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