+
+static int check_L2_lro_capable(u8 *buffer, struct iphdr **ip,
+ struct tcphdr **tcp, RxD_t *rxdp)
+{
+ int ip_off;
+ u8 l2_type = (u8)((rxdp->Control_1 >> 37) & 0x7), ip_len;
+
+ if (!(rxdp->Control_1 & RXD_FRAME_PROTO_TCP)) {
+ DBG_PRINT(INIT_DBG,"%s: Non-TCP frames not supported for LRO\n",
+ __FUNCTION__);
+ return -1;
+ }
+
+ /* TODO:
+ * By default the VLAN field in the MAC is stripped by the card, if this
+ * feature is turned off in rx_pa_cfg register, then the ip_off field
+ * has to be shifted by a further 2 bytes
+ */
+ switch (l2_type) {
+ case 0: /* DIX type */
+ case 4: /* DIX type with VLAN */
+ ip_off = HEADER_ETHERNET_II_802_3_SIZE;
+ break;
+ /* LLC, SNAP etc are considered non-mergeable */
+ default:
+ return -1;
+ }
+
+ *ip = (struct iphdr *)((u8 *)buffer + ip_off);
+ ip_len = (u8)((*ip)->ihl);
+ ip_len <<= 2;
+ *tcp = (struct tcphdr *)((unsigned long)*ip + ip_len);
+
+ return 0;
+}
+
+static int check_for_socket_match(lro_t *lro, struct iphdr *ip,
+ struct tcphdr *tcp)
+{
+ DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
+ if ((lro->iph->saddr != ip->saddr) || (lro->iph->daddr != ip->daddr) ||
+ (lro->tcph->source != tcp->source) || (lro->tcph->dest != tcp->dest))
+ return -1;
+ return 0;
+}
+
+static inline int get_l4_pyld_length(struct iphdr *ip, struct tcphdr *tcp)
+{
+ return(ntohs(ip->tot_len) - (ip->ihl << 2) - (tcp->doff << 2));
+}
+
+static void initiate_new_session(lro_t *lro, u8 *l2h,
+ struct iphdr *ip, struct tcphdr *tcp, u32 tcp_pyld_len)
+{
+ DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
+ lro->l2h = l2h;
+ lro->iph = ip;
+ lro->tcph = tcp;
+ lro->tcp_next_seq = tcp_pyld_len + ntohl(tcp->seq);
+ lro->tcp_ack = ntohl(tcp->ack_seq);
+ lro->sg_num = 1;
+ lro->total_len = ntohs(ip->tot_len);
+ lro->frags_len = 0;
+ /*
+ * check if we saw TCP timestamp. Other consistency checks have
+ * already been done.
+ */
+ if (tcp->doff == 8) {
+ u32 *ptr;
+ ptr = (u32 *)(tcp+1);
+ lro->saw_ts = 1;
+ lro->cur_tsval = *(ptr+1);
+ lro->cur_tsecr = *(ptr+2);
+ }
+ lro->in_use = 1;
+}
+
+static void update_L3L4_header(nic_t *sp, lro_t *lro)
+{
+ struct iphdr *ip = lro->iph;
+ struct tcphdr *tcp = lro->tcph;
+ u16 nchk;
+ StatInfo_t *statinfo = sp->mac_control.stats_info;
+ DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
+
+ /* Update L3 header */
+ ip->tot_len = htons(lro->total_len);
+ ip->check = 0;
+ nchk = ip_fast_csum((u8 *)lro->iph, ip->ihl);
+ ip->check = nchk;
+
+ /* Update L4 header */
+ tcp->ack_seq = lro->tcp_ack;
+ tcp->window = lro->window;
+
+ /* Update tsecr field if this session has timestamps enabled */
+ if (lro->saw_ts) {
+ u32 *ptr = (u32 *)(tcp + 1);
+ *(ptr+2) = lro->cur_tsecr;
+ }
+
+ /* Update counters required for calculation of
+ * average no. of packets aggregated.
+ */
+ statinfo->sw_stat.sum_avg_pkts_aggregated += lro->sg_num;
+ statinfo->sw_stat.num_aggregations++;
+}
+
+static void aggregate_new_rx(lro_t *lro, struct iphdr *ip,
+ struct tcphdr *tcp, u32 l4_pyld)
+{
+ DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
+ lro->total_len += l4_pyld;
+ lro->frags_len += l4_pyld;
+ lro->tcp_next_seq += l4_pyld;
+ lro->sg_num++;
+
+ /* Update ack seq no. and window ad(from this pkt) in LRO object */
+ lro->tcp_ack = tcp->ack_seq;
+ lro->window = tcp->window;
+
+ if (lro->saw_ts) {
+ u32 *ptr;
+ /* Update tsecr and tsval from this packet */
+ ptr = (u32 *) (tcp + 1);
+ lro->cur_tsval = *(ptr + 1);
+ lro->cur_tsecr = *(ptr + 2);
+ }
+}
+
+static int verify_l3_l4_lro_capable(lro_t *l_lro, struct iphdr *ip,
+ struct tcphdr *tcp, u32 tcp_pyld_len)
+{
+ u8 *ptr;
+
+ DBG_PRINT(INFO_DBG,"%s: Been here...\n", __FUNCTION__);
+
+ if (!tcp_pyld_len) {
+ /* Runt frame or a pure ack */
+ return -1;
+ }
+
+ if (ip->ihl != 5) /* IP has options */
+ return -1;
+
+ if (tcp->urg || tcp->psh || tcp->rst || tcp->syn || tcp->fin ||
+ !tcp->ack) {
+ /*
+ * Currently recognize only the ack control word and
+ * any other control field being set would result in
+ * flushing the LRO session
+ */
+ return -1;
+ }
+
+ /*
+ * Allow only one TCP timestamp option. Don't aggregate if
+ * any other options are detected.
+ */
+ if (tcp->doff != 5 && tcp->doff != 8)
+ return -1;
+
+ if (tcp->doff == 8) {
+ ptr = (u8 *)(tcp + 1);
+ while (*ptr == TCPOPT_NOP)
+ ptr++;
+ if (*ptr != TCPOPT_TIMESTAMP || *(ptr+1) != TCPOLEN_TIMESTAMP)
+ return -1;
+
+ /* Ensure timestamp value increases monotonically */
+ if (l_lro)
+ if (l_lro->cur_tsval > *((u32 *)(ptr+2)))
+ return -1;
+
+ /* timestamp echo reply should be non-zero */
+ if (*((u32 *)(ptr+6)) == 0)
+ return -1;
+ }
+
+ return 0;
+}
+
+static int
+s2io_club_tcp_session(u8 *buffer, u8 **tcp, u32 *tcp_len, lro_t **lro,
+ RxD_t *rxdp, nic_t *sp)
+{
+ struct iphdr *ip;
+ struct tcphdr *tcph;
+ int ret = 0, i;
+
+ if (!(ret = check_L2_lro_capable(buffer, &ip, (struct tcphdr **)tcp,
+ rxdp))) {
+ DBG_PRINT(INFO_DBG,"IP Saddr: %x Daddr: %x\n",
+ ip->saddr, ip->daddr);
+ } else {
+ return ret;
+ }
+
+ tcph = (struct tcphdr *)*tcp;
+ *tcp_len = get_l4_pyld_length(ip, tcph);
+ for (i=0; i<MAX_LRO_SESSIONS; i++) {
+ lro_t *l_lro = &sp->lro0_n[i];
+ if (l_lro->in_use) {
+ if (check_for_socket_match(l_lro, ip, tcph))
+ continue;
+ /* Sock pair matched */
+ *lro = l_lro;
+
+ if ((*lro)->tcp_next_seq != ntohl(tcph->seq)) {
+ DBG_PRINT(INFO_DBG, "%s:Out of order. expected "
+ "0x%x, actual 0x%x\n", __FUNCTION__,
+ (*lro)->tcp_next_seq,
+ ntohl(tcph->seq));
+
+ sp->mac_control.stats_info->
+ sw_stat.outof_sequence_pkts++;
+ ret = 2;
+ break;
+ }
+
+ if (!verify_l3_l4_lro_capable(l_lro, ip, tcph,*tcp_len))
+ ret = 1; /* Aggregate */
+ else
+ ret = 2; /* Flush both */
+ break;
+ }
+ }
+
+ if (ret == 0) {
+ /* Before searching for available LRO objects,
+ * check if the pkt is L3/L4 aggregatable. If not
+ * don't create new LRO session. Just send this
+ * packet up.
+ */
+ if (verify_l3_l4_lro_capable(NULL, ip, tcph, *tcp_len)) {
+ return 5;
+ }
+
+ for (i=0; i<MAX_LRO_SESSIONS; i++) {
+ lro_t *l_lro = &sp->lro0_n[i];
+ if (!(l_lro->in_use)) {
+ *lro = l_lro;
+ ret = 3; /* Begin anew */
+ break;
+ }
+ }
+ }
+
+ if (ret == 0) { /* sessions exceeded */
+ DBG_PRINT(INFO_DBG,"%s:All LRO sessions already in use\n",
+ __FUNCTION__);
+ *lro = NULL;
+ return ret;
+ }
+
+ switch (ret) {
+ case 3:
+ initiate_new_session(*lro, buffer, ip, tcph, *tcp_len);
+ break;
+ case 2:
+ update_L3L4_header(sp, *lro);
+ break;
+ case 1:
+ aggregate_new_rx(*lro, ip, tcph, *tcp_len);
+ if ((*lro)->sg_num == sp->lro_max_aggr_per_sess) {
+ update_L3L4_header(sp, *lro);
+ ret = 4; /* Flush the LRO */
+ }
+ break;
+ default:
+ DBG_PRINT(ERR_DBG,"%s:Dont know, can't say!!\n",
+ __FUNCTION__);
+ break;
+ }
+
+ return ret;
+}
+
+static void clear_lro_session(lro_t *lro)
+{
+ static u16 lro_struct_size = sizeof(lro_t);
+
+ memset(lro, 0, lro_struct_size);
+}
+
+static void queue_rx_frame(struct sk_buff *skb)
+{
+ struct net_device *dev = skb->dev;
+
+ skb->protocol = eth_type_trans(skb, dev);
+#ifdef CONFIG_S2IO_NAPI
+ netif_receive_skb(skb);
+#else
+ netif_rx(skb);
+#endif
+}
+
+static void lro_append_pkt(nic_t *sp, lro_t *lro, struct sk_buff *skb,
+ u32 tcp_len)
+{
+ struct sk_buff *tmp, *first = lro->parent;
+
+ first->len += tcp_len;
+ first->data_len = lro->frags_len;
+ skb_pull(skb, (skb->len - tcp_len));
+ if ((tmp = skb_shinfo(first)->frag_list)) {
+ while (tmp->next)
+ tmp = tmp->next;
+ tmp->next = skb;
+ }
+ else
+ skb_shinfo(first)->frag_list = skb;
+ sp->mac_control.stats_info->sw_stat.clubbed_frms_cnt++;
+ return;
+}