1 /*- 2 * Copyright (c) 2012 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 34 #define KTR_CXGBE KTR_SPARE3 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 }; 78 79 struct ofld_tx_sdesc { 80 uint32_t plen; /* payload length */ 81 uint8_t tx_credits; /* firmware tx credits (unit is 16B) */ 82 }; 83 84 struct ppod_region { 85 TAILQ_ENTRY(ppod_region) link; 86 int used; /* # of pods used by this region */ 87 int free; /* # of contiguous pods free right after this region */ 88 }; 89 90 struct ddp_buffer { 91 uint32_t tag; /* includes color, page pod addr, and DDP page size */ 92 int nppods; 93 int offset; 94 int len; 95 struct ppod_region ppod_region; 96 int npages; 97 vm_page_t *pages; 98 }; 99 100 struct toepcb { 101 TAILQ_ENTRY(toepcb) link; /* toep_list */ 102 u_int flags; /* miscellaneous flags */ 103 struct tom_data *td; 104 struct inpcb *inp; /* backpointer to host stack's PCB */ 105 struct port_info *port; /* physical port */ 106 struct sge_wrq *ofld_txq; 107 struct sge_ofld_rxq *ofld_rxq; 108 struct sge_wrq *ctrlq; 109 struct l2t_entry *l2te; /* L2 table entry used by this connection */ 110 struct clip_entry *ce; /* CLIP table entry used by this tid */ 111 int tid; /* Connection identifier */ 112 113 /* tx credit handling */ 114 u_int tx_total; /* total tx WR credits (in 16B units) */ 115 u_int tx_credits; /* tx WR credits (in 16B units) available */ 116 u_int tx_nocompl; /* tx WR credits since last compl request */ 117 u_int plen_nocompl; /* payload since last compl request */ 118 119 /* rx credit handling */ 120 u_int sb_cc; /* last noted value of so_rcv->sb_cc */ 121 int rx_credits; /* rx credits (in bytes) to be returned to hw */ 122 123 u_int ulp_mode; /* ULP mode */ 124 125 u_int ddp_flags; 126 struct ddp_buffer *db[2]; 127 time_t ddp_disabled; 128 uint8_t ddp_score; 129 130 /* Tx software descriptor */ 131 uint8_t txsd_total; 132 uint8_t txsd_pidx; 133 uint8_t txsd_cidx; 134 uint8_t txsd_avail; 135 struct ofld_tx_sdesc txsd[]; 136 }; 137 138 struct flowc_tx_params { 139 uint32_t snd_nxt; 140 uint32_t rcv_nxt; 141 unsigned int snd_space; 142 unsigned int mss; 143 }; 144 145 #define DDP_RETRY_WAIT 5 /* seconds to wait before re-enabling DDP */ 146 #define DDP_LOW_SCORE 1 147 #define DDP_HIGH_SCORE 3 148 149 /* 150 * Compressed state for embryonic connections for a listener. Barely fits in 151 * 64B, try not to grow it further. 152 */ 153 struct synq_entry { 154 TAILQ_ENTRY(synq_entry) link; /* listen_ctx's synq link */ 155 int flags; /* same as toepcb's tp_flags */ 156 int tid; 157 struct listen_ctx *lctx; /* backpointer to listen ctx */ 158 struct mbuf *syn; 159 uint32_t iss; 160 uint32_t ts; 161 volatile uintptr_t wr; 162 volatile u_int refcnt; 163 uint16_t l2e_idx; 164 uint16_t rcv_bufsize; 165 }; 166 167 /* listen_ctx flags */ 168 #define LCTX_RPL_PENDING 1 /* waiting for a CPL_PASS_OPEN_RPL */ 169 170 struct listen_ctx { 171 LIST_ENTRY(listen_ctx) link; /* listen hash linkage */ 172 volatile int refcount; 173 int stid; 174 struct stid_region stid_region; 175 int flags; 176 struct inpcb *inp; /* listening socket's inp */ 177 struct sge_wrq *ctrlq; 178 struct sge_ofld_rxq *ofld_rxq; 179 struct clip_entry *ce; 180 TAILQ_HEAD(, synq_entry) synq; 181 }; 182 183 TAILQ_HEAD(ppod_head, ppod_region); 184 185 struct clip_entry { 186 TAILQ_ENTRY(clip_entry) link; 187 struct in6_addr lip; /* local IPv6 address */ 188 u_int refcount; 189 }; 190 191 TAILQ_HEAD(clip_head, clip_entry); 192 struct tom_data { 193 struct toedev tod; 194 195 /* toepcb's associated with this TOE device */ 196 struct mtx toep_list_lock; 197 TAILQ_HEAD(, toepcb) toep_list; 198 199 struct mtx lctx_hash_lock; 200 LIST_HEAD(, listen_ctx) *listen_hash; 201 u_long listen_mask; 202 int lctx_count; /* # of lctx in the hash table */ 203 204 struct mtx ppod_lock; 205 int nppods; 206 int nppods_free; /* # of available ppods */ 207 int nppods_free_head; /* # of available ppods at the begining */ 208 struct ppod_head ppods; 209 210 struct mtx clip_table_lock; 211 struct clip_head clip_table; 212 int clip_gen; 213 214 /* WRs that will not be sent to the chip because L2 resolution failed */ 215 struct mtx unsent_wr_lock; 216 STAILQ_HEAD(, wrqe) unsent_wr_list; 217 struct task reclaim_wr_resources; 218 }; 219 220 static inline struct tom_data * 221 tod_td(struct toedev *tod) 222 { 223 224 return (__containerof(tod, struct tom_data, tod)); 225 } 226 227 static inline struct adapter * 228 td_adapter(struct tom_data *td) 229 { 230 231 return (td->tod.tod_softc); 232 } 233 234 /* t4_tom.c */ 235 struct toepcb *alloc_toepcb(struct port_info *, int, int, int); 236 void free_toepcb(struct toepcb *); 237 void offload_socket(struct socket *, struct toepcb *); 238 void undo_offload_socket(struct socket *); 239 void final_cpl_received(struct toepcb *); 240 void insert_tid(struct adapter *, int, void *); 241 void *lookup_tid(struct adapter *, int); 242 void update_tid(struct adapter *, int, void *); 243 void remove_tid(struct adapter *, int); 244 void release_tid(struct adapter *, int, struct sge_wrq *); 245 int find_best_mtu_idx(struct adapter *, struct in_conninfo *, int); 246 u_long select_rcv_wnd(struct socket *); 247 int select_rcv_wscale(void); 248 uint64_t calc_opt0(struct socket *, struct port_info *, struct l2t_entry *, 249 int, int, int, int); 250 uint64_t select_ntuple(struct port_info *, struct l2t_entry *); 251 void set_tcpddp_ulp_mode(struct toepcb *); 252 int negative_advice(int); 253 struct clip_entry *hold_lip(struct tom_data *, struct in6_addr *); 254 void release_lip(struct tom_data *, struct clip_entry *); 255 256 /* t4_connect.c */ 257 void t4_init_connect_cpl_handlers(struct adapter *); 258 int t4_connect(struct toedev *, struct socket *, struct rtentry *, 259 struct sockaddr *); 260 void act_open_failure_cleanup(struct adapter *, u_int, u_int); 261 262 /* t4_listen.c */ 263 void t4_init_listen_cpl_handlers(struct adapter *); 264 int t4_listen_start(struct toedev *, struct tcpcb *); 265 int t4_listen_stop(struct toedev *, struct tcpcb *); 266 void t4_syncache_added(struct toedev *, void *); 267 void t4_syncache_removed(struct toedev *, void *); 268 int t4_syncache_respond(struct toedev *, void *, struct mbuf *); 269 int do_abort_req_synqe(struct sge_iq *, const struct rss_header *, 270 struct mbuf *); 271 int do_abort_rpl_synqe(struct sge_iq *, const struct rss_header *, 272 struct mbuf *); 273 void t4_offload_socket(struct toedev *, void *, struct socket *); 274 275 /* t4_cpl_io.c */ 276 void t4_init_cpl_io_handlers(struct adapter *); 277 void t4_uninit_cpl_io_handlers(struct adapter *); 278 void send_abort_rpl(struct adapter *, struct sge_wrq *, int , int); 279 void send_flowc_wr(struct toepcb *, struct flowc_tx_params *); 280 void send_reset(struct adapter *, struct toepcb *, uint32_t); 281 void make_established(struct toepcb *, uint32_t, uint32_t, uint16_t); 282 void t4_rcvd(struct toedev *, struct tcpcb *); 283 int t4_tod_output(struct toedev *, struct tcpcb *); 284 int t4_send_fin(struct toedev *, struct tcpcb *); 285 int t4_send_rst(struct toedev *, struct tcpcb *); 286 void t4_set_tcb_field(struct adapter *, struct toepcb *, int, uint16_t, 287 uint64_t, uint64_t); 288 void t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop); 289 290 /* t4_ddp.c */ 291 void t4_init_ddp(struct adapter *, struct tom_data *); 292 void t4_uninit_ddp(struct adapter *, struct tom_data *); 293 int t4_soreceive_ddp(struct socket *, struct sockaddr **, struct uio *, 294 struct mbuf **, struct mbuf **, int *); 295 struct mbuf *get_ddp_mbuf(int); 296 void enable_ddp(struct adapter *, struct toepcb *toep); 297 void release_ddp_resources(struct toepcb *toep); 298 void insert_ddp_data(struct toepcb *, uint32_t); 299 300 /* ULP related */ 301 #define CXGBE_ISCSI_MBUF_TAG 50 302 int t4tom_cpl_handler_registered(struct adapter *, unsigned int); 303 void t4tom_register_cpl_iscsi_callback(void (*fp)(struct tom_data *, 304 struct socket *, void *, unsigned int)); 305 void t4tom_register_queue_iscsi_callback(struct mbuf *(*fp)(struct socket *, 306 unsigned int, int *)); 307 void t4_ulp_push_frames(struct adapter *sc, struct toepcb *toep, int); 308 int t4_cpl_iscsi_callback(struct tom_data *, struct toepcb *, void *, uint32_t); 309 struct mbuf *t4_queue_iscsi_callback(struct socket *, struct toepcb *, uint32_t, 310 int *); 311 extern void (*tom_cpl_iscsi_callback)(struct tom_data *, struct socket *, 312 void *, unsigned int); 313 extern struct mbuf *(*tom_queue_iscsi_callback)(struct socket*, unsigned int, 314 int *); 315 #endif 316