2 * Copyright (c) 2012, 2015 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
35 #define LISTEN_HASH_SIZE 32
38 * Min receive window. We want it to be large enough to accommodate receive
39 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance.
41 #define MIN_RCV_WND (24 * 1024U)
44 * Max receive window supported by HW in bytes. Only a small part of it can
45 * be set through option0, the rest needs to be set through RX_DATA_ACK.
47 #define MAX_RCV_WND ((1U << 27) - 1)
49 #define DDP_RSVD_WIN (16 * 1024U)
50 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */
52 #define USE_DDP_RX_FLOW_CONTROL
56 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */
57 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */
58 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */
59 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */
60 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */
61 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */
62 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */
63 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */
64 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */
65 TPF_SYNQE_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */
66 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */
67 TPF_SYNQE_EXPANDED = (1 << 11), /* toepcb ready, tid context updated */
68 TPF_SYNQE_HAS_L2TE = (1 << 12), /* we've replied to PASS_ACCEPT_REQ */
72 DDP_OK = (1 << 0), /* OK to turn on DDP */
73 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */
74 DDP_ON = (1 << 2), /* DDP is turned on */
75 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */
76 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */
79 struct ofld_tx_sdesc {
80 uint32_t plen; /* payload length */
81 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */
85 uint32_t tag; /* includes color, page pod addr, and DDP page size */
95 TAILQ_ENTRY(toepcb) link; /* toep_list */
96 u_int flags; /* miscellaneous flags */
98 struct inpcb *inp; /* backpointer to host stack's PCB */
99 struct vi_info *vi; /* virtual interface */
100 struct sge_wrq *ofld_txq;
101 struct sge_ofld_rxq *ofld_rxq;
102 struct sge_wrq *ctrlq;
103 struct l2t_entry *l2te; /* L2 table entry used by this connection */
104 struct clip_entry *ce; /* CLIP table entry used by this tid */
105 int tid; /* Connection identifier */
107 /* tx credit handling */
108 u_int tx_total; /* total tx WR credits (in 16B units) */
109 u_int tx_credits; /* tx WR credits (in 16B units) available */
110 u_int tx_nocompl; /* tx WR credits since last compl request */
111 u_int plen_nocompl; /* payload since last compl request */
113 /* rx credit handling */
114 u_int sb_cc; /* last noted value of so_rcv->sb_cc */
115 int rx_credits; /* rx credits (in bytes) to be returned to hw */
117 u_int ulp_mode; /* ULP mode */
120 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */
121 struct mbufq ulp_pdu_reclaimq;
124 struct ddp_buffer *db[2];
128 /* Tx software descriptor */
133 struct ofld_tx_sdesc txsd[];
136 struct flowc_tx_params {
139 unsigned int snd_space;
143 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */
144 #define DDP_LOW_SCORE 1
145 #define DDP_HIGH_SCORE 3
148 * Compressed state for embryonic connections for a listener. Barely fits in
149 * 64B, try not to grow it further.
152 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */
153 int flags; /* same as toepcb's tp_flags */
155 struct listen_ctx *lctx; /* backpointer to listen ctx */
159 volatile uintptr_t wr;
160 volatile u_int refcnt;
162 uint16_t rcv_bufsize;
165 /* listen_ctx flags */
166 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */
169 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */
170 volatile int refcount;
172 struct stid_region stid_region;
174 struct inpcb *inp; /* listening socket's inp */
175 struct sge_wrq *ctrlq;
176 struct sge_ofld_rxq *ofld_rxq;
177 struct clip_entry *ce;
178 TAILQ_HEAD(, synq_entry) synq;
182 TAILQ_ENTRY(clip_entry) link;
183 struct in6_addr lip; /* local IPv6 address */
187 TAILQ_HEAD(clip_head, clip_entry);
191 /* toepcb's associated with this TOE device */
192 struct mtx toep_list_lock;
193 TAILQ_HEAD(, toepcb) toep_list;
195 struct mtx lctx_hash_lock;
196 LIST_HEAD(, listen_ctx) *listen_hash;
198 int lctx_count; /* # of lctx in the hash table */
203 struct mtx clip_table_lock;
204 struct clip_head clip_table;
207 /* WRs that will not be sent to the chip because L2 resolution failed */
208 struct mtx unsent_wr_lock;
209 STAILQ_HEAD(, wrqe) unsent_wr_list;
210 struct task reclaim_wr_resources;
213 static inline struct tom_data *
214 tod_td(struct toedev *tod)
217 return (__containerof(tod, struct tom_data, tod));
220 static inline struct adapter *
221 td_adapter(struct tom_data *td)
224 return (td->tod.tod_softc);
228 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode)
232 m->m_pkthdr.PH_per.eight[0] = ulp_submode;
235 static inline uint8_t
236 mbuf_ulp_submode(struct mbuf *m)
240 return (m->m_pkthdr.PH_per.eight[0]);
244 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int);
245 void free_toepcb(struct toepcb *);
246 void offload_socket(struct socket *, struct toepcb *);
247 void undo_offload_socket(struct socket *);
248 void final_cpl_received(struct toepcb *);
249 void insert_tid(struct adapter *, int, void *);
250 void *lookup_tid(struct adapter *, int);
251 void update_tid(struct adapter *, int, void *);
252 void remove_tid(struct adapter *, int);
253 void release_tid(struct adapter *, int, struct sge_wrq *);
254 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int);
255 u_long select_rcv_wnd(struct socket *);
256 int select_rcv_wscale(void);
257 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *,
259 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *);
260 void set_tcpddp_ulp_mode(struct toepcb *);
261 int negative_advice(int);
262 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *);
263 void release_lip(struct tom_data *, struct clip_entry *);
266 void t4_init_connect_cpl_handlers(struct adapter *);
267 int t4_connect(struct toedev *, struct socket *, struct rtentry *,
269 void act_open_failure_cleanup(struct adapter *, u_int, u_int);
272 void t4_init_listen_cpl_handlers(struct adapter *);
273 int t4_listen_start(struct toedev *, struct tcpcb *);
274 int t4_listen_stop(struct toedev *, struct tcpcb *);
275 void t4_syncache_added(struct toedev *, void *);
276 void t4_syncache_removed(struct toedev *, void *);
277 int t4_syncache_respond(struct toedev *, void *, struct mbuf *);
278 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *,
280 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *,
282 void t4_offload_socket(struct toedev *, void *, struct socket *);
285 void t4_init_cpl_io_handlers(struct adapter *);
286 void t4_uninit_cpl_io_handlers(struct adapter *);
287 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int);
288 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *);
289 void send_reset(struct adapter *, struct toepcb *, uint32_t);
290 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t);
291 void t4_rcvd(struct toedev *, struct tcpcb *);
292 int t4_tod_output(struct toedev *, struct tcpcb *);
293 int t4_send_fin(struct toedev *, struct tcpcb *);
294 int t4_send_rst(struct toedev *, struct tcpcb *);
295 void t4_set_tcb_field(struct adapter *, struct toepcb *, int, uint16_t,
297 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop);
298 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop);
301 void t4_init_ddp(struct adapter *, struct tom_data *);
302 void t4_uninit_ddp(struct adapter *, struct tom_data *);
303 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *,
304 struct mbuf **, struct mbuf **, int *);
305 void enable_ddp(struct adapter *, struct toepcb *toep);
306 void release_ddp_resources(struct toepcb *toep);
307 void handle_ddp_close(struct toepcb *, struct tcpcb *, struct sockbuf *,
309 void insert_ddp_data(struct toepcb *, uint32_t);