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"
34 #include <sys/param.h>
35 #include <sys/kernel.h>
36 #include <sys/systm.h>
38 #include <sys/module.h>
39 #include <sys/types.h>
40 #include <sys/sockopt.h>
41 #include <sys/sysctl.h>
42 #include <sys/socket.h>
44 #include <net/ethernet.h>
46 #include <net/if_types.h>
47 #include <net/if_vlan_var.h>
48 #include <net/if_llatbl.h>
49 #include <net/route.h>
51 #include <netinet/if_ether.h>
52 #include <netinet/in.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_var.h>
55 #include <netinet6/nd6.h>
57 #include <netinet/tcp.h>
58 #include <netinet/tcp_fsm.h>
59 #include <netinet/tcp_timer.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/tcp_syncache.h>
62 #include <netinet/tcp_offload.h>
63 #include <netinet/toecore.h>
65 static struct mtx toedev_lock;
66 static TAILQ_HEAD(, toedev) toedev_list;
67 static eventhandler_tag listen_start_eh;
68 static eventhandler_tag listen_stop_eh;
69 static eventhandler_tag lle_event_eh;
70 static eventhandler_tag route_redirect_eh;
73 toedev_connect(struct toedev *tod __unused, struct socket *so __unused,
74 struct rtentry *rt __unused, struct sockaddr *nam __unused)
81 toedev_listen_start(struct toedev *tod __unused, struct tcpcb *tp __unused)
88 toedev_listen_stop(struct toedev *tod __unused, struct tcpcb *tp __unused)
95 toedev_input(struct toedev *tod __unused, struct tcpcb *tp __unused,
104 toedev_rcvd(struct toedev *tod __unused, struct tcpcb *tp __unused)
111 toedev_output(struct toedev *tod __unused, struct tcpcb *tp __unused)
118 toedev_pcb_detach(struct toedev *tod __unused, struct tcpcb *tp __unused)
125 toedev_l2_update(struct toedev *tod __unused, struct ifnet *ifp __unused,
126 struct sockaddr *sa __unused, uint8_t *lladdr __unused,
127 uint16_t vtag __unused)
134 toedev_route_redirect(struct toedev *tod __unused, struct ifnet *ifp __unused,
135 struct rtentry *rt0 __unused, struct rtentry *rt1 __unused)
142 toedev_syncache_added(struct toedev *tod __unused, void *ctx __unused)
149 toedev_syncache_removed(struct toedev *tod __unused, void *ctx __unused)
156 toedev_syncache_respond(struct toedev *tod __unused, void *ctx __unused,
165 toedev_offload_socket(struct toedev *tod __unused, void *ctx __unused,
166 struct socket *so __unused)
173 toedev_ctloutput(struct toedev *tod __unused, struct tcpcb *tp __unused,
174 int sopt_dir __unused, int sopt_name __unused)
181 * Inform one or more TOE devices about a listening socket.
184 toe_listen_start(struct inpcb *inp, void *arg)
186 struct toedev *t, *tod;
189 INP_WLOCK_ASSERT(inp);
190 KASSERT(inp->inp_pcbinfo == &V_tcbinfo,
191 ("%s: inp is not a TCP inp", __func__));
193 if (inp->inp_flags & (INP_DROPPED | INP_TIMEWAIT))
197 if (tp->t_state != TCPS_LISTEN)
201 mtx_lock(&toedev_lock);
202 TAILQ_FOREACH(tod, &toedev_list, link) {
203 if (t == NULL || t == tod)
204 tod->tod_listen_start(tod, tp);
206 mtx_unlock(&toedev_lock);
210 toe_listen_start_event(void *arg __unused, struct tcpcb *tp)
212 struct inpcb *inp = tp->t_inpcb;
214 INP_WLOCK_ASSERT(inp);
215 KASSERT(tp->t_state == TCPS_LISTEN,
216 ("%s: t_state %s", __func__, tcpstates[tp->t_state]));
218 toe_listen_start(inp, NULL);
222 toe_listen_stop_event(void *arg __unused, struct tcpcb *tp)
226 struct inpcb *inp = tp->t_inpcb;
229 INP_WLOCK_ASSERT(inp);
230 KASSERT(tp->t_state == TCPS_LISTEN,
231 ("%s: t_state %s", __func__, tcpstates[tp->t_state]));
233 mtx_lock(&toedev_lock);
234 TAILQ_FOREACH(tod, &toedev_list, link)
235 tod->tod_listen_stop(tod, tp);
236 mtx_unlock(&toedev_lock);
240 * Fill up a freshly allocated toedev struct with reasonable defaults.
243 init_toedev(struct toedev *tod)
246 tod->tod_softc = NULL;
249 * Provide no-op defaults so that the kernel can call any toedev
250 * function without having to check whether the TOE driver supplied one
253 tod->tod_connect = toedev_connect;
254 tod->tod_listen_start = toedev_listen_start;
255 tod->tod_listen_stop = toedev_listen_stop;
256 tod->tod_input = toedev_input;
257 tod->tod_rcvd = toedev_rcvd;
258 tod->tod_output = toedev_output;
259 tod->tod_send_rst = toedev_output;
260 tod->tod_send_fin = toedev_output;
261 tod->tod_pcb_detach = toedev_pcb_detach;
262 tod->tod_l2_update = toedev_l2_update;
263 tod->tod_route_redirect = toedev_route_redirect;
264 tod->tod_syncache_added = toedev_syncache_added;
265 tod->tod_syncache_removed = toedev_syncache_removed;
266 tod->tod_syncache_respond = toedev_syncache_respond;
267 tod->tod_offload_socket = toedev_offload_socket;
268 tod->tod_ctloutput = toedev_ctloutput;
272 * Register an active TOE device with the system. This allows it to receive
273 * notifications from the kernel.
276 register_toedev(struct toedev *tod)
280 mtx_lock(&toedev_lock);
281 TAILQ_FOREACH(t, &toedev_list, link) {
283 mtx_unlock(&toedev_lock);
288 TAILQ_INSERT_TAIL(&toedev_list, tod, link);
289 registered_toedevs++;
290 mtx_unlock(&toedev_lock);
292 inp_apply_all(toe_listen_start, tod);
298 * Remove the TOE device from the global list of active TOE devices. It is the
299 * caller's responsibility to ensure that the TOE device is quiesced prior to
303 unregister_toedev(struct toedev *tod)
305 struct toedev *t, *t2;
308 mtx_lock(&toedev_lock);
309 TAILQ_FOREACH_SAFE(t, &toedev_list, link, t2) {
311 TAILQ_REMOVE(&toedev_list, tod, link);
312 registered_toedevs--;
317 KASSERT(registered_toedevs >= 0,
318 ("%s: registered_toedevs (%d) < 0", __func__, registered_toedevs));
319 mtx_unlock(&toedev_lock);
324 toe_syncache_add(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th,
325 struct inpcb *inp, void *tod, void *todctx)
327 struct socket *lso = inp->inp_socket;
329 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
330 INP_WLOCK_ASSERT(inp);
332 syncache_add(inc, to, th, inp, &lso, NULL, tod, todctx);
336 toe_syncache_expand(struct in_conninfo *inc, struct tcpopt *to,
337 struct tcphdr *th, struct socket **lsop)
340 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
342 return (syncache_expand(inc, to, th, lsop, NULL));
346 * General purpose check to see if a 4-tuple is in use by the kernel. If a TCP
347 * header (presumably for an incoming SYN) is also provided, an existing 4-tuple
348 * in TIME_WAIT may be assassinated freeing it up for re-use.
350 * Note that the TCP header must have been run through tcp_fields_to_host() or
354 toe_4tuple_check(struct in_conninfo *inc, struct tcphdr *th, struct ifnet *ifp)
358 if (inc->inc_flags & INC_ISIPV6)
359 return (ENOSYS); /* XXX: implement */
361 inp = in_pcblookup(&V_tcbinfo, inc->inc_faddr, inc->inc_fport,
362 inc->inc_laddr, inc->inc_lport, INPLOOKUP_WLOCKPCB, ifp);
364 INP_WLOCK_ASSERT(inp);
366 if ((inp->inp_flags & INP_TIMEWAIT) && th != NULL) {
368 INP_INFO_WLOCK_ASSERT(&V_tcbinfo); /* for twcheck */
369 if (!tcp_twcheck(inp, NULL, th, NULL, 0))
381 toe_lle_event(void *arg __unused, struct llentry *lle, int evt)
389 LLE_WLOCK_ASSERT(lle);
391 ifp = lle->lle_tbl->llt_ifp;
394 KASSERT(sa->sa_family == AF_INET || sa->sa_family == AF_INET6,
395 ("%s: lle_event %d for lle %p but sa %p !INET && !INET6",
396 __func__, evt, lle, sa));
399 * Not interested if the interface's TOE capability is not enabled.
401 if ((sa->sa_family == AF_INET && !(ifp->if_capenable & IFCAP_TOE4)) ||
402 (sa->sa_family == AF_INET6 && !(ifp->if_capenable & IFCAP_TOE6)))
410 if (evt != LLENTRY_RESOLVED) {
413 * LLENTRY_TIMEDOUT, LLENTRY_DELETED, LLENTRY_EXPIRED all mean
414 * this entry is going to be deleted.
420 KASSERT(lle->la_flags & LLE_VALID,
421 ("%s: %p resolved but not valid?", __func__, lle));
423 lladdr = (uint8_t *)&lle->ll_addr;
425 VLAN_TAG(ifp, &vtag);
429 tod->tod_l2_update(tod, ifp, sa, lladdr, vtag);
436 toe_route_redirect_event(void *arg __unused, struct rtentry *rt0,
437 struct rtentry *rt1, struct sockaddr *sa)
444 * Returns 0 or EWOULDBLOCK on success (any other value is an error). 0 means
445 * lladdr and vtag are valid on return, EWOULDBLOCK means the TOE driver's
446 * tod_l2_update will be called later, when the entry is resolved or times out.
449 toe_l2_resolve(struct toedev *tod, struct ifnet *ifp, struct sockaddr *sa,
450 uint8_t *lladdr, uint16_t *vtag)
455 switch (sa->sa_family) {
458 rc = arpresolve(ifp, NULL, NULL, sa, lladdr, &lle);
463 rc = nd6_storelladdr(ifp, NULL, sa, lladdr, &lle);
467 return (EPROTONOSUPPORT);
472 if (VLAN_TAG(ifp, vtag) != 0)
481 toe_connect_failed(struct toedev *tod, struct tcpcb *tp, int err)
483 struct inpcb *inp = tp->t_inpcb;
485 INP_WLOCK_ASSERT(inp);
486 KASSERT(tp->t_flags & TF_TOE,
487 ("%s: tp %p not offloaded.", __func__, tp));
489 if (!(inp->inp_flags & INP_DROPPED)) {
493 * Temporary failure during offload, take this PCB back.
494 * Detach from the TOE driver and do the rest of what
495 * TCP's pru_connect would have done if the connection
499 tod->tod_pcb_detach(tod, tp);
500 KASSERT(!(tp->t_flags & TF_TOE),
501 ("%s: tp %p still offloaded.", __func__, tp));
502 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
503 (void) tcp_output(tp);
506 INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
507 tp = tcp_drop(tp, err);
509 INP_WLOCK(inp); /* re-acquire */
512 INP_WLOCK_ASSERT(inp);
519 mtx_init(&toedev_lock, "toedev lock", NULL, MTX_DEF);
520 TAILQ_INIT(&toedev_list);
522 listen_start_eh = EVENTHANDLER_REGISTER(tcp_offload_listen_start,
523 toe_listen_start_event, NULL, EVENTHANDLER_PRI_ANY);
524 listen_stop_eh = EVENTHANDLER_REGISTER(tcp_offload_listen_stop,
525 toe_listen_stop_event, NULL, EVENTHANDLER_PRI_ANY);
526 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, toe_lle_event, NULL,
527 EVENTHANDLER_PRI_ANY);
528 route_redirect_eh = EVENTHANDLER_REGISTER(route_redirect_event,
529 toe_route_redirect_event, NULL, EVENTHANDLER_PRI_ANY);
538 mtx_lock(&toedev_lock);
539 if (!TAILQ_EMPTY(&toedev_list)) {
540 mtx_unlock(&toedev_lock);
544 EVENTHANDLER_DEREGISTER(tcp_offload_listen_start, listen_start_eh);
545 EVENTHANDLER_DEREGISTER(tcp_offload_listen_stop, listen_stop_eh);
546 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
547 EVENTHANDLER_DEREGISTER(route_redirect_event, route_redirect_eh);
549 mtx_unlock(&toedev_lock);
550 mtx_destroy(&toedev_lock);
556 toecore_mod_handler(module_t mod, int cmd, void *arg)
560 return (toecore_load());
562 if (cmd == MOD_UNLOAD)
563 return (toecore_unload());
568 static moduledata_t mod_data= {
574 MODULE_VERSION(toecore, 1);
575 DECLARE_MODULE(toecore, mod_data, SI_SUB_EXEC, SI_ORDER_ANY);