void
an_start(struct ifnet *ifp)
{
struct an_softc *sc = (struct an_softc *)ifp->if_softc;
struct ieee80211com *ic = &sc->sc_ic;
struct ieee80211_node *ni;
struct ieee80211_frame *wh;
struct an_txframe frmhdr;
struct mbuf *m;
u_int16_t len;
int cur, fid;
if (!sc->sc_enabled || sc->sc_invalid) {
DPRINTF(("an_start: noop: enabled %d invalid %d\n",
sc->sc_enabled, sc->sc_invalid));
return;
}
memset(&frmhdr, 0, sizeof(frmhdr));
cur = sc->sc_txnext;
for (;;) {
if (ic->ic_state != IEEE80211_S_RUN) {
DPRINTF(("an_start: not running %d\n", ic->ic_state));
break;
}
m = ifq_deq_begin(&ifp->if_snd);
if (m == NULL) {
DPRINTF2(("an_start: no pending mbuf\n"));
break;
}
if (sc->sc_txd[cur].d_inuse) {
ifq_deq_rollback(&ifp->if_snd, m);
DPRINTF2(("an_start: %x/%d busy\n",
sc->sc_txd[cur].d_fid, cur));
ifq_set_oactive(&ifp->if_snd);
break;
}
ifq_deq_commit(&ifp->if_snd, m);
ifp->if_opackets++;
#if NBPFILTER > 0
if (ifp->if_bpf)
bpf_mtap(ifp->if_bpf, m, BPF_DIRECTION_OUT);
#endif
if ((m = ieee80211_encap(ifp, m, &ni)) == NULL) {
ifp->if_oerrors++;
continue;
}
if (ni != NULL)
ieee80211_release_node(ic, ni);
#if NBPFILTER > 0
if (ic->ic_rawbpf)
bpf_mtap(ic->ic_rawbpf, m, BPF_DIRECTION_OUT);
#endif
wh = mtod(m, struct ieee80211_frame *);
if (ic->ic_flags & IEEE80211_F_WEPON)
wh->i_fc[1] |= IEEE80211_FC1_WEP;
m_copydata(m, 0, sizeof(struct ieee80211_frame),
(caddr_t)&frmhdr.an_whdr);
an_swap16((u_int16_t *)&frmhdr.an_whdr, sizeof(struct ieee80211_frame)/2);
/* insert payload length in front of llc/snap */
len = htons(m->m_pkthdr.len - sizeof(struct ieee80211_frame));
m_adj(m, sizeof(struct ieee80211_frame) - sizeof(len));
if (mtod(m, u_long) & 0x01)
memcpy(mtod(m, caddr_t), &len, sizeof(len));
else
*mtod(m, u_int16_t *) = len;
/*
* XXX Aironet firmware apparently convert the packet
* with longer than 1500 bytes in length into LLC/SNAP.
* If we have 1500 bytes in ethernet payload, it is
* 1508 bytes including LLC/SNAP and will be inserted
* additional LLC/SNAP header with 1501-1508 in its
* ethertype !!
* So we skip LLC/SNAP header and force firmware to
* convert it to LLC/SNAP again.
*/
m_adj(m, sizeof(struct llc));
frmhdr.an_tx_ctl = AN_TXCTL_80211;
frmhdr.an_tx_payload_len = m->m_pkthdr.len;
frmhdr.an_gaplen = AN_TXGAP_802_11;
if (ic->ic_fixed_rate != -1)
frmhdr.an_tx_rate =
ic->ic_sup_rates[IEEE80211_MODE_11B].rs_rates[
ic->ic_fixed_rate] & IEEE80211_RATE_VAL;
else
frmhdr.an_tx_rate = 0;
if (sizeof(frmhdr) + AN_TXGAP_802_11 + sizeof(len) +
m->m_pkthdr.len > AN_TX_MAX_LEN) {
ifp->if_oerrors++;
m_freem(m);
continue;
}
#if NBPFILTER > 0
if (sc->sc_drvbpf) {
struct mbuf mb;
struct an_tx_radiotap_header *tap = &sc->sc_txtap;
tap->at_rate =
ic->ic_bss->ni_rates.rs_rates[ic->ic_bss->ni_txrate];
tap->at_chan_freq =
ic->ic_bss->ni_chan->ic_freq;
tap->at_chan_flags =
ic->ic_bss->ni_chan->ic_flags;
mb.m_data = (caddr_t)tap;
mb.m_len = sizeof(sc->sc_txtapu);
mb.m_next = m;
mb.m_nextpkt = NULL;
mb.m_type = 0;
mb.m_flags = 0;
bpf_mtap(sc->sc_drvbpf, m, BPF_DIRECTION_OUT);
}
#endif
fid = sc->sc_txd[cur].d_fid;
if (an_write_bap(sc, fid, 0, &frmhdr, sizeof(frmhdr)) != 0) {
ifp->if_oerrors++;
m_freem(m);
continue;
}
/* dummy write to avoid seek. */
an_write_bap(sc, fid, -1, &frmhdr, AN_TXGAP_802_11);
an_mwrite_bap(sc, fid, -1, m, m->m_pkthdr.len);
m_freem(m);
DPRINTF2(("an_start: send %d byte via %x/%d\n",
ntohs(len) + sizeof(struct ieee80211_frame),
fid, cur));
sc->sc_txd[cur].d_inuse = 1;
if (an_cmd(sc, AN_CMD_TX, fid)) {
printf("%s: xmit failed\n", ifp->if_xname);
sc->sc_txd[cur].d_inuse = 0;
continue;
}
sc->sc_tx_timer = 5;
ifp->if_timer = 1;
AN_INC(cur, AN_TX_RING_CNT);
sc->sc_txnext = cur;
}
}
/*
* Help break down an mbuf chain by setting the first siz bytes contiguous
* pointed to by returned val.
* This is used by the macros nfsm_dissect for tough
* cases. (The macros use the vars. dpos and dpos2)
*/
void *
nfsm_disct(struct mbuf **mdp, caddr_t *dposp, int siz, int left, int how)
{
struct mbuf *mp, *mp2;
int siz2, xfer;
caddr_t ptr, npos = NULL;
void *ret;
mp = *mdp;
while (left == 0) {
*mdp = mp = mp->m_next;
if (mp == NULL)
return (NULL);
left = mp->m_len;
*dposp = mtod(mp, caddr_t);
}
if (left >= siz) {
ret = *dposp;
*dposp += siz;
} else if (mp->m_next == NULL) {
return (NULL);
} else if (siz > MHLEN) {
panic("nfs S too big");
} else {
MGET(mp2, how, MT_DATA);
if (mp2 == NULL)
return (NULL);
mp2->m_len = siz;
mp2->m_next = mp->m_next;
mp->m_next = mp2;
mp->m_len -= left;
mp = mp2;
ptr = mtod(mp, caddr_t);
ret = ptr;
bcopy(*dposp, ptr, left); /* Copy what was left */
siz2 = siz-left;
ptr += left;
mp2 = mp->m_next;
npos = mtod(mp2, caddr_t);
/* Loop around copying up the siz2 bytes */
while (siz2 > 0) {
if (mp2 == NULL)
return (NULL);
xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
if (xfer > 0) {
bcopy(mtod(mp2, caddr_t), ptr, xfer);
mp2->m_data += xfer;
mp2->m_len -= xfer;
ptr += xfer;
siz2 -= xfer;
}
if (siz2 > 0) {
mp2 = mp2->m_next;
if (mp2 != NULL)
npos = mtod(mp2, caddr_t);
}
}
*mdp = mp2;
*dposp = mtod(mp2, caddr_t);
if (!nfsm_aligned(*dposp, u_int32_t)) {
bcopy(*dposp, npos, mp2->m_len);
mp2->m_data = npos;
*dposp = npos;
}
}
return (ret);
}
static
void
ip6_input(netmsg_t msg)
{
struct mbuf *m = msg->packet.nm_packet;
struct ip6_hdr *ip6;
int off = sizeof(struct ip6_hdr), nest;
u_int32_t plen;
u_int32_t rtalert = ~0;
int nxt, ours = 0, rh_present = 0;
struct ifnet *deliverifp = NULL;
struct in6_addr odst;
int srcrt = 0;
#ifdef IPSEC
/*
* should the inner packet be considered authentic?
* see comment in ah4_input().
*/
if (m) {
m->m_flags &= ~M_AUTHIPHDR;
m->m_flags &= ~M_AUTHIPDGM;
}
#endif
/*
* make sure we don't have onion peering information into m_aux.
*/
ip6_delaux(m);
/*
* mbuf statistics
*/
if (m->m_flags & M_EXT) {
if (m->m_next)
ip6stat.ip6s_mext2m++;
else
ip6stat.ip6s_mext1++;
} else {
#define M2MMAX NELEM(ip6stat.ip6s_m2m)
if (m->m_next) {
if (m->m_flags & M_LOOP) {
ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
else
ip6stat.ip6s_m2m[0]++;
} else
ip6stat.ip6s_m1++;
#undef M2MMAX
}
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
ip6stat.ip6s_total++;
#ifndef PULLDOWN_TEST
/*
* L2 bridge code and some other code can return mbuf chain
* that does not conform to KAME requirement. too bad.
* XXX: fails to join if interface MTU > MCLBYTES. jumbogram?
*/
if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
struct mbuf *n;
n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
if (n == NULL)
goto bad;
M_MOVE_PKTHDR(n, m);
m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
n->m_len = n->m_pkthdr.len;
m_freem(m);
m = n;
}
IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
#endif
if (m->m_len < sizeof(struct ip6_hdr)) {
struct ifnet *inifp;
inifp = m->m_pkthdr.rcvif;
if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
ip6stat.ip6s_toosmall++;
in6_ifstat_inc(inifp, ifs6_in_hdrerr);
goto bad2;
}
}
ip6 = mtod(m, struct ip6_hdr *);
if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
ip6stat.ip6s_badvers++;
in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
goto bad;
}
/*
* Run through list of hooks for input packets.
*
* NB: Beware of the destination address changing
* (e.g. by NAT rewriting). When this happens,
* tell ip6_forward to do the right thing.
*/
if (pfil_has_hooks(&inet6_pfil_hook)) {
odst = ip6->ip6_dst;
if (pfil_run_hooks(&inet6_pfil_hook, &m,
m->m_pkthdr.rcvif, PFIL_IN)) {
goto bad2;
}
if (m == NULL) /* consumed by filter */
goto bad2;
ip6 = mtod(m, struct ip6_hdr *);
srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
}
static int
udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
struct mbuf *control, struct thread *td)
{
u_int32_t ulen = m->m_pkthdr.len;
u_int32_t plen = sizeof(struct udphdr) + ulen;
struct ip6_hdr *ip6;
struct udphdr *udp6;
struct in6_addr *laddr, *faddr, in6a;
struct sockaddr_in6 *sin6 = NULL;
int cscov_partial = 0;
int scope_ambiguous = 0;
u_short fport;
int error = 0;
uint8_t nxt;
uint16_t cscov = 0;
struct ip6_pktopts *optp, opt;
int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
int flags;
struct sockaddr_in6 tmp;
INP_WLOCK_ASSERT(inp);
INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
if (addr6) {
/* addr6 has been validated in udp6_send(). */
sin6 = (struct sockaddr_in6 *)addr6;
/* protect *sin6 from overwrites */
tmp = *sin6;
sin6 = &tmp;
/*
* Application should provide a proper zone ID or the use of
* default zone IDs should be enabled. Unfortunately, some
* applications do not behave as it should, so we need a
* workaround. Even if an appropriate ID is not determined,
* we'll see if we can determine the outgoing interface. If we
* can, determine the zone ID based on the interface below.
*/
if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
scope_ambiguous = 1;
if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
return (error);
}
nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
IPPROTO_UDP : IPPROTO_UDPLITE;
if (control) {
if ((error = ip6_setpktopts(control, &opt,
inp->in6p_outputopts, td->td_ucred, nxt)) != 0)
goto release;
optp = &opt;
} else
optp = inp->in6p_outputopts;
if (sin6) {
faddr = &sin6->sin6_addr;
/*
* Since we saw no essential reason for calling in_pcbconnect,
* we get rid of such kind of logic, and call in6_selectsrc
* and in6_pcbsetport in order to fill in the local address
* and the local port.
*/
if (sin6->sin6_port == 0) {
error = EADDRNOTAVAIL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
/* how about ::ffff:0.0.0.0 case? */
error = EISCONN;
goto release;
}
fport = sin6->sin6_port; /* allow 0 port */
if (IN6_IS_ADDR_V4MAPPED(faddr)) {
if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
/*
* I believe we should explicitly discard the
* packet when mapped addresses are disabled,
* rather than send the packet as an IPv6 one.
* If we chose the latter approach, the packet
* might be sent out on the wire based on the
* default route, the situation which we'd
* probably want to avoid.
* (20010421 [email protected])
*/
error = EINVAL;
goto release;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
!IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
/*
* when remote addr is an IPv4-mapped address,
* local addr should not be an IPv6 address,
* since you cannot determine how to map IPv6
* source address to IPv4.
*/
error = EINVAL;
goto release;
}
af = AF_INET;
}
if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
error = in6_selectsrc_socket(sin6, optp, inp,
td->td_ucred, scope_ambiguous, &in6a, NULL);
if (error)
goto release;
laddr = &in6a;
} else
laddr = &inp->in6p_laddr; /* XXX */
if (laddr == NULL) {
if (error == 0)
error = EADDRNOTAVAIL;
goto release;
}
if (inp->inp_lport == 0 &&
(error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
/* Undo an address bind that may have occurred. */
inp->in6p_laddr = in6addr_any;
goto release;
}
} else {
if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
error = ENOTCONN;
goto release;
}
if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
/*
* XXX: this case would happen when the
* application sets the V6ONLY flag after
* connecting the foreign address.
* Such applications should be fixed,
* so we bark here.
*/
log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
"option was set for a connected socket\n");
error = EINVAL;
goto release;
} else
af = AF_INET;
}
laddr = &inp->in6p_laddr;
faddr = &inp->in6p_faddr;
fport = inp->inp_fport;
}
if (af == AF_INET)
hlen = sizeof(struct ip);
/*
* Calculate data length and get a mbuf
* for UDP and IP6 headers.
*/
M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
if (m == NULL) {
error = ENOBUFS;
goto release;
}
/*
* Stuff checksum and output datagram.
*/
udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
udp6->uh_dport = fport;
if (nxt == IPPROTO_UDPLITE) {
struct udpcb *up;
up = intoudpcb(inp);
cscov = up->u_txcslen;
if (cscov >= plen)
cscov = 0;
udp6->uh_ulen = htons(cscov);
/*
* For UDP-Lite, checksum coverage length of zero means
* the entire UDPLite packet is covered by the checksum.
*/
cscov_partial = (cscov == 0) ? 0 : 1;
} else if (plen <= 0xffff)
udp6->uh_ulen = htons((u_short)plen);
else
udp6->uh_ulen = 0;
udp6->uh_sum = 0;
switch (af) {
case AF_INET6:
ip6 = mtod(m, struct ip6_hdr *);
ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
ip6->ip6_vfc |= IPV6_VERSION;
ip6->ip6_plen = htons((u_short)plen);
ip6->ip6_nxt = nxt;
ip6->ip6_hlim = in6_selecthlim(inp, NULL);
ip6->ip6_src = *laddr;
ip6->ip6_dst = *faddr;
if (cscov_partial) {
if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
sizeof(struct ip6_hdr), plen, cscov)) == 0)
udp6->uh_sum = 0xffff;
} else {
udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
}
#ifdef RSS
{
uint32_t hash_val, hash_type;
uint8_t pr;
pr = inp->inp_socket->so_proto->pr_protocol;
/*
* Calculate an appropriate RSS hash for UDP and
* UDP Lite.
*
* The called function will take care of figuring out
* whether a 2-tuple or 4-tuple hash is required based
* on the currently configured scheme.
*
* Later later on connected socket values should be
* cached in the inpcb and reused, rather than constantly
* re-calculating it.
*
* UDP Lite is a different protocol number and will
* likely end up being hashed as a 2-tuple until
* RSS / NICs grow UDP Lite protocol awareness.
*/
if (rss_proto_software_hash_v6(faddr, laddr, fport,
inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
m->m_pkthdr.flowid = hash_val;
M_HASHTYPE_SET(m, hash_type);
}
}
#endif
flags = 0;
#ifdef RSS
/*
* Don't override with the inp cached flowid.
*
* Until the whole UDP path is vetted, it may actually
* be incorrect.
*/
flags |= IP_NODEFAULTFLOWID;
#endif
UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
UDPSTAT_INC(udps_opackets);
error = ip6_output(m, optp, &inp->inp_route6, flags,
inp->in6p_moptions, NULL, inp);
break;
case AF_INET:
error = EAFNOSUPPORT;
goto release;
}
goto releaseopt;
release:
m_freem(m);
releaseopt:
if (control) {
ip6_clearpktopts(&opt, -1);
m_freem(control);
}
return (error);
}
/*
* Create a writable copy of the mbuf chain. While doing this
* we compact the chain with a goal of producing a chain with
* at most two mbufs. The second mbuf in this chain is likely
* to be a cluster. The primary purpose of this work is to create
* a writable packet for encryption, compression, etc. The
* secondary goal is to linearize the data so the data can be
* passed to crypto hardware in the most efficient manner possible.
*/
struct mbuf *
m_clone(struct mbuf *m0)
{
struct mbuf *m, *mprev;
struct mbuf *n, *mfirst, *mlast;
int len, off;
KASSERT(m0 != NULL, ("m_clone: null mbuf"));
mprev = NULL;
for (m = m0; m != NULL; m = mprev->m_next) {
/*
* Regular mbufs are ignored unless there's a cluster
* in front of it that we can use to coalesce. We do
* the latter mainly so later clusters can be coalesced
* also w/o having to handle them specially (i.e. convert
* mbuf+cluster -> cluster). This optimization is heavily
* influenced by the assumption that we're running over
* Ethernet where MCLBYTES is large enough that the max
* packet size will permit lots of coalescing into a
* single cluster. This in turn permits efficient
* crypto operations, especially when using hardware.
*/
if ((m->m_flags & M_EXT) == 0) {
if (mprev && (mprev->m_flags & M_EXT) &&
m->m_len <= M_TRAILINGSPACE(mprev)) {
/* XXX: this ignores mbuf types */
memcpy(mtod(mprev, caddr_t) + mprev->m_len,
mtod(m, caddr_t), m->m_len);
mprev->m_len += m->m_len;
mprev->m_next = m->m_next; /* unlink from chain */
m_free(m); /* reclaim mbuf */
newipsecstat.ips_mbcoalesced++;
} else {
mprev = m;
}
continue;
}
/*
* Writable mbufs are left alone (for now). Note
* that for 4.x systems it's not possible to identify
* whether or not mbufs with external buffers are
* writable unless they use clusters.
*/
if (M_EXT_WRITABLE(m)) {
mprev = m;
continue;
}
/*
* Not writable, replace with a copy or coalesce with
* the previous mbuf if possible (since we have to copy
* it anyway, we try to reduce the number of mbufs and
* clusters so that future work is easier).
*/
KASSERT(m->m_flags & M_EXT,
("m_clone: m_flags 0x%x", m->m_flags));
/* NB: we only coalesce into a cluster or larger */
if (mprev != NULL && (mprev->m_flags & M_EXT) &&
m->m_len <= M_TRAILINGSPACE(mprev)) {
/* XXX: this ignores mbuf types */
memcpy(mtod(mprev, caddr_t) + mprev->m_len,
mtod(m, caddr_t), m->m_len);
mprev->m_len += m->m_len;
mprev->m_next = m->m_next; /* unlink from chain */
m_free(m); /* reclaim mbuf */
newipsecstat.ips_clcoalesced++;
continue;
}
/*
* Allocate new space to hold the copy...
*/
/* XXX why can M_PKTHDR be set past the first mbuf? */
if (mprev == NULL && (m->m_flags & M_PKTHDR)) {
/*
* NB: if a packet header is present we must
* allocate the mbuf separately from any cluster
* because M_MOVE_PKTHDR will smash the data
* pointer and drop the M_EXT marker.
*/
MGETHDR(n, MB_DONTWAIT, m->m_type);
if (n == NULL) {
m_freem(m0);
return (NULL);
}
M_MOVE_PKTHDR(n, m);
MCLGET(n, MB_DONTWAIT);
if ((n->m_flags & M_EXT) == 0) {
m_free(n);
m_freem(m0);
return (NULL);
}
} else {
n = m_getcl(MB_DONTWAIT, m->m_type, m->m_flags);
if (n == NULL) {
m_freem(m0);
return (NULL);
}
}
/*
* ... and copy the data. We deal with jumbo mbufs
* (i.e. m_len > MCLBYTES) by splitting them into
* clusters. We could just malloc a buffer and make
* it external but too many device drivers don't know
* how to break up the non-contiguous memory when
* doing DMA.
*/
len = m->m_len;
off = 0;
mfirst = n;
mlast = NULL;
for (;;) {
int cc = min(len, MCLBYTES);
memcpy(mtod(n, caddr_t), mtod(m, caddr_t) + off, cc);
n->m_len = cc;
if (mlast != NULL)
mlast->m_next = n;
mlast = n;
newipsecstat.ips_clcopied++;
len -= cc;
if (len <= 0)
break;
off += cc;
n = m_getcl(MB_DONTWAIT, m->m_type, m->m_flags);
if (n == NULL) {
m_freem(mfirst);
m_freem(m0);
return (NULL);
}
}
n->m_next = m->m_next;
if (mprev == NULL)
m0 = mfirst; /* new head of chain */
else
mprev->m_next = mfirst; /* replace old mbuf */
m_free(m); /* release old mbuf */
mprev = mfirst;
}
return (m0);
}
/*
* Ethernet output routine.
* Encapsulate a packet of type family for the local net.
* Assumes that ifp is actually pointer to ethercom structure.
*/
int
ssh_interceptor_ether_output(struct ifnet *ifp, struct mbuf *m0,
struct sockaddr *dst, struct rtentry *rt0)
{
u_int16_t etype = 0;
int s, error = 0, hdrcmplt = 0;
u_char esrc[6], edst[6];
struct mbuf *m = m0;
struct rtentry *rt;
struct mbuf *mcopy = (struct mbuf *)0;
struct ether_header *eh, ehd;
#ifdef INET
struct arphdr *ah;
#endif /* INET */
#ifdef NETATALK
struct at_ifaddr *aa;
#endif /* NETATALK */
if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
senderr(ENETDOWN);
ifp->if_lastchange = time;
if ((rt = rt0) != NULL) {
if ((rt->rt_flags & RTF_UP) == 0) {
if ((rt0 = rt = rtalloc1(dst, 1)) != NULL) {
rt->rt_refcnt--;
if (rt->rt_ifp != ifp)
return (*rt->rt_ifp->if_output)
(ifp, m0, dst, rt);
} else
senderr(EHOSTUNREACH);
}
if ((rt->rt_flags & RTF_GATEWAY) && dst->sa_family != AF_NS) {
if (rt->rt_gwroute == 0)
goto lookup;
if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
rtfree(rt); rt = rt0;
lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1);
if ((rt = rt->rt_gwroute) == 0)
senderr(EHOSTUNREACH);
/* the "G" test below also prevents rt == rt0 */
if ((rt->rt_flags & RTF_GATEWAY) ||
(rt->rt_ifp != ifp)) {
rt->rt_refcnt--;
rt0->rt_gwroute = 0;
senderr(EHOSTUNREACH);
}
}
}
if (rt->rt_flags & RTF_REJECT)
if (rt->rt_rmx.rmx_expire == 0 ||
time.tv_sec < rt->rt_rmx.rmx_expire)
senderr(rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
}
switch (dst->sa_family) {
#ifdef INET
case AF_INET:
if (m->m_flags & M_BCAST)
bcopy((caddr_t)etherbroadcastaddr, (caddr_t)edst,
sizeof(edst));
else if (m->m_flags & M_MCAST) {
ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr,
(caddr_t)edst)
} else if (!arpresolve(ifp, rt, m, dst, edst))
return (0); /* if not yet resolved */
/* If broadcasting on a simplex interface, loopback a copy */
if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
mcopy = m_copy(m, 0, (int)M_COPYALL);
etype = htons(ETHERTYPE_IP);
break;
case AF_ARP:
ah = mtod(m, struct arphdr *);
if (m->m_flags & M_BCAST)
bcopy((caddr_t)etherbroadcastaddr, (caddr_t)edst,
sizeof(edst));
else
bcopy((caddr_t)ar_tha(ah),
(caddr_t)edst, sizeof(edst));
ah->ar_hrd = htons(ARPHRD_ETHER);
switch(ntohs(ah->ar_op)) {
case ARPOP_REVREQUEST:
case ARPOP_REVREPLY:
etype = htons(ETHERTYPE_REVARP);
break;
case ARPOP_REQUEST:
case ARPOP_REPLY:
default:
etype = htons(ETHERTYPE_ARP);
}
break;
#endif
#ifdef INET6
case AF_INET6:
#ifdef OLDIP6OUTPUT
if (!nd6_resolve(ifp, rt, m, dst, (u_char *)edst))
return(0); /* if not yet resolves */
#else
if (!nd6_storelladdr(ifp, rt, m, dst, (u_char *)edst)){
/* this must be impossible, so we bark */
printf("nd6_storelladdr failed\n");
return(0);
}
#endif /* OLDIP6OUTPUT */
etype = htons(ETHERTYPE_IPV6);
break;
#endif
#ifdef NETATALK
case AF_APPLETALK:
if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst)) {
#ifdef NETATALKDEBUG
printf("aarpresolv failed\n");
#endif /* NETATALKDEBUG */
return (0);
}
/*
* ifaddr is the first thing in at_ifaddr
*/
aa = (struct at_ifaddr *) at_ifawithnet(
(struct sockaddr_at *)dst, ifp);
if (aa == NULL)
goto bad;
/*
* In the phase 2 case, we need to prepend an mbuf for the
* llc header. Since we must preserve the value of m,
* which is passed to us by value, we m_copy() the first
* mbuf, and use it for our llc header.
*/
if (aa->aa_flags & AFA_PHASE2) {
struct llc llc;
M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
llc.llc_control = LLC_UI;
bcopy(at_org_code, llc.llc_snap_org_code,
sizeof(llc.llc_snap_org_code));
llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
bcopy(&llc, mtod(m, caddr_t), sizeof(struct llc));
} else {
etype = htons(ETHERTYPE_ATALK);
}
break;
#endif /* NETATALK */
#ifdef NS
case AF_NS:
etype = htons(ETHERTYPE_NS);
bcopy((caddr_t)&(((struct sockaddr_ns *)dst)->sns_addr.x_host),
(caddr_t)edst, sizeof (edst));
if (!bcmp((caddr_t)edst, (caddr_t)&ns_thishost, sizeof(edst)))
return (looutput(ifp, m, dst, rt));
/* If broadcasting on a simplex interface, loopback a copy */
if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
mcopy = m_copy(m, 0, (int)M_COPYALL);
break;
#endif
#ifdef IPX
case AF_IPX:
etype = htons(ETHERTYPE_IPX);
bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
(caddr_t)edst, sizeof (edst));
/* If broadcasting on a simplex interface, loopback a copy */
if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
mcopy = m_copy(m, 0, (int)M_COPYALL);
break;
#endif
#ifdef ISO
case AF_ISO: {
int snpalen;
struct llc *l;
struct sockaddr_dl *sdl;
if (rt && (sdl = (struct sockaddr_dl *)rt->rt_gateway) &&
sdl->sdl_family == AF_LINK && sdl->sdl_alen > 0) {
bcopy(LLADDR(sdl), (caddr_t)edst, sizeof(edst));
} else {
error = iso_snparesolve(ifp, (struct sockaddr_iso *)dst,
(char *)edst, &snpalen);
if (error)
goto bad; /* Not Resolved */
}
/* If broadcasting on a simplex interface, loopback a copy */
if (*edst & 1)
m->m_flags |= (M_BCAST|M_MCAST);
if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX) &&
(mcopy = m_copy(m, 0, (int)M_COPYALL))) {
M_PREPEND(mcopy, sizeof (*eh), M_DONTWAIT);
if (mcopy) {
eh = mtod(mcopy, struct ether_header *);
bcopy((caddr_t)edst,
(caddr_t)eh->ether_dhost, sizeof (edst));
bcopy(LLADDR(ifp->if_sadl),
(caddr_t)eh->ether_shost, sizeof (edst));
}
}
M_PREPEND(m, 3, M_DONTWAIT);
if (m == NULL)
return (0);
l = mtod(m, struct llc *);
l->llc_dsap = l->llc_ssap = LLC_ISO_LSAP;
l->llc_control = LLC_UI;
#ifdef ARGO_DEBUG
if (argo_debug[D_ETHER]) {
int i;
printf("unoutput: sending pkt to: ");
for (i=0; i<6; i++)
printf("%x ", edst[i] & 0xff);
printf("\n");
}
#endif
} break;
#endif /* ISO */
#ifdef LLC
/* case AF_NSAP: */
case AF_CCITT: {
struct sockaddr_dl *sdl =
(struct sockaddr_dl *) rt -> rt_gateway;
if (sdl && sdl->sdl_family == AF_LINK
&& sdl->sdl_alen > 0) {
bcopy(LLADDR(sdl), (char *)edst,
sizeof(edst));
} else goto bad; /* Not a link interface ? Funny ... */
if ((ifp->if_flags & IFF_SIMPLEX) && (*edst & 1) &&
(mcopy = m_copy(m, 0, (int)M_COPYALL))) {
M_PREPEND(mcopy, sizeof (*eh), M_DONTWAIT);
if (mcopy) {
eh = mtod(mcopy, struct ether_header *);
bcopy((caddr_t)edst,
(caddr_t)eh->ether_dhost, sizeof (edst));
bcopy(LLADDR(ifp->if_sadl),
(caddr_t)eh->ether_shost, sizeof (edst));
}
}
#ifdef LLC_DEBUG
{
int i;
struct llc *l = mtod(m, struct llc *);
printf("ether_output: sending LLC2 pkt to: ");
for (i=0; i<6; i++)
printf("%x ", edst[i] & 0xff);
printf(" len 0x%x dsap 0x%x ssap 0x%x control 0x%x\n",
m->m_pkthdr.len, l->llc_dsap & 0xff, l->llc_ssap &0xff,
l->llc_control & 0xff);
}
#endif /* LLC_DEBUG */
} break;
struct mbuf *
ieee80211_ccmp_encrypt(struct ieee80211com *ic, struct mbuf *m0,
struct ieee80211_key *k)
{
struct ieee80211_ccmp_ctx *ctx = k->k_priv;
const struct ieee80211_frame *wh;
const u_int8_t *src;
u_int8_t *ivp, *mic, *dst;
u_int8_t a[16], b[16], s0[16], s[16];
struct mbuf *n0, *m, *n;
int hdrlen, left, moff, noff, len;
u_int16_t ctr;
int i, j;
MGET(n0, M_DONTWAIT, m0->m_type);
if (n0 == NULL)
goto nospace;
if (m_dup_pkthdr(n0, m0, M_DONTWAIT))
goto nospace;
n0->m_pkthdr.len += IEEE80211_CCMP_HDRLEN;
n0->m_len = MHLEN;
if (n0->m_pkthdr.len >= MINCLSIZE - IEEE80211_CCMP_MICLEN) {
MCLGET(n0, M_DONTWAIT);
if (n0->m_flags & M_EXT)
n0->m_len = n0->m_ext.ext_size;
}
if (n0->m_len > n0->m_pkthdr.len)
n0->m_len = n0->m_pkthdr.len;
/* copy 802.11 header */
wh = mtod(m0, struct ieee80211_frame *);
hdrlen = ieee80211_get_hdrlen(wh);
memcpy(mtod(n0, caddr_t), wh, hdrlen);
k->k_tsc++; /* increment the 48-bit PN */
/* construct CCMP header */
ivp = mtod(n0, u_int8_t *) + hdrlen;
ivp[0] = k->k_tsc; /* PN0 */
ivp[1] = k->k_tsc >> 8; /* PN1 */
ivp[2] = 0; /* Rsvd */
ivp[3] = k->k_id << 6 | IEEE80211_WEP_EXTIV; /* KeyID | ExtIV */
ivp[4] = k->k_tsc >> 16; /* PN2 */
ivp[5] = k->k_tsc >> 24; /* PN3 */
ivp[6] = k->k_tsc >> 32; /* PN4 */
ivp[7] = k->k_tsc >> 40; /* PN5 */
/* construct initial B, A and S_0 blocks */
ieee80211_ccmp_phase1(&ctx->rijndael, wh, k->k_tsc,
m0->m_pkthdr.len - hdrlen, b, a, s0);
/* construct S_1 */
ctr = 1;
a[14] = ctr >> 8;
a[15] = ctr & 0xff;
rijndael_encrypt(&ctx->rijndael, a, s);
/* encrypt frame body and compute MIC */
j = 0;
m = m0;
n = n0;
moff = hdrlen;
noff = hdrlen + IEEE80211_CCMP_HDRLEN;
left = m0->m_pkthdr.len - moff;
while (left > 0) {
if (moff == m->m_len) {
/* nothing left to copy from m */
m = m->m_next;
moff = 0;
}
if (noff == n->m_len) {
/* n is full and there's more data to copy */
MGET(n->m_next, M_DONTWAIT, n->m_type);
if (n->m_next == NULL)
goto nospace;
n = n->m_next;
n->m_len = MLEN;
if (left >= MINCLSIZE - IEEE80211_CCMP_MICLEN) {
MCLGET(n, M_DONTWAIT);
if (n->m_flags & M_EXT)
n->m_len = n->m_ext.ext_size;
}
if (n->m_len > left)
n->m_len = left;
noff = 0;
}
len = min(m->m_len - moff, n->m_len - noff);
src = mtod(m, u_int8_t *) + moff;
dst = mtod(n, u_int8_t *) + noff;
for (i = 0; i < len; i++) {
/* update MIC with clear text */
b[j] ^= src[i];
/* encrypt message */
dst[i] = src[i] ^ s[j];
if (++j < 16)
continue;
/* we have a full block, encrypt MIC */
rijndael_encrypt(&ctx->rijndael, b, b);
/* construct a new S_ctr block */
ctr++;
a[14] = ctr >> 8;
a[15] = ctr & 0xff;
rijndael_encrypt(&ctx->rijndael, a, s);
j = 0;
}
moff += len;
noff += len;
left -= len;
}
if (j != 0) /* partial block, encrypt MIC */
rijndael_encrypt(&ctx->rijndael, b, b);
/* reserve trailing space for MIC */
if (M_TRAILINGSPACE(n) < IEEE80211_CCMP_MICLEN) {
MGET(n->m_next, M_DONTWAIT, n->m_type);
if (n->m_next == NULL)
goto nospace;
n = n->m_next;
n->m_len = 0;
}
/* finalize MIC, U := T XOR first-M-bytes( S_0 ) */
mic = mtod(n, u_int8_t *) + n->m_len;
for (i = 0; i < IEEE80211_CCMP_MICLEN; i++)
mic[i] = b[i] ^ s0[i];
n->m_len += IEEE80211_CCMP_MICLEN;
n0->m_pkthdr.len += IEEE80211_CCMP_MICLEN;
m_freem(m0);
return n0;
nospace:
ic->ic_stats.is_tx_nombuf++;
m_freem(m0);
if (n0 != NULL)
m_freem(n0);
return NULL;
}
int
sosend(struct socket *so,
struct mbuf *nam, /* sockaddr, if UDP socket, NULL if TCP */
char *data, /* data to send */
int *data_length, /* IN/OUT length of (remaining) data */
int flags)
{
struct mbuf *head = (struct mbuf *)NULL;
struct mbuf *m;
int space;
int resid;
int len;
int error = 0;
int dontroute;
int first = 1;
resid = *data_length;
/*
* In theory resid should be unsigned.
* However, space must be signed, as it might be less than 0
* if we over-committed, and we must use a signed comparison
* of space and resid. On the other hand, a negative resid
* causes us to loop sending 0-length segments to the protocol.
*/
if (resid < 0)
return (EINVAL);
INET_TRACE (INETM_IO, ("INET:sosend: so %lx resid %d sb_hiwat %d so_state %x\n",
so, resid, so->so_snd.sb_hiwat, so->so_state));
if (sosendallatonce(so) && (resid > (int)so->so_snd.sb_hiwat))
return (EMSGSIZE);
dontroute = (flags & MSG_DONTROUTE) &&
((so->so_options & SO_DONTROUTE) == 0) &&
(so->so_proto->pr_flags & PR_ATOMIC);
#define snderr(errno) { error = errno; goto release; }
restart:
sblock(&so->so_snd);
do
{
if (so->so_error)
{
error = so->so_error;
so->so_error = 0; /* ??? */
goto release;
}
if (so->so_state & SS_CANTSENDMORE)
snderr(EPIPE);
if ((so->so_state & SS_ISCONNECTED) == 0)
{
if (so->so_proto->pr_flags & PR_CONNREQUIRED)
snderr(ENOTCONN);
if (nam == 0)
snderr(EDESTADDRREQ);
}
if (flags & MSG_OOB)
space = 1024;
else
{
space = (int)sbspace(&so->so_snd);
if ((sosendallatonce(so) && (space < resid)) ||
((resid >= CLBYTES) && (space < CLBYTES) &&
(so->so_snd.sb_cc >= CLBYTES) &&
((so->so_state & SS_NBIO) == 0) &&
((flags & MSG_DONTWAIT) == 0)))
{
if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT))
{
if (first)
error = EWOULDBLOCK;
goto release;
}
sbunlock(&so->so_snd);
sbwait(&so->so_snd);
goto restart;
}
}
if ( space <= 0 )
{
/* no space in socket send buffer - see if we can wait */
if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT))
{
if (first) /* report first error */
error = EWOULDBLOCK;
goto release;
}
/* If blocking socket, let someone else run */
sbunlock(&so->so_snd);
sbwait(&so->so_snd);
goto restart;
}
while (space > 0)
{
len = resid;
if ( so->so_type == SOCK_STREAM )
{
m = m_getwithdata(MT_TXDATA, len);
if (!m)
snderr(ENOBUFS);
MEMCPY(m->m_data, data, len);
so->so_snd.sb_flags |= SB_MBCOMP; /* allow compression */
}
else
{
m = m_get (M_WAIT, MT_TXDATA);
m->m_data = data;
}
INET_TRACE (INETM_IO,
("sosend:got %d bytes so %lx mlen %d, off %d mtod %x\n",
len, so, m->m_len, m->m_off, mtod (m, caddr_t)));
*data_length -= len;
resid -= len;
data += len;
m->m_len = len;
if (head == (struct mbuf *)NULL)
head = m;
if (error)
goto release;
if (*data_length <= 0)
break;
}
if (dontroute)
so->so_options |= SO_DONTROUTE;
so->so_req = (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND;
error = (*so->so_proto->pr_usrreq)(so, head, nam);
if (dontroute)
so->so_options &= ~SO_DONTROUTE;
head = (struct mbuf *)NULL;
first = 0;
} while ((resid != 0) && (error == 0));
release:
sbunlock(&so->so_snd);
if (head)
m_freem(head);
return error;
}
static void
nfssvc_program(struct svc_req *rqst, SVCXPRT *xprt)
{
struct nfsrv_descript nd;
struct nfsrvcache *rp = NULL;
int cacherep, credflavor;
memset(&nd, 0, sizeof(nd));
if (rqst->rq_vers == NFS_VER2) {
if (rqst->rq_proc > NFSV2PROC_STATFS) {
svcerr_noproc(rqst);
svc_freereq(rqst);
goto out;
}
nd.nd_procnum = newnfs_nfsv3_procid[rqst->rq_proc];
nd.nd_flag = ND_NFSV2;
} else if (rqst->rq_vers == NFS_VER3) {
if (rqst->rq_proc >= NFS_V3NPROCS) {
svcerr_noproc(rqst);
svc_freereq(rqst);
goto out;
}
nd.nd_procnum = rqst->rq_proc;
nd.nd_flag = ND_NFSV3;
} else {
if (rqst->rq_proc != NFSPROC_NULL &&
rqst->rq_proc != NFSV4PROC_COMPOUND) {
svcerr_noproc(rqst);
svc_freereq(rqst);
goto out;
}
nd.nd_procnum = rqst->rq_proc;
nd.nd_flag = ND_NFSV4;
}
/*
* Note: we want rq_addr, not svc_getrpccaller for nd_nam2 -
* NFS_SRVMAXDATA uses a NULL value for nd_nam2 to detect TCP
* mounts.
*/
nd.nd_mrep = rqst->rq_args;
rqst->rq_args = NULL;
newnfs_realign(&nd.nd_mrep);
nd.nd_md = nd.nd_mrep;
nd.nd_dpos = mtod(nd.nd_md, caddr_t);
nd.nd_nam = svc_getrpccaller(rqst);
nd.nd_nam2 = rqst->rq_addr;
nd.nd_mreq = NULL;
nd.nd_cred = NULL;
if (nfs_privport && (nd.nd_flag & ND_NFSV4) == 0) {
/* Check if source port is privileged */
u_short port;
struct sockaddr *nam = nd.nd_nam;
struct sockaddr_in *sin;
sin = (struct sockaddr_in *)nam;
/*
* INET/INET6 - same code:
* sin_port and sin6_port are at same offset
*/
port = ntohs(sin->sin_port);
if (port >= IPPORT_RESERVED &&
nd.nd_procnum != NFSPROC_NULL) {
#ifdef INET6
char b6[INET6_ADDRSTRLEN];
#if defined(KLD_MODULE)
/* Do not use ip6_sprintf: the nfs module should work without INET6. */
#define ip6_sprintf(buf, a) \
(sprintf((buf), "%x:%x:%x:%x:%x:%x:%x:%x", \
(a)->s6_addr16[0], (a)->s6_addr16[1], \
(a)->s6_addr16[2], (a)->s6_addr16[3], \
(a)->s6_addr16[4], (a)->s6_addr16[5], \
(a)->s6_addr16[6], (a)->s6_addr16[7]), \
(buf))
#endif
#endif
printf("NFS request from unprivileged port (%s:%d)\n",
#ifdef INET6
sin->sin_family == AF_INET6 ?
ip6_sprintf(b6, &satosin6(sin)->sin6_addr) :
#if defined(KLD_MODULE)
#undef ip6_sprintf
#endif
#endif
inet_ntoa(sin->sin_addr), port);
svcerr_weakauth(rqst);
svc_freereq(rqst);
m_freem(nd.nd_mrep);
goto out;
}
}
if (nd.nd_procnum != NFSPROC_NULL) {
if (!svc_getcred(rqst, &nd.nd_cred, &credflavor)) {
svcerr_weakauth(rqst);
svc_freereq(rqst);
m_freem(nd.nd_mrep);
goto out;
}
/* Set the flag based on credflavor */
if (credflavor == RPCSEC_GSS_KRB5) {
nd.nd_flag |= ND_GSS;
} else if (credflavor == RPCSEC_GSS_KRB5I) {
nd.nd_flag |= (ND_GSS | ND_GSSINTEGRITY);
} else if (credflavor == RPCSEC_GSS_KRB5P) {
nd.nd_flag |= (ND_GSS | ND_GSSPRIVACY);
} else if (credflavor != AUTH_SYS) {
svcerr_weakauth(rqst);
svc_freereq(rqst);
m_freem(nd.nd_mrep);
goto out;
}
#ifdef MAC
mac_cred_associate_nfsd(nd.nd_cred);
#endif
if ((nd.nd_flag & ND_NFSV4) != 0) {
nd.nd_repstat = nfsvno_v4rootexport(&nd);
if (nd.nd_repstat != 0) {
svcerr_weakauth(rqst);
svc_freereq(rqst);
m_freem(nd.nd_mrep);
goto out;
}
}
cacherep = nfs_proc(&nd, rqst->rq_xid, xprt->xp_socket,
xprt->xp_sockref, &rp);
} else {
NFSMGET(nd.nd_mreq);
nd.nd_mreq->m_len = 0;
cacherep = RC_REPLY;
}
if (nd.nd_mrep != NULL)
m_freem(nd.nd_mrep);
if (nd.nd_cred != NULL)
crfree(nd.nd_cred);
if (cacherep == RC_DROPIT) {
if (nd.nd_mreq != NULL)
m_freem(nd.nd_mreq);
svc_freereq(rqst);
goto out;
}
if (nd.nd_mreq == NULL) {
svcerr_decode(rqst);
svc_freereq(rqst);
goto out;
}
if (nd.nd_repstat & NFSERR_AUTHERR) {
svcerr_auth(rqst, nd.nd_repstat & ~NFSERR_AUTHERR);
if (nd.nd_mreq != NULL)
m_freem(nd.nd_mreq);
} else if (!svc_sendreply_mbuf(rqst, nd.nd_mreq)) {
svcerr_systemerr(rqst);
}
if (rp != NULL)
nfsrvd_sentcache(rp, xprt->xp_socket, 0);
svc_freereq(rqst);
out:
NFSEXITCODE(0);
}
/*
* Set up nameidata for a lookup() call and do it.
*
* If pubflag is set, this call is done for a lookup operation on the
* public filehandle. In that case we allow crossing mountpoints and
* absolute pathnames. However, the caller is expected to check that
* the lookup result is within the public fs, and deny access if
* it is not.
*
* nfs_namei() clears out garbage fields that namei() might leave garbage.
* This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
* error occurs but the parent was not requested.
*
* dirp may be set whether an error is returned or not, and must be
* released by the caller.
*/
int
nfs_namei(struct nameidata *ndp, struct nfsrv_descript *nfsd,
fhandle_t *fhp, int len, struct nfssvc_sock *slp,
struct sockaddr *nam, struct mbuf **mdp,
caddr_t *dposp, struct vnode **retdirp, int v3, struct vattr *retdirattrp,
int *retdirattr_retp, int pubflag)
{
int i, rem;
struct mbuf *md;
char *fromcp, *tocp, *cp;
struct iovec aiov;
struct uio auio;
struct vnode *dp;
int error, rdonly, linklen;
struct componentname *cnp = &ndp->ni_cnd;
int lockleaf = (cnp->cn_flags & LOCKLEAF) != 0;
int dvfslocked;
int vfslocked;
vfslocked = 0;
dvfslocked = 0;
*retdirp = NULL;
cnp->cn_flags |= NOMACCHECK;
cnp->cn_pnbuf = uma_zalloc(namei_zone, M_WAITOK);
/*
* Copy the name from the mbuf list to ndp->ni_pnbuf
* and set the various ndp fields appropriately.
*/
fromcp = *dposp;
tocp = cnp->cn_pnbuf;
md = *mdp;
rem = mtod(md, caddr_t) + md->m_len - fromcp;
for (i = 0; i < len; i++) {
while (rem == 0) {
md = md->m_next;
if (md == NULL) {
error = EBADRPC;
goto out;
}
fromcp = mtod(md, caddr_t);
rem = md->m_len;
}
if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
error = EACCES;
goto out;
}
*tocp++ = *fromcp++;
rem--;
}
*tocp = '\0';
*mdp = md;
*dposp = fromcp;
len = nfsm_rndup(len)-len;
if (len > 0) {
if (rem >= len)
*dposp += len;
else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
goto out;
}
/*
* Extract and set starting directory.
*/
error = nfsrv_fhtovp(fhp, FALSE, &dp, &dvfslocked,
nfsd, slp, nam, &rdonly, pubflag);
if (error)
goto out;
vfslocked = VFS_LOCK_GIANT(dp->v_mount);
if (dp->v_type != VDIR) {
vrele(dp);
error = ENOTDIR;
goto out;
}
if (rdonly)
cnp->cn_flags |= RDONLY;
/*
* Set return directory. Reference to dp is implicitly transfered
* to the returned pointer
*/
*retdirp = dp;
if (v3) {
vn_lock(dp, LK_EXCLUSIVE | LK_RETRY);
*retdirattr_retp = VOP_GETATTR(dp, retdirattrp,
ndp->ni_cnd.cn_cred);
VOP_UNLOCK(dp, 0);
}
if (pubflag) {
/*
* Oh joy. For WebNFS, handle those pesky '%' escapes,
* and the 'native path' indicator.
*/
cp = uma_zalloc(namei_zone, M_WAITOK);
fromcp = cnp->cn_pnbuf;
tocp = cp;
if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
switch ((unsigned char)*fromcp) {
case WEBNFS_NATIVE_CHAR:
/*
* 'Native' path for us is the same
* as a path according to the NFS spec,
* just skip the escape char.
*/
fromcp++;
break;
/*
* More may be added in the future, range 0x80-0xff
*/
default:
error = EIO;
uma_zfree(namei_zone, cp);
goto out;
}
}
/*
* Translate the '%' escapes, URL-style.
*/
while (*fromcp != '\0') {
if (*fromcp == WEBNFS_ESC_CHAR) {
if (fromcp[1] != '\0' && fromcp[2] != '\0') {
fromcp++;
*tocp++ = HEXSTRTOI(fromcp);
fromcp += 2;
continue;
} else {
error = ENOENT;
uma_zfree(namei_zone, cp);
goto out;
}
} else
*tocp++ = *fromcp++;
}
*tocp = '\0';
uma_zfree(namei_zone, cnp->cn_pnbuf);
cnp->cn_pnbuf = cp;
}
ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
ndp->ni_segflg = UIO_SYSSPACE;
if (pubflag) {
ndp->ni_rootdir = rootvnode;
ndp->ni_loopcnt = 0;
if (cnp->cn_pnbuf[0] == '/') {
int tvfslocked;
tvfslocked = VFS_LOCK_GIANT(rootvnode->v_mount);
VFS_UNLOCK_GIANT(vfslocked);
dp = rootvnode;
vfslocked = tvfslocked;
}
} else {
cnp->cn_flags |= NOCROSSMOUNT;
}
/*
* Initialize for scan, set ni_startdir and bump ref on dp again
* because lookup() will dereference ni_startdir.
*/
cnp->cn_thread = curthread;
VREF(dp);
ndp->ni_startdir = dp;
if (!lockleaf)
cnp->cn_flags |= LOCKLEAF;
for (;;) {
cnp->cn_nameptr = cnp->cn_pnbuf;
/*
* Call lookup() to do the real work. If an error occurs,
* ndp->ni_vp and ni_dvp are left uninitialized or NULL and
* we do not have to dereference anything before returning.
* In either case ni_startdir will be dereferenced and NULLed
* out.
*/
if (vfslocked)
ndp->ni_cnd.cn_flags |= GIANTHELD;
error = lookup(ndp);
vfslocked = (ndp->ni_cnd.cn_flags & GIANTHELD) != 0;
ndp->ni_cnd.cn_flags &= ~GIANTHELD;
if (error)
break;
/*
* Check for encountering a symbolic link. Trivial
* termination occurs if no symlink encountered.
* Note: zfree is safe because error is 0, so we will
* not zfree it again when we break.
*/
if ((cnp->cn_flags & ISSYMLINK) == 0) {
if (cnp->cn_flags & (SAVENAME | SAVESTART))
cnp->cn_flags |= HASBUF;
else
uma_zfree(namei_zone, cnp->cn_pnbuf);
if (ndp->ni_vp && !lockleaf)
VOP_UNLOCK(ndp->ni_vp, 0);
break;
}
/*
* Validate symlink
*/
if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
VOP_UNLOCK(ndp->ni_dvp, 0);
if (!pubflag) {
error = EINVAL;
goto badlink2;
}
if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
error = ELOOP;
goto badlink2;
}
if (ndp->ni_pathlen > 1)
cp = uma_zalloc(namei_zone, M_WAITOK);
else
cp = cnp->cn_pnbuf;
aiov.iov_base = cp;
aiov.iov_len = MAXPATHLEN;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = 0;
auio.uio_rw = UIO_READ;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_td = NULL;
auio.uio_resid = MAXPATHLEN;
error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
if (error) {
badlink1:
if (ndp->ni_pathlen > 1)
uma_zfree(namei_zone, cp);
badlink2:
vput(ndp->ni_vp);
vrele(ndp->ni_dvp);
break;
}
linklen = MAXPATHLEN - auio.uio_resid;
if (linklen == 0) {
error = ENOENT;
goto badlink1;
}
if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
error = ENAMETOOLONG;
goto badlink1;
}
/*
* Adjust or replace path
*/
if (ndp->ni_pathlen > 1) {
bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
uma_zfree(namei_zone, cnp->cn_pnbuf);
cnp->cn_pnbuf = cp;
} else
cnp->cn_pnbuf[linklen] = '\0';
ndp->ni_pathlen += linklen;
/*
* Cleanup refs for next loop and check if root directory
* should replace current directory. Normally ni_dvp
* becomes the new base directory and is cleaned up when
* we loop. Explicitly null pointers after invalidation
* to clarify operation.
*/
vput(ndp->ni_vp);
ndp->ni_vp = NULL;
if (cnp->cn_pnbuf[0] == '/') {
vrele(ndp->ni_dvp);
ndp->ni_dvp = ndp->ni_rootdir;
VREF(ndp->ni_dvp);
}
ndp->ni_startdir = ndp->ni_dvp;
ndp->ni_dvp = NULL;
}
if (!lockleaf)
cnp->cn_flags &= ~LOCKLEAF;
if (cnp->cn_flags & GIANTHELD) {
mtx_unlock(&Giant);
cnp->cn_flags &= ~GIANTHELD;
}
/*
* nfs_namei() guarentees that fields will not contain garbage
* whether an error occurs or not. This allows the caller to track
* cleanup state trivially.
*/
out:
if (error) {
uma_zfree(namei_zone, cnp->cn_pnbuf);
ndp->ni_vp = NULL;
ndp->ni_dvp = NULL;
ndp->ni_startdir = NULL;
cnp->cn_flags &= ~HASBUF;
VFS_UNLOCK_GIANT(vfslocked);
vfslocked = 0;
} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
ndp->ni_dvp = NULL;
}
/*
* This differs from normal namei() in that even on failure we may
* return with Giant held due to the dirp return. Make sure we only
* have not recursed however. The calling code only expects to drop
* one acquire.
*/
if (vfslocked || dvfslocked)
ndp->ni_cnd.cn_flags |= GIANTHELD;
if (vfslocked && dvfslocked)
VFS_UNLOCK_GIANT(vfslocked);
return (error);
}
/*
* ESP output routine, called by ipsec[46]_process_packet().
*/
static int
esp_output(
struct mbuf *m,
struct ipsecrequest *isr,
struct mbuf **mp,
int skip,
int protoff
)
{
struct enc_xform *espx;
struct auth_hash *esph;
int hlen, rlen, plen, padding, blks, alen, i, roff;
struct mbuf *mo = (struct mbuf *) NULL;
struct tdb_crypto *tc;
struct secasvar *sav;
struct secasindex *saidx;
unsigned char *pad;
u_int8_t prot;
int error, maxpacketsize;
struct cryptodesc *crde = NULL, *crda = NULL;
struct cryptop *crp;
SPLASSERT(net, "esp_output");
sav = isr->sav;
KASSERT(sav != NULL, ("esp_output: null SA"));
esph = sav->tdb_authalgxform;
espx = sav->tdb_encalgxform;
KASSERT(espx != NULL, ("esp_output: null encoding xform"));
if (sav->flags & SADB_X_EXT_OLD)
hlen = sizeof (struct esp) + sav->ivlen;
else
hlen = sizeof (struct newesp) + sav->ivlen;
rlen = m->m_pkthdr.len - skip; /* Raw payload length. */
/*
* NB: The null encoding transform has a blocksize of 4
* so that headers are properly aligned.
*/
blks = espx->blocksize; /* IV blocksize */
/* XXX clamp padding length a la KAME??? */
padding = ((blks - ((rlen + 2) % blks)) % blks) + 2;
plen = rlen + padding; /* Padded payload length. */
if (esph)
alen = AH_HMAC_HASHLEN;
else
alen = 0;
espstat.esps_output++;
saidx = &sav->sah->saidx;
/* Check for maximum packet size violations. */
switch (saidx->dst.sa.sa_family) {
#ifdef INET
case AF_INET:
maxpacketsize = IP_MAXPACKET;
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
maxpacketsize = IPV6_MAXPACKET;
break;
#endif /* INET6 */
default:
DPRINTF(("esp_output: unknown/unsupported protocol "
"family %d, SA %s/%08lx\n",
saidx->dst.sa.sa_family, ipsec_address(&saidx->dst),
(u_long) ntohl(sav->spi)));
espstat.esps_nopf++;
error = EPFNOSUPPORT;
goto bad;
}
if (skip + hlen + rlen + padding + alen > maxpacketsize) {
DPRINTF(("esp_output: packet in SA %s/%08lx got too big "
"(len %u, max len %u)\n",
ipsec_address(&saidx->dst), (u_long) ntohl(sav->spi),
skip + hlen + rlen + padding + alen, maxpacketsize));
espstat.esps_toobig++;
error = EMSGSIZE;
goto bad;
}
/* Update the counters. */
espstat.esps_obytes += m->m_pkthdr.len - skip;
m = m_clone(m);
if (m == NULL) {
DPRINTF(("esp_output: cannot clone mbuf chain, SA %s/%08lx\n",
ipsec_address(&saidx->dst), (u_long) ntohl(sav->spi)));
espstat.esps_hdrops++;
error = ENOBUFS;
goto bad;
}
/* Inject ESP header. */
mo = m_makespace(m, skip, hlen, &roff);
if (mo == NULL) {
DPRINTF(("esp_output: failed to inject %u byte ESP hdr for SA "
"%s/%08lx\n",
hlen, ipsec_address(&saidx->dst),
(u_long) ntohl(sav->spi)));
espstat.esps_hdrops++; /* XXX diffs from openbsd */
error = ENOBUFS;
goto bad;
}
/* Initialize ESP header. */
bcopy((caddr_t) &sav->spi, mtod(mo, caddr_t) + roff, sizeof(u_int32_t));
if (sav->replay) {
u_int32_t replay = htonl(++(sav->replay->count));
bcopy((caddr_t) &replay,
mtod(mo, caddr_t) + roff + sizeof(u_int32_t),
sizeof(u_int32_t));
}
/*
* Add padding -- better to do it ourselves than use the crypto engine,
* although if/when we support compression, we'd have to do that.
*/
pad = (u_char *) m_pad(m, padding + alen);
if (pad == NULL) {
DPRINTF(("esp_output: m_pad failed for SA %s/%08lx\n",
ipsec_address(&saidx->dst), (u_long) ntohl(sav->spi)));
m = NULL; /* NB: free'd by m_pad */
error = ENOBUFS;
goto bad;
}
/*
* Add padding: random, zero, or self-describing.
* XXX catch unexpected setting
*/
switch (sav->flags & SADB_X_EXT_PMASK) {
case SADB_X_EXT_PRAND:
(void) read_random(pad, padding - 2);
break;
case SADB_X_EXT_PZERO:
bzero(pad, padding - 2);
break;
case SADB_X_EXT_PSEQ:
for (i = 0; i < padding - 2; i++)
pad[i] = i+1;
break;
}
/* Fix padding length and Next Protocol in padding itself. */
pad[padding - 2] = padding - 2;
m_copydata(m, protoff, sizeof(u_int8_t), pad + padding - 1);
/* Fix Next Protocol in IPv4/IPv6 header. */
prot = IPPROTO_ESP;
m_copyback(m, protoff, sizeof(u_int8_t), (u_char *) &prot);
/* Get crypto descriptors. */
crp = crypto_getreq(esph && espx ? 2 : 1);
if (crp == NULL) {
DPRINTF(("esp_output: failed to acquire crypto descriptors\n"));
espstat.esps_crypto++;
error = ENOBUFS;
goto bad;
}
if (espx) {
crde = crp->crp_desc;
crda = crde->crd_next;
/* Encryption descriptor. */
crde->crd_skip = skip + hlen;
crde->crd_len = m->m_pkthdr.len - (skip + hlen + alen);
crde->crd_flags = CRD_F_ENCRYPT;
crde->crd_inject = skip + hlen - sav->ivlen;
/* Encryption operation. */
crde->crd_alg = espx->type;
crde->crd_key = _KEYBUF(sav->key_enc);
crde->crd_klen = _KEYBITS(sav->key_enc);
/* XXX Rounds ? */
} else
crda = crp->crp_desc;
/* IPsec-specific opaque crypto info. */
tc = (struct tdb_crypto *) malloc(sizeof(struct tdb_crypto),
M_XDATA, M_NOWAIT|M_ZERO);
if (tc == NULL) {
crypto_freereq(crp);
DPRINTF(("esp_output: failed to allocate tdb_crypto\n"));
espstat.esps_crypto++;
error = ENOBUFS;
goto bad;
}
/* Callback parameters */
tc->tc_isr = isr;
tc->tc_spi = sav->spi;
tc->tc_dst = saidx->dst;
tc->tc_proto = saidx->proto;
/* Crypto operation descriptor. */
crp->crp_ilen = m->m_pkthdr.len; /* Total input length. */
crp->crp_flags = CRYPTO_F_IMBUF;
crp->crp_buf = (caddr_t) m;
crp->crp_callback = esp_output_cb;
crp->crp_opaque = (caddr_t) tc;
crp->crp_sid = sav->tdb_cryptoid;
if (esph) {
/* Authentication descriptor. */
crda->crd_skip = skip;
crda->crd_len = m->m_pkthdr.len - (skip + alen);
crda->crd_inject = m->m_pkthdr.len - alen;
/* Authentication operation. */
crda->crd_alg = esph->type;
crda->crd_key = _KEYBUF(sav->key_auth);
crda->crd_klen = _KEYBITS(sav->key_auth);
}
return crypto_dispatch(crp);
bad:
if (m)
m_freem(m);
return (error);
}
int
bpf_movein(struct uio *uio, u_int linktype, struct mbuf **mp,
struct sockaddr *sockp, struct bpf_insn *filter)
{
struct mbuf *m;
struct m_tag *mtag;
int error;
u_int hlen;
u_int len;
u_int slen;
/*
* Build a sockaddr based on the data link layer type.
* We do this at this level because the ethernet header
* is copied directly into the data field of the sockaddr.
* In the case of SLIP, there is no header and the packet
* is forwarded as is.
* Also, we are careful to leave room at the front of the mbuf
* for the link level header.
*/
switch (linktype) {
case DLT_SLIP:
sockp->sa_family = AF_INET;
hlen = 0;
break;
case DLT_PPP:
sockp->sa_family = AF_UNSPEC;
hlen = 0;
break;
case DLT_EN10MB:
sockp->sa_family = AF_UNSPEC;
/* XXX Would MAXLINKHDR be better? */
hlen = ETHER_HDR_LEN;
break;
case DLT_FDDI:
sockp->sa_family = AF_UNSPEC;
/* XXX 4(FORMAC)+6(dst)+6(src)+3(LLC)+5(SNAP) */
hlen = 24;
break;
case DLT_IEEE802_11:
case DLT_IEEE802_11_RADIO:
sockp->sa_family = AF_UNSPEC;
hlen = 0;
break;
case DLT_RAW:
case DLT_NULL:
sockp->sa_family = AF_UNSPEC;
hlen = 0;
break;
case DLT_ATM_RFC1483:
/*
* en atm driver requires 4-byte atm pseudo header.
* though it isn't standard, vpi:vci needs to be
* specified anyway.
*/
sockp->sa_family = AF_UNSPEC;
hlen = 12; /* XXX 4(ATM_PH) + 3(LLC) + 5(SNAP) */
break;
default:
return (EIO);
}
if (uio->uio_resid > MCLBYTES)
return (EIO);
len = uio->uio_resid;
MGETHDR(m, M_WAIT, MT_DATA);
m->m_pkthdr.rcvif = 0;
m->m_pkthdr.len = len - hlen;
if (len > MHLEN) {
MCLGET(m, M_WAIT);
if ((m->m_flags & M_EXT) == 0) {
error = ENOBUFS;
goto bad;
}
}
m->m_len = len;
*mp = m;
error = uiomove(mtod(m, caddr_t), len, uio);
if (error)
goto bad;
slen = bpf_filter(filter, mtod(m, u_char *), len, len);
if (slen < len) {
error = EPERM;
goto bad;
}
if (m->m_len < hlen) {
error = EPERM;
goto bad;
}
/*
* Make room for link header, and copy it to sockaddr
*/
if (hlen != 0) {
bcopy(m->m_data, sockp->sa_data, hlen);
m->m_len -= hlen;
m->m_data += hlen; /* XXX */
}
/*
* Prepend the data link type as a mbuf tag
*/
mtag = m_tag_get(PACKET_TAG_DLT, sizeof(u_int), M_NOWAIT);
if (mtag == NULL)
return (ENOMEM);
*(u_int *)(mtag + 1) = linktype;
m_tag_prepend(m, mtag);
return (0);
bad:
m_freem(m);
return (error);
}
static int bsd_accept ( cyg_file *fp, cyg_file *new_fp,
struct sockaddr *name, socklen_t *anamelen )
{
struct mbuf *nam;
socklen_t namelen = 0;
int error = 0, s;
register struct socket *so;
if( anamelen != NULL )
namelen = *anamelen;
s = splsoftnet();
so = (struct socket *)fp->f_data;
if ((so->so_options & SO_ACCEPTCONN) == 0) {
splx(s);
return (EINVAL);
}
if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
splx(s);
return (EWOULDBLOCK);
}
while (so->so_qlen == 0 && so->so_error == 0) {
if (so->so_state & SS_CANTRCVMORE) {
so->so_error = ECONNABORTED;
break;
}
error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
netcon, 0);
if (error) {
splx(s);
return (error);
}
}
if (so->so_error) {
error = so->so_error;
so->so_error = 0;
splx(s);
return (error);
}
{
struct socket *aso = so->so_q;
if (soqremque(aso, 1) == 0)
panic("accept");
so = aso;
}
cyg_selinit(&so->so_rcv.sb_sel);
cyg_selinit(&so->so_snd.sb_sel);
new_fp->f_type = DTYPE_SOCKET;
new_fp->f_flag |= FREAD|FWRITE;
new_fp->f_offset = 0;
new_fp->f_ops = &bsd_sock_fileops;
new_fp->f_data = (CYG_ADDRWORD)so;
new_fp->f_xops = (CYG_ADDRWORD)&bsd_sockops;
nam = m_get(M_WAIT, MT_SONAME);
(void) soaccept(so, nam);
if (name) {
if (namelen > nam->m_len)
namelen = nam->m_len;
/* SHOULD COPY OUT A CHAIN HERE */
if ((error = copyout(mtod(nam, caddr_t),
(caddr_t)name, namelen)) == 0)
*anamelen = namelen;
}
m_freem(nam);
splx(s);
return (error);
}
/*
* Potentially decap ESP in UDP frame. Check for an ESP header
* and optional marker; if present, strip the UDP header and
* push the result through IPSec.
*
* Returns mbuf to be processed (potentially re-allocated) or
* NULL if consumed and/or processed.
*/
static struct mbuf *
udp4_espdecap(struct inpcb *inp, struct mbuf *m, int off)
{
size_t minlen, payload, skip, iphlen;
caddr_t data;
struct udpcb *up;
struct m_tag *tag;
struct udphdr *udphdr;
struct ip *ip;
INP_RLOCK_ASSERT(inp);
/*
* Pull up data so the longest case is contiguous:
* IP/UDP hdr + non ESP marker + ESP hdr.
*/
minlen = off + sizeof(uint64_t) + sizeof(struct esp);
if (minlen > m->m_pkthdr.len)
minlen = m->m_pkthdr.len;
if ((m = m_pullup(m, minlen)) == NULL) {
IPSECSTAT_INC(ips_in_inval);
return (NULL); /* Bypass caller processing. */
}
data = mtod(m, caddr_t); /* Points to ip header. */
payload = m->m_len - off; /* Size of payload. */
if (payload == 1 && data[off] == '\xff')
return (m); /* NB: keepalive packet, no decap. */
up = intoudpcb(inp);
KASSERT(up != NULL, ("%s: udpcb NULL", __func__));
KASSERT((up->u_flags & UF_ESPINUDP_ALL) != 0,
("u_flags 0x%x", up->u_flags));
/*
* Check that the payload is large enough to hold an
* ESP header and compute the amount of data to remove.
*
* NB: the caller has already done a pullup for us.
* XXX can we assume alignment and eliminate bcopys?
*/
if (up->u_flags & UF_ESPINUDP_NON_IKE) {
/*
* draft-ietf-ipsec-nat-t-ike-0[01].txt and
* draft-ietf-ipsec-udp-encaps-(00/)01.txt, ignoring
* possible AH mode non-IKE marker+non-ESP marker
* from draft-ietf-ipsec-udp-encaps-00.txt.
*/
uint64_t marker;
if (payload <= sizeof(uint64_t) + sizeof(struct esp))
return (m); /* NB: no decap. */
bcopy(data + off, &marker, sizeof(uint64_t));
if (marker != 0) /* Non-IKE marker. */
return (m); /* NB: no decap. */
skip = sizeof(uint64_t) + sizeof(struct udphdr);
} else {
uint32_t spi;
if (payload <= sizeof(struct esp)) {
IPSECSTAT_INC(ips_in_inval);
m_freem(m);
return (NULL); /* Discard. */
}
bcopy(data + off, &spi, sizeof(uint32_t));
if (spi == 0) /* Non-ESP marker. */
return (m); /* NB: no decap. */
skip = sizeof(struct udphdr);
}
/*
* Setup a PACKET_TAG_IPSEC_NAT_T_PORT tag to remember
* the UDP ports. This is required if we want to select
* the right SPD for multiple hosts behind same NAT.
*
* NB: ports are maintained in network byte order everywhere
* in the NAT-T code.
*/
tag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
2 * sizeof(uint16_t), M_NOWAIT);
if (tag == NULL) {
IPSECSTAT_INC(ips_in_nomem);
m_freem(m);
return (NULL); /* Discard. */
}
iphlen = off - sizeof(struct udphdr);
udphdr = (struct udphdr *)(data + iphlen);
((uint16_t *)(tag + 1))[0] = udphdr->uh_sport;
((uint16_t *)(tag + 1))[1] = udphdr->uh_dport;
m_tag_prepend(m, tag);
/*
* Remove the UDP header (and possibly the non ESP marker)
* IP header length is iphlen
* Before:
* <--- off --->
* +----+------+-----+
* | IP | UDP | ESP |
* +----+------+-----+
* <-skip->
* After:
* +----+-----+
* | IP | ESP |
* +----+-----+
* <-skip->
*/
ovbcopy(data, data + skip, iphlen);
m_adj(m, skip);
ip = mtod(m, struct ip *);
ip->ip_len = htons(ntohs(ip->ip_len) - skip);
ip->ip_p = IPPROTO_ESP;
/*
* We cannot yet update the cksums so clear any
* h/w cksum flags as they are no longer valid.
*/
if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID)
m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
(void) ipsec4_common_input(m, iphlen, ip->ip_p);
return (NULL); /* NB: consumed, bypass processing. */
}
/*
* nfs_request - goes something like this
* - fill in request struct
* - links it into list
* - calls nfs_send() for first transmit
* - calls nfs_receive() to get reply
* - break down rpc header and return with nfs reply pointed to
* by mrep or error
* nb: always frees up mreq mbuf list
*/
int
nfs_request(struct vnode *vp, struct mbuf *mreq, int procnum,
struct thread *td, struct ucred *cred, struct mbuf **mrp,
struct mbuf **mdp, caddr_t *dposp)
{
struct mbuf *mrep;
u_int32_t *tl;
struct nfsmount *nmp;
struct mbuf *md;
time_t waituntil;
caddr_t dpos;
int error = 0, timeo;
AUTH *auth = NULL;
enum nfs_rto_timer_t timer;
struct nfs_feedback_arg nf;
struct rpc_callextra ext;
enum clnt_stat stat;
struct timeval timo;
/* Reject requests while attempting a forced unmount. */
if (vp->v_mount->mnt_kern_flag & MNTK_UNMOUNTF) {
m_freem(mreq);
return (ESTALE);
}
nmp = VFSTONFS(vp->v_mount);
bzero(&nf, sizeof(struct nfs_feedback_arg));
nf.nf_mount = nmp;
nf.nf_td = td;
nf.nf_lastmsg = time_uptime -
((nmp->nm_tprintf_delay) - (nmp->nm_tprintf_initial_delay));
/*
* XXX if not already connected call nfs_connect now. Longer
* term, change nfs_mount to call nfs_connect unconditionally
* and let clnt_reconnect_create handle reconnects.
*/
if (!nmp->nm_client)
nfs_connect(nmp);
auth = nfs_getauth(nmp, cred);
if (!auth) {
m_freem(mreq);
return (EACCES);
}
bzero(&ext, sizeof(ext));
ext.rc_auth = auth;
ext.rc_feedback = nfs_feedback;
ext.rc_feedback_arg = &nf;
/*
* Use a conservative timeout for RPCs other than getattr,
* lookup, read or write. The justification for doing "other"
* this way is that these RPCs happen so infrequently that
* timer est. would probably be stale. Also, since many of
* these RPCs are non-idempotent, a conservative timeout is
* desired.
*/
timer = nfs_rto_timer(procnum);
if (timer != NFS_DEFAULT_TIMER)
ext.rc_timers = &nmp->nm_timers[timer - 1];
else
ext.rc_timers = NULL;
#ifdef KDTRACE_HOOKS
if (dtrace_nfsclient_nfs23_start_probe != NULL) {
uint32_t probe_id;
int probe_procnum;
if (nmp->nm_flag & NFSMNT_NFSV3) {
probe_id = nfsclient_nfs3_start_probes[procnum];
probe_procnum = procnum;
} else {
probe_id = nfsclient_nfs2_start_probes[procnum];
probe_procnum = nfsv2_procid[procnum];
}
if (probe_id != 0)
(dtrace_nfsclient_nfs23_start_probe)(probe_id, vp,
mreq, cred, probe_procnum);
}
#endif
nfsstats.rpcrequests++;
tryagain:
/*
* This timeout specifies when a new socket should be created,
* along with new xid values. For UDP, this should be done
* infrequently, since retransmits of RPC requests should normally
* use the same xid.
*/
if (nmp->nm_sotype == SOCK_DGRAM) {
if ((nmp->nm_flag & NFSMNT_SOFT) != 0) {
/*
* CLSET_RETRIES is set to 2, so this should be half
* of the total timeout required.
*/
timeo = nmp->nm_retry * nmp->nm_timeo / 2;
if (timeo < 1)
timeo = 1;
timo.tv_sec = timeo / NFS_HZ;
timo.tv_usec = (timeo % NFS_HZ) * 1000000 / NFS_HZ;
} else {
/* For UDP hard mounts, use a large value. */
timo.tv_sec = NFS_MAXTIMEO / NFS_HZ;
timo.tv_usec = 0;
}
} else {
timo.tv_sec = nmp->nm_timeo / NFS_HZ;
timo.tv_usec = (nmp->nm_timeo % NFS_HZ) * 1000000 / NFS_HZ;
}
mrep = NULL;
stat = CLNT_CALL_MBUF(nmp->nm_client, &ext,
(nmp->nm_flag & NFSMNT_NFSV3) ? procnum : nfsv2_procid[procnum],
mreq, &mrep, timo);
/*
* If there was a successful reply and a tprintf msg.
* tprintf a response.
*/
if (stat == RPC_SUCCESS)
error = 0;
else if (stat == RPC_TIMEDOUT) {
nfsstats.rpctimeouts++;
error = ETIMEDOUT;
} else if (stat == RPC_VERSMISMATCH) {
nfsstats.rpcinvalid++;
error = EOPNOTSUPP;
} else if (stat == RPC_PROGVERSMISMATCH) {
nfsstats.rpcinvalid++;
error = EPROTONOSUPPORT;
} else if (stat == RPC_INTR) {
error = EINTR;
} else {
nfsstats.rpcinvalid++;
error = EACCES;
}
if (error)
goto nfsmout;
KASSERT(mrep != NULL, ("mrep shouldn't be NULL if no error\n"));
/*
* Search for any mbufs that are not a multiple of 4 bytes long
* or with m_data not longword aligned.
* These could cause pointer alignment problems, so copy them to
* well aligned mbufs.
*/
error = nfs_realign(&mrep, M_NOWAIT);
if (error == ENOMEM) {
m_freem(mrep);
AUTH_DESTROY(auth);
nfsstats.rpcinvalid++;
return (error);
}
md = mrep;
dpos = mtod(mrep, caddr_t);
tl = nfsm_dissect(u_int32_t *, NFSX_UNSIGNED);
if (*tl != 0) {
error = fxdr_unsigned(int, *tl);
if ((nmp->nm_flag & NFSMNT_NFSV3) &&
error == NFSERR_TRYLATER) {
m_freem(mrep);
error = 0;
waituntil = time_second + nfs3_jukebox_delay;
while (time_second < waituntil)
(void)tsleep(&fake_wchan, PSOCK, "nqnfstry",
hz);
goto tryagain;
}
/*
* Make sure NFSERR_RETERR isn't bogusly set by a server
* such as amd. (No actual NFS error has bit 31 set.)
*/
error &= ~NFSERR_RETERR;
/*
* If the File Handle was stale, invalidate the lookup
* cache, just in case.
*/
if (error == ESTALE)
nfs_purgecache(vp);
/*
* Skip wcc data on non-ENOENT NFS errors for now.
* NetApp filers return corrupt postop attrs in the
* wcc data for NFS err EROFS. Not sure if they could
* return corrupt postop attrs for others errors.
* Blocking ENOENT post-op attributes breaks negative
* name caching, so always allow it through.
*/
if ((nmp->nm_flag & NFSMNT_NFSV3) &&
(!nfs_skip_wcc_data_onerr || error == ENOENT)) {
*mrp = mrep;
*mdp = md;
*dposp = dpos;
error |= NFSERR_RETERR;
} else
m_freem(mrep);
goto nfsmout;
}
/*
* ae_start: [ifnet interface function]
*
* Start packet transmission on the interface.
*/
static void
ae_start(struct ifnet *ifp)
{
struct ae_softc *sc = ifp->if_softc;
struct mbuf *m0, *m;
struct ae_txsoft *txs, *last_txs = NULL;
bus_dmamap_t dmamap;
int error, firsttx, nexttx, lasttx = 1, ofree, seg;
DPRINTF(sc, ("%s: ae_start: sc_flags 0x%08x, if_flags 0x%08x\n",
device_xname(sc->sc_dev), sc->sc_flags, ifp->if_flags));
if ((ifp->if_flags & (IFF_RUNNING|IFF_OACTIVE)) != IFF_RUNNING)
return;
/*
* Remember the previous number of free descriptors and
* the first descriptor we'll use.
*/
ofree = sc->sc_txfree;
firsttx = sc->sc_txnext;
DPRINTF(sc, ("%s: ae_start: txfree %d, txnext %d\n",
device_xname(sc->sc_dev), ofree, firsttx));
/*
* Loop through the send queue, setting up transmit descriptors
* until we drain the queue, or use up all available transmit
* descriptors.
*/
while ((txs = SIMPLEQ_FIRST(&sc->sc_txfreeq)) != NULL &&
sc->sc_txfree != 0) {
/*
* Grab a packet off the queue.
*/
IFQ_POLL(&ifp->if_snd, m0);
if (m0 == NULL)
break;
m = NULL;
dmamap = txs->txs_dmamap;
/*
* Load the DMA map. If this fails, the packet either
* didn't fit in the alloted number of segments, or we were
* short on resources. In this case, we'll copy and try
* again.
*/
if (((mtod(m0, uintptr_t) & 3) != 0) ||
bus_dmamap_load_mbuf(sc->sc_dmat, dmamap, m0,
BUS_DMA_WRITE|BUS_DMA_NOWAIT) != 0) {
MGETHDR(m, M_DONTWAIT, MT_DATA);
if (m == NULL) {
printf("%s: unable to allocate Tx mbuf\n",
device_xname(sc->sc_dev));
break;
}
MCLAIM(m, &sc->sc_ethercom.ec_tx_mowner);
if (m0->m_pkthdr.len > MHLEN) {
MCLGET(m, M_DONTWAIT);
if ((m->m_flags & M_EXT) == 0) {
printf("%s: unable to allocate Tx "
"cluster\n", device_xname(sc->sc_dev));
m_freem(m);
break;
}
}
m_copydata(m0, 0, m0->m_pkthdr.len, mtod(m, void *));
m->m_pkthdr.len = m->m_len = m0->m_pkthdr.len;
error = bus_dmamap_load_mbuf(sc->sc_dmat, dmamap,
m, BUS_DMA_WRITE|BUS_DMA_NOWAIT);
if (error) {
printf("%s: unable to load Tx buffer, "
"error = %d\n", device_xname(sc->sc_dev),
error);
break;
}
}
/*
* Ensure we have enough descriptors free to describe
* the packet.
*/
if (dmamap->dm_nsegs > sc->sc_txfree) {
/*
* Not enough free descriptors to transmit this
* packet. We haven't committed to anything yet,
* so just unload the DMA map, put the packet
* back on the queue, and punt. Notify the upper
* layer that there are no more slots left.
*
* XXX We could allocate an mbuf and copy, but
* XXX it is worth it?
*/
ifp->if_flags |= IFF_OACTIVE;
bus_dmamap_unload(sc->sc_dmat, dmamap);
if (m != NULL)
m_freem(m);
break;
}
IFQ_DEQUEUE(&ifp->if_snd, m0);
if (m != NULL) {
m_freem(m0);
m0 = m;
}
/*
* WE ARE NOW COMMITTED TO TRANSMITTING THE PACKET.
*/
/* Sync the DMA map. */
bus_dmamap_sync(sc->sc_dmat, dmamap, 0, dmamap->dm_mapsize,
BUS_DMASYNC_PREWRITE);
/*
* Initialize the transmit descriptors.
*/
for (nexttx = sc->sc_txnext, seg = 0;
seg < dmamap->dm_nsegs;
seg++, nexttx = AE_NEXTTX(nexttx)) {
/*
* If this is the first descriptor we're
* enqueueing, don't set the OWN bit just
* yet. That could cause a race condition.
* We'll do it below.
*/
sc->sc_txdescs[nexttx].ad_status =
(nexttx == firsttx) ? 0 : ADSTAT_OWN;
sc->sc_txdescs[nexttx].ad_bufaddr1 =
dmamap->dm_segs[seg].ds_addr;
sc->sc_txdescs[nexttx].ad_ctl =
(dmamap->dm_segs[seg].ds_len <<
ADCTL_SIZE1_SHIFT) |
(nexttx == (AE_NTXDESC - 1) ?
ADCTL_ER : 0);
lasttx = nexttx;
}
KASSERT(lasttx != -1);
/* Set `first segment' and `last segment' appropriately. */
sc->sc_txdescs[sc->sc_txnext].ad_ctl |= ADCTL_Tx_FS;
sc->sc_txdescs[lasttx].ad_ctl |= ADCTL_Tx_LS;
#ifdef AE_DEBUG
if (ifp->if_flags & IFF_DEBUG) {
printf(" txsoft %p transmit chain:\n", txs);
for (seg = sc->sc_txnext;; seg = AE_NEXTTX(seg)) {
printf(" descriptor %d:\n", seg);
printf(" ad_status: 0x%08x\n",
sc->sc_txdescs[seg].ad_status);
printf(" ad_ctl: 0x%08x\n",
sc->sc_txdescs[seg].ad_ctl);
printf(" ad_bufaddr1: 0x%08x\n",
sc->sc_txdescs[seg].ad_bufaddr1);
printf(" ad_bufaddr2: 0x%08x\n",
sc->sc_txdescs[seg].ad_bufaddr2);
if (seg == lasttx)
break;
}
}
#endif
/* Sync the descriptors we're using. */
AE_CDTXSYNC(sc, sc->sc_txnext, dmamap->dm_nsegs,
BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
/*
* Store a pointer to the packet so we can free it later,
* and remember what txdirty will be once the packet is
* done.
*/
txs->txs_mbuf = m0;
txs->txs_firstdesc = sc->sc_txnext;
txs->txs_lastdesc = lasttx;
txs->txs_ndescs = dmamap->dm_nsegs;
/* Advance the tx pointer. */
sc->sc_txfree -= dmamap->dm_nsegs;
sc->sc_txnext = nexttx;
SIMPLEQ_REMOVE_HEAD(&sc->sc_txfreeq, txs_q);
SIMPLEQ_INSERT_TAIL(&sc->sc_txdirtyq, txs, txs_q);
last_txs = txs;
/*
* Pass the packet to any BPF listeners.
*/
bpf_mtap(ifp, m0);
}
*np = m->m_next;
m->m_next = n0;
n->m_len = hlen;
m->m_len = skip;
m = n; /* header is at front ... */
*off = 0; /* ... of new mbuf */
}
IPSECSTAT_INC(ips_mbinserted);
} else {
/*
* Copy the remainder to the back of the mbuf
* so there's space to write the new header.
*/
bcopy(mtod(m, caddr_t) + skip,
mtod(m, caddr_t) + skip + hlen, remain);
m->m_len += hlen;
*off = skip;
}
m0->m_pkthdr.len += hlen; /* adjust packet length */
return m;
}
/*
* m_pad(m, n) pads <m> with <n> bytes at the end. The packet header
* length is updated, and a pointer to the first byte of the padding
* (which is guaranteed to be all in one mbuf) is returned.
*/
caddr_t
m_pad(struct mbuf *m, int n)
/*
* Defragment a mbuf chain, returning the shortest possible
* chain of mbufs and clusters. If allocation fails and
* this cannot be completed, NULL will be returned, but
* the passed in chain will be unchanged. Upon success,
* the original chain will be freed, and the new chain
* will be returned.
*
* If a non-packet header is passed in, the original
* mbuf (chain?) will be returned unharmed.
*/
struct mbuf *
m_defrag(struct mbuf *m0, int how)
{
struct mbuf *m_new = NULL, *m_final = NULL;
int progress = 0, length;
MBUF_CHECKSLEEP(how);
if (!(m0->m_flags & M_PKTHDR))
return (m0);
m_fixhdr(m0); /* Needed sanity check */
#ifdef MBUF_STRESS_TEST
if (m_defragrandomfailures) {
int temp = arc4random() & 0xff;
if (temp == 0xba)
goto nospace;
}
#endif
if (m0->m_pkthdr.len > MHLEN)
m_final = m_getcl(how, MT_DATA, M_PKTHDR);
else
m_final = m_gethdr(how, MT_DATA);
if (m_final == NULL)
goto nospace;
if (m_dup_pkthdr(m_final, m0, how) == 0)
goto nospace;
m_new = m_final;
while (progress < m0->m_pkthdr.len) {
length = m0->m_pkthdr.len - progress;
if (length > MCLBYTES)
length = MCLBYTES;
if (m_new == NULL) {
if (length > MLEN)
m_new = m_getcl(how, MT_DATA, 0);
else
m_new = m_get(how, MT_DATA);
if (m_new == NULL)
goto nospace;
}
m_copydata(m0, progress, length, mtod(m_new, caddr_t));
progress += length;
m_new->m_len = length;
if (m_new != m_final)
m_cat(m_final, m_new);
m_new = NULL;
}
#ifdef MBUF_STRESS_TEST
if (m0->m_next == NULL)
m_defraguseless++;
#endif
m_freem(m0);
m0 = m_final;
#ifdef MBUF_STRESS_TEST
m_defragpackets++;
m_defragbytes += m0->m_pkthdr.len;
#endif
return (m0);
nospace:
#ifdef MBUF_STRESS_TEST
m_defragfailure++;
#endif
if (m_final)
m_freem(m_final);
return (NULL);
}
/*
* ESP output routine, called by ipsp_process_packet().
*/
int
esp_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int skip,
int protoff)
{
struct enc_xform *espx = (struct enc_xform *) tdb->tdb_encalgxform;
struct auth_hash *esph = (struct auth_hash *) tdb->tdb_authalgxform;
int ilen, hlen, rlen, padding, blks, alen;
struct mbuf *mi, *mo = (struct mbuf *) NULL;
struct tdb_crypto *tc;
unsigned char *pad;
u_int8_t prot;
struct cryptodesc *crde = NULL, *crda = NULL;
struct cryptop *crp;
#if NBPFILTER > 0
struct ifnet *ifn = &(encif[0].sc_if);
ifn->if_opackets++;
ifn->if_obytes += m->m_pkthdr.len;
if (ifn->if_bpf) {
struct enchdr hdr;
bzero (&hdr, sizeof(hdr));
hdr.af = tdb->tdb_dst.sa.sa_family;
hdr.spi = tdb->tdb_spi;
if (espx)
hdr.flags |= M_CONF;
if (esph)
hdr.flags |= M_AUTH;
bpf_mtap_hdr(ifn->if_bpf, (char *)&hdr, ENC_HDRLEN, m,
BPF_DIRECTION_OUT);
}
#endif
if (tdb->tdb_flags & TDBF_NOREPLAY)
hlen = sizeof(u_int32_t) + tdb->tdb_ivlen;
else
hlen = 2 * sizeof(u_int32_t) + tdb->tdb_ivlen;
rlen = m->m_pkthdr.len - skip; /* Raw payload length. */
if (espx)
blks = espx->blocksize;
else
blks = 4; /* If no encryption, we have to be 4-byte aligned. */
padding = ((blks - ((rlen + 2) % blks)) % blks) + 2;
if (esph)
alen = AH_HMAC_HASHLEN;
else
alen = 0;
espstat.esps_output++;
switch (tdb->tdb_dst.sa.sa_family) {
#ifdef INET
case AF_INET:
/* Check for IP maximum packet size violations. */
if (skip + hlen + rlen + padding + alen > IP_MAXPACKET) {
DPRINTF(("esp_output(): packet in SA %s/%08x got "
"too big\n", ipsp_address(tdb->tdb_dst),
ntohl(tdb->tdb_spi)));
m_freem(m);
espstat.esps_toobig++;
return EMSGSIZE;
}
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
/* Check for IPv6 maximum packet size violations. */
if (skip + hlen + rlen + padding + alen > IPV6_MAXPACKET) {
DPRINTF(("esp_output(): packet in SA %s/%08x got too "
"big\n", ipsp_address(tdb->tdb_dst),
ntohl(tdb->tdb_spi)));
m_freem(m);
espstat.esps_toobig++;
return EMSGSIZE;
}
break;
#endif /* INET6 */
default:
DPRINTF(("esp_output(): unknown/unsupported protocol "
"family %d, SA %s/%08x\n", tdb->tdb_dst.sa.sa_family
, ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
m_freem(m);
espstat.esps_nopf++;
return EPFNOSUPPORT;
}
/* Update the counters. */
tdb->tdb_cur_bytes += m->m_pkthdr.len - skip;
espstat.esps_obytes += m->m_pkthdr.len - skip;
/* Hard byte expiration. */
if (tdb->tdb_flags & TDBF_BYTES &&
tdb->tdb_cur_bytes >= tdb->tdb_exp_bytes) {
pfkeyv2_expire(tdb, SADB_EXT_LIFETIME_HARD);
tdb_delete(tdb);
m_freem(m);
return EINVAL;
}
/* Soft byte expiration. */
if (tdb->tdb_flags & TDBF_SOFT_BYTES &&
tdb->tdb_cur_bytes >= tdb->tdb_soft_bytes) {
pfkeyv2_expire(tdb, SADB_EXT_LIFETIME_SOFT);
tdb->tdb_flags &= ~TDBF_SOFT_BYTES; /* Turn off checking. */
}
/*
* Loop through mbuf chain; if we find a readonly mbuf,
* replace the rest of the chain.
*/
mo = NULL;
mi = m;
while (mi != NULL && !M_READONLY(mi)) {
mo = mi;
mi = mi->m_next;
}
if (mi != NULL) {
/* Replace the rest of the mbuf chain. */
struct mbuf *n = m_copym2(mi, 0, M_COPYALL, M_DONTWAIT);
if (n == NULL) {
DPRINTF(("esp_output(): bad mbuf chain, SA %s/%08x\n",
ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
espstat.esps_hdrops++;
m_freem(m);
return ENOBUFS;
}
if (mo != NULL)
mo->m_next = n;
else
m = n;
m_freem(mi);
}
/* Inject ESP header. */
mo = m_inject(m, skip, hlen, M_DONTWAIT);
if (mo == NULL) {
DPRINTF(("esp_output(): failed to inject ESP header for "
"SA %s/%08x\n", ipsp_address(tdb->tdb_dst),
ntohl(tdb->tdb_spi)));
m_freem(m);
espstat.esps_hdrops++;
return ENOBUFS;
}
/* Initialize ESP header. */
bcopy((caddr_t) &tdb->tdb_spi, mtod(mo, caddr_t), sizeof(u_int32_t));
if (!(tdb->tdb_flags & TDBF_NOREPLAY)) {
u_int32_t replay = htonl(tdb->tdb_rpl++);
bcopy((caddr_t) &replay, mtod(mo, caddr_t) + sizeof(u_int32_t),
sizeof(u_int32_t));
#if NPFSYNC > 0
pfsync_update_tdb(tdb,1);
#endif
}
/*
* Add padding -- better to do it ourselves than use the crypto engine,
* although if/when we support compression, we'd have to do that.
*/
mo = m_inject(m, m->m_pkthdr.len, padding + alen, M_DONTWAIT);
if (mo == NULL) {
DPRINTF(("esp_output(): m_inject failed for SA %s/%08x\n",
ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
m_freem(m);
return ENOBUFS;
}
pad = mtod(mo, u_char *);
/* Self-describing or random padding ? */
if (!(tdb->tdb_flags & TDBF_RANDOMPADDING))
for (ilen = 0; ilen < padding - 2; ilen++)
pad[ilen] = ilen + 1;
else
arc4random_buf((void *) pad, padding - 2);
/* Fix padding length and Next Protocol in padding itself. */
pad[padding - 2] = padding - 2;
m_copydata(m, protoff, sizeof(u_int8_t), pad + padding - 1);
/* Fix Next Protocol in IPv4/IPv6 header. */
prot = IPPROTO_ESP;
m_copyback(m, protoff, sizeof(u_int8_t), &prot);
/* Get crypto descriptors. */
crp = crypto_getreq(esph && espx ? 2 : 1);
if (crp == NULL) {
m_freem(m);
DPRINTF(("esp_output(): failed to acquire crypto "
"descriptors\n"));
espstat.esps_crypto++;
return ENOBUFS;
}
if (espx) {
crde = crp->crp_desc;
crda = crde->crd_next;
/* Encryption descriptor. */
crde->crd_skip = skip + hlen;
crde->crd_len = m->m_pkthdr.len - (skip + hlen + alen);
crde->crd_flags = CRD_F_ENCRYPT;
crde->crd_inject = skip + hlen - tdb->tdb_ivlen;
if (tdb->tdb_flags & TDBF_HALFIV) {
/* Copy half-iv in the packet. */
m_copyback(m, crde->crd_inject, tdb->tdb_ivlen,
tdb->tdb_iv);
/* Cook half-iv. */
bcopy(tdb->tdb_iv, crde->crd_iv, tdb->tdb_ivlen);
for (ilen = 0; ilen < tdb->tdb_ivlen; ilen++)
crde->crd_iv[tdb->tdb_ivlen + ilen] =
~crde->crd_iv[ilen];
crde->crd_flags |=
CRD_F_IV_PRESENT | CRD_F_IV_EXPLICIT;
}
/* Encryption operation. */
crde->crd_alg = espx->type;
crde->crd_key = tdb->tdb_emxkey;
crde->crd_klen = tdb->tdb_emxkeylen * 8;
/* XXX Rounds ? */
} else
crda = crp->crp_desc;
/* IPsec-specific opaque crypto info. */
tc = malloc(sizeof(*tc), M_XDATA, M_NOWAIT | M_ZERO);
if (tc == NULL) {
m_freem(m);
crypto_freereq(crp);
DPRINTF(("esp_output(): failed to allocate tdb_crypto\n"));
espstat.esps_crypto++;
return ENOBUFS;
}
tc->tc_spi = tdb->tdb_spi;
tc->tc_proto = tdb->tdb_sproto;
bcopy(&tdb->tdb_dst, &tc->tc_dst, sizeof(union sockaddr_union));
/* Crypto operation descriptor. */
crp->crp_ilen = m->m_pkthdr.len; /* Total input length. */
crp->crp_flags = CRYPTO_F_IMBUF;
crp->crp_buf = (caddr_t) m;
crp->crp_callback = (int (*) (struct cryptop *)) esp_output_cb;
crp->crp_opaque = (caddr_t) tc;
crp->crp_sid = tdb->tdb_cryptoid;
if (esph) {
/* Authentication descriptor. */
crda->crd_skip = skip;
crda->crd_len = m->m_pkthdr.len - (skip + alen);
crda->crd_inject = m->m_pkthdr.len - alen;
/* Authentication operation. */
crda->crd_alg = esph->type;
crda->crd_key = tdb->tdb_amxkey;
crda->crd_klen = tdb->tdb_amxkeylen * 8;
}
if ((tdb->tdb_flags & TDBF_SKIPCRYPTO) == 0)
return crypto_dispatch(crp);
else
return esp_output_cb(crp);
}
/*
* Make a copy of an mbuf chain starting "off0" bytes from the beginning,
* continuing for "len" bytes. If len is M_COPYALL, copy to end of mbuf.
* The wait parameter is a choice of M_WAITOK/M_NOWAIT from caller.
* Note that the copy is read-only, because clusters are not copied,
* only their reference counts are incremented.
*/
struct mbuf *
m_copym(struct mbuf *m, int off0, int len, int wait)
{
struct mbuf *n, **np;
int off = off0;
struct mbuf *top;
int copyhdr = 0;
KASSERT(off >= 0, ("m_copym, negative off %d", off));
KASSERT(len >= 0, ("m_copym, negative len %d", len));
MBUF_CHECKSLEEP(wait);
if (off == 0 && m->m_flags & M_PKTHDR)
copyhdr = 1;
while (off > 0) {
KASSERT(m != NULL, ("m_copym, offset > size of mbuf chain"));
if (off < m->m_len)
break;
off -= m->m_len;
m = m->m_next;
}
np = ⊤
top = 0;
while (len > 0) {
if (m == NULL) {
KASSERT(len == M_COPYALL,
("m_copym, length > size of mbuf chain"));
break;
}
if (copyhdr)
n = m_gethdr(wait, m->m_type);
else
n = m_get(wait, m->m_type);
*np = n;
if (n == NULL)
goto nospace;
if (copyhdr) {
if (!m_dup_pkthdr(n, m, wait))
goto nospace;
if (len == M_COPYALL)
n->m_pkthdr.len -= off0;
else
n->m_pkthdr.len = len;
copyhdr = 0;
}
n->m_len = min(len, m->m_len - off);
if (m->m_flags & M_EXT) {
n->m_data = m->m_data + off;
mb_dupcl(n, m);
} else
bcopy(mtod(m, caddr_t)+off, mtod(n, caddr_t),
(u_int)n->m_len);
if (len != M_COPYALL)
len -= n->m_len;
off = 0;
m = m->m_next;
np = &n->m_next;
}
return (top);
nospace:
m_freem(top);
return (NULL);
}
struct mbuf *
ieee80211_ccmp_decrypt(struct ieee80211com *ic, struct mbuf *m0,
struct ieee80211_key *k)
{
struct ieee80211_ccmp_ctx *ctx = k->k_priv;
struct ieee80211_frame *wh;
u_int64_t pn, *prsc;
const u_int8_t *ivp, *src;
u_int8_t *dst;
u_int8_t mic0[IEEE80211_CCMP_MICLEN];
u_int8_t a[16], b[16], s0[16], s[16];
struct mbuf *n0, *m, *n;
int hdrlen, left, moff, noff, len;
u_int16_t ctr;
int i, j;
wh = mtod(m0, struct ieee80211_frame *);
hdrlen = ieee80211_get_hdrlen(wh);
ivp = (u_int8_t *)wh + hdrlen;
if (m0->m_pkthdr.len < hdrlen + IEEE80211_CCMP_HDRLEN +
IEEE80211_CCMP_MICLEN) {
m_freem(m0);
return NULL;
}
/* check that ExtIV bit is set */
if (!(ivp[3] & IEEE80211_WEP_EXTIV)) {
m_freem(m0);
return NULL;
}
/* retrieve last seen packet number for this frame type/priority */
if ((wh->i_fc[0] & IEEE80211_FC0_TYPE_MASK) ==
IEEE80211_FC0_TYPE_DATA) {
u_int8_t tid = ieee80211_has_qos(wh) ?
ieee80211_get_qos(wh) & IEEE80211_QOS_TID : 0;
prsc = &k->k_rsc[tid];
} else /* 11w: management frames have their own counters */
prsc = &k->k_mgmt_rsc;
/* extract the 48-bit PN from the CCMP header */
pn = (u_int64_t)ivp[0] |
(u_int64_t)ivp[1] << 8 |
(u_int64_t)ivp[4] << 16 |
(u_int64_t)ivp[5] << 24 |
(u_int64_t)ivp[6] << 32 |
(u_int64_t)ivp[7] << 40;
if (pn <= *prsc) {
/* replayed frame, discard */
ic->ic_stats.is_ccmp_replays++;
m_freem(m0);
return NULL;
}
MGET(n0, M_DONTWAIT, m0->m_type);
if (n0 == NULL)
goto nospace;
if (m_dup_pkthdr(n0, m0, M_DONTWAIT))
goto nospace;
n0->m_pkthdr.len -= IEEE80211_CCMP_HDRLEN + IEEE80211_CCMP_MICLEN;
n0->m_len = MHLEN;
if (n0->m_pkthdr.len >= MINCLSIZE) {
MCLGET(n0, M_DONTWAIT);
if (n0->m_flags & M_EXT)
n0->m_len = n0->m_ext.ext_size;
}
if (n0->m_len > n0->m_pkthdr.len)
n0->m_len = n0->m_pkthdr.len;
/* construct initial B, A and S_0 blocks */
ieee80211_ccmp_phase1(&ctx->rijndael, wh, pn,
n0->m_pkthdr.len - hdrlen, b, a, s0);
/* copy 802.11 header and clear protected bit */
memcpy(mtod(n0, caddr_t), wh, hdrlen);
wh = mtod(n0, struct ieee80211_frame *);
wh->i_fc[1] &= ~IEEE80211_FC1_PROTECTED;
/* construct S_1 */
ctr = 1;
a[14] = ctr >> 8;
a[15] = ctr & 0xff;
rijndael_encrypt(&ctx->rijndael, a, s);
/* decrypt frame body and compute MIC */
j = 0;
m = m0;
n = n0;
moff = hdrlen + IEEE80211_CCMP_HDRLEN;
noff = hdrlen;
left = n0->m_pkthdr.len - noff;
while (left > 0) {
if (moff == m->m_len) {
/* nothing left to copy from m */
m = m->m_next;
moff = 0;
}
if (noff == n->m_len) {
/* n is full and there's more data to copy */
MGET(n->m_next, M_DONTWAIT, n->m_type);
if (n->m_next == NULL)
goto nospace;
n = n->m_next;
n->m_len = MLEN;
if (left >= MINCLSIZE) {
MCLGET(n, M_DONTWAIT);
if (n->m_flags & M_EXT)
n->m_len = n->m_ext.ext_size;
}
if (n->m_len > left)
n->m_len = left;
noff = 0;
}
len = min(m->m_len - moff, n->m_len - noff);
src = mtod(m, u_int8_t *) + moff;
dst = mtod(n, u_int8_t *) + noff;
for (i = 0; i < len; i++) {
/* decrypt message */
dst[i] = src[i] ^ s[j];
/* update MIC with clear text */
b[j] ^= dst[i];
if (++j < 16)
continue;
/* we have a full block, encrypt MIC */
rijndael_encrypt(&ctx->rijndael, b, b);
/* construct a new S_ctr block */
ctr++;
a[14] = ctr >> 8;
a[15] = ctr & 0xff;
rijndael_encrypt(&ctx->rijndael, a, s);
j = 0;
}
moff += len;
noff += len;
left -= len;
}
if (j != 0) /* partial block, encrypt MIC */
rijndael_encrypt(&ctx->rijndael, b, b);
/* finalize MIC, U := T XOR first-M-bytes( S_0 ) */
for (i = 0; i < IEEE80211_CCMP_MICLEN; i++)
b[i] ^= s0[i];
/* check that it matches the MIC in received frame */
m_copydata(m, moff, IEEE80211_CCMP_MICLEN, mic0);
if (timingsafe_bcmp(mic0, b, IEEE80211_CCMP_MICLEN) != 0) {
ic->ic_stats.is_ccmp_dec_errs++;
m_freem(m0);
m_freem(n0);
return NULL;
}
/* update last seen packet number (MIC is validated) */
*prsc = pn;
m_freem(m0);
return n0;
nospace:
ic->ic_stats.is_rx_nombuf++;
m_freem(m0);
if (n0 != NULL)
m_freem(n0);
return NULL;
}
/*
* Partition an mbuf chain in two pieces, returning the tail --
* all but the first len0 bytes. In case of failure, it returns NULL and
* attempts to restore the chain to its original state.
*
* Note that the resulting mbufs might be read-only, because the new
* mbuf can end up sharing an mbuf cluster with the original mbuf if
* the "breaking point" happens to lie within a cluster mbuf. Use the
* M_WRITABLE() macro to check for this case.
*/
struct mbuf *
m_split(struct mbuf *m0, int len0, int wait)
{
struct mbuf *m, *n;
u_int len = len0, remain;
MBUF_CHECKSLEEP(wait);
for (m = m0; m && len > m->m_len; m = m->m_next)
len -= m->m_len;
if (m == NULL)
return (NULL);
remain = m->m_len - len;
if (m0->m_flags & M_PKTHDR && remain == 0) {
n = m_gethdr(wait, m0->m_type);
if (n == NULL)
return (NULL);
n->m_next = m->m_next;
m->m_next = NULL;
n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
n->m_pkthdr.len = m0->m_pkthdr.len - len0;
m0->m_pkthdr.len = len0;
return (n);
} else if (m0->m_flags & M_PKTHDR) {
n = m_gethdr(wait, m0->m_type);
if (n == NULL)
return (NULL);
n->m_pkthdr.rcvif = m0->m_pkthdr.rcvif;
n->m_pkthdr.len = m0->m_pkthdr.len - len0;
m0->m_pkthdr.len = len0;
if (m->m_flags & M_EXT)
goto extpacket;
if (remain > MHLEN) {
/* m can't be the lead packet */
M_ALIGN(n, 0);
n->m_next = m_split(m, len, wait);
if (n->m_next == NULL) {
(void) m_free(n);
return (NULL);
} else {
n->m_len = 0;
return (n);
}
} else
M_ALIGN(n, remain);
} else if (remain == 0) {
n = m->m_next;
m->m_next = NULL;
return (n);
} else {
n = m_get(wait, m->m_type);
if (n == NULL)
return (NULL);
M_ALIGN(n, remain);
}
extpacket:
if (m->m_flags & M_EXT) {
n->m_data = m->m_data + len;
mb_dupcl(n, m);
} else {
bcopy(mtod(m, caddr_t) + len, mtod(n, caddr_t), remain);
}
n->m_len = remain;
m->m_len = len;
n->m_next = m->m_next;
m->m_next = NULL;
return (n);
}
/*
* Make space for a new header of length hlen at skip bytes
* into the packet. When doing this we allocate new mbufs only
* when absolutely necessary. The mbuf where the new header
* is to go is returned together with an offset into the mbuf.
* If NULL is returned then the mbuf chain may have been modified;
* the caller is assumed to always free the chain.
*/
struct mbuf *
m_makespace(struct mbuf *m0, int skip, int hlen, int *off)
{
struct mbuf *m;
unsigned remain;
KASSERT(m0 != NULL, ("m_dmakespace: null mbuf"));
KASSERT(hlen < MHLEN, ("m_makespace: hlen too big: %u", hlen));
for (m = m0; m && skip > m->m_len; m = m->m_next)
skip -= m->m_len;
if (m == NULL)
return (NULL);
/*
* At this point skip is the offset into the mbuf m
* where the new header should be placed. Figure out
* if there's space to insert the new header. If so,
* and copying the remainder makese sense then do so.
* Otherwise insert a new mbuf in the chain, splitting
* the contents of m as needed.
*/
remain = m->m_len - skip; /* data to move */
if (hlen > M_TRAILINGSPACE(m)) {
struct mbuf *n;
/* XXX code doesn't handle clusters XXX */
KASSERT(remain < MLEN,
("m_makespace: remainder too big: %u", remain));
/*
* Not enough space in m, split the contents
* of m, inserting new mbufs as required.
*
* NB: this ignores mbuf types.
*/
MGET(n, MB_DONTWAIT, MT_DATA);
if (n == NULL)
return (NULL);
n->m_next = m->m_next; /* splice new mbuf */
m->m_next = n;
newipsecstat.ips_mbinserted++;
if (hlen <= M_TRAILINGSPACE(m) + remain) {
/*
* New header fits in the old mbuf if we copy
* the remainder; just do the copy to the new
* mbuf and we're good to go.
*/
memcpy(mtod(n, caddr_t),
mtod(m, caddr_t) + skip, remain);
n->m_len = remain;
m->m_len = skip + hlen;
*off = skip;
} else {
/*
* No space in the old mbuf for the new header.
* Make space in the new mbuf and check the
* remainder'd data fits too. If not then we
* must allocate an additional mbuf (yech).
*/
n->m_len = 0;
if (remain + hlen > M_TRAILINGSPACE(n)) {
struct mbuf *n2;
MGET(n2, MB_DONTWAIT, MT_DATA);
/* NB: new mbuf is on chain, let caller free */
if (n2 == NULL)
return (NULL);
n2->m_len = 0;
memcpy(mtod(n2, caddr_t),
mtod(m, caddr_t) + skip, remain);
n2->m_len = remain;
/* splice in second mbuf */
n2->m_next = n->m_next;
n->m_next = n2;
newipsecstat.ips_mbinserted++;
} else {
memcpy(mtod(n, caddr_t) + hlen,
mtod(m, caddr_t) + skip, remain);
n->m_len += remain;
}
m->m_len -= remain;
n->m_len += hlen;
m = n; /* header is at front ... */
*off = 0; /* ... of new mbuf */
}
} else {
/*
* Copy the remainder to the back of the mbuf
* so there's space to write the new header.
*/
/* XXX can this be memcpy? does it handle overlap? */
ovbcopy(mtod(m, caddr_t) + skip,
mtod(m, caddr_t) + skip + hlen, remain);
m->m_len += hlen;
*off = skip;
}
m0->m_pkthdr.len += hlen; /* adjust packet length */
return m;
}
/*
* TCP input routine, follows pages 65-76 of the
* protocol specification dated September, 1981 very closely.
*/
void
tcp_input(struct mbuf *m, int iphlen, struct socket *inso)
{
struct ip save_ip, *ip;
register struct tcpiphdr *ti;
caddr_t optp = NULL;
int optlen = 0;
int len, tlen, off;
register struct tcpcb *tp = NULL;
register int tiflags;
struct socket *so = NULL;
int todrop, acked, ourfinisacked, needoutput = 0;
int iss = 0;
u_long tiwin;
int ret;
struct ex_list *ex_ptr;
Slirp *slirp;
DEBUG_CALL("tcp_input");
DEBUG_ARGS((dfd, " m = %8lx iphlen = %2d inso = %lx\n",
(long )m, iphlen, (long )inso ));
/*
* If called with m == 0, then we're continuing the connect
*/
if (m == NULL) {
so = inso;
slirp = so->slirp;
/* Re-set a few variables */
tp = sototcpcb(so);
m = so->so_m;
so->so_m = NULL;
ti = so->so_ti;
tiwin = ti->ti_win;
tiflags = ti->ti_flags;
goto cont_conn;
}
slirp = m->slirp;
/*
* Get IP and TCP header together in first mbuf.
* Note: IP leaves IP header in first mbuf.
*/
ti = mtod(m, struct tcpiphdr *);
if (iphlen > sizeof(struct ip )) {
ip_stripoptions(m, (struct mbuf *)0);
iphlen=sizeof(struct ip );
}
/* XXX Check if too short */
/*
* Save a copy of the IP header in case we want restore it
* for sending an ICMP error message in response.
*/
ip=mtod(m, struct ip *);
save_ip = *ip;
save_ip.ip_len+= iphlen;
/*
* Checksum extended TCP header and data.
*/
tlen = ((struct ip *)ti)->ip_len;
tcpiphdr2qlink(ti)->next = tcpiphdr2qlink(ti)->prev = NULL;
memset(&ti->ti_i.ih_mbuf, 0 , sizeof(struct mbuf_ptr));
ti->ti_x1 = 0;
ti->ti_len = htons((uint16_t)tlen);
len = sizeof(struct ip ) + tlen;
if(cksum(m, len)) {
goto drop;
}
/*
* Check that TCP offset makes sense,
* pull out TCP options and adjust length. XXX
*/
off = ti->ti_off << 2;
if (off < sizeof (struct tcphdr) || off > tlen) {
goto drop;
}
tlen -= off;
ti->ti_len = tlen;
if (off > sizeof (struct tcphdr)) {
optlen = off - sizeof (struct tcphdr);
optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
}
tiflags = ti->ti_flags;
/*
* Convert TCP protocol specific fields to host format.
*/
NTOHL(ti->ti_seq);
NTOHL(ti->ti_ack);
NTOHS(ti->ti_win);
NTOHS(ti->ti_urp);
/*
* Drop TCP, IP headers and TCP options.
*/
m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
if (slirp->restricted) {
for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
if (ex_ptr->ex_fport == ti->ti_dport &&
ti->ti_dst.s_addr == ex_ptr->ex_addr.s_addr) {
break;
}
}
if (!ex_ptr)
goto drop;
}
/*
* Locate pcb for segment.
*/
findso:
so = slirp->tcp_last_so;
if (so->so_fport != ti->ti_dport ||
so->so_lport != ti->ti_sport ||
so->so_laddr.s_addr != ti->ti_src.s_addr ||
so->so_faddr.s_addr != ti->ti_dst.s_addr) {
so = solookup(&slirp->tcb, ti->ti_src, ti->ti_sport,
ti->ti_dst, ti->ti_dport);
if (so)
slirp->tcp_last_so = so;
}
/*
* If the state is CLOSED (i.e., TCB does not exist) then
* all data in the incoming segment is discarded.
* If the TCB exists but is in CLOSED state, it is embryonic,
* but should either do a listen or a connect soon.
*
* state == CLOSED means we've done socreate() but haven't
* attached it to a protocol yet...
*
* XXX If a TCB does not exist, and the TH_SYN flag is
* the only flag set, then create a session, mark it
* as if it was LISTENING, and continue...
*/
if (so == NULL) {
if ((tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) != TH_SYN)
goto dropwithreset;
if ((so = socreate(slirp)) == NULL)
goto dropwithreset;
if (tcp_attach(so) < 0) {
free(so); /* Not sofree (if it failed, it's not insqued) */
goto dropwithreset;
}
sbreserve(&so->so_snd, TCP_SNDSPACE);
sbreserve(&so->so_rcv, TCP_RCVSPACE);
so->so_laddr = ti->ti_src;
so->so_lport = ti->ti_sport;
so->so_faddr = ti->ti_dst;
so->so_fport = ti->ti_dport;
if ((so->so_iptos = tcp_tos(so)) == 0)
so->so_iptos = ((struct ip *)ti)->ip_tos;
tp = sototcpcb(so);
tp->t_state = TCPS_LISTEN;
}
/*
* If this is a still-connecting socket, this probably
* a retransmit of the SYN. Whether it's a retransmit SYN
* or something else, we nuke it.
*/
if (so->so_state & SS_ISFCONNECTING)
goto drop;
tp = sototcpcb(so);
/* XXX Should never fail */
if (tp == NULL)
goto dropwithreset;
if (tp->t_state == TCPS_CLOSED)
goto drop;
tiwin = ti->ti_win;
/*
* Segment received on connection.
* Reset idle time and keep-alive timer.
*/
tp->t_idle = 0;
if (SO_OPTIONS)
tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL;
else
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE;
/*
* Process options if not in LISTEN state,
* else do it below (after getting remote address).
*/
if (optp && tp->t_state != TCPS_LISTEN)
tcp_dooptions(tp, (u_char *)optp, optlen, ti);
/*
* Header prediction: check for the two common cases
* of a uni-directional data xfer. If the packet has
* no control flags, is in-sequence, the window didn't
* change and we're not retransmitting, it's a
* candidate. If the length is zero and the ack moved
* forward, we're the sender side of the xfer. Just
* free the data acked & wake any higher level process
* that was blocked waiting for space. If the length
* is non-zero and the ack didn't move, we're the
* receiver side. If we're getting packets in-order
* (the reassembly queue is empty), add the data to
* the socket buffer and note that we need a delayed ack.
*
* XXX Some of these tests are not needed
* eg: the tiwin == tp->snd_wnd prevents many more
* predictions.. with no *real* advantage..
*/
if (tp->t_state == TCPS_ESTABLISHED &&
(tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
ti->ti_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
tp->snd_nxt == tp->snd_max) {
if (ti->ti_len == 0) {
if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
tp->snd_cwnd >= tp->snd_wnd) {
/*
* this is a pure ack for outstanding data.
*/
if (tp->t_rtt &&
SEQ_GT(ti->ti_ack, tp->t_rtseq))
tcp_xmit_timer(tp, tp->t_rtt);
acked = ti->ti_ack - tp->snd_una;
sbdrop(&so->so_snd, acked);
tp->snd_una = ti->ti_ack;
m_free(m);
/*
* If all outstanding data are acked, stop
* retransmit timer, otherwise restart timer
* using current (possibly backed-off) value.
* If process is waiting for space,
* wakeup/selwakeup/signal. If data
* are ready to send, let tcp_output
* decide between more output or persist.
*/
if (tp->snd_una == tp->snd_max)
tp->t_timer[TCPT_REXMT] = 0;
else if (tp->t_timer[TCPT_PERSIST] == 0)
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
/*
* This is called because sowwakeup might have
* put data into so_snd. Since we don't so sowwakeup,
* we don't need this.. XXX???
*/
if (so->so_snd.sb_cc)
(void) tcp_output(tp);
return;
}
} else if (ti->ti_ack == tp->snd_una &&
tcpfrag_list_empty(tp) &&
ti->ti_len <= sbspace(&so->so_rcv)) {
/*
* this is a pure, in-sequence data packet
* with nothing on the reassembly queue and
* we have enough buffer space to take it.
*/
tp->rcv_nxt += ti->ti_len;
/*
* Add data to socket buffer.
*/
if (so->so_emu) {
if (tcp_emu(so,m)) sbappend(so, m);
} else
sbappend(so, m);
/*
* If this is a short packet, then ACK now - with Nagel
* congestion avoidance sender won't send more until
* he gets an ACK.
*
* It is better to not delay acks at all to maximize
* TCP throughput. See RFC 2581.
*/
tp->t_flags |= TF_ACKNOW;
tcp_output(tp);
return;
}
} /* header prediction */
/*
* Calculate amount of space in receive window,
* and then do TCP input processing.
* Receive window is amount of space in rcv queue,
* but not less than advertised window.
*/
{ int win;
win = sbspace(&so->so_rcv);
if (win < 0)
win = 0;
tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
}
switch (tp->t_state) {
/*
* If the state is LISTEN then ignore segment if it contains an RST.
* If the segment contains an ACK then it is bad and send a RST.
* If it does not contain a SYN then it is not interesting; drop it.
* Don't bother responding if the destination was a broadcast.
* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
* tp->iss, and send a segment:
* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
* Fill in remote peer address fields if not previously specified.
* Enter SYN_RECEIVED state, and process any other fields of this
* segment in this state.
*/
case TCPS_LISTEN: {
if (tiflags & TH_RST)
goto drop;
if (tiflags & TH_ACK)
goto dropwithreset;
if ((tiflags & TH_SYN) == 0)
goto drop;
/*
* This has way too many gotos...
* But a bit of spaghetti code never hurt anybody :)
*/
/*
* If this is destined for the control address, then flag to
* tcp_ctl once connected, otherwise connect
*/
if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
slirp->vnetwork_addr.s_addr) {
if (so->so_faddr.s_addr != slirp->vhost_addr.s_addr &&
so->so_faddr.s_addr != slirp->vnameserver_addr.s_addr) {
/* May be an add exec */
for (ex_ptr = slirp->exec_list; ex_ptr;
ex_ptr = ex_ptr->ex_next) {
if(ex_ptr->ex_fport == so->so_fport &&
so->so_faddr.s_addr == ex_ptr->ex_addr.s_addr) {
so->so_state |= SS_CTL;
break;
}
}
if (so->so_state & SS_CTL) {
goto cont_input;
}
}
/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
}
if (so->so_emu & EMU_NOCONNECT) {
so->so_emu &= ~EMU_NOCONNECT;
goto cont_input;
}
if((tcp_fconnect(so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
u_char code=ICMP_UNREACH_NET;
DEBUG_MISC((dfd, " tcp fconnect errno = %d-%s\n",
errno,strerror(errno)));
if(errno == ECONNREFUSED) {
/* ACK the SYN, send RST to refuse the connection */
tcp_respond(tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
TH_RST|TH_ACK);
} else {
if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
HTONL(ti->ti_seq); /* restore tcp header */
HTONL(ti->ti_ack);
HTONS(ti->ti_win);
HTONS(ti->ti_urp);
m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
*ip=save_ip;
icmp_error(m, ICMP_UNREACH,code, 0,strerror(errno));
}
tcp_close(tp);
m_free(m);
} else {
/*
* Haven't connected yet, save the current mbuf
* and ti, and return
* XXX Some OS's don't tell us whether the connect()
* succeeded or not. So we must time it out.
*/
so->so_m = m;
so->so_ti = ti;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
tp->t_state = TCPS_SYN_RECEIVED;
tcp_template(tp);
}
return;
cont_conn:
/* m==NULL
* Check if the connect succeeded
*/
if (so->so_state & SS_NOFDREF) {
tp = tcp_close(tp);
goto dropwithreset;
}
cont_input:
tcp_template(tp);
if (optp)
tcp_dooptions(tp, (u_char *)optp, optlen, ti);
if (iss)
tp->iss = iss;
else
tp->iss = slirp->tcp_iss;
slirp->tcp_iss += TCP_ISSINCR/2;
tp->irs = ti->ti_seq;
tcp_sendseqinit(tp);
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
tp->t_state = TCPS_SYN_RECEIVED;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
goto trimthenstep6;
} /* case TCPS_LISTEN */
/*
* If the state is SYN_SENT:
* if seg contains an ACK, but not for our SYN, drop the input.
* if seg contains a RST, then drop the connection.
* if seg does not contain SYN, then drop it.
* Otherwise this is an acceptable SYN segment
* initialize tp->rcv_nxt and tp->irs
* if seg contains ack then advance tp->snd_una
* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
* arrange for segment to be acked (eventually)
* continue processing rest of data/controls, beginning with URG
*/
case TCPS_SYN_SENT:
if ((tiflags & TH_ACK) &&
(SEQ_LEQ(ti->ti_ack, tp->iss) ||
SEQ_GT(ti->ti_ack, tp->snd_max)))
goto dropwithreset;
if (tiflags & TH_RST) {
if (tiflags & TH_ACK) {
tcp_drop(tp, 0); /* XXX Check t_softerror! */
}
goto drop;
}
if ((tiflags & TH_SYN) == 0)
goto drop;
if (tiflags & TH_ACK) {
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
}
tp->t_timer[TCPT_REXMT] = 0;
tp->irs = ti->ti_seq;
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
soisfconnected(so);
tp->t_state = TCPS_ESTABLISHED;
(void) tcp_reass(tp, (struct tcpiphdr *)0,
(struct mbuf *)0);
/*
* if we didn't have to retransmit the SYN,
* use its rtt as our initial srtt & rtt var.
*/
if (tp->t_rtt)
tcp_xmit_timer(tp, tp->t_rtt);
} else
tp->t_state = TCPS_SYN_RECEIVED;
trimthenstep6:
/*
* Advance ti->ti_seq to correspond to first data byte.
* If data, trim to stay within window,
* dropping FIN if necessary.
*/
ti->ti_seq++;
if (ti->ti_len > tp->rcv_wnd) {
todrop = ti->ti_len - tp->rcv_wnd;
m_adj(m, -todrop);
ti->ti_len = tp->rcv_wnd;
tiflags &= ~TH_FIN;
}
tp->snd_wl1 = ti->ti_seq - 1;
tp->rcv_up = ti->ti_seq;
goto step6;
} /* switch tp->t_state */
/*
* States other than LISTEN or SYN_SENT.
* Check that at least some bytes of segment are within
* receive window. If segment begins before rcv_nxt,
* drop leading data (and SYN); if nothing left, just ack.
*/
todrop = tp->rcv_nxt - ti->ti_seq;
if (todrop > 0) {
if (tiflags & TH_SYN) {
tiflags &= ~TH_SYN;
ti->ti_seq++;
if (ti->ti_urp > 1)
ti->ti_urp--;
else
tiflags &= ~TH_URG;
todrop--;
}
/*
* Following if statement from Stevens, vol. 2, p. 960.
*/
if (todrop > ti->ti_len
|| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
/*
* Any valid FIN must be to the left of the window.
* At this point the FIN must be a duplicate or out
* of sequence; drop it.
*/
tiflags &= ~TH_FIN;
/*
* Send an ACK to resynchronize and drop any data.
* But keep on processing for RST or ACK.
*/
tp->t_flags |= TF_ACKNOW;
todrop = ti->ti_len;
}
m_adj(m, todrop);
ti->ti_seq += todrop;
ti->ti_len -= todrop;
if (ti->ti_urp > todrop)
ti->ti_urp -= todrop;
else {
tiflags &= ~TH_URG;
ti->ti_urp = 0;
}
}
/*
* If new data are received on a connection after the
* user processes are gone, then RST the other end.
*/
if ((so->so_state & SS_NOFDREF) &&
tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
tp = tcp_close(tp);
goto dropwithreset;
}
/*
* If segment ends after window, drop trailing data
* (and PUSH and FIN); if nothing left, just ACK.
*/
todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
if (todrop > 0) {
if (todrop >= ti->ti_len) {
/*
* If a new connection request is received
* while in TIME_WAIT, drop the old connection
* and start over if the sequence numbers
* are above the previous ones.
*/
if (tiflags & TH_SYN &&
tp->t_state == TCPS_TIME_WAIT &&
SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
iss = tp->rcv_nxt + TCP_ISSINCR;
tp = tcp_close(tp);
goto findso;
}
/*
* If window is closed can only take segments at
* window edge, and have to drop data and PUSH from
* incoming segments. Continue processing, but
* remember to ack. Otherwise, drop segment
* and ack.
*/
if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
tp->t_flags |= TF_ACKNOW;
} else {
goto dropafterack;
}
}
m_adj(m, -todrop);
ti->ti_len -= todrop;
tiflags &= ~(TH_PUSH|TH_FIN);
}
/*
* If the RST bit is set examine the state:
* SYN_RECEIVED STATE:
* If passive open, return to LISTEN state.
* If active open, inform user that connection was refused.
* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
* Inform user that connection was reset, and close tcb.
* CLOSING, LAST_ACK, TIME_WAIT STATES
* Close the tcb.
*/
if (tiflags&TH_RST) switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
tp->t_state = TCPS_CLOSED;
tcp_close(tp);
goto drop;
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
tcp_close(tp);
goto drop;
}
/*
* If a SYN is in the window, then this is an
* error and we send an RST and drop the connection.
*/
if (tiflags & TH_SYN) {
tp = tcp_drop(tp,0);
goto dropwithreset;
}
/*
* If the ACK bit is off we drop the segment and return.
*/
if ((tiflags & TH_ACK) == 0) goto drop;
/*
* Ack processing.
*/
switch (tp->t_state) {
/*
* In SYN_RECEIVED state if the ack ACKs our SYN then enter
* ESTABLISHED state and continue processing, otherwise
* send an RST. una<=ack<=max
*/
case TCPS_SYN_RECEIVED:
if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
SEQ_GT(ti->ti_ack, tp->snd_max))
goto dropwithreset;
tp->t_state = TCPS_ESTABLISHED;
/*
* The sent SYN is ack'ed with our sequence number +1
* The first data byte already in the buffer will get
* lost if no correction is made. This is only needed for
* SS_CTL since the buffer is empty otherwise.
* tp->snd_una++; or:
*/
tp->snd_una=ti->ti_ack;
if (so->so_state & SS_CTL) {
/* So tcp_ctl reports the right state */
ret = tcp_ctl(so);
if (ret == 1) {
soisfconnected(so);
so->so_state &= ~SS_CTL; /* success XXX */
} else if (ret == 2) {
so->so_state &= SS_PERSISTENT_MASK;
so->so_state |= SS_NOFDREF; /* CTL_CMD */
} else {
needoutput = 1;
tp->t_state = TCPS_FIN_WAIT_1;
}
} else {
soisfconnected(so);
}
(void) tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0);
tp->snd_wl1 = ti->ti_seq - 1;
/* Avoid ack processing; snd_una==ti_ack => dup ack */
goto synrx_to_est;
/* fall into ... */
/*
* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
* ACKs. If the ack is in the range
* tp->snd_una < ti->ti_ack <= tp->snd_max
* then advance tp->snd_una to ti->ti_ack and drop
* data from the retransmission queue. If this ACK reflects
* more up to date window information we update our window information.
*/
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
DEBUG_MISC((dfd, " dup ack m = %lx so = %lx\n",
(long )m, (long )so));
/*
* If we have outstanding data (other than
* a window probe), this is a completely
* duplicate ack (ie, window info didn't
* change), the ack is the biggest we've
* seen and we've seen exactly our rexmt
* threshold of them, assume a packet
* has been dropped and retransmit it.
* Kludge snd_nxt & the congestion
* window so we send only this one
* packet.
*
* We know we're losing at the current
* window size so do congestion avoidance
* (set ssthresh to half the current window
* and pull our congestion window back to
* the new ssthresh).
*
* Dup acks mean that packets have left the
* network (they're now cached at the receiver)
* so bump cwnd by the amount in the receiver
* to keep a constant cwnd packets in the
* network.
*/
if (tp->t_timer[TCPT_REXMT] == 0 ||
ti->ti_ack != tp->snd_una)
tp->t_dupacks = 0;
else if (++tp->t_dupacks == TCPREXMTTHRESH) {
tcp_seq onxt = tp->snd_nxt;
u_int win =
min(tp->snd_wnd, tp->snd_cwnd) / 2 /
tp->t_maxseg;
if (win < 2)
win = 2;
tp->snd_ssthresh = win * tp->t_maxseg;
tp->t_timer[TCPT_REXMT] = 0;
tp->t_rtt = 0;
tp->snd_nxt = ti->ti_ack;
tp->snd_cwnd = tp->t_maxseg;
(void) tcp_output(tp);
tp->snd_cwnd = tp->snd_ssthresh +
tp->t_maxseg * tp->t_dupacks;
if (SEQ_GT(onxt, tp->snd_nxt))
tp->snd_nxt = onxt;
goto drop;
} else if (tp->t_dupacks > TCPREXMTTHRESH) {
tp->snd_cwnd += tp->t_maxseg;
(void) tcp_output(tp);
goto drop;
}
} else
tp->t_dupacks = 0;
break;
}
synrx_to_est:
/*
* If the congestion window was inflated to account
* for the other side's cached packets, retract it.
*/
if (tp->t_dupacks > TCPREXMTTHRESH &&
tp->snd_cwnd > tp->snd_ssthresh)
tp->snd_cwnd = tp->snd_ssthresh;
tp->t_dupacks = 0;
if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
goto dropafterack;
}
acked = ti->ti_ack - tp->snd_una;
/*
* If transmit timer is running and timed sequence
* number was acked, update smoothed round trip time.
* Since we now have an rtt measurement, cancel the
* timer backoff (cf., Phil Karn's retransmit alg.).
* Recompute the initial retransmit timer.
*/
if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
tcp_xmit_timer(tp,tp->t_rtt);
/*
* If all outstanding data is acked, stop retransmit
* timer and remember to restart (more output or persist).
* If there is more data to be acked, restart retransmit
* timer, using current (possibly backed-off) value.
*/
if (ti->ti_ack == tp->snd_max) {
tp->t_timer[TCPT_REXMT] = 0;
needoutput = 1;
} else if (tp->t_timer[TCPT_PERSIST] == 0)
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
/*
* When new data is acked, open the congestion window.
* If the window gives us less than ssthresh packets
* in flight, open exponentially (maxseg per packet).
* Otherwise open linearly: maxseg per window
* (maxseg^2 / cwnd per packet).
*/
{
register u_int cw = tp->snd_cwnd;
register u_int incr = tp->t_maxseg;
if (cw > tp->snd_ssthresh)
incr = incr * incr / cw;
tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
}
if (acked > so->so_snd.sb_cc) {
tp->snd_wnd -= so->so_snd.sb_cc;
sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
ourfinisacked = 1;
} else {
sbdrop(&so->so_snd, acked);
tp->snd_wnd -= acked;
ourfinisacked = 0;
}
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
switch (tp->t_state) {
/*
* In FIN_WAIT_1 STATE in addition to the processing
* for the ESTABLISHED state if our FIN is now acknowledged
* then enter FIN_WAIT_2.
*/
case TCPS_FIN_WAIT_1:
if (ourfinisacked) {
/*
* If we can't receive any more
* data, then closing user can proceed.
* Starting the timer is contrary to the
* specification, but if we don't get a FIN
* we'll hang forever.
*/
if (so->so_state & SS_FCANTRCVMORE) {
tp->t_timer[TCPT_2MSL] = TCP_MAXIDLE;
}
tp->t_state = TCPS_FIN_WAIT_2;
}
break;
/*
* In CLOSING STATE in addition to the processing for
* the ESTABLISHED state if the ACK acknowledges our FIN
* then enter the TIME-WAIT state, otherwise ignore
* the segment.
*/
case TCPS_CLOSING:
if (ourfinisacked) {
tp->t_state = TCPS_TIME_WAIT;
tcp_canceltimers(tp);
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
}
break;
/*
* In LAST_ACK, we may still be waiting for data to drain
* and/or to be acked, as well as for the ack of our FIN.
* If our FIN is now acknowledged, delete the TCB,
* enter the closed state and return.
*/
case TCPS_LAST_ACK:
if (ourfinisacked) {
tcp_close(tp);
goto drop;
}
break;
/*
* In TIME_WAIT state the only thing that should arrive
* is a retransmission of the remote FIN. Acknowledge
* it and restart the finack timer.
*/
case TCPS_TIME_WAIT:
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
goto dropafterack;
}
} /* switch(tp->t_state) */
step6:
/*
* Update window information.
* Don't look at window if no ACK: TAC's send garbage on first SYN.
*/
if ((tiflags & TH_ACK) &&
(SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
tp->snd_wnd = tiwin;
tp->snd_wl1 = ti->ti_seq;
tp->snd_wl2 = ti->ti_ack;
if (tp->snd_wnd > tp->max_sndwnd)
tp->max_sndwnd = tp->snd_wnd;
needoutput = 1;
}
/*
* Process segments with URG.
*/
if ((tiflags & TH_URG) && ti->ti_urp &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
/*
* This is a kludge, but if we receive and accept
* random urgent pointers, we'll crash in
* soreceive. It's hard to imagine someone
* actually wanting to send this much urgent data.
*/
if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
ti->ti_urp = 0;
tiflags &= ~TH_URG;
goto dodata;
}
/*
* If this segment advances the known urgent pointer,
* then mark the data stream. This should not happen
* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
* a FIN has been received from the remote side.
* In these states we ignore the URG.
*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section as the original
* spec states (in one of two places).
*/
if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
tp->rcv_up = ti->ti_seq + ti->ti_urp;
so->so_urgc = so->so_rcv.sb_cc +
(tp->rcv_up - tp->rcv_nxt); /* -1; */
tp->rcv_up = ti->ti_seq + ti->ti_urp;
}
} else
/*
* If no out of band data is expected,
* pull receive urgent pointer along
* with the receive window.
*/
if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
tp->rcv_up = tp->rcv_nxt;
dodata:
/*
* Process the segment text, merging it into the TCP sequencing queue,
* and arranging for acknowledgment of receipt if necessary.
* This process logically involves adjusting tp->rcv_wnd as data
* is presented to the user (this happens in tcp_usrreq.c,
* case PRU_RCVD). If a FIN has already been received on this
* connection then we just ignore the text.
*/
if ((ti->ti_len || (tiflags&TH_FIN)) &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
TCP_REASS(tp, ti, m, so, tiflags);
} else {
m_free(m);
tiflags &= ~TH_FIN;
}
/*
* If FIN is received ACK the FIN and let the user know
* that the connection is closing.
*/
if (tiflags & TH_FIN) {
if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
/*
* If we receive a FIN we can't send more data,
* set it SS_FDRAIN
* Shutdown the socket if there is no rx data in the
* buffer.
* soread() is called on completion of shutdown() and
* will got to TCPS_LAST_ACK, and use tcp_output()
* to send the FIN.
*/
sofwdrain(so);
tp->t_flags |= TF_ACKNOW;
tp->rcv_nxt++;
}
switch (tp->t_state) {
/*
* In SYN_RECEIVED and ESTABLISHED STATES
* enter the CLOSE_WAIT state.
*/
case TCPS_SYN_RECEIVED:
case TCPS_ESTABLISHED:
if(so->so_emu == EMU_CTL) /* no shutdown on socket */
tp->t_state = TCPS_LAST_ACK;
else
tp->t_state = TCPS_CLOSE_WAIT;
break;
/*
* If still in FIN_WAIT_1 STATE FIN has not been acked so
* enter the CLOSING state.
*/
case TCPS_FIN_WAIT_1:
tp->t_state = TCPS_CLOSING;
break;
/*
* In FIN_WAIT_2 state enter the TIME_WAIT state,
* starting the time-wait timer, turning off the other
* standard timers.
*/
case TCPS_FIN_WAIT_2:
tp->t_state = TCPS_TIME_WAIT;
tcp_canceltimers(tp);
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
break;
/*
* In TIME_WAIT state restart the 2 MSL time_wait timer.
*/
case TCPS_TIME_WAIT:
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
break;
}
}
/*
* If this is a small packet, then ACK now - with Nagel
* congestion avoidance sender won't send more until
* he gets an ACK.
*
* See above.
*/
if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
tp->t_flags |= TF_ACKNOW;
}
/*
* Return any desired output.
*/
if (needoutput || (tp->t_flags & TF_ACKNOW)) {
(void) tcp_output(tp);
}
return;
dropafterack:
/*
* Generate an ACK dropping incoming segment if it occupies
* sequence space, where the ACK reflects our state.
*/
if (tiflags & TH_RST)
goto drop;
m_free(m);
tp->t_flags |= TF_ACKNOW;
(void) tcp_output(tp);
return;
dropwithreset:
/* reuses m if m!=NULL, m_free() unnecessary */
if (tiflags & TH_ACK)
tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
else {
if (tiflags & TH_SYN) ti->ti_len++;
tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
TH_RST|TH_ACK);
}
return;
drop:
/*
* Drop space held by incoming segment and return.
*/
m_free(m);
return;
}
/*
* copies mbuf chain to the uio scatter/gather list
*/
int
nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, caddr_t *dpos)
{
char *mbufcp, *uiocp;
int xfer, left, len;
struct mbuf *mp;
long uiosiz, rem;
int error = 0;
mp = *mrep;
mbufcp = *dpos;
len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
rem = nfsm_rndup(siz)-siz;
while (siz > 0) {
if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
return (EFBIG);
left = uiop->uio_iov->iov_len;
uiocp = uiop->uio_iov->iov_base;
if (left > siz)
left = siz;
uiosiz = left;
while (left > 0) {
while (len == 0) {
mp = mp->m_next;
if (mp == NULL)
return (EBADRPC);
mbufcp = mtod(mp, caddr_t);
len = mp->m_len;
}
xfer = (left > len) ? len : left;
#ifdef notdef
/* Not Yet.. */
if (uiop->uio_iov->iov_op != NULL)
(*(uiop->uio_iov->iov_op))
(mbufcp, uiocp, xfer);
else
#endif
if (uiop->uio_segflg == UIO_SYSSPACE)
bcopy(mbufcp, uiocp, xfer);
else
copyout(mbufcp, uiocp, xfer);
left -= xfer;
len -= xfer;
mbufcp += xfer;
uiocp += xfer;
uiop->uio_offset += xfer;
uiop->uio_resid -= xfer;
}
if (uiop->uio_iov->iov_len <= siz) {
uiop->uio_iovcnt--;
uiop->uio_iov++;
} else {
uiop->uio_iov->iov_base =
(char *)uiop->uio_iov->iov_base + uiosiz;
uiop->uio_iov->iov_len -= uiosiz;
}
siz -= uiosiz;
}
*dpos = mbufcp;
*mrep = mp;
if (rem > 0) {
if (len < rem)
error = nfs_adv(mrep, dpos, rem, len);
else
*dpos += rem;
}
return (error);
}
/*
* The output routine. Takes a packet and encapsulates it in the protocol
* given by sc->g_proto. See also RFC 1701 and RFC 2004
*/
static int
gre_output_serialized(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
struct rtentry *rt)
{
int error = 0;
struct gre_softc *sc = ifp->if_softc;
struct greip *gh;
struct ip *ip;
u_short etype = 0;
struct mobile_h mob_h;
struct route *ro;
struct sockaddr_in *ro_dst;
ASSERT_NETISR_NCPUS(mycpuid);
/*
* gre may cause infinite recursion calls when misconfigured.
* We'll prevent this by introducing upper limit.
*/
if (++(sc->called) > max_gre_nesting) {
kprintf("%s: gre_output: recursively called too many "
"times(%d)\n", if_name(&sc->sc_if), sc->called);
m_freem(m);
error = EIO; /* is there better errno? */
goto end;
}
if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) == 0 ||
sc->g_src.s_addr == INADDR_ANY || sc->g_dst.s_addr == INADDR_ANY) {
m_freem(m);
error = ENETDOWN;
goto end;
}
ro = &sc->route_pcpu[mycpuid];
ro_dst = (struct sockaddr_in *)&ro->ro_dst;
if (ro->ro_rt != NULL &&
((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
ro_dst->sin_addr.s_addr != sc->g_dst.s_addr)) {
RTFREE(ro->ro_rt);
ro->ro_rt = NULL;
}
if (ro->ro_rt == NULL) {
error = gre_compute_route(sc, ro);
if (error) {
m_freem(m);
goto end;
}
}
gh = NULL;
ip = NULL;
if (ifp->if_bpf) {
bpf_gettoken();
if (ifp->if_bpf) {
uint32_t af = dst->sa_family;
bpf_ptap(ifp->if_bpf, m, &af, sizeof(af));
}
bpf_reltoken();
}
m->m_flags &= ~(M_BCAST|M_MCAST);
if (sc->g_proto == IPPROTO_MOBILE) {
if (dst->sa_family == AF_INET) {
struct mbuf *m0;
int msiz;
ip = mtod(m, struct ip *);
/*
* RFC2004 specifies that fragmented datagrams shouldn't
* be encapsulated.
*/
if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
m_freem(m);
error = EINVAL; /* is there better errno? */
goto end;
}
memset(&mob_h, 0, MOB_H_SIZ_L);
mob_h.proto = (ip->ip_p) << 8;
mob_h.odst = ip->ip_dst.s_addr;
ip->ip_dst.s_addr = sc->g_dst.s_addr;
/*
* If the packet comes from our host, we only change
* the destination address in the IP header.
* Else we also need to save and change the source
*/
if (in_hosteq(ip->ip_src, sc->g_src)) {
msiz = MOB_H_SIZ_S;
} else {
mob_h.proto |= MOB_H_SBIT;
mob_h.osrc = ip->ip_src.s_addr;
ip->ip_src.s_addr = sc->g_src.s_addr;
msiz = MOB_H_SIZ_L;
}
mob_h.proto = htons(mob_h.proto);
mob_h.hcrc = gre_in_cksum((u_short *)&mob_h, msiz);
if ((m->m_data - msiz) < m->m_pktdat) {
/* need new mbuf */
MGETHDR(m0, M_NOWAIT, MT_HEADER);
if (m0 == NULL) {
m_freem(m);
error = ENOBUFS;
goto end;
}
m0->m_next = m;
m->m_data += sizeof(struct ip);
m->m_len -= sizeof(struct ip);
m0->m_pkthdr.len = m->m_pkthdr.len + msiz;
m0->m_len = msiz + sizeof(struct ip);
m0->m_data += max_linkhdr;
memcpy(mtod(m0, caddr_t), (caddr_t)ip,
sizeof(struct ip));
m = m0;
} else { /* we have some space left in the old one */
m->m_data -= msiz;
m->m_len += msiz;
m->m_pkthdr.len += msiz;
bcopy(ip, mtod(m, caddr_t),
sizeof(struct ip));
}
ip = mtod(m, struct ip *);
memcpy((caddr_t)(ip + 1), &mob_h, (unsigned)msiz);
ip->ip_len = ntohs(ip->ip_len) + msiz;
} else { /* AF_INET */
m_freem(m);
error = EINVAL;
goto end;
}
} else if (sc->g_proto == IPPROTO_GRE) {
/*
* Save an outbound packet for a node in power-save sleep state.
* The new packet is placed on the node's saved queue, and the TIM
* is changed, if necessary.
*/
int
ieee80211_pwrsave(struct ieee80211_node *ni, struct mbuf *m)
{
struct ieee80211_psq *psq = &ni->ni_psq;
struct ieee80211vap *vap = ni->ni_vap;
struct ieee80211com *ic = ni->ni_ic;
struct ieee80211_psq_head *qhead;
int qlen, age;
IEEE80211_PSQ_LOCK(psq);
if (psq->psq_len >= psq->psq_maxlen) {
psq->psq_drops++;
IEEE80211_PSQ_UNLOCK(psq);
IEEE80211_NOTE(vap, IEEE80211_MSG_ANY, ni,
"pwr save q overflow, drops %d (size %d)",
psq->psq_drops, psq->psq_len);
#ifdef IEEE80211_DEBUG
if (ieee80211_msg_dumppkts(vap))
ieee80211_dump_pkt(ni->ni_ic, mtod(m, caddr_t),
m->m_len, -1, -1);
#endif
psq_mfree(m);
return ENOSPC;
}
/*
* Tag the frame with it's expiry time and insert it in
* the appropriate queue. The aging interval is 4 times
* the listen interval specified by the station. Frames
* that sit around too long are reclaimed using this
* information.
*/
/* TU -> secs. XXX handle overflow? */
age = IEEE80211_TU_TO_MS((ni->ni_intval * ic->ic_bintval) << 2) / 1000;
/*
* Encapsulated frames go on the high priority queue,
* other stuff goes on the low priority queue. We use
* this to order frames returned out of the driver
* ahead of frames we collect in ieee80211_start.
*/
if (m->m_flags & M_ENCAP)
qhead = &psq->psq_head[0];
else
qhead = &psq->psq_head[1];
if (qhead->tail == NULL) {
struct mbuf *mh;
qhead->head = m;
/*
* Take care to adjust age when inserting the first
* frame of a queue and the other queue already has
* frames. We need to preserve the age difference
* relationship so ieee80211_node_psq_age works.
*/
if (qhead == &psq->psq_head[1]) {
mh = psq->psq_head[0].head;
if (mh != NULL)
age-= M_AGE_GET(mh);
} else {
mh = psq->psq_head[1].head;
if (mh != NULL) {
int nage = M_AGE_GET(mh) - age;
/* XXX is clamping to zero good 'nuf? */
M_AGE_SET(mh, nage < 0 ? 0 : nage);
}
}
} else {
qhead->tail->m_nextpkt = m;
age -= M_AGE_GET(qhead->head);
}
KASSERT(age >= 0, ("age %d", age));
M_AGE_SET(m, age);
m->m_nextpkt = NULL;
qhead->tail = m;
qhead->len++;
qlen = ++(psq->psq_len);
IEEE80211_PSQ_UNLOCK(psq);
IEEE80211_NOTE(vap, IEEE80211_MSG_POWER, ni,
"save frame with age %d, %u now queued", age, qlen);
if (qlen == 1 && vap->iv_set_tim != NULL)
vap->iv_set_tim(ni, 1);
return 0;
}
int
ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
{
struct tdb_ident *tdbi;
struct m_tag *mtag;
struct secasvar *sav;
struct secasindex *saidx;
int error;
IPSEC_ASSERT(m != NULL, ("null mbuf"));
IPSEC_ASSERT(isr != NULL, ("null ISR"));
IPSEC_ASSERT(isr->sp != NULL, ("NULL isr->sp"));
sav = isr->sav;
IPSEC_ASSERT(sav != NULL, ("null SA"));
IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
saidx = &sav->sah->saidx;
switch (saidx->dst.sa.sa_family) {
#ifdef INET
case AF_INET:
/* Fix the header length, for AH processing. */
mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
/* Fix the header length, for AH processing. */
if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
error = ENXIO;
goto bad;
}
if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
/* No jumbogram support. */
error = ENXIO; /*?*/
goto bad;
}
mtod(m, struct ip6_hdr *)->ip6_plen =
htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
break;
#endif /* INET6 */
default:
DPRINTF(("%s: unknown protocol family %u\n", __func__,
saidx->dst.sa.sa_family));
error = ENXIO;
goto bad;
}
/*
* Add a record of what we've done or what needs to be done to the
* packet.
*/
mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
sizeof(struct tdb_ident), M_NOWAIT);
if (mtag == NULL) {
DPRINTF(("%s: could not get packet tag\n", __func__));
error = ENOMEM;
goto bad;
}
tdbi = (struct tdb_ident *)(mtag + 1);
tdbi->dst = saidx->dst;
tdbi->proto = saidx->proto;
tdbi->spi = sav->spi;
m_tag_prepend(m, mtag);
key_sa_recordxfer(sav, m); /* record data transfer */
/*
* If there's another (bundled) SA to apply, do so.
* Note that this puts a burden on the kernel stack size.
* If this is a problem we'll need to introduce a queue
* to set the packet on so we can unwind the stack before
* doing further processing.
*/
if (isr->next) {
/* XXX-BZ currently only support same AF bundles. */
switch (saidx->dst.sa.sa_family) {
#ifdef INET
case AF_INET:
IPSECSTAT_INC(ips_out_bundlesa);
return (ipsec4_process_packet(m, isr->next));
/* NOTREACHED */
#endif
#ifdef notyet
#ifdef INET6
case AF_INET6:
/* XXX */
IPSEC6STAT_INC(ips_out_bundlesa);
return (ipsec6_process_packet(m, isr->next));
/* NOTREACHED */
#endif /* INET6 */
#endif
default:
DPRINTF(("%s: unknown protocol family %u\n", __func__,
saidx->dst.sa.sa_family));
error = ENXIO;
goto bad;
}
}
/*
* We're done with IPsec processing, transmit the packet using the
* appropriate network protocol (IP or IPv6).
*/
switch (saidx->dst.sa.sa_family) {
#ifdef INET
case AF_INET:
#ifdef IPSEC_NAT_T
/*
* If NAT-T is enabled, now that all IPsec processing is done
* insert UDP encapsulation header after IP header.
*/
if (sav->natt_type) {
struct ip *ip = mtod(m, struct ip *);
const int hlen = (ip->ip_hl << 2);
int size, off;
struct mbuf *mi;
struct udphdr *udp;
size = sizeof(struct udphdr);
if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) {
/*
* draft-ietf-ipsec-nat-t-ike-0[01].txt and
* draft-ietf-ipsec-udp-encaps-(00/)01.txt,
* ignoring possible AH mode
* non-IKE marker + non-ESP marker
* from draft-ietf-ipsec-udp-encaps-00.txt.
*/
size += sizeof(u_int64_t);
}
mi = m_makespace(m, hlen, size, &off);
if (mi == NULL) {
DPRINTF(("%s: m_makespace for udphdr failed\n",
__func__));
error = ENOBUFS;
goto bad;
}
udp = (struct udphdr *)(mtod(mi, caddr_t) + off);
if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT);
else
udp->uh_sport =
KEY_PORTFROMSADDR(&sav->sah->saidx.src);
udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst);
udp->uh_sum = 0;
udp->uh_ulen = htons(m->m_pkthdr.len - hlen);
ip->ip_len = htons(m->m_pkthdr.len);
ip->ip_p = IPPROTO_UDP;
if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
*(u_int64_t *)(udp + 1) = 0;
}
#endif /* IPSEC_NAT_T */
return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
#endif /* INET */
#ifdef INET6
case AF_INET6:
/*
* We don't need massage, IPv6 header fields are always in
* net endian.
*/
return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
#endif /* INET6 */
}
panic("ipsec_process_done");
bad:
m_freem(m);
return (error);
}
/*
* Write an mbuf chain to the destination NIC memory address using
* programmed I/O.
*/
u_short
ed_pio_write_mbufs(struct ed_softc *sc, struct mbuf *m, bus_size_t dst)
{
struct ifnet *ifp = sc->ifp;
unsigned short total_len, dma_len;
struct mbuf *mp;
int maxwait = 200; /* about 240us */
ED_ASSERT_LOCKED(sc);
/* Regular Novell cards */
/* First, count up the total number of bytes to copy */
for (total_len = 0, mp = m; mp; mp = mp->m_next)
total_len += mp->m_len;
dma_len = total_len;
if (sc->isa16bit && (dma_len & 1))
dma_len++;
/* select page 0 registers */
ed_nic_outb(sc, ED_P0_CR, ED_CR_RD2 | ED_CR_STA);
/* reset remote DMA complete flag */
ed_nic_outb(sc, ED_P0_ISR, ED_ISR_RDC);
/* set up DMA byte count */
ed_nic_outb(sc, ED_P0_RBCR0, dma_len);
ed_nic_outb(sc, ED_P0_RBCR1, dma_len >> 8);
/* set up destination address in NIC mem */
ed_nic_outb(sc, ED_P0_RSAR0, dst);
ed_nic_outb(sc, ED_P0_RSAR1, dst >> 8);
/* set remote DMA write */
ed_nic_outb(sc, ED_P0_CR, ED_CR_RD1 | ED_CR_STA);
/*
* Transfer the mbuf chain to the NIC memory.
* 16-bit cards require that data be transferred as words, and only words.
* So that case requires some extra code to patch over odd-length mbufs.
*/
if (!sc->isa16bit) {
/* NE1000s are easy */
while (m) {
if (m->m_len)
ed_asic_outsb(sc, ED_NOVELL_DATA,
m->m_data, m->m_len);
m = m->m_next;
}
} else {
/* NE2000s are a pain */
uint8_t *data;
int len, wantbyte;
union {
uint16_t w;
uint8_t b[2];
} saveword;
wantbyte = 0;
while (m) {
len = m->m_len;
if (len) {
data = mtod(m, caddr_t);
/* finish the last word */
if (wantbyte) {
saveword.b[1] = *data;
ed_asic_outw(sc, ED_NOVELL_DATA,
saveword.w);
data++;
len--;
wantbyte = 0;
}
/* output contiguous words */
if (len > 1) {
ed_asic_outsw(sc, ED_NOVELL_DATA,
data, len >> 1);
data += len & ~1;
len &= 1;
}
/* save last byte, if necessary */
if (len == 1) {
saveword.b[0] = *data;
wantbyte = 1;
}
}
m = m->m_next;
}
/* spit last byte */
if (wantbyte)
ed_asic_outw(sc, ED_NOVELL_DATA, saveword.w);
}
/*
* Destination options header processing.
*/
int
dest6_input(struct mbuf **mp, int *offp, int proto)
{
struct mbuf *m = *mp;
int off = *offp, dstoptlen, optlen;
struct ip6_dest *dstopts;
u_int8_t *opt;
/* validation of the length of the header */
#ifndef PULLDOWN_TEST
IP6_EXTHDR_CHECK(m, off, sizeof(*dstopts), IPPROTO_DONE);
dstopts = (struct ip6_dest *)(mtod(m, caddr_t) + off);
#else
IP6_EXTHDR_GET(dstopts, struct ip6_dest *, m, off, sizeof(*dstopts));
if (dstopts == NULL)
return IPPROTO_DONE;
#endif
dstoptlen = (dstopts->ip6d_len + 1) << 3;
#ifndef PULLDOWN_TEST
IP6_EXTHDR_CHECK(m, off, dstoptlen, IPPROTO_DONE);
dstopts = (struct ip6_dest *)(mtod(m, caddr_t) + off);
#else
IP6_EXTHDR_GET(dstopts, struct ip6_dest *, m, off, dstoptlen);
if (dstopts == NULL)
return IPPROTO_DONE;
#endif
off += dstoptlen;
dstoptlen -= sizeof(struct ip6_dest);
opt = (u_int8_t *)dstopts + sizeof(struct ip6_dest);
/* search header for all options. */
for (optlen = 0; dstoptlen > 0; dstoptlen -= optlen, opt += optlen) {
if (*opt != IP6OPT_PAD1 &&
(dstoptlen < IP6OPT_MINLEN || *(opt + 1) + 2 > dstoptlen)) {
V_ip6stat.ip6s_toosmall++;
goto bad;
}
switch (*opt) {
case IP6OPT_PAD1:
optlen = 1;
break;
case IP6OPT_PADN:
optlen = *(opt + 1) + 2;
break;
default: /* unknown option */
optlen = ip6_unknown_opt(opt, m,
opt - mtod(m, u_int8_t *));
if (optlen == -1)
return (IPPROTO_DONE);
optlen += 2;
break;
}
}
*offp = off;
return (dstopts->ip6d_nxt);
bad:
m_freem(m);
return (IPPROTO_DONE);
}