2 * Copyright (c) 2012 Chelsio Communications, Inc.
4 * Written by: Navdeep Parhar <np@FreeBSD.org>
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
32 #include "opt_inet6.h"
35 #include <sys/param.h>
36 #include <sys/types.h>
37 #include <sys/kernel.h>
39 #include <sys/module.h>
40 #include <sys/protosw.h>
41 #include <sys/domain.h>
42 #include <sys/socket.h>
43 #include <sys/socketvar.h>
44 #include <net/ethernet.h>
46 #include <net/if_types.h>
47 #include <net/if_vlan_var.h>
48 #include <net/route.h>
49 #include <netinet/in.h>
50 #include <netinet/in_pcb.h>
51 #include <netinet/ip.h>
52 #include <netinet/tcp_var.h>
54 #include <netinet/tcp_fsm.h>
55 #include <netinet/toecore.h>
57 #include "common/common.h"
58 #include "common/t4_msg.h"
59 #include "common/t4_regs.h"
60 #include "common/t4_regs_values.h"
61 #include "tom/t4_tom_l2t.h"
62 #include "tom/t4_tom.h"
65 static int alloc_atid(struct adapter *, void *);
66 static void *lookup_atid(struct adapter *, int);
67 static void free_atid(struct adapter *, int);
70 alloc_atid(struct adapter *sc, void *ctx)
72 struct tid_info *t = &sc->tids;
75 mtx_lock(&t->atid_lock);
77 union aopen_entry *p = t->afree;
79 atid = p - t->atid_tab;
84 mtx_unlock(&t->atid_lock);
89 lookup_atid(struct adapter *sc, int atid)
91 struct tid_info *t = &sc->tids;
93 return (t->atid_tab[atid].data);
97 free_atid(struct adapter *sc, int atid)
99 struct tid_info *t = &sc->tids;
100 union aopen_entry *p = &t->atid_tab[atid];
102 mtx_lock(&t->atid_lock);
106 mtx_unlock(&t->atid_lock);
110 * Active open succeeded.
113 do_act_establish(struct sge_iq *iq, const struct rss_header *rss,
116 struct adapter *sc = iq->adapter;
117 const struct cpl_act_establish *cpl = (const void *)(rss + 1);
118 u_int tid = GET_TID(cpl);
119 u_int atid = G_TID_TID(ntohl(cpl->tos_atid));
120 struct toepcb *toep = lookup_atid(sc, atid);
121 struct inpcb *inp = toep->inp;
123 KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__));
124 KASSERT(toep->tid == atid, ("%s: toep tid/atid mismatch", __func__));
126 CTR3(KTR_CXGBE, "%s: atid %u, tid %u", __func__, atid, tid);
129 CURVNET_SET(toep->vnet);
132 insert_tid(sc, tid, toep, inp->inp_vflag & INP_IPV6 ? 2 : 1);
133 if (inp->inp_flags & INP_DROPPED) {
135 /* socket closed by the kernel before hw told us it connected */
137 send_flowc_wr(toep, NULL);
138 send_reset(sc, toep, be32toh(cpl->snd_isn));
142 make_established(toep, cpl->snd_isn, cpl->rcv_isn, cpl->tcp_opt);
150 * Convert an ACT_OPEN_RPL status to an errno.
153 act_open_rpl_status_to_errno(int status)
157 case CPL_ERR_CONN_RESET:
158 return (ECONNREFUSED);
159 case CPL_ERR_ARP_MISS:
160 return (EHOSTUNREACH);
161 case CPL_ERR_CONN_TIMEDOUT:
163 case CPL_ERR_TCAM_FULL:
165 case CPL_ERR_CONN_EXIST:
166 log(LOG_ERR, "ACTIVE_OPEN_RPL: 4-tuple in use\n");
174 act_open_failure_cleanup(struct adapter *sc, u_int atid, u_int status)
176 struct toepcb *toep = lookup_atid(sc, atid);
177 struct inpcb *inp = toep->inp;
178 struct toedev *tod = &toep->td->tod;
183 CURVNET_SET(toep->vnet);
184 if (status != EAGAIN)
185 INP_INFO_RLOCK(&V_tcbinfo);
187 toe_connect_failed(tod, inp, status);
188 final_cpl_received(toep); /* unlocks inp */
189 if (status != EAGAIN)
190 INP_INFO_RUNLOCK(&V_tcbinfo);
195 * Active open failed.
198 do_act_open_rpl(struct sge_iq *iq, const struct rss_header *rss,
201 struct adapter *sc = iq->adapter;
202 const struct cpl_act_open_rpl *cpl = (const void *)(rss + 1);
203 u_int atid = G_TID_TID(G_AOPEN_ATID(be32toh(cpl->atid_status)));
204 u_int status = G_AOPEN_STATUS(be32toh(cpl->atid_status));
205 struct toepcb *toep = lookup_atid(sc, atid);
208 KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__));
209 KASSERT(toep->tid == atid, ("%s: toep tid/atid mismatch", __func__));
211 CTR3(KTR_CXGBE, "%s: atid %u, status %u ", __func__, atid, status);
213 /* Ignore negative advice */
214 if (negative_advice(status))
217 if (status && act_open_has_tid(status))
218 release_tid(sc, GET_TID(cpl), toep->ctrlq);
220 rc = act_open_rpl_status_to_errno(status);
221 act_open_failure_cleanup(sc, atid, rc);
227 * Options2 for active open.
230 calc_opt2a(struct socket *so, struct toepcb *toep)
232 struct tcpcb *tp = so_sototcpcb(so);
233 struct port_info *pi = toep->vi->pi;
234 struct adapter *sc = pi->adapter;
237 opt2 = V_TX_QUEUE(sc->params.tp.tx_modq[pi->tx_chan]) |
238 F_RSS_QUEUE_VALID | V_RSS_QUEUE(toep->ofld_rxq->iq.abs_id);
240 if (tp->t_flags & TF_SACK_PERMIT)
243 if (tp->t_flags & TF_REQ_TSTMP)
244 opt2 |= F_TSTAMPS_EN;
246 if (tp->t_flags & TF_REQ_SCALE)
247 opt2 |= F_WND_SCALE_EN;
252 /* RX_COALESCE is always a valid value (M_RX_COALESCE). */
254 opt2 |= F_RX_COALESCE_VALID;
256 opt2 |= F_T5_OPT_2_VALID;
259 if (sc->tt.rx_coalesce)
260 opt2 |= V_RX_COALESCE(M_RX_COALESCE);
262 #ifdef USE_DDP_RX_FLOW_CONTROL
263 if (toep->ulp_mode == ULP_MODE_TCPDDP)
264 opt2 |= F_RX_FC_VALID | F_RX_FC_DDP;
267 return (htobe32(opt2));
271 t4_init_connect_cpl_handlers(void)
274 t4_register_cpl_handler(CPL_ACT_ESTABLISH, do_act_establish);
275 t4_register_cpl_handler(CPL_ACT_OPEN_RPL, do_act_open_rpl);
278 #define DONT_OFFLOAD_ACTIVE_OPEN(x) do { \
285 act_open_cpl_size(struct adapter *sc, int isipv6)
288 static const int sz_table[3][2] = {
290 sizeof (struct cpl_act_open_req),
291 sizeof (struct cpl_act_open_req6)
294 sizeof (struct cpl_t5_act_open_req),
295 sizeof (struct cpl_t5_act_open_req6)
298 sizeof (struct cpl_t6_act_open_req),
299 sizeof (struct cpl_t6_act_open_req6)
303 MPASS(chip_id(sc) >= CHELSIO_T4);
304 idx = min(chip_id(sc) - CHELSIO_T4, 2);
306 return (sz_table[idx][!!isipv6]);
310 * active open (soconnect).
312 * State of affairs on entry:
313 * soisconnecting (so_state |= SS_ISCONNECTING)
314 * tcbinfo not locked (This has changed - used to be WLOCKed)
316 * tp->t_state = TCPS_SYN_SENT
317 * rtalloc1, RT_UNLOCK on rt.
320 t4_connect(struct toedev *tod, struct socket *so, struct rtentry *rt,
321 struct sockaddr *nam)
323 struct adapter *sc = tod->tod_softc;
324 struct tom_data *td = tod_td(tod);
325 struct toepcb *toep = NULL;
326 struct wrqe *wr = NULL;
327 struct ifnet *rt_ifp = rt->rt_ifp;
329 int mtu_idx, rscale, qid_atid, rc, isipv6;
330 struct inpcb *inp = sotoinpcb(so);
331 struct tcpcb *tp = intotcpcb(inp);
334 INP_WLOCK_ASSERT(inp);
335 KASSERT(nam->sa_family == AF_INET || nam->sa_family == AF_INET6,
336 ("%s: dest addr %p has family %u", __func__, nam, nam->sa_family));
338 if (rt_ifp->if_type == IFT_ETHER)
339 vi = rt_ifp->if_softc;
340 else if (rt_ifp->if_type == IFT_L2VLAN) {
341 struct ifnet *ifp = VLAN_COOKIE(rt_ifp);
344 } else if (rt_ifp->if_type == IFT_IEEE8023ADLAG)
345 DONT_OFFLOAD_ACTIVE_OPEN(ENOSYS); /* XXX: implement lagg+TOE */
347 DONT_OFFLOAD_ACTIVE_OPEN(ENOTSUP);
349 toep = alloc_toepcb(vi, -1, -1, M_NOWAIT | M_ZERO);
351 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM);
353 toep->tid = alloc_atid(sc, toep);
355 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM);
357 toep->l2te = t4_l2t_get(vi->pi, rt_ifp,
358 rt->rt_flags & RTF_GATEWAY ? rt->rt_gateway : nam);
359 if (toep->l2te == NULL)
360 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM);
362 isipv6 = nam->sa_family == AF_INET6;
363 wr = alloc_wrqe(act_open_cpl_size(sc, isipv6), toep->ctrlq);
365 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM);
367 toep->vnet = so->so_vnet;
368 if (sc->tt.ddp && (so->so_options & SO_NO_DDP) == 0)
369 set_tcpddp_ulp_mode(toep);
371 toep->ulp_mode = ULP_MODE_NONE;
372 SOCKBUF_LOCK(&so->so_rcv);
373 /* opt0 rcv_bufsiz initially, assumes its normal meaning later */
374 toep->rx_credits = min(select_rcv_wnd(so) >> 10, M_RCV_BUFSIZ);
375 SOCKBUF_UNLOCK(&so->so_rcv);
378 * The kernel sets request_r_scale based on sb_max whereas we need to
379 * take hardware's MAX_RCV_WND into account too. This is normally a
380 * no-op as MAX_RCV_WND is much larger than the default sb_max.
382 if (tp->t_flags & TF_REQ_SCALE)
383 rscale = tp->request_r_scale = select_rcv_wscale();
386 mtu_idx = find_best_mtu_idx(sc, &inp->inp_inc, 0);
387 qid_atid = (toep->ofld_rxq->iq.abs_id << 14) | toep->tid;
390 struct cpl_act_open_req6 *cpl = wrtod(wr);
391 struct cpl_t5_act_open_req6 *cpl5 = (void *)cpl;
392 struct cpl_t6_act_open_req6 *cpl6 = (void *)cpl;
394 if ((inp->inp_vflag & INP_IPV6) == 0)
395 DONT_OFFLOAD_ACTIVE_OPEN(ENOTSUP);
397 toep->ce = hold_lip(td, &inp->in6p_laddr);
398 if (toep->ce == NULL)
399 DONT_OFFLOAD_ACTIVE_OPEN(ENOENT);
401 switch (chip_id(sc)) {
404 cpl->params = select_ntuple(vi, toep->l2te);
408 cpl5->iss = htobe32(tp->iss);
409 cpl5->params = select_ntuple(vi, toep->l2te);
414 cpl6->iss = htobe32(tp->iss);
415 cpl6->params = select_ntuple(vi, toep->l2te);
418 OPCODE_TID(cpl) = htobe32(MK_OPCODE_TID(CPL_ACT_OPEN_REQ6,
420 cpl->local_port = inp->inp_lport;
421 cpl->local_ip_hi = *(uint64_t *)&inp->in6p_laddr.s6_addr[0];
422 cpl->local_ip_lo = *(uint64_t *)&inp->in6p_laddr.s6_addr[8];
423 cpl->peer_port = inp->inp_fport;
424 cpl->peer_ip_hi = *(uint64_t *)&inp->in6p_faddr.s6_addr[0];
425 cpl->peer_ip_lo = *(uint64_t *)&inp->in6p_faddr.s6_addr[8];
426 cpl->opt0 = calc_opt0(so, vi, toep->l2te, mtu_idx, rscale,
427 toep->rx_credits, toep->ulp_mode);
428 cpl->opt2 = calc_opt2a(so, toep);
430 struct cpl_act_open_req *cpl = wrtod(wr);
431 struct cpl_t5_act_open_req *cpl5 = (void *)cpl;
432 struct cpl_t6_act_open_req *cpl6 = (void *)cpl;
434 switch (chip_id(sc)) {
437 cpl->params = select_ntuple(vi, toep->l2te);
441 cpl5->iss = htobe32(tp->iss);
442 cpl5->params = select_ntuple(vi, toep->l2te);
447 cpl6->iss = htobe32(tp->iss);
448 cpl6->params = select_ntuple(vi, toep->l2te);
451 OPCODE_TID(cpl) = htobe32(MK_OPCODE_TID(CPL_ACT_OPEN_REQ,
453 inp_4tuple_get(inp, &cpl->local_ip, &cpl->local_port,
454 &cpl->peer_ip, &cpl->peer_port);
455 cpl->opt0 = calc_opt0(so, vi, toep->l2te, mtu_idx, rscale,
456 toep->rx_credits, toep->ulp_mode);
457 cpl->opt2 = calc_opt2a(so, toep);
460 CTR5(KTR_CXGBE, "%s: atid %u (%s), toep %p, inp %p", __func__,
461 toep->tid, tcpstates[tp->t_state], toep, inp);
463 offload_socket(so, toep);
464 rc = t4_l2t_send(sc, wr, toep->l2te);
466 toep->flags |= TPF_CPL_PENDING;
470 undo_offload_socket(so);
473 CTR3(KTR_CXGBE, "%s: not offloading (%d), rc %d", __func__, reason, rc);
480 free_atid(sc, toep->tid);
482 t4_l2t_release(toep->l2te);
484 release_lip(td, toep->ce);