1 /*- 2 * Copyright (c) 2012, 2015 Chelsio Communications, Inc. 3 * All rights reserved. 4 * Written by: Navdeep Parhar <np@FreeBSD.org> 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 * 27 * $FreeBSD$ 28 * 29 */ 30 31 #ifndef __T4_TOM_H__ 32 #define __T4_TOM_H__ 33 #include <sys/vmem.h> 34 35 #define LISTEN_HASH_SIZE 32 36 37 /* 38 * Min receive window. We want it to be large enough to accommodate receive 39 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 40 */ 41 #define MIN_RCV_WND (24 * 1024U) 42 43 /* 44 * Max receive window supported by HW in bytes. Only a small part of it can 45 * be set through option0, the rest needs to be set through RX_DATA_ACK. 46 */ 47 #define MAX_RCV_WND ((1U << 27) - 1) 48 49 #define DDP_RSVD_WIN (16 * 1024U) 50 #define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 51 52 #define USE_DDP_RX_FLOW_CONTROL 53 54 /* TOE PCB flags */ 55 enum { 56 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 57 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 58 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 59 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 60 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 61 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 62 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 63 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 64 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 65 TPF_SYNQE_NEEDFREE = (1 << 9), /* synq_entry was malloc'd separately */ 66 TPF_SYNQE_TCPDDP = (1 << 10), /* ulp_mode TCPDDP in toepcb */ 67 TPF_SYNQE_EXPANDED = (1 << 11), /* toepcb ready, tid context updated */ 68 TPF_SYNQE_HAS_L2TE = (1 << 12), /* we've replied to PASS_ACCEPT_REQ */ 69 }; 70 71 enum { 72 DDP_OK = (1 << 0), /* OK to turn on DDP */ 73 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */ 74 DDP_ON = (1 << 2), /* DDP is turned on */ 75 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */ 76 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */ 77 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */ 78 DDP_DEAD = (1 << 6), /* toepcb is shutting down */ 79 }; 80 81 struct ofld_tx_sdesc { 82 uint32_t plen; /* payload length */ 83 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 84 }; 85 86 struct pageset { 87 TAILQ_ENTRY(pageset) link; 88 vm_page_t *pages; 89 int npages; 90 int flags; 91 u_int ppod_addr; 92 int nppods; 93 uint32_t tag; /* includes color, page pod addr, and DDP page size */ 94 int offset; /* offset in first page */ 95 int len; 96 struct vmspace *vm; 97 u_int vm_timestamp; 98 }; 99 100 TAILQ_HEAD(pagesetq, pageset); 101 102 #define PS_WIRED 0x0001 /* Pages wired rather than held. */ 103 #define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 104 105 #define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 106 107 struct ddp_buffer { 108 struct pageset *ps; 109 110 struct kaiocb *job; 111 int cancel_pending; 112 }; 113 114 struct aiotx_buffer { 115 struct pageset ps; 116 struct kaiocb *job; 117 int refcount; 118 }; 119 120 struct toepcb { 121 TAILQ_ENTRY(toepcb) link; /* toep_list */ 122 u_int flags; /* miscellaneous flags */ 123 int refcount; 124 struct tom_data *td; 125 struct inpcb *inp; /* backpointer to host stack's PCB */ 126 struct vi_info *vi; /* virtual interface */ 127 struct sge_wrq *ofld_txq; 128 struct sge_ofld_rxq *ofld_rxq; 129 struct sge_wrq *ctrlq; 130 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 131 struct clip_entry *ce; /* CLIP table entry used by this tid */ 132 int tid; /* Connection identifier */ 133 134 /* tx credit handling */ 135 u_int tx_total; /* total tx WR credits (in 16B units) */ 136 u_int tx_credits; /* tx WR credits (in 16B units) available */ 137 u_int tx_nocompl; /* tx WR credits since last compl request */ 138 u_int plen_nocompl; /* payload since last compl request */ 139 140 /* rx credit handling */ 141 u_int sb_cc; /* last noted value of so_rcv->sb_cc */ 142 int rx_credits; /* rx credits (in bytes) to be returned to hw */ 143 144 u_int ulp_mode; /* ULP mode */ 145 void *ulpcb; 146 void *ulpcb2; 147 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 148 struct mbufq ulp_pdu_reclaimq; 149 150 u_int ddp_flags; 151 struct ddp_buffer db[2]; 152 TAILQ_HEAD(, pageset) ddp_cached_pagesets; 153 TAILQ_HEAD(, kaiocb) ddp_aiojobq; 154 u_int ddp_waiting_count; 155 u_int ddp_active_count; 156 u_int ddp_cached_count; 157 int ddp_active_id; /* the currently active DDP buffer */ 158 struct task ddp_requeue_task; 159 struct kaiocb *ddp_queueing; 160 struct mtx ddp_lock; 161 162 TAILQ_HEAD(, kaiocb) aiotx_jobq; 163 struct task aiotx_task; 164 bool aiotx_task_active; 165 166 /* Tx software descriptor */ 167 uint8_t txsd_total; 168 uint8_t txsd_pidx; 169 uint8_t txsd_cidx; 170 uint8_t txsd_avail; 171 struct ofld_tx_sdesc txsd[]; 172 }; 173 174 #define DDP_LOCK(toep) mtx_lock(&(toep)->ddp_lock) 175 #define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp_lock) 176 #define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp_lock, MA_OWNED) 177 178 struct flowc_tx_params { 179 uint32_t snd_nxt; 180 uint32_t rcv_nxt; 181 unsigned int snd_space; 182 unsigned int mss; 183 }; 184 185 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */ 186 #define DDP_LOW_SCORE 1 187 #define DDP_HIGH_SCORE 3 188 189 /* 190 * Compressed state for embryonic connections for a listener. Barely fits in 191 * 64B, try not to grow it further. 192 */ 193 struct synq_entry { 194 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */ 195 int flags; /* same as toepcb's tp_flags */ 196 int tid; 197 struct listen_ctx *lctx; /* backpointer to listen ctx */ 198 struct mbuf *syn; 199 uint32_t iss; 200 uint32_t ts; 201 volatile uintptr_t wr; 202 volatile u_int refcnt; 203 uint16_t l2e_idx; 204 uint16_t rcv_bufsize; 205 }; 206 207 /* listen_ctx flags */ 208 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 209 210 struct listen_ctx { 211 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 212 volatile int refcount; 213 int stid; 214 struct stid_region stid_region; 215 int flags; 216 struct inpcb *inp; /* listening socket's inp */ 217 struct sge_wrq *ctrlq; 218 struct sge_ofld_rxq *ofld_rxq; 219 struct clip_entry *ce; 220 TAILQ_HEAD(, synq_entry) synq; 221 }; 222 223 struct clip_entry { 224 TAILQ_ENTRY(clip_entry) link; 225 struct in6_addr lip; /* local IPv6 address */ 226 u_int refcount; 227 }; 228 229 TAILQ_HEAD(clip_head, clip_entry); 230 struct tom_data { 231 struct toedev tod; 232 233 /* toepcb's associated with this TOE device */ 234 struct mtx toep_list_lock; 235 TAILQ_HEAD(, toepcb) toep_list; 236 237 struct mtx lctx_hash_lock; 238 LIST_HEAD(, listen_ctx) *listen_hash; 239 u_long listen_mask; 240 int lctx_count; /* # of lctx in the hash table */ 241 242 u_int ppod_start; 243 u_int ddp_pgsz[4]; 244 vmem_t *ppod_arena; 245 246 struct mtx clip_table_lock; 247 struct clip_head clip_table; 248 int clip_gen; 249 250 /* WRs that will not be sent to the chip because L2 resolution failed */ 251 struct mtx unsent_wr_lock; 252 STAILQ_HEAD(, wrqe) unsent_wr_list; 253 struct task reclaim_wr_resources; 254 }; 255 256 static inline struct tom_data * 257 tod_td(struct toedev *tod) 258 { 259 260 return (__containerof(tod, struct tom_data, tod)); 261 } 262 263 static inline struct adapter * 264 td_adapter(struct tom_data *td) 265 { 266 267 return (td->tod.tod_softc); 268 } 269 270 static inline void 271 set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 272 { 273 274 M_ASSERTPKTHDR(m); 275 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 276 } 277 278 static inline uint8_t 279 mbuf_ulp_submode(struct mbuf *m) 280 { 281 282 M_ASSERTPKTHDR(m); 283 return (m->m_pkthdr.PH_per.eight[0]); 284 } 285 286 /* t4_tom.c */ 287 struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 288 struct toepcb *hold_toepcb(struct toepcb *); 289 void free_toepcb(struct toepcb *); 290 void offload_socket(struct socket *, struct toepcb *); 291 void undo_offload_socket(struct socket *); 292 void final_cpl_received(struct toepcb *); 293 void insert_tid(struct adapter *, int, void *); 294 void *lookup_tid(struct adapter *, int); 295 void update_tid(struct adapter *, int, void *); 296 void remove_tid(struct adapter *, int); 297 void release_tid(struct adapter *, int, struct sge_wrq *); 298 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int); 299 u_long select_rcv_wnd(struct socket *); 300 int select_rcv_wscale(void); 301 uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 302 int, int, int, int); 303 uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 304 void set_tcpddp_ulp_mode(struct toepcb *); 305 int negative_advice(int); 306 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *); 307 void release_lip(struct tom_data *, struct clip_entry *); 308 309 /* t4_connect.c */ 310 void t4_init_connect_cpl_handlers(void); 311 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 312 struct sockaddr *); 313 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 314 315 /* t4_listen.c */ 316 void t4_init_listen_cpl_handlers(void); 317 int t4_listen_start(struct toedev *, struct tcpcb *); 318 int t4_listen_stop(struct toedev *, struct tcpcb *); 319 void t4_syncache_added(struct toedev *, void *); 320 void t4_syncache_removed(struct toedev *, void *); 321 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 322 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 323 struct mbuf *); 324 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 325 struct mbuf *); 326 void t4_offload_socket(struct toedev *, void *, struct socket *); 327 328 /* t4_cpl_io.c */ 329 void aiotx_init_toep(struct toepcb *); 330 int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 331 void t4_init_cpl_io_handlers(void); 332 void t4_uninit_cpl_io_handlers(void); 333 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 334 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 335 void send_reset(struct adapter *, struct toepcb *, uint32_t); 336 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 337 void t4_rcvd(struct toedev *, struct tcpcb *); 338 void t4_rcvd_locked(struct toedev *, struct tcpcb *); 339 int t4_tod_output(struct toedev *, struct tcpcb *); 340 int t4_send_fin(struct toedev *, struct tcpcb *); 341 int t4_send_rst(struct toedev *, struct tcpcb *); 342 void t4_set_tcb_field(struct adapter *, struct sge_wrq *, int, uint16_t, 343 uint64_t, uint64_t, int, int, int); 344 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 345 void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 346 int do_set_tcb_rpl(struct sge_iq *, const struct rss_header *, struct mbuf *); 347 348 /* t4_ddp.c */ 349 void t4_init_ddp(struct adapter *, struct tom_data *); 350 void t4_uninit_ddp(struct adapter *, struct tom_data *); 351 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 352 struct mbuf **, struct mbuf **, int *); 353 int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 354 int t4_ddp_mod_load(void); 355 void t4_ddp_mod_unload(void); 356 void ddp_assert_empty(struct toepcb *); 357 void ddp_init_toep(struct toepcb *); 358 void ddp_uninit_toep(struct toepcb *); 359 void ddp_queue_toep(struct toepcb *); 360 void release_ddp_resources(struct toepcb *toep); 361 void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 362 void handle_ddp_indicate(struct toepcb *); 363 void handle_ddp_tcb_rpl(struct toepcb *, const struct cpl_set_tcb_rpl *); 364 void insert_ddp_data(struct toepcb *, uint32_t); 365 366 #endif 367