1/*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2012, 2015 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 * $FreeBSD$ 30 * 31 */ 32 33#ifndef __T4_TOM_H__ 34#define __T4_TOM_H__ 35#include <sys/vmem.h> 36#include "common/t4_hw.h" 37#include "common/t4_msg.h" 38#include "tom/t4_tls.h" 39 40#define LISTEN_HASH_SIZE 32 41 42/* 43 * Min receive window. We want it to be large enough to accommodate receive 44 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 45 */ 46#define MIN_RCV_WND (24 * 1024U) 47 48/* 49 * Max receive window supported by HW in bytes. Only a small part of it can 50 * be set through option0, the rest needs to be set through RX_DATA_ACK. 51 */ 52#define MAX_RCV_WND ((1U << 27) - 1) 53 54#define DDP_RSVD_WIN (16 * 1024U) 55#define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 56 57#define USE_DDP_RX_FLOW_CONTROL 58 59#define PPOD_SZ(n) ((n) * sizeof(struct pagepod)) 60#define PPOD_SIZE (PPOD_SZ(1)) 61 62/* TOE PCB flags */ 63enum { 64 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 65 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 66 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 67 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 68 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 69 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 70 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 71 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 72 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 73 TPF_SYNQE_EXPANDED = (1 << 9), /* toepcb ready, tid context updated */ 74 TPF_FORCE_CREDITS = (1 << 10), /* always send credits */ 75 TPF_INITIALIZED = (1 << 12), /* init_toepcb has been called */ 76 TPF_TLS_RECEIVE = (1 << 13), /* should receive TLS records */ 77}; 78 79enum { 80 DDP_OK = (1 << 0), /* OK to turn on DDP */ 81 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */ 82 DDP_ON = (1 << 2), /* DDP is turned on */ 83 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */ 84 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */ 85 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */ 86 DDP_DEAD = (1 << 6), /* toepcb is shutting down */ 87}; 88 89struct sockopt; 90struct offload_settings; 91 92/* 93 * Connection parameters for an offloaded connection. These are mostly (but not 94 * all) hardware TOE parameters. 95 */ 96struct conn_params { 97 int8_t rx_coalesce; 98 int8_t cong_algo; 99 int8_t tc_idx; 100 int8_t tstamp; 101 int8_t sack; 102 int8_t nagle; 103 int8_t keepalive; 104 int8_t wscale; 105 int8_t ecn; 106 int8_t mtu_idx; 107 int8_t ulp_mode; 108 int8_t tx_align; 109 int16_t txq_idx; /* ofld_txq = &sc->sge.ofld_txq[txq_idx] */ 110 int16_t rxq_idx; /* ofld_rxq = &sc->sge.ofld_rxq[rxq_idx] */ 111 int16_t l2t_idx; 112 uint16_t emss; 113 uint16_t opt0_bufsize; 114 u_int sndbuf; /* controls TP tx pages */ 115}; 116 117struct ofld_tx_sdesc { 118 uint32_t plen; /* payload length */ 119 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 120 void *iv_buffer; /* optional buffer holding IVs for TLS */ 121}; 122 123struct ppod_region { 124 u_int pr_start; 125 u_int pr_len; 126 u_int pr_page_shift[4]; 127 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 128 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 129 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 130 u_int pr_alias_shift; /* shift this much for first alias bit. */ 131 vmem_t *pr_arena; 132}; 133 134struct ppod_reservation { 135 struct ppod_region *prsv_pr; 136 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 137 u_int prsv_nppods; 138}; 139 140struct pageset { 141 TAILQ_ENTRY(pageset) link; 142 vm_page_t *pages; 143 int npages; 144 int flags; 145 int offset; /* offset in first page */ 146 int len; 147 struct ppod_reservation prsv; 148 struct vmspace *vm; 149 vm_offset_t start; 150 u_int vm_timestamp; 151}; 152 153TAILQ_HEAD(pagesetq, pageset); 154 155#define PS_WIRED 0x0001 /* Pages wired rather than held. */ 156#define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 157 158#define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 159 160#define IS_AIOTX_MBUF(m) \ 161 ((m)->m_flags & M_EXT && (m)->m_ext.ext_flags & EXT_FLAG_AIOTX) 162 163struct ddp_buffer { 164 struct pageset *ps; 165 166 struct kaiocb *job; 167 int cancel_pending; 168}; 169 170struct ddp_pcb { 171 u_int flags; 172 struct ddp_buffer db[2]; 173 TAILQ_HEAD(, pageset) cached_pagesets; 174 TAILQ_HEAD(, kaiocb) aiojobq; 175 u_int waiting_count; 176 u_int active_count; 177 u_int cached_count; 178 int active_id; /* the currently active DDP buffer */ 179 struct task requeue_task; 180 struct kaiocb *queueing; 181 struct mtx lock; 182}; 183 184struct aiotx_buffer { 185 struct pageset ps; 186 struct kaiocb *job; 187 int refcount; 188}; 189 190struct toepcb { 191 struct tom_data *td; 192 struct inpcb *inp; /* backpointer to host stack's PCB */ 193 u_int flags; /* miscellaneous flags */ 194 TAILQ_ENTRY(toepcb) link; /* toep_list */ 195 int refcount; 196 struct vnet *vnet; 197 struct vi_info *vi; /* virtual interface */ 198 struct sge_wrq *ofld_txq; 199 struct sge_ofld_rxq *ofld_rxq; 200 struct sge_wrq *ctrlq; 201 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 202 struct clip_entry *ce; /* CLIP table entry used by this tid */ 203 int tid; /* Connection identifier */ 204 205 /* tx credit handling */ 206 u_int tx_total; /* total tx WR credits (in 16B units) */ 207 u_int tx_credits; /* tx WR credits (in 16B units) available */ 208 u_int tx_nocompl; /* tx WR credits since last compl request */ 209 u_int plen_nocompl; /* payload since last compl request */ 210 211 struct conn_params params; 212 213 void *ulpcb; 214 void *ulpcb2; 215 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 216 struct mbufq ulp_pdu_reclaimq; 217 218 struct ddp_pcb ddp; 219 struct tls_ofld_info tls; 220 221 TAILQ_HEAD(, kaiocb) aiotx_jobq; 222 struct task aiotx_task; 223 struct socket *aiotx_so; 224 225 /* Tx software descriptor */ 226 uint8_t txsd_total; 227 uint8_t txsd_pidx; 228 uint8_t txsd_cidx; 229 uint8_t txsd_avail; 230 struct ofld_tx_sdesc txsd[]; 231}; 232 233static inline int 234ulp_mode(struct toepcb *toep) 235{ 236 237 return (toep->params.ulp_mode); 238} 239 240#define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 241#define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 242#define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 243 244/* 245 * Compressed state for embryonic connections for a listener. 246 */ 247struct synq_entry { 248 struct listen_ctx *lctx; /* backpointer to listen ctx */ 249 struct mbuf *syn; 250 int flags; /* same as toepcb's tp_flags */ 251 volatile int ok_to_respond; 252 volatile u_int refcnt; 253 int tid; 254 uint32_t iss; 255 uint32_t irs; 256 uint32_t ts; 257 __be16 tcp_opt; /* from cpl_pass_establish */ 258 struct toepcb *toep; 259 260 struct conn_params params; 261}; 262 263/* listen_ctx flags */ 264#define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 265 266struct listen_ctx { 267 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 268 volatile int refcount; 269 int stid; 270 struct stid_region stid_region; 271 int flags; 272 struct inpcb *inp; /* listening socket's inp */ 273 struct vnet *vnet; 274 struct sge_wrq *ctrlq; 275 struct sge_ofld_rxq *ofld_rxq; 276 struct clip_entry *ce; 277}; 278 279/* tcb_histent flags */ 280#define TE_RPL_PENDING 1 281#define TE_ACTIVE 2 282 283/* bits in one 8b tcb_histent sample. */ 284#define TS_RTO (1 << 0) 285#define TS_DUPACKS (1 << 1) 286#define TS_FASTREXMT (1 << 2) 287#define TS_SND_BACKLOGGED (1 << 3) 288#define TS_CWND_LIMITED (1 << 4) 289#define TS_ECN_ECE (1 << 5) 290#define TS_ECN_CWR (1 << 6) 291#define TS_RESERVED (1 << 7) /* Unused. */ 292 293struct tcb_histent { 294 struct mtx te_lock; 295 struct callout te_callout; 296 uint64_t te_tcb[TCB_SIZE / sizeof(uint64_t)]; 297 struct adapter *te_adapter; 298 u_int te_flags; 299 u_int te_tid; 300 uint8_t te_pidx; 301 uint8_t te_sample[100]; 302}; 303 304struct tom_data { 305 struct toedev tod; 306 307 /* toepcb's associated with this TOE device */ 308 struct mtx toep_list_lock; 309 TAILQ_HEAD(, toepcb) toep_list; 310 311 struct mtx lctx_hash_lock; 312 LIST_HEAD(, listen_ctx) *listen_hash; 313 u_long listen_mask; 314 int lctx_count; /* # of lctx in the hash table */ 315 316 struct ppod_region pr; 317 318 struct rwlock tcb_history_lock __aligned(CACHE_LINE_SIZE); 319 struct tcb_histent **tcb_history; 320 int dupack_threshold; 321 322 /* WRs that will not be sent to the chip because L2 resolution failed */ 323 struct mtx unsent_wr_lock; 324 STAILQ_HEAD(, wrqe) unsent_wr_list; 325 struct task reclaim_wr_resources; 326}; 327 328static inline struct tom_data * 329tod_td(struct toedev *tod) 330{ 331 332 return (__containerof(tod, struct tom_data, tod)); 333} 334 335static inline struct adapter * 336td_adapter(struct tom_data *td) 337{ 338 339 return (td->tod.tod_softc); 340} 341 342static inline void 343set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 344{ 345 346 M_ASSERTPKTHDR(m); 347 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 348} 349 350static inline uint8_t 351mbuf_ulp_submode(struct mbuf *m) 352{ 353 354 M_ASSERTPKTHDR(m); 355 return (m->m_pkthdr.PH_per.eight[0]); 356} 357 358/* t4_tom.c */ 359struct toepcb *alloc_toepcb(struct vi_info *, int); 360int init_toepcb(struct vi_info *, struct toepcb *); 361struct toepcb *hold_toepcb(struct toepcb *); 362void free_toepcb(struct toepcb *); 363void offload_socket(struct socket *, struct toepcb *); 364void undo_offload_socket(struct socket *); 365void final_cpl_received(struct toepcb *); 366void insert_tid(struct adapter *, int, void *, int); 367void *lookup_tid(struct adapter *, int); 368void update_tid(struct adapter *, int, void *); 369void remove_tid(struct adapter *, int, int); 370u_long select_rcv_wnd(struct socket *); 371int select_rcv_wscale(void); 372void init_conn_params(struct vi_info *, struct offload_settings *, 373 struct in_conninfo *, struct socket *, const struct tcp_options *, int16_t, 374 struct conn_params *cp); 375__be64 calc_options0(struct vi_info *, struct conn_params *); 376__be32 calc_options2(struct vi_info *, struct conn_params *); 377uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 378int negative_advice(int); 379int add_tid_to_history(struct adapter *, u_int); 380 381/* t4_connect.c */ 382void t4_init_connect_cpl_handlers(void); 383void t4_uninit_connect_cpl_handlers(void); 384int t4_connect(struct toedev *, struct socket *, struct rtentry *, 385 struct sockaddr *); 386void act_open_failure_cleanup(struct adapter *, u_int, u_int); 387 388/* t4_listen.c */ 389void t4_init_listen_cpl_handlers(void); 390void t4_uninit_listen_cpl_handlers(void); 391int t4_listen_start(struct toedev *, struct tcpcb *); 392int t4_listen_stop(struct toedev *, struct tcpcb *); 393void t4_syncache_added(struct toedev *, void *); 394void t4_syncache_removed(struct toedev *, void *); 395int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 396int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 397 struct mbuf *); 398int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 399 struct mbuf *); 400void t4_offload_socket(struct toedev *, void *, struct socket *); 401void synack_failure_cleanup(struct adapter *, int); 402 403/* t4_cpl_io.c */ 404void aiotx_init_toep(struct toepcb *); 405int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 406void t4_init_cpl_io_handlers(void); 407void t4_uninit_cpl_io_handlers(void); 408void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 409void send_flowc_wr(struct toepcb *, struct tcpcb *); 410void send_reset(struct adapter *, struct toepcb *, uint32_t); 411int send_rx_credits(struct adapter *, struct toepcb *, int); 412void send_rx_modulate(struct adapter *, struct toepcb *); 413void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 414int t4_close_conn(struct adapter *, struct toepcb *); 415void t4_rcvd(struct toedev *, struct tcpcb *); 416void t4_rcvd_locked(struct toedev *, struct tcpcb *); 417int t4_tod_output(struct toedev *, struct tcpcb *); 418int t4_send_fin(struct toedev *, struct tcpcb *); 419int t4_send_rst(struct toedev *, struct tcpcb *); 420void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *, 421 uint16_t, uint64_t, uint64_t, int, int); 422void t4_push_frames(struct adapter *, struct toepcb *, int); 423void t4_push_pdus(struct adapter *, struct toepcb *, int); 424 425/* t4_ddp.c */ 426int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 427 const char *); 428void t4_free_ppod_region(struct ppod_region *); 429int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 430int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 431 struct ppod_reservation *); 432int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 433 struct pageset *); 434int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 435 struct ppod_reservation *, vm_offset_t, int); 436void t4_free_page_pods(struct ppod_reservation *); 437int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 438 struct mbuf **, struct mbuf **, int *); 439int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 440void t4_ddp_mod_load(void); 441void t4_ddp_mod_unload(void); 442void ddp_assert_empty(struct toepcb *); 443void ddp_init_toep(struct toepcb *); 444void ddp_uninit_toep(struct toepcb *); 445void ddp_queue_toep(struct toepcb *); 446void release_ddp_resources(struct toepcb *toep); 447void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 448void handle_ddp_indicate(struct toepcb *); 449void insert_ddp_data(struct toepcb *, uint32_t); 450const struct offload_settings *lookup_offload_policy(struct adapter *, int, 451 struct mbuf *, uint16_t, struct inpcb *); 452 453/* t4_tls.c */ 454bool can_tls_offload(struct adapter *); 455void do_rx_data_tls(const struct cpl_rx_data *, struct toepcb *, struct mbuf *); 456int t4_ctloutput_tls(struct socket *, struct sockopt *); 457void t4_push_tls_records(struct adapter *, struct toepcb *, int); 458void t4_tls_mod_load(void); 459void t4_tls_mod_unload(void); 460void tls_establish(struct toepcb *); 461void tls_init_toep(struct toepcb *); 462int tls_rx_key(struct toepcb *); 463void tls_stop_handshake_timer(struct toepcb *); 464int tls_tx_key(struct toepcb *); 465void tls_uninit_toep(struct toepcb *); 466 467#endif 468