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 37 #define LISTEN_HASH_SIZE 32 38 39 /* 40 * Min receive window. We want it to be large enough to accommodate receive 41 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 42 */ 43 #define MIN_RCV_WND (24 * 1024U) 44 45 /* 46 * Max receive window supported by HW in bytes. Only a small part of it can 47 * be set through option0, the rest needs to be set through RX_DATA_ACK. 48 */ 49 #define MAX_RCV_WND ((1U << 27) - 1) 50 51 #define DDP_RSVD_WIN (16 * 1024U) 52 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 53 54 #define USE_DDP_RX_FLOW_CONTROL 55 56 #define PPOD_SZ(n) ((n) * sizeof(struct pagepod)) 57 #define PPOD_SIZE (PPOD_SZ(1)) 58 59 /* TOE PCB flags */ 60 enum { 61 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 62 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 63 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 64 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 65 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 66 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 67 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 68 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 69 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 70 TPF_SYNQE_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */ 71 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */ 72 TPF_SYNQE_EXPANDED = (1 << 11), /* toepcb ready, tid context updated */ 73 TPF_SYNQE_HAS_L2TE = (1 << 12), /* we've replied to PASS_ACCEPT_REQ */ 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 ofld_tx_sdesc { 87 uint32_t plen; /* payload length */ 88 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 89 }; 90 91 struct ppod_region { 92 u_int pr_start; 93 u_int pr_len; 94 u_int pr_page_shift[4]; 95 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 96 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 97 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 98 u_int pr_alias_shift; /* shift this much for first alias bit. */ 99 vmem_t *pr_arena; 100 }; 101 102 struct ppod_reservation { 103 struct ppod_region *prsv_pr; 104 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 105 u_int prsv_nppods; 106 }; 107 108 struct pageset { 109 TAILQ_ENTRY(pageset) link; 110 vm_page_t *pages; 111 int npages; 112 int flags; 113 int offset; /* offset in first page */ 114 int len; 115 struct ppod_reservation prsv; 116 struct vmspace *vm; 117 vm_offset_t start; 118 u_int vm_timestamp; 119 }; 120 121 TAILQ_HEAD(pagesetq, pageset); 122 123 #define PS_WIRED 0x0001 /* Pages wired rather than held. */ 124 #define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 125 126 #define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 127 128 struct ddp_buffer { 129 struct pageset *ps; 130 131 struct kaiocb *job; 132 int cancel_pending; 133 }; 134 135 struct ddp_pcb { 136 u_int flags; 137 struct ddp_buffer db[2]; 138 TAILQ_HEAD(, pageset) cached_pagesets; 139 TAILQ_HEAD(, kaiocb) aiojobq; 140 u_int waiting_count; 141 u_int active_count; 142 u_int cached_count; 143 int active_id; /* the currently active DDP buffer */ 144 struct task requeue_task; 145 struct kaiocb *queueing; 146 struct mtx lock; 147 }; 148 149 struct aiotx_buffer { 150 struct pageset ps; 151 struct kaiocb *job; 152 int refcount; 153 }; 154 155 struct toepcb { 156 TAILQ_ENTRY(toepcb) link; /* toep_list */ 157 u_int flags; /* miscellaneous flags */ 158 int refcount; 159 struct tom_data *td; 160 struct inpcb *inp; /* backpointer to host stack's PCB */ 161 struct vnet *vnet; 162 struct vi_info *vi; /* virtual interface */ 163 struct sge_wrq *ofld_txq; 164 struct sge_ofld_rxq *ofld_rxq; 165 struct sge_wrq *ctrlq; 166 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 167 struct clip_entry *ce; /* CLIP table entry used by this tid */ 168 int tid; /* Connection identifier */ 169 int tc_idx; /* traffic class that this tid is bound to */ 170 171 /* tx credit handling */ 172 u_int tx_total; /* total tx WR credits (in 16B units) */ 173 u_int tx_credits; /* tx WR credits (in 16B units) available */ 174 u_int tx_nocompl; /* tx WR credits since last compl request */ 175 u_int plen_nocompl; /* payload since last compl request */ 176 177 /* rx credit handling */ 178 u_int sb_cc; /* last noted value of so_rcv->sb_cc */ 179 int rx_credits; /* rx credits (in bytes) to be returned to hw */ 180 181 u_int ulp_mode; /* ULP mode */ 182 void *ulpcb; 183 void *ulpcb2; 184 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 185 struct mbufq ulp_pdu_reclaimq; 186 187 struct ddp_pcb ddp; 188 189 TAILQ_HEAD(, kaiocb) aiotx_jobq; 190 struct task aiotx_task; 191 bool aiotx_task_active; 192 193 /* Tx software descriptor */ 194 uint8_t txsd_total; 195 uint8_t txsd_pidx; 196 uint8_t txsd_cidx; 197 uint8_t txsd_avail; 198 struct ofld_tx_sdesc txsd[]; 199 }; 200 201 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 202 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 203 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 204 205 struct flowc_tx_params { 206 uint32_t snd_nxt; 207 uint32_t rcv_nxt; 208 unsigned int snd_space; 209 unsigned int mss; 210 }; 211 212 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */ 213 #define DDP_LOW_SCORE 1 214 #define DDP_HIGH_SCORE 3 215 216 /* 217 * Compressed state for embryonic connections for a listener. Barely fits in 218 * 64B, try not to grow it further. 219 */ 220 struct synq_entry { 221 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */ 222 int flags; /* same as toepcb's tp_flags */ 223 int tid; 224 struct listen_ctx *lctx; /* backpointer to listen ctx */ 225 struct mbuf *syn; 226 uint32_t iss; 227 uint32_t ts; 228 volatile uintptr_t wr; 229 volatile u_int refcnt; 230 uint16_t l2e_idx; 231 uint16_t rcv_bufsize; 232 }; 233 234 /* listen_ctx flags */ 235 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 236 237 struct listen_ctx { 238 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 239 volatile int refcount; 240 int stid; 241 struct stid_region stid_region; 242 int flags; 243 struct inpcb *inp; /* listening socket's inp */ 244 struct vnet *vnet; 245 struct sge_wrq *ctrlq; 246 struct sge_ofld_rxq *ofld_rxq; 247 struct clip_entry *ce; 248 TAILQ_HEAD(, synq_entry) synq; 249 }; 250 251 struct clip_entry { 252 TAILQ_ENTRY(clip_entry) link; 253 struct in6_addr lip; /* local IPv6 address */ 254 u_int refcount; 255 }; 256 257 TAILQ_HEAD(clip_head, clip_entry); 258 struct tom_data { 259 struct toedev tod; 260 261 /* toepcb's associated with this TOE device */ 262 struct mtx toep_list_lock; 263 TAILQ_HEAD(, toepcb) toep_list; 264 265 struct mtx lctx_hash_lock; 266 LIST_HEAD(, listen_ctx) *listen_hash; 267 u_long listen_mask; 268 int lctx_count; /* # of lctx in the hash table */ 269 270 struct ppod_region pr; 271 272 struct mtx clip_table_lock; 273 struct clip_head clip_table; 274 int clip_gen; 275 276 /* WRs that will not be sent to the chip because L2 resolution failed */ 277 struct mtx unsent_wr_lock; 278 STAILQ_HEAD(, wrqe) unsent_wr_list; 279 struct task reclaim_wr_resources; 280 }; 281 282 static inline struct tom_data * 283 tod_td(struct toedev *tod) 284 { 285 286 return (__containerof(tod, struct tom_data, tod)); 287 } 288 289 static inline struct adapter * 290 td_adapter(struct tom_data *td) 291 { 292 293 return (td->tod.tod_softc); 294 } 295 296 static inline void 297 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 298 { 299 300 M_ASSERTPKTHDR(m); 301 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 302 } 303 304 static inline uint8_t 305 mbuf_ulp_submode(struct mbuf *m) 306 { 307 308 M_ASSERTPKTHDR(m); 309 return (m->m_pkthdr.PH_per.eight[0]); 310 } 311 312 /* t4_tom.c */ 313 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 314 struct toepcb *hold_toepcb(struct toepcb *); 315 void free_toepcb(struct toepcb *); 316 void offload_socket(struct socket *, struct toepcb *); 317 void undo_offload_socket(struct socket *); 318 void final_cpl_received(struct toepcb *); 319 void insert_tid(struct adapter *, int, void *, int); 320 void *lookup_tid(struct adapter *, int); 321 void update_tid(struct adapter *, int, void *); 322 void remove_tid(struct adapter *, int, int); 323 void release_tid(struct adapter *, int, struct sge_wrq *); 324 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int); 325 u_long select_rcv_wnd(struct socket *); 326 int select_rcv_wscale(void); 327 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 328 int, int, int, int); 329 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 330 void set_tcpddp_ulp_mode(struct toepcb *); 331 int negative_advice(int); 332 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *, 333 struct clip_entry *); 334 void release_lip(struct tom_data *, struct clip_entry *); 335 336 /* t4_connect.c */ 337 void t4_init_connect_cpl_handlers(void); 338 void t4_uninit_connect_cpl_handlers(void); 339 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 340 struct sockaddr *); 341 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 342 343 /* t4_listen.c */ 344 void t4_init_listen_cpl_handlers(void); 345 void t4_uninit_listen_cpl_handlers(void); 346 int t4_listen_start(struct toedev *, struct tcpcb *); 347 int t4_listen_stop(struct toedev *, struct tcpcb *); 348 void t4_syncache_added(struct toedev *, void *); 349 void t4_syncache_removed(struct toedev *, void *); 350 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 351 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 352 struct mbuf *); 353 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 354 struct mbuf *); 355 void t4_offload_socket(struct toedev *, void *, struct socket *); 356 357 /* t4_cpl_io.c */ 358 void aiotx_init_toep(struct toepcb *); 359 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 360 void t4_init_cpl_io_handlers(void); 361 void t4_uninit_cpl_io_handlers(void); 362 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 363 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 364 void send_reset(struct adapter *, struct toepcb *, uint32_t); 365 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 366 void t4_rcvd(struct toedev *, struct tcpcb *); 367 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 368 int t4_tod_output(struct toedev *, struct tcpcb *); 369 int t4_send_fin(struct toedev *, struct tcpcb *); 370 int t4_send_rst(struct toedev *, struct tcpcb *); 371 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, int, uint16_t, 372 uint64_t, uint64_t, int, int, int); 373 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 374 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 375 int do_set_tcb_rpl(struct sge_iq *, const struct rss_header *, struct mbuf *); 376 377 /* t4_ddp.c */ 378 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 379 const char *); 380 void t4_free_ppod_region(struct ppod_region *); 381 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 382 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 383 struct ppod_reservation *); 384 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 385 struct pageset *); 386 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 387 struct ppod_reservation *, vm_offset_t, int); 388 void t4_free_page_pods(struct ppod_reservation *); 389 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 390 struct mbuf **, struct mbuf **, int *); 391 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 392 int t4_ddp_mod_load(void); 393 void t4_ddp_mod_unload(void); 394 void ddp_assert_empty(struct toepcb *); 395 void ddp_init_toep(struct toepcb *); 396 void ddp_uninit_toep(struct toepcb *); 397 void ddp_queue_toep(struct toepcb *); 398 void release_ddp_resources(struct toepcb *toep); 399 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 400 void handle_ddp_indicate(struct toepcb *); 401 void handle_ddp_tcb_rpl(struct toepcb *, const struct cpl_set_tcb_rpl *); 402 void insert_ddp_data(struct toepcb *, uint32_t); 403 404 #endif 405