kconfig: use pkg-config for ncurses detection
[akaros.git] / kern / src / net / ip.c
index b2ade7d..afb47f9 100644 (file)
-#include <ros/common.h>
+/* Copyright © 1994-1999 Lucent Technologies Inc.  All rights reserved.
+ * Portions Copyright © 1997-1999 Vita Nuova Limited
+ * Portions Copyright © 2000-2007 Vita Nuova Holdings Limited
+ *                                (www.vitanuova.com)
+ * Revisions Copyright © 2000-2007 Lucent Technologies Inc. and others
+ *
+ * Modified for the Akaros operating system:
+ * Copyright (c) 2013-2014 The Regents of the University of California
+ * Copyright (c) 2013-2015 Google Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE. */
+
+#include <slab.h>
+#include <kmalloc.h>
+#include <kref.h>
+#include <string.h>
+#include <stdio.h>
 #include <assert.h>
-#include <socket.h>
-#include <bits/netinet.h>
-#include <net.h>
+#include <error.h>
+#include <cpio.h>
+#include <pmap.h>
+#include <smp.h>
 #include <net/ip.h>
-#include <net/udp.h>
-#include <net/tcp_impl.h>
-#include <ros/errno.h>
-#include <net/nic_common.h>
-
-/* statically configured next gateway */
-const uint8_t GTWAY[6] = {0xda, 0x76, 0xe7, 0x4c, 0xca, 0x7e};
-
-/* TODO: ip id unique for all ip packets? or is it unique for a flow? */
-// can do atomic increment at a minimum
-static uint16_t ip_id = 0;
-
-/* TODO: build arp table, and look up */
-int eth_send(struct pbuf *p, struct in_addr *dest) {
-       uint32_t bytes_sent; 
-       printk("size of pbuf_header movement %d\n", sizeof(struct ethernet_hdr));
-       if (pbuf_header(p, sizeof(struct ethernet_hdr)) != 0){
-               warn("eth_send buffer ran out");
-               /* unsuccessful, needs to allocate */   
-               return -ENOBUFS;
-       }
-
-       struct ethernet_hdr *ethhdr = (struct ethernet_hdr *)p->payload; 
-       // TODO: for now just forward to gateway
-       memcpy(ethhdr->dst_mac, GTWAY, 6);
-       memcpy(ethhdr->src_mac, device_mac, 6);
-       ethhdr->eth_type = htons(IP_ETH_TYPE);
-       /* The reason for not sending to send_nic for each pbuf in the chain
-        * is so that we can send from multi-buffer later.
+
+typedef struct IP IP;
+typedef struct Fragment4 Fragment4;
+typedef struct Fragment6 Fragment6;
+typedef struct Ipfrag Ipfrag;
+
+enum {
+       IP4HDR = 20,            /* sizeof(Ip4hdr) */
+       IP6HDR = 40,            /* sizeof(Ip6hdr) */
+       IP_HLEN4 = 0x05,        /* Header length in words */
+       IP_DF = 0x4000,         /* Don't fragment */
+       IP_MF = 0x2000,         /* More fragments */
+       IP6FHDR = 8,            /* sizeof(Fraghdr6) */
+       IP_MAX = 64 * 1024,     /* Maximum Internet packet size */
+};
+
+#define BLKIPVER(xp)   (((struct Ip4hdr*)((xp)->rp))->vihl&0xF0)
+#define NEXT_ID(x) (__sync_add_and_fetch(&(x), 1))
+
+/* MIB II counters */
+enum {
+       Forwarding,
+       DefaultTTL,
+       InReceives,
+       InHdrErrors,
+       InAddrErrors,
+       ForwDatagrams,
+       InUnknownProtos,
+       InDiscards,
+       InDelivers,
+       OutRequests,
+       OutDiscards,
+       OutNoRoutes,
+       ReasmTimeout,
+       ReasmReqds,
+       ReasmOKs,
+       ReasmFails,
+       FragOKs,
+       FragFails,
+       FragCreates,
+
+       Nstats,
+};
+
+struct fragment4 {
+       struct block *blist;
+       struct fragment4 *next;
+       uint32_t src;
+       uint32_t dst;
+       uint16_t id;
+       uint64_t age;
+};
+
+struct fragment6 {
+       struct block *blist;
+       struct fragment6 *next;
+       uint8_t src[IPaddrlen];
+       uint8_t dst[IPaddrlen];
+       unsigned int id;
+       uint64_t age;
+};
+
+struct Ipfrag {
+       uint16_t foff;
+       uint16_t flen;
+};
+
+/* an instance of IP */
+struct IP {
+       uint32_t stats[Nstats];
+
+       qlock_t fraglock4;
+       struct fragment4 *flisthead4;
+       struct fragment4 *fragfree4;
+       int id4;
+
+       qlock_t fraglock6;
+       struct fragment6 *flisthead6;
+       struct fragment6 *fragfree6;
+       int id6;
+
+       int iprouting;          /* true if we route like a gateway */
+};
+
+static char *statnames[] = {
+       [Forwarding] "Forwarding",
+       [DefaultTTL] "DefaultTTL",
+       [InReceives] "InReceives",
+       [InHdrErrors] "InHdrErrors",
+       [InAddrErrors] "InAddrErrors",
+       [ForwDatagrams] "ForwDatagrams",
+       [InUnknownProtos] "InUnknownProtos",
+       [InDiscards] "InDiscards",
+       [InDelivers] "InDelivers",
+       [OutRequests] "OutRequests",
+       [OutDiscards] "OutDiscards",
+       [OutNoRoutes] "OutNoRoutes",
+       [ReasmTimeout] "ReasmTimeout",
+       [ReasmReqds] "ReasmReqds",
+       [ReasmOKs] "ReasmOKs",
+       [ReasmFails] "ReasmFails",
+       [FragOKs] "FragOKs",
+       [FragFails] "FragFails",
+       [FragCreates] "FragCreates",
+};
+
+#define BLKIP(xp)      ((struct Ip4hdr*)((xp)->rp))
+/*
+ * This sleazy macro relies on the media header size being
+ * larger than sizeof(Ipfrag). ipreassemble checks this is true
+ */
+#define BKFG(xp)       ((struct Ipfrag*)((xp)->base))
+
+uint16_t ipcsum(uint8_t * unused_uint8_p_t);
+struct block *ip4reassemble(struct IP *, int unused_int, struct block *,
+                           struct Ip4hdr *);
+void ipfragfree4(struct IP *, struct fragment4 *);
+struct fragment4 *ipfragallo4(struct IP *);
+
+void ip_init_6(struct Fs *f)
+{
+       struct V6params *v6p;
+
+       v6p = kzmalloc(sizeof(struct V6params), 0);
+
+       v6p->rp.mflag = 0;      // default not managed
+       v6p->rp.oflag = 0;
+       v6p->rp.maxraint = 600000;      // millisecs
+       v6p->rp.minraint = 200000;
+       v6p->rp.linkmtu = 0;    // no mtu sent
+       v6p->rp.reachtime = 0;
+       v6p->rp.rxmitra = 0;
+       v6p->rp.ttl = MAXTTL;
+       v6p->rp.routerlt = 3 * (v6p->rp.maxraint);
+
+       v6p->hp.rxmithost = 1000;       // v6 RETRANS_TIMER
+
+       v6p->cdrouter = -1;
+
+       f->v6p = v6p;
+
+}
+
+void initfrag(struct IP *ip, int size)
+{
+       struct fragment4 *fq4, *eq4;
+       struct fragment6 *fq6, *eq6;
+
+       ip->fragfree4 =
+           (struct fragment4 *)kzmalloc(sizeof(struct fragment4) * size, 0);
+       if (ip->fragfree4 == NULL)
+               panic("initfrag");
+
+       eq4 = &ip->fragfree4[size];
+       for (fq4 = ip->fragfree4; fq4 < eq4; fq4++)
+               fq4->next = fq4 + 1;
+
+       ip->fragfree4[size - 1].next = NULL;
+
+       ip->fragfree6 =
+           (struct fragment6 *)kzmalloc(sizeof(struct fragment6) * size, 0);
+       if (ip->fragfree6 == NULL)
+               panic("initfrag");
+
+       eq6 = &ip->fragfree6[size];
+       for (fq6 = ip->fragfree6; fq6 < eq6; fq6++)
+               fq6->next = fq6 + 1;
+
+       ip->fragfree6[size - 1].next = NULL;
+}
+
+void ip_init(struct Fs *f)
+{
+       struct IP *ip;
+
+       ip = kzmalloc(sizeof(struct IP), 0);
+       qlock_init(&ip->fraglock4);
+       qlock_init(&ip->fraglock6);
+       initfrag(ip, 100);
+       f->ip = ip;
+
+       ip_init_6(f);
+}
+
+void iprouting(struct Fs *f, int on)
+{
+       f->ip->iprouting = on;
+       if (f->ip->iprouting == 0)
+               f->ip->stats[Forwarding] = 2;
+       else
+               f->ip->stats[Forwarding] = 1;
+}
+
+int ipoput4(struct Fs *f, struct block *bp, int gating, int ttl, int tos, struct
+           conv *c)
+{
+       ERRSTACK(1);
+       struct Ipifc *ifc;
+       uint8_t *gate;
+       uint32_t fragoff;
+       struct block *xp, *nb;
+       struct Ip4hdr *eh, *feh;
+       int lid, len, seglen, chunk, dlen, blklen, offset, medialen;
+       struct route *r, *sr;
+       struct IP *ip;
+       int rv = 0;
+
+       ip = f->ip;
+
+       /* Sanity check for our transport protocols. */
+       if (bp->mss)
+               assert(bp->flag & Btso);
+       /* Fill out the ip header */
+       eh = (struct Ip4hdr *)(bp->rp);
+
+       ip->stats[OutRequests]++;
+
+       /* Number of uint8_ts in data and ip header to write */
+       len = blocklen(bp);
+
+       if (gating) {
+               chunk = nhgets(eh->length);
+               if (chunk > len) {
+                       ip->stats[OutDiscards]++;
+                       netlog(f, Logip, "short gated packet\n");
+                       goto free;
+               }
+               if (chunk < len)
+                       len = chunk;
+       }
+       if (len >= IP_MAX) {
+               ip->stats[OutDiscards]++;
+               netlog(f, Logip, "exceeded ip max size %V\n", eh->dst);
+               goto free;
+       }
+
+       r = v4lookup(f, eh->dst, c);
+       if (r == NULL) {
+               ip->stats[OutNoRoutes]++;
+               netlog(f, Logip, "no interface %V\n", eh->dst);
+               rv = -1;
+               goto free;
+       }
+
+       ifc = r->rt.ifc;
+       if (r->rt.type & (Rifc | Runi))
+               gate = eh->dst;
+       else if (r->rt.type & (Rbcast | Rmulti)) {
+               gate = eh->dst;
+               sr = v4lookup(f, eh->src, NULL);
+               if (sr != NULL && (sr->rt.type & Runi))
+                       ifc = sr->rt.ifc;
+       } else
+               gate = r->v4.gate;
+
+       if (!gating)
+               eh->vihl = IP_VER4 | IP_HLEN4;
+       eh->ttl = ttl;
+       if (!gating)
+               eh->tos = tos;
+
+       if (!canrlock(&ifc->rwlock))
+               goto free;
+       if (waserror()) {
+               runlock(&ifc->rwlock);
+               nexterror();
+       }
+       if (ifc->m == NULL)
+               goto raise;
+
+       /* If we dont need to fragment just send it */
+       medialen = ifc->maxtu - ifc->m->hsize;
+       if (bp->flag & Btso || len <= medialen) {
+               if (!gating)
+                       hnputs(eh->id, NEXT_ID(ip->id4));
+               hnputs(eh->length, len);
+               if (!gating) {
+                       eh->frag[0] = 0x40;
+                       eh->frag[1] = 0;
+               }
+               eh->cksum[0] = 0;
+               eh->cksum[1] = 0;
+               hnputs(eh->cksum, ipcsum(&eh->vihl));
+               ifc->m->bwrite(ifc, bp, V4, gate);
+               runlock(&ifc->rwlock);
+               poperror();
+               return 0;
+       }
+
+       if ((eh->frag[0] & (IP_DF >> 8)) && !gating)
+               printd("%V: DF set\n", eh->dst);
+
+       if (eh->frag[0] & (IP_DF >> 8)) {
+               ip->stats[FragFails]++;
+               ip->stats[OutDiscards]++;
+               icmpcantfrag(f, bp, medialen);
+               netlog(f, Logip, "%V: eh->frag[0] & (IP_DF>>8)\n", eh->dst);
+               goto raise;
+       }
+
+       seglen = (medialen - IP4HDR) & ~7;
+       if (seglen < 8) {
+               ip->stats[FragFails]++;
+               ip->stats[OutDiscards]++;
+               netlog(f, Logip, "%V seglen < 8\n", eh->dst);
+               goto raise;
+       }
+
+       /* compute tcp/udp checksum in software before fragmenting */
+       ptclcsum_finalize(bp, 0);
+
+       dlen = len - IP4HDR;
+       xp = bp;
+       if (gating)
+               lid = nhgets(eh->id);
+       else
+               lid = NEXT_ID(ip->id4);
+
+       /* advance through the blist enough to drop IP4HDR size.  this should
+        * usually just be the first block. */
+       offset = IP4HDR;
+       while (xp != NULL && offset && offset >= BLEN(xp)) {
+               offset -= BLEN(xp);
+               xp = xp->next;
+       }
+       xp->rp += offset;
+
+       if (gating)
+               fragoff = nhgets(eh->frag) << 3;
+       else
+               fragoff = 0;
+       dlen += fragoff;
+       for (; fragoff < dlen; fragoff += seglen) {
+               nb = blist_clone(xp, IP4HDR, seglen, fragoff);
+               feh = (struct Ip4hdr *)(nb->rp);
+
+               memmove(nb->wp, eh, IP4HDR);
+               nb->wp += IP4HDR;
+
+               if ((fragoff + seglen) >= dlen) {
+                       seglen = dlen - fragoff;
+                       hnputs(feh->frag, fragoff >> 3);
+               } else
+                       hnputs(feh->frag, (fragoff >> 3) | IP_MF);
+
+               hnputs(feh->length, seglen + IP4HDR);
+               hnputs(feh->id, lid);
+
+               feh->cksum[0] = 0;
+               feh->cksum[1] = 0;
+               hnputs(feh->cksum, ipcsum(&feh->vihl));
+               ifc->m->bwrite(ifc, nb, V4, gate);
+               ip->stats[FragCreates]++;
+       }
+       ip->stats[FragOKs]++;
+raise:
+       runlock(&ifc->rwlock);
+       poperror();
+free:
+       freeblist(bp);
+       return rv;
+}
+
+void ipiput4(struct Fs *f, struct Ipifc *ifc, struct block *bp)
+{
+       int hl;
+       int hop, tos, proto, olen;
+       struct Ip4hdr *h;
+       struct Proto *p;
+       uint16_t frag;
+       int notforme;
+       uint8_t *dp, v6dst[IPaddrlen];
+       struct IP *ip;
+       struct route *r;
+
+       bp = pullupblock(bp, 1);
+       if (BLKIPVER(bp) != IP_VER4) {
+               ipiput6(f, ifc, bp);
+               return;
+       }
+
+       ip = f->ip;
+       ip->stats[InReceives]++;
+
+       /*
+        *  Ensure we have all the header info in the first
+        *  block.  Make life easier for other protocols by
+        *  collecting up to the first 64 bytes in the first block.
         */
-       if (send_pbuf){
-               bytes_sent = send_pbuf(p);
-               return bytes_sent;
+       if (BLEN(bp) < 64) {
+               hl = blocklen(bp);
+               if (hl < IP4HDR)
+                       hl = IP4HDR;
+               if (hl > 64)
+                       hl = 64;
+               bp = pullupblock(bp, hl);
+               if (bp == NULL)
+                       return;
+       }
+
+       h = (struct Ip4hdr *)(bp->rp);
+
+       /* dump anything that whose header doesn't checksum */
+       if ((bp->flag & Bipck) == 0 && ipcsum(&h->vihl)) {
+               ip->stats[InHdrErrors]++;
+               netlog(f, Logip, "ip: checksum error %V\n", h->src);
+               freeblist(bp);
+               return;
+       }
+       v4tov6(v6dst, h->dst);
+       notforme = ipforme(f, v6dst) == 0;
+
+       /* Check header length and version */
+       if ((h->vihl & 0x0F) != IP_HLEN4) {
+               hl = (h->vihl & 0xF) << 2;
+               if (hl < (IP_HLEN4 << 2)) {
+                       ip->stats[InHdrErrors]++;
+                       netlog(f, Logip, "ip: %V bad hivl 0x%x\n", h->src,
+                              h->vihl);
+                       freeblist(bp);
+                       return;
+               }
+               /* If this is not routed strip off the options */
+               if (notforme == 0) {
+                       olen = nhgets(h->length);
+                       dp = bp->rp + (hl - (IP_HLEN4 << 2));
+                       memmove(dp, h, IP_HLEN4 << 2);
+                       bp->rp = dp;
+                       h = (struct Ip4hdr *)(bp->rp);
+                       h->vihl = (IP_VER4 | IP_HLEN4);
+                       hnputs(h->length, olen - hl + (IP_HLEN4 << 2));
+               }
+       }
+
+       /* route */
+       if (notforme) {
+               struct conv conv;
+
+               if (!ip->iprouting) {
+                       freeb(bp);
+                       return;
+               }
+
+               /* don't forward to source's network */
+               conv.r = NULL;
+               r = v4lookup(f, h->dst, &conv);
+               if (r == NULL || r->rt.ifc == ifc) {
+                       ip->stats[OutDiscards]++;
+                       freeblist(bp);
+                       return;
+               }
+
+               /* don't forward if packet has timed out */
+               hop = h->ttl;
+               if (hop < 1) {
+                       ip->stats[InHdrErrors]++;
+                       icmpttlexceeded(f, ifc->lifc->local, bp);
+                       freeblist(bp);
+                       return;
+               }
+
+               /* reassemble if the interface expects it */
+               if (r->rt.ifc == NULL)
+                       panic("NULL route rfc");
+               if (r->rt.ifc->reassemble) {
+                       frag = nhgets(h->frag);
+                       if (frag) {
+                               h->tos = 0;
+                               if (frag & IP_MF)
+                                       h->tos = 1;
+                               bp = ip4reassemble(ip, frag, bp, h);
+                               if (bp == NULL)
+                                       return;
+                               h = (struct Ip4hdr *)(bp->rp);
+                       }
+               }
+
+               ip->stats[ForwDatagrams]++;
+               tos = h->tos;
+               hop = h->ttl;
+               ipoput4(f, bp, 1, hop - 1, tos, &conv);
+               return;
        }
-       else {
-               warn("no pbuf send function \n");
-               return -1;
+
+       frag = nhgets(h->frag);
+       if (frag && frag != IP_DF) {
+               h->tos = 0;
+               if (frag & IP_MF)
+                       h->tos = 1;
+               bp = ip4reassemble(ip, frag, bp, h);
+               if (bp == NULL)
+                       return;
+               h = (struct Ip4hdr *)(bp->rp);
        }
-       /* is the address local , if no, search for MAC of the gateway and dest to gateway */
-       /* if address is local, use arp etc */
 
+       /* don't let any frag info go up the stack */
+       h->frag[0] = 0;
+       h->frag[1] = 0;
+
+       proto = h->proto;
+       p = Fsrcvpcol(f, proto);
+       if (p != NULL && p->rcv != NULL) {
+               ip->stats[InDelivers]++;
+               (*p->rcv) (p, ifc, bp);
+               return;
+       }
+       ip->stats[InDiscards]++;
+       ip->stats[InUnknownProtos]++;
+       freeblist(bp);
 }
 
-/* while it would be nice to write a generic send_pbuf it is impossible to do so in
- * efficiently.
- */
-/* Assume no ip options */
-int ip_output(struct pbuf *p, struct in_addr *src, struct in_addr *dest, 
-                                                       uint8_t ttl, uint8_t tos, uint8_t proto) {
-       struct pbuf *q;
-       struct ip_hdr *iphdr;
-       printk ("ip output reached\n");
-       /* TODO: Check for IP_HDRINCL */
-       if (dest->s_addr == IP_HDRINCL) {
-               /*send right away since */
-               warn("header included in the ip packets");
-               return -1;
-       }
-       if (pbuf_header(p, IP_HDR_SZ)) {
-               warn("buffer ran out");
-               /* unsuccessful, needs to allocate */   
-               return -ENOBUFS;
-       }
-       iphdr = (struct ip_hdr *) p->payload;
-
-       /* successful */
-       iphdr->version = IPPROTO_IPV4;
-       /* assume no IP options */
-       iphdr->hdr_len = IP_HDR_SZ >> 2;
-       if (tos != 0) {
-               iphdr->tos = htons(tos);
-       }
-       else {
-               iphdr->tos = 0;
-       }
-       iphdr->packet_len = htons(p->tot_len);
-       // TODO: NET_LOCK
-       iphdr->id = htons (ip_id); // 1
-       ip_id++;
-       iphdr->flags_frags = htons(0); // 4000  may fragment
-       iphdr->protocol = proto;
-       iphdr->ttl = ttl; //DEFAULT_TTL;
-       /* Eventually if we support more than one device this may change */
-       printk("src ip %x, dest ip %x \n", src->s_addr, dest->s_addr);
-       iphdr->src_addr = htonl(src->s_addr);
-       iphdr->dst_addr = (dest->s_addr);
-       /* force hardware checksum
-        * TODO: provide option to do both hardware/software checksum
+int ipstats(struct Fs *f, char *buf, int len)
+{
+       struct IP *ip;
+       char *p, *e;
+       int i;
+
+       ip = f->ip;
+       ip->stats[DefaultTTL] = MAXTTL;
+
+       p = buf;
+       e = p + len;
+       for (i = 0; i < Nstats; i++)
+               p = seprintf(p, e, "%s: %u\n", statnames[i], ip->stats[i]);
+       return p - buf;
+}
+
+struct block *ip4reassemble(struct IP *ip, int offset, struct block *bp,
+                                                       struct Ip4hdr *ih)
+{
+       int fend;
+       uint16_t id;
+       struct fragment4 *f, *fnext;
+       uint32_t src, dst;
+       struct block *bl, **l, *last, *prev;
+       int ovlap, len, fragsize, pktposn;
+
+       src = nhgetl(ih->src);
+       dst = nhgetl(ih->dst);
+       id = nhgets(ih->id);
+
+       /*
+        *  block lists are too hard, pullupblock into a single block
         */
-       /* Since the IP header is set already, we can compute the checksum. */
-       /* TODO: Use the card to calculate the checksum */
-       iphdr->checksum = 0;
-       iphdr->checksum = ip_checksum(iphdr); //7ab6
-       if (p->tot_len > DEFAULT_MTU) /*MAX MTU? header included */
-               return -1;//ip_frag(p, dest);
-       else
-               return eth_send(p, dest);
+       if (bp->next) {
+               bp = pullupblock(bp, blocklen(bp));
+               ih = (struct Ip4hdr *)(bp->rp);
+       }
+
+       qlock(&ip->fraglock4);
+
+       /*
+        *  find a reassembly queue for this fragment
+        */
+       for (f = ip->flisthead4; f; f = fnext) {
+               fnext = f->next;/* because ipfragfree4 changes the list */
+               if (f->src == src && f->dst == dst && f->id == id)
+                       break;
+               if (f->age < NOW) {
+                       ip->stats[ReasmTimeout]++;
+                       ipfragfree4(ip, f);
+               }
+       }
+
+       /*
+        *  if this isn't a fragmented packet, accept it
+        *  and get rid of any fragments that might go
+        *  with it.
+        */
+       if (!ih->tos && (offset & ~(IP_MF | IP_DF)) == 0) {
+               if (f != NULL) {
+                       ipfragfree4(ip, f);
+                       ip->stats[ReasmFails]++;
+               }
+               qunlock(&ip->fraglock4);
+               return bp;
+       }
+
+       if (bp->base + sizeof(struct Ipfrag) >= bp->rp) {
+               bp = padblock(bp, sizeof(struct Ipfrag));
+               bp->rp += sizeof(struct Ipfrag);
+       }
+
+       BKFG(bp)->foff = offset << 3;
+       BKFG(bp)->flen = nhgets(ih->length) - IP4HDR;
+
+       /* First fragment allocates a reassembly queue */
+       if (f == NULL) {
+               f = ipfragallo4(ip);
+               f->id = id;
+               f->src = src;
+               f->dst = dst;
+
+               f->blist = bp;
+
+               qunlock(&ip->fraglock4);
+               ip->stats[ReasmReqds]++;
+               return NULL;
+       }
+
+       /*
+        *  find the new fragment's position in the queue
+        */
+       prev = NULL;
+       l = &f->blist;
+       bl = f->blist;
+       while (bl != NULL && BKFG(bp)->foff > BKFG(bl)->foff) {
+               prev = bl;
+               l = &bl->next;
+               bl = bl->next;
+       }
+
+       /* Check overlap of a previous fragment - trim away as necessary */
+       if (prev) {
+               ovlap = BKFG(prev)->foff + BKFG(prev)->flen - BKFG(bp)->foff;
+               if (ovlap > 0) {
+                       if (ovlap >= BKFG(bp)->flen) {
+                               freeblist(bp);
+                               qunlock(&ip->fraglock4);
+                               return NULL;
+                       }
+                       BKFG(prev)->flen -= ovlap;
+               }
+       }
+
+       /* Link onto assembly queue */
+       bp->next = *l;
+       *l = bp;
+
+       /* Check to see if succeeding segments overlap */
+       if (bp->next) {
+               l = &bp->next;
+               fend = BKFG(bp)->foff + BKFG(bp)->flen;
+               /* Take completely covered segments out */
+               while (*l) {
+                       ovlap = fend - BKFG(*l)->foff;
+                       if (ovlap <= 0)
+                               break;
+                       if (ovlap < BKFG(*l)->flen) {
+                               BKFG(*l)->flen -= ovlap;
+                               BKFG(*l)->foff += ovlap;
+                               /* move up ih hdrs */
+                               memmove((*l)->rp + ovlap, (*l)->rp, IP4HDR);
+                               (*l)->rp += ovlap;
+                               break;
+                       }
+                       last = (*l)->next;
+                       (*l)->next = NULL;
+                       freeblist(*l);
+                       *l = last;
+               }
+       }
+
+       /*
+        *  look for a complete packet.  if we get to a fragment
+        *  without IP_MF set, we're done.
+        */
+       pktposn = 0;
+       for (bl = f->blist; bl; bl = bl->next) {
+               if (BKFG(bl)->foff != pktposn)
+                       break;
+               if ((BLKIP(bl)->frag[0] & (IP_MF >> 8)) == 0) {
+                       bl = f->blist;
+                       len = nhgets(BLKIP(bl)->length);
+                       bl->wp = bl->rp + len;
+
+                       /* Pullup all the fragment headers and
+                        * return a complete packet
+                        */
+                       for (bl = bl->next; bl; bl = bl->next) {
+                               fragsize = BKFG(bl)->flen;
+                               len += fragsize;
+                               bl->rp += IP4HDR;
+                               bl->wp = bl->rp + fragsize;
+                       }
+
+                       bl = f->blist;
+                       f->blist = NULL;
+                       ipfragfree4(ip, f);
+                       ih = BLKIP(bl);
+                       hnputs(ih->length, len);
+                       qunlock(&ip->fraglock4);
+                       ip->stats[ReasmOKs]++;
+                       return bl;
+               }
+               pktposn += BKFG(bl)->flen;
+       }
+       qunlock(&ip->fraglock4);
+       return NULL;
+}
+
+/*
+ * ipfragfree4 - Free a list of fragments - assume hold fraglock4
+ */
+void ipfragfree4(struct IP *ip, struct fragment4 *frag)
+{
+       struct fragment4 *fl, **l;
+
+       if (frag->blist)
+               freeblist(frag->blist);
+
+       frag->src = 0;
+       frag->id = 0;
+       frag->blist = NULL;
+
+       l = &ip->flisthead4;
+       for (fl = *l; fl; fl = fl->next) {
+               if (fl == frag) {
+                       *l = frag->next;
+                       break;
+               }
+               l = &fl->next;
+       }
+
+       frag->next = ip->fragfree4;
+       ip->fragfree4 = frag;
+
 }
 
-int ip_input(struct pbuf *p) {
-       uint32_t iphdr_hlen, iphdr_len;
-       struct ip_hdr *iphdr = (struct ip_hdr *)p->payload;
-       //printk("start of ip %p \n", p->payload);
-       //print_pbuf(p);
-       /* use that info to build arp table */
-  if (iphdr->version != 4) {
-               warn("ip version not 4!\n");
-    pbuf_free(p);
-               return -1;
-       }
-       iphdr_hlen = iphdr->hdr_len * 4;
-       iphdr_len = ntohs(iphdr->packet_len);
-       // printk("ip input coming from %x of size %d", ntohs(iphdr->dst_addr), iphdr_len);
-  /* header length exceeds first pbuf length, or ip length exceeds total pbuf length? */
-  if ((iphdr_hlen > p->len) || (iphdr_len > p->tot_len)) {
-    if (iphdr_hlen > p->len) {
-        warn("IP header (len 0x%X) does not fit in first pbuf (len %X), IP packet dropped.\n",
-        iphdr_hlen, p->len);
-    }
-    if (iphdr_len > p->tot_len) {
-        warn("IP (len %X) is longer than pbuf (len %X), IP packet dropped.\n",
-        iphdr_len, p->tot_len);
-    }
-    /* free (drop) packet pbufs */
-    pbuf_free(p);
-    return -1;
-  }
-       if (ip_checksum(iphdr) != 0) {
-               warn("checksum failed \n");
-               pbuf_free(p);
-               return -1;
-       }
-
-       /* check if it is destined for me? */
-       /* XXX: IP address for the interface is IP_ANY */
-       if (ntohl(iphdr->dst_addr) != LOCAL_IP_ADDR.s_addr){
-               printk("dest ip in network order%x\n", ntohl(iphdr->dst_addr));
-               printk("dest ip in network order%x\n", LOCAL_IP_ADDR.s_addr);
-               warn("ip mismatch \n");
-               pbuf_free(p);
-               /* TODO:forward packets */
-               // ip_forward(p, iphdr, inp);
-       }
-
-       if ((ntohs(iphdr->flags_frags) & (IP_OFFMASK | IP_MF)) != 0){
-               panic ("ip fragment detected\n");
-               pbuf_free(p);
-       }
-
-       //printk ("loc head %p, loc protocol %p\n", iphdr, &iphdr->protocol);
-       /* currently a noop, compared to the memory wasted, cutting out ipheader is not really saving much */
-       // pbuf_realloc(p, iphdr_len);
-       switch (iphdr->protocol) {
-               case IPPROTO_UDP:
-                       return udp_input(p);
-               case IPPROTO_TCP:
-                       tcp_input(p);
-                       // XXX: error handling for tcp
-                       return 0;
-               default:
-                       printk("IP protocol type %02x\n", iphdr->protocol);
-                       warn("protocol not supported! \n");
-       }
-       return -1;
+/*
+ * ipfragallo4 - allocate a reassembly queue - assume hold fraglock4
+ */
+struct fragment4 *ipfragallo4(struct IP *ip)
+{
+       struct fragment4 *f;
+
+       while (ip->fragfree4 == NULL) {
+               /* free last entry on fraglist */
+               for (f = ip->flisthead4; f->next; f = f->next) ;
+               ipfragfree4(ip, f);
+       }
+       f = ip->fragfree4;
+       ip->fragfree4 = f->next;
+       f->next = ip->flisthead4;
+       ip->flisthead4 = f;
+       f->age = NOW + 30000;
+
+       return f;
 }
 
-void print_ipheader(struct ip_hdr* iph){
+/* coreboot.c among other things needs this
+ * type of checksum.
+ */
+uint16_t ipchecksum(uint8_t *addr, int len)
+{
+       uint16_t sum = 0;
+
+       while (len > 0) {
+               sum += addr[0] << 8 | addr[1];
+               len -= 2;
+               addr += 2;
+       }
+
+       sum = (sum & 0xffff) + (sum >> 16);
+       sum = (sum & 0xffff) + (sum >> 16);
+
+       return (sum ^ 0xffff);
 
-       
 }
 
+/* change this to call ipchecksum later.
+ * but we have to be sure we're not doing something bad
+ * that violates some ip stack assumption (such as
+ * boundaries etc.)
+ */
+uint16_t ipcsum(uint8_t * addr)
+{
+       int len;
+       uint32_t sum;
+
+       sum = 0;
+       len = (addr[0] & 0xf) << 2;
 
+       while (len > 0) {
+               sum += addr[0] << 8 | addr[1];
+               len -= 2;
+               addr += 2;
+       }
+
+       sum = (sum & 0xffff) + (sum >> 16);
+       sum = (sum & 0xffff) + (sum >> 16);
+
+       return (sum ^ 0xffff);
+}