t4_tom.h revision 348704
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: stable/11/sys/dev/cxgbe/tom/t4_tom.h 348704 2019-06-05 21:46:56Z np $ 28 * 29 */ 30 31#ifndef __T4_TOM_H__ 32#define __T4_TOM_H__ 33#include <sys/vmem.h> 34#include "tom/t4_tls.h" 35 36#define LISTEN_HASH_SIZE 32 37 38/* 39 * Min receive window. We want it to be large enough to accommodate receive 40 * coalescing, handle jumbo frames, and not trigger sender SWS avoidance. 41 */ 42#define MIN_RCV_WND (24 * 1024U) 43 44/* 45 * Max receive window supported by HW in bytes. Only a small part of it can 46 * be set through option0, the rest needs to be set through RX_DATA_ACK. 47 */ 48#define MAX_RCV_WND ((1U << 27) - 1) 49 50#define DDP_RSVD_WIN (16 * 1024U) 51#define SB_DDP_INDICATE SB_IN_TOE /* soreceive must respond to indicate */ 52 53#define USE_DDP_RX_FLOW_CONTROL 54 55#define PPOD_SZ(n) ((n) * sizeof(struct pagepod)) 56#define PPOD_SIZE (PPOD_SZ(1)) 57 58/* TOE PCB flags */ 59enum { 60 TPF_ATTACHED = (1 << 0), /* a tcpcb refers to this toepcb */ 61 TPF_FLOWC_WR_SENT = (1 << 1), /* firmware flow context WR sent */ 62 TPF_TX_DATA_SENT = (1 << 2), /* some data sent */ 63 TPF_TX_SUSPENDED = (1 << 3), /* tx suspended for lack of resources */ 64 TPF_SEND_FIN = (1 << 4), /* send FIN after all pending data */ 65 TPF_FIN_SENT = (1 << 5), /* FIN has been sent */ 66 TPF_ABORT_SHUTDOWN = (1 << 6), /* connection abort is in progress */ 67 TPF_CPL_PENDING = (1 << 7), /* haven't received the last CPL */ 68 TPF_SYNQE = (1 << 8), /* synq_entry, not really a toepcb */ 69 TPF_SYNQE_EXPANDED = (1 << 9), /* toepcb ready, tid context updated */ 70 TPF_FORCE_CREDITS = (1 << 10), /* always send credits */ 71}; 72 73enum { 74 DDP_OK = (1 << 0), /* OK to turn on DDP */ 75 DDP_SC_REQ = (1 << 1), /* state change (on/off) requested */ 76 DDP_ON = (1 << 2), /* DDP is turned on */ 77 DDP_BUF0_ACTIVE = (1 << 3), /* buffer 0 in use (not invalidated) */ 78 DDP_BUF1_ACTIVE = (1 << 4), /* buffer 1 in use (not invalidated) */ 79 DDP_TASK_ACTIVE = (1 << 5), /* requeue task is queued / running */ 80 DDP_DEAD = (1 << 6), /* toepcb is shutting down */ 81}; 82 83struct sockopt; 84struct offload_settings; 85 86struct ofld_tx_sdesc { 87 uint32_t plen; /* payload length */ 88 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 89 void *iv_buffer; /* optional buffer holding IVs for TLS */ 90}; 91 92struct ppod_region { 93 u_int pr_start; 94 u_int pr_len; 95 u_int pr_page_shift[4]; 96 uint32_t pr_tag_mask; /* hardware tagmask for this region. */ 97 uint32_t pr_invalid_bit; /* OR with this to invalidate tag. */ 98 uint32_t pr_alias_mask; /* AND with tag to get alias bits. */ 99 u_int pr_alias_shift; /* shift this much for first alias bit. */ 100 vmem_t *pr_arena; 101}; 102 103struct ppod_reservation { 104 struct ppod_region *prsv_pr; 105 uint32_t prsv_tag; /* Full tag: pgsz, alias, tag, color */ 106 u_int prsv_nppods; 107}; 108 109struct pageset { 110 TAILQ_ENTRY(pageset) link; 111 vm_page_t *pages; 112 int npages; 113 int flags; 114 int offset; /* offset in first page */ 115 int len; 116 struct ppod_reservation prsv; 117 struct vmspace *vm; 118 vm_offset_t start; 119 u_int vm_timestamp; 120}; 121 122TAILQ_HEAD(pagesetq, pageset); 123 124#define PS_WIRED 0x0001 /* Pages wired rather than held. */ 125#define PS_PPODS_WRITTEN 0x0002 /* Page pods written to the card. */ 126 127#define EXT_FLAG_AIOTX EXT_FLAG_VENDOR1 128 129#define IS_AIOTX_MBUF(m) \ 130 ((m)->m_flags & M_EXT && (m)->m_ext.ext_flags & EXT_FLAG_AIOTX) 131 132struct ddp_buffer { 133 struct pageset *ps; 134 135 struct kaiocb *job; 136 int cancel_pending; 137}; 138 139struct ddp_pcb { 140 u_int flags; 141 struct ddp_buffer db[2]; 142 TAILQ_HEAD(, pageset) cached_pagesets; 143 TAILQ_HEAD(, kaiocb) aiojobq; 144 u_int waiting_count; 145 u_int active_count; 146 u_int cached_count; 147 int active_id; /* the currently active DDP buffer */ 148 struct task requeue_task; 149 struct kaiocb *queueing; 150 struct mtx lock; 151}; 152 153struct aiotx_buffer { 154 struct pageset ps; 155 struct kaiocb *job; 156 int refcount; 157}; 158 159struct toepcb { 160 TAILQ_ENTRY(toepcb) link; /* toep_list */ 161 u_int flags; /* miscellaneous flags */ 162 int refcount; 163 struct tom_data *td; 164 struct inpcb *inp; /* backpointer to host stack's PCB */ 165 struct vnet *vnet; 166 struct vi_info *vi; /* virtual interface */ 167 struct sge_wrq *ofld_txq; 168 struct sge_ofld_rxq *ofld_rxq; 169 struct sge_wrq *ctrlq; 170 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 171 struct clip_entry *ce; /* CLIP table entry used by this tid */ 172 int tid; /* Connection identifier */ 173 int tc_idx; /* traffic class that this tid is bound to */ 174 175 /* tx credit handling */ 176 u_int tx_total; /* total tx WR credits (in 16B units) */ 177 u_int tx_credits; /* tx WR credits (in 16B units) available */ 178 u_int tx_nocompl; /* tx WR credits since last compl request */ 179 u_int plen_nocompl; /* payload since last compl request */ 180 181 int opt0_rcv_bufsize; /* XXX: save full opt0/opt2 for later? */ 182 183 u_int ulp_mode; /* ULP mode */ 184 void *ulpcb; 185 void *ulpcb2; 186 struct mbufq ulp_pduq; /* PDUs waiting to be sent out. */ 187 struct mbufq ulp_pdu_reclaimq; 188 189 struct ddp_pcb ddp; 190 struct tls_ofld_info tls; 191 192 TAILQ_HEAD(, kaiocb) aiotx_jobq; 193 struct task aiotx_task; 194 bool aiotx_task_active; 195 196 /* Tx software descriptor */ 197 uint8_t txsd_total; 198 uint8_t txsd_pidx; 199 uint8_t txsd_cidx; 200 uint8_t txsd_avail; 201 struct ofld_tx_sdesc txsd[]; 202}; 203 204#define DDP_LOCK(toep) mtx_lock(&(toep)->ddp.lock) 205#define DDP_UNLOCK(toep) mtx_unlock(&(toep)->ddp.lock) 206#define DDP_ASSERT_LOCKED(toep) mtx_assert(&(toep)->ddp.lock, MA_OWNED) 207 208struct flowc_tx_params { 209 uint32_t snd_nxt; 210 uint32_t rcv_nxt; 211 unsigned int snd_space; 212 unsigned int mss; 213}; 214 215/* 216 * Compressed state for embryonic connections for a listener. 217 */ 218struct synq_entry { 219 struct listen_ctx *lctx; /* backpointer to listen ctx */ 220 struct mbuf *syn; 221 int flags; /* same as toepcb's tp_flags */ 222 volatile int ok_to_respond; 223 volatile u_int refcnt; 224 int tid; 225 uint32_t iss; 226 uint32_t irs; 227 uint32_t ts; 228 uint16_t txqid; 229 uint16_t rxqid; 230 uint16_t l2e_idx; 231 uint16_t ulp_mode; 232 uint16_t rcv_bufsize; 233 __be16 tcp_opt; /* from cpl_pass_establish */ 234 struct toepcb *toep; 235}; 236 237/* listen_ctx flags */ 238#define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 239 240struct listen_ctx { 241 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 242 volatile int refcount; 243 int stid; 244 struct stid_region stid_region; 245 int flags; 246 struct inpcb *inp; /* listening socket's inp */ 247 struct vnet *vnet; 248 struct sge_wrq *ctrlq; 249 struct sge_ofld_rxq *ofld_rxq; 250 struct clip_entry *ce; 251}; 252 253struct tom_data { 254 struct toedev tod; 255 256 /* toepcb's associated with this TOE device */ 257 struct mtx toep_list_lock; 258 TAILQ_HEAD(, toepcb) toep_list; 259 260 struct mtx lctx_hash_lock; 261 LIST_HEAD(, listen_ctx) *listen_hash; 262 u_long listen_mask; 263 int lctx_count; /* # of lctx in the hash table */ 264 265 struct ppod_region pr; 266 267 /* WRs that will not be sent to the chip because L2 resolution failed */ 268 struct mtx unsent_wr_lock; 269 STAILQ_HEAD(, wrqe) unsent_wr_list; 270 struct task reclaim_wr_resources; 271}; 272 273static inline struct tom_data * 274tod_td(struct toedev *tod) 275{ 276 277 return (__containerof(tod, struct tom_data, tod)); 278} 279 280static inline struct adapter * 281td_adapter(struct tom_data *td) 282{ 283 284 return (td->tod.tod_softc); 285} 286 287static inline void 288set_mbuf_ulp_submode(struct mbuf *m, uint8_t ulp_submode) 289{ 290 291 M_ASSERTPKTHDR(m); 292 m->m_pkthdr.PH_per.eight[0] = ulp_submode; 293} 294 295static inline uint8_t 296mbuf_ulp_submode(struct mbuf *m) 297{ 298 299 M_ASSERTPKTHDR(m); 300 return (m->m_pkthdr.PH_per.eight[0]); 301} 302 303/* t4_tom.c */ 304struct toepcb *alloc_toepcb(struct vi_info *, int, int, int); 305struct toepcb *hold_toepcb(struct toepcb *); 306void free_toepcb(struct toepcb *); 307void offload_socket(struct socket *, struct toepcb *); 308void undo_offload_socket(struct socket *); 309void final_cpl_received(struct toepcb *); 310void insert_tid(struct adapter *, int, void *, int); 311void *lookup_tid(struct adapter *, int); 312void update_tid(struct adapter *, int, void *); 313void remove_tid(struct adapter *, int, int); 314int find_best_mtu_idx(struct adapter *, struct in_conninfo *, 315 struct offload_settings *); 316u_long select_rcv_wnd(struct socket *); 317int select_rcv_wscale(void); 318uint64_t calc_opt0(struct socket *, struct vi_info *, struct l2t_entry *, 319 int, int, int, int, struct offload_settings *); 320uint64_t select_ntuple(struct vi_info *, struct l2t_entry *); 321int select_ulp_mode(struct socket *, struct adapter *, 322 struct offload_settings *); 323void set_ulp_mode(struct toepcb *, int); 324int negative_advice(int); 325 326/* t4_connect.c */ 327void t4_init_connect_cpl_handlers(void); 328void t4_uninit_connect_cpl_handlers(void); 329int t4_connect(struct toedev *, struct socket *, struct rtentry *, 330 struct sockaddr *); 331void act_open_failure_cleanup(struct adapter *, u_int, u_int); 332 333/* t4_listen.c */ 334void t4_init_listen_cpl_handlers(void); 335void t4_uninit_listen_cpl_handlers(void); 336int t4_listen_start(struct toedev *, struct tcpcb *); 337int t4_listen_stop(struct toedev *, struct tcpcb *); 338void t4_syncache_added(struct toedev *, void *); 339void t4_syncache_removed(struct toedev *, void *); 340int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 341int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 342 struct mbuf *); 343int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 344 struct mbuf *); 345void t4_offload_socket(struct toedev *, void *, struct socket *); 346void synack_failure_cleanup(struct adapter *, int); 347 348/* t4_cpl_io.c */ 349void aiotx_init_toep(struct toepcb *); 350int t4_aio_queue_aiotx(struct socket *, struct kaiocb *); 351void t4_init_cpl_io_handlers(void); 352void t4_uninit_cpl_io_handlers(void); 353void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 354void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 355void send_reset(struct adapter *, struct toepcb *, uint32_t); 356int send_rx_credits(struct adapter *, struct toepcb *, int); 357void send_rx_modulate(struct adapter *, struct toepcb *); 358void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 359int t4_close_conn(struct adapter *, struct toepcb *); 360void t4_rcvd(struct toedev *, struct tcpcb *); 361void t4_rcvd_locked(struct toedev *, struct tcpcb *); 362int t4_tod_output(struct toedev *, struct tcpcb *); 363int t4_send_fin(struct toedev *, struct tcpcb *); 364int t4_send_rst(struct toedev *, struct tcpcb *); 365void t4_set_tcb_field(struct adapter *, struct sge_wrq *, struct toepcb *, 366 uint16_t, uint64_t, uint64_t, int, int); 367void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 368void t4_push_pdus(struct adapter *sc, struct toepcb *toep, int drop); 369 370/* t4_ddp.c */ 371int t4_init_ppod_region(struct ppod_region *, struct t4_range *, u_int, 372 const char *); 373void t4_free_ppod_region(struct ppod_region *); 374int t4_alloc_page_pods_for_ps(struct ppod_region *, struct pageset *); 375int t4_alloc_page_pods_for_buf(struct ppod_region *, vm_offset_t, int, 376 struct ppod_reservation *); 377int t4_write_page_pods_for_ps(struct adapter *, struct sge_wrq *, int, 378 struct pageset *); 379int t4_write_page_pods_for_buf(struct adapter *, struct sge_wrq *, int tid, 380 struct ppod_reservation *, vm_offset_t, int); 381void t4_free_page_pods(struct ppod_reservation *); 382int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 383 struct mbuf **, struct mbuf **, int *); 384int t4_aio_queue_ddp(struct socket *, struct kaiocb *); 385void t4_ddp_mod_load(void); 386void t4_ddp_mod_unload(void); 387void ddp_assert_empty(struct toepcb *); 388void ddp_init_toep(struct toepcb *); 389void ddp_uninit_toep(struct toepcb *); 390void ddp_queue_toep(struct toepcb *); 391void release_ddp_resources(struct toepcb *toep); 392void handle_ddp_close(struct toepcb *, struct tcpcb *, uint32_t); 393void handle_ddp_indicate(struct toepcb *); 394void insert_ddp_data(struct toepcb *, uint32_t); 395const struct offload_settings *lookup_offload_policy(struct adapter *, int, 396 struct mbuf *, uint16_t, struct inpcb *); 397 398/* t4_tls.c */ 399bool can_tls_offload(struct adapter *); 400int t4_ctloutput_tls(struct socket *, struct sockopt *); 401void t4_push_tls_records(struct adapter *, struct toepcb *, int); 402void t4_tls_mod_load(void); 403void t4_tls_mod_unload(void); 404void tls_establish(struct toepcb *); 405void tls_init_toep(struct toepcb *); 406int tls_rx_key(struct toepcb *); 407void tls_stop_handshake_timer(struct toepcb *); 408int tls_tx_key(struct toepcb *); 409void tls_uninit_toep(struct toepcb *); 410 411#endif 412