;
+
+					a = ip->ip_tos;
+					ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
+					a += ntohs(ip->ip_sum) - ip->ip_tos;
+					ip->ip_sum = htons(a);
+				} else if (is_ipv6) {
+					uint8_t *v;
+
+					v = &((struct ip6_hdr *)ip)->ip6_vfc;
+					*v = (*v & 0xF0) | (code >> 2);
+					v++;
+					*v = (*v & 0x3F) | ((code & 0x03) << 6);
+				} else
+					break;
+
+				IPFW_INC_RULE_COUNTER(f, pktlen);
+				break;
+			}
+
 			case O_NAT:
  				if (!IPFW_NAT_LOADED) {
 				    retval = IP_FW_DENY;
diff --git a/sys/netpfil/ipfw/ip_fw_log.c b/sys/netpfil/ipfw/ip_fw_log.c
index 48e6791..0f3ab9d 100644
--- a/sys/netpfil/ipfw/ip_fw_log.c
+++ b/sys/netpfil/ipfw/ip_fw_log.c
@@ -292,10 +292,8 @@ ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args,
 				altq->qid);
 			cmd += F_LEN(cmd);
 		}
-		if (cmd->opcode == O_PROB)
-			cmd += F_LEN(cmd);
-
-		if (cmd->opcode == O_TAG)
+		if (cmd->opcode == O_PROB || cmd->opcode == O_TAG ||
+		    cmd->opcode == O_SETDSCP)
 			cmd += F_LEN(cmd);
 
 		action = action2;
diff --git a/sys/netpfil/ipfw/ip_fw_sockopt.c b/sys/netpfil/ipfw/ip_fw_sockopt.c
index 3363cf9..45da394 100644
--- a/sys/netpfil/ipfw/ip_fw_sockopt.c
+++ b/sys/netpfil/ipfw/ip_fw_sockopt.c
@@ -671,6 +671,10 @@ check_ipfw_struct(struct ip_fw *rule, int size)
 		case O_IPID:
 		case O_IPTTL:
 		case O_IPLEN:
+		case O_DSCP:
+			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1)
+				goto bad_size;
+			break;
 		case O_TCPDATALEN:
 		case O_TCPWIN:
 		case O_TAGGED:
@@ -738,6 +742,7 @@ check_ipfw_struct(struct ip_fw *rule, int size)
 		case O_ACCEPT:
 		case O_DENY:
 		case O_REJECT:
+		case O_SETDSCP:
 #ifdef INET6
 		case O_UNREACH6:
 #endif