int
tcp_inpinfo_lock_add(struct inpcb *inp)
{
in_pcbref(inp);
INP_WUNLOCK(inp);
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
return(1);
}
return(0);
}
/*
* RTO timer activated
*/
void
tcplike_rto_timeout(void *ccb)
{
struct tcplike_send_ccb *cb = (struct tcplike_send_ccb *) ccb;
/*struct inpcb *inp;*/
int s;
mutex_enter(&(cb->mutex));
cb->ssthresh = cb->cwnd >>1;
cb->cwnd = 1; /* allowing 1 packet to be sent */
cb->outstanding = 0; /* is this correct? */
cb->rto_timer_callout = 0;
cb->rto = cb->rto << 1;
TIMEOUT_DEBUG((LOG_INFO, "RTO Timeout. New RTO = %u\n", cb->rto));
cb->sample_rtt = 0;
cb->ack_last = 0;
cb->ack_miss = 0;
cb->rcvr_ackratio = 1; /* Constraint 2 & 3. We need ACKs asap */
dccp_remove_feature(cb->pcb, DCCP_OPT_CHANGE_R, DCCP_FEATURE_ACKRATIO);
dccp_add_feature(cb->pcb, DCCP_OPT_CHANGE_R, DCCP_FEATURE_ACKRATIO,
(char *) &cb->rcvr_ackratio, 1);
cb->acked_in_win = 0;
cb->acked_windows = 0;
cb->oldcwnd_ts = cb->pcb->seq_snd;
LOSS_DEBUG((LOG_INFO, "Timeout. CWND value: %u , OUTSTANDING value: %u\n",
cb->cwnd, cb->outstanding));
mutex_exit(&(cb->mutex));
/* lock'n run dccp_output */
s = splnet();
INP_INFO_RLOCK(&dccpbinfo);
/*inp = cb->pcb->d_inpcb;*/
INP_LOCK(inp);
INP_INFO_RUNLOCK(&dccpbinfo);
dccp_output(cb->pcb, 1);
INP_UNLOCK(inp);
splx(s);
}
void
act_open_failure_cleanup(struct adapter *sc, u_int atid, u_int status)
{
struct toepcb *toep = lookup_atid(sc, atid);
struct inpcb *inp = toep->inp;
struct toedev *tod = &toep->td->tod;
free_atid(sc, atid);
toep->tid = -1;
if (status != EAGAIN)
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
toe_connect_failed(tod, inp, status);
final_cpl_received(toep); /* unlocks inp */
if (status != EAGAIN)
INP_INFO_RUNLOCK(&V_tcbinfo);
}
/*
* Accept a connection. Essentially all the work is
* done at higher levels; just return the address
* of the peer, storing through addr.
*/
static int
tcp_usr_accept(struct socket *so, struct sockaddr **nam)
{
int error = 0;
struct inpcb *inp = NULL;
struct tcpcb *tp = NULL;
struct in_addr addr;
in_port_t port = 0;
TCPDEBUG0;
//printf("%s: called \n", __FUNCTION__);
if (so->so_state & SS_ISDISCONNECTED)
return (ECONNABORTED);
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
INP_INFO_RLOCK(&tcbinfo);
INP_LOCK(inp);
if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNABORTED;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
/*
* We inline in_getpeeraddr and COMMON_END here, so that we can
* copy the data of interest and defer the malloc until after we
* release the lock.
*/
port = inp->inp_fport;
addr = inp->inp_faddr;
//printf("%s: called 5\n", __FUNCTION__);
out:
//printf("%s: called out\n", __FUNCTION__);
TCPDEBUG2(PRU_ACCEPT);
INP_UNLOCK(inp);
INP_INFO_RUNLOCK(&tcbinfo);
if (error == 0)
*nam = in_sockaddr(port, &addr);
return error;
}
void
tcp_timer_delack(void *xtp)
{
struct tcpcb *tp = xtp;
struct inpcb *inp;
CURVNET_SET(tp->t_vnet);
INP_INFO_RLOCK(&V_tcbinfo);
inp = tp->t_inpcb;
/*
* XXXRW: While this assert is in fact correct, bugs in the tcpcb
* tear-down mean we need it as a work-around for races between
* timers and tcp_discardcb().
*
* KASSERT(inp != NULL, ("tcp_timer_delack: inp == NULL"));
*/
if (inp == NULL) {
tcp_timer_race++;
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
INP_WLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_delack)
|| !callout_active(&tp->t_timers->tt_delack)) {
INP_WUNLOCK(inp);
CURVNET_RESTORE();
return;
}
callout_deactivate(&tp->t_timers->tt_delack);
tp->t_flags |= TF_ACKNOW;
TCPSTAT_INC(tcps_delack);
(void) tcp_output(tp);
INP_WUNLOCK(inp);
CURVNET_RESTORE();
}
static int
udp_pcblist(SYSCTL_HANDLER_ARGS)
{
int error, i, n;
struct inpcb *inp, **inp_list;
inp_gen_t gencnt;
struct xinpgen xig;
/*
* The process of preparing the PCB list is too time-consuming and
* resource-intensive to repeat twice on every request.
*/
if (req->oldptr == 0) {
n = V_udbinfo.ipi_count;
n += imax(n / 8, 10);
req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
return (0);
}
if (req->newptr != 0)
return (EPERM);
/*
* OK, now we're committed to doing something.
*/
INP_INFO_RLOCK(&V_udbinfo);
gencnt = V_udbinfo.ipi_gencnt;
n = V_udbinfo.ipi_count;
INP_INFO_RUNLOCK(&V_udbinfo);
error = sysctl_wire_old_buffer(req, 2 * (sizeof xig)
+ n * sizeof(struct xinpcb));
if (error != 0)
return (error);
xig.xig_len = sizeof xig;
xig.xig_count = n;
xig.xig_gen = gencnt;
xig.xig_sogen = so_gencnt;
error = SYSCTL_OUT(req, &xig, sizeof xig);
if (error)
return (error);
inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
if (inp_list == 0)
return (ENOMEM);
INP_INFO_RLOCK(&V_udbinfo);
for (inp = LIST_FIRST(V_udbinfo.ipi_listhead), i = 0; inp && i < n;
inp = LIST_NEXT(inp, inp_list)) {
INP_WLOCK(inp);
if (inp->inp_gencnt <= gencnt &&
cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
in_pcbref(inp);
inp_list[i++] = inp;
}
INP_WUNLOCK(inp);
}
INP_INFO_RUNLOCK(&V_udbinfo);
n = i;
error = 0;
for (i = 0; i < n; i++) {
inp = inp_list[i];
INP_RLOCK(inp);
if (inp->inp_gencnt <= gencnt) {
struct xinpcb xi;
bzero(&xi, sizeof(xi));
xi.xi_len = sizeof xi;
/* XXX should avoid extra copy */
bcopy(inp, &xi.xi_inp, sizeof *inp);
if (inp->inp_socket)
sotoxsocket(inp->inp_socket, &xi.xi_socket);
xi.xi_inp.inp_gencnt = inp->inp_gencnt;
INP_RUNLOCK(inp);
error = SYSCTL_OUT(req, &xi, sizeof xi);
} else
INP_RUNLOCK(inp);
}
INP_INFO_WLOCK(&V_udbinfo);
for (i = 0; i < n; i++) {
inp = inp_list[i];
INP_RLOCK(inp);
if (!in_pcbrele_rlocked(inp))
INP_RUNLOCK(inp);
}
INP_INFO_WUNLOCK(&V_udbinfo);
if (!error) {
/*
* Give the user an updated idea of our state. If the
* generation differs from what we told her before, she knows
* that something happened while we were processing this
* request, and it might be necessary to retry.
*/
INP_INFO_RLOCK(&V_udbinfo);
xig.xig_gen = V_udbinfo.ipi_gencnt;
xig.xig_sogen = so_gencnt;
xig.xig_count = V_udbinfo.ipi_count;
INP_INFO_RUNLOCK(&V_udbinfo);
error = SYSCTL_OUT(req, &xig, sizeof xig);
}
free(inp_list, M_TEMP);
return (error);
}
static int
do_rx_iscsi_ddp(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m)
{
struct adapter *sc = iq->adapter;
struct cxgbei_data *ci = sc->iscsi_ulp_softc;
const struct cpl_rx_data_ddp *cpl = (const void *)(rss + 1);
u_int tid = GET_TID(cpl);
struct toepcb *toep = lookup_tid(sc, tid);
struct inpcb *inp = toep->inp;
struct socket *so;
struct sockbuf *sb;
struct tcpcb *tp;
struct icl_cxgbei_conn *icc;
struct icl_conn *ic;
struct icl_cxgbei_pdu *icp = toep->ulpcb2;
struct icl_pdu *ip;
u_int pdu_len, val;
MPASS(m == NULL);
/* Must already be assembling a PDU. */
MPASS(icp != NULL);
MPASS(icp->icp_flags & ICPF_RX_HDR); /* Data is optional. */
MPASS((icp->icp_flags & ICPF_RX_STATUS) == 0);
pdu_len = be16toh(cpl->len); /* includes everything. */
val = be32toh(cpl->ddpvld);
#if 0
CTR5(KTR_CXGBE,
"%s: tid %u, cpl->len %u, ddpvld 0x%08x, icp_flags 0x%08x",
__func__, tid, pdu_len, val, icp->icp_flags);
#endif
icp->icp_flags |= ICPF_RX_STATUS;
ip = &icp->ip;
if (val & F_DDP_PADDING_ERR)
icp->icp_flags |= ICPF_PAD_ERR;
if (val & F_DDP_HDRCRC_ERR)
icp->icp_flags |= ICPF_HCRC_ERR;
if (val & F_DDP_DATACRC_ERR)
icp->icp_flags |= ICPF_DCRC_ERR;
if (val & F_DDP_PDU && ip->ip_data_mbuf == NULL) {
MPASS((icp->icp_flags & ICPF_RX_FLBUF) == 0);
MPASS(ip->ip_data_len > 0);
icp->icp_flags |= ICPF_RX_DDP;
counter_u64_add(ci->ddp_pdus, 1);
counter_u64_add(ci->ddp_bytes, ip->ip_data_len);
}
INP_WLOCK(inp);
if (__predict_false(inp->inp_flags & (INP_DROPPED | INP_TIMEWAIT))) {
CTR4(KTR_CXGBE, "%s: tid %u, rx (%d bytes), inp_flags 0x%x",
__func__, tid, pdu_len, inp->inp_flags);
INP_WUNLOCK(inp);
icl_cxgbei_conn_pdu_free(NULL, ip);
#ifdef INVARIANTS
toep->ulpcb2 = NULL;
#endif
return (0);
}
tp = intotcpcb(inp);
MPASS(icp->icp_seq == tp->rcv_nxt);
MPASS(tp->rcv_wnd >= pdu_len);
tp->rcv_nxt += pdu_len;
tp->rcv_wnd -= pdu_len;
tp->t_rcvtime = ticks;
/* update rx credits */
toep->rx_credits += pdu_len;
t4_rcvd(&toep->td->tod, tp); /* XXX: sc->tom_softc.tod */
so = inp->inp_socket;
sb = &so->so_rcv;
SOCKBUF_LOCK(sb);
icc = toep->ulpcb;
if (__predict_false(icc == NULL || sb->sb_state & SBS_CANTRCVMORE)) {
CTR5(KTR_CXGBE,
"%s: tid %u, excess rx (%d bytes), icc %p, sb_state 0x%x",
__func__, tid, pdu_len, icc, sb->sb_state);
SOCKBUF_UNLOCK(sb);
INP_WUNLOCK(inp);
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
tp = tcp_drop(tp, ECONNRESET);
if (tp)
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
icl_cxgbei_conn_pdu_free(NULL, ip);
#ifdef INVARIANTS
toep->ulpcb2 = NULL;
#endif
return (0);
}
MPASS(icc->icc_signature == CXGBEI_CONN_SIGNATURE);
ic = &icc->ic;
icl_cxgbei_new_pdu_set_conn(ip, ic);
MPASS(m == NULL); /* was unused, we'll use it now. */
m = sbcut_locked(sb, sbused(sb)); /* XXXNP: toep->sb_cc accounting? */
if (__predict_false(m != NULL)) {
int len = m_length(m, NULL);
/*
* PDUs were received before the tid transitioned to ULP mode.
* Convert them to icl_cxgbei_pdus and send them to ICL before
* the PDU in icp/ip.
*/
CTR3(KTR_CXGBE, "%s: tid %u, %u bytes in so_rcv", __func__, tid,
len);
/* XXXNP: needs to be rewritten. */
if (len == sizeof(struct iscsi_bhs) || len == 4 + sizeof(struct
iscsi_bhs)) {
struct icl_cxgbei_pdu *icp0;
struct icl_pdu *ip0;
ip0 = icl_cxgbei_new_pdu(M_NOWAIT);
icl_cxgbei_new_pdu_set_conn(ip0, ic);
if (ip0 == NULL)
CXGBE_UNIMPLEMENTED("PDU allocation failure");
icp0 = ip_to_icp(ip0);
icp0->icp_seq = 0; /* XXX */
icp0->icp_flags = ICPF_RX_HDR | ICPF_RX_STATUS;
m_copydata(m, 0, sizeof(struct iscsi_bhs), (void *)ip0->ip_bhs);
STAILQ_INSERT_TAIL(&icc->rcvd_pdus, ip0, ip_next);
}
m_freem(m);
}
STAILQ_INSERT_TAIL(&icc->rcvd_pdus, ip, ip_next);
if ((icc->rx_flags & RXF_ACTIVE) == 0) {
struct cxgbei_worker_thread_softc *cwt = &cwt_softc[icc->cwt];
mtx_lock(&cwt->cwt_lock);
icc->rx_flags |= RXF_ACTIVE;
TAILQ_INSERT_TAIL(&cwt->rx_head, icc, rx_link);
if (cwt->cwt_state == CWT_SLEEPING) {
cwt->cwt_state = CWT_RUNNING;
cv_signal(&cwt->cwt_cv);
}
mtx_unlock(&cwt->cwt_lock);
}
SOCKBUF_UNLOCK(sb);
INP_WUNLOCK(inp);
#ifdef INVARIANTS
toep->ulpcb2 = NULL;
#endif
return (0);
}
void
udp_input(struct mbuf *m, int off)
{
int iphlen = off;
struct ip *ip;
struct udphdr *uh;
struct ifnet *ifp;
struct inpcb *inp;
uint16_t len, ip_len;
struct inpcbinfo *pcbinfo;
struct ip save_ip;
struct sockaddr_in udp_in;
struct m_tag *fwd_tag;
int cscov_partial;
uint8_t pr;
ifp = m->m_pkthdr.rcvif;
UDPSTAT_INC(udps_ipackets);
/*
* Strip IP options, if any; should skip this, make available to
* user, and use on returned packets, but we don't yet have a way to
* check the checksum with options still present.
*/
if (iphlen > sizeof (struct ip)) {
ip_stripoptions(m);
iphlen = sizeof(struct ip);
}
/*
* Get IP and UDP header together in first mbuf.
*/
ip = mtod(m, struct ip *);
if (m->m_len < iphlen + sizeof(struct udphdr)) {
if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == NULL) {
UDPSTAT_INC(udps_hdrops);
return;
}
ip = mtod(m, struct ip *);
}
uh = (struct udphdr *)((caddr_t)ip + iphlen);
pr = ip->ip_p;
cscov_partial = (pr == IPPROTO_UDPLITE) ? 1 : 0;
/*
* Destination port of 0 is illegal, based on RFC768.
*/
if (uh->uh_dport == 0)
goto badunlocked;
/*
* Construct sockaddr format source address. Stuff source address
* and datagram in user buffer.
*/
bzero(&udp_in, sizeof(udp_in));
udp_in.sin_len = sizeof(udp_in);
udp_in.sin_family = AF_INET;
udp_in.sin_port = uh->uh_sport;
udp_in.sin_addr = ip->ip_src;
/*
* Make mbuf data length reflect UDP length. If not enough data to
* reflect UDP length, drop.
*/
len = ntohs((u_short)uh->uh_ulen);
ip_len = ntohs(ip->ip_len) - iphlen;
if (pr == IPPROTO_UDPLITE && len == 0) {
/* Zero means checksum over the complete packet. */
len = ip_len;
cscov_partial = 0;
}
if (ip_len != len) {
if (len > ip_len || len < sizeof(struct udphdr)) {
UDPSTAT_INC(udps_badlen);
goto badunlocked;
}
if (pr == IPPROTO_UDP)
m_adj(m, len - ip_len);
}
/*
* Save a copy of the IP header in case we want restore it for
* sending an ICMP error message in response.
*/
if (!V_udp_blackhole)
save_ip = *ip;
else
memset(&save_ip, 0, sizeof(save_ip));
/*
* Checksum extended UDP header and data.
*/
if (uh->uh_sum) {
u_short uh_sum;
if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
!cscov_partial) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
uh_sum = m->m_pkthdr.csum_data;
else
uh_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htonl((u_short)len +
m->m_pkthdr.csum_data + pr));
uh_sum ^= 0xffff;
} else {
char b[9];
bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
bzero(((struct ipovly *)ip)->ih_x1, 9);
((struct ipovly *)ip)->ih_len = (pr == IPPROTO_UDP) ?
uh->uh_ulen : htons(ip_len);
uh_sum = in_cksum(m, len + sizeof (struct ip));
bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
}
if (uh_sum) {
UDPSTAT_INC(udps_badsum);
m_freem(m);
return;
}
} else
UDPSTAT_INC(udps_nosum);
pcbinfo = get_inpcbinfo(pr);
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
in_broadcast(ip->ip_dst, ifp)) {
struct inpcb *last;
struct inpcbhead *pcblist;
struct ip_moptions *imo;
INP_INFO_RLOCK(pcbinfo);
pcblist = get_pcblist(pr);
last = NULL;
LIST_FOREACH(inp, pcblist, inp_list) {
if (inp->inp_lport != uh->uh_dport)
continue;
#ifdef INET6
if ((inp->inp_vflag & INP_IPV4) == 0)
continue;
#endif
if (inp->inp_laddr.s_addr != INADDR_ANY &&
inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
continue;
if (inp->inp_faddr.s_addr != INADDR_ANY &&
inp->inp_faddr.s_addr != ip->ip_src.s_addr)
continue;
if (inp->inp_fport != 0 &&
inp->inp_fport != uh->uh_sport)
continue;
INP_RLOCK(inp);
/*
* XXXRW: Because we weren't holding either the inpcb
* or the hash lock when we checked for a match
* before, we should probably recheck now that the
* inpcb lock is held.
*/
/*
* Handle socket delivery policy for any-source
* and source-specific multicast. [RFC3678]
*/
imo = inp->inp_moptions;
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
struct sockaddr_in group;
int blocked;
if (imo == NULL) {
INP_RUNLOCK(inp);
continue;
}
bzero(&group, sizeof(struct sockaddr_in));
group.sin_len = sizeof(struct sockaddr_in);
group.sin_family = AF_INET;
group.sin_addr = ip->ip_dst;
blocked = imo_multi_filter(imo, ifp,
(struct sockaddr *)&group,
(struct sockaddr *)&udp_in);
if (blocked != MCAST_PASS) {
if (blocked == MCAST_NOTGMEMBER)
IPSTAT_INC(ips_notmember);
if (blocked == MCAST_NOTSMEMBER ||
blocked == MCAST_MUTED)
UDPSTAT_INC(udps_filtermcast);
INP_RUNLOCK(inp);
continue;
}
}
if (last != NULL) {
struct mbuf *n;
n = m_copy(m, 0, M_COPYALL);
udp_append(last, ip, n, iphlen, &udp_in);
INP_RUNLOCK(last);
}
last = inp;
/*
* Don't look for additional matches if this one does
* not have either the SO_REUSEPORT or SO_REUSEADDR
* socket options set. This heuristic avoids
* searching through all pcbs in the common case of a
* non-shared port. It assumes that an application
* will never clear these options after setting them.
*/
if ((last->inp_socket->so_options &
(SO_REUSEPORT|SO_REUSEADDR)) == 0)
break;
}
if (last == NULL) {
/*
* No matching pcb found; discard datagram. (No need
* to send an ICMP Port Unreachable for a broadcast
* or multicast datgram.)
*/
UDPSTAT_INC(udps_noportbcast);
if (inp)
INP_RUNLOCK(inp);
INP_INFO_RUNLOCK(pcbinfo);
goto badunlocked;
}
udp_append(last, ip, m, iphlen, &udp_in);
INP_RUNLOCK(last);
INP_INFO_RUNLOCK(pcbinfo);
return;
}
/*
* Locate pcb for datagram.
*/
/*
* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
*/
if ((m->m_flags & M_IP_NEXTHOP) &&
(fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
struct sockaddr_in *next_hop;
next_hop = (struct sockaddr_in *)(fwd_tag + 1);
/*
* Transparently forwarded. Pretend to be the destination.
* Already got one like this?
*/
inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport,
ip->ip_dst, uh->uh_dport, INPLOOKUP_RLOCKPCB, ifp, m);
if (!inp) {
/*
* It's new. Try to find the ambushing socket.
* Because we've rewritten the destination address,
* any hardware-generated hash is ignored.
*/
inp = in_pcblookup(pcbinfo, ip->ip_src,
uh->uh_sport, next_hop->sin_addr,
next_hop->sin_port ? htons(next_hop->sin_port) :
uh->uh_dport, INPLOOKUP_WILDCARD |
INPLOOKUP_RLOCKPCB, ifp);
}
/* Remove the tag from the packet. We don't need it anymore. */
m_tag_delete(m, fwd_tag);
m->m_flags &= ~M_IP_NEXTHOP;
} else
inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport,
ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD |
INPLOOKUP_RLOCKPCB, ifp, m);
if (inp == NULL) {
if (udp_log_in_vain) {
char buf[4*sizeof "123"];
strcpy(buf, inet_ntoa(ip->ip_dst));
log(LOG_INFO,
"Connection attempt to UDP %s:%d from %s:%d\n",
buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
ntohs(uh->uh_sport));
}
UDPSTAT_INC(udps_noport);
if (m->m_flags & (M_BCAST | M_MCAST)) {
UDPSTAT_INC(udps_noportbcast);
goto badunlocked;
}
if (V_udp_blackhole)
goto badunlocked;
if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
goto badunlocked;
*ip = save_ip;
icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
return;
}
/*
* Check the minimum TTL for socket.
*/
INP_RLOCK_ASSERT(inp);
if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl) {
INP_RUNLOCK(inp);
m_freem(m);
return;
}
if (cscov_partial) {
struct udpcb *up;
up = intoudpcb(inp);
if (up->u_rxcslen > len) {
INP_RUNLOCK(inp);
m_freem(m);
return;
}
}
UDP_PROBE(receive, NULL, inp, ip, inp, uh);
udp_append(inp, ip, m, iphlen, &udp_in);
INP_RUNLOCK(inp);
return;
badunlocked:
m_freem(m);
}
void
tcp_timer_rexmt(void * xtp)
{
struct tcpcb *tp = xtp;
CURVNET_SET(tp->t_vnet);
int rexmt;
int headlocked;
struct inpcb *inp;
#ifdef TCPDEBUG
int ostate;
ostate = tp->t_state;
#endif
INP_INFO_RLOCK(&V_tcbinfo);
inp = tp->t_inpcb;
/*
* XXXRW: While this assert is in fact correct, bugs in the tcpcb
* tear-down mean we need it as a work-around for races between
* timers and tcp_discardcb().
*
* KASSERT(inp != NULL, ("tcp_timer_rexmt: inp == NULL"));
*/
if (inp == NULL) {
tcp_timer_race++;
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
INP_WLOCK(inp);
if ((inp->inp_flags & INP_DROPPED) || callout_pending(&tp->t_timers->tt_rexmt)
|| !callout_active(&tp->t_timers->tt_rexmt)) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
callout_deactivate(&tp->t_timers->tt_rexmt);
tcp_free_sackholes(tp);
/*
* Retransmission timer went off. Message has not
* been acked within retransmit interval. Back off
* to a longer retransmit interval and retransmit one segment.
*/
if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
tp->t_rxtshift = TCP_MAXRXTSHIFT;
TCPSTAT_INC(tcps_timeoutdrop);
in_pcbref(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
INP_WUNLOCK(inp);
INP_INFO_WLOCK(&V_tcbinfo);
INP_WLOCK(inp);
if (in_pcbrele_wlocked(inp)) {
INP_INFO_WUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
if (inp->inp_flags & INP_DROPPED) {
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
tp = tcp_drop(tp, tp->t_softerror ?
tp->t_softerror : ETIMEDOUT);
headlocked = 1;
goto out;
}
INP_INFO_RUNLOCK(&V_tcbinfo);
headlocked = 0;
if (tp->t_rxtshift == 1) {
/*
* first retransmit; record ssthresh and cwnd so they can
* be recovered if this turns out to be a "bad" retransmit.
* A retransmit is considered "bad" if an ACK for this
* segment is received within RTT/2 interval; the assumption
* here is that the ACK was already in flight. See
* "On Estimating End-to-End Network Path Properties" by
* Allman and Paxson for more details.
*/
tp->snd_cwnd_prev = tp->snd_cwnd;
tp->snd_ssthresh_prev = tp->snd_ssthresh;
tp->snd_recover_prev = tp->snd_recover;
if (IN_FASTRECOVERY(tp->t_flags))
tp->t_flags |= TF_WASFRECOVERY;
else
tp->t_flags &= ~TF_WASFRECOVERY;
if (IN_CONGRECOVERY(tp->t_flags))
tp->t_flags |= TF_WASCRECOVERY;
else
tp->t_flags &= ~TF_WASCRECOVERY;
tp->t_badrxtwin = ticks + (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
tp->t_flags |= TF_PREVVALID;
} else
static int
do_rx_data(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m)
{
struct adapter *sc = iq->adapter;
const struct cpl_rx_data *cpl = mtod(m, const void *);
unsigned int tid = GET_TID(cpl);
struct toepcb *toep = lookup_tid(sc, tid);
struct inpcb *inp = toep->inp;
struct tcpcb *tp;
struct socket *so;
struct sockbuf *sb;
int len;
uint32_t ddp_placed = 0;
if (__predict_false(toep->flags & TPF_SYNQE)) {
#ifdef INVARIANTS
struct synq_entry *synqe = (void *)toep;
INP_WLOCK(synqe->lctx->inp);
if (synqe->flags & TPF_SYNQE_HAS_L2TE) {
KASSERT(synqe->flags & TPF_ABORT_SHUTDOWN,
("%s: listen socket closed but tid %u not aborted.",
__func__, tid));
} else {
/*
* do_pass_accept_req is still running and will
* eventually take care of this tid.
*/
}
INP_WUNLOCK(synqe->lctx->inp);
#endif
CTR4(KTR_CXGBE, "%s: tid %u, synqe %p (0x%x)", __func__, tid,
toep, toep->flags);
m_freem(m);
return (0);
}
KASSERT(toep->tid == tid, ("%s: toep tid mismatch", __func__));
/* strip off CPL header */
m_adj(m, sizeof(*cpl));
len = m->m_pkthdr.len;
INP_WLOCK(inp);
if (inp->inp_flags & (INP_DROPPED | INP_TIMEWAIT)) {
CTR4(KTR_CXGBE, "%s: tid %u, rx (%d bytes), inp_flags 0x%x",
__func__, tid, len, inp->inp_flags);
INP_WUNLOCK(inp);
m_freem(m);
return (0);
}
tp = intotcpcb(inp);
if (__predict_false(tp->rcv_nxt != be32toh(cpl->seq)))
ddp_placed = be32toh(cpl->seq) - tp->rcv_nxt;
tp->rcv_nxt += len;
KASSERT(tp->rcv_wnd >= len, ("%s: negative window size", __func__));
tp->rcv_wnd -= len;
tp->t_rcvtime = ticks;
so = inp_inpcbtosocket(inp);
sb = &so->so_rcv;
SOCKBUF_LOCK(sb);
if (__predict_false(sb->sb_state & SBS_CANTRCVMORE)) {
CTR3(KTR_CXGBE, "%s: tid %u, excess rx (%d bytes)",
__func__, tid, len);
m_freem(m);
SOCKBUF_UNLOCK(sb);
INP_WUNLOCK(inp);
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
tp = tcp_drop(tp, ECONNRESET);
if (tp)
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
return (0);
}
/* receive buffer autosize */
if (sb->sb_flags & SB_AUTOSIZE &&
V_tcp_do_autorcvbuf &&
sb->sb_hiwat < V_tcp_autorcvbuf_max &&
len > (sbspace(sb) / 8 * 7)) {
unsigned int hiwat = sb->sb_hiwat;
unsigned int newsize = min(hiwat + V_tcp_autorcvbuf_inc,
V_tcp_autorcvbuf_max);
if (!sbreserve_locked(sb, newsize, so, NULL))
sb->sb_flags &= ~SB_AUTOSIZE;
else
toep->rx_credits += newsize - hiwat;
}
if (toep->ulp_mode == ULP_MODE_TCPDDP) {
int changed = !(toep->ddp_flags & DDP_ON) ^ cpl->ddp_off;
if (changed) {
if (toep->ddp_flags & DDP_SC_REQ)
toep->ddp_flags ^= DDP_ON | DDP_SC_REQ;
else {
KASSERT(cpl->ddp_off == 1,
("%s: DDP switched on by itself.",
__func__));
/* Fell out of DDP mode */
toep->ddp_flags &= ~(DDP_ON | DDP_BUF0_ACTIVE |
DDP_BUF1_ACTIVE);
if (ddp_placed)
insert_ddp_data(toep, ddp_placed);
}
}
if ((toep->ddp_flags & DDP_OK) == 0 &&
time_uptime >= toep->ddp_disabled + DDP_RETRY_WAIT) {
toep->ddp_score = DDP_LOW_SCORE;
toep->ddp_flags |= DDP_OK;
CTR3(KTR_CXGBE, "%s: tid %u DDP_OK @ %u",
__func__, tid, time_uptime);
}
if (toep->ddp_flags & DDP_ON) {
/*
* CPL_RX_DATA with DDP on can only be an indicate. Ask
* soreceive to post a buffer or disable DDP. The
* payload that arrived in this indicate is appended to
* the socket buffer as usual.
*/
#if 0
CTR5(KTR_CXGBE,
"%s: tid %u (0x%x) DDP indicate (seq 0x%x, len %d)",
__func__, tid, toep->flags, be32toh(cpl->seq), len);
#endif
sb->sb_flags |= SB_DDP_INDICATE;
} else if ((toep->ddp_flags & (DDP_OK|DDP_SC_REQ)) == DDP_OK &&
tp->rcv_wnd > DDP_RSVD_WIN && len >= sc->tt.ddp_thres) {
/*
* DDP allowed but isn't on (and a request to switch it
* on isn't pending either), and conditions are ripe for
* it to work. Switch it on.
*/
enable_ddp(sc, toep);
}
}
KASSERT(toep->sb_cc >= sbused(sb),
("%s: sb %p has more data (%d) than last time (%d).",
__func__, sb, sbused(sb), toep->sb_cc));
toep->rx_credits += toep->sb_cc - sbused(sb);
sbappendstream_locked(sb, m, 0);
toep->sb_cc = sbused(sb);
if (toep->rx_credits > 0 && toep->sb_cc + tp->rcv_wnd < sb->sb_lowat) {
int credits;
credits = send_rx_credits(sc, toep, toep->rx_credits);
toep->rx_credits -= credits;
tp->rcv_wnd += credits;
tp->rcv_adv += credits;
}
sorwakeup_locked(so);
SOCKBUF_UNLOCK_ASSERT(sb);
INP_WUNLOCK(inp);
return (0);
}
void
tcp_timer_persist(void *xtp)
{
struct tcpcb *tp = xtp;
struct inpcb *inp;
CURVNET_SET(tp->t_vnet);
#ifdef TCPDEBUG
int ostate;
ostate = tp->t_state;
#endif
INP_INFO_RLOCK(&V_tcbinfo);
inp = tp->t_inpcb;
KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp));
INP_WLOCK(inp);
if (callout_pending(&tp->t_timers->tt_persist) ||
!callout_active(&tp->t_timers->tt_persist)) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
callout_deactivate(&tp->t_timers->tt_persist);
if ((inp->inp_flags & INP_DROPPED) != 0) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
KASSERT((tp->t_timers->tt_flags & TT_STOPPED) == 0,
("%s: tp %p tcpcb can't be stopped here", __func__, tp));
KASSERT((tp->t_timers->tt_flags & TT_PERSIST) != 0,
("%s: tp %p persist callout should be running", __func__, tp));
/*
* Persistance timer into zero window.
* Force a byte to be output, if possible.
*/
TCPSTAT_INC(tcps_persisttimeo);
/*
* Hack: if the peer is dead/unreachable, we do not
* time out if the window is closed. After a full
* backoff, drop the connection if the idle time
* (no responses to probes) reaches the maximum
* backoff that we would use if retransmitting.
*/
if (tp->t_rxtshift == TCP_MAXRXTSHIFT &&
(ticks - tp->t_rcvtime >= tcp_maxpersistidle ||
ticks - tp->t_rcvtime >= TCP_REXMTVAL(tp) * tcp_totbackoff)) {
TCPSTAT_INC(tcps_persistdrop);
tp = tcp_drop(tp, ETIMEDOUT);
goto out;
}
/*
* If the user has closed the socket then drop a persisting
* connection after a much reduced timeout.
*/
if (tp->t_state > TCPS_CLOSE_WAIT &&
(ticks - tp->t_rcvtime) >= TCPTV_PERSMAX) {
TCPSTAT_INC(tcps_persistdrop);
tp = tcp_drop(tp, ETIMEDOUT);
goto out;
}
tcp_setpersist(tp);
tp->t_flags |= TF_FORCEDATA;
(void) tp->t_fb->tfb_tcp_output(tp);
tp->t_flags &= ~TF_FORCEDATA;
out:
#ifdef TCPDEBUG
if (tp != NULL && tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO);
#endif
TCP_PROBE2(debug__user, tp, PRU_SLOWTIMO);
if (tp != NULL)
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
}
void
tcp_timer_rexmt(void * xtp)
{
struct tcpcb *tp = xtp;
CURVNET_SET(tp->t_vnet);
int rexmt;
int headlocked;
struct inpcb *inp;
#ifdef TCPDEBUG
int ostate;
ostate = tp->t_state;
#endif
INP_INFO_RLOCK(&V_tcbinfo);
inp = tp->t_inpcb;
KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp));
INP_WLOCK(inp);
if (callout_pending(&tp->t_timers->tt_rexmt) ||
!callout_active(&tp->t_timers->tt_rexmt)) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
callout_deactivate(&tp->t_timers->tt_rexmt);
if ((inp->inp_flags & INP_DROPPED) != 0) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
KASSERT((tp->t_timers->tt_flags & TT_STOPPED) == 0,
("%s: tp %p tcpcb can't be stopped here", __func__, tp));
KASSERT((tp->t_timers->tt_flags & TT_REXMT) != 0,
("%s: tp %p rexmt callout should be running", __func__, tp));
tcp_free_sackholes(tp);
/*
* Retransmission timer went off. Message has not
* been acked within retransmit interval. Back off
* to a longer retransmit interval and retransmit one segment.
*/
if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
tp->t_rxtshift = TCP_MAXRXTSHIFT;
TCPSTAT_INC(tcps_timeoutdrop);
tp = tcp_drop(tp, tp->t_softerror ?
tp->t_softerror : ETIMEDOUT);
headlocked = 1;
goto out;
}
INP_INFO_RUNLOCK(&V_tcbinfo);
headlocked = 0;
if (tp->t_state == TCPS_SYN_SENT) {
/*
* If the SYN was retransmitted, indicate CWND to be
* limited to 1 segment in cc_conn_init().
*/
tp->snd_cwnd = 1;
} else if (tp->t_rxtshift == 1) {
/*
* first retransmit; record ssthresh and cwnd so they can
* be recovered if this turns out to be a "bad" retransmit.
* A retransmit is considered "bad" if an ACK for this
* segment is received within RTT/2 interval; the assumption
* here is that the ACK was already in flight. See
* "On Estimating End-to-End Network Path Properties" by
* Allman and Paxson for more details.
*/
tp->snd_cwnd_prev = tp->snd_cwnd;
tp->snd_ssthresh_prev = tp->snd_ssthresh;
tp->snd_recover_prev = tp->snd_recover;
if (IN_FASTRECOVERY(tp->t_flags))
tp->t_flags |= TF_WASFRECOVERY;
else
tp->t_flags &= ~TF_WASFRECOVERY;
if (IN_CONGRECOVERY(tp->t_flags))
tp->t_flags |= TF_WASCRECOVERY;
else
tp->t_flags &= ~TF_WASCRECOVERY;
tp->t_badrxtwin = ticks + (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
tp->t_flags |= TF_PREVVALID;
} else
void
tcp_timer_2msl(void *xtp)
{
struct tcpcb *tp = xtp;
struct inpcb *inp;
CURVNET_SET(tp->t_vnet);
#ifdef TCPDEBUG
int ostate;
ostate = tp->t_state;
#endif
INP_INFO_RLOCK(&V_tcbinfo);
inp = tp->t_inpcb;
KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp));
INP_WLOCK(inp);
tcp_free_sackholes(tp);
if (callout_pending(&tp->t_timers->tt_2msl) ||
!callout_active(&tp->t_timers->tt_2msl)) {
INP_WUNLOCK(tp->t_inpcb);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
callout_deactivate(&tp->t_timers->tt_2msl);
if ((inp->inp_flags & INP_DROPPED) != 0) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
KASSERT((tp->t_timers->tt_flags & TT_STOPPED) == 0,
("%s: tp %p tcpcb can't be stopped here", __func__, tp));
KASSERT((tp->t_timers->tt_flags & TT_2MSL) != 0,
("%s: tp %p 2msl callout should be running", __func__, tp));
/*
* 2 MSL timeout in shutdown went off. If we're closed but
* still waiting for peer to close and connection has been idle
* too long delete connection control block. Otherwise, check
* again in a bit.
*
* If in TIME_WAIT state just ignore as this timeout is handled in
* tcp_tw_2msl_scan().
*
* If fastrecycle of FIN_WAIT_2, in FIN_WAIT_2 and receiver has closed,
* there's no point in hanging onto FIN_WAIT_2 socket. Just close it.
* Ignore fact that there were recent incoming segments.
*/
if ((inp->inp_flags & INP_TIMEWAIT) != 0) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
if (tcp_fast_finwait2_recycle && tp->t_state == TCPS_FIN_WAIT_2 &&
tp->t_inpcb && tp->t_inpcb->inp_socket &&
(tp->t_inpcb->inp_socket->so_rcv.sb_state & SBS_CANTRCVMORE)) {
TCPSTAT_INC(tcps_finwait2_drops);
tp = tcp_close(tp);
} else {
if (ticks - tp->t_rcvtime <= TP_MAXIDLE(tp)) {
if (!callout_reset(&tp->t_timers->tt_2msl,
TP_KEEPINTVL(tp), tcp_timer_2msl, tp)) {
tp->t_timers->tt_flags &= ~TT_2MSL_RST;
}
} else
tp = tcp_close(tp);
}
#ifdef TCPDEBUG
if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0,
PRU_SLOWTIMO);
#endif
TCP_PROBE2(debug__user, tp, PRU_SLOWTIMO);
if (tp != NULL)
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
}
void
tcp_timer_keep(void *xtp)
{
struct tcpcb *tp = xtp;
struct tcptemp *t_template;
struct inpcb *inp;
CURVNET_SET(tp->t_vnet);
#ifdef TCPDEBUG
int ostate;
ostate = tp->t_state;
#endif
INP_INFO_RLOCK(&V_tcbinfo);
inp = tp->t_inpcb;
KASSERT(inp != NULL, ("%s: tp %p tp->t_inpcb == NULL", __func__, tp));
INP_WLOCK(inp);
if (callout_pending(&tp->t_timers->tt_keep) ||
!callout_active(&tp->t_timers->tt_keep)) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
callout_deactivate(&tp->t_timers->tt_keep);
if ((inp->inp_flags & INP_DROPPED) != 0) {
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
}
KASSERT((tp->t_timers->tt_flags & TT_STOPPED) == 0,
("%s: tp %p tcpcb can't be stopped here", __func__, tp));
KASSERT((tp->t_timers->tt_flags & TT_KEEP) != 0,
("%s: tp %p keep callout should be running", __func__, tp));
/*
* Keep-alive timer went off; send something
* or drop connection if idle for too long.
*/
TCPSTAT_INC(tcps_keeptimeo);
if (tp->t_state < TCPS_ESTABLISHED)
goto dropit;
if ((always_keepalive || inp->inp_socket->so_options & SO_KEEPALIVE) &&
tp->t_state <= TCPS_CLOSING) {
if (ticks - tp->t_rcvtime >= TP_KEEPIDLE(tp) + TP_MAXIDLE(tp))
goto dropit;
/*
* Send a packet designed to force a response
* if the peer is up and reachable:
* either an ACK if the connection is still alive,
* or an RST if the peer has closed the connection
* due to timeout or reboot.
* Using sequence number tp->snd_una-1
* causes the transmitted zero-length segment
* to lie outside the receive window;
* by the protocol spec, this requires the
* correspondent TCP to respond.
*/
TCPSTAT_INC(tcps_keepprobe);
t_template = tcpip_maketemplate(inp);
if (t_template) {
tcp_respond(tp, t_template->tt_ipgen,
&t_template->tt_t, (struct mbuf *)NULL,
tp->rcv_nxt, tp->snd_una - 1, 0);
free(t_template, M_TEMP);
}
if (!callout_reset(&tp->t_timers->tt_keep, TP_KEEPINTVL(tp),
tcp_timer_keep, tp)) {
tp->t_timers->tt_flags &= ~TT_KEEP_RST;
}
} else if (!callout_reset(&tp->t_timers->tt_keep, TP_KEEPIDLE(tp),
tcp_timer_keep, tp)) {
tp->t_timers->tt_flags &= ~TT_KEEP_RST;
}
#ifdef TCPDEBUG
if (inp->inp_socket->so_options & SO_DEBUG)
tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0,
PRU_SLOWTIMO);
#endif
TCP_PROBE2(debug__user, tp, PRU_SLOWTIMO);
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
return;
dropit:
TCPSTAT_INC(tcps_keepdrops);
tp = tcp_drop(tp, ETIMEDOUT);
#ifdef TCPDEBUG
if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0,
PRU_SLOWTIMO);
#endif
TCP_PROBE2(debug__user, tp, PRU_SLOWTIMO);
if (tp != NULL)
INP_WUNLOCK(tp->t_inpcb);
INP_INFO_RUNLOCK(&V_tcbinfo);
CURVNET_RESTORE();
}
/*
* Peer has sent us a FIN.
*/
static int
do_peer_close(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m)
{
struct adapter *sc = iq->adapter;
const struct cpl_peer_close *cpl = (const void *)(rss + 1);
unsigned int tid = GET_TID(cpl);
struct toepcb *toep = lookup_tid(sc, tid);
struct inpcb *inp = toep->inp;
struct tcpcb *tp = NULL;
struct socket *so;
#ifdef INVARIANTS
unsigned int opcode = G_CPL_OPCODE(be32toh(OPCODE_TID(cpl)));
#endif
KASSERT(opcode == CPL_PEER_CLOSE,
("%s: unexpected opcode 0x%x", __func__, opcode));
KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__));
if (__predict_false(toep->flags & TPF_SYNQE)) {
#ifdef INVARIANTS
struct synq_entry *synqe = (void *)toep;
INP_WLOCK(synqe->lctx->inp);
if (synqe->flags & TPF_SYNQE_HAS_L2TE) {
KASSERT(synqe->flags & TPF_ABORT_SHUTDOWN,
("%s: listen socket closed but tid %u not aborted.",
__func__, tid));
} else {
/*
* do_pass_accept_req is still running and will
* eventually take care of this tid.
*/
}
INP_WUNLOCK(synqe->lctx->inp);
#endif
CTR4(KTR_CXGBE, "%s: tid %u, synqe %p (0x%x)", __func__, tid,
toep, toep->flags);
return (0);
}
KASSERT(toep->tid == tid, ("%s: toep tid mismatch", __func__));
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
tp = intotcpcb(inp);
CTR5(KTR_CXGBE, "%s: tid %u (%s), toep_flags 0x%x, inp %p", __func__,
tid, tp ? tcpstates[tp->t_state] : "no tp", toep->flags, inp);
if (toep->flags & TPF_ABORT_SHUTDOWN)
goto done;
tp->rcv_nxt++; /* FIN */
so = inp->inp_socket;
if (toep->ulp_mode == ULP_MODE_TCPDDP) {
DDP_LOCK(toep);
if (__predict_false(toep->ddp_flags &
(DDP_BUF0_ACTIVE | DDP_BUF1_ACTIVE)))
handle_ddp_close(toep, tp, cpl->rcv_nxt);
DDP_UNLOCK(toep);
}
socantrcvmore(so);
if (toep->ulp_mode != ULP_MODE_RDMA) {
KASSERT(tp->rcv_nxt == be32toh(cpl->rcv_nxt),
("%s: rcv_nxt mismatch: %u %u", __func__, tp->rcv_nxt,
be32toh(cpl->rcv_nxt)));
}
switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
tp->t_starttime = ticks;
/* FALLTHROUGH */
case TCPS_ESTABLISHED:
tp->t_state = TCPS_CLOSE_WAIT;
break;
case TCPS_FIN_WAIT_1:
tp->t_state = TCPS_CLOSING;
break;
case TCPS_FIN_WAIT_2:
tcp_twstart(tp);
INP_UNLOCK_ASSERT(inp); /* safe, we have a ref on the inp */
INP_INFO_RUNLOCK(&V_tcbinfo);
INP_WLOCK(inp);
final_cpl_received(toep);
return (0);
default:
log(LOG_ERR, "%s: TID %u received CPL_PEER_CLOSE in state %d\n",
__func__, tid, tp->t_state);
}
done:
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
return (0);
}
int
udp6_input(struct mbuf **mp, int *offp, int proto)
{
struct mbuf *m = *mp;
struct ifnet *ifp;
struct ip6_hdr *ip6;
struct udphdr *uh;
struct inpcb *inp;
struct inpcbinfo *pcbinfo;
struct udpcb *up;
int off = *offp;
int cscov_partial;
int plen, ulen;
struct sockaddr_in6 fromsa;
struct m_tag *fwd_tag;
uint16_t uh_sum;
uint8_t nxt;
ifp = m->m_pkthdr.rcvif;
ip6 = mtod(m, struct ip6_hdr *);
#ifndef PULLDOWN_TEST
IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
ip6 = mtod(m, struct ip6_hdr *);
uh = (struct udphdr *)((caddr_t)ip6 + off);
#else
IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
if (!uh)
return (IPPROTO_DONE);
#endif
UDPSTAT_INC(udps_ipackets);
/*
* Destination port of 0 is illegal, based on RFC768.
*/
if (uh->uh_dport == 0)
goto badunlocked;
plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
ulen = ntohs((u_short)uh->uh_ulen);
nxt = proto;
cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
if (nxt == IPPROTO_UDPLITE) {
/* Zero means checksum over the complete packet. */
if (ulen == 0)
ulen = plen;
if (ulen == plen)
cscov_partial = 0;
if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
/* XXX: What is the right UDPLite MIB counter? */
goto badunlocked;
}
if (uh->uh_sum == 0) {
/* XXX: What is the right UDPLite MIB counter? */
goto badunlocked;
}
} else {
if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
UDPSTAT_INC(udps_badlen);
goto badunlocked;
}
if (uh->uh_sum == 0) {
UDPSTAT_INC(udps_nosum);
goto badunlocked;
}
}
if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
!cscov_partial) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
uh_sum = m->m_pkthdr.csum_data;
else
uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
m->m_pkthdr.csum_data);
uh_sum ^= 0xffff;
} else
uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
if (uh_sum != 0) {
UDPSTAT_INC(udps_badsum);
goto badunlocked;
}
/*
* Construct sockaddr format source address.
*/
init_sin6(&fromsa, m);
fromsa.sin6_port = uh->uh_sport;
pcbinfo = udp_get_inpcbinfo(nxt);
if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
struct inpcb *last;
struct inpcbhead *pcblist;
struct ip6_moptions *imo;
INP_INFO_RLOCK(pcbinfo);
/*
* In the event that laddr should be set to the link-local
* address (this happens in RIPng), the multicast address
* specified in the received packet will not match laddr. To
* handle this situation, matching is relaxed if the
* receiving interface is the same as one specified in the
* socket and if the destination multicast address matches
* one of the multicast groups specified in the socket.
*/
/*
* KAME note: traditionally we dropped udpiphdr from mbuf
* here. We need udphdr for IPsec processing so we do that
* later.
*/
pcblist = udp_get_pcblist(nxt);
last = NULL;
LIST_FOREACH(inp, pcblist, inp_list) {
if ((inp->inp_vflag & INP_IPV6) == 0)
continue;
if (inp->inp_lport != uh->uh_dport)
continue;
if (inp->inp_fport != 0 &&
inp->inp_fport != uh->uh_sport)
continue;
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
&ip6->ip6_dst))
continue;
}
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
&ip6->ip6_src) ||
inp->inp_fport != uh->uh_sport)
continue;
}
/*
* XXXRW: Because we weren't holding either the inpcb
* or the hash lock when we checked for a match
* before, we should probably recheck now that the
* inpcb lock is (supposed to be) held.
*/
/*
* Handle socket delivery policy for any-source
* and source-specific multicast. [RFC3678]
*/
imo = inp->in6p_moptions;
if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
struct sockaddr_in6 mcaddr;
int blocked;
INP_RLOCK(inp);
bzero(&mcaddr, sizeof(struct sockaddr_in6));
mcaddr.sin6_len = sizeof(struct sockaddr_in6);
mcaddr.sin6_family = AF_INET6;
mcaddr.sin6_addr = ip6->ip6_dst;
blocked = im6o_mc_filter(imo, ifp,
(struct sockaddr *)&mcaddr,
(struct sockaddr *)&fromsa);
if (blocked != MCAST_PASS) {
if (blocked == MCAST_NOTGMEMBER)
IP6STAT_INC(ip6s_notmember);
if (blocked == MCAST_NOTSMEMBER ||
blocked == MCAST_MUTED)
UDPSTAT_INC(udps_filtermcast);
INP_RUNLOCK(inp); /* XXX */
continue;
}
INP_RUNLOCK(inp);
}
if (last != NULL) {
struct mbuf *n;
if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
INP_RLOCK(last);
UDP_PROBE(receive, NULL, last, ip6,
last, uh);
if (udp6_append(last, n, off, &fromsa))
goto inp_lost;
INP_RUNLOCK(last);
}
}
last = inp;
/*
* Don't look for additional matches if this one does
* not have either the SO_REUSEPORT or SO_REUSEADDR
* socket options set. This heuristic avoids
* searching through all pcbs in the common case of a
* non-shared port. It assumes that an application
* will never clear these options after setting them.
*/
if ((last->inp_socket->so_options &
(SO_REUSEPORT|SO_REUSEADDR)) == 0)
break;
}
if (last == NULL) {
/*
* No matching pcb found; discard datagram. (No need
* to send an ICMP Port Unreachable for a broadcast
* or multicast datgram.)
*/
UDPSTAT_INC(udps_noport);
UDPSTAT_INC(udps_noportmcast);
goto badheadlocked;
}
INP_RLOCK(last);
INP_INFO_RUNLOCK(pcbinfo);
UDP_PROBE(receive, NULL, last, ip6, last, uh);
if (udp6_append(last, m, off, &fromsa) == 0)
INP_RUNLOCK(last);
inp_lost:
return (IPPROTO_DONE);
}
/*
* Locate pcb for datagram.
*/
/*
* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
*/
if ((m->m_flags & M_IP6_NEXTHOP) &&
(fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
struct sockaddr_in6 *next_hop6;
next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
/*
* Transparently forwarded. Pretend to be the destination.
* Already got one like this?
*/
inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
if (!inp) {
/*
* It's new. Try to find the ambushing socket.
* Because we've rewritten the destination address,
* any hardware-generated hash is ignored.
*/
inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
uh->uh_sport, &next_hop6->sin6_addr,
next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
uh->uh_dport, INPLOOKUP_WILDCARD |
INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
}
/* Remove the tag from the packet. We don't need it anymore. */
m_tag_delete(m, fwd_tag);
m->m_flags &= ~M_IP6_NEXTHOP;
} else
inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
m->m_pkthdr.rcvif, m);
if (inp == NULL) {
if (udp_log_in_vain) {
char ip6bufs[INET6_ADDRSTRLEN];
char ip6bufd[INET6_ADDRSTRLEN];
log(LOG_INFO,
"Connection attempt to UDP [%s]:%d from [%s]:%d\n",
ip6_sprintf(ip6bufd, &ip6->ip6_dst),
ntohs(uh->uh_dport),
ip6_sprintf(ip6bufs, &ip6->ip6_src),
ntohs(uh->uh_sport));
}
UDPSTAT_INC(udps_noport);
if (m->m_flags & M_MCAST) {
printf("UDP6: M_MCAST is set in a unicast packet.\n");
UDPSTAT_INC(udps_noportmcast);
goto badunlocked;
}
if (V_udp_blackhole)
goto badunlocked;
if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
goto badunlocked;
icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
return (IPPROTO_DONE);
}
INP_RLOCK_ASSERT(inp);
up = intoudpcb(inp);
if (cscov_partial) {
if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
INP_RUNLOCK(inp);
m_freem(m);
return (IPPROTO_DONE);
}
}
UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
if (udp6_append(inp, m, off, &fromsa) == 0)
INP_RUNLOCK(inp);
return (IPPROTO_DONE);
badheadlocked:
INP_INFO_RUNLOCK(pcbinfo);
badunlocked:
if (m)
m_freem(m);
return (IPPROTO_DONE);
}
/*
* Peer has ACK'd our FIN.
*/
static int
do_close_con_rpl(struct sge_iq *iq, const struct rss_header *rss,
struct mbuf *m)
{
struct adapter *sc = iq->adapter;
const struct cpl_close_con_rpl *cpl = (const void *)(rss + 1);
unsigned int tid = GET_TID(cpl);
struct toepcb *toep = lookup_tid(sc, tid);
struct inpcb *inp = toep->inp;
struct tcpcb *tp = NULL;
struct socket *so = NULL;
#ifdef INVARIANTS
unsigned int opcode = G_CPL_OPCODE(be32toh(OPCODE_TID(cpl)));
#endif
KASSERT(opcode == CPL_CLOSE_CON_RPL,
("%s: unexpected opcode 0x%x", __func__, opcode));
KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__));
KASSERT(toep->tid == tid, ("%s: toep tid mismatch", __func__));
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
tp = intotcpcb(inp);
CTR4(KTR_CXGBE, "%s: tid %u (%s), toep_flags 0x%x",
__func__, tid, tp ? tcpstates[tp->t_state] : "no tp", toep->flags);
if (toep->flags & TPF_ABORT_SHUTDOWN)
goto done;
so = inp->inp_socket;
tp->snd_una = be32toh(cpl->snd_nxt) - 1; /* exclude FIN */
switch (tp->t_state) {
case TCPS_CLOSING: /* see TCPS_FIN_WAIT_2 in do_peer_close too */
tcp_twstart(tp);
release:
INP_UNLOCK_ASSERT(inp); /* safe, we have a ref on the inp */
INP_INFO_RUNLOCK(&V_tcbinfo);
INP_WLOCK(inp);
final_cpl_received(toep); /* no more CPLs expected */
return (0);
case TCPS_LAST_ACK:
if (tcp_close(tp))
INP_WUNLOCK(inp);
goto release;
case TCPS_FIN_WAIT_1:
if (so->so_rcv.sb_state & SBS_CANTRCVMORE)
soisdisconnected(so);
tp->t_state = TCPS_FIN_WAIT_2;
break;
default:
log(LOG_ERR,
"%s: TID %u received CPL_CLOSE_CON_RPL in state %s\n",
__func__, tid, tcpstates[tp->t_state]);
}
done:
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
return (0);
}
static int
do_rx_data(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m)
{
struct adapter *sc = iq->adapter;
const struct cpl_rx_data *cpl = mtod(m, const void *);
unsigned int tid = GET_TID(cpl);
struct toepcb *toep = lookup_tid(sc, tid);
struct inpcb *inp = toep->inp;
struct tcpcb *tp;
struct socket *so;
struct sockbuf *sb;
int len;
uint32_t ddp_placed = 0;
if (__predict_false(toep->flags & TPF_SYNQE)) {
#ifdef INVARIANTS
struct synq_entry *synqe = (void *)toep;
INP_WLOCK(synqe->lctx->inp);
if (synqe->flags & TPF_SYNQE_HAS_L2TE) {
KASSERT(synqe->flags & TPF_ABORT_SHUTDOWN,
("%s: listen socket closed but tid %u not aborted.",
__func__, tid));
} else {
/*
* do_pass_accept_req is still running and will
* eventually take care of this tid.
*/
}
INP_WUNLOCK(synqe->lctx->inp);
#endif
CTR4(KTR_CXGBE, "%s: tid %u, synqe %p (0x%x)", __func__, tid,
toep, toep->flags);
m_freem(m);
return (0);
}
KASSERT(toep->tid == tid, ("%s: toep tid mismatch", __func__));
/* strip off CPL header */
m_adj(m, sizeof(*cpl));
len = m->m_pkthdr.len;
INP_WLOCK(inp);
if (inp->inp_flags & (INP_DROPPED | INP_TIMEWAIT)) {
CTR4(KTR_CXGBE, "%s: tid %u, rx (%d bytes), inp_flags 0x%x",
__func__, tid, len, inp->inp_flags);
INP_WUNLOCK(inp);
m_freem(m);
return (0);
}
tp = intotcpcb(inp);
if (__predict_false(tp->rcv_nxt != be32toh(cpl->seq)))
ddp_placed = be32toh(cpl->seq) - tp->rcv_nxt;
tp->rcv_nxt += len;
if (tp->rcv_wnd < len) {
KASSERT(toep->ulp_mode == ULP_MODE_RDMA,
("%s: negative window size", __func__));
}
tp->rcv_wnd -= len;
tp->t_rcvtime = ticks;
if (toep->ulp_mode == ULP_MODE_TCPDDP)
DDP_LOCK(toep);
so = inp_inpcbtosocket(inp);
sb = &so->so_rcv;
SOCKBUF_LOCK(sb);
if (__predict_false(sb->sb_state & SBS_CANTRCVMORE)) {
CTR3(KTR_CXGBE, "%s: tid %u, excess rx (%d bytes)",
__func__, tid, len);
m_freem(m);
SOCKBUF_UNLOCK(sb);
if (toep->ulp_mode == ULP_MODE_TCPDDP)
DDP_UNLOCK(toep);
INP_WUNLOCK(inp);
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
tp = tcp_drop(tp, ECONNRESET);
if (tp)
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
return (0);
}
/* receive buffer autosize */
CURVNET_SET(so->so_vnet);
if (sb->sb_flags & SB_AUTOSIZE &&
V_tcp_do_autorcvbuf &&
sb->sb_hiwat < V_tcp_autorcvbuf_max &&
len > (sbspace(sb) / 8 * 7)) {
unsigned int hiwat = sb->sb_hiwat;
unsigned int newsize = min(hiwat + V_tcp_autorcvbuf_inc,
V_tcp_autorcvbuf_max);
if (!sbreserve_locked(sb, newsize, so, NULL))
sb->sb_flags &= ~SB_AUTOSIZE;
else
toep->rx_credits += newsize - hiwat;
}
if (toep->ddp_waiting_count != 0 || toep->ddp_active_count != 0)
CTR3(KTR_CXGBE, "%s: tid %u, non-ddp rx (%d bytes)", __func__,
tid, len);
if (toep->ulp_mode == ULP_MODE_TCPDDP) {
int changed = !(toep->ddp_flags & DDP_ON) ^ cpl->ddp_off;
if (changed) {
if (toep->ddp_flags & DDP_SC_REQ)
toep->ddp_flags ^= DDP_ON | DDP_SC_REQ;
else {
KASSERT(cpl->ddp_off == 1,
("%s: DDP switched on by itself.",
__func__));
/* Fell out of DDP mode */
toep->ddp_flags &= ~DDP_ON;
CTR1(KTR_CXGBE, "%s: fell out of DDP mode",
__func__);
insert_ddp_data(toep, ddp_placed);
}
}
if (toep->ddp_flags & DDP_ON) {
/*
* CPL_RX_DATA with DDP on can only be an indicate.
* Start posting queued AIO requests via DDP. The
* payload that arrived in this indicate is appended
* to the socket buffer as usual.
*/
handle_ddp_indicate(toep);
}
}
KASSERT(toep->sb_cc >= sbused(sb),
("%s: sb %p has more data (%d) than last time (%d).",
__func__, sb, sbused(sb), toep->sb_cc));
toep->rx_credits += toep->sb_cc - sbused(sb);
sbappendstream_locked(sb, m, 0);
toep->sb_cc = sbused(sb);
if (toep->rx_credits > 0 && toep->sb_cc + tp->rcv_wnd < sb->sb_lowat) {
int credits;
credits = send_rx_credits(sc, toep, toep->rx_credits);
toep->rx_credits -= credits;
tp->rcv_wnd += credits;
tp->rcv_adv += credits;
}
if (toep->ddp_waiting_count > 0 && sbavail(sb) != 0) {
CTR2(KTR_CXGBE, "%s: tid %u queueing AIO task", __func__,
tid);
ddp_queue_toep(toep);
}
sorwakeup_locked(so);
SOCKBUF_UNLOCK_ASSERT(sb);
if (toep->ulp_mode == ULP_MODE_TCPDDP)
DDP_UNLOCK(toep);
INP_WUNLOCK(inp);
CURVNET_RESTORE();
return (0);
}
/*
* TCP RST from the peer, timeout, or some other such critical error.
*/
static int
do_abort_req(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m)
{
struct adapter *sc = iq->adapter;
const struct cpl_abort_req_rss *cpl = (const void *)(rss + 1);
unsigned int tid = GET_TID(cpl);
struct toepcb *toep = lookup_tid(sc, tid);
struct sge_wrq *ofld_txq = toep->ofld_txq;
struct inpcb *inp;
struct tcpcb *tp;
#ifdef INVARIANTS
unsigned int opcode = G_CPL_OPCODE(be32toh(OPCODE_TID(cpl)));
#endif
KASSERT(opcode == CPL_ABORT_REQ_RSS,
("%s: unexpected opcode 0x%x", __func__, opcode));
KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__));
if (toep->flags & TPF_SYNQE)
return (do_abort_req_synqe(iq, rss, m));
KASSERT(toep->tid == tid, ("%s: toep tid mismatch", __func__));
if (negative_advice(cpl->status)) {
CTR4(KTR_CXGBE, "%s: negative advice %d for tid %d (0x%x)",
__func__, cpl->status, tid, toep->flags);
return (0); /* Ignore negative advice */
}
inp = toep->inp;
INP_INFO_RLOCK(&V_tcbinfo); /* for tcp_close */
INP_WLOCK(inp);
tp = intotcpcb(inp);
CTR6(KTR_CXGBE,
"%s: tid %d (%s), toep_flags 0x%x, inp_flags 0x%x, status %d",
__func__, tid, tp ? tcpstates[tp->t_state] : "no tp", toep->flags,
inp->inp_flags, cpl->status);
/*
* If we'd initiated an abort earlier the reply to it is responsible for
* cleaning up resources. Otherwise we tear everything down right here
* right now. We owe the T4 a CPL_ABORT_RPL no matter what.
*/
if (toep->flags & TPF_ABORT_SHUTDOWN) {
INP_WUNLOCK(inp);
goto done;
}
toep->flags |= TPF_ABORT_SHUTDOWN;
if ((inp->inp_flags & (INP_DROPPED | INP_TIMEWAIT)) == 0) {
struct socket *so = inp->inp_socket;
if (so != NULL)
so_error_set(so, abort_status_to_errno(tp,
cpl->status));
tp = tcp_close(tp);
if (tp == NULL)
INP_WLOCK(inp); /* re-acquire */
}
final_cpl_received(toep);
done:
INP_INFO_RUNLOCK(&V_tcbinfo);
send_abort_rpl(sc, ofld_txq, tid, CPL_ABORT_NO_RST);
return (0);
}