1 /* 2 * Copyright (c) 2009-2010 Chelsio, Inc. All rights reserved. 3 * 4 * This software is available to you under a choice of one of two 5 * licenses. You may choose to be licensed under the terms of the GNU 6 * General Public License (GPL) Version 2, available from the file 7 * COPYING in the main directory of this source tree, or the 8 * OpenIB.org BSD license below: 9 * 10 * Redistribution and use in source and binary forms, with or 11 * without modification, are permitted provided that the following 12 * conditions are met: 13 * 14 * - Redistributions of source code must retain the above 15 * copyright notice, this list of conditions and the following 16 * disclaimer. 17 * - Redistributions in binary form must reproduce the above 18 * copyright notice, this list of conditions and the following 19 * disclaimer in the documentation and/or other materials 20 * provided with the distribution. 21 * 22 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 23 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 24 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 25 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 26 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 27 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 28 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 29 * SOFTWARE. 30 */ 31 #ifndef __T4_H__ 32 #define __T4_H__ 33 34 #include "t4_hw.h" 35 #include "t4_regs.h" 36 #include "t4_values.h" 37 #include "t4_msg.h" 38 #include "t4fw_ri_api.h" 39 40 #define T4_MAX_NUM_PD 65536 41 #define T4_MAX_MR_SIZE (~0ULL) 42 #define T4_PAGESIZE_MASK 0xffff000 /* 4KB-128MB */ 43 #define T4_STAG_UNSET 0xffffffff 44 #define T4_FW_MAJ 0 45 #define PCIE_MA_SYNC_A 0x30b4 46 47 struct t4_status_page { 48 __be32 rsvd1; /* flit 0 - hw owns */ 49 __be16 rsvd2; 50 __be16 qid; 51 __be16 cidx; 52 __be16 pidx; 53 u8 qp_err; /* flit 1 - sw owns */ 54 u8 db_off; 55 u8 pad; 56 u16 host_wq_pidx; 57 u16 host_cidx; 58 u16 host_pidx; 59 }; 60 61 #define T4_EQ_ENTRY_SIZE 64 62 63 #define T4_SQ_NUM_SLOTS 5 64 #define T4_SQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_SQ_NUM_SLOTS) 65 #define T4_MAX_SEND_SGE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \ 66 sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge)) 67 #define T4_MAX_SEND_INLINE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \ 68 sizeof(struct fw_ri_immd))) 69 #define T4_MAX_WRITE_INLINE ((T4_SQ_NUM_BYTES - \ 70 sizeof(struct fw_ri_rdma_write_wr) - \ 71 sizeof(struct fw_ri_immd))) 72 #define T4_MAX_WRITE_SGE ((T4_SQ_NUM_BYTES - \ 73 sizeof(struct fw_ri_rdma_write_wr) - \ 74 sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge)) 75 #define T4_MAX_FR_IMMD ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_fr_nsmr_wr) - \ 76 sizeof(struct fw_ri_immd)) & ~31UL) 77 #define T4_MAX_FR_IMMD_DEPTH (T4_MAX_FR_IMMD / sizeof(u64)) 78 #define T4_MAX_FR_DSGL 1024 79 #define T4_MAX_FR_DSGL_DEPTH (T4_MAX_FR_DSGL / sizeof(u64)) 80 81 static inline int t4_max_fr_depth(int use_dsgl) 82 { 83 return use_dsgl ? T4_MAX_FR_DSGL_DEPTH : T4_MAX_FR_IMMD_DEPTH; 84 } 85 86 #define T4_RQ_NUM_SLOTS 2 87 #define T4_RQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_RQ_NUM_SLOTS) 88 #define T4_MAX_RECV_SGE 4 89 90 union t4_wr { 91 struct fw_ri_res_wr res; 92 struct fw_ri_wr ri; 93 struct fw_ri_rdma_write_wr write; 94 struct fw_ri_send_wr send; 95 struct fw_ri_rdma_read_wr read; 96 struct fw_ri_bind_mw_wr bind; 97 struct fw_ri_fr_nsmr_wr fr; 98 struct fw_ri_fr_nsmr_tpte_wr fr_tpte; 99 struct fw_ri_inv_lstag_wr inv; 100 struct t4_status_page status; 101 __be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_SQ_NUM_SLOTS]; 102 }; 103 104 union t4_recv_wr { 105 struct fw_ri_recv_wr recv; 106 struct t4_status_page status; 107 __be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_RQ_NUM_SLOTS]; 108 }; 109 110 static inline void init_wr_hdr(union t4_wr *wqe, u16 wrid, 111 enum fw_wr_opcodes opcode, u8 flags, u8 len16) 112 { 113 wqe->send.opcode = (u8)opcode; 114 wqe->send.flags = flags; 115 wqe->send.wrid = wrid; 116 wqe->send.r1[0] = 0; 117 wqe->send.r1[1] = 0; 118 wqe->send.r1[2] = 0; 119 wqe->send.len16 = len16; 120 } 121 122 /* CQE/AE status codes */ 123 #define T4_ERR_SUCCESS 0x0 124 #define T4_ERR_STAG 0x1 /* STAG invalid: either the */ 125 /* STAG is offlimt, being 0, */ 126 /* or STAG_key mismatch */ 127 #define T4_ERR_PDID 0x2 /* PDID mismatch */ 128 #define T4_ERR_QPID 0x3 /* QPID mismatch */ 129 #define T4_ERR_ACCESS 0x4 /* Invalid access right */ 130 #define T4_ERR_WRAP 0x5 /* Wrap error */ 131 #define T4_ERR_BOUND 0x6 /* base and bounds voilation */ 132 #define T4_ERR_INVALIDATE_SHARED_MR 0x7 /* attempt to invalidate a */ 133 /* shared memory region */ 134 #define T4_ERR_INVALIDATE_MR_WITH_MW_BOUND 0x8 /* attempt to invalidate a */ 135 /* shared memory region */ 136 #define T4_ERR_ECC 0x9 /* ECC error detected */ 137 #define T4_ERR_ECC_PSTAG 0xA /* ECC error detected when */ 138 /* reading PSTAG for a MW */ 139 /* Invalidate */ 140 #define T4_ERR_PBL_ADDR_BOUND 0xB /* pbl addr out of bounds: */ 141 /* software error */ 142 #define T4_ERR_SWFLUSH 0xC /* SW FLUSHED */ 143 #define T4_ERR_CRC 0x10 /* CRC error */ 144 #define T4_ERR_MARKER 0x11 /* Marker error */ 145 #define T4_ERR_PDU_LEN_ERR 0x12 /* invalid PDU length */ 146 #define T4_ERR_OUT_OF_RQE 0x13 /* out of RQE */ 147 #define T4_ERR_DDP_VERSION 0x14 /* wrong DDP version */ 148 #define T4_ERR_RDMA_VERSION 0x15 /* wrong RDMA version */ 149 #define T4_ERR_OPCODE 0x16 /* invalid rdma opcode */ 150 #define T4_ERR_DDP_QUEUE_NUM 0x17 /* invalid ddp queue number */ 151 #define T4_ERR_MSN 0x18 /* MSN error */ 152 #define T4_ERR_TBIT 0x19 /* tag bit not set correctly */ 153 #define T4_ERR_MO 0x1A /* MO not 0 for TERMINATE */ 154 /* or READ_REQ */ 155 #define T4_ERR_MSN_GAP 0x1B 156 #define T4_ERR_MSN_RANGE 0x1C 157 #define T4_ERR_IRD_OVERFLOW 0x1D 158 #define T4_ERR_RQE_ADDR_BOUND 0x1E /* RQE addr out of bounds: */ 159 /* software error */ 160 #define T4_ERR_INTERNAL_ERR 0x1F /* internal error (opcode */ 161 /* mismatch) */ 162 /* 163 * CQE defs 164 */ 165 struct t4_cqe { 166 __be32 header; 167 __be32 len; 168 union { 169 struct { 170 __be32 stag; 171 __be32 msn; 172 } rcqe; 173 struct { 174 u32 stag; 175 u16 nada2; 176 u16 cidx; 177 } scqe; 178 struct { 179 __be32 wrid_hi; 180 __be32 wrid_low; 181 } gen; 182 u64 drain_cookie; 183 } u; 184 __be64 reserved; 185 __be64 bits_type_ts; 186 }; 187 188 /* macros for flit 0 of the cqe */ 189 190 #define CQE_QPID_S 12 191 #define CQE_QPID_M 0xFFFFF 192 #define CQE_QPID_G(x) ((((x) >> CQE_QPID_S)) & CQE_QPID_M) 193 #define CQE_QPID_V(x) ((x)<<CQE_QPID_S) 194 195 #define CQE_SWCQE_S 11 196 #define CQE_SWCQE_M 0x1 197 #define CQE_SWCQE_G(x) ((((x) >> CQE_SWCQE_S)) & CQE_SWCQE_M) 198 #define CQE_SWCQE_V(x) ((x)<<CQE_SWCQE_S) 199 200 #define CQE_STATUS_S 5 201 #define CQE_STATUS_M 0x1F 202 #define CQE_STATUS_G(x) ((((x) >> CQE_STATUS_S)) & CQE_STATUS_M) 203 #define CQE_STATUS_V(x) ((x)<<CQE_STATUS_S) 204 205 #define CQE_TYPE_S 4 206 #define CQE_TYPE_M 0x1 207 #define CQE_TYPE_G(x) ((((x) >> CQE_TYPE_S)) & CQE_TYPE_M) 208 #define CQE_TYPE_V(x) ((x)<<CQE_TYPE_S) 209 210 #define CQE_OPCODE_S 0 211 #define CQE_OPCODE_M 0xF 212 #define CQE_OPCODE_G(x) ((((x) >> CQE_OPCODE_S)) & CQE_OPCODE_M) 213 #define CQE_OPCODE_V(x) ((x)<<CQE_OPCODE_S) 214 215 #define SW_CQE(x) (CQE_SWCQE_G(be32_to_cpu((x)->header))) 216 #define CQE_QPID(x) (CQE_QPID_G(be32_to_cpu((x)->header))) 217 #define CQE_TYPE(x) (CQE_TYPE_G(be32_to_cpu((x)->header))) 218 #define SQ_TYPE(x) (CQE_TYPE((x))) 219 #define RQ_TYPE(x) (!CQE_TYPE((x))) 220 #define CQE_STATUS(x) (CQE_STATUS_G(be32_to_cpu((x)->header))) 221 #define CQE_OPCODE(x) (CQE_OPCODE_G(be32_to_cpu((x)->header))) 222 223 #define CQE_SEND_OPCODE(x)( \ 224 (CQE_OPCODE_G(be32_to_cpu((x)->header)) == FW_RI_SEND) || \ 225 (CQE_OPCODE_G(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE) || \ 226 (CQE_OPCODE_G(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_INV) || \ 227 (CQE_OPCODE_G(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE_INV)) 228 229 #define CQE_LEN(x) (be32_to_cpu((x)->len)) 230 231 /* used for RQ completion processing */ 232 #define CQE_WRID_STAG(x) (be32_to_cpu((x)->u.rcqe.stag)) 233 #define CQE_WRID_MSN(x) (be32_to_cpu((x)->u.rcqe.msn)) 234 235 /* used for SQ completion processing */ 236 #define CQE_WRID_SQ_IDX(x) ((x)->u.scqe.cidx) 237 #define CQE_WRID_FR_STAG(x) (be32_to_cpu((x)->u.scqe.stag)) 238 239 /* generic accessor macros */ 240 #define CQE_WRID_HI(x) (be32_to_cpu((x)->u.gen.wrid_hi)) 241 #define CQE_WRID_LOW(x) (be32_to_cpu((x)->u.gen.wrid_low)) 242 #define CQE_DRAIN_COOKIE(x) ((x)->u.drain_cookie) 243 244 /* macros for flit 3 of the cqe */ 245 #define CQE_GENBIT_S 63 246 #define CQE_GENBIT_M 0x1 247 #define CQE_GENBIT_G(x) (((x) >> CQE_GENBIT_S) & CQE_GENBIT_M) 248 #define CQE_GENBIT_V(x) ((x)<<CQE_GENBIT_S) 249 250 #define CQE_OVFBIT_S 62 251 #define CQE_OVFBIT_M 0x1 252 #define CQE_OVFBIT_G(x) ((((x) >> CQE_OVFBIT_S)) & CQE_OVFBIT_M) 253 254 #define CQE_IQTYPE_S 60 255 #define CQE_IQTYPE_M 0x3 256 #define CQE_IQTYPE_G(x) ((((x) >> CQE_IQTYPE_S)) & CQE_IQTYPE_M) 257 258 #define CQE_TS_M 0x0fffffffffffffffULL 259 #define CQE_TS_G(x) ((x) & CQE_TS_M) 260 261 #define CQE_OVFBIT(x) ((unsigned)CQE_OVFBIT_G(be64_to_cpu((x)->bits_type_ts))) 262 #define CQE_GENBIT(x) ((unsigned)CQE_GENBIT_G(be64_to_cpu((x)->bits_type_ts))) 263 #define CQE_TS(x) (CQE_TS_G(be64_to_cpu((x)->bits_type_ts))) 264 265 struct t4_swsqe { 266 u64 wr_id; 267 struct t4_cqe cqe; 268 int read_len; 269 int opcode; 270 int complete; 271 int signaled; 272 u16 idx; 273 int flushed; 274 struct timespec host_ts; 275 u64 sge_ts; 276 }; 277 278 static inline pgprot_t t4_pgprot_wc(pgprot_t prot) 279 { 280 #if defined(__i386__) || defined(__x86_64__) || defined(CONFIG_PPC64) 281 return pgprot_writecombine(prot); 282 #else 283 return pgprot_noncached(prot); 284 #endif 285 } 286 287 enum { 288 T4_SQ_ONCHIP = (1<<0), 289 }; 290 291 struct t4_sq { 292 union t4_wr *queue; 293 dma_addr_t dma_addr; 294 DEFINE_DMA_UNMAP_ADDR(mapping); 295 unsigned long phys_addr; 296 struct t4_swsqe *sw_sq; 297 struct t4_swsqe *oldest_read; 298 void __iomem *bar2_va; 299 u64 bar2_pa; 300 size_t memsize; 301 u32 bar2_qid; 302 u32 qid; 303 u16 in_use; 304 u16 size; 305 u16 cidx; 306 u16 pidx; 307 u16 wq_pidx; 308 u16 wq_pidx_inc; 309 u16 flags; 310 short flush_cidx; 311 }; 312 313 struct t4_swrqe { 314 u64 wr_id; 315 struct timespec host_ts; 316 u64 sge_ts; 317 }; 318 319 struct t4_rq { 320 union t4_recv_wr *queue; 321 dma_addr_t dma_addr; 322 DEFINE_DMA_UNMAP_ADDR(mapping); 323 struct t4_swrqe *sw_rq; 324 void __iomem *bar2_va; 325 u64 bar2_pa; 326 size_t memsize; 327 u32 bar2_qid; 328 u32 qid; 329 u32 msn; 330 u32 rqt_hwaddr; 331 u16 rqt_size; 332 u16 in_use; 333 u16 size; 334 u16 cidx; 335 u16 pidx; 336 u16 wq_pidx; 337 u16 wq_pidx_inc; 338 }; 339 340 struct t4_wq { 341 struct t4_sq sq; 342 struct t4_rq rq; 343 void __iomem *db; 344 struct c4iw_rdev *rdev; 345 int flushed; 346 }; 347 348 static inline int t4_rqes_posted(struct t4_wq *wq) 349 { 350 return wq->rq.in_use; 351 } 352 353 static inline int t4_rq_empty(struct t4_wq *wq) 354 { 355 return wq->rq.in_use == 0; 356 } 357 358 static inline int t4_rq_full(struct t4_wq *wq) 359 { 360 return wq->rq.in_use == (wq->rq.size - 1); 361 } 362 363 static inline u32 t4_rq_avail(struct t4_wq *wq) 364 { 365 return wq->rq.size - 1 - wq->rq.in_use; 366 } 367 368 static inline void t4_rq_produce(struct t4_wq *wq, u8 len16) 369 { 370 wq->rq.in_use++; 371 if (++wq->rq.pidx == wq->rq.size) 372 wq->rq.pidx = 0; 373 wq->rq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE); 374 if (wq->rq.wq_pidx >= wq->rq.size * T4_RQ_NUM_SLOTS) 375 wq->rq.wq_pidx %= wq->rq.size * T4_RQ_NUM_SLOTS; 376 } 377 378 static inline void t4_rq_consume(struct t4_wq *wq) 379 { 380 wq->rq.in_use--; 381 wq->rq.msn++; 382 if (++wq->rq.cidx == wq->rq.size) 383 wq->rq.cidx = 0; 384 } 385 386 static inline u16 t4_rq_host_wq_pidx(struct t4_wq *wq) 387 { 388 return wq->rq.queue[wq->rq.size].status.host_wq_pidx; 389 } 390 391 static inline u16 t4_rq_wq_size(struct t4_wq *wq) 392 { 393 return wq->rq.size * T4_RQ_NUM_SLOTS; 394 } 395 396 static inline int t4_sq_onchip(struct t4_sq *sq) 397 { 398 return sq->flags & T4_SQ_ONCHIP; 399 } 400 401 static inline int t4_sq_empty(struct t4_wq *wq) 402 { 403 return wq->sq.in_use == 0; 404 } 405 406 static inline int t4_sq_full(struct t4_wq *wq) 407 { 408 return wq->sq.in_use == (wq->sq.size - 1); 409 } 410 411 static inline u32 t4_sq_avail(struct t4_wq *wq) 412 { 413 return wq->sq.size - 1 - wq->sq.in_use; 414 } 415 416 static inline void t4_sq_produce(struct t4_wq *wq, u8 len16) 417 { 418 wq->sq.in_use++; 419 if (++wq->sq.pidx == wq->sq.size) 420 wq->sq.pidx = 0; 421 wq->sq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE); 422 if (wq->sq.wq_pidx >= wq->sq.size * T4_SQ_NUM_SLOTS) 423 wq->sq.wq_pidx %= wq->sq.size * T4_SQ_NUM_SLOTS; 424 } 425 426 static inline void t4_sq_consume(struct t4_wq *wq) 427 { 428 BUG_ON(wq->sq.in_use < 1); 429 if (wq->sq.cidx == wq->sq.flush_cidx) 430 wq->sq.flush_cidx = -1; 431 wq->sq.in_use--; 432 if (++wq->sq.cidx == wq->sq.size) 433 wq->sq.cidx = 0; 434 } 435 436 static inline u16 t4_sq_host_wq_pidx(struct t4_wq *wq) 437 { 438 return wq->sq.queue[wq->sq.size].status.host_wq_pidx; 439 } 440 441 static inline u16 t4_sq_wq_size(struct t4_wq *wq) 442 { 443 return wq->sq.size * T4_SQ_NUM_SLOTS; 444 } 445 446 /* This function copies 64 byte coalesced work request to memory 447 * mapped BAR2 space. For coalesced WRs, the SGE fetches data 448 * from the FIFO instead of from Host. 449 */ 450 static inline void pio_copy(u64 __iomem *dst, u64 *src) 451 { 452 int count = 8; 453 454 while (count) { 455 writeq(*src, dst); 456 src++; 457 dst++; 458 count--; 459 } 460 } 461 462 static inline void t4_ring_sq_db(struct t4_wq *wq, u16 inc, union t4_wr *wqe) 463 { 464 465 /* Flush host queue memory writes. */ 466 wmb(); 467 if (wq->sq.bar2_va) { 468 if (inc == 1 && wq->sq.bar2_qid == 0 && wqe) { 469 pr_debug("%s: WC wq->sq.pidx = %d\n", 470 __func__, wq->sq.pidx); 471 pio_copy((u64 __iomem *) 472 (wq->sq.bar2_va + SGE_UDB_WCDOORBELL), 473 (u64 *)wqe); 474 } else { 475 pr_debug("%s: DB wq->sq.pidx = %d\n", 476 __func__, wq->sq.pidx); 477 writel(PIDX_T5_V(inc) | QID_V(wq->sq.bar2_qid), 478 wq->sq.bar2_va + SGE_UDB_KDOORBELL); 479 } 480 481 /* Flush user doorbell area writes. */ 482 wmb(); 483 return; 484 } 485 writel(QID_V(wq->sq.qid) | PIDX_V(inc), wq->db); 486 } 487 488 static inline void t4_ring_rq_db(struct t4_wq *wq, u16 inc, 489 union t4_recv_wr *wqe) 490 { 491 492 /* Flush host queue memory writes. */ 493 wmb(); 494 if (wq->rq.bar2_va) { 495 if (inc == 1 && wq->rq.bar2_qid == 0 && wqe) { 496 pr_debug("%s: WC wq->rq.pidx = %d\n", 497 __func__, wq->rq.pidx); 498 pio_copy((u64 __iomem *) 499 (wq->rq.bar2_va + SGE_UDB_WCDOORBELL), 500 (void *)wqe); 501 } else { 502 pr_debug("%s: DB wq->rq.pidx = %d\n", 503 __func__, wq->rq.pidx); 504 writel(PIDX_T5_V(inc) | QID_V(wq->rq.bar2_qid), 505 wq->rq.bar2_va + SGE_UDB_KDOORBELL); 506 } 507 508 /* Flush user doorbell area writes. */ 509 wmb(); 510 return; 511 } 512 writel(QID_V(wq->rq.qid) | PIDX_V(inc), wq->db); 513 } 514 515 static inline int t4_wq_in_error(struct t4_wq *wq) 516 { 517 return wq->rq.queue[wq->rq.size].status.qp_err; 518 } 519 520 static inline void t4_set_wq_in_error(struct t4_wq *wq) 521 { 522 wq->rq.queue[wq->rq.size].status.qp_err = 1; 523 } 524 525 static inline void t4_disable_wq_db(struct t4_wq *wq) 526 { 527 wq->rq.queue[wq->rq.size].status.db_off = 1; 528 } 529 530 static inline void t4_enable_wq_db(struct t4_wq *wq) 531 { 532 wq->rq.queue[wq->rq.size].status.db_off = 0; 533 } 534 535 static inline int t4_wq_db_enabled(struct t4_wq *wq) 536 { 537 return !wq->rq.queue[wq->rq.size].status.db_off; 538 } 539 540 enum t4_cq_flags { 541 CQ_ARMED = 1, 542 }; 543 544 struct t4_cq { 545 struct t4_cqe *queue; 546 dma_addr_t dma_addr; 547 DEFINE_DMA_UNMAP_ADDR(mapping); 548 struct t4_cqe *sw_queue; 549 void __iomem *gts; 550 void __iomem *bar2_va; 551 u64 bar2_pa; 552 u32 bar2_qid; 553 struct c4iw_rdev *rdev; 554 size_t memsize; 555 __be64 bits_type_ts; 556 u32 cqid; 557 u32 qid_mask; 558 int vector; 559 u16 size; /* including status page */ 560 u16 cidx; 561 u16 sw_pidx; 562 u16 sw_cidx; 563 u16 sw_in_use; 564 u16 cidx_inc; 565 u8 gen; 566 u8 error; 567 unsigned long flags; 568 }; 569 570 static inline void write_gts(struct t4_cq *cq, u32 val) 571 { 572 if (cq->bar2_va) 573 writel(val | INGRESSQID_V(cq->bar2_qid), 574 cq->bar2_va + SGE_UDB_GTS); 575 else 576 writel(val | INGRESSQID_V(cq->cqid), cq->gts); 577 } 578 579 static inline int t4_clear_cq_armed(struct t4_cq *cq) 580 { 581 return test_and_clear_bit(CQ_ARMED, &cq->flags); 582 } 583 584 static inline int t4_arm_cq(struct t4_cq *cq, int se) 585 { 586 u32 val; 587 588 set_bit(CQ_ARMED, &cq->flags); 589 while (cq->cidx_inc > CIDXINC_M) { 590 val = SEINTARM_V(0) | CIDXINC_V(CIDXINC_M) | TIMERREG_V(7); 591 write_gts(cq, val); 592 cq->cidx_inc -= CIDXINC_M; 593 } 594 val = SEINTARM_V(se) | CIDXINC_V(cq->cidx_inc) | TIMERREG_V(6); 595 write_gts(cq, val); 596 cq->cidx_inc = 0; 597 return 0; 598 } 599 600 static inline void t4_swcq_produce(struct t4_cq *cq) 601 { 602 cq->sw_in_use++; 603 if (cq->sw_in_use == cq->size) { 604 pr_debug("%s cxgb4 sw cq overflow cqid %u\n", 605 __func__, cq->cqid); 606 cq->error = 1; 607 BUG_ON(1); 608 } 609 if (++cq->sw_pidx == cq->size) 610 cq->sw_pidx = 0; 611 } 612 613 static inline void t4_swcq_consume(struct t4_cq *cq) 614 { 615 BUG_ON(cq->sw_in_use < 1); 616 cq->sw_in_use--; 617 if (++cq->sw_cidx == cq->size) 618 cq->sw_cidx = 0; 619 } 620 621 static inline void t4_hwcq_consume(struct t4_cq *cq) 622 { 623 cq->bits_type_ts = cq->queue[cq->cidx].bits_type_ts; 624 if (++cq->cidx_inc == (cq->size >> 4) || cq->cidx_inc == CIDXINC_M) { 625 u32 val; 626 627 val = SEINTARM_V(0) | CIDXINC_V(cq->cidx_inc) | TIMERREG_V(7); 628 write_gts(cq, val); 629 cq->cidx_inc = 0; 630 } 631 if (++cq->cidx == cq->size) { 632 cq->cidx = 0; 633 cq->gen ^= 1; 634 } 635 } 636 637 static inline int t4_valid_cqe(struct t4_cq *cq, struct t4_cqe *cqe) 638 { 639 return (CQE_GENBIT(cqe) == cq->gen); 640 } 641 642 static inline int t4_cq_notempty(struct t4_cq *cq) 643 { 644 return cq->sw_in_use || t4_valid_cqe(cq, &cq->queue[cq->cidx]); 645 } 646 647 static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe) 648 { 649 int ret; 650 u16 prev_cidx; 651 652 if (cq->cidx == 0) 653 prev_cidx = cq->size - 1; 654 else 655 prev_cidx = cq->cidx - 1; 656 657 if (cq->queue[prev_cidx].bits_type_ts != cq->bits_type_ts) { 658 ret = -EOVERFLOW; 659 cq->error = 1; 660 pr_err("cq overflow cqid %u\n", cq->cqid); 661 BUG_ON(1); 662 } else if (t4_valid_cqe(cq, &cq->queue[cq->cidx])) { 663 664 /* Ensure CQE is flushed to memory */ 665 rmb(); 666 *cqe = &cq->queue[cq->cidx]; 667 ret = 0; 668 } else 669 ret = -ENODATA; 670 return ret; 671 } 672 673 static inline struct t4_cqe *t4_next_sw_cqe(struct t4_cq *cq) 674 { 675 if (cq->sw_in_use == cq->size) { 676 pr_debug("%s cxgb4 sw cq overflow cqid %u\n", 677 __func__, cq->cqid); 678 cq->error = 1; 679 BUG_ON(1); 680 return NULL; 681 } 682 if (cq->sw_in_use) 683 return &cq->sw_queue[cq->sw_cidx]; 684 return NULL; 685 } 686 687 static inline int t4_next_cqe(struct t4_cq *cq, struct t4_cqe **cqe) 688 { 689 int ret = 0; 690 691 if (cq->error) 692 ret = -ENODATA; 693 else if (cq->sw_in_use) 694 *cqe = &cq->sw_queue[cq->sw_cidx]; 695 else 696 ret = t4_next_hw_cqe(cq, cqe); 697 return ret; 698 } 699 700 static inline int t4_cq_in_error(struct t4_cq *cq) 701 { 702 return ((struct t4_status_page *)&cq->queue[cq->size])->qp_err; 703 } 704 705 static inline void t4_set_cq_in_error(struct t4_cq *cq) 706 { 707 ((struct t4_status_page *)&cq->queue[cq->size])->qp_err = 1; 708 } 709 #endif 710 711 struct t4_dev_status_page { 712 u8 db_off; 713 u8 pad1; 714 u16 pad2; 715 u32 pad3; 716 u64 qp_start; 717 u64 qp_size; 718 u64 cq_start; 719 u64 cq_size; 720 }; 721