1 /* 2 * Copyright(c) 2015, 2016 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 u8 r_adefered; /* number of acks defered */ 129 struct iowait s_iowait; 130 struct rvt_qp *owner; 131 }; 132 133 /* 134 * This structure is used to hold commonly lookedup and computed values during 135 * the send engine progress. 136 */ 137 struct hfi1_pkt_state { 138 struct hfi1_ibdev *dev; 139 struct hfi1_ibport *ibp; 140 struct hfi1_pportdata *ppd; 141 struct verbs_txreq *s_txreq; 142 unsigned long flags; 143 }; 144 145 #define HFI1_PSN_CREDIT 16 146 147 struct hfi1_opcode_stats { 148 u64 n_packets; /* number of packets */ 149 u64 n_bytes; /* total number of bytes */ 150 }; 151 152 struct hfi1_opcode_stats_perctx { 153 struct hfi1_opcode_stats stats[256]; 154 }; 155 156 static inline void inc_opstats( 157 u32 tlen, 158 struct hfi1_opcode_stats *stats) 159 { 160 #ifdef CONFIG_DEBUG_FS 161 stats->n_bytes += tlen; 162 stats->n_packets++; 163 #endif 164 } 165 166 struct hfi1_ibport { 167 struct rvt_qp __rcu *qp[2]; 168 struct rvt_ibport rvp; 169 170 /* the first 16 entries are sl_to_vl for !OPA */ 171 u8 sl_to_sc[32]; 172 u8 sc_to_sl[32]; 173 }; 174 175 struct hfi1_ibdev { 176 struct rvt_dev_info rdi; /* Must be first */ 177 178 /* QP numbers are shared by all IB ports */ 179 /* protect txwait list */ 180 seqlock_t txwait_lock ____cacheline_aligned_in_smp; 181 struct list_head txwait; /* list for wait verbs_txreq */ 182 struct list_head memwait; /* list for wait kernel memory */ 183 struct kmem_cache *verbs_txreq_cache; 184 u64 n_txwait; 185 u64 n_kmem_wait; 186 187 /* protect iowait lists */ 188 seqlock_t iowait_lock ____cacheline_aligned_in_smp; 189 u64 n_piowait; 190 u64 n_piodrain; 191 struct timer_list mem_timer; 192 193 #ifdef CONFIG_DEBUG_FS 194 /* per HFI debugfs */ 195 struct dentry *hfi1_ibdev_dbg; 196 /* per HFI symlinks to above */ 197 struct dentry *hfi1_ibdev_link; 198 #endif 199 }; 200 201 static inline struct hfi1_ibdev *to_idev(struct ib_device *ibdev) 202 { 203 struct rvt_dev_info *rdi; 204 205 rdi = container_of(ibdev, struct rvt_dev_info, ibdev); 206 return container_of(rdi, struct hfi1_ibdev, rdi); 207 } 208 209 static inline struct rvt_qp *iowait_to_qp(struct iowait *s_iowait) 210 { 211 struct hfi1_qp_priv *priv; 212 213 priv = container_of(s_iowait, struct hfi1_qp_priv, s_iowait); 214 return priv->owner; 215 } 216 217 /* 218 * Send if not busy or waiting for I/O and either 219 * a RC response is pending or we can process send work requests. 220 */ 221 static inline int hfi1_send_ok(struct rvt_qp *qp) 222 { 223 return !(qp->s_flags & (RVT_S_BUSY | RVT_S_ANY_WAIT_IO)) && 224 (qp->s_hdrwords || (qp->s_flags & RVT_S_RESP_PENDING) || 225 !(qp->s_flags & RVT_S_ANY_WAIT_SEND)); 226 } 227 228 /* 229 * This must be called with s_lock held. 230 */ 231 void hfi1_bad_pqkey(struct hfi1_ibport *ibp, __be16 trap_num, u32 key, u32 sl, 232 u32 qp1, u32 qp2, u16 lid1, u16 lid2); 233 void hfi1_cap_mask_chg(struct rvt_dev_info *rdi, u8 port_num); 234 void hfi1_sys_guid_chg(struct hfi1_ibport *ibp); 235 void hfi1_node_desc_chg(struct hfi1_ibport *ibp); 236 int hfi1_process_mad(struct ib_device *ibdev, int mad_flags, u8 port, 237 const struct ib_wc *in_wc, const struct ib_grh *in_grh, 238 const struct ib_mad_hdr *in_mad, size_t in_mad_size, 239 struct ib_mad_hdr *out_mad, size_t *out_mad_size, 240 u16 *out_mad_pkey_index); 241 242 /* 243 * The PSN_MASK and PSN_SHIFT allow for 244 * 1) comparing two PSNs 245 * 2) returning the PSN with any upper bits masked 246 * 3) returning the difference between to PSNs 247 * 248 * The number of significant bits in the PSN must 249 * necessarily be at least one bit less than 250 * the container holding the PSN. 251 */ 252 #ifndef CONFIG_HFI1_VERBS_31BIT_PSN 253 #define PSN_MASK 0xFFFFFF 254 #define PSN_SHIFT 8 255 #else 256 #define PSN_MASK 0x7FFFFFFF 257 #define PSN_SHIFT 1 258 #endif 259 #define PSN_MODIFY_MASK 0xFFFFFF 260 261 /* 262 * Compare two PSNs 263 * Returns an integer <, ==, or > than zero. 264 */ 265 static inline int cmp_psn(u32 a, u32 b) 266 { 267 return (((int)a) - ((int)b)) << PSN_SHIFT; 268 } 269 270 /* 271 * Return masked PSN 272 */ 273 static inline u32 mask_psn(u32 a) 274 { 275 return a & PSN_MASK; 276 } 277 278 /* 279 * Return delta between two PSNs 280 */ 281 static inline u32 delta_psn(u32 a, u32 b) 282 { 283 return (((int)a - (int)b) << PSN_SHIFT) >> PSN_SHIFT; 284 } 285 286 struct verbs_txreq; 287 void hfi1_put_txreq(struct verbs_txreq *tx); 288 289 int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 290 291 void hfi1_copy_sge(struct rvt_sge_state *ss, void *data, u32 length, 292 bool release, bool copy_last); 293 294 void hfi1_cnp_rcv(struct hfi1_packet *packet); 295 296 void hfi1_uc_rcv(struct hfi1_packet *packet); 297 298 void hfi1_rc_rcv(struct hfi1_packet *packet); 299 300 void hfi1_rc_hdrerr( 301 struct hfi1_ctxtdata *rcd, 302 struct ib_header *hdr, 303 u32 rcv_flags, 304 struct rvt_qp *qp); 305 306 u8 ah_to_sc(struct ib_device *ibdev, struct ib_ah_attr *ah_attr); 307 308 struct ib_ah *hfi1_create_qp0_ah(struct hfi1_ibport *ibp, u16 dlid); 309 310 void hfi1_rc_send_complete(struct rvt_qp *qp, struct ib_header *hdr); 311 312 void hfi1_ud_rcv(struct hfi1_packet *packet); 313 314 int hfi1_lookup_pkey_idx(struct hfi1_ibport *ibp, u16 pkey); 315 316 int hfi1_rvt_get_rwqe(struct rvt_qp *qp, int wr_id_only); 317 318 void hfi1_migrate_qp(struct rvt_qp *qp); 319 320 int hfi1_check_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr, 321 int attr_mask, struct ib_udata *udata); 322 323 void hfi1_modify_qp(struct rvt_qp *qp, struct ib_qp_attr *attr, 324 int attr_mask, struct ib_udata *udata); 325 void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait); 326 int hfi1_check_send_wqe(struct rvt_qp *qp, struct rvt_swqe *wqe); 327 328 extern const u32 rc_only_opcode; 329 extern const u32 uc_only_opcode; 330 331 static inline u8 get_opcode(struct ib_header *h) 332 { 333 u16 lnh = be16_to_cpu(h->lrh[0]) & 3; 334 335 if (lnh == IB_LNH_IBA_LOCAL) 336 return be32_to_cpu(h->u.oth.bth[0]) >> 24; 337 else 338 return be32_to_cpu(h->u.l.oth.bth[0]) >> 24; 339 } 340 341 int hfi1_ruc_check_hdr(struct hfi1_ibport *ibp, struct ib_header *hdr, 342 int has_grh, struct rvt_qp *qp, u32 bth0); 343 344 u32 hfi1_make_grh(struct hfi1_ibport *ibp, struct ib_grh *hdr, 345 struct ib_global_route *grh, u32 hwords, u32 nwords); 346 347 void hfi1_make_ruc_header(struct rvt_qp *qp, struct ib_other_headers *ohdr, 348 u32 bth0, u32 bth2, int middle, 349 struct hfi1_pkt_state *ps); 350 351 void _hfi1_do_send(struct work_struct *work); 352 353 void hfi1_do_send(struct rvt_qp *qp); 354 355 void hfi1_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, 356 enum ib_wc_status status); 357 358 void hfi1_send_rc_ack(struct hfi1_ctxtdata *, struct rvt_qp *qp, int is_fecn); 359 360 int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 361 362 int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 363 364 int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps); 365 366 int hfi1_register_ib_device(struct hfi1_devdata *); 367 368 void hfi1_unregister_ib_device(struct hfi1_devdata *); 369 370 void hfi1_ib_rcv(struct hfi1_packet *packet); 371 372 unsigned hfi1_get_npkeys(struct hfi1_devdata *); 373 374 int hfi1_verbs_send_dma(struct rvt_qp *qp, struct hfi1_pkt_state *ps, 375 u64 pbc); 376 377 int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps, 378 u64 pbc); 379 380 int hfi1_wss_init(void); 381 void hfi1_wss_exit(void); 382 383 /* platform specific: return the lowest level cache (llc) size, in KiB */ 384 static inline int wss_llc_size(void) 385 { 386 /* assume that the boot CPU value is universal for all CPUs */ 387 return boot_cpu_data.x86_cache_size; 388 } 389 390 /* platform specific: cacheless copy */ 391 static inline void cacheless_memcpy(void *dst, void *src, size_t n) 392 { 393 /* 394 * Use the only available X64 cacheless copy. Add a __user cast 395 * to quiet sparse. The src agument is already in the kernel so 396 * there are no security issues. The extra fault recovery machinery 397 * is not invoked. 398 */ 399 __copy_user_nocache(dst, (void __user *)src, n, 0); 400 } 401 402 extern const enum ib_wc_opcode ib_hfi1_wc_opcode[]; 403 404 extern const u8 hdr_len_by_opcode[]; 405 406 extern const int ib_rvt_state_ops[]; 407 408 extern __be64 ib_hfi1_sys_image_guid; /* in network order */ 409 410 extern unsigned int hfi1_max_cqes; 411 412 extern unsigned int hfi1_max_cqs; 413 414 extern unsigned int hfi1_max_qp_wrs; 415 416 extern unsigned int hfi1_max_qps; 417 418 extern unsigned int hfi1_max_sges; 419 420 extern unsigned int hfi1_max_mcast_grps; 421 422 extern unsigned int hfi1_max_mcast_qp_attached; 423 424 extern unsigned int hfi1_max_srqs; 425 426 extern unsigned int hfi1_max_srq_sges; 427 428 extern unsigned int hfi1_max_srq_wrs; 429 430 extern unsigned short piothreshold; 431 432 extern const u32 ib_hfi1_rnr_table[]; 433 434 #endif /* HFI1_VERBS_H */ 435