1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2012 Chelsio Communications, Inc. 5 * All rights reserved. 6 * Written by: Navdeep Parhar <np@FreeBSD.org> 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 */ 29 30 #include <sys/cdefs.h> 31 __FBSDID("$FreeBSD$"); 32 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 36 #ifdef TCP_OFFLOAD 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/kernel.h> 40 #include <sys/ktr.h> 41 #include <sys/module.h> 42 #include <sys/protosw.h> 43 #include <sys/domain.h> 44 #include <sys/socket.h> 45 #include <sys/socketvar.h> 46 #include <sys/sysctl.h> 47 #include <net/ethernet.h> 48 #include <net/if.h> 49 #include <net/if_types.h> 50 #include <net/if_vlan_var.h> 51 #include <net/route.h> 52 #include <net/route/nhop.h> 53 #include <netinet/in.h> 54 #include <netinet/in_pcb.h> 55 #include <netinet/ip.h> 56 #define TCPSTATES 57 #include <netinet/tcp_fsm.h> 58 #include <netinet/tcp_var.h> 59 #include <netinet/toecore.h> 60 #include <netinet/cc/cc.h> 61 62 #include "common/common.h" 63 #include "common/t4_msg.h" 64 #include "common/t4_regs.h" 65 #include "common/t4_regs_values.h" 66 #include "t4_clip.h" 67 #include "tom/t4_tom_l2t.h" 68 #include "tom/t4_tom.h" 69 70 /* 71 * Active open succeeded. 72 */ 73 static int 74 do_act_establish(struct sge_iq *iq, const struct rss_header *rss, 75 struct mbuf *m) 76 { 77 struct adapter *sc = iq->adapter; 78 const struct cpl_act_establish *cpl = (const void *)(rss + 1); 79 u_int tid = GET_TID(cpl); 80 u_int atid = G_TID_TID(ntohl(cpl->tos_atid)); 81 struct toepcb *toep = lookup_atid(sc, atid); 82 struct inpcb *inp = toep->inp; 83 84 KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__)); 85 KASSERT(toep->tid == atid, ("%s: toep tid/atid mismatch", __func__)); 86 87 CTR3(KTR_CXGBE, "%s: atid %u, tid %u", __func__, atid, tid); 88 free_atid(sc, atid); 89 90 CURVNET_SET(toep->vnet); 91 INP_WLOCK(inp); 92 toep->tid = tid; 93 insert_tid(sc, tid, toep, inp->inp_vflag & INP_IPV6 ? 2 : 1); 94 if (inp->inp_flags & INP_DROPPED) { 95 96 /* socket closed by the kernel before hw told us it connected */ 97 98 send_flowc_wr(toep, NULL); 99 send_reset(sc, toep, be32toh(cpl->snd_isn)); 100 goto done; 101 } 102 103 make_established(toep, be32toh(cpl->snd_isn) - 1, 104 be32toh(cpl->rcv_isn) - 1, cpl->tcp_opt); 105 inp->inp_flowtype = M_HASHTYPE_OPAQUE; 106 inp->inp_flowid = tid; 107 108 done: 109 INP_WUNLOCK(inp); 110 CURVNET_RESTORE(); 111 return (0); 112 } 113 114 void 115 act_open_failure_cleanup(struct adapter *sc, u_int atid, u_int status) 116 { 117 struct toepcb *toep = lookup_atid(sc, atid); 118 struct inpcb *inp = toep->inp; 119 struct toedev *tod = &toep->td->tod; 120 struct epoch_tracker et; 121 122 free_atid(sc, atid); 123 toep->tid = -1; 124 125 CURVNET_SET(toep->vnet); 126 if (status != EAGAIN) 127 NET_EPOCH_ENTER(et); 128 INP_WLOCK(inp); 129 toe_connect_failed(tod, inp, status); 130 final_cpl_received(toep); /* unlocks inp */ 131 if (status != EAGAIN) 132 NET_EPOCH_EXIT(et); 133 CURVNET_RESTORE(); 134 } 135 136 /* 137 * Active open failed. 138 */ 139 static int 140 do_act_open_rpl(struct sge_iq *iq, const struct rss_header *rss, 141 struct mbuf *m) 142 { 143 struct adapter *sc = iq->adapter; 144 const struct cpl_act_open_rpl *cpl = (const void *)(rss + 1); 145 u_int atid = G_TID_TID(G_AOPEN_ATID(be32toh(cpl->atid_status))); 146 u_int status = G_AOPEN_STATUS(be32toh(cpl->atid_status)); 147 struct toepcb *toep = lookup_atid(sc, atid); 148 int rc; 149 150 KASSERT(m == NULL, ("%s: wasn't expecting payload", __func__)); 151 KASSERT(toep->tid == atid, ("%s: toep tid/atid mismatch", __func__)); 152 153 CTR3(KTR_CXGBE, "%s: atid %u, status %u ", __func__, atid, status); 154 155 /* Ignore negative advice */ 156 if (negative_advice(status)) 157 return (0); 158 159 if (status && act_open_has_tid(status)) 160 release_tid(sc, GET_TID(cpl), toep->ctrlq); 161 162 rc = act_open_rpl_status_to_errno(status); 163 act_open_failure_cleanup(sc, atid, rc); 164 165 return (0); 166 } 167 168 void 169 t4_init_connect_cpl_handlers(void) 170 { 171 172 t4_register_cpl_handler(CPL_ACT_ESTABLISH, do_act_establish); 173 t4_register_shared_cpl_handler(CPL_ACT_OPEN_RPL, do_act_open_rpl, 174 CPL_COOKIE_TOM); 175 } 176 177 void 178 t4_uninit_connect_cpl_handlers(void) 179 { 180 181 t4_register_cpl_handler(CPL_ACT_ESTABLISH, NULL); 182 t4_register_shared_cpl_handler(CPL_ACT_OPEN_RPL, NULL, CPL_COOKIE_TOM); 183 } 184 185 #ifdef KTR 186 #define DONT_OFFLOAD_ACTIVE_OPEN(x) do { \ 187 reason = __LINE__; \ 188 rc = (x); \ 189 goto failed; \ 190 } while (0) 191 #else 192 #define DONT_OFFLOAD_ACTIVE_OPEN(x) do { \ 193 rc = (x); \ 194 goto failed; \ 195 } while (0) 196 #endif 197 198 static inline int 199 act_open_cpl_size(struct adapter *sc, int isipv6) 200 { 201 int idx; 202 static const int sz_table[3][2] = { 203 { 204 sizeof (struct cpl_act_open_req), 205 sizeof (struct cpl_act_open_req6) 206 }, 207 { 208 sizeof (struct cpl_t5_act_open_req), 209 sizeof (struct cpl_t5_act_open_req6) 210 }, 211 { 212 sizeof (struct cpl_t6_act_open_req), 213 sizeof (struct cpl_t6_act_open_req6) 214 }, 215 }; 216 217 MPASS(chip_id(sc) >= CHELSIO_T4); 218 idx = min(chip_id(sc) - CHELSIO_T4, 2); 219 220 return (sz_table[idx][!!isipv6]); 221 } 222 223 /* 224 * active open (soconnect). 225 * 226 * State of affairs on entry: 227 * soisconnecting (so_state |= SS_ISCONNECTING) 228 * tcbinfo not locked (This has changed - used to be WLOCKed) 229 * inp WLOCKed 230 * tp->t_state = TCPS_SYN_SENT 231 * rtalloc1, RT_UNLOCK on rt. 232 */ 233 int 234 t4_connect(struct toedev *tod, struct socket *so, struct nhop_object *nh, 235 struct sockaddr *nam) 236 { 237 struct adapter *sc = tod->tod_softc; 238 struct toepcb *toep = NULL; 239 struct wrqe *wr = NULL; 240 struct ifnet *rt_ifp = nh->nh_ifp; 241 struct vi_info *vi; 242 int qid_atid, rc, isipv6; 243 struct inpcb *inp = sotoinpcb(so); 244 struct tcpcb *tp = intotcpcb(inp); 245 #ifdef KTR 246 int reason; 247 #endif 248 struct offload_settings settings; 249 struct epoch_tracker et; 250 uint16_t vid = 0xfff, pcp = 0; 251 252 INP_WLOCK_ASSERT(inp); 253 KASSERT(nam->sa_family == AF_INET || nam->sa_family == AF_INET6, 254 ("%s: dest addr %p has family %u", __func__, nam, nam->sa_family)); 255 256 if (rt_ifp->if_type == IFT_ETHER) 257 vi = rt_ifp->if_softc; 258 else if (rt_ifp->if_type == IFT_L2VLAN) { 259 struct ifnet *ifp = VLAN_TRUNKDEV(rt_ifp); 260 261 vi = ifp->if_softc; 262 VLAN_TAG(rt_ifp, &vid); 263 VLAN_PCP(rt_ifp, &pcp); 264 } else if (rt_ifp->if_type == IFT_IEEE8023ADLAG) 265 DONT_OFFLOAD_ACTIVE_OPEN(ENOSYS); /* XXX: implement lagg+TOE */ 266 else 267 DONT_OFFLOAD_ACTIVE_OPEN(ENOTSUP); 268 if (sc->flags & KERN_TLS_ON) 269 DONT_OFFLOAD_ACTIVE_OPEN(ENOTSUP); 270 271 rw_rlock(&sc->policy_lock); 272 settings = *lookup_offload_policy(sc, OPEN_TYPE_ACTIVE, NULL, 273 EVL_MAKETAG(vid, pcp, 0), inp); 274 rw_runlock(&sc->policy_lock); 275 if (!settings.offload) 276 DONT_OFFLOAD_ACTIVE_OPEN(EPERM); 277 278 toep = alloc_toepcb(vi, M_NOWAIT); 279 if (toep == NULL) 280 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM); 281 282 toep->tid = alloc_atid(sc, toep); 283 if (toep->tid < 0) 284 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM); 285 286 toep->l2te = t4_l2t_get(vi->pi, rt_ifp, 287 nh->nh_flags & NHF_GATEWAY ? &nh->gw_sa : nam); 288 if (toep->l2te == NULL) 289 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM); 290 291 toep->vnet = so->so_vnet; 292 init_conn_params(vi, &settings, &inp->inp_inc, so, NULL, 293 toep->l2te->idx, &toep->params); 294 init_toepcb(vi, toep); 295 296 isipv6 = nam->sa_family == AF_INET6; 297 wr = alloc_wrqe(act_open_cpl_size(sc, isipv6), toep->ctrlq); 298 if (wr == NULL) 299 DONT_OFFLOAD_ACTIVE_OPEN(ENOMEM); 300 301 qid_atid = V_TID_QID(toep->ofld_rxq->iq.abs_id) | V_TID_TID(toep->tid) | 302 V_TID_COOKIE(CPL_COOKIE_TOM); 303 304 if (isipv6) { 305 struct cpl_act_open_req6 *cpl = wrtod(wr); 306 struct cpl_t5_act_open_req6 *cpl5 = (void *)cpl; 307 struct cpl_t6_act_open_req6 *cpl6 = (void *)cpl; 308 309 if ((inp->inp_vflag & INP_IPV6) == 0) 310 DONT_OFFLOAD_ACTIVE_OPEN(ENOTSUP); 311 312 toep->ce = t4_get_clip_entry(sc, &inp->in6p_laddr, true); 313 if (toep->ce == NULL) 314 DONT_OFFLOAD_ACTIVE_OPEN(ENOENT); 315 316 switch (chip_id(sc)) { 317 case CHELSIO_T4: 318 INIT_TP_WR(cpl, 0); 319 cpl->params = select_ntuple(vi, toep->l2te); 320 break; 321 case CHELSIO_T5: 322 INIT_TP_WR(cpl5, 0); 323 cpl5->iss = htobe32(tp->iss); 324 cpl5->params = select_ntuple(vi, toep->l2te); 325 break; 326 case CHELSIO_T6: 327 default: 328 INIT_TP_WR(cpl6, 0); 329 cpl6->iss = htobe32(tp->iss); 330 cpl6->params = select_ntuple(vi, toep->l2te); 331 break; 332 } 333 OPCODE_TID(cpl) = htobe32(MK_OPCODE_TID(CPL_ACT_OPEN_REQ6, 334 qid_atid)); 335 cpl->local_port = inp->inp_lport; 336 cpl->local_ip_hi = *(uint64_t *)&inp->in6p_laddr.s6_addr[0]; 337 cpl->local_ip_lo = *(uint64_t *)&inp->in6p_laddr.s6_addr[8]; 338 cpl->peer_port = inp->inp_fport; 339 cpl->peer_ip_hi = *(uint64_t *)&inp->in6p_faddr.s6_addr[0]; 340 cpl->peer_ip_lo = *(uint64_t *)&inp->in6p_faddr.s6_addr[8]; 341 cpl->opt0 = calc_options0(vi, &toep->params); 342 cpl->opt2 = calc_options2(vi, &toep->params); 343 344 CTR6(KTR_CXGBE, 345 "%s: atid %u, toep %p, inp %p, opt0 %#016lx, opt2 %#08x", 346 __func__, toep->tid, toep, inp, be64toh(cpl->opt0), 347 be32toh(cpl->opt2)); 348 } else { 349 struct cpl_act_open_req *cpl = wrtod(wr); 350 struct cpl_t5_act_open_req *cpl5 = (void *)cpl; 351 struct cpl_t6_act_open_req *cpl6 = (void *)cpl; 352 353 switch (chip_id(sc)) { 354 case CHELSIO_T4: 355 INIT_TP_WR(cpl, 0); 356 cpl->params = select_ntuple(vi, toep->l2te); 357 break; 358 case CHELSIO_T5: 359 INIT_TP_WR(cpl5, 0); 360 cpl5->iss = htobe32(tp->iss); 361 cpl5->params = select_ntuple(vi, toep->l2te); 362 break; 363 case CHELSIO_T6: 364 default: 365 INIT_TP_WR(cpl6, 0); 366 cpl6->iss = htobe32(tp->iss); 367 cpl6->params = select_ntuple(vi, toep->l2te); 368 break; 369 } 370 OPCODE_TID(cpl) = htobe32(MK_OPCODE_TID(CPL_ACT_OPEN_REQ, 371 qid_atid)); 372 inp_4tuple_get(inp, &cpl->local_ip, &cpl->local_port, 373 &cpl->peer_ip, &cpl->peer_port); 374 cpl->opt0 = calc_options0(vi, &toep->params); 375 cpl->opt2 = calc_options2(vi, &toep->params); 376 377 CTR6(KTR_CXGBE, 378 "%s: atid %u, toep %p, inp %p, opt0 %#016lx, opt2 %#08x", 379 __func__, toep->tid, toep, inp, be64toh(cpl->opt0), 380 be32toh(cpl->opt2)); 381 } 382 383 offload_socket(so, toep); 384 NET_EPOCH_ENTER(et); 385 rc = t4_l2t_send(sc, wr, toep->l2te); 386 NET_EPOCH_EXIT(et); 387 if (rc == 0) { 388 toep->flags |= TPF_CPL_PENDING; 389 return (0); 390 } 391 392 undo_offload_socket(so); 393 #if defined(KTR) 394 reason = __LINE__; 395 #endif 396 failed: 397 CTR3(KTR_CXGBE, "%s: not offloading (%d), rc %d", __func__, reason, rc); 398 399 if (wr) 400 free_wrqe(wr); 401 402 if (toep) { 403 if (toep->tid >= 0) 404 free_atid(sc, toep->tid); 405 if (toep->l2te) 406 t4_l2t_release(toep->l2te); 407 if (toep->ce) 408 t4_release_clip_entry(sc, toep->ce); 409 free_toepcb(toep); 410 } 411 412 return (rc); 413 } 414 #endif 415