[tcpip] Add IP statistics collection as per RFC 4293

Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/include/ipxe/fragment.h

 	size_t hdrlen;
 	/** Reassembly timer */
 	struct retry_timer timer;
+	/** Fragment reassembler */
+	struct fragment_reassembler *fragments;
 };
 
 /** A fragment reassembler */
 	 * @ret more_frags	More fragments exist
 	 */
 	int ( * more_fragments ) ( struct io_buffer *iobuf, size_t hdrlen );
+	/** Associated IP statistics */
+	struct ip_statistics *stats;
 };
 
 extern struct io_buffer *

src/include/ipxe/ipstat.h

+#ifndef _IPXE_IPSTATS_H
+#define _IPXE_IPSTATS_H
+
+/** @file
+ *
+ * IP statistics
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <ipxe/tables.h>
+
+struct io_buffer;
+
+/** IP system statistics
+ *
+ * Definitions are taken from the RFC4293 section 5
+ * "ipSystemStatsEntry" table.
+ *
+ * To minimise code size, we use "unsigned long" as the counter
+ * variable type regardless of whether this type is 32-bit or 64-bit.
+ * On a 32-bit build (e.g. the standard BIOS build), this means that
+ * we omit the "high capacity" 64-bit counters (prefixed with "HC").
+ * This reduces the code size required to maintain the counter values,
+ * and avoids the need to support the "%lld" format in vsprintf.c
+ * (which would require dragging in the 64-bit division library on a
+ * standard 32-bit build).  Since total available memory in a 32-bit
+ * environment is limited to 4GB, it is unlikely that we will overflow
+ * even the 32-bit octet counters under normal operation.
+ *
+ * Counters relating to packet forwarding are omitted, since iPXE
+ * includes no functionality for acting as a router.
+ *
+ * Counters related to output fragmentation are omitted, since iPXE
+ * has no support for fragmenting transmitted packets.
+ *
+ * The ipSystemStatsInDiscards and ipSystemStatsOutDiscards counters
+ * are omitted, since they will always be zero.
+ *
+ * Separate octet counters for multicast packets are omitted to save
+ * code size.
+ */
+struct ip_statistics {
+	/** ipSystemStatsInReceives
+	 *
+	 * The total number of input IP datagrams received, including
+	 * those received in error.
+	 */
+	unsigned long in_receives;
+	/** ipSystemStatsInOctets
+	 *
+	 * The total number of octets received in input IP datagrams,
+	 * including those received in error.  Octets from datagrams
+	 * counted in ipSystemStatsInReceives MUST be counted here.
+	 */
+	unsigned long in_octets;
+	/** ipSystemStatsInHdrErrors
+	 *
+	 * The number of input IP datagrams discarded due to errors in
+	 * their IP headers, including version number mismatch, other
+	 * format errors, hop count exceeded, errors discovered in
+	 * processing their IP options, etc.
+	 */
+	unsigned long in_hdr_errors;
+	/** ipSystemStatsInAddrErrors
+	 *
+	 * The number of input IP datagrams discarded because the IP
+	 * address in their IP header's destination field was not a
+	 * valid address to be received at this entity.  This count
+	 * includes invalid addresses (e.g., ::0).  For entities that
+	 * are not IP routers and therefore do not forward datagrams,
+	 * this counter includes datagrams discarded because the
+	 * destination address was not a local address.
+	 */
+	unsigned long in_addr_errors;
+	/** ipSystemStatsInUnknownProtos
+	 *
+	 * The number of locally-addressed IP datagrams received
+	 * successfully but discarded because of an unknown or
+	 * unsupported protocol.
+	 */
+	unsigned long in_unknown_protos;
+	/** ipSystemStatsInTruncatedPkts
+	 *
+	 * The number of input IP datagrams discarded because the
+	 * datagram frame didn't carry enough data.
+	 */
+	unsigned long in_truncated_pkts;
+	/** ipSystemStatsReasmReqds
+	 *
+	 * The number of IP fragments received that needed to be
+	 * reassembled at this interface.
+	 */
+	unsigned long reasm_reqds;
+	/** ipSystemStatsReasmOks
+	 *
+	 * The number of IP datagrams successfully reassembled.
+	 */
+	unsigned long reasm_oks;
+	/** ipSystemStatsReasmFails
+	 *
+	 * The number of failures detected by the IP re-assembly
+	 * algorithm (for whatever reason: timed out, errors, etc.).
+	 * Note that this is not necessarily a count of discarded IP
+	 * fragments since some algorithms (notably the algorithm in
+	 * RFC 815) can lose track of the number of fragments by
+	 * combining them as they are received.
+	 */
+	unsigned long reasm_fails;
+	/** ipSystemStatsInDelivers
+	 *
+	 * The total number of datagrams successfully delivered to IP
+	 * user-protocols (including ICMP).
+	 */
+	unsigned long in_delivers;
+	/** ipSystemStatsOutRequests
+	 *
+	 * The total number of IP datagrams that local IP user-
+	 * protocols (including ICMP) supplied to IP in requests for
+	 * transmission.
+	 */
+	unsigned long out_requests;
+	/** ipSystemStatsOutNoRoutes
+	 *
+	 * The number of locally generated IP datagrams discarded
+	 * because no route could be found to transmit them to their
+	 * destination.
+	 */
+	unsigned long out_no_routes;
+	/** ipSystemStatsOutTransmits
+	 *
+	 * The total number of IP datagrams that this entity supplied
+	 * to the lower layers for transmission.  This includes
+	 * datagrams generated locally and those forwarded by this
+	 * entity.
+	 */
+	unsigned long out_transmits;
+	/** ipSystemStatsOutOctets
+	 *
+	 * The total number of octets in IP datagrams delivered to the
+	 * lower layers for transmission.  Octets from datagrams
+	 * counted in ipSystemStatsOutTransmits MUST be counted here.
+	 */
+	unsigned long out_octets;
+	/** ipSystemStatsInMcastPkts
+	 *
+	 * The number of IP multicast datagrams received.
+	 */
+	unsigned long in_mcast_pkts;
+	/** ipSystemStatsOutMcastPkts
+	 *
+	 * The number of IP multicast datagrams transmitted.
+	 */
+	unsigned long out_mcast_pkts;
+	/** ipSystemStatsInBcastPkts
+	 *
+	 * The number of IP broadcast datagrams received.
+	 */
+	unsigned long in_bcast_pkts;
+	/** ipSystemStatsOutBcastPkts
+	 *
+	 * The number of IP broadcast datagrams transmitted.
+	 */
+	unsigned long out_bcast_pkts;
+};
+
+/** An IP system statistics family */
+struct ip_statistics_family {
+	/** IP version */
+	unsigned int version;
+	/** Statistics */
+	struct ip_statistics *stats;
+};
+
+/** IP system statistics family table */
+#define IP_STATISTICS_FAMILIES \
+	__table ( struct ip_statistics_family, "ip_statistics_families" )
+
+/** Declare an IP system statistics family */
+#define __ip_statistics_family( order ) \
+	__table_entry ( IP_STATISTICS_FAMILIES, order )
+
+#define IP_STATISTICS_IPV4 01
+#define IP_STATISTICS_IPV6 02
+
+#endif /* _IPXE_IPSTATS_H */

src/include/ipxe/tcpip.h

 
 struct io_buffer;
 struct net_device;
+struct ip_statistics;
 
 /** Empty checksum value
  *
 
 extern int tcpip_rx ( struct io_buffer *iobuf, struct net_device *netdev,
 		      uint8_t tcpip_proto, struct sockaddr_tcpip *st_src,
-		      struct sockaddr_tcpip *st_dest, uint16_t pshdr_csum );
+		      struct sockaddr_tcpip *st_dest, uint16_t pshdr_csum,
+		      struct ip_statistics *stats );
 extern int tcpip_tx ( struct io_buffer *iobuf, struct tcpip_protocol *tcpip,
 		      struct sockaddr_tcpip *st_src,
 		      struct sockaddr_tcpip *st_dest,

src/net/fragment.c

 #include <string.h>
 #include <ipxe/retry.h>
 #include <ipxe/timer.h>
+#include <ipxe/ipstat.h>
 #include <ipxe/fragment.h>
 
 /** @file
 	DBGC ( fragment, "FRAG %p expired\n", fragment );
 	free_iob ( fragment->iobuf );
 	list_del ( &fragment->list );
+	fragment->fragments->stats->reasm_fails++;
 	free ( fragment );
 }
 
 	size_t expected_offset;
 	int more_frags;
 
+	/* Update statistics */
+	fragments->stats->reasm_reqds++;
+
 	/* Find matching fragment reassembly buffer, if any */
 	fragment = fragment_find ( fragments, iobuf, *hdrlen );
 
 		fragment->iobuf = iobuf;
 		fragment->hdrlen = *hdrlen;
 		timer_init ( &fragment->timer, fragment_expired, NULL );
+		fragment->fragments = fragments;
 		DBGC ( fragment, "FRAG %p [0,%zd)\n", fragment,
 		       ( iob_len ( iobuf ) - *hdrlen ) );
 
 			*hdrlen = fragment->hdrlen;
 			list_del ( &fragment->list );
 			free ( fragment );
+			fragments->stats->reasm_oks++;
 			return iobuf;
 		}
 	}
 	return NULL;
 
  drop:
+	fragments->stats->reasm_fails++;
 	free_iob ( iobuf );
 	return NULL;
 }

src/net/ipv4.c

 #include <ipxe/dhcp.h>
 #include <ipxe/settings.h>
 #include <ipxe/fragment.h>
+#include <ipxe/ipstat.h>
 
 /** @file
  *
 /** List of IPv4 miniroutes */
 struct list_head ipv4_miniroutes = LIST_HEAD_INIT ( ipv4_miniroutes );
 
+/** IPv4 statistics */
+static struct ip_statistics ipv4_stats;
+
+/** IPv4 statistics family */
+struct ip_statistics_family
+ipv4_stats_family __ip_statistics_family ( IP_STATISTICS_IPV4 ) = {
+	.version = 4,
+	.stats = &ipv4_stats,
+};
+
 /**
  * Add IPv4 minirouting table entry
  *
 	.is_fragment = ipv4_is_fragment,
 	.fragment_offset = ipv4_fragment_offset,
 	.more_fragments = ipv4_more_fragments,
+	.stats = &ipv4_stats,
 };
 
 /**
 	const void *ll_dest;
 	int rc;
 
+	/* Update statistics */
+	ipv4_stats.out_requests++;
+
 	/* Fill up the IP header, except source address */
 	memset ( iphdr, 0, sizeof ( *iphdr ) );
 	iphdr->verhdrlen = ( IP_VER | ( sizeof ( *iphdr ) / 4 ) );
 	if ( ! netdev ) {
 		DBGC ( sin_dest->sin_addr, "IPv4 has no route to %s\n",
 		       inet_ntoa ( iphdr->dest ) );
+		ipv4_stats.out_no_routes++;
 		rc = -ENETUNREACH;
 		goto err;
 	}
 	/* Calculate link-layer destination address, if possible */
 	if ( ( ( next_hop.s_addr ^ INADDR_BROADCAST ) & ~netmask.s_addr ) == 0){
 		/* Broadcast address */
+		ipv4_stats.out_bcast_pkts++;
 		ll_dest = netdev->ll_broadcast;
 	} else if ( IN_MULTICAST ( ntohl ( next_hop.s_addr ) ) ) {
 		/* Multicast address */
+		ipv4_stats.out_mcast_pkts++;
 		if ( ( rc = netdev->ll_protocol->mc_hash ( AF_INET, &next_hop,
 							   ll_dest_buf ) ) !=0){
 			DBGC ( sin_dest->sin_addr, "IPv4 could not hash "
 		ll_dest = NULL;
 	}
 
+	/* Update statistics */
+	ipv4_stats.out_transmits++;
+	ipv4_stats.out_octets += iob_len ( iobuf );
+
 	/* Hand off to link layer (via ARP if applicable) */
 	if ( ll_dest ) {
 		if ( ( rc = net_tx ( iobuf, netdev, &ipv4_protocol, ll_dest,
 	uint16_t pshdr_csum;
 	int rc;
 
+	/* Update statistics */
+	ipv4_stats.in_receives++;
+	ipv4_stats.in_octets += iob_len ( iobuf );
+	if ( flags & LL_BROADCAST ) {
+		ipv4_stats.in_bcast_pkts++;
+	} else if ( flags & LL_MULTICAST ) {
+		ipv4_stats.in_mcast_pkts++;
+	}
+
 	/* Sanity check the IPv4 header */
 	if ( iob_len ( iobuf ) < sizeof ( *iphdr ) ) {
 		DBGC ( iphdr->src, "IPv4 packet too short at %zd bytes (min "
 		       "%zd bytes)\n", iob_len ( iobuf ), sizeof ( *iphdr ) );
-		goto err;
+		goto err_header;
 	}
 	if ( ( iphdr->verhdrlen & IP_MASK_VER ) != IP_VER ) {
 		DBGC ( iphdr->src, "IPv4 version %#02x not supported\n",
 		       iphdr->verhdrlen );
-		goto err;
+		goto err_header;
 	}
 	hdrlen = ( ( iphdr->verhdrlen & IP_MASK_HLEN ) * 4 );
 	if ( hdrlen < sizeof ( *iphdr ) ) {
 		DBGC ( iphdr->src, "IPv4 header too short at %zd bytes (min "
 		       "%zd bytes)\n", hdrlen, sizeof ( *iphdr ) );
-		goto err;
+		goto err_header;
 	}
 	if ( hdrlen > iob_len ( iobuf ) ) {
 		DBGC ( iphdr->src, "IPv4 header too long at %zd bytes "
 		       "(packet is %zd bytes)\n", hdrlen, iob_len ( iobuf ) );
-		goto err;
+		goto err_header;
 	}
 	if ( ( csum = tcpip_chksum ( iphdr, hdrlen ) ) != 0 ) {
 		DBGC ( iphdr->src, "IPv4 checksum incorrect (is %04x "
 		       "including checksum field, should be 0000)\n", csum );
-		goto err;
+		goto err_header;
 	}
 	len = ntohs ( iphdr->len );
 	if ( len < hdrlen ) {
 		DBGC ( iphdr->src, "IPv4 length too short at %zd bytes "
 		       "(header is %zd bytes)\n", len, hdrlen );
-		goto err;
+		goto err_header;
 	}
 	if ( len > iob_len ( iobuf ) ) {
 		DBGC ( iphdr->src, "IPv4 length too long at %zd bytes "
 		       "(packet is %zd bytes)\n", len, iob_len ( iobuf ) );
-		goto err;
+		ipv4_stats.in_truncated_pkts++;
+		goto err_other;
 	}
 
 	/* Truncate packet to correct length */
 	     ( ! ipv4_has_addr ( netdev, iphdr->dest ) ) ) {
 		DBGC ( iphdr->src, "IPv4 discarding non-local unicast packet "
 		       "for %s\n", inet_ntoa ( iphdr->dest ) );
-		goto err;
+		ipv4_stats.in_addr_errors++;
+		goto err_other;
 	}
 
 	/* Perform fragment reassembly if applicable */
 	pshdr_csum = ipv4_pshdr_chksum ( iobuf, TCPIP_EMPTY_CSUM );
 	iob_pull ( iobuf, hdrlen );
 	if ( ( rc = tcpip_rx ( iobuf, netdev, iphdr->protocol, &src.st,
-			       &dest.st, pshdr_csum ) ) != 0 ) {
+			       &dest.st, pshdr_csum, &ipv4_stats ) ) != 0 ) {
 		DBGC ( src.sin.sin_addr, "IPv4 received packet rejected by "
 		       "stack: %s\n", strerror ( rc ) );
 		return rc;
 
 	return 0;
 
- err:
+ err_header:
+	ipv4_stats.in_hdr_errors++;
+ err_other:
 	free_iob ( iobuf );
 	return -EINVAL;
 }

src/net/ipv6.c

 #include <ipxe/if_ether.h>
 #include <ipxe/crc32.h>
 #include <ipxe/fragment.h>
+#include <ipxe/ipstat.h>
 #include <ipxe/ndp.h>
 #include <ipxe/ipv6.h>
 
 /** List of IPv6 miniroutes */
 struct list_head ipv6_miniroutes = LIST_HEAD_INIT ( ipv6_miniroutes );
 
+/** IPv6 statistics */
+static struct ip_statistics ipv6_stats;
+
+/** IPv6 statistics family */
+struct ip_statistics_family
+ipv6_statistics_family __ip_statistics_family ( IP_STATISTICS_IPV6 ) = {
+	.version = 6,
+	.stats = &ipv6_stats,
+};
+
 /**
  * Determine debugging colour for IPv6 debug messages
  *
 	.is_fragment = ipv6_is_fragment,
 	.fragment_offset = ipv6_fragment_offset,
 	.more_fragments = ipv6_more_fragments,
+	.stats = &ipv6_stats,
 };
 
 /**
 	size_t len;
 	int rc;
 
+	/* Update statistics */
+	ipv6_stats.out_requests++;
+
 	/* Fill up the IPv6 header, except source address */
 	len = iob_len ( iobuf );
 	iphdr = iob_push ( iobuf, sizeof ( *iphdr ) );
 	if ( ! netdev ) {
 		DBGC ( ipv6col ( &iphdr->dest ), "IPv6 has no route to %s\n",
 		       inet6_ntoa ( &iphdr->dest ) );
+		ipv6_stats.out_no_routes++;
 		rc = -ENETUNREACH;
 		goto err;
 	}
 	/* Calculate link-layer destination address, if possible */
 	if ( IN6_IS_ADDR_MULTICAST ( next_hop ) ) {
 		/* Multicast address */
+		ipv6_stats.out_mcast_pkts++;
 		if ( ( rc = netdev->ll_protocol->mc_hash ( AF_INET6, next_hop,
 							   ll_dest_buf ) ) !=0){
 			DBGC ( ipv6col ( &iphdr->dest ), "IPv6 could not hash "
 		ll_dest = NULL;
 	}
 
+	/* Update statistics */
+	ipv6_stats.out_transmits++;
+	ipv6_stats.out_octets += iob_len ( iobuf );
+
 	/* Hand off to link layer (via NDP if applicable) */
 	if ( ll_dest ) {
 		if ( ( rc = net_tx ( iobuf, netdev, &ipv6_protocol, ll_dest,
 	int next_header;
 	int rc;
 
+	/* Update statistics */
+	ipv6_stats.in_receives++;
+	ipv6_stats.in_octets += iob_len ( iobuf );
+	if ( flags & LL_BROADCAST ) {
+		ipv6_stats.in_bcast_pkts++;
+	} else if ( flags & LL_MULTICAST ) {
+		ipv6_stats.in_mcast_pkts++;
+	}
+
 	/* Sanity check the IPv6 header */
 	if ( iob_len ( iobuf ) < sizeof ( *iphdr ) ) {
 		DBGC ( ipv6col ( &iphdr->src ), "IPv6 packet too short at %zd "
 		       "bytes (min %zd bytes)\n", iob_len ( iobuf ),
 		       sizeof ( *iphdr ) );
 		rc = -EINVAL_LEN;
-		goto err;
+		goto err_header;
 	}
 	if ( ( iphdr->ver_tc_label & htonl ( IPV6_MASK_VER ) ) !=
 	     htonl ( IPV6_VER ) ) {
 		DBGC ( ipv6col ( &iphdr->src ), "IPv6 version %#08x not "
 		       "supported\n", ntohl ( iphdr->ver_tc_label ) );
 		rc = -ENOTSUP_VER;
-		goto err;
+		goto err_header;
 	}
 
 	/* Truncate packet to specified length */
 	if ( len > iob_len ( iobuf ) ) {
 		DBGC ( ipv6col ( &iphdr->src ), "IPv6 length too long at %zd "
 		       "bytes (packet is %zd bytes)\n", len, iob_len ( iobuf ));
+		ipv6_stats.in_truncated_pkts++;
 		rc = -EINVAL_LEN;
-		goto err;
+		goto err_other;
 	}
 	iob_unput ( iobuf, ( iob_len ( iobuf ) - len - sizeof ( *iphdr ) ) );
 	hdrlen = sizeof ( *iphdr );
 	     ( ! ipv6_has_addr ( netdev, &iphdr->dest ) ) ) {
 		DBGC ( ipv6col ( &iphdr->src ), "IPv6 discarding non-local "
 		       "unicast packet for %s\n", inet6_ntoa ( &iphdr->dest ) );
+		ipv6_stats.in_addr_errors++;
 		rc = -EPIPE;
-		goto err;
+		goto err_other;
 	}
 
 	/* Process any extension headers */
 			       "%zd bytes)\n", this_header,
 			       ( iob_len ( iobuf ) - hdrlen ), extlen );
 			rc = -EINVAL_LEN;
-			goto err;
+			goto err_header;
 		}
 
 		/* Determine size of extension header (if applicable) */
 			       "%zd bytes)\n", this_header,
 			       ( iob_len ( iobuf ) - hdrlen ), extlen );
 			rc = -EINVAL_LEN;
-			goto err;
+			goto err_header;
 		}
 		hdrlen += extlen;
 		next_header = ext->common.next_header;
 			/* Check that all options can be ignored */
 			if ( ( rc = ipv6_check_options ( iphdr, &ext->options,
 							 extlen ) ) != 0 )
-				goto err;
+				goto err_header;
 
 		} else if ( this_header == IPV6_FRAGMENT ) {
 
 	pshdr_csum = ipv6_pshdr_chksum ( iphdr, iob_len ( iobuf ),
 					 next_header, TCPIP_EMPTY_CSUM );
 	if ( ( rc = tcpip_rx ( iobuf, netdev, next_header, &src.st, &dest.st,
-			       pshdr_csum ) ) != 0 ) {
+			       pshdr_csum, &ipv6_stats ) ) != 0 ) {
 		DBGC ( ipv6col ( &src.sin6.sin6_addr ), "IPv6 received packet "
 				"rejected by stack: %s\n", strerror ( rc ) );
 		return rc;
 
 	return 0;
 
- err:
+ err_header:
+	ipv6_stats.in_hdr_errors++;
+ err_other:
 	free_iob ( iobuf );
 	return rc;
 }

src/net/tcpip.c

 #include <byteswap.h>
 #include <ipxe/iobuf.h>
 #include <ipxe/tables.h>
+#include <ipxe/ipstat.h>
 #include <ipxe/tcpip.h>
 
 /** @file
  * @v st_src		Partially-filled source address
  * @v st_dest		Partially-filled destination address
  * @v pshdr_csum	Pseudo-header checksum
+ * @v stats		IP statistics
  * @ret rc		Return status code
  *
  * This function expects a transport-layer segment from the network
  */
 int tcpip_rx ( struct io_buffer *iobuf, struct net_device *netdev,
 	       uint8_t tcpip_proto, struct sockaddr_tcpip *st_src,
-	       struct sockaddr_tcpip *st_dest,
-	       uint16_t pshdr_csum ) {
+	       struct sockaddr_tcpip *st_dest, uint16_t pshdr_csum,
+	       struct ip_statistics *stats ) {
 	struct tcpip_protocol *tcpip;
 
 	/* Hand off packet to the appropriate transport-layer protocol */
 	for_each_table_entry ( tcpip, TCPIP_PROTOCOLS ) {
 		if ( tcpip->tcpip_proto == tcpip_proto ) {
 			DBG ( "TCP/IP received %s packet\n", tcpip->name );
+			stats->in_delivers++;
 			return tcpip->rx ( iobuf, netdev, st_src, st_dest,
 					   pshdr_csum );
 		}
 	}
 
 	DBG ( "Unrecognised TCP/IP protocol %d\n", tcpip_proto );
+	stats->in_unknown_protos++;
 	free_iob ( iobuf );
 	return -EPROTONOSUPPORT;
 }