Added fragment reassembly code

3 files changed, 155 insertions, 15 deletions

diff --git a/src/include/gpxe/ip.h b/src/include/gpxe/ip.h
index a6c590643..4f0f42f85 100644
--- a/src/include/gpxe/ip.h
+++ b/src/include/gpxe/ip.h
@@ -8,18 +8,25 @@
  */
 
 #include <ip.h>
+#include <gpxe/retry.h>
 
 /* IP constants */
 
 #define IP_VER		4
 #define IP_MASK_VER	0xf0
 #define IP_MASK_HLEN 	0x0f
+#define IP_MASK_OFFSET	0x1fff
+#define IP_MASK_DONOTFRAG	0x4000
+#define IP_MASK_MOREFRAGS	0x2000
 #define IP_PSHLEN 	12
 
 /* IP header defaults */
 #define IP_TOS		0
 #define IP_TTL		64
 
+#define IP_FRAG_PKB_SIZE	1500
+#define IP_FRAG_TIMEOUT		50
+
 /* IP4 pseudo header */
 struct ipv4_pseudo_header {
 	struct in_addr src;
@@ -29,6 +36,22 @@ struct ipv4_pseudo_header {
 	uint16_t len;
 };
 
+/* Fragment reassembly buffer */
+struct frag_buffer {
+	/* Identification number */
+	uint16_t ident;
+	/* Source network address */
+	struct in_addr src;
+	/* Destination network address */
+	struct in_addr dest;
+	/* Reassembled packet buffer */
+	struct pk_buff *frag_pkb;
+	/* Reassembly timer */
+	struct retry_timer frag_timer;
+	/* List of fragment reassembly buffers */
+	struct list_head list;
+};
+
 struct pk_buff;
 struct net_device;
 struct net_protocol;
diff --git a/src/net/ipv4.c b/src/net/ipv4.c
index 9669b07db..19e564401 100644
--- a/src/net/ipv4.c
+++ b/src/net/ipv4.c
@@ -47,6 +47,9 @@ struct ipv4_miniroute {
 /** List of IPv4 miniroutes */
 static LIST_HEAD ( miniroutes );
 
+/** List of fragment reassembly buffers */
+static LIST_HEAD ( frag_buffers );
+
 /**
  * Add IPv4 interface
  *
@@ -120,6 +123,110 @@ static void ipv4_dump ( struct iphdr *iphdr __unused ) {
 }
 
 /**
+ * Fragment reassembly counter timeout
+ *
+ * @v timer	Retry timer
+ * @v over	If asserted, the timer is greater than @c MAX_TIMEOUT 
+ */
+void ipv4_frag_expired ( struct retry_timer *timer __unused , int over ) {
+	if ( over ) {
+		DBG ( "Fragment reassembly timeout" );
+		/* Free the fragment buffer */
+	}
+}
+
+/**
+ * Free fragment buffer
+ *
+ * @v fragbug	Fragment buffer
+ */
+void free_fragbuf ( struct frag_buffer *fragbuf ) {
+	if ( fragbuf ) {
+		free_dma ( fragbuf, sizeof ( *fragbuf ) );
+	}
+}
+
+/**
+ * Fragment reassembler
+ *
+ * @v pkb		Packet buffer, fragment of the datagram
+ * @ret frag_pkb	Reassembled packet, or NULL
+ */
+struct pk_buff * ipv4_reassemble ( struct pk_buff * pkb ) {
+	struct iphdr *iphdr = pkb->data;
+	struct frag_buffer *fragbuf;
+	
+	/**
+	 * Check if the fragment belongs to any fragment series
+	 */
+	list_for_each_entry ( fragbuf, &frag_buffers, list ) {
+		if ( fragbuf->ident == iphdr->ident &&
+		     fragbuf->src.s_addr == iphdr->src.s_addr ) {
+			/**
+			 * Check if the packet is the expected fragment
+			 * 
+			 * The offset of the new packet must be equal to the
+			 * length of the data accumulated so far (the length of
+			 * the reassembled packet buffer
+			 */
+			if ( pkb_len ( fragbuf->frag_pkb ) == 
+			      ( iphdr->frags & IP_MASK_OFFSET ) ) {
+				/**
+				 * Append the contents of the fragment to the
+				 * reassembled packet buffer
+				 */
+				pkb_pull ( pkb, sizeof ( *iphdr ) );
+				memcpy ( pkb_put ( fragbuf->frag_pkb,
+							pkb_len ( pkb ) ),
+					 pkb->data, pkb_len ( pkb ) );
+				free_pkb ( pkb );
+
+				/** Check if the fragment series is over */
+				if ( !iphdr->frags & IP_MASK_MOREFRAGS ) {
+					pkb = fragbuf->frag_pkb;
+					free_fragbuf ( fragbuf );
+					return pkb;
+				}
+
+			} else {
+				/* Discard the fragment series */
+				free_fragbuf ( fragbuf );
+				free_pkb ( pkb );
+			}
+			return NULL;
+		}
+	}
+	
+	/** Check if the fragment is the first in the fragment series */
+	if ( iphdr->frags & IP_MASK_MOREFRAGS &&
+			( ( iphdr->frags & IP_MASK_OFFSET ) == 0 ) ) {
+	
+		/** Create a new fragment buffer */
+		fragbuf = ( struct frag_buffer* ) malloc ( sizeof( *fragbuf ) );
+		fragbuf->ident = iphdr->ident;
+		fragbuf->src = iphdr->src;
+
+		/* Set up the reassembly packet buffer */
+		fragbuf->frag_pkb = alloc_pkb ( IP_FRAG_PKB_SIZE );
+		pkb_pull ( pkb, sizeof ( *iphdr ) );
+		memcpy ( pkb_put ( fragbuf->frag_pkb, pkb_len ( pkb ) ),
+			 pkb->data, pkb_len ( pkb ) );
+		free_pkb ( pkb );
+
+		/* Set the reassembly timer */
+		fragbuf->frag_timer.timeout = IP_FRAG_TIMEOUT;
+		fragbuf->frag_timer.expired = ipv4_frag_expired;
+		start_timer ( &fragbuf->frag_timer );
+
+		/* Add the fragment buffer to the list of fragment buffers */
+		list_add ( &fragbuf->list, &frag_buffers );
+	}
+	
+	return NULL;
+}
+
+
+/**
  * Complete the transport-layer checksum
  *
  * @v pkb	Packet buffer
@@ -294,7 +401,9 @@ int ipv4_tx ( struct pk_buff *pkb, struct tcpip_protocol *tcpip,
 	}
 
 	/* Calculate the transport layer checksum */
-	ipv4_tx_csum ( pkb, tcpip );
+	if ( tcpip->csum_offset > 0 ) {
+		ipv4_tx_csum ( pkb, tcpip );
+	}
 
 	/* Calculate header checksum, in network byte order */
 	iphdr->chksum = 0;
@@ -416,6 +525,18 @@ void ipv4_rx ( struct pk_buff *pkb, struct net_device *netdev __unused,
 	if ( ( chksum = ipv4_rx_csum ( pkb, iphdr->protocol ) )	!= 0xffff ) {
 		DBG ( "Bad checksum %x\n", chksum );
 	}
+	/* Fragment reassembly */
+	if ( iphdr->frags & IP_MASK_MOREFRAGS || 
+		( !iphdr->frags & IP_MASK_MOREFRAGS &&
+			iphdr->frags & IP_MASK_OFFSET != 0 ) ) {
+		/* Pass the fragment to the reassembler ipv4_ressable() which
+		 * either returns a fully reassembled packet buffer or NULL.
+		 */
+		pkb = ipv4_reassemble ( pkb );
+		if ( !pkb ) {
+			return;
+		}
+	}
 
 	/* To reduce code size, the following functions are not implemented:
 	 * 1. Check the destination address
diff --git a/src/net/udp.c b/src/net/udp.c
index 4a82e9761..b84d51610 100644
--- a/src/net/udp.c
+++ b/src/net/udp.c
@@ -22,19 +22,14 @@ static inline void copy_sockaddr ( struct sockaddr *source, struct sockaddr *des
 	memcpy ( dest, source, sizeof ( *dest ) );
 }
 
-static inline uint16_t dest_port ( struct sockaddr *sock, uint16_t *dest ) {
+static inline uint16_t * dest_port ( struct sockaddr *sock ) {
 	switch ( sock->sa_family ) {
 	case AF_INET:
-		dest = &sock->sin.sin_port;
-		break;
+		return &sock->sin.sin_port;
 	case AF_INET6:
-		dest = &sock->sin6.sin6_port;
-		break;
-	default:
-		DBG ( "Network family %d not supported\n", sock->sa_family );
-		return -EAFNOSUPPORT;
+		return &sock->sin6.sin6_port;
 	}
-	return 0;
+	return NULL;
 }
 
 /**
@@ -49,7 +44,7 @@ void udp_dump ( struct udp_header *udphdr ) {
 	DBG ( "\tSource Port = %d\n", ntohs ( udphdr->source_port ) );
 	DBG ( "\tDestination Port = %d\n", ntohs ( udphdr->dest_port ) );
 	DBG ( "\tLength = %d\n", ntohs ( udphdr->len ) );
-	DBG ( "\tChecksum = %d\n", ntohs ( udphdr->chksum ) );
+	DBG ( "\tChecksum = %x\n", ntohs ( udphdr->chksum ) );
 	DBG ( "\tChecksum located at %#x\n", &udphdr->chksum );
 }
 
@@ -139,11 +134,12 @@ int udp_send ( struct udp_connection *conn, const void *data, size_t len ) {
 	 * sending it over the network
 	 */
 	udphdr = pkb_push ( conn->tx_pkb, sizeof ( *udphdr ) );
-	if ( (rc = dest_port ( sock, dest ) ) != 0 ) {
-		return rc;
+	if ( (dest = dest_port ( sock ) ) == NULL ) {
+		DBG ( "Network family %d not supported\n", sock->sa_family );
+		return -EAFNOSUPPORT;
 	}
-	udphdr->dest_port = htons ( *dest );
-	udphdr->source_port = htons ( conn->local_port );
+	udphdr->dest_port = *dest;
+	udphdr->source_port = conn->local_port;
 	udphdr->len = htons ( pkb_len ( conn->tx_pkb ) );
 	udphdr->chksum = htons ( calc_chksum ( udphdr, sizeof ( *udphdr ) ) );