Linux 3.11-rc3
[cris-mirror.git] / drivers / infiniband / hw / cxgb4 / t4.h
blobebcb03bd1b72ed6003bd7ae85e88dd8ce9c5595d
1 /*
2 * Copyright (c) 2009-2010 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
12 * conditions are met:
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.
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.
31 #ifndef __T4_H__
32 #define __T4_H__
34 #include "t4_hw.h"
35 #include "t4_regs.h"
36 #include "t4_msg.h"
37 #include "t4fw_ri_api.h"
39 #define T4_MAX_NUM_QP (1<<16)
40 #define T4_MAX_NUM_CQ (1<<15)
41 #define T4_MAX_NUM_PD (1<<15)
42 #define T4_EQ_STATUS_ENTRIES (L1_CACHE_BYTES > 64 ? 2 : 1)
43 #define T4_MAX_EQ_SIZE (65520 - T4_EQ_STATUS_ENTRIES)
44 #define T4_MAX_IQ_SIZE (65520 - 1)
45 #define T4_MAX_RQ_SIZE (8192 - T4_EQ_STATUS_ENTRIES)
46 #define T4_MAX_SQ_SIZE (T4_MAX_EQ_SIZE - 1)
47 #define T4_MAX_QP_DEPTH (T4_MAX_RQ_SIZE - 1)
48 #define T4_MAX_CQ_DEPTH (T4_MAX_IQ_SIZE - 1)
49 #define T4_MAX_NUM_STAG (1<<15)
50 #define T4_MAX_MR_SIZE (~0ULL - 1)
51 #define T4_PAGESIZE_MASK 0xffff000 /* 4KB-128MB */
52 #define T4_STAG_UNSET 0xffffffff
53 #define T4_FW_MAJ 0
54 #define T4_EQ_STATUS_ENTRIES (L1_CACHE_BYTES > 64 ? 2 : 1)
55 #define A_PCIE_MA_SYNC 0x30b4
57 struct t4_status_page {
58 __be32 rsvd1; /* flit 0 - hw owns */
59 __be16 rsvd2;
60 __be16 qid;
61 __be16 cidx;
62 __be16 pidx;
63 u8 qp_err; /* flit 1 - sw owns */
64 u8 db_off;
65 u8 pad;
66 u16 host_wq_pidx;
67 u16 host_cidx;
68 u16 host_pidx;
71 #define T4_EQ_ENTRY_SIZE 64
73 #define T4_SQ_NUM_SLOTS 5
74 #define T4_SQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_SQ_NUM_SLOTS)
75 #define T4_MAX_SEND_SGE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \
76 sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
77 #define T4_MAX_SEND_INLINE ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_send_wr) - \
78 sizeof(struct fw_ri_immd)))
79 #define T4_MAX_WRITE_INLINE ((T4_SQ_NUM_BYTES - \
80 sizeof(struct fw_ri_rdma_write_wr) - \
81 sizeof(struct fw_ri_immd)))
82 #define T4_MAX_WRITE_SGE ((T4_SQ_NUM_BYTES - \
83 sizeof(struct fw_ri_rdma_write_wr) - \
84 sizeof(struct fw_ri_isgl)) / sizeof(struct fw_ri_sge))
85 #define T4_MAX_FR_IMMD ((T4_SQ_NUM_BYTES - sizeof(struct fw_ri_fr_nsmr_wr) - \
86 sizeof(struct fw_ri_immd)) & ~31UL)
87 #define T4_MAX_FR_DEPTH (1024 / sizeof(u64))
89 #define T4_RQ_NUM_SLOTS 2
90 #define T4_RQ_NUM_BYTES (T4_EQ_ENTRY_SIZE * T4_RQ_NUM_SLOTS)
91 #define T4_MAX_RECV_SGE 4
93 union t4_wr {
94 struct fw_ri_res_wr res;
95 struct fw_ri_wr ri;
96 struct fw_ri_rdma_write_wr write;
97 struct fw_ri_send_wr send;
98 struct fw_ri_rdma_read_wr read;
99 struct fw_ri_bind_mw_wr bind;
100 struct fw_ri_fr_nsmr_wr fr;
101 struct fw_ri_inv_lstag_wr inv;
102 struct t4_status_page status;
103 __be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_SQ_NUM_SLOTS];
106 union t4_recv_wr {
107 struct fw_ri_recv_wr recv;
108 struct t4_status_page status;
109 __be64 flits[T4_EQ_ENTRY_SIZE / sizeof(__be64) * T4_RQ_NUM_SLOTS];
112 static inline void init_wr_hdr(union t4_wr *wqe, u16 wrid,
113 enum fw_wr_opcodes opcode, u8 flags, u8 len16)
115 wqe->send.opcode = (u8)opcode;
116 wqe->send.flags = flags;
117 wqe->send.wrid = wrid;
118 wqe->send.r1[0] = 0;
119 wqe->send.r1[1] = 0;
120 wqe->send.r1[2] = 0;
121 wqe->send.len16 = len16;
124 /* CQE/AE status codes */
125 #define T4_ERR_SUCCESS 0x0
126 #define T4_ERR_STAG 0x1 /* STAG invalid: either the */
127 /* STAG is offlimt, being 0, */
128 /* or STAG_key mismatch */
129 #define T4_ERR_PDID 0x2 /* PDID mismatch */
130 #define T4_ERR_QPID 0x3 /* QPID mismatch */
131 #define T4_ERR_ACCESS 0x4 /* Invalid access right */
132 #define T4_ERR_WRAP 0x5 /* Wrap error */
133 #define T4_ERR_BOUND 0x6 /* base and bounds voilation */
134 #define T4_ERR_INVALIDATE_SHARED_MR 0x7 /* attempt to invalidate a */
135 /* shared memory region */
136 #define T4_ERR_INVALIDATE_MR_WITH_MW_BOUND 0x8 /* attempt to invalidate a */
137 /* shared memory region */
138 #define T4_ERR_ECC 0x9 /* ECC error detected */
139 #define T4_ERR_ECC_PSTAG 0xA /* ECC error detected when */
140 /* reading PSTAG for a MW */
141 /* Invalidate */
142 #define T4_ERR_PBL_ADDR_BOUND 0xB /* pbl addr out of bounds: */
143 /* software error */
144 #define T4_ERR_SWFLUSH 0xC /* SW FLUSHED */
145 #define T4_ERR_CRC 0x10 /* CRC error */
146 #define T4_ERR_MARKER 0x11 /* Marker error */
147 #define T4_ERR_PDU_LEN_ERR 0x12 /* invalid PDU length */
148 #define T4_ERR_OUT_OF_RQE 0x13 /* out of RQE */
149 #define T4_ERR_DDP_VERSION 0x14 /* wrong DDP version */
150 #define T4_ERR_RDMA_VERSION 0x15 /* wrong RDMA version */
151 #define T4_ERR_OPCODE 0x16 /* invalid rdma opcode */
152 #define T4_ERR_DDP_QUEUE_NUM 0x17 /* invalid ddp queue number */
153 #define T4_ERR_MSN 0x18 /* MSN error */
154 #define T4_ERR_TBIT 0x19 /* tag bit not set correctly */
155 #define T4_ERR_MO 0x1A /* MO not 0 for TERMINATE */
156 /* or READ_REQ */
157 #define T4_ERR_MSN_GAP 0x1B
158 #define T4_ERR_MSN_RANGE 0x1C
159 #define T4_ERR_IRD_OVERFLOW 0x1D
160 #define T4_ERR_RQE_ADDR_BOUND 0x1E /* RQE addr out of bounds: */
161 /* software error */
162 #define T4_ERR_INTERNAL_ERR 0x1F /* internal error (opcode */
163 /* mismatch) */
165 * CQE defs
167 struct t4_cqe {
168 __be32 header;
169 __be32 len;
170 union {
171 struct {
172 __be32 stag;
173 __be32 msn;
174 } rcqe;
175 struct {
176 u32 nada1;
177 u16 nada2;
178 u16 cidx;
179 } scqe;
180 struct {
181 __be32 wrid_hi;
182 __be32 wrid_low;
183 } gen;
184 } u;
185 __be64 reserved;
186 __be64 bits_type_ts;
189 /* macros for flit 0 of the cqe */
191 #define S_CQE_QPID 12
192 #define M_CQE_QPID 0xFFFFF
193 #define G_CQE_QPID(x) ((((x) >> S_CQE_QPID)) & M_CQE_QPID)
194 #define V_CQE_QPID(x) ((x)<<S_CQE_QPID)
196 #define S_CQE_SWCQE 11
197 #define M_CQE_SWCQE 0x1
198 #define G_CQE_SWCQE(x) ((((x) >> S_CQE_SWCQE)) & M_CQE_SWCQE)
199 #define V_CQE_SWCQE(x) ((x)<<S_CQE_SWCQE)
201 #define S_CQE_STATUS 5
202 #define M_CQE_STATUS 0x1F
203 #define G_CQE_STATUS(x) ((((x) >> S_CQE_STATUS)) & M_CQE_STATUS)
204 #define V_CQE_STATUS(x) ((x)<<S_CQE_STATUS)
206 #define S_CQE_TYPE 4
207 #define M_CQE_TYPE 0x1
208 #define G_CQE_TYPE(x) ((((x) >> S_CQE_TYPE)) & M_CQE_TYPE)
209 #define V_CQE_TYPE(x) ((x)<<S_CQE_TYPE)
211 #define S_CQE_OPCODE 0
212 #define M_CQE_OPCODE 0xF
213 #define G_CQE_OPCODE(x) ((((x) >> S_CQE_OPCODE)) & M_CQE_OPCODE)
214 #define V_CQE_OPCODE(x) ((x)<<S_CQE_OPCODE)
216 #define SW_CQE(x) (G_CQE_SWCQE(be32_to_cpu((x)->header)))
217 #define CQE_QPID(x) (G_CQE_QPID(be32_to_cpu((x)->header)))
218 #define CQE_TYPE(x) (G_CQE_TYPE(be32_to_cpu((x)->header)))
219 #define SQ_TYPE(x) (CQE_TYPE((x)))
220 #define RQ_TYPE(x) (!CQE_TYPE((x)))
221 #define CQE_STATUS(x) (G_CQE_STATUS(be32_to_cpu((x)->header)))
222 #define CQE_OPCODE(x) (G_CQE_OPCODE(be32_to_cpu((x)->header)))
224 #define CQE_SEND_OPCODE(x)( \
225 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND) || \
226 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE) || \
227 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_INV) || \
228 (G_CQE_OPCODE(be32_to_cpu((x)->header)) == FW_RI_SEND_WITH_SE_INV))
230 #define CQE_LEN(x) (be32_to_cpu((x)->len))
232 /* used for RQ completion processing */
233 #define CQE_WRID_STAG(x) (be32_to_cpu((x)->u.rcqe.stag))
234 #define CQE_WRID_MSN(x) (be32_to_cpu((x)->u.rcqe.msn))
236 /* used for SQ completion processing */
237 #define CQE_WRID_SQ_IDX(x) ((x)->u.scqe.cidx)
239 /* generic accessor macros */
240 #define CQE_WRID_HI(x) ((x)->u.gen.wrid_hi)
241 #define CQE_WRID_LOW(x) ((x)->u.gen.wrid_low)
243 /* macros for flit 3 of the cqe */
244 #define S_CQE_GENBIT 63
245 #define M_CQE_GENBIT 0x1
246 #define G_CQE_GENBIT(x) (((x) >> S_CQE_GENBIT) & M_CQE_GENBIT)
247 #define V_CQE_GENBIT(x) ((x)<<S_CQE_GENBIT)
249 #define S_CQE_OVFBIT 62
250 #define M_CQE_OVFBIT 0x1
251 #define G_CQE_OVFBIT(x) ((((x) >> S_CQE_OVFBIT)) & M_CQE_OVFBIT)
253 #define S_CQE_IQTYPE 60
254 #define M_CQE_IQTYPE 0x3
255 #define G_CQE_IQTYPE(x) ((((x) >> S_CQE_IQTYPE)) & M_CQE_IQTYPE)
257 #define M_CQE_TS 0x0fffffffffffffffULL
258 #define G_CQE_TS(x) ((x) & M_CQE_TS)
260 #define CQE_OVFBIT(x) ((unsigned)G_CQE_OVFBIT(be64_to_cpu((x)->bits_type_ts)))
261 #define CQE_GENBIT(x) ((unsigned)G_CQE_GENBIT(be64_to_cpu((x)->bits_type_ts)))
262 #define CQE_TS(x) (G_CQE_TS(be64_to_cpu((x)->bits_type_ts)))
264 struct t4_swsqe {
265 u64 wr_id;
266 struct t4_cqe cqe;
267 int read_len;
268 int opcode;
269 int complete;
270 int signaled;
271 u16 idx;
274 static inline pgprot_t t4_pgprot_wc(pgprot_t prot)
276 #if defined(__i386__) || defined(__x86_64__) || defined(CONFIG_PPC64)
277 return pgprot_writecombine(prot);
278 #else
279 return pgprot_noncached(prot);
280 #endif
283 enum {
284 T4_SQ_ONCHIP = (1<<0),
287 struct t4_sq {
288 union t4_wr *queue;
289 dma_addr_t dma_addr;
290 DEFINE_DMA_UNMAP_ADDR(mapping);
291 unsigned long phys_addr;
292 struct t4_swsqe *sw_sq;
293 struct t4_swsqe *oldest_read;
294 u64 udb;
295 size_t memsize;
296 u32 qid;
297 u16 in_use;
298 u16 size;
299 u16 cidx;
300 u16 pidx;
301 u16 wq_pidx;
302 u16 flags;
305 struct t4_swrqe {
306 u64 wr_id;
309 struct t4_rq {
310 union t4_recv_wr *queue;
311 dma_addr_t dma_addr;
312 DEFINE_DMA_UNMAP_ADDR(mapping);
313 struct t4_swrqe *sw_rq;
314 u64 udb;
315 size_t memsize;
316 u32 qid;
317 u32 msn;
318 u32 rqt_hwaddr;
319 u16 rqt_size;
320 u16 in_use;
321 u16 size;
322 u16 cidx;
323 u16 pidx;
324 u16 wq_pidx;
327 struct t4_wq {
328 struct t4_sq sq;
329 struct t4_rq rq;
330 void __iomem *db;
331 void __iomem *gts;
332 struct c4iw_rdev *rdev;
335 static inline int t4_rqes_posted(struct t4_wq *wq)
337 return wq->rq.in_use;
340 static inline int t4_rq_empty(struct t4_wq *wq)
342 return wq->rq.in_use == 0;
345 static inline int t4_rq_full(struct t4_wq *wq)
347 return wq->rq.in_use == (wq->rq.size - 1);
350 static inline u32 t4_rq_avail(struct t4_wq *wq)
352 return wq->rq.size - 1 - wq->rq.in_use;
355 static inline void t4_rq_produce(struct t4_wq *wq, u8 len16)
357 wq->rq.in_use++;
358 if (++wq->rq.pidx == wq->rq.size)
359 wq->rq.pidx = 0;
360 wq->rq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE);
361 if (wq->rq.wq_pidx >= wq->rq.size * T4_RQ_NUM_SLOTS)
362 wq->rq.wq_pidx %= wq->rq.size * T4_RQ_NUM_SLOTS;
365 static inline void t4_rq_consume(struct t4_wq *wq)
367 wq->rq.in_use--;
368 wq->rq.msn++;
369 if (++wq->rq.cidx == wq->rq.size)
370 wq->rq.cidx = 0;
373 static inline u16 t4_rq_host_wq_pidx(struct t4_wq *wq)
375 return wq->rq.queue[wq->rq.size].status.host_wq_pidx;
378 static inline u16 t4_rq_wq_size(struct t4_wq *wq)
380 return wq->rq.size * T4_RQ_NUM_SLOTS;
383 static inline int t4_sq_onchip(struct t4_sq *sq)
385 return sq->flags & T4_SQ_ONCHIP;
388 static inline int t4_sq_empty(struct t4_wq *wq)
390 return wq->sq.in_use == 0;
393 static inline int t4_sq_full(struct t4_wq *wq)
395 return wq->sq.in_use == (wq->sq.size - 1);
398 static inline u32 t4_sq_avail(struct t4_wq *wq)
400 return wq->sq.size - 1 - wq->sq.in_use;
403 static inline void t4_sq_produce(struct t4_wq *wq, u8 len16)
405 wq->sq.in_use++;
406 if (++wq->sq.pidx == wq->sq.size)
407 wq->sq.pidx = 0;
408 wq->sq.wq_pidx += DIV_ROUND_UP(len16*16, T4_EQ_ENTRY_SIZE);
409 if (wq->sq.wq_pidx >= wq->sq.size * T4_SQ_NUM_SLOTS)
410 wq->sq.wq_pidx %= wq->sq.size * T4_SQ_NUM_SLOTS;
413 static inline void t4_sq_consume(struct t4_wq *wq)
415 wq->sq.in_use--;
416 if (++wq->sq.cidx == wq->sq.size)
417 wq->sq.cidx = 0;
420 static inline u16 t4_sq_host_wq_pidx(struct t4_wq *wq)
422 return wq->sq.queue[wq->sq.size].status.host_wq_pidx;
425 static inline u16 t4_sq_wq_size(struct t4_wq *wq)
427 return wq->sq.size * T4_SQ_NUM_SLOTS;
430 static inline void t4_ring_sq_db(struct t4_wq *wq, u16 inc)
432 wmb();
433 writel(QID(wq->sq.qid) | PIDX(inc), wq->db);
436 static inline void t4_ring_rq_db(struct t4_wq *wq, u16 inc)
438 wmb();
439 writel(QID(wq->rq.qid) | PIDX(inc), wq->db);
442 static inline int t4_wq_in_error(struct t4_wq *wq)
444 return wq->rq.queue[wq->rq.size].status.qp_err;
447 static inline void t4_set_wq_in_error(struct t4_wq *wq)
449 wq->rq.queue[wq->rq.size].status.qp_err = 1;
452 static inline void t4_disable_wq_db(struct t4_wq *wq)
454 wq->rq.queue[wq->rq.size].status.db_off = 1;
457 static inline void t4_enable_wq_db(struct t4_wq *wq)
459 wq->rq.queue[wq->rq.size].status.db_off = 0;
462 static inline int t4_wq_db_enabled(struct t4_wq *wq)
464 return !wq->rq.queue[wq->rq.size].status.db_off;
467 struct t4_cq {
468 struct t4_cqe *queue;
469 dma_addr_t dma_addr;
470 DEFINE_DMA_UNMAP_ADDR(mapping);
471 struct t4_cqe *sw_queue;
472 void __iomem *gts;
473 struct c4iw_rdev *rdev;
474 u64 ugts;
475 size_t memsize;
476 __be64 bits_type_ts;
477 u32 cqid;
478 u16 size; /* including status page */
479 u16 cidx;
480 u16 sw_pidx;
481 u16 sw_cidx;
482 u16 sw_in_use;
483 u16 cidx_inc;
484 u8 gen;
485 u8 error;
488 static inline int t4_arm_cq(struct t4_cq *cq, int se)
490 u32 val;
492 while (cq->cidx_inc > CIDXINC_MASK) {
493 val = SEINTARM(0) | CIDXINC(CIDXINC_MASK) | TIMERREG(7) |
494 INGRESSQID(cq->cqid);
495 writel(val, cq->gts);
496 cq->cidx_inc -= CIDXINC_MASK;
498 val = SEINTARM(se) | CIDXINC(cq->cidx_inc) | TIMERREG(6) |
499 INGRESSQID(cq->cqid);
500 writel(val, cq->gts);
501 cq->cidx_inc = 0;
502 return 0;
505 static inline void t4_swcq_produce(struct t4_cq *cq)
507 cq->sw_in_use++;
508 if (++cq->sw_pidx == cq->size)
509 cq->sw_pidx = 0;
512 static inline void t4_swcq_consume(struct t4_cq *cq)
514 cq->sw_in_use--;
515 if (++cq->sw_cidx == cq->size)
516 cq->sw_cidx = 0;
519 static inline void t4_hwcq_consume(struct t4_cq *cq)
521 cq->bits_type_ts = cq->queue[cq->cidx].bits_type_ts;
522 if (++cq->cidx_inc == (cq->size >> 4)) {
523 u32 val;
525 val = SEINTARM(0) | CIDXINC(cq->cidx_inc) | TIMERREG(7) |
526 INGRESSQID(cq->cqid);
527 writel(val, cq->gts);
528 cq->cidx_inc = 0;
530 if (++cq->cidx == cq->size) {
531 cq->cidx = 0;
532 cq->gen ^= 1;
536 static inline int t4_valid_cqe(struct t4_cq *cq, struct t4_cqe *cqe)
538 return (CQE_GENBIT(cqe) == cq->gen);
541 static inline int t4_next_hw_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
543 int ret;
544 u16 prev_cidx;
546 if (cq->cidx == 0)
547 prev_cidx = cq->size - 1;
548 else
549 prev_cidx = cq->cidx - 1;
551 if (cq->queue[prev_cidx].bits_type_ts != cq->bits_type_ts) {
552 ret = -EOVERFLOW;
553 cq->error = 1;
554 printk(KERN_ERR MOD "cq overflow cqid %u\n", cq->cqid);
555 } else if (t4_valid_cqe(cq, &cq->queue[cq->cidx])) {
556 *cqe = &cq->queue[cq->cidx];
557 ret = 0;
558 } else
559 ret = -ENODATA;
560 return ret;
563 static inline struct t4_cqe *t4_next_sw_cqe(struct t4_cq *cq)
565 if (cq->sw_in_use)
566 return &cq->sw_queue[cq->sw_cidx];
567 return NULL;
570 static inline int t4_next_cqe(struct t4_cq *cq, struct t4_cqe **cqe)
572 int ret = 0;
574 if (cq->error)
575 ret = -ENODATA;
576 else if (cq->sw_in_use)
577 *cqe = &cq->sw_queue[cq->sw_cidx];
578 else
579 ret = t4_next_hw_cqe(cq, cqe);
580 return ret;
583 static inline int t4_cq_in_error(struct t4_cq *cq)
585 return ((struct t4_status_page *)&cq->queue[cq->size])->qp_err;
588 static inline void t4_set_cq_in_error(struct t4_cq *cq)
590 ((struct t4_status_page *)&cq->queue[cq->size])->qp_err = 1;
592 #endif