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 "tom/t4_tls.h" 38 39 #define LISTEN_HASH_SIZE 32 40 41 /* 42 * Min receive window. We want it to be large enough to accommodate receive 43 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 44 */ 45 #define MIN_RCV_WND (24 * 1024U) 46 47 /* 48 * Max receive window supported by HW in bytes. Only a small part of it can 49 * be set through option0, the rest needs to be set through RX_DATA_ACK. 50 */ 51 #define MAX_RCV_WND ((1U << 27) - 1) 52 53 #define DDP_RSVD_WIN (16 * 1024U) 54 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 55 56 #define USE_DDP_RX_FLOW_CONTROL 57 58 #define PPOD_SZ(n) ((n) * sizeof(struct pagepod)) 59 #define PPOD_SIZE (PPOD_SZ(1)) 60 61 /* TOE PCB flags */ 62 enum { 63 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 64 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 65 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 66 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 67 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 68 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 69 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 70 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 71 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 72 TPF_SYNQE_EXPANDED = (1 << 9), /* toepcb ready, tid context updated */ 73 TPF_FORCE_CREDITS = (1 << 10), /* always send credits */ 74 }; 75 76 enum { 77 DDP_OK = (1 << 0), /* OK to turn on DDP */ 78 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */ 79 DDP_ON = (1 << 2), /* DDP is turned on */ 80 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */ 81 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */ 82 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */ 83 DDP_DEAD = (1 << 6), /* toepcb is shutting down */ 84 }; 85 86 struct sockopt; 87 struct offload_settings; 88 89 struct ofld_tx_sdesc { 90 uint32_t plen; /* payload length */ 91 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 92 void *iv_buffer; /* optional buffer holding IVs for TLS */ 93 }; 94 95 struct ppod_region { 96 u_int pr_start; 97 u_int pr_len; 98 u_int pr_page_shift[4]; 99 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 100 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 101 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 102 u_int pr_alias_shift; /* shift this much for first alias bit. */ 103 vmem_t *pr_arena; 104 }; 105 106 struct ppod_reservation { 107 struct ppod_region *prsv_pr; 108 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 109 u_int prsv_nppods; 110 }; 111 112 struct pageset { 113 TAILQ_ENTRY(pageset) link; 114 vm_page_t *pages; 115 int npages; 116 int flags; 117 int offset; /* offset in first page */ 118 int len; 119 struct ppod_reservation prsv; 120 struct vmspace *vm; 121 vm_offset_t start; 122 u_int vm_timestamp; 123 }; 124 125 TAILQ_HEAD(pagesetq, pageset); 126 127 #define PS_WIRED 0x0001 /* Pages wired rather than held. */ 128 #define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 129 130 #define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 131 132 #define IS_AIOTX_MBUF(m) \ 133 ((m)->m_flags & M_EXT && (m)->m_ext.ext_flags & EXT_FLAG_AIOTX) 134 135 struct ddp_buffer { 136 struct pageset *ps; 137 138 struct kaiocb *job; 139 int cancel_pending; 140 }; 141 142 struct ddp_pcb { 143 u_int flags; 144 struct ddp_buffer db[2]; 145 TAILQ_HEAD(, pageset) cached_pagesets; 146 TAILQ_HEAD(, kaiocb) aiojobq; 147 u_int waiting_count; 148 u_int active_count; 149 u_int cached_count; 150 int active_id; /* the currently active DDP buffer */ 151 struct task requeue_task; 152 struct kaiocb *queueing; 153 struct mtx lock; 154 }; 155 156 struct aiotx_buffer { 157 struct pageset ps; 158 struct kaiocb *job; 159 int refcount; 160 }; 161 162 struct toepcb { 163 TAILQ_ENTRY(toepcb) link; /* toep_list */ 164 u_int flags; /* miscellaneous flags */ 165 int refcount; 166 struct tom_data *td; 167 struct inpcb *inp; /* backpointer to host stack's PCB */ 168 struct vnet *vnet; 169 struct vi_info *vi; /* virtual interface */ 170 struct sge_wrq *ofld_txq; 171 struct sge_ofld_rxq *ofld_rxq; 172 struct sge_wrq *ctrlq; 173 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 174 struct clip_entry *ce; /* CLIP table entry used by this tid */ 175 int tid; /* Connection identifier */ 176 int tc_idx; /* traffic class that this tid is bound to */ 177 178 /* tx credit handling */ 179 u_int tx_total; /* total tx WR credits (in 16B units) */ 180 u_int tx_credits; /* tx WR credits (in 16B units) available */ 181 u_int tx_nocompl; /* tx WR credits since last compl request */ 182 u_int plen_nocompl; /* payload since last compl request */ 183 184 uint16_t opt0_rcv_bufsize; /* XXX: save full opt0/opt2 for later? */ 185 uint16_t mtu_idx; 186 uint16_t emss; 187 uint16_t tcp_opt; 188 189 u_int ulp_mode; /* ULP mode */ 190 void *ulpcb; 191 void *ulpcb2; 192 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 193 struct mbufq ulp_pdu_reclaimq; 194 195 struct ddp_pcb ddp; 196 struct tls_ofld_info tls; 197 198 TAILQ_HEAD(, kaiocb) aiotx_jobq; 199 struct task aiotx_task; 200 struct socket *aiotx_so; 201 202 /* Tx software descriptor */ 203 uint8_t txsd_total; 204 uint8_t txsd_pidx; 205 uint8_t txsd_cidx; 206 uint8_t txsd_avail; 207 struct ofld_tx_sdesc txsd[]; 208 }; 209 210 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 211 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 212 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 213 214 struct flowc_tx_params { 215 uint32_t snd_nxt; 216 uint32_t rcv_nxt; 217 unsigned int snd_space; 218 unsigned int mss; 219 }; 220 221 /* 222 * Compressed state for embryonic connections for a listener. 223 */ 224 struct synq_entry { 225 struct listen_ctx *lctx; /* backpointer to listen ctx */ 226 struct mbuf *syn; 227 int flags; /* same as toepcb's tp_flags */ 228 volatile int ok_to_respond; 229 volatile u_int refcnt; 230 int tid; 231 uint32_t iss; 232 uint32_t irs; 233 uint32_t ts; 234 uint16_t txqid; 235 uint16_t rxqid; 236 uint16_t l2e_idx; 237 uint16_t ulp_mode; 238 uint16_t rcv_bufsize; 239 __be16 tcp_opt; /* from cpl_pass_establish */ 240 struct toepcb *toep; 241 }; 242 243 /* listen_ctx flags */ 244 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 245 246 struct listen_ctx { 247 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 248 volatile int refcount; 249 int stid; 250 struct stid_region stid_region; 251 int flags; 252 struct inpcb *inp; /* listening socket's inp */ 253 struct vnet *vnet; 254 struct sge_wrq *ctrlq; 255 struct sge_ofld_rxq *ofld_rxq; 256 struct clip_entry *ce; 257 }; 258 259 /* tcb_histent flags */ 260 #define TE_RPL_PENDING 1 261 #define TE_ACTIVE 2 262 263 /* bits in one 8b tcb_histent sample. */ 264 #define TS_RTO (1 << 0) 265 #define TS_DUPACKS (1 << 1) 266 #define TS_FASTREXMT (1 << 2) 267 #define TS_SND_BACKLOGGED (1 << 3) 268 #define TS_CWND_LIMITED (1 << 4) 269 #define TS_ECN_ECE (1 << 5) 270 #define TS_ECN_CWR (1 << 6) 271 #define TS_RESERVED (1 << 7) /* Unused. */ 272 273 struct tcb_histent { 274 struct mtx te_lock; 275 struct callout te_callout; 276 uint64_t te_tcb[TCB_SIZE / sizeof(uint64_t)]; 277 struct adapter *te_adapter; 278 u_int te_flags; 279 u_int te_tid; 280 uint8_t te_pidx; 281 uint8_t te_sample[100]; 282 }; 283 284 struct tom_data { 285 struct toedev tod; 286 287 /* toepcb's associated with this TOE device */ 288 struct mtx toep_list_lock; 289 TAILQ_HEAD(, toepcb) toep_list; 290 291 struct mtx lctx_hash_lock; 292 LIST_HEAD(, listen_ctx) *listen_hash; 293 u_long listen_mask; 294 int lctx_count; /* # of lctx in the hash table */ 295 296 struct ppod_region pr; 297 298 struct rwlock tcb_history_lock __aligned(CACHE_LINE_SIZE); 299 struct tcb_histent **tcb_history; 300 int dupack_threshold; 301 302 /* WRs that will not be sent to the chip because L2 resolution failed */ 303 struct mtx unsent_wr_lock; 304 STAILQ_HEAD(, wrqe) unsent_wr_list; 305 struct task reclaim_wr_resources; 306 }; 307 308 static inline struct tom_data * 309 tod_td(struct toedev *tod) 310 { 311 312 return (__containerof(tod, struct tom_data, tod)); 313 } 314 315 static inline struct adapter * 316 td_adapter(struct tom_data *td) 317 { 318 319 return (td->tod.tod_softc); 320 } 321 322 static inline void 323 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 324 { 325 326 M_ASSERTPKTHDR(m); 327 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 328 } 329 330 static inline uint8_t 331 mbuf_ulp_submode(struct mbuf *m) 332 { 333 334 M_ASSERTPKTHDR(m); 335 return (m->m_pkthdr.PH_per.eight[0]); 336 } 337 338 /* t4_tom.c */ 339 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 340 struct toepcb *hold_toepcb(struct toepcb *); 341 void free_toepcb(struct toepcb *); 342 void offload_socket(struct socket *, struct toepcb *); 343 void undo_offload_socket(struct socket *); 344 void final_cpl_received(struct toepcb *); 345 void insert_tid(struct adapter *, int, void *, int); 346 void *lookup_tid(struct adapter *, int); 347 void update_tid(struct adapter *, int, void *); 348 void remove_tid(struct adapter *, int, int); 349 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, 350 struct offload_settings *); 351 u_long select_rcv_wnd(struct socket *); 352 int select_rcv_wscale(void); 353 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 354 int, int, int, int, struct offload_settings *); 355 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 356 int select_ulp_mode(struct socket *, struct adapter *, 357 struct offload_settings *); 358 void set_ulp_mode(struct toepcb *, int); 359 int negative_advice(int); 360 int add_tid_to_history(struct adapter *, u_int); 361 362 /* t4_connect.c */ 363 void t4_init_connect_cpl_handlers(void); 364 void t4_uninit_connect_cpl_handlers(void); 365 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 366 struct sockaddr *); 367 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 368 369 /* t4_listen.c */ 370 void t4_init_listen_cpl_handlers(void); 371 void t4_uninit_listen_cpl_handlers(void); 372 int t4_listen_start(struct toedev *, struct tcpcb *); 373 int t4_listen_stop(struct toedev *, struct tcpcb *); 374 void t4_syncache_added(struct toedev *, void *); 375 void t4_syncache_removed(struct toedev *, void *); 376 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 377 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 378 struct mbuf *); 379 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 380 struct mbuf *); 381 void t4_offload_socket(struct toedev *, void *, struct socket *); 382 void synack_failure_cleanup(struct adapter *, int); 383 384 /* t4_cpl_io.c */ 385 void aiotx_init_toep(struct toepcb *); 386 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 387 void t4_init_cpl_io_handlers(void); 388 void t4_uninit_cpl_io_handlers(void); 389 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 390 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 391 void send_reset(struct adapter *, struct toepcb *, uint32_t); 392 int send_rx_credits(struct adapter *, struct toepcb *, int); 393 void send_rx_modulate(struct adapter *, struct toepcb *); 394 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 395 int t4_close_conn(struct adapter *, struct toepcb *); 396 void t4_rcvd(struct toedev *, struct tcpcb *); 397 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 398 int t4_tod_output(struct toedev *, struct tcpcb *); 399 int t4_send_fin(struct toedev *, struct tcpcb *); 400 int t4_send_rst(struct toedev *, struct tcpcb *); 401 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *, 402 uint16_t, uint64_t, uint64_t, int, int); 403 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 404 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 405 406 /* t4_ddp.c */ 407 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 408 const char *); 409 void t4_free_ppod_region(struct ppod_region *); 410 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 411 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 412 struct ppod_reservation *); 413 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 414 struct pageset *); 415 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 416 struct ppod_reservation *, vm_offset_t, int); 417 void t4_free_page_pods(struct ppod_reservation *); 418 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 419 struct mbuf **, struct mbuf **, int *); 420 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 421 void t4_ddp_mod_load(void); 422 void t4_ddp_mod_unload(void); 423 void ddp_assert_empty(struct toepcb *); 424 void ddp_init_toep(struct toepcb *); 425 void ddp_uninit_toep(struct toepcb *); 426 void ddp_queue_toep(struct toepcb *); 427 void release_ddp_resources(struct toepcb *toep); 428 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 429 void handle_ddp_indicate(struct toepcb *); 430 void insert_ddp_data(struct toepcb *, uint32_t); 431 const struct offload_settings *lookup_offload_policy(struct adapter *, int, 432 struct mbuf *, uint16_t, struct inpcb *); 433 434 /* t4_tls.c */ 435 bool can_tls_offload(struct adapter *); 436 int t4_ctloutput_tls(struct socket *, struct sockopt *); 437 void t4_push_tls_records(struct adapter *, struct toepcb *, int); 438 void t4_tls_mod_load(void); 439 void t4_tls_mod_unload(void); 440 void tls_establish(struct toepcb *); 441 void tls_init_toep(struct toepcb *); 442 int tls_rx_key(struct toepcb *); 443 void tls_stop_handshake_timer(struct toepcb *); 444 int tls_tx_key(struct toepcb *); 445 void tls_uninit_toep(struct toepcb *); 446 447 #endif 448