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 aiotx_buffer { 136 struct pageset ps; 137 struct kaiocb *job; 138 int refcount; 139 }; 140 141 struct toepcb { 142 TAILQ_ENTRY(toepcb) link; /* toep_list */ 143 u_int flags; /* miscellaneous flags */ 144 int refcount; 145 struct tom_data *td; 146 struct inpcb *inp; /* backpointer to host stack's PCB */ 147 struct vnet *vnet; 148 struct vi_info *vi; /* virtual interface */ 149 struct sge_wrq *ofld_txq; 150 struct sge_ofld_rxq *ofld_rxq; 151 struct sge_wrq *ctrlq; 152 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 153 struct clip_entry *ce; /* CLIP table entry used by this tid */ 154 int tid; /* Connection identifier */ 155 int tc_idx; /* traffic class that this tid is bound to */ 156 157 /* tx credit handling */ 158 u_int tx_total; /* total tx WR credits (in 16B units) */ 159 u_int tx_credits; /* tx WR credits (in 16B units) available */ 160 u_int tx_nocompl; /* tx WR credits since last compl request */ 161 u_int plen_nocompl; /* payload since last compl request */ 162 163 /* rx credit handling */ 164 u_int sb_cc; /* last noted value of so_rcv->sb_cc */ 165 int rx_credits; /* rx credits (in bytes) to be returned to hw */ 166 167 u_int ulp_mode; /* ULP mode */ 168 void *ulpcb; 169 void *ulpcb2; 170 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 171 struct mbufq ulp_pdu_reclaimq; 172 173 u_int ddp_flags; 174 struct ddp_buffer db[2]; 175 TAILQ_HEAD(, pageset) ddp_cached_pagesets; 176 TAILQ_HEAD(, kaiocb) ddp_aiojobq; 177 u_int ddp_waiting_count; 178 u_int ddp_active_count; 179 u_int ddp_cached_count; 180 int ddp_active_id; /* the currently active DDP buffer */ 181 struct task ddp_requeue_task; 182 struct kaiocb *ddp_queueing; 183 struct mtx ddp_lock; 184 185 TAILQ_HEAD(, kaiocb) aiotx_jobq; 186 struct task aiotx_task; 187 bool aiotx_task_active; 188 189 /* Tx software descriptor */ 190 uint8_t txsd_total; 191 uint8_t txsd_pidx; 192 uint8_t txsd_cidx; 193 uint8_t txsd_avail; 194 struct ofld_tx_sdesc txsd[]; 195 }; 196 197 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp_lock) 198 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp_lock) 199 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp_lock, MA_OWNED) 200 201 struct flowc_tx_params { 202 uint32_t snd_nxt; 203 uint32_t rcv_nxt; 204 unsigned int snd_space; 205 unsigned int mss; 206 }; 207 208 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */ 209 #define DDP_LOW_SCORE 1 210 #define DDP_HIGH_SCORE 3 211 212 /* 213 * Compressed state for embryonic connections for a listener. Barely fits in 214 * 64B, try not to grow it further. 215 */ 216 struct synq_entry { 217 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */ 218 int flags; /* same as toepcb's tp_flags */ 219 int tid; 220 struct listen_ctx *lctx; /* backpointer to listen ctx */ 221 struct mbuf *syn; 222 uint32_t iss; 223 uint32_t ts; 224 volatile uintptr_t wr; 225 volatile u_int refcnt; 226 uint16_t l2e_idx; 227 uint16_t rcv_bufsize; 228 }; 229 230 /* listen_ctx flags */ 231 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 232 233 struct listen_ctx { 234 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 235 volatile int refcount; 236 int stid; 237 struct stid_region stid_region; 238 int flags; 239 struct inpcb *inp; /* listening socket's inp */ 240 struct vnet *vnet; 241 struct sge_wrq *ctrlq; 242 struct sge_ofld_rxq *ofld_rxq; 243 struct clip_entry *ce; 244 TAILQ_HEAD(, synq_entry) synq; 245 }; 246 247 struct clip_entry { 248 TAILQ_ENTRY(clip_entry) link; 249 struct in6_addr lip; /* local IPv6 address */ 250 u_int refcount; 251 }; 252 253 TAILQ_HEAD(clip_head, clip_entry); 254 struct tom_data { 255 struct toedev tod; 256 257 /* toepcb's associated with this TOE device */ 258 struct mtx toep_list_lock; 259 TAILQ_HEAD(, toepcb) toep_list; 260 261 struct mtx lctx_hash_lock; 262 LIST_HEAD(, listen_ctx) *listen_hash; 263 u_long listen_mask; 264 int lctx_count; /* # of lctx in the hash table */ 265 266 struct ppod_region pr; 267 268 struct mtx clip_table_lock; 269 struct clip_head clip_table; 270 int clip_gen; 271 272 /* WRs that will not be sent to the chip because L2 resolution failed */ 273 struct mtx unsent_wr_lock; 274 STAILQ_HEAD(, wrqe) unsent_wr_list; 275 struct task reclaim_wr_resources; 276 }; 277 278 static inline struct tom_data * 279 tod_td(struct toedev *tod) 280 { 281 282 return (__containerof(tod, struct tom_data, tod)); 283 } 284 285 static inline struct adapter * 286 td_adapter(struct tom_data *td) 287 { 288 289 return (td->tod.tod_softc); 290 } 291 292 static inline void 293 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 294 { 295 296 M_ASSERTPKTHDR(m); 297 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 298 } 299 300 static inline uint8_t 301 mbuf_ulp_submode(struct mbuf *m) 302 { 303 304 M_ASSERTPKTHDR(m); 305 return (m->m_pkthdr.PH_per.eight[0]); 306 } 307 308 /* t4_tom.c */ 309 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 310 struct toepcb *hold_toepcb(struct toepcb *); 311 void free_toepcb(struct toepcb *); 312 void offload_socket(struct socket *, struct toepcb *); 313 void undo_offload_socket(struct socket *); 314 void final_cpl_received(struct toepcb *); 315 void insert_tid(struct adapter *, int, void *, int); 316 void *lookup_tid(struct adapter *, int); 317 void update_tid(struct adapter *, int, void *); 318 void remove_tid(struct adapter *, int, int); 319 void release_tid(struct adapter *, int, struct sge_wrq *); 320 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int); 321 u_long select_rcv_wnd(struct socket *); 322 int select_rcv_wscale(void); 323 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 324 int, int, int, int); 325 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 326 void set_tcpddp_ulp_mode(struct toepcb *); 327 int negative_advice(int); 328 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *, 329 struct clip_entry *); 330 void release_lip(struct tom_data *, struct clip_entry *); 331 332 /* t4_connect.c */ 333 void t4_init_connect_cpl_handlers(void); 334 void t4_uninit_connect_cpl_handlers(void); 335 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 336 struct sockaddr *); 337 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 338 339 /* t4_listen.c */ 340 void t4_init_listen_cpl_handlers(void); 341 void t4_uninit_listen_cpl_handlers(void); 342 int t4_listen_start(struct toedev *, struct tcpcb *); 343 int t4_listen_stop(struct toedev *, struct tcpcb *); 344 void t4_syncache_added(struct toedev *, void *); 345 void t4_syncache_removed(struct toedev *, void *); 346 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 347 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 348 struct mbuf *); 349 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 350 struct mbuf *); 351 void t4_offload_socket(struct toedev *, void *, struct socket *); 352 353 /* t4_cpl_io.c */ 354 void aiotx_init_toep(struct toepcb *); 355 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 356 void t4_init_cpl_io_handlers(void); 357 void t4_uninit_cpl_io_handlers(void); 358 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 359 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 360 void send_reset(struct adapter *, struct toepcb *, uint32_t); 361 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 362 void t4_rcvd(struct toedev *, struct tcpcb *); 363 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 364 int t4_tod_output(struct toedev *, struct tcpcb *); 365 int t4_send_fin(struct toedev *, struct tcpcb *); 366 int t4_send_rst(struct toedev *, struct tcpcb *); 367 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, int, uint16_t, 368 uint64_t, uint64_t, int, int, int); 369 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 370 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 371 int do_set_tcb_rpl(struct sge_iq *, const struct rss_header *, struct mbuf *); 372 373 /* t4_ddp.c */ 374 int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 375 const char *); 376 void t4_free_ppod_region(struct ppod_region *); 377 int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 378 int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 379 struct ppod_reservation *); 380 int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 381 struct pageset *); 382 int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 383 struct ppod_reservation *, vm_offset_t, int); 384 void t4_free_page_pods(struct ppod_reservation *); 385 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 386 struct mbuf **, struct mbuf **, int *); 387 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 388 int t4_ddp_mod_load(void); 389 void t4_ddp_mod_unload(void); 390 void ddp_assert_empty(struct toepcb *); 391 void ddp_init_toep(struct toepcb *); 392 void ddp_uninit_toep(struct toepcb *); 393 void ddp_queue_toep(struct toepcb *); 394 void release_ddp_resources(struct toepcb *toep); 395 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 396 void handle_ddp_indicate(struct toepcb *); 397 void handle_ddp_tcb_rpl(struct toepcb *, const struct cpl_set_tcb_rpl *); 398 void insert_ddp_data(struct toepcb *, uint32_t); 399 400 #endif 401