2 * Copyright (c) 2009-2013 Chelsio, Inc. All rights reserved.
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:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
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.
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
37 * Fixme: Adding missing defines
39 #define SGE_PF_KDOORBELL 0x0
40 #define QID_MASK 0xffff8000U
42 #define QID(x) ((x) << QID_SHIFT)
43 #define DBPRIO 0x00004000U
44 #define PIDX_MASK 0x00003fffU
46 #define PIDX(x) ((x) << PIDX_SHIFT)
48 #define SGE_PF_GTS 0x4
49 #define INGRESSQID_MASK 0xffff0000U
50 #define INGRESSQID_SHIFT 16
51 #define INGRESSQID(x) ((x) << INGRESSQID_SHIFT)
52 #define TIMERREG_MASK 0x0000e000U
53 #define TIMERREG_SHIFT 13
54 #define TIMERREG(x) ((x) << TIMERREG_SHIFT)
55 #define SEINTARM_MASK 0x00001000U
56 #define SEINTARM_SHIFT 12
57 #define SEINTARM(x) ((x) << SEINTARM_SHIFT)
58 #define CIDXINC_MASK 0x00000fffU
59 #define CIDXINC_SHIFT 0
60 #define CIDXINC(x) ((x) << CIDXINC_SHIFT)
62 #define T4_MAX_NUM_PD 65536
63 #define T4_MAX_EQ_SIZE 65520
64 #define T4_MAX_IQ_SIZE 65520
65 #define T4_MAX_RQ_SIZE(n) (8192 - (n) - 1)
66 #define T4_MAX_SQ_SIZE(n) (T4_MAX_EQ_SIZE - (n) - 1)
67 #define T4_MAX_QP_DEPTH(n) (T4_MAX_RQ_SIZE(n))
68 #define T4_MAX_CQ_DEPTH (T4_MAX_IQ_SIZE - 2)
69 #define T4_MAX_MR_SIZE (~0ULL - 1)
70 #define T4_PAGESIZE_MASK 0xffffffff000 /* 4KB-8TB */
71 #define T4_STAG_UNSET 0xffffffff
73 #define A_PCIE_MA_SYNC 0x30b4
75 struct t4_status_page {
76 __be32 rsvd1; /* flit 0 - hw owns */
81 u8 qp_err; /* flit 1 - sw owns */
89 #define T4_EQ_ENTRY_SIZE 64
91 #define T4_SQ_NUM_SLOTS 5
92 #define T4_SQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_SQ_NUM_SLOTS)
93 #define T4_MAX_SEND_SGE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \
94 sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
95 #define T4_MAX_SEND_INLINE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \
96 sizeof(struct fw_ri_immd)))
97 #define T4_MAX_WRITE_INLINE ((T4_SQ_NUM_BYTES - \
98 sizeof(struct fw_ri_rdma_write_wr) - \
99 sizeof(struct fw_ri_immd)))
100 #define T4_MAX_WRITE_SGE ((T4_SQ_NUM_BYTES - \
101 sizeof(struct fw_ri_rdma_write_wr) - \
102 sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
103 #define T4_MAX_FR_IMMD ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_fr_nsmr_wr) - \
104 sizeof(struct fw_ri_immd)) & ~31UL)
105 #define T4_MAX_FR_DEPTH (T4_MAX_FR_IMMD / sizeof(u64))
107 #define T4_RQ_NUM_SLOTS 2
108 #define T4_RQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_RQ_NUM_SLOTS)
109 #define T4_MAX_RECV_SGE 4
112 struct fw_ri_res_wr res;
114 struct fw_ri_rdma_write_wr write;
115 struct fw_ri_send_wr send;
116 struct fw_ri_rdma_read_wr read;
117 struct fw_ri_bind_mw_wr bind;
118 struct fw_ri_fr_nsmr_wr fr;
119 struct fw_ri_inv_lstag_wr inv;
120 struct t4_status_page status;
121 __be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_SQ_NUM_SLOTS];
125 struct fw_ri_recv_wr recv;
126 struct t4_status_page status;
127 __be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_RQ_NUM_SLOTS];
130 static inline void init_wr_hdr(union t4_wr *wqe, u16 wrid,
131 enum fw_wr_opcodes opcode, u8 flags, u8 len16)
133 wqe->send.opcode = (u8)opcode;
134 wqe->send.flags = flags;
135 wqe->send.wrid = wrid;
139 wqe->send.len16 = len16;
142 /* CQE/AE status codes */
143 #define T4_ERR_SUCCESS 0x0
144 #define T4_ERR_STAG 0x1 /* STAG invalid: either the */
145 /* STAG is offlimt, being 0, */
146 /* or STAG_key mismatch */
147 #define T4_ERR_PDID 0x2 /* PDID mismatch */
148 #define T4_ERR_QPID 0x3 /* QPID mismatch */
149 #define T4_ERR_ACCESS 0x4 /* Invalid access right */
150 #define T4_ERR_WRAP 0x5 /* Wrap error */
151 #define T4_ERR_BOUND 0x6 /* base and bounds voilation */
152 #define T4_ERR_INVALIDATE_SHARED_MR 0x7 /* attempt to invalidate a */
153 /* shared memory region */
154 #define T4_ERR_INVALIDATE_MR_WITH_MW_BOUND 0x8 /* attempt to invalidate a */
155 /* shared memory region */
156 #define T4_ERR_ECC 0x9 /* ECC error detected */
157 #define T4_ERR_ECC_PSTAG 0xA /* ECC error detected when */
158 /* reading PSTAG for a MW */
160 #define T4_ERR_PBL_ADDR_BOUND 0xB /* pbl addr out of bounds: */
162 #define T4_ERR_SWFLUSH 0xC /* SW FLUSHED */
163 #define T4_ERR_CRC 0x10 /* CRC error */
164 #define T4_ERR_MARKER 0x11 /* Marker error */
165 #define T4_ERR_PDU_LEN_ERR 0x12 /* invalid PDU length */
166 #define T4_ERR_OUT_OF_RQE 0x13 /* out of RQE */
167 #define T4_ERR_DDP_VERSION 0x14 /* wrong DDP version */
168 #define T4_ERR_RDMA_VERSION 0x15 /* wrong RDMA version */
169 #define T4_ERR_OPCODE 0x16 /* invalid rdma opcode */
170 #define T4_ERR_DDP_QUEUE_NUM 0x17 /* invalid ddp queue number */
171 #define T4_ERR_MSN 0x18 /* MSN error */
172 #define T4_ERR_TBIT 0x19 /* tag bit not set correctly */
173 #define T4_ERR_MO 0x1A /* MO not 0 for TERMINATE */
175 #define T4_ERR_MSN_GAP 0x1B
176 #define T4_ERR_MSN_RANGE 0x1C
177 #define T4_ERR_IRD_OVERFLOW 0x1D
178 #define T4_ERR_RQE_ADDR_BOUND 0x1E /* RQE addr out of bounds: */
180 #define T4_ERR_INTERNAL_ERR 0x1F /* internal error (opcode */
208 /* macros for flit 0 of the cqe */
210 #define S_CQE_QPID 12
211 #define M_CQE_QPID 0xFFFFF
212 #define G_CQE_QPID(x) ((((x) >> S_CQE_QPID)) & M_CQE_QPID)
213 #define V_CQE_QPID(x) ((x)<<S_CQE_QPID)
215 #define S_CQE_SWCQE 11
216 #define M_CQE_SWCQE 0x1
217 #define G_CQE_SWCQE(x) ((((x) >> S_CQE_SWCQE)) & M_CQE_SWCQE)
218 #define V_CQE_SWCQE(x) ((x)<<S_CQE_SWCQE)
220 #define S_CQE_STATUS 5
221 #define M_CQE_STATUS 0x1F
222 #define G_CQE_STATUS(x) ((((x) >> S_CQE_STATUS)) & M_CQE_STATUS)
223 #define V_CQE_STATUS(x) ((x)<<S_CQE_STATUS)
226 #define M_CQE_TYPE 0x1
227 #define G_CQE_TYPE(x) ((((x) >> S_CQE_TYPE)) & M_CQE_TYPE)
228 #define V_CQE_TYPE(x) ((x)<<S_CQE_TYPE)
230 #define S_CQE_OPCODE 0
231 #define M_CQE_OPCODE 0xF
232 #define G_CQE_OPCODE(x) ((((x) >> S_CQE_OPCODE)) & M_CQE_OPCODE)
233 #define V_CQE_OPCODE(x) ((x)<<S_CQE_OPCODE)
235 #define SW_CQE(x) (G_CQE_SWCQE(be32_to_cpu((x)->header)))
236 #define CQE_QPID(x) (G_CQE_QPID(be32_to_cpu((x)->header)))
237 #define CQE_TYPE(x) (G_CQE_TYPE(be32_to_cpu((x)->header)))
238 #define SQ_TYPE(x) (CQE_TYPE((x)))
239 #define RQ_TYPE(x) (!CQE_TYPE((x)))
240 #define CQE_STATUS(x) (G_CQE_STATUS(be32_to_cpu((x)->header)))
241 #define CQE_OPCODE(x) (G_CQE_OPCODE(be32_to_cpu((x)->header)))
243 #define CQE_SEND_OPCODE(x)(\
244 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND) || \
245 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE) || \
246 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_INV) || \
247 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE_INV))
249 #define CQE_LEN(x) (be32_to_cpu((x)->len))
251 /* used for RQ completion processing */
252 #define CQE_WRID_STAG(x) (be32_to_cpu((x)->u.rcqe.stag))
253 #define CQE_WRID_MSN(x) (be32_to_cpu((x)->u.rcqe.msn))
255 /* used for SQ completion processing */
256 #define CQE_WRID_SQ_IDX(x) ((x)->u.scqe.cidx)
258 /* generic accessor macros */
259 #define CQE_WRID_HI(x) ((x)->u.gen.wrid_hi)
260 #define CQE_WRID_LOW(x) ((x)->u.gen.wrid_low)
261 #define CQE_DRAIN_COOKIE(x) (x)->u.drain_cookie;
263 /* macros for flit 3 of the cqe */
264 #define S_CQE_GENBIT 63
265 #define M_CQE_GENBIT 0x1
266 #define G_CQE_GENBIT(x) (((x) >> S_CQE_GENBIT) & M_CQE_GENBIT)
267 #define V_CQE_GENBIT(x) ((x)<<S_CQE_GENBIT)
269 #define S_CQE_OVFBIT 62
270 #define M_CQE_OVFBIT 0x1
271 #define G_CQE_OVFBIT(x) ((((x) >> S_CQE_OVFBIT)) & M_CQE_OVFBIT)
273 #define S_CQE_IQTYPE 60
274 #define M_CQE_IQTYPE 0x3
275 #define G_CQE_IQTYPE(x) ((((x) >> S_CQE_IQTYPE)) & M_CQE_IQTYPE)
277 #define M_CQE_TS 0x0fffffffffffffffULL
278 #define G_CQE_TS(x) ((x) & M_CQE_TS)
280 #define CQE_OVFBIT(x) ((unsigned)G_CQE_OVFBIT(be64_to_cpu((x)->bits_type_ts)))
281 #define CQE_GENBIT(x) ((unsigned)G_CQE_GENBIT(be64_to_cpu((x)->bits_type_ts)))
282 #define CQE_TS(x) (G_CQE_TS(be64_to_cpu((x)->bits_type_ts)))
297 DECLARE_PCI_UNMAP_ADDR(mapping);
298 unsigned long phys_addr;
299 struct t4_swsqe *sw_sq;
300 struct t4_swsqe *oldest_read;
317 union t4_recv_wr *queue;
319 DECLARE_PCI_UNMAP_ADDR(mapping);
320 struct t4_swrqe *sw_rq;
339 struct c4iw_rdev *rdev;
342 static inline int t4_rqes_posted(struct t4_wq *wq)
344 return wq->rq.in_use;
347 static inline int t4_rq_empty(struct t4_wq *wq)
349 return wq->rq.in_use == 0;
352 static inline int t4_rq_full(struct t4_wq *wq)
354 return wq->rq.in_use == (wq->rq.size - 1);
357 static inline u32 t4_rq_avail(struct t4_wq *wq)
359 return wq->rq.size - 1 - wq->rq.in_use;
362 static inline void t4_rq_produce(struct t4_wq *wq, u8 len16)
365 if (++wq->rq.pidx == wq->rq.size)
367 wq->rq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE);
368 if (wq->rq.wq_pidx >= wq->rq.size * T4_RQ_NUM_SLOTS)
369 wq->rq.wq_pidx %= wq->rq.size * T4_RQ_NUM_SLOTS;
372 static inline void t4_rq_consume(struct t4_wq *wq)
376 if (++wq->rq.cidx == wq->rq.size)
380 static inline u16 t4_rq_host_wq_pidx(struct t4_wq *wq)
382 return wq->rq.queue[wq->rq.size].status.host_wq_pidx;
385 static inline u16 t4_rq_wq_size(struct t4_wq *wq)
387 return wq->rq.size * T4_RQ_NUM_SLOTS;
390 static inline int t4_sq_empty(struct t4_wq *wq)
392 return wq->sq.in_use == 0;
395 static inline int t4_sq_full(struct t4_wq *wq)
397 return wq->sq.in_use == (wq->sq.size - 1);
400 static inline u32 t4_sq_avail(struct t4_wq *wq)
402 return wq->sq.size - 1 - wq->sq.in_use;
405 static inline void t4_sq_produce(struct t4_wq *wq, u8 len16)
408 if (++wq->sq.pidx == wq->sq.size)
410 wq->sq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE);
411 if (wq->sq.wq_pidx >= wq->sq.size * T4_SQ_NUM_SLOTS)
412 wq->sq.wq_pidx %= wq->sq.size * T4_SQ_NUM_SLOTS;
415 static inline void t4_sq_consume(struct t4_wq *wq)
418 if (++wq->sq.cidx == wq->sq.size)
422 static inline u16 t4_sq_host_wq_pidx(struct t4_wq *wq)
424 return wq->sq.queue[wq->sq.size].status.host_wq_pidx;
427 static inline u16 t4_sq_wq_size(struct t4_wq *wq)
429 return wq->sq.size * T4_SQ_NUM_SLOTS;
432 static inline void t4_ring_sq_db(struct t4_wq *wq, u16 inc)
435 writel(QID(wq->sq.qid) | PIDX(inc), wq->db);
438 static inline void t4_ring_rq_db(struct t4_wq *wq, u16 inc)
441 writel(QID(wq->rq.qid) | PIDX(inc), wq->db);
444 static inline int t4_wq_in_error(struct t4_wq *wq)
446 return wq->rq.queue[wq->rq.size].status.qp_err;
449 static inline void t4_set_wq_in_error(struct t4_wq *wq)
451 wq->rq.queue[wq->rq.size].status.qp_err = 1;
455 struct t4_cqe *queue;
457 DECLARE_PCI_UNMAP_ADDR(mapping);
458 struct t4_cqe *sw_queue;
460 struct c4iw_rdev *rdev;
465 u16 size; /* including status page */
475 static inline int t4_arm_cq(struct t4_cq *cq, int se)
479 while (cq->cidx_inc > CIDXINC_MASK) {
480 val = SEINTARM(0) | CIDXINC(CIDXINC_MASK) | TIMERREG(7) |
481 INGRESSQID(cq->cqid);
482 writel(val, cq->gts);
483 cq->cidx_inc -= CIDXINC_MASK;
485 val = SEINTARM(se) | CIDXINC(cq->cidx_inc) | TIMERREG(6) |
486 INGRESSQID(cq->cqid);
487 writel(val, cq->gts);
492 static inline void t4_swcq_produce(struct t4_cq *cq)
495 if (++cq->sw_pidx == cq->size)
499 static inline void t4_swcq_consume(struct t4_cq *cq)
502 if (++cq->sw_cidx == cq->size)
506 static inline void t4_hwcq_consume(struct t4_cq *cq)
508 cq->bits_type_ts = cq->queue[cq->cidx].bits_type_ts;
509 if (++cq->cidx_inc == (cq->size >> 4) || cq->cidx_inc == M_CIDXINC) {
512 val = SEINTARM(0) | CIDXINC(cq->cidx_inc) | TIMERREG(7) |
513 INGRESSQID(cq->cqid);
514 writel(val, cq->gts);
517 if (++cq->cidx == cq->size) {
523 static inline int t4_valid_cqe(struct t4_cq *cq, struct t4_cqe *cqe)
525 return (CQE_GENBIT(cqe) == cq->gen);
528 static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
534 prev_cidx = cq->size - 1;
536 prev_cidx = cq->cidx - 1;
538 if (cq->queue[prev_cidx].bits_type_ts != cq->bits_type_ts) {
541 printk(KERN_ERR MOD "cq overflow cqid %u\n", cq->cqid);
542 } else if (t4_valid_cqe(cq, &cq->queue[cq->cidx])) {
543 *cqe = &cq->queue[cq->cidx];
550 static inline struct t4_cqe *t4_next_sw_cqe(struct t4_cq *cq)
553 return &cq->sw_queue[cq->sw_cidx];
557 static inline int t4_next_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
563 else if (cq->sw_in_use)
564 *cqe = &cq->sw_queue[cq->sw_cidx];
566 ret = t4_next_hw_cqe(cq, cqe);
570 static inline int t4_cq_in_error(struct t4_cq *cq)
572 return ((struct t4_status_page *)&cq->queue[cq->size])->qp_err;
575 static inline void t4_set_cq_in_error(struct t4_cq *cq)
577 ((struct t4_status_page *)&cq->queue[cq->size])->qp_err = 1;