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 "tom/t4_tls.h" 37 38 #define LISTEN_HASH_SIZE 32 39 40 /* 41 * Min receive window. We want it to be large enough to accommodate receive 42 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 43 */ 44 #define MIN_RCV_WND (24 * 1024U) 45 46 /* 47 * Max receive window supported by HW in bytes. Only a small part of it can 48 * be set through option0, the rest needs to be set through RX_DATA_ACK. 49 */ 50 #define MAX_RCV_WND ((1U << 27) - 1) 51 52 #define DDP_RSVD_WIN (16 * 1024U) 53 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 54 55 #define USE_DDP_RX_FLOW_CONTROL 56 57 #define PPOD_SZ(n) ((n) * sizeof(struct pagepod)) 58 #define PPOD_SIZE (PPOD_SZ(1)) 59 60 /* TOE PCB flags */ 61 enum { 62 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 63 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 64 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 65 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 66 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 67 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 68 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 69 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 70 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 71 TPF_SYNQE_EXPANDED = (1 << 9), /* toepcb ready, tid context updated */ 72 TPF_FORCE_CREDITS = (1 << 10), /* always send credits */ 73 }; 74 75 enum { 76 DDP_OK = (1 << 0), /* OK to turn on DDP */ 77 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */ 78 DDP_ON = (1 << 2), /* DDP is turned on */ 79 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */ 80 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */ 81 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */ 82 DDP_DEAD = (1 << 6), /* toepcb is shutting down */ 83 }; 84 85 struct sockopt; 86 struct offload_settings; 87 88 struct ofld_tx_sdesc { 89 uint32_t plen; /* payload length */ 90 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 91 void *iv_buffer; /* optional buffer holding IVs for TLS */ 92 }; 93 94 struct ppod_region { 95 u_int pr_start; 96 u_int pr_len; 97 u_int pr_page_shift[4]; 98 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 99 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 100 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 101 u_int pr_alias_shift; /* shift this much for first alias bit. */ 102 vmem_t *pr_arena; 103 }; 104 105 struct ppod_reservation { 106 struct ppod_region *prsv_pr; 107 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 108 u_int prsv_nppods; 109 }; 110 111 struct pageset { 112 TAILQ_ENTRY(pageset) link; 113 vm_page_t *pages; 114 int npages; 115 int flags; 116 int offset; /* offset in first page */ 117 int len; 118 struct ppod_reservation prsv; 119 struct vmspace *vm; 120 vm_offset_t start; 121 u_int vm_timestamp; 122 }; 123 124 TAILQ_HEAD(pagesetq, pageset); 125 126 #define PS_WIRED 0x0001 /* Pages wired rather than held. */ 127 #define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 128 129 #define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 130 131 #define IS_AIOTX_MBUF(m) \ 132 ((m)->m_flags & M_EXT && (m)->m_ext.ext_flags & EXT_FLAG_AIOTX) 133 134 struct ddp_buffer { 135 struct pageset *ps; 136 137 struct kaiocb *job; 138 int cancel_pending; 139 }; 140 141 struct ddp_pcb { 142 u_int flags; 143 struct ddp_buffer db[2]; 144 TAILQ_HEAD(, pageset) cached_pagesets; 145 TAILQ_HEAD(, kaiocb) aiojobq; 146 u_int waiting_count; 147 u_int active_count; 148 u_int cached_count; 149 int active_id; /* the currently active DDP buffer */ 150 struct task requeue_task; 151 struct kaiocb *queueing; 152 struct mtx lock; 153 }; 154 155 struct aiotx_buffer { 156 struct pageset ps; 157 struct kaiocb *job; 158 int refcount; 159 }; 160 161 struct toepcb { 162 TAILQ_ENTRY(toepcb) link; /* toep_list */ 163 u_int flags; /* miscellaneous flags */ 164 int refcount; 165 struct tom_data *td; 166 struct inpcb *inp; /* backpointer to host stack's PCB */ 167 struct vnet *vnet; 168 struct vi_info *vi; /* virtual interface */ 169 struct sge_wrq *ofld_txq; 170 struct sge_ofld_rxq *ofld_rxq; 171 struct sge_wrq *ctrlq; 172 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 173 struct clip_entry *ce; /* CLIP table entry used by this tid */ 174 int tid; /* Connection identifier */ 175 int tc_idx; /* traffic class that this tid is bound to */ 176 177 /* tx credit handling */ 178 u_int tx_total; /* total tx WR credits (in 16B units) */ 179 u_int tx_credits; /* tx WR credits (in 16B units) available */ 180 u_int tx_nocompl; /* tx WR credits since last compl request */ 181 u_int plen_nocompl; /* payload since last compl request */ 182 183 /* rx credit handling */ 184 u_int sb_cc; /* last noted value of so_rcv->sb_cc */ 185 int rx_credits; /* rx credits (in bytes) to be returned to hw */ 186 187 u_int ulp_mode; /* ULP mode */ 188 void *ulpcb; 189 void *ulpcb2; 190 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 191 struct mbufq ulp_pdu_reclaimq; 192 193 struct ddp_pcb ddp; 194 struct tls_ofld_info tls; 195 196 TAILQ_HEAD(, kaiocb) aiotx_jobq; 197 struct task aiotx_task; 198 bool aiotx_task_active; 199 200 /* Tx software descriptor */ 201 uint8_t txsd_total; 202 uint8_t txsd_pidx; 203 uint8_t txsd_cidx; 204 uint8_t txsd_avail; 205 struct ofld_tx_sdesc txsd[]; 206 }; 207 208 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 209 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 210 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 211 212 struct flowc_tx_params { 213 uint32_t snd_nxt; 214 uint32_t rcv_nxt; 215 unsigned int snd_space; 216 unsigned int mss; 217 }; 218 219 /* 220 * Compressed state for embryonic connections for a listener. 221 */ 222 struct synq_entry { 223 struct listen_ctx *lctx; /* backpointer to listen ctx */ 224 struct mbuf *syn; 225 int flags; /* same as toepcb's tp_flags */ 226 volatile int ok_to_respond; 227 volatile u_int refcnt; 228 int tid; 229 uint32_t iss; 230 uint32_t irs; 231 uint32_t ts; 232 uint16_t txqid; 233 uint16_t rxqid; 234 uint16_t l2e_idx; 235 uint16_t ulp_mode; 236 uint16_t rcv_bufsize; 237 __be16 tcp_opt; /* from cpl_pass_establish */ 238 struct toepcb *toep; 239 }; 240 241 /* listen_ctx flags */ 242 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 243 244 struct listen_ctx { 245 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 246 volatile int refcount; 247 int stid; 248 struct stid_region stid_region; 249 int flags; 250 struct inpcb *inp; /* listening socket's inp */ 251 struct vnet *vnet; 252 struct sge_wrq *ctrlq; 253 struct sge_ofld_rxq *ofld_rxq; 254 struct clip_entry *ce; 255 }; 256 257 struct tom_data { 258 struct toedev tod; 259 260 /* toepcb's associated with this TOE device */ 261 struct mtx toep_list_lock; 262 TAILQ_HEAD(, toepcb) toep_list; 263 264 struct mtx lctx_hash_lock; 265 LIST_HEAD(, listen_ctx) *listen_hash; 266 u_long listen_mask; 267 int lctx_count; /* # of lctx in the hash table */ 268 269 struct ppod_region pr; 270 271 /* WRs that will not be sent to the chip because L2 resolution failed */ 272 struct mtx unsent_wr_lock; 273 STAILQ_HEAD(, wrqe) unsent_wr_list; 274 struct task reclaim_wr_resources; 275 }; 276 277 static inline struct tom_data * 278 tod_td(struct toedev *tod) 279 { 280 281 return (__containerof(tod, struct tom_data, tod)); 282 } 283 284 static inline struct adapter * 285 td_adapter(struct tom_data *td) 286 { 287 288 return (td->tod.tod_softc); 289 } 290 291 static inline void 292 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 293 { 294 295 M_ASSERTPKTHDR(m); 296 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 297 } 298 299 static inline uint8_t 300 mbuf_ulp_submode(struct mbuf *m) 301 { 302 303 M_ASSERTPKTHDR(m); 304 return (m->m_pkthdr.PH_per.eight[0]); 305 } 306 307 /* t4_tom.c */ 308 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 309 struct toepcb *hold_toepcb(struct toepcb *); 310 void free_toepcb(struct toepcb *); 311 void offload_socket(struct socket *, struct toepcb *); 312 void undo_offload_socket(struct socket *); 313 void final_cpl_received(struct toepcb *); 314 void insert_tid(struct adapter *, int, void *, int); 315 void *lookup_tid(struct adapter *, int); 316 void update_tid(struct adapter *, int, void *); 317 void remove_tid(struct adapter *, int, int); 318 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, 319 struct offload_settings *); 320 u_long select_rcv_wnd(struct socket *); 321 int select_rcv_wscale(void); 322 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 323 int, int, int, int, struct offload_settings *); 324 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 325 int select_ulp_mode(struct socket *, struct adapter *, 326 struct offload_settings *); 327 void set_ulp_mode(struct toepcb *, int); 328 int negative_advice(int); 329 330 /* t4_connect.c */ 331 void t4_init_connect_cpl_handlers(void); 332 void t4_uninit_connect_cpl_handlers(void); 333 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 334 struct sockaddr *); 335 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 336 337 /* t4_listen.c */ 338 void t4_init_listen_cpl_handlers(void); 339 void t4_uninit_listen_cpl_handlers(void); 340 int t4_listen_start(struct toedev *, struct tcpcb *); 341 int t4_listen_stop(struct toedev *, struct tcpcb *); 342 void t4_syncache_added(struct toedev *, void *); 343 void t4_syncache_removed(struct toedev *, void *); 344 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 345 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 346 struct mbuf *); 347 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 348 struct mbuf *); 349 void t4_offload_socket(struct toedev *, void *, struct socket *); 350 void synack_failure_cleanup(struct adapter *, int); 351 352 /* t4_cpl_io.c */ 353 void aiotx_init_toep(struct toepcb *); 354 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 355 void t4_init_cpl_io_handlers(void); 356 void t4_uninit_cpl_io_handlers(void); 357 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 358 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 359 void send_reset(struct adapter *, struct toepcb *, uint32_t); 360 int send_rx_credits(struct adapter *, struct toepcb *, int); 361 void send_rx_modulate(struct adapter *, struct toepcb *); 362 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 363 int t4_close_conn(struct adapter *, struct toepcb *); 364 void t4_rcvd(struct toedev *, struct tcpcb *); 365 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 366 int t4_tod_output(struct toedev *, struct tcpcb *); 367 int t4_send_fin(struct toedev *, struct tcpcb *); 368 int t4_send_rst(struct toedev *, struct tcpcb *); 369 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *, 370 uint16_t, uint64_t, uint64_t, int, int); 371 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 372 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 373 374 /* t4_ddp.c */ 375 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 376 const char *); 377 void t4_free_ppod_region(struct ppod_region *); 378 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 379 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 380 struct ppod_reservation *); 381 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 382 struct pageset *); 383 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 384 struct ppod_reservation *, vm_offset_t, int); 385 void t4_free_page_pods(struct ppod_reservation *); 386 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 387 struct mbuf **, struct mbuf **, int *); 388 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 389 void t4_ddp_mod_load(void); 390 void t4_ddp_mod_unload(void); 391 void ddp_assert_empty(struct toepcb *); 392 void ddp_init_toep(struct toepcb *); 393 void ddp_uninit_toep(struct toepcb *); 394 void ddp_queue_toep(struct toepcb *); 395 void release_ddp_resources(struct toepcb *toep); 396 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 397 void handle_ddp_indicate(struct toepcb *); 398 void insert_ddp_data(struct toepcb *, uint32_t); 399 const struct offload_settings *lookup_offload_policy(struct adapter *, int, 400 struct mbuf *, uint16_t, struct inpcb *); 401 402 /* t4_tls.c */ 403 bool can_tls_offload(struct adapter *); 404 int t4_ctloutput_tls(struct socket *, struct sockopt *); 405 void t4_push_tls_records(struct adapter *, struct toepcb *, int); 406 void t4_tls_mod_load(void); 407 void t4_tls_mod_unload(void); 408 void tls_establish(struct toepcb *); 409 void tls_init_toep(struct toepcb *); 410 int tls_rx_key(struct toepcb *); 411 void tls_stop_handshake_timer(struct toepcb *); 412 int tls_tx_key(struct toepcb *); 413 void tls_uninit_toep(struct toepcb *); 414 415 #endif 416