1 /* 2 * Copyright(c) 2015 - 2017 Intel Corporation. 3 * 4 * This file is provided under a dual BSD/GPLv2 license. When using or 5 * redistributing this file, you may do so under either license. 6 * 7 * GPL LICENSE SUMMARY 8 * 9 * This program is free software; you can redistribute it and/or modify 10 * it under the terms of version 2 of the GNU General Public License as 11 * published by the Free Software Foundation. 12 * 13 * This program is distributed in the hope that it will be useful, but 14 * WITHOUT ANY WARRANTY; without even the implied warranty of 15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 16 * General Public License for more details. 17 * 18 * BSD LICENSE 19 * 20 * Redistribution and use in source and binary forms, with or without 21 * modification, are permitted provided that the following conditions 22 * are met: 23 * 24 * - Redistributions of source code must retain the above copyright 25 * notice, this list of conditions and the following disclaimer. 26 * - Redistributions in binary form must reproduce the above copyright 27 * notice, this list of conditions and the following disclaimer in 28 * the documentation and/or other materials provided with the 29 * distribution. 30 * - Neither the name of Intel Corporation nor the names of its 31 * contributors may be used to endorse or promote products derived 32 * from this software without specific prior written permission. 33 * 34 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 35 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 36 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 37 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 38 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 39 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 40 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 41 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 42 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 43 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 44 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 45 * 46 */ 47 48 #ifndef HFI1_VERBS_H 49 #define HFI1_VERBS_H 50 51 #include <linux/types.h> 52 #include <linux/seqlock.h> 53 #include <linux/kernel.h> 54 #include <linux/interrupt.h> 55 #include <linux/kref.h> 56 #include <linux/workqueue.h> 57 #include <linux/kthread.h> 58 #include <linux/completion.h> 59 #include <linux/slab.h> 60 #include <rdma/ib_pack.h> 61 #include <rdma/ib_user_verbs.h> 62 #include <rdma/ib_mad.h> 63 #include <rdma/ib_hdrs.h> 64 #include <rdma/rdma_vt.h> 65 #include <rdma/rdmavt_qp.h> 66 #include <rdma/rdmavt_cq.h> 67 68 struct hfi1_ctxtdata; 69 struct hfi1_pportdata; 70 struct hfi1_devdata; 71 struct hfi1_packet; 72 73 #include "iowait.h" 74 75 #define HFI1_MAX_RDMA_ATOMIC 16 76 77 /* 78 * Increment this value if any changes that break userspace ABI 79 * compatibility are made. 80 */ 81 #define HFI1_UVERBS_ABI_VERSION 2 82 83 /* IB Performance Manager status values */ 84 #define IB_PMA_SAMPLE_STATUS_DONE 0x00 85 #define IB_PMA_SAMPLE_STATUS_STARTED 0x01 86 #define IB_PMA_SAMPLE_STATUS_RUNNING 0x02 87 88 /* Mandatory IB performance counter select values. */ 89 #define IB_PMA_PORT_XMIT_DATA cpu_to_be16(0x0001) 90 #define IB_PMA_PORT_RCV_DATA cpu_to_be16(0x0002) 91 #define IB_PMA_PORT_XMIT_PKTS cpu_to_be16(0x0003) 92 #define IB_PMA_PORT_RCV_PKTS cpu_to_be16(0x0004) 93 #define IB_PMA_PORT_XMIT_WAIT cpu_to_be16(0x0005) 94 95 #define HFI1_VENDOR_IPG cpu_to_be16(0xFFA0) 96 97 #define IB_DEFAULT_GID_PREFIX cpu_to_be64(0xfe80000000000000ULL) 98 99 #define RC_OP(x) IB_OPCODE_RC_##x 100 #define UC_OP(x) IB_OPCODE_UC_##x 101 102 /* flags passed by hfi1_ib_rcv() */ 103 enum { 104 HFI1_HAS_GRH = (1 << 0), 105 }; 106 107 struct hfi1_ahg_info { 108 u32 ahgdesc[2]; 109 u16 tx_flags; 110 u8 ahgcount; 111 u8 ahgidx; 112 }; 113 114 struct hfi1_sdma_header { 115 __le64 pbc; 116 struct ib_header hdr; 117 } __packed; 118 119 /* 120 * hfi1 specific data structures that will be hidden from rvt after the queue 121 * pair is made common 122 */ 123 struct hfi1_qp_priv { 124 struct hfi1_ahg_info *s_ahg; /* ahg info for next header */ 125 struct sdma_engine *s_sde; /* current sde */ 126 struct send_context *s_sendcontext; /* current sendcontext */ 127 u8 s_sc; /* SC[0..4] for next packet */ 128 struct iowait s_iowait; 129 struct rvt_qp *owner; 130 }; 131 132 /* 133 * This structure is used to hold commonly lookedup and computed values during 134 * the send engine progress. 135 */ 136 struct hfi1_pkt_state { 137 struct hfi1_ibdev *dev; 138 struct hfi1_ibport *ibp; 139 struct hfi1_pportdata *ppd; 140 struct verbs_txreq *s_txreq; 141 unsigned long flags; 142 unsigned long timeout; 143 unsigned long timeout_int; 144 int cpu; 145 bool in_thread; 146 }; 147 148 #define HFI1_PSN_CREDIT 16 149 150 struct hfi1_opcode_stats { 151 u64 n_packets; /* number of packets */ 152 u64 n_bytes; /* total number of bytes */ 153 }; 154 155 struct hfi1_opcode_stats_perctx { 156 struct hfi1_opcode_stats stats[256]; 157 }; 158 159 static inline void inc_opstats( 160 u32 tlen, 161 struct hfi1_opcode_stats *stats) 162 { 163 #ifdef CONFIG_DEBUG_FS 164 stats->n_bytes += tlen; 165 stats->n_packets++; 166 #endif 167 } 168 169 struct hfi1_ibport { 170 struct rvt_qp __rcu *qp[2]; 171 struct rvt_ibport rvp; 172 173 /* the first 16 entries are sl_to_vl for !OPA */ 174 u8 sl_to_sc[32]; 175 u8 sc_to_sl[32]; 176 }; 177 178 struct hfi1_ibdev { 179 struct rvt_dev_info rdi; /* Must be first */ 180 181 /* QP numbers are shared by all IB ports */ 182 /* protect txwait list */ 183 seqlock_t txwait_lock ____cacheline_aligned_in_smp; 184 struct list_head txwait; /* list for wait verbs_txreq */ 185 struct list_head memwait; /* list for wait kernel memory */ 186 struct kmem_cache *verbs_txreq_cache; 187 u64 n_txwait; 188 u64 n_kmem_wait; 189 190 /* protect iowait lists */ 191 seqlock_t iowait_lock ____cacheline_aligned_in_smp; 192 u64 n_piowait; 193 u64 n_piodrain; 194 struct timer_list mem_timer; 195 196 #ifdef CONFIG_DEBUG_FS 197 /* per HFI debugfs */ 198 struct dentry *hfi1_ibdev_dbg; 199 /* per HFI symlinks to above */ 200 struct dentry *hfi1_ibdev_link; 201 #ifdef CONFIG_FAULT_INJECTION 202 struct fault_opcode *fault_opcode; 203 struct fault_packet *fault_packet; 204 bool fault_suppress_err; 205 #endif 206 #endif 207 }; 208 209 static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev) 210 { 211 struct rvt_dev_info *rdi; 212 213 rdi = container_of(ibdev, struct rvt_dev_info, ibdev); 214 return container_of(rdi, struct hfi1_ibdev, rdi); 215 } 216 217 static inline struct rvt_qp *iowait_to_qp(struct iowait *s_iowait) 218 { 219 struct hfi1_qp_priv *priv; 220 221 priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait); 222 return priv->owner; 223 } 224 225 /* 226 * Send if not busy or waiting for I/O and either 227 * a RC response is pending or we can process send work requests. 228 */ 229 static inline int hfi1_send_ok(struct rvt_qp *qp) 230 { 231 return !(qp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT_IO)) && 232 (qp->s_hdrwords || (qp->s_flags & RVT_S_RESP_PENDING) || 233 !(qp->s_flags & RVT_S_ANY_WAIT_SEND)); 234 } 235 236 /* 237 * This must be called with s_lock held. 238 */ 239 void hfi1_bad_pqkey(struct hfi1_ibport *ibp, __be16 trap_num, u32 key, u32 sl, 240 u32 qp1, u32 qp2, u16 lid1, u16 lid2); 241 void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num); 242 void hfi1_sys_guid_chg(struct hfi1_ibport *ibp); 243 void hfi1_node_desc_chg(struct hfi1_ibport *ibp); 244 int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port, 245 const struct ib_wc *in_wc, const struct ib_grh *in_grh, 246 const struct ib_mad_hdr *in_mad, size_t in_mad_size, 247 struct ib_mad_hdr *out_mad, size_t *out_mad_size, 248 u16 *out_mad_pkey_index); 249 250 /* 251 * The PSN_MASK and PSN_SHIFT allow for 252 * 1) comparing two PSNs 253 * 2) returning the PSN with any upper bits masked 254 * 3) returning the difference between to PSNs 255 * 256 * The number of significant bits in the PSN must 257 * necessarily be at least one bit less than 258 * the container holding the PSN. 259 */ 260 #ifndef CONFIG_HFI1_VERBS_31BIT_PSN 261 #define PSN_MASK 0xFFFFFF 262 #define PSN_SHIFT 8 263 #else 264 #define PSN_MASK 0x7FFFFFFF 265 #define PSN_SHIFT 1 266 #endif 267 #define PSN_MODIFY_MASK 0xFFFFFF 268 269 /* 270 * Compare two PSNs 271 * Returns an integer <, ==, or > than zero. 272 */ 273 static inline int cmp_psn(u32 a, u32 b) 274 { 275 return (((int)a) - ((int)b)) << PSN_SHIFT; 276 } 277 278 /* 279 * Return masked PSN 280 */ 281 static inline u32 mask_psn(u32 a) 282 { 283 return a & PSN_MASK; 284 } 285 286 /* 287 * Return delta between two PSNs 288 */ 289 static inline u32 delta_psn(u32 a, u32 b) 290 { 291 return (((int)a - (int)b) << PSN_SHIFT) >> PSN_SHIFT; 292 } 293 294 struct verbs_txreq; 295 void hfi1_put_txreq(struct verbs_txreq *tx); 296 297 int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 298 299 void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length, 300 bool release, bool copy_last); 301 302 void hfi1_cnp_rcv(struct hfi1_packet *packet); 303 304 void hfi1_uc_rcv(struct hfi1_packet *packet); 305 306 void hfi1_rc_rcv(struct hfi1_packet *packet); 307 308 void hfi1_rc_hdrerr( 309 struct hfi1_ctxtdata *rcd, 310 struct ib_header *hdr, 311 u32 rcv_flags, 312 struct rvt_qp *qp); 313 314 u8 ah_to_sc(struct ib_device *ibdev, struct rdma_ah_attr *ah_attr); 315 316 struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid); 317 318 void hfi1_rc_send_complete(struct rvt_qp *qp, struct ib_header *hdr); 319 320 void hfi1_ud_rcv(struct hfi1_packet *packet); 321 322 int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey); 323 324 int hfi1_rvt_get_rwqe(struct rvt_qp *qp, int wr_id_only); 325 326 void hfi1_migrate_qp(struct rvt_qp *qp); 327 328 int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr, 329 int attr_mask, struct ib_udata *udata); 330 331 void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr, 332 int attr_mask, struct ib_udata *udata); 333 void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait); 334 int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe); 335 336 extern const u32 rc_only_opcode; 337 extern const u32 uc_only_opcode; 338 339 static inline u8 get_opcode(struct ib_header *h) 340 { 341 u16 lnh = be16_to_cpu(h->lrh[0]) & 3; 342 343 if (lnh == IB_LNH_IBA_LOCAL) 344 return be32_to_cpu(h->u.oth.bth[0]) >> 24; 345 else 346 return be32_to_cpu(h->u.l.oth.bth[0]) >> 24; 347 } 348 349 int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct ib_header *hdr, 350 int has_grh, struct rvt_qp *qp, u32 bth0); 351 352 u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr, 353 const struct ib_global_route *grh, u32 hwords, u32 nwords); 354 355 void hfi1_make_ruc_header(struct rvt_qp *qp, struct ib_other_headers *ohdr, 356 u32 bth0, u32 bth2, int middle, 357 struct hfi1_pkt_state *ps); 358 359 void _hfi1_do_send(struct work_struct *work); 360 361 void hfi1_do_send_from_rvt(struct rvt_qp *qp); 362 363 void hfi1_do_send(struct rvt_qp *qp, bool in_thread); 364 365 void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, 366 enum ib_wc_status status); 367 368 void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct rvt_qp *qp, int is_fecn); 369 370 int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 371 372 int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 373 374 int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 375 376 int hfi1_register_ib_device(struct hfi1_devdata *); 377 378 void hfi1_unregister_ib_device(struct hfi1_devdata *); 379 380 void hfi1_ib_rcv(struct hfi1_packet *packet); 381 382 unsigned hfi1_get_npkeys(struct hfi1_devdata *); 383 384 int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps, 385 u64 pbc); 386 387 int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps, 388 u64 pbc); 389 390 int hfi1_wss_init(void); 391 void hfi1_wss_exit(void); 392 393 /* platform specific: return the lowest level cache (llc) size, in KiB */ 394 static inline int wss_llc_size(void) 395 { 396 /* assume that the boot CPU value is universal for all CPUs */ 397 return boot_cpu_data.x86_cache_size; 398 } 399 400 /* platform specific: cacheless copy */ 401 static inline void cacheless_memcpy(void *dst, void *src, size_t n) 402 { 403 /* 404 * Use the only available X64 cacheless copy. Add a __user cast 405 * to quiet sparse. The src agument is already in the kernel so 406 * there are no security issues. The extra fault recovery machinery 407 * is not invoked. 408 */ 409 __copy_user_nocache(dst, (void __user *)src, n, 0); 410 } 411 412 extern const enum ib_wc_opcode ib_hfi1_wc_opcode[]; 413 414 extern const u8 hdr_len_by_opcode[]; 415 416 extern const int ib_rvt_state_ops[]; 417 418 extern __be64 ib_hfi1_sys_image_guid; /* in network order */ 419 420 extern unsigned int hfi1_max_cqes; 421 422 extern unsigned int hfi1_max_cqs; 423 424 extern unsigned int hfi1_max_qp_wrs; 425 426 extern unsigned int hfi1_max_qps; 427 428 extern unsigned int hfi1_max_sges; 429 430 extern unsigned int hfi1_max_mcast_grps; 431 432 extern unsigned int hfi1_max_mcast_qp_attached; 433 434 extern unsigned int hfi1_max_srqs; 435 436 extern unsigned int hfi1_max_srq_sges; 437 438 extern unsigned int hfi1_max_srq_wrs; 439 440 extern unsigned short piothreshold; 441 442 extern const u32 ib_hfi1_rnr_table[]; 443 444 #endif /* HFI1_VERBS_H */ 445