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_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */ 72 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */ 73 TPF_SYNQE_EXPANDED = (1 << 11), /* toepcb ready, tid context updated */ 74 TPF_SYNQE_HAS_L2TE = (1 << 12), /* we've replied to PASS_ACCEPT_REQ */ 75 TPF_SYNQE_TLS = (1 << 13), /* ulp_mode TLS in toepcb */ 76 TPF_FORCE_CREDITS = (1 << 14), /* always send credits */ 77 }; 78 79 enum { 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 89 struct sockopt; 90 91 struct ofld_tx_sdesc { 92 uint32_t plen; /* payload length */ 93 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 94 void *iv_buffer; /* optional buffer holding IVs for TLS */ 95 }; 96 97 struct ppod_region { 98 u_int pr_start; 99 u_int pr_len; 100 u_int pr_page_shift[4]; 101 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 102 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 103 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 104 u_int pr_alias_shift; /* shift this much for first alias bit. */ 105 vmem_t *pr_arena; 106 }; 107 108 struct ppod_reservation { 109 struct ppod_region *prsv_pr; 110 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 111 u_int prsv_nppods; 112 }; 113 114 struct pageset { 115 TAILQ_ENTRY(pageset) link; 116 vm_page_t *pages; 117 int npages; 118 int flags; 119 int offset; /* offset in first page */ 120 int len; 121 struct ppod_reservation prsv; 122 struct vmspace *vm; 123 vm_offset_t start; 124 u_int vm_timestamp; 125 }; 126 127 TAILQ_HEAD(pagesetq, pageset); 128 129 #define PS_WIRED 0x0001 /* Pages wired rather than held. */ 130 #define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 131 132 #define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 133 134 #define IS_AIOTX_MBUF(m) \ 135 ((m)->m_flags & M_EXT && (m)->m_ext.ext_flags & EXT_FLAG_AIOTX) 136 137 struct ddp_buffer { 138 struct pageset *ps; 139 140 struct kaiocb *job; 141 int cancel_pending; 142 }; 143 144 struct ddp_pcb { 145 u_int flags; 146 struct ddp_buffer db[2]; 147 TAILQ_HEAD(, pageset) cached_pagesets; 148 TAILQ_HEAD(, kaiocb) aiojobq; 149 u_int waiting_count; 150 u_int active_count; 151 u_int cached_count; 152 int active_id; /* the currently active DDP buffer */ 153 struct task requeue_task; 154 struct kaiocb *queueing; 155 struct mtx lock; 156 }; 157 158 struct aiotx_buffer { 159 struct pageset ps; 160 struct kaiocb *job; 161 int refcount; 162 }; 163 164 struct toepcb { 165 TAILQ_ENTRY(toepcb) link; /* toep_list */ 166 u_int flags; /* miscellaneous flags */ 167 int refcount; 168 struct tom_data *td; 169 struct inpcb *inp; /* backpointer to host stack's PCB */ 170 struct vnet *vnet; 171 struct vi_info *vi; /* virtual interface */ 172 struct sge_wrq *ofld_txq; 173 struct sge_ofld_rxq *ofld_rxq; 174 struct sge_wrq *ctrlq; 175 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 176 struct clip_entry *ce; /* CLIP table entry used by this tid */ 177 int tid; /* Connection identifier */ 178 int tc_idx; /* traffic class that this tid is bound to */ 179 180 /* tx credit handling */ 181 u_int tx_total; /* total tx WR credits (in 16B units) */ 182 u_int tx_credits; /* tx WR credits (in 16B units) available */ 183 u_int tx_nocompl; /* tx WR credits since last compl request */ 184 u_int plen_nocompl; /* payload since last compl request */ 185 186 /* rx credit handling */ 187 u_int sb_cc; /* last noted value of so_rcv->sb_cc */ 188 int rx_credits; /* rx credits (in bytes) to be returned to hw */ 189 190 u_int ulp_mode; /* ULP mode */ 191 void *ulpcb; 192 void *ulpcb2; 193 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 194 struct mbufq ulp_pdu_reclaimq; 195 196 struct ddp_pcb ddp; 197 struct tls_ofld_info tls; 198 199 TAILQ_HEAD(, kaiocb) aiotx_jobq; 200 struct task aiotx_task; 201 bool aiotx_task_active; 202 203 /* Tx software descriptor */ 204 uint8_t txsd_total; 205 uint8_t txsd_pidx; 206 uint8_t txsd_cidx; 207 uint8_t txsd_avail; 208 struct ofld_tx_sdesc txsd[]; 209 }; 210 211 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 212 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 213 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 214 215 struct flowc_tx_params { 216 uint32_t snd_nxt; 217 uint32_t rcv_nxt; 218 unsigned int snd_space; 219 unsigned int mss; 220 }; 221 222 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */ 223 #define DDP_LOW_SCORE 1 224 #define DDP_HIGH_SCORE 3 225 226 /* 227 * Compressed state for embryonic connections for a listener. Barely fits in 228 * 64B, try not to grow it further. 229 */ 230 struct synq_entry { 231 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */ 232 int flags; /* same as toepcb's tp_flags */ 233 int tid; 234 struct listen_ctx *lctx; /* backpointer to listen ctx */ 235 struct mbuf *syn; 236 uint32_t iss; 237 uint32_t ts; 238 volatile uintptr_t wr; 239 volatile u_int refcnt; 240 uint16_t l2e_idx; 241 uint16_t rcv_bufsize; 242 }; 243 244 /* listen_ctx flags */ 245 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 246 247 struct listen_ctx { 248 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 249 volatile int refcount; 250 int stid; 251 struct stid_region stid_region; 252 int flags; 253 struct inpcb *inp; /* listening socket's inp */ 254 struct vnet *vnet; 255 struct sge_wrq *ctrlq; 256 struct sge_ofld_rxq *ofld_rxq; 257 struct clip_entry *ce; 258 TAILQ_HEAD(, synq_entry) synq; 259 }; 260 261 struct clip_entry { 262 TAILQ_ENTRY(clip_entry) link; 263 struct in6_addr lip; /* local IPv6 address */ 264 u_int refcount; 265 }; 266 267 TAILQ_HEAD(clip_head, clip_entry); 268 struct tom_data { 269 struct toedev tod; 270 271 /* toepcb's associated with this TOE device */ 272 struct mtx toep_list_lock; 273 TAILQ_HEAD(, toepcb) toep_list; 274 275 struct mtx lctx_hash_lock; 276 LIST_HEAD(, listen_ctx) *listen_hash; 277 u_long listen_mask; 278 int lctx_count; /* # of lctx in the hash table */ 279 280 struct ppod_region pr; 281 282 vmem_t *key_map; 283 284 struct mtx clip_table_lock; 285 struct clip_head clip_table; 286 int clip_gen; 287 288 /* WRs that will not be sent to the chip because L2 resolution failed */ 289 struct mtx unsent_wr_lock; 290 STAILQ_HEAD(, wrqe) unsent_wr_list; 291 struct task reclaim_wr_resources; 292 }; 293 294 static inline struct tom_data * 295 tod_td(struct toedev *tod) 296 { 297 298 return (__containerof(tod, struct tom_data, tod)); 299 } 300 301 static inline struct adapter * 302 td_adapter(struct tom_data *td) 303 { 304 305 return (td->tod.tod_softc); 306 } 307 308 static inline void 309 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 310 { 311 312 M_ASSERTPKTHDR(m); 313 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 314 } 315 316 static inline uint8_t 317 mbuf_ulp_submode(struct mbuf *m) 318 { 319 320 M_ASSERTPKTHDR(m); 321 return (m->m_pkthdr.PH_per.eight[0]); 322 } 323 324 /* t4_tom.c */ 325 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 326 struct toepcb *hold_toepcb(struct toepcb *); 327 void free_toepcb(struct toepcb *); 328 void offload_socket(struct socket *, struct toepcb *); 329 void undo_offload_socket(struct socket *); 330 void final_cpl_received(struct toepcb *); 331 void insert_tid(struct adapter *, int, void *, int); 332 void *lookup_tid(struct adapter *, int); 333 void update_tid(struct adapter *, int, void *); 334 void remove_tid(struct adapter *, int, int); 335 void release_tid(struct adapter *, int, struct sge_wrq *); 336 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int); 337 u_long select_rcv_wnd(struct socket *); 338 int select_rcv_wscale(void); 339 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 340 int, int, int, int); 341 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 342 int select_ulp_mode(struct socket *, struct adapter *); 343 void set_ulp_mode(struct toepcb *, int); 344 int negative_advice(int); 345 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *, 346 struct clip_entry *); 347 void release_lip(struct tom_data *, struct clip_entry *); 348 349 /* t4_connect.c */ 350 void t4_init_connect_cpl_handlers(void); 351 void t4_uninit_connect_cpl_handlers(void); 352 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 353 struct sockaddr *); 354 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 355 356 /* t4_listen.c */ 357 void t4_init_listen_cpl_handlers(void); 358 void t4_uninit_listen_cpl_handlers(void); 359 int t4_listen_start(struct toedev *, struct tcpcb *); 360 int t4_listen_stop(struct toedev *, struct tcpcb *); 361 void t4_syncache_added(struct toedev *, void *); 362 void t4_syncache_removed(struct toedev *, void *); 363 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 364 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 365 struct mbuf *); 366 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 367 struct mbuf *); 368 void t4_offload_socket(struct toedev *, void *, struct socket *); 369 370 /* t4_cpl_io.c */ 371 void aiotx_init_toep(struct toepcb *); 372 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 373 void t4_init_cpl_io_handlers(void); 374 void t4_uninit_cpl_io_handlers(void); 375 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 376 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 377 void send_reset(struct adapter *, struct toepcb *, uint32_t); 378 int send_rx_credits(struct adapter *, struct toepcb *, int); 379 void send_rx_modulate(struct adapter *, struct toepcb *); 380 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 381 int t4_close_conn(struct adapter *, struct toepcb *); 382 void t4_rcvd(struct toedev *, struct tcpcb *); 383 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 384 int t4_tod_output(struct toedev *, struct tcpcb *); 385 int t4_send_fin(struct toedev *, struct tcpcb *); 386 int t4_send_rst(struct toedev *, struct tcpcb *); 387 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *, 388 uint16_t, uint64_t, uint64_t, int, int); 389 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 390 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 391 int do_set_tcb_rpl(struct sge_iq *, const struct rss_header *, struct mbuf *); 392 393 /* t4_ddp.c */ 394 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 395 const char *); 396 void t4_free_ppod_region(struct ppod_region *); 397 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 398 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 399 struct ppod_reservation *); 400 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 401 struct pageset *); 402 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 403 struct ppod_reservation *, vm_offset_t, int); 404 void t4_free_page_pods(struct ppod_reservation *); 405 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 406 struct mbuf **, struct mbuf **, int *); 407 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 408 void t4_ddp_mod_load(void); 409 void t4_ddp_mod_unload(void); 410 void ddp_assert_empty(struct toepcb *); 411 void ddp_init_toep(struct toepcb *); 412 void ddp_uninit_toep(struct toepcb *); 413 void ddp_queue_toep(struct toepcb *); 414 void release_ddp_resources(struct toepcb *toep); 415 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 416 void handle_ddp_indicate(struct toepcb *); 417 void handle_ddp_tcb_rpl(struct toepcb *, const struct cpl_set_tcb_rpl *); 418 void insert_ddp_data(struct toepcb *, uint32_t); 419 420 /* t4_tls.c */ 421 bool can_tls_offload(struct adapter *); 422 int t4_ctloutput_tls(struct socket *, struct sockopt *); 423 void t4_push_tls_records(struct adapter *, struct toepcb *, int); 424 void t4_tls_mod_load(void); 425 void t4_tls_mod_unload(void); 426 void tls_establish(struct toepcb *); 427 void tls_free_kmap(struct tom_data *); 428 int tls_init_kmap(struct adapter *, struct tom_data *); 429 void tls_init_toep(struct toepcb *); 430 int tls_rx_key(struct toepcb *); 431 void tls_stop_handshake_timer(struct toepcb *); 432 int tls_tx_key(struct toepcb *); 433 void tls_uninit_toep(struct toepcb *); 434 435 #endif 436